| Message ID | 1482843136-12838-1-git-send-email-pali.rohar@gmail.com (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Headers | show |
> Dell platform team told us that some (DMI whitelisted) Dell Latitude > machines have ST microelectronics accelerometer at i2c address 0x29. > That > i2c address is not specified in DMI or ACPI, so runtime detection > without > whitelist which is below is not possible. I'd think this should rather live somewhere in drivers/platform/x86/dell*.c? -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tuesday 27 December 2016 14:43:49 Wolfram Sang wrote: > > Dell platform team told us that some (DMI whitelisted) Dell > > Latitude machines have ST microelectronics accelerometer at i2c > > address 0x29. That > > i2c address is not specified in DMI or ACPI, so runtime detection > > without > > whitelist which is below is not possible. > > I'd think this should rather live somewhere in > drivers/platform/x86/dell*.c? i2c_new_device() with lis3lv02d for i801 i2c bus needs to be called after initializing i2c-i801 bus driver. I have no idea how to do it (properly) outside of i2c-i801.c file. Same thing is done for Fujitsu machines, see function i801_probe_optional_slaves() in i2c-i801.c file. So I did similar approach for Dell machines.
On Tue, Dec 27, 2016 at 3:51 PM, Pali Rohár <pali.rohar@gmail.com> wrote: > On Tuesday 27 December 2016 14:43:49 Wolfram Sang wrote: >> > Dell platform team told us that some (DMI whitelisted) Dell >> > Latitude machines have ST microelectronics accelerometer at i2c >> > address 0x29. That >> > i2c address is not specified in DMI or ACPI, so runtime detection >> > without >> > whitelist which is below is not possible. >> >> I'd think this should rather live somewhere in >> drivers/platform/x86/dell*.c? > > i2c_new_device() with lis3lv02d for i801 i2c bus needs to be called > after initializing i2c-i801 bus driver. > > I have no idea how to do it (properly) outside of i2c-i801.c file. I doubt we need a single line of code for this. See [1] and perhaps create an EFI variable with necessary upgrade device node. > Same thing is done for Fujitsu machines, see function > i801_probe_optional_slaves() in i2c-i801.c file. So I did similar > approach for Dell machines. Perhaps, this also needs to be converted to use EFI variable. [1] https://lwn.net/Articles/693212/
On Wed, 28 Dec 2016 00:15:30 +0200, Andy Shevchenko said: > On Tue, Dec 27, 2016 at 3:51 PM, Pali Rohár <pali.rohar@gmail.com> wrote: > > I have no idea how to do it (properly) outside of i2c-i801.c file. > > I doubt we need a single line of code for this. See [1] and perhaps > create an EFI variable with necessary upgrade device node. > > > Same thing is done for Fujitsu machines, see function > > i801_probe_optional_slaves() in i2c-i801.c file. So I did similar > > approach for Dell machines. > > Perhaps, this also needs to be converted to use EFI variable. > > [1] https://lwn.net/Articles/693212/ There's no guarantee that the laptops in question are booted with UEFI, as Dell still supports legacy boot. So assuming the presence of EFI variables is somewhat problematic. In addition, it requires the user (or something in userspace) to set the UEFI variable or configfs tweak, rather than Just Working Out Of The Box.
On Wed, Dec 28, 2016 at 12:41 AM, <Valdis.Kletnieks@vt.edu> wrote: > On Wed, 28 Dec 2016 00:15:30 +0200, Andy Shevchenko said: >> On Tue, Dec 27, 2016 at 3:51 PM, Pali Rohár <pali.rohar@gmail.com> wrote: > >> > I have no idea how to do it (properly) outside of i2c-i801.c file. >> >> I doubt we need a single line of code for this. See [1] and perhaps >> create an EFI variable with necessary upgrade device node. >> >> > Same thing is done for Fujitsu machines, see function >> > i801_probe_optional_slaves() in i2c-i801.c file. So I did similar >> > approach for Dell machines. >> >> Perhaps, this also needs to be converted to use EFI variable. >> >> [1] https://lwn.net/Articles/693212/ > > There's no guarantee that the laptops in question are booted with UEFI, > as Dell still supports legacy boot. So assuming the presence of EFI variables > is somewhat problematic. > In addition, it requires the user (or something in userspace) to set the UEFI > variable or configfs tweak, rather than Just Working Out Of The Box. I have no strong opinion, though I don't support the idea to put all hacks in the world to the kernel. For example, we have user space tool to switch USB modem from storage to actual communication device and that is just working out of the box. Mika, Darren, what are your opinions?
On Tuesday 27 December 2016 23:15:30 Andy Shevchenko wrote: > On Tue, Dec 27, 2016 at 3:51 PM, Pali Rohár <pali.rohar@gmail.com> > wrote: > > On Tuesday 27 December 2016 14:43:49 Wolfram Sang wrote: > >> > Dell platform team told us that some (DMI whitelisted) Dell > >> > Latitude machines have ST microelectronics accelerometer at i2c > >> > address 0x29. That > >> > i2c address is not specified in DMI or ACPI, so runtime > >> > detection without > >> > whitelist which is below is not possible. > >> > >> I'd think this should rather live somewhere in > >> drivers/platform/x86/dell*.c? > > > > i2c_new_device() with lis3lv02d for i801 i2c bus needs to be called > > after initializing i2c-i801 bus driver. > > > > I have no idea how to do it (properly) outside of i2c-i801.c file. > > I doubt we need a single line of code for this. See [1] and perhaps > create an EFI variable with necessary upgrade device node. Sorry, but EFI variable is not accessible from BIOS booted kernel. So such thing will not work.
On Wednesday 28 December 2016 08:55:18 Andy Shevchenko wrote: > On Wed, Dec 28, 2016 at 12:41 AM, <Valdis.Kletnieks@vt.edu> wrote: > > On Wed, 28 Dec 2016 00:15:30 +0200, Andy Shevchenko said: > >> On Tue, Dec 27, 2016 at 3:51 PM, Pali Rohár <pali.rohar@gmail.com> wrote: > > > >> > I have no idea how to do it (properly) outside of i2c-i801.c file. > >> > >> I doubt we need a single line of code for this. See [1] and perhaps > >> create an EFI variable with necessary upgrade device node. > >> > >> > Same thing is done for Fujitsu machines, see function > >> > i801_probe_optional_slaves() in i2c-i801.c file. So I did similar > >> > approach for Dell machines. > >> > >> Perhaps, this also needs to be converted to use EFI variable. > >> > >> [1] https://lwn.net/Articles/693212/ > > > > There's no guarantee that the laptops in question are booted with UEFI, > > as Dell still supports legacy boot. So assuming the presence of EFI variables > > is somewhat problematic. > > > > In addition, it requires the user (or something in userspace) to set the UEFI > > variable or configfs tweak, rather than Just Working Out Of The Box. > > I have no strong opinion, though I don't support the idea to put all > hacks in the world to the kernel. For example, we have user space tool > to switch USB modem from storage to actual communication device and > that is just working out of the box. > > Mika, Darren, what are your opinions? I have absolutely no idea how to you want to achieve calling that i2c_new_device() registration without kernel patches. So before starting discussion which option to use (EFI, kernel patch, userspace script, etc...) please describe how would you implement such logic with different options.
On Tue, Dec 27, 2016 at 02:51:01PM +0100, Pali Rohár wrote: > On Tuesday 27 December 2016 14:43:49 Wolfram Sang wrote: > > > Dell platform team told us that some (DMI whitelisted) Dell > > > Latitude machines have ST microelectronics accelerometer at i2c > > > address 0x29. That > > > i2c address is not specified in DMI or ACPI, so runtime detection > > > without > > > whitelist which is below is not possible. > > > > I'd think this should rather live somewhere in > > drivers/platform/x86/dell*.c? > > i2c_new_device() with lis3lv02d for i801 i2c bus needs to be called > after initializing i2c-i801 bus driver. > > I have no idea how to do it (properly) outside of i2c-i801.c file. I once used bus_notifiers to achieve something similar. You could check arch/arm/mach-shmobile/regulator-quirk-rcar-gen2.c to see an action triggered once a client device got added, but you could act on another action like BUS_NOTIFY_BOUND_DRIVER. I used exactly that, too, somewhen somewhere. Haven't checked if that helps here, too. And since we have a precedence (Fujitsu case), I'll leave it to Jean who is the maintainer of this driver. Thanks, Wolfram -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
> I have absolutely no idea how to you want to achieve calling that > i2c_new_device() registration > without kernel patches. Documentation/i2c/instantiating-devices lists all supported methods. Method 4 is userspace instantiation. -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, 28 Dec 2016 15:03:02 +0100, Wolfram Sang said: > > I have absolutely no idea how to you want to achieve calling that > > i2c_new_device() registration > > without kernel patches. > > Documentation/i2c/instantiating-devices lists all supported methods. > Method 4 is userspace instantiation. I'd be totally OK with userspace doing it, except for the question "How good will distros be about shipping it"? I don't have any sense of how good Fedora and Ubuntu and so on will be about making sure the userspace part is already done for the user. Anybody got evidence one way or another?
> Dell platform team told us that some (DMI whitelisted) Dell Latitude > machines have ST microelectronics accelerometer at i2c address 0x29. That > i2c address is not specified in DMI or ACPI, so runtime detection without > whitelist which is below is not possible. > > Presence of that ST microelectronics accelerometer is verified by existence > of SMO88xx ACPI device which represent that accelerometer. Unfortunately > without i2c address. This part of the commit message sounded a bit confusing to me at first because there is already an ACPI driver which handles SMO88xx devices (dell-smo8800). My understanding is that: * the purpose of this patch is to expose a richer interface (as provided by lis3lv02d) to these devices on some machines, * on whitelisted machines, dell-smo8800 and lis3lv02d can work simultaneously (even though dell-smo8800 effectively duplicates the work that lis3lv02d does). If I got something wrong, please correct me. If I got it right, it might make sense to rephrase the commit message a bit so that the first bullet point above is immediately clear to the reader. > > This patch registers lis3lv02d device at i2c address 0x29 if is detected. > > Finally commit a7ae81952cda ("i2c: i801: Allow ACPI SystemIO OpRegion to > conflict with PCI BAR") allowed to use i2c-i801 driver on Dell machines so > lis3lv02d correctly initialize accelerometer. > > Tested on Dell Latitude E6440. > > Signed-off-by: Pali Rohár <pali.rohar@gmail.com> > --- > drivers/i2c/busses/i2c-i801.c | 98 +++++++++++++++++++++++++++++++++++++++++ > 1 file changed, 98 insertions(+) > > diff --git a/drivers/i2c/busses/i2c-i801.c b/drivers/i2c/busses/i2c-i801.c > index eb3627f..188cfd4 100644 > --- a/drivers/i2c/busses/i2c-i801.c > +++ b/drivers/i2c/busses/i2c-i801.c > @@ -1118,6 +1118,101 @@ static void dmi_check_onboard_devices(const struct dmi_header *dm, void *adap) > } > } > > +static acpi_status check_acpi_smo88xx_device(acpi_handle obj_handle, > + u32 nesting_level, > + void *context, > + void **return_value) > +{ > + struct acpi_device_info *info; > + acpi_status status; > + char *hid; > + > + status = acpi_get_object_info(obj_handle, &info); acpi_get_object_info() allocates the returned buffer, which the caller has to free. > + if (!ACPI_SUCCESS(status) || !(info->valid & ACPI_VALID_HID)) > + return AE_OK; > + > + hid = info->hardware_id.string; > + if (!hid) > + return AE_OK; > + > + if (strlen(hid) < 7) > + return AE_OK; > + > + if (memcmp(hid, "SMO88", 5) != 0) > + return AE_OK; > + > + *((bool *)return_value) = true; > + return AE_CTRL_TERMINATE; > +} > + > +static bool is_dell_system_with_lis3lv02d(void) > +{ > + bool found; > + acpi_status status; > + const char *vendor; > + > + vendor = dmi_get_system_info(DMI_SYS_VENDOR); > + if (strcmp(vendor, "Dell Inc.") != 0) > + return false; > + > + /* > + * Check if ACPI device SMO88xx exists and if is enabled. That ACPI > + * device represent our ST microelectronics lis3lv02d accelerometer but > + * unfortunately without any other additional information. > + */ > + found = false; > + status = acpi_get_devices(NULL, check_acpi_smo88xx_device, NULL, > + (void **)&found); > + if (!ACPI_SUCCESS(status) || !found) > + return false; > + > + return true; > +} > + > +/* > + * Dell platform team told us that these Latitude devices have > + * ST microelectronics accelerometer at i2c address 0x29. > + * That i2c address is not specified in DMI or ACPI, so runtime > + * detection without whitelist which is below is not possible. > + */ > +static const char * const dmi_dell_product_names[] = { > + "Latitude E5250", > + "Latitude E5450", > + "Latitude E5550", > + "Latitude E6440", > + "Latitude E6440 ATG", > + "Latitude E6540", > +}; > + > +static void register_dell_lis3lv02d_i2c_device(struct i801_priv *priv) > +{ > + struct i2c_board_info info; > + const char *product_name; > + bool known_i2c_address; > + int i; > + > + known_i2c_address = false; > + product_name = dmi_get_system_info(DMI_PRODUCT_NAME); > + for (i = 0; i < ARRAY_SIZE(dmi_dell_product_names); ++i) { > + if (strcmp(product_name, dmi_dell_product_names[i]) == 0) { > + known_i2c_address = true; > + break; > + } > + } > + > + if (!known_i2c_address) { > + dev_warn(&priv->pci_dev->dev, > + "Accelerometer lis3lv02d i2c device is present " > + "but its i2c address is unknown, skipping ...\n"); You are probably well aware of this, but checkpatch prefers keeping long log messages in one line. I am pointing it out just in case. > + return; > + } > + > + memset(&info, 0, sizeof(struct i2c_board_info)); How about just doing "struct i2c_board_info info = { 0 };" instead? > + info.addr = 0x29; > + strlcpy(info.type, "lis3lv02d", I2C_NAME_SIZE); > + i2c_new_device(&priv->adapter, &info); > +} > + > /* Register optional slaves */ > static void i801_probe_optional_slaves(struct i801_priv *priv) > { > @@ -1136,6 +1231,9 @@ static void i801_probe_optional_slaves(struct i801_priv *priv) > > if (dmi_name_in_vendors("FUJITSU")) > dmi_walk(dmi_check_onboard_devices, &priv->adapter); > + > + if (is_dell_system_with_lis3lv02d()) > + register_dell_lis3lv02d_i2c_device(priv); > } > #else > static void __init input_apanel_init(void) {} > -- > 1.7.9.5 > I tested this patch on a Vostro V131, which is not on the whitelist, so all I got was the warning message, but to this extent, it works for me.
On Thursday 29 December 2016 09:29:36 Michał Kępień wrote: > > Dell platform team told us that some (DMI whitelisted) Dell > > Latitude machines have ST microelectronics accelerometer at i2c > > address 0x29. That i2c address is not specified in DMI or ACPI, so > > runtime detection without whitelist which is below is not > > possible. > > > > Presence of that ST microelectronics accelerometer is verified by > > existence of SMO88xx ACPI device which represent that > > accelerometer. Unfortunately without i2c address. > > This part of the commit message sounded a bit confusing to me at > first because there is already an ACPI driver which handles SMO88xx > devices (dell-smo8800). My understanding is that: > > * the purpose of this patch is to expose a richer interface (as > provided by lis3lv02d) to these devices on some machines, > > * on whitelisted machines, dell-smo8800 and lis3lv02d can work > simultaneously (even though dell-smo8800 effectively duplicates > the work that lis3lv02d does). No. dell-smo8800 reads from ACPI irq number and exports /dev/freefall device which notify userspace about falls. lis3lv02d is i2c driver which exports axes of accelerometer. Additionaly lis3lv02d can export also /dev/freefall if registerer of i2c device provides irq number -- which is not case of this patch. So both drivers are doing different things and both are useful. IIRC both dell-smo8800 and lis3lv02d represent one HW device (that ST microelectronics accelerometer) but due to complicated HW abstraction and layers on Dell laptops it is handled by two drivers, one ACPI and one i2c. Yes, in ideal world irq number should be passed to lis3lv02d driver and that would export whole device (with /dev/freefall too), but due to HW abstraction it is too much complicated... > If I got something wrong, please correct me. If I got it right, it > might make sense to rephrase the commit message a bit so that the > first bullet point above is immediately clear to the reader. > > > This patch registers lis3lv02d device at i2c address 0x29 if is > > detected. > > > > Finally commit a7ae81952cda ("i2c: i801: Allow ACPI SystemIO > > OpRegion to conflict with PCI BAR") allowed to use i2c-i801 driver > > on Dell machines so lis3lv02d correctly initialize accelerometer. > > > > Tested on Dell Latitude E6440. > > > > Signed-off-by: Pali Rohár <pali.rohar@gmail.com> > > --- > > > > drivers/i2c/busses/i2c-i801.c | 98 > > +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 98 > > insertions(+) > > > > diff --git a/drivers/i2c/busses/i2c-i801.c > > b/drivers/i2c/busses/i2c-i801.c index eb3627f..188cfd4 100644 > > --- a/drivers/i2c/busses/i2c-i801.c > > +++ b/drivers/i2c/busses/i2c-i801.c > > @@ -1118,6 +1118,101 @@ static void dmi_check_onboard_devices(const > > struct dmi_header *dm, void *adap) > > > > } > > > > } > > > > +static acpi_status check_acpi_smo88xx_device(acpi_handle > > obj_handle, + u32 nesting_level, > > + void *context, > > + void **return_value) > > +{ > > + struct acpi_device_info *info; > > + acpi_status status; > > + char *hid; > > + > > + status = acpi_get_object_info(obj_handle, &info); > > acpi_get_object_info() allocates the returned buffer, which the > caller has to free. Ok, I will fix it in next patch iteration. > > + if (!ACPI_SUCCESS(status) || !(info->valid & ACPI_VALID_HID)) > > + return AE_OK; > > + > > + hid = info->hardware_id.string; > > + if (!hid) > > + return AE_OK; > > + > > + if (strlen(hid) < 7) > > + return AE_OK; > > + > > + if (memcmp(hid, "SMO88", 5) != 0) > > + return AE_OK; > > + > > + *((bool *)return_value) = true; > > + return AE_CTRL_TERMINATE; > > +} > > + > > +static bool is_dell_system_with_lis3lv02d(void) > > +{ > > + bool found; > > + acpi_status status; > > + const char *vendor; > > + > > + vendor = dmi_get_system_info(DMI_SYS_VENDOR); > > + if (strcmp(vendor, "Dell Inc.") != 0) > > + return false; > > + > > + /* > > + * Check if ACPI device SMO88xx exists and if is enabled. That > > ACPI + * device represent our ST microelectronics lis3lv02d > > accelerometer but + * unfortunately without any other additional > > information. + */ > > + found = false; > > + status = acpi_get_devices(NULL, check_acpi_smo88xx_device, NULL, > > + (void **)&found); > > + if (!ACPI_SUCCESS(status) || !found) > > + return false; > > + > > + return true; > > +} > > + > > +/* > > + * Dell platform team told us that these Latitude devices have > > + * ST microelectronics accelerometer at i2c address 0x29. > > + * That i2c address is not specified in DMI or ACPI, so runtime > > + * detection without whitelist which is below is not possible. > > + */ > > +static const char * const dmi_dell_product_names[] = { > > + "Latitude E5250", > > + "Latitude E5450", > > + "Latitude E5550", > > + "Latitude E6440", > > + "Latitude E6440 ATG", > > + "Latitude E6540", > > +}; > > + > > +static void register_dell_lis3lv02d_i2c_device(struct i801_priv > > *priv) +{ > > + struct i2c_board_info info; > > + const char *product_name; > > + bool known_i2c_address; > > + int i; > > + > > + known_i2c_address = false; > > + product_name = dmi_get_system_info(DMI_PRODUCT_NAME); > > + for (i = 0; i < ARRAY_SIZE(dmi_dell_product_names); ++i) { > > + if (strcmp(product_name, dmi_dell_product_names[i]) == 0) { > > + known_i2c_address = true; > > + break; > > + } > > + } > > + > > + if (!known_i2c_address) { > > + dev_warn(&priv->pci_dev->dev, > > + "Accelerometer lis3lv02d i2c device is present " > > + "but its i2c address is unknown, skipping ...\n"); > > You are probably well aware of this, but checkpatch prefers keeping > long log messages in one line. I am pointing it out just in case. Yes, but I do not know how to fix it. Splitting message into two lines generates warning. Having long line generates warning too. > > + return; > > + } > > + > > + memset(&info, 0, sizeof(struct i2c_board_info)); > > How about just doing "struct i2c_board_info info = { 0 };" instead? Ok. > > + info.addr = 0x29; > > + strlcpy(info.type, "lis3lv02d", I2C_NAME_SIZE); > > + i2c_new_device(&priv->adapter, &info); > > +} > > + > > > > /* Register optional slaves */ > > static void i801_probe_optional_slaves(struct i801_priv *priv) > > { > > > > @@ -1136,6 +1231,9 @@ static void i801_probe_optional_slaves(struct > > i801_priv *priv) > > > > if (dmi_name_in_vendors("FUJITSU")) > > > > dmi_walk(dmi_check_onboard_devices, &priv->adapter); > > > > + > > + if (is_dell_system_with_lis3lv02d()) > > + register_dell_lis3lv02d_i2c_device(priv); > > > > } > > #else > > static void __init input_apanel_init(void) {} > > I tested this patch on a Vostro V131, which is not on the whitelist, > so all I got was the warning message, but to this extent, it works > for me. Hm... That means your notebook has ST microelectronics accelerometer too. You could try to find it on i2c-i801 bus with userspace i2cdetect program (part of i2c-tools) and get i2c address. If it will work we can extend DMI name --> i2c address mapping and include your notebook too. I have that list of confirmed Latitude devices since April 2015 but due to problem with ACPI resource conflicts it was not possible to load i2c- i801.ko bus driver. It was fixed only this year by commit a7ae81952cda. So list of devices is probably not up-to-date and new appeared.
> On Thursday 29 December 2016 09:29:36 Michał Kępień wrote: > > > Dell platform team told us that some (DMI whitelisted) Dell > > > Latitude machines have ST microelectronics accelerometer at i2c > > > address 0x29. That i2c address is not specified in DMI or ACPI, so > > > runtime detection without whitelist which is below is not > > > possible. > > > > > > Presence of that ST microelectronics accelerometer is verified by > > > existence of SMO88xx ACPI device which represent that > > > accelerometer. Unfortunately without i2c address. > > > > This part of the commit message sounded a bit confusing to me at > > first because there is already an ACPI driver which handles SMO88xx > > devices (dell-smo8800). My understanding is that: > > > > * the purpose of this patch is to expose a richer interface (as > > provided by lis3lv02d) to these devices on some machines, > > > > * on whitelisted machines, dell-smo8800 and lis3lv02d can work > > simultaneously (even though dell-smo8800 effectively duplicates > > the work that lis3lv02d does). > > No. dell-smo8800 reads from ACPI irq number and exports /dev/freefall > device which notify userspace about falls. lis3lv02d is i2c driver which > exports axes of accelerometer. Additionaly lis3lv02d can export also > /dev/freefall if registerer of i2c device provides irq number -- which > is not case of this patch. > > So both drivers are doing different things and both are useful. > > IIRC both dell-smo8800 and lis3lv02d represent one HW device (that ST > microelectronics accelerometer) but due to complicated HW abstraction > and layers on Dell laptops it is handled by two drivers, one ACPI and > one i2c. > > Yes, in ideal world irq number should be passed to lis3lv02d driver and > that would export whole device (with /dev/freefall too), but due to HW > abstraction it is too much complicated... Why? AFAICT, all that is required to pass that IRQ number all the way down to lis3lv02d is to set the irq field of the struct i2c_board_info you are passing to i2c_new_device(). And you can extract that IRQ number e.g. in check_acpi_smo88xx_device(). However, you would then need to make sure dell-smo8800 does not attempt to request the same IRQ on whitelisted machines. This got me thinking about a way to somehow incorporate your changes into dell-smo8800 using Wolfram's bus_notifier suggestion, but I do not have a working solution for now. What is tempting about this approach is that you would not have to scan the ACPI namespace in search of SMO88xx devices, because smo8800_add() is automatically called for them. However, I fear that the resulting solution may be more complicated than the one you submitted. > > > If I got something wrong, please correct me. If I got it right, it > > might make sense to rephrase the commit message a bit so that the > > first bullet point above is immediately clear to the reader. > > > > > This patch registers lis3lv02d device at i2c address 0x29 if is > > > detected. > > > > > > Finally commit a7ae81952cda ("i2c: i801: Allow ACPI SystemIO > > > OpRegion to conflict with PCI BAR") allowed to use i2c-i801 driver > > > on Dell machines so lis3lv02d correctly initialize accelerometer. > > > > > > Tested on Dell Latitude E6440. > > > > > > Signed-off-by: Pali Rohár <pali.rohar@gmail.com> > > > --- > > > > > > drivers/i2c/busses/i2c-i801.c | 98 > > > +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 98 > > > insertions(+) > > > > > > diff --git a/drivers/i2c/busses/i2c-i801.c > > > b/drivers/i2c/busses/i2c-i801.c index eb3627f..188cfd4 100644 > > > --- a/drivers/i2c/busses/i2c-i801.c > > > +++ b/drivers/i2c/busses/i2c-i801.c > > > @@ -1118,6 +1118,101 @@ static void dmi_check_onboard_devices(const > > > struct dmi_header *dm, void *adap) > > > > > > } > > > > > > } > > > > > > +static acpi_status check_acpi_smo88xx_device(acpi_handle > > > obj_handle, + u32 nesting_level, > > > + void *context, > > > + void **return_value) > > > +{ > > > + struct acpi_device_info *info; > > > + acpi_status status; > > > + char *hid; > > > + > > > + status = acpi_get_object_info(obj_handle, &info); > > > > acpi_get_object_info() allocates the returned buffer, which the > > caller has to free. > > Ok, I will fix it in next patch iteration. > > > > + if (!ACPI_SUCCESS(status) || !(info->valid & ACPI_VALID_HID)) > > > + return AE_OK; > > > + > > > + hid = info->hardware_id.string; > > > + if (!hid) > > > + return AE_OK; > > > + > > > + if (strlen(hid) < 7) > > > + return AE_OK; > > > + > > > + if (memcmp(hid, "SMO88", 5) != 0) > > > + return AE_OK; > > > + > > > + *((bool *)return_value) = true; > > > + return AE_CTRL_TERMINATE; > > > +} > > > + > > > +static bool is_dell_system_with_lis3lv02d(void) > > > +{ > > > + bool found; > > > + acpi_status status; > > > + const char *vendor; > > > + > > > + vendor = dmi_get_system_info(DMI_SYS_VENDOR); > > > + if (strcmp(vendor, "Dell Inc.") != 0) > > > + return false; > > > + > > > + /* > > > + * Check if ACPI device SMO88xx exists and if is enabled. That > > > ACPI + * device represent our ST microelectronics lis3lv02d > > > accelerometer but + * unfortunately without any other additional > > > information. + */ > > > + found = false; > > > + status = acpi_get_devices(NULL, check_acpi_smo88xx_device, NULL, > > > + (void **)&found); > > > + if (!ACPI_SUCCESS(status) || !found) > > > + return false; > > > + > > > + return true; > > > +} > > > + > > > +/* > > > + * Dell platform team told us that these Latitude devices have > > > + * ST microelectronics accelerometer at i2c address 0x29. > > > + * That i2c address is not specified in DMI or ACPI, so runtime > > > + * detection without whitelist which is below is not possible. > > > + */ > > > +static const char * const dmi_dell_product_names[] = { > > > + "Latitude E5250", > > > + "Latitude E5450", > > > + "Latitude E5550", > > > + "Latitude E6440", > > > + "Latitude E6440 ATG", > > > + "Latitude E6540", > > > +}; > > > + > > > +static void register_dell_lis3lv02d_i2c_device(struct i801_priv > > > *priv) +{ > > > + struct i2c_board_info info; > > > + const char *product_name; > > > + bool known_i2c_address; > > > + int i; > > > + > > > + known_i2c_address = false; > > > + product_name = dmi_get_system_info(DMI_PRODUCT_NAME); > > > + for (i = 0; i < ARRAY_SIZE(dmi_dell_product_names); ++i) { > > > + if (strcmp(product_name, dmi_dell_product_names[i]) == 0) { > > > + known_i2c_address = true; > > > + break; > > > + } > > > + } > > > + > > > + if (!known_i2c_address) { > > > + dev_warn(&priv->pci_dev->dev, > > > + "Accelerometer lis3lv02d i2c device is present " > > > + "but its i2c address is unknown, skipping ...\n"); > > > > You are probably well aware of this, but checkpatch prefers keeping > > long log messages in one line. I am pointing it out just in case. > > Yes, but I do not know how to fix it. Splitting message into two lines > generates warning. Having long line generates warning too. Weird, checkpatch does not protest on my machine when the log message is written on a single line... > > > > + return; > > > + } > > > + > > > + memset(&info, 0, sizeof(struct i2c_board_info)); > > > > How about just doing "struct i2c_board_info info = { 0 };" instead? > > Ok. > > > > + info.addr = 0x29; > > > + strlcpy(info.type, "lis3lv02d", I2C_NAME_SIZE); > > > + i2c_new_device(&priv->adapter, &info); > > > +} > > > + > > > > > > /* Register optional slaves */ > > > static void i801_probe_optional_slaves(struct i801_priv *priv) > > > { > > > > > > @@ -1136,6 +1231,9 @@ static void i801_probe_optional_slaves(struct > > > i801_priv *priv) > > > > > > if (dmi_name_in_vendors("FUJITSU")) > > > > > > dmi_walk(dmi_check_onboard_devices, &priv->adapter); > > > > > > + > > > + if (is_dell_system_with_lis3lv02d()) > > > + register_dell_lis3lv02d_i2c_device(priv); > > > > > > } > > > #else > > > static void __init input_apanel_init(void) {} > > > > I tested this patch on a Vostro V131, which is not on the whitelist, > > so all I got was the warning message, but to this extent, it works > > for me. > > Hm... That means your notebook has ST microelectronics accelerometer > too. You could try to find it on i2c-i801 bus with userspace i2cdetect > program (part of i2c-tools) and get i2c address. Bingo, it is at 0x1d. I modified your patch to set the i2c address to 0x1d and at least free fall detection seems to be working correctly.
On Thursday 29 December 2016 14:47:19 Michał Kępień wrote: > > On Thursday 29 December 2016 09:29:36 Michał Kępień wrote: > > > > Dell platform team told us that some (DMI whitelisted) Dell > > > > Latitude machines have ST microelectronics accelerometer at i2c > > > > address 0x29. That i2c address is not specified in DMI or ACPI, > > > > so runtime detection without whitelist which is below is not > > > > possible. > > > > > > > > Presence of that ST microelectronics accelerometer is verified > > > > by existence of SMO88xx ACPI device which represent that > > > > accelerometer. Unfortunately without i2c address. > > > > > > This part of the commit message sounded a bit confusing to me at > > > first because there is already an ACPI driver which handles > > > SMO88xx > > > > > > devices (dell-smo8800). My understanding is that: > > > * the purpose of this patch is to expose a richer interface (as > > > > > > provided by lis3lv02d) to these devices on some machines, > > > > > > * on whitelisted machines, dell-smo8800 and lis3lv02d can work > > > > > > simultaneously (even though dell-smo8800 effectively > > > duplicates the work that lis3lv02d does). > > > > No. dell-smo8800 reads from ACPI irq number and exports > > /dev/freefall device which notify userspace about falls. lis3lv02d > > is i2c driver which exports axes of accelerometer. Additionaly > > lis3lv02d can export also /dev/freefall if registerer of i2c > > device provides irq number -- which is not case of this patch. > > > > So both drivers are doing different things and both are useful. > > > > IIRC both dell-smo8800 and lis3lv02d represent one HW device (that > > ST microelectronics accelerometer) but due to complicated HW > > abstraction and layers on Dell laptops it is handled by two > > drivers, one ACPI and one i2c. > > > > Yes, in ideal world irq number should be passed to lis3lv02d driver > > and that would export whole device (with /dev/freefall too), but > > due to HW abstraction it is too much complicated... > > Why? AFAICT, all that is required to pass that IRQ number all the > way down to lis3lv02d is to set the irq field of the struct > i2c_board_info you are passing to i2c_new_device(). And you can > extract that IRQ number e.g. in check_acpi_smo88xx_device(). > However, you would then need to make sure dell-smo8800 does not > attempt to request the same IRQ on whitelisted machines. This got > me thinking about a way to somehow incorporate your changes into > dell-smo8800 using Wolfram's bus_notifier suggestion, but I do not > have a working solution for now. What is tempting about this > approach is that you would not have to scan the ACPI namespace in > search of SMO88xx devices, because smo8800_add() is automatically > called for them. However, I fear that the resulting solution may be > more complicated than the one you submitted. Then we need to deal with lot of problems. Order of loading .ko modules is undefined. Binding devices to drivers registered by .ko module is also in "random" order. At any time any of those .ko module can be unloaded or at least device unbind (via sysfs) from driver... And there can be some pathological situation (thanks to adding ACPI layer as Andy pointed) that there will be more SMO88xx devices in ACPI. Plus you can compile kernel with and without those modules and also you can blacklist loading them (so compile time check is not enough). And still some correct message notifier must be used. I think such solution is much much more complicated, there are lot of combinations of kernel configuration and available dell devices... > > > If I got something wrong, please correct me. If I got it right, > > > it might make sense to rephrase the commit message a bit so that > > > the first bullet point above is immediately clear to the reader. > > > > > > > This patch registers lis3lv02d device at i2c address 0x29 if is > > > > detected. > > > > > > > > Finally commit a7ae81952cda ("i2c: i801: Allow ACPI SystemIO > > > > OpRegion to conflict with PCI BAR") allowed to use i2c-i801 > > > > driver on Dell machines so lis3lv02d correctly initialize > > > > accelerometer. > > > > > > > > Tested on Dell Latitude E6440. > > > > > > > > Signed-off-by: Pali Rohár <pali.rohar@gmail.com> > > > > --- > > > > > > > > drivers/i2c/busses/i2c-i801.c | 98 > > > > +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 98 > > > > insertions(+) > > > > > > > > diff --git a/drivers/i2c/busses/i2c-i801.c > > > > b/drivers/i2c/busses/i2c-i801.c index eb3627f..188cfd4 100644 > > > > --- a/drivers/i2c/busses/i2c-i801.c > > > > +++ b/drivers/i2c/busses/i2c-i801.c > > > > @@ -1118,6 +1118,101 @@ static void > > > > dmi_check_onboard_devices(const struct dmi_header *dm, void > > > > *adap) > > > > > > > > } > > > > > > > > } > > > > > > > > +static acpi_status check_acpi_smo88xx_device(acpi_handle > > > > obj_handle, + u32 nesting_level, > > > > + void *context, > > > > + void **return_value) > > > > +{ > > > > + struct acpi_device_info *info; > > > > + acpi_status status; > > > > + char *hid; > > > > + > > > > + status = acpi_get_object_info(obj_handle, &info); > > > > > > acpi_get_object_info() allocates the returned buffer, which the > > > caller has to free. > > > > Ok, I will fix it in next patch iteration. > > > > > > + if (!ACPI_SUCCESS(status) || !(info->valid & ACPI_VALID_HID)) > > > > + return AE_OK; > > > > + > > > > + hid = info->hardware_id.string; > > > > + if (!hid) > > > > + return AE_OK; > > > > + > > > > + if (strlen(hid) < 7) > > > > + return AE_OK; > > > > + > > > > + if (memcmp(hid, "SMO88", 5) != 0) > > > > + return AE_OK; > > > > + > > > > + *((bool *)return_value) = true; > > > > + return AE_CTRL_TERMINATE; > > > > +} > > > > + > > > > +static bool is_dell_system_with_lis3lv02d(void) > > > > +{ > > > > + bool found; > > > > + acpi_status status; > > > > + const char *vendor; > > > > + > > > > + vendor = dmi_get_system_info(DMI_SYS_VENDOR); > > > > + if (strcmp(vendor, "Dell Inc.") != 0) > > > > + return false; > > > > + > > > > + /* > > > > + * Check if ACPI device SMO88xx exists and if is enabled. > > > > That ACPI + * device represent our ST microelectronics > > > > lis3lv02d accelerometer but + * unfortunately without any > > > > other additional information. + */ > > > > + found = false; > > > > + status = acpi_get_devices(NULL, check_acpi_smo88xx_device, > > > > NULL, + (void **)&found); > > > > + if (!ACPI_SUCCESS(status) || !found) > > > > + return false; > > > > + > > > > + return true; > > > > +} > > > > + > > > > +/* > > > > + * Dell platform team told us that these Latitude devices have > > > > + * ST microelectronics accelerometer at i2c address 0x29. > > > > + * That i2c address is not specified in DMI or ACPI, so > > > > runtime + * detection without whitelist which is below is not > > > > possible. + */ > > > > +static const char * const dmi_dell_product_names[] = { > > > > + "Latitude E5250", > > > > + "Latitude E5450", > > > > + "Latitude E5550", > > > > + "Latitude E6440", > > > > + "Latitude E6440 ATG", > > > > + "Latitude E6540", > > > > +}; > > > > + > > > > +static void register_dell_lis3lv02d_i2c_device(struct > > > > i801_priv *priv) +{ > > > > + struct i2c_board_info info; > > > > + const char *product_name; > > > > + bool known_i2c_address; > > > > + int i; > > > > + > > > > + known_i2c_address = false; > > > > + product_name = dmi_get_system_info(DMI_PRODUCT_NAME); > > > > + for (i = 0; i < ARRAY_SIZE(dmi_dell_product_names); ++i) { > > > > + if (strcmp(product_name, dmi_dell_product_names[i]) == 0) > > > > { > > > > + known_i2c_address = true; > > > > + break; > > > > + } > > > > + } > > > > + > > > > + if (!known_i2c_address) { > > > > + dev_warn(&priv->pci_dev->dev, > > > > + "Accelerometer lis3lv02d i2c device is present " > > > > + "but its i2c address is unknown, skipping ...\n"); > > > > > > You are probably well aware of this, but checkpatch prefers > > > keeping long log messages in one line. I am pointing it out > > > just in case. > > > > Yes, but I do not know how to fix it. Splitting message into two > > lines generates warning. Having long line generates warning too. > > Weird, checkpatch does not protest on my machine when the log message > is written on a single line... I hope that i2c maintainers decide how to format that line. > > > > + return; > > > > + } > > > > + > > > > + memset(&info, 0, sizeof(struct i2c_board_info)); > > > > > > How about just doing "struct i2c_board_info info = { 0 };" > > > instead? > > > > Ok. > > > > > > + info.addr = 0x29; > > > > + strlcpy(info.type, "lis3lv02d", I2C_NAME_SIZE); > > > > + i2c_new_device(&priv->adapter, &info); > > > > +} > > > > + > > > > > > > > /* Register optional slaves */ > > > > static void i801_probe_optional_slaves(struct i801_priv *priv) > > > > { > > > > > > > > @@ -1136,6 +1231,9 @@ static void > > > > i801_probe_optional_slaves(struct i801_priv *priv) > > > > > > > > if (dmi_name_in_vendors("FUJITSU")) > > > > > > > > dmi_walk(dmi_check_onboard_devices, &priv->adapter); > > > > > > > > + > > > > + if (is_dell_system_with_lis3lv02d()) > > > > + register_dell_lis3lv02d_i2c_device(priv); > > > > > > > > } > > > > #else > > > > static void __init input_apanel_init(void) {} > > > > > > I tested this patch on a Vostro V131, which is not on the > > > whitelist, so all I got was the warning message, but to this > > > extent, it works for me. > > > > Hm... That means your notebook has ST microelectronics > > accelerometer too. You could try to find it on i2c-i801 bus with > > userspace i2cdetect program (part of i2c-tools) and get i2c > > address. > > Bingo, it is at 0x1d. I modified your patch to set the i2c address > to 0x1d and at least free fall detection seems to be working > correctly. lis3lv02d exports input device, you should find its number lsinput. You can then test accelerometer with e.g. program input-events. If it is working fine, I can add your machine to whitelist with i2c address 0x1d.
> On Thursday 29 December 2016 14:47:19 Michał Kępień wrote: > > > On Thursday 29 December 2016 09:29:36 Michał Kępień wrote: > > > > > Dell platform team told us that some (DMI whitelisted) Dell > > > > > Latitude machines have ST microelectronics accelerometer at i2c > > > > > address 0x29. That i2c address is not specified in DMI or ACPI, > > > > > so runtime detection without whitelist which is below is not > > > > > possible. > > > > > > > > > > Presence of that ST microelectronics accelerometer is verified > > > > > by existence of SMO88xx ACPI device which represent that > > > > > accelerometer. Unfortunately without i2c address. > > > > > > > > This part of the commit message sounded a bit confusing to me at > > > > first because there is already an ACPI driver which handles > > > > SMO88xx > > > > > > > > devices (dell-smo8800). My understanding is that: > > > > * the purpose of this patch is to expose a richer interface (as > > > > > > > > provided by lis3lv02d) to these devices on some machines, > > > > > > > > * on whitelisted machines, dell-smo8800 and lis3lv02d can work > > > > > > > > simultaneously (even though dell-smo8800 effectively > > > > duplicates the work that lis3lv02d does). > > > > > > No. dell-smo8800 reads from ACPI irq number and exports > > > /dev/freefall device which notify userspace about falls. lis3lv02d > > > is i2c driver which exports axes of accelerometer. Additionaly > > > lis3lv02d can export also /dev/freefall if registerer of i2c > > > device provides irq number -- which is not case of this patch. > > > > > > So both drivers are doing different things and both are useful. > > > > > > IIRC both dell-smo8800 and lis3lv02d represent one HW device (that > > > ST microelectronics accelerometer) but due to complicated HW > > > abstraction and layers on Dell laptops it is handled by two > > > drivers, one ACPI and one i2c. > > > > > > Yes, in ideal world irq number should be passed to lis3lv02d driver > > > and that would export whole device (with /dev/freefall too), but > > > due to HW abstraction it is too much complicated... > > > > Why? AFAICT, all that is required to pass that IRQ number all the > > way down to lis3lv02d is to set the irq field of the struct > > i2c_board_info you are passing to i2c_new_device(). And you can > > extract that IRQ number e.g. in check_acpi_smo88xx_device(). > > However, you would then need to make sure dell-smo8800 does not > > attempt to request the same IRQ on whitelisted machines. This got > > me thinking about a way to somehow incorporate your changes into > > dell-smo8800 using Wolfram's bus_notifier suggestion, but I do not > > have a working solution for now. What is tempting about this > > approach is that you would not have to scan the ACPI namespace in > > search of SMO88xx devices, because smo8800_add() is automatically > > called for them. However, I fear that the resulting solution may be > > more complicated than the one you submitted. > > Then we need to deal with lot of problems. Order of loading .ko modules > is undefined. Binding devices to drivers registered by .ko module is > also in "random" order. At any time any of those .ko module can be > unloaded or at least device unbind (via sysfs) from driver... And there > can be some pathological situation (thanks to adding ACPI layer as Andy > pointed) that there will be more SMO88xx devices in ACPI. Plus you can > compile kernel with and without those modules and also you can blacklist > loading them (so compile time check is not enough). And still some > correct message notifier must be used. > > I think such solution is much much more complicated, there are lot of > combinations of kernel configuration and available dell devices... I tried a few more things, but ultimately failed to find a nice way to implement this. Another issue popped up, though. Linus' master branch contains a recent commit by Benjamin Tissoires (CC'ed), 4d5538f5882a ("i2c: use an IRQ to report Host Notify events, not alert") which breaks your patch. The reason for that is that lis3lv02d relies on the i2c client's IRQ being 0 to detect that it should not create /dev/freefall. Benjamin's patch causes the Host Notify IRQ to be assigned to the i2c client your patch creates, thus causing lis3lv02d to create /dev/freefall, which in turn conflicts with dell-smo8800 which is trying to create /dev/freefall itself. Also, just to make sure we do not overthink this, I understand that not every unit of the models from the whitelist has an accelerometer, correct? In other words, could we perhaps skip the part where we are making sure the SMO88xx ACPI device is there? > > > > > If I got something wrong, please correct me. If I got it right, > > > > it might make sense to rephrase the commit message a bit so that > > > > the first bullet point above is immediately clear to the reader. > > > > > > > > > This patch registers lis3lv02d device at i2c address 0x29 if is > > > > > detected. > > > > > > > > > > Finally commit a7ae81952cda ("i2c: i801: Allow ACPI SystemIO > > > > > OpRegion to conflict with PCI BAR") allowed to use i2c-i801 > > > > > driver on Dell machines so lis3lv02d correctly initialize > > > > > accelerometer. > > > > > > > > > > Tested on Dell Latitude E6440. > > > > > > > > > > Signed-off-by: Pali Rohár <pali.rohar@gmail.com> > > > > > --- > > > > > > > > > > drivers/i2c/busses/i2c-i801.c | 98 > > > > > +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 98 > > > > > insertions(+) > > > > > > > > > > diff --git a/drivers/i2c/busses/i2c-i801.c > > > > > b/drivers/i2c/busses/i2c-i801.c index eb3627f..188cfd4 100644 > > > > > --- a/drivers/i2c/busses/i2c-i801.c > > > > > +++ b/drivers/i2c/busses/i2c-i801.c > > > > > @@ -1118,6 +1118,101 @@ static void > > > > > dmi_check_onboard_devices(const struct dmi_header *dm, void > > > > > *adap) > > > > > > > > > > } > > > > > > > > > > } > > > > > > > > > > +static acpi_status check_acpi_smo88xx_device(acpi_handle > > > > > obj_handle, + u32 nesting_level, > > > > > + void *context, > > > > > + void **return_value) > > > > > +{ > > > > > + struct acpi_device_info *info; > > > > > + acpi_status status; > > > > > + char *hid; > > > > > + > > > > > + status = acpi_get_object_info(obj_handle, &info); > > > > > > > > acpi_get_object_info() allocates the returned buffer, which the > > > > caller has to free. > > > > > > Ok, I will fix it in next patch iteration. > > > > > > > > + if (!ACPI_SUCCESS(status) || !(info->valid & ACPI_VALID_HID)) > > > > > + return AE_OK; > > > > > + > > > > > + hid = info->hardware_id.string; > > > > > + if (!hid) > > > > > + return AE_OK; > > > > > + > > > > > + if (strlen(hid) < 7) > > > > > + return AE_OK; > > > > > + > > > > > + if (memcmp(hid, "SMO88", 5) != 0) > > > > > + return AE_OK; > > > > > + > > > > > + *((bool *)return_value) = true; > > > > > + return AE_CTRL_TERMINATE; > > > > > +} > > > > > + > > > > > +static bool is_dell_system_with_lis3lv02d(void) > > > > > +{ > > > > > + bool found; > > > > > + acpi_status status; > > > > > + const char *vendor; > > > > > + > > > > > + vendor = dmi_get_system_info(DMI_SYS_VENDOR); > > > > > + if (strcmp(vendor, "Dell Inc.") != 0) > > > > > + return false; > > > > > + > > > > > + /* > > > > > + * Check if ACPI device SMO88xx exists and if is enabled. > > > > > That ACPI + * device represent our ST microelectronics > > > > > lis3lv02d accelerometer but + * unfortunately without any > > > > > other additional information. + */ > > > > > + found = false; > > > > > + status = acpi_get_devices(NULL, check_acpi_smo88xx_device, > > > > > NULL, + (void **)&found); > > > > > + if (!ACPI_SUCCESS(status) || !found) > > > > > + return false; > > > > > + > > > > > + return true; > > > > > +} > > > > > + > > > > > +/* > > > > > + * Dell platform team told us that these Latitude devices have > > > > > + * ST microelectronics accelerometer at i2c address 0x29. > > > > > + * That i2c address is not specified in DMI or ACPI, so > > > > > runtime + * detection without whitelist which is below is not > > > > > possible. + */ > > > > > +static const char * const dmi_dell_product_names[] = { > > > > > + "Latitude E5250", > > > > > + "Latitude E5450", > > > > > + "Latitude E5550", > > > > > + "Latitude E6440", > > > > > + "Latitude E6440 ATG", > > > > > + "Latitude E6540", > > > > > +}; > > > > > + > > > > > +static void register_dell_lis3lv02d_i2c_device(struct > > > > > i801_priv *priv) +{ > > > > > + struct i2c_board_info info; > > > > > + const char *product_name; > > > > > + bool known_i2c_address; > > > > > + int i; > > > > > + > > > > > + known_i2c_address = false; > > > > > + product_name = dmi_get_system_info(DMI_PRODUCT_NAME); > > > > > + for (i = 0; i < ARRAY_SIZE(dmi_dell_product_names); ++i) { > > > > > + if (strcmp(product_name, dmi_dell_product_names[i]) == 0) > > > > > { > > > > > + known_i2c_address = true; > > > > > + break; > > > > > + } > > > > > + } > > > > > + > > > > > + if (!known_i2c_address) { > > > > > + dev_warn(&priv->pci_dev->dev, > > > > > + "Accelerometer lis3lv02d i2c device is present " > > > > > + "but its i2c address is unknown, skipping ...\n"); > > > > > > > > You are probably well aware of this, but checkpatch prefers > > > > keeping long log messages in one line. I am pointing it out > > > > just in case. > > > > > > Yes, but I do not know how to fix it. Splitting message into two > > > lines generates warning. Having long line generates warning too. > > > > Weird, checkpatch does not protest on my machine when the log message > > is written on a single line... > > I hope that i2c maintainers decide how to format that line. > > > > > > + return; > > > > > + } > > > > > + > > > > > + memset(&info, 0, sizeof(struct i2c_board_info)); > > > > > > > > How about just doing "struct i2c_board_info info = { 0 };" > > > > instead? > > > > > > Ok. > > > > > > > > + info.addr = 0x29; > > > > > + strlcpy(info.type, "lis3lv02d", I2C_NAME_SIZE); > > > > > + i2c_new_device(&priv->adapter, &info); > > > > > +} > > > > > + > > > > > > > > > > /* Register optional slaves */ > > > > > static void i801_probe_optional_slaves(struct i801_priv *priv) > > > > > { > > > > > > > > > > @@ -1136,6 +1231,9 @@ static void > > > > > i801_probe_optional_slaves(struct i801_priv *priv) > > > > > > > > > > if (dmi_name_in_vendors("FUJITSU")) > > > > > > > > > > dmi_walk(dmi_check_onboard_devices, &priv->adapter); > > > > > > > > > > + > > > > > + if (is_dell_system_with_lis3lv02d()) > > > > > + register_dell_lis3lv02d_i2c_device(priv); > > > > > > > > > > } > > > > > #else > > > > > static void __init input_apanel_init(void) {} > > > > > > > > I tested this patch on a Vostro V131, which is not on the > > > > whitelist, so all I got was the warning message, but to this > > > > extent, it works for me. > > > > > > Hm... That means your notebook has ST microelectronics > > > accelerometer too. You could try to find it on i2c-i801 bus with > > > userspace i2cdetect program (part of i2c-tools) and get i2c > > > address. > > > > Bingo, it is at 0x1d. I modified your patch to set the i2c address > > to 0x1d and at least free fall detection seems to be working > > correctly. > > lis3lv02d exports input device, you should find its number lsinput. You > can then test accelerometer with e.g. program input-events. > > If it is working fine, I can add your machine to whitelist with i2c > address 0x1d. I did some tests with evtest and it seems that axis values are consistent with laptop's movements, so I think it is safe to whitelist Vostro V131 with i2c address 0x1d.
On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > On Thursday 29 December 2016 14:47:19 Michał Kępień wrote: > > > > On Thursday 29 December 2016 09:29:36 Michał Kępień wrote: > > > > > > Dell platform team told us that some (DMI whitelisted) Dell > > > > > > Latitude machines have ST microelectronics accelerometer at > > > > > > i2c address 0x29. That i2c address is not specified in DMI > > > > > > or ACPI, so runtime detection without whitelist which is > > > > > > below is not possible. > > > > > > > > > > > > Presence of that ST microelectronics accelerometer is > > > > > > verified by existence of SMO88xx ACPI device which > > > > > > represent that accelerometer. Unfortunately without i2c > > > > > > address. > > > > > > > > > > This part of the commit message sounded a bit confusing to me > > > > > at first because there is already an ACPI driver which > > > > > handles SMO88xx > > > > > > > > > > devices (dell-smo8800). My understanding is that: > > > > > * the purpose of this patch is to expose a richer interface > > > > > (as > > > > > > > > > > provided by lis3lv02d) to these devices on some machines, > > > > > > > > > > * on whitelisted machines, dell-smo8800 and lis3lv02d can > > > > > work > > > > > > > > > > simultaneously (even though dell-smo8800 effectively > > > > > duplicates the work that lis3lv02d does). > > > > > > > > No. dell-smo8800 reads from ACPI irq number and exports > > > > /dev/freefall device which notify userspace about falls. > > > > lis3lv02d is i2c driver which exports axes of accelerometer. > > > > Additionaly lis3lv02d can export also /dev/freefall if > > > > registerer of i2c device provides irq number -- which is not > > > > case of this patch. > > > > > > > > So both drivers are doing different things and both are useful. > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > (that ST microelectronics accelerometer) but due to > > > > complicated HW abstraction and layers on Dell laptops it is > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > Yes, in ideal world irq number should be passed to lis3lv02d > > > > driver and that would export whole device (with /dev/freefall > > > > too), but due to HW abstraction it is too much complicated... > > > > > > Why? AFAICT, all that is required to pass that IRQ number all > > > the way down to lis3lv02d is to set the irq field of the struct > > > i2c_board_info you are passing to i2c_new_device(). And you can > > > extract that IRQ number e.g. in check_acpi_smo88xx_device(). > > > However, you would then need to make sure dell-smo8800 does not > > > attempt to request the same IRQ on whitelisted machines. This > > > got me thinking about a way to somehow incorporate your changes > > > into dell-smo8800 using Wolfram's bus_notifier suggestion, but I > > > do not have a working solution for now. What is tempting about > > > this approach is that you would not have to scan the ACPI > > > namespace in search of SMO88xx devices, because smo8800_add() is > > > automatically called for them. However, I fear that the > > > resulting solution may be more complicated than the one you > > > submitted. > > > > Then we need to deal with lot of problems. Order of loading .ko > > modules is undefined. Binding devices to drivers registered by .ko > > module is also in "random" order. At any time any of those .ko > > module can be unloaded or at least device unbind (via sysfs) from > > driver... And there can be some pathological situation (thanks to > > adding ACPI layer as Andy pointed) that there will be more SMO88xx > > devices in ACPI. Plus you can compile kernel with and without > > those modules and also you can blacklist loading them (so compile > > time check is not enough). And still some correct message notifier > > must be used. > > > > I think such solution is much much more complicated, there are lot > > of combinations of kernel configuration and available dell > > devices... > > I tried a few more things, but ultimately failed to find a nice way > to implement this. > > Another issue popped up, though. Linus' master branch contains a > recent commit by Benjamin Tissoires (CC'ed), 4d5538f5882a ("i2c: use > an IRQ to report Host Notify events, not alert") which breaks your > patch. The reason for that is that lis3lv02d relies on the i2c > client's IRQ being 0 to detect that it should not create > /dev/freefall. Benjamin's patch causes the Host Notify IRQ to be > assigned to the i2c client your patch creates, thus causing > lis3lv02d to create /dev/freefall, which in turn conflicts with > dell-smo8800 which is trying to create /dev/freefall itself. So 4d5538f5882a is breaking lis3lv02d driver... > Also, just to make sure we do not overthink this, I understand that > not every unit of the models from the whitelist has an > accelerometer, correct? In other words, could we perhaps skip the > part where we are making sure the SMO88xx ACPI device is there? Good question... At least for E6440 I'm did not thing it was possible to configure notebook without "3 axes free fall sensor". But! In BIOS SETUP it is possible to disable free fall sensor. I will try to disable it there and will check what happen. My guess is that it will be disabled in ACPI. > > > > > If I got something wrong, please correct me. If I got it > > > > > right, it might make sense to rephrase the commit message a > > > > > bit so that the first bullet point above is immediately > > > > > clear to the reader. > > > > > > > > > > > This patch registers lis3lv02d device at i2c address 0x29 > > > > > > if is detected. > > > > > > > > > > > > Finally commit a7ae81952cda ("i2c: i801: Allow ACPI > > > > > > SystemIO OpRegion to conflict with PCI BAR") allowed to > > > > > > use i2c-i801 driver on Dell machines so lis3lv02d > > > > > > correctly initialize accelerometer. > > > > > > > > > > > > Tested on Dell Latitude E6440. > > > > > > > > > > > > Signed-off-by: Pali Rohár <pali.rohar@gmail.com> > > > > > > --- > > > > > > > > > > > > drivers/i2c/busses/i2c-i801.c | 98 > > > > > > +++++++++++++++++++++++++++++++++++++++++ 1 file changed, > > > > > > 98 insertions(+) > > > > > > > > > > > > diff --git a/drivers/i2c/busses/i2c-i801.c > > > > > > b/drivers/i2c/busses/i2c-i801.c index eb3627f..188cfd4 > > > > > > 100644 --- a/drivers/i2c/busses/i2c-i801.c > > > > > > +++ b/drivers/i2c/busses/i2c-i801.c > > > > > > @@ -1118,6 +1118,101 @@ static void > > > > > > dmi_check_onboard_devices(const struct dmi_header *dm, void > > > > > > *adap) > > > > > > > > > > > > } > > > > > > > > > > > > } > > > > > > > > > > > > +static acpi_status check_acpi_smo88xx_device(acpi_handle > > > > > > obj_handle, + u32 nesting_level, > > > > > > + void *context, > > > > > > + void **return_value) > > > > > > +{ > > > > > > + struct acpi_device_info *info; > > > > > > + acpi_status status; > > > > > > + char *hid; > > > > > > + > > > > > > + status = acpi_get_object_info(obj_handle, &info); > > > > > > > > > > acpi_get_object_info() allocates the returned buffer, which > > > > > the caller has to free. > > > > > > > > Ok, I will fix it in next patch iteration. > > > > > > > > > > + if (!ACPI_SUCCESS(status) || !(info->valid & > > > > > > ACPI_VALID_HID)) + return AE_OK; > > > > > > + > > > > > > + hid = info->hardware_id.string; > > > > > > + if (!hid) > > > > > > + return AE_OK; > > > > > > + > > > > > > + if (strlen(hid) < 7) > > > > > > + return AE_OK; > > > > > > + > > > > > > + if (memcmp(hid, "SMO88", 5) != 0) > > > > > > + return AE_OK; > > > > > > + > > > > > > + *((bool *)return_value) = true; > > > > > > + return AE_CTRL_TERMINATE; > > > > > > +} > > > > > > + > > > > > > +static bool is_dell_system_with_lis3lv02d(void) > > > > > > +{ > > > > > > + bool found; > > > > > > + acpi_status status; > > > > > > + const char *vendor; > > > > > > + > > > > > > + vendor = dmi_get_system_info(DMI_SYS_VENDOR); > > > > > > + if (strcmp(vendor, "Dell Inc.") != 0) > > > > > > + return false; > > > > > > + > > > > > > + /* > > > > > > + * Check if ACPI device SMO88xx exists and if is enabled. > > > > > > That ACPI + * device represent our ST microelectronics > > > > > > lis3lv02d accelerometer but + * unfortunately without any > > > > > > other additional information. + */ > > > > > > + found = false; > > > > > > + status = acpi_get_devices(NULL, > > > > > > check_acpi_smo88xx_device, NULL, + (void > > > > > > **)&found); > > > > > > + if (!ACPI_SUCCESS(status) || !found) > > > > > > + return false; > > > > > > + > > > > > > + return true; > > > > > > +} > > > > > > + > > > > > > +/* > > > > > > + * Dell platform team told us that these Latitude devices > > > > > > have + * ST microelectronics accelerometer at i2c address > > > > > > 0x29. + * That i2c address is not specified in DMI or > > > > > > ACPI, so runtime + * detection without whitelist which is > > > > > > below is not possible. + */ > > > > > > +static const char * const dmi_dell_product_names[] = { > > > > > > + "Latitude E5250", > > > > > > + "Latitude E5450", > > > > > > + "Latitude E5550", > > > > > > + "Latitude E6440", > > > > > > + "Latitude E6440 ATG", > > > > > > + "Latitude E6540", > > > > > > +}; > > > > > > + > > > > > > +static void register_dell_lis3lv02d_i2c_device(struct > > > > > > i801_priv *priv) +{ > > > > > > + struct i2c_board_info info; > > > > > > + const char *product_name; > > > > > > + bool known_i2c_address; > > > > > > + int i; > > > > > > + > > > > > > + known_i2c_address = false; > > > > > > + product_name = dmi_get_system_info(DMI_PRODUCT_NAME); > > > > > > + for (i = 0; i < ARRAY_SIZE(dmi_dell_product_names); ++i) > > > > > > { + if (strcmp(product_name, dmi_dell_product_names[i]) > > > > > > == 0) { > > > > > > + known_i2c_address = true; > > > > > > + break; > > > > > > + } > > > > > > + } > > > > > > + > > > > > > + if (!known_i2c_address) { > > > > > > + dev_warn(&priv->pci_dev->dev, > > > > > > + "Accelerometer lis3lv02d i2c device is present " > > > > > > + "but its i2c address is unknown, skipping ... > > > > > > \n"); > > > > > > > > > > You are probably well aware of this, but checkpatch prefers > > > > > keeping long log messages in one line. I am pointing it out > > > > > just in case. > > > > > > > > Yes, but I do not know how to fix it. Splitting message into > > > > two lines generates warning. Having long line generates > > > > warning too. > > > > > > Weird, checkpatch does not protest on my machine when the log > > > message is written on a single line... > > > > I hope that i2c maintainers decide how to format that line. > > > > > > > > + return; > > > > > > + } > > > > > > + > > > > > > + memset(&info, 0, sizeof(struct i2c_board_info)); > > > > > > > > > > How about just doing "struct i2c_board_info info = { 0 };" > > > > > instead? > > > > > > > > Ok. > > > > > > > > > > + info.addr = 0x29; > > > > > > + strlcpy(info.type, "lis3lv02d", I2C_NAME_SIZE); > > > > > > + i2c_new_device(&priv->adapter, &info); > > > > > > +} > > > > > > + > > > > > > > > > > > > /* Register optional slaves */ > > > > > > static void i801_probe_optional_slaves(struct i801_priv > > > > > > *priv) { > > > > > > > > > > > > @@ -1136,6 +1231,9 @@ static void > > > > > > i801_probe_optional_slaves(struct i801_priv *priv) > > > > > > > > > > > > if (dmi_name_in_vendors("FUJITSU")) > > > > > > > > > > > > dmi_walk(dmi_check_onboard_devices, &priv->adapter); > > > > > > > > > > > > + > > > > > > + if (is_dell_system_with_lis3lv02d()) > > > > > > + register_dell_lis3lv02d_i2c_device(priv); > > > > > > > > > > > > } > > > > > > #else > > > > > > static void __init input_apanel_init(void) {} > > > > > > > > > > I tested this patch on a Vostro V131, which is not on the > > > > > whitelist, so all I got was the warning message, but to this > > > > > extent, it works for me. > > > > > > > > Hm... That means your notebook has ST microelectronics > > > > accelerometer too. You could try to find it on i2c-i801 bus > > > > with userspace i2cdetect program (part of i2c-tools) and get > > > > i2c address. > > > > > > Bingo, it is at 0x1d. I modified your patch to set the i2c > > > address to 0x1d and at least free fall detection seems to be > > > working correctly. > > > > lis3lv02d exports input device, you should find its number lsinput. > > You can then test accelerometer with e.g. program input-events. > > > > If it is working fine, I can add your machine to whitelist with i2c > > address 0x1d. > > I did some tests with evtest and it seems that axis values are > consistent with laptop's movements, so I think it is safe to > whitelist Vostro V131 with i2c address 0x1d. Ok.
On Dec 29 2016 or thereabouts, Pali Rohár wrote: > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > On Thursday 29 December 2016 14:47:19 Michał Kępień wrote: > > > > > On Thursday 29 December 2016 09:29:36 Michał Kępień wrote: > > > > > > > Dell platform team told us that some (DMI whitelisted) Dell > > > > > > > Latitude machines have ST microelectronics accelerometer at > > > > > > > i2c address 0x29. That i2c address is not specified in DMI > > > > > > > or ACPI, so runtime detection without whitelist which is > > > > > > > below is not possible. > > > > > > > > > > > > > > Presence of that ST microelectronics accelerometer is > > > > > > > verified by existence of SMO88xx ACPI device which > > > > > > > represent that accelerometer. Unfortunately without i2c > > > > > > > address. > > > > > > > > > > > > This part of the commit message sounded a bit confusing to me > > > > > > at first because there is already an ACPI driver which > > > > > > handles SMO88xx > > > > > > > > > > > > devices (dell-smo8800). My understanding is that: > > > > > > * the purpose of this patch is to expose a richer interface > > > > > > (as > > > > > > > > > > > > provided by lis3lv02d) to these devices on some machines, > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and lis3lv02d can > > > > > > work > > > > > > > > > > > > simultaneously (even though dell-smo8800 effectively > > > > > > duplicates the work that lis3lv02d does). > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and exports > > > > > /dev/freefall device which notify userspace about falls. > > > > > lis3lv02d is i2c driver which exports axes of accelerometer. > > > > > Additionaly lis3lv02d can export also /dev/freefall if > > > > > registerer of i2c device provides irq number -- which is not > > > > > case of this patch. > > > > > > > > > > So both drivers are doing different things and both are useful. > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > (that ST microelectronics accelerometer) but due to > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > Yes, in ideal world irq number should be passed to lis3lv02d > > > > > driver and that would export whole device (with /dev/freefall > > > > > too), but due to HW abstraction it is too much complicated... > > > > > > > > Why? AFAICT, all that is required to pass that IRQ number all > > > > the way down to lis3lv02d is to set the irq field of the struct > > > > i2c_board_info you are passing to i2c_new_device(). And you can > > > > extract that IRQ number e.g. in check_acpi_smo88xx_device(). > > > > However, you would then need to make sure dell-smo8800 does not > > > > attempt to request the same IRQ on whitelisted machines. This > > > > got me thinking about a way to somehow incorporate your changes > > > > into dell-smo8800 using Wolfram's bus_notifier suggestion, but I > > > > do not have a working solution for now. What is tempting about > > > > this approach is that you would not have to scan the ACPI > > > > namespace in search of SMO88xx devices, because smo8800_add() is > > > > automatically called for them. However, I fear that the > > > > resulting solution may be more complicated than the one you > > > > submitted. > > > > > > Then we need to deal with lot of problems. Order of loading .ko > > > modules is undefined. Binding devices to drivers registered by .ko > > > module is also in "random" order. At any time any of those .ko > > > module can be unloaded or at least device unbind (via sysfs) from > > > driver... And there can be some pathological situation (thanks to > > > adding ACPI layer as Andy pointed) that there will be more SMO88xx > > > devices in ACPI. Plus you can compile kernel with and without > > > those modules and also you can blacklist loading them (so compile > > > time check is not enough). And still some correct message notifier > > > must be used. > > > > > > I think such solution is much much more complicated, there are lot > > > of combinations of kernel configuration and available dell > > > devices... > > > > I tried a few more things, but ultimately failed to find a nice way > > to implement this. > > > > Another issue popped up, though. Linus' master branch contains a > > recent commit by Benjamin Tissoires (CC'ed), 4d5538f5882a ("i2c: use > > an IRQ to report Host Notify events, not alert") which breaks your > > patch. The reason for that is that lis3lv02d relies on the i2c > > client's IRQ being 0 to detect that it should not create > > /dev/freefall. Benjamin's patch causes the Host Notify IRQ to be > > assigned to the i2c client your patch creates, thus causing > > lis3lv02d to create /dev/freefall, which in turn conflicts with > > dell-smo8800 which is trying to create /dev/freefall itself. > > So 4d5538f5882a is breaking lis3lv02d driver... Apologies for that. I could easily fix this by adding a kernel API to know whether the provided irq is from Host Notify or if it was coming from an actual declaration. However, I have no idea how many other drivers would require this (hopefully only this one). One other solution would be to reserve the Host Notify IRQ and let the actual drivers that need it to set it, but this was not the best solution according to Dmitri. On my side, I am not entirely against this given that it's a chip feature, so the driver should be able to know that it's available. Dmitri, Wolfram, Jean, any preferences? > > > Also, just to make sure we do not overthink this, I understand that > > not every unit of the models from the whitelist has an > > accelerometer, correct? In other words, could we perhaps skip the > > part where we are making sure the SMO88xx ACPI device is there? > > Good question... At least for E6440 I'm did not thing it was possible to > configure notebook without "3 axes free fall sensor". > > But! In BIOS SETUP it is possible to disable free fall sensor. I will > try to disable it there and will check what happen. My guess is that it > will be disabled in ACPI. Just adding my 2 cents regarding the whitelist and interaction between those 2 drivers. I find this very fragile to have only one available /dev/freefall node and to rely on the fairness of each driver to not bind one. It would have been much simpler to have /dev/freefallXX and a proper misc class device for it. This way, we don't even need to mutually exclude the drivers. But this is already 8 years old code, so I guess userspace expects this... (why isn't that using the input subsystem at all?). Cheers, Benjamin. -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tuesday 03 January 2017 10:06:41 Benjamin Tissoires wrote: > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > Also, just to make sure we do not overthink this, I understand that > > > not every unit of the models from the whitelist has an > > > accelerometer, correct? In other words, could we perhaps skip the > > > part where we are making sure the SMO88xx ACPI device is there? > > > > Good question... At least for E6440 I'm did not thing it was possible to > > configure notebook without "3 axes free fall sensor". > > > > But! In BIOS SETUP it is possible to disable free fall sensor. I will > > try to disable it there and will check what happen. My guess is that it > > will be disabled in ACPI. > > Just adding my 2 cents regarding the whitelist and interaction between > those 2 drivers. I find this very fragile to have only one available > /dev/freefall node and to rely on the fairness of each driver to not bind > one. It would have been much simpler to have /dev/freefallXX and a > proper misc class device for it. This way, we don't even need to > mutually exclude the drivers. But this is already 8 years old code, so I > guess userspace expects this... (why isn't that using the input subsystem > at all?). > > Cheers, > Benjamin. > I think there is no problem with more /dev/freefall devices. With these Dell drivers it should not happen as only one driver can request IRQ which is associated with /dev/freefall. And /dev/freefal is registered after acquiring IRQ. But... there are other problems with it as wrote in previous emails.
On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires wrote: > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień wrote: > > > > > > On Thursday 29 December 2016 09:29:36 Michał Kępień wrote: > > > > > > > > Dell platform team told us that some (DMI whitelisted) Dell > > > > > > > > Latitude machines have ST microelectronics accelerometer at > > > > > > > > i2c address 0x29. That i2c address is not specified in DMI > > > > > > > > or ACPI, so runtime detection without whitelist which is > > > > > > > > below is not possible. > > > > > > > > > > > > > > > > Presence of that ST microelectronics accelerometer is > > > > > > > > verified by existence of SMO88xx ACPI device which > > > > > > > > represent that accelerometer. Unfortunately without i2c > > > > > > > > address. > > > > > > > > > > > > > > This part of the commit message sounded a bit confusing to me > > > > > > > at first because there is already an ACPI driver which > > > > > > > handles SMO88xx > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is that: > > > > > > > * the purpose of this patch is to expose a richer interface > > > > > > > (as > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on some machines, > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and lis3lv02d can > > > > > > > work > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 effectively > > > > > > > duplicates the work that lis3lv02d does). > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and exports > > > > > > /dev/freefall device which notify userspace about falls. > > > > > > lis3lv02d is i2c driver which exports axes of accelerometer. > > > > > > Additionaly lis3lv02d can export also /dev/freefall if > > > > > > registerer of i2c device provides irq number -- which is not > > > > > > case of this patch. > > > > > > > > > > > > So both drivers are doing different things and both are useful. > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > > (that ST microelectronics accelerometer) but due to > > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > Yes, in ideal world irq number should be passed to lis3lv02d > > > > > > driver and that would export whole device (with /dev/freefall > > > > > > too), but due to HW abstraction it is too much complicated... > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ number all > > > > > the way down to lis3lv02d is to set the irq field of the struct > > > > > i2c_board_info you are passing to i2c_new_device(). And you can > > > > > extract that IRQ number e.g. in check_acpi_smo88xx_device(). > > > > > However, you would then need to make sure dell-smo8800 does not > > > > > attempt to request the same IRQ on whitelisted machines. This > > > > > got me thinking about a way to somehow incorporate your changes > > > > > into dell-smo8800 using Wolfram's bus_notifier suggestion, but I > > > > > do not have a working solution for now. What is tempting about > > > > > this approach is that you would not have to scan the ACPI > > > > > namespace in search of SMO88xx devices, because smo8800_add() is > > > > > automatically called for them. However, I fear that the > > > > > resulting solution may be more complicated than the one you > > > > > submitted. > > > > > > > > Then we need to deal with lot of problems. Order of loading .ko > > > > modules is undefined. Binding devices to drivers registered by .ko > > > > module is also in "random" order. At any time any of those .ko > > > > module can be unloaded or at least device unbind (via sysfs) from > > > > driver... And there can be some pathological situation (thanks to > > > > adding ACPI layer as Andy pointed) that there will be more SMO88xx > > > > devices in ACPI. Plus you can compile kernel with and without > > > > those modules and also you can blacklist loading them (so compile > > > > time check is not enough). And still some correct message notifier > > > > must be used. > > > > > > > > I think such solution is much much more complicated, there are lot > > > > of combinations of kernel configuration and available dell > > > > devices... > > > > > > I tried a few more things, but ultimately failed to find a nice way > > > to implement this. > > > > > > Another issue popped up, though. Linus' master branch contains a > > > recent commit by Benjamin Tissoires (CC'ed), 4d5538f5882a ("i2c: use > > > an IRQ to report Host Notify events, not alert") which breaks your > > > patch. The reason for that is that lis3lv02d relies on the i2c > > > client's IRQ being 0 to detect that it should not create > > > /dev/freefall. Benjamin's patch causes the Host Notify IRQ to be > > > assigned to the i2c client your patch creates, thus causing > > > lis3lv02d to create /dev/freefall, which in turn conflicts with > > > dell-smo8800 which is trying to create /dev/freefall itself. > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > Apologies for that. > > I could easily fix this by adding a kernel API to know whether the > provided irq is from Host Notify or if it was coming from an actual > declaration. However, I have no idea how many other drivers would > require this (hopefully only this one). > > One other solution would be to reserve the Host Notify IRQ and let the > actual drivers that need it to set it, but this was not the best > solution according to Dmitri. On my side, I am not entirely against this > given that it's a chip feature, so the driver should be able to know > that it's available. > > Dmitri, Wolfram, Jean, any preferences? I read this: "IIRC both dell-smo8800 and lis3lv02d represent one HW device (that ST microelectronics accelerometer) but due to complicated HW abstraction and layers on Dell laptops it is handled by two drivers, one ACPI and one i2c." and that is the core of the issue. You have 2 drivers fighting over the same device. Fix this and it will all work. As far as I can see hp_accel instantiates lis3lv02d and accesses it via ACPI methods, can the same be done for Dell? > > > > > > Also, just to make sure we do not overthink this, I understand that > > > not every unit of the models from the whitelist has an > > > accelerometer, correct? In other words, could we perhaps skip the > > > part where we are making sure the SMO88xx ACPI device is there? > > > > Good question... At least for E6440 I'm did not thing it was possible to > > configure notebook without "3 axes free fall sensor". > > > > But! In BIOS SETUP it is possible to disable free fall sensor. I will > > try to disable it there and will check what happen. My guess is that it > > will be disabled in ACPI. > > Just adding my 2 cents regarding the whitelist and interaction between > those 2 drivers. I find this very fragile to have only one available > /dev/freefall node and to rely on the fairness of each driver to not bind > one. It would have been much simpler to have /dev/freefallXX and a > proper misc class device for it. This way, we don't even need to > mutually exclude the drivers. But this is already 8 years old code, so I > guess userspace expects this... (why isn't that using the input subsystem > at all?). I do not consider throwing a unit down an ordinary user interaction ;) So there is no input event code for this. Userspace should really use IIO accelerometer interface here. And kernel could provide composite IIO->/dev/freefall bridge, like we did for /dev/input/mice when all userspace wanted only PS/2. Thanks.
On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires wrote: > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień wrote: > > > > > > > On Thursday 29 December 2016 09:29:36 Michał Kępień wrote: > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > microelectronics accelerometer at i2c address 0x29. > > > > > > > > > That i2c address is not specified in DMI or ACPI, so > > > > > > > > > runtime detection without whitelist which is below > > > > > > > > > is not possible. > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics accelerometer is > > > > > > > > > verified by existence of SMO88xx ACPI device which > > > > > > > > > represent that accelerometer. Unfortunately without > > > > > > > > > i2c address. > > > > > > > > > > > > > > > > This part of the commit message sounded a bit confusing > > > > > > > > to me at first because there is already an ACPI driver > > > > > > > > which handles SMO88xx > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is that: > > > > > > > > * the purpose of this patch is to expose a richer > > > > > > > > interface (as > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on some > > > > > > > > machines, > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and lis3lv02d > > > > > > > > can work > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > effectively duplicates the work that lis3lv02d > > > > > > > > does). > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and exports > > > > > > > /dev/freefall device which notify userspace about falls. > > > > > > > lis3lv02d is i2c driver which exports axes of > > > > > > > accelerometer. Additionaly lis3lv02d can export also > > > > > > > /dev/freefall if registerer of i2c device provides irq > > > > > > > number -- which is not case of this patch. > > > > > > > > > > > > > > So both drivers are doing different things and both are > > > > > > > useful. > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent one HW > > > > > > > device (that ST microelectronics accelerometer) but due > > > > > > > to complicated HW abstraction and layers on Dell laptops > > > > > > > it is handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed to > > > > > > > lis3lv02d driver and that would export whole device > > > > > > > (with /dev/freefall too), but due to HW abstraction it > > > > > > > is too much complicated... > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ number > > > > > > all the way down to lis3lv02d is to set the irq field of > > > > > > the struct i2c_board_info you are passing to > > > > > > i2c_new_device(). And you can extract that IRQ number > > > > > > e.g. in check_acpi_smo88xx_device(). However, you would > > > > > > then need to make sure dell-smo8800 does not attempt to > > > > > > request the same IRQ on whitelisted machines. This got me > > > > > > thinking about a way to somehow incorporate your changes > > > > > > into dell-smo8800 using Wolfram's bus_notifier suggestion, > > > > > > but I do not have a working solution for now. What is > > > > > > tempting about this approach is that you would not have to > > > > > > scan the ACPI namespace in search of SMO88xx devices, > > > > > > because smo8800_add() is automatically called for them. > > > > > > However, I fear that the resulting solution may be more > > > > > > complicated than the one you submitted. > > > > > > > > > > Then we need to deal with lot of problems. Order of loading > > > > > .ko modules is undefined. Binding devices to drivers > > > > > registered by .ko module is also in "random" order. At any > > > > > time any of those .ko module can be unloaded or at least > > > > > device unbind (via sysfs) from driver... And there can be > > > > > some pathological situation (thanks to adding ACPI layer as > > > > > Andy pointed) that there will be more SMO88xx devices in > > > > > ACPI. Plus you can compile kernel with and without those > > > > > modules and also you can blacklist loading them (so compile > > > > > time check is not enough). And still some correct message > > > > > notifier must be used. > > > > > > > > > > I think such solution is much much more complicated, there > > > > > are lot of combinations of kernel configuration and > > > > > available dell devices... > > > > > > > > I tried a few more things, but ultimately failed to find a nice > > > > way to implement this. > > > > > > > > Another issue popped up, though. Linus' master branch contains > > > > a recent commit by Benjamin Tissoires (CC'ed), 4d5538f5882a > > > > ("i2c: use an IRQ to report Host Notify events, not alert") > > > > which breaks your patch. The reason for that is that > > > > lis3lv02d relies on the i2c client's IRQ being 0 to detect > > > > that it should not create /dev/freefall. Benjamin's patch > > > > causes the Host Notify IRQ to be assigned to the i2c client > > > > your patch creates, thus causing lis3lv02d to create > > > > /dev/freefall, which in turn conflicts with dell-smo8800 which > > > > is trying to create /dev/freefall itself. > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > Apologies for that. > > > > I could easily fix this by adding a kernel API to know whether the > > provided irq is from Host Notify or if it was coming from an actual > > declaration. However, I have no idea how many other drivers would > > require this (hopefully only this one). > > > > One other solution would be to reserve the Host Notify IRQ and let > > the actual drivers that need it to set it, but this was not the > > best solution according to Dmitri. On my side, I am not entirely > > against this given that it's a chip feature, so the driver should > > be able to know that it's available. > > > > Dmitri, Wolfram, Jean, any preferences? > > I read this: > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device (that > ST microelectronics accelerometer) but due to complicated HW > abstraction and layers on Dell laptops it is handled by two drivers, > one ACPI and one i2c." > > and that is the core of the issue. You have 2 drivers fighting over > the same device. Fix this and it will all work. With my current implementation (which I sent in this patch), they are not fighting. dell-smo8800 exports /dev/freefall (and nothing more) and lis3lv02d only accelerometer device as lis3lv02d driver does not get IRQ number in platform data. > As far as I can see hp_accel instantiates lis3lv02d and accesses it > via ACPI methods, can the same be done for Dell? No, Dell does not have any ACPI methods. And as I wrote in ACPI or DMI is even not i2c address of device, so it needs to be specified in code itself. Really there is no other way... :-( > > > > Also, just to make sure we do not overthink this, I understand > > > > that not every unit of the models from the whitelist has an > > > > accelerometer, correct? In other words, could we perhaps skip > > > > the part where we are making sure the SMO88xx ACPI device is > > > > there? > > > > > > Good question... At least for E6440 I'm did not thing it was > > > possible to configure notebook without "3 axes free fall > > > sensor". > > > > > > But! In BIOS SETUP it is possible to disable free fall sensor. I > > > will try to disable it there and will check what happen. My > > > guess is that it will be disabled in ACPI. > > > > Just adding my 2 cents regarding the whitelist and interaction > > between those 2 drivers. I find this very fragile to have only one > > available /dev/freefall node and to rely on the fairness of each > > driver to not bind one. It would have been much simpler to have > > /dev/freefallXX and a proper misc class device for it. This way, > > we don't even need to mutually exclude the drivers. But this is > > already 8 years old code, so I guess userspace expects this... > > (why isn't that using the input subsystem at all?). > > I do not consider throwing a unit down an ordinary user interaction > ;) So there is no input event code for this. > > Userspace should really use IIO accelerometer interface here. And > kernel could provide composite IIO->/dev/freefall bridge, like we Such "generic" bridge is probably not possible, as /dev/freefall is connected to specific lis3lv02d IRQ. > did for /dev/input/mice when all userspace wanted only PS/2. > > Thanks.
> -----Original Message----- > From: Pali Rohár [mailto:pali.rohar@gmail.com] > Sent: Tuesday, January 3, 2017 12:50 PM > To: Dmitry Torokhov <dmitry.torokhov@gmail.com> > Cc: Benjamin Tissoires <benjamin.tissoires@redhat.com>; Michał Kępień > <kernel@kempniu.pl>; Jean Delvare <jdelvare@suse.com>; Wolfram Sang > <wsa@the-dreams.de>; Steven Honeyman <stevenhoneyman@gmail.com>; > Valdis.Kletnieks@vt.edu; Jochen Eisinger <jochen@penguin-breeder.org>; > Gabriele Mazzotta <gabriele.mzt@gmail.com>; Andy Lutomirski > <luto@kernel.org>; Limonciello, Mario <Mario_Limonciello@Dell.com>; Alex > Hung <alex.hung@canonical.com>; Takashi Iwai <tiwai@suse.de>; linux- > i2c@vger.kernel.org; linux-kernel@vger.kernel.org; platform-driver- > x86@vger.kernel.org > Subject: Re: [PATCH] i2c: i801: Register optional lis3lv02d i2c device on Dell > machines > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires wrote: > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień wrote: > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał Kępień wrote: > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > microelectronics accelerometer at i2c address 0x29. > > > > > > > > > > That i2c address is not specified in DMI or ACPI, so > > > > > > > > > > runtime detection without whitelist which is below is > > > > > > > > > > not possible. > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics accelerometer is > > > > > > > > > > verified by existence of SMO88xx ACPI device which > > > > > > > > > > represent that accelerometer. Unfortunately without > > > > > > > > > > i2c address. > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit confusing > > > > > > > > > to me at first because there is already an ACPI driver > > > > > > > > > which handles SMO88xx > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is that: > > > > > > > > > * the purpose of this patch is to expose a richer > > > > > > > > > interface (as > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on some > > > > > > > > > machines, > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and lis3lv02d > > > > > > > > > can work > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > effectively duplicates the work that lis3lv02d > > > > > > > > > does). > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and exports > > > > > > > > /dev/freefall device which notify userspace about falls. > > > > > > > > lis3lv02d is i2c driver which exports axes of > > > > > > > > accelerometer. Additionaly lis3lv02d can export also > > > > > > > > /dev/freefall if registerer of i2c device provides irq > > > > > > > > number -- which is not case of this patch. > > > > > > > > > > > > > > > > So both drivers are doing different things and both are > > > > > > > > useful. > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent one HW > > > > > > > > device (that ST microelectronics accelerometer) but due to > > > > > > > > complicated HW abstraction and layers on Dell laptops it > > > > > > > > is handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed to > > > > > > > > lis3lv02d driver and that would export whole device (with > > > > > > > > /dev/freefall too), but due to HW abstraction it is too > > > > > > > > much complicated... > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ number > > > > > > > all the way down to lis3lv02d is to set the irq field of the > > > > > > > struct i2c_board_info you are passing to i2c_new_device(). > > > > > > > And you can extract that IRQ number e.g. in > > > > > > > check_acpi_smo88xx_device(). However, you would then need to > > > > > > > make sure dell-smo8800 does not attempt to request the same > > > > > > > IRQ on whitelisted machines. This got me thinking about a > > > > > > > way to somehow incorporate your changes into dell-smo8800 > > > > > > > using Wolfram's bus_notifier suggestion, but I do not have a > > > > > > > working solution for now. What is tempting about this > > > > > > > approach is that you would not have to scan the ACPI > > > > > > > namespace in search of SMO88xx devices, because > > > > > > > smo8800_add() is automatically called for them. > > > > > > > However, I fear that the resulting solution may be more > > > > > > > complicated than the one you submitted. > > > > > > > > > > > > Then we need to deal with lot of problems. Order of loading > > > > > > .ko modules is undefined. Binding devices to drivers > > > > > > registered by .ko module is also in "random" order. At any > > > > > > time any of those .ko module can be unloaded or at least > > > > > > device unbind (via sysfs) from driver... And there can be some > > > > > > pathological situation (thanks to adding ACPI layer as Andy > > > > > > pointed) that there will be more SMO88xx devices in ACPI. Plus > > > > > > you can compile kernel with and without those modules and also > > > > > > you can blacklist loading them (so compile time check is not > > > > > > enough). And still some correct message notifier must be used. > > > > > > > > > > > > I think such solution is much much more complicated, there are > > > > > > lot of combinations of kernel configuration and available dell > > > > > > devices... > > > > > > > > > > I tried a few more things, but ultimately failed to find a nice > > > > > way to implement this. > > > > > > > > > > Another issue popped up, though. Linus' master branch contains > > > > > a recent commit by Benjamin Tissoires (CC'ed), 4d5538f5882a > > > > > ("i2c: use an IRQ to report Host Notify events, not alert") > > > > > which breaks your patch. The reason for that is that lis3lv02d > > > > > relies on the i2c client's IRQ being 0 to detect that it should > > > > > not create /dev/freefall. Benjamin's patch causes the Host > > > > > Notify IRQ to be assigned to the i2c client your patch creates, > > > > > thus causing lis3lv02d to create /dev/freefall, which in turn > > > > > conflicts with dell-smo8800 which is trying to create > > > > > /dev/freefall itself. > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > Apologies for that. > > > > > > I could easily fix this by adding a kernel API to know whether the > > > provided irq is from Host Notify or if it was coming from an actual > > > declaration. However, I have no idea how many other drivers would > > > require this (hopefully only this one). > > > > > > One other solution would be to reserve the Host Notify IRQ and let > > > the actual drivers that need it to set it, but this was not the best > > > solution according to Dmitri. On my side, I am not entirely against > > > this given that it's a chip feature, so the driver should be able to > > > know that it's available. > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > I read this: > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device (that ST > > microelectronics accelerometer) but due to complicated HW abstraction > > and layers on Dell laptops it is handled by two drivers, one ACPI and > > one i2c." > > > > and that is the core of the issue. You have 2 drivers fighting over > > the same device. Fix this and it will all work. > > With my current implementation (which I sent in this patch), they are not > fighting. > > dell-smo8800 exports /dev/freefall (and nothing more) and lis3lv02d only > accelerometer device as lis3lv02d driver does not get IRQ number in platform > data. > > > As far as I can see hp_accel instantiates lis3lv02d and accesses it > > via ACPI methods, can the same be done for Dell? > > No, Dell does not have any ACPI methods. And as I wrote in ACPI or DMI is even > not i2c address of device, so it needs to be specified in code itself. > > Really there is no other way... :-( Dell doesn't export the general purpose accelerometer data up to the OS. Pali wanted it however and came up with a way to access it though from the information we have shared. That's the reason that there is no ACPI method for this and it needs to be whitelisted in this driver on the platforms that have it wired up this way. > > > > > > Also, just to make sure we do not overthink this, I understand > > > > > that not every unit of the models from the whitelist has an > > > > > accelerometer, correct? In other words, could we perhaps skip > > > > > the part where we are making sure the SMO88xx ACPI device is > > > > > there? > > > > > > > > Good question... At least for E6440 I'm did not thing it was > > > > possible to configure notebook without "3 axes free fall sensor". > > > > > > > > But! In BIOS SETUP it is possible to disable free fall sensor. I > > > > will try to disable it there and will check what happen. My guess > > > > is that it will be disabled in ACPI. > > > > > > Just adding my 2 cents regarding the whitelist and interaction > > > between those 2 drivers. I find this very fragile to have only one > > > available /dev/freefall node and to rely on the fairness of each > > > driver to not bind one. It would have been much simpler to have > > > /dev/freefallXX and a proper misc class device for it. This way, we > > > don't even need to mutually exclude the drivers. But this is already > > > 8 years old code, so I guess userspace expects this... > > > (why isn't that using the input subsystem at all?). > > > > I do not consider throwing a unit down an ordinary user interaction > > ;) So there is no input event code for this. > > > > Userspace should really use IIO accelerometer interface here. And > > kernel could provide composite IIO->/dev/freefall bridge, like we > > Such "generic" bridge is probably not possible, as /dev/freefall is connected to > specific lis3lv02d IRQ. > > > did for /dev/input/mice when all userspace wanted only PS/2. > > > > Thanks. > > -- > Pali Rohár > pali.rohar@gmail.com
On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires wrote: > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień wrote: > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał Kępień wrote: > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > microelectronics accelerometer at i2c address 0x29. > > > > > > > > > > That i2c address is not specified in DMI or ACPI, so > > > > > > > > > > runtime detection without whitelist which is below > > > > > > > > > > is not possible. > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics accelerometer is > > > > > > > > > > verified by existence of SMO88xx ACPI device which > > > > > > > > > > represent that accelerometer. Unfortunately without > > > > > > > > > > i2c address. > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit confusing > > > > > > > > > to me at first because there is already an ACPI driver > > > > > > > > > which handles SMO88xx > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is that: > > > > > > > > > * the purpose of this patch is to expose a richer > > > > > > > > > interface (as > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on some > > > > > > > > > machines, > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and lis3lv02d > > > > > > > > > can work > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > effectively duplicates the work that lis3lv02d > > > > > > > > > does). > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and exports > > > > > > > > /dev/freefall device which notify userspace about falls. > > > > > > > > lis3lv02d is i2c driver which exports axes of > > > > > > > > accelerometer. Additionaly lis3lv02d can export also > > > > > > > > /dev/freefall if registerer of i2c device provides irq > > > > > > > > number -- which is not case of this patch. > > > > > > > > > > > > > > > > So both drivers are doing different things and both are > > > > > > > > useful. > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent one HW > > > > > > > > device (that ST microelectronics accelerometer) but due > > > > > > > > to complicated HW abstraction and layers on Dell laptops > > > > > > > > it is handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed to > > > > > > > > lis3lv02d driver and that would export whole device > > > > > > > > (with /dev/freefall too), but due to HW abstraction it > > > > > > > > is too much complicated... > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ number > > > > > > > all the way down to lis3lv02d is to set the irq field of > > > > > > > the struct i2c_board_info you are passing to > > > > > > > i2c_new_device(). And you can extract that IRQ number > > > > > > > e.g. in check_acpi_smo88xx_device(). However, you would > > > > > > > then need to make sure dell-smo8800 does not attempt to > > > > > > > request the same IRQ on whitelisted machines. This got me > > > > > > > thinking about a way to somehow incorporate your changes > > > > > > > into dell-smo8800 using Wolfram's bus_notifier suggestion, > > > > > > > but I do not have a working solution for now. What is > > > > > > > tempting about this approach is that you would not have to > > > > > > > scan the ACPI namespace in search of SMO88xx devices, > > > > > > > because smo8800_add() is automatically called for them. > > > > > > > However, I fear that the resulting solution may be more > > > > > > > complicated than the one you submitted. > > > > > > > > > > > > Then we need to deal with lot of problems. Order of loading > > > > > > .ko modules is undefined. Binding devices to drivers > > > > > > registered by .ko module is also in "random" order. At any > > > > > > time any of those .ko module can be unloaded or at least > > > > > > device unbind (via sysfs) from driver... And there can be > > > > > > some pathological situation (thanks to adding ACPI layer as > > > > > > Andy pointed) that there will be more SMO88xx devices in > > > > > > ACPI. Plus you can compile kernel with and without those > > > > > > modules and also you can blacklist loading them (so compile > > > > > > time check is not enough). And still some correct message > > > > > > notifier must be used. > > > > > > > > > > > > I think such solution is much much more complicated, there > > > > > > are lot of combinations of kernel configuration and > > > > > > available dell devices... > > > > > > > > > > I tried a few more things, but ultimately failed to find a nice > > > > > way to implement this. > > > > > > > > > > Another issue popped up, though. Linus' master branch contains > > > > > a recent commit by Benjamin Tissoires (CC'ed), 4d5538f5882a > > > > > ("i2c: use an IRQ to report Host Notify events, not alert") > > > > > which breaks your patch. The reason for that is that > > > > > lis3lv02d relies on the i2c client's IRQ being 0 to detect > > > > > that it should not create /dev/freefall. Benjamin's patch > > > > > causes the Host Notify IRQ to be assigned to the i2c client > > > > > your patch creates, thus causing lis3lv02d to create > > > > > /dev/freefall, which in turn conflicts with dell-smo8800 which > > > > > is trying to create /dev/freefall itself. > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > Apologies for that. > > > > > > I could easily fix this by adding a kernel API to know whether the > > > provided irq is from Host Notify or if it was coming from an actual > > > declaration. However, I have no idea how many other drivers would > > > require this (hopefully only this one). > > > > > > One other solution would be to reserve the Host Notify IRQ and let > > > the actual drivers that need it to set it, but this was not the > > > best solution according to Dmitri. On my side, I am not entirely > > > against this given that it's a chip feature, so the driver should > > > be able to know that it's available. > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > I read this: > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device (that > > ST microelectronics accelerometer) but due to complicated HW > > abstraction and layers on Dell laptops it is handled by two drivers, > > one ACPI and one i2c." > > > > and that is the core of the issue. You have 2 drivers fighting over > > the same device. Fix this and it will all work. > > With my current implementation (which I sent in this patch), they are > not fighting. > > dell-smo8800 exports /dev/freefall (and nothing more) and lis3lv02d only > accelerometer device as lis3lv02d driver does not get IRQ number in > platform data. > > > As far as I can see hp_accel instantiates lis3lv02d and accesses it > > via ACPI methods, can the same be done for Dell? > > No, Dell does not have any ACPI methods. And as I wrote in ACPI or DMI > is even not i2c address of device, so it needs to be specified in code > itself. > > Really there is no other way... :-( Sure there is: 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". 2. dell-smo8800 provides read/write functions for lis3lv02d that simply forward requests to dell-smo8800-accel i2c client. 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel does. Alternatively, can lis3lv02d be tasked to create /dev/freefall? Yet another option: can we add a new flag to i2c_board_info controlling whether we want to enable/disable wiring up host notify interrupt? Benjamin, is there anything "special" in RMI SMbus ACPI descriptors we could use? Thanks.
On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > wrote: > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień wrote: > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał Kępień > > > > > > > > > wrote: > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > 0x29. That i2c address is not specified in DMI > > > > > > > > > > > or ACPI, so runtime detection without whitelist > > > > > > > > > > > which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > confusing to me at first because there is already > > > > > > > > > > an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is that: > > > > > > > > > > * the purpose of this patch is to expose a richer > > > > > > > > > > interface (as > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on some > > > > > > > > > > machines, > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > effectively duplicates the work that lis3lv02d > > > > > > > > > > does). > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > exports /dev/freefall device which notify userspace > > > > > > > > > about falls. lis3lv02d is i2c driver which exports > > > > > > > > > axes of accelerometer. Additionaly lis3lv02d can > > > > > > > > > export also /dev/freefall if registerer of i2c > > > > > > > > > device provides irq number -- which is not case of > > > > > > > > > this patch. > > > > > > > > > > > > > > > > > > So both drivers are doing different things and both > > > > > > > > > are useful. > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent one HW > > > > > > > > > device (that ST microelectronics accelerometer) but > > > > > > > > > due to complicated HW abstraction and layers on Dell > > > > > > > > > laptops it is handled by two drivers, one ACPI and > > > > > > > > > one i2c. > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed to > > > > > > > > > lis3lv02d driver and that would export whole device > > > > > > > > > (with /dev/freefall too), but due to HW abstraction > > > > > > > > > it is too much complicated... > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > number all the way down to lis3lv02d is to set the irq > > > > > > > > field of the struct i2c_board_info you are passing to > > > > > > > > i2c_new_device(). And you can extract that IRQ number > > > > > > > > e.g. in check_acpi_smo88xx_device(). However, you would > > > > > > > > then need to make sure dell-smo8800 does not attempt to > > > > > > > > request the same IRQ on whitelisted machines. This got > > > > > > > > me thinking about a way to somehow incorporate your > > > > > > > > changes into dell-smo8800 using Wolfram's bus_notifier > > > > > > > > suggestion, but I do not have a working solution for > > > > > > > > now. What is tempting about this approach is that you > > > > > > > > would not have to scan the ACPI namespace in search of > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > automatically called for them. However, I fear that > > > > > > > > the resulting solution may be more complicated than > > > > > > > > the one you submitted. > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > order. At any time any of those .ko module can be > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > driver... And there can be some pathological situation > > > > > > > (thanks to adding ACPI layer as Andy pointed) that there > > > > > > > will be more SMO88xx devices in ACPI. Plus you can > > > > > > > compile kernel with and without those modules and also > > > > > > > you can blacklist loading them (so compile time check is > > > > > > > not enough). And still some correct message notifier > > > > > > > must be used. > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > there are lot of combinations of kernel configuration > > > > > > > and available dell devices... > > > > > > > > > > > > I tried a few more things, but ultimately failed to find a > > > > > > nice way to implement this. > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > events, not alert") which breaks your patch. The reason > > > > > > for that is that lis3lv02d relies on the i2c client's IRQ > > > > > > being 0 to detect that it should not create /dev/freefall. > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > assigned to the i2c client your patch creates, thus > > > > > > causing lis3lv02d to create /dev/freefall, which in turn > > > > > > conflicts with dell-smo8800 which is trying to create > > > > > > /dev/freefall itself. > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > Apologies for that. > > > > > > > > I could easily fix this by adding a kernel API to know whether > > > > the provided irq is from Host Notify or if it was coming from > > > > an actual declaration. However, I have no idea how many other > > > > drivers would require this (hopefully only this one). > > > > > > > > One other solution would be to reserve the Host Notify IRQ and > > > > let the actual drivers that need it to set it, but this was > > > > not the best solution according to Dmitri. On my side, I am > > > > not entirely against this given that it's a chip feature, so > > > > the driver should be able to know that it's available. > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > I read this: > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > (that ST microelectronics accelerometer) but due to complicated > > > HW abstraction and layers on Dell laptops it is handled by two > > > drivers, one ACPI and one i2c." > > > > > > and that is the core of the issue. You have 2 drivers fighting > > > over the same device. Fix this and it will all work. > > > > With my current implementation (which I sent in this patch), they > > are not fighting. > > > > dell-smo8800 exports /dev/freefall (and nothing more) and lis3lv02d > > only accelerometer device as lis3lv02d driver does not get IRQ > > number in platform data. > > > > > As far as I can see hp_accel instantiates lis3lv02d and accesses > > > it via ACPI methods, can the same be done for Dell? > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI or > > DMI is even not i2c address of device, so it needs to be specified > > in code itself. > > > > Really there is no other way... :-( > > Sure there is: > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > 2. dell-smo8800 provides read/write functions for lis3lv02d that > simply forward requests to dell-smo8800-accel i2c client. > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel does. Sorry, but I do not understand how you mean it... Why to provides new read/write i2c functions which are already implemented by i2c-i801 bus and lis3lv02d i2c driver? > Alternatively, can lis3lv02d be tasked to create /dev/freefall? If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall device. But... what is problem with current implementation? Accelerometer HW provides two functions: 1) 3 axes reports 2) Disk freefall detection And 1) is handled by i2c driver lis3lv02d and 2) is by dell-smo8800. Both functions are independent here. I think you just trying to complicate this situation even more to be more complicated as currently is. > Yet another option: can we add a new flag to i2c_board_info > controlling whether we want to enable/disable wiring up host notify > interrupt? Benjamin, is there anything "special" in RMI SMbus ACPI > descriptors we could use? > > Thanks.
On Tue, Jan 3, 2017 at 8:50 PM, Pali Rohár <pali.rohar@gmail.com> wrote: > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: >> "IIRC both dell-smo8800 and lis3lv02d represent one HW device (that >> ST microelectronics accelerometer) but due to complicated HW >> abstraction and layers on Dell laptops it is handled by two drivers, >> one ACPI and one i2c." >> >> and that is the core of the issue. You have 2 drivers fighting over >> the same device. Fix this and it will all work. > > With my current implementation (which I sent in this patch), they are > not fighting. > > dell-smo8800 exports /dev/freefall (and nothing more) and lis3lv02d only > accelerometer device as lis3lv02d driver does not get IRQ number in > platform data. > >> As far as I can see hp_accel instantiates lis3lv02d and accesses it >> via ACPI methods, can the same be done for Dell? > > No, Dell does not have any ACPI methods. > And as I wrote in ACPI or DMI > is even not i2c address of device, so it needs to be specified in code > itself. And as I wrote there is still a way to provide it without hardcoding on model basis. > Really there is no other way... :-(
On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > wrote: > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień wrote: > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał Kępień > > > > > > > > > > wrote: > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > 0x29. That i2c address is not specified in DMI > > > > > > > > > > > > or ACPI, so runtime detection without whitelist > > > > > > > > > > > > which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > confusing to me at first because there is already > > > > > > > > > > > an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is that: > > > > > > > > > > > * the purpose of this patch is to expose a richer > > > > > > > > > > > interface (as > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on some > > > > > > > > > > > machines, > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > effectively duplicates the work that lis3lv02d > > > > > > > > > > > does). > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > exports /dev/freefall device which notify userspace > > > > > > > > > > about falls. lis3lv02d is i2c driver which exports > > > > > > > > > > axes of accelerometer. Additionaly lis3lv02d can > > > > > > > > > > export also /dev/freefall if registerer of i2c > > > > > > > > > > device provides irq number -- which is not case of > > > > > > > > > > this patch. > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and both > > > > > > > > > > are useful. > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent one HW > > > > > > > > > > device (that ST microelectronics accelerometer) but > > > > > > > > > > due to complicated HW abstraction and layers on Dell > > > > > > > > > > laptops it is handled by two drivers, one ACPI and > > > > > > > > > > one i2c. > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed to > > > > > > > > > > lis3lv02d driver and that would export whole device > > > > > > > > > > (with /dev/freefall too), but due to HW abstraction > > > > > > > > > > it is too much complicated... > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > number all the way down to lis3lv02d is to set the irq > > > > > > > > > field of the struct i2c_board_info you are passing to > > > > > > > > > i2c_new_device(). And you can extract that IRQ number > > > > > > > > > e.g. in check_acpi_smo88xx_device(). However, you would > > > > > > > > > then need to make sure dell-smo8800 does not attempt to > > > > > > > > > request the same IRQ on whitelisted machines. This got > > > > > > > > > me thinking about a way to somehow incorporate your > > > > > > > > > changes into dell-smo8800 using Wolfram's bus_notifier > > > > > > > > > suggestion, but I do not have a working solution for > > > > > > > > > now. What is tempting about this approach is that you > > > > > > > > > would not have to scan the ACPI namespace in search of > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > driver... And there can be some pathological situation > > > > > > > > (thanks to adding ACPI layer as Andy pointed) that there > > > > > > > > will be more SMO88xx devices in ACPI. Plus you can > > > > > > > > compile kernel with and without those modules and also > > > > > > > > you can blacklist loading them (so compile time check is > > > > > > > > not enough). And still some correct message notifier > > > > > > > > must be used. > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > and available dell devices... > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to find a > > > > > > > nice way to implement this. > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > events, not alert") which breaks your patch. The reason > > > > > > > for that is that lis3lv02d relies on the i2c client's IRQ > > > > > > > being 0 to detect that it should not create /dev/freefall. > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > causing lis3lv02d to create /dev/freefall, which in turn > > > > > > > conflicts with dell-smo8800 which is trying to create > > > > > > > /dev/freefall itself. > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > Apologies for that. > > > > > > > > > > I could easily fix this by adding a kernel API to know whether > > > > > the provided irq is from Host Notify or if it was coming from > > > > > an actual declaration. However, I have no idea how many other > > > > > drivers would require this (hopefully only this one). > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ and > > > > > let the actual drivers that need it to set it, but this was > > > > > not the best solution according to Dmitri. On my side, I am > > > > > not entirely against this given that it's a chip feature, so > > > > > the driver should be able to know that it's available. > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > I read this: > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > (that ST microelectronics accelerometer) but due to complicated > > > > HW abstraction and layers on Dell laptops it is handled by two > > > > drivers, one ACPI and one i2c." > > > > > > > > and that is the core of the issue. You have 2 drivers fighting > > > > over the same device. Fix this and it will all work. > > > > > > With my current implementation (which I sent in this patch), they > > > are not fighting. > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and lis3lv02d > > > only accelerometer device as lis3lv02d driver does not get IRQ > > > number in platform data. > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and accesses > > > > it via ACPI methods, can the same be done for Dell? > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI or > > > DMI is even not i2c address of device, so it needs to be specified > > > in code itself. > > > > > > Really there is no other way... :-( > > > > Sure there is: > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > simply forward requests to dell-smo8800-accel i2c client. > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel does. > > Sorry, but I do not understand how you mean it... Why to provides new > read/write i2c functions which are already implemented by i2c-i801 bus > and lis3lv02d i2c driver? Because that would allow you to avoid clashes with i2c creating interrupt mapping for client residing on host-notify-capable controller. > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > device. > > But... what is problem with current implementation? Accelerometer HW > provides two functions: > > 1) 3 axes reports > 2) Disk freefall detection > > And 1) is handled by i2c driver lis3lv02d and 2) is by dell-smo8800. > Both functions are independent here. > > I think you just trying to complicate this situation even more to be > more complicated as currently is. Because this apparently does not work for you, does it? In general, if you want the same hardware be handled by 2 different drivers you are going to have bad time. It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are the same, right? So, instead of having 2 drivers split the functionality, can you forego registering smo8800 ACPI driver on your whitelisted boxes and instead instantiate full i2c client device with properly assigned both address and IRQ and let lis3lv02d handle it (providing both accelerometer data and /dev/freefall)? Thanks.
On Tuesday 03 January 2017 21:24:18 Dmitry Torokhov wrote: > On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > > > > > > > wrote: > > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień > > > > > > > > > wrote: > > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał > > > > > > > > > > > Kępień > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > > 0x29. That i2c address is not specified in > > > > > > > > > > > > > DMI or ACPI, so runtime detection without > > > > > > > > > > > > > whitelist which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > > confusing to me at first because there is > > > > > > > > > > > > already an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is > > > > > > > > > > > > that: > > > > > > > > > > > > * the purpose of this patch is to expose a > > > > > > > > > > > > richer interface (as > > > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on > > > > > > > > > > > > some machines, > > > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > > effectively duplicates the work that > > > > > > > > > > > > lis3lv02d does). > > > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > > exports /dev/freefall device which notify > > > > > > > > > > > userspace about falls. lis3lv02d is i2c driver > > > > > > > > > > > which exports axes of accelerometer. Additionaly > > > > > > > > > > > lis3lv02d can export also /dev/freefall if > > > > > > > > > > > registerer of i2c device provides irq number -- > > > > > > > > > > > which is not case of this patch. > > > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and > > > > > > > > > > > both are useful. > > > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent > > > > > > > > > > > one HW device (that ST microelectronics > > > > > > > > > > > accelerometer) but due to complicated HW > > > > > > > > > > > abstraction and layers on Dell laptops it is > > > > > > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed > > > > > > > > > > > to lis3lv02d driver and that would export whole > > > > > > > > > > > device (with /dev/freefall too), but due to HW > > > > > > > > > > > abstraction it is too much complicated... > > > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > > number all the way down to lis3lv02d is to set the > > > > > > > > > > irq field of the struct i2c_board_info you are > > > > > > > > > > passing to i2c_new_device(). And you can extract > > > > > > > > > > that IRQ number e.g. in > > > > > > > > > > check_acpi_smo88xx_device(). However, you would > > > > > > > > > > then need to make sure dell-smo8800 does not > > > > > > > > > > attempt to request the same IRQ on whitelisted > > > > > > > > > > machines. This got me thinking about a way to > > > > > > > > > > somehow incorporate your changes into dell-smo8800 > > > > > > > > > > using Wolfram's bus_notifier suggestion, but I do > > > > > > > > > > not have a working solution for now. What is > > > > > > > > > > tempting about this approach is that you would not > > > > > > > > > > have to scan the ACPI namespace in search of > > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > > driver... And there can be some pathological > > > > > > > > > situation (thanks to adding ACPI layer as Andy > > > > > > > > > pointed) that there will be more SMO88xx devices in > > > > > > > > > ACPI. Plus you can compile kernel with and without > > > > > > > > > those modules and also you can blacklist loading > > > > > > > > > them (so compile time check is not enough). And > > > > > > > > > still some correct message notifier must be used. > > > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > > and available dell devices... > > > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to > > > > > > > > find a nice way to implement this. > > > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > > events, not alert") which breaks your patch. The > > > > > > > > reason for that is that lis3lv02d relies on the i2c > > > > > > > > client's IRQ being 0 to detect that it should not > > > > > > > > create /dev/freefall. > > > > > > > > > > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > > > > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > > causing lis3lv02d to create /dev/freefall, which in > > > > > > > > turn conflicts with dell-smo8800 which is trying to > > > > > > > > create /dev/freefall itself. > > > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > > > Apologies for that. > > > > > > > > > > > > I could easily fix this by adding a kernel API to know > > > > > > whether the provided irq is from Host Notify or if it was > > > > > > coming from an actual declaration. However, I have no idea > > > > > > how many other drivers would require this (hopefully only > > > > > > this one). > > > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ > > > > > > and let the actual drivers that need it to set it, but > > > > > > this was not the best solution according to Dmitri. On my > > > > > > side, I am not entirely against this given that it's a > > > > > > chip feature, so the driver should be able to know that > > > > > > it's available. > > > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > > > I read this: > > > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > (that ST microelectronics accelerometer) but due to > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > handled by two drivers, one ACPI and one i2c." > > > > > > > > > > and that is the core of the issue. You have 2 drivers > > > > > fighting over the same device. Fix this and it will all > > > > > work. > > > > > > > > With my current implementation (which I sent in this patch), > > > > they are not fighting. > > > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and > > > > lis3lv02d only accelerometer device as lis3lv02d driver does > > > > not get IRQ number in platform data. > > > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and > > > > > accesses it via ACPI methods, can the same be done for Dell? > > > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI > > > > or DMI is even not i2c address of device, so it needs to be > > > > specified in code itself. > > > > > > > > Really there is no other way... :-( > > > > > > Sure there is: > > > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > > simply forward requests to dell-smo8800-accel i2c client. > > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel > > > does. > > > > Sorry, but I do not understand how you mean it... Why to provides > > new read/write i2c functions which are already implemented by > > i2c-i801 bus and lis3lv02d i2c driver? > > Because that would allow you to avoid clashes with i2c creating > interrupt mapping for client residing on host-notify-capable > controller. > > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > > device. > > > > But... what is problem with current implementation? Accelerometer > > HW provides two functions: > > > > 1) 3 axes reports > > 2) Disk freefall detection > > > > And 1) is handled by i2c driver lis3lv02d and 2) is by > > dell-smo8800. Both functions are independent here. > > > > I think you just trying to complicate this situation even more to > > be more complicated as currently is. > > Because this apparently does not work for you, does it? It is working fine. I do not see any problem. > In general, > if you want the same hardware be handled by 2 different drivers you > are going to have bad time. Yes, but in this case half of device is ACPI based and other half i2c based. This is problem of ACPI and Dell design. > It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are > the same, right? Yes. I understand that clean solution is to have one driver which provides everything. But because half of data are ACPI and half i2c, you still needs to create two drivers (one ACPI and one i2c). You can put both drivers into one .ko module, but still these will be two drivers due to how ACPI and i2c linux abstractions are different. > So, instead of having 2 drivers split the > functionality, can you forego registering smo8800 ACPI driver on > your whitelisted boxes and instead instantiate full i2c client > device with properly assigned both address and IRQ and let lis3lv02d > handle it (providing both accelerometer data and /dev/freefall)? With Michał we already discussed about it, see emails. Basically you can enable/disable kernel modules at compile time or blacklist at runtime (or even chose what will be compiled into vmlinux and what as external .ko module). Some distributions blacklist i2c-i801.ko module... And there can be also problem with initialization of i2c-i801 driver (fix is in commit a7ae81952cda, but does not have to work at every time!). So that move on whitelisted machines can potentially cause disappearance of /dev/freefall and users will not have hdd protection which is currently working.
On Tue, Jan 03, 2017 at 09:39:13PM +0100, Pali Rohár wrote: > On Tuesday 03 January 2017 21:24:18 Dmitry Torokhov wrote: > > On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > > > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > > > > > > > > > wrote: > > > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień > > > > > > > > > > wrote: > > > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał > > > > > > > > > > > > Kępień > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > > > 0x29. That i2c address is not specified in > > > > > > > > > > > > > > DMI or ACPI, so runtime detection without > > > > > > > > > > > > > > whitelist which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > > > confusing to me at first because there is > > > > > > > > > > > > > already an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is > > > > > > > > > > > > > that: > > > > > > > > > > > > > * the purpose of this patch is to expose a > > > > > > > > > > > > > richer interface (as > > > > > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on > > > > > > > > > > > > > some machines, > > > > > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > > > effectively duplicates the work that > > > > > > > > > > > > > lis3lv02d does). > > > > > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > > > exports /dev/freefall device which notify > > > > > > > > > > > > userspace about falls. lis3lv02d is i2c driver > > > > > > > > > > > > which exports axes of accelerometer. Additionaly > > > > > > > > > > > > lis3lv02d can export also /dev/freefall if > > > > > > > > > > > > registerer of i2c device provides irq number -- > > > > > > > > > > > > which is not case of this patch. > > > > > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and > > > > > > > > > > > > both are useful. > > > > > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent > > > > > > > > > > > > one HW device (that ST microelectronics > > > > > > > > > > > > accelerometer) but due to complicated HW > > > > > > > > > > > > abstraction and layers on Dell laptops it is > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed > > > > > > > > > > > > to lis3lv02d driver and that would export whole > > > > > > > > > > > > device (with /dev/freefall too), but due to HW > > > > > > > > > > > > abstraction it is too much complicated... > > > > > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > > > number all the way down to lis3lv02d is to set the > > > > > > > > > > > irq field of the struct i2c_board_info you are > > > > > > > > > > > passing to i2c_new_device(). And you can extract > > > > > > > > > > > that IRQ number e.g. in > > > > > > > > > > > check_acpi_smo88xx_device(). However, you would > > > > > > > > > > > then need to make sure dell-smo8800 does not > > > > > > > > > > > attempt to request the same IRQ on whitelisted > > > > > > > > > > > machines. This got me thinking about a way to > > > > > > > > > > > somehow incorporate your changes into dell-smo8800 > > > > > > > > > > > using Wolfram's bus_notifier suggestion, but I do > > > > > > > > > > > not have a working solution for now. What is > > > > > > > > > > > tempting about this approach is that you would not > > > > > > > > > > > have to scan the ACPI namespace in search of > > > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > > > driver... And there can be some pathological > > > > > > > > > > situation (thanks to adding ACPI layer as Andy > > > > > > > > > > pointed) that there will be more SMO88xx devices in > > > > > > > > > > ACPI. Plus you can compile kernel with and without > > > > > > > > > > those modules and also you can blacklist loading > > > > > > > > > > them (so compile time check is not enough). And > > > > > > > > > > still some correct message notifier must be used. > > > > > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > > > and available dell devices... > > > > > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to > > > > > > > > > find a nice way to implement this. > > > > > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > > > events, not alert") which breaks your patch. The > > > > > > > > > reason for that is that lis3lv02d relies on the i2c > > > > > > > > > client's IRQ being 0 to detect that it should not > > > > > > > > > create /dev/freefall. > > > > > > > > > > > > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > > > > > > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > > > causing lis3lv02d to create /dev/freefall, which in > > > > > > > > > turn conflicts with dell-smo8800 which is trying to > > > > > > > > > create /dev/freefall itself. > > > > > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > > > > > Apologies for that. > > > > > > > > > > > > > > I could easily fix this by adding a kernel API to know > > > > > > > whether the provided irq is from Host Notify or if it was > > > > > > > coming from an actual declaration. However, I have no idea > > > > > > > how many other drivers would require this (hopefully only > > > > > > > this one). > > > > > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ > > > > > > > and let the actual drivers that need it to set it, but > > > > > > > this was not the best solution according to Dmitri. On my > > > > > > > side, I am not entirely against this given that it's a > > > > > > > chip feature, so the driver should be able to know that > > > > > > > it's available. > > > > > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > > > > > I read this: > > > > > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > > (that ST microelectronics accelerometer) but due to > > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > > handled by two drivers, one ACPI and one i2c." > > > > > > > > > > > > and that is the core of the issue. You have 2 drivers > > > > > > fighting over the same device. Fix this and it will all > > > > > > work. > > > > > > > > > > With my current implementation (which I sent in this patch), > > > > > they are not fighting. > > > > > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and > > > > > lis3lv02d only accelerometer device as lis3lv02d driver does > > > > > not get IRQ number in platform data. > > > > > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and > > > > > > accesses it via ACPI methods, can the same be done for Dell? > > > > > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI > > > > > or DMI is even not i2c address of device, so it needs to be > > > > > specified in code itself. > > > > > > > > > > Really there is no other way... :-( > > > > > > > > Sure there is: > > > > > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > > > simply forward requests to dell-smo8800-accel i2c client. > > > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel > > > > does. > > > > > > Sorry, but I do not understand how you mean it... Why to provides > > > new read/write i2c functions which are already implemented by > > > i2c-i801 bus and lis3lv02d i2c driver? > > > > Because that would allow you to avoid clashes with i2c creating > > interrupt mapping for client residing on host-notify-capable > > controller. > > > > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > > > > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > > > device. > > > > > > But... what is problem with current implementation? Accelerometer > > > HW provides two functions: > > > > > > 1) 3 axes reports > > > 2) Disk freefall detection > > > > > > And 1) is handled by i2c driver lis3lv02d and 2) is by > > > dell-smo8800. Both functions are independent here. > > > > > > I think you just trying to complicate this situation even more to > > > be more complicated as currently is. > > > > Because this apparently does not work for you, does it? > > It is working fine. I do not see any problem. > > > In general, > > if you want the same hardware be handled by 2 different drivers you > > are going to have bad time. > > Yes, but in this case half of device is ACPI based and other half i2c > based. This is problem of ACPI and Dell design. > > > It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are > > the same, right? > > Yes. I understand that clean solution is to have one driver which > provides everything. > > But because half of data are ACPI and half i2c, you still needs to > create two drivers (one ACPI and one i2c). You can put both drivers into > one .ko module, but still these will be two drivers due to how ACPI and > i2c linux abstractions are different. > > > So, instead of having 2 drivers split the > > functionality, can you forego registering smo8800 ACPI driver on > > your whitelisted boxes and instead instantiate full i2c client > > device with properly assigned both address and IRQ and let lis3lv02d > > handle it (providing both accelerometer data and /dev/freefall)? > > With Michał we already discussed about it, see emails. Basically you can > enable/disable kernel modules at compile time or blacklist at runtime > (or even chose what will be compiled into vmlinux and what as external > .ko module). This can be solved with a bit of Kconfig/IS_ENABLED() code. > Some distributions blacklist i2c-i801.ko module... And Any particular reason for that? > there can be also problem with initialization of i2c-i801 driver (fix is > in commit a7ae81952cda, but does not have to work at every time!). So > that move on whitelisted machines can potentially cause disappearance of > /dev/freefall and users will not have hdd protection which is currently > working. Well, I gave you 2 possible solutions (roll your own i2c read/write, forward them to i2c client) or have faith in your implementation and let lis3lv02d handle it. The 3rd one is to possibly add a flag to disable host notify to IRQ mapping for given client (if Wolfram/Jean OK with it). Oh, the 4th one: change the irq in lis3lv02d.h to be "int" and change the check in lis3lv02d.c to be "lis->irq <= 0" and instantiate your i2c_client with board_info->irq = -1. Pick whichever you prefer. By the way, what do you need accelerometer for on these devices? They don't appear to be tablets that could use one... Thanks.
On Jan 03 2017 or thereabouts, Dmitry Torokhov wrote: > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires wrote: > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień wrote: > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał Kępień wrote: > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > microelectronics accelerometer at i2c address 0x29. > > > > > > > > > > > That i2c address is not specified in DMI or ACPI, so > > > > > > > > > > > runtime detection without whitelist which is below > > > > > > > > > > > is not possible. > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics accelerometer is > > > > > > > > > > > verified by existence of SMO88xx ACPI device which > > > > > > > > > > > represent that accelerometer. Unfortunately without > > > > > > > > > > > i2c address. > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit confusing > > > > > > > > > > to me at first because there is already an ACPI driver > > > > > > > > > > which handles SMO88xx > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is that: > > > > > > > > > > * the purpose of this patch is to expose a richer > > > > > > > > > > interface (as > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on some > > > > > > > > > > machines, > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and lis3lv02d > > > > > > > > > > can work > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > effectively duplicates the work that lis3lv02d > > > > > > > > > > does). > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and exports > > > > > > > > > /dev/freefall device which notify userspace about falls. > > > > > > > > > lis3lv02d is i2c driver which exports axes of > > > > > > > > > accelerometer. Additionaly lis3lv02d can export also > > > > > > > > > /dev/freefall if registerer of i2c device provides irq > > > > > > > > > number -- which is not case of this patch. > > > > > > > > > > > > > > > > > > So both drivers are doing different things and both are > > > > > > > > > useful. > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent one HW > > > > > > > > > device (that ST microelectronics accelerometer) but due > > > > > > > > > to complicated HW abstraction and layers on Dell laptops > > > > > > > > > it is handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed to > > > > > > > > > lis3lv02d driver and that would export whole device > > > > > > > > > (with /dev/freefall too), but due to HW abstraction it > > > > > > > > > is too much complicated... > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ number > > > > > > > > all the way down to lis3lv02d is to set the irq field of > > > > > > > > the struct i2c_board_info you are passing to > > > > > > > > i2c_new_device(). And you can extract that IRQ number > > > > > > > > e.g. in check_acpi_smo88xx_device(). However, you would > > > > > > > > then need to make sure dell-smo8800 does not attempt to > > > > > > > > request the same IRQ on whitelisted machines. This got me > > > > > > > > thinking about a way to somehow incorporate your changes > > > > > > > > into dell-smo8800 using Wolfram's bus_notifier suggestion, > > > > > > > > but I do not have a working solution for now. What is > > > > > > > > tempting about this approach is that you would not have to > > > > > > > > scan the ACPI namespace in search of SMO88xx devices, > > > > > > > > because smo8800_add() is automatically called for them. > > > > > > > > However, I fear that the resulting solution may be more > > > > > > > > complicated than the one you submitted. > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of loading > > > > > > > .ko modules is undefined. Binding devices to drivers > > > > > > > registered by .ko module is also in "random" order. At any > > > > > > > time any of those .ko module can be unloaded or at least > > > > > > > device unbind (via sysfs) from driver... And there can be > > > > > > > some pathological situation (thanks to adding ACPI layer as > > > > > > > Andy pointed) that there will be more SMO88xx devices in > > > > > > > ACPI. Plus you can compile kernel with and without those > > > > > > > modules and also you can blacklist loading them (so compile > > > > > > > time check is not enough). And still some correct message > > > > > > > notifier must be used. > > > > > > > > > > > > > > I think such solution is much much more complicated, there > > > > > > > are lot of combinations of kernel configuration and > > > > > > > available dell devices... > > > > > > > > > > > > I tried a few more things, but ultimately failed to find a nice > > > > > > way to implement this. > > > > > > > > > > > > Another issue popped up, though. Linus' master branch contains > > > > > > a recent commit by Benjamin Tissoires (CC'ed), 4d5538f5882a > > > > > > ("i2c: use an IRQ to report Host Notify events, not alert") > > > > > > which breaks your patch. The reason for that is that > > > > > > lis3lv02d relies on the i2c client's IRQ being 0 to detect > > > > > > that it should not create /dev/freefall. Benjamin's patch > > > > > > causes the Host Notify IRQ to be assigned to the i2c client > > > > > > your patch creates, thus causing lis3lv02d to create > > > > > > /dev/freefall, which in turn conflicts with dell-smo8800 which > > > > > > is trying to create /dev/freefall itself. > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > Apologies for that. > > > > > > > > I could easily fix this by adding a kernel API to know whether the > > > > provided irq is from Host Notify or if it was coming from an actual > > > > declaration. However, I have no idea how many other drivers would > > > > require this (hopefully only this one). > > > > > > > > One other solution would be to reserve the Host Notify IRQ and let > > > > the actual drivers that need it to set it, but this was not the > > > > best solution according to Dmitri. On my side, I am not entirely > > > > against this given that it's a chip feature, so the driver should > > > > be able to know that it's available. > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > I read this: > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device (that > > > ST microelectronics accelerometer) but due to complicated HW > > > abstraction and layers on Dell laptops it is handled by two drivers, > > > one ACPI and one i2c." > > > > > > and that is the core of the issue. You have 2 drivers fighting over > > > the same device. Fix this and it will all work. > > > > With my current implementation (which I sent in this patch), they are > > not fighting. > > > > dell-smo8800 exports /dev/freefall (and nothing more) and lis3lv02d only > > accelerometer device as lis3lv02d driver does not get IRQ number in > > platform data. > > > > > As far as I can see hp_accel instantiates lis3lv02d and accesses it > > > via ACPI methods, can the same be done for Dell? > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI or DMI > > is even not i2c address of device, so it needs to be specified in code > > itself. > > > > Really there is no other way... :-( > > Sure there is: > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > 2. dell-smo8800 provides read/write functions for lis3lv02d that simply > forward requests to dell-smo8800-accel i2c client. > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel does. > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > Yet another option: can we add a new flag to i2c_board_info controlling > whether we want to enable/disable wiring up host notify interrupt? That should be fairly easy to implement. For now, given that only Elan and Synaptics are the one in need for Host Notify, it could be better to request the Host Notify IRQ instead of disabling it unconditionally (which would make the current yet 8 years old lis3lv02d driver happy again). > Benjamin, is there anything "special" in RMI SMbus ACPI descriptors we > could use? > No, there is nothing special. Same situation for Elan with their latest touchpads over PS/2. There is just a knowledge from the driver that there is a device connected on a Host Notify capable bus on a specific address. To give you an idea, on Windows, the Synaptics (and Elan) driver even ships the equivalent of i2c-i801 to be sure to have one driver for it... So the knowledge is all in the driver. Cheers, Benjamin -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, Jan 03, 2017 at 10:29:34PM +0100, Benjamin Tissoires wrote: > On Jan 03 2017 or thereabouts, Dmitry Torokhov wrote: > > > Yet another option: can we add a new flag to i2c_board_info controlling > > whether we want to enable/disable wiring up host notify interrupt? > > That should be fairly easy to implement. For now, given that only Elan > and Synaptics are the one in need for Host Notify, it could be better to > request the Host Notify IRQ instead of disabling it unconditionally > (which would make the current yet 8 years old lis3lv02d driver happy > again). I like that we have it done in i2c core instead of having drivers implementing it individually. Since you are saying that handling host notify is property of the slave/driver maybe we should be adding a flag to the *i2c_driver* structure to let i2c core that we want to have host notify mapped as interrupt if "native" interrupt is not supplied by the platform code? Thanks.
On Tuesday 03 January 2017 12:59:37 Dmitry Torokhov wrote: > On Tue, Jan 03, 2017 at 09:39:13PM +0100, Pali Rohár wrote: > > On Tuesday 03 January 2017 21:24:18 Dmitry Torokhov wrote: > > > On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > > > > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > > > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > > > > > > > > > > > wrote: > > > > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień > > > > > > > > > > > wrote: > > > > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał > > > > > > > > > > > > > Kępień > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > > > > 0x29. That i2c address is not specified in > > > > > > > > > > > > > > > DMI or ACPI, so runtime detection without > > > > > > > > > > > > > > > whitelist which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > > > > confusing to me at first because there is > > > > > > > > > > > > > > already an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is > > > > > > > > > > > > > > that: > > > > > > > > > > > > > > * the purpose of this patch is to expose a > > > > > > > > > > > > > > richer interface (as > > > > > > > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on > > > > > > > > > > > > > > some machines, > > > > > > > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > > > > effectively duplicates the work that > > > > > > > > > > > > > > lis3lv02d does). > > > > > > > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > > > > exports /dev/freefall device which notify > > > > > > > > > > > > > userspace about falls. lis3lv02d is i2c driver > > > > > > > > > > > > > which exports axes of accelerometer. Additionaly > > > > > > > > > > > > > lis3lv02d can export also /dev/freefall if > > > > > > > > > > > > > registerer of i2c device provides irq number -- > > > > > > > > > > > > > which is not case of this patch. > > > > > > > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and > > > > > > > > > > > > > both are useful. > > > > > > > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent > > > > > > > > > > > > > one HW device (that ST microelectronics > > > > > > > > > > > > > accelerometer) but due to complicated HW > > > > > > > > > > > > > abstraction and layers on Dell laptops it is > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed > > > > > > > > > > > > > to lis3lv02d driver and that would export whole > > > > > > > > > > > > > device (with /dev/freefall too), but due to HW > > > > > > > > > > > > > abstraction it is too much complicated... > > > > > > > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > > > > number all the way down to lis3lv02d is to set the > > > > > > > > > > > > irq field of the struct i2c_board_info you are > > > > > > > > > > > > passing to i2c_new_device(). And you can extract > > > > > > > > > > > > that IRQ number e.g. in > > > > > > > > > > > > check_acpi_smo88xx_device(). However, you would > > > > > > > > > > > > then need to make sure dell-smo8800 does not > > > > > > > > > > > > attempt to request the same IRQ on whitelisted > > > > > > > > > > > > machines. This got me thinking about a way to > > > > > > > > > > > > somehow incorporate your changes into dell-smo8800 > > > > > > > > > > > > using Wolfram's bus_notifier suggestion, but I do > > > > > > > > > > > > not have a working solution for now. What is > > > > > > > > > > > > tempting about this approach is that you would not > > > > > > > > > > > > have to scan the ACPI namespace in search of > > > > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > > > > driver... And there can be some pathological > > > > > > > > > > > situation (thanks to adding ACPI layer as Andy > > > > > > > > > > > pointed) that there will be more SMO88xx devices in > > > > > > > > > > > ACPI. Plus you can compile kernel with and without > > > > > > > > > > > those modules and also you can blacklist loading > > > > > > > > > > > them (so compile time check is not enough). And > > > > > > > > > > > still some correct message notifier must be used. > > > > > > > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > > > > and available dell devices... > > > > > > > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to > > > > > > > > > > find a nice way to implement this. > > > > > > > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > > > > events, not alert") which breaks your patch. The > > > > > > > > > > reason for that is that lis3lv02d relies on the i2c > > > > > > > > > > client's IRQ being 0 to detect that it should not > > > > > > > > > > create /dev/freefall. > > > > > > > > > > > > > > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > > > > > > > > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > > > > causing lis3lv02d to create /dev/freefall, which in > > > > > > > > > > turn conflicts with dell-smo8800 which is trying to > > > > > > > > > > create /dev/freefall itself. > > > > > > > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > > > > > > > Apologies for that. > > > > > > > > > > > > > > > > I could easily fix this by adding a kernel API to know > > > > > > > > whether the provided irq is from Host Notify or if it was > > > > > > > > coming from an actual declaration. However, I have no idea > > > > > > > > how many other drivers would require this (hopefully only > > > > > > > > this one). > > > > > > > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ > > > > > > > > and let the actual drivers that need it to set it, but > > > > > > > > this was not the best solution according to Dmitri. On my > > > > > > > > side, I am not entirely against this given that it's a > > > > > > > > chip feature, so the driver should be able to know that > > > > > > > > it's available. > > > > > > > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > > > > > > > I read this: > > > > > > > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > > > (that ST microelectronics accelerometer) but due to > > > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > > > handled by two drivers, one ACPI and one i2c." > > > > > > > > > > > > > > and that is the core of the issue. You have 2 drivers > > > > > > > fighting over the same device. Fix this and it will all > > > > > > > work. > > > > > > > > > > > > With my current implementation (which I sent in this patch), > > > > > > they are not fighting. > > > > > > > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and > > > > > > lis3lv02d only accelerometer device as lis3lv02d driver does > > > > > > not get IRQ number in platform data. > > > > > > > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and > > > > > > > accesses it via ACPI methods, can the same be done for Dell? > > > > > > > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI > > > > > > or DMI is even not i2c address of device, so it needs to be > > > > > > specified in code itself. > > > > > > > > > > > > Really there is no other way... :-( > > > > > > > > > > Sure there is: > > > > > > > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > > > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > > > > simply forward requests to dell-smo8800-accel i2c client. > > > > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel > > > > > does. > > > > > > > > Sorry, but I do not understand how you mean it... Why to provides > > > > new read/write i2c functions which are already implemented by > > > > i2c-i801 bus and lis3lv02d i2c driver? > > > > > > Because that would allow you to avoid clashes with i2c creating > > > interrupt mapping for client residing on host-notify-capable > > > controller. > > > > > > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > > > > > > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > > > > device. > > > > > > > > But... what is problem with current implementation? Accelerometer > > > > HW provides two functions: > > > > > > > > 1) 3 axes reports > > > > 2) Disk freefall detection > > > > > > > > And 1) is handled by i2c driver lis3lv02d and 2) is by > > > > dell-smo8800. Both functions are independent here. > > > > > > > > I think you just trying to complicate this situation even more to > > > > be more complicated as currently is. > > > > > > Because this apparently does not work for you, does it? > > > > It is working fine. I do not see any problem. > > > > > In general, > > > if you want the same hardware be handled by 2 different drivers you > > > are going to have bad time. > > > > Yes, but in this case half of device is ACPI based and other half i2c > > based. This is problem of ACPI and Dell design. > > > > > It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are > > > the same, right? > > > > Yes. I understand that clean solution is to have one driver which > > provides everything. > > > > But because half of data are ACPI and half i2c, you still needs to > > create two drivers (one ACPI and one i2c). You can put both drivers into > > one .ko module, but still these will be two drivers due to how ACPI and > > i2c linux abstractions are different. > > > > > So, instead of having 2 drivers split the > > > functionality, can you forego registering smo8800 ACPI driver on > > > your whitelisted boxes and instead instantiate full i2c client > > > device with properly assigned both address and IRQ and let lis3lv02d > > > handle it (providing both accelerometer data and /dev/freefall)? > > > > With Michał we already discussed about it, see emails. Basically you can > > enable/disable kernel modules at compile time or blacklist at runtime > > (or even chose what will be compiled into vmlinux and what as external > > .ko module). > > This can be solved with a bit of Kconfig/IS_ENABLED() code. > > > Some distributions blacklist i2c-i801.ko module... And > > Any particular reason for that? > > > there can be also problem with initialization of i2c-i801 driver (fix is > > in commit a7ae81952cda, but does not have to work at every time!). So > > that move on whitelisted machines can potentially cause disappearance of > > /dev/freefall and users will not have hdd protection which is currently > > working. > > Well, I gave you 2 possible solutions (roll your own i2c read/write, > forward them to i2c client) or have faith in your implementation and let > lis3lv02d handle it. > > The 3rd one is to possibly add a flag to disable host notify to IRQ > mapping for given client (if Wolfram/Jean OK with it). > > Oh, the 4th one: change the irq in lis3lv02d.h to be "int" and change > the check in lis3lv02d.c to be "lis->irq <= 0" and instantiate your > i2c_client with board_info->irq = -1. > > Pick whichever you prefer. > > By the way, what do you need accelerometer for on these devices? They > don't appear to be tablets that could use one... Ah, you are talking about problem that after 4d5538f5882a lis3lv02d will not work... I thought that discussion is about different mechanism how to implement bus registration notification to smo8800 driver (or different solution to not have registration in i801).
On Jan 04 2017 or thereabouts, Pali Rohár wrote: > On Tuesday 03 January 2017 12:59:37 Dmitry Torokhov wrote: > > On Tue, Jan 03, 2017 at 09:39:13PM +0100, Pali Rohár wrote: > > > On Tuesday 03 January 2017 21:24:18 Dmitry Torokhov wrote: > > > > On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > > > > > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > > > > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > > > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > > > > > > > > > > > > > wrote: > > > > > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał > > > > > > > > > > > > > > Kępień > > > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > > > > > 0x29. That i2c address is not specified in > > > > > > > > > > > > > > > > DMI or ACPI, so runtime detection without > > > > > > > > > > > > > > > > whitelist which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > > > > > confusing to me at first because there is > > > > > > > > > > > > > > > already an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is > > > > > > > > > > > > > > > that: > > > > > > > > > > > > > > > * the purpose of this patch is to expose a > > > > > > > > > > > > > > > richer interface (as > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on > > > > > > > > > > > > > > > some machines, > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > > > > > effectively duplicates the work that > > > > > > > > > > > > > > > lis3lv02d does). > > > > > > > > > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > > > > > exports /dev/freefall device which notify > > > > > > > > > > > > > > userspace about falls. lis3lv02d is i2c driver > > > > > > > > > > > > > > which exports axes of accelerometer. Additionaly > > > > > > > > > > > > > > lis3lv02d can export also /dev/freefall if > > > > > > > > > > > > > > registerer of i2c device provides irq number -- > > > > > > > > > > > > > > which is not case of this patch. > > > > > > > > > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and > > > > > > > > > > > > > > both are useful. > > > > > > > > > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent > > > > > > > > > > > > > > one HW device (that ST microelectronics > > > > > > > > > > > > > > accelerometer) but due to complicated HW > > > > > > > > > > > > > > abstraction and layers on Dell laptops it is > > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed > > > > > > > > > > > > > > to lis3lv02d driver and that would export whole > > > > > > > > > > > > > > device (with /dev/freefall too), but due to HW > > > > > > > > > > > > > > abstraction it is too much complicated... > > > > > > > > > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > > > > > number all the way down to lis3lv02d is to set the > > > > > > > > > > > > > irq field of the struct i2c_board_info you are > > > > > > > > > > > > > passing to i2c_new_device(). And you can extract > > > > > > > > > > > > > that IRQ number e.g. in > > > > > > > > > > > > > check_acpi_smo88xx_device(). However, you would > > > > > > > > > > > > > then need to make sure dell-smo8800 does not > > > > > > > > > > > > > attempt to request the same IRQ on whitelisted > > > > > > > > > > > > > machines. This got me thinking about a way to > > > > > > > > > > > > > somehow incorporate your changes into dell-smo8800 > > > > > > > > > > > > > using Wolfram's bus_notifier suggestion, but I do > > > > > > > > > > > > > not have a working solution for now. What is > > > > > > > > > > > > > tempting about this approach is that you would not > > > > > > > > > > > > > have to scan the ACPI namespace in search of > > > > > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > > > > > driver... And there can be some pathological > > > > > > > > > > > > situation (thanks to adding ACPI layer as Andy > > > > > > > > > > > > pointed) that there will be more SMO88xx devices in > > > > > > > > > > > > ACPI. Plus you can compile kernel with and without > > > > > > > > > > > > those modules and also you can blacklist loading > > > > > > > > > > > > them (so compile time check is not enough). And > > > > > > > > > > > > still some correct message notifier must be used. > > > > > > > > > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > > > > > and available dell devices... > > > > > > > > > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to > > > > > > > > > > > find a nice way to implement this. > > > > > > > > > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > > > > > events, not alert") which breaks your patch. The > > > > > > > > > > > reason for that is that lis3lv02d relies on the i2c > > > > > > > > > > > client's IRQ being 0 to detect that it should not > > > > > > > > > > > create /dev/freefall. > > > > > > > > > > > > > > > > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > > > > > > > > > > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > > > > > causing lis3lv02d to create /dev/freefall, which in > > > > > > > > > > > turn conflicts with dell-smo8800 which is trying to > > > > > > > > > > > create /dev/freefall itself. > > > > > > > > > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > > > > > > > > > Apologies for that. > > > > > > > > > > > > > > > > > > I could easily fix this by adding a kernel API to know > > > > > > > > > whether the provided irq is from Host Notify or if it was > > > > > > > > > coming from an actual declaration. However, I have no idea > > > > > > > > > how many other drivers would require this (hopefully only > > > > > > > > > this one). > > > > > > > > > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ > > > > > > > > > and let the actual drivers that need it to set it, but > > > > > > > > > this was not the best solution according to Dmitri. On my > > > > > > > > > side, I am not entirely against this given that it's a > > > > > > > > > chip feature, so the driver should be able to know that > > > > > > > > > it's available. > > > > > > > > > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > > > > > > > > > I read this: > > > > > > > > > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > > > > (that ST microelectronics accelerometer) but due to > > > > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > > > > handled by two drivers, one ACPI and one i2c." > > > > > > > > > > > > > > > > and that is the core of the issue. You have 2 drivers > > > > > > > > fighting over the same device. Fix this and it will all > > > > > > > > work. > > > > > > > > > > > > > > With my current implementation (which I sent in this patch), > > > > > > > they are not fighting. > > > > > > > > > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and > > > > > > > lis3lv02d only accelerometer device as lis3lv02d driver does > > > > > > > not get IRQ number in platform data. > > > > > > > > > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and > > > > > > > > accesses it via ACPI methods, can the same be done for Dell? > > > > > > > > > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI > > > > > > > or DMI is even not i2c address of device, so it needs to be > > > > > > > specified in code itself. > > > > > > > > > > > > > > Really there is no other way... :-( > > > > > > > > > > > > Sure there is: > > > > > > > > > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > > > > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > > > > > simply forward requests to dell-smo8800-accel i2c client. > > > > > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel > > > > > > does. > > > > > > > > > > Sorry, but I do not understand how you mean it... Why to provides > > > > > new read/write i2c functions which are already implemented by > > > > > i2c-i801 bus and lis3lv02d i2c driver? > > > > > > > > Because that would allow you to avoid clashes with i2c creating > > > > interrupt mapping for client residing on host-notify-capable > > > > controller. > > > > > > > > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > > > > > > > > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > > > > > device. > > > > > > > > > > But... what is problem with current implementation? Accelerometer > > > > > HW provides two functions: > > > > > > > > > > 1) 3 axes reports > > > > > 2) Disk freefall detection > > > > > > > > > > And 1) is handled by i2c driver lis3lv02d and 2) is by > > > > > dell-smo8800. Both functions are independent here. > > > > > > > > > > I think you just trying to complicate this situation even more to > > > > > be more complicated as currently is. > > > > > > > > Because this apparently does not work for you, does it? > > > > > > It is working fine. I do not see any problem. > > > > > > > In general, > > > > if you want the same hardware be handled by 2 different drivers you > > > > are going to have bad time. > > > > > > Yes, but in this case half of device is ACPI based and other half i2c > > > based. This is problem of ACPI and Dell design. > > > > > > > It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are > > > > the same, right? > > > > > > Yes. I understand that clean solution is to have one driver which > > > provides everything. > > > > > > But because half of data are ACPI and half i2c, you still needs to > > > create two drivers (one ACPI and one i2c). You can put both drivers into > > > one .ko module, but still these will be two drivers due to how ACPI and > > > i2c linux abstractions are different. > > > > > > > So, instead of having 2 drivers split the > > > > functionality, can you forego registering smo8800 ACPI driver on > > > > your whitelisted boxes and instead instantiate full i2c client > > > > device with properly assigned both address and IRQ and let lis3lv02d > > > > handle it (providing both accelerometer data and /dev/freefall)? > > > > > > With Michał we already discussed about it, see emails. Basically you can > > > enable/disable kernel modules at compile time or blacklist at runtime > > > (or even chose what will be compiled into vmlinux and what as external > > > .ko module). > > > > This can be solved with a bit of Kconfig/IS_ENABLED() code. > > > > > Some distributions blacklist i2c-i801.ko module... And > > > > Any particular reason for that? > > > > > there can be also problem with initialization of i2c-i801 driver (fix is > > > in commit a7ae81952cda, but does not have to work at every time!). So > > > that move on whitelisted machines can potentially cause disappearance of > > > /dev/freefall and users will not have hdd protection which is currently > > > working. > > > > Well, I gave you 2 possible solutions (roll your own i2c read/write, > > forward them to i2c client) or have faith in your implementation and let > > lis3lv02d handle it. > > > > The 3rd one is to possibly add a flag to disable host notify to IRQ > > mapping for given client (if Wolfram/Jean OK with it). > > > > Oh, the 4th one: change the irq in lis3lv02d.h to be "int" and change > > the check in lis3lv02d.c to be "lis->irq <= 0" and instantiate your > > i2c_client with board_info->irq = -1. > > > > Pick whichever you prefer. > > > > By the way, what do you need accelerometer for on these devices? They > > don't appear to be tablets that could use one... > > Ah, you are talking about problem that after 4d5538f5882a lis3lv02d will > not work... I thought that discussion is about different mechanism how > to implement bus registration notification to smo8800 driver (or > different solution to not have registration in i801). > Just because I am not sure I got everything right, could you confirm that: - in the current upstream tree, the dell-smo8800 driver is now broken after 4d5538f5882a (i2c: use an IRQ to report Host Notify events, not alert) - this series adds an extra lis3lv02d on some machines and you have problem fighting for the irq (but this is not upstream yet). The extra lis3lv02d node is added from dell-smo8800 If the first point is not correct (by default, dell-smo8800 will not be loaded at the same time than lis3lv02d), then it's a design issue with the interactions between those 2 drivers. If the first point is correct because ACPI declares both devices, then there is an urgent fix to propose to not enable Host Notify by default on Host Notifier capable adapters. (even though the design between the 2 drivers is wrong, it's considered as a regression). Cheers, Benjamin -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Jan 03 2017 or thereabouts, Dmitry Torokhov wrote: > On Tue, Jan 03, 2017 at 10:29:34PM +0100, Benjamin Tissoires wrote: > > On Jan 03 2017 or thereabouts, Dmitry Torokhov wrote: > > > > > Yet another option: can we add a new flag to i2c_board_info controlling > > > whether we want to enable/disable wiring up host notify interrupt? > > > > That should be fairly easy to implement. For now, given that only Elan > > and Synaptics are the one in need for Host Notify, it could be better to > > request the Host Notify IRQ instead of disabling it unconditionally > > (which would make the current yet 8 years old lis3lv02d driver happy > > again). > > I like that we have it done in i2c core instead of having drivers > implementing it individually. Since you are saying that handling host > notify is property of the slave/driver maybe we should be adding a flag > to the *i2c_driver* structure to let i2c core that we want to have host > notify mapped as interrupt if "native" interrupt is not supplied by the > platform code? I don't think this is a good idea. It's still a property of the I2C device, not the driver. It's crappy under Windows, but that doesn't prevent us to do the right thing. I think the idea of having it at the i2c_board_info level is the good one. It's a property of the device node and it wouldn't hurt me to have a device tree property for that too (not entering the DT field now). There is no ACPI prop for it too, but I wouldn't be surprised if it comes in a later revision. The advantage of having it turned on unconditionally is that we can instantiate it from userspace without breaking the sysfs ABI. Note that in the 2 uses I have seen so far of Host Notify, in both cases I need to instantiate the I2C device from an other driver (psmouse) so I can control the content of i2c_board_info. Cheers, Benjamin -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wednesday 04 January 2017 10:05:22 Benjamin Tissoires wrote: > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > On Tuesday 03 January 2017 12:59:37 Dmitry Torokhov wrote: > > > On Tue, Jan 03, 2017 at 09:39:13PM +0100, Pali Rohár wrote: > > > > On Tuesday 03 January 2017 21:24:18 Dmitry Torokhov wrote: > > > > > On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > > > > > > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > > > > > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > > > > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > > > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał > > > > > > > > > > > > > > > Kępień > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > > > > > > 0x29. That i2c address is not specified in > > > > > > > > > > > > > > > > > DMI or ACPI, so runtime detection without > > > > > > > > > > > > > > > > > whitelist which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > > > > > > confusing to me at first because there is > > > > > > > > > > > > > > > > already an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is > > > > > > > > > > > > > > > > that: > > > > > > > > > > > > > > > > * the purpose of this patch is to expose a > > > > > > > > > > > > > > > > richer interface (as > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on > > > > > > > > > > > > > > > > some machines, > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > > > > > > effectively duplicates the work that > > > > > > > > > > > > > > > > lis3lv02d does). > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > > > > > > exports /dev/freefall device which notify > > > > > > > > > > > > > > > userspace about falls. lis3lv02d is i2c driver > > > > > > > > > > > > > > > which exports axes of accelerometer. Additionaly > > > > > > > > > > > > > > > lis3lv02d can export also /dev/freefall if > > > > > > > > > > > > > > > registerer of i2c device provides irq number -- > > > > > > > > > > > > > > > which is not case of this patch. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and > > > > > > > > > > > > > > > both are useful. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent > > > > > > > > > > > > > > > one HW device (that ST microelectronics > > > > > > > > > > > > > > > accelerometer) but due to complicated HW > > > > > > > > > > > > > > > abstraction and layers on Dell laptops it is > > > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed > > > > > > > > > > > > > > > to lis3lv02d driver and that would export whole > > > > > > > > > > > > > > > device (with /dev/freefall too), but due to HW > > > > > > > > > > > > > > > abstraction it is too much complicated... > > > > > > > > > > > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > > > > > > number all the way down to lis3lv02d is to set the > > > > > > > > > > > > > > irq field of the struct i2c_board_info you are > > > > > > > > > > > > > > passing to i2c_new_device(). And you can extract > > > > > > > > > > > > > > that IRQ number e.g. in > > > > > > > > > > > > > > check_acpi_smo88xx_device(). However, you would > > > > > > > > > > > > > > then need to make sure dell-smo8800 does not > > > > > > > > > > > > > > attempt to request the same IRQ on whitelisted > > > > > > > > > > > > > > machines. This got me thinking about a way to > > > > > > > > > > > > > > somehow incorporate your changes into dell-smo8800 > > > > > > > > > > > > > > using Wolfram's bus_notifier suggestion, but I do > > > > > > > > > > > > > > not have a working solution for now. What is > > > > > > > > > > > > > > tempting about this approach is that you would not > > > > > > > > > > > > > > have to scan the ACPI namespace in search of > > > > > > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > > > > > > driver... And there can be some pathological > > > > > > > > > > > > > situation (thanks to adding ACPI layer as Andy > > > > > > > > > > > > > pointed) that there will be more SMO88xx devices in > > > > > > > > > > > > > ACPI. Plus you can compile kernel with and without > > > > > > > > > > > > > those modules and also you can blacklist loading > > > > > > > > > > > > > them (so compile time check is not enough). And > > > > > > > > > > > > > still some correct message notifier must be used. > > > > > > > > > > > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > > > > > > and available dell devices... > > > > > > > > > > > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to > > > > > > > > > > > > find a nice way to implement this. > > > > > > > > > > > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > > > > > > events, not alert") which breaks your patch. The > > > > > > > > > > > > reason for that is that lis3lv02d relies on the i2c > > > > > > > > > > > > client's IRQ being 0 to detect that it should not > > > > > > > > > > > > create /dev/freefall. > > > > > > > > > > > > > > > > > > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > > > > > > > > > > > > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > > > > > > causing lis3lv02d to create /dev/freefall, which in > > > > > > > > > > > > turn conflicts with dell-smo8800 which is trying to > > > > > > > > > > > > create /dev/freefall itself. > > > > > > > > > > > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > > > > > > > > > > > Apologies for that. > > > > > > > > > > > > > > > > > > > > I could easily fix this by adding a kernel API to know > > > > > > > > > > whether the provided irq is from Host Notify or if it was > > > > > > > > > > coming from an actual declaration. However, I have no idea > > > > > > > > > > how many other drivers would require this (hopefully only > > > > > > > > > > this one). > > > > > > > > > > > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ > > > > > > > > > > and let the actual drivers that need it to set it, but > > > > > > > > > > this was not the best solution according to Dmitri. On my > > > > > > > > > > side, I am not entirely against this given that it's a > > > > > > > > > > chip feature, so the driver should be able to know that > > > > > > > > > > it's available. > > > > > > > > > > > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > > > > > > > > > > > I read this: > > > > > > > > > > > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > > > > > (that ST microelectronics accelerometer) but due to > > > > > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > > > > > handled by two drivers, one ACPI and one i2c." > > > > > > > > > > > > > > > > > > and that is the core of the issue. You have 2 drivers > > > > > > > > > fighting over the same device. Fix this and it will all > > > > > > > > > work. > > > > > > > > > > > > > > > > With my current implementation (which I sent in this patch), > > > > > > > > they are not fighting. > > > > > > > > > > > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and > > > > > > > > lis3lv02d only accelerometer device as lis3lv02d driver does > > > > > > > > not get IRQ number in platform data. > > > > > > > > > > > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and > > > > > > > > > accesses it via ACPI methods, can the same be done for Dell? > > > > > > > > > > > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI > > > > > > > > or DMI is even not i2c address of device, so it needs to be > > > > > > > > specified in code itself. > > > > > > > > > > > > > > > > Really there is no other way... :-( > > > > > > > > > > > > > > Sure there is: > > > > > > > > > > > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > > > > > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > > > > > > simply forward requests to dell-smo8800-accel i2c client. > > > > > > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel > > > > > > > does. > > > > > > > > > > > > Sorry, but I do not understand how you mean it... Why to provides > > > > > > new read/write i2c functions which are already implemented by > > > > > > i2c-i801 bus and lis3lv02d i2c driver? > > > > > > > > > > Because that would allow you to avoid clashes with i2c creating > > > > > interrupt mapping for client residing on host-notify-capable > > > > > controller. > > > > > > > > > > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > > > > > > > > > > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > > > > > > device. > > > > > > > > > > > > But... what is problem with current implementation? Accelerometer > > > > > > HW provides two functions: > > > > > > > > > > > > 1) 3 axes reports > > > > > > 2) Disk freefall detection > > > > > > > > > > > > And 1) is handled by i2c driver lis3lv02d and 2) is by > > > > > > dell-smo8800. Both functions are independent here. > > > > > > > > > > > > I think you just trying to complicate this situation even more to > > > > > > be more complicated as currently is. > > > > > > > > > > Because this apparently does not work for you, does it? > > > > > > > > It is working fine. I do not see any problem. > > > > > > > > > In general, > > > > > if you want the same hardware be handled by 2 different drivers you > > > > > are going to have bad time. > > > > > > > > Yes, but in this case half of device is ACPI based and other half i2c > > > > based. This is problem of ACPI and Dell design. > > > > > > > > > It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are > > > > > the same, right? > > > > > > > > Yes. I understand that clean solution is to have one driver which > > > > provides everything. > > > > > > > > But because half of data are ACPI and half i2c, you still needs to > > > > create two drivers (one ACPI and one i2c). You can put both drivers into > > > > one .ko module, but still these will be two drivers due to how ACPI and > > > > i2c linux abstractions are different. > > > > > > > > > So, instead of having 2 drivers split the > > > > > functionality, can you forego registering smo8800 ACPI driver on > > > > > your whitelisted boxes and instead instantiate full i2c client > > > > > device with properly assigned both address and IRQ and let lis3lv02d > > > > > handle it (providing both accelerometer data and /dev/freefall)? > > > > > > > > With Michał we already discussed about it, see emails. Basically you can > > > > enable/disable kernel modules at compile time or blacklist at runtime > > > > (or even chose what will be compiled into vmlinux and what as external > > > > .ko module). > > > > > > This can be solved with a bit of Kconfig/IS_ENABLED() code. > > > > > > > Some distributions blacklist i2c-i801.ko module... And > > > > > > Any particular reason for that? > > > > > > > there can be also problem with initialization of i2c-i801 driver (fix is > > > > in commit a7ae81952cda, but does not have to work at every time!). So > > > > that move on whitelisted machines can potentially cause disappearance of > > > > /dev/freefall and users will not have hdd protection which is currently > > > > working. > > > > > > Well, I gave you 2 possible solutions (roll your own i2c read/write, > > > forward them to i2c client) or have faith in your implementation and let > > > lis3lv02d handle it. > > > > > > The 3rd one is to possibly add a flag to disable host notify to IRQ > > > mapping for given client (if Wolfram/Jean OK with it). > > > > > > Oh, the 4th one: change the irq in lis3lv02d.h to be "int" and change > > > the check in lis3lv02d.c to be "lis->irq <= 0" and instantiate your > > > i2c_client with board_info->irq = -1. > > > > > > Pick whichever you prefer. > > > > > > By the way, what do you need accelerometer for on these devices? They > > > don't appear to be tablets that could use one... > > > > Ah, you are talking about problem that after 4d5538f5882a lis3lv02d will > > not work... I thought that discussion is about different mechanism how > > to implement bus registration notification to smo8800 driver (or > > different solution to not have registration in i801). > > > > Just because I am not sure I got everything right, could you confirm > that: > - in the current upstream tree, the dell-smo8800 driver is now broken > after 4d5538f5882a (i2c: use an IRQ to report Host Notify events, not > alert) No, dell-smo8800 it is working fine. It is fully independent from i2c and lis3lv02d. It is pure ACPI driver which does not share anything with i2c. > - this series adds an extra lis3lv02d on some machines and you have > problem fighting for the irq (but this is not upstream yet). Yes, this series (not merged yet) adds extra lis3lv02d device but is not working because of 4d5538f5882a. > The extra lis3lv02d node is added from dell-smo8800 No, dell-smo8800 does not add new node in this patch. > If the first point is not correct (by default, dell-smo8800 will not be > loaded at the same time than lis3lv02d), then it's a design issue with > the interactions between those 2 drivers. No, there is no interactions between these two drivers (dell-smo8800 and lis3lv02d). dell-smo8800 is pure ACPI driver and exports just /dev/freefall device based on IRQ (and nothing more). And lis3lv02d in *current* configuration in this patch exports only accelerometer input device, not /dev/freefall. It does not use IRQ. (Just there is problem with 4d5538f5882a which tells lis3lv02d IRQ number which is not freefall report, therefore lis3lv02d does not work). To make it clear, ST Accelerometer provides two operations: * report free fall * report 3 axes Free fall is reported by IRQ, state of 3 axes via i2c bus. Free fall IRQ is handled by dell-smo8800, state of 3 axes via i2c lis3lv02d driver. lis3lv02d can handle also free fall IRQ is platform i2c data provides IRQ number for it -- but this is not case in our Dell configuration. But commit 4d5538f5882a inject some IRQ number to lis3lv02d driver which is not free fall detection and so is breaking lis3lv02 driver. In our Dell configuration (by this patch) there should be no IRQ number. It is clear now? > If the first point is correct because ACPI declares both devices, then > there is an urgent fix to propose to not enable Host Notify by default > on Host Notifier capable adapters. (even though the design between the > 2 drivers is wrong, it's considered as a regression). > > Cheers, > Benjamin
On Wednesday 04 January 2017 10:13:38 Benjamin Tissoires wrote: > On Jan 03 2017 or thereabouts, Dmitry Torokhov wrote: > > On Tue, Jan 03, 2017 at 10:29:34PM +0100, Benjamin Tissoires wrote: > > > On Jan 03 2017 or thereabouts, Dmitry Torokhov wrote: > > > > > > > Yet another option: can we add a new flag to i2c_board_info controlling > > > > whether we want to enable/disable wiring up host notify interrupt? > > > > > > That should be fairly easy to implement. For now, given that only Elan > > > and Synaptics are the one in need for Host Notify, it could be better to > > > request the Host Notify IRQ instead of disabling it unconditionally > > > (which would make the current yet 8 years old lis3lv02d driver happy > > > again). > > > > I like that we have it done in i2c core instead of having drivers > > implementing it individually. Since you are saying that handling host > > notify is property of the slave/driver maybe we should be adding a flag > > to the *i2c_driver* structure to let i2c core that we want to have host > > notify mapped as interrupt if "native" interrupt is not supplied by the > > platform code? > > I don't think this is a good idea. It's still a property of the I2C > device, not the driver. It's crappy under Windows, but that doesn't > prevent us to do the right thing. > > I think the idea of having it at the i2c_board_info level is the good > one. It's a property of the device node and it wouldn't hurt me to have > a device tree property for that too (not entering the DT field now). > There is no ACPI prop for it too, but I wouldn't be surprised if it > comes in a later revision. The advantage of having it turned on > unconditionally is that we can instantiate it from userspace without > breaking the sysfs ABI. > > Note that in the 2 uses I have seen so far of Host Notify, in both cases > I need to instantiate the I2C device from an other driver (psmouse) so I > can control the content of i2c_board_info. If I understand it correctly, there is problem that i2c lis3lv02d driver needs to get IRQ number for freefall and in current structure you pass host notify IRQ number. It means that one property in lis3lv02d driver is used for two things: free fall IRQ and host notify IRQ. So the only way how to fix it is to use two different IRQ properties. Probably free fall IRQ is lis3lv02d specific and should be in platform data structure, not in generic i2c_board_info shared across all i2c drivers?
On Wed, 28 Dec 2016 15:02:52 +0100, Wolfram Sang wrote: > On Tue, Dec 27, 2016 at 02:51:01PM +0100, Pali Rohár wrote: > > i2c_new_device() with lis3lv02d for i801 i2c bus needs to be called > > after initializing i2c-i801 bus driver. > > > > I have no idea how to do it (properly) outside of i2c-i801.c file. > > I once used bus_notifiers to achieve something similar. You could check > arch/arm/mach-shmobile/regulator-quirk-rcar-gen2.c to see an action > triggered once a client device got added, but you could act on another > action like BUS_NOTIFY_BOUND_DRIVER. I used exactly that, too, somewhen > somewhere. Haven't checked if that helps here, too. And since we have > a > precedence (Fujitsu case), I'll leave it to Jean who is the maintainer > of this driver. I don't have a strong opinion on the matter, whatever works is fine with me.
On Tue, Jan 3, 2017 at 10:39 PM, Pali Rohár <pali.rohar@gmail.com> wrote: > > With Michał we already discussed about it, see emails. Basically you can > enable/disable kernel modules at compile time or blacklist at runtime > (or even chose what will be compiled into vmlinux and what as external > .ko module). Some distributions blacklist i2c-i801.ko module... But you understand that any of compile/not compile is not an option, right? The case which we face will be both of them, if possible, will be compiled as modules. Blacklisting means making your problem the actual user's one. Not good. > And > there can be also problem with initialization of i2c-i801 driver (fix is > in commit a7ae81952cda, but does not have to work at every time!). So > that move on whitelisted machines can potentially cause disappearance of > /dev/freefall and users will not have hdd protection which is currently > working.
On Jan 04 2017 or thereabouts, Pali Rohár wrote: > On Wednesday 04 January 2017 10:05:22 Benjamin Tissoires wrote: > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > On Tuesday 03 January 2017 12:59:37 Dmitry Torokhov wrote: > > > > On Tue, Jan 03, 2017 at 09:39:13PM +0100, Pali Rohár wrote: > > > > > On Tuesday 03 January 2017 21:24:18 Dmitry Torokhov wrote: > > > > > > On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > > > > > > > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > > > > > > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > > > > > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > > > > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał > > > > > > > > > > > > > > > > Kępień > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > > > > > > > 0x29. That i2c address is not specified in > > > > > > > > > > > > > > > > > > DMI or ACPI, so runtime detection without > > > > > > > > > > > > > > > > > > whitelist which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > > > > > > > confusing to me at first because there is > > > > > > > > > > > > > > > > > already an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is > > > > > > > > > > > > > > > > > that: > > > > > > > > > > > > > > > > > * the purpose of this patch is to expose a > > > > > > > > > > > > > > > > > richer interface (as > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on > > > > > > > > > > > > > > > > > some machines, > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > > > > > > > effectively duplicates the work that > > > > > > > > > > > > > > > > > lis3lv02d does). > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > > > > > > > exports /dev/freefall device which notify > > > > > > > > > > > > > > > > userspace about falls. lis3lv02d is i2c driver > > > > > > > > > > > > > > > > which exports axes of accelerometer. Additionaly > > > > > > > > > > > > > > > > lis3lv02d can export also /dev/freefall if > > > > > > > > > > > > > > > > registerer of i2c device provides irq number -- > > > > > > > > > > > > > > > > which is not case of this patch. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and > > > > > > > > > > > > > > > > both are useful. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent > > > > > > > > > > > > > > > > one HW device (that ST microelectronics > > > > > > > > > > > > > > > > accelerometer) but due to complicated HW > > > > > > > > > > > > > > > > abstraction and layers on Dell laptops it is > > > > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed > > > > > > > > > > > > > > > > to lis3lv02d driver and that would export whole > > > > > > > > > > > > > > > > device (with /dev/freefall too), but due to HW > > > > > > > > > > > > > > > > abstraction it is too much complicated... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > > > > > > > number all the way down to lis3lv02d is to set the > > > > > > > > > > > > > > > irq field of the struct i2c_board_info you are > > > > > > > > > > > > > > > passing to i2c_new_device(). And you can extract > > > > > > > > > > > > > > > that IRQ number e.g. in > > > > > > > > > > > > > > > check_acpi_smo88xx_device(). However, you would > > > > > > > > > > > > > > > then need to make sure dell-smo8800 does not > > > > > > > > > > > > > > > attempt to request the same IRQ on whitelisted > > > > > > > > > > > > > > > machines. This got me thinking about a way to > > > > > > > > > > > > > > > somehow incorporate your changes into dell-smo8800 > > > > > > > > > > > > > > > using Wolfram's bus_notifier suggestion, but I do > > > > > > > > > > > > > > > not have a working solution for now. What is > > > > > > > > > > > > > > > tempting about this approach is that you would not > > > > > > > > > > > > > > > have to scan the ACPI namespace in search of > > > > > > > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > > > > > > > driver... And there can be some pathological > > > > > > > > > > > > > > situation (thanks to adding ACPI layer as Andy > > > > > > > > > > > > > > pointed) that there will be more SMO88xx devices in > > > > > > > > > > > > > > ACPI. Plus you can compile kernel with and without > > > > > > > > > > > > > > those modules and also you can blacklist loading > > > > > > > > > > > > > > them (so compile time check is not enough). And > > > > > > > > > > > > > > still some correct message notifier must be used. > > > > > > > > > > > > > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > > > > > > > and available dell devices... > > > > > > > > > > > > > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to > > > > > > > > > > > > > find a nice way to implement this. > > > > > > > > > > > > > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > > > > > > > events, not alert") which breaks your patch. The > > > > > > > > > > > > > reason for that is that lis3lv02d relies on the i2c > > > > > > > > > > > > > client's IRQ being 0 to detect that it should not > > > > > > > > > > > > > create /dev/freefall. > > > > > > > > > > > > > > > > > > > > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > > > > > > > > > > > > > > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > > > > > > > causing lis3lv02d to create /dev/freefall, which in > > > > > > > > > > > > > turn conflicts with dell-smo8800 which is trying to > > > > > > > > > > > > > create /dev/freefall itself. > > > > > > > > > > > > > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > > > > > > > > > > > > > Apologies for that. > > > > > > > > > > > > > > > > > > > > > > I could easily fix this by adding a kernel API to know > > > > > > > > > > > whether the provided irq is from Host Notify or if it was > > > > > > > > > > > coming from an actual declaration. However, I have no idea > > > > > > > > > > > how many other drivers would require this (hopefully only > > > > > > > > > > > this one). > > > > > > > > > > > > > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ > > > > > > > > > > > and let the actual drivers that need it to set it, but > > > > > > > > > > > this was not the best solution according to Dmitri. On my > > > > > > > > > > > side, I am not entirely against this given that it's a > > > > > > > > > > > chip feature, so the driver should be able to know that > > > > > > > > > > > it's available. > > > > > > > > > > > > > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > > > > > > > > > > > > > I read this: > > > > > > > > > > > > > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > > > > > > (that ST microelectronics accelerometer) but due to > > > > > > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > > > > > > handled by two drivers, one ACPI and one i2c." > > > > > > > > > > > > > > > > > > > > and that is the core of the issue. You have 2 drivers > > > > > > > > > > fighting over the same device. Fix this and it will all > > > > > > > > > > work. > > > > > > > > > > > > > > > > > > With my current implementation (which I sent in this patch), > > > > > > > > > they are not fighting. > > > > > > > > > > > > > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and > > > > > > > > > lis3lv02d only accelerometer device as lis3lv02d driver does > > > > > > > > > not get IRQ number in platform data. > > > > > > > > > > > > > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and > > > > > > > > > > accesses it via ACPI methods, can the same be done for Dell? > > > > > > > > > > > > > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI > > > > > > > > > or DMI is even not i2c address of device, so it needs to be > > > > > > > > > specified in code itself. > > > > > > > > > > > > > > > > > > Really there is no other way... :-( > > > > > > > > > > > > > > > > Sure there is: > > > > > > > > > > > > > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > > > > > > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > > > > > > > simply forward requests to dell-smo8800-accel i2c client. > > > > > > > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel > > > > > > > > does. > > > > > > > > > > > > > > Sorry, but I do not understand how you mean it... Why to provides > > > > > > > new read/write i2c functions which are already implemented by > > > > > > > i2c-i801 bus and lis3lv02d i2c driver? > > > > > > > > > > > > Because that would allow you to avoid clashes with i2c creating > > > > > > interrupt mapping for client residing on host-notify-capable > > > > > > controller. > > > > > > > > > > > > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > > > > > > > > > > > > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > > > > > > > device. > > > > > > > > > > > > > > But... what is problem with current implementation? Accelerometer > > > > > > > HW provides two functions: > > > > > > > > > > > > > > 1) 3 axes reports > > > > > > > 2) Disk freefall detection > > > > > > > > > > > > > > And 1) is handled by i2c driver lis3lv02d and 2) is by > > > > > > > dell-smo8800. Both functions are independent here. > > > > > > > > > > > > > > I think you just trying to complicate this situation even more to > > > > > > > be more complicated as currently is. > > > > > > > > > > > > Because this apparently does not work for you, does it? > > > > > > > > > > It is working fine. I do not see any problem. > > > > > > > > > > > In general, > > > > > > if you want the same hardware be handled by 2 different drivers you > > > > > > are going to have bad time. > > > > > > > > > > Yes, but in this case half of device is ACPI based and other half i2c > > > > > based. This is problem of ACPI and Dell design. > > > > > > > > > > > It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are > > > > > > the same, right? > > > > > > > > > > Yes. I understand that clean solution is to have one driver which > > > > > provides everything. > > > > > > > > > > But because half of data are ACPI and half i2c, you still needs to > > > > > create two drivers (one ACPI and one i2c). You can put both drivers into > > > > > one .ko module, but still these will be two drivers due to how ACPI and > > > > > i2c linux abstractions are different. > > > > > > > > > > > So, instead of having 2 drivers split the > > > > > > functionality, can you forego registering smo8800 ACPI driver on > > > > > > your whitelisted boxes and instead instantiate full i2c client > > > > > > device with properly assigned both address and IRQ and let lis3lv02d > > > > > > handle it (providing both accelerometer data and /dev/freefall)? > > > > > > > > > > With Michał we already discussed about it, see emails. Basically you can > > > > > enable/disable kernel modules at compile time or blacklist at runtime > > > > > (or even chose what will be compiled into vmlinux and what as external > > > > > .ko module). > > > > > > > > This can be solved with a bit of Kconfig/IS_ENABLED() code. > > > > > > > > > Some distributions blacklist i2c-i801.ko module... And > > > > > > > > Any particular reason for that? > > > > > > > > > there can be also problem with initialization of i2c-i801 driver (fix is > > > > > in commit a7ae81952cda, but does not have to work at every time!). So > > > > > that move on whitelisted machines can potentially cause disappearance of > > > > > /dev/freefall and users will not have hdd protection which is currently > > > > > working. > > > > > > > > Well, I gave you 2 possible solutions (roll your own i2c read/write, > > > > forward them to i2c client) or have faith in your implementation and let > > > > lis3lv02d handle it. > > > > > > > > The 3rd one is to possibly add a flag to disable host notify to IRQ > > > > mapping for given client (if Wolfram/Jean OK with it). > > > > > > > > Oh, the 4th one: change the irq in lis3lv02d.h to be "int" and change > > > > the check in lis3lv02d.c to be "lis->irq <= 0" and instantiate your > > > > i2c_client with board_info->irq = -1. > > > > > > > > Pick whichever you prefer. > > > > > > > > By the way, what do you need accelerometer for on these devices? They > > > > don't appear to be tablets that could use one... > > > > > > Ah, you are talking about problem that after 4d5538f5882a lis3lv02d will > > > not work... I thought that discussion is about different mechanism how > > > to implement bus registration notification to smo8800 driver (or > > > different solution to not have registration in i801). > > > > > > > Just because I am not sure I got everything right, could you confirm > > that: > > - in the current upstream tree, the dell-smo8800 driver is now broken > > after 4d5538f5882a (i2c: use an IRQ to report Host Notify events, not > > alert) > > No, dell-smo8800 it is working fine. It is fully independent from i2c > and lis3lv02d. It is pure ACPI driver which does not share anything with > i2c. > > > - this series adds an extra lis3lv02d on some machines and you have > > problem fighting for the irq (but this is not upstream yet). > > Yes, this series (not merged yet) adds extra lis3lv02d device but is not > working because of 4d5538f5882a. > > > The extra lis3lv02d node is added from dell-smo8800 > > No, dell-smo8800 does not add new node in this patch. > > > If the first point is not correct (by default, dell-smo8800 will not be > > loaded at the same time than lis3lv02d), then it's a design issue with > > the interactions between those 2 drivers. > > No, there is no interactions between these two drivers (dell-smo8800 and > lis3lv02d). dell-smo8800 is pure ACPI driver and exports just > /dev/freefall device based on IRQ (and nothing more). > > And lis3lv02d in *current* configuration in this patch exports only > accelerometer input device, not /dev/freefall. It does not use IRQ. > (Just there is problem with 4d5538f5882a which tells lis3lv02d IRQ > number which is not freefall report, therefore lis3lv02d does not work). > > To make it clear, ST Accelerometer provides two operations: > * report free fall > * report 3 axes > > Free fall is reported by IRQ, state of 3 axes via i2c bus. Free fall IRQ > is handled by dell-smo8800, state of 3 axes via i2c lis3lv02d driver. > > lis3lv02d can handle also free fall IRQ is platform i2c data provides > IRQ number for it -- but this is not case in our Dell configuration. But > commit 4d5538f5882a inject some IRQ number to lis3lv02d driver which is > not free fall detection and so is breaking lis3lv02 driver. In our Dell > configuration (by this patch) there should be no IRQ number. > > It is clear now? I think I am starting to understand the problem. Currently (upstream tree), 4d5538f5882a doesn't break anything. On the mentioned Dell platforms, the dell-smo8800 gets loaded and provides /dev/freefall. lis3lv02d is not loaded so everything just works. The problem comes from this patch which doesn't set the irq in i2c_board_info and so i2c_core sets the IRQ to Host Notify. I think Dmitri already gave you the solution: set the irq to -1 (or -ENOENT) in i2c_board_info in your patch and only forward it to lis3lv02d_init_device() only if it is positive (valid). So, to me 4d5538f5882a doesn't introduce a regression so we should keep it in its current state. Cheers, Benjamin > > > If the first point is correct because ACPI declares both devices, then > > there is an urgent fix to propose to not enable Host Notify by default > > on Host Notifier capable adapters. (even though the design between the > > 2 drivers is wrong, it's considered as a regression). > > > > Cheers, > > Benjamin > > -- > Pali Rohár > pali.rohar@gmail.com -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue, 3 Jan 2017 10:06:41 +0100, Benjamin Tissoires wrote: > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > So 4d5538f5882a is breaking lis3lv02d driver... > > Apologies for that. > > I could easily fix this by adding a kernel API to know whether the > provided irq is from Host Notify or if it was coming from an actual > declaration. However, I have no idea how many other drivers would > require this (hopefully only this one). > > One other solution would be to reserve the Host Notify IRQ and let the > actual drivers that need it to set it, but this was not the best > solution according to Dmitri. On my side, I am not entirely against this > given that it's a chip feature, so the driver should be able to know > that it's available. > > Dmitri, Wolfram, Jean, any preferences? I did not look at your work so I do not feel qualified to give advice on the proper way. I trust you to make the right decision :-)
On Wednesday 04 January 2017 11:13:06 Benjamin Tissoires wrote: > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > On Wednesday 04 January 2017 10:05:22 Benjamin Tissoires wrote: > > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > > On Tuesday 03 January 2017 12:59:37 Dmitry Torokhov wrote: > > > > > On Tue, Jan 03, 2017 at 09:39:13PM +0100, Pali Rohár wrote: > > > > > > On Tuesday 03 January 2017 21:24:18 Dmitry Torokhov wrote: > > > > > > > On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > > > > > > > > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > > > > > > > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > > > > > > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > > > > > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > > > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał > > > > > > > > > > > > > > > > > Kępień > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > > > > > > > > 0x29. That i2c address is not specified in > > > > > > > > > > > > > > > > > > > DMI or ACPI, so runtime detection without > > > > > > > > > > > > > > > > > > > whitelist which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > > > > > > > > confusing to me at first because there is > > > > > > > > > > > > > > > > > > already an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is > > > > > > > > > > > > > > > > > > that: > > > > > > > > > > > > > > > > > > * the purpose of this patch is to expose a > > > > > > > > > > > > > > > > > > richer interface (as > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on > > > > > > > > > > > > > > > > > > some machines, > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > > > > > > > > effectively duplicates the work that > > > > > > > > > > > > > > > > > > lis3lv02d does). > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > > > > > > > > exports /dev/freefall device which notify > > > > > > > > > > > > > > > > > userspace about falls. lis3lv02d is i2c driver > > > > > > > > > > > > > > > > > which exports axes of accelerometer. Additionaly > > > > > > > > > > > > > > > > > lis3lv02d can export also /dev/freefall if > > > > > > > > > > > > > > > > > registerer of i2c device provides irq number -- > > > > > > > > > > > > > > > > > which is not case of this patch. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and > > > > > > > > > > > > > > > > > both are useful. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent > > > > > > > > > > > > > > > > > one HW device (that ST microelectronics > > > > > > > > > > > > > > > > > accelerometer) but due to complicated HW > > > > > > > > > > > > > > > > > abstraction and layers on Dell laptops it is > > > > > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed > > > > > > > > > > > > > > > > > to lis3lv02d driver and that would export whole > > > > > > > > > > > > > > > > > device (with /dev/freefall too), but due to HW > > > > > > > > > > > > > > > > > abstraction it is too much complicated... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > > > > > > > > number all the way down to lis3lv02d is to set the > > > > > > > > > > > > > > > > irq field of the struct i2c_board_info you are > > > > > > > > > > > > > > > > passing to i2c_new_device(). And you can extract > > > > > > > > > > > > > > > > that IRQ number e.g. in > > > > > > > > > > > > > > > > check_acpi_smo88xx_device(). However, you would > > > > > > > > > > > > > > > > then need to make sure dell-smo8800 does not > > > > > > > > > > > > > > > > attempt to request the same IRQ on whitelisted > > > > > > > > > > > > > > > > machines. This got me thinking about a way to > > > > > > > > > > > > > > > > somehow incorporate your changes into dell-smo8800 > > > > > > > > > > > > > > > > using Wolfram's bus_notifier suggestion, but I do > > > > > > > > > > > > > > > > not have a working solution for now. What is > > > > > > > > > > > > > > > > tempting about this approach is that you would not > > > > > > > > > > > > > > > > have to scan the ACPI namespace in search of > > > > > > > > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > > > > > > > > driver... And there can be some pathological > > > > > > > > > > > > > > > situation (thanks to adding ACPI layer as Andy > > > > > > > > > > > > > > > pointed) that there will be more SMO88xx devices in > > > > > > > > > > > > > > > ACPI. Plus you can compile kernel with and without > > > > > > > > > > > > > > > those modules and also you can blacklist loading > > > > > > > > > > > > > > > them (so compile time check is not enough). And > > > > > > > > > > > > > > > still some correct message notifier must be used. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > > > > > > > > and available dell devices... > > > > > > > > > > > > > > > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to > > > > > > > > > > > > > > find a nice way to implement this. > > > > > > > > > > > > > > > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > > > > > > > > events, not alert") which breaks your patch. The > > > > > > > > > > > > > > reason for that is that lis3lv02d relies on the i2c > > > > > > > > > > > > > > client's IRQ being 0 to detect that it should not > > > > > > > > > > > > > > create /dev/freefall. > > > > > > > > > > > > > > > > > > > > > > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > > > > > > > > > > > > > > > > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > > > > > > > > causing lis3lv02d to create /dev/freefall, which in > > > > > > > > > > > > > > turn conflicts with dell-smo8800 which is trying to > > > > > > > > > > > > > > create /dev/freefall itself. > > > > > > > > > > > > > > > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > > > > > > > > > > > > > > > Apologies for that. > > > > > > > > > > > > > > > > > > > > > > > > I could easily fix this by adding a kernel API to know > > > > > > > > > > > > whether the provided irq is from Host Notify or if it was > > > > > > > > > > > > coming from an actual declaration. However, I have no idea > > > > > > > > > > > > how many other drivers would require this (hopefully only > > > > > > > > > > > > this one). > > > > > > > > > > > > > > > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ > > > > > > > > > > > > and let the actual drivers that need it to set it, but > > > > > > > > > > > > this was not the best solution according to Dmitri. On my > > > > > > > > > > > > side, I am not entirely against this given that it's a > > > > > > > > > > > > chip feature, so the driver should be able to know that > > > > > > > > > > > > it's available. > > > > > > > > > > > > > > > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > > > > > > > > > > > > > > > I read this: > > > > > > > > > > > > > > > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > > > > > > > (that ST microelectronics accelerometer) but due to > > > > > > > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > > > > > > > handled by two drivers, one ACPI and one i2c." > > > > > > > > > > > > > > > > > > > > > > and that is the core of the issue. You have 2 drivers > > > > > > > > > > > fighting over the same device. Fix this and it will all > > > > > > > > > > > work. > > > > > > > > > > > > > > > > > > > > With my current implementation (which I sent in this patch), > > > > > > > > > > they are not fighting. > > > > > > > > > > > > > > > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and > > > > > > > > > > lis3lv02d only accelerometer device as lis3lv02d driver does > > > > > > > > > > not get IRQ number in platform data. > > > > > > > > > > > > > > > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and > > > > > > > > > > > accesses it via ACPI methods, can the same be done for Dell? > > > > > > > > > > > > > > > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI > > > > > > > > > > or DMI is even not i2c address of device, so it needs to be > > > > > > > > > > specified in code itself. > > > > > > > > > > > > > > > > > > > > Really there is no other way... :-( > > > > > > > > > > > > > > > > > > Sure there is: > > > > > > > > > > > > > > > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > > > > > > > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > > > > > > > > simply forward requests to dell-smo8800-accel i2c client. > > > > > > > > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel > > > > > > > > > does. > > > > > > > > > > > > > > > > Sorry, but I do not understand how you mean it... Why to provides > > > > > > > > new read/write i2c functions which are already implemented by > > > > > > > > i2c-i801 bus and lis3lv02d i2c driver? > > > > > > > > > > > > > > Because that would allow you to avoid clashes with i2c creating > > > > > > > interrupt mapping for client residing on host-notify-capable > > > > > > > controller. > > > > > > > > > > > > > > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > > > > > > > > > > > > > > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > > > > > > > > device. > > > > > > > > > > > > > > > > But... what is problem with current implementation? Accelerometer > > > > > > > > HW provides two functions: > > > > > > > > > > > > > > > > 1) 3 axes reports > > > > > > > > 2) Disk freefall detection > > > > > > > > > > > > > > > > And 1) is handled by i2c driver lis3lv02d and 2) is by > > > > > > > > dell-smo8800. Both functions are independent here. > > > > > > > > > > > > > > > > I think you just trying to complicate this situation even more to > > > > > > > > be more complicated as currently is. > > > > > > > > > > > > > > Because this apparently does not work for you, does it? > > > > > > > > > > > > It is working fine. I do not see any problem. > > > > > > > > > > > > > In general, > > > > > > > if you want the same hardware be handled by 2 different drivers you > > > > > > > are going to have bad time. > > > > > > > > > > > > Yes, but in this case half of device is ACPI based and other half i2c > > > > > > based. This is problem of ACPI and Dell design. > > > > > > > > > > > > > It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are > > > > > > > the same, right? > > > > > > > > > > > > Yes. I understand that clean solution is to have one driver which > > > > > > provides everything. > > > > > > > > > > > > But because half of data are ACPI and half i2c, you still needs to > > > > > > create two drivers (one ACPI and one i2c). You can put both drivers into > > > > > > one .ko module, but still these will be two drivers due to how ACPI and > > > > > > i2c linux abstractions are different. > > > > > > > > > > > > > So, instead of having 2 drivers split the > > > > > > > functionality, can you forego registering smo8800 ACPI driver on > > > > > > > your whitelisted boxes and instead instantiate full i2c client > > > > > > > device with properly assigned both address and IRQ and let lis3lv02d > > > > > > > handle it (providing both accelerometer data and /dev/freefall)? > > > > > > > > > > > > With Michał we already discussed about it, see emails. Basically you can > > > > > > enable/disable kernel modules at compile time or blacklist at runtime > > > > > > (or even chose what will be compiled into vmlinux and what as external > > > > > > .ko module). > > > > > > > > > > This can be solved with a bit of Kconfig/IS_ENABLED() code. > > > > > > > > > > > Some distributions blacklist i2c-i801.ko module... And > > > > > > > > > > Any particular reason for that? > > > > > > > > > > > there can be also problem with initialization of i2c-i801 driver (fix is > > > > > > in commit a7ae81952cda, but does not have to work at every time!). So > > > > > > that move on whitelisted machines can potentially cause disappearance of > > > > > > /dev/freefall and users will not have hdd protection which is currently > > > > > > working. > > > > > > > > > > Well, I gave you 2 possible solutions (roll your own i2c read/write, > > > > > forward them to i2c client) or have faith in your implementation and let > > > > > lis3lv02d handle it. > > > > > > > > > > The 3rd one is to possibly add a flag to disable host notify to IRQ > > > > > mapping for given client (if Wolfram/Jean OK with it). > > > > > > > > > > Oh, the 4th one: change the irq in lis3lv02d.h to be "int" and change > > > > > the check in lis3lv02d.c to be "lis->irq <= 0" and instantiate your > > > > > i2c_client with board_info->irq = -1. > > > > > > > > > > Pick whichever you prefer. > > > > > > > > > > By the way, what do you need accelerometer for on these devices? They > > > > > don't appear to be tablets that could use one... > > > > > > > > Ah, you are talking about problem that after 4d5538f5882a lis3lv02d will > > > > not work... I thought that discussion is about different mechanism how > > > > to implement bus registration notification to smo8800 driver (or > > > > different solution to not have registration in i801). > > > > > > > > > > Just because I am not sure I got everything right, could you confirm > > > that: > > > - in the current upstream tree, the dell-smo8800 driver is now broken > > > after 4d5538f5882a (i2c: use an IRQ to report Host Notify events, not > > > alert) > > > > No, dell-smo8800 it is working fine. It is fully independent from i2c > > and lis3lv02d. It is pure ACPI driver which does not share anything with > > i2c. > > > > > - this series adds an extra lis3lv02d on some machines and you have > > > problem fighting for the irq (but this is not upstream yet). > > > > Yes, this series (not merged yet) adds extra lis3lv02d device but is not > > working because of 4d5538f5882a. > > > > > The extra lis3lv02d node is added from dell-smo8800 > > > > No, dell-smo8800 does not add new node in this patch. > > > > > If the first point is not correct (by default, dell-smo8800 will not be > > > loaded at the same time than lis3lv02d), then it's a design issue with > > > the interactions between those 2 drivers. > > > > No, there is no interactions between these two drivers (dell-smo8800 and > > lis3lv02d). dell-smo8800 is pure ACPI driver and exports just > > /dev/freefall device based on IRQ (and nothing more). > > > > And lis3lv02d in *current* configuration in this patch exports only > > accelerometer input device, not /dev/freefall. It does not use IRQ. > > (Just there is problem with 4d5538f5882a which tells lis3lv02d IRQ > > number which is not freefall report, therefore lis3lv02d does not work). > > > > To make it clear, ST Accelerometer provides two operations: > > * report free fall > > * report 3 axes > > > > Free fall is reported by IRQ, state of 3 axes via i2c bus. Free fall IRQ > > is handled by dell-smo8800, state of 3 axes via i2c lis3lv02d driver. > > > > lis3lv02d can handle also free fall IRQ is platform i2c data provides > > IRQ number for it -- but this is not case in our Dell configuration. But > > commit 4d5538f5882a inject some IRQ number to lis3lv02d driver which is > > not free fall detection and so is breaking lis3lv02 driver. In our Dell > > configuration (by this patch) there should be no IRQ number. > > > > It is clear now? > > I think I am starting to understand the problem. > > Currently (upstream tree), 4d5538f5882a doesn't break anything. I cannot prove it... lis3lv02d i2c driver itself uses this IRQ logic (missing = no freefall) and there could be already hardware/boards which register lis3lv02d i2c device without IRQ number. And definitions could be also in device tree and so outside of kernel sources... So there can be potentially break. But at least not case of Dell. > On the > mentioned Dell platforms, the dell-smo8800 gets loaded and provides > /dev/freefall. lis3lv02d is not loaded so everything just works. Yes. > The problem comes from this patch which doesn't set the irq in > i2c_board_info and so i2c_core sets the IRQ to Host Notify. I think > Dmitri already gave you the solution: set the irq to -1 (or -ENOENT) > in i2c_board_info in your patch and only forward it to > lis3lv02d_init_device() only if it is positive (valid). Now I understood that Dmitri means. For this Dell platforms it is OK, but we have a problem for device tree platforms. Normally in device tree if device does not support something you just do not specify it. But with this approach you need to explicitly specify IRQ to -1 in device tree. And I think this is really not a clean solution. > So, to me 4d5538f5882a doesn't introduce a regression so we should keep > it in its current state. > > Cheers, > Benjamin > > > > > > If the first point is correct because ACPI declares both devices, then > > > there is an urgent fix to propose to not enable Host Notify by default > > > on Host Notifier capable adapters. (even though the design between the > > > 2 drivers is wrong, it's considered as a regression). > > > > > > Cheers, > > > Benjamin > > > > -- > > Pali Rohár > > pali.rohar@gmail.com
On Jan 04 2017 or thereabouts, Andy Shevchenko wrote: > On Tue, Jan 3, 2017 at 10:39 PM, Pali Rohár <pali.rohar@gmail.com> wrote: > > > > With Michał we already discussed about it, see emails. Basically you can > > enable/disable kernel modules at compile time or blacklist at runtime > > (or even chose what will be compiled into vmlinux and what as external > > .ko module). Some distributions blacklist i2c-i801.ko module... > > But you understand that any of compile/not compile is not an option, right? > The case which we face will be both of them, if possible, will be > compiled as modules. > > Blacklisting means making your problem the actual user's one. Not good. > > > And > > there can be also problem with initialization of i2c-i801 driver (fix is > > in commit a7ae81952cda, but does not have to work at every time!). So > > that move on whitelisted machines can potentially cause disappearance of > > /dev/freefall and users will not have hdd protection which is currently > > working. > I am seeing the same issues with psmouse and SMBus touchpads. The PS/2 device knows about the availability of a better but unlisted device at the ACPI level. The way I solved this to not have to deal with compile/not compile and runtime errors is the same way Wolfram told you about: bus notifiers. I also use an intermediate platform driver to not add i2c dependency on psmouse. For you the solution would be: - In dell-smo8800, after checking the whitelist, add a platform driver "dell-lis3lv02d-platform", and add in the platform_data the I2C address of the chip. - create a new driver dell-lis3lv02d-platform.ko which listens for the i2c bus creation and registers the lis3lv02d I2C node when it sees a matching adapter. (see [1] for my solution) - in dell-lis3lv02d-platform.ko make sure to set the irq to -ENOENT so that lis3lv02d.ko doesn't create /dev/freefall which will still be handled by ACPI. How does that sound? Cheers, Benjamin [1] https://github.com/bentiss/linux/blob/synaptics-rmi4-v4.9-rc7+/drivers/input/rmi4/rmi_platform.c -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Jan 04 2017 or thereabouts, Pali Rohár wrote: > On Wednesday 04 January 2017 11:13:06 Benjamin Tissoires wrote: > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > On Wednesday 04 January 2017 10:05:22 Benjamin Tissoires wrote: > > > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > > > On Tuesday 03 January 2017 12:59:37 Dmitry Torokhov wrote: > > > > > > On Tue, Jan 03, 2017 at 09:39:13PM +0100, Pali Rohár wrote: > > > > > > > On Tuesday 03 January 2017 21:24:18 Dmitry Torokhov wrote: > > > > > > > > On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > > > > > > > > > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > > > > > > > > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > > > > > > > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > > > > > > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > > > > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał > > > > > > > > > > > > > > > > > > Kępień > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > > > > > > > > > 0x29. That i2c address is not specified in > > > > > > > > > > > > > > > > > > > > DMI or ACPI, so runtime detection without > > > > > > > > > > > > > > > > > > > > whitelist which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > > > > > > > > > confusing to me at first because there is > > > > > > > > > > > > > > > > > > > already an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is > > > > > > > > > > > > > > > > > > > that: > > > > > > > > > > > > > > > > > > > * the purpose of this patch is to expose a > > > > > > > > > > > > > > > > > > > richer interface (as > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on > > > > > > > > > > > > > > > > > > > some machines, > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > > > > > > > > > effectively duplicates the work that > > > > > > > > > > > > > > > > > > > lis3lv02d does). > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > > > > > > > > > exports /dev/freefall device which notify > > > > > > > > > > > > > > > > > > userspace about falls. lis3lv02d is i2c driver > > > > > > > > > > > > > > > > > > which exports axes of accelerometer. Additionaly > > > > > > > > > > > > > > > > > > lis3lv02d can export also /dev/freefall if > > > > > > > > > > > > > > > > > > registerer of i2c device provides irq number -- > > > > > > > > > > > > > > > > > > which is not case of this patch. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and > > > > > > > > > > > > > > > > > > both are useful. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent > > > > > > > > > > > > > > > > > > one HW device (that ST microelectronics > > > > > > > > > > > > > > > > > > accelerometer) but due to complicated HW > > > > > > > > > > > > > > > > > > abstraction and layers on Dell laptops it is > > > > > > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed > > > > > > > > > > > > > > > > > > to lis3lv02d driver and that would export whole > > > > > > > > > > > > > > > > > > device (with /dev/freefall too), but due to HW > > > > > > > > > > > > > > > > > > abstraction it is too much complicated... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > > > > > > > > > number all the way down to lis3lv02d is to set the > > > > > > > > > > > > > > > > > irq field of the struct i2c_board_info you are > > > > > > > > > > > > > > > > > passing to i2c_new_device(). And you can extract > > > > > > > > > > > > > > > > > that IRQ number e.g. in > > > > > > > > > > > > > > > > > check_acpi_smo88xx_device(). However, you would > > > > > > > > > > > > > > > > > then need to make sure dell-smo8800 does not > > > > > > > > > > > > > > > > > attempt to request the same IRQ on whitelisted > > > > > > > > > > > > > > > > > machines. This got me thinking about a way to > > > > > > > > > > > > > > > > > somehow incorporate your changes into dell-smo8800 > > > > > > > > > > > > > > > > > using Wolfram's bus_notifier suggestion, but I do > > > > > > > > > > > > > > > > > not have a working solution for now. What is > > > > > > > > > > > > > > > > > tempting about this approach is that you would not > > > > > > > > > > > > > > > > > have to scan the ACPI namespace in search of > > > > > > > > > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > > > > > > > > > driver... And there can be some pathological > > > > > > > > > > > > > > > > situation (thanks to adding ACPI layer as Andy > > > > > > > > > > > > > > > > pointed) that there will be more SMO88xx devices in > > > > > > > > > > > > > > > > ACPI. Plus you can compile kernel with and without > > > > > > > > > > > > > > > > those modules and also you can blacklist loading > > > > > > > > > > > > > > > > them (so compile time check is not enough). And > > > > > > > > > > > > > > > > still some correct message notifier must be used. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > > > > > > > > > and available dell devices... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to > > > > > > > > > > > > > > > find a nice way to implement this. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > > > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > > > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > > > > > > > > > events, not alert") which breaks your patch. The > > > > > > > > > > > > > > > reason for that is that lis3lv02d relies on the i2c > > > > > > > > > > > > > > > client's IRQ being 0 to detect that it should not > > > > > > > > > > > > > > > create /dev/freefall. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > > > > > > > > > causing lis3lv02d to create /dev/freefall, which in > > > > > > > > > > > > > > > turn conflicts with dell-smo8800 which is trying to > > > > > > > > > > > > > > > create /dev/freefall itself. > > > > > > > > > > > > > > > > > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > > > > > > > > > > > > > > > > > Apologies for that. > > > > > > > > > > > > > > > > > > > > > > > > > > I could easily fix this by adding a kernel API to know > > > > > > > > > > > > > whether the provided irq is from Host Notify or if it was > > > > > > > > > > > > > coming from an actual declaration. However, I have no idea > > > > > > > > > > > > > how many other drivers would require this (hopefully only > > > > > > > > > > > > > this one). > > > > > > > > > > > > > > > > > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ > > > > > > > > > > > > > and let the actual drivers that need it to set it, but > > > > > > > > > > > > > this was not the best solution according to Dmitri. On my > > > > > > > > > > > > > side, I am not entirely against this given that it's a > > > > > > > > > > > > > chip feature, so the driver should be able to know that > > > > > > > > > > > > > it's available. > > > > > > > > > > > > > > > > > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > > > > > > > > > > > > > > > > > I read this: > > > > > > > > > > > > > > > > > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > > > > > > > > (that ST microelectronics accelerometer) but due to > > > > > > > > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c." > > > > > > > > > > > > > > > > > > > > > > > > and that is the core of the issue. You have 2 drivers > > > > > > > > > > > > fighting over the same device. Fix this and it will all > > > > > > > > > > > > work. > > > > > > > > > > > > > > > > > > > > > > With my current implementation (which I sent in this patch), > > > > > > > > > > > they are not fighting. > > > > > > > > > > > > > > > > > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and > > > > > > > > > > > lis3lv02d only accelerometer device as lis3lv02d driver does > > > > > > > > > > > not get IRQ number in platform data. > > > > > > > > > > > > > > > > > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and > > > > > > > > > > > > accesses it via ACPI methods, can the same be done for Dell? > > > > > > > > > > > > > > > > > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI > > > > > > > > > > > or DMI is even not i2c address of device, so it needs to be > > > > > > > > > > > specified in code itself. > > > > > > > > > > > > > > > > > > > > > > Really there is no other way... :-( > > > > > > > > > > > > > > > > > > > > Sure there is: > > > > > > > > > > > > > > > > > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > > > > > > > > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > > > > > > > > > simply forward requests to dell-smo8800-accel i2c client. > > > > > > > > > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel > > > > > > > > > > does. > > > > > > > > > > > > > > > > > > Sorry, but I do not understand how you mean it... Why to provides > > > > > > > > > new read/write i2c functions which are already implemented by > > > > > > > > > i2c-i801 bus and lis3lv02d i2c driver? > > > > > > > > > > > > > > > > Because that would allow you to avoid clashes with i2c creating > > > > > > > > interrupt mapping for client residing on host-notify-capable > > > > > > > > controller. > > > > > > > > > > > > > > > > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > > > > > > > > > > > > > > > > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > > > > > > > > > device. > > > > > > > > > > > > > > > > > > But... what is problem with current implementation? Accelerometer > > > > > > > > > HW provides two functions: > > > > > > > > > > > > > > > > > > 1) 3 axes reports > > > > > > > > > 2) Disk freefall detection > > > > > > > > > > > > > > > > > > And 1) is handled by i2c driver lis3lv02d and 2) is by > > > > > > > > > dell-smo8800. Both functions are independent here. > > > > > > > > > > > > > > > > > > I think you just trying to complicate this situation even more to > > > > > > > > > be more complicated as currently is. > > > > > > > > > > > > > > > > Because this apparently does not work for you, does it? > > > > > > > > > > > > > > It is working fine. I do not see any problem. > > > > > > > > > > > > > > > In general, > > > > > > > > if you want the same hardware be handled by 2 different drivers you > > > > > > > > are going to have bad time. > > > > > > > > > > > > > > Yes, but in this case half of device is ACPI based and other half i2c > > > > > > > based. This is problem of ACPI and Dell design. > > > > > > > > > > > > > > > It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are > > > > > > > > the same, right? > > > > > > > > > > > > > > Yes. I understand that clean solution is to have one driver which > > > > > > > provides everything. > > > > > > > > > > > > > > But because half of data are ACPI and half i2c, you still needs to > > > > > > > create two drivers (one ACPI and one i2c). You can put both drivers into > > > > > > > one .ko module, but still these will be two drivers due to how ACPI and > > > > > > > i2c linux abstractions are different. > > > > > > > > > > > > > > > So, instead of having 2 drivers split the > > > > > > > > functionality, can you forego registering smo8800 ACPI driver on > > > > > > > > your whitelisted boxes and instead instantiate full i2c client > > > > > > > > device with properly assigned both address and IRQ and let lis3lv02d > > > > > > > > handle it (providing both accelerometer data and /dev/freefall)? > > > > > > > > > > > > > > With Michał we already discussed about it, see emails. Basically you can > > > > > > > enable/disable kernel modules at compile time or blacklist at runtime > > > > > > > (or even chose what will be compiled into vmlinux and what as external > > > > > > > .ko module). > > > > > > > > > > > > This can be solved with a bit of Kconfig/IS_ENABLED() code. > > > > > > > > > > > > > Some distributions blacklist i2c-i801.ko module... And > > > > > > > > > > > > Any particular reason for that? > > > > > > > > > > > > > there can be also problem with initialization of i2c-i801 driver (fix is > > > > > > > in commit a7ae81952cda, but does not have to work at every time!). So > > > > > > > that move on whitelisted machines can potentially cause disappearance of > > > > > > > /dev/freefall and users will not have hdd protection which is currently > > > > > > > working. > > > > > > > > > > > > Well, I gave you 2 possible solutions (roll your own i2c read/write, > > > > > > forward them to i2c client) or have faith in your implementation and let > > > > > > lis3lv02d handle it. > > > > > > > > > > > > The 3rd one is to possibly add a flag to disable host notify to IRQ > > > > > > mapping for given client (if Wolfram/Jean OK with it). > > > > > > > > > > > > Oh, the 4th one: change the irq in lis3lv02d.h to be "int" and change > > > > > > the check in lis3lv02d.c to be "lis->irq <= 0" and instantiate your > > > > > > i2c_client with board_info->irq = -1. > > > > > > > > > > > > Pick whichever you prefer. > > > > > > > > > > > > By the way, what do you need accelerometer for on these devices? They > > > > > > don't appear to be tablets that could use one... > > > > > > > > > > Ah, you are talking about problem that after 4d5538f5882a lis3lv02d will > > > > > not work... I thought that discussion is about different mechanism how > > > > > to implement bus registration notification to smo8800 driver (or > > > > > different solution to not have registration in i801). > > > > > > > > > > > > > Just because I am not sure I got everything right, could you confirm > > > > that: > > > > - in the current upstream tree, the dell-smo8800 driver is now broken > > > > after 4d5538f5882a (i2c: use an IRQ to report Host Notify events, not > > > > alert) > > > > > > No, dell-smo8800 it is working fine. It is fully independent from i2c > > > and lis3lv02d. It is pure ACPI driver which does not share anything with > > > i2c. > > > > > > > - this series adds an extra lis3lv02d on some machines and you have > > > > problem fighting for the irq (but this is not upstream yet). > > > > > > Yes, this series (not merged yet) adds extra lis3lv02d device but is not > > > working because of 4d5538f5882a. > > > > > > > The extra lis3lv02d node is added from dell-smo8800 > > > > > > No, dell-smo8800 does not add new node in this patch. > > > > > > > If the first point is not correct (by default, dell-smo8800 will not be > > > > loaded at the same time than lis3lv02d), then it's a design issue with > > > > the interactions between those 2 drivers. > > > > > > No, there is no interactions between these two drivers (dell-smo8800 and > > > lis3lv02d). dell-smo8800 is pure ACPI driver and exports just > > > /dev/freefall device based on IRQ (and nothing more). > > > > > > And lis3lv02d in *current* configuration in this patch exports only > > > accelerometer input device, not /dev/freefall. It does not use IRQ. > > > (Just there is problem with 4d5538f5882a which tells lis3lv02d IRQ > > > number which is not freefall report, therefore lis3lv02d does not work). > > > > > > To make it clear, ST Accelerometer provides two operations: > > > * report free fall > > > * report 3 axes > > > > > > Free fall is reported by IRQ, state of 3 axes via i2c bus. Free fall IRQ > > > is handled by dell-smo8800, state of 3 axes via i2c lis3lv02d driver. > > > > > > lis3lv02d can handle also free fall IRQ is platform i2c data provides > > > IRQ number for it -- but this is not case in our Dell configuration. But > > > commit 4d5538f5882a inject some IRQ number to lis3lv02d driver which is > > > not free fall detection and so is breaking lis3lv02 driver. In our Dell > > > configuration (by this patch) there should be no IRQ number. > > > > > > It is clear now? > > > > I think I am starting to understand the problem. > > > > Currently (upstream tree), 4d5538f5882a doesn't break anything. > > I cannot prove it... lis3lv02d i2c driver itself uses this IRQ logic > (missing = no freefall) and there could be already hardware/boards which > register lis3lv02d i2c device without IRQ number. And definitions could > be also in device tree and so outside of kernel sources... > > So there can be potentially break. But at least not case of Dell. > > > On the > > mentioned Dell platforms, the dell-smo8800 gets loaded and provides > > /dev/freefall. lis3lv02d is not loaded so everything just works. > > Yes. > > > The problem comes from this patch which doesn't set the irq in > > i2c_board_info and so i2c_core sets the IRQ to Host Notify. I think > > Dmitri already gave you the solution: set the irq to -1 (or -ENOENT) > > in i2c_board_info in your patch and only forward it to > > lis3lv02d_init_device() only if it is positive (valid). > > Now I understood that Dmitri means. For this Dell platforms it is OK, > but we have a problem for device tree platforms. Normally in device tree > if device does not support something you just do not specify it. But > with this approach you need to explicitly specify IRQ to -1 in device > tree. And I think this is really not a clean solution. > No. DT platforms won't have an issue: they don't change anything, and there will be a new /dev/freefall misc device for the platforms that don't have the irq set *and* if the device is on a Host Notify capable adapter, currently only i2c-i801. But given that they are DT and not ACPI, this won't conflict with dell-smo8800, so there won't be any errors, just a dangling unused device node. This approach is IMO the best if you want to have this in the kernel. Cheers, Benjamin -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wednesday 04 January 2017 11:32:33 Benjamin Tissoires wrote: > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > On Wednesday 04 January 2017 11:13:06 Benjamin Tissoires wrote: > > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > > On Wednesday 04 January 2017 10:05:22 Benjamin Tissoires wrote: > > > > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > > > > On Tuesday 03 January 2017 12:59:37 Dmitry Torokhov wrote: > > > > > > > On Tue, Jan 03, 2017 at 09:39:13PM +0100, Pali Rohár wrote: > > > > > > > > On Tuesday 03 January 2017 21:24:18 Dmitry Torokhov wrote: > > > > > > > > > On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > > > > > > > > > > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > > > > > > > > > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > > > > > > > > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > > > > > > > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > > > > > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > > > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał > > > > > > > > > > > > > > > > > > > Kępień > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > > > > > > > > > > 0x29. That i2c address is not specified in > > > > > > > > > > > > > > > > > > > > > DMI or ACPI, so runtime detection without > > > > > > > > > > > > > > > > > > > > > whitelist which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > > > > > > > > > > confusing to me at first because there is > > > > > > > > > > > > > > > > > > > > already an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is > > > > > > > > > > > > > > > > > > > > that: > > > > > > > > > > > > > > > > > > > > * the purpose of this patch is to expose a > > > > > > > > > > > > > > > > > > > > richer interface (as > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on > > > > > > > > > > > > > > > > > > > > some machines, > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > > > > > > > > > > effectively duplicates the work that > > > > > > > > > > > > > > > > > > > > lis3lv02d does). > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > > > > > > > > > > exports /dev/freefall device which notify > > > > > > > > > > > > > > > > > > > userspace about falls. lis3lv02d is i2c driver > > > > > > > > > > > > > > > > > > > which exports axes of accelerometer. Additionaly > > > > > > > > > > > > > > > > > > > lis3lv02d can export also /dev/freefall if > > > > > > > > > > > > > > > > > > > registerer of i2c device provides irq number -- > > > > > > > > > > > > > > > > > > > which is not case of this patch. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and > > > > > > > > > > > > > > > > > > > both are useful. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent > > > > > > > > > > > > > > > > > > > one HW device (that ST microelectronics > > > > > > > > > > > > > > > > > > > accelerometer) but due to complicated HW > > > > > > > > > > > > > > > > > > > abstraction and layers on Dell laptops it is > > > > > > > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed > > > > > > > > > > > > > > > > > > > to lis3lv02d driver and that would export whole > > > > > > > > > > > > > > > > > > > device (with /dev/freefall too), but due to HW > > > > > > > > > > > > > > > > > > > abstraction it is too much complicated... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > > > > > > > > > > number all the way down to lis3lv02d is to set the > > > > > > > > > > > > > > > > > > irq field of the struct i2c_board_info you are > > > > > > > > > > > > > > > > > > passing to i2c_new_device(). And you can extract > > > > > > > > > > > > > > > > > > that IRQ number e.g. in > > > > > > > > > > > > > > > > > > check_acpi_smo88xx_device(). However, you would > > > > > > > > > > > > > > > > > > then need to make sure dell-smo8800 does not > > > > > > > > > > > > > > > > > > attempt to request the same IRQ on whitelisted > > > > > > > > > > > > > > > > > > machines. This got me thinking about a way to > > > > > > > > > > > > > > > > > > somehow incorporate your changes into dell-smo8800 > > > > > > > > > > > > > > > > > > using Wolfram's bus_notifier suggestion, but I do > > > > > > > > > > > > > > > > > > not have a working solution for now. What is > > > > > > > > > > > > > > > > > > tempting about this approach is that you would not > > > > > > > > > > > > > > > > > > have to scan the ACPI namespace in search of > > > > > > > > > > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > > > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > > > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > > > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > > > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > > > > > > > > > > driver... And there can be some pathological > > > > > > > > > > > > > > > > > situation (thanks to adding ACPI layer as Andy > > > > > > > > > > > > > > > > > pointed) that there will be more SMO88xx devices in > > > > > > > > > > > > > > > > > ACPI. Plus you can compile kernel with and without > > > > > > > > > > > > > > > > > those modules and also you can blacklist loading > > > > > > > > > > > > > > > > > them (so compile time check is not enough). And > > > > > > > > > > > > > > > > > still some correct message notifier must be used. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > > > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > > > > > > > > > > and available dell devices... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to > > > > > > > > > > > > > > > > find a nice way to implement this. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > > > > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > > > > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > > > > > > > > > > events, not alert") which breaks your patch. The > > > > > > > > > > > > > > > > reason for that is that lis3lv02d relies on the i2c > > > > > > > > > > > > > > > > client's IRQ being 0 to detect that it should not > > > > > > > > > > > > > > > > create /dev/freefall. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > > > > > > > > > > causing lis3lv02d to create /dev/freefall, which in > > > > > > > > > > > > > > > > turn conflicts with dell-smo8800 which is trying to > > > > > > > > > > > > > > > > create /dev/freefall itself. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > > > > > > > > > > > > > > > > > > > Apologies for that. > > > > > > > > > > > > > > > > > > > > > > > > > > > > I could easily fix this by adding a kernel API to know > > > > > > > > > > > > > > whether the provided irq is from Host Notify or if it was > > > > > > > > > > > > > > coming from an actual declaration. However, I have no idea > > > > > > > > > > > > > > how many other drivers would require this (hopefully only > > > > > > > > > > > > > > this one). > > > > > > > > > > > > > > > > > > > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ > > > > > > > > > > > > > > and let the actual drivers that need it to set it, but > > > > > > > > > > > > > > this was not the best solution according to Dmitri. On my > > > > > > > > > > > > > > side, I am not entirely against this given that it's a > > > > > > > > > > > > > > chip feature, so the driver should be able to know that > > > > > > > > > > > > > > it's available. > > > > > > > > > > > > > > > > > > > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > > > > > > > > > > > > > > > > > > > I read this: > > > > > > > > > > > > > > > > > > > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > > > > > > > > > (that ST microelectronics accelerometer) but due to > > > > > > > > > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c." > > > > > > > > > > > > > > > > > > > > > > > > > > and that is the core of the issue. You have 2 drivers > > > > > > > > > > > > > fighting over the same device. Fix this and it will all > > > > > > > > > > > > > work. > > > > > > > > > > > > > > > > > > > > > > > > With my current implementation (which I sent in this patch), > > > > > > > > > > > > they are not fighting. > > > > > > > > > > > > > > > > > > > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and > > > > > > > > > > > > lis3lv02d only accelerometer device as lis3lv02d driver does > > > > > > > > > > > > not get IRQ number in platform data. > > > > > > > > > > > > > > > > > > > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and > > > > > > > > > > > > > accesses it via ACPI methods, can the same be done for Dell? > > > > > > > > > > > > > > > > > > > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI > > > > > > > > > > > > or DMI is even not i2c address of device, so it needs to be > > > > > > > > > > > > specified in code itself. > > > > > > > > > > > > > > > > > > > > > > > > Really there is no other way... :-( > > > > > > > > > > > > > > > > > > > > > > Sure there is: > > > > > > > > > > > > > > > > > > > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > > > > > > > > > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > > > > > > > > > > simply forward requests to dell-smo8800-accel i2c client. > > > > > > > > > > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel > > > > > > > > > > > does. > > > > > > > > > > > > > > > > > > > > Sorry, but I do not understand how you mean it... Why to provides > > > > > > > > > > new read/write i2c functions which are already implemented by > > > > > > > > > > i2c-i801 bus and lis3lv02d i2c driver? > > > > > > > > > > > > > > > > > > Because that would allow you to avoid clashes with i2c creating > > > > > > > > > interrupt mapping for client residing on host-notify-capable > > > > > > > > > controller. > > > > > > > > > > > > > > > > > > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > > > > > > > > > > > > > > > > > > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > > > > > > > > > > device. > > > > > > > > > > > > > > > > > > > > But... what is problem with current implementation? Accelerometer > > > > > > > > > > HW provides two functions: > > > > > > > > > > > > > > > > > > > > 1) 3 axes reports > > > > > > > > > > 2) Disk freefall detection > > > > > > > > > > > > > > > > > > > > And 1) is handled by i2c driver lis3lv02d and 2) is by > > > > > > > > > > dell-smo8800. Both functions are independent here. > > > > > > > > > > > > > > > > > > > > I think you just trying to complicate this situation even more to > > > > > > > > > > be more complicated as currently is. > > > > > > > > > > > > > > > > > > Because this apparently does not work for you, does it? > > > > > > > > > > > > > > > > It is working fine. I do not see any problem. > > > > > > > > > > > > > > > > > In general, > > > > > > > > > if you want the same hardware be handled by 2 different drivers you > > > > > > > > > are going to have bad time. > > > > > > > > > > > > > > > > Yes, but in this case half of device is ACPI based and other half i2c > > > > > > > > based. This is problem of ACPI and Dell design. > > > > > > > > > > > > > > > > > It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are > > > > > > > > > the same, right? > > > > > > > > > > > > > > > > Yes. I understand that clean solution is to have one driver which > > > > > > > > provides everything. > > > > > > > > > > > > > > > > But because half of data are ACPI and half i2c, you still needs to > > > > > > > > create two drivers (one ACPI and one i2c). You can put both drivers into > > > > > > > > one .ko module, but still these will be two drivers due to how ACPI and > > > > > > > > i2c linux abstractions are different. > > > > > > > > > > > > > > > > > So, instead of having 2 drivers split the > > > > > > > > > functionality, can you forego registering smo8800 ACPI driver on > > > > > > > > > your whitelisted boxes and instead instantiate full i2c client > > > > > > > > > device with properly assigned both address and IRQ and let lis3lv02d > > > > > > > > > handle it (providing both accelerometer data and /dev/freefall)? > > > > > > > > > > > > > > > > With Michał we already discussed about it, see emails. Basically you can > > > > > > > > enable/disable kernel modules at compile time or blacklist at runtime > > > > > > > > (or even chose what will be compiled into vmlinux and what as external > > > > > > > > .ko module). > > > > > > > > > > > > > > This can be solved with a bit of Kconfig/IS_ENABLED() code. > > > > > > > > > > > > > > > Some distributions blacklist i2c-i801.ko module... And > > > > > > > > > > > > > > Any particular reason for that? > > > > > > > > > > > > > > > there can be also problem with initialization of i2c-i801 driver (fix is > > > > > > > > in commit a7ae81952cda, but does not have to work at every time!). So > > > > > > > > that move on whitelisted machines can potentially cause disappearance of > > > > > > > > /dev/freefall and users will not have hdd protection which is currently > > > > > > > > working. > > > > > > > > > > > > > > Well, I gave you 2 possible solutions (roll your own i2c read/write, > > > > > > > forward them to i2c client) or have faith in your implementation and let > > > > > > > lis3lv02d handle it. > > > > > > > > > > > > > > The 3rd one is to possibly add a flag to disable host notify to IRQ > > > > > > > mapping for given client (if Wolfram/Jean OK with it). > > > > > > > > > > > > > > Oh, the 4th one: change the irq in lis3lv02d.h to be "int" and change > > > > > > > the check in lis3lv02d.c to be "lis->irq <= 0" and instantiate your > > > > > > > i2c_client with board_info->irq = -1. > > > > > > > > > > > > > > Pick whichever you prefer. > > > > > > > > > > > > > > By the way, what do you need accelerometer for on these devices? They > > > > > > > don't appear to be tablets that could use one... > > > > > > > > > > > > Ah, you are talking about problem that after 4d5538f5882a lis3lv02d will > > > > > > not work... I thought that discussion is about different mechanism how > > > > > > to implement bus registration notification to smo8800 driver (or > > > > > > different solution to not have registration in i801). > > > > > > > > > > > > > > > > Just because I am not sure I got everything right, could you confirm > > > > > that: > > > > > - in the current upstream tree, the dell-smo8800 driver is now broken > > > > > after 4d5538f5882a (i2c: use an IRQ to report Host Notify events, not > > > > > alert) > > > > > > > > No, dell-smo8800 it is working fine. It is fully independent from i2c > > > > and lis3lv02d. It is pure ACPI driver which does not share anything with > > > > i2c. > > > > > > > > > - this series adds an extra lis3lv02d on some machines and you have > > > > > problem fighting for the irq (but this is not upstream yet). > > > > > > > > Yes, this series (not merged yet) adds extra lis3lv02d device but is not > > > > working because of 4d5538f5882a. > > > > > > > > > The extra lis3lv02d node is added from dell-smo8800 > > > > > > > > No, dell-smo8800 does not add new node in this patch. > > > > > > > > > If the first point is not correct (by default, dell-smo8800 will not be > > > > > loaded at the same time than lis3lv02d), then it's a design issue with > > > > > the interactions between those 2 drivers. > > > > > > > > No, there is no interactions between these two drivers (dell-smo8800 and > > > > lis3lv02d). dell-smo8800 is pure ACPI driver and exports just > > > > /dev/freefall device based on IRQ (and nothing more). > > > > > > > > And lis3lv02d in *current* configuration in this patch exports only > > > > accelerometer input device, not /dev/freefall. It does not use IRQ. > > > > (Just there is problem with 4d5538f5882a which tells lis3lv02d IRQ > > > > number which is not freefall report, therefore lis3lv02d does not work). > > > > > > > > To make it clear, ST Accelerometer provides two operations: > > > > * report free fall > > > > * report 3 axes > > > > > > > > Free fall is reported by IRQ, state of 3 axes via i2c bus. Free fall IRQ > > > > is handled by dell-smo8800, state of 3 axes via i2c lis3lv02d driver. > > > > > > > > lis3lv02d can handle also free fall IRQ is platform i2c data provides > > > > IRQ number for it -- but this is not case in our Dell configuration. But > > > > commit 4d5538f5882a inject some IRQ number to lis3lv02d driver which is > > > > not free fall detection and so is breaking lis3lv02 driver. In our Dell > > > > configuration (by this patch) there should be no IRQ number. > > > > > > > > It is clear now? > > > > > > I think I am starting to understand the problem. > > > > > > Currently (upstream tree), 4d5538f5882a doesn't break anything. > > > > I cannot prove it... lis3lv02d i2c driver itself uses this IRQ logic > > (missing = no freefall) and there could be already hardware/boards which > > register lis3lv02d i2c device without IRQ number. And definitions could > > be also in device tree and so outside of kernel sources... > > > > So there can be potentially break. But at least not case of Dell. > > > > > On the > > > mentioned Dell platforms, the dell-smo8800 gets loaded and provides > > > /dev/freefall. lis3lv02d is not loaded so everything just works. > > > > Yes. > > > > > The problem comes from this patch which doesn't set the irq in > > > i2c_board_info and so i2c_core sets the IRQ to Host Notify. I think > > > Dmitri already gave you the solution: set the irq to -1 (or -ENOENT) > > > in i2c_board_info in your patch and only forward it to > > > lis3lv02d_init_device() only if it is positive (valid). > > > > Now I understood that Dmitri means. For this Dell platforms it is OK, > > but we have a problem for device tree platforms. Normally in device tree > > if device does not support something you just do not specify it. But > > with this approach you need to explicitly specify IRQ to -1 in device > > tree. And I think this is really not a clean solution. > > Here I was talking about other platforms/devices (not this Dell one). > No. DT platforms won't have an issue: they don't change anything, and > there will be a new /dev/freefall misc device for the platforms that And this is wrong! There should not be any /dev/freefall device connected with signal host notify. /dev/freefall is for hardware which supports free fall hdd detection. > don't have the irq set *and* if the device is on a Host Notify capable > adapter, currently only i2c-i801. I understood, but in future more bus drivers could support host notify. This is basically incorrect if we use one property for two different things (host notify and hdd free fall). For me it looks like that we should separate these two things into two IRQ properties. It is really wrong design if one property is used for two different things. > But given that they are DT and not > ACPI, this won't conflict with dell-smo8800, so there won't be any > errors, just a dangling unused device node. Yes, for upcoming Dell in this patch (with IRQ -1) it is not a problem. > This approach is IMO the best if you want to have this in the kernel. > > Cheers, > Benjamin
On Wednesday 04 January 2017 11:25:44 Benjamin Tissoires wrote: > On Jan 04 2017 or thereabouts, Andy Shevchenko wrote: > > On Tue, Jan 3, 2017 at 10:39 PM, Pali Rohár <pali.rohar@gmail.com> wrote: > > > > > > With Michał we already discussed about it, see emails. Basically you can > > > enable/disable kernel modules at compile time or blacklist at runtime > > > (or even chose what will be compiled into vmlinux and what as external > > > .ko module). Some distributions blacklist i2c-i801.ko module... > > > > But you understand that any of compile/not compile is not an option, right? > > The case which we face will be both of them, if possible, will be > > compiled as modules. > > > > Blacklisting means making your problem the actual user's one. Not good. > > > > > And > > > there can be also problem with initialization of i2c-i801 driver (fix is > > > in commit a7ae81952cda, but does not have to work at every time!). So > > > that move on whitelisted machines can potentially cause disappearance of > > > /dev/freefall and users will not have hdd protection which is currently > > > working. > > > > I am seeing the same issues with psmouse and SMBus touchpads. The PS/2 > device knows about the availability of a better but unlisted device at > the ACPI level. > > The way I solved this to not have to deal with compile/not compile and > runtime errors is the same way Wolfram told you about: bus notifiers. > I also use an intermediate platform driver to not add i2c dependency on > psmouse. > > For you the solution would be: > - In dell-smo8800, after checking the whitelist, add a platform driver > "dell-lis3lv02d-platform", and add in the platform_data the I2C address > of the chip. > - create a new driver dell-lis3lv02d-platform.ko which listens for the > i2c bus creation and registers the lis3lv02d I2C node when it sees a > matching adapter. (see [1] for my solution) > - in dell-lis3lv02d-platform.ko make sure to set the irq to -ENOENT so > that lis3lv02d.ko doesn't create /dev/freefall which will still be > handled by ACPI. > > How does that sound? Yes, something like this was already suggested. But it is more complicated as my approach and less error prone... See my notes in previous emails. My current path (after fixing IRQ to -1) is smaller more intuitive and do not introduce new complicated parts like bus notifier and new "fake" i2c driver... > Cheers, > Benjamin > > [1] https://github.com/bentiss/linux/blob/synaptics-rmi4-v4.9-rc7+/drivers/input/rmi4/rmi_platform.c
On Wednesday 04 January 2017 12:22:23 Pali Rohár wrote: > On Wednesday 04 January 2017 11:32:33 Benjamin Tissoires wrote: > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > On Wednesday 04 January 2017 11:13:06 Benjamin Tissoires wrote: > > > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > > > On Wednesday 04 January 2017 10:05:22 Benjamin Tissoires wrote: > > > > > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > > > > > On Tuesday 03 January 2017 12:59:37 Dmitry Torokhov wrote: > > > > > > > > On Tue, Jan 03, 2017 at 09:39:13PM +0100, Pali Rohár wrote: > > > > > > > > > On Tuesday 03 January 2017 21:24:18 Dmitry Torokhov wrote: > > > > > > > > > > On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > > > > > > > > > > > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > > > > > > > > > > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > > > > > > > > > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > > > > > > > > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > > > > > > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > > > > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał > > > > > > > > > > > > > > > > > > > > Kępień > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > > > > > > > > > > > 0x29. That i2c address is not specified in > > > > > > > > > > > > > > > > > > > > > > DMI or ACPI, so runtime detection without > > > > > > > > > > > > > > > > > > > > > > whitelist which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > > > > > > > > > > > confusing to me at first because there is > > > > > > > > > > > > > > > > > > > > > already an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is > > > > > > > > > > > > > > > > > > > > > that: > > > > > > > > > > > > > > > > > > > > > * the purpose of this patch is to expose a > > > > > > > > > > > > > > > > > > > > > richer interface (as > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on > > > > > > > > > > > > > > > > > > > > > some machines, > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > > > > > > > > > > > effectively duplicates the work that > > > > > > > > > > > > > > > > > > > > > lis3lv02d does). > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > > > > > > > > > > > exports /dev/freefall device which notify > > > > > > > > > > > > > > > > > > > > userspace about falls. lis3lv02d is i2c driver > > > > > > > > > > > > > > > > > > > > which exports axes of accelerometer. Additionaly > > > > > > > > > > > > > > > > > > > > lis3lv02d can export also /dev/freefall if > > > > > > > > > > > > > > > > > > > > registerer of i2c device provides irq number -- > > > > > > > > > > > > > > > > > > > > which is not case of this patch. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and > > > > > > > > > > > > > > > > > > > > both are useful. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent > > > > > > > > > > > > > > > > > > > > one HW device (that ST microelectronics > > > > > > > > > > > > > > > > > > > > accelerometer) but due to complicated HW > > > > > > > > > > > > > > > > > > > > abstraction and layers on Dell laptops it is > > > > > > > > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed > > > > > > > > > > > > > > > > > > > > to lis3lv02d driver and that would export whole > > > > > > > > > > > > > > > > > > > > device (with /dev/freefall too), but due to HW > > > > > > > > > > > > > > > > > > > > abstraction it is too much complicated... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > > > > > > > > > > > number all the way down to lis3lv02d is to set the > > > > > > > > > > > > > > > > > > > irq field of the struct i2c_board_info you are > > > > > > > > > > > > > > > > > > > passing to i2c_new_device(). And you can extract > > > > > > > > > > > > > > > > > > > that IRQ number e.g. in > > > > > > > > > > > > > > > > > > > check_acpi_smo88xx_device(). However, you would > > > > > > > > > > > > > > > > > > > then need to make sure dell-smo8800 does not > > > > > > > > > > > > > > > > > > > attempt to request the same IRQ on whitelisted > > > > > > > > > > > > > > > > > > > machines. This got me thinking about a way to > > > > > > > > > > > > > > > > > > > somehow incorporate your changes into dell-smo8800 > > > > > > > > > > > > > > > > > > > using Wolfram's bus_notifier suggestion, but I do > > > > > > > > > > > > > > > > > > > not have a working solution for now. What is > > > > > > > > > > > > > > > > > > > tempting about this approach is that you would not > > > > > > > > > > > > > > > > > > > have to scan the ACPI namespace in search of > > > > > > > > > > > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > > > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > > > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > > > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > > > > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > > > > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > > > > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > > > > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > > > > > > > > > > > driver... And there can be some pathological > > > > > > > > > > > > > > > > > > situation (thanks to adding ACPI layer as Andy > > > > > > > > > > > > > > > > > > pointed) that there will be more SMO88xx devices in > > > > > > > > > > > > > > > > > > ACPI. Plus you can compile kernel with and without > > > > > > > > > > > > > > > > > > those modules and also you can blacklist loading > > > > > > > > > > > > > > > > > > them (so compile time check is not enough). And > > > > > > > > > > > > > > > > > > still some correct message notifier must be used. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > > > > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > > > > > > > > > > > and available dell devices... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to > > > > > > > > > > > > > > > > > find a nice way to implement this. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > > > > > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > > > > > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > > > > > > > > > > > events, not alert") which breaks your patch. The > > > > > > > > > > > > > > > > > reason for that is that lis3lv02d relies on the i2c > > > > > > > > > > > > > > > > > client's IRQ being 0 to detect that it should not > > > > > > > > > > > > > > > > > create /dev/freefall. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > > > > > > > > > > > causing lis3lv02d to create /dev/freefall, which in > > > > > > > > > > > > > > > > > turn conflicts with dell-smo8800 which is trying to > > > > > > > > > > > > > > > > > create /dev/freefall itself. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Apologies for that. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I could easily fix this by adding a kernel API to know > > > > > > > > > > > > > > > whether the provided irq is from Host Notify or if it was > > > > > > > > > > > > > > > coming from an actual declaration. However, I have no idea > > > > > > > > > > > > > > > how many other drivers would require this (hopefully only > > > > > > > > > > > > > > > this one). > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ > > > > > > > > > > > > > > > and let the actual drivers that need it to set it, but > > > > > > > > > > > > > > > this was not the best solution according to Dmitri. On my > > > > > > > > > > > > > > > side, I am not entirely against this given that it's a > > > > > > > > > > > > > > > chip feature, so the driver should be able to know that > > > > > > > > > > > > > > > it's available. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > > > > > > > > > > > > > > > > > > > > > I read this: > > > > > > > > > > > > > > > > > > > > > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > > > > > > > > > > (that ST microelectronics accelerometer) but due to > > > > > > > > > > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c." > > > > > > > > > > > > > > > > > > > > > > > > > > > > and that is the core of the issue. You have 2 drivers > > > > > > > > > > > > > > fighting over the same device. Fix this and it will all > > > > > > > > > > > > > > work. > > > > > > > > > > > > > > > > > > > > > > > > > > With my current implementation (which I sent in this patch), > > > > > > > > > > > > > they are not fighting. > > > > > > > > > > > > > > > > > > > > > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and > > > > > > > > > > > > > lis3lv02d only accelerometer device as lis3lv02d driver does > > > > > > > > > > > > > not get IRQ number in platform data. > > > > > > > > > > > > > > > > > > > > > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and > > > > > > > > > > > > > > accesses it via ACPI methods, can the same be done for Dell? > > > > > > > > > > > > > > > > > > > > > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI > > > > > > > > > > > > > or DMI is even not i2c address of device, so it needs to be > > > > > > > > > > > > > specified in code itself. > > > > > > > > > > > > > > > > > > > > > > > > > > Really there is no other way... :-( > > > > > > > > > > > > > > > > > > > > > > > > Sure there is: > > > > > > > > > > > > > > > > > > > > > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > > > > > > > > > > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > > > > > > > > > > > simply forward requests to dell-smo8800-accel i2c client. > > > > > > > > > > > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel > > > > > > > > > > > > does. > > > > > > > > > > > > > > > > > > > > > > Sorry, but I do not understand how you mean it... Why to provides > > > > > > > > > > > new read/write i2c functions which are already implemented by > > > > > > > > > > > i2c-i801 bus and lis3lv02d i2c driver? > > > > > > > > > > > > > > > > > > > > Because that would allow you to avoid clashes with i2c creating > > > > > > > > > > interrupt mapping for client residing on host-notify-capable > > > > > > > > > > controller. > > > > > > > > > > > > > > > > > > > > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > > > > > > > > > > > > > > > > > > > > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > > > > > > > > > > > device. > > > > > > > > > > > > > > > > > > > > > > But... what is problem with current implementation? Accelerometer > > > > > > > > > > > HW provides two functions: > > > > > > > > > > > > > > > > > > > > > > 1) 3 axes reports > > > > > > > > > > > 2) Disk freefall detection > > > > > > > > > > > > > > > > > > > > > > And 1) is handled by i2c driver lis3lv02d and 2) is by > > > > > > > > > > > dell-smo8800. Both functions are independent here. > > > > > > > > > > > > > > > > > > > > > > I think you just trying to complicate this situation even more to > > > > > > > > > > > be more complicated as currently is. > > > > > > > > > > > > > > > > > > > > Because this apparently does not work for you, does it? > > > > > > > > > > > > > > > > > > It is working fine. I do not see any problem. > > > > > > > > > > > > > > > > > > > In general, > > > > > > > > > > if you want the same hardware be handled by 2 different drivers you > > > > > > > > > > are going to have bad time. > > > > > > > > > > > > > > > > > > Yes, but in this case half of device is ACPI based and other half i2c > > > > > > > > > based. This is problem of ACPI and Dell design. > > > > > > > > > > > > > > > > > > > It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are > > > > > > > > > > the same, right? > > > > > > > > > > > > > > > > > > Yes. I understand that clean solution is to have one driver which > > > > > > > > > provides everything. > > > > > > > > > > > > > > > > > > But because half of data are ACPI and half i2c, you still needs to > > > > > > > > > create two drivers (one ACPI and one i2c). You can put both drivers into > > > > > > > > > one .ko module, but still these will be two drivers due to how ACPI and > > > > > > > > > i2c linux abstractions are different. > > > > > > > > > > > > > > > > > > > So, instead of having 2 drivers split the > > > > > > > > > > functionality, can you forego registering smo8800 ACPI driver on > > > > > > > > > > your whitelisted boxes and instead instantiate full i2c client > > > > > > > > > > device with properly assigned both address and IRQ and let lis3lv02d > > > > > > > > > > handle it (providing both accelerometer data and /dev/freefall)? > > > > > > > > > > > > > > > > > > With Michał we already discussed about it, see emails. Basically you can > > > > > > > > > enable/disable kernel modules at compile time or blacklist at runtime > > > > > > > > > (or even chose what will be compiled into vmlinux and what as external > > > > > > > > > .ko module). > > > > > > > > > > > > > > > > This can be solved with a bit of Kconfig/IS_ENABLED() code. > > > > > > > > > > > > > > > > > Some distributions blacklist i2c-i801.ko module... And > > > > > > > > > > > > > > > > Any particular reason for that? > > > > > > > > > > > > > > > > > there can be also problem with initialization of i2c-i801 driver (fix is > > > > > > > > > in commit a7ae81952cda, but does not have to work at every time!). So > > > > > > > > > that move on whitelisted machines can potentially cause disappearance of > > > > > > > > > /dev/freefall and users will not have hdd protection which is currently > > > > > > > > > working. > > > > > > > > > > > > > > > > Well, I gave you 2 possible solutions (roll your own i2c read/write, > > > > > > > > forward them to i2c client) or have faith in your implementation and let > > > > > > > > lis3lv02d handle it. > > > > > > > > > > > > > > > > The 3rd one is to possibly add a flag to disable host notify to IRQ > > > > > > > > mapping for given client (if Wolfram/Jean OK with it). > > > > > > > > > > > > > > > > Oh, the 4th one: change the irq in lis3lv02d.h to be "int" and change > > > > > > > > the check in lis3lv02d.c to be "lis->irq <= 0" and instantiate your > > > > > > > > i2c_client with board_info->irq = -1. > > > > > > > > > > > > > > > > Pick whichever you prefer. > > > > > > > > > > > > > > > > By the way, what do you need accelerometer for on these devices? They > > > > > > > > don't appear to be tablets that could use one... > > > > > > > > > > > > > > Ah, you are talking about problem that after 4d5538f5882a lis3lv02d will > > > > > > > not work... I thought that discussion is about different mechanism how > > > > > > > to implement bus registration notification to smo8800 driver (or > > > > > > > different solution to not have registration in i801). > > > > > > > > > > > > > > > > > > > Just because I am not sure I got everything right, could you confirm > > > > > > that: > > > > > > - in the current upstream tree, the dell-smo8800 driver is now broken > > > > > > after 4d5538f5882a (i2c: use an IRQ to report Host Notify events, not > > > > > > alert) > > > > > > > > > > No, dell-smo8800 it is working fine. It is fully independent from i2c > > > > > and lis3lv02d. It is pure ACPI driver which does not share anything with > > > > > i2c. > > > > > > > > > > > - this series adds an extra lis3lv02d on some machines and you have > > > > > > problem fighting for the irq (but this is not upstream yet). > > > > > > > > > > Yes, this series (not merged yet) adds extra lis3lv02d device but is not > > > > > working because of 4d5538f5882a. > > > > > > > > > > > The extra lis3lv02d node is added from dell-smo8800 > > > > > > > > > > No, dell-smo8800 does not add new node in this patch. > > > > > > > > > > > If the first point is not correct (by default, dell-smo8800 will not be > > > > > > loaded at the same time than lis3lv02d), then it's a design issue with > > > > > > the interactions between those 2 drivers. > > > > > > > > > > No, there is no interactions between these two drivers (dell-smo8800 and > > > > > lis3lv02d). dell-smo8800 is pure ACPI driver and exports just > > > > > /dev/freefall device based on IRQ (and nothing more). > > > > > > > > > > And lis3lv02d in *current* configuration in this patch exports only > > > > > accelerometer input device, not /dev/freefall. It does not use IRQ. > > > > > (Just there is problem with 4d5538f5882a which tells lis3lv02d IRQ > > > > > number which is not freefall report, therefore lis3lv02d does not work). > > > > > > > > > > To make it clear, ST Accelerometer provides two operations: > > > > > * report free fall > > > > > * report 3 axes > > > > > > > > > > Free fall is reported by IRQ, state of 3 axes via i2c bus. Free fall IRQ > > > > > is handled by dell-smo8800, state of 3 axes via i2c lis3lv02d driver. > > > > > > > > > > lis3lv02d can handle also free fall IRQ is platform i2c data provides > > > > > IRQ number for it -- but this is not case in our Dell configuration. But > > > > > commit 4d5538f5882a inject some IRQ number to lis3lv02d driver which is > > > > > not free fall detection and so is breaking lis3lv02 driver. In our Dell > > > > > configuration (by this patch) there should be no IRQ number. > > > > > > > > > > It is clear now? > > > > > > > > I think I am starting to understand the problem. > > > > > > > > Currently (upstream tree), 4d5538f5882a doesn't break anything. > > > > > > I cannot prove it... lis3lv02d i2c driver itself uses this IRQ logic > > > (missing = no freefall) and there could be already hardware/boards which > > > register lis3lv02d i2c device without IRQ number. And definitions could > > > be also in device tree and so outside of kernel sources... > > > > > > So there can be potentially break. But at least not case of Dell. > > > > > > > On the > > > > mentioned Dell platforms, the dell-smo8800 gets loaded and provides > > > > /dev/freefall. lis3lv02d is not loaded so everything just works. > > > > > > Yes. > > > > > > > The problem comes from this patch which doesn't set the irq in > > > > i2c_board_info and so i2c_core sets the IRQ to Host Notify. I think > > > > Dmitri already gave you the solution: set the irq to -1 (or -ENOENT) > > > > in i2c_board_info in your patch and only forward it to > > > > lis3lv02d_init_device() only if it is positive (valid). > > > > > > Now I understood that Dmitri means. For this Dell platforms it is OK, > > > but we have a problem for device tree platforms. Normally in device tree > > > if device does not support something you just do not specify it. But > > > with this approach you need to explicitly specify IRQ to -1 in device > > > tree. And I think this is really not a clean solution. > > > > > Here I was talking about other platforms/devices (not this Dell one). > > > No. DT platforms won't have an issue: they don't change anything, and > > there will be a new /dev/freefall misc device for the platforms that > > And this is wrong! There should not be any /dev/freefall device > connected with signal host notify. /dev/freefall is for hardware which > supports free fall hdd detection. > > > don't have the irq set *and* if the device is on a Host Notify capable > > adapter, currently only i2c-i801. > > I understood, but in future more bus drivers could support host notify. This is really fragile. lis3lv02d will empty IRQ (0) will work on some i2c buses and on some not. I would really expect that i2c driver (lis3lv02d) will work in same way with any i2c bus driver. And not that on i801 will acts differently as on e.g. omap3 i2c bus. > This is basically incorrect if we use one property for two different > things (host notify and hdd free fall). > > For me it looks like that we should separate these two things into two > IRQ properties. It is really wrong design if one property is used for > two different things. > > > But given that they are DT and not > > ACPI, this won't conflict with dell-smo8800, so there won't be any > > errors, just a dangling unused device node. > > Yes, for upcoming Dell in this patch (with IRQ -1) it is not a problem. > > > This approach is IMO the best if you want to have this in the kernel. > > > > Cheers, > > Benjamin >
On Jan 04 2017 or thereabouts, Pali Rohár wrote: > On Wednesday 04 January 2017 11:25:44 Benjamin Tissoires wrote: > > On Jan 04 2017 or thereabouts, Andy Shevchenko wrote: > > > On Tue, Jan 3, 2017 at 10:39 PM, Pali Rohár <pali.rohar@gmail.com> wrote: > > > > > > > > With Michał we already discussed about it, see emails. Basically you can > > > > enable/disable kernel modules at compile time or blacklist at runtime > > > > (or even chose what will be compiled into vmlinux and what as external > > > > .ko module). Some distributions blacklist i2c-i801.ko module... > > > > > > But you understand that any of compile/not compile is not an option, right? > > > The case which we face will be both of them, if possible, will be > > > compiled as modules. > > > > > > Blacklisting means making your problem the actual user's one. Not good. > > > > > > > And > > > > there can be also problem with initialization of i2c-i801 driver (fix is > > > > in commit a7ae81952cda, but does not have to work at every time!). So > > > > that move on whitelisted machines can potentially cause disappearance of > > > > /dev/freefall and users will not have hdd protection which is currently > > > > working. > > > > > > > I am seeing the same issues with psmouse and SMBus touchpads. The PS/2 > > device knows about the availability of a better but unlisted device at > > the ACPI level. > > > > The way I solved this to not have to deal with compile/not compile and > > runtime errors is the same way Wolfram told you about: bus notifiers. > > I also use an intermediate platform driver to not add i2c dependency on > > psmouse. > > > > For you the solution would be: > > - In dell-smo8800, after checking the whitelist, add a platform driver > > "dell-lis3lv02d-platform", and add in the platform_data the I2C address > > of the chip. > > - create a new driver dell-lis3lv02d-platform.ko which listens for the > > i2c bus creation and registers the lis3lv02d I2C node when it sees a > > matching adapter. (see [1] for my solution) > > - in dell-lis3lv02d-platform.ko make sure to set the irq to -ENOENT so > > that lis3lv02d.ko doesn't create /dev/freefall which will still be > > handled by ACPI. > > > > How does that sound? > > Yes, something like this was already suggested. But it is more > complicated as my approach and less error prone... See my notes in > previous emails. Sorry but I can't find your notes about errors in your previous emails. This solution indeed is a little bit more complex than just adding the bits in i2c-i801, but I don't really see the reason for an *adapter* driver to instantiate specific *clients* depending on some DMI matching. It's a platform issue, so it should be solved at a platform level, not at the i2c adapter level. Plus the bus notifier solutions guarantees plain separation between the elements and prevents any conflicts, with or without compile guards. > > My current path (after fixing IRQ to -1) is smaller more intuitive and This will only fix the fact that you have 2 concurrent drivers on the same resource (freefall), not the fact that the global design of having 2 drivers which do not cooperate is wrong. > do not introduce new complicated parts like bus notifier and new "fake" > i2c driver... It's not a "fake i2c driver". It's a platform driver. If you don't like the idea of the platform driver, just add the bus notifier code in dell-smo8800, but this will pull a new dependency on I2C in this ACPI driver. Cheers, Benjamin > > [1] https://github.com/bentiss/linux/blob/synaptics-rmi4-v4.9-rc7+/drivers/input/rmi4/rmi_platform.c > > -- > Pali Rohár > pali.rohar@gmail.com -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Jan 04 2017 or thereabouts, Pali Rohár wrote: > On Wednesday 04 January 2017 12:22:23 Pali Rohár wrote: > > On Wednesday 04 January 2017 11:32:33 Benjamin Tissoires wrote: > > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > > On Wednesday 04 January 2017 11:13:06 Benjamin Tissoires wrote: > > > > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > > > > On Wednesday 04 January 2017 10:05:22 Benjamin Tissoires wrote: > > > > > > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > > > > > > On Tuesday 03 January 2017 12:59:37 Dmitry Torokhov wrote: > > > > > > > > > On Tue, Jan 03, 2017 at 09:39:13PM +0100, Pali Rohár wrote: > > > > > > > > > > On Tuesday 03 January 2017 21:24:18 Dmitry Torokhov wrote: > > > > > > > > > > > On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > > > > > > > > > > > > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > > > > > > > > > > > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > > > > > > > > > > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > > > > > > > > > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > > > > > > > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > > > > > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał > > > > > > > > > > > > > > > > > > > > > Kępień > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > > > > > > > > > > > > 0x29. That i2c address is not specified in > > > > > > > > > > > > > > > > > > > > > > > DMI or ACPI, so runtime detection without > > > > > > > > > > > > > > > > > > > > > > > whitelist which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > > > > > > > > > > > > confusing to me at first because there is > > > > > > > > > > > > > > > > > > > > > > already an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is > > > > > > > > > > > > > > > > > > > > > > that: > > > > > > > > > > > > > > > > > > > > > > * the purpose of this patch is to expose a > > > > > > > > > > > > > > > > > > > > > > richer interface (as > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on > > > > > > > > > > > > > > > > > > > > > > some machines, > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > > > > > > > > > > > > effectively duplicates the work that > > > > > > > > > > > > > > > > > > > > > > lis3lv02d does). > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > > > > > > > > > > > > exports /dev/freefall device which notify > > > > > > > > > > > > > > > > > > > > > userspace about falls. lis3lv02d is i2c driver > > > > > > > > > > > > > > > > > > > > > which exports axes of accelerometer. Additionaly > > > > > > > > > > > > > > > > > > > > > lis3lv02d can export also /dev/freefall if > > > > > > > > > > > > > > > > > > > > > registerer of i2c device provides irq number -- > > > > > > > > > > > > > > > > > > > > > which is not case of this patch. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and > > > > > > > > > > > > > > > > > > > > > both are useful. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent > > > > > > > > > > > > > > > > > > > > > one HW device (that ST microelectronics > > > > > > > > > > > > > > > > > > > > > accelerometer) but due to complicated HW > > > > > > > > > > > > > > > > > > > > > abstraction and layers on Dell laptops it is > > > > > > > > > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed > > > > > > > > > > > > > > > > > > > > > to lis3lv02d driver and that would export whole > > > > > > > > > > > > > > > > > > > > > device (with /dev/freefall too), but due to HW > > > > > > > > > > > > > > > > > > > > > abstraction it is too much complicated... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > > > > > > > > > > > > number all the way down to lis3lv02d is to set the > > > > > > > > > > > > > > > > > > > > irq field of the struct i2c_board_info you are > > > > > > > > > > > > > > > > > > > > passing to i2c_new_device(). And you can extract > > > > > > > > > > > > > > > > > > > > that IRQ number e.g. in > > > > > > > > > > > > > > > > > > > > check_acpi_smo88xx_device(). However, you would > > > > > > > > > > > > > > > > > > > > then need to make sure dell-smo8800 does not > > > > > > > > > > > > > > > > > > > > attempt to request the same IRQ on whitelisted > > > > > > > > > > > > > > > > > > > > machines. This got me thinking about a way to > > > > > > > > > > > > > > > > > > > > somehow incorporate your changes into dell-smo8800 > > > > > > > > > > > > > > > > > > > > using Wolfram's bus_notifier suggestion, but I do > > > > > > > > > > > > > > > > > > > > not have a working solution for now. What is > > > > > > > > > > > > > > > > > > > > tempting about this approach is that you would not > > > > > > > > > > > > > > > > > > > > have to scan the ACPI namespace in search of > > > > > > > > > > > > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > > > > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > > > > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > > > > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > > > > > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > > > > > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > > > > > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > > > > > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > > > > > > > > > > > > driver... And there can be some pathological > > > > > > > > > > > > > > > > > > > situation (thanks to adding ACPI layer as Andy > > > > > > > > > > > > > > > > > > > pointed) that there will be more SMO88xx devices in > > > > > > > > > > > > > > > > > > > ACPI. Plus you can compile kernel with and without > > > > > > > > > > > > > > > > > > > those modules and also you can blacklist loading > > > > > > > > > > > > > > > > > > > them (so compile time check is not enough). And > > > > > > > > > > > > > > > > > > > still some correct message notifier must be used. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > > > > > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > > > > > > > > > > > > and available dell devices... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to > > > > > > > > > > > > > > > > > > find a nice way to implement this. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > > > > > > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > > > > > > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > > > > > > > > > > > > events, not alert") which breaks your patch. The > > > > > > > > > > > > > > > > > > reason for that is that lis3lv02d relies on the i2c > > > > > > > > > > > > > > > > > > client's IRQ being 0 to detect that it should not > > > > > > > > > > > > > > > > > > create /dev/freefall. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > > > > > > > > > > > > causing lis3lv02d to create /dev/freefall, which in > > > > > > > > > > > > > > > > > > turn conflicts with dell-smo8800 which is trying to > > > > > > > > > > > > > > > > > > create /dev/freefall itself. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Apologies for that. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I could easily fix this by adding a kernel API to know > > > > > > > > > > > > > > > > whether the provided irq is from Host Notify or if it was > > > > > > > > > > > > > > > > coming from an actual declaration. However, I have no idea > > > > > > > > > > > > > > > > how many other drivers would require this (hopefully only > > > > > > > > > > > > > > > > this one). > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ > > > > > > > > > > > > > > > > and let the actual drivers that need it to set it, but > > > > > > > > > > > > > > > > this was not the best solution according to Dmitri. On my > > > > > > > > > > > > > > > > side, I am not entirely against this given that it's a > > > > > > > > > > > > > > > > chip feature, so the driver should be able to know that > > > > > > > > > > > > > > > > it's available. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I read this: > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > > > > > > > > > > > (that ST microelectronics accelerometer) but due to > > > > > > > > > > > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c." > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > and that is the core of the issue. You have 2 drivers > > > > > > > > > > > > > > > fighting over the same device. Fix this and it will all > > > > > > > > > > > > > > > work. > > > > > > > > > > > > > > > > > > > > > > > > > > > > With my current implementation (which I sent in this patch), > > > > > > > > > > > > > > they are not fighting. > > > > > > > > > > > > > > > > > > > > > > > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and > > > > > > > > > > > > > > lis3lv02d only accelerometer device as lis3lv02d driver does > > > > > > > > > > > > > > not get IRQ number in platform data. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and > > > > > > > > > > > > > > > accesses it via ACPI methods, can the same be done for Dell? > > > > > > > > > > > > > > > > > > > > > > > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI > > > > > > > > > > > > > > or DMI is even not i2c address of device, so it needs to be > > > > > > > > > > > > > > specified in code itself. > > > > > > > > > > > > > > > > > > > > > > > > > > > > Really there is no other way... :-( > > > > > > > > > > > > > > > > > > > > > > > > > > Sure there is: > > > > > > > > > > > > > > > > > > > > > > > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > > > > > > > > > > > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > > > > > > > > > > > > simply forward requests to dell-smo8800-accel i2c client. > > > > > > > > > > > > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel > > > > > > > > > > > > > does. > > > > > > > > > > > > > > > > > > > > > > > > Sorry, but I do not understand how you mean it... Why to provides > > > > > > > > > > > > new read/write i2c functions which are already implemented by > > > > > > > > > > > > i2c-i801 bus and lis3lv02d i2c driver? > > > > > > > > > > > > > > > > > > > > > > Because that would allow you to avoid clashes with i2c creating > > > > > > > > > > > interrupt mapping for client residing on host-notify-capable > > > > > > > > > > > controller. > > > > > > > > > > > > > > > > > > > > > > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > > > > > > > > > > > > > > > > > > > > > > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > > > > > > > > > > > > device. > > > > > > > > > > > > > > > > > > > > > > > > But... what is problem with current implementation? Accelerometer > > > > > > > > > > > > HW provides two functions: > > > > > > > > > > > > > > > > > > > > > > > > 1) 3 axes reports > > > > > > > > > > > > 2) Disk freefall detection > > > > > > > > > > > > > > > > > > > > > > > > And 1) is handled by i2c driver lis3lv02d and 2) is by > > > > > > > > > > > > dell-smo8800. Both functions are independent here. > > > > > > > > > > > > > > > > > > > > > > > > I think you just trying to complicate this situation even more to > > > > > > > > > > > > be more complicated as currently is. > > > > > > > > > > > > > > > > > > > > > > Because this apparently does not work for you, does it? > > > > > > > > > > > > > > > > > > > > It is working fine. I do not see any problem. > > > > > > > > > > > > > > > > > > > > > In general, > > > > > > > > > > > if you want the same hardware be handled by 2 different drivers you > > > > > > > > > > > are going to have bad time. > > > > > > > > > > > > > > > > > > > > Yes, but in this case half of device is ACPI based and other half i2c > > > > > > > > > > based. This is problem of ACPI and Dell design. > > > > > > > > > > > > > > > > > > > > > It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are > > > > > > > > > > > the same, right? > > > > > > > > > > > > > > > > > > > > Yes. I understand that clean solution is to have one driver which > > > > > > > > > > provides everything. > > > > > > > > > > > > > > > > > > > > But because half of data are ACPI and half i2c, you still needs to > > > > > > > > > > create two drivers (one ACPI and one i2c). You can put both drivers into > > > > > > > > > > one .ko module, but still these will be two drivers due to how ACPI and > > > > > > > > > > i2c linux abstractions are different. > > > > > > > > > > > > > > > > > > > > > So, instead of having 2 drivers split the > > > > > > > > > > > functionality, can you forego registering smo8800 ACPI driver on > > > > > > > > > > > your whitelisted boxes and instead instantiate full i2c client > > > > > > > > > > > device with properly assigned both address and IRQ and let lis3lv02d > > > > > > > > > > > handle it (providing both accelerometer data and /dev/freefall)? > > > > > > > > > > > > > > > > > > > > With Michał we already discussed about it, see emails. Basically you can > > > > > > > > > > enable/disable kernel modules at compile time or blacklist at runtime > > > > > > > > > > (or even chose what will be compiled into vmlinux and what as external > > > > > > > > > > .ko module). > > > > > > > > > > > > > > > > > > This can be solved with a bit of Kconfig/IS_ENABLED() code. > > > > > > > > > > > > > > > > > > > Some distributions blacklist i2c-i801.ko module... And > > > > > > > > > > > > > > > > > > Any particular reason for that? > > > > > > > > > > > > > > > > > > > there can be also problem with initialization of i2c-i801 driver (fix is > > > > > > > > > > in commit a7ae81952cda, but does not have to work at every time!). So > > > > > > > > > > that move on whitelisted machines can potentially cause disappearance of > > > > > > > > > > /dev/freefall and users will not have hdd protection which is currently > > > > > > > > > > working. > > > > > > > > > > > > > > > > > > Well, I gave you 2 possible solutions (roll your own i2c read/write, > > > > > > > > > forward them to i2c client) or have faith in your implementation and let > > > > > > > > > lis3lv02d handle it. > > > > > > > > > > > > > > > > > > The 3rd one is to possibly add a flag to disable host notify to IRQ > > > > > > > > > mapping for given client (if Wolfram/Jean OK with it). > > > > > > > > > > > > > > > > > > Oh, the 4th one: change the irq in lis3lv02d.h to be "int" and change > > > > > > > > > the check in lis3lv02d.c to be "lis->irq <= 0" and instantiate your > > > > > > > > > i2c_client with board_info->irq = -1. > > > > > > > > > > > > > > > > > > Pick whichever you prefer. > > > > > > > > > > > > > > > > > > By the way, what do you need accelerometer for on these devices? They > > > > > > > > > don't appear to be tablets that could use one... > > > > > > > > > > > > > > > > Ah, you are talking about problem that after 4d5538f5882a lis3lv02d will > > > > > > > > not work... I thought that discussion is about different mechanism how > > > > > > > > to implement bus registration notification to smo8800 driver (or > > > > > > > > different solution to not have registration in i801). > > > > > > > > > > > > > > > > > > > > > > Just because I am not sure I got everything right, could you confirm > > > > > > > that: > > > > > > > - in the current upstream tree, the dell-smo8800 driver is now broken > > > > > > > after 4d5538f5882a (i2c: use an IRQ to report Host Notify events, not > > > > > > > alert) > > > > > > > > > > > > No, dell-smo8800 it is working fine. It is fully independent from i2c > > > > > > and lis3lv02d. It is pure ACPI driver which does not share anything with > > > > > > i2c. > > > > > > > > > > > > > - this series adds an extra lis3lv02d on some machines and you have > > > > > > > problem fighting for the irq (but this is not upstream yet). > > > > > > > > > > > > Yes, this series (not merged yet) adds extra lis3lv02d device but is not > > > > > > working because of 4d5538f5882a. > > > > > > > > > > > > > The extra lis3lv02d node is added from dell-smo8800 > > > > > > > > > > > > No, dell-smo8800 does not add new node in this patch. > > > > > > > > > > > > > If the first point is not correct (by default, dell-smo8800 will not be > > > > > > > loaded at the same time than lis3lv02d), then it's a design issue with > > > > > > > the interactions between those 2 drivers. > > > > > > > > > > > > No, there is no interactions between these two drivers (dell-smo8800 and > > > > > > lis3lv02d). dell-smo8800 is pure ACPI driver and exports just > > > > > > /dev/freefall device based on IRQ (and nothing more). > > > > > > > > > > > > And lis3lv02d in *current* configuration in this patch exports only > > > > > > accelerometer input device, not /dev/freefall. It does not use IRQ. > > > > > > (Just there is problem with 4d5538f5882a which tells lis3lv02d IRQ > > > > > > number which is not freefall report, therefore lis3lv02d does not work). > > > > > > > > > > > > To make it clear, ST Accelerometer provides two operations: > > > > > > * report free fall > > > > > > * report 3 axes > > > > > > > > > > > > Free fall is reported by IRQ, state of 3 axes via i2c bus. Free fall IRQ > > > > > > is handled by dell-smo8800, state of 3 axes via i2c lis3lv02d driver. > > > > > > > > > > > > lis3lv02d can handle also free fall IRQ is platform i2c data provides > > > > > > IRQ number for it -- but this is not case in our Dell configuration. But > > > > > > commit 4d5538f5882a inject some IRQ number to lis3lv02d driver which is > > > > > > not free fall detection and so is breaking lis3lv02 driver. In our Dell > > > > > > configuration (by this patch) there should be no IRQ number. > > > > > > > > > > > > It is clear now? > > > > > > > > > > I think I am starting to understand the problem. > > > > > > > > > > Currently (upstream tree), 4d5538f5882a doesn't break anything. > > > > > > > > I cannot prove it... lis3lv02d i2c driver itself uses this IRQ logic > > > > (missing = no freefall) and there could be already hardware/boards which > > > > register lis3lv02d i2c device without IRQ number. And definitions could > > > > be also in device tree and so outside of kernel sources... > > > > > > > > So there can be potentially break. But at least not case of Dell. > > > > > > > > > On the > > > > > mentioned Dell platforms, the dell-smo8800 gets loaded and provides > > > > > /dev/freefall. lis3lv02d is not loaded so everything just works. > > > > > > > > Yes. > > > > > > > > > The problem comes from this patch which doesn't set the irq in > > > > > i2c_board_info and so i2c_core sets the IRQ to Host Notify. I think > > > > > Dmitri already gave you the solution: set the irq to -1 (or -ENOENT) > > > > > in i2c_board_info in your patch and only forward it to > > > > > lis3lv02d_init_device() only if it is positive (valid). > > > > > > > > Now I understood that Dmitri means. For this Dell platforms it is OK, > > > > but we have a problem for device tree platforms. Normally in device tree > > > > if device does not support something you just do not specify it. But > > > > with this approach you need to explicitly specify IRQ to -1 in device > > > > tree. And I think this is really not a clean solution. > > > > > > > > Here I was talking about other platforms/devices (not this Dell one). > > > > > No. DT platforms won't have an issue: they don't change anything, and > > > there will be a new /dev/freefall misc device for the platforms that > > > > And this is wrong! There should not be any /dev/freefall device > > connected with signal host notify. /dev/freefall is for hardware which > > supports free fall hdd detection. On the principle, I agree. But let's be realistic: this device will be created but no events will be generated from it. It's a capability for the chip to provide freefall, so I don't see why it's such an issue to have one created which says nothing. (talking about non Dell platforms here too) > > > > > don't have the irq set *and* if the device is on a Host Notify capable > > > adapter, currently only i2c-i801. > > > > I understood, but in future more bus drivers could support host notify. > > This is really fragile. lis3lv02d will empty IRQ (0) will work on some > i2c buses and on some not. I would really expect that i2c driver > (lis3lv02d) will work in same way with any i2c bus driver. And not that > on i801 will acts differently as on e.g. omap3 i2c bus. I really don't follow you here. On a bus with no Host Notify, the IRQ will be 0 and it will work in the same way. On a bus with Host Notify, the IRQ should be set to something negative in the i2c_board_info to be sure to not have an Host Notify irq assigned. In both case the lis3lv02d_i2c driver will just do: if (client->irq > 0) lis3_dev.irq = client->irq; So there is no bus adapter specifics in the driver. > > > This is basically incorrect if we use one property for two different > > things (host notify and hdd free fall). Host Notify is a way for a I2C device to report an interrupt without having a dedicated line assigned (and so an IRQ). So the chip could use Host Notify to report hdd free fall if it supports the feature. > > > > For me it looks like that we should separate these two things into two > > IRQ properties. It is really wrong design if one property is used for > > two different things. I won't say the design is wrong because Host Notify really is just an IRQ. Where the issue lies now is that lis3lv02d makes the assumption that an irq attached to an i2c_client means that it will report freefall. It was correct before, but now it's not true anymore. So maybe this series should also add a way to provide the source of the IRQ (Host Notify or ACPI or DT). Then, lis3lv02d could ignore or not the incoming IRQ. Cheers, Benjamin > > > > > But given that they are DT and not > > > ACPI, this won't conflict with dell-smo8800, so there won't be any > > > errors, just a dangling unused device node. > > > > Yes, for upcoming Dell in this patch (with IRQ -1) it is not a problem. > > > > > This approach is IMO the best if you want to have this in the kernel. > > > > > > Cheers, > > > Benjamin > > > > -- > Pali Rohár > pali.rohar@gmail.com -- To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wednesday 04 January 2017 14:02:52 Benjamin Tissoires wrote: > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > On Wednesday 04 January 2017 12:22:23 Pali Rohár wrote: > > > On Wednesday 04 January 2017 11:32:33 Benjamin Tissoires wrote: > > > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > > > On Wednesday 04 January 2017 11:13:06 Benjamin Tissoires wrote: > > > > > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > > > > > On Wednesday 04 January 2017 10:05:22 Benjamin Tissoires wrote: > > > > > > > > On Jan 04 2017 or thereabouts, Pali Rohár wrote: > > > > > > > > > On Tuesday 03 January 2017 12:59:37 Dmitry Torokhov wrote: > > > > > > > > > > On Tue, Jan 03, 2017 at 09:39:13PM +0100, Pali Rohár wrote: > > > > > > > > > > > On Tuesday 03 January 2017 21:24:18 Dmitry Torokhov wrote: > > > > > > > > > > > > On Tue, Jan 03, 2017 at 09:05:51PM +0100, Pali Rohár wrote: > > > > > > > > > > > > > On Tuesday 03 January 2017 20:48:12 Dmitry Torokhov wrote: > > > > > > > > > > > > > > On Tue, Jan 03, 2017 at 07:50:17PM +0100, Pali Rohár wrote: > > > > > > > > > > > > > > > On Tuesday 03 January 2017 19:38:43 Dmitry Torokhov wrote: > > > > > > > > > > > > > > > > On Tue, Jan 03, 2017 at 10:06:41AM +0100, Benjamin Tissoires > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > On Dec 29 2016 or thereabouts, Pali Rohár wrote: > > > > > > > > > > > > > > > > > > On Thursday 29 December 2016 22:09:32 Michał Kępień wrote: > > > > > > > > > > > > > > > > > > > > On Thursday 29 December 2016 14:47:19 Michał Kępień > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > > > > On Thursday 29 December 2016 09:29:36 Michał > > > > > > > > > > > > > > > > > > > > > > Kępień > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > > > > > > Dell platform team told us that some (DMI > > > > > > > > > > > > > > > > > > > > > > > > whitelisted) Dell Latitude machines have ST > > > > > > > > > > > > > > > > > > > > > > > > microelectronics accelerometer at i2c address > > > > > > > > > > > > > > > > > > > > > > > > 0x29. That i2c address is not specified in > > > > > > > > > > > > > > > > > > > > > > > > DMI or ACPI, so runtime detection without > > > > > > > > > > > > > > > > > > > > > > > > whitelist which is below is not possible. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Presence of that ST microelectronics > > > > > > > > > > > > > > > > > > > > > > > > accelerometer is verified by existence of > > > > > > > > > > > > > > > > > > > > > > > > SMO88xx ACPI device which represent that > > > > > > > > > > > > > > > > > > > > > > > > accelerometer. Unfortunately without i2c > > > > > > > > > > > > > > > > > > > > > > > > address. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > This part of the commit message sounded a bit > > > > > > > > > > > > > > > > > > > > > > > confusing to me at first because there is > > > > > > > > > > > > > > > > > > > > > > > already an ACPI driver which handles SMO88xx > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > devices (dell-smo8800). My understanding is > > > > > > > > > > > > > > > > > > > > > > > that: > > > > > > > > > > > > > > > > > > > > > > > * the purpose of this patch is to expose a > > > > > > > > > > > > > > > > > > > > > > > richer interface (as > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > provided by lis3lv02d) to these devices on > > > > > > > > > > > > > > > > > > > > > > > some machines, > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > * on whitelisted machines, dell-smo8800 and > > > > > > > > > > > > > > > > > > > > > > > lis3lv02d can work > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > simultaneously (even though dell-smo8800 > > > > > > > > > > > > > > > > > > > > > > > effectively duplicates the work that > > > > > > > > > > > > > > > > > > > > > > > lis3lv02d does). > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > No. dell-smo8800 reads from ACPI irq number and > > > > > > > > > > > > > > > > > > > > > > exports /dev/freefall device which notify > > > > > > > > > > > > > > > > > > > > > > userspace about falls. lis3lv02d is i2c driver > > > > > > > > > > > > > > > > > > > > > > which exports axes of accelerometer. Additionaly > > > > > > > > > > > > > > > > > > > > > > lis3lv02d can export also /dev/freefall if > > > > > > > > > > > > > > > > > > > > > > registerer of i2c device provides irq number -- > > > > > > > > > > > > > > > > > > > > > > which is not case of this patch. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > So both drivers are doing different things and > > > > > > > > > > > > > > > > > > > > > > both are useful. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > IIRC both dell-smo8800 and lis3lv02d represent > > > > > > > > > > > > > > > > > > > > > > one HW device (that ST microelectronics > > > > > > > > > > > > > > > > > > > > > > accelerometer) but due to complicated HW > > > > > > > > > > > > > > > > > > > > > > abstraction and layers on Dell laptops it is > > > > > > > > > > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Yes, in ideal world irq number should be passed > > > > > > > > > > > > > > > > > > > > > > to lis3lv02d driver and that would export whole > > > > > > > > > > > > > > > > > > > > > > device (with /dev/freefall too), but due to HW > > > > > > > > > > > > > > > > > > > > > > abstraction it is too much complicated... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Why? AFAICT, all that is required to pass that IRQ > > > > > > > > > > > > > > > > > > > > > number all the way down to lis3lv02d is to set the > > > > > > > > > > > > > > > > > > > > > irq field of the struct i2c_board_info you are > > > > > > > > > > > > > > > > > > > > > passing to i2c_new_device(). And you can extract > > > > > > > > > > > > > > > > > > > > > that IRQ number e.g. in > > > > > > > > > > > > > > > > > > > > > check_acpi_smo88xx_device(). However, you would > > > > > > > > > > > > > > > > > > > > > then need to make sure dell-smo8800 does not > > > > > > > > > > > > > > > > > > > > > attempt to request the same IRQ on whitelisted > > > > > > > > > > > > > > > > > > > > > machines. This got me thinking about a way to > > > > > > > > > > > > > > > > > > > > > somehow incorporate your changes into dell-smo8800 > > > > > > > > > > > > > > > > > > > > > using Wolfram's bus_notifier suggestion, but I do > > > > > > > > > > > > > > > > > > > > > not have a working solution for now. What is > > > > > > > > > > > > > > > > > > > > > tempting about this approach is that you would not > > > > > > > > > > > > > > > > > > > > > have to scan the ACPI namespace in search of > > > > > > > > > > > > > > > > > > > > > SMO88xx devices, because smo8800_add() is > > > > > > > > > > > > > > > > > > > > > automatically called for them. However, I fear that > > > > > > > > > > > > > > > > > > > > > the resulting solution may be more complicated than > > > > > > > > > > > > > > > > > > > > > the one you submitted. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Then we need to deal with lot of problems. Order of > > > > > > > > > > > > > > > > > > > > loading .ko modules is undefined. Binding devices to > > > > > > > > > > > > > > > > > > > > drivers registered by .ko module is also in "random" > > > > > > > > > > > > > > > > > > > > order. At any time any of those .ko module can be > > > > > > > > > > > > > > > > > > > > unloaded or at least device unbind (via sysfs) from > > > > > > > > > > > > > > > > > > > > driver... And there can be some pathological > > > > > > > > > > > > > > > > > > > > situation (thanks to adding ACPI layer as Andy > > > > > > > > > > > > > > > > > > > > pointed) that there will be more SMO88xx devices in > > > > > > > > > > > > > > > > > > > > ACPI. Plus you can compile kernel with and without > > > > > > > > > > > > > > > > > > > > those modules and also you can blacklist loading > > > > > > > > > > > > > > > > > > > > them (so compile time check is not enough). And > > > > > > > > > > > > > > > > > > > > still some correct message notifier must be used. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I think such solution is much much more complicated, > > > > > > > > > > > > > > > > > > > > there are lot of combinations of kernel configuration > > > > > > > > > > > > > > > > > > > > and available dell devices... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I tried a few more things, but ultimately failed to > > > > > > > > > > > > > > > > > > > find a nice way to implement this. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Another issue popped up, though. Linus' master branch > > > > > > > > > > > > > > > > > > > contains a recent commit by Benjamin Tissoires (CC'ed), > > > > > > > > > > > > > > > > > > > 4d5538f5882a ("i2c: use an IRQ to report Host Notify > > > > > > > > > > > > > > > > > > > events, not alert") which breaks your patch. The > > > > > > > > > > > > > > > > > > > reason for that is that lis3lv02d relies on the i2c > > > > > > > > > > > > > > > > > > > client's IRQ being 0 to detect that it should not > > > > > > > > > > > > > > > > > > > create /dev/freefall. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Benjamin's patch causes the Host Notify IRQ to be > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > assigned to the i2c client your patch creates, thus > > > > > > > > > > > > > > > > > > > causing lis3lv02d to create /dev/freefall, which in > > > > > > > > > > > > > > > > > > > turn conflicts with dell-smo8800 which is trying to > > > > > > > > > > > > > > > > > > > create /dev/freefall itself. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > So 4d5538f5882a is breaking lis3lv02d driver... > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Apologies for that. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I could easily fix this by adding a kernel API to know > > > > > > > > > > > > > > > > > whether the provided irq is from Host Notify or if it was > > > > > > > > > > > > > > > > > coming from an actual declaration. However, I have no idea > > > > > > > > > > > > > > > > > how many other drivers would require this (hopefully only > > > > > > > > > > > > > > > > > this one). > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > One other solution would be to reserve the Host Notify IRQ > > > > > > > > > > > > > > > > > and let the actual drivers that need it to set it, but > > > > > > > > > > > > > > > > > this was not the best solution according to Dmitri. On my > > > > > > > > > > > > > > > > > side, I am not entirely against this given that it's a > > > > > > > > > > > > > > > > > chip feature, so the driver should be able to know that > > > > > > > > > > > > > > > > > it's available. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Dmitri, Wolfram, Jean, any preferences? > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > I read this: > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > "IIRC both dell-smo8800 and lis3lv02d represent one HW device > > > > > > > > > > > > > > > > (that ST microelectronics accelerometer) but due to > > > > > > > > > > > > > > > > complicated HW abstraction and layers on Dell laptops it is > > > > > > > > > > > > > > > > handled by two drivers, one ACPI and one i2c." > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > and that is the core of the issue. You have 2 drivers > > > > > > > > > > > > > > > > fighting over the same device. Fix this and it will all > > > > > > > > > > > > > > > > work. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > With my current implementation (which I sent in this patch), > > > > > > > > > > > > > > > they are not fighting. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > dell-smo8800 exports /dev/freefall (and nothing more) and > > > > > > > > > > > > > > > lis3lv02d only accelerometer device as lis3lv02d driver does > > > > > > > > > > > > > > > not get IRQ number in platform data. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > As far as I can see hp_accel instantiates lis3lv02d and > > > > > > > > > > > > > > > > accesses it via ACPI methods, can the same be done for Dell? > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > No, Dell does not have any ACPI methods. And as I wrote in ACPI > > > > > > > > > > > > > > > or DMI is even not i2c address of device, so it needs to be > > > > > > > > > > > > > > > specified in code itself. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > Really there is no other way... :-( > > > > > > > > > > > > > > > > > > > > > > > > > > > > Sure there is: > > > > > > > > > > > > > > > > > > > > > > > > > > > > 1. dell-smo8800 instantiates I2C device as "dell-smo8800-accel". > > > > > > > > > > > > > > 2. dell-smo8800 provides read/write functions for lis3lv02d that > > > > > > > > > > > > > > simply forward requests to dell-smo8800-accel i2c client. > > > > > > > > > > > > > > 3. dell-smo8800 instantiates lis3lv02d instance like hp_accel > > > > > > > > > > > > > > does. > > > > > > > > > > > > > > > > > > > > > > > > > > Sorry, but I do not understand how you mean it... Why to provides > > > > > > > > > > > > > new read/write i2c functions which are already implemented by > > > > > > > > > > > > > i2c-i801 bus and lis3lv02d i2c driver? > > > > > > > > > > > > > > > > > > > > > > > > Because that would allow you to avoid clashes with i2c creating > > > > > > > > > > > > interrupt mapping for client residing on host-notify-capable > > > > > > > > > > > > controller. > > > > > > > > > > > > > > > > > > > > > > > > > > Alternatively, can lis3lv02d be tasked to create /dev/freefall? > > > > > > > > > > > > > > > > > > > > > > > > > > If i2c_board_info contains IRQ then lis3lv02d create /dev/freefall > > > > > > > > > > > > > device. > > > > > > > > > > > > > > > > > > > > > > > > > > But... what is problem with current implementation? Accelerometer > > > > > > > > > > > > > HW provides two functions: > > > > > > > > > > > > > > > > > > > > > > > > > > 1) 3 axes reports > > > > > > > > > > > > > 2) Disk freefall detection > > > > > > > > > > > > > > > > > > > > > > > > > > And 1) is handled by i2c driver lis3lv02d and 2) is by > > > > > > > > > > > > > dell-smo8800. Both functions are independent here. > > > > > > > > > > > > > > > > > > > > > > > > > > I think you just trying to complicate this situation even more to > > > > > > > > > > > > > be more complicated as currently is. > > > > > > > > > > > > > > > > > > > > > > > > Because this apparently does not work for you, does it? > > > > > > > > > > > > > > > > > > > > > > It is working fine. I do not see any problem. > > > > > > > > > > > > > > > > > > > > > > > In general, > > > > > > > > > > > > if you want the same hardware be handled by 2 different drivers you > > > > > > > > > > > > are going to have bad time. > > > > > > > > > > > > > > > > > > > > > > Yes, but in this case half of device is ACPI based and other half i2c > > > > > > > > > > > based. This is problem of ACPI and Dell design. > > > > > > > > > > > > > > > > > > > > > > > It seems to be that /dev/freefall in dell-smo8800 and lis3lv02d are > > > > > > > > > > > > the same, right? > > > > > > > > > > > > > > > > > > > > > > Yes. I understand that clean solution is to have one driver which > > > > > > > > > > > provides everything. > > > > > > > > > > > > > > > > > > > > > > But because half of data are ACPI and half i2c, you still needs to > > > > > > > > > > > create two drivers (one ACPI and one i2c). You can put both drivers into > > > > > > > > > > > one .ko module, but still these will be two drivers due to how ACPI and > > > > > > > > > > > i2c linux abstractions are different. > > > > > > > > > > > > > > > > > > > > > > > So, instead of having 2 drivers split the > > > > > > > > > > > > functionality, can you forego registering smo8800 ACPI driver on > > > > > > > > > > > > your whitelisted boxes and instead instantiate full i2c client > > > > > > > > > > > > device with properly assigned both address and IRQ and let lis3lv02d > > > > > > > > > > > > handle it (providing both accelerometer data and /dev/freefall)? > > > > > > > > > > > > > > > > > > > > > > With Michał we already discussed about it, see emails. Basically you can > > > > > > > > > > > enable/disable kernel modules at compile time or blacklist at runtime > > > > > > > > > > > (or even chose what will be compiled into vmlinux and what as external > > > > > > > > > > > .ko module). > > > > > > > > > > > > > > > > > > > > This can be solved with a bit of Kconfig/IS_ENABLED() code. > > > > > > > > > > > > > > > > > > > > > Some distributions blacklist i2c-i801.ko module... And > > > > > > > > > > > > > > > > > > > > Any particular reason for that? > > > > > > > > > > > > > > > > > > > > > there can be also problem with initialization of i2c-i801 driver (fix is > > > > > > > > > > > in commit a7ae81952cda, but does not have to work at every time!). So > > > > > > > > > > > that move on whitelisted machines can potentially cause disappearance of > > > > > > > > > > > /dev/freefall and users will not have hdd protection which is currently > > > > > > > > > > > working. > > > > > > > > > > > > > > > > > > > > Well, I gave you 2 possible solutions (roll your own i2c read/write, > > > > > > > > > > forward them to i2c client) or have faith in your implementation and let > > > > > > > > > > lis3lv02d handle it. > > > > > > > > > > > > > > > > > > > > The 3rd one is to possibly add a flag to disable host notify to IRQ > > > > > > > > > > mapping for given client (if Wolfram/Jean OK with it). > > > > > > > > > > > > > > > > > > > > Oh, the 4th one: change the irq in lis3lv02d.h to be "int" and change > > > > > > > > > > the check in lis3lv02d.c to be "lis->irq <= 0" and instantiate your > > > > > > > > > > i2c_client with board_info->irq = -1. > > > > > > > > > > > > > > > > > > > > Pick whichever you prefer. > > > > > > > > > > > > > > > > > > > > By the way, what do you need accelerometer for on these devices? They > > > > > > > > > > don't appear to be tablets that could use one... > > > > > > > > > > > > > > > > > > Ah, you are talking about problem that after 4d5538f5882a lis3lv02d will > > > > > > > > > not work... I thought that discussion is about different mechanism how > > > > > > > > > to implement bus registration notification to smo8800 driver (or > > > > > > > > > different solution to not have registration in i801). > > > > > > > > > > > > > > > > > > > > > > > > > Just because I am not sure I got everything right, could you confirm > > > > > > > > that: > > > > > > > > - in the current upstream tree, the dell-smo8800 driver is now broken > > > > > > > > after 4d5538f5882a (i2c: use an IRQ to report Host Notify events, not > > > > > > > > alert) > > > > > > > > > > > > > > No, dell-smo8800 it is working fine. It is fully independent from i2c > > > > > > > and lis3lv02d. It is pure ACPI driver which does not share anything with > > > > > > > i2c. > > > > > > > > > > > > > > > - this series adds an extra lis3lv02d on some machines and you have > > > > > > > > problem fighting for the irq (but this is not upstream yet). > > > > > > > > > > > > > > Yes, this series (not merged yet) adds extra lis3lv02d device but is not > > > > > > > working because of 4d5538f5882a. > > > > > > > > > > > > > > > The extra lis3lv02d node is added from dell-smo8800 > > > > > > > > > > > > > > No, dell-smo8800 does not add new node in this patch. > > > > > > > > > > > > > > > If the first point is not correct (by default, dell-smo8800 will not be > > > > > > > > loaded at the same time than lis3lv02d), then it's a design issue with > > > > > > > > the interactions between those 2 drivers. > > > > > > > > > > > > > > No, there is no interactions between these two drivers (dell-smo8800 and > > > > > > > lis3lv02d). dell-smo8800 is pure ACPI driver and exports just > > > > > > > /dev/freefall device based on IRQ (and nothing more). > > > > > > > > > > > > > > And lis3lv02d in *current* configuration in this patch exports only > > > > > > > accelerometer input device, not /dev/freefall. It does not use IRQ. > > > > > > > (Just there is problem with 4d5538f5882a which tells lis3lv02d IRQ > > > > > > > number which is not freefall report, therefore lis3lv02d does not work). > > > > > > > > > > > > > > To make it clear, ST Accelerometer provides two operations: > > > > > > > * report free fall > > > > > > > * report 3 axes > > > > > > > > > > > > > > Free fall is reported by IRQ, state of 3 axes via i2c bus. Free fall IRQ > > > > > > > is handled by dell-smo8800, state of 3 axes via i2c lis3lv02d driver. > > > > > > > > > > > > > > lis3lv02d can handle also free fall IRQ is platform i2c data provides > > > > > > > IRQ number for it -- but this is not case in our Dell configuration. But > > > > > > > commit 4d5538f5882a inject some IRQ number to lis3lv02d driver which is > > > > > > > not free fall detection and so is breaking lis3lv02 driver. In our Dell > > > > > > > configuration (by this patch) there should be no IRQ number. > > > > > > > > > > > > > > It is clear now? > > > > > > > > > > > > I think I am starting to understand the problem. > > > > > > > > > > > > Currently (upstream tree), 4d5538f5882a doesn't break anything. > > > > > > > > > > I cannot prove it... lis3lv02d i2c driver itself uses this IRQ logic > > > > > (missing = no freefall) and there could be already hardware/boards which > > > > > register lis3lv02d i2c device without IRQ number. And definitions could > > > > > be also in device tree and so outside of kernel sources... > > > > > > > > > > So there can be potentially break. But at least not case of Dell. > > > > > > > > > > > On the > > > > > > mentioned Dell platforms, the dell-smo8800 gets loaded and provides > > > > > > /dev/freefall. lis3lv02d is not loaded so everything just works. > > > > > > > > > > Yes. > > > > > > > > > > > The problem comes from this patch which doesn't set the irq in > > > > > > i2c_board_info and so i2c_core sets the IRQ to Host Notify. I think > > > > > > Dmitri already gave you the solution: set the irq to -1 (or -ENOENT) > > > > > > in i2c_board_info in your patch and only forward it to > > > > > > lis3lv02d_init_device() only if it is positive (valid). > > > > > > > > > > Now I understood that Dmitri means. For this Dell platforms it is OK, > > > > > but we have a problem for device tree platforms. Normally in device tree > > > > > if device does not support something you just do not specify it. But > > > > > with this approach you need to explicitly specify IRQ to -1 in device > > > > > tree. And I think this is really not a clean solution. > > > > > > > > > > > Here I was talking about other platforms/devices (not this Dell one). > > > > > > > No. DT platforms won't have an issue: they don't change anything, and > > > > there will be a new /dev/freefall misc device for the platforms that > > > > > > And this is wrong! There should not be any /dev/freefall device > > > connected with signal host notify. /dev/freefall is for hardware which > > > supports free fall hdd detection. > > On the principle, I agree. But let's be realistic: this device will be > created but no events will be generated from it. And I think this is a problem. This provides false information to kernel and also to userspace applications that there is working free fall sensor (even there is not). > It's a capability for the chip to provide freefall, That it is not truth. Not every chip has it. Platform data are there to provides this information for driver, if HW has freefall capability or not. > so I don't see why it's such an issue to > have one created which says nothing. (talking about non Dell platforms > here too) > > > > > > > > don't have the irq set *and* if the device is on a Host Notify capable > > > > adapter, currently only i2c-i801. > > > > > > I understood, but in future more bus drivers could support host notify. > > > > This is really fragile. lis3lv02d will empty IRQ (0) will work on some > > i2c buses and on some not. I would really expect that i2c driver > > (lis3lv02d) will work in same way with any i2c bus driver. And not that > > on i801 will acts differently as on e.g. omap3 i2c bus. > > I really don't follow you here. On a bus with no Host Notify, the IRQ > will be 0 and it will work in the same way. That it OK. > On a bus with Host Notify, > the IRQ should be set to something negative in the i2c_board_info Not only in i2c_board_info but also in DTS. Which means every one DTS needs to be updates -- also those outside of linux kernel. Which is for me wrong. > to be > sure to not have an Host Notify irq assigned. In both case the > lis3lv02d_i2c driver will just do: > if (client->irq > 0) > lis3_dev.irq = client->irq; > > So there is no bus adapter specifics in the driver. > > > > > > This is basically incorrect if we use one property for two different > > > things (host notify and hdd free fall). > > Host Notify is a way for a I2C device to report an interrupt without > having a dedicated line assigned (and so an IRQ). Yes. But lis3lv02d i2c devices which I have uses dedicated IRQ. They does not work in way like you describe. > So the chip could use > Host Notify to report hdd free fall if it supports the feature. Old existing HW could not. As those were not designed for such purpose. Yes, new HW can be designed in this way, but linux kernel should not drop (or modify) correct support for old HW devices just because there is new HW design. > > > > > > For me it looks like that we should separate these two things into two > > > IRQ properties. It is really wrong design if one property is used for > > > two different things. > > I won't say the design is wrong because Host Notify really is just an > IRQ. Yes it is IRQ, but different. And lis3lv02d does not issue host notify IRQ, so it should not be assigned to lis3lv02d. > Where the issue lies now is that lis3lv02d makes the assumption > that an irq attached to an i2c_client means that it will report > freefall. Yes. > It was correct before, but now it's not true anymore. Agree. > So maybe this series should also add a way to provide the source of > the IRQ (Host Notify or ACPI or DT). Then, lis3lv02d could ignore or > not the incoming IRQ. Yes, this is an option how to fix this problem. I would rather introduce new IRQ property in both DTS and lis3lv02d platform data (e.g. irq-freefall) which will unambiguously identify freefall IRQ. But I do not think it should be part of my Dell patch. It is independent fix for lis3lv02d driver for bug which was introduced by commit 4d5538f5882a. > > Cheers, > Benjamin > > > > > > > > But given that they are DT and not > > > > ACPI, this won't conflict with dell-smo8800, so there won't be any > > > > errors, just a dangling unused device node. > > > > > > Yes, for upcoming Dell in this patch (with IRQ -1) it is not a problem. > > > > > > > This approach is IMO the best if you want to have this in the kernel. > > > > > > > > Cheers, > > > > Benjamin > > > > > > > -- > > Pali Rohár > > pali.rohar@gmail.com
On Tue, 3 Jan 2017 12:59:37 -0800, Dmitry Torokhov wrote: > On Tue, Jan 03, 2017 at 09:39:13PM +0100, Pali Rohár wrote: > > Some distributions blacklist i2c-i801.ko module... And > > Any particular reason for that? At some point in time, the i2c-i801 driver caused problems on a few systems. They decided that blacklisting the driver for everybody was the easiest way to fix the problem. Of course this doesn't make any sense and should be reverted. The root cause of the problem should have been investigated back then. Maybe it's fixed by now and they will never now...
> > > > > + if (!known_i2c_address) { > > > > > + dev_warn(&priv->pci_dev->dev, > > > > > + "Accelerometer lis3lv02d i2c device is present " > > > > > + "but its i2c address is unknown, skipping ...\n"); > > > > > > > > You are probably well aware of this, but checkpatch prefers > > > > keeping long log messages in one line. I am pointing it out > > > > just in case. > > > > > > Yes, but I do not know how to fix it. Splitting message into two > > > lines generates warning. Having long line generates warning too. > > > > Weird, checkpatch does not protest on my machine when the log message > > is written on a single line... > > I hope that i2c maintainers decide how to format that line. Make it one line saying "accelerometer address unknown"?
diff --git a/drivers/i2c/busses/i2c-i801.c b/drivers/i2c/busses/i2c-i801.c index eb3627f..188cfd4 100644 --- a/drivers/i2c/busses/i2c-i801.c +++ b/drivers/i2c/busses/i2c-i801.c @@ -1118,6 +1118,101 @@ static void dmi_check_onboard_devices(const struct dmi_header *dm, void *adap) } } +static acpi_status check_acpi_smo88xx_device(acpi_handle obj_handle, + u32 nesting_level, + void *context, + void **return_value) +{ + struct acpi_device_info *info; + acpi_status status; + char *hid; + + status = acpi_get_object_info(obj_handle, &info); + if (!ACPI_SUCCESS(status) || !(info->valid & ACPI_VALID_HID)) + return AE_OK; + + hid = info->hardware_id.string; + if (!hid) + return AE_OK; + + if (strlen(hid) < 7) + return AE_OK; + + if (memcmp(hid, "SMO88", 5) != 0) + return AE_OK; + + *((bool *)return_value) = true; + return AE_CTRL_TERMINATE; +} + +static bool is_dell_system_with_lis3lv02d(void) +{ + bool found; + acpi_status status; + const char *vendor; + + vendor = dmi_get_system_info(DMI_SYS_VENDOR); + if (strcmp(vendor, "Dell Inc.") != 0) + return false; + + /* + * Check if ACPI device SMO88xx exists and if is enabled. That ACPI + * device represent our ST microelectronics lis3lv02d accelerometer but + * unfortunately without any other additional information. + */ + found = false; + status = acpi_get_devices(NULL, check_acpi_smo88xx_device, NULL, + (void **)&found); + if (!ACPI_SUCCESS(status) || !found) + return false; + + return true; +} + +/* + * Dell platform team told us that these Latitude devices have + * ST microelectronics accelerometer at i2c address 0x29. + * That i2c address is not specified in DMI or ACPI, so runtime + * detection without whitelist which is below is not possible. + */ +static const char * const dmi_dell_product_names[] = { + "Latitude E5250", + "Latitude E5450", + "Latitude E5550", + "Latitude E6440", + "Latitude E6440 ATG", + "Latitude E6540", +}; + +static void register_dell_lis3lv02d_i2c_device(struct i801_priv *priv) +{ + struct i2c_board_info info; + const char *product_name; + bool known_i2c_address; + int i; + + known_i2c_address = false; + product_name = dmi_get_system_info(DMI_PRODUCT_NAME); + for (i = 0; i < ARRAY_SIZE(dmi_dell_product_names); ++i) { + if (strcmp(product_name, dmi_dell_product_names[i]) == 0) { + known_i2c_address = true; + break; + } + } + + if (!known_i2c_address) { + dev_warn(&priv->pci_dev->dev, + "Accelerometer lis3lv02d i2c device is present " + "but its i2c address is unknown, skipping ...\n"); + return; + } + + memset(&info, 0, sizeof(struct i2c_board_info)); + info.addr = 0x29; + strlcpy(info.type, "lis3lv02d", I2C_NAME_SIZE); + i2c_new_device(&priv->adapter, &info); +} + /* Register optional slaves */ static void i801_probe_optional_slaves(struct i801_priv *priv) { @@ -1136,6 +1231,9 @@ static void i801_probe_optional_slaves(struct i801_priv *priv) if (dmi_name_in_vendors("FUJITSU")) dmi_walk(dmi_check_onboard_devices, &priv->adapter); + + if (is_dell_system_with_lis3lv02d()) + register_dell_lis3lv02d_i2c_device(priv); } #else static void __init input_apanel_init(void) {}
Dell platform team told us that some (DMI whitelisted) Dell Latitude machines have ST microelectronics accelerometer at i2c address 0x29. That i2c address is not specified in DMI or ACPI, so runtime detection without whitelist which is below is not possible. Presence of that ST microelectronics accelerometer is verified by existence of SMO88xx ACPI device which represent that accelerometer. Unfortunately without i2c address. This patch registers lis3lv02d device at i2c address 0x29 if is detected. Finally commit a7ae81952cda ("i2c: i801: Allow ACPI SystemIO OpRegion to conflict with PCI BAR") allowed to use i2c-i801 driver on Dell machines so lis3lv02d correctly initialize accelerometer. Tested on Dell Latitude E6440. Signed-off-by: Pali Rohár <pali.rohar@gmail.com> --- drivers/i2c/busses/i2c-i801.c | 98 +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 98 insertions(+)