| Message ID | 20170510161240.13229-2-benjamin.tissoires@redhat.com (mailing list archive) |
|---|---|
| State | Superseded, archived |
| Headers | show |
Hi, Benjiamin > From: Benjamin Tissoires [mailto:benjamin.tissoires@redhat.com] > Sent: Thursday, May 11, 2017 12:13 AM > To: Rafael J . Wysocki <rjw@rjwysocki.net>; Zheng, Lv <lv.zheng@intel.com> > Cc: Jiri Eischmann <jeischma@redhat.com>; linux-acpi@vger.kernel.org; linux-kernel@vger.kernel.org > Subject: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > > This reverts commit ecb10b694b72ca5ea51b3c90a71ff2a11963425a. > > Even if the method can be buggy some times, it's still a need > when you boot a laptop with a lid closed and an external monitor > connected (typical situation when using a docking station) > > Note: this option has been removed without being deprecated, which > is terrible in term of distribution handling. The new default "open" > is plain wrong and we don't even have the chance to keep the current > default option because it's not there anymore. I have reverted this: https://patchwork.kernel.org/patch/9717109/ Also I noticed one thing: https://patchwork.kernel.org/patch/9717111/ After I changed kernel lid notification to always send lid return value to other drivers. In order to find out a single driver mode (without platform quirks) to handle both cases. I failed. I should still send close init state to the user space program to work around the external monitor issues. So as you know, we need to send "open" init state to the user space program to work around suspend/resume loop issue. Then for such platforms, how can ACPI button driver automatically detect if an external monitor is there? Unless we fix the BIOS code to make lid return value work as user space's expectation. OK, then this creates an endless business in ACPI community to "re-develop" BIOS tables if they cannot meat user space's expectation. That sucks. It sound the best way is the user space program equipped with hwdb quirks who knows everything to alter acpi button driver mode from user space to work around this. For example: If hwdb is hit, and there is no external monitor, then Echo "open" > /sys/module/button/parameters/lid_init_state If hwdb is not hit or there is an external monitor, then If hwdb is hit, and there is no external monitor, then Echo "close" > /sys/module/button/parameters/lid_init_state So PATCH 2 is not useful. Reverting that can trigger a regression loop we surely do not want to handle. Thanks and best regards Lv > > Cc: stable@vger.kernel.org > Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com> > --- > Documentation/acpi/acpi-lid.txt | 16 ++++++++++++---- > drivers/acpi/button.c | 9 +++++++++ > 2 files changed, 21 insertions(+), 4 deletions(-) > > diff --git a/Documentation/acpi/acpi-lid.txt b/Documentation/acpi/acpi-lid.txt > index 22cb309..effe7af 100644 > --- a/Documentation/acpi/acpi-lid.txt > +++ b/Documentation/acpi/acpi-lid.txt > @@ -59,20 +59,28 @@ button driver uses the following 3 modes in order not to trigger issues. > If the userspace hasn't been prepared to ignore the unreliable "opened" > events and the unreliable initial state notification, Linux users can use > the following kernel parameters to handle the possible issues: > -A. button.lid_init_state=open: > +A. button.lid_init_state=method: > + When this option is specified, the ACPI button driver reports the > + initial lid state using the returning value of the _LID control method > + and whether the "opened"/"closed" events are paired fully relies on the > + firmware implementation. > + This option can be used to fix some platforms where the returning value > + of the _LID control method is reliable but the initial lid state > + notification is missing. > + This option is the default behavior during the period the userspace > + isn't ready to handle the buggy AML tables. > +B. button.lid_init_state=open: > When this option is specified, the ACPI button driver always reports the > initial lid state as "opened" and whether the "opened"/"closed" events > are paired fully relies on the firmware implementation. > This may fix some platforms where the returning value of the _LID > control method is not reliable and the initial lid state notification is > missing. > - This option is the default behavior during the period the userspace > - isn't ready to handle the buggy AML tables. > > If the userspace has been prepared to ignore the unreliable "opened" events > and the unreliable initial state notification, Linux users should always > use the following kernel parameter: > -B. button.lid_init_state=ignore: > +C. button.lid_init_state=ignore: > When this option is specified, the ACPI button driver never reports the > initial lid state and there is a compensation mechanism implemented to > ensure that the reliable "closed" notifications can always be delievered > diff --git a/drivers/acpi/button.c b/drivers/acpi/button.c > index 668137e..6d5a8c1 100644 > --- a/drivers/acpi/button.c > +++ b/drivers/acpi/button.c > @@ -57,6 +57,7 @@ > > #define ACPI_BUTTON_LID_INIT_IGNORE 0x00 > #define ACPI_BUTTON_LID_INIT_OPEN 0x01 > +#define ACPI_BUTTON_LID_INIT_METHOD 0x02 > > #define _COMPONENT ACPI_BUTTON_COMPONENT > ACPI_MODULE_NAME("button"); > @@ -376,6 +377,9 @@ static void acpi_lid_initialize_state(struct acpi_device *device) > case ACPI_BUTTON_LID_INIT_OPEN: > (void)acpi_lid_notify_state(device, 1); > break; > + case ACPI_BUTTON_LID_INIT_METHOD: > + (void)acpi_lid_update_state(device); > + break; > case ACPI_BUTTON_LID_INIT_IGNORE: > default: > break; > @@ -559,6 +563,9 @@ static int param_set_lid_init_state(const char *val, struct kernel_param *kp) > if (!strncmp(val, "open", sizeof("open") - 1)) { > lid_init_state = ACPI_BUTTON_LID_INIT_OPEN; > pr_info("Notify initial lid state as open\n"); > + } else if (!strncmp(val, "method", sizeof("method") - 1)) { > + lid_init_state = ACPI_BUTTON_LID_INIT_METHOD; > + pr_info("Notify initial lid state with _LID return value\n"); > } else if (!strncmp(val, "ignore", sizeof("ignore") - 1)) { > lid_init_state = ACPI_BUTTON_LID_INIT_IGNORE; > pr_info("Do not notify initial lid state\n"); > @@ -572,6 +579,8 @@ static int param_get_lid_init_state(char *buffer, struct kernel_param *kp) > switch (lid_init_state) { > case ACPI_BUTTON_LID_INIT_OPEN: > return sprintf(buffer, "open"); > + case ACPI_BUTTON_LID_INIT_METHOD: > + return sprintf(buffer, "method"); > case ACPI_BUTTON_LID_INIT_IGNORE: > return sprintf(buffer, "ignore"); > default: > -- > 2.9.3 -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi, Benjamin > From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-owner@vger.kernel.org] On Behalf Of Zheng, > Lv > Subject: RE: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > > Hi, Benjiamin > > > From: Benjamin Tissoires [mailto:benjamin.tissoires@redhat.com] > > Sent: Thursday, May 11, 2017 12:13 AM > > To: Rafael J . Wysocki <rjw@rjwysocki.net>; Zheng, Lv <lv.zheng@intel.com> > > Cc: Jiri Eischmann <jeischma@redhat.com>; linux-acpi@vger.kernel.org; linux-kernel@vger.kernel.org > > Subject: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > > > > This reverts commit ecb10b694b72ca5ea51b3c90a71ff2a11963425a. > > > > Even if the method can be buggy some times, it's still a need > > when you boot a laptop with a lid closed and an external monitor > > connected (typical situation when using a docking station) > > > > Note: this option has been removed without being deprecated, which > > is terrible in term of distribution handling. The new default "open" > > is plain wrong and we don't even have the chance to keep the current > > default option because it's not there anymore. > > I have reverted this: > https://patchwork.kernel.org/patch/9717109/ > > Also I noticed one thing: > https://patchwork.kernel.org/patch/9717111/ > > After I changed kernel lid notification to always send lid return value to other drivers. > In order to find out a single driver mode (without platform quirks) to handle both cases. > I failed. > I should still send close init state to the user space program to work around the external monitor > issues. > > So as you know, we need to send "open" init state to the user space program to work around > suspend/resume loop issue. > > Then for such platforms, how can ACPI button driver automatically detect if an external monitor is > there? > Unless we fix the BIOS code to make lid return value work as user space's expectation. > OK, then this creates an endless business in ACPI community to "re-develop" BIOS tables if they cannot > meat user space's expectation. > That sucks. > > It sound the best way is the user space program equipped with hwdb quirks who knows everything to > alter acpi button driver mode from user space to work around this. > For example: > > If hwdb is hit, and there is no external monitor, then > Echo "open" > /sys/module/button/parameters/lid_init_state > If hwdb is not hit or there is an external monitor, then > If hwdb is hit, and there is no external monitor, then > Echo "close" > /sys/module/button/parameters/lid_init_state Let me do re-wording. If hwdb is not hit echo "method" > /sys/module/button/parameters/lid_init_state If hwdb is hit, and there is no external monitor, then echo "open" > /sys/module/button/parameters/lid_init_state If hwdb is hit, and there is an external monitor echo "close" > /sys/module/button/parameters/lid_init_state Then this always assumes the hard requirements of the platform quirks. And it then looks it's better to do such switches in the user space as ACPI button driver doesn't know and doesn't have to know the existence of the external monitor. However, is that possible to not introduce platform quirks? For example, for systemd: If it detected ACPI lid device, automatically switch to an "IgnoreLidOpenEvent" mode (like nouveau drivers' ignorelid=Y option). BTW, which program is responsible for lighting on internal/external monitors? Can it determine that by its own without listening to the lid key events? This is what we preferred. If all of above usage models are corrected, we'll change acpi button driver default mode to "ignore". Another problem for changing default mode back to "method" is: If we changed button driver default mode back to "method", then ACPI community will be flooded of suspend/resume loop bug reports. After we root cause that's not a kernel problem, do we have mean in other community to handle such reports? For example, to collect hwdb entries. Thanks and best regards Lv > > So PATCH 2 is not useful. > Reverting that can trigger a regression loop we surely do not want to handle. > > Thanks and best regards > Lv > > > > > Cc: stable@vger.kernel.org > > Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com> > > --- > > Documentation/acpi/acpi-lid.txt | 16 ++++++++++++---- > > drivers/acpi/button.c | 9 +++++++++ > > 2 files changed, 21 insertions(+), 4 deletions(-) > > > > diff --git a/Documentation/acpi/acpi-lid.txt b/Documentation/acpi/acpi-lid.txt > > index 22cb309..effe7af 100644 > > --- a/Documentation/acpi/acpi-lid.txt > > +++ b/Documentation/acpi/acpi-lid.txt > > @@ -59,20 +59,28 @@ button driver uses the following 3 modes in order not to trigger issues. > > If the userspace hasn't been prepared to ignore the unreliable "opened" > > events and the unreliable initial state notification, Linux users can use > > the following kernel parameters to handle the possible issues: > > -A. button.lid_init_state=open: > > +A. button.lid_init_state=method: > > + When this option is specified, the ACPI button driver reports the > > + initial lid state using the returning value of the _LID control method > > + and whether the "opened"/"closed" events are paired fully relies on the > > + firmware implementation. > > + This option can be used to fix some platforms where the returning value > > + of the _LID control method is reliable but the initial lid state > > + notification is missing. > > + This option is the default behavior during the period the userspace > > + isn't ready to handle the buggy AML tables. > > +B. button.lid_init_state=open: > > When this option is specified, the ACPI button driver always reports the > > initial lid state as "opened" and whether the "opened"/"closed" events > > are paired fully relies on the firmware implementation. > > This may fix some platforms where the returning value of the _LID > > control method is not reliable and the initial lid state notification is > > missing. > > - This option is the default behavior during the period the userspace > > - isn't ready to handle the buggy AML tables. > > > > If the userspace has been prepared to ignore the unreliable "opened" events > > and the unreliable initial state notification, Linux users should always > > use the following kernel parameter: > > -B. button.lid_init_state=ignore: > > +C. button.lid_init_state=ignore: > > When this option is specified, the ACPI button driver never reports the > > initial lid state and there is a compensation mechanism implemented to > > ensure that the reliable "closed" notifications can always be delievered > > diff --git a/drivers/acpi/button.c b/drivers/acpi/button.c > > index 668137e..6d5a8c1 100644 > > --- a/drivers/acpi/button.c > > +++ b/drivers/acpi/button.c > > @@ -57,6 +57,7 @@ > > > > #define ACPI_BUTTON_LID_INIT_IGNORE 0x00 > > #define ACPI_BUTTON_LID_INIT_OPEN 0x01 > > +#define ACPI_BUTTON_LID_INIT_METHOD 0x02 > > > > #define _COMPONENT ACPI_BUTTON_COMPONENT > > ACPI_MODULE_NAME("button"); > > @@ -376,6 +377,9 @@ static void acpi_lid_initialize_state(struct acpi_device *device) > > case ACPI_BUTTON_LID_INIT_OPEN: > > (void)acpi_lid_notify_state(device, 1); > > break; > > + case ACPI_BUTTON_LID_INIT_METHOD: > > + (void)acpi_lid_update_state(device); > > + break; > > case ACPI_BUTTON_LID_INIT_IGNORE: > > default: > > break; > > @@ -559,6 +563,9 @@ static int param_set_lid_init_state(const char *val, struct kernel_param *kp) > > if (!strncmp(val, "open", sizeof("open") - 1)) { > > lid_init_state = ACPI_BUTTON_LID_INIT_OPEN; > > pr_info("Notify initial lid state as open\n"); > > + } else if (!strncmp(val, "method", sizeof("method") - 1)) { > > + lid_init_state = ACPI_BUTTON_LID_INIT_METHOD; > > + pr_info("Notify initial lid state with _LID return value\n"); > > } else if (!strncmp(val, "ignore", sizeof("ignore") - 1)) { > > lid_init_state = ACPI_BUTTON_LID_INIT_IGNORE; > > pr_info("Do not notify initial lid state\n"); > > @@ -572,6 +579,8 @@ static int param_get_lid_init_state(char *buffer, struct kernel_param *kp) > > switch (lid_init_state) { > > case ACPI_BUTTON_LID_INIT_OPEN: > > return sprintf(buffer, "open"); > > + case ACPI_BUTTON_LID_INIT_METHOD: > > + return sprintf(buffer, "method"); > > case ACPI_BUTTON_LID_INIT_IGNORE: > > return sprintf(buffer, "ignore"); > > default: > > -- > > 2.9.3 > > -- > To unsubscribe from this list: send the line "unsubscribe linux-acpi" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On May 11 2017 or thereabouts, Zheng, Lv wrote: > Hi, Benjamin > > > From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-owner@vger.kernel.org] On Behalf Of Zheng, > > Lv > > Subject: RE: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > > > > Hi, Benjiamin > > > > > From: Benjamin Tissoires [mailto:benjamin.tissoires@redhat.com] > > > Sent: Thursday, May 11, 2017 12:13 AM > > > To: Rafael J . Wysocki <rjw@rjwysocki.net>; Zheng, Lv <lv.zheng@intel.com> > > > Cc: Jiri Eischmann <jeischma@redhat.com>; linux-acpi@vger.kernel.org; linux-kernel@vger.kernel.org > > > Subject: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > > > > > > This reverts commit ecb10b694b72ca5ea51b3c90a71ff2a11963425a. > > > > > > Even if the method can be buggy some times, it's still a need > > > when you boot a laptop with a lid closed and an external monitor > > > connected (typical situation when using a docking station) > > > > > > Note: this option has been removed without being deprecated, which > > > is terrible in term of distribution handling. The new default "open" > > > is plain wrong and we don't even have the chance to keep the current > > > default option because it's not there anymore. > > > > I have reverted this: > > https://patchwork.kernel.org/patch/9717109/ Yeah, sorry, as mentioned to Rafael, I only saw it after I sent my series. > > > > Also I noticed one thing: > > https://patchwork.kernel.org/patch/9717111/ > > > > After I changed kernel lid notification to always send lid return value to other drivers. > > In order to find out a single driver mode (without platform quirks) to handle both cases. > > I failed. Yes. As long as you do not have more information on the device, this is not something you can solve with hacks in the kernel. > > I should still send close init state to the user space program to work around the external monitor > > issues. > > > > So as you know, we need to send "open" init state to the user space program to work around > > suspend/resume loop issue. I disagree. You are solving a user space issue (suspend boot loop), with the wrong tool. The user space expects the kernel to provide accurate events, but if the hardware is wrong, we should either fix it (overwriting AML tables), make reasonable assumptions based on the exact model capabilities (is the power button accessible with the LID closed), or teach user space about these situations. There is no point in assuming wrong states in the kernel "to fix user space" when user space is obviously not doing the right thing. > > > > Then for such platforms, how can ACPI button driver automatically detect if an external monitor is > > there? That's not the ACPI button role. However, user space can write to the switch to overwrite it's value when it judges that the kernel is doing things wrong. Libinput is a pretty good candidate, given it has a view of all input devices. And guess what, the code is already there. > > Unless we fix the BIOS code to make lid return value work as user space's expectation. That would be great. > > OK, then this creates an endless business in ACPI community to "re-develop" BIOS tables if they cannot > > meat user space's expectation. > > That sucks. Yes, but unless you teach OEM to not do crap, that's our daily burden. I'd love to not have to quirk endlessly all the drivers I maintain, but each generation of new devices has a new creative way of breaking the existing code, "because it works under Windows". > > > > It sound the best way is the user space program equipped with hwdb quirks who knows everything to > > alter acpi button driver mode from user space to work around this. Yes, the hwdb entry is the solution (or a quirk list in acpi/button.c). The advantage of the hwdb entry is that it will be asynchronous from the kernel releases, and users can just drop a file in their /etc folder and they solved the issue. Distribution will also be able to carry the list of quirked devices, and hopefully end users won't see the boot loop. > > For example: > > > > If hwdb is hit, and there is no external monitor, then > > Echo "open" > /sys/module/button/parameters/lid_init_state > > If hwdb is not hit or there is an external monitor, then > > If hwdb is hit, and there is no external monitor, then > > Echo "close" > /sys/module/button/parameters/lid_init_state Hum, no. This is too late. acpi/button.c is loaded before udev hits, so the initial state will already be evaluated. > > Let me do re-wording. > If hwdb is not hit > echo "method" > /sys/module/button/parameters/lid_init_state > If hwdb is hit, and there is no external monitor, then > echo "open" > /sys/module/button/parameters/lid_init_state > If hwdb is hit, and there is an external monitor > echo "close" > /sys/module/button/parameters/lid_init_state > Then this always assumes the hard requirements of the platform quirks. > And it then looks it's better to do such switches in the user space as ACPI button driver doesn't know and doesn't have to know the existence of the external monitor. Again, the external monitor doesn't matter here. The external monitor issue is a user space choice to: - not suspend if the LID is closed and a monitor is plugged - only show the greater on the internal monitor if both are turned on. The issue we are fixing here is the fact that the switch state is wrong, which makes user space assumptions wrong too (and users angry). But given that the LID switch is an actual input switch device, user space can overwrite it by simply writing to the input node. > > However, is that possible to not introduce platform quirks? > For example, for systemd: > If it detected ACPI lid device, automatically switch to an "IgnoreLidOpenEvent" mode (like nouveau drivers' ignorelid=Y option). Well, given that 99.9% of laptops have this ACPI lid button, you'll just remove the feature from all laptops. > BTW, which program is responsible for lighting on internal/external monitors? I would say the compositor or X, so gdm, kdm, gnome, kde, etc... > Can it determine that by its own without listening to the lid key events? Basically no. This switch is there for a reason. However, I am convinced that a good heuristic is to say that if you are using the internal keyboard, touchscreen or touchpad, unless the user has very very thin fingers, the lid is open. > This is what we preferred. > If all of above usage models are corrected, we'll change acpi button driver default mode to "ignore". No, we need to report accurate state, or explicitly mark the platform as not reliable, and so we need "method" and a hwdb of problematic ones. > > Another problem for changing default mode back to "method" is: > If we changed button driver default mode back to "method", then ACPI community will be flooded of suspend/resume loop bug reports. But that's your job to fix bugs. If there is a user space problem that can be solved in user space, you just need to redirect the users to the correct solution and close the bug. But you are talking about "flood", and I don't think we ever talked about more than 4 devices. So could you point me at a list of bugs that you actually had to fix? > After we root cause that's not a kernel problem, do we have mean in other community to handle such reports? > For example, to collect hwdb entries. libinput, systemd are good candidate. Libinput already has the bits in place, so I'd say we should probably ask the users to report a bug on the wayland/libinput component of bugs.freedesktop.org. But this will only work if the default initial lid state is "method". Sorry for showing I am angry. But I thought we solved this months ago and this bites back. Cheers, Benjamin > > Thanks and best regards > Lv > > > > > So PATCH 2 is not useful. > > Reverting that can trigger a regression loop we surely do not want to handle. > > > > Thanks and best regards > > Lv > > > > > > > > Cc: stable@vger.kernel.org > > > Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com> > > > --- > > > Documentation/acpi/acpi-lid.txt | 16 ++++++++++++---- > > > drivers/acpi/button.c | 9 +++++++++ > > > 2 files changed, 21 insertions(+), 4 deletions(-) > > > > > > diff --git a/Documentation/acpi/acpi-lid.txt b/Documentation/acpi/acpi-lid.txt > > > index 22cb309..effe7af 100644 > > > --- a/Documentation/acpi/acpi-lid.txt > > > +++ b/Documentation/acpi/acpi-lid.txt > > > @@ -59,20 +59,28 @@ button driver uses the following 3 modes in order not to trigger issues. > > > If the userspace hasn't been prepared to ignore the unreliable "opened" > > > events and the unreliable initial state notification, Linux users can use > > > the following kernel parameters to handle the possible issues: > > > -A. button.lid_init_state=open: > > > +A. button.lid_init_state=method: > > > + When this option is specified, the ACPI button driver reports the > > > + initial lid state using the returning value of the _LID control method > > > + and whether the "opened"/"closed" events are paired fully relies on the > > > + firmware implementation. > > > + This option can be used to fix some platforms where the returning value > > > + of the _LID control method is reliable but the initial lid state > > > + notification is missing. > > > + This option is the default behavior during the period the userspace > > > + isn't ready to handle the buggy AML tables. > > > +B. button.lid_init_state=open: > > > When this option is specified, the ACPI button driver always reports the > > > initial lid state as "opened" and whether the "opened"/"closed" events > > > are paired fully relies on the firmware implementation. > > > This may fix some platforms where the returning value of the _LID > > > control method is not reliable and the initial lid state notification is > > > missing. > > > - This option is the default behavior during the period the userspace > > > - isn't ready to handle the buggy AML tables. > > > > > > If the userspace has been prepared to ignore the unreliable "opened" events > > > and the unreliable initial state notification, Linux users should always > > > use the following kernel parameter: > > > -B. button.lid_init_state=ignore: > > > +C. button.lid_init_state=ignore: > > > When this option is specified, the ACPI button driver never reports the > > > initial lid state and there is a compensation mechanism implemented to > > > ensure that the reliable "closed" notifications can always be delievered > > > diff --git a/drivers/acpi/button.c b/drivers/acpi/button.c > > > index 668137e..6d5a8c1 100644 > > > --- a/drivers/acpi/button.c > > > +++ b/drivers/acpi/button.c > > > @@ -57,6 +57,7 @@ > > > > > > #define ACPI_BUTTON_LID_INIT_IGNORE 0x00 > > > #define ACPI_BUTTON_LID_INIT_OPEN 0x01 > > > +#define ACPI_BUTTON_LID_INIT_METHOD 0x02 > > > > > > #define _COMPONENT ACPI_BUTTON_COMPONENT > > > ACPI_MODULE_NAME("button"); > > > @@ -376,6 +377,9 @@ static void acpi_lid_initialize_state(struct acpi_device *device) > > > case ACPI_BUTTON_LID_INIT_OPEN: > > > (void)acpi_lid_notify_state(device, 1); > > > break; > > > + case ACPI_BUTTON_LID_INIT_METHOD: > > > + (void)acpi_lid_update_state(device); > > > + break; > > > case ACPI_BUTTON_LID_INIT_IGNORE: > > > default: > > > break; > > > @@ -559,6 +563,9 @@ static int param_set_lid_init_state(const char *val, struct kernel_param *kp) > > > if (!strncmp(val, "open", sizeof("open") - 1)) { > > > lid_init_state = ACPI_BUTTON_LID_INIT_OPEN; > > > pr_info("Notify initial lid state as open\n"); > > > + } else if (!strncmp(val, "method", sizeof("method") - 1)) { > > > + lid_init_state = ACPI_BUTTON_LID_INIT_METHOD; > > > + pr_info("Notify initial lid state with _LID return value\n"); > > > } else if (!strncmp(val, "ignore", sizeof("ignore") - 1)) { > > > lid_init_state = ACPI_BUTTON_LID_INIT_IGNORE; > > > pr_info("Do not notify initial lid state\n"); > > > @@ -572,6 +579,8 @@ static int param_get_lid_init_state(char *buffer, struct kernel_param *kp) > > > switch (lid_init_state) { > > > case ACPI_BUTTON_LID_INIT_OPEN: > > > return sprintf(buffer, "open"); > > > + case ACPI_BUTTON_LID_INIT_METHOD: > > > + return sprintf(buffer, "method"); > > > case ACPI_BUTTON_LID_INIT_IGNORE: > > > return sprintf(buffer, "ignore"); > > > default: > > > -- > > > 2.9.3 > > > > -- > > To unsubscribe from this list: send the line "unsubscribe linux-acpi" in > > the body of a message to majordomo@vger.kernel.org > > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi, > From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-owner@vger.kernel.org] On Behalf Of Benjamin > Tissoires > Subject: Re: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > > On May 11 2017 or thereabouts, Zheng, Lv wrote: > > Hi, Benjamin > > > > > From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-owner@vger.kernel.org] On Behalf Of > Zheng, > > > Lv > > > Subject: RE: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > > > > > > Hi, Benjiamin > > > > > > > From: Benjamin Tissoires [mailto:benjamin.tissoires@redhat.com] > > > > Sent: Thursday, May 11, 2017 12:13 AM > > > > To: Rafael J . Wysocki <rjw@rjwysocki.net>; Zheng, Lv <lv.zheng@intel.com> > > > > Cc: Jiri Eischmann <jeischma@redhat.com>; linux-acpi@vger.kernel.org; linux- > kernel@vger.kernel.org > > > > Subject: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > > > > > > > > This reverts commit ecb10b694b72ca5ea51b3c90a71ff2a11963425a. > > > > > > > > Even if the method can be buggy some times, it's still a need > > > > when you boot a laptop with a lid closed and an external monitor > > > > connected (typical situation when using a docking station) > > > > > > > > Note: this option has been removed without being deprecated, which > > > > is terrible in term of distribution handling. The new default "open" > > > > is plain wrong and we don't even have the chance to keep the current > > > > default option because it's not there anymore. > > > > > > I have reverted this: > > > https://patchwork.kernel.org/patch/9717109/ > > Yeah, sorry, as mentioned to Rafael, I only saw it after I sent my > series. > > > > > > > Also I noticed one thing: > > > https://patchwork.kernel.org/patch/9717111/ > > > > > > After I changed kernel lid notification to always send lid return value to other drivers. > > > In order to find out a single driver mode (without platform quirks) to handle both cases. > > > I failed. > > Yes. As long as you do not have more information on the device, this is > not something you can solve with hacks in the kernel. > > > > I should still send close init state to the user space program to work around the external monitor > > > issues. > > > > > > So as you know, we need to send "open" init state to the user space program to work around > > > suspend/resume loop issue. > > I disagree. You are solving a user space issue (suspend boot loop), with > the wrong tool. I'm not trying to solve it. I'm trying to confirm, it’s the input event not the lid notifier event triggered the problem. > The user space expects the kernel to provide accurate > events, but if the hardware is wrong, we should either fix it > (overwriting AML tables), I don't think it's hardware related. If it's hardware related, we should always be able to fix it. IMO, it's firmware related. And the firmware only provides functionalities Windows requires. It cares no about what the Linux user space requires. We have no mean to prevent future firmwares from breaking Linux user space again. So if it trends to break in AML, introducing quirks to fix the old AMLs doesn't help to solve the problem. > make reasonable assumptions based on the exact > model capabilities (is the power button accessible with the LID closed), Sounds it's always accessible. > or teach user space about these situations. We are trying. But if the end users don't buy it, no user space programs will be interested in looking into the details of these issues. So we need to teach end users first IMO. Telling them to raise the issues to the right responsibles rather than acpi button driver. Some end users even make reports according to some BKMs (it has been changed now after discussed on ACPI Bugzilla bug): https://wiki.archlinux.org/index.php/Razer#Suspend_Loop > There is no point in assuming wrong states in the kernel "to fix user > space" when user space is obviously not doing the right thing. Agreed, as we've already synchronized. :) But the only problem is: We can see flooded ACPI button driver problems reported on kernel Bugzilla. That's why we changed the default behavior of button driver. It really helped to mute all suspend/resume loop issue reports. > > > > > > > Then for such platforms, how can ACPI button driver automatically detect if an external monitor is > > > there? > > That's not the ACPI button role. > However, user space can write to the switch to overwrite it's value when > it judges that the kernel is doing things wrong. Libinput is a pretty > good candidate, given it has a view of all input devices. And guess > what, the code is already there. You only told us the way the user space should use to fix the issue. But didn't tell us who should be responsible of fixing this issue. > > > > Unless we fix the BIOS code to make lid return value work as user space's expectation. > > That would be great. That's not the real fix. Just another kind of work around and cannot prevent future bug reports. > > > > OK, then this creates an endless business in ACPI community to "re-develop" BIOS tables if they > cannot > > > meat user space's expectation. > > > That sucks. > > Yes, but unless you teach OEM to not do crap, that's our daily burden. > I'd love to not have to quirk endlessly all the drivers I maintain, but > each generation of new devices has a new creative way of breaking the > existing code, "because it works under Windows". Or we can teach user space to not do any expectation other than what Windows supports on OEMs. > > > > > > > It sound the best way is the user space program equipped with hwdb quirks who knows everything to > > > alter acpi button driver mode from user space to work around this. > > Yes, the hwdb entry is the solution (or a quirk list in acpi/button.c). I'd rather to see such quirks accumulated in the user side not in the button driver. > The advantage of the hwdb entry is that it will be asynchronous from the > kernel releases, and users can just drop a file in their /etc folder and > they solved the issue. Distribution will also be able to carry the list > of quirked devices, and hopefully end users won't see the boot loop. The advantage sounds good. > > > > For example: > > > > > > If hwdb is hit, and there is no external monitor, then > > > Echo "open" > /sys/module/button/parameters/lid_init_state > > > If hwdb is not hit or there is an external monitor, then > > > If hwdb is hit, and there is no external monitor, then > > > Echo "close" > /sys/module/button/parameters/lid_init_state > > Hum, no. This is too late. acpi/button.c is loaded before udev hits, so > the initial state will already be evaluated. But this can affect post-resume behavior. And the bug reports are related to the post-resume behaviors. > > > > > Let me do re-wording. > > If hwdb is not hit > > echo "method" > /sys/module/button/parameters/lid_init_state > > If hwdb is hit, and there is no external monitor, then > > echo "open" > /sys/module/button/parameters/lid_init_state > > If hwdb is hit, and there is an external monitor > > echo "close" > /sys/module/button/parameters/lid_init_state > > Then this always assumes the hard requirements of the platform quirks. > > And it then looks it's better to do such switches in the user space as ACPI button driver doesn't > know and doesn't have to know the existence of the external monitor. > > Again, the external monitor doesn't matter here. The external monitor > issue is a user space choice to: > - not suspend if the LID is closed and a monitor is plugged > - only show the greater on the internal monitor if both are turned on. Which program turned the lid off when the lid was closed? > > The issue we are fixing here is the fact that the switch state is wrong, > which makes user space assumptions wrong too (and users angry). Considering no platform quirks. If ACPI button driver sends SW_LID, users are likely angry. Unless the user space programs are changed to "ignore open event". If ACPI button driver doesn't send switch events, but key events. The user space programs need to change to handle the new events. So finally whatever we do, user space need to change a bit related to ACPI control method lid device. > > But given that the LID switch is an actual input switch device, user > space can overwrite it by simply writing to the input node. I can see what you mean here. You are suggesting to use input overwrite rather than changing lid_init_state to fix the issue. This also means to me: user space is able to fix everything on its own, ACPI button driver needn't participant in. > > > > > > However, is that possible to not introduce platform quirks? > > For example, for systemd: > > If it detected ACPI lid device, automatically switch to an "IgnoreLidOpenEvent" mode (like nouveau > drivers' ignorelid=Y option). > > Well, given that 99.9% of laptops have this ACPI lid button, you'll just > remove the feature from all laptops. No, I only removed the wrong usage models related to the strict "open" event. All laptops are still capable of sending correct "close" event. > > > BTW, which program is responsible for lighting on internal/external monitors? > > I would say the compositor or X, so gdm, kdm, gnome, kde, etc... So I already raised this to freedesktop: https://bugs.freedesktop.org/show_bug.cgi?id=100923 But couldn't see any one there responding. > > > Can it determine that by its own without listening to the lid key events? > > Basically no. This switch is there for a reason. However, I am convinced > that a good heuristic is to say that if you are using the internal keyboard, > touchscreen or touchpad, unless the user has very very thin fingers, the > lid is open. I'm also convinced that the benefit of having a file in sysfs/procfs to allow user to know if the lid is open is marginal. > > > This is what we preferred. > > If all of above usage models are corrected, we'll change acpi button driver default mode to "ignore". > > No, we need to report accurate state, or explicitly mark the platform as > not reliable, and so we need "method" and a hwdb of problematic ones. I'm actually OK with any default value. But like previously I don't want to send a patch to change the default behavior to "open". I don't want to be the one to change the default behavior back to "method". > > > > > Another problem for changing default mode back to "method" is: > > If we changed button driver default mode back to "method", then ACPI community will be flooded of > suspend/resume loop bug reports. > > But that's your job to fix bugs. If there is a user space problem that > can be solved in user space, you just need to redirect the users to the > correct solution and close the bug. > > But you are talking about "flood", and I don't think we ever talked > about more than 4 devices. So could you point me at a list of bugs that > you actually had to fix? I just opened my Bugzilla filter, and copied top most lid related ones here: https://bugzilla.kernel.org/show_bug.cgi?id=187271 https://bugzilla.kernel.org/show_bug.cgi?id=192231 https://bugzilla.kernel.org/show_bug.cgi?id=191211 https://bugzilla.kernel.org/show_bug.cgi?id=189171 https://bugzilla.kernel.org/show_bug.cgi?id=116001 ... And old reports here, occasionally opened by different users: https://bugzilla.kernel.org/show_bug.cgi?id=89211 https://bugzilla.kernel.org/show_bug.cgi?id=106151 https://bugzilla.kernel.org/show_bug.cgi?id=106941 https://bugzilla.novell.com/show_bug.cgi?id=326814 Do you think the kernel community should prepare a candidate to handle such bug flood? I'm unfortunately the one changed button driver recently and have to handle them. That's why I changed default behavior from method to open. After upstream merged that default behavior change, no such reports could be seen again. But it's just 1-2 months Bugzilla silence before seeing a different bug flood trend: This time, they are related to the external monitor usage model: https://bugzilla.kernel.org/show_bug.cgi?id=187271 https://bugzilla.redhat.com/show_bug.cgi?id=1430259 https://bugzilla.kernel.org/show_bug.cgi?id=195455 Now that you can see why I didn't send a patch to change the default behavior to "open" at the time we were discussing. That's also why I think we needn't revert back to "method" as the default behavior. I'm afraid you may need to be prepared to handle some of the reports if you revert the default behavior back to "method". So if you insist, I can put my Acked-by to your PATCH 2/2. > > > After we root cause that's not a kernel problem, do we have mean in other community to handle such > reports? > > For example, to collect hwdb entries. > > libinput, systemd are good candidate. > Libinput already has the bits in place, so I'd say we should probably > ask the users to report a bug on the wayland/libinput component of > bugs.freedesktop.org. But this will only work if the default initial lid > state is "method". Good to know. :) I've just changed the category of the forwarded report. > > Sorry for showing I am angry. But I thought we solved this months ago > and this bites back. You always help me a lot, appreciated! Cheers Lv > > Cheers, > Benjamin > > > > > Thanks and best regards > > Lv > > > > > > > > So PATCH 2 is not useful. > > > Reverting that can trigger a regression loop we surely do not want to handle. > > > > > > Thanks and best regards > > > Lv > > > > > > > > > > > Cc: stable@vger.kernel.org > > > > Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com> > > > > --- > > > > Documentation/acpi/acpi-lid.txt | 16 ++++++++++++---- > > > > drivers/acpi/button.c | 9 +++++++++ > > > > 2 files changed, 21 insertions(+), 4 deletions(-) > > > > > > > > diff --git a/Documentation/acpi/acpi-lid.txt b/Documentation/acpi/acpi-lid.txt > > > > index 22cb309..effe7af 100644 > > > > --- a/Documentation/acpi/acpi-lid.txt > > > > +++ b/Documentation/acpi/acpi-lid.txt > > > > @@ -59,20 +59,28 @@ button driver uses the following 3 modes in order not to trigger issues. > > > > If the userspace hasn't been prepared to ignore the unreliable "opened" > > > > events and the unreliable initial state notification, Linux users can use > > > > the following kernel parameters to handle the possible issues: > > > > -A. button.lid_init_state=open: > > > > +A. button.lid_init_state=method: > > > > + When this option is specified, the ACPI button driver reports the > > > > + initial lid state using the returning value of the _LID control method > > > > + and whether the "opened"/"closed" events are paired fully relies on the > > > > + firmware implementation. > > > > + This option can be used to fix some platforms where the returning value > > > > + of the _LID control method is reliable but the initial lid state > > > > + notification is missing. > > > > + This option is the default behavior during the period the userspace > > > > + isn't ready to handle the buggy AML tables. > > > > +B. button.lid_init_state=open: > > > > When this option is specified, the ACPI button driver always reports the > > > > initial lid state as "opened" and whether the "opened"/"closed" events > > > > are paired fully relies on the firmware implementation. > > > > This may fix some platforms where the returning value of the _LID > > > > control method is not reliable and the initial lid state notification is > > > > missing. > > > > - This option is the default behavior during the period the userspace > > > > - isn't ready to handle the buggy AML tables. > > > > > > > > If the userspace has been prepared to ignore the unreliable "opened" events > > > > and the unreliable initial state notification, Linux users should always > > > > use the following kernel parameter: > > > > -B. button.lid_init_state=ignore: > > > > +C. button.lid_init_state=ignore: > > > > When this option is specified, the ACPI button driver never reports the > > > > initial lid state and there is a compensation mechanism implemented to > > > > ensure that the reliable "closed" notifications can always be delievered > > > > diff --git a/drivers/acpi/button.c b/drivers/acpi/button.c > > > > index 668137e..6d5a8c1 100644 > > > > --- a/drivers/acpi/button.c > > > > +++ b/drivers/acpi/button.c > > > > @@ -57,6 +57,7 @@ > > > > > > > > #define ACPI_BUTTON_LID_INIT_IGNORE 0x00 > > > > #define ACPI_BUTTON_LID_INIT_OPEN 0x01 > > > > +#define ACPI_BUTTON_LID_INIT_METHOD 0x02 > > > > > > > > #define _COMPONENT ACPI_BUTTON_COMPONENT > > > > ACPI_MODULE_NAME("button"); > > > > @@ -376,6 +377,9 @@ static void acpi_lid_initialize_state(struct acpi_device *device) > > > > case ACPI_BUTTON_LID_INIT_OPEN: > > > > (void)acpi_lid_notify_state(device, 1); > > > > break; > > > > + case ACPI_BUTTON_LID_INIT_METHOD: > > > > + (void)acpi_lid_update_state(device); > > > > + break; > > > > case ACPI_BUTTON_LID_INIT_IGNORE: > > > > default: > > > > break; > > > > @@ -559,6 +563,9 @@ static int param_set_lid_init_state(const char *val, struct kernel_param *kp) > > > > if (!strncmp(val, "open", sizeof("open") - 1)) { > > > > lid_init_state = ACPI_BUTTON_LID_INIT_OPEN; > > > > pr_info("Notify initial lid state as open\n"); > > > > + } else if (!strncmp(val, "method", sizeof("method") - 1)) { > > > > + lid_init_state = ACPI_BUTTON_LID_INIT_METHOD; > > > > + pr_info("Notify initial lid state with _LID return value\n"); > > > > } else if (!strncmp(val, "ignore", sizeof("ignore") - 1)) { > > > > lid_init_state = ACPI_BUTTON_LID_INIT_IGNORE; > > > > pr_info("Do not notify initial lid state\n"); > > > > @@ -572,6 +579,8 @@ static int param_get_lid_init_state(char *buffer, struct kernel_param *kp) > > > > switch (lid_init_state) { > > > > case ACPI_BUTTON_LID_INIT_OPEN: > > > > return sprintf(buffer, "open"); > > > > + case ACPI_BUTTON_LID_INIT_METHOD: > > > > + return sprintf(buffer, "method"); > > > > case ACPI_BUTTON_LID_INIT_IGNORE: > > > > return sprintf(buffer, "ignore"); > > > > default: > > > > -- > > > > 2.9.3 > > > > > > -- > > > To unsubscribe from this list: send the line "unsubscribe linux-acpi" in > > > the body of a message to majordomo@vger.kernel.org > > > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- > To unsubscribe from this list: send the line "unsubscribe linux-acpi" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html
On May 12 2017 or thereabouts, Zheng, Lv wrote: > Hi, > > > From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-owner@vger.kernel.org] On Behalf Of Benjamin > > Tissoires > > Subject: Re: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > > > > On May 11 2017 or thereabouts, Zheng, Lv wrote: > > > Hi, Benjamin > > > > > > > From: linux-acpi-owner@vger.kernel.org [mailto:linux-acpi-owner@vger.kernel.org] On Behalf Of > > Zheng, > > > > Lv > > > > Subject: RE: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > > > > > > > > Hi, Benjiamin > > > > > > > > > From: Benjamin Tissoires [mailto:benjamin.tissoires@redhat.com] > > > > > Sent: Thursday, May 11, 2017 12:13 AM > > > > > To: Rafael J . Wysocki <rjw@rjwysocki.net>; Zheng, Lv <lv.zheng@intel.com> > > > > > Cc: Jiri Eischmann <jeischma@redhat.com>; linux-acpi@vger.kernel.org; linux- > > kernel@vger.kernel.org > > > > > Subject: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > > > > > > > > > > This reverts commit ecb10b694b72ca5ea51b3c90a71ff2a11963425a. > > > > > > > > > > Even if the method can be buggy some times, it's still a need > > > > > when you boot a laptop with a lid closed and an external monitor > > > > > connected (typical situation when using a docking station) > > > > > > > > > > Note: this option has been removed without being deprecated, which > > > > > is terrible in term of distribution handling. The new default "open" > > > > > is plain wrong and we don't even have the chance to keep the current > > > > > default option because it's not there anymore. > > > > > > > > I have reverted this: > > > > https://patchwork.kernel.org/patch/9717109/ > > > > Yeah, sorry, as mentioned to Rafael, I only saw it after I sent my > > series. > > > > > > > > > > Also I noticed one thing: > > > > https://patchwork.kernel.org/patch/9717111/ > > > > > > > > After I changed kernel lid notification to always send lid return value to other drivers. > > > > In order to find out a single driver mode (without platform quirks) to handle both cases. > > > > I failed. > > > > Yes. As long as you do not have more information on the device, this is > > not something you can solve with hacks in the kernel. > > > > > > I should still send close init state to the user space program to work around the external monitor > > > > issues. > > > > > > > > So as you know, we need to send "open" init state to the user space program to work around > > > > suspend/resume loop issue. > > > > I disagree. You are solving a user space issue (suspend boot loop), with > > the wrong tool. > > I'm not trying to solve it. > I'm trying to confirm, it’s the input event not the lid notifier event triggered the problem. You would have asked, I would have answered without you spending too much time figuring this out. The reason is simple: the kernel should report an accurate state, no matter what. Given this statement, user space made some design decisions, and in that situation, the design differs from the Windows behavior. Which means OEM only check Windows and might break Linux. End of the story. The only viable way of not having boot/resume loops is to report accurate events/states, and for that you need a way to fix it. > > > The user space expects the kernel to provide accurate > > events, but if the hardware is wrong, we should either fix it > > (overwriting AML tables), > > I don't think it's hardware related. > If it's hardware related, we should always be able to fix it. Well, yeah, hardware/firmware, for me it's all before the kernel. And hardware would mean an actual failure of the sensor, so definitely not something we can fix at the software level. > > IMO, it's firmware related. > And the firmware only provides functionalities Windows requires. > It cares no about what the Linux user space requires. > We have no mean to prevent future firmwares from breaking Linux user space again. agree. > So if it trends to break in AML, introducing quirks to fix the old AMLs doesn't help to solve the problem. > > > make reasonable assumptions based on the exact > > model capabilities (is the power button accessible with the LID closed), > > Sounds it's always accessible. hmm, I am not sure we have the same fingers.... For me on all the laptops I have seen, if the LID is actually (physically) closed, I can not press the button. It's a design choice to not have anything powering on the laptop when it's in your bag. The reported state is something different, and it's software only. Remember that the _LID acpi method corresponds to a physical state of the laptop. So if the ACPI returned value is crap, that means user space should either ignore it, or it should be fixed. But we can't have a situation where we say, yes, it's reliable except few cases, without telling which cases are unreliable. > > > or teach user space about these situations. > > We are trying. > But if the end users don't buy it, no user space programs will be interested in looking into the details of these issues. > So we need to teach end users first IMO. > Telling them to raise the issues to the right responsibles rather than acpi button driver. I believe they are correct raising it in acpi button. Again, kernel should provide accurate state. It's not the case. Then, we can redirect them to the band-aid provided by libinput. > > Some end users even make reports according to some BKMs (it has been changed now after discussed on ACPI Bugzilla bug): > https://wiki.archlinux.org/index.php/Razer#Suspend_Loop > > > There is no point in assuming wrong states in the kernel "to fix user > > space" when user space is obviously not doing the right thing. > > Agreed, as we've already synchronized. :) > > But the only problem is: > We can see flooded ACPI button driver problems reported on kernel Bugzilla. > That's why we changed the default behavior of button driver. ... and you broke thousands to fix a handful :) > It really helped to mute all suspend/resume loop issue reports. No. It helped you in your daily work routine. It broke thousands of customers. > > > > > > > > > > > Then for such platforms, how can ACPI button driver automatically detect if an external monitor is > > > > there? > > > > That's not the ACPI button role. > > However, user space can write to the switch to overwrite it's value when > > it judges that the kernel is doing things wrong. Libinput is a pretty > > good candidate, given it has a view of all input devices. And guess > > what, the code is already there. > > You only told us the way the user space should use to fix the issue. > But didn't tell us who should be responsible of fixing this issue. "Who" is libinput if you saw above. And if you are interested in the actual names, that's Peter Hutterer and myself mostly. > > > > > > > Unless we fix the BIOS code to make lid return value work as user space's expectation. > > > > That would be great. > > That's not the real fix. According to the kernel policy, that's the only viable fix. > Just another kind of work around and cannot prevent future bug reports. Agree, that's a workaround, but you can not prevent OEMs to do crap. > > > > > > > OK, then this creates an endless business in ACPI community to "re-develop" BIOS tables if they > > cannot > > > > meat user space's expectation. > > > > That sucks. > > > > Yes, but unless you teach OEM to not do crap, that's our daily burden. > > I'd love to not have to quirk endlessly all the drivers I maintain, but > > each generation of new devices has a new creative way of breaking the > > existing code, "because it works under Windows". > > Or we can teach user space to not do any expectation other than what Windows supports on OEMs. > > > > > > > > > > > It sound the best way is the user space program equipped with hwdb quirks who knows everything to > > > > alter acpi button driver mode from user space to work around this. > > > > Yes, the hwdb entry is the solution (or a quirk list in acpi/button.c). > > I'd rather to see such quirks accumulated in the user side not in the button driver. > > > The advantage of the hwdb entry is that it will be asynchronous from the > > kernel releases, and users can just drop a file in their /etc folder and > > they solved the issue. Distribution will also be able to carry the list > > of quirked devices, and hopefully end users won't see the boot loop. > > The advantage sounds good. > > > > > > > For example: > > > > > > > > If hwdb is hit, and there is no external monitor, then > > > > Echo "open" > /sys/module/button/parameters/lid_init_state > > > > If hwdb is not hit or there is an external monitor, then > > > > If hwdb is hit, and there is no external monitor, then > > > > Echo "close" > /sys/module/button/parameters/lid_init_state > > > > Hum, no. This is too late. acpi/button.c is loaded before udev hits, so > > the initial state will already be evaluated. > > But this can affect post-resume behavior. > And the bug reports are related to the post-resume behaviors. Nothing prevents libinput to fix that. And I think it'll fix it too, because there is no differences between the boot false state and the resume false state: the state is wrong, and this can be detected. Also, nothing prevents to have the rule you wrote (overwriting lid_init_state) to have resume work even better. > > > > > > > > > Let me do re-wording. > > > If hwdb is not hit > > > echo "method" > /sys/module/button/parameters/lid_init_state > > > If hwdb is hit, and there is no external monitor, then > > > echo "open" > /sys/module/button/parameters/lid_init_state > > > If hwdb is hit, and there is an external monitor > > > echo "close" > /sys/module/button/parameters/lid_init_state > > > Then this always assumes the hard requirements of the platform quirks. > > > And it then looks it's better to do such switches in the user space as ACPI button driver doesn't > > know and doesn't have to know the existence of the external monitor. > > > > Again, the external monitor doesn't matter here. The external monitor > > issue is a user space choice to: > > - not suspend if the LID is closed and a monitor is plugged > > - only show the greater on the internal monitor if both are turned on. > > Which program turned the lid off when the lid was closed? I am not remotely sure of what you are asking here... > > > > > The issue we are fixing here is the fact that the switch state is wrong, > > which makes user space assumptions wrong too (and users angry). > > Considering no platform quirks. > > If ACPI button driver sends SW_LID, users are likely angry. > Unless the user space programs are changed to "ignore open event". > > If ACPI button driver doesn't send switch events, but key events. > The user space programs need to change to handle the new events. > > So finally whatever we do, user space need to change a bit related to ACPI control method lid device. Or we fix the switch to report accurate events/state. You do realise that all the energy you are spending, answering to me, talking to user space maintainers, users, all comes down to the fact that you refuse to have hardware quirks? If you don't want to write/maintain the code, fine. I can do it, but please stop trying to make everyone else change because you just don't want the burden of quirking a handful of laptops. > > > > > But given that the LID switch is an actual input switch device, user > > space can overwrite it by simply writing to the input node. > > I can see what you mean here. > You are suggesting to use input overwrite rather than changing lid_init_state to fix the issue. > > This also means to me: user space is able to fix everything on its own, ACPI button driver needn't participant in. Well, fixing it in the kernel means you don't need user space to fix it, which is nice too on systems where there is no libinput. But on such systems, maybe they just don't care and do something different. Having a fix in user space has advantages too (see above), but for it to be working, you need to stop changing the kernel behavior and the kernel API. > > > > > > > > > > > However, is that possible to not introduce platform quirks? > > > For example, for systemd: > > > If it detected ACPI lid device, automatically switch to an "IgnoreLidOpenEvent" mode (like nouveau > > drivers' ignorelid=Y option). > > > > Well, given that 99.9% of laptops have this ACPI lid button, you'll just > > remove the feature from all laptops. > > No, I only removed the wrong usage models related to the strict "open" event. > All laptops are still capable of sending correct "close" event. My bad, I read too fast and missed the "...Open..." part of "IgnoreLidOpenEvent". Though I am not sure IgnoreLidOpenEvent is accurate. "OpenEventNeverSent" seems to reflect more the reality. But again, that's not of our (kernel developers) business. > > > > > > BTW, which program is responsible for lighting on internal/external monitors? > > > > I would say the compositor or X, so gdm, kdm, gnome, kde, etc... > > So I already raised this to freedesktop: > https://bugs.freedesktop.org/show_bug.cgi?id=100923 > But couldn't see any one there responding. Well, Joachim answered, saying that there is likely a regression in acpi/button.c, not in i915. I guess you presented the problem in the wrong term: what you want is developer to fix the acpi_lid_open() call. All you need to do is explaining them that the call might not be accurate. And I also believe the solution is not to tweak i915 or nouveau driver, but make sure acpi/button does the right thing. > > > > > > Can it determine that by its own without listening to the lid key events? > > > > Basically no. This switch is there for a reason. However, I am convinced > > that a good heuristic is to say that if you are using the internal keyboard, > > touchscreen or touchpad, unless the user has very very thin fingers, the > > lid is open. > > I'm also convinced that the benefit of having a file in sysfs/procfs to allow user to know if the lid is open is marginal. I am convinced I don't get your point. We are obviously not talking about the same thing here. I was talking about the physical world, where the user interact with the laptop... > > > > > > This is what we preferred. > > > If all of above usage models are corrected, we'll change acpi button driver default mode to "ignore". > > > > No, we need to report accurate state, or explicitly mark the platform as > > not reliable, and so we need "method" and a hwdb of problematic ones. > > I'm actually OK with any default value. Thanks. > But like previously I don't want to send a patch to change the default behavior to "open". > I don't want to be the one to change the default behavior back to "method". I already did. So feel free to point finger at me if somebody else complains. > > > > > > > > > Another problem for changing default mode back to "method" is: > > > If we changed button driver default mode back to "method", then ACPI community will be flooded of > > suspend/resume loop bug reports. > > > > But that's your job to fix bugs. If there is a user space problem that > > can be solved in user space, you just need to redirect the users to the > > correct solution and close the bug. > > > > But you are talking about "flood", and I don't think we ever talked > > about more than 4 devices. So could you point me at a list of bugs that > > you actually had to fix? > > I just opened my Bugzilla filter, and copied top most lid related ones here: > https://bugzilla.kernel.org/show_bug.cgi?id=187271 > https://bugzilla.kernel.org/show_bug.cgi?id=192231 > https://bugzilla.kernel.org/show_bug.cgi?id=191211 > https://bugzilla.kernel.org/show_bug.cgi?id=189171 > https://bugzilla.kernel.org/show_bug.cgi?id=116001 That's 3 new laptops I wasn't aware of: - Razer Blade Stealth 2016 - SAMSUNG 530U3C/530U4C - Razerblade 2014 The Dell one is different as it seems the EC is doing things wrong, and do not forward the notification even if the _LID acpi method returns the proper value. > ... > > And old reports here, occasionally opened by different users: > https://bugzilla.kernel.org/show_bug.cgi?id=89211 MS Surface Pro 1, already known > https://bugzilla.kernel.org/show_bug.cgi?id=106151 Samsung N210 Plus something like 8 or 10 year old laptop, already known > https://bugzilla.kernel.org/show_bug.cgi?id=106941 Surface 3, already fixed in the kernel by myself (drivers/platform/x86/surface3-wmi.c) > https://bugzilla.novell.com/show_bug.cgi?id=326814 Lenovo X60, at least 5 years old one, recent models don't have any issue. > > Do you think the kernel community should prepare a candidate to handle such bug flood? Well, I don't mind receiving such "flood" in my face: 7 laptops in 10 years seems something manageable. And also looks like Razer has a common way of being wrong, so we can just match on RazerBlade to fix all current and future generations. > I'm unfortunately the one changed button driver recently and have to handle them. > That's why I changed default behavior from method to open. > After upstream merged that default behavior change, no such reports could be seen again. Well, that is not true. Once a change gets in linux-acpi, you need to wait for the next merge window for it to be in linus' tree to get wider testing. Then you need to wait for the release to have distributions taking it and actually have broader testers and most use cases tackled. And this corresponds to the 2 months you had between your submission and the v4.11 release, where fedora picked it in Fedora 26, and where our users started testing it intensively. > > But it's just 1-2 months Bugzilla silence before seeing a different bug flood trend: > This time, they are related to the external monitor usage model: > https://bugzilla.kernel.org/show_bug.cgi?id=187271 > https://bugzilla.redhat.com/show_bug.cgi?id=1430259 > https://bugzilla.kernel.org/show_bug.cgi?id=195455 > Now that you can see why I didn't send a patch to change the default behavior to "open" at the time we were discussing. > That's also why I think we needn't revert back to "method" as the default behavior. Still, I strongly disagree. We do not fix 7 devices by breaking hundreds. That's just not how the kernel work. > > I'm afraid you may need to be prepared to handle some of the reports if you revert the default behavior back to "method". > So if you insist, I can put my Acked-by to your PATCH 2/2. I insist, and I am not afraid. I'd rather be facing those than having someone else breaking a single switch in the system in a way we can't do anything else to solve the issue in user space. > > > > > > After we root cause that's not a kernel problem, do we have mean in other community to handle such > > reports? > > > For example, to collect hwdb entries. > > > > libinput, systemd are good candidate. > > Libinput already has the bits in place, so I'd say we should probably > > ask the users to report a bug on the wayland/libinput component of > > bugs.freedesktop.org. But this will only work if the default initial lid > > state is "method". > > Good to know. :) > I've just changed the category of the forwarded report. Well, I am not sure your bug report should have been changed. The bug report was regarding i915 being confident in the _LID acpi call, and that is not something libinput can change. > > > > > Sorry for showing I am angry. But I thought we solved this months ago > > and this bites back. > > You always help me a lot, appreciated! > Thanks for your kindness :) Would you agree on: - acking both my patches so Rafael can take them ans stable can apply them ASAP - give me maintainership of drivers/acpi/button.c (ACK a following patch where I add me to MAINTAINERS) - redirect all kernel bugs to me related to LID switch (or having me the default assignee for a new acpi/button component) - allow me to add a list of quirk in acpi/button.c if I judge this as fit If you agree on that, that would be good for both our sake I think. Cheers, Benjamin > Cheers > Lv > > > > > Cheers, > > Benjamin > > > > > > > > Thanks and best regards > > > Lv > > > > > > > > > > > So PATCH 2 is not useful. > > > > Reverting that can trigger a regression loop we surely do not want to handle. > > > > > > > > Thanks and best regards > > > > Lv > > > > > > > > > > > > > > Cc: stable@vger.kernel.org > > > > > Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com> > > > > > --- > > > > > Documentation/acpi/acpi-lid.txt | 16 ++++++++++++---- > > > > > drivers/acpi/button.c | 9 +++++++++ > > > > > 2 files changed, 21 insertions(+), 4 deletions(-) > > > > > > > > > > diff --git a/Documentation/acpi/acpi-lid.txt b/Documentation/acpi/acpi-lid.txt > > > > > index 22cb309..effe7af 100644 > > > > > --- a/Documentation/acpi/acpi-lid.txt > > > > > +++ b/Documentation/acpi/acpi-lid.txt > > > > > @@ -59,20 +59,28 @@ button driver uses the following 3 modes in order not to trigger issues. > > > > > If the userspace hasn't been prepared to ignore the unreliable "opened" > > > > > events and the unreliable initial state notification, Linux users can use > > > > > the following kernel parameters to handle the possible issues: > > > > > -A. button.lid_init_state=open: > > > > > +A. button.lid_init_state=method: > > > > > + When this option is specified, the ACPI button driver reports the > > > > > + initial lid state using the returning value of the _LID control method > > > > > + and whether the "opened"/"closed" events are paired fully relies on the > > > > > + firmware implementation. > > > > > + This option can be used to fix some platforms where the returning value > > > > > + of the _LID control method is reliable but the initial lid state > > > > > + notification is missing. > > > > > + This option is the default behavior during the period the userspace > > > > > + isn't ready to handle the buggy AML tables. > > > > > +B. button.lid_init_state=open: > > > > > When this option is specified, the ACPI button driver always reports the > > > > > initial lid state as "opened" and whether the "opened"/"closed" events > > > > > are paired fully relies on the firmware implementation. > > > > > This may fix some platforms where the returning value of the _LID > > > > > control method is not reliable and the initial lid state notification is > > > > > missing. > > > > > - This option is the default behavior during the period the userspace > > > > > - isn't ready to handle the buggy AML tables. > > > > > > > > > > If the userspace has been prepared to ignore the unreliable "opened" events > > > > > and the unreliable initial state notification, Linux users should always > > > > > use the following kernel parameter: > > > > > -B. button.lid_init_state=ignore: > > > > > +C. button.lid_init_state=ignore: > > > > > When this option is specified, the ACPI button driver never reports the > > > > > initial lid state and there is a compensation mechanism implemented to > > > > > ensure that the reliable "closed" notifications can always be delievered > > > > > diff --git a/drivers/acpi/button.c b/drivers/acpi/button.c > > > > > index 668137e..6d5a8c1 100644 > > > > > --- a/drivers/acpi/button.c > > > > > +++ b/drivers/acpi/button.c > > > > > @@ -57,6 +57,7 @@ > > > > > > > > > > #define ACPI_BUTTON_LID_INIT_IGNORE 0x00 > > > > > #define ACPI_BUTTON_LID_INIT_OPEN 0x01 > > > > > +#define ACPI_BUTTON_LID_INIT_METHOD 0x02 > > > > > > > > > > #define _COMPONENT ACPI_BUTTON_COMPONENT > > > > > ACPI_MODULE_NAME("button"); > > > > > @@ -376,6 +377,9 @@ static void acpi_lid_initialize_state(struct acpi_device *device) > > > > > case ACPI_BUTTON_LID_INIT_OPEN: > > > > > (void)acpi_lid_notify_state(device, 1); > > > > > break; > > > > > + case ACPI_BUTTON_LID_INIT_METHOD: > > > > > + (void)acpi_lid_update_state(device); > > > > > + break; > > > > > case ACPI_BUTTON_LID_INIT_IGNORE: > > > > > default: > > > > > break; > > > > > @@ -559,6 +563,9 @@ static int param_set_lid_init_state(const char *val, struct kernel_param *kp) > > > > > if (!strncmp(val, "open", sizeof("open") - 1)) { > > > > > lid_init_state = ACPI_BUTTON_LID_INIT_OPEN; > > > > > pr_info("Notify initial lid state as open\n"); > > > > > + } else if (!strncmp(val, "method", sizeof("method") - 1)) { > > > > > + lid_init_state = ACPI_BUTTON_LID_INIT_METHOD; > > > > > + pr_info("Notify initial lid state with _LID return value\n"); > > > > > } else if (!strncmp(val, "ignore", sizeof("ignore") - 1)) { > > > > > lid_init_state = ACPI_BUTTON_LID_INIT_IGNORE; > > > > > pr_info("Do not notify initial lid state\n"); > > > > > @@ -572,6 +579,8 @@ static int param_get_lid_init_state(char *buffer, struct kernel_param *kp) > > > > > switch (lid_init_state) { > > > > > case ACPI_BUTTON_LID_INIT_OPEN: > > > > > return sprintf(buffer, "open"); > > > > > + case ACPI_BUTTON_LID_INIT_METHOD: > > > > > + return sprintf(buffer, "method"); > > > > > case ACPI_BUTTON_LID_INIT_IGNORE: > > > > > return sprintf(buffer, "ignore"); > > > > > default: > > > > > -- > > > > > 2.9.3 > > > > > > > > -- > > > > To unsubscribe from this list: send the line "unsubscribe linux-acpi" in > > > > the body of a message to majordomo@vger.kernel.org > > > > More majordomo info at http://vger.kernel.org/majordomo-info.html > > -- > > To unsubscribe from this list: send the line "unsubscribe linux-acpi" in > > the body of a message to majordomo@vger.kernel.org > > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi, Benjamin Let's stop endless discussing and focus on our needs. I just copied my questions here. You can ask them directly. For the below inlined replies, I'll stop replying if they are based on dependent on our basic agreements. And I'll reply if something is really bad from my point of view. My point of view is: There is a gap between (BIOS ensured/Windows expected) acpi control method lid device behavior and Linux user space expected acpi control method lid device behavior. Button driver default behavior should be: button.lid_init_state=ignore If user space programs have special needs, they can fix problems on their own, via the following mean: echo -n "open" > /sys/modules/button/parameters/lid_init_state echo -n "close" > /sys/modules/button/parameters/lid_init_state Keeping open/close modes is because I don't think there is any bug in button driver. So I need to prepare quirk modes from button driver's point of view and use them as a response to related bug reports reported in acpi community. Your point of view is: There is a gap between (BIOS ensured/Windows expected) acpi control method lid device behavior and Linux user space expected acpi control method lid device behavior. Button driver default behavior should be (not 100% sure if this is your opinion): button.lid_init_state=method If user space programs have special needs, they can fix them on their own, via the following mean: libinput:name:*Lid Switch*:dmi:*svnMicrosoftCorporation:pnSurface3:* LIBINPUT_ATTR_LID_SWITCH_RELIABILITY=write_open From this point of view, we actually don't need open/close modes. It seems we just need to determine the following first: 1. Who should be responsible for solving bugs triggered by the conflict between bios and linux user space expectations: button driver? libinput? Some other user space programs? Users? 2. What should be the default button driver behavior? button.lid_init_state=ignore? button.lid_init_state=method? 3. If non button driver quirks are working, button driver quirk modes are useless. The question is: Should button.lid_init_state=open/close be kept? 4. From button driver's point of view, button.lid_init_state=ignore seems to be always correct, so we won't abandon it. If we can use libinput to manage platform quirks, then button.lid_init_state=method also looks useless. The question is: Should button.lid_init_state=method be kept? I should also let you know my preference: 1. using button.lid_init_state=ignore and button.lid_init_state=method as default behavior is ok for me if answer to 1 is not button driver, otherwise using button.lid_init_state=method is not ok for me 2. deletion of button.lid_init_state=open/close is ok for me if answer to 1 is not button driver, otherwise deletion of button.lid_init_state=open/close is not ok for me 3. deletion of button.lid_init_state=method is always ok for me See the base line from my side is very clear: If acpi community need to handle such bug reports, button.lid_init_state=method cannot be the default behavior. We are just using a different default behavior than "method" to drive things to reach the final root caused solution. Could you let me know your preference so that we can figure out an agreement between us. Though I don't know if end users will buy it (they may keep on filing regression reports in ACPI community). > From: Benjamin Tissoires [mailto:benjamin.tissoires@redhat.com] > Subject: Re: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > ... Skip as it looks the result of debating here can also be concluded from the answers of the questions. Let's just stop arguing same thing again and again using different contexts. > > > make reasonable assumptions based on the exact > > > model capabilities (is the power button accessible with the LID closed), > > > > Sounds it's always accessible. > > hmm, I am not sure we have the same fingers.... For me on all the > laptops I have seen, if the LID is actually (physically) closed, I can > not press the button. It's a design choice to not have anything powering > on the laptop when it's in your bag. Wow... I can see that recent laptops are having their power buttons on the edge, not on the keypad. For recent laptop/tablet 2-in-1 PCs or touch laptops which can have their keyboards folded and act notebooks, the power buttons are obviously not on the detachable keypads. Also for the laptops supporting external monitors, obviously allow users to push power buttons while the lid is closed. > ... Skip as it looks the result of debating here can also be concluded from the answers of the questions. > > > The issue we are fixing here is the fact that the switch state is wrong, > > > which makes user space assumptions wrong too (and users angry). > > > > Considering no platform quirks. > > > > If ACPI button driver sends SW_LID, users are likely angry. > > Unless the user space programs are changed to "ignore open event". > > > > If ACPI button driver doesn't send switch events, but key events. > > The user space programs need to change to handle the new events. > > > > So finally whatever we do, user space need to change a bit related to ACPI control method lid device. > > Or we fix the switch to report accurate events/state. > You do realise that all the energy you are spending, answering to me, > talking to user space maintainers, users, all comes down to the fact > that you refuse to have hardware quirks? > > If you don't want to write/maintain the code, fine. I can do it, but > please stop trying to make everyone else change because you just don't > want the burden of quirking a handful of laptops. But I have a very interesting question to ask: Can button driver/libinput actually fix this without modifying BIOS: It seems we both rely on user's decision to use proper quirk modes for a specific usage model. For example, for this case: A BIOS never sends lid notification after boot/resume, and _LID return value is close after boot/resume. Let's see which quirk mode should the user choose before suspending: 1. with no external monitor attached, user should: write_open/button.lid_init_state=open 2. with external monitor attached, if user wants the lid to be lit on write_open/button.lid_init_state=open 3. with external monitor attached, if user wants the lid not to be lit on write_close/button.lid_init_state=close See there is no single default value for all usage models. Thus there is no possible __FIX__ for acpi button driver and libinput. While user space programs can just fix their usage models. So finally we actually cannot fix anything by maintaining the quirk table, but just create a business of maintaining the quirk table. > ... Skip as it looks the result of debating here can also be concluded from the answers of the questions. > > > > However, is that possible to not introduce platform quirks? > > > > For example, for systemd: > > > > If it detected ACPI lid device, automatically switch to an "IgnoreLidOpenEvent" mode (like > nouveau > > > drivers' ignorelid=Y option). > > > > > > Well, given that 99.9% of laptops have this ACPI lid button, you'll just > > > remove the feature from all laptops. > > > > No, I only removed the wrong usage models related to the strict "open" event. > > All laptops are still capable of sending correct "close" event. > > My bad, I read too fast and missed the "...Open..." part of > "IgnoreLidOpenEvent". > > Though I am not sure IgnoreLidOpenEvent is accurate. > "OpenEventNeverSent" seems to reflect more the reality. But again, > that's not of our (kernel developers) business. IMO this is the only root cause fix. :) It's the only way that the user can use without changing its quirk modes for different usage models. > > So I already raised this to freedesktop: > > https://bugs.freedesktop.org/show_bug.cgi?id=100923 > > But couldn't see any one there responding. > > Well, Joachim answered, saying that there is likely a regression in > acpi/button.c, not in i915. He surely can add more details as he is responsible of triage of the referenced bug reported on redhat. But there is no agreement reached yet. > > > > Can it determine that by its own without listening to the lid key events? > > > > > > Basically no. This switch is there for a reason. However, I am convinced > > > that a good heuristic is to say that if you are using the internal keyboard, > > > touchscreen or touchpad, unless the user has very very thin fingers, the > > > lid is open. > > > > I'm also convinced that the benefit of having a file in sysfs/procfs to allow user to know if the > lid is open is marginal. > > I am convinced I don't get your point. We are obviously not talking > about the same thing here. I was talking about the physical world, where > the user interact with the laptop... See my previous reply against accessing power buttons when lid is closed. > ... Skip as it looks the result of debating here can also be concluded from the answers of the questions. > > But it's just 1-2 months Bugzilla silence before seeing a different bug flood trend: > > This time, they are related to the external monitor usage model: > > https://bugzilla.kernel.org/show_bug.cgi?id=187271 > > https://bugzilla.redhat.com/show_bug.cgi?id=1430259 > > https://bugzilla.kernel.org/show_bug.cgi?id=195455 > > Now that you can see why I didn't send a patch to change the default behavior to "open" at the time > we were discussing. > > That's also why I think we needn't revert back to "method" as the default behavior. > > Still, I strongly disagree. We do not fix 7 devices by breaking > hundreds. That's just not how the kernel work. No. The word of "broken" is entirely wrong and emotional here. If both user space and kernel space are changed to act according to button.lid_init_state=ignore. No one will be broken. And no future laptops will be broken. > ... Skip as it looks the result of debating here can also be concluded from the answers of the questions. > > > > After we root cause that's not a kernel problem, do we have mean in other community to handle > such > > > reports? > > > > For example, to collect hwdb entries. > > > > > > libinput, systemd are good candidate. > > > Libinput already has the bits in place, so I'd say we should probably > > > ask the users to report a bug on the wayland/libinput component of > > > bugs.freedesktop.org. But this will only work if the default initial lid > > > state is "method". > > > > Good to know. :) > > I've just changed the category of the forwarded report. > > Well, I am not sure your bug report should have been changed. The bug > report was regarding i915 being confident in the _LID acpi call, and > that is not something libinput can change. OK, I see. > ... Skip as it looks the result of debating here can also be concluded from the answers of the questions. Cheers, Lv
Hi Lv, On May 15 2017 or thereabouts, Zheng, Lv wrote: > Hi, Benjamin > > Let's stop endless discussing and focus on our needs. > > I just copied my questions here. > You can ask them directly. > For the below inlined replies, I'll stop replying if they are based on dependent on our basic agreements. > And I'll reply if something is really bad from my point of view. > > My point of view is: > There is a gap between (BIOS ensured/Windows expected) acpi control method lid device behavior and Linux user space expected acpi control method lid device behavior. > Button driver default behavior should be: button.lid_init_state=ignore > If user space programs have special needs, they can fix problems on their own, via the following mean: > echo -n "open" > /sys/modules/button/parameters/lid_init_state > echo -n "close" > /sys/modules/button/parameters/lid_init_state > Keeping open/close modes is because I don't think there is any bug in button driver. > So I need to prepare quirk modes from button driver's point of view and use them as a response to related bug reports reported in acpi community. > Your point of view is: > There is a gap between (BIOS ensured/Windows expected) acpi control method lid device behavior and Linux user space expected acpi control method lid device behavior. Yep. But our role, as kernel developers, is to not break user space even if they made wrong design choices. > Button driver default behavior should be (not 100% sure if this is your opinion): button.lid_init_state=method Yes, I'd like to revert to the old behavior (see below for a rationale). > If user space programs have special needs, they can fix them on their own, via the following mean: > libinput:name:*Lid Switch*:dmi:*svnMicrosoftCorporation:pnSurface3:* > LIBINPUT_ATTR_LID_SWITCH_RELIABILITY=write_open > From this point of view, we actually don't need open/close modes. We don't need close, but we need to keep open, for 2 reasons I'll detail below. > > It seems we just need to determine the following first: > 1. Who should be responsible for solving bugs triggered by the conflict between bios and linux user space expectations: > button driver? libinput? Some other user space programs? Users? Hopefully libinput or systemd (through a udev rule). If things gets worse a acpi/button quirk might be used, but in a second time. > 2. What should be the default button driver behavior? > button.lid_init_state=ignore? button.lid_init_state=method? button.lid_init_state=method: - this was the old default state, and switching to something else creates regression - the lid switch input node is an EV_SWITCH. And this has a state we need to synchronise with the hardware first. The default EV_SW state is closed, so if we do not sync it first, we might report wrong state to user space. > 3. If non button driver quirks are working, button driver quirk modes are useless. > The question is: Should button.lid_init_state=open/close be kept? We should keep button.lid_init_state=open: - it's kernel API now, so you can't remove it without breaking users configuration - it helps booting laptops that are not known to be working, the time the user installs the distribution and add the hwdb entry to fix it. we should not keep button.lid_init_state=close: - I don't see a good use case where it is needed, besides debugging drivers that should be debugged in an other way. > 4. From button driver's point of view, button.lid_init_state=ignore seems to be always correct, so we won't abandon it. We can keep it, and we should, it's kernel API > If we can use libinput to manage platform quirks, then button.lid_init_state=method also looks useless. No. 'method' is the only way to guarantee the exported input node is synced properly with the actual physical world. > The question is: Should button.lid_init_state=method be kept? Definitively. > > I should also let you know my preference: > 1. using button.lid_init_state=ignore and button.lid_init_state=method as default behavior is ok for me if answer to 1 is not button driver, otherwise using button.lid_init_state=method is not ok for me > 2. deletion of button.lid_init_state=open/close is ok for me if answer to 1 is not button driver, otherwise deletion of button.lid_init_state=open/close is not ok for me > 3. deletion of button.lid_init_state=method is always ok for me > > See the base line from my side is very clear: > If acpi community need to handle such bug reports, button.lid_init_state=method cannot be the default behavior. I already proposed to handle those reports. I don't see why you are concerned about those future reports. > We are just using a different default behavior than "method" to drive things to reach the final root caused solution. > > Could you let me know your preference so that we can figure out an agreement between us. > Though I don't know if end users will buy it (they may keep on filing regression reports in ACPI community). Switching back to button.lid_init_state=method will prevent regressions to be filled. If a user buys a new machine and the LID switch doesn't work as expected, it's not a regression, it's a bug. Regression is something we absolutely need to avoid, and switching to button.lid_init_state=open introduced far too many regressions on systems. > > > From: Benjamin Tissoires [mailto:benjamin.tissoires@redhat.com] > > Subject: Re: [PATCH 1/2] Revert "ACPI / button: Remove lid_init_state=method mode" > > > ... > Skip as it looks the result of debating here can also be concluded from the answers of the questions. > Let's just stop arguing same thing again and again using different contexts. > > > > > make reasonable assumptions based on the exact > > > > model capabilities (is the power button accessible with the LID closed), > > > > > > Sounds it's always accessible. > > > > hmm, I am not sure we have the same fingers.... For me on all the > > laptops I have seen, if the LID is actually (physically) closed, I can > > not press the button. It's a design choice to not have anything powering > > on the laptop when it's in your bag. > > Wow... > I can see that recent laptops are having their power buttons on the edge, not on the keypad. > For recent laptop/tablet 2-in-1 PCs or touch laptops which can have their keyboards folded and act notebooks, the power buttons are obviously not on the detachable keypads. Right. I forgot about this new class of tablets/convertible. I was probably too focused on the professional laptops. > Also for the laptops supporting external monitors, obviously allow users to push power buttons while the lid is closed. Well, on the Thinkpads (and Fujitsu from what I can tell), this situation can happen because there is a physical power button on the docking station itself. > > > ... > Skip as it looks the result of debating here can also be concluded from the answers of the questions. > > > > > > The issue we are fixing here is the fact that the switch state is wrong, > > > > which makes user space assumptions wrong too (and users angry). > > > > > > Considering no platform quirks. > > > > > > If ACPI button driver sends SW_LID, users are likely angry. > > > Unless the user space programs are changed to "ignore open event". > > > > > > If ACPI button driver doesn't send switch events, but key events. > > > The user space programs need to change to handle the new events. > > > > > > So finally whatever we do, user space need to change a bit related to ACPI control method lid device. > > > > Or we fix the switch to report accurate events/state. > > You do realise that all the energy you are spending, answering to me, > > talking to user space maintainers, users, all comes down to the fact > > that you refuse to have hardware quirks? > > > > If you don't want to write/maintain the code, fine. I can do it, but > > please stop trying to make everyone else change because you just don't > > want the burden of quirking a handful of laptops. > > But I have a very interesting question to ask: > Can button driver/libinput actually fix this without modifying BIOS: > > It seems we both rely on user's decision to use proper quirk modes for a specific usage model. > For example, for this case: > A BIOS never sends lid notification after boot/resume, and _LID return value is close after boot/resume. It used to work for years (_LID return value is correct on most laptops after boot/resume). > Let's see which quirk mode should the user choose before suspending: > 1. with no external monitor attached, user should: > write_open/button.lid_init_state=open > 2. with external monitor attached, if user wants the lid to be lit on > write_open/button.lid_init_state=open > 3. with external monitor attached, if user wants the lid not to be lit on > write_close/button.lid_init_state=close > See there is no single default value for all usage models. No. Users shouldn't interact with kernel parameters. The kernel parameter can help them to boot the system if the user space is not able to fix the situation. But in any case, end users should not have to deal with kernel parameters. > > Thus there is no possible __FIX__ for acpi button driver and libinput. I never talked about a fix. I know the situation is unsolvable, which is why I talked about quirks or workarounds. > While user space programs can just fix their usage models. You can't expect user space to change anything from the kernel point of view without a long enough warning. > > So finally we actually cannot fix anything by maintaining the quirk table, but just create a business of maintaining the quirk table. Yes, but I never asked you to maintain such table. I am volunteering to do that if needed. > > > ... > Skip as it looks the result of debating here can also be concluded from the answers of the questions. > > > > > > However, is that possible to not introduce platform quirks? > > > > > For example, for systemd: > > > > > If it detected ACPI lid device, automatically switch to an "IgnoreLidOpenEvent" mode (like > > nouveau > > > > drivers' ignorelid=Y option). > > > > > > > > Well, given that 99.9% of laptops have this ACPI lid button, you'll just > > > > remove the feature from all laptops. > > > > > > No, I only removed the wrong usage models related to the strict "open" event. > > > All laptops are still capable of sending correct "close" event. > > > > My bad, I read too fast and missed the "...Open..." part of > > "IgnoreLidOpenEvent". > > > > Though I am not sure IgnoreLidOpenEvent is accurate. > > "OpenEventNeverSent" seems to reflect more the reality. But again, > > that's not of our (kernel developers) business. > > IMO this is the only root cause fix. :) > It's the only way that the user can use without changing its quirk modes for different usage models. Yes and no. There is a design issue in logind, but this is based on the design choice made in acpi/button: we use an input switch. A switch has a state, and the kernel ought to report correct state. With this in mind, the design choice in logind is correct. But now we are seeing that some OEM are not providing everything we need, and we need to find the solution. > > > > So I already raised this to freedesktop: > > > https://bugs.freedesktop.org/show_bug.cgi?id=100923 > > > But couldn't see any one there responding. > > > > Well, Joachim answered, saying that there is likely a regression in > > acpi/button.c, not in i915. > > He surely can add more details as he is responsible of triage of the referenced bug reported on redhat. (Joachim doesn't work for Red Hat, from what I can tell). > But there is no agreement reached yet. > > > > > > Can it determine that by its own without listening to the lid key events? > > > > > > > > Basically no. This switch is there for a reason. However, I am convinced > > > > that a good heuristic is to say that if you are using the internal keyboard, > > > > touchscreen or touchpad, unless the user has very very thin fingers, the > > > > lid is open. > > > > > > I'm also convinced that the benefit of having a file in sysfs/procfs to allow user to know if the > > lid is open is marginal. > > > > I am convinced I don't get your point. We are obviously not talking > > about the same thing here. I was talking about the physical world, where > > the user interact with the laptop... > > See my previous reply against accessing power buttons when lid is closed. > > > ... > Skip as it looks the result of debating here can also be concluded from the answers of the questions. > > > > But it's just 1-2 months Bugzilla silence before seeing a different bug flood trend: > > > This time, they are related to the external monitor usage model: > > > https://bugzilla.kernel.org/show_bug.cgi?id=187271 > > > https://bugzilla.redhat.com/show_bug.cgi?id=1430259 > > > https://bugzilla.kernel.org/show_bug.cgi?id=195455 > > > Now that you can see why I didn't send a patch to change the default behavior to "open" at the time > > we were discussing. > > > That's also why I think we needn't revert back to "method" as the default behavior. > > > > Still, I strongly disagree. We do not fix 7 devices by breaking > > hundreds. That's just not how the kernel work. > > No. The word of "broken" is entirely wrong and emotional here. Yes. There was bug before, and now there are regressions. > If both user space and kernel space are changed to act according to button.lid_init_state=ignore. Again, you can't expect a good synchronisation between kernel space and user space without regressing some configurations (yes, it's a pain). > No one will be broken. 'ignore' is not the solution, see above for the rationale. > And no future laptops will be broken. no comments. > > > ... > Skip as it looks the result of debating here can also be concluded from the answers of the questions. > > > > > > After we root cause that's not a kernel problem, do we have mean in other community to handle > > such > > > > reports? > > > > > For example, to collect hwdb entries. > > > > > > > > libinput, systemd are good candidate. > > > > Libinput already has the bits in place, so I'd say we should probably > > > > ask the users to report a bug on the wayland/libinput component of > > > > bugs.freedesktop.org. But this will only work if the default initial lid > > > > state is "method". > > > > > > Good to know. :) > > > I've just changed the category of the forwarded report. > > > > Well, I am not sure your bug report should have been changed. The bug > > report was regarding i915 being confident in the _LID acpi call, and > > that is not something libinput can change. > > OK, I see. > > > ... > Skip as it looks the result of debating here can also be concluded from the answers of the questions. > Thanks for your answer, however, you did not answered my questions: Would you agree on: - acking both my patches so Rafael can take them ans stable can apply them ASAP - give me maintainership of drivers/acpi/button.c (ACK a following patch where I add me to MAINTAINERS) - redirect all kernel bugs to me related to LID switch (or having me the default assignee for a new acpi/button component) - allow me to add a list of quirk in acpi/button.c if I judge this as fit If you agree on that, that would be good for both our sake I think. Cheers, Benjamin > Cheers, > Lv -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi, Benjamin I reordered the discussion to collect topics and delete things to make discussion shorter. 1. root caused issue: > > It seems we just need to determine the following first: > > 1. Who should be responsible for solving bugs triggered by the conflict between bios and linux user > space expectations: > > button driver? libinput? Some other user space programs? Users? > Hopefully libinput or systemd (through a udev rule). If things gets > worse a acpi/button quirk might be used, but in a second time. I have concerns about what's in your mind. :) So let me high light a root caused issue: https://bugzilla.kernel.org/show_bug.cgi?id=106151 If we use any "open" modes, the suspend/resume loop can be fixed. Both "ignore/method" modes cannot fix the problem. In this bug, lid open event has a huge delay. But it can correctly arrive. However systemd will force 2nd suspend if it cannot see "open" event instantly after resume. So why don't systemd fix the issue (the enforcement) prior than letting us (input layer/button driver) to invent workarounds? IMO, this is a root caused problem and should be the first priority. Before seeing it is addressed in systemd, any changes made to libinput/button driver may not be proper. So the order of fixing all troubles are the followings in my mind: 1. system -> should eliminate the enforcement first 2. libinput -> may change lid event type (see my reply below for topic 2) 3. button driver > > > > > The issue we are fixing here is the fact that the switch state is wrong, > > > > > which makes user space assumptions wrong too (and users angry). > > > > Considering no platform quirks. > > > > If ACPI button driver sends SW_LID, users are likely angry. > > > > Unless the user space programs are changed to "ignore open event". > > > > If ACPI button driver doesn't send switch events, but key events. > > > > The user space programs need to change to handle the new events. > > > > So finally whatever we do, user space need to change a bit related to ACPI control method lid > device. > > > Or we fix the switch to report accurate events/state. > > > You do realise that all the energy you are spending, answering to me, > > > talking to user space maintainers, users, all comes down to the fact > > > that you refuse to have hardware quirks? Yes, as we have a root caused but unfixed issue in systemd first. It's pointless to introduce hardware quirks at this point. > > Thus there is no possible __FIX__ for acpi button driver and libinput. > I never talked about a fix. I know the situation is unsolvable, which is > why I talked about quirks or workarounds. > > While user space programs can just fix their usage models. > You can't expect user space to change anything from the kernel point of > view without a long enough warning. Why cannot we expect so? The above issue has already been root caused. > > > > > > However, is that possible to not introduce platform quirks? > > > > > > For example, for systemd: > > > > > > If it detected ACPI lid device, automatically switch to an "IgnoreLidOpenEvent" mode (like > > > nouveau > > > > > drivers' ignorelid=Y option). > > > > > Well, given that 99.9% of laptops have this ACPI lid button, you'll just > > > > > remove the feature from all laptops. > > > > No, I only removed the wrong usage models related to the strict "open" event. > > > > All laptops are still capable of sending correct "close" event. > > > My bad, I read too fast and missed the "...Open..." part of > > > "IgnoreLidOpenEvent". > > > Though I am not sure IgnoreLidOpenEvent is accurate. > > > "OpenEventNeverSent" seems to reflect more the reality. But again, > > > that's not of our (kernel developers) business. > > IMO this is the only root cause fix. :) > > It's the only way that the user can use without changing its quirk modes for different usage models. > Yes and no. There is a design issue in logind, but this is based on the > design choice made in acpi/button: we use an input switch. A switch has > a state, and the kernel ought to report correct state. With this in > mind, the design choice in logind is correct. But now we are > seeing that some OEM are not providing everything we need, and we need > to find the solution. We can stop arguing this. First of all, we just need to check if systemd can remove the enforcement mentioned above. And what will happen next. It's likely that there will be no user issues or button driver design issues then. Thus finally we may needn't introduce any "hardware quirks". 2. keep "method" mode: > > Button driver default behavior should be (not 100% sure if this is your opinion): > button.lid_init_state=method > > Yes, I'd like to revert to the old behavior (see below for a rationale). ... > > 2. What should be the default button driver behavior? > > button.lid_init_state=ignore? button.lid_init_state=method? > button.lid_init_state=method: > - this was the old default state, and switching to something else > creates regression > - the lid switch input node is an EV_SWITCH. And this has a state we > need to synchronise with the hardware first. The default EV_SW state > is closed, so if we do not sync it first, we might report wrong state > to user space. ... > > If we can use libinput to manage platform quirks, then button.lid_init_state=method also looks > useless. > No. 'method' is the only way to guarantee the exported input node is > synced properly with the actual physical world. > > The question is: Should button.lid_init_state=method be kept? > Definitively. Don't you think we actually have a new feature here? See, many years ago, all power buttons are on keyboard. Now all power buttons are on the edge, and many laptops have docking station. Don't you think we need to re-define the input event type of lid switch? Or we need to drive users of SW_LID/ACPI lid event to be not so strict to its switch event nature. So if SW_LID itself or its users are changed in this direction, will that really be useful to keep "method"? > Switching back to button.lid_init_state=method will prevent regressions > to be filled. If a user buys a new machine and the LID switch doesn't > work as expected, it's not a regression, it's a bug. Regression is > something we absolutely need to avoid, and switching to > button.lid_init_state=open introduced far too many regressions on > systems. IMO, whether "method" can be the default behavior depending on 2 factors: 1. We are discussing on freedesktop to see if things can be improved in graphics world; 2. topic 1 issue can be fixed. If graphics world is not changed and issue 1 is fixed, we can use "method" mode. If graphics world is changed and issue 1 is fixed, we can use "ignore" mode. If graphics world is changed and issue 1 is not fixed, we should stay using "open" mode. So it might be pointless to determine the default behavior such early. Given users can work things around using boot parameters. Aggressively making such a change now just make bug fixing work going back and forth. 3. keep "open/close" mode > We should keep button.lid_init_state=open: > - it's kernel API now, so you can't remove it without breaking users > configuration > - it helps booting laptops that are not known to be working, the time > the user installs the distribution and add the hwdb entry to fix it. > > If user space programs have special needs, they can fix them on their own, via the following mean: > > libinput:name:*Lid Switch*:dmi:*svnMicrosoftCorporation:pnSurface3:* > > LIBINPUT_ATTR_LID_SWITCH_RELIABILITY=write_open > > From this point of view, we actually don't need open/close modes. > We don't need close, but we need to keep open, for 2 reasons I'll detail > below. > > 3. If non button driver quirks are working, button driver quirk modes are useless. > > The question is: Should button.lid_init_state=open/close be kept? > we should not keep button.lid_init_state=close: > - I don't see a good use case where it is needed, besides debugging > drivers that should be debugged in an other way. If we presuppose that SW_LID cannot change, we surely need quirk modes for some new usage model issues. IMO, If quirks are done in libinput, then we needn't "open/close". If button driver need to provide quirks on its own, then we need both "open/close" to be able to deal with all cases. I already listed a use case that "open/method" cannot deal with: No lid notification and _LID return a cached "open" value after boot/resume. External monitor is connected and lid is closed. But we haven't reached an agreement yet. So please answer the following question. Could you tell me what the quirk mode (button driver mode or libinput node writes, whatever) should we use for the following cases: 1. No lid notification and _LID return value is a cached open after boot/resume 1.1. External monitor is connected, lid is closed 1.2. External monitor is connected, lid is opened 2. No lid notification and _LID return value is a cached close after boot/resume 2.1. External monitor is connected, lid is closed 2.2. External monitor is connected, lid is opened 4. quirk tables > > > If you don't want to write/maintain the code, fine. I can do it, but > > > please stop trying to make everyone else change because you just don't > > > want the burden of quirking a handful of laptops. > > See the base line from my side is very clear: > > If acpi community need to handle such bug reports, button.lid_init_state=method cannot be the > default behavior. > I already proposed to handle those reports. I don't see why you are > concerned about those future reports. > > So finally we actually cannot fix anything by maintaining the quirk table, but just create a > business of maintaining the quirk table. > Yes, but I never asked you to maintain such table. I am volunteering to > do that if needed. > Would you agree on: > - acking both my patches so Rafael can take them ans stable can apply > them ASAP > - give me maintainership of drivers/acpi/button.c (ACK a following patch > where I add me to MAINTAINERS) > - redirect all kernel bugs to me related to LID switch (or having me the > default assignee for a new acpi/button component) > - allow me to add a list of quirk in acpi/button.c if I judge this as > fit As reasons mentioned in topic 1 and 2. For now, I'd rather to see progresses made in systemd/libinput first. Actually there is no special maintainership related to the button driver. All drivers of ACPI spec defined devices (PNPxxx stuffs) are maintained as a whole by Rafael. And we are helping him by doing triage on kernel Bugzilla. There isn't a category on kernel Bugzilla related to lid issues. Probably you can help to create one under ACPI product. If this can be achieved, you can be the default assignee for such issues. If this cannot be achieved, we'll just forward some lid reports to you first. Please tell me which product/component you'd prefer for us to forward. For modifying acpi/button.c, it's open source, you can send patches and ACPI community can help to review. Thanks and best regards Lv
diff --git a/Documentation/acpi/acpi-lid.txt b/Documentation/acpi/acpi-lid.txt index 22cb309..effe7af 100644 --- a/Documentation/acpi/acpi-lid.txt +++ b/Documentation/acpi/acpi-lid.txt @@ -59,20 +59,28 @@ button driver uses the following 3 modes in order not to trigger issues. If the userspace hasn't been prepared to ignore the unreliable "opened" events and the unreliable initial state notification, Linux users can use the following kernel parameters to handle the possible issues: -A. button.lid_init_state=open: +A. button.lid_init_state=method: + When this option is specified, the ACPI button driver reports the + initial lid state using the returning value of the _LID control method + and whether the "opened"/"closed" events are paired fully relies on the + firmware implementation. + This option can be used to fix some platforms where the returning value + of the _LID control method is reliable but the initial lid state + notification is missing. + This option is the default behavior during the period the userspace + isn't ready to handle the buggy AML tables. +B. button.lid_init_state=open: When this option is specified, the ACPI button driver always reports the initial lid state as "opened" and whether the "opened"/"closed" events are paired fully relies on the firmware implementation. This may fix some platforms where the returning value of the _LID control method is not reliable and the initial lid state notification is missing. - This option is the default behavior during the period the userspace - isn't ready to handle the buggy AML tables. If the userspace has been prepared to ignore the unreliable "opened" events and the unreliable initial state notification, Linux users should always use the following kernel parameter: -B. button.lid_init_state=ignore: +C. button.lid_init_state=ignore: When this option is specified, the ACPI button driver never reports the initial lid state and there is a compensation mechanism implemented to ensure that the reliable "closed" notifications can always be delievered diff --git a/drivers/acpi/button.c b/drivers/acpi/button.c index 668137e..6d5a8c1 100644 --- a/drivers/acpi/button.c +++ b/drivers/acpi/button.c @@ -57,6 +57,7 @@ #define ACPI_BUTTON_LID_INIT_IGNORE 0x00 #define ACPI_BUTTON_LID_INIT_OPEN 0x01 +#define ACPI_BUTTON_LID_INIT_METHOD 0x02 #define _COMPONENT ACPI_BUTTON_COMPONENT ACPI_MODULE_NAME("button"); @@ -376,6 +377,9 @@ static void acpi_lid_initialize_state(struct acpi_device *device) case ACPI_BUTTON_LID_INIT_OPEN: (void)acpi_lid_notify_state(device, 1); break; + case ACPI_BUTTON_LID_INIT_METHOD: + (void)acpi_lid_update_state(device); + break; case ACPI_BUTTON_LID_INIT_IGNORE: default: break; @@ -559,6 +563,9 @@ static int param_set_lid_init_state(const char *val, struct kernel_param *kp) if (!strncmp(val, "open", sizeof("open") - 1)) { lid_init_state = ACPI_BUTTON_LID_INIT_OPEN; pr_info("Notify initial lid state as open\n"); + } else if (!strncmp(val, "method", sizeof("method") - 1)) { + lid_init_state = ACPI_BUTTON_LID_INIT_METHOD; + pr_info("Notify initial lid state with _LID return value\n"); } else if (!strncmp(val, "ignore", sizeof("ignore") - 1)) { lid_init_state = ACPI_BUTTON_LID_INIT_IGNORE; pr_info("Do not notify initial lid state\n"); @@ -572,6 +579,8 @@ static int param_get_lid_init_state(char *buffer, struct kernel_param *kp) switch (lid_init_state) { case ACPI_BUTTON_LID_INIT_OPEN: return sprintf(buffer, "open"); + case ACPI_BUTTON_LID_INIT_METHOD: + return sprintf(buffer, "method"); case ACPI_BUTTON_LID_INIT_IGNORE: return sprintf(buffer, "ignore"); default:
This reverts commit ecb10b694b72ca5ea51b3c90a71ff2a11963425a. Even if the method can be buggy some times, it's still a need when you boot a laptop with a lid closed and an external monitor connected (typical situation when using a docking station) Note: this option has been removed without being deprecated, which is terrible in term of distribution handling. The new default "open" is plain wrong and we don't even have the chance to keep the current default option because it's not there anymore. Cc: stable@vger.kernel.org Signed-off-by: Benjamin Tissoires <benjamin.tissoires@redhat.com> --- Documentation/acpi/acpi-lid.txt | 16 ++++++++++++---- drivers/acpi/button.c | 9 +++++++++ 2 files changed, 21 insertions(+), 4 deletions(-)