| Message ID | 72f375b46a34702fc9f8be419fa83c219fbc28c5.1413481880.git.luto@amacapital.net (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Takashi Iwai |
| Headers | show |
On Thu, 16 Oct 2014, Andy Lutomirski wrote: > ThinkPads have hardware volume controls and three buttons to control > them. (These are separate from the standard mixer.) By default, > the buttons are: > > - Mute: Mutes the hardware volume control and generates KEY_MUTE. > - Up: Unmutes, generates KEY_VOLUMEUP, and increases volume if > applicable. (Newer thinkpads only have hardware mute/unmute.) > - Down: Unmutes, generates KEY_VOLUMEDOWN, and decreases volume > if applicable. On IBM ThinkPads, the "generates KEY_foo" thing is controlable by the driver. In fact, thinkpad_acpi defaults to NOT send these, all breakage is entirely the fault of userspace that enables it instead of doing OSD out of the extra mixer, and then fails to deal with the side-effects. On Lenovo ThinkPads, it depends on BIOS/EC version for the X60/T60. Later versions will behave like the X61/T61. Since family 200 you can tell the EC to switch modes (latch, toggle, none), but Lenovo didn't test it well enough. > This behavior is unfortunate, since modern userspace will also > handle the hotkeys and change the other mixer. If the software Like I said, we disable that by default, and userspace enables them back because they want to use it as a keyboard. We could override it entirely on the IBM thinkpads, and refuse to issue any key events no matter what, I suppose. > This should also allow us to remove _OSI(Linux) for all ThinkPads. No, it doesn't. Only after an exaustive search for what _OSI(Linux) changes across a brickload of thinkpad ACPI dumps could we say that for sure. > For future reference: It turns out that we can ask ACPI for one of > three behaviors directly on very new models. They are "latch" (the > default), "none" (no automatic control), and "toggle" (mute unmutes > when muted). All of the modes besides "none" seem to be a bit buggy, > though, and there is no known way to get any notification that the > hardware state has changed other than listening for a mute button > press on the i8042 port. Yeah. > + * The EC or perhaps SMM firmware can optionally automatically change > + * the volume in response to user input. Unfortunately, this rarely It is the EC, which also emulates the i8042 KBC. I don't see any problems with the code, will have to test it in a IBM thinkpad, though (I have one).
On Thu, Oct 16, 2014 at 11:31 AM, Henrique de Moraes Holschuh <hmh@hmh.eng.br> wrote: > On Thu, 16 Oct 2014, Andy Lutomirski wrote: >> ThinkPads have hardware volume controls and three buttons to control >> them. (These are separate from the standard mixer.) By default, >> the buttons are: >> >> - Mute: Mutes the hardware volume control and generates KEY_MUTE. >> - Up: Unmutes, generates KEY_VOLUMEUP, and increases volume if >> applicable. (Newer thinkpads only have hardware mute/unmute.) >> - Down: Unmutes, generates KEY_VOLUMEDOWN, and decreases volume >> if applicable. > > On IBM ThinkPads, the "generates KEY_foo" thing is controlable by the > driver. In fact, thinkpad_acpi defaults to NOT send these, all breakage > is entirely the fault of userspace that enables it instead of doing OSD > out of the extra mixer, and then fails to deal with the side-effects. Are you sure? I'm reasonably confident that the KEY_MUTE comes from i8042, not from an ACPI GPE. Even without this patch, pressing the mute button sends KEY_MUTE. Maybe there's a control I missed, but if we turn off KEY_MUTE, then I don't know how we'd be notified at all that the HW mixer changed state. (In any case, with this patch, everything seems to work.) > > On Lenovo ThinkPads, it depends on BIOS/EC version for the X60/T60. Later > versions will behave like the X61/T61. Since family 200 you can tell the > EC to switch modes (latch, toggle, none), but Lenovo didn't test it well > enough. "none" seems to work well on all ThinkPads that I've tried. The other modes seem to have spotty support. > >> This behavior is unfortunate, since modern userspace will also >> handle the hotkeys and change the other mixer. If the software > > Like I said, we disable that by default, and userspace enables them back > because they want to use it as a keyboard. We could override it entirely > on the IBM thinkpads, and refuse to issue any key events no matter what, I > suppose. How? The only way I ever found to do that was using the i8042_filter in the older incarnations of this driver, and that was incredibly complex, added little obvious value, and *still* didn't give a fully consistent UI, since userspace doesn't understand the extra thinkpad mixer. > >> This should also allow us to remove _OSI(Linux) for all ThinkPads. > > No, it doesn't. Only after an exaustive search for what _OSI(Linux) > changes across a brickload of thinkpad ACPI dumps could we say that for > sure. I think we only need that for the four models in question. Also, all the commit messages for those quirk additions seem to suggest that they were done to improve volume controls. And I'm pretty sure I already checked at least one of those models (that's how I discovered HAUM/SAUM in the first place). Anyway, patch 2 is optional. > >> For future reference: It turns out that we can ask ACPI for one of >> three behaviors directly on very new models. They are "latch" (the >> default), "none" (no automatic control), and "toggle" (mute unmutes >> when muted). All of the modes besides "none" seem to be a bit buggy, >> though, and there is no known way to get any notification that the >> hardware state has changed other than listening for a mute button >> press on the i8042 port. > > Yeah. > >> + * The EC or perhaps SMM firmware can optionally automatically change >> + * the volume in response to user input. Unfortunately, this rarely > > It is the EC, which also emulates the i8042 KBC. > > I don't see any problems with the code, will have to test it in a IBM > thinkpad, though (I have one). Please test v2 instead. :) --Andy > > -- > "One disk to rule them all, One disk to find them. One disk to bring > them all and in the darkness grind them. In the Land of Redmond > where the shadows lie." -- The Silicon Valley Tarot > Henrique Holschuh
On Thu, 16 Oct 2014, Andy Lutomirski wrote: > > On IBM ThinkPads, the "generates KEY_foo" thing is controlable by the > > driver. In fact, thinkpad_acpi defaults to NOT send these, all breakage > > is entirely the fault of userspace that enables it instead of doing OSD > > out of the extra mixer, and then fails to deal with the side-effects. > > Are you sure? I'm reasonably confident that the KEY_MUTE comes from Yes. > i8042, not from an ACPI GPE. Even without this patch, pressing the The EC *is* the i8042 in a ThinkPad. Really. You talk to the same chip and to the same firmware over the ACPI EC ports, the i8042 ports, and the Renesas extra LPC3 channel (used by HDAPS and SBS). That said, on a **IBM** ThinkPad, _nothing_ related to the digital volume control comes from the i8042 port. In fact, with older BIOSes, you couldn't even get it from ACPI unless you hooked the GPE directly because the required ACPI L/E methods were missing and were added by later versions of the T4x/R4x/X3x BIOS. The EC firmware in a thinkpad is often more capable than what the BIOS exposes. KEY_MUTE comes from the i8042 in Lenovo Thinkpads, only. And they started doing that when they removed the hardware volume control chip. It was retrofitted in later versions of the T60 BIOS (v2.x), I don't know how well though. > > On Lenovo ThinkPads, it depends on BIOS/EC version for the X60/T60. Later > > versions will behave like the X61/T61. Since family 200 you can tell the > > EC to switch modes (latch, toggle, none), but Lenovo didn't test it well > > enough. > > "none" seems to work well on all ThinkPads that I've tried. The other > modes seem to have spotty support. I am fine that we go with "none" and emulate everything in the driver where supported, yes. > >> This behavior is unfortunate, since modern userspace will also > >> handle the hotkeys and change the other mixer. If the software > > > > Like I said, we disable that by default, and userspace enables them back > > because they want to use it as a keyboard. We could override it entirely > > on the IBM thinkpads, and refuse to issue any key events no matter what, I > > suppose. > > How? The only way I ever found to do that was using the i8042_filter > in the older incarnations of this driver, and that was incredibly You didn't try it on a IBM thinkpad, just on Lenovo thinkpads, then... > >> This should also allow us to remove _OSI(Linux) for all ThinkPads. > > > > No, it doesn't. Only after an exaustive search for what _OSI(Linux) > > changes across a brickload of thinkpad ACPI dumps could we say that for > > sure. > > I think we only need that for the four models in question. Also, all > the commit messages for those quirk additions seem to suggest that > they were done to improve volume controls. And I'm pretty sure I > already checked at least one of those models (that's how I discovered > HAUM/SAUM in the first place). > > Anyway, patch 2 is optional. Since you did due diligency, I'm fine with that. It is already in a separate patch, so it is easy to bissect and revert anyway. > >> + * The EC or perhaps SMM firmware can optionally automatically change > >> + * the volume in response to user input. Unfortunately, this rarely > > > > It is the EC, which also emulates the i8042 KBC. > > > > I don't see any problems with the code, will have to test it in a IBM > > thinkpad, though (I have one). > > Please test v2 instead. :) I will only be able to do that during the weekend anyway, so feel free to send a v3 if you need :-)
On Thu, Oct 16, 2014 at 11:54 AM, Henrique de Moraes Holschuh <hmh@hmh.eng.br> wrote: > On Thu, 16 Oct 2014, Andy Lutomirski wrote: >> > On IBM ThinkPads, the "generates KEY_foo" thing is controlable by the >> > driver. In fact, thinkpad_acpi defaults to NOT send these, all breakage >> > is entirely the fault of userspace that enables it instead of doing OSD >> > out of the extra mixer, and then fails to deal with the side-effects. >> >> Are you sure? I'm reasonably confident that the KEY_MUTE comes from > > Yes. > >> i8042, not from an ACPI GPE. Even without this patch, pressing the > > The EC *is* the i8042 in a ThinkPad. Really. You talk to the same chip and > to the same firmware over the ACPI EC ports, the i8042 ports, and the > Renesas extra LPC3 channel (used by HDAPS and SBS). Huh, interesting. That makes sense. > > That said, on a **IBM** ThinkPad, _nothing_ related to the digital volume > control comes from the i8042 port. In fact, with older BIOSes, you couldn't > even get it from ACPI unless you hooked the GPE directly because the > required ACPI L/E methods were missing and were added by later versions of > the T4x/R4x/X3x BIOS. > > The EC firmware in a thinkpad is often more capable than what the BIOS > exposes. > > KEY_MUTE comes from the i8042 in Lenovo Thinkpads, only. And they started > doing that when they removed the hardware volume control chip. It was > retrofitted in later versions of the T60 BIOS (v2.x), I don't know how well > though. So the question is: are there IBM models (or models that don't send KEY_MUTE, anyway) that nonetheless expose HAUM and SAUM? If not, then my patch should be okay. If so, we'll need further filtering. > >> > On Lenovo ThinkPads, it depends on BIOS/EC version for the X60/T60. Later >> > versions will behave like the X61/T61. Since family 200 you can tell the >> > EC to switch modes (latch, toggle, none), but Lenovo didn't test it well >> > enough. >> >> "none" seems to work well on all ThinkPads that I've tried. The other >> modes seem to have spotty support. > > I am fine that we go with "none" and emulate everything in the driver where > supported, yes. > >> >> This behavior is unfortunate, since modern userspace will also >> >> handle the hotkeys and change the other mixer. If the software >> > >> > Like I said, we disable that by default, and userspace enables them back >> > because they want to use it as a keyboard. We could override it entirely >> > on the IBM thinkpads, and refuse to issue any key events no matter what, I >> > suppose. >> >> How? The only way I ever found to do that was using the i8042_filter >> in the older incarnations of this driver, and that was incredibly > > You didn't try it on a IBM thinkpad, just on Lenovo thinkpads, then... I'm not sure I ever owned an IBM thinkpad. I used Acer back then, and my next laptop is quite likely to be > >> >> This should also allow us to remove _OSI(Linux) for all ThinkPads. >> > >> > No, it doesn't. Only after an exaustive search for what _OSI(Linux) >> > changes across a brickload of thinkpad ACPI dumps could we say that for >> > sure. >> >> I think we only need that for the four models in question. Also, all >> the commit messages for those quirk additions seem to suggest that >> they were done to improve volume controls. And I'm pretty sure I >> already checked at least one of those models (that's how I discovered >> HAUM/SAUM in the first place). >> >> Anyway, patch 2 is optional. > > Since you did due diligency, I'm fine with that. It is already in a > separate patch, so it is easy to bissect and revert anyway. > >> >> + * The EC or perhaps SMM firmware can optionally automatically change >> >> + * the volume in response to user input. Unfortunately, this rarely >> > >> > It is the EC, which also emulates the i8042 KBC. >> > >> > I don't see any problems with the code, will have to test it in a IBM >> > thinkpad, though (I have one). >> >> Please test v2 instead. :) > > I will only be able to do that during the weekend anyway, so feel free to > send a v3 if you need :-) > > -- > "One disk to rule them all, One disk to find them. One disk to bring > them all and in the darkness grind them. In the Land of Redmond > where the shadows lie." -- The Silicon Valley Tarot > Henrique Holschuh
On Thu, 16 Oct 2014, Andy Lutomirski wrote: > On Thu, Oct 16, 2014 at 11:54 AM, Henrique de Moraes Holschuh > <hmh@hmh.eng.br> wrote: > So the question is: are there IBM models (or models that don't send > KEY_MUTE, anyway) that nonetheless expose HAUM and SAUM? If not, then > my patch should be okay. If so, we'll need further filtering. No, these were introduced in the X61/T61. Still, the standard paranoia applies, and you should check for HAUM/SAUM only on VENDOR_LENOVO, and assume they don't exist (because if they do, they are something else) for VENDOR_IBM.
On Thu, Oct 16, 2014 at 12:58 PM, Henrique de Moraes Holschuh <hmh@hmh.eng.br> wrote: > On Thu, 16 Oct 2014, Andy Lutomirski wrote: >> On Thu, Oct 16, 2014 at 11:54 AM, Henrique de Moraes Holschuh >> <hmh@hmh.eng.br> wrote: >> So the question is: are there IBM models (or models that don't send >> KEY_MUTE, anyway) that nonetheless expose HAUM and SAUM? If not, then >> my patch should be okay. If so, we'll need further filtering. > > No, these were introduced in the X61/T61. Still, the standard paranoia > applies, and you should check for HAUM/SAUM only on VENDOR_LENOVO, and > assume they don't exist (because if they do, they are something else) for > VENDOR_IBM. Done in v3. My X220 seems to be alive again for the time being -- I suspect that there's a marginal connection somewhere. It seems to be fully functional with this match, and the mute light works as expected (thanks, Takashi et al!). --Andy
On Thu, Oct 16, 2014 at 11:54 AM, Henrique de Moraes Holschuh <hmh@hmh.eng.br> wrote: > On Thu, 16 Oct 2014, Andy Lutomirski wrote: >> > On IBM ThinkPads, the "generates KEY_foo" thing is controlable by the >> > driver. In fact, thinkpad_acpi defaults to NOT send these, all breakage >> > is entirely the fault of userspace that enables it instead of doing OSD >> > out of the extra mixer, and then fails to deal with the side-effects. >> >> Are you sure? I'm reasonably confident that the KEY_MUTE comes from > > Yes. > >> i8042, not from an ACPI GPE. Even without this patch, pressing the > > The EC *is* the i8042 in a ThinkPad. Really. You talk to the same chip and > to the same firmware over the ACPI EC ports, the i8042 ports, and the > Renesas extra LPC3 channel (used by HDAPS and SBS). > > That said, on a **IBM** ThinkPad, _nothing_ related to the digital volume > control comes from the i8042 port. In fact, with older BIOSes, you couldn't > even get it from ACPI unless you hooked the GPE directly because the > required ACPI L/E methods were missing and were added by later versions of > the T4x/R4x/X3x BIOS. > > The EC firmware in a thinkpad is often more capable than what the BIOS > exposes. > > KEY_MUTE comes from the i8042 in Lenovo Thinkpads, only. And they started > doing that when they removed the hardware volume control chip. It was > retrofitted in later versions of the T60 BIOS (v2.x), I don't know how well > though. > >> > On Lenovo ThinkPads, it depends on BIOS/EC version for the X60/T60. Later >> > versions will behave like the X61/T61. Since family 200 you can tell the >> > EC to switch modes (latch, toggle, none), but Lenovo didn't test it well >> > enough. >> >> "none" seems to work well on all ThinkPads that I've tried. The other >> modes seem to have spotty support. > > I am fine that we go with "none" and emulate everything in the driver where > supported, yes. > >> >> This behavior is unfortunate, since modern userspace will also >> >> handle the hotkeys and change the other mixer. If the software >> > >> > Like I said, we disable that by default, and userspace enables them back >> > because they want to use it as a keyboard. We could override it entirely >> > on the IBM thinkpads, and refuse to issue any key events no matter what, I >> > suppose. >> >> How? The only way I ever found to do that was using the i8042_filter >> in the older incarnations of this driver, and that was incredibly > > You didn't try it on a IBM thinkpad, just on Lenovo thinkpads, then... > >> >> This should also allow us to remove _OSI(Linux) for all ThinkPads. >> > >> > No, it doesn't. Only after an exaustive search for what _OSI(Linux) >> > changes across a brickload of thinkpad ACPI dumps could we say that for >> > sure. >> >> I think we only need that for the four models in question. Also, all >> the commit messages for those quirk additions seem to suggest that >> they were done to improve volume controls. And I'm pretty sure I >> already checked at least one of those models (that's how I discovered >> HAUM/SAUM in the first place). >> >> Anyway, patch 2 is optional. > > Since you did due diligency, I'm fine with that. It is already in a > separate patch, so it is easy to bissect and revert anyway. > >> >> + * The EC or perhaps SMM firmware can optionally automatically change >> >> + * the volume in response to user input. Unfortunately, this rarely >> > >> > It is the EC, which also emulates the i8042 KBC. >> > >> > I don't see any problems with the code, will have to test it in a IBM >> > thinkpad, though (I have one). >> >> Please test v2 instead. :) > > I will only be able to do that during the weekend anyway, so feel free to > send a v3 if you need :-) > Any update? --Andy > -- > "One disk to rule them all, One disk to find them. One disk to bring > them all and in the darkness grind them. In the Land of Redmond > where the shadows lie." -- The Silicon Valley Tarot > Henrique Holschuh
On Sat, 25 Oct 2014, Andy Lutomirski wrote: > > I will only be able to do that during the weekend anyway, so feel free to > > send a v3 if you need :-) > > > > Any update? Will test tomorrow, or with some luck, tonight... Had some nasty issues to deal with.
diff --git a/drivers/platform/x86/thinkpad_acpi.c b/drivers/platform/x86/thinkpad_acpi.c index 3bbc6eb60de5..9c8e17ec4b14 100644 --- a/drivers/platform/x86/thinkpad_acpi.c +++ b/drivers/platform/x86/thinkpad_acpi.c @@ -6559,6 +6559,16 @@ static struct ibm_struct brightness_driver_data = { * bits 3-0 (volume). Other bits in NVRAM may have other functions, * such as bit 7 which is used to detect repeated presses of MUTE, * and we leave them unchanged. + * + * The EC or perhaps SMM firmware can optionally automatically change + * the volume in response to user input. Unfortunately, this rarely + * works well. The laptop changes the state of its internal MUTE gate + * *and* sends KEY_MUTE, causing any user code that responds to the mute + * button to get confused. The hardware MUTE gate is also unnecessary, + * since user code can handle the mute button without kernel or EC help. + * + * To avoid confusing userspace, we simply disable all EC-based mute + * and volume controls when possible. */ #ifdef CONFIG_THINKPAD_ACPI_ALSA_SUPPORT @@ -6613,11 +6623,20 @@ enum tpacpi_volume_capabilities { TPACPI_VOL_CAP_MAX }; +enum tpacpi_mute_btn_mode { + TP_EC_MUTE_BTN_LATCH = 0, /* Mute mutes; up/down unmutes */ + /* We don't know what mode 1 is. */ + TP_EC_MUTE_BTN_NONE = 2, /* Mute and up/down are just keys */ + TP_EC_MUTE_BTN_TOGGLE = 3, /* Mute toggles; up/down unmutes */ +}; + static enum tpacpi_volume_access_mode volume_mode = TPACPI_VOL_MODE_MAX; static enum tpacpi_volume_capabilities volume_capabilities; static bool volume_control_allowed; +static bool software_mute_requested = true; +static bool software_mute_active; /* * Used to syncronize writers to TP_EC_AUDIO and @@ -6635,6 +6654,8 @@ static void tpacpi_volume_checkpoint_nvram(void) return; if (!volume_control_allowed) return; + if (software_mute_active) + return; vdbg_printk(TPACPI_DBG_MIXER, "trying to checkpoint mixer state to NVRAM...\n"); @@ -6696,6 +6717,12 @@ static int volume_set_status_ec(const u8 status) dbg_printk(TPACPI_DBG_MIXER, "set EC mixer to 0x%02x\n", status); + /* + * On X200s, and possibly on others, it can take a while for + * reads to become correct. + */ + msleep(1); + return 0; } @@ -6778,6 +6805,38 @@ unlock: return rc; } +static int volume_set_software_mute(bool verbose) +{ + int rc = 0; + int result; + + if (mutex_lock_killable(&volume_mutex) < 0) + return -EINTR; + + if (verbose) { + if (!acpi_evalf(ec_handle, &result, "HAUM", "qd")) + return -EIO; + + dbg_printk(TPACPI_DBG_INIT | TPACPI_DBG_MIXER, + "Initial HAUM setting was %d\n", + result); + } + + if (!acpi_evalf(ec_handle, &result, "SAUM", "qdd", + (int)TP_EC_MUTE_BTN_NONE)) { + rc = -EIO; + goto out; + } + + if (result != TP_EC_MUTE_BTN_NONE) + pr_warn("thinkpad_acpi: Unexpected SAUM result %d\n", + result); + +out: + mutex_unlock(&volume_mutex); + return rc; +} + static int volume_alsa_set_volume(const u8 vol) { dbg_printk(TPACPI_DBG_MIXER, @@ -6885,7 +6944,12 @@ static void volume_suspend(void) static void volume_resume(void) { - volume_alsa_notify_change(); + if (software_mute_active) { + if (volume_set_software_mute(false) < 0) + pr_warn("thinkpad_acpi: failed to restore software mute\n"); + } else { + volume_alsa_notify_change(); + } } static void volume_shutdown(void) @@ -7004,6 +7068,8 @@ static const struct tpacpi_quirk volume_quirk_table[] __initconst = { .quirks = TPACPI_VOL_Q_MUTEONLY } }; +static int volume_write(char *buf); + static int __init volume_init(struct ibm_init_struct *iibm) { unsigned long quirks; @@ -7085,16 +7151,33 @@ static int __init volume_init(struct ibm_init_struct *iibm) "mute is supported, volume control is %s\n", str_supported(!tp_features.mixer_no_level_control)); - rc = volume_create_alsa_mixer(); - if (rc) { - pr_err("Could not create the ALSA mixer interface\n"); - return rc; - } + if (software_mute_requested && volume_set_software_mute(true) == 0) { + char unmute_cmd[] = "unmute,level 14"; /* must be non-const */ + BUILD_BUG_ON(TP_EC_VOLUME_MAX != 14); + + software_mute_active = true; + + /* + * In software mute mode, the standard codec controls + * take precendence, so we unmute the ThinkPad HW switch + * at startup. Just on case there are SAUM-capable + * ThinkPads with level controls, set max HW volume as + * well. + */ + if (volume_write(unmute_cmd) != 0) + pr_warn("Failed to unmute ThinkPad HW\n"); + } else { + rc = volume_create_alsa_mixer(); + if (rc) { + pr_err("Could not create the ALSA mixer interface\n"); + return rc; + } - pr_info("Console audio control enabled, mode: %s\n", - (volume_control_allowed) ? - "override (read/write)" : - "monitor (read only)"); + pr_info("Console audio control enabled, mode: %s\n", + (volume_control_allowed) ? + "override (read/write)" : + "monitor (read only)"); + } vdbg_printk(TPACPI_DBG_INIT | TPACPI_DBG_MIXER, "registering volume hotkeys as change notification\n"); @@ -9091,6 +9174,10 @@ MODULE_PARM_DESC(volume_control, "Enables software override for the console audio " "control when true"); +module_param_named(software_mute, software_mute_requested, bool, 0444); +MODULE_PARM_DESC(software_mute, + "Request full software mute control"); + /* ALSA module API parameters */ module_param_named(index, alsa_index, int, 0444); MODULE_PARM_DESC(index, "ALSA index for the ACPI EC Mixer");