suspend-to-mem broken by eed4d47efe95 (ACPI / sleep: Ignore spurious SCI wakeups from suspend-to-idle)

suspend-to-mem broken by eed4d47efe95 (ACPI / sleep: Ignore spurious SCI wakeups from suspend-to-idle)

[Dominik Brodowski]

[-- Attachment #1: Type: text/plain, Size: 849 bytes --]

Rafael,

unfortunately, commit eed4d47efe95 (ACPI / sleep: Ignore spurious SCI
wakeups from suspend-to-idle) breaks suspend-to-mem on my Dell XPS 13
(9343) up to v4.12-rc4: Issuing

	$ echo "mem" > /sys/power/state

in the initramfs returns, after a while, with "write error: Resource
busy", and the system *not* having entered the sleep state in between.

In contrast thereto, 8a537ece3d94 (PM / wakeup: Integrate mechanism to abort
transitions in progress) still works fine, and allows to enter
suspend-to-mem. No real difference is to be seen in dmesg, with the notable
exception of

	ACPI: Low-level resume complete
	ACPI : EC: EC started
	PM: Restoring platform NVS memory
	Suspended for N.NNNN seconds

only showing up on working kernels. Reverting eed4d47efe95 on top of
v4.12-rc4 restores suspend-to-mem to work as expected.

Best,
	Dominik

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: suspend-to-mem broken by eed4d47efe95 (ACPI / sleep: Ignore spurious SCI wakeups from suspend-to-idle)

[Rafael J. Wysocki]

On Monday, June 05, 2017 11:05:21 AM Dominik Brodowski wrote:
> Rafael,

Thanks for the report!

> unfortunately, commit eed4d47efe95 (ACPI / sleep: Ignore spurious SCI
> wakeups from suspend-to-idle) breaks suspend-to-mem on my Dell XPS 13
> (9343) up to v4.12-rc4: Issuing
> 
> 	$ echo "mem" > /sys/power/state
> 
> in the initramfs returns, after a while, with "write error: Resource
> busy", and the system *not* having entered the sleep state in between.

Why initramfs?

> In contrast thereto, 8a537ece3d94 (PM / wakeup: Integrate mechanism to abort
> transitions in progress) still works fine, and allows to enter
> suspend-to-mem. No real difference is to be seen in dmesg, with the notable
> exception of
> 
> 	ACPI: Low-level resume complete
> 	ACPI : EC: EC started
> 	PM: Restoring platform NVS memory
> 	Suspended for N.NNNN seconds
> 
> only showing up on working kernels. Reverting eed4d47efe95 on top of
> v4.12-rc4 restores suspend-to-mem to work as expected.

I'm sure it is not necessary to revert all of it.

I guess what happens is that you get a wakeup event during suspend which is
aborted as a result.

Please apply the partial revert below and see if it makes the issue go away.

---
 drivers/acpi/battery.c   |    2 +-
 drivers/acpi/button.c    |    4 ++--
 drivers/acpi/device_pm.c |    2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)

Index: linux-pm/drivers/acpi/battery.c
===================================================================
--- linux-pm.orig/drivers/acpi/battery.c
+++ linux-pm/drivers/acpi/battery.c
@@ -782,7 +782,7 @@ static int acpi_battery_update(struct ac
 	if ((battery->state & ACPI_BATTERY_STATE_CRITICAL) ||
 	    (test_bit(ACPI_BATTERY_ALARM_PRESENT, &battery->flags) &&
             (battery->capacity_now <= battery->alarm)))
-		pm_wakeup_hard_event(&battery->device->dev);
+		pm_wakeup_event(&battery->device->dev, 0);
 
 	return result;
 }
Index: linux-pm/drivers/acpi/device_pm.c
===================================================================
--- linux-pm.orig/drivers/acpi/device_pm.c
+++ linux-pm/drivers/acpi/device_pm.c
@@ -400,7 +400,7 @@ static void acpi_pm_notify_handler(acpi_
 	mutex_lock(&acpi_pm_notifier_lock);
 
 	if (adev->wakeup.flags.notifier_present) {
-		pm_wakeup_ws_event(adev->wakeup.ws, 0, true);
+		__pm_wakeup_event(adev->wakeup.ws, 0);
 		if (adev->wakeup.context.work.func)
 			queue_pm_work(&adev->wakeup.context.work);
 	}
Index: linux-pm/drivers/acpi/button.c
===================================================================
--- linux-pm.orig/drivers/acpi/button.c
+++ linux-pm/drivers/acpi/button.c
@@ -217,7 +217,7 @@ static int acpi_lid_notify_state(struct
 	}
 
 	if (state)
-		pm_wakeup_hard_event(&device->dev);
+		pm_wakeup_event(&device->dev, 0);
 
 	ret = blocking_notifier_call_chain(&acpi_lid_notifier, state, device);
 	if (ret == NOTIFY_DONE)
@@ -402,7 +402,7 @@ static void acpi_button_notify(struct ac
 		} else {
 			int keycode;
 
-			pm_wakeup_hard_event(&device->dev);
+			pm_wakeup_event(&device->dev, 0);
 			if (button->suspended)
 				break;
 

Re: suspend-to-mem broken by eed4d47efe95 (ACPI / sleep: Ignore spurious SCI wakeups from suspend-to-idle)

[Dominik Brodowski]

[-- Attachment #1: Type: text/plain, Size: 4671 bytes --]

On Mon, Jun 05, 2017 at 03:08:57PM +0200, Rafael J. Wysocki wrote:
> On Monday, June 05, 2017 11:05:21 AM Dominik Brodowski wrote:
> > unfortunately, commit eed4d47efe95 (ACPI / sleep: Ignore spurious SCI
> > wakeups from suspend-to-idle) breaks suspend-to-mem on my Dell XPS 13
> > (9343) up to v4.12-rc4: Issuing
> > 
> > 	$ echo "mem" > /sys/power/state
> > 
> > in the initramfs returns, after a while, with "write error: Resource
> > busy", and the system *not* having entered the sleep state in between.
> 
> Why initramfs?

Easiest and quickest to test, with no userspace in between to interfere.
During normal operations, I see suspend-to-mem fail, then a fall-back to
s2idle...

> > In contrast thereto, 8a537ece3d94 (PM / wakeup: Integrate mechanism to abort
> > transitions in progress) still works fine, and allows to enter
> > suspend-to-mem. No real difference is to be seen in dmesg, with the notable
> > exception of
> > 
> > 	ACPI: Low-level resume complete
> > 	ACPI : EC: EC started
> > 	PM: Restoring platform NVS memory
> > 	Suspended for N.NNNN seconds
> > 
> > only showing up on working kernels. Reverting eed4d47efe95 on top of
> > v4.12-rc4 restores suspend-to-mem to work as expected.
> 
> I'm sure it is not necessary to revert all of it.

Thought so, just wanted to confirm that eed4d47efe95 is indeed the culprit.

> I guess what happens is that you get a wakeup event during suspend which is
> aborted as a result.
> 
> Please apply the partial revert below and see if it makes the issue go away.

Yes it does -- with this patch applied on top of v4.12-rc4, everything works
as expected. Both in initramfs and during normal runtime operations. Only an
(probably unrelated) issue still appears during resume, namely "cache"
complaining that the non-boot CPUs should not be sleeping:

[  140.773931] Suspended for 2.244 seconds
[  140.776090] Enabling non-boot CPUs ...
[  140.780214] x86: Booting SMP configuration:
[  140.780216] smpboot: Booting Node 0 Processor 1 APIC 0x2
[  140.781116]  cache: parent cpu1 should not be sleeping
[  140.781325] microcode: sig=0x306d4, pf=0x40, revision=0x1f
[  140.782400] CPU1 is up
[  140.788172] smpboot: Booting Node 0 Processor 2 APIC 0x1
[  140.788865]  cache: parent cpu2 should not be sleeping
[  140.788987] microcode: sig=0x306d4, pf=0x40, revision=0x24
[  140.789095] CPU2 is up
[  140.797493] smpboot: Booting Node 0 Processor 3 APIC 0x3
[  140.798208]  cache: parent cpu3 should not be sleeping
[  140.798551] CPU3 is up
[  140.802735] ACPI: Waking up from system sleep state S3


Thanks!

	Dominik


>  drivers/acpi/battery.c   |    2 +-
>  drivers/acpi/button.c    |    4 ++--
>  drivers/acpi/device_pm.c |    2 +-
>  3 files changed, 4 insertions(+), 4 deletions(-)
> 
> Index: linux-pm/drivers/acpi/battery.c
> ===================================================================
> --- linux-pm.orig/drivers/acpi/battery.c
> +++ linux-pm/drivers/acpi/battery.c
> @@ -782,7 +782,7 @@ static int acpi_battery_update(struct ac
>  	if ((battery->state & ACPI_BATTERY_STATE_CRITICAL) ||
>  	    (test_bit(ACPI_BATTERY_ALARM_PRESENT, &battery->flags) &&
>              (battery->capacity_now <= battery->alarm)))
> -		pm_wakeup_hard_event(&battery->device->dev);
> +		pm_wakeup_event(&battery->device->dev, 0);
>  
>  	return result;
>  }
> Index: linux-pm/drivers/acpi/device_pm.c
> ===================================================================
> --- linux-pm.orig/drivers/acpi/device_pm.c
> +++ linux-pm/drivers/acpi/device_pm.c
> @@ -400,7 +400,7 @@ static void acpi_pm_notify_handler(acpi_
>  	mutex_lock(&acpi_pm_notifier_lock);
>  
>  	if (adev->wakeup.flags.notifier_present) {
> -		pm_wakeup_ws_event(adev->wakeup.ws, 0, true);
> +		__pm_wakeup_event(adev->wakeup.ws, 0);
>  		if (adev->wakeup.context.work.func)
>  			queue_pm_work(&adev->wakeup.context.work);
>  	}
> Index: linux-pm/drivers/acpi/button.c
> ===================================================================
> --- linux-pm.orig/drivers/acpi/button.c
> +++ linux-pm/drivers/acpi/button.c
> @@ -217,7 +217,7 @@ static int acpi_lid_notify_state(struct
>  	}
>  
>  	if (state)
> -		pm_wakeup_hard_event(&device->dev);
> +		pm_wakeup_event(&device->dev, 0);
>  
>  	ret = blocking_notifier_call_chain(&acpi_lid_notifier, state, device);
>  	if (ret == NOTIFY_DONE)
> @@ -402,7 +402,7 @@ static void acpi_button_notify(struct ac
>  		} else {
>  			int keycode;
>  
> -			pm_wakeup_hard_event(&device->dev);
> +			pm_wakeup_event(&device->dev, 0);
>  			if (button->suspended)
>  				break;
>  

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: suspend-to-mem broken by eed4d47efe95 (ACPI / sleep: Ignore spurious SCI wakeups from suspend-to-idle)

[Rafael J. Wysocki]

On Mon, Jun 5, 2017 at 5:09 PM, Dominik Brodowski
<linux@dominikbrodowski.net> wrote:
> On Mon, Jun 05, 2017 at 03:08:57PM +0200, Rafael J. Wysocki wrote:
>> On Monday, June 05, 2017 11:05:21 AM Dominik Brodowski wrote:
>> > unfortunately, commit eed4d47efe95 (ACPI / sleep: Ignore spurious SCI
>> > wakeups from suspend-to-idle) breaks suspend-to-mem on my Dell XPS 13
>> > (9343) up to v4.12-rc4: Issuing
>> >
>> >     $ echo "mem" > /sys/power/state
>> >
>> > in the initramfs returns, after a while, with "write error: Resource
>> > busy", and the system *not* having entered the sleep state in between.
>>
>> Why initramfs?
>
> Easiest and quickest to test, with no userspace in between to interfere.
> During normal operations, I see suspend-to-mem fail, then a fall-back to
> s2idle...

I see, OK.

>> > In contrast thereto, 8a537ece3d94 (PM / wakeup: Integrate mechanism to abort
>> > transitions in progress) still works fine, and allows to enter
>> > suspend-to-mem. No real difference is to be seen in dmesg, with the notable
>> > exception of
>> >
>> >     ACPI: Low-level resume complete
>> >     ACPI : EC: EC started
>> >     PM: Restoring platform NVS memory
>> >     Suspended for N.NNNN seconds
>> >
>> > only showing up on working kernels. Reverting eed4d47efe95 on top of
>> > v4.12-rc4 restores suspend-to-mem to work as expected.
>>
>> I'm sure it is not necessary to revert all of it.
>
> Thought so, just wanted to confirm that eed4d47efe95 is indeed the culprit.
>
>> I guess what happens is that you get a wakeup event during suspend which is
>> aborted as a result.
>>
>> Please apply the partial revert below and see if it makes the issue go away.
>
> Yes it does -- with this patch applied on top of v4.12-rc4, everything works
> as expected. Both in initramfs and during normal runtime operations. Only an
> (probably unrelated) issue still appears during resume, namely "cache"
> complaining that the non-boot CPUs should not be sleeping:

That's unrelated.

OK, this still leaves us with three possibilities.

Can you please try to apply the device_pm.c part of the partial revert
alone and see if that helps (and if not, same for the button.c and
battery.c changes)?

Thanks,
Rafael

Re: suspend-to-mem broken by eed4d47efe95 (ACPI / sleep: Ignore spurious SCI wakeups from suspend-to-idle)

[Dominik Brodowski]

[-- Attachment #1: Type: text/plain, Size: 5479 bytes --]

On Mon, Jun 05, 2017 at 10:32:15PM +0200, Rafael J. Wysocki wrote:
> On Mon, Jun 5, 2017 at 5:09 PM, Dominik Brodowski
> <linux@dominikbrodowski.net> wrote:
> > On Mon, Jun 05, 2017 at 03:08:57PM +0200, Rafael J. Wysocki wrote:
> >> On Monday, June 05, 2017 11:05:21 AM Dominik Brodowski wrote:
> >> > unfortunately, commit eed4d47efe95 (ACPI / sleep: Ignore spurious SCI
> >> > wakeups from suspend-to-idle) breaks suspend-to-mem on my Dell XPS 13
> >> > (9343) up to v4.12-rc4: Issuing
> >> >
> >> >     $ echo "mem" > /sys/power/state
> >> >
> >> > in the initramfs returns, after a while, with "write error: Resource
> >> > busy", and the system *not* having entered the sleep state in between.
> >>
> >> Why initramfs?
> >
> > Easiest and quickest to test, with no userspace in between to interfere.
> > During normal operations, I see suspend-to-mem fail, then a fall-back to
> > s2idle...
> 
> I see, OK.
> 
> >> > In contrast thereto, 8a537ece3d94 (PM / wakeup: Integrate mechanism to abort
> >> > transitions in progress) still works fine, and allows to enter
> >> > suspend-to-mem. No real difference is to be seen in dmesg, with the notable
> >> > exception of
> >> >
> >> >     ACPI: Low-level resume complete
> >> >     ACPI : EC: EC started
> >> >     PM: Restoring platform NVS memory
> >> >     Suspended for N.NNNN seconds
> >> >
> >> > only showing up on working kernels. Reverting eed4d47efe95 on top of
> >> > v4.12-rc4 restores suspend-to-mem to work as expected.
> >>
> >> I'm sure it is not necessary to revert all of it.
> >
> > Thought so, just wanted to confirm that eed4d47efe95 is indeed the culprit.
> >
> >> I guess what happens is that you get a wakeup event during suspend which is
> >> aborted as a result.
> >>
> >> Please apply the partial revert below and see if it makes the issue go away.
> >
> > Yes it does -- with this patch applied on top of v4.12-rc4, everything works
> > as expected. Both in initramfs and during normal runtime operations. Only an
> > (probably unrelated) issue still appears during resume, namely "cache"
> > complaining that the non-boot CPUs should not be sleeping:
> 
> That's unrelated.
> 
> OK, this still leaves us with three possibilities.
> 
> Can you please try to apply the device_pm.c part of the partial revert
> alone and see if that helps (and if not, same for the button.c and
> battery.c changes)?

Indeed, the device_pm.c part of the partial revert helps on its own.
For completeness, applying only the chunks for button.c or battery.c
does not help; whether the following stack trace (with battery.c applied)
is of any help, seems doubtful, but here it is:

[   11.312194] PM: Suspending system (mem)
[   11.312301] Suspending console(s) (use no_console_suspend to debug)
[   15.879041] psmouse serio1: Failed to deactivate mouse on isa0060/serio1
[   16.351046] psmouse serio1: Failed to enable mouse on isa0060/serio1
[   20.695329] input: PS/2 Generic Mouse as /devices/platform/i8042/serio1/input/input7
[   20.903110] psmouse serio1: Failed to enable mouse on isa0060/serio1
[   20.923324] sd 3:0:0:0: [sda] Synchronizing SCSI cache
[   20.925306] sd 3:0:0:0: [sda] Stopping disk
[   21.126993] psmouse serio1: Failed to disable mouse on isa0060/serio1
[   21.572782] ACPI : EC: event blocked
[   22.745084] PM: suspend of devices complete after 1821.805 msecs
[   22.745085] PM: suspend devices took 11.433 seconds
[   22.745086] Component: suspend devices, time: 11433
[   22.745092] ------------[ cut here ]------------
[   22.745096] WARNING: CPU: 1 PID: 515 at /home/brodo/local/kernel/git/linux/kernel/power/suspend_test.c:55 suspend_test_finish+0x76/0x80
[   22.745096] Modules linked in:
[   22.745099] CPU: 1 PID: 515 Comm: sh Not tainted 4.12.0-rc4+ #5
[   22.745100] Hardware name: Dell Inc. XPS 13 9343/0TM99H, BIOS A11 12/08/2016
<snip>
[   22.745111] Call Trace:
[   22.745114]  suspend_devices_and_enter+0xdb/0x860
[   22.745116]  ? console_unlock.part.8+0x261/0x4a0
[   22.745118]  pm_suspend+0x3c3/0x590
[   22.745119]  state_store+0x85/0x100
[   22.745122]  kobj_attr_store+0xf/0x20
[   22.745124]  sysfs_kf_write+0x37/0x40
[   22.745125]  kernfs_fop_write+0x11c/0x1a0
[   22.745128]  __vfs_write+0x37/0x140
[   22.745130]  ? preempt_count_sub+0x9b/0x100
[   22.745132]  vfs_write+0xb1/0x170
[   22.745134]  SyS_write+0x55/0xc0
[   22.745137]  entry_SYSCALL_64_fastpath+0x18/0xa8
<snip>
[   22.745173] ---[ end trace beda9cb50fa83d31 ]---
[   22.758152] PM: late suspend of devices complete after 12.961 msecs

A second suspend then does not take as long, so I presume that there is much
noise in the messages above:

[   36.538507] PM: Preparing system for sleep (mem)
[   36.538846] Freezing user space processes ... (elapsed 0.001 seconds) done.
[   36.540248] OOM killer disabled.
[   36.540250] Freezing remaining freezable tasks ... (elapsed 0.001 seconds) done.
[   36.541671] PM: Suspending system (mem)
[   36.541759] Suspending console(s) (use no_console_suspend to debug)
[   36.561114] sd 3:0:0:0: [sda] Synchronizing SCSI cache
[   36.563755] sd 3:0:0:0: [sda] Stopping disk
[   36.593838] ACPI : EC: event blocked
[   37.785272] PM: suspend of devices complete after 1224.215 msecs
[   37.785274] PM: suspend devices took 1.244 seconds
[   37.798224] PM: late suspend of devices complete after 12.946 msecs


> Thanks,

Thank you!

	Dominik

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: suspend-to-mem broken by eed4d47efe95 (ACPI / sleep: Ignore spurious SCI wakeups from suspend-to-idle)

[Rafael J. Wysocki]

On Mon, Jun 5, 2017 at 11:14 PM, Dominik Brodowski
<linux@dominikbrodowski.net> wrote:
> On Mon, Jun 05, 2017 at 10:32:15PM +0200, Rafael J. Wysocki wrote:
>> On Mon, Jun 5, 2017 at 5:09 PM, Dominik Brodowski
>> <linux@dominikbrodowski.net> wrote:
>> > On Mon, Jun 05, 2017 at 03:08:57PM +0200, Rafael J. Wysocki wrote:
>> >> On Monday, June 05, 2017 11:05:21 AM Dominik Brodowski wrote:
>> >> > unfortunately, commit eed4d47efe95 (ACPI / sleep: Ignore spurious SCI
>> >> > wakeups from suspend-to-idle) breaks suspend-to-mem on my Dell XPS 13
>> >> > (9343) up to v4.12-rc4: Issuing
>> >> >
>> >> >     $ echo "mem" > /sys/power/state
>> >> >
>> >> > in the initramfs returns, after a while, with "write error: Resource
>> >> > busy", and the system *not* having entered the sleep state in between.
>> >>
>> >> Why initramfs?
>> >
>> > Easiest and quickest to test, with no userspace in between to interfere.
>> > During normal operations, I see suspend-to-mem fail, then a fall-back to
>> > s2idle...
>>
>> I see, OK.
>>
>> >> > In contrast thereto, 8a537ece3d94 (PM / wakeup: Integrate mechanism to abort
>> >> > transitions in progress) still works fine, and allows to enter
>> >> > suspend-to-mem. No real difference is to be seen in dmesg, with the notable
>> >> > exception of
>> >> >
>> >> >     ACPI: Low-level resume complete
>> >> >     ACPI : EC: EC started
>> >> >     PM: Restoring platform NVS memory
>> >> >     Suspended for N.NNNN seconds
>> >> >
>> >> > only showing up on working kernels. Reverting eed4d47efe95 on top of
>> >> > v4.12-rc4 restores suspend-to-mem to work as expected.
>> >>
>> >> I'm sure it is not necessary to revert all of it.
>> >
>> > Thought so, just wanted to confirm that eed4d47efe95 is indeed the culprit.
>> >
>> >> I guess what happens is that you get a wakeup event during suspend which is
>> >> aborted as a result.
>> >>
>> >> Please apply the partial revert below and see if it makes the issue go away.
>> >
>> > Yes it does -- with this patch applied on top of v4.12-rc4, everything works
>> > as expected. Both in initramfs and during normal runtime operations. Only an
>> > (probably unrelated) issue still appears during resume, namely "cache"
>> > complaining that the non-boot CPUs should not be sleeping:
>>
>> That's unrelated.
>>
>> OK, this still leaves us with three possibilities.
>>
>> Can you please try to apply the device_pm.c part of the partial revert
>> alone and see if that helps (and if not, same for the button.c and
>> battery.c changes)?
>
> Indeed, the device_pm.c part of the partial revert helps on its own.
> For completeness, applying only the chunks for button.c or battery.c
> does not help;

That's kind of expected. :-)

Anyway, I decided to revert commit eed4d47efe95 entirely as it turns
out to be premature and causes too many problems to happen all over.

Thanks,
Rafael