Re: -rc6: something (probably active usb bluetooth) blocks suspend

Re: -rc6: something (probably active usb bluetooth) blocks suspend

[Pavel Machek]

On Mon 2009-11-16 12:48:29, Oliver Neukum wrote:
> Am Montag, 16. November 2009 12:22:54 schrieb Pavel Machek:
> 
> Hi Pavle,

:-)

> first the obvious question. Does it work if you down the interface?
> (not by killswitch)

Well, I'm not sure how repeatable it is. It happens once in two tries
today.

> > usb 5-1: usb suspend
> > usb 5-1: usb auto-resume
> > uhci_hcd 0000:00:1d.3: port 1 portsc 0095,01
> > usb 5-1: finish resume
> > uhci_hcd 0000:00:1d.3: reserve dev 2 ep81-INT, period 1, phase 0, 23
> > us
> 
> Why is the device autoresumed here? Are you using autosuspend?
> This is odd, as the freezer should block the send path and remote
> wakeups should be deferred.

I believe I have it enabled in kconfig but not during runtime.

(Another problem is that I need to turn the kill switch on/off to get
my bt usb after resume. That used not be the case long time ago.)
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

-rc6: something (probably active usb bluetooth) blocks suspend

[Pavel Machek]

Linux amd 2.6.32-rc6-00498-g05b553a-dirty #89 SMP Fri Nov 13 10:43:31
CET 2009 i686 GNU/Linux

...and yes, it is a regression. Hw is thinkpad x60.

Tried to suspend, machine came right back, used radio kill switch, and
it suspended ok. Relevant messages seem to be:

								Pavel

PM: Preparing system for mem sleep
Freezing user space processes ... (elapsed 0.02 seconds) done.
Freezing remaining freezable tasks ... (elapsed 0.00 seconds) done.
PM: Entering mem sleep
usb 5-2: usb suspend
btusb_intr_complete: hci0 urb f3b57420 failed to resubmit (1)
uhci_hcd 0000:00:1d.3: release dev 2 ep81-INT, period 1, phase 0, 23
us
btusb_bulk_complete: hci0 urb f3b57680 failed to resubmit (1)
btusb_bulk_complete: hci0 urb f3b57718 failed to resubmit (1)
usb 5-1: usb suspend
usb 5-1: usb auto-resume
uhci_hcd 0000:00:1d.3: port 1 portsc 0095,01
usb 5-1: finish resume
uhci_hcd 0000:00:1d.3: reserve dev 2 ep81-INT, period 1, phase 0, 23
us
hub 5-0:1.0: port 1 nyet suspended
hub 5-0:1.0: suspend error -16
pm_op(): usb_dev_suspend+0x0/0x10 returns -16
PM: Device usb5 failed to suspend: error -16
PM: Some devices failed to suspend
thinkpad_acpi: ACPI backlight control delay disabled
usb 5-2: usb resume
uhci_hcd 0000:00:1d.3: port 2 portsc 00a5,01
usb 5-2: finish resume
usb usb4: usb resume
usb usb4: wakeup_rh

									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: -rc6: something (probably active usb bluetooth) blocks suspend

[Oliver Neukum]

Am Montag, 16. November 2009 12:22:54 schrieb Pavel Machek:

Hi Pavle,

first the obvious question. Does it work if you down the interface?
(not by killswitch)

> usb 5-1: usb suspend
> usb 5-1: usb auto-resume
> uhci_hcd 0000:00:1d.3: port 1 portsc 0095,01
> usb 5-1: finish resume
> uhci_hcd 0000:00:1d.3: reserve dev 2 ep81-INT, period 1, phase 0, 23
> us

Why is the device autoresumed here? Are you using autosuspend?
This is odd, as the freezer should block the send path and remote
wakeups should be deferred.

	Regards
		Oliver

Re: -rc6: something (probably active usb bluetooth) blocks suspend

[Oliver Neukum]

Am Montag, 16. November 2009 12:22:54 schrieb Pavel Machek:
> Linux amd 2.6.32-rc6-00498-g05b553a-dirty #89 SMP Fri Nov 13 10:43:31
> CET 2009 i686 GNU/Linux
> 
> ...and yes, it is a regression. Hw is thinkpad x60.
> 
> Tried to suspend, machine came right back, used radio kill switch, and
> it suspended ok. Relevant messages seem to be:

I forgot to ask. Is this repeatable? The code is build on the assumption
that workqueues are frozen as the system calls suspend for system
sleep. If that assumption is wrong, a race can lead to this failure.
Is it wrong?

	Regards
		Oliver

Re: -rc6: something (probably active usb bluetooth) blocks suspend

[Oliver Neukum]

Am Samstag, 14. November 2009 12:10:21 schrieb Pavel Machek:
> On Mon 2009-11-16 12:48:29, Oliver Neukum wrote:
> > Am Montag, 16. November 2009 12:22:54 schrieb Pavel Machek:
> >
> > Hi Pavle,
> >
> :-)
> :
> > first the obvious question. Does it work if you down the interface?
> > (not by killswitch)
> 
> Well, I'm not sure how repeatable it is. It happens once in two tries
> today.

Ok, so it looks like a race.

> > > usb 5-1: usb suspend
> > > usb 5-1: usb auto-resume
> > > uhci_hcd 0000:00:1d.3: port 1 portsc 0095,01
> > > usb 5-1: finish resume
> > > uhci_hcd 0000:00:1d.3: reserve dev 2 ep81-INT, period 1, phase 0, 23
> > > us
> >
> > Why is the device autoresumed here? Are you using autosuspend?
> > This is odd, as the freezer should block the send path and remote
> > wakeups should be deferred.
> 
> I believe I have it enabled in kconfig but not during runtime.

Please check.
 
> (Another problem is that I need to turn the kill switch on/off to get
> my bt usb after resume. That used not be the case long time ago.)

That is possibly another problem.

	Regards
		Oliver

Re: -rc6: something (probably active usb bluetooth) blocks suspend

[Pavel Machek]

On Mon 2009-11-16 14:43:20, Oliver Neukum wrote:
> Am Montag, 16. November 2009 12:22:54 schrieb Pavel Machek:
> > Linux amd 2.6.32-rc6-00498-g05b553a-dirty #89 SMP Fri Nov 13 10:43:31
> > CET 2009 i686 GNU/Linux
> > 
> > ...and yes, it is a regression. Hw is thinkpad x60.
> > 
> > Tried to suspend, machine came right back, used radio kill switch, and
> > it suspended ok. Relevant messages seem to be:
> 
> I forgot to ask. Is this repeatable? The code is build on the

It happened once...  

> assumption
> that workqueues are frozen as the system calls suspend for system
> sleep. If that assumption is wrong, a race can lead to this failure.
> Is it wrong?

There are two kinds of workqueues -- AFAICT -- kind one is frozen,
kind two is not.
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: -rc6: something (probably active usb bluetooth) blocks suspend

[Alan Stern]

On Mon, 16 Nov 2009, Oliver Neukum wrote:

> Am Samstag, 14. November 2009 12:10:21 schrieb Pavel Machek:
> > On Mon 2009-11-16 12:48:29, Oliver Neukum wrote:
> > > Am Montag, 16. November 2009 12:22:54 schrieb Pavel Machek:
> > >
> > > Hi Pavle,
> > >
> > :-)
> > :
> > > first the obvious question. Does it work if you down the interface?
> > > (not by killswitch)
> > 
> > Well, I'm not sure how repeatable it is. It happens once in two tries
> > today.
> 
> Ok, so it looks like a race.
> 
> > > > usb 5-1: usb suspend
> > > > usb 5-1: usb auto-resume
> > > > uhci_hcd 0000:00:1d.3: port 1 portsc 0095,01
> > > > usb 5-1: finish resume
> > > > uhci_hcd 0000:00:1d.3: reserve dev 2 ep81-INT, period 1, phase 0, 23
> > > > us
> > >
> > > Why is the device autoresumed here? Are you using autosuspend?
> > > This is odd, as the freezer should block the send path and remote
> > > wakeups should be deferred.
> > 
> > I believe I have it enabled in kconfig but not during runtime.
> 
> Please check.

If you want to learn the source of the autoresume, do something like
this:

Index: usb-2.6/drivers/usb/core/hub.c
===================================================================
--- usb-2.6.orig/drivers/usb/core/hub.c
+++ usb-2.6/drivers/usb/core/hub.c
@@ -2334,6 +2334,8 @@ int usb_port_resume(struct usb_device *u
 		/* drive resume for at least 20 msec */
 		dev_dbg(&udev->dev, "usb %sresume\n",
 				(msg.event & PM_EVENT_AUTO ? "auto-" : ""));
+		if ((msg.event & PM_EVENT_AUTO) && udev->maxchild == 0)
+			dump_stack();
 		msleep(25);
 
 		/* Virtual root hubs can trigger on GET_PORT_STATUS to

This will show you what is responsible for all autoresumes of non-hub 
USB devices.

Alan Stern

Re: -rc6: something (probably active usb bluetooth) blocks suspend

[Pavel Machek]

On Mon 2009-11-16 14:43:20, Oliver Neukum wrote:
> Am Montag, 16. November 2009 12:22:54 schrieb Pavel Machek:
> > Linux amd 2.6.32-rc6-00498-g05b553a-dirty #89 SMP Fri Nov 13 10:43:31
> > CET 2009 i686 GNU/Linux
> > 
> > ...and yes, it is a regression. Hw is thinkpad x60.
> > 
> > Tried to suspend, machine came right back, used radio kill switch, and
> > it suspended ok. Relevant messages seem to be:
> 
> I forgot to ask. Is this repeatable? The code is build on the assumption
> that workqueues are frozen as the system calls suspend for system
> sleep. If that assumption is wrong, a race can lead to this failure.
> Is it wrong?

I (shortly) tried to repeat it but could not.
									Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: -rc6: something (probably active usb bluetooth) blocks suspend

[Pavel Machek]

Hi!

> > > Why is the device autoresumed here? Are you using autosuspend?
> > > This is odd, as the freezer should block the send path and remote
> > > wakeups should be deferred.
> > 
> > I believe I have it enabled in kconfig but not during runtime.
> 
> Please check.

CONFIG_USB_SUSPEND=y

I'm not aware of  boot scripts enabling anything.

> > (Another problem is that I need to turn the kill switch on/off to get
> > my bt usb after resume. That used not be the case long time ago.)
> 
> That is possibly another problem.

Yes, I guess so. With those many USB experts around, I thought I'd
mention it :-).
									Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html