regression: a0d4922 causes PCI host constroller problems

regression: a0d4922 causes PCI host constroller problems

[Simon Holm Thøgersen]

commit a0d4922 ("USB: fix up suspend and resume for PCI host
controllers") causes a lot of these two lines to be logged:

uhci_hcd 0000:00:1d.2: host system error, PCI problems?
uhci_hcd 0000:00:1d.2: host controller process error, something bad
happened!

Normally they appear on resume after suspend to disk, but sometimes I'm
not even sure I had to suspend to disk to see them. Also, it seems like
they don't trigger if X has not been started, or at least they are much
less likely to.

Reverting the patch restores normal operation.

No USB devices were attached BTW.


Simon Holm Thøgersen

Re: regression: a0d4922 causes PCI host constroller problems

[Simon Holm Thøgersen]

fre, 16 01 2009 kl. 09:05 +0100, skrev Simon Holm Thøgersen:
> commit a0d4922 ("USB: fix up suspend and resume for PCI host
> controllers") causes a lot of these two lines to be logged:
> 
> uhci_hcd 0000:00:1d.2: host system error, PCI problems?
> uhci_hcd 0000:00:1d.2: host controller process error, something bad
> happened!
> 
> Normally they appear on resume after suspend to disk, but sometimes I'm
> not even sure I had to suspend to disk to see them. Also, it seems like
> they don't trigger if X has not been started, or at least they are much
> less likely to.
> 
> Reverting the patch restores normal operation.
> 
> No USB devices were attached BTW.

So this one has already been reported [1]. There was some mentioning of
a patch to try out instead of reverting, but I couldn't find it.

Rafael, did you have something else to suggest or can we please get a
revert of this one upstream asap before too many are wasting their time
on this one?

[1] http://lkml.org/lkml/2009/1/11/98


Simon Holm Thøgersen

Re: regression: a0d4922 causes PCI host constroller problems

[Rafael J. Wysocki]

On Friday 16 January 2009, Simon Holm Thøgersen wrote:
> fre, 16 01 2009 kl. 09:05 +0100, skrev Simon Holm Thøgersen:
> > commit a0d4922 ("USB: fix up suspend and resume for PCI host
> > controllers") causes a lot of these two lines to be logged:
> > 
> > uhci_hcd 0000:00:1d.2: host system error, PCI problems?
> > uhci_hcd 0000:00:1d.2: host controller process error, something bad
> > happened!
> > 
> > Normally they appear on resume after suspend to disk, but sometimes I'm
> > not even sure I had to suspend to disk to see them. Also, it seems like
> > they don't trigger if X has not been started, or at least they are much
> > less likely to.
> > 
> > Reverting the patch restores normal operation.
> > 
> > No USB devices were attached BTW.
> 
> So this one has already been reported [1]. There was some mentioning of
> a patch to try out instead of reverting, but I couldn't find it.
> 
> Rafael, did you have something else to suggest or can we please get a
> revert of this one upstream asap before too many are wasting their time
> on this one?

Yes, please try the patch from http://lkml.org/lkml/2009/1/13/144 and report
back.

Thanks,
Rafael

Re: regression: a0d4922 causes PCI host constroller problems

[Simon Holm Thøgersen]

fre, 16 01 2009 kl. 13:06 +0100, skrev Rafael J. Wysocki:
> On Friday 16 January 2009, Simon Holm Thøgersen wrote:
> > fre, 16 01 2009 kl. 09:05 +0100, skrev Simon Holm Thøgersen:
> > > commit a0d4922 ("USB: fix up suspend and resume for PCI host
> > > controllers") causes a lot of these two lines to be logged:
> > > 
> > > uhci_hcd 0000:00:1d.2: host system error, PCI problems?
> > > uhci_hcd 0000:00:1d.2: host controller process error, something bad
> > > happened!
> > > 
> > > Normally they appear on resume after suspend to disk, but sometimes I'm
> > > not even sure I had to suspend to disk to see them. Also, it seems like
> > > they don't trigger if X has not been started, or at least they are much
> > > less likely to.
> > > 
> > > Reverting the patch restores normal operation.
> > > 
> > > No USB devices were attached BTW.
> > 
> > So this one has already been reported [1]. There was some mentioning of
> > a patch to try out instead of reverting, but I couldn't find it.
> > 
> > Rafael, did you have something else to suggest or can we please get a
> > revert of this one upstream asap before too many are wasting their time
> > on this one?
> 
> Yes, please try the patch from http://lkml.org/lkml/2009/1/13/144 and report
> back.

That fixes suspend to disk for me.

Thanks Rafael, you really do an important and fantastic job on the
regression work.


Simon Holm Thøgersen