Re: [PATCH] Workaround for a PCI restoring bug

[Andrew Morton <akpm@linux-foundation.org>]

On Sun, 13 May 2007 10:57:18 +0200 Lukas Hejtmanek <xhejtman@ics.muni.cz> wrote:

> Hello,
> 
> On Sat, May 12, 2007 at 11:47:43PM -0700, Andrew Morton wrote:
> > This change might indeed be a suitable workaround for some busted hardware,
> > but we'd need to know quite a bit about the problem before we could merge
> > anything like this
> > 
> > So, again, please send a full bug report.  An emailed one would be OK in
> > this case.
> 
> I've reported this some time ago. I have been recommended to use -mm tree
> instead of the mainline.
> 
> I've also noticed that someone pointed out that for some reason, PCI config
> space is saved twice, the first is OK, the second saves already disabled
> devices. Unfortunately, I'm unable to find this discussion in LKM.

Yeah, that sounds risky.

Can you put a dump_stack() call into the PCI saving function?  That way
we'll see what the two call paths are.


> This is not a regression, this problem has been always present in the kernel.
> 
> These devices are not saved correctly:
> 01:01.0 CardBus bridge: Ricoh Co Ltd RL5c476 II (rev b3)
> 01:01.1 FireWire (IEEE 1394): Ricoh Co Ltd R5C552 IEEE 1394 Controller (rev 08)
> 01:01.2 Generic system peripheral [0805]: Ricoh Co Ltd R5C822 SD/SDIO/MMC/MS/MSPro Host Adapter (rev 17)
> 01:01.3 System peripheral: Ricoh Co Ltd R5C592 Memory Stick Bus Host Adapter (rev 08)
> 01:01.4 System peripheral: Ricoh Co Ltd xD-Picture Card Controller (rev 03)
> 
> A bit strange, the devices:
> 01:00.0 Ethernet controller 
> 01:02.0 Network controller 
> seem to be OK. These two devices are behind the same PCI bridge as the devices
> above.
> 
> The log contains the following and similar messages for all these devices:
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset f (was 800100, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset e (was d4fc, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset d (was d400, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset c (was 0, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset b (was 0, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset a (was 3bfff000, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset 9 (was 0, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset 8 (was 0, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset 7 (was 0, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset 6 (was 40020201, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset 5 (was 20000dc, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset 4 (was 0, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset 3 (was 822000, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset 2 (was 60700b3, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset 1 (was 2100107, writing ffffffff)
> May 12 22:07:13 anubis kernel: PM: Writing back config space on device
> 0000:01:01.0 at offset 0 (was 4761180, writing ffffffff)

It does seem wrong.