Nouveau - Oct 2013 - Resource map sanity check fails after GRUB "keeps" the gfx mode

Hi

On Fri, Sep 27, 2013 at 5:37 PM, Pavel Roskin <proski at gnu.org>
wrote:
> On Fri, 27 Sep 2013 10:33:26 +0100
> Emil Velikov <emil.l.velikov at gmail.com> wrote:
>
>> FWIW another nvc0 user is experiencing the same/similar issue and the
>> patch seems to resolve the problem in his case.
>>
>> https://bugs.freedesktop.org/show_bug.cgi?id=69488
>
> Good to hear that my patch actually makes more than cosmetic difference
> for somebody.  My system doesn't hang even with the stack trace.
I am aware of this and there is already a fix pending from 2 weeks
ago. Unfortunately, I have been in the US for the last 2 weeks and
haven't really been able to push on that. Anyhow, it should really be
just a cosmetic issue. We need to mark the sysfb FB as busy to avoid
the nasty warnings. Other than that, this shouldn't cause any issues.

Thanks
David
> I posted another patch on the bug page.  It widens the BOOTFB resource
> to match the PCI BAR instead of making it busy. With the new patch,
> nvidiafb doesn't report that the resource is busy, it reports
> unsupported hardware, which is true.
>
> Old patch:
> nvidiafb 0000:01:00.0: BAR 3: can't reserve [mem 0xf0000000-0xf1ffffff
> 64bit pref]
> nvidiafb: cannot request PCI regions
>
> New patch:
> nvidiafb: Device ID: 10de0ffc
> nvidiafb: unknown NV_ARCH
>
> I'll send the patch to x86 at kernel.org, it's clearly not an
nvidia issue.
>
> --
> Regards,
> Pavel Roskin

Quoting David Herrmann <dh.herrmann at gmail.com>:
> I am aware of this and there is already a fix pending from 2 weeks
> ago. Unfortunately, I have been in the US for the last 2 weeks and
> haven't really been able to push on that. Anyhow, it should really be
> just a cosmetic issue. We need to mark the sysfb FB as busy to avoid
> the nasty warnings. Other than that, this shouldn't cause any issues.
OK, you are the one who must be better aware of the whole picture, so  
if the intention is to make the resource busy, that's my first patch  
and whatever you are trying to push.

You still may want to have a look at my second patch that widens the  
resource to match an appropriate PCI BAR.

I'm too puzzled why the resource conflict warning causes a GPU lockup  
for another user.  Maybe it's an unrelated issue.  Also, I don't use  
UEFI.  I plan to test it eventually if I figure out how to migrate to  
it without reinstalling everything.

-- 
Regards,
Pavel Roskin

Hi Pavel

On Fri, Sep 27, 2013 at 9:40 PM, Pavel Roskin <proski at gnu.org>
wrote:
> Quoting David Herrmann <dh.herrmann at gmail.com>:
>
>> I am aware of this and there is already a fix pending from 2 weeks
>> ago. Unfortunately, I have been in the US for the last 2 weeks and
>> haven't really been able to push on that. Anyhow, it should really
be
>> just a cosmetic issue. We need to mark the sysfb FB as busy to avoid
>> the nasty warnings. Other than that, this shouldn't cause any
issues.
>
>
> OK, you are the one who must be better aware of the whole picture, so if
the
> intention is to make the resource busy, that's my first patch and
whatever
> you are trying to push.
>
> You still may want to have a look at my second patch that widens the
> resource to match an appropriate PCI BAR.
>
> I'm too puzzled why the resource conflict warning causes a GPU lockup
for
> another user.  Maybe it's an unrelated issue.  Also, I don't use
UEFI.  I
> plan to test it eventually if I figure out how to migrate to it without
> reinstalling everything.
Thanks for your investigations. I finally sent my patch and put you on
CC. Sorry for the delay, but I have mostly catched up with all emails
now.

This really just suppresses the warning and nothing else. It does not
fix any lockup/crash/... And your PCI-BAR adjustment doesn't change
anything either, sorry. There is no reason to size the resource
according to PCI tables as it is totally fine to remap partial
segments. resource management is only used for sanity checks and
bookkeeping. If there are real bugs (other than just "Your kernel is
fine." warnings) then we need to bisect them.

Let me know if there are still related issues. I will gladly try to
resolve them.
Thanks!
David