Bug 32422

(In reply to comment #0)
> Created an attachment (id=41150) [details]
> Picture describing a thousand words.
> 
> "A picture says more than a thousand words" (aka see attachment)
> 
> This is what happens when I try to view the XKCD "Online Communities 2" map
> (http://xkcd.com/802_large/) or other large images in firefox.

Additional information:
This only happens with 'Option NoAccel="false"' in xorg.conf. Enabling NoAccel (e.g. turning acceleration OFF, using 'softpipe') the image renders as expected.

What GPU is this?  Can you attach your xorg log and dmesg?

Created attachment 41151 [details]
xorg log

Created attachment 41152 [details] [review]
output of /var/log/messages since bootup

Hardware is an IBM Thinkpad T42, which has a ATI Radeon 9600 Mobility M10 / rv350 graphics card.

*** This bug has been marked as a duplicate of bug 31708 ***

(In reply to comment #6)
> 
> *** This bug has been marked as a duplicate of bug 31708 ***

I don't think this bug is related to 31708:  Even with the patch for bug 31708
applied, I see the same corruption in firefox (and only firefox; chrome, imagemagick, the KDE image viewer for example all display these large images correctly).

(In reply to comment #7)
> Even with the patch for bug 31708 applied, I see the same corruption in firefox [...]

That's with the latest version (v3) of the patch? (Jerome even submitted v4 later, but I don't think the difference from v3 should be relevant for this problem; wouldn't hurt to try of course)

(In reply to comment #8)
> (In reply to comment #7)
> > Even with the patch for bug 31708 applied, I see the same corruption in firefox [...]
> 
> That's with the latest version (v3) of the patch? (Jerome even submitted v4
> later, but I don't think the difference from v3 should be relevant for this
> problem; wouldn't hurt to try of course)

I have tried with the version present in 2.6.37-rc6, which according to the changelog (http://www.kernel.org/pub/linux/kernel/v2.6/testing/ChangeLog-2.6.37-rc6) is v3. 

ps.
I only noticed that there were new posts on this bug after submitting my comment on the other bug.

Really looks like a separate bug.

Created attachment 41339 [details] [review]
Always include CPU domain for BO eviction

Does this patch help?

Though even if it does, I don't understand why it fails to evict to GTT with Jerome's v3 fix, might be interesting to add some debugging output to find out why ttm_bo_mem_space() fails.

(In reply to comment #11)
> Created an attachment (id=41339) [details]
> Always include CPU domain for BO eviction
> 
> Does this patch help?
> 
> Though even if it does, I don't understand why it fails to evict to GTT with
> Jerome's v3 fix, might be interesting to add some debugging output to find out
> why ttm_bo_mem_space() fails.

Sorry, it still fails to render properly. I did notice that at least for the xkcd image it sometimes displays properly, even with only the v3 patch. However I haven't yet been able to determine when this happens. I just happened to notice this because I left the xkcd image open in a tab, and when I later came back to that tab I could see the whole image and pan around without issue. I also don't know yet if this also happens on 2.6.36, I would need to 'test'(try) that.

I've had this same issue for ages on X1400. I has been seemingly random, but always with large images. With this one I can finally reproduce it reliably! Just pan around the xkcd-image for a while, keep switching between tabs and it will happen eventually.

Created attachment 50166 [details]
Image corruption on Firefox

linux-3.0.1
xf86-video-ati-6.14.2
xorg-server-1.10.3.901
firefox-5.0

Created attachment 50224 [details] [review]
Some additional debugging output for the failure

Please attach dmesg from reproducing the problem with this kernel patch applied.

Ah. Sorry. I examined the original reporter's logs and compared them to mine. I have no errors in dmesg related to TTM (or anything else for that matter) whatsoever. Or do I need to enable some particular debugging?

Might be that my problem is different - however the symptoms are identical.

asasassadadasdasdasdasdasdadasdad

(In reply to comment #17)
> asasassadadasdasdasdasdasdadasdad

I'm terribly sorry, please ignore that.

-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/drm/amd/issues/965.