Bug 27571 - Regression: r600 CS checker rejects GL_DEPTH_TEST w/o depth buffer.
Summary: Regression: r600 CS checker rejects GL_DEPTH_TEST w/o depth buffer.
Status: RESOLVED FIXED
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Radeon (show other bugs)
Version: unspecified
Hardware: Other All
: medium major
Assignee: Default DRI bug account
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2010-04-09 21:57 UTC by Nick Bowler
Modified: 2010-04-23 03:15 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
Test case. (703 bytes, text/x-c)
2010-04-09 21:57 UTC, Nick Bowler
no flags Details
Full kernel log. (16.80 KB, text/plain)
2010-04-09 21:58 UTC, Nick Bowler
no flags Details
Actual full kernel log. (31.29 KB, text/plain)
2010-04-12 18:54 UTC, Nick Bowler
no flags Details

Description Nick Bowler 2010-04-09 21:57:23 UTC
Created attachment 34858 [details]
Test case.

The command stream checker recently added to the radeon driver kills
programs which render to a target which does not have a depth buffer
whilst GL_DEPTH_TEST is enabled.  Instead of simply not performing
depth buffer operations, the result is program termination and some
rather cryptic error messages:

  drmRadeonCmdBuffer: -22. Kernel failed to parse or rejected command
  stream.  See dmesg for more info.

and in dmesg:
  radeon 0000:03:00.0: z/stencil with no depth buffer
  radeon 0000:03:00.0: r600_packet3_check:1109 invalid cmd stream 815
  [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !

I first noticed this when a program that works using 2.6.33 died
rendering to a FBO, but it is trivially reproducible without using
anything fancy.  A test case is attached which works normally with
either Linux 2.6.33, the software rasterizer, or the intel driver.
The failure is reproducible with both mesa 7.8.1 and git master.
Comment 1 Nick Bowler 2010-04-09 21:58:27 UTC
Created attachment 34859 [details]
Full kernel log.

Just in case it's needed.
Comment 2 Nick Bowler 2010-04-11 17:00:36 UTC
The fbo-nodepth piglit test also triggers this issue.
Comment 3 Nick Bowler 2010-04-12 18:54:00 UTC
Created attachment 34947 [details]
Actual full kernel log.

I just realized that the previous "Full kernel log" was truncated and hence wasn't so full, after all.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.