Bug 105127 - [CI][DRMTIP] igt@kms_rotation_crc@*| igt@kms_chv_cursor_fail@pipe - dmesg-warn - *ERROR* Overflow of CRC buffer, userspace reads too slow.
Summary: [CI][DRMTIP] igt@kms_rotation_crc@*| igt@kms_chv_cursor_fail@pipe - dmesg-war...
Status: NEW
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: XOrg git
Hardware: Other All
: medium normal
Assignee: Intel GFX Bugs mailing list
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard: ReadyForDev
Keywords:
Depends on:
Blocks:
 
Reported: 2018-02-16 12:16 UTC by Martin Peres
Modified: 2019-09-09 18:09 UTC (History)
2 users (show)

See Also:
i915 platform: BXT, BYT, CFL, GLK, SNB
i915 features: display/Other


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Martin Peres 2018-02-16 12:16:14 UTC
[ 110.245656] hpet1: lost 7160 rtc interrupts
[ 110.868837] [drm:drm_crtc_add_crc_entry] *ERROR* Overflow of CRC buffer, userspace reads too slow.

https://intel-gfx-ci.01.org/tree/drm-tip/IGT_4257/shard-glkb4/igt@kms_rotation_crc@bad-tiling.html
Comment 1 Martin Peres 2018-04-25 09:39:29 UTC
Also seen on APL: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_4089/shard-apl3/igt@kms_rotation_crc@sprite-rotation-90-pos-100-0.html

[  215.962250] [drm:drm_crtc_add_crc_entry] *ERROR* Overflow of CRC buffer, userspace reads too slow.
Comment 3 CI Bug Log 2019-02-21 14:56:11 UTC
The CI Bug Log issue associated to this bug has been updated.

### New filters associated

* CFL: igt@kms_chv_cursor_fail@pipe-* - dmesg-warn-*ERROR* Overflow of CRC buffer, userspace reads too slow.
  - https://intel-gfx-ci.01.org/tree/drm-tip/drmtip_225/fi-cfl-8109u/igt@kms_chv_cursor_fail@pipe-b-64x64-right-edge.html
Comment 4 Matt Roper 2019-09-09 16:37:56 UTC
kms_rotation_crc and kms_cursor_edge_walk (the new name for kms_chv_cursor_fail) use igt_pipe_crc_start/stop to turn CRC collection on and off.  Between these two calls, an internal buffer of 128 entries will capture CRC's for each frame and, if the buffer fills up and overflows, the message here will be reported.

For kms_rotation_crc at least, the manual use of start/stop is unnecessary since the test is only interested in reading single CRC's at various times and doesn't care about sequences of frame-by-frame CRC's.  Thus switching it over to the igt_pipe_crc_collect_crc() which will start/stop the CRC collection around the reading of a single CRC value should solve the problem since the test won't be collecting a bunch of unwanted CRC values while it's busy doing other work.  I've submitted a patch for that here:  https://patchwork.freedesktop.org/series/66448/

From a quick glance, kms_cursor_edge_walk looks like it probably does need a continuous sequence of CRC values; it may just be that it's doing too long a sequence of operations between start/stop.  I'll look more into the details of that test later.

Ultimately this defect should have no end-user impact since end user software doesn't utilize display CRC's.  The main impact here is to our CI system, which will trip over these messages.


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.