All machines: All tests - map pfn RAM range req write-combining for [mem 0x.*], got write-back

What's the right component to be assigned here?

changed title from All: All tests - map pfn RAM range req write-combining for [mem 0x.*], got write-back to All machines: All tests - map pfn RAM range req write-combining for [mem 0x.*], got write-back

The CI Bug Log issue associated to this bug has been updated by Lakshmi Vudum.

New filters associated

TGL: igt@runner@aborted - fail - Previous test: kms_cursor_crc (pipe-b-cursor-128x128-sliding)
- https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10089/shard-tglb5/igt@runner@aborted.html

added GPU hang label

We see this failure in Mesa on uprevving from 5.11 to v5.13-rc1. https://gitlab.freedesktop.org/mesa/mesa/-/jobs/9813893#L723 is an associated job log.

This seems also to be on GLK, and I guess you don't see GPU hangs in "wflinfo" with an earlier kernel version?

Right.

This regression i.e. getting these messages, and a lot more GPU hangs started to happen between following commits:

drm-tip: 2021y-05m-10d-14h-18m-33s UTC integration manifest
drm-tip: 2021y-05m-11d-14h-40m-05s UTC integration manifest

There are GPU hangs even when machine boots to idle desktop:

...
[    4.023615] r8169 0000:01:00.0 enp1s0: Link is Down
[    6.169007] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x112f96000-0x112f96fff], got write-back
[    6.204104] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x11281d000-0x112820fff], got write-back
[    6.204685] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x112821000-0x112821fff], got write-back
[    6.204839] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x112833000-0x112833fff], got write-back
[    6.204925] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x112823000-0x112832fff], got write-back
[    6.204938] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x112835000-0x112844fff], got write-back
[    6.278811] i915 0000:00:02.0: [drm] Resetting rcs0 for CS error
[    6.278866] i915 0000:00:02.0: [drm] X[551] context reset due to GPU hang
[    6.284855] i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:8cbf0db6, in X [551]
[    6.313762] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x112143000-0x112146fff], got write-back
[    6.340899] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x11214a000-0x11214dfff], got write-back
[    6.349089] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x11214e000-0x112151fff], got write-back
[    6.481893] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x112152000-0x112152fff], got write-back
[    6.504618] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x112153000-0x112157fff], got write-back
[    6.509959] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x1121d3000-0x1122d2fff], got write-back
[    6.609051] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x1120f7000-0x112106fff], got write-back
[    6.610526] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x1123ee000-0x1123f1fff], got write-back
[    6.610556] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x11280d000-0x11281cfff], got write-back
[    7.102179] r8169 0000:01:00.0 enp1s0: Link is Up - 1Gbps/Full - flow control off
[    7.102197] IPv6: ADDRCONF(NETDEV_CHANGE): enp1s0: link becomes ready
[    7.167628] fuse: init (API version 7.33)
[    7.742709] x86/PAT: ibus-extension-:751 map pfn RAM range req write-combining for [mem 0x111d83000-0x111d83fff], got write-back
[    7.745256] x86/PAT: ibus-x11:754 map pfn RAM range req write-combining for [mem 0x111df3000-0x111df3fff], got write-back
[    7.766427] x86/PAT: ibus-ui-gtk3:747 map pfn RAM range req write-combining for [mem 0x111df4000-0x111df4fff], got write-back
[   10.398731] x86/PAT: unity_support_t:1163 map pfn RAM range req write-combining for [mem 0x110a44000-0x110a44fff], got write-back
[   10.433328] x86/PAT: gnome-session-c:1183 map pfn RAM range req write-combining for [mem 0x110abf000-0x110abffff], got write-back
[   10.504834] x86/PAT: unity_support_t:1163 map pfn RAM range req write-combining for [mem 0x110b42000-0x110b45fff], got write-back
[   10.504881] x86/PAT: unity_support_t:1163 map pfn RAM range req write-combining for [mem 0x110b38000-0x110b38fff], got write-back
[   10.506182] x86/PAT: unity_support_t:1163 map pfn RAM range req write-combining for [mem 0x110a02000-0x110a02fff], got write-back
[   10.508744] x86/PAT: gnome-session-c:1183 map pfn RAM range req write-combining for [mem 0x110b0e000-0x110b11fff], got write-back
[   10.508796] x86/PAT: gnome-session-c:1183 map pfn RAM range req write-combining for [mem 0x110b12000-0x110b12fff], got write-back
[   10.508938] x86/PAT: gnome-session-c:1183 map pfn RAM range req write-combining for [mem 0x110b24000-0x110b24fff], got write-back
[   10.509057] x86/PAT: gnome-session-c:1183 map pfn RAM range req write-combining for [mem 0x110b14000-0x110b23fff], got write-back
[   10.512139] x86/PAT: gnome-session-c:1183 map pfn RAM range req write-combining for [mem 0x10e47f000-0x10e480fff], got write-back
[   10.539140] i915 0000:00:02.0: [drm] Resetting rcs0 for CS error
[   10.539184] i915 0000:00:02.0: [drm] unity_support_t[1163] context reset due to GPU hang
[   10.545552] i915 0000:00:02.0: [drm] GPU HANG: ecode 9:1:8cbf8db6, in unity_support_t [1163]
[   11.392935] x86/PAT: compiz:1257 map pfn RAM range req write-combining for [mem 0x10e4aa000-0x10e4aafff], got write-back
[   11.582224] x86/PAT: compiz:1257 map pfn RAM range req write-combining for [mem 0x10e480000-0x10e480fff], got write-back
[   11.582402] x86/PAT: compiz:1257 map pfn RAM range req write-combining for [mem 0x10d004000-0x10d004fff], got write-back
[   11.582522] x86/PAT: compiz:1257 map pfn RAM range req write-combining for [mem 0x10d006000-0x10d015fff], got write-back
[   11.613518] x86/PAT: compiz:1257 map pfn RAM range req write-combining for [mem 0x10bdb0000-0x10bdb3fff], got write-back
[   11.765453] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x122de4000-0x122df3fff], got write-back
[   11.772608] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x122dfa000-0x122e09fff], got write-back
[   11.774630] x86/PAT: X:551 map pfn RAM range req write-combining for [mem 0x122e10000-0x122e1ffff], got write-back
...

I'm seeing them on BXT & GLK machines, but not on SKL, KBL or CML. I.e. it seems to be specific to LP machines.

This regression i.e. getting these messages, and a lot more GPU hangs started to happen between following commits:

drm-tip: 2021y-05m-10d-14h-18m-33s UTC integration manifest

drm-tip: 2021y-05m-11d-14h-40m-05s UTC integration manifest

Looks like the patch culprit has been identified:

https://patchwork.kernel.org/project/intel-gfx/patch/20210326055505.1424432-5-hch@lst.de/

We also see it on BXT and GLX, but not AML-Y.

mentioned in merge request mesa/mesa!10785 (merged)

http://lore.kernel.org/r/CAHk-=whcr5M=4Mz2ydu4XtxTL_34WkXPnmFmA4f8r+ELXDC6hg@mail.gmail.com

I got now also TGL-H results, and I'm not seeing this issue there, although original report above specifically linked TGL results (my TGL test machine has just a lot of other GPU hangs for which I've filed bugs months ago to mesa & drm-intel).

I.e. on my set of GEN9 + TGL machines, these errors, and the sharply increased[1] number of GPU hangs, happen only on BXT & GLK.

[1] Before this, there were only the daily #1205 (closed) GPU hang + recovery failures on BXT & GLK. That issue doesn't happen with ClearLinux + Weston + performance governor, only with Ubuntu + X/unity/compiz + powersave/ondemand governor. For this issue, the distro/governor and Weston/X distinctions don't have any impact though.

commit 293837b9ac8d3021657f44c9d7a14948ec01c5d0
Author: Linus Torvalds <torvalds@linux-foundation.org>
Date:   Wed May 19 05:55:57 2021 -1000

    Revert "i915: fix remap_io_sg to verify the pgprot"
    
    This reverts commit b12d691ea5e01db42ccf3b4207e57cb3ce7cfe91.
    
    It turns out this is not ready for primetime yet.  The intentions are
    good, but using remap_pfn_range() requires that there is nothing already
    mapped in the area, and the i915 code seems to very much intentionally
    remap the same area multiple times.
    
    That will then just trigger the
    
                    BUG_ON(!pte_none(*pte));
    
    in mm/memory.c: remap_pte_range().
    
    There are also reports of mapping type inconsistencies, resulting in
    warnings and in screen corruption.
    
    Link: https://lore.kernel.org/lkml/20210519024322.GA29704@xsang-OptiPlex-9020/
    Link: https://lore.kernel.org/lkml/YKUjvoaKKggAmpIR@sf/
    Link: https://lore.kernel.org/lkml/b6b61cf0-5874-f4c0-1fcc-4b3848451c31@redhat.com/
    Reported-by: kernel test robot <oliver.sang@intel.com>
    Reported-by: Kalle Valo <kvalo@codeaurora.org>
    Reported-by: Hans de Goede <hdegoede@redhat.com>
    Reported-by: Sergei Trofimovich <slyfox@gentoo.org>
    Acked-by: Christoph Hellwig <hch@lst.de>
    Cc: Chris Wilson <chris@chris-wilson.co.uk>
    Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
    Cc: Jani Nikula <jani.nikula@linux.intel.com>
    Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
    Cc: Peter Zijlstra <peterz@infradead.org>
    Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
    Cc: Andrew Morton <akpm@linux-foundation.org>
    Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

This is in current Linus' master, expected to be released in v5.13-rc3 on Sunday, and make its way to drm-tip right after that.

Thanks Jani, should we close only when removed from core-for-ci and merged officially?

A CI Bug Log filter associated to this bug has been updated by :

Description: ELK TGL: igt@runner@aborted - fail - Previous test: kms_cursor_crc (pipe-b-cursor-128x128-sliding)|gem_wait (busy@all)

Equivalent query: runconfig_tag IS IN ["DRM-TIP"] AND (machine_name IS IN ["fi-tgl-u", "re-tgl-u", "fi-tgl-u2", "shard-tglb1", "shard-tglb2", "shard-tglb3", "shard-tglb4", "shard-tglb5", "shard-tglb6", "re-tgl1-display", "re-tgl2-display", "shard-tglb7", "shard-tglb8", "fi-elk-e7500", "shard-tglb0", "fi-tgl-y", "re-tgl-y", "shard-tglb9", "re-tgl-cham", "fi-tgl-guc", "re-tgl-u2", "fi-tgl-dsi", "ba-tgl-1", "custom-tgl-joshikun", "tgl-u-lpddr-ba", "re-tgl-u3", "fi-tgl-h", "fi-tgl-1115g4"] OR machine_tag IS IN ["ELK", "TGL"]) AND ((testsuite_name = "IGT" AND test_name IS IN ["igt@runner@aborted"])) AND ((testsuite_name = "IGT" AND status_name IS IN ["fail"])) AND stdout ~= 'Previous test: kms_cursor_crc \(pipe-b-cursor-128x128-sliding\)|gem_wait \(busy@all\)'

New failures caught by the filter:

The ELK failures above are still using the old kernel baseline for whatever reason.

Closing.

closed

mentioned in issue #3519 (closed)

The CI Bug Log issue associated to this bug has been archived.

New failures matching the above filters will not be associated to this bug anymore.

All machines: All tests - map pfn RAM range req write-combining for [mem 0x.*], got write-back

Child items 0

Activity

New filters associated

Admin message

Admin message

All machines: All tests - map pfn RAM range req write-combining for [mem 0x.*], got write-back

Activity

New filters associated