Same issue here, on a Zen 3 / Renoir-family APU (so not limited to
Phoenix/Strix).
Hardware: Lenovo ThinkPad P14s Gen 2 AMD (21A0CTO1WW), Ryzen 5000 "Cezanne",
amdgpu RENOIR 0x1002:0x1638, BIOS R1MET64W (1.34) 04/16/2026 (latest)
OS: Ubuntu 26.04, fresh install (no upgrade)
Kernel: 7.0.0-31-generic (Ubuntu 7.0.0-31.31-generic 7.0.14)
Setup: LUKS + LVM, swap LV used for hibernation, suspend-then-hibernate
Non-default kernel parameters at the time (set for debugging):
initcall_debug ignore_loglevel processor.max_cstate=1 amd_iommu=off
amdgpu.gpu_recovery=1
What happened (single boot, 18.09. - 23.09., several hibernate/resume
cycles):
1. 23.09. 01:23 suspend-then-hibernate enters hibernation. During the freeze
phase ttm_resource_add_bulk_move() hits the WARN at ttm_resource.c:235,
call trace:
ttm_resource_add_bulk_move <- ttm_resource_alloc <- ttm_bo_swapout_cb
<- ttm_lru_walk_for_evict <- ttm_bo_swapout <- ttm_device_swapout
<- ttm_device_prepare_hibernation <- amdgpu_device_evict_resources
<- amdgpu_device_suspend <- amdgpu_pmops_freeze
(journald additionally reports "Missed 2973 kernel messages" there.)
Since this already happens while freezing, the dangling bulk_move cursor
most likely stems from one of the earlier hibernate/resume cycles.
2. Resume at 10:48 succeeds, but the same WARN continues from userspace
GEM activity. 121 x ttm_resource.c:235 WARN in total.
3. From 12:44 on: 257 x "list_del corruption" (lib/list_debug.c:62/65) in
ttm_resource_move_to_lru_tail / ttm_bo_move_to_lru_tail, ttm_tt_populate,
ttm_sys_man_alloc and ttm_resource_fini (ttm_bo_delayed_delete), from
gnome-shell, chrome, thunderbird, ptyxis, firefox, mattermost and kworkers.
4. 19:14 suspend-then-hibernate: two 1h S3 suspend cycles work fine; at 21:14:54
the hibernation entry is the last message. The screen goes black, power LED
and fan keep running for hours; hard power-off needed.
5. Next boot: "PM: Image not found (code -22)" - the hibernation image was never
completed, so the hang most likely occurred during freeze/eviction, not at
power-off.
This matches the analysis in drm/amd#5387 and the upstream fix
"drm/ttm: fix swapped-out resources never leaving their bulk_move range".
Please note that the commit that landed in drm-misc-fixes (3db7d7d58341)
applied the one-line change to the wrong "if" (in ttm_bo_alloc_at_place());
the follow-up "drm/ttm: apply the swapout bulk_move fix to the intended
condition" is required as well. Both are tagged Cc: stable # v7.1+, but the
Ubuntu 7.0 kernel carries the backport of b2ed01e7ad3d and is affected, so
please pick up both commits for the 7.0 SRU.
Attached: kernel log excerpt (all relevant traces, plus timeline).
** Attachment added: "launchpad-2166778-kernel-excerpt.txt"
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2166778/+attachment/6002259/+files/launchpad-2166778-kernel-excerpt.txt
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/2166778
Title:
Ubuntu 26.04 / kernel 7.0.0-31: AMDGPU/TTM list_del corruption after
hibernate/resume causes complete system freeze
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2166778/+subscriptions
--
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
