marfrit/marfrit-packages
0
0
Fork 1
Code Issues 2 Pull Requests Actions Packages Projects Releases Wiki Activity

daedalus-v4l2-dkms: 79256dc -> 6ffe92b — fix kernel panic regression from #67 #69

Merged
marfrit merged 1 commits from claude-noether/marfrit-packages:noether/daedalus-dkms-bump-6ffe92b into main 2026-05-21 12:00:16 +00:00
Conversation 0 Commits 1 Files Changed 3
+35 -10
marfrit commented 2026-05-21 11:57:20 +00:00
Owner
Copy Link
Copy Source

Kernel-only bump. Fixes the hard-reboot regression introduced by daedalus-v4l2 PR #7 (the split-completion design landed via marfrit-packages PR #67).

What broke

On higgs (Pi CM5, 6.18.29+rpt-rpi-2712), the first mpv --hwdec=vaapi-copy playback of a 720p H.264 clip after installing daedalus-v4l2{,-dkms} 0.1.0+r28+g79256dc-1 hard-rebooted the box. No persistent journal on this host, no recoverable trace.

Root cause

device_run previously left both src and dst in m2m_ctx's rdy_queue between pickup and buf_done, relying on the old v4l2_m2m_buf_done_and_job_finish helper to remove them at completion. The new split-completion path (HAS_PIXELS / SRC_CONSUMED separate) calls v4l2_m2m_job_finish on SRC_CONSUMED even when the dst_buf is still parked. job_finish moves the m2m_ctx back to IDLE; the scheduler dispatches the next device_run; v4l2_m2m_next_dst_buf returns the head of the rdy_queue — which is still our parked dst_buf. Two inflight entries on the same vb2_buffer. Later HAS_PIXELS triggers v4l2_m2m_dst_buf_remove_by_buf → list_del on an already-detached list_head → memory corruption, panic.

Fix (daedalus-v4l2#8, merged at 6ffe92b)

device_run now calls v4l2_m2m_src_buf_remove_by_buf + v4l2_m2m_dst_buf_remove_by_buf as soon as it has peeked the buffers and all validation passes, BEFORE adding them to the inflight list. After that, the daemon owns both halves; the m2m scheduler can't re-issue them. Completion path drops the redundant _remove_by_buf calls.

Matches the amphion vdec.c/venc.c pattern.

Wire compat

DAEDALUS_PROTO_VERSION stays at 1. The daemon doesn't need to bump — daedalus-v4l2 0.1.0+r28+g79256dc-1 (already deployed on higgs) is wire-compatible with this new DKMS module. apt install daedalus-v4l2-dkms alone is sufficient.

Deployment on higgs

sudo apt update
sudo systemctl stop daedalus-v4l2  # may already be stopped/masked
sudo apt install daedalus-v4l2-dkms
sudo rmmod daedalus_v4l2
sudo modprobe daedalus_v4l2
sudo systemctl unmask daedalus-v4l2   # if previously masked
sudo systemctl start daedalus-v4l2

Refs

**Kernel-only bump.** Fixes the hard-reboot regression introduced by daedalus-v4l2 PR #7 (the split-completion design landed via marfrit-packages PR #67). ## What broke On higgs (Pi CM5, 6.18.29+rpt-rpi-2712), the first `mpv --hwdec=vaapi-copy` playback of a 720p H.264 clip after installing `daedalus-v4l2{,-dkms} 0.1.0+r28+g79256dc-1` hard-rebooted the box. No persistent journal on this host, no recoverable trace. ## Root cause `device_run` previously left both src and dst in `m2m_ctx`'s rdy_queue between pickup and `buf_done`, relying on the old `v4l2_m2m_buf_done_and_job_finish` helper to remove them at completion. The new split-completion path (`HAS_PIXELS` / `SRC_CONSUMED` separate) calls `v4l2_m2m_job_finish` on `SRC_CONSUMED` even when the dst_buf is still parked. `job_finish` moves the m2m_ctx back to IDLE; the scheduler dispatches the next `device_run`; `v4l2_m2m_next_dst_buf` returns the head of the rdy_queue — which is still our parked dst_buf. Two inflight entries on the same `vb2_buffer`. Later `HAS_PIXELS` triggers `v4l2_m2m_dst_buf_remove_by_buf` → `list_del` on an already-detached `list_head` → memory corruption, panic. ## Fix (daedalus-v4l2#8, merged at 6ffe92b) `device_run` now calls `v4l2_m2m_src_buf_remove_by_buf` + `v4l2_m2m_dst_buf_remove_by_buf` as soon as it has peeked the buffers and all validation passes, BEFORE adding them to the inflight list. After that, the daemon owns both halves; the m2m scheduler can't re-issue them. Completion path drops the redundant `_remove_by_buf` calls. Matches the amphion `vdec.c`/`venc.c` pattern. ## Wire compat `DAEDALUS_PROTO_VERSION` stays at 1. **The daemon doesn't need to bump** — `daedalus-v4l2 0.1.0+r28+g79256dc-1` (already deployed on higgs) is wire-compatible with this new DKMS module. `apt install daedalus-v4l2-dkms` alone is sufficient. ## Deployment on higgs ```sh sudo apt update sudo systemctl stop daedalus-v4l2 # may already be stopped/masked sudo apt install daedalus-v4l2-dkms sudo rmmod daedalus_v4l2 sudo modprobe daedalus_v4l2 sudo systemctl unmask daedalus-v4l2 # if previously masked sudo systemctl start daedalus-v4l2 ``` ## Refs - reauktion/daedalus-v4l2 PR #8: https://git.reauktion.de/reauktion/daedalus-v4l2/pulls/8 (merged) - Predecessor bump (the one that introduced the panic): marfrit-packages PR #67
marfrit added 1 commit 2026-05-21 11:57:21 +00:00
daedalus-v4l2-dkms: 79256dc -> 6ffe92b — fix kernel panic from #7 a0be2dcc9f 
Kernel-only bump.  Fixes the hard-reboot regression introduced by
the daedalus-v4l2#7 split-completion design and observed on higgs
(Pi CM5) during the first mpv vaapi-copy playback of 720p H.264:
device_run now removes src + dst from m2m_ctx's rdy_queue at the
moment it picks them up, not at buf_done time.  Without this, a
parked dst_buf (waiting for libavcodec's display-order release)
stayed in the rdy_queue and got re-picked by the next device_run
after SRC_CONSUMED's job_finish released the scheduler — two
inflight entries on the same vb2_buffer, later HAS_PIXELS calls
list_del on an already-detached list_head, panic.

DAEDALUS_PROTO_VERSION stays at 1 — daemon (userspace
daedalus-v4l2) need NOT bump in lockstep with this DKMS update.
The existing daedalus-v4l2 0.1.0+r28+g79256dc is wire-compatible
with daedalus-v4l2-dkms 0.1.0+r30+g6ffe92b.

Refs:
  * reauktion/daedalus-v4l2#8
marfrit merged commit 1c8c186681 into main 2026-05-21 12:00:16 +00:00
marfrit deleted branch noether/daedalus-dkms-bump-6ffe92b 2026-05-21 12:00:16 +00:00
Sign in to join this conversation.
Reviewers
No Reviewers
Labels
No items
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
marfrit (Markus Fritsche)
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: marfrit/marfrit-packages#69
Delete Branch "claude-noether/marfrit-packages:noether/daedalus-dkms-bump-6ffe92b"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?