| created | 2025-03-23T08:41:25Z |
|---|---|
| begin | 2025-03-21T00:00:00Z |
| end | 2025-03-22T00:00:00Z |
| path | src/sys |
| commits | 4 |
| date | 2025-03-21T13:19:33Z | |||
|---|---|---|---|---|
| author | mpi | |||
| files | src/sys/uvm/uvm_glue.c | log | diff | annotate |
| message |
Make vslock(9) similar to mlock(2): silently ignore non-mapped ranges. Prevents panics triggerable by multi-threaded processes where a thread A munmap(2)s a page being used and wired, by sysctl(2) executed in thread B. This turns vslock(9) into a best-effort mechanism similar to doing mlock(2) on sysctl(2)'s arguments. Note that with this change, thread B might still fault during its copyout(9) if a sibling thread does munmap(2) or munlock(2) multiple times. Tested by and ok bluhm@, ok kettenis@ Reported-by: [email protected] Reported-by: [email protected] Reported-by: [email protected] |
|||
| date | 2025-03-21T13:27:37Z | |||
|---|---|---|---|---|
| author | kirill | |||
| files | src/sys/dev/video.c | log | diff | annotate |
| message |
sys/video: switch to use device_lookup() Using device_lookup() instead of manual access looks ideomatically correct and protecs from calling video_stop() when a device almost gone. OK: mglocker@ mpi@ |
|||
| date | 2025-03-21T13:34:41Z | |||
|---|---|---|---|---|
| author | kirill | |||
| files | src/sys/dev/usb/uvideo.c | log | diff | annotate |
| message |
sys/uvideo: call uvideo_close() from uvideo_detach() It also drop useless usbd_delay_ms() in uvideo_detach() and clean usage of usbd_is_dying() to avoid uvm fault panic on detach of isoc webcam. OK and collaboration: mglocker@ and mpi@ |
|||
| date | 2025-03-21T22:47:08Z | |||
|---|---|---|---|---|
| author | mglocker | |||
| files | src/sys/dev/usb/uvideo.c | log | diff | annotate |
| message |
Remove UVIDEO_DUMP code. Writing frame data in to a file for debugging was required in the very early days, but not anymore. ok kirill@ |
|||