[Bug 2171286] Re: ipu-bridge: HIMX1092 entry without a driver breaks IPU7 camera (ov02c10)

Hi, just want to add more information on this after looking into kernel lore ----------------------------------- The HIMX1092 entry comes from mainline: - commit 4fdb0342f05e "media: ipu-bridge: Add Himax HM1092 IR sensor", authored by Jake Steinman, applied by Sakari Ailus, first released in v7.3-rc1. There is NO HM1092 sensor driver in mainline as of v7.3-rc7. The driver series ("media: i2c: Add HM1092 camera sensor driver", v6, August 2026) is still under review. The posting of the ipu-bridge patch itself said it depended on that driver and "should wait for it". Removing the single HIMX1092 line from ipu_supported_sensors[] restores normal behaviour. Since an upstream driver is unlikely to arrive before the final 7.3 release, consider reverting 4fdb0342f05e (or drop the entry) or wait for it to be available for backporting to 7.3 on Stonking. -- You received this bug notification because you are subscribed to linux in Ubuntu. Matching subscriptions: Bgg, Bmail, Nb https://bugs.launchpad.net/bugs/2171286 Title: ipu-bridge: HIMX1092 entry without a driver breaks IPU7 camera (ov02c10) Status in linux package in Ubuntu: New Bug description: Kernel 7.3's ipu-bridge has a sensor entry for the IR sensor:     IPU_SENSOR_CONFIG("HIMX1092", 2, 180000000, 180480000) but 7.3 ships no driver for the HM1092. ipu-bridge therefore creates a graph endpoint for it that is never bound, and the IPU7 ISYS async notifier waits for it forever. The comment above the table itself says "Do not add an entry for a sensor that is not actually supported." My setup -------- Ubuntu 26.10 on XPS 13 (9350), pre-installed with Ubuntu 24.04 LTS, IPU7 ov02c10 camera. Effect ------ - The RGB sensor (ov02c10 12-0036) binds through Intel CVS's sub-notifier   rather than ISYS's bound() callback, so no /dev/v4l-subdev node is created   for it while the ISYS notifier is incomplete. - Intel's camera HAL (libcamhal-ipu7x) cannot set the sensor's format or   controls, so the picture comes out with wrong colours / exposure. Steps to reproduce ------------------ 1. Boot 7.3.0-9-generic on a Lunar Lake laptop whose ACPI tables list HIMX1092    (e.g. Dell XPS 13 9350). 2. `journalctl -k -b | grep "sensor registration"` → nothing. 3. `sudo cat /sys/kernel/debug/v4l2-async/pending_async_subdevices` → the    HIMX1092 endpoint above is pending. 4. `ls /dev/v4l-subdev*` and check for ov02c10 → no subdev for the sensor. Temporary workaround -------------------- Rebuilding ipu-bridge from the 7.3.0-9.9 source with only the HIMX1092 entry removed (out-of-tree DKMS module): "All sensor registration completed." returns, `v4l-subdev5: ov02c10 12-0036` appears, and the camera's colours are correct. No other change was needed for this part. AI assistance declaration ------------------------- This bug was investigated with the help of an AI assistant. That said, I ran the tests on my own hardware and confirmed the results above (the missing "All sensor registration completed." message and pending HIMX1092 endpoint on the stock kernel, and the camera working correctly with the workaround). To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2171286/+subscriptions