OV2740 intermittently fails after another camera application; frame timeouts and low FPS on ThinkPad X1 Carbon Gen 10

[dtia-sudo]

Description

I am experiencing intermittent problems with the built-in OV2740 camera on a ThinkPad X1 Carbon Gen 10 using Intel IPU6.

Sakari Ailus from linux-media suggested that I report the Intel Camera HAL portion of the problem to this repository.

Hardware

• Laptop: Lenovo ThinkPad X1 Carbon Gen 10
• Camera sensor: OV2740
• Camera interface: Intel IPU6
• IPU6 hardware version: 5
• Sensor: ov2740 17-0036

Software

• OS: CachyOS
• Kernel: 7.1.8-1-cachyos
• libcamera: 0.7.0+ov01a10.2
• Application: qcam
• Intel Camera HAL / libcamhal

Problems

There appear to be two related problems.

1. Camera state-dependent failure

qcam can successfully initialize and stream from the camera when it is the first camera application to access it.

However, after another camera application has accessed the camera, qcam can sometimes fail to open or configure the camera.

One failure produced:

Failed to open camera device 0: -38 (errno=13)
Failed to configure camera

In another failure, qcam successfully opened and configured the camera and started streaming, but then stopped receiving frames:

Camera device 0 opened successfully
Configured HAL stream: 1280x720 NV12
Streaming started for camera 0

followed by repeated:

CamHAL[WAR] <id0>@waitFrame, time out happens, wait recovery

Eventually qcam terminated with SIGABRT.

2. Low frame rate

When the camera does work, 1280x720 video is only approximately 10–11 FPS.

I am not sure whether this is caused by the Camera HAL, IPU6 configuration, sensor configuration, or another part of the pipeline.

Successful qcam run

The same system can successfully initialize and stream the camera.

For example:

Found IPU6 camera 0: ov2740-uf (facing=0)
Adding camera 'ipu6-ov2740-uf-0'
Opening camera device 0
Camera device 0 opened successfully
Configured HAL stream: 1280x720 NV12
Streaming started for camera 0
Streaming stopped for camera 0

This suggests that the camera is supported and can function correctly, but its behavior is not consistent between camera sessions.

Kernel messages during a failed streaming attempt

During one failed streaming attempt, the kernel reported CSI2 errors including:

Payload checksum (CRC) error
Single packet header error corrected
Multiple packet header errors detected
Transfer FIFO overflow
Incomplete long packet detected
Frame sync error
Inter-frame long packet discarded

This was followed by:

stream stop time out
stream close error: 3

Additional information

The media topology shows the OV2740 connected to Intel IPU6 CSI2 1 and the enabled capture path is:

ov2740 17-0036
    -> Intel IPU6 CSI2 1
    -> Intel IPU6 ISYS Capture 8
    -> /dev/video8

The sensor reports:

SGRBG10_1X10/1932x1092

The ISYS capture node reports support for several Bayer and converted formats, including BG10, UYVY, YUYV, RGBP and BGR3.

I initially tested /dev/video8 directly using v4l2-ctl. This resulted in VIDIOC_STREAMON failing with Broken pipe.

I understand from the discussion on linux-media that this direct test does not demonstrate a kernel streaming failure because the media-controller pipeline has not been configured. qcam configures the pipeline through libcamera before streaming.

Question

Could you advise whether the intermittent camera state/failure and the approximately 10–11 FPS performance are known Camera HAL issues?

In particular, I would like to determine:

1. Why the camera can work when qcam is started first but fail after another camera application has accessed it.
2. Whether the CSI2 errors and frame timeouts indicate a Camera HAL/IPU6 configuration problem.
3. Whether the approximately 10–11 FPS at 1280x720 is expected for this configuration or indicates a problem.
4. What additional tests or logs would be useful to diagnose this.

I can provide the complete dmesg, media topology, and qcam logs if needed.

qcam-cachyos-success.txt
dmesg-cachyos-qcam-failure.txt

v4l2-qcam-failure.txt
media-qcam-failure.txt