6by9

Yes, it's at https://github.com/raspberrypi/linux/blob/rpi-6.6.y/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_core.c#L3227-L3240

So switching to `idr_alloc_cyclic` does give me a number of log messages of "VC released a buffer that is already released" (technically it's that we can't match the kernel_id value...

Something is certainly going wrong and stopping messages being handled. I have the VPU stuck waiting to send MSG_TYPE_RELEASED message, which is blocking the thread that processes any of the...

It's something in the handling of VCHI. It's most likely that "slots" aren't getting freed under some conditions, and we end up with none available. I'll be talking to pelwell...

I'm now either getting lead astray due to incorrect debug output, or weirder things are going on. The VPU side is implying that with more than 4 threads running we...

So at least some of the leaks are due to interrupting the VCHIQ call. `vc_sm_cma_ioctl_alloc` has called `vc_sm_cma_vchiq_import` to call `vc_sm_cma_vchiq_send_msg`, and the message has been sent to the VPU....

No typo. https://github.com/raspberrypi/linux/blob/rpi-6.6.y/arch/arm/boot/dts/overlays/pciex1-compat-pi5-overlay.dts It was only merged 3 weeks ago, so you can either use `sudo rpi-update` (normal warnings apply over potential regressions), or download just that one overlay with...

It would have been useful for you to have put in the PR description that I had asked you to create this for test purposes in https://forums.raspberrypi.com/viewtopic.php?t=389125 I'll give it...

Tested against the analyser. Images are fine. Timings are reported as: - link freq of 500Msps. - horizontally line time 12.57us, Hactive 800. Hsync 16 pixels / 0.26us. HBP 0.96usc...

Having looked again at your image from the forum post, please test with kmstest rather than the desktop. The desktop could have rendering errors in what is presented to the...