Hi Folks!
I'm pretty sure I'm experiencing some variant of the problem described in https://community.nxp.com/t5/i-MX-Processors/i-MX8MQ-LCDIF-FIFO-starved-by-PCIe-traffic/m-p/1229152 except in my case there are some key differences:
- The hardware in question uses eDP instead of LCDIF
- The break only happens when playing a h264 video
- The break persists even after the high PCIe traffic workload is stopped. So once the graphics output disappears it stays broken.
I applied proposed magic code for QoSC adjustment and it works and the problem goes away (hence me believing this is a variant of the linked problem), so what I'm hoping I could get some advise/help with is the following:
- it's not clear to me which IP block is being starved in this scenario. I suspect it is HDMI_TX, but I've no evidence yet. Any good indicators in the register file I could use to confirm this?
- Is there any documentation for that QoSC block that I can look at to better understand the knobs I'm tweaking?
NOTE: The HW in question is https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/arch/arm64/boot/dts/freescale/... and we were running vanilla 5.17 + custom patch delta for experiments, but we've also been able to reproduce the issue using 5.10 NXP kernel on i.MX8MQ EVK board with external HDMI monitor.
