15:53 <calebccff> vkoul: i spent a while trying to make dma work with the touchscreen on my OnePlus 6 - I'm pretty sure I had it working around a year ago. The rmi4 touch driver fails to send the first command to the device

15:53 <calebccff> this is the diff I'm appllying: https://p.calebs.dev/7962a9

16:28 <vkoul> calebccff: this looks okay to me

16:28 <vkoul> cat /sys/kernel/debug/dmaengine/summary?

16:29 <calebccff> vkoul: https://p.calebs.dev/367139

16:30 <vkoul> looks like channel were not assigned/failed to assign... anything in dmesg

16:30 <vkoul> dmesg |egrep -i "dma|gpi|i2c"

16:35 <calebccff> vkoul: this is what i get https://p.calebs.dev/e1e856, everything seems fine it just can't talk to the touchscreen

16:36 <vkoul> calebccff: [ 21.586941] geni_i2c 890000.i2c: CA:: i2c fifo/se-dma mode. fifo depth:16

16:37 <vkoul> looks like it not using gsi but se dma mode

16:37 <calebccff> oh, hm

16:37 <calebccff> it's from fifo_disable = readl_relaxed(gi2c->se.base + GENI_IF_DISABLE_RO) & FIFO_IF_DISABLE;

16:38 <vkoul> right, that tells us which mode is supported there, gpi or fifo

16:39 <calebccff> is that based on the qup firmware?

16:53 <vkoul> calebccff: that is correct

16:54 <calebccff> vkoul: so downstream didn't use GPI for the touchscreen on my device?

16:55 <vkoul> I guess so, the firmware (qup) for a board controls this...

16:56 <calebccff> huh i see. I looked through downstream DTS and GPI DMA is enabled and on the touchscreen i2c node, so is it a bug on mainline that it doesn't properly fall back to FIFO mode?

16:57 <Mis012[m]> sounds more like hacks on downstream so it doesn't act according to the dts :P