00:00 <iscle> could be either of the two, not really sure

00:08 <apritzel> iscle: are there interrupt status bits you could poll to check this?

00:08 <apritzel> and maybe inspect the DMA buffer memory to see if the controller put something in there?

00:12 <iscle> I am printing the irq status register every time there's an usb interrupt (just in case the dma interrupt does not get triggered, but at least I get to read them after another usb interrupt is triggered) and the status is always 0.

00:12 <iscle> I think the issue is that the DMA transfer is never started because I never receive anything in the host side (I'm testing as gadget/peripheral mode)

00:13 <iscle> Is it possible that the BUS_DRQ_SEL has to be configured beforehand for the DMA interrupt to start? Datasheet PDF page 853 (https://linux-sunxi.org/images/c/c1/V853-amp-v853s_datasheet_v1.1.pdf)

00:13 <iscle> It does make sense for me, but I did not see the original allwinner code do this anywhere

00:15 <iscle> I've uploaded my current code here: https://github.com/iscle/linux-v853/tree/v853-usb-dma

00:43 <iscle> nice, finally got an interrupt!

00:48 <apritzel> looks odd (read: embedded) to have this a build time option. Why would one not want this? And who would make this decision? The distro?

00:50 <apritzel> I guess a kernel command line option would be more useful?

00:50 <apritzel> does the original Allwinner code actually use DMA? Or do they just have the code in, but never enable or use DMA?

00:51 <apritzel> Because this is what I remember from previous investigations: that the BSP does not use DMA, and so people were wary it would work at all

00:52 <apritzel> anyway, good job, let's hope it works now (and it's not just a spurious IRQ) ;-)

00:54 <iscle> Not a spurious IRQ: https://pastecode.io/s/0phmszod it actually sets the DMA IRQ status bits for the appropriate channels

00:55 <iscle> Now I just need to figure out how to handle it

00:56 <iscle> And yes, the Allwinner code seems to use DMA, I checked build flags and everything and it should all be connected. That's why I did not know if those drq bits was needed or not

00:58 <iscle> BTW, the code I'm writing does not support all SoCs, according to a comment in Allwinner's code: only SUN8IW5 and later ic support inner dma, former ic(eg. SUN8IW1, SUN8IW3, SUN8IW2 etc) use outer dma.

01:05 <iscle> apritzel: Okay, so, I'm starting to think Allwinner's code does not use DMA, because without the drq bits correctly set, the DMA interrupt is never triggered (I'm guessing because it never starts)

01:05 <iscle> However, with the drq bits being set, it does trigger the irq

01:16 <iscle> Okay, forget what I said xD The IRQ gets triggered when I write something from host to peripheral (remember, DUT is running as peripheral)

01:16 <iscle> But we're a bit closer, I think... Will continue tomorrow

23:48 <iscle> So, I'm about to give the USB DMA driver another go, and I found something which might explain what was happening... I had the DMA transfer direction the other way around...

23:57 <iscle> It sure looks like that was the issue! Nice! :)