Re: xhci_hcd / ASMedia ASM4242: Bulk-OUT -EPROTO with Logitec 0789:0308 during DVD+RW recording

[wakasio]

Hi Michal,

I looked for the offending TRB in the dump, attached is the full log
(dyndbg_capture4_trb_dump.log, 1061 lines).

The 11 xhci_dbg lines all say "Transfer error for slot 1 ep 3", same
address ffffd140425178e0, resid 23552, each followed by "Soft-reset ep
3". After the 11th one the driver printed "event ring:" and dumped it,
then "endpoint ring:" and dumped that too.

I searched the endpoint ring dump for the DMA address that the event
ring entries point to. I could not find it. The endpoint ring dump
covers two segments, 0xfffc00000 to 0xfffc00ff0 and 0xfffbf0000 to
0xfffbf0ff0, linked back to each other with a Link TRB. The address in
question, 0xfffbf8e0, falls between these two ranges, not inside
either one.

I also noticed the event ring entries pointing to that address are all
labeled "ep 4", not "ep 3". There are 12 of them, all status "USB
Transaction Error", len 23552, same address 0xfffbf8e0:

  ffffc09f0: TRB 00000000fffbf8e0 status 'USB Transaction Error' len 23552 slot 1 ep 4 type 'Transfer Event' flags e:c
  (11 more identical entries, addresses ffffc0a10 through ffffc0b50)

No "Transfer error for slot 1 ep 4" line appears anywhere in the log.
The only ep 4 xhci_dbg lines in the whole capture are plain Success
completions.

I checked the source for the code that calls dump_ring(). The endpoint
ring it dumps is ep_ring, the same variable used in the xhci_dbg call
right above it that prints "ep %u" as ep 3. So the endpoint ring dump
should be for ep 3, which may be why ep 4's address does not show up
in it.

Right before the first ep 4 error entry, there are 5 consecutive ep 4
Success entries at addresses fffbf890, fffbf8a0, fffbf8b0, fffbf8c0,
fffbf8d0, each 0x10 higher than the last, then fffbf8e0 where the
errors start. That part looks like a normal ring progressing forward.

Let me know if you want me to dump ep 4's ring as well, or add
anything else to the patch.

Thanks,
wakasio