Frank,
As I had mentioned we at TI are working towards enabling the DaVinci GIT more
aggressivey.
I will try to fix the USB DMA issue at the earliest (1-2 weeks).
regards
swami
________________________________
From: Frank Xue [EMAIL PROTECTED]
Sent: Tuesday, May 06, 2008 11:06 PM
To: davinci-linux-open-source
Cc: Subbrathnam, Swaminathan; Michael Gao; will.zheng; Kridner, Jason
Subject: RE: Davinci-EVM USB issue at git tree 2.6.23
Dear Swami,
I am glad to know you are expert of TI. I dug up some threads from this
mailing list.
and found nobody solved this issue about USB DMA mode support on git
2.6.23-davinci.
You mentioned TI is making attempts porting some patches from mvl to GIT tree.
would you like let me know your progress about usb support?
Is there some pending patches available for usb on git 2.6.23-davinc?
>From the last week, i made a conscious effort to trace this issue.
I tested USB performance on PIO mode. The speed benchmark result is:
Write: 1.8MB/s (PIO Mode)
But it's unfortunate that when i tested USB with DMA support. I got following
error message:
usb 1-1: reset high speed USB device using musb_hdrc and address 2
usb 1-1: reset high speed USB device using musb_hdrc and address 2
usb 1-1: reset high speed USB device using musb_hdrc and address 2
usb 1-1: reset high speed USB device using musb_hdrc and address 2
Then i turned on kernel debug option and musb debug option. More detail console
snapshot is attached. Pls have a look.
>From the debug message, we can see
usb_stor_msg_common() function in transport.c sometimes failed.
usb_submit_urb is ok, but wait_for_completion_interruptible_timeout() is
timeout.
I guess there is a race condition between kernel threads.
And i also got following error message, It seems RX endpoint 1 is STALL?
<7>musb_ep_program 651: <-- hw1 urb c3a87ca0 spd3 dev2 ep1in h_addr00 h_port00
bytes 13
<7>musb_host_rx 1420: RX end 1 STALL
<7>cppi_dump_rx 374: RX DMA0/abort: 3 left, csr 2240, 00000000 H831e7820
S831e7800 C831e7800, B832a600d L000d0000 0000000d .. 831e7800
<7>RXBD/last 831e7800: nxt 831e7820 buf 832a6000 off.blen 0000000d opt.plen
d000000d
<7>RXBD/active 831e7820: nxt 00000000 buf 832a6000 off.blen 0000000d opt.plen
e000000d
<7>__musb_giveback 304: complete c3a87ca0 (-32), dev2 ep1in, 0/13
So many errors drives me crazy :(
OK, anyway, I would appreciate any comment regarding this issue.
Thank you.
Regards
/Frank
Lee,
It is good to note that patch45 is delivering the performance
improvements that we intended to deliver.
We are making attempts within TI to enable our customers more on the
GIT tree based development. More concrete steps will follow and I will keep
everyone updated of the same including porting patch 45 to latest DaVinci GIT.
Regards
Swami
_______________________________________________
Davinci-linux-open-source mailing list
[email protected]
http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
