> -----Original Message-----
> From: Petko Manolov [mailto:petkan@spct.net]
> Sent: Thursday, December 16, 1999 9:04 AM
> To: Dunlap, Randy
> Cc: linux-kernel@vger.rutgers.edu
> Subject: Re: strange usb behavior
>
> "Dunlap, Randy" wrote:
> >
> > This says that you did a control msg read/INput,
> > requesting 4 bytes, and 0 bytes were actually read
> > from the device (i.e., the device sent 0 bytes, not 4).
>
> I can guess this, but is not true because registers i
> modify changed their values.
So are you saying that you do a control msg read/INput
to modify device registers? Or could it be that
you did a control msg write/OUTput to modify registers
and then did a control msg read/INput to read those
registers, and we are not seeing the entire log here?
>
> > I don't see how this would have anything to do with
> > the registers being modified. Wouldn't that usually
> > be done with a control pipe write/OUTput message?
>
> I use vendor specific call that to change MII and
> device specific registers. So the change succeed,
> the values are set correctly (the read show so), but
> warning appears.
So is the vendor-specific [control msg] a write/Output
or a read/Input?
>
> > usb_bulk_msg() is a synchronous I/O function. It
> > waits until I/O complete (or timeout) to return to
> > your driver. usb_request_bulk() is an async I/O
> > request. Your driver will continue execution
> > immediately after the I/O to your device has been
> > setup. For any decent performance, I expect that
> > you would want to use usb_request_bulk().
>
> Hm, that looks ok, but kernel oopses bad blaming me
> that do request_bulk in interrupt. Actually i use
> request_bulk() from my start_xmit() routine. Speaking
> of which i have always thought that start_xmit() is
> the top half of the driver and executes in process
> context. Is that true?
You certainly couldn't do a usb_bulk_msg() from interrupt
context since it calls schedule_timeout(). However,
I don't know of any reason that you can't call
usb_request_bulk() from interrupt context.
I haven't written a network driver for Linux (I have for
NetWare, DOS, and some Windows). In all of those, I
agree with you. start_xmit() would be in process
context, not interrupt context.
~Randy
>
> > BTW, linux-usb has a web site (www.linux-usb.org) and
> > a mailing list (linux-usb@suse.com).
>
> ;-)) Thanks, this simplifies my job very much.
>
> Best regards,
> Petkan
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/