Re: [PATCH 0/6] get rid of on-stack dma buffers

From: Oliver Neukum
Date: Tue Mar 22 2011 - 09:12:16 EST

Am Dienstag, 22. März 2011, 14:08:17 schrieb Florian Mickler:
> Am 22.03.2011 12:10 schrieb "Roedel, Joerg" <Joerg.Roedel@xxxxxxx>:
> >
> > On Mon, Mar 21, 2011 at 05:03:15PM -0400, Florian Mickler wrote:
> > > I guess (not verified), that the dma api takes sufficient precautions
> > > to abort the dma transfer if a timeout happens. So freeing _should_
> > > not be an issue. (At least, I would expect big fat warnings everywhere
> > > if that were the case)
> >
> > Freeing is very well an issue. All you can expect from the DMA-API is to
> > give you a valid DMA handle for your device. But it can not prevent that
> > a device uses this handle after you returned it. You need to make sure
> > yourself that any pending DMA is canceled before calling kfree().
> Does usb_control_msg do this? It waits for completion but takes also a
> timeout parameter. I will recheck this once I'm home.

It uses usb_start_wait_urb() which upon a timeout kills the URB. The
buffer is unused after usb_control_msg() returns.

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at