PL330 alignment requirement questions

PL330 alignment requirement questions

[johnlinn at comcast.net]

I'm testing our PL330 with the driver/dma/pl330 driver and see some strangeness that's probably my own setup issue.

I'm also using drivers/dmatest to test it with memory to memory copies. I'm seeing that the copy alignment (copy_align) is not setup (it's 0) such that I see errors on some transfers at the end of the transfer.

If I setup the copy_align to be 3 or larger then I don't see these errors. This number doesn't make much sense to me and I'm digging into the details (seems like 4 would make sense).  I just wanted to see if anyone had tested with dmatest and seen anything similar.

Thanks,
John

PL330 alignment requirement questions

[Rob Herring]

On 07/08/2011 04:56 PM, johnlinn at comcast.net wrote:
> I'm testing our PL330 with the driver/dma/pl330 driver and see some
> strangeness that's probably my own setup issue.
> 
> I'm also using drivers/dmatest to test it with memory to memory
> copies. I'm seeing that the copy alignment (copy_align) is not setup
> (it's 0) such that I see errors on some transfers at the end of the
> transfer.
> 
> If I setup the copy_align to be 3 or larger then I don't see these
> errors. This number doesn't make much sense to me and I'm digging
> into the details (seems like 4 would make sense).  I just wanted to
> see if anyone had tested with dmatest and seen anything similar.
> 
I ran this dmatest with Versatile Express and the example FPGA image
from ARM with the PL330 and also saw errors. You've narrowed it down
farther than I did and I don't have it setup currently.

Rob

PL330 alignment requirement questions

[johnlinn at comcast.net]

> On 07/08/2011 04:56 PM, johnlinn at comcast.net wrote:
> > I'm testing our PL330 with the driver/dma/pl330 driver and see some
> > strangeness that's probably my own setup issue.
> >
> > I'm also using drivers/dmatest to test it with memory to memory
> > copies. I'm seeing that the copy alignment (copy_align) is not setup
> > (it's 0) such that I see errors on some transfers at the end of the
> > transfer.
> >
> > If I setup the copy_align to be 3 or larger then I don't see these
> > errors. This number doesn't make much sense to me and I'm digging
> > into the details (seems like 4 would make sense). I just wanted to
> > see if anyone had tested with dmatest and seen anything similar.
> >
> I ran this dmatest with Versatile Express and the example FPGA image
> from ARM with the PL330 and also saw errors. You've narrowed it down
> farther than I did and I don't have it setup currently.
> 
> Rob

Thanks Rob, appreciate the input.  Let's see if the Samsung guys have any input.

PL330 alignment requirement questions

[Jassi Brar]

On Sat, Jul 9, 2011 at 3:26 AM,  <johnlinn@comcast.net> wrote:
> I'm testing our PL330 with the driver/dma/pl330 driver and see some strangeness that's probably my own setup issue.
>
> I'm also using drivers/dmatest to test it with memory to memory copies. I'm seeing that the copy alignment (copy_align) is not setup (it's 0) such that I see errors on some transfers at the end of the transfer.
>
> If I setup the copy_align to be 3 or larger then I don't see these errors. This number doesn't make much sense to me and I'm digging into the details (seems like 4 would make sense). ?I just wanted to see if anyone had tested with dmatest and seen anything similar.
>
It might help if you dump length, src_addr and dst_addr of failed transfers.
drivers/dma/pl330.c was not rigorously tested because Samsung drivers
use their own DMA API.
Btw, please remember to CC the author.

PL330 alignment requirement questions

[John Linn]

> -----Original Message-----
> From: Jassi Brar [mailto:jassisinghbrar at gmail.com]
> Sent: Friday, July 08, 2011 7:57 PM
> To: johnlinn at comcast.net
> Cc: linux-arm-kernel at lists.infradead.org
> Subject: Re: PL330 alignment requirement questions
> 
> On Sat, Jul 9, 2011 at 3:26 AM,  <johnlinn@comcast.net> wrote:
> > I'm testing our PL330 with the driver/dma/pl330 driver and see some
> strangeness that's probably my own setup issue.
> >
> > I'm also using drivers/dmatest to test it with memory to memory
> copies. I'm seeing that the copy alignment (copy_align) is not setup
> (it's 0) such that I see errors on some transfers at the end of the
> transfer.
> >
> > If I setup the copy_align to be 3 or larger then I don't see these
> errors. This number doesn't make much sense to me and I'm digging into
> the details (seems like 4 would make sense). ?I just wanted to see if
> anyone had tested with dmatest and seen anything similar.
> >
> It might help if you dump length, src_addr and dst_addr of failed
> transfers.
> drivers/dma/pl330.c was not rigorously tested because Samsung drivers
> use their own DMA API.
> Btw, please remember to CC the author.

Thanks Jassi for the input, sounds good on CCing you :).

This email and any attachments are intended for the sole use of the named recipient(s) and contain(s) confidential information that may be proprietary, privileged or copyrighted under applicable law. If you are not the intended recipient, do not read, copy, or forward this email message or any attachments. Delete this email message and any attachments immediately.