On Thu, Feb 18, 2016 at 02:10:56PM +0100, David Engraf wrote:
> Hi Ludovic,
>
> Am 18.02.2016 um 11:48 schrieb Ludovic Desroches:
> >---
> > drivers/dma/at_xdmac.c | 25 ++++++++++++++++++++++---
> > 1 file changed, 22 insertions(+), 3 deletions(-)
> >
> >Hi David,
> >
> >I tried to adpat the fix Cyrille did for at_hdmac. I think it was exactly the
> >same issue. We have to read two registers, one to know the descriptor currently
> >used and one to know the residue for this descriptor. Unfortunately, between
> >the two reads, the descriptor currently used can change.
> >
> >Could you try this draft patch?
>
> It works as well with your patch. I did several runs without loosing any
> bytes or a wrong residue. I've also checked how often the loop needs to
> retry when check_nda and cur_nda is different and in all cases one more read
> was enough. So trying up to 5 times is a good value.
Good news, thanks for testing it. I will send a new version with
comments, commit message, etc.
>
> Tested-by: David Engraf <david.engraf@xxxxxxxxx>
>
> Best regards
> - David
>
>
> >Regards
> >
> >Ludovic
> >
> >diff --git a/drivers/dma/at_xdmac.c b/drivers/dma/at_xdmac.c
> >index 6b512a2..a38356e 100644
> >--- a/drivers/dma/at_xdmac.c
> >+++ b/drivers/dma/at_xdmac.c
> >@@ -1395,8 +1395,8 @@ at_xdmac_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
> > struct at_xdmac_desc *desc, *_desc;
> > struct list_head *descs_list;
> > enum dma_status ret;
> >- int residue;
> >- u32 cur_nda, mask, value;
> >+ int residue, i;
> >+ u32 cur_nda, check_nda, cur_ubc, mask, value;
> > u8 dwidth = 0;
> > unsigned long flags;
> >
> >@@ -1434,6 +1434,25 @@ at_xdmac_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
> > }
> >
> > cur_nda = at_xdmac_chan_read(atchan, AT_XDMAC_CNDA) & 0xfffffffc;
> >+ rmb();
> >+ cur_ubc = at_xdmac_chan_read(atchan, AT_XDMAC_CUBC);
> >+ for (i = 0; i < 5; i++) {
> >+ rmb();
> >+ check_nda = at_xdmac_chan_read(atchan, AT_XDMAC_CNDA) & 0xfffffffc;
> >+
> >+ if (likely (cur_nda == check_nda))
> >+ break;
> >+
> >+ cur_nda = check_nda;
> >+ rmb();
> >+ cur_ubc = at_xdmac_chan_read(atchan, AT_XDMAC_CUBC);
> >+ }
> >+
> >+ if (unlikely(i >= 5)) {
> >+ ret = DMA_ERROR;
> >+ goto spin_unlock;
> >+ }
> >+
> > /*
> > * Remove size of all microblocks already transferred and the current
> > * one. Then add the remaining size to transfer of the current
> >@@ -1446,7 +1465,7 @@ at_xdmac_tx_status(struct dma_chan *chan, dma_cookie_t cookie,
> > if ((desc->lld.mbr_nda & 0xfffffffc) == cur_nda)
> > break;
> > }
> >- residue += at_xdmac_chan_read(atchan, AT_XDMAC_CUBC) << dwidth;
> >+ residue += cur_ubc << dwidth;
> >
> > dma_set_residue(txstate, residue);
> >
> >
>
> --
> To unsubscribe from this list: send the line "unsubscribe dmaengine" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe dmaengine" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
