Message ID | 20200202222854.18409-14-digetx@gmail.com (mailing list archive) |
---|---|
State | Superseded |
Headers | show |
Series | NVIDIA Tegra APB DMA driver fixes and improvements | expand |
On 02/02/2020 22:28, Dmitry Osipenko wrote: > There is no harm in keeping DMA active in the case of error condition, > which should never happen in practice anyways. This will become useful > for the next patch, which will keep RPM enabled only during of DMA > transfer, and thus, it will be much nicer if cyclic DMA handler could > not touch the DMA-enable state. > > Signed-off-by: Dmitry Osipenko <digetx@gmail.com> > --- > drivers/dma/tegra20-apb-dma.c | 9 +++++---- > 1 file changed, 5 insertions(+), 4 deletions(-) > > diff --git a/drivers/dma/tegra20-apb-dma.c b/drivers/dma/tegra20-apb-dma.c > index c7dc27ef1856..50abce608318 100644 > --- a/drivers/dma/tegra20-apb-dma.c > +++ b/drivers/dma/tegra20-apb-dma.c > @@ -571,9 +571,7 @@ static bool handle_continuous_head_request(struct tegra_dma_channel *tdc, > */ > hsgreq = list_first_entry(&tdc->pending_sg_req, typeof(*hsgreq), node); > if (!hsgreq->configured) { > - tegra_dma_stop(tdc); > - dev_err(tdc2dev(tdc), "Error in DMA transfer, aborting DMA\n"); > - tegra_dma_abort_all(tdc); > + dev_err_ratelimited(tdc2dev(tdc), "Error in DMA transfer\n"); While we are at it, a more descriptive error message could be good here. I believe that this condition would indicate a potential underrun condition. > return false; > } > > @@ -772,7 +770,10 @@ static int tegra_dma_terminate_all(struct dma_chan *dc) > if (!list_empty(&tdc->pending_sg_req) && was_busy) { > sgreq = list_first_entry(&tdc->pending_sg_req, typeof(*sgreq), > node); > - sgreq->dma_desc->bytes_transferred += > + dma_desc = sgreq->dma_desc; > + > + if (dma_desc->dma_status != DMA_ERROR) > + dma_desc->bytes_transferred += > get_current_xferred_count(tdc, sgreq, wcount); I am wondering if we need to check this here? I assume that the transfer count would still reflect the amount of data transferred, even if some was dropped. We will never know how much data was lost. Jon
04.02.2020 15:02, Jon Hunter пишет: > > On 02/02/2020 22:28, Dmitry Osipenko wrote: >> There is no harm in keeping DMA active in the case of error condition, >> which should never happen in practice anyways. This will become useful >> for the next patch, which will keep RPM enabled only during of DMA >> transfer, and thus, it will be much nicer if cyclic DMA handler could >> not touch the DMA-enable state. >> >> Signed-off-by: Dmitry Osipenko <digetx@gmail.com> >> --- >> drivers/dma/tegra20-apb-dma.c | 9 +++++---- >> 1 file changed, 5 insertions(+), 4 deletions(-) >> >> diff --git a/drivers/dma/tegra20-apb-dma.c b/drivers/dma/tegra20-apb-dma.c >> index c7dc27ef1856..50abce608318 100644 >> --- a/drivers/dma/tegra20-apb-dma.c >> +++ b/drivers/dma/tegra20-apb-dma.c >> @@ -571,9 +571,7 @@ static bool handle_continuous_head_request(struct tegra_dma_channel *tdc, >> */ >> hsgreq = list_first_entry(&tdc->pending_sg_req, typeof(*hsgreq), node); >> if (!hsgreq->configured) { >> - tegra_dma_stop(tdc); >> - dev_err(tdc2dev(tdc), "Error in DMA transfer, aborting DMA\n"); >> - tegra_dma_abort_all(tdc); >> + dev_err_ratelimited(tdc2dev(tdc), "Error in DMA transfer\n"); > > While we are at it, a more descriptive error message could be good here. > I believe that this condition would indicate a potential underrun condition. Yes, this error indicates the underrun and indeed the error message could be improved. I'll change it in v8. >> return false; >> } >> >> @@ -772,7 +770,10 @@ static int tegra_dma_terminate_all(struct dma_chan *dc) >> if (!list_empty(&tdc->pending_sg_req) && was_busy) { >> sgreq = list_first_entry(&tdc->pending_sg_req, typeof(*sgreq), >> node); >> - sgreq->dma_desc->bytes_transferred += >> + dma_desc = sgreq->dma_desc; >> + >> + if (dma_desc->dma_status != DMA_ERROR) >> + dma_desc->bytes_transferred += >> get_current_xferred_count(tdc, sgreq, wcount); > > I am wondering if we need to check this here? I assume that the transfer > count would still reflect the amount of data transferred, even if some > was dropped. We will never know how much data was lost. I'm wondering too.. stopping DMA in a error case removes this ambiguity and that's why in my previous answer to v6 I suggested to drop this patch. Do you think it's worth to keep this patch?
diff --git a/drivers/dma/tegra20-apb-dma.c b/drivers/dma/tegra20-apb-dma.c index c7dc27ef1856..50abce608318 100644 --- a/drivers/dma/tegra20-apb-dma.c +++ b/drivers/dma/tegra20-apb-dma.c @@ -571,9 +571,7 @@ static bool handle_continuous_head_request(struct tegra_dma_channel *tdc, */ hsgreq = list_first_entry(&tdc->pending_sg_req, typeof(*hsgreq), node); if (!hsgreq->configured) { - tegra_dma_stop(tdc); - dev_err(tdc2dev(tdc), "Error in DMA transfer, aborting DMA\n"); - tegra_dma_abort_all(tdc); + dev_err_ratelimited(tdc2dev(tdc), "Error in DMA transfer\n"); return false; } @@ -772,7 +770,10 @@ static int tegra_dma_terminate_all(struct dma_chan *dc) if (!list_empty(&tdc->pending_sg_req) && was_busy) { sgreq = list_first_entry(&tdc->pending_sg_req, typeof(*sgreq), node); - sgreq->dma_desc->bytes_transferred += + dma_desc = sgreq->dma_desc; + + if (dma_desc->dma_status != DMA_ERROR) + dma_desc->bytes_transferred += get_current_xferred_count(tdc, sgreq, wcount); } tegra_dma_resume(tdc);
There is no harm in keeping DMA active in the case of error condition, which should never happen in practice anyways. This will become useful for the next patch, which will keep RPM enabled only during of DMA transfer, and thus, it will be much nicer if cyclic DMA handler could not touch the DMA-enable state. Signed-off-by: Dmitry Osipenko <digetx@gmail.com> --- drivers/dma/tegra20-apb-dma.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)