| Message ID | 20220610125215.240539-6-nicolas.dufresne@collabora.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | media: rkvdec: Fix H.264 error resilience | expand |
Hey Nicolas, Tested with: python3 fluster.py run -d GStreamer-H.264-V4L2SL-Gst1.0 -ts JVT-AVC_ V1 -so /tmp/h264_test.csv -f csv -j1 Ran 129/135 tests successfully in 82.280 secs On 10.06.2022 08:52, Nicolas Dufresne wrote: >There is two way decode errors can occur. In one case, the ready s/There is two way decode/There are two ways decoding/ >status is not set and nothing have been written into the destination, s/nothing have been/nothing has been/ >while in the other case, the buffer is written but may contain a >certain amount of errors. In order to differentiate these, we set >the payload for the first case to 0. > >Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com> Tested-by: Sebastian Fricke <sebastian.fricke@collabora.com> >--- > drivers/staging/media/rkvdec/rkvdec.c | 26 ++++++++++++++++++++++---- > 1 file changed, 22 insertions(+), 4 deletions(-) > >diff --git a/drivers/staging/media/rkvdec/rkvdec.c b/drivers/staging/media/rkvdec/rkvdec.c >index 7e76f8b72885..27f1f7276dd2 100644 >--- a/drivers/staging/media/rkvdec/rkvdec.c >+++ b/drivers/staging/media/rkvdec/rkvdec.c >@@ -954,14 +954,32 @@ static irqreturn_t rkvdec_irq_handler(int irq, void *priv) > enum vb2_buffer_state state; > u32 status; > >+ ctx = v4l2_m2m_get_curr_priv(rkvdec->m2m_dev); > status = readl(rkvdec->regs + RKVDEC_REG_INTERRUPT); >- state = (status & RKVDEC_RDY_STA) ? >- VB2_BUF_STATE_DONE : VB2_BUF_STATE_ERROR; >+ >+ if (!(status & RKVDEC_RDY_STA)) { >+ struct vb2_v4l2_buffer *dst_buf = NULL; >+ >+ if (status & RKVDEC_TIMEOUT_STA) >+ pr_debug("Decoder stopped due to internal timeout."); >+ else >+ pr_debug("Decoder stopped due to internal error."); (Just personal preference.. I would prefer "due to an internal" over "due to internal") >+ >+ /* >+ * When this happens, the buffer is left unmodified. As it >+ * contains no meaningful data we mark is a empty. s/is a empty/it as empty/ The rest looks nice. Thanks. Greetings, Sebastian >+ */ >+ dst_buf = v4l2_m2m_next_dst_buf(ctx->fh.m2m_ctx); >+ vb2_set_plane_payload(&dst_buf->vb2_buf, 0, 0); >+ state = VB2_BUF_STATE_ERROR; >+ } else { >+ state = VB2_BUF_STATE_DONE; >+ } > > writel(0, rkvdec->regs + RKVDEC_REG_INTERRUPT); >- ctx = v4l2_m2m_get_curr_priv(rkvdec->m2m_dev); > >- if (ctx->coded_fmt_desc->ops->check_error_info) >+ if (ctx->coded_fmt_desc->ops->check_error_info && >+ state == VB2_BUF_STATE_DONE) > state = ctx->coded_fmt_desc->ops->check_error_info(ctx); > > if (cancel_delayed_work(&rkvdec->watchdog_work)) >-- >2.36.1 >
diff --git a/drivers/staging/media/rkvdec/rkvdec.c b/drivers/staging/media/rkvdec/rkvdec.c index 7e76f8b72885..27f1f7276dd2 100644 --- a/drivers/staging/media/rkvdec/rkvdec.c +++ b/drivers/staging/media/rkvdec/rkvdec.c @@ -954,14 +954,32 @@ static irqreturn_t rkvdec_irq_handler(int irq, void *priv) enum vb2_buffer_state state; u32 status; + ctx = v4l2_m2m_get_curr_priv(rkvdec->m2m_dev); status = readl(rkvdec->regs + RKVDEC_REG_INTERRUPT); - state = (status & RKVDEC_RDY_STA) ? - VB2_BUF_STATE_DONE : VB2_BUF_STATE_ERROR; + + if (!(status & RKVDEC_RDY_STA)) { + struct vb2_v4l2_buffer *dst_buf = NULL; + + if (status & RKVDEC_TIMEOUT_STA) + pr_debug("Decoder stopped due to internal timeout."); + else + pr_debug("Decoder stopped due to internal error."); + + /* + * When this happens, the buffer is left unmodified. As it + * contains no meaningful data we mark is a empty. + */ + dst_buf = v4l2_m2m_next_dst_buf(ctx->fh.m2m_ctx); + vb2_set_plane_payload(&dst_buf->vb2_buf, 0, 0); + state = VB2_BUF_STATE_ERROR; + } else { + state = VB2_BUF_STATE_DONE; + } writel(0, rkvdec->regs + RKVDEC_REG_INTERRUPT); - ctx = v4l2_m2m_get_curr_priv(rkvdec->m2m_dev); - if (ctx->coded_fmt_desc->ops->check_error_info) + if (ctx->coded_fmt_desc->ops->check_error_info && + state == VB2_BUF_STATE_DONE) state = ctx->coded_fmt_desc->ops->check_error_info(ctx); if (cancel_delayed_work(&rkvdec->watchdog_work))
There is two way decode errors can occur. In one case, the ready status is not set and nothing have been written into the destination, while in the other case, the buffer is written but may contain a certain amount of errors. In order to differentiate these, we set the payload for the first case to 0. Signed-off-by: Nicolas Dufresne <nicolas.dufresne@collabora.com> --- drivers/staging/media/rkvdec/rkvdec.c | 26 ++++++++++++++++++++++---- 1 file changed, 22 insertions(+), 4 deletions(-)