* [Patch v3] media: s5p-mfc: Corrected NV12M/NV21M plane-sizes [not found] <CGME20250305055904epcas5p45828c732019a847a0852c26d9ef9f2db@epcas5p4.samsung.com> @ 2025-03-05 5:53 ` Aakarsh Jain [not found] ` <CGME20250305055908epcas5p222e192cc21512ac178543445184a2519@epcas5p2.samsung.com> 0 siblings, 1 reply; 3+ messages in thread From: Aakarsh Jain @ 2025-03-05 5:53 UTC (permalink / raw) To: linux-arm-kernel, linux-media, linux-kernel Cc: m.szyprowski, andrzej.hajda, mchehab, hverkuil-cisco, krzysztof.kozlowski+dt, linux-samsung-soc, gost.dev, aswani.reddy, pankaj.dubey, Aakarsh Jain There is a possibility of getting page fault if the overall buffer size is not aligned to 256bytes. Since MFC does read operation only and it won't corrupt the data values even if it reads the extra bytes. Corrected luma and chroma plane sizes for V4L2_PIX_FMT_NV12M and V4L2_PIX_FMT_NV21M pixel format. Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Reviewed-by: Nicolas Dufresne <nicolas.dufresne@collabora.com> Suggested-by: Hans Verkuil <hverkuil-cisco@xs4all.nl> Signed-off-by: Aakarsh Jain <aakarsh.jain@samsung.com> --- changelog: v2->v3 Patch link:https://patchwork.kernel.org/project/linux-media/patch/20250226102251.9040-1-aakarsh.jain@samsung.com/ Fixed checkpatch warning. v1->v2 Patch link: https://patchwork.kernel.org/project/linux-media/patch/20240806115714.29828-1-aakarsh.jain@samsung.com/ Removed duplicate code and aligned luma and chroma size to multiple of 256bytes as suggested by Hans. drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c index 73f7af674c01..4cf12f33d706 100644 --- a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c +++ b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c @@ -549,8 +549,9 @@ static void s5p_mfc_enc_calc_src_size_v6(struct s5p_mfc_ctx *ctx) case V4L2_PIX_FMT_NV21M: ctx->stride[0] = ALIGN(ctx->img_width, S5P_FIMV_NV12M_HALIGN_V6); ctx->stride[1] = ALIGN(ctx->img_width, S5P_FIMV_NV12M_HALIGN_V6); - ctx->luma_size = ctx->stride[0] * ALIGN(ctx->img_height, 16); - ctx->chroma_size = ctx->stride[0] * ALIGN(ctx->img_height / 2, 16); + ctx->luma_size = ALIGN(ctx->stride[0] * ALIGN(ctx->img_height, 16), 256); + ctx->chroma_size = ALIGN(ctx->stride[0] * + ALIGN(ctx->img_height / 2, 16), 256); break; case V4L2_PIX_FMT_YUV420M: case V4L2_PIX_FMT_YVU420M: -- 2.17.1 ^ permalink raw reply [flat|nested] 3+ messages in thread
[parent not found: <CGME20250305055908epcas5p222e192cc21512ac178543445184a2519@epcas5p2.samsung.com>]
* [Patch v2] media: s5p-mfc: Support for handling RET_ENC_BUFFER_FULL interrupt [not found] ` <CGME20250305055908epcas5p222e192cc21512ac178543445184a2519@epcas5p2.samsung.com> @ 2025-03-05 5:53 ` Aakarsh Jain 2025-03-21 7:35 ` Marek Szyprowski 0 siblings, 1 reply; 3+ messages in thread From: Aakarsh Jain @ 2025-03-05 5:53 UTC (permalink / raw) To: linux-arm-kernel, linux-media, linux-kernel Cc: m.szyprowski, andrzej.hajda, mchehab, hverkuil-cisco, krzysztof.kozlowski+dt, linux-samsung-soc, gost.dev, aswani.reddy, pankaj.dubey, Aakarsh Jain When output encoded buffer size provided by userspace is insufficient with current encoding parameters, it leads to RET_ENC_BUFFER_FULL interrupt which was not handled in IRQ handler. On handling of RET_ENC_BUFFER_FULL interrupt leads to NAL_ABORT command from host to risc which in turn leads to RET_NAL_ABORT interrupt. On receiving RET_NAL_ABORT driver clears workbit and VB2 queues for cleaner closing of MFC instance. When user encounters "Call on DQBUF after unrecoverable error", userspace should close fd and restart with larger output encoder buffer size. Signed-off-by: Aakarsh Jain <aakarsh.jain@samsung.com> --- changelog: v1->v2 Patch link: https://patchwork.kernel.org/project/linux-media/patch/20250228065952.14375-1-aakarsh.jain@samsung.com/ Fixed build errors reported by Media CI robot. .../media/platform/samsung/s5p-mfc/regs-mfc-v6.h | 1 + drivers/media/platform/samsung/s5p-mfc/s5p_mfc.c | 14 ++++++++++++++ .../platform/samsung/s5p-mfc/s5p_mfc_common.h | 1 + .../platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c | 5 +++++ 4 files changed, 21 insertions(+) diff --git a/drivers/media/platform/samsung/s5p-mfc/regs-mfc-v6.h b/drivers/media/platform/samsung/s5p-mfc/regs-mfc-v6.h index fa49fe580e1a..075a58b50b8c 100644 --- a/drivers/media/platform/samsung/s5p-mfc/regs-mfc-v6.h +++ b/drivers/media/platform/samsung/s5p-mfc/regs-mfc-v6.h @@ -45,6 +45,7 @@ #define S5P_FIMV_H2R_CMD_WAKEUP_V6 8 #define S5P_FIMV_CH_LAST_FRAME_V6 9 #define S5P_FIMV_H2R_CMD_FLUSH_V6 10 +#define S5P_FIMV_H2R_CMD_NAL_ABORT_V6 11 /* RMVME: REALLOC used? */ #define S5P_FIMV_CH_FRAME_START_REALLOC_V6 5 diff --git a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc.c b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc.c index 5f80931f056d..e764cab2485c 100644 --- a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc.c +++ b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc.c @@ -739,6 +739,20 @@ static irqreturn_t s5p_mfc_irq(int irq, void *priv) ctx->state = MFCINST_RUNNING; goto irq_cleanup_hw; + case S5P_MFC_R2H_CMD_ENC_BUFFER_FUL_RET: + ctx->state = MFCINST_NAL_ABORT; + s5p_mfc_hw_call(dev->mfc_ops, clear_int_flags, dev); + set_work_bit(ctx); + WARN_ON(test_and_clear_bit(0, &dev->hw_lock) == 0); + s5p_mfc_hw_call(dev->mfc_ops, try_run, dev); + break; + + case S5P_MFC_R2H_CMD_NAL_ABORT_RET: + ctx->state = MFCINST_ERROR; + s5p_mfc_cleanup_queue(&ctx->dst_queue, &ctx->vq_dst); + s5p_mfc_cleanup_queue(&ctx->src_queue, &ctx->vq_src); + goto irq_cleanup_hw; + default: mfc_debug(2, "Unknown int reason\n"); s5p_mfc_hw_call(dev->mfc_ops, clear_int_flags, dev); diff --git a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_common.h b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_common.h index 3cc2a4f5c40a..86c316c1ff8f 100644 --- a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_common.h +++ b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_common.h @@ -141,6 +141,7 @@ enum s5p_mfc_inst_state { MFCINST_RES_CHANGE_INIT, MFCINST_RES_CHANGE_FLUSH, MFCINST_RES_CHANGE_END, + MFCINST_NAL_ABORT, }; /* diff --git a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c index 4cf12f33d706..356adfddcfcf 100644 --- a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c +++ b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c @@ -2229,6 +2229,11 @@ static void s5p_mfc_try_run_v6(struct s5p_mfc_dev *dev) case MFCINST_HEAD_PRODUCED: ret = s5p_mfc_run_init_enc_buffers(ctx); break; + case MFCINST_NAL_ABORT: + mfc_write(dev, ctx->inst_no, S5P_FIMV_INSTANCE_ID_V6); + s5p_mfc_hw_call(dev->mfc_cmds, cmd_host2risc, + dev, S5P_FIMV_H2R_CMD_NAL_ABORT_V6, NULL); + break; default: ret = -EAGAIN; } -- 2.17.1 ^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [Patch v2] media: s5p-mfc: Support for handling RET_ENC_BUFFER_FULL interrupt 2025-03-05 5:53 ` [Patch v2] media: s5p-mfc: Support for handling RET_ENC_BUFFER_FULL interrupt Aakarsh Jain @ 2025-03-21 7:35 ` Marek Szyprowski 0 siblings, 0 replies; 3+ messages in thread From: Marek Szyprowski @ 2025-03-21 7:35 UTC (permalink / raw) To: Aakarsh Jain, linux-arm-kernel, linux-media, linux-kernel Cc: andrzej.hajda, mchehab, hverkuil-cisco, krzysztof.kozlowski+dt, linux-samsung-soc, gost.dev, aswani.reddy, pankaj.dubey On 05.03.2025 06:53, Aakarsh Jain wrote: > When output encoded buffer size provided by userspace > is insufficient with current encoding parameters, it > leads to RET_ENC_BUFFER_FULL interrupt which was not > handled in IRQ handler. > > On handling of RET_ENC_BUFFER_FULL interrupt leads to > NAL_ABORT command from host to risc which in turn leads > to RET_NAL_ABORT interrupt. On receiving RET_NAL_ABORT > driver clears workbit and VB2 queues for cleaner closing > of MFC instance. > > When user encounters "Call on DQBUF after unrecoverable > error", userspace should close fd and restart with larger > output encoder buffer size. > > Signed-off-by: Aakarsh Jain <aakarsh.jain@samsung.com> Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> > --- > changelog: > v1->v2 > Patch link: https://patchwork.kernel.org/project/linux-media/patch/20250228065952.14375-1-aakarsh.jain@samsung.com/ > Fixed build errors reported by Media CI robot. > .../media/platform/samsung/s5p-mfc/regs-mfc-v6.h | 1 + > drivers/media/platform/samsung/s5p-mfc/s5p_mfc.c | 14 ++++++++++++++ > .../platform/samsung/s5p-mfc/s5p_mfc_common.h | 1 + > .../platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c | 5 +++++ > 4 files changed, 21 insertions(+) > > diff --git a/drivers/media/platform/samsung/s5p-mfc/regs-mfc-v6.h b/drivers/media/platform/samsung/s5p-mfc/regs-mfc-v6.h > index fa49fe580e1a..075a58b50b8c 100644 > --- a/drivers/media/platform/samsung/s5p-mfc/regs-mfc-v6.h > +++ b/drivers/media/platform/samsung/s5p-mfc/regs-mfc-v6.h > @@ -45,6 +45,7 @@ > #define S5P_FIMV_H2R_CMD_WAKEUP_V6 8 > #define S5P_FIMV_CH_LAST_FRAME_V6 9 > #define S5P_FIMV_H2R_CMD_FLUSH_V6 10 > +#define S5P_FIMV_H2R_CMD_NAL_ABORT_V6 11 > /* RMVME: REALLOC used? */ > #define S5P_FIMV_CH_FRAME_START_REALLOC_V6 5 > > diff --git a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc.c b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc.c > index 5f80931f056d..e764cab2485c 100644 > --- a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc.c > +++ b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc.c > @@ -739,6 +739,20 @@ static irqreturn_t s5p_mfc_irq(int irq, void *priv) > ctx->state = MFCINST_RUNNING; > goto irq_cleanup_hw; > > + case S5P_MFC_R2H_CMD_ENC_BUFFER_FUL_RET: > + ctx->state = MFCINST_NAL_ABORT; > + s5p_mfc_hw_call(dev->mfc_ops, clear_int_flags, dev); > + set_work_bit(ctx); > + WARN_ON(test_and_clear_bit(0, &dev->hw_lock) == 0); > + s5p_mfc_hw_call(dev->mfc_ops, try_run, dev); > + break; > + > + case S5P_MFC_R2H_CMD_NAL_ABORT_RET: > + ctx->state = MFCINST_ERROR; > + s5p_mfc_cleanup_queue(&ctx->dst_queue, &ctx->vq_dst); > + s5p_mfc_cleanup_queue(&ctx->src_queue, &ctx->vq_src); > + goto irq_cleanup_hw; > + > default: > mfc_debug(2, "Unknown int reason\n"); > s5p_mfc_hw_call(dev->mfc_ops, clear_int_flags, dev); > diff --git a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_common.h b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_common.h > index 3cc2a4f5c40a..86c316c1ff8f 100644 > --- a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_common.h > +++ b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_common.h > @@ -141,6 +141,7 @@ enum s5p_mfc_inst_state { > MFCINST_RES_CHANGE_INIT, > MFCINST_RES_CHANGE_FLUSH, > MFCINST_RES_CHANGE_END, > + MFCINST_NAL_ABORT, > }; > > /* > diff --git a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c > index 4cf12f33d706..356adfddcfcf 100644 > --- a/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c > +++ b/drivers/media/platform/samsung/s5p-mfc/s5p_mfc_opr_v6.c > @@ -2229,6 +2229,11 @@ static void s5p_mfc_try_run_v6(struct s5p_mfc_dev *dev) > case MFCINST_HEAD_PRODUCED: > ret = s5p_mfc_run_init_enc_buffers(ctx); > break; > + case MFCINST_NAL_ABORT: > + mfc_write(dev, ctx->inst_no, S5P_FIMV_INSTANCE_ID_V6); > + s5p_mfc_hw_call(dev->mfc_cmds, cmd_host2risc, > + dev, S5P_FIMV_H2R_CMD_NAL_ABORT_V6, NULL); > + break; > default: > ret = -EAGAIN; > } Best regards -- Marek Szyprowski, PhD Samsung R&D Institute Poland ^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2025-03-21 7:35 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
[not found] <CGME20250305055904epcas5p45828c732019a847a0852c26d9ef9f2db@epcas5p4.samsung.com>
2025-03-05 5:53 ` [Patch v3] media: s5p-mfc: Corrected NV12M/NV21M plane-sizes Aakarsh Jain
[not found] ` <CGME20250305055908epcas5p222e192cc21512ac178543445184a2519@epcas5p2.samsung.com>
2025-03-05 5:53 ` [Patch v2] media: s5p-mfc: Support for handling RET_ENC_BUFFER_FULL interrupt Aakarsh Jain
2025-03-21 7:35 ` Marek Szyprowski
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®