答复: 答复: [PATCH 4/5] media: mtk-vcodec: Add two error cases upon vpu irq handling
Irui Wang
irui.wang at mediatek.com
Wed Nov 3 18:38:48 PDT 2021
Hi,
On Wed, 2021-11-03 at 22:50 +0200, Dafna Hirschfeld wrote:
>
> On 09.08.21 12:12, Irui Wang (王瑞) wrote:
> > Hi Dafna,
> >
> > > > 2. Always happened issue ? timeout at the beginning or in
> > > > processing ?
> > > The commands that I run is:
> > > > sudo --user=#1000 /usr/local/libexec/chrome-binary-
> > > > tests/video_encode_accelerator_tests --gtest_filter=-
> > > > *NV12Dmabuf* -->>codec=vp8
> > > > >/usr/local/share/tast/data/chromiumos/tast/local/bundles/cros/
> > > > video/data/tulip2-320x180.yuv --disable_validator
> > > > The command sometime succeed but when I run it sequentially
> > > > then at some point after few attempts I start to get those
> > > > timeout errors.
> >
> > It seems mean VP8 encoding function OK, but failed sometimes, did
> > you have check VENC clock info during encoding:
> >
> > cat /sys/kernel/debug/clk/clk_summary | grep venc:
> >
> > venc_sel > it's H.264 clock
> > venclt_sel > it's VP8 clock
> >
> > the enable&prepare count is not 0 during encoding process.
> >
> > Thanks
> > Best Regards
> >
> > -----邮件原件-----
> > 发件人: Dafna Hirschfeld [mailto:dafna.hirschfeld at collabora.com]
> > 发送时间: 2021年8月9日 16:27
> > 收件人: Irui Wang (王瑞)
> > 抄送: Linux Media Mailing List; moderated list:ARM/Mediatek SoC
> > support; Enric Balletbo i Serra
> > 主题: Re: 答复: [PATCH 4/5] media: mtk-vcodec: Add two error cases upon
> > vpu irq handling
> >
> >
> >
> > On 09.08.21 09:37, Irui Wang (王瑞) wrote:
> > > Hi Dafna,
> > >
> > > > > I am testing the vp8 encoder on chromeos and at some point
> > > > > the encoder interrupts stop arriving so I try to figure out
> > > > > why and report any possible error.
> > >
> > > 1. Log shows wait IRQ timeout ?
> >
> > Hi, yes, I get timeout when waiting to the encoder interrupt. The
> > timeout is on vp8_enc_wait_venc_done
> >
> >
> > > 2. Always happened issue ? timeout at the beginning or in
> > > processing ?
> >
> > The commands that I run is:
> >
> > sudo --user=#1000 /usr/local/libexec/chrome-binary-
> > tests/video_encode_accelerator_tests --gtest_filter=-
> > *NV12Dmabuf* --codec=vp8
> > /usr/local/share/tast/data/chromiumos/tast/local/bundles/cros/video
> > /data/tulip2-320x180.yuv --disable_validator
> >
> > The command sometime succeed but when I run it sequentially then at
> > some point after few attempts I start to get those timeout errors.
> >
> > > 3. how about IRQ infos?
> > > cat /proc/interrupts | grep vcodec
> > > 18002000.vcodec >> it's H.264 encoder
> > > 19002000.vcodec >> it's VP8 encoder
> > >
> > > I was told you have met another H.264 encoding failed before,
> > > did you find reasons about that ?
> >
> > No,
> > But since I see that the google meetup uses the vp8 encoder I
> > decided to test the vp8 first.
> >
> > > > [ 81.918747] [MTK_V4L2][ERROR]
> > > > mtk_vcodec_wait_for_done_ctx:32: [3] ctx->type=1, cmd=1,
> > > > wait_event_interruptible_timeout time=1000ms out 0 0!
> > > > [ 81.931392] [MTK_VCODEC][ERROR][3]: h264_encode_frame()
> > > > irq_status=0 failed
> > > > [ 81.938470] [MTK_V4L2][ERROR] mtk_venc_worker:1219:
> > > > venc_if_encode failed=-5
>
> Hi,
> I test the driver on debian now, I stream several instances of vp8
> and h264 in parallel.
> I see that those errors always occur after an iommu page fault:
>
> [gst-master] root at debian:~/gst-build# [ 5743.206014] mtk-iommu
> 10205000.iommu: fault type=0x5 iova=0xac2ff003 pa=0x0 larb=3 port=0
> layer=1 write
> [ 5744.204964] [MTK_V4L2][ERROR] mtk_vcodec_wait_for_done_ctx:32:
> [2706] ctx->type=1, cmd=1, wait_event_interruptible_timeout
> time=1000ms out 0 0!
> [ 5744.217849] [MTK_VCODEC][ERROR][2706]: vp8_enc_encode_frame()
> irq_status=0 failed
> [ 5744.225359] [MTK_V4L2][ERROR] mtk_venc_worker:1243: venc_if_encode
> failed=-5
>
> I suspect that maybe this is because the iova of the working_buffers
> is handed to the vpu in the function vp8_enc_alloc_work_buf:
> wb[i].iova = inst->work_bufs[i].dma_addr;
>
> Maybe the vpu keep writing to those addresses after they are already
> unmapped?
>
> Thanks,
> Dafna
failed larb=3, it's for h264 encoder, vp8 encoder is larb5, is there
any h264 encoding failed at that time? I have no ideas why these errors
happened on your platform, but when we got the "iommu: fault", the
possible reason is related to power or clock.
>
>
> >
> >
> > >
> > > MT8173 latest VPUD firmware:
> > >
https://git.kernel.org/pub/scm/linux/kernel/git/firmware/linux-firmwar
> > > e.git/commit/?id=aaed4a8bf9a77ec68376e8d92fb218d5fdd88b59
> > >
> >
> > I uses the latest firmware.
> >
> > Thanks,
> > Dafna
> >
> > > Thanks
> > > Best Regards
> > >
> > > -----邮件原件-----
> > > 发件人: Dafna Hirschfeld [mailto:dafna.hirschfeld at collabora.com]
> > > 发送时间: 2021年8月6日 15:49
> > > 收件人: Irui Wang (王瑞); linux-kernel at vger.kernel.org;
> > > linux-media at vger.kernel.org; linux-mediatek at lists.infradead.org
> > > 抄送: dafna3 at gmail.com; tfiga at chromium.org; Tiffany Lin (林慧珊);
> > > eizan at chromium.org; Maoguang Meng (孟毛广); kernel at collabora.com;
> > > mchehab at kernel.org; hverkuil at xs4all.nl; Yunfei Dong (董云飞); Yong
> > > Wu
> > > (吴勇); hsinyi at chromium.org; matthias.bgg at gmail.com; Andrew-CT Chen
> > > (陳智迪); acourbot at chromium.org
> > > 主题: Re: [PATCH 4/5] media: mtk-vcodec: Add two error cases upon
> > > vpu
> > > irq handling
> > >
> > >
> > >
> > > On 06.08.21 08:58, Irui Wang (王瑞) wrote:
> > > > On Wed, 2021-08-04 at 16:27 +0200, Dafna Hirschfeld wrote:
> > > > > 1. Fail if the function mtk_vcodec_fw_map_dm_addr returns ERR
> > > > > pointer.
> > > > > 2. Fail if the state from the vpu msg is either
> > > > > VEN_IPI_MSG_ENC_STATE_ERROR or VEN_IPI_MSG_ENC_STATE_PART
> > > > >
> > > > > Signed-off-by: Dafna Hirschfeld <
> > > > > dafna.hirschfeld at collabora.com>
> > > > > ---
> > > > > drivers/media/platform/mtk-vcodec/venc_vpu_if.c | 8
> > > > > ++++++++
> > > > > 1 file changed, 8 insertions(+)
> > > > >
> > > > > diff --git a/drivers/media/platform/mtk-vcodec/venc_vpu_if.c
> > > > > b/drivers/media/platform/mtk-vcodec/venc_vpu_if.c
> > > > > index 32dc844d16f9..234705ba7cd6 100644
> > > > > --- a/drivers/media/platform/mtk-vcodec/venc_vpu_if.c
> > > > > +++ b/drivers/media/platform/mtk-vcodec/venc_vpu_if.c
> > > > > @@ -17,6 +17,8 @@ static int handle_enc_init_msg(struct
> > > > > venc_vpu_inst *vpu, const void *data)
> > > > > vpu->vsi = mtk_vcodec_fw_map_dm_addr(vpu->ctx->dev-
> > > > > >fw_handler,
> > > > > msg->vpu_inst_addr);
> > > > >
> > > > > +if (IS_ERR(vpu->vsi))
> > > > > +return PTR_ERR(vpu->vsi);
> > > > > /* Firmware version field value is unspecified on MT8173.
> > > > > */
> > > > > if (vpu->ctx->dev->venc_pdata->chip == MTK_MT8173)
> > > > > return 0;
> > > > > @@ -42,6 +44,12 @@ static int handle_enc_encode_msg(struct
> > > > > venc_vpu_inst *vpu, const void *data)
> > > > > vpu->state = msg->state;
> > > > > vpu->bs_size = msg->bs_size;
> > > > > vpu->is_key_frm = msg->is_key_frm;
> > > > > +if (vpu->state == VEN_IPI_MSG_ENC_STATE_ERROR ||
> > > > > + vpu->state == VEN_IPI_MSG_ENC_STATE_PART) {
> > > > > mtk_vcodec_err(vpu,
> > > > > +"bad ipi-enc-state: %s",
> > > > > + vpu->state ==
> > > > > VEN_IPI_MSG_ENC_STATE_ERROR ? "ERR" : "PART");
> > > > > +return -EINVAL;
> > > > > +}
> > > >
> > > > Hi Dafna,
> > > >
> > > > This state check is useless, the enc result will check in
> > > > "vpu_enc_ipi_handler".
> > > >
> > >
> > > Hi, thanks for reviewing. I see that the vpu_enc_ipi_handler only
> > > test the msg->status and I see that the states are not tested
> > > anywhere except of "skip" state in the h264 enc.
> > >
> > > Can't there be a scenario where msg->status is ok but the state
> > > is error?
> > > I am testing the vp8 encoder on chromeos and at some point the
> > > encoder interrupts stop arriving so I try to figure out why and
> > > report any possible error.
> > >
> > > Thanks,
> > > Dafna
> > >
> > > > Thanks
> > > >
> > > > > return 0;
> > > > > }
> > > > >
> > >
> > > ************* MEDIATEK Confidentiality Notice
> > > ******************** The
> > > information contained in this e-mail message (including any
> > > attachments) may be confidential, proprietary, privileged, or
> > > otherwise exempt from disclosure under applicable laws. It is
> > > intended
> > > to be conveyed only to the designated recipient(s). Any use,
> > > dissemination, distribution, printing, retaining or copying of
> > > this
> > > e-mail (including its
> > > attachments) by unintended recipient(s) is strictly prohibited
> > > and may
> > > be unlawful. If you are not an intended recipient of this e-mail,
> > > or
> > > believe that you have received this e-mail in error, please
> > > notify the
> > > sender immediately (by replying to this e-mail), delete any and
> > > all
> > > copies of this e-mail (including any attachments) from your
> > > system,
> > > and do not disclose the content of this e-mail to any other
> > > person. Thank you!
> > >
More information about the Linux-mediatek
mailing list