On Thu, 2021-08-05 at 15:22 +0200, Dafna Hirschfeld wrote:
On 30.07.21 04:52, Yong Wu wrote:
MediaTek IOMMU-SMI diagram is like below. all the consumer connect with smi-larb, then connect with smi-common.
M4U | smi-common |
| | ... | | larb1 larb2 | | vdec venc
When the consumer works, it should enable the smi-larb's power which also need enable the smi-common's power firstly.
Thus, First of all, use the device link connect the consumer and the smi-larbs. then add device link between the smi-larb and smi- common.
This patch adds device_link between the consumer and the larbs.
When device_link_add, I add the flag DL_FLAG_STATELESS to avoid calling pm_runtime_xx to keep the original status of clocks. It can avoid two issues:
- Display HW show fastlogo abnormally reported in [1]. At the
beggining, all the clocks are enabled before entering kernel, but the clocks for display HW(always in larb0) will be gated after clk_enable and clk_disable called from device_link_add(->pm_runtime_resume) and rpm_idle. The clock operation happened before display driver probe. At that time, the display HW will be abnormal.
- A deadlock issue reported in [2]. Use DL_FLAG_STATELESS to skip
pm_runtime_xx to avoid the deadlock.
Corresponding, DL_FLAG_AUTOREMOVE_CONSUMER can't be added, then device_link_removed should be added explicitly.
[1] https://lore.kernel.org/linux-mediatek/1564213888.22908.4.camel@mhfsdcap03/ [2] https://lore.kernel.org/patchwork/patch/1086569/
Suggested-by: Tomasz Figa tfiga@chromium.org Signed-off-by: Yong Wu yong.wu@mediatek.com Tested-by: Dafna Hirschfeld dafna.hirschfeld@collabora.com # on mt8173
Hi, unfortunately, I have to take back the Tested-by tag.
sorry for inconvenience you.
(and sorry for reply late, there is something wrong about my local mail server.)
I am now testing the mtk-vcodec with latest kernel + patches sent from the mailing list:
https://gitlab.collabora.com/eballetbo/linux/-/commits/topic/chromeos/chrome...
which includes this patchset.
On chromeos I open a video conference with googl-meet which cause the mtk-vcodec vp8 encoder to run. If I kill it with `killall -9 chrome` I get some page fault messages from the iommu:
Does the "git bisect" point to this patch?
If you don't kill it, Does it also have these error below?
I don't know what happen about "killall -9 chrome', Does it cause freeing some buffer?
[ 837.255952] mtk-iommu 10205000.iommu: fault type=0x5 iova=0xfcff0001 pa=0x0 larb=0 port=0 layer=1 read
This means "larb0 port0" translation fault.
If I am not wrong, you work at mt8173, from [0], this is DISP_OVL0.
May be "killall -9 chrome" free the buffer(iova:0xfcff0000) that DISP_OVL is accessing, then iommu complain it is not a valid iova.
[0] https://elixir.bootlin.com/linux/v5.14-rc1/source/include/dt-bindings/memory...
[ 837.265696] mtk-iommu 10205000.iommu: fault type=0x5 iova=0xfcff0001 pa=0x0 larb=0 port=0 layer=1 read [ 837.282367] mtk-iommu 10205000.iommu: fault type=0x5 iova=0xfcff0001 pa=0x0 larb=0 port=0 layer=1 read [ 837.299028] mtk-iommu 10205000.iommu: fault type=0x5 iova=0xfcff0001 pa=0x0 larb=0 port=0 layer=1 read [ 837.315683] mtk-iommu 10205000.iommu: fault type=0x5 iova=0xfcff0001 pa=0x0 larb=0 port=0 layer=1 read [ 837.332345] mtk-iommu 10205000.iommu: fault type=0x5 iova=0xfcff0001 pa=0x0 larb=0 port=0 layer=1 read [ 837.349004] mtk-iommu 10205000.iommu: fault type=0x5 iova=0xfcff0001 pa=0x0 larb=0 port=0 layer=1 read [ 837.365665] mtk-iommu 10205000.iommu: fault type=0x5 iova=0xfcff0001 pa=0x0 larb=0 port=0 layer=1 read [ 837.382329] mtk-iommu 10205000.iommu: fault type=0x5 iova=0xfcff0001 pa=0x0 larb=0 port=0 layer=1 read [ 837.400002] mtk-iommu 10205000.iommu: fault type=0x5 iova=0xfcff0001 pa=0x0 larb=0 port=0 layer=1 read
In addition, running the encoder tests from the shell:
sudo --user=#1000 /usr/local/libexec/chrome-binary- tests/video_encode_accelerator_tests -- gtest_filter=VideoEncoderTest.FlushAtEndOfStream_Multiple* --codec=vp8 /usr/local/share/tast/data/chromiumos/tast/local/bundles/cros/video/d ata/tulip2-320x180.yuv --disable_validator
At some point it fails with the error
[ 5472.161821] [MTK_V4L2][ERROR] mtk_vcodec_wait_for_done_ctx:32: [290] ctx->type=1, cmd=1, wait_event_interruptible_timeout time=1000ms out 0 0! [ 5472.174678] [MTK_VCODEC][ERROR][290]: vp8_enc_encode_frame() irq_status=0 failed [ 5472.182687] [MTK_V4L2][ERROR] mtk_venc_worker:1239: venc_if_encode failed=-5
+our venc guy Irui.
This looks VENC HW don't start to work. Does this caused by this patchset? this patchset only change the flow of power.
I guess we should check if the power/clock for venc here is enable or not?
If you have any idea of what might be the problem or how to debug?
Thanks, Dafna
--