Get it. I'll figure it out how to split up.

Thanks, Zheng

Dmitry Osipenko dmitry.osipenko@collabora.com 于2023年10月24日周二 21:18写道：

On 10/20/23 07:07, Zheng Wang wrote:

...

In mtk_jpeg_probe, &jpeg->job_timeout_work is bound with mtk_jpeg_job_timeout_work.

In mtk_jpeg_dec_device_run, if error happens in mtk_jpeg_set_dec_dst, it will finally start the worker while mark the job as finished by invoking v4l2_m2m_job_finish.

There are two methods to trigger the bug. If we remove the module, it which will call mtk_jpeg_remove to make cleanup. The possible sequence is as follows, which will cause a use-after-free bug.

CPU0 CPU1 mtk_jpeg_dec_... | start worker | |mtk_jpeg_job_timeout_work mtk_jpeg_remove | v4l2_m2m_release | kfree(m2m_dev); | | | v4l2_m2m_get_curr_priv | m2m_dev->curr_ctx //use

If we close the file descriptor, which will call mtk_jpeg_release, it will have a similar sequence.

Fix this bug by start timeout worker only if started jpegdec worker successfully so the v4l2_m2m_job_finish will only be called on either mtk_jpeg_job_timeout_work or mtk_jpeg_dec_device_run.

This patch also reverts commit c677d7ae8314 ("media: mtk-jpeg: Fix use after free bug due to uncanceled work") for this patch also fixed the use-after-free bug mentioned before. Before mtk_jpeg_remove is invoked, mtk_jpeg_release must be invoked to close opened files. And it will call v4l2_m2m_cancel_job to wait for the timeout worker finished so the canceling in mtk_jpeg_remove is unnecessary.

Fixes: b2f0d2724ba4 ("[media] vcodec: mediatek: Add Mediatek JPEG Decoder Driver") Signed-off-by: Zheng Wang zyytlz.wz@163.com Signed-off-by: Dmitry Osipenko dmitry.osipenko@collabora.com Cc: stable@vger.kernel.org

---

.../media/platform/mediatek/jpeg/mtk_jpeg_core.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/drivers/media/platform/mediatek/jpeg/mtk_jpeg_core.c b/drivers/media/platform/mediatek/jpeg/mtk_jpeg_core.c index 7194f88edc0f..c3456c700c07 100644 --- a/drivers/media/platform/mediatek/jpeg/mtk_jpeg_core.c +++ b/drivers/media/platform/mediatek/jpeg/mtk_jpeg_core.c @@ -1021,13 +1021,13 @@ static void mtk_jpeg_dec_device_run(void *priv) if (ret < 0) goto dec_end;

• schedule_delayed_work(&jpeg->job_timeout_work,
  

•                       msecs_to_jiffies(MTK_JPEG_HW_TIMEOUT_MSEC));
  

• mtk_jpeg_set_dec_src(ctx, &src_buf->vb2_buf, &bs);
  if (mtk_jpeg_set_dec_dst(ctx, &jpeg_src_buf->dec_param, &dst_buf->vb2_buf, &fb))
          goto dec_end;
  
  

• schedule_delayed_work(&jpeg->job_timeout_work,
  

•                       msecs_to_jiffies(MTK_JPEG_HW_TIMEOUT_MSEC));
  

• spin_lock_irqsave(&jpeg->hw_lock, flags);
  mtk_jpeg_dec_reset(jpeg->reg_base);
  mtk_jpeg_dec_set_config(jpeg->reg_base,
  

@@ -1403,7 +1403,6 @@ static void mtk_jpeg_remove(struct platform_device *pdev) { struct mtk_jpeg_dev *jpeg = platform_get_drvdata(pdev);

• cancel_delayed_work_sync(&jpeg->job_timeout_work);
  pm_runtime_disable(&pdev->dev);
  video_unregister_device(jpeg->vdev);
  v4l2_m2m_release(jpeg->m2m_dev);
  

@@ -1750,9 +1749,6 @@ static void mtk_jpegdec_worker(struct work_struct *work) v4l2_m2m_src_buf_remove(ctx->fh.m2m_ctx); v4l2_m2m_dst_buf_remove(ctx->fh.m2m_ctx);

• schedule_delayed_work(&comp_jpeg[hw_id]->job_timeout_work,
  

•                       msecs_to_jiffies(MTK_JPEG_HW_TIMEOUT_MSEC));
  

• mtk_jpeg_set_dec_src(ctx, &src_buf->vb2_buf, &bs);
  if (mtk_jpeg_set_dec_dst(ctx,
                           &jpeg_src_buf->dec_param,
  

@@ -1762,6 +1758,9 @@ static void mtk_jpegdec_worker(struct work_struct *work) goto setdst_end; }

• schedule_delayed_work(&comp_jpeg[hw_id]->job_timeout_work,
  

•                       msecs_to_jiffies(MTK_JPEG_HW_TIMEOUT_MSEC));
  

• spin_lock_irqsave(&comp_jpeg[hw_id]->hw_lock, flags);
  ctx->total_frame_num++;
  mtk_jpeg_dec_reset(comp_jpeg[hw_id]->reg_base);
  

What about to split this patch into 3 patches:

1. will remove cancel_delayed_work_sync()
2. will update mtk_jpeg_dec_device_run()
3. will update mtk_jpegdec_worker()

The reason for splitting is because the multi-core mtk_jpegdec_worker() doesn't present in older stable kernels, and thus, the patch isn't backportable as-is.

-- Best regards, Dmitry