media: coda: fix deadlock between decoder picture run and start command
authorPhilipp Zabel <p.zabel@pengutronix.de>
Thu, 24 Oct 2019 10:32:11 +0000 (07:32 -0300)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 17 Jan 2020 18:48:57 +0000 (19:48 +0100)
commit a3fd80198de6ab98a205cf7fb148d88e9e1c44bb upstream.

The BIT decoder picture run temporarily locks the bitstream mutex while
the coda device mutex is locked, to refill the bitstream ring buffer.
Consequently, the decoder start command, which locks both mutexes when
flushing the bitstream ring buffer, must lock the coda device mutex
first as well, to avoid an ABBA deadlock.

Fixes: e7fd95849b3c ("media: coda: flush bitstream ring buffer on decoder restart")
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/media/platform/coda/coda-common.c

index 73222c0..834f11f 100644 (file)
@@ -1084,16 +1084,16 @@ static int coda_decoder_cmd(struct file *file, void *fh,
 
        switch (dc->cmd) {
        case V4L2_DEC_CMD_START:
-               mutex_lock(&ctx->bitstream_mutex);
                mutex_lock(&dev->coda_mutex);
+               mutex_lock(&ctx->bitstream_mutex);
                coda_bitstream_flush(ctx);
-               mutex_unlock(&dev->coda_mutex);
                dst_vq = v4l2_m2m_get_vq(ctx->fh.m2m_ctx,
                                         V4L2_BUF_TYPE_VIDEO_CAPTURE);
                vb2_clear_last_buffer_dequeued(dst_vq);
                ctx->bit_stream_param &= ~CODA_BIT_STREAM_END_FLAG;
                coda_fill_bitstream(ctx, NULL);
                mutex_unlock(&ctx->bitstream_mutex);
+               mutex_unlock(&dev->coda_mutex);
                break;
        case V4L2_DEC_CMD_STOP:
                stream_end = false;