Skip to contentSkip to conversation

s5p-mfc: use MFC_BUF_FLAG_EOS to identify last buffers in decoder capture queue

bug report Dan Carpenter 2 participants 3 messages Apr 4, 2019 – Apr 5, 2019 Media (V4L/DVB)

lore
Share & tools
lore
#0 lore raw

Hello Andrzej Hajda,

The patch 4d0b0ed63660: "[media] s5p-mfc: use MFC_BUF_FLAG_EOS to identify last buffers in decoder capture queue" from Oct 7, 2015, leads to the following static checker warning:

	drivers/media/platform/s5p-mfc/s5p_mfc_dec.c:664 vidioc_dqbuf()
	warn: uncapped user index 'ctx->dst_bufs[buf->index]'

drivers/media/platform/s5p-mfc/s5p_mfc_dec.c

    642 static int vidioc_dqbuf(struct file *file, void *priv, struct v4l2_buffer *buf)
    643 {
    644 	const struct v4l2_event ev = {
    645 		.type = V4L2_EVENT_EOS
    646 	};
    647 	struct s5p_mfc_ctx *ctx = fh_to_ctx(priv);
    648 	int ret;
    649 
    650 	if (ctx->state == MFCINST_ERROR) {
    651 		mfc_err_limited("Call on DQBUF after unrecoverable error\n");
    652 		return -EIO;
    653 	}
    654 
    655 	switch (buf->type) {
    656 	case V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE:
    657 		return vb2_dqbuf(&ctx->vq_src, buf, file->f_flags & O_NONBLOCK);
    658 	case V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE:
    659 		ret = vb2_dqbuf(&ctx->vq_dst, buf, file->f_flags & O_NONBLOCK);
    660 		if (ret)
    661 			return ret;
    662 
    663 		if (ctx->state == MFCINST_FINISHED &&
--> 664 		    (ctx->dst_bufs[buf->index].flags & MFC_BUF_FLAG_EOS))
                                           ^^^^^^^^^^

Smatch is saying that this isn't capped. The truth is that v4l2 code is a bit complicated for Smatch, but in this case I can't see where "buf->index" does get capped. I would have expected it to be capped in check_array_args() where we check "buf->length" but it's not.

I've been going through these warnings really carefully in the past couple weeks trying to fix false positives so let me know what I'm missing and I will update the check. Even if I have to manually muck in the DB.

    665 			v4l2_event_queue_fh(&ctx->fh, &ev);
    666 		return 0;
    667 	default:
    668 		return -EINVAL;
    669 	}
    670 }

regards,
dan carpenter

#1 lore raw

Hi Dan,

Quoted 8 lines · Dan Carpenter

Hello Andrzej Hajda,

The patch 4d0b0ed63660: "[media] s5p-mfc: use MFC_BUF_FLAG_EOS to identify last buffers in decoder capture queue" from Oct 7, 2015, leads to the following static checker warning:

	drivers/media/platform/s5p-mfc/s5p_mfc_dec.c:664 vidioc_dqbuf()
	warn: uncapped user index 'ctx->dst_bufs[buf->index]'

Almost identical e-mail you have sent about year ago, and me and Hans have explained you that it is false positive.

Has something changed?

Regards

Andrzej

Quoted 47 lines

drivers/media/platform/s5p-mfc/s5p_mfc_dec.c

    642 static int vidioc_dqbuf(struct file *file, void *priv, struct v4l2_buffer *buf)
    643 {
    644 	const struct v4l2_event ev = {
    645 		.type = V4L2_EVENT_EOS
    646 	};
    647 	struct s5p_mfc_ctx *ctx = fh_to_ctx(priv);
    648 	int ret;
    649 
    650 	if (ctx->state == MFCINST_ERROR) {
    651 		mfc_err_limited("Call on DQBUF after unrecoverable error\n");
    652 		return -EIO;
    653 	}
    654 
    655 	switch (buf->type) {
    656 	case V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE:
    657 		return vb2_dqbuf(&ctx->vq_src, buf, file->f_flags & O_NONBLOCK);
    658 	case V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE:
    659 		ret = vb2_dqbuf(&ctx->vq_dst, buf, file->f_flags & O_NONBLOCK);
    660 		if (ret)
    661 			return ret;
    662 
    663 		if (ctx->state == MFCINST_FINISHED &&
--> 664 		    (ctx->dst_bufs[buf->index].flags & MFC_BUF_FLAG_EOS))
                                           ^^^^^^^^^^

Smatch is saying that this isn't capped. The truth is that v4l2 code is a bit complicated for Smatch, but in this case I can't see where "buf->index" does get capped. I would have expected it to be capped in check_array_args() where we check "buf->length" but it's not.

I've been going through these warnings really carefully in the past couple weeks trying to fix false positives so let me know what I'm missing and I will update the check. Even if I have to manually muck in the DB.

    665 			v4l2_event_queue_fh(&ctx->fh, &ev);
    666 		return 0;
    667 	default:
    668 		return -EINVAL;
    669 	}
    670 }

regards,
dan carpenter

#2 lore raw
Quoted 9 lines · Andrzej Hajda

Hi Dan,

Quoted 6 lines · Dan Carpenter

Hello Andrzej Hajda,

The patch 4d0b0ed63660: "[media] s5p-mfc: use MFC_BUF_FLAG_EOS to identify last buffers in decoder capture queue" from Oct 7, 2015, leads to the following static checker warning:

Quoted 8 lines
Quoted 2 lines
	drivers/media/platform/s5p-mfc/s5p_mfc_dec.c:664 vidioc_dqbuf()
	warn: uncapped user index 'ctx->dst_bufs[buf->index]'

Almost identical e-mail you have sent about year ago, and me and Hans have explained you that it is false positive.

Has something changed?

Wow! It's remarkable how similar the emails are. Thank you for your patience on this. I am testing a patch which should silence this false positive in Smatch.

regards,
dan carpenter