Re: [PATCH 1/3] media: mediatek: vcodec: fix h264 multi statless decoder smatch warning

[Andrzej Pietrasiewicz]

Hi,

W dniu 3.04.2024 o 05:45, Yunfei Dong (董云飞) pisze:
> Hi AngeloGioacchino,
>
> Thanks for your reviewing.
> On Tue, 2024-04-02 at 11:50 +0200, AngeloGioacchino Del Regno wrote:
> > Il 29/02/24 10:56, Yunfei Dong ha scritto:
> > > Fix smatch static checker warning for vdec_h264_req_multi_if.c.
> > > Leading to kernel crash when fb is NULL.
> > >
> > > Fixes: 397edc703a10 ("media: mediatek: vcodec: add h264 decoder")
> > > Signed-off-by: Yunfei Dong <yunfei.dong@xxxxxxxxxxxx>
> > > ---
> > >    .../vcodec/decoder/vdec/vdec_h264_req_multi_if.c         | 9
> > > +++++++--
> > >    1 file changed, 7 insertions(+), 2 deletions(-)
> > >
> > > diff --git
> > > a/drivers/media/platform/mediatek/vcodec/decoder/vdec/vdec_h264_req
> > > _multi_if.c
> > > b/drivers/media/platform/mediatek/vcodec/decoder/vdec/vdec_h264_req
> > > _multi_if.c
> > > index 0e741e0dc8ba..ab8e708e0df1 100644
> > > ---
> > > a/drivers/media/platform/mediatek/vcodec/decoder/vdec/vdec_h264_req
> > > _multi_if.c
> > > +++
> > > b/drivers/media/platform/mediatek/vcodec/decoder/vdec/vdec_h264_req
> > > _multi_if.c
> > > @@ -724,11 +724,16 @@ static int vdec_h264_slice_single_decode(void
> > > *h_vdec, struct mtk_vcodec_mem *bs
> > >    		return vpu_dec_reset(vpu);
> > >    
> > >    	fb = inst->ctx->dev->vdec_pdata->get_cap_buffer(inst->ctx);
> > > +	if (!fb) {
> > > +		mtk_vdec_err(inst->ctx, "fb buffer is NULL");
> > > +		return -EBUSY;
> > > +	}
> > > +
> > >    	src_buf_info = container_of(bs, struct mtk_video_dec_buf,
> > > bs_buffer);
> > >    	dst_buf_info = container_of(fb, struct mtk_video_dec_buf,
> > > frame_buffer);
> > >    
> > > -	y_fb_dma = fb ? (u64)fb->base_y.dma_addr : 0;
> > > -	c_fb_dma = fb ? (u64)fb->base_c.dma_addr : 0;
> >
> > You're changing the behavior here, can you please explain why this
> > change is valid
> > into the commit description?
> The driver already add the condition to check whether fb is NULL at the
> front, no need these two lines again.

Maybe Angelo refers to the function never returning -EBUSY before?
While at it, if fb is a kind of a buffer, why not -ENOMEM
when get_cap_buffer() fails?

Regards,

Andrzej

> > Thanks,
> > Angelo
> Best Regards,
> Yunfei Dong
> > > +	y_fb_dma = (u64)fb->base_y.dma_addr;
> > > +	c_fb_dma = (u64)fb->base_c.dma_addr;
> > >    	mtk_vdec_debug(inst->ctx, "[h264-dec] [%d] y_dma=%llx
> > > c_dma=%llx",
> > >    		       inst->ctx->decoded_frame_cnt, y_fb_dma,
> > > c_fb_dma);
> > >