2017-01-13 13:16 GMT+01:00 Dan Carpenter <dan.carpenter@xxxxxxxxxx>:
> Hello M'boumba Cedric Madianga,
>
> The patch 2b12c5580e0a: "dmaengine: stm32-dma: Fix residue
> computation issue in cyclic mode" from Dec 13, 2016, leads to the
> following static checker warning:
>
> drivers/dma/stm32-dma.c:914 stm32_dma_tx_status()
> error: we previously assumed 'chan->desc' could be null (see line 910)
>
> drivers/dma/stm32-dma.c
> 868 static size_t stm32_dma_desc_residue(struct stm32_dma_chan *chan,
> 869 struct stm32_dma_desc *desc,
> 870 u32 next_sg)
> 871 {
> 872 u32 residue = 0;
> 873 int i;
> 874
> 875 /*
> 876 * In cyclic mode, for the last period, residue = remaining bytes from
> 877 * NDTR
> 878 */
> 879 if (chan->desc->cyclic && next_sg == 0)
>
> Should this just be "desc->cyclic"?
Yes you are right. Thanks
>
> 880 return stm32_dma_get_remaining_bytes(chan);
> 881
> 882 /*
> 883 * For all other periods in cyclic mode, and in sg mode,
> 884 * residue = remaining bytes from NDTR + remaining periods/sg to be
> 885 * transferred
> 886 */
> 887 for (i = next_sg; i < desc->num_sgs; i++)
> 888 residue += desc->sg_req[i].len;
> 889 residue += stm32_dma_get_remaining_bytes(chan);
> 890
> 891 return residue;
> 892 }
>
> [ snip ]
>
> 894 static enum dma_status stm32_dma_tx_status(struct dma_chan *c,
> 895 dma_cookie_t cookie,
> 896 struct dma_tx_state *state)
> 897 {
> 898 struct stm32_dma_chan *chan = to_stm32_dma_chan(c);
> 899 struct virt_dma_desc *vdesc;
> 900 enum dma_status status;
> 901 unsigned long flags;
> 902 u32 residue = 0;
> 903
> 904 status = dma_cookie_status(c, cookie, state);
> 905 if ((status == DMA_COMPLETE) || (!state))
> 906 return status;
> 907
> 908 spin_lock_irqsave(&chan->vchan.lock, flags);
> 909 vdesc = vchan_find_desc(&chan->vchan, cookie);
> 910 if (chan->desc && cookie == chan->desc->vdesc.tx.cookie)
> ^^^^^^^^^^
> Check for NULL
>
> 911 residue = stm32_dma_desc_residue(chan, chan->desc,
> 912 chan->next_sg);
> 913 else if (vdesc)
> 914 residue = stm32_dma_desc_residue(chan,
> ^^^^
> Unchecked dereference.
if vdesc is not NULL, chan->desc could not be NULL because they are
completely linked.
>
> 915 to_stm32_dma_desc(vdesc), 0);
> 916 dma_set_residue(state, residue);
> 917
> 918 spin_unlock_irqrestore(&chan->vchan.lock, flags);
> 919
> 920 return status;
> 921 }
> 922
>
> regards,
> dan carpenter
Could you please tell me how I can check this kind of warning before
submitting a patch ?
Thanks in advance.
regards,
Cedric
--
To unsubscribe from this list: send the line "unsubscribe dmaengine" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
