Re: [PATCH] dmaengine: xgene-dma: Fix sparse wannings and coccinelle warnings

From: Arnd Bergmann
Date: Tue Apr 14 2015 - 04:43:55 EST


On Tuesday 14 April 2015 12:34:00 Rameshwar Prasad Sahu wrote:
> diff --git a/drivers/dma/xgene-dma.c b/drivers/dma/xgene-dma.c
> index aa61935..59f95db 100755
> --- a/drivers/dma/xgene-dma.c
> +++ b/drivers/dma/xgene-dma.c
> @@ -238,10 +238,10 @@
> dev_err(chan->dev, "%s: " fmt, chan->name, ##arg)
>
> struct xgene_dma_desc_hw {
> - u64 m0;
> - u64 m1;
> - u64 m2;
> - u64 m3;
> + __le64 m0;
> + __le64 m1;
> + __le64 m2;
> + __le64 m3;
> };

This part looks good.

> enum xgene_dma_ring_cfgsize {
> @@ -388,12 +388,12 @@ static bool is_pq_enabled(struct xgene_dma *pdma)
> return !(val & XGENE_DMA_PQ_DISABLE_MASK);
> }
>
> -static void xgene_dma_cpu_to_le64(u64 *desc, int count)
> +static void xgene_dma_cpu_to_le64(struct xgene_dma_desc_hw *desc)
> {
> - int i;
> -
> - for (i = 0; i < count; i++)
> - desc[i] = cpu_to_le64(desc[i]);
> + desc->m0 = cpu_to_le64(((u64 *)desc)[0]);
> + desc->m1 = cpu_to_le64(((u64 *)desc)[1]);
> + desc->m2 = cpu_to_le64(((u64 *)desc)[2]);
> + desc->m3 = cpu_to_le64(((u64 *)desc)[3]);
> }

This part does not: you are circumventing the checks that are supposed
to help you here, and make things harder to read in the process.

> static u16 xgene_dma_encode_len(u32 len)
> @@ -499,9 +499,9 @@ static void xgene_dma_prep_cpy_desc(struct xgene_dma_chan *chan,
>
> skip_additional_src:
> /* Hardware stores descriptor in little endian format */
> - xgene_dma_cpu_to_le64(desc1, 4);
> + xgene_dma_cpu_to_le64(desc1);
> if (desc2)
> - xgene_dma_cpu_to_le64(desc2, 4);
> + xgene_dma_cpu_to_le64(desc2);
> }
>
> static void xgene_dma_prep_xor_desc(struct xgene_dma_chan *chan,
> @@ -540,8 +540,8 @@ static void xgene_dma_prep_xor_desc(struct xgene_dma_chan *chan,
> }
>
> /* Hardware stores descriptor in little endian format */
> - xgene_dma_cpu_to_le64(desc1, 4);
> - xgene_dma_cpu_to_le64(desc2, 4);
> + xgene_dma_cpu_to_le64(desc1);
> + xgene_dma_cpu_to_le64(desc2);
>
> /* Update meta data */
> *nbytes = len;

All these calls should just be removed, and the accesses to the descriptor
get changed to be little-endian. You can use the opportunity to remove
a lot of the macros that make the code harder to understand, and open-code
them like this:

diff --git a/drivers/dma/xgene-dma.c b/drivers/dma/xgene-dma.c
index aa61935ee706..3e3854559ecc 100755
--- a/drivers/dma/xgene-dma.c
+++ b/drivers/dma/xgene-dma.c
@@ -446,12 +446,12 @@ static void *xgene_dma_lookup_ext8(u64 *desc, int idx)
return (idx % 2) ? (desc + idx - 1) : (desc + idx + 1);
}

-static void xgene_dma_init_desc(void *desc, u16 dst_ring_num)
+static void xgene_dma_init_desc(struct xgene_dma_desc_hw *desc, u16 dst_ring_num)
{
- XGENE_DMA_DESC_C_SET(desc); /* Coherent IO */
- XGENE_DMA_DESC_IN_SET(desc);
- XGENE_DMA_DESC_H0ENQ_NUM_SET(desc, dst_ring_num);
- XGENE_DMA_DESC_RTYPE_SET(desc, XGENE_DMA_RING_OWNER_DMA);
+ desc->m1 |= cpu_to_le64(XGENE_DMA_DESC_C_BIT);
+ desc->m0 |= cpu_to_le64(XGENE_DMA_DESC_IN_BIT);
+ desc->m3 |= cpu_to_le64(dst_ring_num << XGENE_DMA_DESC_HOENQ_NUM_POS);
+ desc->m0 |= cpu_to_le64(dst_ring_num << XGENE_DMA_RING_OWNER_DMA);
}


which will store the descriptors in the right format with correct endianess,
make use of the sparse checking and let the reader see what's actually
going on.

Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/