On Tue, 10 Apr 2018 21:54:19 +0800
Jia-Ju Bai <baijiaju1990@xxxxxxxxx> wrote:
> dma_tx_fragment() is never called in atomic context.
>
> dma_tx_fragment() is only called by b43legacy_dma_tx(), which is
> only called by b43legacy_tx_work().
> b43legacy_tx_work() is only set a parameter of INIT_WORK() in
> b43legacy_wireless_init().
>
> Despite never getting called from atomic context,
> dma_tx_fragment() calls alloc_skb() with GFP_ATOMIC,
> which does not sleep for allocation.
> GFP_ATOMIC is not necessary and can be replaced with GFP_KERNEL,
> which can sleep and improve the possibility of sucessful allocation.
>
> This is found by a static analysis tool named DCNS written by myself.
> And I also manually check it.
>
> Signed-off-by: Jia-Ju Bai <baijiaju1990@xxxxxxxxx>
> ---
> drivers/net/wireless/broadcom/b43legacy/dma.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/net/wireless/broadcom/b43legacy/dma.c b/drivers/net/wireless/broadcom/b43legacy/dma.c
> index cfa617d..2f0c64c 100644
> --- a/drivers/net/wireless/broadcom/b43legacy/dma.c
> +++ b/drivers/net/wireless/broadcom/b43legacy/dma.c
> @@ -1064,7 +1064,7 @@ static int dma_tx_fragment(struct b43legacy_dmaring *ring,
> meta->dmaaddr = map_descbuffer(ring, skb->data, skb->len, 1);
> /* create a bounce buffer in zone_dma on mapping failure. */
> if (b43legacy_dma_mapping_error(ring, meta->dmaaddr, skb->len, 1)) {
> - bounce_skb = alloc_skb(skb->len, GFP_ATOMIC | GFP_DMA);
> + bounce_skb = alloc_skb(skb->len, GFP_KERNEL | GFP_DMA);
> if (!bounce_skb) {
> ring->current_slot = old_top_slot;
> ring->used_slots = old_used_slots;
Ack.
I think the GFP_ATOMIC came from the days where we did DMA operations
under spinlock instead of mutex.
The same thing can be done in b43.
Also
setup_rx_descbuffer(ring, desc, meta, GFP_ATOMIC)
could be GFP_KERNEL in dma_rx().
This function is called from IRQ thread context.
--
Michael
Attachment:
pgp3pVtYl8fNM.pgp
Description: OpenPGP digital signature
