Re: [PATCH] forcedeth: Fix WARNINGS that result when DMA mappingis not checked

[Larry Finger]

On 12/27/2012 02:05 PM, Eric Dumazet wrote:
> On Thu, 2012-12-27 at 13:42 -0600, Larry Finger wrote:
> > With 3.8-rc1, the first call of pci_map_single() that is not checked
> > with a corresponding pci_dma_mapping_error() call results in a warning
> > with a splat as follows:
> >
> > WARNING: at lib/dma-debug.c:933 check_unmap+0x480/0x950()
> > Hardware name: HP Pavilion dv2700 Notebook PC
> > forcedeth 0000:00:0a.0: DMA-API: device driver failed to check
> >   map error[device address=0x00000000b176e002] [size=90 bytes] [mapped as single]
> >
> > Signed-off-by: Larry Finger <Larry.Finger@xxxxxxxxxxxx>
> > ---
> >   drivers/net/ethernet/nvidia/forcedeth.c |   23 +++++++++++++++++++++++
> >   1 file changed, 23 insertions(+)
> >
> > diff --git a/drivers/net/ethernet/nvidia/forcedeth.c b/drivers/net/ethernet/nvidia/forcedeth.c
> > index 653487d..de39cf2 100644
> > --- a/drivers/net/ethernet/nvidia/forcedeth.c
> > +++ b/drivers/net/ethernet/nvidia/forcedeth.c
> > @@ -1821,6 +1821,11 @@ static int nv_alloc_rx(struct net_device *dev)
> >   							     skb->data,
> >   							     skb_tailroom(skb),
> >   							     PCI_DMA_FROMDEVICE);
> > +			if (pci_dma_mapping_error(np->pci_dev,
> > +						  np->put_rx_ctx->dma)) {
> > +				dev_kfree_skb_any(skb);
>
> skb has no destructor yet, kfree_skb(skb) should be fine

OK.

> > +				goto packet_dropped;
> > +			}
> >   			np->put_rx_ctx->dma_len = skb_tailroom(skb);
> >   			np->put_rx.orig->buf = cpu_to_le32(np->put_rx_ctx->dma);
> >   			wmb();
> > @@ -1830,6 +1835,7 @@ static int nv_alloc_rx(struct net_device *dev)
> >   			if (unlikely(np->put_rx_ctx++ == np->last_rx_ctx))
> >   				np->put_rx_ctx = np->first_rx_ctx;
> >   		} else {
> > +packet_dropped:
> >   			u64_stats_update_begin(&np->swstats_rx_syncp);
> >   			np->stat_rx_dropped++;
> >   			u64_stats_update_end(&np->swstats_rx_syncp);
> > @@ -1856,6 +1862,11 @@ static int nv_alloc_rx_optimized(struct net_device *dev)
> >   							     skb->data,
> >   							     skb_tailroom(skb),
> >   							     PCI_DMA_FROMDEVICE);
> > +			if (pci_dma_mapping_error(np->pci_dev,
> > +						  np->put_rx_ctx->dma)) {
> > +				dev_kfree_skb_any(skb);
> > +				goto packet_dropped;
> > +			}
> >   			np->put_rx_ctx->dma_len = skb_tailroom(skb);
> >   			np->put_rx.ex->bufhigh = cpu_to_le32(dma_high(np->put_rx_ctx->dma));
> >   			np->put_rx.ex->buflow = cpu_to_le32(dma_low(np->put_rx_ctx->dma));
> > @@ -1866,6 +1877,7 @@ static int nv_alloc_rx_optimized(struct net_device *dev)
> >   			if (unlikely(np->put_rx_ctx++ == np->last_rx_ctx))
> >   				np->put_rx_ctx = np->first_rx_ctx;
> >   		} else {
> > +packet_dropped:
> >   			u64_stats_update_begin(&np->swstats_rx_syncp);
> >   			np->stat_rx_dropped++;
> >   			u64_stats_update_end(&np->swstats_rx_syncp);
> > @@ -2217,6 +2229,9 @@ static netdev_tx_t nv_start_xmit(struct sk_buff *skb, struct net_device *dev)
> >   		bcnt = (size > NV_TX2_TSO_MAX_SIZE) ? NV_TX2_TSO_MAX_SIZE : size;
> >   		np->put_tx_ctx->dma = pci_map_single(np->pci_dev, skb->data + offset, bcnt,
> >   						PCI_DMA_TODEVICE);
> > +		if (pci_dma_mapping_error(np->pci_dev,
> > +					  np->put_tx_ctx->dma))
> > +			return NETDEV_TX_BUSY;
>
> Really this is not going to work very well : caller will call this in a
> loop.

Any suggestions on what value should be returned, or does the caller need to be 
modified?

Thanks for the review,

Larry

--
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/