Message ID | 572a7d510ace5e5a5ba41c4774d330133291c82a.camel@kernel.crashing.org |
---|---|
State | Accepted |
Delegated to: | David Miller |
Headers | show |
Series | net: ethernet: ftgmac100: Fix DMA coherency issue with SW checksum | expand |
Looks good to me. And I have tested it. It works perfectly fine. Reviewed-by: Vijay Khemka <vijaykhemka@fb.com> Tested-by: Vijay Khemka <vijaykhemka@fb.com> On 10/24/19, 7:47 PM, "Benjamin Herrenschmidt" <benh@kernel.crashing.org> wrote: We are calling the checksum helper after the dma_map_single() call to map the packet. This is incorrect as the checksumming code will touch the packet from the CPU. This means the cache won't be properly flushes (or the bounce buffering will leave us with the unmodified packet to DMA). This moves the calculation of the checksum & vlan tags to before the DMA mapping. This also has the side effect of fixing another bug: If the checksum helper fails, we goto "drop" to drop the packet, which will not unmap the DMA mapping. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Fixes: 05690d633f30 ftgmac100: Upgrade to NETIF_F_HW_CSUM CC: stable@vger.kernel.org [v4.12+] --- drivers/net/ethernet/faraday/ftgmac100.c | 25 ++++++++++++------------ 1 file changed, 12 insertions(+), 13 deletions(-) diff --git a/drivers/net/ethernet/faraday/ftgmac100.c b/drivers/net/ethernet/faraday/ftgmac100.c index 9b7af94a40bb..96e9565f1e08 100644 --- a/drivers/net/ethernet/faraday/ftgmac100.c +++ b/drivers/net/ethernet/faraday/ftgmac100.c @@ -727,6 +727,18 @@ static netdev_tx_t ftgmac100_hard_start_xmit(struct sk_buff *skb, */ nfrags = skb_shinfo(skb)->nr_frags; + /* Setup HW checksumming */ + csum_vlan = 0; + if (skb->ip_summed == CHECKSUM_PARTIAL && + !ftgmac100_prep_tx_csum(skb, &csum_vlan)) + goto drop; + + /* Add VLAN tag */ + if (skb_vlan_tag_present(skb)) { + csum_vlan |= FTGMAC100_TXDES1_INS_VLANTAG; + csum_vlan |= skb_vlan_tag_get(skb) & 0xffff; + } + /* Get header len */ len = skb_headlen(skb); @@ -753,19 +765,6 @@ static netdev_tx_t ftgmac100_hard_start_xmit(struct sk_buff *skb, if (nfrags == 0) f_ctl_stat |= FTGMAC100_TXDES0_LTS; txdes->txdes3 = cpu_to_le32(map); - - /* Setup HW checksumming */ - csum_vlan = 0; - if (skb->ip_summed == CHECKSUM_PARTIAL && - !ftgmac100_prep_tx_csum(skb, &csum_vlan)) - goto drop; - - /* Add VLAN tag */ - if (skb_vlan_tag_present(skb)) { - csum_vlan |= FTGMAC100_TXDES1_INS_VLANTAG; - csum_vlan |= skb_vlan_tag_get(skb) & 0xffff; - } - txdes->txdes1 = cpu_to_le32(csum_vlan); /* Next descriptor */
From: Benjamin Herrenschmidt <benh@kernel.crashing.org> Date: Fri, 25 Oct 2019 13:47:24 +1100 > Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> > Fixes: 05690d633f30 ftgmac100: Upgrade to NETIF_F_HW_CSUM Please put the commit header string inside double quotes and parenthesis (" ") > CC: stable@vger.kernel.org [v4.12+] Do not CC: stable for networking submissions as per the netdev FAQ. All fixed up, and queued up for -stable, thanks Ben.
On Mon, 2019-10-28 at 16:24 -0700, David Miller wrote: > From: Benjamin Herrenschmidt <benh@kernel.crashing.org> > Date: Fri, 25 Oct 2019 13:47:24 +1100 > > > Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> > > Fixes: 05690d633f30 ftgmac100: Upgrade to NETIF_F_HW_CSUM > > Please put the commit header string inside double quotes and > parenthesis > (" ") Will do next time, I had just copy/pasted the git shortlog :-) > > CC: stable@vger.kernel.org [v4.12+] > > Do not CC: stable for networking submissions as per the netdev FAQ. Ah oops, learned. > All fixed up, and queued up for -stable, thanks Ben. Thanks. Cheers, Ben.
diff --git a/drivers/net/ethernet/faraday/ftgmac100.c b/drivers/net/ethernet/faraday/ftgmac100.c index 9b7af94a40bb..96e9565f1e08 100644 --- a/drivers/net/ethernet/faraday/ftgmac100.c +++ b/drivers/net/ethernet/faraday/ftgmac100.c @@ -727,6 +727,18 @@ static netdev_tx_t ftgmac100_hard_start_xmit(struct sk_buff *skb, */ nfrags = skb_shinfo(skb)->nr_frags; + /* Setup HW checksumming */ + csum_vlan = 0; + if (skb->ip_summed == CHECKSUM_PARTIAL && + !ftgmac100_prep_tx_csum(skb, &csum_vlan)) + goto drop; + + /* Add VLAN tag */ + if (skb_vlan_tag_present(skb)) { + csum_vlan |= FTGMAC100_TXDES1_INS_VLANTAG; + csum_vlan |= skb_vlan_tag_get(skb) & 0xffff; + } + /* Get header len */ len = skb_headlen(skb); @@ -753,19 +765,6 @@ static netdev_tx_t ftgmac100_hard_start_xmit(struct sk_buff *skb, if (nfrags == 0) f_ctl_stat |= FTGMAC100_TXDES0_LTS; txdes->txdes3 = cpu_to_le32(map); - - /* Setup HW checksumming */ - csum_vlan = 0; - if (skb->ip_summed == CHECKSUM_PARTIAL && - !ftgmac100_prep_tx_csum(skb, &csum_vlan)) - goto drop; - - /* Add VLAN tag */ - if (skb_vlan_tag_present(skb)) { - csum_vlan |= FTGMAC100_TXDES1_INS_VLANTAG; - csum_vlan |= skb_vlan_tag_get(skb) & 0xffff; - } - txdes->txdes1 = cpu_to_le32(csum_vlan); /* Next descriptor */
We are calling the checksum helper after the dma_map_single() call to map the packet. This is incorrect as the checksumming code will touch the packet from the CPU. This means the cache won't be properly flushes (or the bounce buffering will leave us with the unmodified packet to DMA). This moves the calculation of the checksum & vlan tags to before the DMA mapping. This also has the side effect of fixing another bug: If the checksum helper fails, we goto "drop" to drop the packet, which will not unmap the DMA mapping. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Fixes: 05690d633f30 ftgmac100: Upgrade to NETIF_F_HW_CSUM CC: stable@vger.kernel.org [v4.12+] --- drivers/net/ethernet/faraday/ftgmac100.c | 25 ++++++++++++------------ 1 file changed, 12 insertions(+), 13 deletions(-)