[v2] iov_iter: don't revert iov buffer if csum error

Message ID d1ef9185-0956-aed9-e44b-84458cf28764@huawei.com
State New
Headers show

Commit Message

Ding Tianhong April 29, 2017, 2:38 a.m.
The patch 327868212381 (make skb_copy_datagram_msg() et.al. preserve
->msg_iter on error) will revert the iov buffer if copy to iter
failed, but it didn't copy any datagram if the skb_checksum_complete
error, so no need to revert any data at this place.

v2: Sabrina notice that return -EFAULT when checksum error is not correct
    here, it would confuse the caller about the return value, so fix it.

Fixes: 327868212381 ("make skb_copy_datagram_msg() et.al. preserve->msg_iter on error")
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>

Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>

---
 net/core/datagram.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

-- 
1.8.3.1

Comments

Al Viro April 29, 2017, 2:46 a.m. | #1
On Sat, Apr 29, 2017 at 10:38:48AM +0800, Ding Tianhong wrote:
> The patch 327868212381 (make skb_copy_datagram_msg() et.al. preserve

> ->msg_iter on error) will revert the iov buffer if copy to iter

> failed, but it didn't copy any datagram if the skb_checksum_complete

> error, so no need to revert any data at this place.


The bug is real, but I would suggest a simpler fix:
                if (__skb_checksum_complete(skb))
                        return -EINVAL;
leaving the rest as-is.
Ding Tianhong April 29, 2017, 9:37 a.m. | #2
On 2017/4/29 10:46, Al Viro wrote:
> On Sat, Apr 29, 2017 at 10:38:48AM +0800, Ding Tianhong wrote:

>> The patch 327868212381 (make skb_copy_datagram_msg() et.al. preserve

>> ->msg_iter on error) will revert the iov buffer if copy to iter

>> failed, but it didn't copy any datagram if the skb_checksum_complete

>> error, so no need to revert any data at this place.

> 

> The bug is real, but I would suggest a simpler fix:

>                 if (__skb_checksum_complete(skb))

>                         return -EINVAL;

> leaving the rest as-is.

> 

Looks good, if so, we don't need the csum_error any more,

-		if (csum_fold(csum))
+
+		if (csum_fold(csum)) {
+			iov_iter_revert(&msg->msg_iter, chunk);
+ 			return -EINVAL;
+		}
+
 		if (unlikely(skb->ip_summed == CHECKSUM_COMPLETE))
 			netdev_rx_csum_fault(skb->dev);
 	}
 	return 0;
- csum_error:
-	iov_iter_revert(&msg->msg_iter, chunk);
- 	return -EINVAL;
 fault:
 	return -EFAULT;

DO you agree this way? :)

Thanks
Ding

> .

>

Patch hide | download patch | download mbox

diff --git a/net/core/datagram.c b/net/core/datagram.c
index f4947e7..0e6a9a9 100644
--- a/net/core/datagram.c
+++ b/net/core/datagram.c
@@ -768,14 +768,17 @@  int skb_copy_and_csum_datagram_msg(struct sk_buff *skb,
 		if (skb_copy_and_csum_datagram(skb, hlen, &msg->msg_iter,
 					       chunk, &csum))
 			goto fault;
-		if (csum_fold(csum))
+
+		if (csum_fold(csum)) {
+			iov_iter_revert(&msg->msg_iter, chunk);
 			goto csum_error;
+		}
+
 		if (unlikely(skb->ip_summed == CHECKSUM_COMPLETE))
 			netdev_rx_csum_fault(skb->dev);
 	}
 	return 0;
 csum_error:
-	iov_iter_revert(&msg->msg_iter, chunk);
 	return -EINVAL;
 fault:
 	return -EFAULT;