[PATCH] net: fix tp->undo_retrans accounting in tcp_sacktag_one()

Commit a71d77e6be1e ("tcp: fix segment accounting when DSACK range covers
multiple segments") fix some DSACK accounting for multiple segments.
In tcp_sacktag_one(), we should also use the actual DSACK rang(pcount)
for tp->undo_retrans accounting.

Signed-off-by: zhenggy <redacted>
---
 net/ipv4/tcp_input.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 3f7bd7a..141e85e 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c

@@ -1346,7 +1346,7 @@ static u8 tcp_sacktag_one(struct sock *sk,
 	if (dup_sack && (sacked & TCPCB_RETRANS)) {
 		if (tp->undo_marker && tp->undo_retrans > 0 &&
 		    after(end_seq, tp->undo_marker))
-			tp->undo_retrans--;
+			tp->undo_retrans = max_t(int, 0, tp->undo_retrans - pcount);
 		if ((sacked & TCPCB_SACKED_ACKED) &&
 		    before(start_seq, state->reord))
 				state->reord = start_seq;

-- 
1.8.3.1

On Thu, Sep 9, 2021 at 6:34 AM zhenggy [off-list ref] wrote:

Thanks for the fix!

I think it would be useful to have a Fixes: footer to help maintainers
know how far back to backport the fix.

I think an appropriate fixes footer might be the following:

Fixes: 10d3be569243 ("tcp-tso: do not split TSO packets at retransmit time")

Before that commit, the assumption underlying the tp->undo_retrans--
seems correct, AFAICT.

thanks,
neal

On Thu, Sep 9, 2021 at 9:38 AM Neal Cardwell [off-list ref] wrote:

On Thu, Sep 9, 2021 at 6:34 AM zhenggy [off-list ref] wrote:

quoted

Commit a71d77e6be1e ("tcp: fix segment accounting when DSACK range covers
multiple segments") fix some DSACK accounting for multiple segments.
In tcp_sacktag_one(), we should also use the actual DSACK rang(pcount)
for tp->undo_retrans accounting.

Signed-off-by: zhenggy <redacted>

Another nit: in the commit title, rather than "net":

  net: fix tp->undo_retrans accounting in tcp_sacktag_one()

...I would suggest the more specific "tcp", which is more typical for
commits fixing tcp*.c files:

  tcp: fix tp->undo_retrans accounting in tcp_sacktag_one()

thanks,
neal

Hi，Neal：

Thanks very much for your checking, I will send a v2 patch to make it.

在 2021/9/9 21:38, Neal Cardwell 写道:

On Thu, Sep 9, 2021 at 6:34 AM zhenggy [off-list ref] wrote:

quoted

Commit a71d77e6be1e ("tcp: fix segment accounting when DSACK range covers
multiple segments") fix some DSACK accounting for multiple segments.
In tcp_sacktag_one(), we should also use the actual DSACK rang(pcount)
for tp->undo_retrans accounting.

Signed-off-by: zhenggy <redacted>
---
 net/ipv4/tcp_input.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index 3f7bd7a..141e85e 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c

@@ -1346,7 +1346,7 @@ static u8 tcp_sacktag_one(struct sock *sk,
        if (dup_sack && (sacked & TCPCB_RETRANS)) {
                if (tp->undo_marker && tp->undo_retrans > 0 &&
                    after(end_seq, tp->undo_marker))
-                       tp->undo_retrans--;
+                       tp->undo_retrans = max_t(int, 0, tp->undo_retrans - pcount);
                if ((sacked & TCPCB_SACKED_ACKED) &&
                    before(start_seq, state->reord))
                                state->reord = start_seq;
--

Thanks for the fix!

I think it would be useful to have a Fixes: footer to help maintainers
know how far back to backport the fix.

I think an appropriate fixes footer might be the following:

Fixes: 10d3be569243 ("tcp-tso: do not split TSO packets at retransmit time")

Before that commit, the assumption underlying the tp->undo_retrans--
seems correct, AFAICT.

thanks,
neal