2022-11-24 12:06:08

by Richard Gobert

[permalink] [raw]
Subject: [PATCH net-next 1/2] gro: simplify flush logic

In tcp_gro_receive, the variable mss is sometimes assigned 1 to make the check
"flush = len < mss" fail, and indirectly set flush to 0.

This somewhat obfuscates the meaning of the code. Also, setting mss to 1 doesn't
influence the value of flush in the case when skb_is_gso(skb) is true.

This patch changes the code to set flush to 0 directly. The out_check_final can
then be moved, which avoids unnecessary conditions.

Signed-off-by: Richard Gobert <[email protected]>
---
net/ipv4/tcp_offload.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/net/ipv4/tcp_offload.c b/net/ipv4/tcp_offload.c
index 45dda7889387..f17271e5c7c5 100644
--- a/net/ipv4/tcp_offload.c
+++ b/net/ipv4/tcp_offload.c
@@ -187,7 +187,7 @@ struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb)
unsigned int len;
unsigned int thlen;
__be32 flags;
- unsigned int mss = 1;
+ unsigned int mss;
unsigned int hlen;
unsigned int off;
int flush = 1;
@@ -229,6 +229,7 @@ struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb)
goto found;
}
p = NULL;
+ flush = 0;
goto out_check_final;

found:
@@ -270,19 +271,19 @@ struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb)
#endif

if (flush || skb_gro_receive(p, skb)) {
- mss = 1;
+ flush = 0;
goto out_check_final;
}

tcp_flag_word(th2) |= flags & (TCP_FLAG_FIN | TCP_FLAG_PSH);

-out_check_final:
/* Force a flush if last segment is smaller than mss. */
if (unlikely(skb_is_gso(skb)))
flush = len != NAPI_GRO_CB(skb)->count * skb_shinfo(skb)->gso_size;
else
flush = len < mss;

+out_check_final:
flush |= (__force int)(flags & (TCP_FLAG_URG | TCP_FLAG_PSH |
TCP_FLAG_RST | TCP_FLAG_SYN |
TCP_FLAG_FIN));
--
2.36.1