tcp_metrics: fix data-race in tcpm_suck_dst() vs fastopen

Whenever tcpm_new() reclaims an old entry, tcpm_suck_dst()
would overwrite data that could be read from tcp_fastopen_cache_get()
or tcp_metrics_fill_info().

We need to acquire fastopen_seqlock to maintain consistency.

For newly allocated objects, tcpm_new() can switch to kzalloc()
to avoid an extra fastopen_seqlock acquisition.

Fixes: 1fe4c481ba63 ("net-tcp: Fast Open client - cookie cache")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://lore.kernel.org/r/20230802131500.1478140-7-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

diff --git a/net/ipv4/tcp_metrics.c b/net/ipv4/tcp_metrics.c
index 4fd274836a48f73d0b1206adfa14c17c3b28bc30..99ac5efe244d3c654deaa8f8c0fffeeb5d5597b1 100644 (file)
--- a/net/ipv4/tcp_metrics.c
+++ b/net/ipv4/tcp_metrics.c
@@ -93,6 +93,7 @@ static struct tcpm_hash_bucket        *tcp_metrics_hash __read_mostly;
 static unsigned int            tcp_metrics_hash_log __read_mostly;
 
 static DEFINE_SPINLOCK(tcp_metrics_lock);
+static DEFINE_SEQLOCK(fastopen_seqlock);
 
 static void tcpm_suck_dst(struct tcp_metrics_block *tm,
                          const struct dst_entry *dst,
@@ -129,11 +130,13 @@ static void tcpm_suck_dst(struct tcp_metrics_block *tm,
        tcp_metric_set(tm, TCP_METRIC_REORDERING,
                       dst_metric_raw(dst, RTAX_REORDERING));
        if (fastopen_clear) {
+               write_seqlock(&fastopen_seqlock);
                tm->tcpm_fastopen.mss = 0;
                tm->tcpm_fastopen.syn_loss = 0;
                tm->tcpm_fastopen.try_exp = 0;
                tm->tcpm_fastopen.cookie.exp = false;
                tm->tcpm_fastopen.cookie.len = 0;
+               write_sequnlock(&fastopen_seqlock);
        }
 }
 
@@ -194,7 +197,7 @@ static struct tcp_metrics_block *tcpm_new(struct dst_entry *dst,
                }
                tm = oldest;
        } else {
-               tm = kmalloc(sizeof(*tm), GFP_ATOMIC);
+               tm = kzalloc(sizeof(*tm), GFP_ATOMIC);
                if (!tm)
                        goto out_unlock;
        }
@@ -204,7 +207,7 @@ static struct tcp_metrics_block *tcpm_new(struct dst_entry *dst,
        tm->tcpm_saddr = *saddr;
        tm->tcpm_daddr = *daddr;
 
-       tcpm_suck_dst(tm, dst, true);
+       tcpm_suck_dst(tm, dst, reclaim);
 
        if (likely(!reclaim)) {
                tm->tcpm_next = tcp_metrics_hash[hash].chain;
@@ -556,8 +559,6 @@ bool tcp_peer_is_proven(struct request_sock *req, struct dst_entry *dst)
        return ret;
 }
 
-static DEFINE_SEQLOCK(fastopen_seqlock);
-
 void tcp_fastopen_cache_get(struct sock *sk, u16 *mss,
                            struct tcp_fastopen_cookie *cookie)
 {