Eric -
With this patch applied the test ran clean for 2 days.
Thanks for your help.
Quoting Eric Dumazet <eric.duma...@gmail.com>:
On Fri, 2015-12-11 at 07:48 -0800, Eric Dumazet wrote:
On Fri, 2015-12-11 at 06:23 -0800, Eric Dumazet wrote:
> On Sun, 2015-12-06 at 17:58 -0800, Eric Dumazet wrote:
> > On Sun, 2015-12-06 at 13:03 -0800, Eric Dumazet wrote:
> >
> > > But then when later we promote a skb->dst to a refctounted one
> > > (skb_dst_force(), we might make sure we abort the operation
if __refcnt
> > > == 0 ( and DST_NOCACHE is in dst->flags)
> > >
> >
> > Minimum patch would be :
> >
>
> Here is a more complete patch, it should fix the issue I think :
Hmm, I'll send a v3, I forgot to test DST_NOCACHE properly.
David, please test the following patch, thanks !
include/net/dst.h | 33 +++++++++++++++++++++++++++++++++
include/net/sock.h | 2 +-
net/ipv4/tcp_ipv4.c | 5 ++---
net/ipv6/tcp_ipv6.c | 3 +--
4 files changed, 37 insertions(+), 6 deletions(-)
diff --git a/include/net/dst.h b/include/net/dst.h
index 1279f9b09791..c7329dcd90cc 100644
--- a/include/net/dst.h
+++ b/include/net/dst.h
@@ -322,6 +322,39 @@ static inline void skb_dst_force(struct sk_buff *skb)
}
}
+/**
+ * dst_hold_safe - Take a reference on a dst if possible
+ * @dst: pointer to dst entry
+ *
+ * This helper returns false if it could not safely
+ * take a reference on a dst.
+ */
+static inline bool dst_hold_safe(struct dst_entry *dst)
+{
+ if (dst->flags & DST_NOCACHE)
+ return atomic_inc_not_zero(&dst->__refcnt);
+ dst_hold(dst);
+ return true;
+}
+
+/**
+ * skb_dst_force_safe - makes sure skb dst is refcounted
+ * @skb: buffer
+ *
+ * If dst is not yet refcounted and not destroyed, grab a ref on it.
+ */
+static inline void skb_dst_force_safe(struct sk_buff *skb)
+{
+ if (skb_dst_is_noref(skb)) {
+ struct dst_entry *dst = skb_dst(skb);
+
+ if (!dst_hold_safe(dst))
+ dst = NULL;
+
+ skb->_skb_refdst = (unsigned long)dst;
+ }
+}
+
/**
* __skb_tunnel_rx - prepare skb for rx reinsert
diff --git a/include/net/sock.h b/include/net/sock.h
index eaef41433d7a..18322bded064 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -816,7 +816,7 @@ void sk_stream_write_space(struct sock *sk);
static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
{
/* dont let skb dst not refcounted, we are going to leave rcu lock */
- skb_dst_force(skb);
+ skb_dst_force_safe(skb);
if (!sk->sk_backlog.tail)
sk->sk_backlog.head = skb;
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index db003438aaf5..d8841a2f1569 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -1493,7 +1493,7 @@ bool tcp_prequeue(struct sock *sk, struct sk_buff *skb)
if (likely(sk->sk_rx_dst))
skb_dst_drop(skb);
else
- skb_dst_force(skb);
+ skb_dst_force_safe(skb);
__skb_queue_tail(&tp->ucopy.prequeue, skb);
tp->ucopy.memory += skb->truesize;
@@ -1721,8 +1721,7 @@ void inet_sk_rx_dst_set(struct sock *sk, const
struct sk_buff *skb)
{
struct dst_entry *dst = skb_dst(skb);
- if (dst) {
- dst_hold(dst);
+ if (dst && dst_hold_safe(dst)) {
sk->sk_rx_dst = dst;
inet_sk(sk)->rx_dst_ifindex = skb->skb_iif;
}
diff --git a/net/ipv6/tcp_ipv6.c b/net/ipv6/tcp_ipv6.c
index e7aab561b7b4..6b8a8a9091fa 100644
--- a/net/ipv6/tcp_ipv6.c
+++ b/net/ipv6/tcp_ipv6.c
@@ -93,10 +93,9 @@ static void inet6_sk_rx_dst_set(struct sock *sk,
const struct sk_buff *skb)
{
struct dst_entry *dst = skb_dst(skb);
- if (dst) {
+ if (dst && dst_hold_safe(dst)) {
const struct rt6_info *rt = (const struct rt6_info *)dst;
- dst_hold(dst);
sk->sk_rx_dst = dst;
inet_sk(sk)->rx_dst_ifindex = skb->skb_iif;
inet6_sk(sk)->rx_dst_cookie = rt6_get_cookie(rt);
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
