[PATCH v2 net] inet: frags: better deal with smp races

[PATCH v2 net] inet: frags: better deal with smp races

[Eric Dumazet]

Multiple cpus might attempt to insert a new fragment in rhashtable,
if for example RPS is buggy, as reported by 배석진 in
https://patchwork.ozlabs.org/patch/994601/

We use rhashtable_lookup_get_insert_key() instead of
rhashtable_insert_fast() to let cpus losing the race
free their own inet_frag_queue and use the one that
was inserted by another cpu.

Fixes: 648700f76b03 ("inet: frags: use rhashtables for reassembly units")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: 배석진 <soukjin.bae@samsung.com>
---
 net/ipv4/inet_fragment.c | 29 +++++++++++++++--------------
 1 file changed, 15 insertions(+), 14 deletions(-)

diff --git a/net/ipv4/inet_fragment.c b/net/ipv4/inet_fragment.c
index bcb11f3a27c0c34115af05034a5a20f57842eb0a..760a9e52e02b91b36af323c92f7027e150858f88 100644
--- a/net/ipv4/inet_fragment.c
+++ b/net/ipv4/inet_fragment.c
@@ -178,21 +178,22 @@ static struct inet_frag_queue *inet_frag_alloc(struct netns_frags *nf,
 }
 
 static struct inet_frag_queue *inet_frag_create(struct netns_frags *nf,
-						void *arg)
+						void *arg,
+						struct inet_frag_queue **prev)
 {
 	struct inet_frags *f = nf->f;
 	struct inet_frag_queue *q;
-	int err;
 
 	q = inet_frag_alloc(nf, f, arg);
-	if (!q)
+	if (!q) {
+		*prev = ERR_PTR(-ENOMEM);
 		return NULL;
-
+	}
 	mod_timer(&q->timer, jiffies + nf->timeout);
 
-	err = rhashtable_insert_fast(&nf->rhashtable, &q->node,
-				     f->rhash_params);
-	if (err < 0) {
+	*prev = rhashtable_lookup_get_insert_key(&nf->rhashtable, &q->key,
+						 &q->node, f->rhash_params);
+	if (*prev) {
 		q->flags |= INET_FRAG_COMPLETE;
 		inet_frag_kill(q);
 		inet_frag_destroy(q);
@@ -204,22 +205,22 @@ static struct inet_frag_queue *inet_frag_create(struct netns_frags *nf,
 /* TODO : call from rcu_read_lock() and no longer use refcount_inc_not_zero() */
 struct inet_frag_queue *inet_frag_find(struct netns_frags *nf, void *key)
 {
-	struct inet_frag_queue *fq;
+	struct inet_frag_queue *fq = NULL, *prev;
 
 	if (!nf->high_thresh || frag_mem_limit(nf) > nf->high_thresh)
 		return NULL;
 
 	rcu_read_lock();
 
-	fq = rhashtable_lookup(&nf->rhashtable, key, nf->f->rhash_params);
-	if (fq) {
+	prev = rhashtable_lookup(&nf->rhashtable, key, nf->f->rhash_params);
+	if (!prev)
+		fq = inet_frag_create(nf, key, &prev);
+	if (prev && !IS_ERR(prev)) {
+		fq = prev;
 		if (!refcount_inc_not_zero(&fq->refcnt))
 			fq = NULL;
-		rcu_read_unlock();
-		return fq;
 	}
 	rcu_read_unlock();
-
-	return inet_frag_create(nf, key);
+	return fq;
 }
 EXPORT_SYMBOL(inet_frag_find);
-- 
2.19.1.930.g4563a0d9d0-goog

Re: [PATCH v2 net] inet: frags: better deal with smp races

[David Miller]

From: Eric Dumazet <edumazet@google.com>
Date: Thu,  8 Nov 2018 17:34:27 -0800

> Multiple cpus might attempt to insert a new fragment in rhashtable,
> if for example RPS is buggy, as reported by 배석진 in
> https://patchwork.ozlabs.org/patch/994601/
> 
> We use rhashtable_lookup_get_insert_key() instead of
> rhashtable_insert_fast() to let cpus losing the race
> free their own inet_frag_queue and use the one that
> was inserted by another cpu.
> 
> Fixes: 648700f76b03 ("inet: frags: use rhashtables for reassembly units")
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Reported-by: 배석진 <soukjin.bae@samsung.com>

Applied and queued up for -stable, thanks.