5.10-stable review patch. If anyone has any objections, please let me know.
------------------
From: Kuniyuki Iwashima kuniyu@google.com
[ Upstream commit ae8f160e7eb24240a2a79fc4c815c6a0d4ee16cc ]
Netlink has this pattern in some places
if (atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf) atomic_add(skb->truesize, &sk->sk_rmem_alloc);
, which has the same problem fixed by commit 5a465a0da13e ("udp: Fix multiple wraparounds of sk->sk_rmem_alloc.").
For example, if we set INT_MAX to SO_RCVBUFFORCE, the condition is always false as the two operands are of int.
Then, a single socket can eat as many skb as possible until OOM happens, and we can see multiple wraparounds of sk->sk_rmem_alloc.
Let's fix it by using atomic_add_return() and comparing the two variables as unsigned int.
Before: [root@fedora ~]# ss -f netlink Recv-Q Send-Q Local Address:Port Peer Address:Port -1668710080 0 rtnl:nl_wraparound/293 *
After: [root@fedora ~]# ss -f netlink Recv-Q Send-Q Local Address:Port Peer Address:Port 2147483072 0 rtnl:nl_wraparound/290 * ^ `--- INT_MAX - 576
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Reported-by: Jason Baron jbaron@akamai.com Closes: https://lore.kernel.org/netdev/cover.1750285100.git.jbaron@akamai.com/ Signed-off-by: Kuniyuki Iwashima kuniyu@google.com Link: https://patch.msgid.link/20250704054824.1580222-1-kuniyu@google.com Signed-off-by: Jakub Kicinski kuba@kernel.org Signed-off-by: Sasha Levin sashal@kernel.org --- net/netlink/af_netlink.c | 81 ++++++++++++++++++++++++---------------- 1 file changed, 49 insertions(+), 32 deletions(-)
diff --git a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c index 4da043d9f2c7a..6f71b4c84d1d2 100644 --- a/net/netlink/af_netlink.c +++ b/net/netlink/af_netlink.c @@ -379,7 +379,6 @@ static void netlink_skb_set_owner_r(struct sk_buff *skb, struct sock *sk) WARN_ON(skb->sk != NULL); skb->sk = sk; skb->destructor = netlink_skb_destructor; - atomic_add(skb->truesize, &sk->sk_rmem_alloc); sk_mem_charge(sk, skb->truesize); }
@@ -1207,41 +1206,48 @@ static struct sk_buff *netlink_alloc_large_skb(unsigned int size, int netlink_attachskb(struct sock *sk, struct sk_buff *skb, long *timeo, struct sock *ssk) { + DECLARE_WAITQUEUE(wait, current); struct netlink_sock *nlk; + unsigned int rmem;
nlk = nlk_sk(sk); + rmem = atomic_add_return(skb->truesize, &sk->sk_rmem_alloc);
- if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || - test_bit(NETLINK_S_CONGESTED, &nlk->state))) { - DECLARE_WAITQUEUE(wait, current); - if (!*timeo) { - if (!ssk || netlink_is_kernel(ssk)) - netlink_overrun(sk); - sock_put(sk); - kfree_skb(skb); - return -EAGAIN; - } - - __set_current_state(TASK_INTERRUPTIBLE); - add_wait_queue(&nlk->wait, &wait); + if ((rmem == skb->truesize || rmem < READ_ONCE(sk->sk_rcvbuf)) && + !test_bit(NETLINK_S_CONGESTED, &nlk->state)) { + netlink_skb_set_owner_r(skb, sk); + return 0; + }
- if ((atomic_read(&sk->sk_rmem_alloc) > sk->sk_rcvbuf || - test_bit(NETLINK_S_CONGESTED, &nlk->state)) && - !sock_flag(sk, SOCK_DEAD)) - *timeo = schedule_timeout(*timeo); + atomic_sub(skb->truesize, &sk->sk_rmem_alloc);
- __set_current_state(TASK_RUNNING); - remove_wait_queue(&nlk->wait, &wait); + if (!*timeo) { + if (!ssk || netlink_is_kernel(ssk)) + netlink_overrun(sk); sock_put(sk); + kfree_skb(skb); + return -EAGAIN; + }
- if (signal_pending(current)) { - kfree_skb(skb); - return sock_intr_errno(*timeo); - } - return 1; + __set_current_state(TASK_INTERRUPTIBLE); + add_wait_queue(&nlk->wait, &wait); + rmem = atomic_read(&sk->sk_rmem_alloc); + + if (((rmem && rmem + skb->truesize > READ_ONCE(sk->sk_rcvbuf)) || + test_bit(NETLINK_S_CONGESTED, &nlk->state)) && + !sock_flag(sk, SOCK_DEAD)) + *timeo = schedule_timeout(*timeo); + + __set_current_state(TASK_RUNNING); + remove_wait_queue(&nlk->wait, &wait); + sock_put(sk); + + if (signal_pending(current)) { + kfree_skb(skb); + return sock_intr_errno(*timeo); } - netlink_skb_set_owner_r(skb, sk); - return 0; + + return 1; }
static int __netlink_sendskb(struct sock *sk, struct sk_buff *skb) @@ -1301,6 +1307,7 @@ static int netlink_unicast_kernel(struct sock *sk, struct sk_buff *skb, ret = -ECONNREFUSED; if (nlk->netlink_rcv != NULL) { ret = skb->len; + atomic_add(skb->truesize, &sk->sk_rmem_alloc); netlink_skb_set_owner_r(skb, sk); NETLINK_CB(skb).sk = ssk; netlink_deliver_tap_kernel(sk, ssk, skb); @@ -1379,13 +1386,19 @@ EXPORT_SYMBOL_GPL(netlink_strict_get_check); static int netlink_broadcast_deliver(struct sock *sk, struct sk_buff *skb) { struct netlink_sock *nlk = nlk_sk(sk); + unsigned int rmem, rcvbuf;
- if (atomic_read(&sk->sk_rmem_alloc) <= sk->sk_rcvbuf && + rmem = atomic_add_return(skb->truesize, &sk->sk_rmem_alloc); + rcvbuf = READ_ONCE(sk->sk_rcvbuf); + + if ((rmem != skb->truesize || rmem <= rcvbuf) && !test_bit(NETLINK_S_CONGESTED, &nlk->state)) { netlink_skb_set_owner_r(skb, sk); __netlink_sendskb(sk, skb); - return atomic_read(&sk->sk_rmem_alloc) > (sk->sk_rcvbuf >> 1); + return rmem > (rcvbuf >> 1); } + + atomic_sub(skb->truesize, &sk->sk_rmem_alloc); return -1; }
@@ -2202,6 +2215,7 @@ static int netlink_dump(struct sock *sk, bool lock_taken) struct module *module; int err = -ENOBUFS; int alloc_min_size; + unsigned int rmem; int alloc_size;
if (!lock_taken) @@ -2211,9 +2225,6 @@ static int netlink_dump(struct sock *sk, bool lock_taken) goto errout_skb; }
- if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) - goto errout_skb; - /* NLMSG_GOODSIZE is small to avoid high order allocations being * required, but it makes sense to _attempt_ a 16K bytes allocation * to reduce number of system calls on dump operations, if user @@ -2236,6 +2247,12 @@ static int netlink_dump(struct sock *sk, bool lock_taken) if (!skb) goto errout_skb;
+ rmem = atomic_add_return(skb->truesize, &sk->sk_rmem_alloc); + if (rmem >= READ_ONCE(sk->sk_rcvbuf)) { + atomic_sub(skb->truesize, &sk->sk_rmem_alloc); + goto errout_skb; + } + /* Trim skb to allocated size. User is expected to provide buffer as * large as max(min_dump_alloc, 16KiB (mac_recvmsg_len capped at * netlink_recvmsg())). dump will pack as many smaller messages as