From 518fbf9cdf17875d808596afd77fc115a6f942ca Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Sat, 28 Apr 2012 23:21:56 +0000 Subject: [PATCH] net: fix sk_sockets_allocated_read_positive Denys Fedoryshchenko reported frequent crashes on a proxy server and kindly provided a lockdep report that explains it all : [ 762.903868] [ 762.903880] ================================= [ 762.903890] [ INFO: inconsistent lock state ] [ 762.903903] 3.3.4-build-0061 #8 Not tainted [ 762.904133] --------------------------------- [ 762.904344] inconsistent {IN-SOFTIRQ-W} -> {SOFTIRQ-ON-W} usage. [ 762.904542] squid/1603 [HC0[0]:SC0[0]:HE1:SE1] takes: [ 762.904542] (key#3){+.?...}, at: [] __percpu_counter_sum+0xd/0x58 [ 762.904542] {IN-SOFTIRQ-W} state was registered at: [ 762.904542] [] __lock_acquire+0x284/0xc26 [ 762.904542] [] lock_acquire+0x71/0x85 [ 762.904542] [] _raw_spin_lock+0x33/0x40 [ 762.904542] [] __percpu_counter_add+0x58/0x7c [ 762.904542] [] sk_clone_lock+0x1e5/0x200 [ 762.904542] [] inet_csk_clone_lock+0xe/0x78 [ 762.904542] [] tcp_create_openreq_child+0x1b/0x404 [ 762.904542] [] tcp_v4_syn_recv_sock+0x32/0x1c1 [ 762.904542] [] tcp_check_req+0x1fd/0x2d7 [ 762.904542] [] tcp_v4_do_rcv+0xab/0x194 [ 762.904542] [] tcp_v4_rcv+0x3b3/0x5cc [ 762.904542] [] ip_local_deliver_finish+0x13a/0x1e9 [ 762.904542] [] NF_HOOK.clone.11+0x46/0x4d [ 762.904542] [] ip_local_deliver+0x41/0x45 [ 762.904542] [] ip_rcv_finish+0x31a/0x33c [ 762.904542] [] NF_HOOK.clone.11+0x46/0x4d [ 762.904542] [] ip_rcv+0x201/0x23e [ 762.904542] [] __netif_receive_skb+0x319/0x368 [ 762.904542] [] netif_receive_skb+0x4e/0x7d [ 762.904542] [] napi_skb_finish+0x1e/0x34 [ 762.904542] [] napi_gro_receive+0x20/0x24 [ 762.904542] [] e1000_receive_skb+0x3f/0x45 [e1000e] [ 762.904542] [] e1000_clean_rx_irq+0x1f9/0x284 [e1000e] [ 762.904542] [] e1000_clean+0x62/0x1f4 [e1000e] [ 762.904542] [] net_rx_action+0x90/0x160 [ 762.904542] [] __do_softirq+0x7b/0x118 [ 762.904542] irq event stamp: 156915469 [ 762.904542] hardirqs last enabled at (156915469): [] __slab_alloc.clone.58.clone.63+0xc4/0x2de [ 762.904542] hardirqs last disabled at (156915468): [] __slab_alloc.clone.58.clone.63+0x22/0x2de [ 762.904542] softirqs last enabled at (156915466): [] lock_sock_nested+0x64/0x6c [ 762.904542] softirqs last disabled at (156915464): [] _raw_spin_lock_bh+0xe/0x45 [ 762.904542] [ 762.904542] other info that might help us debug this: [ 762.904542] Possible unsafe locking scenario: [ 762.904542] [ 762.904542] CPU0 [ 762.904542] ---- [ 762.904542] lock(key#3); [ 762.904542] [ 762.904542] lock(key#3); [ 762.904542] [ 762.904542] *** DEADLOCK *** [ 762.904542] [ 762.904542] 1 lock held by squid/1603: [ 762.904542] #0: (sk_lock-AF_INET){+.+.+.}, at: [] lock_sock+0xa/0xc [ 762.904542] [ 762.904542] stack backtrace: [ 762.904542] Pid: 1603, comm: squid Not tainted 3.3.4-build-0061 #8 [ 762.904542] Call Trace: [ 762.904542] [] ? printk+0x18/0x1d [ 762.904542] [] valid_state+0x1f6/0x201 [ 762.904542] [] mark_lock+0xd1/0x1bb [ 762.904542] [] ? mark_lock+0x26/0x1bb [ 762.904542] [] ? check_usage_forwards+0x77/0x77 [ 762.904542] [] __lock_acquire+0x2f8/0xc26 [ 762.904542] [] ? mark_held_locks+0x5d/0x7b [ 762.904542] [] ? trace_hardirqs_on+0xb/0xd [ 762.904542] [] ? __lock_acquire+0x4d4/0xc26 [ 762.904542] [] lock_acquire+0x71/0x85 [ 762.904542] [] ? __percpu_counter_sum+0xd/0x58 [ 762.904542] [] _raw_spin_lock+0x33/0x40 [ 762.904542] [] ? __percpu_counter_sum+0xd/0x58 [ 762.904542] [] __percpu_counter_sum+0xd/0x58 [ 762.904542] [] __sk_mem_schedule+0xdd/0x1c7 [ 762.904542] [] ? __alloc_skb+0x76/0x100 [ 762.904542] [] sk_wmem_schedule+0x21/0x2d [ 762.904542] [] sk_stream_alloc_skb+0x42/0xaa [ 762.904542] [] tcp_sendmsg+0x18f/0x68b [ 762.904542] [] ? ip_fast_csum+0x30/0x30 [ 762.904542] [] inet_sendmsg+0x53/0x5a [ 762.904542] [] sock_aio_write+0xd2/0xda [ 762.904542] [] ? mark_lock+0x26/0x1bb [ 762.904542] [] do_sync_write+0x9f/0xd9 [ 762.904542] [] ? file_free_rcu+0x2f/0x2f [ 762.904542] [] vfs_write+0x8f/0xab [ 762.904542] [] ? fget_light+0x75/0x7c [ 762.904542] [] sys_write+0x3d/0x5e [ 762.904542] [] syscall_call+0x7/0xb [ 762.904542] [] ? rp_sidt+0x41/0x83 Bug is that sk_sockets_allocated_read_positive() calls percpu_counter_sum_positive() without BH being disabled. This bug was added in commit 180d8cd942ce33 (foundations of per-cgroup memory pressure controlling.), since previous code was using percpu_counter_read_positive() which is IRQ safe. In __sk_mem_schedule() we dont need the precise count of allocated sockets and can revert to previous behavior. Reported-by: Denys Fedoryshchenko Sined-off-by: Eric Dumazet Cc: Glauber Costa Acked-by: Neal Cardwell Signed-off-by: David S. Miller --- include/net/sock.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/include/net/sock.h b/include/net/sock.h index 188532ee88b6..5a0a58ac4126 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -1129,9 +1129,9 @@ sk_sockets_allocated_read_positive(struct sock *sk) struct proto *prot = sk->sk_prot; if (mem_cgroup_sockets_enabled && sk->sk_cgrp) - return percpu_counter_sum_positive(sk->sk_cgrp->sockets_allocated); + return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated); - return percpu_counter_sum_positive(prot->sockets_allocated); + return percpu_counter_read_positive(prot->sockets_allocated); } static inline int -- 2.34.1