Re: memory leak in create_ctx
From: John Fastabend <john.fastabend@gmail.com>
Date: 2019-06-14 04:21:36
Also in:
bpf, lkml
Hillf Danton wrote:
Hi Dmitry On Tue, 11 Jun 2019 19:45:28 +0800 Dmitry Vyukov wrote:quoted
I've run the repro as "./syz-execprog -repeat=0 -procs=6 repro" and in 10 mins I got the following splat, which indeed suggests a bpf bug. But we of course can have both bpf stack overflow and a memory leak in tls. 2019/06/11 10:26:52 executed programs: 887 2019/06/11 10:26:57 executed programs: 899 2019/06/11 10:27:02 executed programs: 916 [ 429.171049][ T9870] BUG: stack guard page was hit at 00000000a78467b9 (stack is 000000001452e9df..000000004fb93e51) [ 429.173714][ T9870] kernel stack overflow (double-fault): 0000 [#1] PREEMPT SMP [ 429.174819][ T9870] CPU: 3 PID: 9870 Comm: syz-executor Not tainted 5.2.0-rc4+ #6 [ 429.175901][ T9870] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014 [ 429.177215][ T9870] RIP: 0010:tcp_bpf_unhash+0xc/0x80 [ 429.177950][ T9870] Code: 28 4c 89 ee 48 89 df ff 10 e8 30 56 66 fe 5b 41 5c 41 5d 41 5e 5d c3 0f 1f 80 00 00 00 00 55 48 89 e5 41 55 41 54 53 48 89 fb <e8> 0f 56 66 fe e8 6a bb 5f fe 4c 8b a3 80 02 00 00 4d 85 e4 74 2f [ 429.180707][ T9870] RSP: 0018:ffffc90003690000 EFLAGS: 00010293 [ 429.181562][ T9870] RAX: ffff888066a72000 RBX: ffff88806695b640 RCX: ffffffff82c82f80 [ 429.182681][ T9870] RDX: 0000000000000000 RSI: 0000000000000007 RDI: ffff88806695b640 [ 429.183807][ T9870] RBP: ffffc90003690018 R08: 0000000000000000 R09: 0000000000000000 [ 429.184931][ T9870] R10: ffffc90003693e70 R11: 0000000000000000 R12: ffffffff82c82f10 [ 429.186104][ T9870] R13: 0000000000000007 R14: ffff88806695b710 R15: ffff88806695b710 [ 429.187303][ T9870] FS: 00005555569fc940(0000) GS:ffff88807db80000(0000) knlGS:0000000000000000 [ 429.188678][ T9870] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 429.189674][ T9870] CR2: ffffc9000368fff8 CR3: 00000000762bc002 CR4: 00000000007606e0 [ 429.190880][ T9870] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 429.192094][ T9870] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 429.193295][ T9870] PKRU: 55555554 [ 429.193829][ T9870] Call Trace: [ 429.194326][ T9870] ? tcp_bpf_close+0xa0/0xa0 [ 429.195020][ T9870] tcp_bpf_unhash+0x76/0x80 [ 429.195706][ T9870] ? tcp_bpf_close+0xa0/0xa0 [ 429.196400][ T9870] tcp_bpf_unhash+0x76/0x80 [ 429.197079][ T9870] ? tcp_bpf_close+0xa0/0xa0 [ 429.197773][ T9870] tcp_bpf_unhash+0x76/0x80 [ 429.651942][ T9870] ? tcp_bpf_close+0xa0/0xa0... duplicated info trimed ...quoted
[ 429.652512][ T9870] tcp_bpf_unhash+0x76/0x80 [ 429.656467][ T9870] ? tcp_bpf_close+0xa0/0xa0 [ 429.657037][ T9870] tcp_bpf_unhash+0x76/0x80 [ 429.657600][ T9870] tcp_set_state+0x7b/0x220 [ 429.658160][ T9870] ? put_object+0x20/0x30 [ 429.658699][ T9870] ? debug_smp_processor_id+0x2b/0x130 [ 429.659382][ T9870] tcp_disconnect+0x518/0x610 [ 429.659973][ T9870] tcp_close+0x41d/0x540 [ 429.660501][ T9870] ? tcp_check_oom+0x180/0x180 [ 429.661095][ T9870] tls_sk_proto_close+0x86/0x2a0 [ 429.661711][ T9870] ? locks_remove_posix+0x114/0x1c0 [ 429.662359][ T9870] inet_release+0x44/0x80 [ 429.662899][ T9870] inet6_release+0x36/0x50 [ 429.663453][ T9870] __sock_release+0x4b/0x100 [ 429.664024][ T9870] ? __sock_release+0x100/0x100 [ 429.664625][ T9870] sock_close+0x19/0x20 [ 429.665141][ T9870] __fput+0xe7/0x2f0 [ 429.665624][ T9870] ____fput+0x15/0x20 [ 429.666120][ T9870] task_work_run+0xa4/0xd0 [ 429.666671][ T9870] exit_to_usermode_loop+0x16f/0x180 [ 429.667329][ T9870] do_syscall_64+0x187/0x1b0 [ 429.667920][ T9870] entry_SYSCALL_64_after_hwframe+0x44/0xa9 [ 429.668654][ T9870] RIP: 0033:0x412451 [ 429.669141][ T9870] Code: 75 14 b8 03 00 00 00 0f 05 48 3d 01 f0 ff ff 0f 83 94 1a 00 00 c3 48 83 ec 08 e8 0a fc ff ff 48 89 04 24 b8 03 00 00 00 0f 05 <48> 8b 3c 24 48 89 c2 e8 53 fc ff ff 48 89 d0 48 83 c4 08 48 3d 01 [ 429.671586][ T9870] RSP: 002b:00007ffde18b5470 EFLAGS: 00000293 ORIG_RAX: 0000000000000003 [ 429.672636][ T9870] RAX: 0000000000000000 RBX: 0000000000000005 RCX: 0000000000412451 [ 429.673628][ T9870] RDX: 0000000000000000 RSI: 0000000000000081 RDI: 0000000000000004 [ 429.674643][ T9870] RBP: 0000000000000000 R08: 0000000000721170 R09: 0000000000000000 [ 429.675641][ T9870] R10: 00007ffde18b5580 R11: 0000000000000293 R12: 0000000000000000 [ 429.676636][ T9870] R13: 000000000071bf00 R14: 00000000006e3140 R15: ffffffffffffffff [ 429.677630][ T9870] Modules linked in: [ 429.678119][ T9870] ---[ end trace a429c7ce256ca7bb ]--- [ 429.678798][ T9870] RIP: 0010:tcp_bpf_unhash+0xc/0x80 [ 429.679447][ T9870] Code: 28 4c 89 ee 48 89 df ff 10 e8 30 56 66 fe 5b 41 5c 41 5d 41 5e 5d c3 0f 1f 80 00 00 00 00 55 48 89 e5 41 55 41 54 53 48 89 fb <e8> 0f 56 66 fe e8 6a bb 5f fe 4c 8b a3 80 02 00 00 4d 85 e4 74 2f [ 429.681882][ T9870] RSP: 0018:ffffc90003690000 EFLAGS: 00010293 [ 429.682637][ T9870] RAX: ffff888066a72000 RBX: ffff88806695b640 RCX: ffffffff82c82f80 [ 429.683630][ T9870] RDX: 0000000000000000 RSI: 0000000000000007 RDI: ffff88806695b640 [ 429.684622][ T9870] RBP: ffffc90003690018 R08: 0000000000000000 R09: 0000000000000000 [ 429.685611][ T9870] R10: ffffc90003693e70 R11: 0000000000000000 R12: ffffffff82c82f10 [ 429.686601][ T9870] R13: 0000000000000007 R14: ffff88806695b710 R15: ffff88806695b710 [ 429.687592][ T9870] FS: 00005555569fc940(0000) GS:ffff88807db80000(0000) knlGS:0000000000000000 [ 429.688701][ T9870] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 429.689519][ T9870] CR2: ffffc9000368fff8 CR3: 00000000762bc002 CR4: 00000000007606e0 [ 429.690511][ T9870] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 429.691507][ T9870] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 429.692502][ T9870] PKRU: 55555554 [ 429.692941][ T9870] Kernel panic - not syncing: Fatal exception [ 429.694377][ T9870] Kernel Offset: disabled [ 429.694913][ T9870] Rebooting in 86400 seconds..In case of short circuit, bail out after warning on the unusual event to advoid spiraling down the stack without brakes.
Thanks for tracking this down, I was able to reproduce with sys-execprog so will try and come up with some fix. We really shouldn't ever get here, somehow we dropped the psock but left the sk_prot hooked up. At this point just returning is going to leave lots of psock/tls/sock state floating around. I guess we need to find the case where sk_prot is not reset. .John
quoted hunk
Thanks Hillf --- net/ipv4/tcp_bpf.c | 4 ++++ 1 file changed, 4 insertions(+)diff --git a/net/ipv4/tcp_bpf.c b/net/ipv4/tcp_bpf.c index 3d1e154..5022cd3 100644 --- a/net/ipv4/tcp_bpf.c +++ b/net/ipv4/tcp_bpf.c@@ -546,6 +546,8 @@ static void tcp_bpf_unhash(struct sock *sk) psock = sk_psock(sk); if (unlikely(!psock)) { rcu_read_unlock(); + if (WARN_ON(sk->sk_prot->unhash == tcp_bpf_unhash)) + return; if (sk->sk_prot->unhash) sk->sk_prot->unhash(sk); return;@@ -568,6 +570,8 @@ static void tcp_bpf_close(struct sock *sk, long timeout) if (unlikely(!psock)) { rcu_read_unlock(); release_sock(sk); + if (WARN_ON(sk->sk_prot->close == tcp_bpf_close)) + return; return sk->sk_prot->close(sk, timeout); } --