Re: [RFC PATCH net v2] net: fix data-races around sk->sk_forward_alloc

[Simon Horman <horms@kernel.org>]

On Tue, Nov 05, 2024 at 10:52:34AM +0100, Eric Dumazet wrote:
> On Tue, Nov 5, 2024 at 8:46 AM Wang Liang <wangliang74@huawei.com> wrote:
> >
> > Syzkaller reported this warning:
> >  ------------[ cut here ]------------
> >  WARNING: CPU: 0 PID: 16 at net/ipv4/af_inet.c:156 inet_sock_destruct+0x1c5/0x1e0
> >  Modules linked in:
> >  CPU: 0 UID: 0 PID: 16 Comm: ksoftirqd/0 Not tainted 6.12.0-rc5 #26
> >  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.15.0-1 04/01/2014
> >  RIP: 0010:inet_sock_destruct+0x1c5/0x1e0
> >  Code: 24 12 4c 89 e2 5b 48 c7 c7 98 ec bb 82 41 5c e9 d1 18 17 ff 4c 89 e6 5b 48 c7 c7 d0 ec bb 82 41 5c e9 bf 18 17 ff 0f 0b eb 83 <0f> 0b eb 97 0f 0b eb 87 0f 0b e9 68 ff ff ff 66 66 2e 0f 1f 84 00
> >  RSP: 0018:ffffc9000008bd90 EFLAGS: 00010206
> >  RAX: 0000000000000300 RBX: ffff88810b172a90 RCX: 0000000000000007
> >  RDX: 0000000000000002 RSI: 0000000000000300 RDI: ffff88810b172a00
> >  RBP: ffff88810b172a00 R08: ffff888104273c00 R09: 0000000000100007
> >  R10: 0000000000020000 R11: 0000000000000006 R12: ffff88810b172a00
> >  R13: 0000000000000004 R14: 0000000000000000 R15: ffff888237c31f78
> >  FS:  0000000000000000(0000) GS:ffff888237c00000(0000) knlGS:0000000000000000
> >  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> >  CR2: 00007ffc63fecac8 CR3: 000000000342e000 CR4: 00000000000006f0
> >  DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> >  DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> >  Call Trace:
> >   <TASK>
> >   ? __warn+0x88/0x130
> >   ? inet_sock_destruct+0x1c5/0x1e0
> >   ? report_bug+0x18e/0x1a0
> >   ? handle_bug+0x53/0x90
> >   ? exc_invalid_op+0x18/0x70
> >   ? asm_exc_invalid_op+0x1a/0x20
> >   ? inet_sock_destruct+0x1c5/0x1e0
> >   __sk_destruct+0x2a/0x200
> >   rcu_do_batch+0x1aa/0x530
> >   ? rcu_do_batch+0x13b/0x530
> >   rcu_core+0x159/0x2f0
> >   handle_softirqs+0xd3/0x2b0
> >   ? __pfx_smpboot_thread_fn+0x10/0x10
> >   run_ksoftirqd+0x25/0x30
> >   smpboot_thread_fn+0xdd/0x1d0
> >   kthread+0xd3/0x100
> >   ? __pfx_kthread+0x10/0x10
> >   ret_from_fork+0x34/0x50
> >   ? __pfx_kthread+0x10/0x10
> >   ret_from_fork_asm+0x1a/0x30
> >   </TASK>
> >  ---[ end trace 0000000000000000 ]---
> >
> > Its possible that two threads call tcp_v6_do_rcv()/sk_forward_alloc_add()
> > concurrently when sk->sk_state == TCP_LISTEN with sk->sk_lock unlocked,
> > which triggers a data-race around sk->sk_forward_alloc:
> > tcp_v6_rcv
> >     tcp_v6_do_rcv
> >         skb_clone_and_charge_r
> >             sk_rmem_schedule
> >                 __sk_mem_schedule
> >                     sk_forward_alloc_add()
> >             skb_set_owner_r
> >                 sk_mem_charge
> >                     sk_forward_alloc_add()
> >         __kfree_skb
> >             skb_release_all
> >                 skb_release_head_state
> >                     sock_rfree
> >                         sk_mem_uncharge
> >                             sk_forward_alloc_add()
> >                             sk_mem_reclaim
> >                                 // set local var reclaimable
> >                                 __sk_mem_reclaim
> >                                     sk_forward_alloc_add()
> >
> > In this syzkaller testcase, two threads call
> > tcp_v6_do_rcv() with skb->truesize=768, the sk_forward_alloc changes like
> > this:
> >  (cpu 1)             | (cpu 2)             | sk_forward_alloc
> >  ...                 | ...                 | 0
> >  __sk_mem_schedule() |                     | +4096 = 4096
> >                      | __sk_mem_schedule() | +4096 = 8192
> >  sk_mem_charge()     |                     | -768  = 7424
> >                      | sk_mem_charge()     | -768  = 6656
> >  ...                 |    ...              |
> >  sk_mem_uncharge()   |                     | +768  = 7424
> >  reclaimable=7424    |                     |
> >                      | sk_mem_uncharge()   | +768  = 8192
> >                      | reclaimable=8192    |
> >  __sk_mem_reclaim()  |                     | -4096 = 4096
> >                      | __sk_mem_reclaim()  | -8192 = -4096 != 0
> >
> > The skb_clone_and_charge_r() should not be called in tcp_v6_do_rcv() when
> > sk->sk_state is TCP_LISTEN, it happens later in tcp_v6_syn_recv_sock().
> > Fix the same issue in dccp_v6_do_rcv().
> >
> > Suggested-by: Eric Dumazet <edumazet@google.com>
> > Fixes: e994b2f0fb92 ("tcp: do not lock listener to process SYN packets")
> > Signed-off-by: Wang Liang <wangliang74@huawei.com>
> 
> Reviewed-by: Eric Dumazet <edumazet@google.com>

Hi Wang Liang,

Please post a non-RFC variant of this patch so it can be considered for
inclusion in net. And please include Eric's Reviewed-by tag.

Thanks!