Message-ID: <CAN-5tyEsGNTgJkkG4aiTSZqn3QvTQu7hSNg2+5u7Z9Ycvk3GUA@mail.gmail.com>
Date: Tue, 4 Mar 2025 12:36:24 -0500
From: Olga Kornievskaia <aglo@...ch.edu>
To: Andrew Morton <akpm@...ux-foundation.org>, "David S. Miller" <davem@...emloft.net>, 
	David Ahern <dsahern@...nel.org>
Cc: linux-mm@...ck.org, netdev@...r.kernel.org, 
	linux-nfs <linux-nfs@...r.kernel.org>
Subject: circular locking between memory management and network layer (from nfs)

Hello memory management and networking gurus,

While doing NFS testing, we've encountered the following lockdep
warning (included below). While hit in NFS it might be problematic to
other network filesystems. The warning suggests that one mmap thread
acquired a lock doing page fault and calling into NFS which will do a
network call which will grab a socket lock. Elsewhere, there might be
a process trying to send data on a socket and thus holding a socket
lock and then needs to copy userland data which would require
mmap_lock.

Any ideas how this should be resolved/handled?

Thank you.

[ 2720.332903] ======================================================
[ 2720.333945] WARNING: possible circular locking dependency detected
[ 2720.334990] 6.14.0-rc5+ #10 Not tainted
[ 2720.335667] ------------------------------------------------------
[ 2720.336711] xfs_io/48526 is trying to acquire lock:
[ 2720.337553] ffff88817231b778 (sk_lock-AF_INET-RPC){+.+.}-{0:0}, at:
tcp_sock_set_cork+0x17/0x70
[ 2720.339031]
[ 2720.339031] but task is already holding lock:
[ 2720.340013] ffff888111658200 (&mm->mmap_lock){++++}-{4:4}, at:
vm_mmap_pgoff+0x141/0x330
[ 2720.341358]
[ 2720.341358] which lock already depends on the new lock.
[ 2720.341358]
[ 2720.342703]
[ 2720.342703] the existing dependency chain (in reverse order) is:
[ 2720.343944]
[ 2720.343944] -> #1 (&mm->mmap_lock){++++}-{4:4}:
[ 2720.344973]        __lock_acquire+0xcff/0x1fc0
[ 2720.345754]        lock_acquire.part.0+0x11b/0x360
[ 2720.346579]        __might_fault+0xbd/0x120
[ 2720.347311]        _copy_from_iter+0xdf/0x1500
[ 2720.348082]        skb_do_copy_data_nocache+0x160/0x240
[ 2720.348968]        tcp_sendmsg_locked+0x903/0x3920
[ 2720.349796]        tcp_sendmsg+0x2b/0x40
[ 2720.350491]        ____sys_sendmsg+0x8e1/0xc60
[ 2720.351262]        ___sys_sendmsg+0xfd/0x180
[ 2720.352004]        __sys_sendmsg+0x122/0x1b0
[ 2720.352747]        do_syscall_64+0x92/0x180
[ 2720.353475]        entry_SYSCALL_64_after_hwframe+0x76/0x7e
[ 2720.354416]
[ 2720.354416] -> #0 (sk_lock-AF_INET-RPC){+.+.}-{0:0}:
[ 2720.355490]        check_prev_add+0x1b7/0x23e0
[ 2720.356258]        validate_chain+0xa8a/0xf00
[ 2720.357013]        __lock_acquire+0xcff/0x1fc0
[ 2720.357777]        lock_acquire.part.0+0x11b/0x360
[ 2720.358600]        lock_sock_nested+0x3b/0xe0
[ 2720.359359]        tcp_sock_set_cork+0x17/0x70
[ 2720.360129]        xs_tcp_send_request+0x29e/0xaa0 [sunrpc]
[ 2720.361333]        xprt_request_transmit.isra.0+0x23d/0x810 [sunrpc]
[ 2720.362481]        xprt_transmit+0x19a/0x2b0 [sunrpc]
[ 2720.363438]        call_transmit+0x1c7/0x240 [sunrpc]
[ 2720.364386]        __rpc_execute+0x2bb/0xc50 [sunrpc]
[ 2720.365357]        rpc_execute+0xb4/0x110 [sunrpc]
[ 2720.366288]        rpc_run_task+0x363/0x500 [sunrpc]
[ 2720.367224]        rpc_call_sync+0xad/0x150 [sunrpc]
[ 2720.368161]        nfs3_rpc_wrapper+0x46/0x140 [nfsv3]
[ 2720.369041]        nfs3_proc_getattr+0x13d/0x1f0 [nfsv3]
[ 2720.369933]        __nfs_revalidate_inode+0x25e/0x710 [nfs]
[ 2720.371063]        nfs_revalidate_mapping+0xde/0x160 [nfs]
[ 2720.372047]        __mmap_new_vma+0x39c/0x1050
[ 2720.372806]        __mmap_region+0x76a/0xff0
[ 2720.373545]        mmap_region+0x22f/0x300
[ 2720.374258]        do_mmap+0x9d9/0x1020
[ 2720.374925]        vm_mmap_pgoff+0x1b1/0x330
[ 2720.375671]        ksys_mmap_pgoff+0x2ec/0x440
[ 2720.376438]        do_syscall_64+0x92/0x180
[ 2720.377162]        entry_SYSCALL_64_after_hwframe+0x76/0x7e
[ 2720.378097]
[ 2720.378097] other info that might help us debug this:
[ 2720.378097]
[ 2720.379402]  Possible unsafe locking scenario:
[ 2720.379402]
[ 2720.380399]        CPU0                    CPU1
[ 2720.381187]        ----                    ----
[ 2720.381966]   lock(&mm->mmap_lock);
[ 2720.382597]                                lock(sk_lock-AF_INET-RPC);
[ 2720.383667]                                lock(&mm->mmap_lock);
[ 2720.384675]   lock(sk_lock-AF_INET-RPC);
[ 2720.385367]
[ 2720.385367]  *** DEADLOCK ***
[ 2720.385367]
[ 2720.386352] 1 lock held by xfs_io/48526:
[ 2720.387032]  #0: ffff888111658200 (&mm->mmap_lock){++++}-{4:4}, at:
vm_mmap_pgoff+0x141/0x330
[ 2720.388426]
[ 2720.388426] stack backtrace:
[ 2720.389184] CPU: 0 UID: 0 PID: 48526 Comm: xfs_io Not tainted 6.14.0-rc5+ #10
[ 2720.389189] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009),
BIOS 1.16.3-2.fc40 04/01/2014
[ 2720.389196] Call Trace:
[ 2720.389202]  <TASK>
[ 2720.389211]  dump_stack_lvl+0x6f/0xb0
[ 2720.389222]  print_circular_bug.cold+0x38/0x48
[ 2720.389229]  check_noncircular+0x308/0x3f0
[ 2720.389233]  ? rcu_is_watching+0x15/0xb0
[ 2720.389239]  ? __pfx_check_noncircular+0x10/0x10
[ 2720.389243]  ? lock_release+0x288/0x350
[ 2720.389255]  check_prev_add+0x1b7/0x23e0
[ 2720.389260]  ? is_bpf_text_address+0x64/0x100
[ 2720.389264]  ? rcu_is_watching+0x15/0xb0
[ 2720.389270]  validate_chain+0xa8a/0xf00
[ 2720.389277]  ? __pfx_validate_chain+0x10/0x10
[ 2720.389281]  ? mark_lock+0x78/0x860
[ 2720.389288]  __lock_acquire+0xcff/0x1fc0
[ 2720.389298]  lock_acquire.part.0+0x11b/0x360
[ 2720.389302]  ? tcp_sock_set_cork+0x17/0x70
[ 2720.389310]  ? __pfx_lock_acquire.part.0+0x10/0x10
[ 2720.389313]  ? add_lock_to_list+0x18b/0x370
[ 2720.389319]  ? mark_lock+0x78/0x860
[ 2720.389323]  ? rcu_is_watching+0x15/0xb0
[ 2720.389327]  ? lock_acquire+0x234/0x2f0
[ 2720.389333]  lock_sock_nested+0x3b/0xe0
[ 2720.389339]  ? tcp_sock_set_cork+0x17/0x70
[ 2720.389344]  tcp_sock_set_cork+0x17/0x70
[ 2720.389348]  xs_tcp_send_request+0x29e/0xaa0 [sunrpc]
[ 2720.389454]  ? mark_lock+0x78/0x860
[ 2720.389464]  ? __pfx_xs_tcp_send_request+0x10/0x10 [sunrpc]
[ 2720.389567]  ? find_held_lock+0x34/0x120
[ 2720.389572]  ? kvm_sched_clock_read+0x11/0x20
[ 2720.389576]  ? local_clock_noinstr+0xd/0xe0
[ 2720.389581]  ? __lock_release.isra.0+0x4ba/0x9f0
[ 2720.389586]  ? __pfx___lock_release.isra.0+0x10/0x10
[ 2720.389592]  ? rcu_is_watching+0x15/0xb0
[ 2720.389599]  xprt_request_transmit.isra.0+0x23d/0x810 [sunrpc]
[ 2720.389697]  xprt_transmit+0x19a/0x2b0 [sunrpc]
[ 2720.389796]  ? __pfx_call_transmit+0x10/0x10 [sunrpc]
[ 2720.389887]  call_transmit+0x1c7/0x240 [sunrpc]
[ 2720.389986]  __rpc_execute+0x2bb/0xc50 [sunrpc]
[ 2720.390105]  rpc_execute+0xb4/0x110 [sunrpc]
[ 2720.390218]  rpc_run_task+0x363/0x500 [sunrpc]
[ 2720.390312]  rpc_call_sync+0xad/0x150 [sunrpc]
[ 2720.390406]  ? __pfx_rpc_call_sync+0x10/0x10 [sunrpc]
[ 2720.390506]  nfs3_rpc_wrapper+0x46/0x140 [nfsv3]
[ 2720.390518]  nfs3_proc_getattr+0x13d/0x1f0 [nfsv3]
[ 2720.390528]  ? __pfx_nfs3_proc_getattr+0x10/0x10 [nfsv3]
[ 2720.390539]  ? kasan_save_track+0x14/0x30
[ 2720.390543]  ? __kasan_kmalloc+0x8f/0xa0
[ 2720.390550]  __nfs_revalidate_inode+0x25e/0x710 [nfs]
[ 2720.390622]  nfs_revalidate_mapping+0xde/0x160 [nfs]
[ 2720.390692]  __mmap_new_vma+0x39c/0x1050
[ 2720.390701]  __mmap_region+0x76a/0xff0
[ 2720.390706]  ? __pfx___mmap_region+0x10/0x10
[ 2720.390722]  ? __pfx_unmapped_area_topdown+0x10/0x10
[ 2720.390726]  ? __pfx_lock_acquire.part.0+0x10/0x10
[ 2720.390752]  mmap_region+0x22f/0x300
[ 2720.390758]  do_mmap+0x9d9/0x1020
[ 2720.390765]  ? vm_mmap_pgoff+0x141/0x330
[ 2720.390770]  ? __pfx_do_mmap+0x10/0x10
[ 2720.390776]  ? __pfx_down_write_killable+0x10/0x10
[ 2720.390785]  vm_mmap_pgoff+0x1b1/0x330
[ 2720.390792]  ? __pfx_vm_mmap_pgoff+0x10/0x10
[ 2720.390797]  ? __fget_files+0x1a4/0x2e0
[ 2720.390803]  ? __fget_files+0x1ae/0x2e0
[ 2720.390809]  ksys_mmap_pgoff+0x2ec/0x440
[ 2720.390815]  do_syscall_64+0x92/0x180
[ 2720.390820]  ? mark_lock+0x78/0x860
[ 2720.390826]  ? __lock_acquire+0xcff/0x1fc0
[ 2720.390834]  ? find_held_lock+0x34/0x120
[ 2720.390838]  ? kvm_sched_clock_read+0x11/0x20
[ 2720.390842]  ? local_clock_noinstr+0xd/0xe0
[ 2720.390846]  ? __lock_release.isra.0+0x4ba/0x9f0
[ 2720.390851]  ? __pfx___lock_release.isra.0+0x10/0x10
[ 2720.390857]  ? __pfx_do_raw_spin_trylock+0x10/0x10
[ 2720.390864]  ? _raw_spin_unlock_irq+0x28/0x50
[ 2720.390869]  ? lockdep_hardirqs_on+0x78/0x100
[ 2720.390873]  ? _raw_spin_unlock_irq+0x33/0x50
[ 2720.390877]  ? __x64_sys_rt_sigprocmask+0x230/0x390
[ 2720.390883]  ? __pfx___x64_sys_rt_sigprocmask+0x10/0x10
[ 2720.390888]  ? ktime_get_coarse_real_ts64+0x12a/0x190
[ 2720.390895]  ? rcu_is_watching+0x15/0xb0
[ 2720.390900]  ? lockdep_hardirqs_on_prepare+0x171/0x400
[ 2720.390904]  ? do_syscall_64+0x9e/0x180
[ 2720.390908]  ? lockdep_hardirqs_on+0x78/0x100
[ 2720.390912]  ? do_syscall_64+0x9e/0x180
[ 2720.390917]  ? rcu_is_watching+0x15/0xb0
[ 2720.390922]  ? clear_bhb_loop+0x25/0x80
[ 2720.390927]  ? clear_bhb_loop+0x25/0x80
[ 2720.390932]  entry_SYSCALL_64_after_hwframe+0x76/0x7e
[ 2720.390937] RIP: 0033:0x7f61aa0a8b36
[ 2720.390948] Code: 00 00 00 90 f3 0f 1e fa 41 f7 c1 ff 0f 00 00 75
2b 55 89 cd 53 48 89 fb 48 85 ff 74 37 41 89 ea 48 89 df b8 09 00 00
00 0f 05 <48> 3d 00 f0 ff ff 77 72 5b 5d c3 0f 1f 80 00 00 00 00 48 8b
05 a1
[ 2720.390954] RSP: 002b:00007ffffc0efa68 EFLAGS: 00000246 ORIG_RAX:
0000000000000009
[ 2720.390963] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f61aa0a8b36
[ 2720.390966] RDX: 0000000000000002 RSI: 0000000000100000 RDI: 0000000000000000
[ 2720.390969] RBP: 0000000000000001 R08: 0000000000000003 R09: 000000001ec00000
[ 2720.390971] R10: 0000000000000001 R11: 0000000000000246 R12: 0000000000000002
[ 2720.390980] R13: 000000001ec00000 R14: 0000000000000000 R15: 000055844a355ac0
[ 2720.390991]  </TASK>

Powered by blists - more mailing lists