nfs: avoid race that crashes nfs_init_commit
commit691c507ec01fa0cab2a9cfb5bd4398ddd5480a8a
authorWeston Andros Adamson <dros@monkey.org>
Wed, 25 May 2016 14:07:23 +0000 (25 10:07 -0400)
committerSasha Levin <sasha.levin@oracle.com>
Mon, 11 Jul 2016 00:20:05 +0000 (10 20:20 -0400)
tree2f4a78609c22ea92f15307947121a865e62522d2
parent94d06a437eb0ad1a492ed75f929d82715b6c05f8
nfs: avoid race that crashes nfs_init_commit

[ Upstream commit ade8febde0271513360bac44883dbebad44276c3 ]

Since the patch "NFS: Allow multiple commit requests in flight per file"
we can run multiple simultaneous commits on the same inode.  This
introduced a race over collecting pages to commit that made it possible
to call nfs_init_commit() with an empty list - which causes crashes like
the one below.

The fix is to catch this race and avoid calling nfs_init_commit and
initiate_commit when there is no work to do.

Here is the crash:

[600522.076832] BUG: unable to handle kernel NULL pointer dereference at 0000000000000040
[600522.078475] IP: [<ffffffffa0479e72>] nfs_init_commit+0x22/0x130 [nfs]
[600522.078745] PGD 4272b1067 PUD 4272cb067 PMD 0
[600522.078972] Oops: 0000 [#1] SMP
[600522.079204] Modules linked in: nfsv3 nfs_layout_flexfiles rpcsec_gss_krb5 nfsv4 dns_resolver nfs fscache dcdbas ip6t_rpfilter ip6t_REJECT nf_reject_ipv6 xt_conntrack ebtable_nat ebtable_broute bridge stp llc ebtable_filter ebtables ip6table_nat nf_conntrack_ipv6 nf_defrag_ipv6 nf_nat_ipv6 ip6table_mangle ip6table_security ip6table_raw ip6table_filter ip6_tables iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 nf_nat nf_conntrack iptable_mangle iptable_security iptable_raw vmw_vsock_vmci_transport vsock bonding ipmi_devintf ipmi_msghandler coretemp crct10dif_pclmul crc32_pclmul ghash_clmulni_intel ppdev vmw_balloon parport_pc parport acpi_cpufreq vmw_vmci i2c_piix4 shpchp nfsd auth_rpcgss nfs_acl lockd grace sunrpc xfs libcrc32c vmwgfx drm_kms_helper ttm drm crc32c_intel serio_raw vmxnet3
[600522.081380]  vmw_pvscsi ata_generic pata_acpi
[600522.081809] CPU: 3 PID: 15667 Comm: /usr/bin/python Not tainted 4.1.9-100.pd.88.el7.x86_64 #1
[600522.082281] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 09/30/2014
[600522.082814] task: ffff8800bbbfa780 ti: ffff88042ae84000 task.ti: ffff88042ae84000
[600522.083378] RIP: 0010:[<ffffffffa0479e72>]  [<ffffffffa0479e72>] nfs_init_commit+0x22/0x130 [nfs]
[600522.083973] RSP: 0018:ffff88042ae87438  EFLAGS: 00010246
[600522.084571] RAX: 0000000000000000 RBX: ffff880003485e40 RCX: ffff88042ae87588
[600522.085188] RDX: 0000000000000000 RSI: ffff88042ae874b0 RDI: ffff880003485e40
[600522.085756] RBP: ffff88042ae87448 R08: ffff880003486010 R09: ffff88042ae874b0
[600522.086332] R10: 0000000000000000 R11: 0000000000000005 R12: ffff88042ae872d0
[600522.086905] R13: ffff88042ae874b0 R14: ffff880003485e40 R15: ffff88042704c840
[600522.087484] FS:  00007f4728ff2740(0000) GS:ffff88043fd80000(0000) knlGS:0000000000000000
[600522.088070] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[600522.088663] CR2: 0000000000000040 CR3: 000000042b6aa000 CR4: 00000000001406e0
[600522.089327] Stack:
[600522.089926]  0000000000000001 ffff88042ae87588 ffff88042ae874f8 ffffffffa04f09fa
[600522.090549]  0000000000017840 0000000000017840 ffff88042ae87588 ffff8803258d9930
[600522.091169]  ffff88042ae87578 ffffffffa0563d80 0000000000000000 ffff88042704c840
[600522.091789] Call Trace:
[600522.092420]  [<ffffffffa04f09fa>] pnfs_generic_commit_pagelist+0x1da/0x320 [nfsv4]
[600522.093052]  [<ffffffffa0563d80>] ? ff_layout_commit_prepare_v3+0x30/0x30 [nfs_layout_flexfiles]
[600522.093696]  [<ffffffffa0562645>] ff_layout_commit_pagelist+0x15/0x20 [nfs_layout_flexfiles]
[600522.094359]  [<ffffffffa047bc78>] nfs_generic_commit_list+0xe8/0x120 [nfs]
[600522.095032]  [<ffffffffa047bd6a>] nfs_commit_inode+0xba/0x110 [nfs]
[600522.095719]  [<ffffffffa046ac54>] nfs_release_page+0x44/0xd0 [nfs]
[600522.096410]  [<ffffffff811a8122>] try_to_release_page+0x32/0x50
[600522.097109]  [<ffffffff811bd4f1>] shrink_page_list+0x961/0xb30
[600522.097812]  [<ffffffff811bdced>] shrink_inactive_list+0x1cd/0x550
[600522.098530]  [<ffffffff811bea65>] shrink_lruvec+0x635/0x840
[600522.099250]  [<ffffffff811bed60>] shrink_zone+0xf0/0x2f0
[600522.099974]  [<ffffffff811bf312>] do_try_to_free_pages+0x192/0x470
[600522.100709]  [<ffffffff811bf6ca>] try_to_free_pages+0xda/0x170
[600522.101464]  [<ffffffff811b2198>] __alloc_pages_nodemask+0x588/0x970
[600522.102235]  [<ffffffff811fbbd5>] alloc_pages_vma+0xb5/0x230
[600522.103000]  [<ffffffff813a1589>] ? cpumask_any_but+0x39/0x50
[600522.103774]  [<ffffffff811d6115>] wp_page_copy.isra.55+0x95/0x490
[600522.104558]  [<ffffffff810e3438>] ? __wake_up+0x48/0x60
[600522.105357]  [<ffffffff811d7d3b>] do_wp_page+0xab/0x4f0
[600522.106137]  [<ffffffff810a1bbb>] ? release_task+0x36b/0x470
[600522.106902]  [<ffffffff8126dbd7>] ? eventfd_ctx_read+0x67/0x1c0
[600522.107659]  [<ffffffff811da2a8>] handle_mm_fault+0xc78/0x1900
[600522.108431]  [<ffffffff81067ef1>] __do_page_fault+0x181/0x420
[600522.109173]  [<ffffffff811446a6>] ? __audit_syscall_exit+0x1e6/0x280
[600522.109893]  [<ffffffff810681c0>] do_page_fault+0x30/0x80
[600522.110594]  [<ffffffff81024f36>] ? syscall_trace_leave+0xc6/0x120
[600522.111288]  [<ffffffff81790a58>] page_fault+0x28/0x30
[600522.111947] Code: 5d c3 0f 1f 80 00 00 00 00 0f 1f 44 00 00 55 4c 8d 87 d0 01 00 00 48 89 e5 53 48 89 fb 48 83 ec 08 4c 8b 0e 49 8b 41 18 4c 39 ce <48> 8b 40 40 4c 8b 50 30 74 24 48 8b 87 d0 01 00 00 48 8b 7e 08
[600522.113343] RIP  [<ffffffffa0479e72>] nfs_init_commit+0x22/0x130 [nfs]
[600522.114003]  RSP <ffff88042ae87438>
[600522.114636] CR2: 0000000000000040

Fixes: af7cf057 (NFS: Allow multiple commit requests in flight per file)
CC: stable@vger.kernel.org
Signed-off-by: Weston Andros Adamson <dros@primarydata.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
fs/nfs/pnfs_nfs.c
fs/nfs/write.c