]> www.infradead.org Git - users/hch/misc.git/commitdiff
NFSD: fix race between nfsd registration and exports_proc
authorManinder Singh <maninder1.s@samsung.com>
Thu, 6 Mar 2025 09:20:07 +0000 (14:50 +0530)
committerChuck Lever <chuck.lever@oracle.com>
Sun, 11 May 2025 23:48:23 +0000 (19:48 -0400)
As of now nfsd calls create_proc_exports_entry() at start of init_nfsd
and cleanup by remove_proc_entry() at last of exit_nfsd.

Which causes kernel OOPs if there is race between below 2 operations:
(i) exportfs -r
(ii) mount -t nfsd none /proc/fs/nfsd

for 5.4 kernel ARM64:

CPU 1:
el1_irq+0xbc/0x180
arch_counter_get_cntvct+0x14/0x18
running_clock+0xc/0x18
preempt_count_add+0x88/0x110
prep_new_page+0xb0/0x220
get_page_from_freelist+0x2d8/0x1778
__alloc_pages_nodemask+0x15c/0xef0
__vmalloc_node_range+0x28c/0x478
__vmalloc_node_flags_caller+0x8c/0xb0
kvmalloc_node+0x88/0xe0
nfsd_init_net+0x6c/0x108 [nfsd]
ops_init+0x44/0x170
register_pernet_operations+0x114/0x270
register_pernet_subsys+0x34/0x50
init_nfsd+0xa8/0x718 [nfsd]
do_one_initcall+0x54/0x2e0

CPU 2 :
Unable to handle kernel NULL pointer dereference at virtual address 0000000000000010

PC is at : exports_net_open+0x50/0x68 [nfsd]

Call trace:
exports_net_open+0x50/0x68 [nfsd]
exports_proc_open+0x2c/0x38 [nfsd]
proc_reg_open+0xb8/0x198
do_dentry_open+0x1c4/0x418
vfs_open+0x38/0x48
path_openat+0x28c/0xf18
do_filp_open+0x70/0xe8
do_sys_open+0x154/0x248

Sometimes it crashes at exports_net_open() and sometimes cache_seq_next_rcu().

and same is happening on latest 6.14 kernel as well:

[    0.000000] Linux version 6.14.0-rc5-next-20250304-dirty
...
[  285.455918] Unable to handle kernel paging request at virtual address 00001f4800001f48
...
[  285.464902] pc : cache_seq_next_rcu+0x78/0xa4
...
[  285.469695] Call trace:
[  285.470083]  cache_seq_next_rcu+0x78/0xa4 (P)
[  285.470488]  seq_read+0xe0/0x11c
[  285.470675]  proc_reg_read+0x9c/0xf0
[  285.470874]  vfs_read+0xc4/0x2fc
[  285.471057]  ksys_read+0x6c/0xf4
[  285.471231]  __arm64_sys_read+0x1c/0x28
[  285.471428]  invoke_syscall+0x44/0x100
[  285.471633]  el0_svc_common.constprop.0+0x40/0xe0
[  285.471870]  do_el0_svc_compat+0x1c/0x34
[  285.472073]  el0_svc_compat+0x2c/0x80
[  285.472265]  el0t_32_sync_handler+0x90/0x140
[  285.472473]  el0t_32_sync+0x19c/0x1a0
[  285.472887] Code: f9400885 93407c23 937d7c27 11000421 (f86378a3)
[  285.473422] ---[ end trace 0000000000000000 ]---

It reproduced simply with below script:
while [ 1 ]
do
/exportfs -r
done &

while [ 1 ]
do
insmod /nfsd.ko
mount -t nfsd none /proc/fs/nfsd
umount /proc/fs/nfsd
rmmod nfsd
done &

So exporting interfaces to user space shall be done at last and
cleanup at first place.

With change there is no Kernel OOPs.

Co-developed-by: Shubham Rana <s9.rana@samsung.com>
Signed-off-by: Shubham Rana <s9.rana@samsung.com>
Signed-off-by: Maninder Singh <maninder1.s@samsung.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Cc: stable@vger.kernel.org
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
fs/nfsd/nfsctl.c

index d773481bcf1025681402cdbaa78837f073e3eb6f..f9763ced743d2497e88f39182df53f6d31d815ab 100644 (file)
@@ -2291,12 +2291,9 @@ static int __init init_nfsd(void)
        if (retval)
                goto out_free_pnfs;
        nfsd_lockd_init();      /* lockd->nfsd callbacks */
-       retval = create_proc_exports_entry();
-       if (retval)
-               goto out_free_lockd;
        retval = register_pernet_subsys(&nfsd_net_ops);
        if (retval < 0)
-               goto out_free_exports;
+               goto out_free_lockd;
        retval = register_cld_notifier();
        if (retval)
                goto out_free_subsys;
@@ -2307,12 +2304,17 @@ static int __init init_nfsd(void)
        if (retval)
                goto out_free_nfsd4;
        retval = genl_register_family(&nfsd_nl_family);
+       if (retval)
+               goto out_free_filesystem;
+       retval = create_proc_exports_entry();
        if (retval)
                goto out_free_all;
        nfsd_localio_ops_init();
 
        return 0;
 out_free_all:
+       genl_unregister_family(&nfsd_nl_family);
+out_free_filesystem:
        unregister_filesystem(&nfsd_fs_type);
 out_free_nfsd4:
        nfsd4_destroy_laundry_wq();
@@ -2320,9 +2322,6 @@ out_free_cld:
        unregister_cld_notifier();
 out_free_subsys:
        unregister_pernet_subsys(&nfsd_net_ops);
-out_free_exports:
-       remove_proc_entry("fs/nfs/exports", NULL);
-       remove_proc_entry("fs/nfs", NULL);
 out_free_lockd:
        nfsd_lockd_shutdown();
        nfsd_drc_slab_free();
@@ -2335,14 +2334,14 @@ out_free_slabs:
 
 static void __exit exit_nfsd(void)
 {
+       remove_proc_entry("fs/nfs/exports", NULL);
+       remove_proc_entry("fs/nfs", NULL);
        genl_unregister_family(&nfsd_nl_family);
        unregister_filesystem(&nfsd_fs_type);
        nfsd4_destroy_laundry_wq();
        unregister_cld_notifier();
        unregister_pernet_subsys(&nfsd_net_ops);
        nfsd_drc_slab_free();
-       remove_proc_entry("fs/nfs/exports", NULL);
-       remove_proc_entry("fs/nfs", NULL);
        nfsd_lockd_shutdown();
        nfsd4_free_slabs();
        nfsd4_exit_pnfs();