On Thu, 13 Dec 2012 01:05:15 +0000 "Myklebust, Trond"
<Trond.Myklebust@xxxxxxxxxx> wrote:
> On Thu, 2012-12-13 at 11:05 +1100, NeilBrown wrote:
> > Hi,
> >
> > I recently managed to get the following stack trace, though I haven't been
> > able to reproduce it.
> >
> > Dec 12 17:17:04 hp kernel: [22684.894434] BUG: unable to handle kernel NULL pointer dereference at 0000000000000310
> > Dec 12 17:17:04 hp kernel: [22684.894490] IP: [<ffffffff813248f9>] nlmclnt_done+0x9/0x30
> > Dec 12 17:17:04 hp kernel: [22684.894529] PGD 13cc5a067 PUD 13ca0a067 PMD 0
> > Dec 12 17:17:04 hp kernel: [22684.894593] Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC
> > Dec 12 17:17:04 hp kernel: [22684.894660] Modules linked in:
> > Dec 12 17:17:04 hp kernel: [22684.894691] CPU 0
> > Dec 12 17:17:04 hp kernel: [22684.894708] Pid: 6874, comm: ls Not tainted 3.7.0-rc1+ #323 HP ProLiant ML310 G3
> > Dec 12 17:17:04 hp kernel: [22684.894745] RIP: 0010:[<ffffffff813248f9>] [<ffffffff813248f9>] nlmclnt_done+0x9/0x30
> > Dec 12 17:17:04 hp kernel: [22684.894783] RSP: 0018:ffff880139d59958 EFLAGS: 00010292
> > Dec 12 17:17:04 hp kernel: [22684.894804] RAX: 0000000000000000 RBX: ffff88013760e7f0 RCX: 0000000000000000
> > Dec 12 17:17:04 hp kernel: [22684.894829] RDX: 0000000000000046 RSI: 0000000000000001 RDI: 0000000000000000
> > Dec 12 17:17:04 hp kernel: [22684.894852] RBP: ffff880139d59968 R08: 0000000000000000 R09: 0000000000000000
> > Dec 12 17:17:04 hp kernel: [22684.894878] R10: 0000000000000001 R11: 0000000000000000 R12: dead000000200200
> > Dec 12 17:17:04 hp kernel: [22684.894900] R13: ffff88013bdfaed0 R14: dead000000200200 R15: ffff880139fde7f0
> > Dec 12 17:17:04 hp kernel: [22684.894924] FS: 00007f279bbf57c0(0000) GS:ffff880147c00000(0000) knlGS:0000000000000000
> > Dec 12 17:17:04 hp kernel: [22684.894947] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
> > Dec 12 17:17:04 hp kernel: [22684.894968] CR2: 0000000000000310 CR3: 0000000137ea1000 CR4: 00000000000007f0
> > Dec 12 17:17:04 hp kernel: [22684.894995] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> > Dec 12 17:17:04 hp kernel: [22684.895017] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
> > Dec 12 17:17:04 hp kernel: [22684.895039] Process ls (pid: 6874, threadinfo ffff880139d58000, task ffff88013bbdc050)
> > Dec 12 17:17:04 hp kernel: [22684.895061] Stack:
> > Dec 12 17:17:04 hp kernel: [22684.895079] ffff880139d59978 ffff88013760e7f0 ffff880139d59978 ffffffff812c41ff
> > Dec 12 17:17:04 hp kernel: [22684.895148] ffff880139d599b8 ffffffff812c4eab ffffffff812c4d9b ffff88013760e7f0
> > Dec 12 17:17:04 hp kernel: [22684.895220] ffff88013760e7f0 ffffffffffffff8c ffff88013e3a7ef0 00000000ffffff8c
> > Dec 12 17:17:04 hp kernel: [22684.895287] Call Trace:
> > Dec 12 17:17:04 hp kernel: [22684.895312] [<ffffffff812c41ff>] nfs_destroy_server+0x1f/0x30
> > Dec 12 17:17:04 hp kernel: [22684.895337] [<ffffffff812c4eab>] nfs_free_server+0x13b/0x200
> > Dec 12 17:17:04 hp kernel: [22684.895362] [<ffffffff812c4d9b>] ? nfs_free_server+0x2b/0x200
> > Dec 12 17:17:04 hp kernel: [22684.895386] [<ffffffff812c511b>] nfs_clone_server+0x1ab/0x250
> > Dec 12 17:17:04 hp kernel: [22684.895411] [<ffffffff812dc498>] nfs3_clone_server+0x18/0x50
> > Dec 12 17:17:04 hp kernel: [22684.895437] [<ffffffff812ce202>] nfs_xdev_mount+0x82/0x120
> > Dec 12 17:17:04 hp kernel: [22684.895462] [<ffffffff812ce300>] ? nfs_set_super+0x60/0x60
> > Dec 12 17:17:04 hp kernel: [22684.895486] [<ffffffff812cdd60>] ? nfs_set_sb_security+0x10/0x10
> > Dec 12 17:17:04 hp kernel: [22684.895512] [<ffffffff8116a3bb>] mount_fs+0x1b/0xd0
> > Dec 12 17:17:04 hp kernel: [22684.895545] [<ffffffff811843bf>] vfs_kern_mount+0x6f/0x110
> > Dec 12 17:17:04 hp kernel: [22684.895569] [<ffffffff812d8f62>] nfs_do_submount+0xa2/0x150
> > Dec 12 17:17:04 hp kernel: [22684.895593] [<ffffffff812d908e>] nfs_submount+0x7e/0xa0
> > Dec 12 17:17:04 hp kernel: [22684.895617] [<ffffffff812d917c>] nfs_d_automount+0xcc/0x1c0
> > Dec 12 17:17:04 hp kernel: [22684.895643] [<ffffffff811706c0>] follow_managed+0x150/0x310
> > Dec 12 17:17:04 hp kernel: [22684.895668] [<ffffffff811710e1>] lookup_fast+0x1c1/0x310
> > Dec 12 17:17:04 hp kernel: [22684.895693] [<ffffffff811744ac>] do_last.isra.57+0x17c/0xc50
> > Dec 12 17:17:04 hp kernel: [22684.895717] [<ffffffff811718a3>] ? inode_permission+0x13/0x50
> > Dec 12 17:17:04 hp kernel: [22684.895741] [<ffffffff81171d6d>] ? link_path_walk+0x22d/0x8f0
> > Dec 12 17:17:04 hp kernel: [22684.895766] [<ffffffff81175033>] path_openat.isra.58+0xb3/0x4c0
> > Dec 12 17:17:04 hp kernel: [22684.895791] [<ffffffff811716db>] ? getname_flags+0x2b/0x110
> > Dec 12 17:17:04 hp kernel: [22684.895816] [<ffffffff8118285f>] ? __alloc_fd+0x2f/0x130
> > Dec 12 17:17:04 hp kernel: [22684.895842] [<ffffffff8117579c>] do_filp_open+0x3c/0x90
> > Dec 12 17:17:04 hp kernel: [22684.895866] [<ffffffff81182909>] ? __alloc_fd+0xd9/0x130
> > Dec 12 17:17:04 hp kernel: [22684.895891] [<ffffffff8116614f>] do_sys_open+0xef/0x1d0
> >
> >
> > The problem is that nfs_destroy_server is calling nlmclnt_done(NULL).
> > This can happen if nfs_clone_server() is called on a v2/v3 mount but gets an
> > error between
> >
> > server->destroy = source->destroy;
> >
> > (which sets server->destroy to nfs_destroy_dever without setting
> > server->nlm_host) and
> >
> > error = nfs_start_lockd(server);
> >
> > (which sets both server->destroy and server->nlm_host).
> >
> > If this happens then nfs_free_server() calls ->destroy() which crashes as
> > shown.
> >
> > Maybe this patch?
> >
> > Signed-off-by: NeilBrown <neilb@xxxxxxx>
> > diff --git a/fs/nfs/client.c b/fs/nfs/client.c
> > index 8b39a42..b6603bb 100644
> > --- a/fs/nfs/client.c
> > +++ b/fs/nfs/client.c
> > @@ -617,7 +617,8 @@ static void nfs_destroy_server(struct nfs_server *server)
> > {
> > if (!(server->flags & NFS_MOUNT_LOCAL_FLOCK) ||
> > !(server->flags & NFS_MOUNT_LOCAL_FCNTL))
> > - nlmclnt_done(server->nlm_host);
> > + if (server->nlm_host)
> > + nlmclnt_done(server->nlm_host);
> > }
>
> Hmm... Do we need all those tests of server->flags above if we can just
> check server->nlm_host for a NULL value?
>
No, you are right. nlm_host will only be non-NULL if the one of the flags
is set, so the test is pointless.
NeilBrown
diff --git a/fs/nfs/client.c b/fs/nfs/client.c
index 8b39a42..5e8d24d 100644
--- a/fs/nfs/client.c
+++ b/fs/nfs/client.c
@@ -615,8 +615,7 @@ EXPORT_SYMBOL_GPL(nfs_create_rpc_client);
*/
static void nfs_destroy_server(struct nfs_server *server)
{
- if (!(server->flags & NFS_MOUNT_LOCAL_FLOCK) ||
- !(server->flags & NFS_MOUNT_LOCAL_FCNTL))
+ if (server->nlm_host)
nlmclnt_done(server->nlm_host);
}
Attachment:
signature.asc
Description: PGP signature
