Lockdep pop on nfs4_proc_link

Lockdep pop on nfs4_proc_link

[Benjamin Coddington]

I just had this one pop up:

[  439.768290] ====================================
[  439.769027] WARNING: ln/1431 still has locks held!
[  439.769802] 5.3.0-rc5.d1abaeb3be #1 Not tainted
[  439.770586] ------------------------------------
[  439.771362] 3 locks held by ln/1431:
[  439.771925]  #0: 000000007cfdaccc (sb_writers#21){.+.+}, at: 
mnt_want_write+0x20/0x50
[  439.773238]  #1: 0000000082e25c62 
(&type->i_mutex_dir_key#11/1){+.+.}, at: filename_create+0x7f/0x160
[  439.774716]  #2: 00000000980e7edc 
(&sb->s_type->i_mutex_key#22){++++}, at: vfs_link+0x175/0x320
[  439.775996]
                stack backtrace:
[  439.776667] CPU: 1 PID: 1431 Comm: ln Not tainted 
5.3.0-rc5.d1abaeb3be #1
[  439.777653] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), 
BIOS 1.10.2-2.fc27 04/01/2014
[  439.778977] Call Trace:
[  439.779341]  dump_stack+0x85/0xcb
[  439.779837]  nfs4_handle_exception+0x1df/0x250 [nfsv4]
[  439.780585]  nfs4_proc_link+0x7a/0xa0 [nfsv4]
[  439.781234]  nfs_link+0x74/0x230 [nfs]
[  439.781818]  vfs_link+0x22b/0x320
[  439.782338]  do_linkat+0x212/0x330
[  439.782830]  __x64_sys_linkat+0x20/0x30
[  439.783396]  do_syscall_64+0x75/0x320
[  439.783935]  ? trace_hardirqs_off_thunk+0x1a/0x20
[  439.784612]  entry_SYSCALL_64_after_hwframe+0x49/0xbe
[  439.785393] RIP: 0033:0x7f5ed9fb742a
[  439.785943] Code: 48 8b 0d 59 7a 2c 00 f7 d8 64 89 01 48 83 c8 ff c3 
66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 49 89 ca b8 09 01 00 00 0f 
05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d 26 7a 2c 00 f7 d8 64 89 01 48
[  439.788627] RSP: 002b:00007ffec9cddd08 EFLAGS: 00000246 ORIG_RAX: 
0000000000000109
[  439.789722] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 
00007f5ed9fb742a
[  439.790759] RDX: 00000000ffffff9c RSI: 00007ffec9cdf601 RDI: 
00000000ffffff9c
[  439.791807] RBP: 00000000ffffff9c R08: 0000000000000000 R09: 
0000000000000000
[  439.792833] R10: 000055a57e85f430 R11: 0000000000000246 R12: 
000055a57e85f430
[  439.793869] R13: 0000000000000000 R14: 00007ffec9cdf601 R15: 
00000000ffffff9c

When we did commit "0688e64bc600 NFS: Allow signal interruption of
NFS4ERR_DELAYed operations", we moved from
freezable_schedule_timeout_killable_unsafe() to
freezable_schedule_timeout_interruptible().  There isn't a
freezable_schedule_timeout_interruptible_unsafe(), and the comments in
freezer.h suggest we shouldn't really add one, however those comments 
are
from the same commit that added the unsafe helpers for NFS.

So far, this is the only proc that has popped that warning for me, but 
that
is likely because I don't often see NFS4ERR_DELAY.

Should we create a freezable_schedule_timeout_interruptible_unsafe() for
now, or live with the future lockdep warnings?

Ben