hang on shrink_dcache_sb at remount of directories inside container

hang on shrink_dcache_sb at remount of directories inside container

[tiandi huang]

Hi, guys!
Recently, We have met a problem that systemd insides docker always
hang at below stack, stays inside shrink_dcache_sb as long as hours.
 __dentry_kill+0x124/0x170
shrink_dentry_list+0x64/0xb0
shrink_dcache_sb+0xb2/0x140
reconfigure_super+0x85/0x220
do_mount+0x90e/0x980
ksys_mount+0xb6/0xd0
 __x64_sys_mount+0x25/0x30
do_syscall_64+0x47/0x140
entry_SYSCALL_64_after_hwframe+0x44/0xa9

We run many containers onside /data which is a xfs file system.
Directories on /data are bind mounted inside containers, sometimes,
systemd inside containers may do remount operations with these bind
directories. Each remount would cause a reconfigure on /data and
shrink_dcache_sb is called to shrink the s_dentry_lru list, as data is
shared by all containers, processes may still produce dentries on this
list, which may cause a long time to wait shrinking ends.
So, We mailed to wonder  is it possible to skip this shrink operation
for containers as lower xfs file system is still running, or is it
possible to optimize shrink_dcache_sb function to scan only once of
the dentry lru list, do not wait the list to become empty.
We appreciate for a response, Thanks!