]> www.infradead.org Git - users/willy/xarray.git/commit
kernfs: use dumber locking for kernfs_find_and_get_node_by_ino()
authorTejun Heo <tj@kernel.org>
Mon, 4 Nov 2019 23:54:29 +0000 (15:54 -0800)
committerTejun Heo <tj@kernel.org>
Tue, 12 Nov 2019 16:18:03 +0000 (08:18 -0800)
commitb680b08171ebf890a4ebb7f82ada9959f4534ade
treee4dbcd63eb466e0efb09e76239057a2e2a03faab
parentdb53c73a8b5db120cb741d7d932cdf831a576e8f
kernfs: use dumber locking for kernfs_find_and_get_node_by_ino()

kernfs_find_and_get_node_by_ino() uses RCU protection.  It's currently
a bit buggy because it can look up a node which hasn't been activated
yet and thus may end up exposing a node that the kernfs user is still
prepping.

While it can be fixed by pushing it further in the current direction,
it's already complicated and isn't clear whether the complexity is
justified.  The main use of kernfs_find_and_get_node_by_ino() is for
exportfs operations.  They aren't super hot and all the follow-up
operations (e.g. mapping to path) use normal locking anyway.

Let's switch to a dumber locking scheme and protect the lookup with
kernfs_idr_lock.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
fs/kernfs/dir.c
fs/kernfs/mount.c