| Message ID | 20140416040337.10604.86740.stgit@notabene.brown (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Wed, Apr 16, 2014 at 02:03:37PM +1000, NeilBrown wrote: > namespace_sem can be taken while various i_mutex locks are held, so we > need to avoid reclaim from blocking on an FS (particularly loop-back > NFS). I would really prefer to deal with that differently - by explicit change of gfp_t arguments of allocators. The thing is, namespace_sem is held *only* over allocations, and not a lot of them, at that - only mnt_alloc_id(), mnt_alloc_group_id(), alloc_vfsmnt() and new_mountpoint(). That is all that is allowed. Again, actual work with filesystems (setup, shutdown, remount, pathname resolution, etc.) is all done outside of namespace_sem; it's held only for manipulations of fs/{namespace,pnode}.c data structures and the only reason it isn't a spinlock is that we need to do some allocations. So I'd rather slap GFP_NOFS on those few allocations... -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/fs/namespace.c b/fs/namespace.c index 2ffc5a2905d4..83dcd5083dbb 100644 --- a/fs/namespace.c +++ b/fs/namespace.c @@ -63,6 +63,7 @@ static struct hlist_head *mount_hashtable __read_mostly; static struct hlist_head *mountpoint_hashtable __read_mostly; static struct kmem_cache *mnt_cache __read_mostly; static DECLARE_RWSEM(namespace_sem); +static unsigned long namespace_sem_pflags; /* /sys/fs */ struct kobject *fs_kobj; @@ -1196,6 +1197,8 @@ static void namespace_unlock(void) struct mount *mnt; struct hlist_head head = unmounted; + current_restore_flags_nested(&namespace_sem_pflags, PF_FSTRANS); + if (likely(hlist_empty(&head))) { up_write(&namespace_sem); return; @@ -1220,6 +1223,7 @@ static void namespace_unlock(void) static inline void namespace_lock(void) { down_write(&namespace_sem); + current_set_flags_nested(&namespace_sem_pflags, PF_FSTRANS); } /*
namespace_sem can be taken while various i_mutex locks are held, so we need to avoid reclaim from blocking on an FS (particularly loop-back NFS). A memory allocation happens under namespace_sem at least in: [<ffffffff8119d16f>] kmem_cache_alloc+0x4f/0x290 [<ffffffff811c2fff>] alloc_vfsmnt+0x1f/0x1d0 [<ffffffff811c339a>] clone_mnt+0x2a/0x310 [<ffffffff811c57e3>] copy_tree+0x53/0x380 [<ffffffff811c6aef>] copy_mnt_ns+0x7f/0x280 [<ffffffff810c16fc>] create_new_namespaces+0x5c/0x190 [<ffffffff810c1ab9>] unshare_nsproxy_namespaces+0x59/0x90 So set PF_FSTRANS in namespace_lock() and restore in namespace_unlock(). Signed-off-by: NeilBrown <neilb@suse.de> --- fs/namespace.c | 4 ++++ 1 file changed, 4 insertions(+) -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html