From: Ian Kent <raven@themaw.net>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: autofs mailing list <autofs@vger.kernel.org>,
Kernel Mailing List <linux-kernel@vger.kernel.org>,
"Eric W. Biederman" <ebiederm@xmission.com>,
linux-fsdevel <linux-fsdevel@vger.kernel.org>,
Omar Sandoval <osandov@osandov.com>,
Al Viro <viro@ZenIV.linux.org.uk>
Subject: [PATCH 7/7] vfs - make may_umount_tree() mount propogation aware
Date: Mon, 28 Nov 2016 10:12:19 +0800 [thread overview]
Message-ID: <148029913966.27779.10094416563067823851.stgit@pluto.themaw.net> (raw)
In-Reply-To: <148029910861.27779.4517883721395202453.stgit@pluto.themaw.net>
From: Ian Kent <ikent@redhat.com>
Now that autofs has namespace aware mounted checks the expire needs
changes to make it aware of mount propagation.
When checking for expiration may_umount_tree() checks only if the
given mount is in use. This leads to a callback to the automount
daemon to umount the mount which will fail if any propagated mounts
are in use.
To avoid this unnecessary call back may_umount_tree() needs to check
propagated mount trees also.
Signed-off-by: Ian Kent <raven@themaw.net>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Omar Sandoval <osandov@osandov.com>
---
fs/autofs4/expire.c | 4 +--
fs/namespace.c | 71 +++++++++++++++++++++++++++++++++++++++++++++------
fs/pnode.c | 3 +-
fs/pnode.h | 1 +
4 files changed, 66 insertions(+), 13 deletions(-)
diff --git a/fs/autofs4/expire.c b/fs/autofs4/expire.c
index 57725d4..d44ae32 100644
--- a/fs/autofs4/expire.c
+++ b/fs/autofs4/expire.c
@@ -52,7 +52,7 @@ static int autofs4_mount_busy(struct vfsmount *mnt, struct dentry *dentry)
goto done;
}
- /* Update the expiry counter if fs is busy */
+ /* Update the expiry counter if fs is busy in any namespace */
if (!may_umount_tree(path.mnt)) {
struct autofs_info *ino;
@@ -191,7 +191,7 @@ static int autofs4_direct_busy(struct vfsmount *mnt,
{
pr_debug("top %p %pd\n", top, top);
- /* If it's busy update the expiry counters */
+ /* If it's busy in any namespace update the expiry counters */
if (!may_umount_tree(mnt)) {
struct autofs_info *ino;
diff --git a/fs/namespace.c b/fs/namespace.c
index da1cd87..092d75c 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -1311,6 +1311,33 @@ const struct seq_operations mounts_op = {
};
#endif /* CONFIG_PROC_FS */
+struct mnt_tree_refs {
+ struct mount *root;
+ unsigned int refs;
+ unsigned int min_refs;
+};
+
+static void mnt_get_tree_refs(struct mnt_tree_refs *mtr)
+{
+ struct mount *mnt = mtr->root;
+ struct mount *p;
+
+ /*
+ * Each propagated tree contribues 2 * #mounts - 1 to
+ * the minimal reference count. But when a mount is
+ * umounted and connected the mount doesn't hold a
+ * reference to its parent so it contributes a single
+ * reference.
+ */
+ for (p = mnt; p; p = next_mnt(p, mnt)) {
+ mtr->refs += mnt_get_count(p);
+ if (p == mnt || p->mnt.mnt_flags & MNT_UMOUNT)
+ mtr->min_refs++;
+ else
+ mtr->min_refs += 2;
+ }
+}
+
/**
* may_umount_tree - check if a mount tree is busy
* @mnt: root of mount tree
@@ -1322,25 +1349,51 @@ const struct seq_operations mounts_op = {
int may_umount_tree(struct vfsmount *m)
{
struct mount *mnt = real_mount(m);
- int actual_refs = 0;
- int minimum_refs = 0;
- struct mount *p;
+ struct mount *parent = mnt->mnt_parent;
+ struct mnt_tree_refs mtr;
+ struct mount *p, *child;
+
BUG_ON(!m);
- /* write lock needed for mnt_get_count */
+ down_read(&namespace_sem);
lock_mount_hash();
- for (p = mnt; p; p = next_mnt(p, mnt)) {
- actual_refs += mnt_get_count(p);
- minimum_refs += 2;
+
+ mtr.root = mnt;
+ mtr.refs = 0;
+ mtr.min_refs = 0;
+
+ mnt_get_tree_refs(&mtr);
+ /*
+ * Caller holds a mount reference so minimum references
+ * to the tree at mnt is one greater than the minumum
+ * references.
+ */
+ mtr.min_refs++;
+
+ /* The pnode.c propagation_next() function (as used below)
+ * returns each mount propogated from a given mount. Using
+ * the parent of mnt and matching the mnt->mnt_mountpoint
+ * gets the list of mounts propogated from mnt. To work
+ * out if the tree is in use (eg. open file or pwd) the
+ * reference counts of each of these mounts needs to be
+ * checked as well as mnt itself.
+ */
+ for (p = propagation_next(parent, parent); p;
+ p = propagation_next(p, parent)) {
+ child = __lookup_mnt_last(&p->mnt, mnt->mnt_mountpoint);
+ if (child) {
+ mtr.root = child;
+ mnt_get_tree_refs(&mtr);
+ }
}
unlock_mount_hash();
+ up_read(&namespace_sem);
- if (actual_refs > minimum_refs)
+ if (mtr.refs > mtr.min_refs)
return 0;
return 1;
}
-
EXPORT_SYMBOL(may_umount_tree);
/**
diff --git a/fs/pnode.c b/fs/pnode.c
index 234a9ac..ae6f95e 100644
--- a/fs/pnode.c
+++ b/fs/pnode.c
@@ -143,8 +143,7 @@ void change_mnt_propagation(struct mount *mnt, int type)
* vfsmount found while iterating with propagation_next() is
* a peer of one we'd found earlier.
*/
-static struct mount *propagation_next(struct mount *m,
- struct mount *origin)
+struct mount *propagation_next(struct mount *m, struct mount *origin)
{
/* are there any slaves of this mount? */
if (!IS_MNT_NEW(m) && !list_empty(&m->mnt_slave_list))
diff --git a/fs/pnode.h b/fs/pnode.h
index 550f5a8..2accd2d 100644
--- a/fs/pnode.h
+++ b/fs/pnode.h
@@ -38,6 +38,7 @@ static inline void set_mnt_shared(struct mount *mnt)
mnt->mnt.mnt_flags |= MNT_SHARED;
}
+struct mount *propagation_next(struct mount *, struct mount *);
void change_mnt_propagation(struct mount *, int);
int propagate_mnt(struct mount *, struct mountpoint *, struct mount *,
struct hlist_head *);
next prev parent reply other threads:[~2016-11-28 2:12 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-11-28 2:11 [PATCH 1/7] vfs - merge path_is_mountpoint() and path_is_mountpoint_rcu() Ian Kent
2016-11-28 2:11 ` [PATCH 2/7] autofs - make struct path const in autofs4_dir_open() Ian Kent
2016-11-28 2:11 ` [PATCH 3/7] autofs - change struct path to const in autofs4_expire_wait() and autofs4_wait() Ian Kent
2016-11-28 2:12 ` [PATCH 4/7] vfs - change struct path to const in d_manage() Ian Kent
2016-11-28 2:12 ` [PATCH 5/7] vfs - constify path parameter of path_has_submounts() Ian Kent
2016-11-28 2:12 ` [PATCH 6/7] autofs - dont hold spin lock over direct mount expire Ian Kent
2016-11-28 2:12 ` Ian Kent [this message]
2016-11-30 22:22 ` [PATCH 1/7] vfs - merge path_is_mountpoint() and path_is_mountpoint_rcu() Andrew Morton
2016-12-01 0:13 ` Ian Kent
2016-12-03 5:13 ` Al Viro
2016-12-03 23:29 ` Al Viro
2016-12-04 2:18 ` Ian Kent
2016-12-05 2:50 ` Ian Kent
2016-12-04 2:07 ` Ian Kent
2016-12-06 9:37 ` Ian Kent
2016-12-07 21:30 ` Eric W. Biederman
2016-12-08 3:29 ` Ian Kent
2016-12-08 4:28 ` Eric W. Biederman
2016-12-08 6:18 ` Ian Kent
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=148029913966.27779.10094416563067823851.stgit@pluto.themaw.net \
--to=raven@themaw.net \
--cc=akpm@linux-foundation.org \
--cc=autofs@vger.kernel.org \
--cc=ebiederm@xmission.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=osandov@osandov.com \
--cc=viro@ZenIV.linux.org.uk \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®