aboutsummaryrefslogtreecommitdiff
path: root/fs/notify/inode_mark.c
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2010-08-10 11:39:13 -0700
committerLinus Torvalds <torvalds@linux-foundation.org>2010-08-10 11:39:13 -0700
commit8c8946f509a494769a8c602b5ed189df01917d39 (patch)
treedfd96bd6ca5ea6803c6d77f65ba37e04f78b2d3b /fs/notify/inode_mark.c
parent5f248c9c251c60af3403902b26e08de43964ea0b (diff)
parent1968f5eed54ce47bde488fd9a450912e4a2d7138 (diff)
downloadlinux-8c8946f509a494769a8c602b5ed189df01917d39.tar.gz
Merge branch 'for-linus' of git://git.infradead.org/users/eparis/notify
* 'for-linus' of git://git.infradead.org/users/eparis/notify: (132 commits) fanotify: use both marks when possible fsnotify: pass both the vfsmount mark and inode mark fsnotify: walk the inode and vfsmount lists simultaneously fsnotify: rework ignored mark flushing fsnotify: remove global fsnotify groups lists fsnotify: remove group->mask fsnotify: remove the global masks fsnotify: cleanup should_send_event fanotify: use the mark in handler functions audit: use the mark in handler functions dnotify: use the mark in handler functions inotify: use the mark in handler functions fsnotify: send fsnotify_mark to groups in event handling functions fsnotify: Exchange list heads instead of moving elements fsnotify: srcu to protect read side of inode and vfsmount locks fsnotify: use an explicit flag to indicate fsnotify_destroy_mark has been called fsnotify: use _rcu functions for mark list traversal fsnotify: place marks on object in order of group memory address vfs/fsnotify: fsnotify_close can delay the final work in fput fsnotify: store struct file not struct path ... Fix up trivial delete/modify conflict in fs/notify/inotify/inotify.c.
Diffstat (limited to 'fs/notify/inode_mark.c')
-rw-r--r--fs/notify/inode_mark.c331
1 files changed, 103 insertions, 228 deletions
diff --git a/fs/notify/inode_mark.c b/fs/notify/inode_mark.c
index 152b83ec005..33297c00506 100644
--- a/fs/notify/inode_mark.c
+++ b/fs/notify/inode_mark.c
@@ -16,72 +16,6 @@
* the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
*/
-/*
- * fsnotify inode mark locking/lifetime/and refcnting
- *
- * REFCNT:
- * The mark->refcnt tells how many "things" in the kernel currently are
- * referencing this object. The object typically will live inside the kernel
- * with a refcnt of 2, one for each list it is on (i_list, g_list). Any task
- * which can find this object holding the appropriete locks, can take a reference
- * and the object itself is guarenteed to survive until the reference is dropped.
- *
- * LOCKING:
- * There are 3 spinlocks involved with fsnotify inode marks and they MUST
- * be taken in order as follows:
- *
- * entry->lock
- * group->mark_lock
- * inode->i_lock
- *
- * entry->lock protects 2 things, entry->group and entry->inode. You must hold
- * that lock to dereference either of these things (they could be NULL even with
- * the lock)
- *
- * group->mark_lock protects the mark_entries list anchored inside a given group
- * and each entry is hooked via the g_list. It also sorta protects the
- * free_g_list, which when used is anchored by a private list on the stack of the
- * task which held the group->mark_lock.
- *
- * inode->i_lock protects the i_fsnotify_mark_entries list anchored inside a
- * given inode and each entry is hooked via the i_list. (and sorta the
- * free_i_list)
- *
- *
- * LIFETIME:
- * Inode marks survive between when they are added to an inode and when their
- * refcnt==0.
- *
- * The inode mark can be cleared for a number of different reasons including:
- * - The inode is unlinked for the last time. (fsnotify_inode_remove)
- * - The inode is being evicted from cache. (fsnotify_inode_delete)
- * - The fs the inode is on is unmounted. (fsnotify_inode_delete/fsnotify_unmount_inodes)
- * - Something explicitly requests that it be removed. (fsnotify_destroy_mark_by_entry)
- * - The fsnotify_group associated with the mark is going away and all such marks
- * need to be cleaned up. (fsnotify_clear_marks_by_group)
- *
- * Worst case we are given an inode and need to clean up all the marks on that
- * inode. We take i_lock and walk the i_fsnotify_mark_entries safely. For each
- * mark on the list we take a reference (so the mark can't disappear under us).
- * We remove that mark form the inode's list of marks and we add this mark to a
- * private list anchored on the stack using i_free_list; At this point we no
- * longer fear anything finding the mark using the inode's list of marks.
- *
- * We can safely and locklessly run the private list on the stack of everything
- * we just unattached from the original inode. For each mark on the private list
- * we grab the mark-> and can thus dereference mark->group and mark->inode. If
- * we see the group and inode are not NULL we take those locks. Now holding all
- * 3 locks we can completely remove the mark from other tasks finding it in the
- * future. Remember, 10 things might already be referencing this mark, but they
- * better be holding a ref. We drop our reference we took before we unhooked it
- * from the inode. When the ref hits 0 we can free the mark.
- *
- * Very similarly for freeing by group, except we use free_g_list.
- *
- * This has the very interesting property of being able to run concurrently with
- * any (or all) other directions.
- */
-
#include <linux/fs.h>
#include <linux/init.h>
#include <linux/kernel.h>
@@ -95,30 +29,19 @@
#include <linux/fsnotify_backend.h>
#include "fsnotify.h"
-void fsnotify_get_mark(struct fsnotify_mark_entry *entry)
-{
- atomic_inc(&entry->refcnt);
-}
-
-void fsnotify_put_mark(struct fsnotify_mark_entry *entry)
-{
- if (atomic_dec_and_test(&entry->refcnt))
- entry->free_mark(entry);
-}
-
/*
* Recalculate the mask of events relevant to a given inode locked.
*/
static void fsnotify_recalc_inode_mask_locked(struct inode *inode)
{
- struct fsnotify_mark_entry *entry;
+ struct fsnotify_mark *mark;
struct hlist_node *pos;
__u32 new_mask = 0;
assert_spin_locked(&inode->i_lock);
- hlist_for_each_entry(entry, pos, &inode->i_fsnotify_mark_entries, i_list)
- new_mask |= entry->mask;
+ hlist_for_each_entry(mark, pos, &inode->i_fsnotify_marks, i.i_list)
+ new_mask |= mark->mask;
inode->i_fsnotify_mask = new_mask;
}
@@ -135,107 +58,26 @@ void fsnotify_recalc_inode_mask(struct inode *inode)
__fsnotify_update_child_dentry_flags(inode);
}
-/*
- * Any time a mark is getting freed we end up here.
- * The caller had better be holding a reference to this mark so we don't actually
- * do the final put under the entry->lock
- */
-void fsnotify_destroy_mark_by_entry(struct fsnotify_mark_entry *entry)
+void fsnotify_destroy_inode_mark(struct fsnotify_mark *mark)
{
- struct fsnotify_group *group;
- struct inode *inode;
+ struct inode *inode = mark->i.inode;
- spin_lock(&entry->lock);
+ assert_spin_locked(&mark->lock);
+ assert_spin_locked(&mark->group->mark_lock);
- group = entry->group;
- inode = entry->inode;
-
- BUG_ON(group && !inode);
- BUG_ON(!group && inode);
-
- /* if !group something else already marked this to die */
- if (!group) {
- spin_unlock(&entry->lock);
- return;
- }
-
- /* 1 from caller and 1 for being on i_list/g_list */
- BUG_ON(atomic_read(&entry->refcnt) < 2);
-
- spin_lock(&group->mark_lock);
spin_lock(&inode->i_lock);
- hlist_del_init(&entry->i_list);
- entry->inode = NULL;
-
- list_del_init(&entry->g_list);
- entry->group = NULL;
-
- fsnotify_put_mark(entry); /* for i_list and g_list */
+ hlist_del_init_rcu(&mark->i.i_list);
+ mark->i.inode = NULL;
/*
- * this mark is now off the inode->i_fsnotify_mark_entries list and we
+ * this mark is now off the inode->i_fsnotify_marks list and we
* hold the inode->i_lock, so this is the perfect time to update the
* inode->i_fsnotify_mask
*/
fsnotify_recalc_inode_mask_locked(inode);
spin_unlock(&inode->i_lock);
- spin_unlock(&group->mark_lock);
- spin_unlock(&entry->lock);
-
- /*
- * Some groups like to know that marks are being freed. This is a
- * callback to the group function to let it know that this entry
- * is being freed.
- */
- if (group->ops->freeing_mark)
- group->ops->freeing_mark(entry, group);
-
- /*
- * __fsnotify_update_child_dentry_flags(inode);
- *
- * I really want to call that, but we can't, we have no idea if the inode
- * still exists the second we drop the entry->lock.
- *
- * The next time an event arrive to this inode from one of it's children
- * __fsnotify_parent will see that the inode doesn't care about it's
- * children and will update all of these flags then. So really this
- * is just a lazy update (and could be a perf win...)
- */
-
-
- iput(inode);
-
- /*
- * it's possible that this group tried to destroy itself, but this
- * this mark was simultaneously being freed by inode. If that's the
- * case, we finish freeing the group here.
- */
- if (unlikely(atomic_dec_and_test(&group->num_marks)))
- fsnotify_final_destroy_group(group);
-}
-
-/*
- * Given a group, destroy all of the marks associated with that group.
- */
-void fsnotify_clear_marks_by_group(struct fsnotify_group *group)
-{
- struct fsnotify_mark_entry *lentry, *entry;
- LIST_HEAD(free_list);
-
- spin_lock(&group->mark_lock);
- list_for_each_entry_safe(entry, lentry, &group->mark_entries, g_list) {
- list_add(&entry->free_g_list, &free_list);
- list_del_init(&entry->g_list);
- fsnotify_get_mark(entry);
- }
- spin_unlock(&group->mark_lock);
-
- list_for_each_entry_safe(entry, lentry, &free_list, free_g_list) {
- fsnotify_destroy_mark_by_entry(entry);
- fsnotify_put_mark(entry);
- }
}
/*
@@ -243,112 +85,145 @@ void fsnotify_clear_marks_by_group(struct fsnotify_group *group)
*/
void fsnotify_clear_marks_by_inode(struct inode *inode)
{
- struct fsnotify_mark_entry *entry, *lentry;
+ struct fsnotify_mark *mark, *lmark;
struct hlist_node *pos, *n;
LIST_HEAD(free_list);
spin_lock(&inode->i_lock);
- hlist_for_each_entry_safe(entry, pos, n, &inode->i_fsnotify_mark_entries, i_list) {
- list_add(&entry->free_i_list, &free_list);
- hlist_del_init(&entry->i_list);
- fsnotify_get_mark(entry);
+ hlist_for_each_entry_safe(mark, pos, n, &inode->i_fsnotify_marks, i.i_list) {
+ list_add(&mark->i.free_i_list, &free_list);
+ hlist_del_init_rcu(&mark->i.i_list);
+ fsnotify_get_mark(mark);
}
spin_unlock(&inode->i_lock);
- list_for_each_entry_safe(entry, lentry, &free_list, free_i_list) {
- fsnotify_destroy_mark_by_entry(entry);
- fsnotify_put_mark(entry);
+ list_for_each_entry_safe(mark, lmark, &free_list, i.free_i_list) {
+ fsnotify_destroy_mark(mark);
+ fsnotify_put_mark(mark);
}
}
/*
+ * Given a group clear all of the inode marks associated with that group.
+ */
+void fsnotify_clear_inode_marks_by_group(struct fsnotify_group *group)
+{
+ fsnotify_clear_marks_by_group_flags(group, FSNOTIFY_MARK_FLAG_INODE);
+}
+
+/*
* given a group and inode, find the mark associated with that combination.
* if found take a reference to that mark and return it, else return NULL
*/
-struct fsnotify_mark_entry *fsnotify_find_mark_entry(struct fsnotify_group *group,
- struct inode *inode)
+struct fsnotify_mark *fsnotify_find_inode_mark_locked(struct fsnotify_group *group,
+ struct inode *inode)
{
- struct fsnotify_mark_entry *entry;
+ struct fsnotify_mark *mark;
struct hlist_node *pos;
assert_spin_locked(&inode->i_lock);
- hlist_for_each_entry(entry, pos, &inode->i_fsnotify_mark_entries, i_list) {
- if (entry->group == group) {
- fsnotify_get_mark(entry);
- return entry;
+ hlist_for_each_entry(mark, pos, &inode->i_fsnotify_marks, i.i_list) {
+ if (mark->group == group) {
+ fsnotify_get_mark(mark);
+ return mark;
}
}
return NULL;
}
/*
- * Nothing fancy, just initialize lists and locks and counters.
+ * given a group and inode, find the mark associated with that combination.
+ * if found take a reference to that mark and return it, else return NULL
*/
-void fsnotify_init_mark(struct fsnotify_mark_entry *entry,
- void (*free_mark)(struct fsnotify_mark_entry *entry))
+struct fsnotify_mark *fsnotify_find_inode_mark(struct fsnotify_group *group,
+ struct inode *inode)
+{
+ struct fsnotify_mark *mark;
+
+ spin_lock(&inode->i_lock);
+ mark = fsnotify_find_inode_mark_locked(group, inode);
+ spin_unlock(&inode->i_lock);
+ return mark;
+}
+
+/*
+ * If we are setting a mark mask on an inode mark we should pin the inode
+ * in memory.
+ */
+void fsnotify_set_inode_mark_mask_locked(struct fsnotify_mark *mark,
+ __u32 mask)
{
- spin_lock_init(&entry->lock);
- atomic_set(&entry->refcnt, 1);
- INIT_HLIST_NODE(&entry->i_list);
- entry->group = NULL;
- entry->mask = 0;
- entry->inode = NULL;
- entry->free_mark = free_mark;
+ struct inode *inode;
+
+ assert_spin_locked(&mark->lock);
+
+ if (mask &&
+ mark->i.inode &&
+ !(mark->flags & FSNOTIFY_MARK_FLAG_OBJECT_PINNED)) {
+ mark->flags |= FSNOTIFY_MARK_FLAG_OBJECT_PINNED;
+ inode = igrab(mark->i.inode);
+ /*
+ * we shouldn't be able to get here if the inode wasn't
+ * already safely held in memory. But bug in case it
+ * ever is wrong.
+ */
+ BUG_ON(!inode);
+ }
}
/*
- * Attach an initialized mark entry to a given group and inode.
+ * Attach an initialized mark to a given inode.
* These marks may be used for the fsnotify backend to determine which
- * event types should be delivered to which group and for which inodes.
+ * event types should be delivered to which group and for which inodes. These
+ * marks are ordered according to the group's location in memory.
*/
-int fsnotify_add_mark(struct fsnotify_mark_entry *entry,
- struct fsnotify_group *group, struct inode *inode)
+int fsnotify_add_inode_mark(struct fsnotify_mark *mark,
+ struct fsnotify_group *group, struct inode *inode,
+ int allow_dups)
{
- struct fsnotify_mark_entry *lentry;
+ struct fsnotify_mark *lmark;
+ struct hlist_node *node, *last = NULL;
int ret = 0;
- inode = igrab(inode);
- if (unlikely(!inode))
- return -EINVAL;
+ mark->flags |= FSNOTIFY_MARK_FLAG_INODE;
+
+ assert_spin_locked(&mark->lock);
+ assert_spin_locked(&group->mark_lock);
- /*
- * LOCKING ORDER!!!!
- * entry->lock
- * group->mark_lock
- * inode->i_lock
- */
- spin_lock(&entry->lock);
- spin_lock(&group->mark_lock);
spin_lock(&inode->i_lock);
- lentry = fsnotify_find_mark_entry(group, inode);
- if (!lentry) {
- entry->group = group;
- entry->inode = inode;
+ mark->i.inode = inode;
- hlist_add_head(&entry->i_list, &inode->i_fsnotify_mark_entries);
- list_add(&entry->g_list, &group->mark_entries);
+ /* is mark the first mark? */
+ if (hlist_empty(&inode->i_fsnotify_marks)) {
+ hlist_add_head_rcu(&mark->i.i_list, &inode->i_fsnotify_marks);
+ goto out;
+ }
- fsnotify_get_mark(entry); /* for i_list and g_list */
+ /* should mark be in the middle of the current list? */
+ hlist_for_each_entry(lmark, node, &inode->i_fsnotify_marks, i.i_list) {
+ last = node;
+
+ if ((lmark->group == group) && !allow_dups) {
+ ret = -EEXIST;
+ goto out;
+ }
- atomic_inc(&group->num_marks);
+ if (mark->group < lmark->group)
+ continue;
- fsnotify_recalc_inode_mask_locked(inode);
+ hlist_add_before_rcu(&mark->i.i_list, &lmark->i.i_list);
+ goto out;
}
+ BUG_ON(last == NULL);
+ /* mark should be the last entry. last is the current last entry */
+ hlist_add_after_rcu(last, &mark->i.i_list);
+out:
+ fsnotify_recalc_inode_mask_locked(inode);
spin_unlock(&inode->i_lock);
- spin_unlock(&group->mark_lock);
- spin_unlock(&entry->lock);
-
- if (lentry) {
- ret = -EEXIST;
- iput(inode);
- fsnotify_put_mark(lentry);
- } else {
- __fsnotify_update_child_dentry_flags(inode);
- }
return ret;
}