Skip to contentSkip to conversation

fsnotify cleanups

PATCH Superseded Jan Kara 2 participants 5 messages Nov 12, 2014 – Nov 19, 2014

Subsystems (5)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Jan Kara

author maintainer
#0 lore raw

v2 cover fsnotify cleanups

Hello,

this is second version of my fsnotify cleanups. Since previous version
I've dropped patch removing free_list. It was introducing possible use after free issues. I will revisit that later. I've also updated comment about use of g_list for list of marks to destroy and factored out some more duplicated code.

Honza

author maintainer
#1 lore raw

PATCH v2 1/2 fsnotify: Unify inode and mount marks handling

There's lot of common code in inode and mount marks handling. Factor it out to common helper function.

---
 fs/notify/dnotify/dnotify.c          |   4 +-
 fs/notify/fdinfo.c                   |   6 +-
 fs/notify/fsnotify.c                 |   4 +-
 fs/notify/fsnotify.h                 |  12 ++++
 fs/notify/inode_mark.c               | 108 ++++++-----------------------------
 fs/notify/inotify/inotify_fsnotify.c |   2 +-
 fs/notify/inotify/inotify_user.c     |  10 ++--
 fs/notify/mark.c                     |  88 +++++++++++++++++++++++++++-
 fs/notify/vfsmount_mark.c            | 103 ++++++---------------------------
 include/linux/fsnotify_backend.h     |  24 ++------
 kernel/audit_tree.c                  |  16 +++---
 11 files changed, 154 insertions(+), 223 deletions(-)
fs/notify/dnotify/dnotify.c +2 -2
index caaaf9dfe353..44523f4a6084 100644
@@ -69,8 +69,8 @@ static void dnotify_recalc_inode_mask(struct fsnotify_mark *fsn_mark)
 	if (old_mask == new_mask)
 		return;
 
-	if (fsn_mark->i.inode)
-		fsnotify_recalc_inode_mask(fsn_mark->i.inode);
+	if (fsn_mark->inode)
+		fsnotify_recalc_inode_mask(fsn_mark->inode);
 }
 
 /*
fs/notify/fdinfo.c +3 -3
index 9d7e2b9659cb..c7a37d221030 100644
@@ -85,7 +85,7 @@ static int inotify_fdinfo(struct seq_file *m, struct fsnotify_mark *mark)
 		return 0;
 
 	inode_mark = container_of(mark, struct inotify_inode_mark, fsn_mark);
-	inode = igrab(mark->i.inode);
+	inode = igrab(mark->inode);
 	if (inode) {
 		ret = seq_printf(m, "inotify wd:%x ino:%lx sdev:%x "
 				 "mask:%x ignored_mask:%x ",
@@ -122,7 +122,7 @@ static int fanotify_fdinfo(struct seq_file *m, struct fsnotify_mark *mark)
 		mflags |= FAN_MARK_IGNORED_SURV_MODIFY;
 
 	if (mark->flags & FSNOTIFY_MARK_FLAG_INODE) {
-		inode = igrab(mark->i.inode);
+		inode = igrab(mark->inode);
 		if (!inode)
 			goto out;
 		ret = seq_printf(m, "fanotify ino:%lx sdev:%x "
@@ -133,7 +133,7 @@ static int fanotify_fdinfo(struct seq_file *m, struct fsnotify_mark *mark)
 		ret |= seq_putc(m, '\n');
 		iput(inode);
 	} else if (mark->flags & FSNOTIFY_MARK_FLAG_VFSMOUNT) {
-		struct mount *mnt = real_mount(mark->m.mnt);
+		struct mount *mnt = real_mount(mark->mnt);
 
 		ret = seq_printf(m, "fanotify mnt_id:%x mflags:%x mask:%x "
 				 "ignored_mask:%x\n", mnt->mnt_id, mflags,
fs/notify/fsnotify.c +2 -2
index 89326acd4561..31cf7af3618b 100644
@@ -242,13 +242,13 @@ int fsnotify(struct inode *to_tell, __u32 mask, void *data, int data_is,
 
 		if (inode_node) {
 			inode_mark = hlist_entry(srcu_dereference(inode_node, &fsnotify_mark_srcu),
-						 struct fsnotify_mark, i.i_list);
+						 struct fsnotify_mark, obj_list);
 			inode_group = inode_mark->group;
 		}
 
 		if (vfsmount_node) {
 			vfsmount_mark = hlist_entry(srcu_dereference(vfsmount_node, &fsnotify_mark_srcu),
-							struct fsnotify_mark, m.m_list);
+						    struct fsnotify_mark, obj_list);
 			vfsmount_group = vfsmount_mark->group;
 		}
 
fs/notify/fsnotify.h +12 -0
index 3b68b0ae0a97..13a00be516d2 100644
@@ -12,12 +12,19 @@ extern void fsnotify_flush_notify(struct fsnotify_group *group);
 /* protects reads of inode and vfsmount marks list */
 extern struct srcu_struct fsnotify_mark_srcu;
 
+/* Calculate mask of events for a list of marks */
+extern u32 fsnotify_recalc_mask(struct hlist_head *head);
+
 /* compare two groups for sorting of marks lists */
 extern int fsnotify_compare_groups(struct fsnotify_group *a,
 				   struct fsnotify_group *b);
 
 extern void fsnotify_set_inode_mark_mask_locked(struct fsnotify_mark *fsn_mark,
 						__u32 mask);
+/* Add mark to a proper place in mark list */
+extern int fsnotify_add_mark_list(struct hlist_head *head,
+				  struct fsnotify_mark *mark,
+				  int allow_dups);
 /* add a mark to an inode */
 extern int fsnotify_add_inode_mark(struct fsnotify_mark *mark,
 				   struct fsnotify_group *group, struct inode *inode,
@@ -31,6 +38,11 @@ extern int fsnotify_add_vfsmount_mark(struct fsnotify_mark *mark,
 extern void fsnotify_destroy_vfsmount_mark(struct fsnotify_mark *mark);
 /* inode specific destruction of a mark */
 extern void fsnotify_destroy_inode_mark(struct fsnotify_mark *mark);
+/* Destroy all marks in the given list */
+extern void fsnotify_destroy_marks(struct list_head *to_free);
+/* Find mark belonging to given group in the list of marks */
+extern struct fsnotify_mark *fsnotify_find_mark(struct hlist_head *head,
+						struct fsnotify_group *group);
 /* run the list of all marks associated with inode and flag them to be freed */
 extern void fsnotify_clear_marks_by_inode(struct inode *inode);
 /* run the list of all marks associated with vfsmount and flag them to be freed */
fs/notify/inode_mark.c +16 -92

178 lines not on this page · raw message

fs/notify/inotify/inotify_fsnotify.c +1 -1
index 7d888d77d59a..2cd900c2c737 100644
@@ -156,7 +156,7 @@ static int idr_callback(int id, void *p, void *data)
 	 */
 	if (fsn_mark)
 		printk(KERN_WARNING "fsn_mark->group=%p inode=%p wd=%d\n",
-			fsn_mark->group, fsn_mark->i.inode, i_mark->wd);
+			fsn_mark->group, fsn_mark->inode, i_mark->wd);
 	return 0;
 }
 
fs/notify/inotify/inotify_user.c +5 -5
index daf76652fe58..1f1069e3c7f3 100644
@@ -434,7 +434,7 @@ static void inotify_remove_from_idr(struct fsnotify_group *group,
 	if (wd == -1) {
 		WARN_ONCE(1, "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p"
 			" i_mark->inode=%p\n", __func__, i_mark, i_mark->wd,
-			i_mark->fsn_mark.group, i_mark->fsn_mark.i.inode);
+			i_mark->fsn_mark.group, i_mark->fsn_mark.inode);
 		goto out;
 	}
 
@@ -443,7 +443,7 @@ static void inotify_remove_from_idr(struct fsnotify_group *group,
 	if (unlikely(!found_i_mark)) {
 		WARN_ONCE(1, "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p"
 			" i_mark->inode=%p\n", __func__, i_mark, i_mark->wd,
-			i_mark->fsn_mark.group, i_mark->fsn_mark.i.inode);
+			i_mark->fsn_mark.group, i_mark->fsn_mark.inode);
 		goto out;
 	}
 
@@ -457,9 +457,9 @@ static void inotify_remove_from_idr(struct fsnotify_group *group,
 			"mark->inode=%p found_i_mark=%p found_i_mark->wd=%d "
 			"found_i_mark->group=%p found_i_mark->inode=%p\n",
 			__func__, i_mark, i_mark->wd, i_mark->fsn_mark.group,
-			i_mark->fsn_mark.i.inode, found_i_mark, found_i_mark->wd,
+			i_mark->fsn_mark.inode, found_i_mark, found_i_mark->wd,
 			found_i_mark->fsn_mark.group,
-			found_i_mark->fsn_mark.i.inode);
+			found_i_mark->fsn_mark.inode);
 		goto out;
 	}
 
@@ -471,7 +471,7 @@ static void inotify_remove_from_idr(struct fsnotify_group *group,
 	if (unlikely(atomic_read(&i_mark->fsn_mark.refcnt) < 3)) {
 		printk(KERN_ERR "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p"
 			" i_mark->inode=%p\n", __func__, i_mark, i_mark->wd,
-			i_mark->fsn_mark.group, i_mark->fsn_mark.i.inode);
+			i_mark->fsn_mark.group, i_mark->fsn_mark.inode);
 		/* we can't really recover with bad ref cnting.. */
 		BUG();
 	}
fs/notify/mark.c +85 -3

124 lines not on this page · raw message

fs/notify/vfsmount_mark.c +16 -87

168 lines not on this page · raw message

include/linux/fsnotify_backend.h +4 -20
index ca060d7c4fa6..442847a02b8f 100644
@@ -197,24 +197,6 @@ struct fsnotify_group {
 #define FSNOTIFY_EVENT_INODE	2
 
 /*
- * Inode specific fields in an fsnotify_mark
- */
-struct fsnotify_inode_mark {
-	struct inode *inode;		/* inode this mark is associated with */
-	struct hlist_node i_list;	/* list of marks by inode->i_fsnotify_marks */
-	struct list_head free_i_list;	/* tmp list used when freeing this mark */
-};
-
-/*
- * Mount point specific fields in an fsnotify_mark
- */
-struct fsnotify_vfsmount_mark {
-	struct vfsmount *mnt;		/* vfsmount this mark is associated with */
-	struct hlist_node m_list;	/* list of marks by inode->i_fsnotify_marks */
-	struct list_head free_m_list;	/* tmp list used when freeing this mark */
-};
-
-/*
  * a mark is simply an object attached to an in core inode which allows an
  * fsnotify listener to indicate they are either no longer interested in events
  * of a type matching mask or only interested in those events.
@@ -232,9 +214,11 @@ struct fsnotify_mark {
 	struct fsnotify_group *group;	/* group this mark is for */
 	struct list_head g_list;	/* list of marks by group->i_fsnotify_marks */
 	spinlock_t lock;		/* protect group and inode */
+	struct hlist_node obj_list;	/* list of marks for inode / vfsmount */
+	struct list_head free_list;	/* tmp list used when freeing this mark */
 	union {
-		struct fsnotify_inode_mark i;
-		struct fsnotify_vfsmount_mark m;
+		struct inode *inode;	/* inode this mark is associated with */
+		struct vfsmount *mnt;	/* vfsmount this mark is associated with */
 	};
 	__u32 ignored_mask;		/* events types to ignore */
 #define FSNOTIFY_MARK_FLAG_INODE		0x01
kernel/audit_tree.c +8 -8
index e242e3a9864a..33b3089d88dd 100644
@@ -173,9 +173,9 @@ static void insert_hash(struct audit_chunk *chunk)
 	struct fsnotify_mark *entry = &chunk->mark;
 	struct list_head *list;
 
-	if (!entry->i.inode)
+	if (!entry->inode)
 		return;
-	list = chunk_hash(entry->i.inode);
+	list = chunk_hash(entry->inode);
 	list_add_rcu(&chunk->hash, list);
 }
 
@@ -187,7 +187,7 @@ struct audit_chunk *audit_tree_lookup(const struct inode *inode)
 
 	list_for_each_entry_rcu(p, list, hash) {
 		/* mark.inode may have gone NULL, but who cares? */
-		if (p->mark.i.inode == inode) {
+		if (p->mark.inode == inode) {
 			atomic_long_inc(&p->refs);
 			return p;
 		}
@@ -230,7 +230,7 @@ static void untag_chunk(struct node *p)
 		new = alloc_chunk(size);
 
 	spin_lock(&entry->lock);
-	if (chunk->dead || !entry->i.inode) {
+	if (chunk->dead || !entry->inode) {
 		spin_unlock(&entry->lock);
 		if (new)
 			free_chunk(new);
@@ -257,7 +257,7 @@ static void untag_chunk(struct node *p)
 		goto Fallback;
 
 	fsnotify_duplicate_mark(&new->mark, entry);
-	if (fsnotify_add_mark(&new->mark, new->mark.group, new->mark.i.inode, NULL, 1)) {
+	if (fsnotify_add_mark(&new->mark, new->mark.group, new->mark.inode, NULL, 1)) {
 		fsnotify_put_mark(&new->mark);
 		goto Fallback;
 	}
@@ -385,7 +385,7 @@ static int tag_chunk(struct inode *inode, struct audit_tree *tree)
 	chunk_entry = &chunk->mark;
 
 	spin_lock(&old_entry->lock);
-	if (!old_entry->i.inode) {
+	if (!old_entry->inode) {
 		/* old_entry is being shot, lets just lie */
 		spin_unlock(&old_entry->lock);
 		fsnotify_put_mark(old_entry);
@@ -394,7 +394,7 @@ static int tag_chunk(struct inode *inode, struct audit_tree *tree)
 	}
 
 	fsnotify_duplicate_mark(chunk_entry, old_entry);
-	if (fsnotify_add_mark(chunk_entry, chunk_entry->group, chunk_entry->i.inode, NULL, 1)) {
+	if (fsnotify_add_mark(chunk_entry, chunk_entry->group, chunk_entry->inode, NULL, 1)) {
 		spin_unlock(&old_entry->lock);
 		fsnotify_put_mark(chunk_entry);
 		fsnotify_put_mark(old_entry);
@@ -610,7 +610,7 @@ void audit_trim_trees(void)
 		list_for_each_entry(node, &tree->chunks, list) {
 			struct audit_chunk *chunk = find_chunk(node);
 			/* this could be NULL if the watch is dying else where... */
-			struct inode *inode = chunk->mark.i.inode;
+			struct inode *inode = chunk->mark.inode;
 			node->index |= 1U<<31;
 			if (iterate_mounts(compare_root, inode, root_mnt))
 				node->index &= ~(1U<<31);
Signature · 1 line
1.8.1.4
#2 lore raw

Re: [PATCH 1/2] fsnotify: Unify inode and mount marks handling

Quoted 2 lines · On Wed, 12 Nov 2014 17:07:42 +0100 Jan Kara <[email protected]> wrote:

There's lot of common code in inode and mount marks handling. Factor it out to common helper function.

Confused.

Quoted 9 lines
fs/notify/inode_mark.c +0 -3
@@ -31,28 +31,13 @@
 #include "../internal.h"
 
 /*
- * Recalculate the mask of events relevant to a given inode locked.
- */
-static void fsnotify_recalc_inode_mask_locked(struct inode *inode)

fs/notify/inode_mark.c line 46

Quoted 10 lines
fs/notify/inode_mark.c +0 -10
-{
-	struct fsnotify_mark *mark;
-	__u32 new_mask = 0;
-
-	assert_spin_locked(&inode->i_lock);
-
-	hlist_for_each_entry(mark, &inode->i_fsnotify_marks, i.i_list)
-		new_mask |= mark->mask;
-	inode->i_fsnotify_mask = new_mask;
-}

Gone.

Quoted 28 lines
fs/notify/inode_mark.c +0 -24
@@ -202,36 +153,9 @@ int fsnotify_add_inode_mark(struct fsnotify_mark *mark,
 	assert_spin_locked(&mark->lock);
 
 	spin_lock(&inode->i_lock);
-
-	mark->i.inode = inode;
-
-	/* 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;
-	}
-
-	/* should mark be in the middle of the current list? */
-	hlist_for_each_entry(lmark, &inode->i_fsnotify_marks, i.i_list) {
-		last = lmark;
-
-		if ((lmark->group == group) && !allow_dups) {
-			ret = -EEXIST;
-			goto out;
-		}
-
-		cmp = fsnotify_compare_groups(lmark->group, mark->group);
-		if (cmp < 0)
-			continue;
-
-		hlist_add_before_rcu(&mark->i.i_list, &lmark->i.i_list);
-		goto out;

fs/notify/inode_mark.c int fsnotify_add_inode_mark(struct fsnotify_mark *mark, · line 159

Quoted 10 lines
fs/notify/inode_mark.c +3 -6
-	}
-
-	BUG_ON(last == NULL);
-	/* mark should be the last entry.  last is the current last entry */
-	hlist_add_behind_rcu(&mark->i.i_list, &last->i.i_list);
-out:
+	mark->inode = inode;
+	ret = fsnotify_add_mark_list(&inode->i_fsnotify_marks, mark,
+				     allow_dups);
 	fsnotify_recalc_inode_mask_locked(inode);

But not forgotten. The compiler noticed this ;)

I'll drop fsnotify-unify-inode-and-mount-marks-handling.patch and fsnotify-remove-destroy_list-from-fsnotify_mark.patch - could you take a look please?

I had to stage those two patches after linux-next.patch because someone has been fiddling with fs/notify/ in -next, causing conflicts. Perhaps that had something to do with it, but I don't see how.

author maintainer
#3 lore raw

Re: [PATCH 1/2] fsnotify: Unify inode and mount marks handling

Quoted 70 lines · Andrew Morton
Quoted 2 lines · On Wed, 12 Nov 2014 17:07:42 +0100 Jan Kara <[email protected]> wrote:

There's lot of common code in inode and mount marks handling. Factor it out to common helper function.

Confused.

Quoted 9 lines
fs/notify/inode_mark.c +0 -3
@@ -31,28 +31,13 @@
 #include "../internal.h"
 
 /*
- * Recalculate the mask of events relevant to a given inode locked.
- */
-static void fsnotify_recalc_inode_mask_locked(struct inode *inode)

fs/notify/inode_mark.c line 46

Quoted 10 lines
fs/notify/inode_mark.c +0 -10
-{
-	struct fsnotify_mark *mark;
-	__u32 new_mask = 0;
-
-	assert_spin_locked(&inode->i_lock);
-
-	hlist_for_each_entry(mark, &inode->i_fsnotify_marks, i.i_list)
-		new_mask |= mark->mask;
-	inode->i_fsnotify_mask = new_mask;
-}

Gone.

Quoted 28 lines
fs/notify/inode_mark.c +0 -24
@@ -202,36 +153,9 @@ int fsnotify_add_inode_mark(struct fsnotify_mark *mark,
 	assert_spin_locked(&mark->lock);
 
 	spin_lock(&inode->i_lock);
-
-	mark->i.inode = inode;
-
-	/* 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;
-	}
-
-	/* should mark be in the middle of the current list? */
-	hlist_for_each_entry(lmark, &inode->i_fsnotify_marks, i.i_list) {
-		last = lmark;
-
-		if ((lmark->group == group) && !allow_dups) {
-			ret = -EEXIST;
-			goto out;
-		}
-
-		cmp = fsnotify_compare_groups(lmark->group, mark->group);
-		if (cmp < 0)
-			continue;
-
-		hlist_add_before_rcu(&mark->i.i_list, &lmark->i.i_list);
-		goto out;

fs/notify/inode_mark.c int fsnotify_add_inode_mark(struct fsnotify_mark *mark, · line 159

Quoted 10 lines
fs/notify/inode_mark.c +3 -6
-	}
-
-	BUG_ON(last == NULL);
-	/* mark should be the last entry.  last is the current last entry */
-	hlist_add_behind_rcu(&mark->i.i_list, &last->i.i_list);
-out:
+	mark->inode = inode;
+	ret = fsnotify_add_mark_list(&inode->i_fsnotify_marks, mark,
+				     allow_dups);
 	fsnotify_recalc_inode_mask_locked(inode);

But not forgotten. The compiler noticed this ;)

Quoted 7 lines

I'll drop fsnotify-unify-inode-and-mount-marks-handling.patch and fsnotify-remove-destroy_list-from-fsnotify_mark.patch - could you take a look please?

I had to stage those two patches after linux-next.patch because someone has been fiddling with fs/notify/ in -next, causing conflicts. Perhaps that had something to do with it, but I don't see how.

Bah, looks like some stupidity on my side. Not sure how that happened.
I'll fix it up and resend.

Honza

Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR
author maintainer
#4 lore raw

PATCH v2 2/2 fsnotify: Remove destroy_list from fsnotify_mark

In reply to Jan Kara

destroy_list is used to track marks which still need waiting for srcu period end before they can be freed. However by the time mark is added to destroy_list it isn't in group's list of marks anymore and thus we can reuse fsnotify_mark->g_list for queueing into destroy_list. This saves two pointers for each fsnotify_mark.

---
 fs/notify/mark.c                 | 8 ++++----
 include/linux/fsnotify_backend.h | 7 +++++--
 2 files changed, 9 insertions(+), 6 deletions(-)
fs/notify/mark.c +4 -4
index c533f0074da5..765598017e81 100644
@@ -160,7 +160,7 @@ void fsnotify_destroy_mark_locked(struct fsnotify_mark *mark,
 	mutex_unlock(&group->mark_mutex);
 
 	spin_lock(&destroy_lock);
-	list_add(&mark->destroy_list, &destroy_list);
+	list_add(&mark->g_list, &destroy_list);
 	spin_unlock(&destroy_lock);
 	wake_up(&destroy_waitq);
 	/*
@@ -369,7 +369,7 @@ err:
 	spin_unlock(&mark->lock);
 
 	spin_lock(&destroy_lock);
-	list_add(&mark->destroy_list, &destroy_list);
+	list_add(&mark->g_list, &destroy_list);
 	spin_unlock(&destroy_lock);
 	wake_up(&destroy_waitq);
 
@@ -468,8 +468,8 @@ static int fsnotify_mark_destroy(void *ignored)
 
 		synchronize_srcu(&fsnotify_mark_srcu);
 
-		list_for_each_entry_safe(mark, next, &private_destroy_list, destroy_list) {
-			list_del_init(&mark->destroy_list);
+		list_for_each_entry_safe(mark, next, &private_destroy_list, g_list) {
+			list_del_init(&mark->g_list);
 			fsnotify_put_mark(mark);
 		}
 
include/linux/fsnotify_backend.h +5 -2
index 442847a02b8f..0f313f93c586 100644
@@ -212,7 +212,11 @@ struct fsnotify_mark {
 	 * in kernel that found and may be using this mark. */
 	atomic_t refcnt;		/* active things looking at this mark */
 	struct fsnotify_group *group;	/* group this mark is for */
-	struct list_head g_list;	/* list of marks by group->i_fsnotify_marks */
+	struct list_head g_list;	/* list of marks by group->i_fsnotify_marks
+					 * Also reused for queueing mark into
+					 * destroy_list when it's waiting for
+					 * the end of SRCU period before it can
+					 * be freed */
 	spinlock_t lock;		/* protect group and inode */
 	struct hlist_node obj_list;	/* list of marks for inode / vfsmount */
 	struct list_head free_list;	/* tmp list used when freeing this mark */
@@ -227,7 +231,6 @@ struct fsnotify_mark {
 #define FSNOTIFY_MARK_FLAG_IGNORED_SURV_MODIFY	0x08
 #define FSNOTIFY_MARK_FLAG_ALIVE		0x10
 	unsigned int flags;		/* vfsmount or inode mark? */
-	struct list_head destroy_list;
 	void (*free_mark)(struct fsnotify_mark *mark); /* called on final put+free */
 };
 
Signature · 1 line
1.8.1.4