[PATCH v7 0/6] kernfs: proposed locking and concurrency improvement

[PATCH v7 0/6] kernfs: proposed locking and concurrency improvement

[Ian Kent]

There have been a few instances of contention on the kernfs_mutex during
path walks, a case on very large IBM systems seen by myself, a report by
Brice Goglin and followed up by Fox Chen, and I've since seen a couple
of other reports by CoreOS users.

The common thread is a large number of kernfs path walks leading to
slowness of path walks due to kernfs_mutex contention.

The problem being that changes to the VFS over some time have increased
it's concurrency capabilities to an extent that kernfs's use of a mutex
is no longer appropriate. There's also an issue of walks for non-existent
paths causing contention if there are quite a few of them which is a less
common problem.

This patch series is relatively straight forward.

All it does is add the ability to take advantage of VFS negative dentry
caching to avoid needless dentry alloc/free cycles for lookups of paths
that don't exit and change the kernfs_mutex to a read/write semaphore.

This patch series is relatively straight forward.

All it does is add the ability to take advantage of VFS negative dentry
caching to avoid needless dentry alloc/free cycles for lookups of paths
that don't exit and change the kernfs_mutex to a read/write semaphore.

The patch that tried to stay in VFS rcu-walk mode during path walks has
been dropped for two reasons. First, it doesn't actually give very much
improvement and, second, if there's a place where mistakes could go
unnoticed it would be in that path. This makes the patch series simpler
to review and reduces the likelihood of problems going unnoticed and
popping up later.

Changes since v6:
- ensure negative dentry as rename target gets invalidated.
- don't bother checking if node is a directory in revision helpers.
- don't use dget_parent(), just use dentry d_lock to ensure parent
  is stable.
- fix kernfs_iop_getattr() and kernfs_iop_permission() locking.
- drop add kernfs_need_inode_refresh() patch, it can't be used.

Changes since v5:
- change kernfs_dir_changed() comparison.
- move negative dentry out from under kernfs node lock in revalidate.
- only set d_time for negative dentries.
- add patch to move d_splice_alias() out from under kernfs node lock
  in lookup.

Changes since v4:
- fixed kernfs_active() naming.
- added back kernfs_node revision patch to use for negative dentry
  validation.
- minor updates to patch descriptions.

Changes since v3:
- remove unneeded indirection when referencing the super block.
- check if inode attribute update is actually needed.

Changes since v2:
- actually fix the inode attribute update locking.
- drop the patch that tried to stay in rcu-walk mode.
- drop the use a revision to identify if a directory has changed patch.

Changes since v1:
- fix locking in .permission() and .gated() by re-factoring the
  attribute handling code.
---

Ian Kent (6):
      kernfs: move revalidate to be near lookup
      kernfs: add a revision to identify directory node changes
      kernfs: use VFS negative dentry caching
      kernfs: switch kernfs to use an rwsem
      kernfs: use i_lock to protect concurrent inode updates
      kernfs: dont call d_splice_alias() under kernfs node lock


 fs/kernfs/dir.c             | 151 ++++++++++++++++++++----------------
 fs/kernfs/file.c            |   4 +-
 fs/kernfs/inode.c           |  26 ++++---
 fs/kernfs/kernfs-internal.h |  24 +++++-
 fs/kernfs/mount.c           |  12 +--
 fs/kernfs/symlink.c         |   4 +-
 include/linux/kernfs.h      |   7 +-
 7 files changed, 136 insertions(+), 92 deletions(-)

--
Ian

[PATCH v7 1/6] kernfs: move revalidate to be near lookup

[Ian Kent]

While the dentry operation kernfs_dop_revalidate() is grouped with
dentry type functions it also has a strong affinity to the inode
operation ->lookup().

It makes sense to locate this function near to kernfs_iop_lookup()
because we will be adding VFS negative dentry caching to reduce path
lookup overhead for non-existent paths.

There's no functional change from this patch.

Signed-off-by: Ian Kent <raven@themaw.net>
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
---
 fs/kernfs/dir.c |   86 ++++++++++++++++++++++++++++---------------------------
 1 file changed, 43 insertions(+), 43 deletions(-)

diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
index 7e0e62deab53c..33166ec90a112 100644
--- a/fs/kernfs/dir.c
+++ b/fs/kernfs/dir.c
@@ -548,49 +548,6 @@ void kernfs_put(struct kernfs_node *kn)
 }
 EXPORT_SYMBOL_GPL(kernfs_put);
 
-static int kernfs_dop_revalidate(struct dentry *dentry, unsigned int flags)
-{
-	struct kernfs_node *kn;
-
-	if (flags & LOOKUP_RCU)
-		return -ECHILD;
-
-	/* Always perform fresh lookup for negatives */
-	if (d_really_is_negative(dentry))
-		goto out_bad_unlocked;
-
-	kn = kernfs_dentry_node(dentry);
-	mutex_lock(&kernfs_mutex);
-
-	/* The kernfs node has been deactivated */
-	if (!kernfs_active(kn))
-		goto out_bad;
-
-	/* The kernfs node has been moved? */
-	if (kernfs_dentry_node(dentry->d_parent) != kn->parent)
-		goto out_bad;
-
-	/* The kernfs node has been renamed */
-	if (strcmp(dentry->d_name.name, kn->name) != 0)
-		goto out_bad;
-
-	/* The kernfs node has been moved to a different namespace */
-	if (kn->parent && kernfs_ns_enabled(kn->parent) &&
-	    kernfs_info(dentry->d_sb)->ns != kn->ns)
-		goto out_bad;
-
-	mutex_unlock(&kernfs_mutex);
-	return 1;
-out_bad:
-	mutex_unlock(&kernfs_mutex);
-out_bad_unlocked:
-	return 0;
-}
-
-const struct dentry_operations kernfs_dops = {
-	.d_revalidate	= kernfs_dop_revalidate,
-};
-
 /**
  * kernfs_node_from_dentry - determine kernfs_node associated with a dentry
  * @dentry: the dentry in question
@@ -1073,6 +1030,49 @@ struct kernfs_node *kernfs_create_empty_dir(struct kernfs_node *parent,
 	return ERR_PTR(rc);
 }
 
+static int kernfs_dop_revalidate(struct dentry *dentry, unsigned int flags)
+{
+	struct kernfs_node *kn;
+
+	if (flags & LOOKUP_RCU)
+		return -ECHILD;
+
+	/* Always perform fresh lookup for negatives */
+	if (d_really_is_negative(dentry))
+		goto out_bad_unlocked;
+
+	kn = kernfs_dentry_node(dentry);
+	mutex_lock(&kernfs_mutex);
+
+	/* The kernfs node has been deactivated */
+	if (!kernfs_active(kn))
+		goto out_bad;
+
+	/* The kernfs node has been moved? */
+	if (kernfs_dentry_node(dentry->d_parent) != kn->parent)
+		goto out_bad;
+
+	/* The kernfs node has been renamed */
+	if (strcmp(dentry->d_name.name, kn->name) != 0)
+		goto out_bad;
+
+	/* The kernfs node has been moved to a different namespace */
+	if (kn->parent && kernfs_ns_enabled(kn->parent) &&
+	    kernfs_info(dentry->d_sb)->ns != kn->ns)
+		goto out_bad;
+
+	mutex_unlock(&kernfs_mutex);
+	return 1;
+out_bad:
+	mutex_unlock(&kernfs_mutex);
+out_bad_unlocked:
+	return 0;
+}
+
+const struct dentry_operations kernfs_dops = {
+	.d_revalidate	= kernfs_dop_revalidate,
+};
+
 static struct dentry *kernfs_iop_lookup(struct inode *dir,
 					struct dentry *dentry,
 					unsigned int flags)

[PATCH v7 2/6] kernfs: add a revision to identify directory node changes

[Ian Kent]

Add a revision counter to kernfs directory nodes so it can be used
to detect if a directory node has changed during negative dentry
revalidation.

There's an assumption that sizeof(unsigned long) <= sizeof(pointer)
on all architectures and as far as I know that assumption holds.

So adding a revision counter to the struct kernfs_elem_dir variant of
the kernfs_node type union won't increase the size of the kernfs_node
struct. This is because struct kernfs_elem_dir is at least
sizeof(pointer) smaller than the largest union variant. It's tempting
to make the revision counter a u64 but that would increase the size of
kernfs_node on archs where sizeof(pointer) is smaller than the revision
counter.

Signed-off-by: Ian Kent <raven@themaw.net>
---
 fs/kernfs/dir.c             |    3 +++
 fs/kernfs/kernfs-internal.h |   19 +++++++++++++++++++
 include/linux/kernfs.h      |    5 +++++
 3 files changed, 27 insertions(+)

diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
index 33166ec90a112..48704c5b6a072 100644
--- a/fs/kernfs/dir.c
+++ b/fs/kernfs/dir.c
@@ -372,6 +372,7 @@ static int kernfs_link_sibling(struct kernfs_node *kn)
 	/* successfully added, account subdir number */
 	if (kernfs_type(kn) == KERNFS_DIR)
 		kn->parent->dir.subdirs++;
+	kernfs_inc_rev(kn->parent);
 
 	return 0;
 }
@@ -394,6 +395,7 @@ static bool kernfs_unlink_sibling(struct kernfs_node *kn)
 
 	if (kernfs_type(kn) == KERNFS_DIR)
 		kn->parent->dir.subdirs--;
+	kernfs_inc_rev(kn->parent);
 
 	rb_erase(&kn->rb, &kn->parent->dir.children);
 	RB_CLEAR_NODE(&kn->rb);
@@ -1573,6 +1575,7 @@ int kernfs_rename_ns(struct kernfs_node *kn, struct kernfs_node *new_parent,
 	 */
 	kernfs_unlink_sibling(kn);
 	kernfs_get(new_parent);
+	kernfs_inc_rev(new_parent);
 
 	/* rename_lock protects ->parent and ->name accessors */
 	spin_lock_irq(&kernfs_rename_lock);
diff --git a/fs/kernfs/kernfs-internal.h b/fs/kernfs/kernfs-internal.h
index ccc3b44f6306f..6a38897b4d02c 100644
--- a/fs/kernfs/kernfs-internal.h
+++ b/fs/kernfs/kernfs-internal.h
@@ -81,6 +81,25 @@ static inline struct kernfs_node *kernfs_dentry_node(struct dentry *dentry)
 	return d_inode(dentry)->i_private;
 }
 
+static inline void kernfs_set_rev(struct kernfs_node *parent,
+				  struct dentry *dentry)
+{
+	dentry->d_time = parent->dir.rev;
+}
+
+static inline void kernfs_inc_rev(struct kernfs_node *parent)
+{
+	parent->dir.rev++;
+}
+
+static inline bool kernfs_dir_changed(struct kernfs_node *parent,
+				      struct dentry *dentry)
+{
+	if (parent->dir.rev != dentry->d_time)
+			return true;
+	return false;
+}
+
 extern const struct super_operations kernfs_sops;
 extern struct kmem_cache *kernfs_node_cache, *kernfs_iattrs_cache;
 
diff --git a/include/linux/kernfs.h b/include/linux/kernfs.h
index 9e8ca8743c268..d68b4ad095737 100644
--- a/include/linux/kernfs.h
+++ b/include/linux/kernfs.h
@@ -98,6 +98,11 @@ struct kernfs_elem_dir {
 	 * better directly in kernfs_node but is here to save space.
 	 */
 	struct kernfs_root	*root;
+	/*
+	 * Monotonic revision counter, used to identify if a directory
+	 * node has changed during negative dentry revalidation.
+	 */
+	unsigned long		rev;
 };
 
 struct kernfs_elem_symlink {

[PATCH v7 3/6] kernfs: use VFS negative dentry caching

[Ian Kent]

If there are many lookups for non-existent paths these negative lookups
can lead to a lot of overhead during path walks.

The VFS allows dentries to be created as negative and hashed, and caches
them so they can be used to reduce the fairly high overhead alloc/free
cycle that occurs during these lookups.

Use the kernfs node parent revision to identify if a change has been
made to the containing directory so that the negative dentry can be
discarded and the lookup redone.

Signed-off-by: Ian Kent <raven@themaw.net>
---
 fs/kernfs/dir.c |   52 ++++++++++++++++++++++++++++++++--------------------
 1 file changed, 32 insertions(+), 20 deletions(-)

diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
index 48704c5b6a072..f9352c9e90581 100644
--- a/fs/kernfs/dir.c
+++ b/fs/kernfs/dir.c
@@ -1039,9 +1039,28 @@ static int kernfs_dop_revalidate(struct dentry *dentry, unsigned int flags)
 	if (flags & LOOKUP_RCU)
 		return -ECHILD;
 
-	/* Always perform fresh lookup for negatives */
-	if (d_really_is_negative(dentry))
-		goto out_bad_unlocked;
+	/* Negative hashed dentry? */
+	if (d_really_is_negative(dentry)) {
+		struct kernfs_node *parent;
+
+		/* If the kernfs parent node has changed discard and
+		 * proceed to ->lookup.
+		 */
+		spin_lock(&dentry->d_lock);
+		parent = kernfs_dentry_node(dentry->d_parent);
+		if (parent) {
+			if (kernfs_dir_changed(parent, dentry)) {
+				spin_unlock(&dentry->d_lock);
+				return 0;
+			}
+		}
+		spin_unlock(&dentry->d_lock);
+
+		/* The kernfs parent node hasn't changed, leave the
+		 * dentry negative and return success.
+		 */
+		return 1;
+	}
 
 	kn = kernfs_dentry_node(dentry);
 	mutex_lock(&kernfs_mutex);
@@ -1067,7 +1086,6 @@ static int kernfs_dop_revalidate(struct dentry *dentry, unsigned int flags)
 	return 1;
 out_bad:
 	mutex_unlock(&kernfs_mutex);
-out_bad_unlocked:
 	return 0;
 }
 
@@ -1082,33 +1100,27 @@ static struct dentry *kernfs_iop_lookup(struct inode *dir,
 	struct dentry *ret;
 	struct kernfs_node *parent = dir->i_private;
 	struct kernfs_node *kn;
-	struct inode *inode;
+	struct inode *inode = NULL;
 	const void *ns = NULL;
 
 	mutex_lock(&kernfs_mutex);
-
 	if (kernfs_ns_enabled(parent))
 		ns = kernfs_info(dir->i_sb)->ns;
 
 	kn = kernfs_find_ns(parent, dentry->d_name.name, ns);
-
-	/* no such entry */
-	if (!kn || !kernfs_active(kn)) {
-		ret = NULL;
-		goto out_unlock;
-	}
-
 	/* attach dentry and inode */
-	inode = kernfs_get_inode(dir->i_sb, kn);
-	if (!inode) {
-		ret = ERR_PTR(-ENOMEM);
-		goto out_unlock;
+	if (kn && kernfs_active(kn)) {
+		inode = kernfs_get_inode(dir->i_sb, kn);
+		if (!inode)
+			inode = ERR_PTR(-ENOMEM);
 	}
-
-	/* instantiate and hash dentry */
+	/* Needed only for negative dentry validation */
+	if (!inode)
+		kernfs_set_rev(parent, dentry);
+	/* instantiate and hash (possibly negative) dentry */
 	ret = d_splice_alias(inode, dentry);
- out_unlock:
 	mutex_unlock(&kernfs_mutex);
+
 	return ret;
 }
 

[PATCH v7 4/6] kernfs: switch kernfs to use an rwsem

[Ian Kent]

The kernfs global lock restricts the ability to perform kernfs node
lookup operations in parallel during path walks.

Change the kernfs mutex to an rwsem so that, when opportunity arises,
node searches can be done in parallel with path walk lookups.

Signed-off-by: Ian Kent <raven@themaw.net>
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
---
 fs/kernfs/dir.c             |   94 ++++++++++++++++++++++---------------------
 fs/kernfs/file.c            |    4 +-
 fs/kernfs/inode.c           |   16 ++++---
 fs/kernfs/kernfs-internal.h |    5 +-
 fs/kernfs/mount.c           |   12 +++--
 fs/kernfs/symlink.c         |    4 +-
 include/linux/kernfs.h      |    2 -
 7 files changed, 69 insertions(+), 68 deletions(-)

diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
index f9352c9e90581..a5080fb1dfc05 100644
--- a/fs/kernfs/dir.c
+++ b/fs/kernfs/dir.c
@@ -17,7 +17,7 @@
 
 #include "kernfs-internal.h"
 
-DEFINE_MUTEX(kernfs_mutex);
+DECLARE_RWSEM(kernfs_rwsem);
 static DEFINE_SPINLOCK(kernfs_rename_lock);	/* kn->parent and ->name */
 static char kernfs_pr_cont_buf[PATH_MAX];	/* protected by rename_lock */
 static DEFINE_SPINLOCK(kernfs_idr_lock);	/* root->ino_idr */
@@ -26,7 +26,7 @@ static DEFINE_SPINLOCK(kernfs_idr_lock);	/* root->ino_idr */
 
 static bool kernfs_active(struct kernfs_node *kn)
 {
-	lockdep_assert_held(&kernfs_mutex);
+	lockdep_assert_held(&kernfs_rwsem);
 	return atomic_read(&kn->active) >= 0;
 }
 
@@ -340,7 +340,7 @@ static int kernfs_sd_compare(const struct kernfs_node *left,
  *	@kn->parent->dir.children.
  *
  *	Locking:
- *	mutex_lock(kernfs_mutex)
+ *	kernfs_rwsem held exclusive
  *
  *	RETURNS:
  *	0 on susccess -EEXIST on failure.
@@ -386,7 +386,7 @@ static int kernfs_link_sibling(struct kernfs_node *kn)
  *	removed, %false if @kn wasn't on the rbtree.
  *
  *	Locking:
- *	mutex_lock(kernfs_mutex)
+ *	kernfs_rwsem held exclusive
  */
 static bool kernfs_unlink_sibling(struct kernfs_node *kn)
 {
@@ -457,14 +457,14 @@ void kernfs_put_active(struct kernfs_node *kn)
  * return after draining is complete.
  */
 static void kernfs_drain(struct kernfs_node *kn)
-	__releases(&kernfs_mutex) __acquires(&kernfs_mutex)
+	__releases(&kernfs_rwsem) __acquires(&kernfs_rwsem)
 {
 	struct kernfs_root *root = kernfs_root(kn);
 
-	lockdep_assert_held(&kernfs_mutex);
+	lockdep_assert_held_write(&kernfs_rwsem);
 	WARN_ON_ONCE(kernfs_active(kn));
 
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 
 	if (kernfs_lockdep(kn)) {
 		rwsem_acquire(&kn->dep_map, 0, 0, _RET_IP_);
@@ -483,7 +483,7 @@ static void kernfs_drain(struct kernfs_node *kn)
 
 	kernfs_drain_open_files(kn);
 
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 }
 
 /**
@@ -722,7 +722,7 @@ int kernfs_add_one(struct kernfs_node *kn)
 	bool has_ns;
 	int ret;
 
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 
 	ret = -EINVAL;
 	has_ns = kernfs_ns_enabled(parent);
@@ -753,7 +753,7 @@ int kernfs_add_one(struct kernfs_node *kn)
 		ps_iattr->ia_mtime = ps_iattr->ia_ctime;
 	}
 
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 
 	/*
 	 * Activate the new node unless CREATE_DEACTIVATED is requested.
@@ -767,7 +767,7 @@ int kernfs_add_one(struct kernfs_node *kn)
 	return 0;
 
 out_unlock:
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 	return ret;
 }
 
@@ -788,7 +788,7 @@ static struct kernfs_node *kernfs_find_ns(struct kernfs_node *parent,
 	bool has_ns = kernfs_ns_enabled(parent);
 	unsigned int hash;
 
-	lockdep_assert_held(&kernfs_mutex);
+	lockdep_assert_held(&kernfs_rwsem);
 
 	if (has_ns != (bool)ns) {
 		WARN(1, KERN_WARNING "kernfs: ns %s in '%s' for '%s'\n",
@@ -820,7 +820,7 @@ static struct kernfs_node *kernfs_walk_ns(struct kernfs_node *parent,
 	size_t len;
 	char *p, *name;
 
-	lockdep_assert_held(&kernfs_mutex);
+	lockdep_assert_held_read(&kernfs_rwsem);
 
 	/* grab kernfs_rename_lock to piggy back on kernfs_pr_cont_buf */
 	spin_lock_irq(&kernfs_rename_lock);
@@ -860,10 +860,10 @@ struct kernfs_node *kernfs_find_and_get_ns(struct kernfs_node *parent,
 {
 	struct kernfs_node *kn;
 
-	mutex_lock(&kernfs_mutex);
+	down_read(&kernfs_rwsem);
 	kn = kernfs_find_ns(parent, name, ns);
 	kernfs_get(kn);
-	mutex_unlock(&kernfs_mutex);
+	up_read(&kernfs_rwsem);
 
 	return kn;
 }
@@ -884,10 +884,10 @@ struct kernfs_node *kernfs_walk_and_get_ns(struct kernfs_node *parent,
 {
 	struct kernfs_node *kn;
 
-	mutex_lock(&kernfs_mutex);
+	down_read(&kernfs_rwsem);
 	kn = kernfs_walk_ns(parent, path, ns);
 	kernfs_get(kn);
-	mutex_unlock(&kernfs_mutex);
+	up_read(&kernfs_rwsem);
 
 	return kn;
 }
@@ -1063,7 +1063,7 @@ static int kernfs_dop_revalidate(struct dentry *dentry, unsigned int flags)
 	}
 
 	kn = kernfs_dentry_node(dentry);
-	mutex_lock(&kernfs_mutex);
+	down_read(&kernfs_rwsem);
 
 	/* The kernfs node has been deactivated */
 	if (!kernfs_active(kn))
@@ -1082,10 +1082,10 @@ static int kernfs_dop_revalidate(struct dentry *dentry, unsigned int flags)
 	    kernfs_info(dentry->d_sb)->ns != kn->ns)
 		goto out_bad;
 
-	mutex_unlock(&kernfs_mutex);
+	up_read(&kernfs_rwsem);
 	return 1;
 out_bad:
-	mutex_unlock(&kernfs_mutex);
+	up_read(&kernfs_rwsem);
 	return 0;
 }
 
@@ -1103,7 +1103,7 @@ static struct dentry *kernfs_iop_lookup(struct inode *dir,
 	struct inode *inode = NULL;
 	const void *ns = NULL;
 
-	mutex_lock(&kernfs_mutex);
+	down_read(&kernfs_rwsem);
 	if (kernfs_ns_enabled(parent))
 		ns = kernfs_info(dir->i_sb)->ns;
 
@@ -1119,7 +1119,7 @@ static struct dentry *kernfs_iop_lookup(struct inode *dir,
 		kernfs_set_rev(parent, dentry);
 	/* instantiate and hash (possibly negative) dentry */
 	ret = d_splice_alias(inode, dentry);
-	mutex_unlock(&kernfs_mutex);
+	up_read(&kernfs_rwsem);
 
 	return ret;
 }
@@ -1241,7 +1241,7 @@ static struct kernfs_node *kernfs_next_descendant_post(struct kernfs_node *pos,
 {
 	struct rb_node *rbn;
 
-	lockdep_assert_held(&kernfs_mutex);
+	lockdep_assert_held_write(&kernfs_rwsem);
 
 	/* if first iteration, visit leftmost descendant which may be root */
 	if (!pos)
@@ -1277,7 +1277,7 @@ void kernfs_activate(struct kernfs_node *kn)
 {
 	struct kernfs_node *pos;
 
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 
 	pos = NULL;
 	while ((pos = kernfs_next_descendant_post(pos, kn))) {
@@ -1291,14 +1291,14 @@ void kernfs_activate(struct kernfs_node *kn)
 		pos->flags |= KERNFS_ACTIVATED;
 	}
 
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 }
 
 static void __kernfs_remove(struct kernfs_node *kn)
 {
 	struct kernfs_node *pos;
 
-	lockdep_assert_held(&kernfs_mutex);
+	lockdep_assert_held_write(&kernfs_rwsem);
 
 	/*
 	 * Short-circuit if non-root @kn has already finished removal.
@@ -1321,7 +1321,7 @@ static void __kernfs_remove(struct kernfs_node *kn)
 		pos = kernfs_leftmost_descendant(kn);
 
 		/*
-		 * kernfs_drain() drops kernfs_mutex temporarily and @pos's
+		 * kernfs_drain() drops kernfs_rwsem temporarily and @pos's
 		 * base ref could have been put by someone else by the time
 		 * the function returns.  Make sure it doesn't go away
 		 * underneath us.
@@ -1368,9 +1368,9 @@ static void __kernfs_remove(struct kernfs_node *kn)
  */
 void kernfs_remove(struct kernfs_node *kn)
 {
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 	__kernfs_remove(kn);
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 }
 
 /**
@@ -1457,17 +1457,17 @@ bool kernfs_remove_self(struct kernfs_node *kn)
 {
 	bool ret;
 
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 	kernfs_break_active_protection(kn);
 
 	/*
 	 * SUICIDAL is used to arbitrate among competing invocations.  Only
 	 * the first one will actually perform removal.  When the removal
 	 * is complete, SUICIDED is set and the active ref is restored
-	 * while holding kernfs_mutex.  The ones which lost arbitration
-	 * waits for SUICDED && drained which can happen only after the
-	 * enclosing kernfs operation which executed the winning instance
-	 * of kernfs_remove_self() finished.
+	 * while kernfs_rwsem for held exclusive.  The ones which lost
+	 * arbitration waits for SUICIDED && drained which can happen only
+	 * after the enclosing kernfs operation which executed the winning
+	 * instance of kernfs_remove_self() finished.
 	 */
 	if (!(kn->flags & KERNFS_SUICIDAL)) {
 		kn->flags |= KERNFS_SUICIDAL;
@@ -1485,9 +1485,9 @@ bool kernfs_remove_self(struct kernfs_node *kn)
 			    atomic_read(&kn->active) == KN_DEACTIVATED_BIAS)
 				break;
 
-			mutex_unlock(&kernfs_mutex);
+			up_write(&kernfs_rwsem);
 			schedule();
-			mutex_lock(&kernfs_mutex);
+			down_write(&kernfs_rwsem);
 		}
 		finish_wait(waitq, &wait);
 		WARN_ON_ONCE(!RB_EMPTY_NODE(&kn->rb));
@@ -1495,12 +1495,12 @@ bool kernfs_remove_self(struct kernfs_node *kn)
 	}
 
 	/*
-	 * This must be done while holding kernfs_mutex; otherwise, waiting
-	 * for SUICIDED && deactivated could finish prematurely.
+	 * This must be done while kernfs_rwsem held exclusive; otherwise,
+	 * waiting for SUICIDED && deactivated could finish prematurely.
 	 */
 	kernfs_unbreak_active_protection(kn);
 
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 	return ret;
 }
 
@@ -1524,13 +1524,13 @@ int kernfs_remove_by_name_ns(struct kernfs_node *parent, const char *name,
 		return -ENOENT;
 	}
 
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 
 	kn = kernfs_find_ns(parent, name, ns);
 	if (kn)
 		__kernfs_remove(kn);
 
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 
 	if (kn)
 		return 0;
@@ -1556,7 +1556,7 @@ int kernfs_rename_ns(struct kernfs_node *kn, struct kernfs_node *new_parent,
 	if (!kn->parent)
 		return -EINVAL;
 
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 
 	error = -ENOENT;
 	if (!kernfs_active(kn) || !kernfs_active(new_parent) ||
@@ -1611,7 +1611,7 @@ int kernfs_rename_ns(struct kernfs_node *kn, struct kernfs_node *new_parent,
 
 	error = 0;
  out:
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 	return error;
 }
 
@@ -1686,7 +1686,7 @@ static int kernfs_fop_readdir(struct file *file, struct dir_context *ctx)
 
 	if (!dir_emit_dots(file, ctx))
 		return 0;
-	mutex_lock(&kernfs_mutex);
+	down_read(&kernfs_rwsem);
 
 	if (kernfs_ns_enabled(parent))
 		ns = kernfs_info(dentry->d_sb)->ns;
@@ -1703,12 +1703,12 @@ static int kernfs_fop_readdir(struct file *file, struct dir_context *ctx)
 		file->private_data = pos;
 		kernfs_get(pos);
 
-		mutex_unlock(&kernfs_mutex);
+		up_read(&kernfs_rwsem);
 		if (!dir_emit(ctx, name, len, ino, type))
 			return 0;
-		mutex_lock(&kernfs_mutex);
+		down_read(&kernfs_rwsem);
 	}
-	mutex_unlock(&kernfs_mutex);
+	up_read(&kernfs_rwsem);
 	file->private_data = NULL;
 	ctx->pos = INT_MAX;
 	return 0;
diff --git a/fs/kernfs/file.c b/fs/kernfs/file.c
index c757193121475..60e2a86c535eb 100644
--- a/fs/kernfs/file.c
+++ b/fs/kernfs/file.c
@@ -860,7 +860,7 @@ static void kernfs_notify_workfn(struct work_struct *work)
 	spin_unlock_irq(&kernfs_notify_lock);
 
 	/* kick fsnotify */
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 
 	list_for_each_entry(info, &kernfs_root(kn)->supers, node) {
 		struct kernfs_node *parent;
@@ -898,7 +898,7 @@ static void kernfs_notify_workfn(struct work_struct *work)
 		iput(inode);
 	}
 
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 	kernfs_put(kn);
 	goto repeat;
 }
diff --git a/fs/kernfs/inode.c b/fs/kernfs/inode.c
index d73950fc3d57d..3b01e9e61f14e 100644
--- a/fs/kernfs/inode.c
+++ b/fs/kernfs/inode.c
@@ -106,9 +106,9 @@ int kernfs_setattr(struct kernfs_node *kn, const struct iattr *iattr)
 {
 	int ret;
 
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 	ret = __kernfs_setattr(kn, iattr);
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 	return ret;
 }
 
@@ -122,7 +122,7 @@ int kernfs_iop_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 	if (!kn)
 		return -EINVAL;
 
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 	error = setattr_prepare(&init_user_ns, dentry, iattr);
 	if (error)
 		goto out;
@@ -135,7 +135,7 @@ int kernfs_iop_setattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 	setattr_copy(&init_user_ns, inode, iattr);
 
 out:
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 	return error;
 }
 
@@ -191,9 +191,9 @@ int kernfs_iop_getattr(struct user_namespace *mnt_userns,
 	struct inode *inode = d_inode(path->dentry);
 	struct kernfs_node *kn = inode->i_private;
 
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 	kernfs_refresh_inode(kn, inode);
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 
 	generic_fillattr(&init_user_ns, inode, stat);
 	return 0;
@@ -284,9 +284,9 @@ int kernfs_iop_permission(struct user_namespace *mnt_userns,
 
 	kn = inode->i_private;
 
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 	kernfs_refresh_inode(kn, inode);
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 
 	return generic_permission(&init_user_ns, inode, mask);
 }
diff --git a/fs/kernfs/kernfs-internal.h b/fs/kernfs/kernfs-internal.h
index 6a38897b4d02c..eaaa2e1607e8c 100644
--- a/fs/kernfs/kernfs-internal.h
+++ b/fs/kernfs/kernfs-internal.h
@@ -13,6 +13,7 @@
 #include <linux/lockdep.h>
 #include <linux/fs.h>
 #include <linux/mutex.h>
+#include <linux/rwsem.h>
 #include <linux/xattr.h>
 
 #include <linux/kernfs.h>
@@ -69,7 +70,7 @@ struct kernfs_super_info {
 	 */
 	const void		*ns;
 
-	/* anchored at kernfs_root->supers, protected by kernfs_mutex */
+	/* anchored at kernfs_root->supers, protected by kernfs_rwsem */
 	struct list_head	node;
 };
 #define kernfs_info(SB) ((struct kernfs_super_info *)(SB->s_fs_info))
@@ -121,7 +122,7 @@ int __kernfs_setattr(struct kernfs_node *kn, const struct iattr *iattr);
 /*
  * dir.c
  */
-extern struct mutex kernfs_mutex;
+extern struct rw_semaphore kernfs_rwsem;
 extern const struct dentry_operations kernfs_dops;
 extern const struct file_operations kernfs_dir_fops;
 extern const struct inode_operations kernfs_dir_iops;
diff --git a/fs/kernfs/mount.c b/fs/kernfs/mount.c
index 9dc7e7a64e10f..baa4155ba2edf 100644
--- a/fs/kernfs/mount.c
+++ b/fs/kernfs/mount.c
@@ -255,9 +255,9 @@ static int kernfs_fill_super(struct super_block *sb, struct kernfs_fs_context *k
 	sb->s_shrink.seeks = 0;
 
 	/* get root inode, initialize and unlock it */
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 	inode = kernfs_get_inode(sb, info->root->kn);
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 	if (!inode) {
 		pr_debug("kernfs: could not get root inode\n");
 		return -ENOMEM;
@@ -344,9 +344,9 @@ int kernfs_get_tree(struct fs_context *fc)
 		}
 		sb->s_flags |= SB_ACTIVE;
 
-		mutex_lock(&kernfs_mutex);
+		down_write(&kernfs_rwsem);
 		list_add(&info->node, &info->root->supers);
-		mutex_unlock(&kernfs_mutex);
+		up_write(&kernfs_rwsem);
 	}
 
 	fc->root = dget(sb->s_root);
@@ -372,9 +372,9 @@ void kernfs_kill_sb(struct super_block *sb)
 {
 	struct kernfs_super_info *info = kernfs_info(sb);
 
-	mutex_lock(&kernfs_mutex);
+	down_write(&kernfs_rwsem);
 	list_del(&info->node);
-	mutex_unlock(&kernfs_mutex);
+	up_write(&kernfs_rwsem);
 
 	/*
 	 * Remove the superblock from fs_supers/s_instances
diff --git a/fs/kernfs/symlink.c b/fs/kernfs/symlink.c
index 5432883d819f2..c8f8e41b84110 100644
--- a/fs/kernfs/symlink.c
+++ b/fs/kernfs/symlink.c
@@ -116,9 +116,9 @@ static int kernfs_getlink(struct inode *inode, char *path)
 	struct kernfs_node *target = kn->symlink.target_kn;
 	int error;
 
-	mutex_lock(&kernfs_mutex);
+	down_read(&kernfs_rwsem);
 	error = kernfs_get_target_path(parent, target, path);
-	mutex_unlock(&kernfs_mutex);
+	up_read(&kernfs_rwsem);
 
 	return error;
 }
diff --git a/include/linux/kernfs.h b/include/linux/kernfs.h
index d68b4ad095737..1093abf7c28cc 100644
--- a/include/linux/kernfs.h
+++ b/include/linux/kernfs.h
@@ -193,7 +193,7 @@ struct kernfs_root {
 	u32			id_highbits;
 	struct kernfs_syscall_ops *syscall_ops;
 
-	/* list of kernfs_super_info of this root, protected by kernfs_mutex */
+	/* list of kernfs_super_info of this root, protected by kernfs_rwsem */
 	struct list_head	supers;
 
 	wait_queue_head_t	deactivate_waitq;

[PATCH v7 5/6] kernfs: use i_lock to protect concurrent inode updates

[Ian Kent]

The inode operations .permission() and .getattr() use the kernfs node
write lock but all that's needed is the read lock to protect against
partial updates of these kernfs node fields which are all done under
the write lock.

And .permission() is called frequently during path walks and can cause
quite a bit of contention between kernfs node operations and path
walks when the number of concurrent walks is high.

To change kernfs_iop_getattr() and kernfs_iop_permission() to take
the rw sem read lock instead of the write lock an additional lock is
needed to protect against multiple processes concurrently updating
the inode attributes and link count in kernfs_refresh_inode().

The inode i_lock seems like the sensible thing to use to protect these
inode attribute updates so use it in kernfs_refresh_inode().

The last hunk in the patch, applied to kernfs_fill_super(), is possibly
not needed but taking the lock was present originally. I prefer to
continue to take it to protect against a partial update of the source
kernfs fields during the call to kernfs_refresh_inode() made by
kernfs_get_inode().

Signed-off-by: Ian Kent <raven@themaw.net>
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
---
 fs/kernfs/inode.c |   18 ++++++++++++------
 fs/kernfs/mount.c |    4 ++--
 2 files changed, 14 insertions(+), 8 deletions(-)

diff --git a/fs/kernfs/inode.c b/fs/kernfs/inode.c
index 3b01e9e61f14e..6592938511909 100644
--- a/fs/kernfs/inode.c
+++ b/fs/kernfs/inode.c
@@ -191,11 +191,13 @@ int kernfs_iop_getattr(struct user_namespace *mnt_userns,
 	struct inode *inode = d_inode(path->dentry);
 	struct kernfs_node *kn = inode->i_private;
 
-	down_write(&kernfs_rwsem);
+	down_read(&kernfs_rwsem);
+	spin_lock(&inode->i_lock);
 	kernfs_refresh_inode(kn, inode);
-	up_write(&kernfs_rwsem);
-
 	generic_fillattr(&init_user_ns, inode, stat);
+	spin_unlock(&inode->i_lock);
+	up_read(&kernfs_rwsem);
+
 	return 0;
 }
 
@@ -278,17 +280,21 @@ int kernfs_iop_permission(struct user_namespace *mnt_userns,
 			  struct inode *inode, int mask)
 {
 	struct kernfs_node *kn;
+	int ret;
 
 	if (mask & MAY_NOT_BLOCK)
 		return -ECHILD;
 
 	kn = inode->i_private;
 
-	down_write(&kernfs_rwsem);
+	down_read(&kernfs_rwsem);
+	spin_lock(&inode->i_lock);
 	kernfs_refresh_inode(kn, inode);
-	up_write(&kernfs_rwsem);
+	ret = generic_permission(&init_user_ns, inode, mask);
+	spin_unlock(&inode->i_lock);
+	up_read(&kernfs_rwsem);
 
-	return generic_permission(&init_user_ns, inode, mask);
+	return ret;
 }
 
 int kernfs_xattr_get(struct kernfs_node *kn, const char *name,
diff --git a/fs/kernfs/mount.c b/fs/kernfs/mount.c
index baa4155ba2edf..f2f909d09f522 100644
--- a/fs/kernfs/mount.c
+++ b/fs/kernfs/mount.c
@@ -255,9 +255,9 @@ static int kernfs_fill_super(struct super_block *sb, struct kernfs_fs_context *k
 	sb->s_shrink.seeks = 0;
 
 	/* get root inode, initialize and unlock it */
-	down_write(&kernfs_rwsem);
+	down_read(&kernfs_rwsem);
 	inode = kernfs_get_inode(sb, info->root->kn);
-	up_write(&kernfs_rwsem);
+	up_read(&kernfs_rwsem);
 	if (!inode) {
 		pr_debug("kernfs: could not get root inode\n");
 		return -ENOMEM;

[PATCH v7 6/6] kernfs: dont call d_splice_alias() under kernfs node lock

[Ian Kent]

The call to d_splice_alias() in kernfs_iop_lookup() doesn't depend on
any kernfs node so there's no reason to hold the kernfs node lock when
calling it.

Signed-off-by: Ian Kent <raven@themaw.net>
Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
---
 fs/kernfs/dir.c |    6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
index a5080fb1dfc05..723348607033c 100644
--- a/fs/kernfs/dir.c
+++ b/fs/kernfs/dir.c
@@ -1097,7 +1097,6 @@ static struct dentry *kernfs_iop_lookup(struct inode *dir,
 					struct dentry *dentry,
 					unsigned int flags)
 {
-	struct dentry *ret;
 	struct kernfs_node *parent = dir->i_private;
 	struct kernfs_node *kn;
 	struct inode *inode = NULL;
@@ -1117,11 +1116,10 @@ static struct dentry *kernfs_iop_lookup(struct inode *dir,
 	/* Needed only for negative dentry validation */
 	if (!inode)
 		kernfs_set_rev(parent, dentry);
-	/* instantiate and hash (possibly negative) dentry */
-	ret = d_splice_alias(inode, dentry);
 	up_read(&kernfs_rwsem);
 
-	return ret;
+	/* instantiate and hash (possibly negative) dentry */
+	return d_splice_alias(inode, dentry);
 }
 
 static int kernfs_iop_mkdir(struct user_namespace *mnt_userns,

Re: [PATCH v7 1/6] kernfs: move revalidate to be near lookup

[Greg Kroah-Hartman]

On Tue, Jun 15, 2021 at 06:25:53PM +0800, Ian Kent wrote:
> While the dentry operation kernfs_dop_revalidate() is grouped with
> dentry type functions it also has a strong affinity to the inode
> operation ->lookup().
> 
> It makes sense to locate this function near to kernfs_iop_lookup()
> because we will be adding VFS negative dentry caching to reduce path
> lookup overhead for non-existent paths.
> 
> There's no functional change from this patch.
> 
> Signed-off-by: Ian Kent <raven@themaw.net>
> Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>
> ---
>  fs/kernfs/dir.c |   86 ++++++++++++++++++++++++++++---------------------------
>  1 file changed, 43 insertions(+), 43 deletions(-)

As everyone has agreed this patch does nothing wrong, I've applied this
one for now while everyone reviews the other ones :)

thanks,

greg k-h

Re: [PATCH v7 2/6] kernfs: add a revision to identify directory node changes

[Miklos Szeredi]

On Tue, 15 Jun 2021 at 12:26, Ian Kent <raven@themaw.net> wrote:
>
> Add a revision counter to kernfs directory nodes so it can be used
> to detect if a directory node has changed during negative dentry
> revalidation.
>
> There's an assumption that sizeof(unsigned long) <= sizeof(pointer)
> on all architectures and as far as I know that assumption holds.
>
> So adding a revision counter to the struct kernfs_elem_dir variant of
> the kernfs_node type union won't increase the size of the kernfs_node
> struct. This is because struct kernfs_elem_dir is at least
> sizeof(pointer) smaller than the largest union variant. It's tempting
> to make the revision counter a u64 but that would increase the size of
> kernfs_node on archs where sizeof(pointer) is smaller than the revision
> counter.
>
> Signed-off-by: Ian Kent <raven@themaw.net>

Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>

Re: [PATCH v7 3/6] kernfs: use VFS negative dentry caching

[Miklos Szeredi]

On Tue, 15 Jun 2021 at 12:26, Ian Kent <raven@themaw.net> wrote:
>
> If there are many lookups for non-existent paths these negative lookups
> can lead to a lot of overhead during path walks.
>
> The VFS allows dentries to be created as negative and hashed, and caches
> them so they can be used to reduce the fairly high overhead alloc/free
> cycle that occurs during these lookups.
>
> Use the kernfs node parent revision to identify if a change has been
> made to the containing directory so that the negative dentry can be
> discarded and the lookup redone.
>
> Signed-off-by: Ian Kent <raven@themaw.net>

Reviewed-by: Miklos Szeredi <mszeredi@redhat.com>