[PATCH 0/5] afs: Fixes and development

[PATCH 0/5] afs: Fixes and development

[David Howells]

Hi Al,

Here's a set of AFS patches, one fixes, but mostly development.  The fix is:

 (1) Fix AFS to build if CONFIG_PROC_FS=n.

and the development patches:

 (1) Make AFS show all of a server's addresses in /proc/fs/afs/servers.

 (2) Allow DNS upcalls to ask for IPv6 addresses.

 (3) Add directories to the dynamic root pseudo-superblock to represent the
     names of known cells.

     This needs checking as it adds a VFS function to look up a dentry
     without creating one if it doesn't exist.  This is used to find a
     dentry if it exists so that we can delete it.

 (4) Optimise the handling of change notifications from the server
     (callback breaks) to avoid doing a volume lookup for every callback
     broken if there are two or more with the same volume ID.

The patches can be found here also:

	http://git.kernel.org/cgit/linux/kernel/git/dhowells/linux-fs.git/log/?h=afs-next

David
---
David Howells (5):
      afs: Handle CONFIG_PROC_FS=n
      afs: Show all of a server's addresses in /proc/fs/afs/servers
      afs: Enable IPv6 DNS lookups
      afs: Display manually added cells in dynamic root mount
      afs: Optimise callback breaking by not repeating volume lookup


 fs/afs/Makefile       |    4 +-
 fs/afs/addr_list.c    |    2 -
 fs/afs/callback.c     |  110 ++++++++++++++++++++++++++++++++++++-------
 fs/afs/cell.c         |   12 +++--
 fs/afs/dynroot.c      |  126 ++++++++++++++++++++++++++++++++++++++++++++++++-
 fs/afs/internal.h     |   30 +++++++++++-
 fs/afs/main.c         |    2 -
 fs/afs/proc.c         |   10 +++-
 fs/afs/server.c       |    2 -
 fs/afs/super.c        |   19 ++++++-
 fs/namei.c            |   29 +++++++++++
 include/linux/namei.h |    1 
 12 files changed, 310 insertions(+), 37 deletions(-)

[PATCH 1/5] afs: Handle CONFIG_PROC_FS=n

[David Howells]

The AFS filesystem depends at the moment on /proc for configuration and
also presents information that way - however, this causes a compilation
failure if procfs is disabled.

Fix it so that the procfs bits aren't compiled in if procfs is disabled.

This means that you can't configure the AFS filesystem directly, but it is
still usable provided that an up-to-date keyutils is installed to look up
cells by SRV or AFSDB DNS records.

Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: David Howells <dhowells@redhat.com>
---

 fs/afs/Makefile   |    4 ++--
 fs/afs/internal.h |    8 ++++++++
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/fs/afs/Makefile b/fs/afs/Makefile
index 532acae25453..546874057bd3 100644
--- a/fs/afs/Makefile
+++ b/fs/afs/Makefile
@@ -5,7 +5,7 @@
 
 afs-cache-$(CONFIG_AFS_FSCACHE) := cache.o
 
-kafs-objs := \
+kafs-y := \
 	$(afs-cache-y) \
 	addr_list.o \
 	callback.o \
@@ -21,7 +21,6 @@ kafs-objs := \
 	main.o \
 	misc.o \
 	mntpt.o \
-	proc.o \
 	rotate.o \
 	rxrpc.o \
 	security.o \
@@ -34,4 +33,5 @@ kafs-objs := \
 	write.o \
 	xattr.o
 
+kafs-$(CONFIG_PROC_FS) += proc.o
 obj-$(CONFIG_AFS_FS)  := kafs.o
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index b51fb8587166..bc3d8c87a938 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -870,11 +870,19 @@ extern int afs_get_ipv4_interfaces(struct afs_net *, struct afs_interface *,
 /*
  * proc.c
  */
+#ifdef CONFIG_PROC_FS
 extern int __net_init afs_proc_init(struct afs_net *);
 extern void __net_exit afs_proc_cleanup(struct afs_net *);
 extern int afs_proc_cell_setup(struct afs_cell *);
 extern void afs_proc_cell_remove(struct afs_cell *);
 extern void afs_put_sysnames(struct afs_sysnames *);
+#else
+static inline int afs_proc_init(struct afs_net *net) { return 0; }
+static inline void afs_proc_cleanup(struct afs_net *net) {}
+static inline int afs_proc_cell_setup(struct afs_cell *cell) { return 0; }
+static inline void afs_proc_cell_remove(struct afs_cell *cell) {}
+static inline void afs_put_sysnames(struct afs_sysnames *sysnames) {}
+#endif
 
 /*
  * rotate.c

[PATCH 2/5] afs: Show all of a server's addresses in /proc/fs/afs/servers

[David Howells]

Show all of a server's addresses in /proc/fs/afs/servers, placing the
second plus addresses on padded lines of their own.  The current address is
marked with a star.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 fs/afs/proc.c |   10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/fs/afs/proc.c b/fs/afs/proc.c
index 3512b9b66caf..0c3285c8db95 100644
--- a/fs/afs/proc.c
+++ b/fs/afs/proc.c
@@ -326,6 +326,7 @@ static int afs_proc_servers_show(struct seq_file *m, void *v)
 {
 	struct afs_server *server;
 	struct afs_addr_list *alist;
+	int i;
 
 	if (v == SEQ_START_TOKEN) {
 		seq_puts(m, "UUID                                 USE ADDR\n");
@@ -334,10 +335,15 @@ static int afs_proc_servers_show(struct seq_file *m, void *v)
 
 	server = list_entry(v, struct afs_server, proc_link);
 	alist = rcu_dereference(server->addresses);
-	seq_printf(m, "%pU %3d %pISp\n",
+	seq_printf(m, "%pU %3d %pISpc%s\n",
 		   &server->uuid,
 		   atomic_read(&server->usage),
-		   &alist->addrs[alist->index].transport);
+		   &alist->addrs[0].transport,
+		   alist->index == 0 ? "*" : "");
+	for (i = 1; i < alist->nr_addrs; i++)
+		seq_printf(m, "                                         %pISpc%s\n",
+			   &alist->addrs[i].transport,
+			   alist->index == i ? "*" : "");
 	return 0;
 }
 

[PATCH 3/5] afs: Enable IPv6 DNS lookups

[David Howells]

Remove the restriction on DNS lookup upcalls that prevents ipv6 addresses
from being looked up.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 fs/afs/addr_list.c |    2 +-
 fs/afs/dynroot.c   |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/fs/afs/addr_list.c b/fs/afs/addr_list.c
index 7587fb665ff1..b49ccae07875 100644
--- a/fs/afs/addr_list.c
+++ b/fs/afs/addr_list.c
@@ -216,7 +216,7 @@ struct afs_addr_list *afs_dns_query(struct afs_cell *cell, time64_t *_expiry)
 	_enter("%s", cell->name);
 
 	ret = dns_query("afsdb", cell->name, cell->name_len,
-			"ipv4", &vllist, _expiry);
+			"", &vllist, _expiry);
 	if (ret < 0)
 		return ERR_PTR(ret);
 
diff --git a/fs/afs/dynroot.c b/fs/afs/dynroot.c
index 983f3946ab57..7425f416ed73 100644
--- a/fs/afs/dynroot.c
+++ b/fs/afs/dynroot.c
@@ -46,7 +46,7 @@ static int afs_probe_cell_name(struct dentry *dentry)
 		return 0;
 	}
 
-	ret = dns_query("afsdb", name, len, "ipv4", NULL, NULL);
+	ret = dns_query("afsdb", name, len, "", NULL, NULL);
 	if (ret == -ENODATA)
 		ret = -EDESTADDRREQ;
 	return ret;

[PATCH 4/5] afs: Display manually added cells in dynamic root mount

[David Howells]

Alter the dynroot mount so that cells created by manipulation of
/proc/fs/afs/cells and /proc/fs/afs/rootcell and by specification of a root
cell as a module parameter will cause directories for those cells to be
created in the dynamic root superblock for the network namespace[*].

To this end:

 (1) Only one dynamic root superblock is now created per network namespace
     and this is shared between all attempts to mount it.  This makes it
     easier to find the superblock to modify.

 (2) When a dynamic root superblock is created, the list of cells is walked
     and directories created for each cell already defined.

 (3) When a new cell is added, if a dynamic root superblock exists, a
     directory is created for it.

 (4) When a cell is destroyed, the directory is removed.

 (5) These directories are created by calling lookup_one_len() on the root
     dir which automatically creates them if they don't exist.

[*] Inasmuch as network namespaces are currently supported here.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 fs/afs/cell.c         |   12 +++--
 fs/afs/dynroot.c      |  124 +++++++++++++++++++++++++++++++++++++++++++++++++
 fs/afs/internal.h     |    7 ++-
 fs/afs/main.c         |    2 -
 fs/afs/super.c        |   19 ++++++--
 fs/namei.c            |   29 +++++++++++
 include/linux/namei.h |    1 
 7 files changed, 183 insertions(+), 11 deletions(-)

diff --git a/fs/afs/cell.c b/fs/afs/cell.c
index bb92b54d2a4a..f3d0bef16d78 100644
--- a/fs/afs/cell.c
+++ b/fs/afs/cell.c
@@ -15,6 +15,7 @@
 #include <linux/dns_resolver.h>
 #include <linux/sched.h>
 #include <linux/inet.h>
+#include <linux/namei.h>
 #include <keys/rxrpc-type.h>
 #include "internal.h"
 
@@ -531,9 +532,11 @@ static int afs_activate_cell(struct afs_net *net, struct afs_cell *cell)
 	ret = afs_proc_cell_setup(cell);
 	if (ret < 0)
 		return ret;
-	spin_lock(&net->proc_cells_lock);
+
+	mutex_lock(&net->proc_cells_lock);
 	list_add_tail(&cell->proc_link, &net->proc_cells);
-	spin_unlock(&net->proc_cells_lock);
+	afs_dynroot_mkdir(net, cell);
+	mutex_unlock(&net->proc_cells_lock);
 	return 0;
 }
 
@@ -546,9 +549,10 @@ static void afs_deactivate_cell(struct afs_net *net, struct afs_cell *cell)
 
 	afs_proc_cell_remove(cell);
 
-	spin_lock(&net->proc_cells_lock);
+	mutex_lock(&net->proc_cells_lock);
 	list_del_init(&cell->proc_link);
-	spin_unlock(&net->proc_cells_lock);
+	afs_dynroot_rmdir(net, cell);
+	mutex_unlock(&net->proc_cells_lock);
 
 #ifdef CONFIG_AFS_FSCACHE
 	fscache_relinquish_cookie(cell->cache, NULL, false);
diff --git a/fs/afs/dynroot.c b/fs/afs/dynroot.c
index 7425f416ed73..174e843f0633 100644
--- a/fs/afs/dynroot.c
+++ b/fs/afs/dynroot.c
@@ -1,4 +1,4 @@
-/* dir.c: AFS dynamic root handling
+/* AFS dynamic root handling
  *
  * Copyright (C) 2018 Red Hat, Inc. All Rights Reserved.
  * Written by David Howells (dhowells@redhat.com)
@@ -207,3 +207,125 @@ const struct dentry_operations afs_dynroot_dentry_operations = {
 	.d_release	= afs_d_release,
 	.d_automount	= afs_d_automount,
 };
+
+/*
+ * Create a manually added cell mount directory.
+ * - The caller must hold net->proc_cells_lock
+ */
+int afs_dynroot_mkdir(struct afs_net *net, struct afs_cell *cell)
+{
+	struct super_block *sb = net->dynroot_sb;
+	struct dentry *root, *subdir;
+	int ret;
+
+	if (!sb || atomic_read(&sb->s_active) == 0)
+		return 0;
+
+	/* Let the ->lookup op do the creation */
+	root = sb->s_root;
+	inode_lock(root->d_inode);
+	subdir = lookup_one_len(cell->name, root, cell->name_len);
+	if (IS_ERR(subdir)) {
+		ret = PTR_ERR(subdir);
+		goto unlock;
+	}
+
+	/* Note that we're retaining an extra ref on the dentry */
+	subdir->d_fsdata = (void *)1UL;
+	ret = 0;
+unlock:
+	inode_unlock(root->d_inode);
+	return ret;
+}
+
+/*
+ * Remove a manually added cell mount directory.
+ * - The caller must hold net->proc_cells_lock
+ */
+void afs_dynroot_rmdir(struct afs_net *net, struct afs_cell *cell)
+{
+	struct super_block *sb = net->dynroot_sb;
+	struct dentry *root, *subdir;
+
+	if (!sb || atomic_read(&sb->s_active) == 0)
+		return;
+
+	root = sb->s_root;
+	inode_lock(root->d_inode);
+
+	/* Don't want to trigger a lookup call, which will re-add the cell */
+	subdir = try_lookup_one_len(cell->name, root, cell->name_len);
+	if (IS_ERR_OR_NULL(subdir)) {
+		_debug("lookup %ld", PTR_ERR(subdir));
+		goto no_dentry;
+	}
+
+	_debug("rmdir %pd %u", subdir, d_count(subdir));
+
+	if (subdir->d_fsdata) {
+		_debug("unpin %u", d_count(subdir));
+		subdir->d_fsdata = NULL;
+		dput(subdir);
+	}
+	dput(subdir);
+no_dentry:
+	inode_unlock(root->d_inode);
+	_leave("");
+}
+
+/*
+ * Populate a newly created dynamic root with cell names.
+ */
+int afs_dynroot_populate(struct super_block *sb)
+{
+	struct afs_cell *cell;
+	struct afs_net *net = afs_sb2net(sb);
+	int ret;
+
+	if (mutex_lock_interruptible(&net->proc_cells_lock) < 0)
+		return -ERESTARTSYS;
+
+	net->dynroot_sb = sb;
+	list_for_each_entry(cell, &net->proc_cells, proc_link) {
+		ret = afs_dynroot_mkdir(net, cell);
+		if (ret < 0)
+			goto error;
+	}
+
+	ret = 0;
+out:
+	mutex_unlock(&net->proc_cells_lock);
+	return ret;
+
+error:
+	net->dynroot_sb = NULL;
+	goto out;
+}
+
+/*
+ * When a dynamic root that's in the process of being destroyed, depopulate it
+ * of pinned directories.
+ */
+void afs_dynroot_depopulate(struct super_block *sb)
+{
+	struct afs_net *net = afs_sb2net(sb);
+	struct dentry *root = sb->s_root, *subdir, *tmp;
+
+	/* Prevent more subdirs from being created */
+	mutex_lock(&net->proc_cells_lock);
+	if (net->dynroot_sb == sb)
+		net->dynroot_sb = NULL;
+	mutex_unlock(&net->proc_cells_lock);
+
+	inode_lock(root->d_inode);
+
+	/* Remove all the pins for dirs created for manually added cells */
+	list_for_each_entry_safe(subdir, tmp, &root->d_subdirs, d_child) {
+		if (subdir->d_fsdata) {
+			subdir->d_fsdata = NULL;
+			dput(subdir);
+		}
+	}
+
+	inode_unlock(root->d_inode);
+}
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index bc3d8c87a938..ab6bdf456f1a 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -239,7 +239,7 @@ struct afs_net {
 	atomic_t		cells_outstanding;
 	seqlock_t		cells_lock;
 
-	spinlock_t		proc_cells_lock;
+	struct mutex		proc_cells_lock;
 	struct list_head	proc_cells;
 
 	/* Known servers.  Theoretically each fileserver can only be in one
@@ -263,6 +263,7 @@ struct afs_net {
 	struct mutex		lock_manager_mutex;
 
 	/* Misc */
+	struct super_block	*dynroot_sb;	/* Dynamic root mount superblock */
 	struct proc_dir_entry	*proc_afs;	/* /proc/net/afs directory */
 	struct afs_sysnames	*sysnames;
 	rwlock_t		sysnames_lock;
@@ -721,6 +722,10 @@ extern const struct inode_operations afs_dynroot_inode_operations;
 extern const struct dentry_operations afs_dynroot_dentry_operations;
 
 extern struct inode *afs_try_auto_mntpt(struct dentry *, struct inode *);
+extern int afs_dynroot_mkdir(struct afs_net *, struct afs_cell *);
+extern void afs_dynroot_rmdir(struct afs_net *, struct afs_cell *);
+extern int afs_dynroot_populate(struct super_block *);
+extern void afs_dynroot_depopulate(struct super_block *);
 
 /*
  * file.c
diff --git a/fs/afs/main.c b/fs/afs/main.c
index 7d2c1354e2ca..e84fe822a960 100644
--- a/fs/afs/main.c
+++ b/fs/afs/main.c
@@ -86,7 +86,7 @@ static int __net_init afs_net_init(struct net *net_ns)
 	INIT_WORK(&net->cells_manager, afs_manage_cells);
 	timer_setup(&net->cells_timer, afs_cells_timer, 0);
 
-	spin_lock_init(&net->proc_cells_lock);
+	mutex_init(&net->proc_cells_lock);
 	INIT_LIST_HEAD(&net->proc_cells);
 
 	seqlock_init(&net->fs_lock);
diff --git a/fs/afs/super.c b/fs/afs/super.c
index 67135c2c3c9a..d524e4e6f290 100644
--- a/fs/afs/super.c
+++ b/fs/afs/super.c
@@ -365,12 +365,16 @@ static int afs_test_super(struct super_block *sb, struct fs_context *fc)
 
 	return (as->net_ns == fc->net_ns &&
 		as->volume &&
-		as->volume->vid == ctx->volume->vid);
+		as->volume->vid == ctx->volume->vid &&
+		!as->dyn_root);
 }
 
 static int afs_dynroot_test_super(struct super_block *sb, struct fs_context *fc)
 {
-	return false;
+	struct afs_super_info *as = AFS_FS_S(sb);
+
+	return (as->net_ns == fc->net_ns &&
+		as->dyn_root);
 }
 
 static int afs_set_super(struct super_block *sb, struct fs_context *fc)
@@ -426,10 +430,14 @@ static int afs_fill_super(struct super_block *sb, struct afs_fs_context *ctx)
 	if (!sb->s_root)
 		goto error;
 
-	if (as->dyn_root)
+	if (as->dyn_root) {
 		sb->s_d_op = &afs_dynroot_dentry_operations;
-	else
+		ret = afs_dynroot_populate(sb);
+		if (ret < 0)
+			goto error;
+	} else {
 		sb->s_d_op = &afs_fs_dentry_operations;
+	}
 
 	_leave(" = 0");
 	return 0;
@@ -472,6 +480,9 @@ static void afs_kill_super(struct super_block *sb)
 	struct afs_super_info *as = AFS_FS_S(sb);
 	struct afs_net *net = afs_net(as->net_ns);
 
+	if (as->dyn_root)
+		afs_dynroot_depopulate(sb);
+	
 	/* Clear the callback interests (which will do ilookup5) before
 	 * deactivating the superblock.
 	 */
diff --git a/fs/namei.c b/fs/namei.c
index a59968de1636..2794a3b265ee 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -2461,6 +2461,35 @@ static int lookup_one_len_common(const char *name, struct dentry *base,
 	return inode_permission(base->d_inode, MAY_EXEC);
 }
 
+/**
+ * try_lookup_one_len - filesystem helper to lookup single pathname component
+ * @name:	pathname component to lookup
+ * @base:	base directory to lookup from
+ * @len:	maximum length @len should be interpreted to
+ *
+ * Look up a dentry by name in the dcache, returning NULL if it does not
+ * currently exist.  The function does not try to create a dentry.
+ *
+ * Note that this routine is purely a helper for filesystem usage and should
+ * not be called by generic code.
+ *
+ * The caller must hold base->i_mutex.
+ */
+struct dentry *try_lookup_one_len(const char *name, struct dentry *base, int len)
+{
+	struct qstr this;
+	int err;
+
+	WARN_ON_ONCE(!inode_is_locked(base->d_inode));
+
+	err = lookup_one_len_common(name, base, len, &this);
+	if (err)
+		return ERR_PTR(err);
+
+	return lookup_dcache(&this, base, 0);
+}
+EXPORT_SYMBOL(try_lookup_one_len);
+
 /**
  * lookup_one_len - filesystem helper to lookup single pathname component
  * @name:	pathname component to lookup
diff --git a/include/linux/namei.h b/include/linux/namei.h
index a982bb7cd480..a78606e8e3df 100644
--- a/include/linux/namei.h
+++ b/include/linux/namei.h
@@ -81,6 +81,7 @@ extern void done_path_create(struct path *, struct dentry *);
 extern struct dentry *kern_path_locked(const char *, struct path *);
 extern int kern_path_mountpoint(int, const char *, struct path *, unsigned int);
 
+extern struct dentry *try_lookup_one_len(const char *, struct dentry *, int);
 extern struct dentry *lookup_one_len(const char *, struct dentry *, int);
 extern struct dentry *lookup_one_len_unlocked(const char *, struct dentry *, int);
 

[PATCH 5/5] afs: Optimise callback breaking by not repeating volume lookup

[David Howells]

At the moment, afs_break_callbacks calls afs_break_one_callback() for each
separate FID it was given, and the latter looks up the volume individually
for each one.

However, this is inefficient if two or more FIDs have the same vid as we
could reuse the volume.  This is complicated by cell aliasing whereby we
may have multiple cells sharing a volume and can therefore have multiple
callback interests for any particular volume ID.

At the moment afs_break_one_callback() scans the entire list of volumes
we're getting from a server and breaks the appropriate callback in every
matching volume, regardless of cell.  This scan is done for every FID.

Optimise callback breaking by the following means:

 (1) Sort the FID list by vid so that all FIDs belonging to the same volume
     are clumped together.

     This is done through the use of an indirection table as we cannot do
     an insertion sort on the afs_callback_break array as we decode FIDs
     into it as we subsequently also have to decode callback info into it
     that corresponds by array index only.

     We also don't really want to bubblesort afterwards if we can avoid it.

 (2) Sort the server->cb_interests array by vid so that all the matching
     volumes are grouped together.  This permits the scan to stop after
     finding a record that has a higher vid.

 (3) When breaking FIDs, we try to keep server->cb_break_lock as long as
     possible, caching the start point in the array for that volume group
     as long as possible.

     It might make sense to add another layer in that list and have a
     refcounted volume ID anchor that has the matching interests attached
     to it rather than being in the list.  This would allow the lock to be
     dropped without losing the cursor.

Signed-off-by: David Howells <dhowells@redhat.com>
---

 fs/afs/callback.c |  110 +++++++++++++++++++++++++++++++++++++++++++++--------
 fs/afs/internal.h |   15 ++++++-
 fs/afs/server.c   |    2 -
 3 files changed, 107 insertions(+), 20 deletions(-)

diff --git a/fs/afs/callback.c b/fs/afs/callback.c
index 571437dcb252..5f261fbf2182 100644
--- a/fs/afs/callback.c
+++ b/fs/afs/callback.c
@@ -20,6 +20,66 @@
 #include <linux/sched.h>
 #include "internal.h"
 
+/*
+ * Create volume and callback interests on a server.
+ */
+static struct afs_cb_interest *afs_create_interest(struct afs_server *server,
+						   struct afs_vnode *vnode)
+{
+	struct afs_vol_interest *new_vi, *vi;
+	struct afs_cb_interest *new;
+	struct hlist_node **pp;
+
+	new_vi = kzalloc(sizeof(struct afs_vol_interest), GFP_KERNEL);
+	if (!new_vi)
+		return NULL;
+
+	new = kzalloc(sizeof(struct afs_cb_interest), GFP_KERNEL);
+	if (!new) {
+		kfree(new_vi);
+		return NULL;
+	}
+
+	new_vi->usage = 1;
+	new_vi->vid = vnode->volume->vid;
+	INIT_HLIST_NODE(&new_vi->srv_link);
+	INIT_HLIST_HEAD(&new_vi->cb_interests);
+
+	refcount_set(&new->usage, 1);
+	new->sb = vnode->vfs_inode.i_sb;
+	new->vid = vnode->volume->vid;
+	new->server = afs_get_server(server);
+	INIT_HLIST_NODE(&new->cb_vlink);
+
+	write_lock(&server->cb_break_lock);
+
+	for (pp = &server->cb_volumes.first; *pp; pp = &(*pp)->next) {
+		vi = hlist_entry(*pp, struct afs_vol_interest, srv_link);
+		if (vi->vid < new_vi->vid)
+			continue;
+		if (vi->vid > new_vi->vid)
+			break;
+		vi->usage++;
+		goto found_vi;
+	}
+
+	new_vi->srv_link.pprev = pp;
+	new_vi->srv_link.next = *pp;
+	if (*pp)
+		(*pp)->pprev = &new_vi->srv_link.next;
+	*pp = &new_vi->srv_link;
+	vi = new_vi;
+	new_vi = NULL;
+found_vi:
+
+	new->vol_interest = vi;
+	hlist_add_head(&new->cb_vlink, &vi->cb_interests);
+
+	write_unlock(&server->cb_break_lock);
+	kfree(new_vi);
+	return new;
+}
+
 /*
  * Set up an interest-in-callbacks record for a volume on a server and
  * register it with the server.
@@ -77,20 +137,10 @@ int afs_register_server_cb_interest(struct afs_vnode *vnode,
 	}
 
 	if (!cbi) {
-		new = kzalloc(sizeof(struct afs_cb_interest), GFP_KERNEL);
+		new = afs_create_interest(server, vnode);
 		if (!new)
 			return -ENOMEM;
 
-		refcount_set(&new->usage, 1);
-		new->sb = vnode->vfs_inode.i_sb;
-		new->vid = vnode->volume->vid;
-		new->server = afs_get_server(server);
-		INIT_LIST_HEAD(&new->cb_link);
-
-		write_lock(&server->cb_break_lock);
-		list_add_tail(&new->cb_link, &server->cb_interests);
-		write_unlock(&server->cb_break_lock);
-
 		write_lock(&slist->lock);
 		if (!entry->cb_interest) {
 			entry->cb_interest = afs_get_cb_interest(new);
@@ -126,11 +176,22 @@ int afs_register_server_cb_interest(struct afs_vnode *vnode,
  */
 void afs_put_cb_interest(struct afs_net *net, struct afs_cb_interest *cbi)
 {
+	struct afs_vol_interest *vi;
+
 	if (cbi && refcount_dec_and_test(&cbi->usage)) {
-		if (!list_empty(&cbi->cb_link)) {
+		if (!hlist_unhashed(&cbi->cb_vlink)) {
 			write_lock(&cbi->server->cb_break_lock);
-			list_del_init(&cbi->cb_link);
+
+			hlist_del_init(&cbi->cb_vlink);
+			vi = cbi->vol_interest;
+			cbi->vol_interest = NULL;
+			if (--vi->usage == 0)
+				hlist_del(&vi->srv_link);
+			else
+				vi = NULL;
+
 			write_unlock(&cbi->server->cb_break_lock);
+			kfree(vi);
 			afs_put_server(net, cbi->server);
 		}
 		kfree(cbi);
@@ -182,20 +243,34 @@ void afs_break_callback(struct afs_vnode *vnode)
 static void afs_break_one_callback(struct afs_server *server,
 				   struct afs_fid *fid)
 {
+	struct afs_vol_interest *vi;
 	struct afs_cb_interest *cbi;
 	struct afs_iget_data data;
 	struct afs_vnode *vnode;
 	struct inode *inode;
 
 	read_lock(&server->cb_break_lock);
+	hlist_for_each_entry(vi, &server->cb_volumes, srv_link) {
+		if (vi->vid < fid->vid)
+			continue;
+		if (vi->vid > fid->vid) {
+			vi = NULL;
+			break;
+		}
+		//atomic_inc(&vi->usage);
+		break;
+	}
+
+	/* TODO: Find all matching volumes if we couldn't match the server and
+	 * break them anyway.
+	 */
+	if (!vi)
+		goto out;
 
 	/* Step through all interested superblocks.  There may be more than one
 	 * because of cell aliasing.
 	 */
-	list_for_each_entry(cbi, &server->cb_interests, cb_link) {
-		if (cbi->vid != fid->vid)
-			continue;
-
+	hlist_for_each_entry(cbi, &vi->cb_interests, cb_vlink) {
 		if (fid->vnode == 0 && fid->unique == 0) {
 			/* The callback break applies to an entire volume. */
 			struct afs_super_info *as = AFS_FS_S(cbi->sb);
@@ -217,6 +292,7 @@ static void afs_break_one_callback(struct afs_server *server,
 		}
 	}
 
+out:
 	read_unlock(&server->cb_break_lock);
 }
 
diff --git a/fs/afs/internal.h b/fs/afs/internal.h
index ab6bdf456f1a..e35d59761d47 100644
--- a/fs/afs/internal.h
+++ b/fs/afs/internal.h
@@ -406,16 +406,27 @@ struct afs_server {
 	rwlock_t		fs_lock;	/* access lock */
 
 	/* callback promise management */
-	struct list_head	cb_interests;	/* List of superblocks using this server */
+	struct hlist_head	cb_volumes;	/* List of volume interests on this server */
 	unsigned		cb_s_break;	/* Break-everything counter. */
 	rwlock_t		cb_break_lock;	/* Volume finding lock */
 };
 
+/*
+ * Volume collation in the server's callback interest list.
+ */
+struct afs_vol_interest {
+	struct hlist_node	srv_link;	/* Link in server->cb_volumes */
+	struct hlist_head	cb_interests;	/* List of callback interests on the server */
+	afs_volid_t		vid;		/* Volume ID to match */
+	unsigned int		usage;
+};
+
 /*
  * Interest by a superblock on a server.
  */
 struct afs_cb_interest {
-	struct list_head	cb_link;	/* Link in server->cb_interests */
+	struct hlist_node	cb_vlink;	/* Link in vol_interest->cb_interests */
+	struct afs_vol_interest	*vol_interest;
 	struct afs_server	*server;	/* Server on which this interest resides */
 	struct super_block	*sb;		/* Superblock on which inodes reside */
 	afs_volid_t		vid;		/* Volume ID to match */
diff --git a/fs/afs/server.c b/fs/afs/server.c
index 3af4625e2f8c..1d329e6981d5 100644
--- a/fs/afs/server.c
+++ b/fs/afs/server.c
@@ -228,7 +228,7 @@ static struct afs_server *afs_alloc_server(struct afs_net *net,
 	server->flags = (1UL << AFS_SERVER_FL_NEW);
 	server->update_at = ktime_get_real_seconds() + afs_server_update_delay;
 	rwlock_init(&server->fs_lock);
-	INIT_LIST_HEAD(&server->cb_interests);
+	INIT_HLIST_HEAD(&server->cb_volumes);
 	rwlock_init(&server->cb_break_lock);
 
 	afs_inc_servers_outstanding(net);