1 files changed, 120 insertions, 34 deletions

diff --git a/fs/afs/volume.c b/fs/afs/volume.c
index 92ca5e27573b..9bc0509e3634 100644
--- a/fs/afs/volume.c
+++ b/fs/afs/volume.c
@@ -13,6 +13,56 @@ unsigned __read_mostly afs_volume_gc_delay = 10;
 unsigned __read_mostly afs_volume_record_life = 60 * 60;
 
 /*
+ * Insert a volume into a cell.  If there's an existing volume record, that is
+ * returned instead with a ref held.
+ */
+static struct afs_volume *afs_insert_volume_into_cell(struct afs_cell *cell,
+						      struct afs_volume *volume)
+{
+	struct afs_volume *p;
+	struct rb_node *parent = NULL, **pp;
+
+	write_seqlock(&cell->volume_lock);
+
+	pp = &cell->volumes.rb_node;
+	while (*pp) {
+		parent = *pp;
+		p = rb_entry(parent, struct afs_volume, cell_node);
+		if (p->vid < volume->vid) {
+			pp = &(*pp)->rb_left;
+		} else if (p->vid > volume->vid) {
+			pp = &(*pp)->rb_right;
+		} else {
+			volume = afs_get_volume(p, afs_volume_trace_get_cell_insert);
+			goto found;
+		}
+	}
+
+	rb_link_node_rcu(&volume->cell_node, parent, pp);
+	rb_insert_color(&volume->cell_node, &cell->volumes);
+	hlist_add_head_rcu(&volume->proc_link, &cell->proc_volumes);
+
+found:
+	write_sequnlock(&cell->volume_lock);
+	return volume;
+
+}
+
+static void afs_remove_volume_from_cell(struct afs_volume *volume)
+{
+	struct afs_cell *cell = volume->cell;
+
+	if (!hlist_unhashed(&volume->proc_link)) {
+		trace_afs_volume(volume->vid, atomic_read(&volume->usage),
+				 afs_volume_trace_remove);
+		write_seqlock(&cell->volume_lock);
+		hlist_del_rcu(&volume->proc_link);
+		rb_erase(&volume->cell_node, &cell->volumes);
+		write_sequnlock(&cell->volume_lock);
+	}
+}
+
+/*
  * Allocate a volume record and load it up from a vldb record.
  */
 static struct afs_volume *afs_alloc_volume(struct afs_fs_context *params,
@@ -39,7 +89,7 @@ static struct afs_volume *afs_alloc_volume(struct afs_fs_context *params,
 	volume->name_len	= vldb->name_len;
 
 	atomic_set(&volume->usage, 1);
-	INIT_LIST_HEAD(&volume->proc_link);
+	INIT_HLIST_NODE(&volume->proc_link);
 	rwlock_init(&volume->servers_lock);
 	rwlock_init(&volume->cb_v_break_lock);
 	memcpy(volume->name, vldb->name, vldb->name_len + 1);
@@ -51,7 +101,8 @@ static struct afs_volume *afs_alloc_volume(struct afs_fs_context *params,
 	}
 
 	refcount_set(&slist->usage, 1);
-	volume->servers = slist;
+	rcu_assign_pointer(volume->servers, slist);
+	trace_afs_volume(volume->vid, 1, afs_volume_trace_alloc);
 	return volume;
 
 error_1:
@@ -62,6 +113,25 @@ error_0:
 }
 
 /*
+ * Look up or allocate a volume record.
+ */
+static struct afs_volume *afs_lookup_volume(struct afs_fs_context *params,
+					    struct afs_vldb_entry *vldb,
+					    unsigned long type_mask)
+{
+	struct afs_volume *candidate, *volume;
+
+	candidate = afs_alloc_volume(params, vldb, type_mask);
+	if (IS_ERR(candidate))
+		return candidate;
+
+	volume = afs_insert_volume_into_cell(params->cell, candidate);
+	if (volume != candidate)
+		afs_put_volume(params->net, candidate, afs_volume_trace_put_cell_dup);
+	return volume;
+}
+
+/*
  * Look up a VLDB record for a volume.
  */
 static struct afs_vldb_entry *afs_vl_lookup_vldb(struct afs_cell *cell,
@@ -138,7 +208,7 @@ struct afs_volume *afs_create_volume(struct afs_fs_context *params)
 	}
 
 	type_mask = 1UL << params->type;
-	volume = afs_alloc_volume(params, vldb, type_mask);
+	volume = afs_lookup_volume(params, vldb, type_mask);
 
 error:
 	kfree(vldb);
@@ -156,23 +226,42 @@ static void afs_destroy_volume(struct afs_net *net, struct afs_volume *volume)
 	ASSERTCMP(volume->cache, ==, NULL);
 #endif
 
-	afs_put_serverlist(net, volume->servers);
+	afs_remove_volume_from_cell(volume);
+	afs_put_serverlist(net, rcu_access_pointer(volume->servers));
 	afs_put_cell(net, volume->cell);
-	kfree(volume);
+	trace_afs_volume(volume->vid, atomic_read(&volume->usage),
+			 afs_volume_trace_free);
+	kfree_rcu(volume, rcu);
 
 	_leave(" [destroyed]");
 }
 
 /*
- * Drop a reference on a volume record.
+ * Get a reference on a volume record.
  */
-void afs_put_volume(struct afs_cell *cell, struct afs_volume *volume)
+struct afs_volume *afs_get_volume(struct afs_volume *volume,
+				  enum afs_volume_trace reason)
 {
 	if (volume) {
-		_enter("%s", volume->name);
+		int u = atomic_inc_return(&volume->usage);
+		trace_afs_volume(volume->vid, u, reason);
+	}
+	return volume;
+}
+
 
-		if (atomic_dec_and_test(&volume->usage))
-			afs_destroy_volume(cell->net, volume);
+/*
+ * Drop a reference on a volume record.
+ */
+void afs_put_volume(struct afs_net *net, struct afs_volume *volume,
+		    enum afs_volume_trace reason)
+{
+	if (volume) {
+		afs_volid_t vid = volume->vid;
+		int u = atomic_dec_return(&volume->usage);
+		trace_afs_volume(vid, u, reason);
+		if (u == 0)
+			afs_destroy_volume(net, volume);
 	}
 }
 
@@ -188,10 +277,6 @@ void afs_activate_volume(struct afs_volume *volume)
 					       NULL, 0,
 					       volume, 0, true);
 #endif
-
-	write_lock(&volume->cell->proc_lock);
-	list_add_tail(&volume->proc_link, &volume->cell->proc_volumes);
-	write_unlock(&volume->cell->proc_lock);
 }
 
 /*
@@ -201,10 +286,6 @@ void afs_deactivate_volume(struct afs_volume *volume)
 {
 	_enter("%s", volume->name);
 
-	write_lock(&volume->cell->proc_lock);
-	list_del_init(&volume->proc_link);
-	write_unlock(&volume->cell->proc_lock);
-
 #ifdef CONFIG_AFS_FSCACHE
 	fscache_relinquish_cookie(volume->cache, NULL,
 				  test_bit(AFS_VOLUME_DELETED, &volume->flags));
@@ -256,17 +337,17 @@ static int afs_update_volume_status(struct afs_volume *volume, struct key *key)
 	write_lock(&volume->servers_lock);
 
 	discard = new;
-	old = volume->servers;
+	old = rcu_dereference_protected(volume->servers,
+					lockdep_is_held(&volume->servers_lock));
 	if (afs_annotate_server_list(new, old)) {
 		new->seq = volume->servers_seq + 1;
-		volume->servers = new;
+		rcu_assign_pointer(volume->servers, new);
 		smp_wmb();
 		volume->servers_seq++;
 		discard = old;
 	}
 
 	volume->update_at = ktime_get_real_seconds() + afs_volume_record_life;
-	clear_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags);
 	write_unlock(&volume->servers_lock);
 	ret = 0;
 
@@ -281,25 +362,27 @@ error:
 /*
  * Make sure the volume record is up to date.
  */
-int afs_check_volume_status(struct afs_volume *volume, struct key *key)
+int afs_check_volume_status(struct afs_volume *volume, struct afs_operation *op)
 {
-	time64_t now = ktime_get_real_seconds();
 	int ret, retries = 0;
 
 	_enter("");
 
-	if (volume->update_at <= now)
-		set_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags);
-
 retry:
-	if (!test_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags) &&
-	    !test_bit(AFS_VOLUME_WAIT, &volume->flags)) {
-		_leave(" = 0");
-		return 0;
-	}
-
+	if (test_bit(AFS_VOLUME_WAIT, &volume->flags))
+		goto wait;
+	if (volume->update_at <= ktime_get_real_seconds() ||
+	    test_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags))
+		goto update;
+	_leave(" = 0");
+	return 0;
+
+update:
 	if (!test_and_set_bit_lock(AFS_VOLUME_UPDATING, &volume->flags)) {
-		ret = afs_update_volume_status(volume, key);
+		clear_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags);
+		ret = afs_update_volume_status(volume, op->key);
+		if (ret < 0)
+			set_bit(AFS_VOLUME_NEEDS_UPDATE, &volume->flags);
 		clear_bit_unlock(AFS_VOLUME_WAIT, &volume->flags);
 		clear_bit_unlock(AFS_VOLUME_UPDATING, &volume->flags);
 		wake_up_bit(&volume->flags, AFS_VOLUME_WAIT);
@@ -307,12 +390,15 @@ retry:
 		return ret;
 	}
 
+wait:
 	if (!test_bit(AFS_VOLUME_WAIT, &volume->flags)) {
 		_leave(" = 0 [no wait]");
 		return 0;
 	}
 
-	ret = wait_on_bit(&volume->flags, AFS_VOLUME_WAIT, TASK_INTERRUPTIBLE);
+	ret = wait_on_bit(&volume->flags, AFS_VOLUME_WAIT,
+			  (op->flags & AFS_OPERATION_UNINTR) ?
+			  TASK_UNINTERRUPTIBLE : TASK_INTERRUPTIBLE);
 	if (ret == -ERESTARTSYS) {
 		_leave(" = %d", ret);
 		return ret;