This allows namespace data structure to be shared by cephfs inodes
with same layout.
Signed-off-by: Yan, Zheng <zyan@xxxxxxxxxx>
---
include/linux/ceph/libceph.h | 1 +
include/linux/ceph/pool_ns.h | 47 +++++++++++++++
net/ceph/Makefile | 2 +-
net/ceph/ceph_common.c | 2 +
net/ceph/pool_ns.c | 133 +++++++++++++++++++++++++++++++++++++++++++
5 files changed, 184 insertions(+), 1 deletion(-)
create mode 100644 include/linux/ceph/pool_ns.h
create mode 100644 net/ceph/pool_ns.c
diff --git a/include/linux/ceph/libceph.h b/include/linux/ceph/libceph.h
index e7975e4..40e0d84 100644
--- a/include/linux/ceph/libceph.h
+++ b/include/linux/ceph/libceph.h
@@ -21,6 +21,7 @@
#include <linux/ceph/mon_client.h>
#include <linux/ceph/osd_client.h>
#include <linux/ceph/ceph_fs.h>
+#include <linux/ceph/pool_ns.h>
/*
* mount options
diff --git a/include/linux/ceph/pool_ns.h b/include/linux/ceph/pool_ns.h
new file mode 100644
index 0000000..627cf9d
--- /dev/null
+++ b/include/linux/ceph/pool_ns.h
@@ -0,0 +1,47 @@
+#ifndef _FS_CEPH_POOL_NS_H
+#define _FS_CEPH_POOL_NS_H
+
+#include <linux/types.h>
+#include <linux/kref.h>
+#include <linux/rbtree.h>
+#include <linux/rcupdate.h>
+
+struct ceph_pool_ns {
+ struct kref kref;
+ union {
+ struct rb_node node;
+ struct rcu_head rcu;
+ };
+ size_t name_len;
+ char name[];
+};
+
+extern void ceph_release_pool_ns(struct kref *ref);
+extern struct ceph_pool_ns *ceph_find_or_create_pool_ns(const char *str,
+ size_t len);
+extern struct ceph_pool_ns *ceph_try_get_pool_ns(struct ceph_pool_ns **pns);
+extern void ceph_pool_ns_cleanup(void);
+
+static inline void ceph_get_pool_ns(struct ceph_pool_ns *ns)
+{
+ kref_get(&ns->kref);
+}
+
+static inline void ceph_put_pool_ns(struct ceph_pool_ns *ns)
+{
+ if (!ns)
+ return;
+ kref_put(&ns->kref, ceph_release_pool_ns);
+}
+
+static inline int ceph_compare_pool_ns(struct ceph_pool_ns *ns,
+ const char* str, size_t len)
+{
+ size_t nsl = ns ? ns->name_len : 0;
+ if (nsl != len)
+ return nsl - len;
+ if (nsl == 0)
+ return 0;
+ return strncmp(ns->name, str, len);
+}
+#endif
diff --git a/net/ceph/Makefile b/net/ceph/Makefile
index 958d9856..520dab6 100644
--- a/net/ceph/Makefile
+++ b/net/ceph/Makefile
@@ -11,5 +11,5 @@ libceph-y := ceph_common.o messenger.o msgpool.o buffer.o pagelist.o \
crypto.o armor.o \
auth_x.o \
ceph_fs.o ceph_strings.o ceph_hash.o \
- pagevec.o snapshot.o
+ pagevec.o snapshot.o pool_ns.o
diff --git a/net/ceph/ceph_common.c b/net/ceph/ceph_common.c
index dcc18c6..ac60293 100644
--- a/net/ceph/ceph_common.c
+++ b/net/ceph/ceph_common.c
@@ -25,6 +25,7 @@
#include <linux/ceph/decode.h>
#include <linux/ceph/mon_client.h>
#include <linux/ceph/auth.h>
+#include <linux/ceph/pool_ns.h>
#include "crypto.h"
@@ -751,6 +752,7 @@ static void __exit exit_ceph_lib(void)
ceph_msgr_exit();
ceph_crypto_shutdown();
ceph_debugfs_cleanup();
+ ceph_pool_ns_cleanup();
}
module_init(init_ceph_lib);
diff --git a/net/ceph/pool_ns.c b/net/ceph/pool_ns.c
new file mode 100644
index 0000000..64ccd77
--- /dev/null
+++ b/net/ceph/pool_ns.c
@@ -0,0 +1,133 @@
+#include <linux/slab.h>
+#include <linux/gfp.h>
+#include <linux/string.h>
+#include <linux/spinlock.h>
+#include <linux/ceph/pool_ns.h>
+
+static DEFINE_SPINLOCK(pool_ns_lock);
+static struct rb_root pool_ns_tree = RB_ROOT;
+
+struct ceph_pool_ns *ceph_find_or_create_pool_ns(const char* name, size_t len)
+{
+ struct ceph_pool_ns *ns, *exist;
+ struct rb_node **p, *parent;
+ int ret;
+
+ exist = NULL;
+ spin_lock(&pool_ns_lock);
+ p = &pool_ns_tree.rb_node;
+ while (*p) {
+ exist = rb_entry(*p, struct ceph_pool_ns, node);
+ ret = ceph_compare_pool_ns(exist, name, len);
+ if (ret > 0)
+ p = &(*p)->rb_left;
+ else if (ret < 0)
+ p = &(*p)->rb_right;
+ else
+ break;
+ exist = NULL;
+ }
+ if (exist && !kref_get_unless_zero(&exist->kref)) {
+ rb_erase(&exist->node, &pool_ns_tree);
+ RB_CLEAR_NODE(&exist->node);
+ exist = NULL;
+ }
+ spin_unlock(&pool_ns_lock);
+ if (exist)
+ return exist;
+
+ ns = kmalloc(sizeof(*ns) + len + 1, GFP_NOFS);
+ if (!ns)
+ return NULL;
+
+ kref_init(&ns->kref);
+ ns->name_len = len;
+ memcpy(ns->name, name, len);
+ ns->name[len] = 0;
+
+retry:
+ exist = NULL;
+ parent = NULL;
+ p = &pool_ns_tree.rb_node;
+ spin_lock(&pool_ns_lock);
+ while (*p) {
+ parent = *p;
+ exist = rb_entry(*p, struct ceph_pool_ns, node);
+ ret = ceph_compare_pool_ns(exist, name, len);
+ if (ret > 0)
+ p = &(*p)->rb_left;
+ else if (ret < 0)
+ p = &(*p)->rb_right;
+ else
+ break;
+ exist = NULL;
+ }
+ ret = 0;
+ if (!exist) {
+ rb_link_node(&ns->node, parent, p);
+ rb_insert_color(&ns->node, &pool_ns_tree);
+ } else if (!kref_get_unless_zero(&exist->kref)) {
+ rb_erase(&exist->node, &pool_ns_tree);
+ RB_CLEAR_NODE(&exist->node);
+ ret = -EAGAIN;
+ }
+ spin_unlock(&pool_ns_lock);
+ if (ret == -EAGAIN)
+ goto retry;
+
+ if (exist) {
+ kfree(ns);
+ ns = exist;
+ }
+
+ return ns;
+}
+EXPORT_SYMBOL(ceph_find_or_create_pool_ns);
+
+static void ceph_free_pool_ns(struct rcu_head *head)
+{
+ struct ceph_pool_ns *ns = container_of(head, struct ceph_pool_ns, rcu);
+ kfree(ns);
+}
+
+void ceph_release_pool_ns(struct kref *ref)
+{
+ struct ceph_pool_ns *ns = container_of(ref, struct ceph_pool_ns, kref);
+
+ spin_lock(&pool_ns_lock);
+ if (!RB_EMPTY_NODE(&ns->node)) {
+ rb_erase(&ns->node, &pool_ns_tree);
+ RB_CLEAR_NODE(&ns->node);
+ }
+ spin_unlock(&pool_ns_lock);
+
+ call_rcu(&ns->rcu, ceph_free_pool_ns);
+}
+EXPORT_SYMBOL(ceph_release_pool_ns);
+
+struct ceph_pool_ns *ceph_try_get_pool_ns(struct ceph_pool_ns **pns)
+{
+ struct ceph_pool_ns *ns;
+ rcu_read_lock();
+ ns = rcu_dereference(*pns);
+ if (ns && !kref_get_unless_zero(&ns->kref))
+ ns = NULL;
+ rcu_read_unlock();
+ return ns;
+}
+EXPORT_SYMBOL(ceph_try_get_pool_ns);
+
+void ceph_pool_ns_cleanup(void)
+{
+ struct rb_node *p;
+ struct ceph_pool_ns *ns;
+ if (RB_EMPTY_ROOT(&pool_ns_tree))
+ return;
+
+ pr_err("libceph: detect pool ns leaks\n");
+ while ((p = rb_first(&pool_ns_tree))) {
+ ns = rb_entry(p, struct ceph_pool_ns, node);
+ rb_erase(p, &pool_ns_tree);
+ kfree(ns);
+ }
+}
--
2.5.0
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
