ceph: introduce global empty snap context

commit 97c85a828f36bbfffe9d77b977b65a5872b6cad4 upstream.

Current snaphost code does not properly handle moving inode from one
empty snap realm to another empty snap realm. After changing inode's
snap realm, some dirty pages' snap context can be not equal to inode's
i_head_snap. This can trigger BUG() in ceph_put_wrbuffer_cap_refs()

The fix is introduce a global empty snap context for all empty snap
realm. This avoids triggering the BUG() for filesystem with no snapshot.

Fixes: http://tracker.ceph.com/issues/9928

Signed-off-by: Yan, Zheng <zyan@redhat.com>
Reviewed-by: Ilya Dryomov <idryomov@redhat.com>
[bwh: Backported to 3.2:
 - Adjust context
 - As we don't have ceph_create_snap_context(), open-code it in
   ceph_snap_init()]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>

diff --git a/fs/ceph/snap.c b/fs/ceph/snap.c
index e5206fc..d5df940 100644 (file)
--- a/fs/ceph/snap.c
+++ b/fs/ceph/snap.c
@@ -288,6 +288,9 @@ static int cmpu64_rev(const void *a, const void *b)
        return 0;
 }
 
        return 0;
 }
 

+
+static struct ceph_snap_context *empty_snapc;
+

 /*
  * build the snap context for a given realm.
  */
 /*
  * build the snap context for a given realm.
  */


@@ -329,6 +332,12 @@ static int build_snap_context(struct ceph_snap_realm *realm)
                return 0;
        }
 
                return 0;
        }
 

+       if (num == 0 && realm->seq == empty_snapc->seq) {
+               ceph_get_snap_context(empty_snapc);
+               snapc = empty_snapc;
+               goto done;
+       }
+

        /* alloc new snap context */
        err = -ENOMEM;
        if (num > (SIZE_MAX - sizeof(*snapc)) / sizeof(u64))
        /* alloc new snap context */
        err = -ENOMEM;
        if (num > (SIZE_MAX - sizeof(*snapc)) / sizeof(u64))


@@ -364,6 +373,7 @@ static int build_snap_context(struct ceph_snap_realm *realm)
        dout("build_snap_context %llx %p: %p seq %lld (%d snaps)\n",
             realm->ino, realm, snapc, snapc->seq, snapc->num_snaps);
 
        dout("build_snap_context %llx %p: %p seq %lld (%d snaps)\n",
             realm->ino, realm, snapc, snapc->seq, snapc->num_snaps);
 

+done:

        if (realm->cached_context)
                ceph_put_snap_context(realm->cached_context);
        realm->cached_context = snapc;
        if (realm->cached_context)
                ceph_put_snap_context(realm->cached_context);
        realm->cached_context = snapc;


@@ -465,6 +475,9 @@ void ceph_queue_cap_snap(struct ceph_inode_info *ci)
                   cap_snap.  lucky us. */
                dout("queue_cap_snap %p already pending\n", inode);
                kfree(capsnap);
                   cap_snap.  lucky us. */
                dout("queue_cap_snap %p already pending\n", inode);
                kfree(capsnap);

+       } else if (ci->i_snap_realm->cached_context == empty_snapc) {
+               dout("queue_cap_snap %p empty snapc\n", inode);
+               kfree(capsnap);

        } else if (dirty & (CEPH_CAP_AUTH_EXCL|CEPH_CAP_XATTR_EXCL|
                            CEPH_CAP_FILE_EXCL|CEPH_CAP_FILE_WR)) {
                struct ceph_snap_context *snapc = ci->i_head_snapc;
        } else if (dirty & (CEPH_CAP_AUTH_EXCL|CEPH_CAP_XATTR_EXCL|
                            CEPH_CAP_FILE_EXCL|CEPH_CAP_FILE_WR)) {
                struct ceph_snap_context *snapc = ci->i_head_snapc;


@@ -927,5 +940,17 @@ out:
        return;
 }
 
        return;
 }
 

+int __init ceph_snap_init(void)
+{
+       empty_snapc = kzalloc(sizeof(struct ceph_snap_context), GFP_NOFS);
+       if (!empty_snapc)
+               return -ENOMEM;
+       atomic_set(&empty_snapc->nref, 1);
+       empty_snapc->seq = 1;
+       return 0;
+}

 
 

-
+void ceph_snap_exit(void)
+{
+       ceph_put_snap_context(empty_snapc);
+}

diff --git a/fs/ceph/super.c b/fs/ceph/super.c
index de268a8..3c981db 100644 (file)
--- a/fs/ceph/super.c
+++ b/fs/ceph/super.c
@@ -911,14 +911,20 @@ static int __init init_ceph(void)
        if (ret)
                goto out;
 
        if (ret)
                goto out;
 

-       ret = register_filesystem(&ceph_fs_type);
+       ret = ceph_snap_init();

        if (ret)
                goto out_icache;
 
        if (ret)
                goto out_icache;
 

+       ret = register_filesystem(&ceph_fs_type);
+       if (ret)
+               goto out_snap;
+

        pr_info("loaded (mds proto %d)\n", CEPH_MDSC_PROTOCOL);
 
        return 0;
 
        pr_info("loaded (mds proto %d)\n", CEPH_MDSC_PROTOCOL);
 
        return 0;
 

+out_snap:
+       ceph_snap_exit();

 out_icache:
        destroy_caches();
 out:
 out_icache:
        destroy_caches();
 out:


@@ -929,6 +935,7 @@ static void __exit exit_ceph(void)
 {
        dout("exit_ceph\n");
        unregister_filesystem(&ceph_fs_type);
 {
        dout("exit_ceph\n");
        unregister_filesystem(&ceph_fs_type);

+       ceph_snap_exit();

        destroy_caches();
 }
 
        destroy_caches();
 }
 

diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index a097817..242df58 100644 (file)
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -677,6 +677,8 @@ extern void ceph_queue_cap_snap(struct ceph_inode_info *ci);
 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
                                  struct ceph_cap_snap *capsnap);
 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);
 extern int __ceph_finish_cap_snap(struct ceph_inode_info *ci,
                                  struct ceph_cap_snap *capsnap);
 extern void ceph_cleanup_empty_realms(struct ceph_mds_client *mdsc);

+extern int ceph_snap_init(void);
+extern void ceph_snap_exit(void);

 
 /*
  * a cap_snap is "pending" if it is still awaiting an in-progress
 
 /*
  * a cap_snap is "pending" if it is still awaiting an in-progress