ceph: maintain i_head_snapc when any caps are dirty, not just for data

We used to use i_head_snapc to keep track of which snapc the current epoch
of dirty data was dirtied under.  It is used by queue_cap_snap to set up
the cap_snap.  However, since we queue cap snaps for any dirty caps, not
just for dirty file data, we need to keep a valid i_head_snapc anytime
we have dirty|flushing caps.  This fixes a NULL pointer deref in
queue_cap_snap when writing back dirty caps without data (e.g.,
snaptest-authwb.sh).

Signed-off-by: Sage Weil <sage@newdream.net>

diff --git a/fs/ceph/addr.c b/fs/ceph/addr.c
index 420d469..4cfce1e 100644 (file)
--- a/fs/ceph/addr.c
+++ b/fs/ceph/addr.c
@@ -87,7 +87,7 @@ static int ceph_set_page_dirty(struct page *page)
 
        /* dirty the head */
        spin_lock(&inode->i_lock);
-       if (ci->i_wrbuffer_ref_head == 0)
+       if (ci->i_head_snapc == NULL)
                ci->i_head_snapc = ceph_get_snap_context(snapc);
        ++ci->i_wrbuffer_ref_head;
        if (ci->i_wrbuffer_ref == 0)
@@ -346,7 +346,7 @@ static struct ceph_snap_context *get_oldest_context(struct inode *inode,
                        break;
                }
        }
-       if (!snapc && ci->i_head_snapc) {
+       if (!snapc && ci->i_wrbuffer_ref_head) {
                snapc = ceph_get_snap_context(ci->i_head_snapc);
                dout(" head snapc %p has %d dirty pages\n",
                     snapc, ci->i_wrbuffer_ref_head);

diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index ba5bbf3..a2069b6 100644 (file)
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -1143,6 +1143,10 @@ static int __send_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap,
                for (i = 0; i < CEPH_CAP_BITS; i++)
                        if (flushing & (1 << i))
                                ci->i_cap_flush_tid[i] = flush_tid;
+
+               follows = ci->i_head_snapc->seq;
+       } else {
+               follows = 0;
        }
 
        keep = cap->implemented;
@@ -1156,7 +1160,6 @@ static int __send_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap,
        mtime = inode->i_mtime;
        atime = inode->i_atime;
        time_warp_seq = ci->i_time_warp_seq;
-       follows = ci->i_snap_realm->cached_context->seq;
        uid = inode->i_uid;
        gid = inode->i_gid;
        mode = inode->i_mode;
@@ -1332,7 +1335,11 @@ void __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask)
             ceph_cap_string(was | mask));
        ci->i_dirty_caps |= mask;
        if (was == 0) {
-               dout(" inode %p now dirty\n", &ci->vfs_inode);
+               if (!ci->i_head_snapc)
+                       ci->i_head_snapc = ceph_get_snap_context(
+                               ci->i_snap_realm->cached_context);
+               dout(" inode %p now dirty snapc %p\n", &ci->vfs_inode,
+                       ci->i_head_snapc);
                BUG_ON(!list_empty(&ci->i_dirty_item));
                spin_lock(&mdsc->cap_dirty_lock);
                list_add(&ci->i_dirty_item, &mdsc->cap_dirty);
@@ -2190,7 +2197,9 @@ void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
 
        if (ci->i_head_snapc == snapc) {
                ci->i_wrbuffer_ref_head -= nr;
-               if (!ci->i_wrbuffer_ref_head) {
+               if (ci->i_wrbuffer_ref_head == 0 &&
+                   ci->i_dirty_caps == 0 && ci->i_flushing_caps == 0) {
+                       BUG_ON(!ci->i_head_snapc);
                        ceph_put_snap_context(ci->i_head_snapc);
                        ci->i_head_snapc = NULL;
                }
@@ -2483,6 +2492,11 @@ static void handle_cap_flush_ack(struct inode *inode, u64 flush_tid,
                        dout(" inode %p now clean\n", inode);
                        BUG_ON(!list_empty(&ci->i_dirty_item));
                        drop = 1;
+                       if (ci->i_wrbuffer_ref_head == 0) {
+                               BUG_ON(!ci->i_head_snapc);
+                               ceph_put_snap_context(ci->i_head_snapc);
+                               ci->i_head_snapc = NULL;
+                       }
                } else {
                        BUG_ON(list_empty(&ci->i_dirty_item));
                }

diff --git a/fs/ceph/snap.c b/fs/ceph/snap.c
index 6bdbf3a..4868b9d 100644 (file)
--- a/fs/ceph/snap.c
+++ b/fs/ceph/snap.c
@@ -458,6 +458,8 @@ void ceph_queue_cap_snap(struct ceph_inode_info *ci)
                             CEPH_CAP_FILE_EXCL|CEPH_CAP_FILE_WR))) {
                struct ceph_snap_context *snapc = ci->i_head_snapc;
 
+               dout("queue_cap_snap %p cap_snap %p queuing under %p\n", inode,
+                    capsnap, snapc);
                igrab(inode);
                
                atomic_set(&capsnap->nref, 1);
@@ -489,7 +491,9 @@ void ceph_queue_cap_snap(struct ceph_inode_info *ci)
                capsnap->dirty_pages = ci->i_wrbuffer_ref_head;
                ci->i_wrbuffer_ref_head = 0;
                capsnap->context = snapc;
-               ci->i_head_snapc = NULL;
+               ci->i_head_snapc =
+                       ceph_get_snap_context(ci->i_snap_realm->cached_context);
+               dout(" new snapc is %p\n", ci->i_head_snapc);
                list_add_tail(&capsnap->ci_item, &ci->i_cap_snaps);
 
                if (used & CEPH_CAP_FILE_WR) {

diff --git a/fs/ceph/super.h b/fs/ceph/super.h
index b33929d..c33897a 100644 (file)
--- a/fs/ceph/super.h
+++ b/fs/ceph/super.h
@@ -344,7 +344,8 @@ struct ceph_inode_info {
        unsigned i_cap_exporting_issued;
        struct ceph_cap_reservation i_cap_migration_resv;
        struct list_head i_cap_snaps;   /* snapped state pending flush to mds */
-       struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 */
+       struct ceph_snap_context *i_head_snapc;  /* set if wr_buffer_head > 0 or
+                                                   dirty|flushing caps */
        unsigned i_snap_caps;           /* cap bits for snapped files */
 
        int i_nr_by_mode[CEPH_FILE_MODE_NUM];  /* open file counts */