1 files changed, 203 insertions, 101 deletions

diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index 7bdefd0c789a..6c0e52fd0743 100644
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -703,29 +703,12 @@ void ceph_add_cap(struct inode *inode,
 		 */
 		struct ceph_snap_realm *realm = ceph_lookup_snap_realm(mdsc,
 							       realmino);
-		if (realm) {
-			struct ceph_snap_realm *oldrealm = ci->i_snap_realm;
-			if (oldrealm) {
-				spin_lock(&oldrealm->inodes_with_caps_lock);
-				list_del_init(&ci->i_snap_realm_item);
-				spin_unlock(&oldrealm->inodes_with_caps_lock);
-			}
-
-			spin_lock(&realm->inodes_with_caps_lock);
-			list_add(&ci->i_snap_realm_item,
-				 &realm->inodes_with_caps);
-			ci->i_snap_realm = realm;
-			if (realm->ino == ci->i_vino.ino)
-				realm->inode = inode;
-			spin_unlock(&realm->inodes_with_caps_lock);
-
-			if (oldrealm)
-				ceph_put_snap_realm(mdsc, oldrealm);
-		} else {
-			pr_err("ceph_add_cap: couldn't find snap realm %llx\n",
-			       realmino);
-			WARN_ON(!realm);
-		}
+		if (realm)
+			ceph_change_snap_realm(inode, realm);
+		else
+			WARN(1, "%s: couldn't find snap realm 0x%llx (ino 0x%llx oldrealm 0x%llx)\n",
+			     __func__, realmino, ci->i_vino.ino,
+			     ci->i_snap_realm ? ci->i_snap_realm->ino : 0);
 	}
 
 	__check_cap_issue(ci, cap, issued);
@@ -1112,20 +1095,6 @@ int ceph_is_any_caps(struct inode *inode)
 	return ret;
 }
 
-static void drop_inode_snap_realm(struct ceph_inode_info *ci)
-{
-	struct ceph_snap_realm *realm = ci->i_snap_realm;
-	spin_lock(&realm->inodes_with_caps_lock);
-	list_del_init(&ci->i_snap_realm_item);
-	ci->i_snap_realm_counter++;
-	ci->i_snap_realm = NULL;
-	if (realm->ino == ci->i_vino.ino)
-		realm->inode = NULL;
-	spin_unlock(&realm->inodes_with_caps_lock);
-	ceph_put_snap_realm(ceph_sb_to_client(ci->vfs_inode.i_sb)->mdsc,
-			    realm);
-}
-
 /*
  * Remove a cap.  Take steps to deal with a racing iterate_session_caps.
  *
@@ -1145,17 +1114,16 @@ void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release)
 		return;
 	}
 
+	lockdep_assert_held(&ci->i_ceph_lock);
+
 	dout("__ceph_remove_cap %p from %p\n", cap, &ci->vfs_inode);
 
 	mdsc = ceph_inode_to_client(&ci->vfs_inode)->mdsc;
 
 	/* remove from inode's cap rbtree, and clear auth cap */
 	rb_erase(&cap->ci_node, &ci->i_caps);
-	if (ci->i_auth_cap == cap) {
-		WARN_ON_ONCE(!list_empty(&ci->i_dirty_item) &&
-			     !mdsc->fsc->blocklisted);
+	if (ci->i_auth_cap == cap)
 		ci->i_auth_cap = NULL;
-	}
 
 	/* remove from session list */
 	spin_lock(&session->s_cap_lock);
@@ -1201,12 +1169,34 @@ void __ceph_remove_cap(struct ceph_cap *cap, bool queue_release)
 		 * keep i_snap_realm.
 		 */
 		if (ci->i_wr_ref == 0 && ci->i_snap_realm)
-			drop_inode_snap_realm(ci);
+			ceph_change_snap_realm(&ci->vfs_inode, NULL);
 
 		__cap_delay_cancel(mdsc, ci);
 	}
 }
 
+void ceph_remove_cap(struct ceph_cap *cap, bool queue_release)
+{
+	struct ceph_inode_info *ci = cap->ci;
+	struct ceph_fs_client *fsc;
+
+	/* 'ci' being NULL means the remove have already occurred */
+	if (!ci) {
+		dout("%s: cap inode is NULL\n", __func__);
+		return;
+	}
+
+	lockdep_assert_held(&ci->i_ceph_lock);
+
+	fsc = ceph_sb_to_client(ci->vfs_inode.i_sb);
+	WARN_ON_ONCE(ci->i_auth_cap == cap &&
+		     !list_empty(&ci->i_dirty_item) &&
+		     !fsc->blocklisted &&
+		     READ_ONCE(fsc->mount_state) != CEPH_MOUNT_SHUTDOWN);
+
+	__ceph_remove_cap(cap, queue_release);
+}
+
 struct cap_msg_args {
 	struct ceph_mds_session	*session;
 	u64			ino, cid, follows;
@@ -1335,7 +1325,7 @@ void __ceph_remove_caps(struct ceph_inode_info *ci)
 	while (p) {
 		struct ceph_cap *cap = rb_entry(p, struct ceph_cap, ci_node);
 		p = rb_next(p);
-		__ceph_remove_cap(cap, true);
+		ceph_remove_cap(cap, true);
 	}
 	spin_unlock(&ci->i_ceph_lock);
 }
@@ -1743,7 +1733,14 @@ int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask,
 
 struct ceph_cap_flush *ceph_alloc_cap_flush(void)
 {
-	return kmem_cache_alloc(ceph_cap_flush_cachep, GFP_KERNEL);
+	struct ceph_cap_flush *cf;
+
+	cf = kmem_cache_alloc(ceph_cap_flush_cachep, GFP_KERNEL);
+	if (!cf)
+		return NULL;
+
+	cf->is_capsnap = false;
+	return cf;
 }
 
 void ceph_free_cap_flush(struct ceph_cap_flush *cf)
@@ -1778,7 +1775,7 @@ static bool __detach_cap_flush_from_mdsc(struct ceph_mds_client *mdsc,
 		prev->wake = true;
 		wake = false;
 	}
-	list_del(&cf->g_list);
+	list_del_init(&cf->g_list);
 	return wake;
 }
 
@@ -1793,7 +1790,7 @@ static bool __detach_cap_flush_from_ci(struct ceph_inode_info *ci,
 		prev->wake = true;
 		wake = false;
 	}
-	list_del(&cf->i_list);
+	list_del_init(&cf->i_list);
 	return wake;
 }
 
@@ -1852,6 +1849,8 @@ static u64 __mark_caps_flushing(struct inode *inode,
  * try to invalidate mapping pages without blocking.
  */
 static int try_nonblocking_invalidate(struct inode *inode)
+	__releases(ci->i_ceph_lock)
+	__acquires(ci->i_ceph_lock)
 {
 	struct ceph_inode_info *ci = ceph_inode(inode);
 	u32 invalidating_gen = ci->i_rdcache_gen;
@@ -2215,6 +2214,7 @@ static int caps_are_flushed(struct inode *inode, u64 flush_tid)
  */
 static int unsafe_request_wait(struct inode *inode)
 {
+	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
 	struct ceph_inode_info *ci = ceph_inode(inode);
 	struct ceph_mds_request *req1 = NULL, *req2 = NULL;
 	int ret, err = 0;
@@ -2234,6 +2234,81 @@ static int unsafe_request_wait(struct inode *inode)
 	}
 	spin_unlock(&ci->i_unsafe_lock);
 
+	/*
+	 * Trigger to flush the journal logs in all the relevant MDSes
+	 * manually, or in the worst case we must wait at most 5 seconds
+	 * to wait the journal logs to be flushed by the MDSes periodically.
+	 */
+	if (req1 || req2) {
+		struct ceph_mds_session **sessions = NULL;
+		struct ceph_mds_session *s;
+		struct ceph_mds_request *req;
+		unsigned int max;
+		int i;
+
+		/*
+		 * The mdsc->max_sessions is unlikely to be changed
+		 * mostly, here we will retry it by reallocating the
+		 * sessions arrary memory to get rid of the mdsc->mutex
+		 * lock.
+		 */
+retry:
+		max = mdsc->max_sessions;
+		sessions = krealloc(sessions, max * sizeof(s), __GFP_ZERO);
+		if (!sessions)
+			return -ENOMEM;
+
+		spin_lock(&ci->i_unsafe_lock);
+		if (req1) {
+			list_for_each_entry(req, &ci->i_unsafe_dirops,
+					    r_unsafe_dir_item) {
+				s = req->r_session;
+				if (unlikely(s->s_mds > max)) {
+					spin_unlock(&ci->i_unsafe_lock);
+					goto retry;
+				}
+				if (!sessions[s->s_mds]) {
+					s = ceph_get_mds_session(s);
+					sessions[s->s_mds] = s;
+				}
+			}
+		}
+		if (req2) {
+			list_for_each_entry(req, &ci->i_unsafe_iops,
+					    r_unsafe_target_item) {
+				s = req->r_session;
+				if (unlikely(s->s_mds > max)) {
+					spin_unlock(&ci->i_unsafe_lock);
+					goto retry;
+				}
+				if (!sessions[s->s_mds]) {
+					s = ceph_get_mds_session(s);
+					sessions[s->s_mds] = s;
+				}
+			}
+		}
+		spin_unlock(&ci->i_unsafe_lock);
+
+		/* the auth MDS */
+		spin_lock(&ci->i_ceph_lock);
+		if (ci->i_auth_cap) {
+		      s = ci->i_auth_cap->session;
+		      if (!sessions[s->s_mds])
+			      sessions[s->s_mds] = ceph_get_mds_session(s);
+		}
+		spin_unlock(&ci->i_ceph_lock);
+
+		/* send flush mdlog request to MDSes */
+		for (i = 0; i < max; i++) {
+			s = sessions[i];
+			if (s) {
+				send_flush_mdlog(s);
+				ceph_put_mds_session(s);
+			}
+		}
+		kfree(sessions);
+	}
+
 	dout("unsafe_request_wait %p wait on tid %llu %llu\n",
 	     inode, req1 ? req1->r_tid : 0ULL, req2 ? req2->r_tid : 0ULL);
 	if (req1) {
@@ -2352,7 +2427,7 @@ static void __kick_flushing_caps(struct ceph_mds_client *mdsc,
 	ci->i_ceph_flags &= ~CEPH_I_KICK_FLUSH;
 
 	list_for_each_entry_reverse(cf, &ci->i_cap_flush_list, i_list) {
-		if (!cf->caps) {
+		if (cf->is_capsnap) {
 			last_snap_flush = cf->tid;
 			break;
 		}
@@ -2371,7 +2446,7 @@ static void __kick_flushing_caps(struct ceph_mds_client *mdsc,
 
 		first_tid = cf->tid + 1;
 
-		if (cf->caps) {
+		if (!cf->is_capsnap) {
 			struct cap_msg_args arg;
 
 			dout("kick_flushing_caps %p cap %p tid %llu %s\n",
@@ -3004,7 +3079,7 @@ static void __ceph_put_cap_refs(struct ceph_inode_info *ci, int had,
 			}
 			/* see comment in __ceph_remove_cap() */
 			if (!__ceph_is_any_real_caps(ci) && ci->i_snap_realm)
-				drop_inode_snap_realm(ci);
+				ceph_change_snap_realm(inode, NULL);
 		}
 	}
 	if (check_flushsnaps && __ceph_have_pending_cap_snap(ci)) {
@@ -3110,7 +3185,16 @@ void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
 				break;
 			}
 		}
-		BUG_ON(!found);
+
+		if (!found) {
+			/*
+			 * The capsnap should already be removed when removing
+			 * auth cap in the case of a forced unmount.
+			 */
+			WARN_ON_ONCE(ci->i_auth_cap);
+			goto unlock;
+		}
+
 		capsnap->dirty_pages -= nr;
 		if (capsnap->dirty_pages == 0) {
 			complete_capsnap = true;
@@ -3132,6 +3216,7 @@ void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
 		     complete_capsnap ? " (complete capsnap)" : "");
 	}
 
+unlock:
 	spin_unlock(&ci->i_ceph_lock);
 
 	if (last) {
@@ -3516,7 +3601,7 @@ static void handle_cap_flush_ack(struct inode *inode, u64 flush_tid,
 			cleaned = cf->caps;
 
 		/* Is this a capsnap? */
-		if (cf->caps == 0)
+		if (cf->is_capsnap)
 			continue;
 
 		if (cf->tid <= flush_tid) {
@@ -3589,8 +3674,9 @@ out:
 	while (!list_empty(&to_remove)) {
 		cf = list_first_entry(&to_remove,
 				      struct ceph_cap_flush, i_list);
-		list_del(&cf->i_list);
-		ceph_free_cap_flush(cf);
+		list_del_init(&cf->i_list);
+		if (!cf->is_capsnap)
+			ceph_free_cap_flush(cf);
 	}
 
 	if (wake_ci)
@@ -3601,6 +3687,43 @@ out:
 		iput(inode);
 }
 
+void __ceph_remove_capsnap(struct inode *inode, struct ceph_cap_snap *capsnap,
+			   bool *wake_ci, bool *wake_mdsc)
+{
+	struct ceph_inode_info *ci = ceph_inode(inode);
+	struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
+	bool ret;
+
+	lockdep_assert_held(&ci->i_ceph_lock);
+
+	dout("removing capsnap %p, inode %p ci %p\n", capsnap, inode, ci);
+
+	list_del_init(&capsnap->ci_item);
+	ret = __detach_cap_flush_from_ci(ci, &capsnap->cap_flush);
+	if (wake_ci)
+		*wake_ci = ret;
+
+	spin_lock(&mdsc->cap_dirty_lock);
+	if (list_empty(&ci->i_cap_flush_list))
+		list_del_init(&ci->i_flushing_item);
+
+	ret = __detach_cap_flush_from_mdsc(mdsc, &capsnap->cap_flush);
+	if (wake_mdsc)
+		*wake_mdsc = ret;
+	spin_unlock(&mdsc->cap_dirty_lock);
+}
+
+void ceph_remove_capsnap(struct inode *inode, struct ceph_cap_snap *capsnap,
+			 bool *wake_ci, bool *wake_mdsc)
+{
+	struct ceph_inode_info *ci = ceph_inode(inode);
+
+	lockdep_assert_held(&ci->i_ceph_lock);
+
+	WARN_ON_ONCE(capsnap->dirty_pages || capsnap->writing);
+	__ceph_remove_capsnap(inode, capsnap, wake_ci, wake_mdsc);
+}
+
 /*
  * Handle FLUSHSNAP_ACK.  MDS has flushed snap data to disk and we can
  * throw away our cap_snap.
@@ -3638,23 +3761,10 @@ static void handle_cap_flushsnap_ack(struct inode *inode, u64 flush_tid,
 			     capsnap, capsnap->follows);
 		}
 	}
-	if (flushed) {
-		WARN_ON(capsnap->dirty_pages || capsnap->writing);
-		dout(" removing %p cap_snap %p follows %lld\n",
-		     inode, capsnap, follows);
-		list_del(&capsnap->ci_item);
-		wake_ci |= __detach_cap_flush_from_ci(ci, &capsnap->cap_flush);
-
-		spin_lock(&mdsc->cap_dirty_lock);
-
-		if (list_empty(&ci->i_cap_flush_list))
-			list_del_init(&ci->i_flushing_item);
-
-		wake_mdsc |= __detach_cap_flush_from_mdsc(mdsc,
-							  &capsnap->cap_flush);
-		spin_unlock(&mdsc->cap_dirty_lock);
-	}
+	if (flushed)
+		ceph_remove_capsnap(inode, capsnap, &wake_ci, &wake_mdsc);
 	spin_unlock(&ci->i_ceph_lock);
+
 	if (flushed) {
 		ceph_put_snap_context(capsnap->context);
 		ceph_put_cap_snap(capsnap);
@@ -3738,7 +3848,7 @@ retry:
 		goto out_unlock;
 
 	if (target < 0) {
-		__ceph_remove_cap(cap, false);
+		ceph_remove_cap(cap, false);
 		goto out_unlock;
 	}
 
@@ -3773,7 +3883,7 @@ retry:
 				change_auth_cap_ses(ci, tcap->session);
 			}
 		}
-		__ceph_remove_cap(cap, false);
+		ceph_remove_cap(cap, false);
 		goto out_unlock;
 	} else if (tsession) {
 		/* add placeholder for the export tagert */
@@ -3790,7 +3900,7 @@ retry:
 			spin_unlock(&mdsc->cap_dirty_lock);
 		}
 
-		__ceph_remove_cap(cap, false);
+		ceph_remove_cap(cap, false);
 		goto out_unlock;
 	}
 
@@ -3901,7 +4011,7 @@ retry:
 					ocap->mseq, mds, le32_to_cpu(ph->seq),
 					le32_to_cpu(ph->mseq));
 		}
-		__ceph_remove_cap(ocap, (ph->flags & CEPH_CAP_FLAG_RELEASE));
+		ceph_remove_cap(ocap, (ph->flags & CEPH_CAP_FLAG_RELEASE));
 	}
 
 	*old_issued = issued;
@@ -4129,8 +4239,9 @@ void ceph_handle_caps(struct ceph_mds_session *session,
 done:
 	mutex_unlock(&session->s_mutex);
 done_unlocked:
-	ceph_put_string(extra_info.pool_ns);
 	iput(inode);
+out:
+	ceph_put_string(extra_info.pool_ns);
 	return;
 
 flush_cap_releases:
@@ -4145,16 +4256,24 @@ flush_cap_releases:
 bad:
 	pr_err("ceph_handle_caps: corrupt message\n");
 	ceph_msg_dump(msg);
-	return;
+	goto out;
 }
 
 /*
  * Delayed work handler to process end of delayed cap release LRU list.
+ *
+ * If new caps are added to the list while processing it, these won't get
+ * processed in this run.  In this case, the ci->i_hold_caps_max will be
+ * returned so that the work can be scheduled accordingly.
  */
-void ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
+unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
 {
 	struct inode *inode;
 	struct ceph_inode_info *ci;
+	struct ceph_mount_options *opt = mdsc->fsc->mount_options;
+	unsigned long delay_max = opt->caps_wanted_delay_max * HZ;
+	unsigned long loop_start = jiffies;
+	unsigned long delay = 0;
 
 	dout("check_delayed_caps\n");
 	spin_lock(&mdsc->cap_delay_lock);
@@ -4162,6 +4281,11 @@ void ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
 		ci = list_first_entry(&mdsc->cap_delay_list,
 				      struct ceph_inode_info,
 				      i_cap_delay_list);
+		if (time_before(loop_start, ci->i_hold_caps_max - delay_max)) {
+			dout("%s caps added recently.  Exiting loop", __func__);
+			delay = ci->i_hold_caps_max;
+			break;
+		}
 		if ((ci->i_ceph_flags & CEPH_I_FLUSH) == 0 &&
 		    time_before(jiffies, ci->i_hold_caps_max))
 			break;
@@ -4177,6 +4301,8 @@ void ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
 		}
 	}
 	spin_unlock(&mdsc->cap_delay_lock);
+
+	return delay;
 }
 
 /*
@@ -4205,33 +4331,9 @@ static void flush_dirty_session_caps(struct ceph_mds_session *s)
 	dout("flush_dirty_caps done\n");
 }
 
-static void iterate_sessions(struct ceph_mds_client *mdsc,
-			     void (*cb)(struct ceph_mds_session *))
-{
-	int mds;
-
-	mutex_lock(&mdsc->mutex);
-	for (mds = 0; mds < mdsc->max_sessions; ++mds) {
-		struct ceph_mds_session *s;
-
-		if (!mdsc->sessions[mds])
-			continue;
-
-		s = ceph_get_mds_session(mdsc->sessions[mds]);
-		if (!s)
-			continue;
-
-		mutex_unlock(&mdsc->mutex);
-		cb(s);
-		ceph_put_mds_session(s);
-		mutex_lock(&mdsc->mutex);
-	}
-	mutex_unlock(&mdsc->mutex);
-}
-
 void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc)
 {
-	iterate_sessions(mdsc, flush_dirty_session_caps);
+	ceph_mdsc_iterate_sessions(mdsc, flush_dirty_session_caps, true);
 }
 
 void __ceph_touch_fmode(struct ceph_inode_info *ci,