Merge tag 'gfs2-v6.0-rc2-fixes' of git://git.kernel.org/pub/scm/linux…

…/kernel/git/gfs2/linux-gfs2

Pull gfs2 updates from Andreas Gruenbacher:

 - Make sure to initialize the filesystem work queues before registering
   the filesystem; this prevents them from being used uninitialized.

 - On filesystem withdraw: prevent a a double iput() and immediately
   reject pending locking requests that can no longer succeed.

 - Use TRY lock in gfs2_inode_lookup() to prevent a rare glock hang
   during evict.

 - During filesystem mount, explicitly make sure that the sb_bsize and
   sb_bsize_shift super block fields are consistent with each other.
   This prevents messy error messages during fuzz testing.

 - Switch from strlcpy to strscpy.

* tag 'gfs2-v6.0-rc2-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2:
  gfs2: Register fs after creating workqueues
  gfs2: Check sb_bsize_shift after reading superblock
  gfs2: Switch from strlcpy to strscpy
  gfs2: Clear flags when withdraw prevents xmote
  gfs2: Dequeue waiters when withdrawn
  gfs2: Prevent double iput for journal on error
  gfs2: Use TRY lock in gfs2_inode_lookup for UNLINKED inodes

fs/gfs2/glock.c

@@ -59,6 +59,8 @@ typedef void (*glock_examiner) (struct gfs2_glock * gl);
 
 static void do_xmote(struct gfs2_glock *gl, struct gfs2_holder *gh, unsigned int target);
 static void __gfs2_glock_dq(struct gfs2_holder *gh);
+static void handle_callback(struct gfs2_glock *gl, unsigned int state,
+			    unsigned long delay, bool remote);
 
 static struct dentry *gfs2_root;
 static struct workqueue_struct *glock_workqueue;
@@ -730,7 +732,8 @@ static bool is_system_glock(struct gfs2_glock *gl)
  *
  */
 
-static void do_xmote(struct gfs2_glock *gl, struct gfs2_holder *gh, unsigned int target)
+static void do_xmote(struct gfs2_glock *gl, struct gfs2_holder *gh,
+					 unsigned int target)
 __releases(&gl->gl_lockref.lock)
 __acquires(&gl->gl_lockref.lock)
 {
@@ -741,7 +744,8 @@ __acquires(&gl->gl_lockref.lock)
 
 	if (target != LM_ST_UNLOCKED && glock_blocked_by_withdraw(gl) &&
 	    gh && !(gh->gh_flags & LM_FLAG_NOEXP))
-		return;
+		goto skip_inval;
+
 	lck_flags &= (LM_FLAG_TRY | LM_FLAG_TRY_1CB | LM_FLAG_NOEXP |
 		      LM_FLAG_PRIORITY);
 	GLOCK_BUG_ON(gl, gl->gl_state == target);
@@ -826,6 +830,20 @@ __acquires(&gl->gl_lockref.lock)
 	    (target != LM_ST_UNLOCKED ||
 	     test_bit(SDF_WITHDRAW_RECOVERY, &sdp->sd_flags))) {
 		if (!is_system_glock(gl)) {
+			handle_callback(gl, LM_ST_UNLOCKED, 0, false); /* sets demote */
+			/*
+			 * Ordinarily, we would call dlm and its callback would call
+			 * finish_xmote, which would call state_change() to the new state.
+			 * Since we withdrew, we won't call dlm, so call state_change
+			 * manually, but to the UNLOCKED state we desire.
+			 */
+			state_change(gl, LM_ST_UNLOCKED);
+			/*
+			 * We skip telling dlm to do the locking, so we won't get a
+			 * reply that would otherwise clear GLF_LOCK. So we clear it here.
+			 */
+			clear_bit(GLF_LOCK, &gl->gl_flags);
+			clear_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags);
 			gfs2_glock_queue_work(gl, GL_GLOCK_DFT_HOLD);
 			goto out;
 		} else {
@@ -1018,16 +1036,18 @@ static void delete_work_func(struct work_struct *work)
 			if (gfs2_queue_delete_work(gl, 5 * HZ))
 				return;
 		}
-		goto out;
 	}
 
 	inode = gfs2_lookup_by_inum(sdp, no_addr, gl->gl_no_formal_ino,
 				    GFS2_BLKST_UNLINKED);
-	if (!IS_ERR_OR_NULL(inode)) {
+	if (IS_ERR(inode)) {
+		if (PTR_ERR(inode) == -EAGAIN &&
+			(gfs2_queue_delete_work(gl, 5 * HZ)))
+				return;
+	} else {
 		d_prune_aliases(inode);
 		iput(inode);
 	}
-out:
 	gfs2_glock_put(gl);
 }
 
@@ -2194,6 +2214,20 @@ static void dump_glock_func(struct gfs2_glock *gl)
 	dump_glock(NULL, gl, true);
 }
 
+static void withdraw_dq(struct gfs2_glock *gl)
+{
+	spin_lock(&gl->gl_lockref.lock);
+	if (!__lockref_is_dead(&gl->gl_lockref) &&
+	    glock_blocked_by_withdraw(gl))
+		do_error(gl, LM_OUT_ERROR); /* remove pending waiters */
+	spin_unlock(&gl->gl_lockref.lock);
+}
+
+void gfs2_gl_dq_holders(struct gfs2_sbd *sdp)
+{
+	glock_hash_walk(withdraw_dq, sdp);
+}
+
 /**
  * gfs2_gl_hash_clear - Empty out the glock hash table
  * @sdp: the filesystem

fs/gfs2/glock.h

@@ -274,6 +274,7 @@ extern void gfs2_cancel_delete_work(struct gfs2_glock *gl);
 extern bool gfs2_delete_work_queued(const struct gfs2_glock *gl);
 extern void gfs2_flush_delete_work(struct gfs2_sbd *sdp);
 extern void gfs2_gl_hash_clear(struct gfs2_sbd *sdp);
+extern void gfs2_gl_dq_holders(struct gfs2_sbd *sdp);
 extern void gfs2_glock_thaw(struct gfs2_sbd *sdp);
 extern void gfs2_glock_add_to_lru(struct gfs2_glock *gl);
 extern void gfs2_glock_free(struct gfs2_glock *gl);

fs/gfs2/inode.c

@@ -130,6 +130,7 @@ struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
 	if (inode->i_state & I_NEW) {
 		struct gfs2_sbd *sdp = GFS2_SB(inode);
 		struct gfs2_glock *io_gl;
+		int extra_flags = 0;
 
 		error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE,
 				       &ip->i_gl);
@@ -141,9 +142,12 @@ struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
 		if (unlikely(error))
 			goto fail;
 
-		if (blktype != GFS2_BLKST_UNLINKED)
+		if (blktype == GFS2_BLKST_UNLINKED)
+			extra_flags |= LM_FLAG_TRY;
+		else
 			gfs2_cancel_delete_work(io_gl);
-		error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT,
+		error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED,
+					   GL_EXACT | extra_flags,
 					   &ip->i_iopen_gh);
 		gfs2_glock_put(io_gl);
 		if (unlikely(error))
@@ -210,6 +214,8 @@ struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
 	return inode;
 
 fail:
+	if (error == GLR_TRYFAILED)
+		error = -EAGAIN;
 	if (gfs2_holder_initialized(&ip->i_iopen_gh))
 		gfs2_glock_dq_uninit(&ip->i_iopen_gh);
 	if (gfs2_holder_initialized(&i_gh))

fs/gfs2/main.c

@@ -151,14 +151,6 @@ static int __init init_gfs2_fs(void)
 	if (error)
 		goto fail_shrinker;
 
-	error = register_filesystem(&gfs2_fs_type);
-	if (error)
-		goto fail_fs1;
-
-	error = register_filesystem(&gfs2meta_fs_type);
-	if (error)
-		goto fail_fs2;
-
 	error = -ENOMEM;
 	gfs_recovery_wq = alloc_workqueue("gfs_recovery",
 					  WQ_MEM_RECLAIM | WQ_FREEZABLE, 0);
@@ -180,22 +172,30 @@ static int __init init_gfs2_fs(void)
 		goto fail_mempool;
 
 	gfs2_register_debugfs();
+	error = register_filesystem(&gfs2_fs_type);
+	if (error)
+		goto fail_fs1;
+
+	error = register_filesystem(&gfs2meta_fs_type);
+	if (error)
+		goto fail_fs2;
+
 
 	pr_info("GFS2 installed\n");
 
 	return 0;
 
+fail_fs2:
+	unregister_filesystem(&gfs2_fs_type);
+fail_fs1:
+	mempool_destroy(gfs2_page_pool);
 fail_mempool:
 	destroy_workqueue(gfs2_freeze_wq);
 fail_wq3:
 	destroy_workqueue(gfs2_control_wq);
 fail_wq2:
 	destroy_workqueue(gfs_recovery_wq);
 fail_wq1:
-	unregister_filesystem(&gfs2meta_fs_type);
-fail_fs2:
-	unregister_filesystem(&gfs2_fs_type);
-fail_fs1:
 	unregister_shrinker(&gfs2_qd_shrinker);
 fail_shrinker:
 	kmem_cache_destroy(gfs2_trans_cachep);

fs/gfs2/ops_fstype.c

@@ -178,7 +178,10 @@ static int gfs2_check_sb(struct gfs2_sbd *sdp, int silent)
 		pr_warn("Invalid block size\n");
 		return -EINVAL;
 	}
-
+	if (sb->sb_bsize_shift != ffs(sb->sb_bsize) - 1) {
+		pr_warn("Invalid block size shift\n");
+		return -EINVAL;
+	}
 	return 0;
 }
 
@@ -381,8 +384,10 @@ static int init_names(struct gfs2_sbd *sdp, int silent)
 	if (!table[0])
 		table = sdp->sd_vfs->s_id;
 
-	strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
-	strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
+	BUILD_BUG_ON(GFS2_LOCKNAME_LEN > GFS2_FSNAME_LEN);
+
+	strscpy(sdp->sd_proto_name, proto, GFS2_LOCKNAME_LEN);
+	strscpy(sdp->sd_table_name, table, GFS2_LOCKNAME_LEN);
 
 	table = sdp->sd_table_name;
 	while ((table = strchr(table, '/')))
@@ -1439,13 +1444,13 @@ static int gfs2_parse_param(struct fs_context *fc, struct fs_parameter *param)
 
 	switch (o) {
 	case Opt_lockproto:
-		strlcpy(args->ar_lockproto, param->string, GFS2_LOCKNAME_LEN);
+		strscpy(args->ar_lockproto, param->string, GFS2_LOCKNAME_LEN);
 		break;
 	case Opt_locktable:
-		strlcpy(args->ar_locktable, param->string, GFS2_LOCKNAME_LEN);
+		strscpy(args->ar_locktable, param->string, GFS2_LOCKNAME_LEN);
 		break;
 	case Opt_hostdata:
-		strlcpy(args->ar_hostdata, param->string, GFS2_LOCKNAME_LEN);
+		strscpy(args->ar_hostdata, param->string, GFS2_LOCKNAME_LEN);
 		break;
 	case Opt_spectator:
 		args->ar_spectator = 1;

fs/gfs2/util.c

@@ -164,6 +164,11 @@ static void signal_our_withdraw(struct gfs2_sbd *sdp)
 		}
 		if (!ret)
 			gfs2_make_fs_ro(sdp);
+		/*
+		 * Dequeue any pending non-system glock holders that can no
+		 * longer be granted because the file system is withdrawn.
+		 */
+		gfs2_gl_dq_holders(sdp);
 		gfs2_freeze_unlock(&freeze_gh);
 	}
 
@@ -204,6 +209,7 @@ static void signal_our_withdraw(struct gfs2_sbd *sdp)
 	 * exception code in glock_dq.
 	 */
 	iput(inode);
+	sdp->sd_jdesc->jd_inode = NULL;
 	/*
 	 * Wait until the journal inode's glock is freed. This allows try locks
 	 * on other nodes to be successful, otherwise we remain the owner of