gfs2: fix GL_SKIP node_scope problems

Before this patch, when a glock was locked, the very first holder on the
queue would unlock the lockref and call the go_instantiate glops function
(if one existed), unless GL_SKIP was specified. When we introduced the new
node-scope concept, we allowed multiple holders to lock glocks in EX mode
and share the lock.

But node-scope introduced a new problem: if the first holder has GL_SKIP
and the next one does NOT, since it is not the first holder on the queue,
the go_instantiate op was not called. Eventually the GL_SKIP holder may
call the instantiate sub-function (e.g. gfs2_rgrp_bh_get) but there was
still a window of time in which another non-GL_SKIP holder assumes the
instantiate function had been called by the first holder. In the case of
rgrp glocks, this led to a NULL pointer dereference on the buffer_heads.

This patch tries to fix the problem by introducing two new glock flags:

GLF_INSTANTIATE_NEEDED, which keeps track of when the instantiate function
needs to be called to "fill in" or "read in" the object before it is
referenced.

GLF_INSTANTIATE_IN_PROG which is used to determine when a process is
in the process of reading in the object. Whenever a function needs to
reference the object, it checks the GLF_INSTANTIATE_NEEDED flag, and if
set, it sets GLF_INSTANTIATE_IN_PROG and calls the glops "go_instantiate"
function.

As before, the gl_lockref spin_lock is unlocked during the IO operation,
which may take a relatively long amount of time to complete. While
unlocked, if another process determines go_instantiate is still needed,
it sees GLF_INSTANTIATE_IN_PROG is set, and waits for the go_instantiate
glop operation to be completed. Once GLF_INSTANTIATE_IN_PROG is cleared,
it needs to check GLF_INSTANTIATE_NEEDED again because the other process's
go_instantiate operation may not have been successful.

Functions that previously called the instantiate sub-functions now call
directly into gfs2_instantiate so the new bits are managed properly.
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>

fs/gfs2/glock.c

@@ -479,12 +479,42 @@ find_first_strong_holder(struct gfs2_glock *gl)
  * Returns: 0 if instantiate was successful, 2 if type specific operation is
  * underway, or error.
  */
-static int gfs2_instantiate(struct gfs2_holder *gh)
+int gfs2_instantiate(struct gfs2_holder *gh)
 {
 	struct gfs2_glock *gl = gh->gh_gl;
 	const struct gfs2_glock_operations *glops = gl->gl_ops;
+	int ret;
+
+again:
+	if (!test_bit(GLF_INSTANTIATE_NEEDED, &gl->gl_flags))
+		return 0;
 
-	return glops->go_instantiate(gh);
+	/*
+	 * Since we unlock the lockref lock, we set a flag to indicate
+	 * instantiate is in progress.
+	 */
+	if (test_bit(GLF_INSTANTIATE_IN_PROG, &gl->gl_flags)) {
+		wait_on_bit(&gl->gl_flags, GLF_INSTANTIATE_IN_PROG,
+			    TASK_UNINTERRUPTIBLE);
+		/*
+		 * Here we just waited for a different instantiate to finish.
+		 * But that may not have been successful, as when a process
+		 * locks an inode glock _before_ it has an actual inode to
+		 * instantiate into. So we check again. This process might
+		 * have an inode to instantiate, so might be successful.
+		 */
+		goto again;
+	}
+
+	set_bit(GLF_INSTANTIATE_IN_PROG, &gl->gl_flags);
+
+	ret = glops->go_instantiate(gh);
+	if (!ret)
+		clear_bit(GLF_INSTANTIATE_NEEDED, &gl->gl_flags);
+	clear_bit(GLF_INSTANTIATE_IN_PROG, &gl->gl_flags);
+	smp_mb__after_atomic();
+	wake_up_bit(&gl->gl_flags, GLF_INSTANTIATE_IN_PROG);
+	return ret;
 }
 
 /**
@@ -526,7 +556,7 @@ __acquires(&gl->gl_lockref.lock)
 			incompat_holders_demoted = true;
 			first_gh = gh;
 		}
-		if (gh->gh_list.prev == &gl->gl_holders &&
+		if (test_bit(GLF_INSTANTIATE_NEEDED, &gl->gl_flags) &&
 		    !(gh->gh_flags & GL_SKIP) && gl->gl_ops->go_instantiate) {
 			lock_released = true;
 			spin_unlock(&gl->gl_lockref.lock);
@@ -1162,7 +1192,7 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 
 	atomic_inc(&sdp->sd_glock_disposal);
 	gl->gl_node.next = NULL;
-	gl->gl_flags = 0;
+	gl->gl_flags = glops->go_instantiate ? BIT(GLF_INSTANTIATE_NEEDED) : 0;
 	gl->gl_name = name;
 	lockdep_set_subclass(&gl->gl_lockref.lock, glops->go_subclass);
 	gl->gl_lockref.count = 1;
@@ -2326,6 +2356,10 @@ static const char *gflags2str(char *buf, const struct gfs2_glock *gl)
 		*p++ = 'P';
 	if (test_bit(GLF_FREEING, gflags))
 		*p++ = 'x';
+	if (test_bit(GLF_INSTANTIATE_NEEDED, gflags))
+		*p++ = 'n';
+	if (test_bit(GLF_INSTANTIATE_IN_PROG, gflags))
+		*p++ = 'N';
 	*p = 0;
 	return buf;
 }

fs/gfs2/glock.h

@@ -204,6 +204,7 @@ extern void gfs2_holder_reinit(unsigned int state, u16 flags,
 extern void gfs2_holder_uninit(struct gfs2_holder *gh);
 extern int gfs2_glock_nq(struct gfs2_holder *gh);
 extern int gfs2_glock_poll(struct gfs2_holder *gh);
+extern int gfs2_instantiate(struct gfs2_holder *gh);
 extern int gfs2_glock_wait(struct gfs2_holder *gh);
 extern int gfs2_glock_async_wait(unsigned int num_gh, struct gfs2_holder *ghs);
 extern void gfs2_glock_dq(struct gfs2_holder *gh);

fs/gfs2/glops.c

@@ -357,6 +357,7 @@ static void inode_go_inval(struct gfs2_glock *gl, int flags)
 		truncate_inode_pages(mapping, 0);
 		if (ip) {
 			set_bit(GIF_INVALID, &ip->i_flags);
+			set_bit(GLF_INSTANTIATE_NEEDED, &gl->gl_flags);
 			forget_all_cached_acls(&ip->i_inode);
 			security_inode_invalidate_secctx(&ip->i_inode);
 			gfs2_dir_hash_inval(ip);
@@ -495,13 +496,13 @@ static int inode_go_instantiate(struct gfs2_holder *gh)
 	struct gfs2_inode *ip = gl->gl_object;
 	int error = 0;
 
-	if (!ip)
-		return 0;
+	if (!ip) /* no inode to populate - read it in later */
+		goto out;
 
 	if (test_bit(GIF_INVALID, &ip->i_flags)) {
 		error = gfs2_inode_refresh(ip);
 		if (error)
-			return error;
+			goto out;
 	}
 
 	if (gh->gh_state != LM_ST_DEFERRED)
@@ -515,9 +516,10 @@ static int inode_go_instantiate(struct gfs2_holder *gh)
 			list_add(&ip->i_trunc_list, &sdp->sd_trunc_list);
 		spin_unlock(&sdp->sd_trunc_lock);
 		wake_up(&sdp->sd_quota_wait);
-		return 1;
+		error = 1;
 	}
 
+out:
 	return error;
 }
 

fs/gfs2/incore.h

@@ -316,6 +316,7 @@ struct gfs2_alloc_parms {
 
 enum {
 	GLF_LOCK			= 1,
+	GLF_INSTANTIATE_NEEDED		= 2, /* needs instantiate */
 	GLF_DEMOTE			= 3,
 	GLF_PENDING_DEMOTE		= 4,
 	GLF_DEMOTE_IN_PROGRESS		= 5,
@@ -325,6 +326,7 @@ enum {
 	GLF_REPLY_PENDING		= 9,
 	GLF_INITIAL			= 10,
 	GLF_FROZEN			= 11,
+	GLF_INSTANTIATE_IN_PROG		= 12, /* instantiate happening now */
 	GLF_LRU				= 13,
 	GLF_OBJECT			= 14, /* Used only for tracing */
 	GLF_BLOCKING			= 15,

fs/gfs2/inode.c

@@ -183,6 +183,7 @@ struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
 
 		glock_set_object(ip->i_gl, ip);
 		set_bit(GIF_INVALID, &ip->i_flags);
+		set_bit(GLF_INSTANTIATE_NEEDED, &ip->i_gl->gl_flags);
 		error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
 		if (unlikely(error))
 			goto fail;
@@ -196,7 +197,7 @@ struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
 
 		if (type == DT_UNKNOWN) {
 			/* Inode glock must be locked already */
-			error = gfs2_inode_refresh(GFS2_I(inode));
+			error = gfs2_instantiate(&i_gh);
 			if (error)
 				goto fail;
 		} else {

fs/gfs2/rgrp.c

@@ -1238,8 +1238,7 @@ static int gfs2_rgrp_bh_get(struct gfs2_rgrpd *rgd)
 		rgd->rd_rgl->rl_unlinked = cpu_to_be32(count_unlinked(rgd));
 		gfs2_rgrp_ondisk2lvb(rgd->rd_rgl,
 				     rgd->rd_bits[0].bi_bh->b_data);
-	}
-	else if (sdp->sd_args.ar_rgrplvb) {
+	} else if (sdp->sd_args.ar_rgrplvb) {
 		if (!gfs2_rgrp_lvb_valid(rgd)){
 			gfs2_consist_rgrpd(rgd);
 			error = -EIO;
@@ -1257,11 +1256,10 @@ static int gfs2_rgrp_bh_get(struct gfs2_rgrpd *rgd)
 		bi->bi_bh = NULL;
 		gfs2_assert_warn(sdp, !bi->bi_clone);
 	}
-
 	return error;
 }
 
-static int update_rgrp_lvb(struct gfs2_rgrpd *rgd)
+static int update_rgrp_lvb(struct gfs2_rgrpd *rgd, struct gfs2_holder *gh)
 {
 	u32 rl_flags;
 
@@ -1269,7 +1267,7 @@ static int update_rgrp_lvb(struct gfs2_rgrpd *rgd)
 		return 0;
 
 	if (cpu_to_be32(GFS2_MAGIC) != rgd->rd_rgl->rl_magic)
-		return gfs2_rgrp_bh_get(rgd);
+		return gfs2_instantiate(gh);
 
 	rl_flags = be32_to_cpu(rgd->rd_rgl->rl_flags);
 	rl_flags &= ~GFS2_RDF_MASK;
@@ -1312,6 +1310,7 @@ void gfs2_rgrp_brelse(struct gfs2_rgrpd *rgd)
 			bi->bi_bh = NULL;
 		}
 	}
+	set_bit(GLF_INSTANTIATE_NEEDED, &rgd->rd_gl->gl_flags);
 }
 
 int gfs2_rgrp_send_discards(struct gfs2_sbd *sdp, u64 offset,
@@ -2110,7 +2109,8 @@ int gfs2_inplace_reserve(struct gfs2_inode *ip, struct gfs2_alloc_parms *ap)
 			    gfs2_rgrp_congested(rs->rs_rgd, loops))
 				goto skip_rgrp;
 			if (sdp->sd_args.ar_rgrplvb) {
-				error = update_rgrp_lvb(rs->rs_rgd);
+				error = update_rgrp_lvb(rs->rs_rgd,
+							&ip->i_rgd_gh);
 				if (unlikely(error)) {
 					rgrp_unlock_local(rs->rs_rgd);
 					gfs2_glock_dq_uninit(&ip->i_rgd_gh);
@@ -2125,8 +2125,11 @@ int gfs2_inplace_reserve(struct gfs2_inode *ip, struct gfs2_alloc_parms *ap)
 		    (loops == 0 && target > rs->rs_rgd->rd_extfail_pt))
 			goto skip_rgrp;
 
-		if (sdp->sd_args.ar_rgrplvb)
-			gfs2_rgrp_bh_get(rs->rs_rgd);
+		if (sdp->sd_args.ar_rgrplvb) {
+			error = gfs2_instantiate(&ip->i_rgd_gh);
+			if (error)
+				goto skip_rgrp;
+		}
 
 		/* Get a reservation if we don't already have one */
 		if (!gfs2_rs_active(rs))
@@ -2762,8 +2765,6 @@ void gfs2_rlist_free(struct gfs2_rgrp_list *rlist)
 
 void rgrp_lock_local(struct gfs2_rgrpd *rgd)
 {
-	GLOCK_BUG_ON(rgd->rd_gl, !gfs2_glock_is_held_excl(rgd->rd_gl) &&
-		     !test_bit(SDF_NORECOVERY, &rgd->rd_sbd->sd_flags));
 	mutex_lock(&rgd->rd_mutex);
 }
 

fs/gfs2/super.c

@@ -1245,7 +1245,7 @@ static enum dinode_demise evict_should_delete(struct inode *inode,
 		return SHOULD_NOT_DELETE_DINODE;
 
 	if (test_bit(GIF_INVALID, &ip->i_flags)) {
-		ret = gfs2_inode_refresh(ip);
+		ret = gfs2_instantiate(gh);
 		if (ret)
 			return SHOULD_NOT_DELETE_DINODE;
 	}