1 files changed, 102 insertions, 128 deletions

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 269c3bc7fced..ba25b884169e 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -34,8 +34,8 @@
 #include <linux/lockref.h>
 #include <linux/rhashtable.h>
 #include <linux/pid_namespace.h>
-#include <linux/fdtable.h>
 #include <linux/file.h>
+#include <linux/random.h>
 
 #include "gfs2.h"
 #include "incore.h"
@@ -563,11 +563,11 @@ static void state_change(struct gfs2_glock *gl, unsigned int new_state)
 	gl->gl_tchange = jiffies;
 }
 
-static void gfs2_set_demote(struct gfs2_glock *gl)
+static void gfs2_set_demote(int nr, struct gfs2_glock *gl)
 {
 	struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
 
-	set_bit(GLF_DEMOTE, &gl->gl_flags);
+	set_bit(nr, &gl->gl_flags);
 	smp_mb();
 	wake_up(&sdp->sd_async_glock_wait);
 }
@@ -607,14 +607,19 @@ static void finish_xmote(struct gfs2_glock *gl, unsigned int ret)
 		if (gh && (ret & LM_OUT_CANCELED))
 			gfs2_holder_wake(gh);
 		if (gh && !test_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags)) {
-			/* move to back of queue and try next entry */
 			if (ret & LM_OUT_CANCELED) {
-				list_move_tail(&gh->gh_list, &gl->gl_holders);
+				list_del_init(&gh->gh_list);
+				trace_gfs2_glock_queue(gh, 0);
+				gl->gl_target = gl->gl_state;
 				gh = find_first_waiter(gl);
-				gl->gl_target = gh->gh_state;
-				if (do_promote(gl))
-					goto out;
-				goto retry;
+				if (gh) {
+					gl->gl_target = gh->gh_state;
+					if (do_promote(gl))
+						goto out;
+					do_xmote(gl, gh, gl->gl_target);
+					return;
+				}
+				goto out;
 			}
 			/* Some error or failed "try lock" - report it */
 			if ((ret & LM_OUT_ERROR) ||
@@ -627,7 +632,6 @@ static void finish_xmote(struct gfs2_glock *gl, unsigned int ret)
 		switch(state) {
 		/* Unlocked due to conversion deadlock, try again */
 		case LM_ST_UNLOCKED:
-retry:
 			do_xmote(gl, gh, gl->gl_target);
 			break;
 		/* Conversion fails, unlock and try again */
@@ -661,7 +665,8 @@ retry:
 		do_promote(gl);
 	}
 out:
-	clear_bit(GLF_LOCK, &gl->gl_flags);
+	if (!test_bit(GLF_CANCELING, &gl->gl_flags))
+		clear_bit(GLF_LOCK, &gl->gl_flags);
 }
 
 static bool is_system_glock(struct gfs2_glock *gl)
@@ -807,6 +812,7 @@ skip_inval:
 	}
 
 	if (ls->ls_ops->lm_lock) {
+		set_bit(GLF_PENDING_REPLY, &gl->gl_flags);
 		spin_unlock(&gl->gl_lockref.lock);
 		ret = ls->ls_ops->lm_lock(gl, target, lck_flags);
 		spin_lock(&gl->gl_lockref.lock);
@@ -825,6 +831,7 @@ skip_inval:
 			/* The operation will be completed asynchronously. */
 			return;
 		}
+		clear_bit(GLF_PENDING_REPLY, &gl->gl_flags);
 	}
 
 	/* Complete the operation now. */
@@ -843,12 +850,13 @@ static void run_queue(struct gfs2_glock *gl, const int nonblock)
 __releases(&gl->gl_lockref.lock)
 __acquires(&gl->gl_lockref.lock)
 {
-	struct gfs2_holder *gh = NULL;
+	struct gfs2_holder *gh;
 
 	if (test_bit(GLF_LOCK, &gl->gl_flags))
 		return;
 	set_bit(GLF_LOCK, &gl->gl_flags);
 
+	/* While a demote is in progress, the GLF_LOCK flag must be set. */
 	GLOCK_BUG_ON(gl, test_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags));
 
 	if (test_bit(GLF_DEMOTE, &gl->gl_flags) &&
@@ -860,18 +868,22 @@ __acquires(&gl->gl_lockref.lock)
 		set_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags);
 		GLOCK_BUG_ON(gl, gl->gl_demote_state == LM_ST_EXCLUSIVE);
 		gl->gl_target = gl->gl_demote_state;
+		do_xmote(gl, NULL, gl->gl_target);
+		return;
 	} else {
 		if (test_bit(GLF_DEMOTE, &gl->gl_flags))
 			gfs2_demote_wake(gl);
 		if (do_promote(gl))
 			goto out_unlock;
 		gh = find_first_waiter(gl);
+		if (!gh)
+			goto out_unlock;
 		gl->gl_target = gh->gh_state;
 		if (!(gh->gh_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB)))
 			do_error(gl, 0); /* Fail queued try locks */
+		do_xmote(gl, gh, gl->gl_target);
+		return;
 	}
-	do_xmote(gl, gh, gl->gl_target);
-	return;
 
 out_sched:
 	clear_bit(GLF_LOCK, &gl->gl_flags);
@@ -898,12 +910,8 @@ void glock_set_object(struct gfs2_glock *gl, void *object)
 	prev_object = gl->gl_object;
 	gl->gl_object = object;
 	spin_unlock(&gl->gl_lockref.lock);
-	if (gfs2_assert_warn(gl->gl_name.ln_sbd, prev_object == NULL)) {
-		pr_warn("glock=%u/%llx\n",
-			gl->gl_name.ln_type,
-			(unsigned long long)gl->gl_name.ln_number);
+	if (gfs2_assert_warn(gl->gl_name.ln_sbd, prev_object == NULL))
 		gfs2_dump_glock(NULL, gl, true);
-	}
 }
 
 /**
@@ -919,12 +927,8 @@ void glock_clear_object(struct gfs2_glock *gl, void *object)
 	prev_object = gl->gl_object;
 	gl->gl_object = NULL;
 	spin_unlock(&gl->gl_lockref.lock);
-	if (gfs2_assert_warn(gl->gl_name.ln_sbd, prev_object == object)) {
-		pr_warn("glock=%u/%llx\n",
-			gl->gl_name.ln_type,
-			(unsigned long long)gl->gl_name.ln_number);
+	if (gfs2_assert_warn(gl->gl_name.ln_sbd, prev_object == object))
 		gfs2_dump_glock(NULL, gl, true);
-	}
 }
 
 void gfs2_inode_remember_delete(struct gfs2_glock *gl, u64 generation)
@@ -959,48 +963,56 @@ static void gfs2_glock_poke(struct gfs2_glock *gl)
 	gfs2_holder_uninit(&gh);
 }
 
-static bool gfs2_try_evict(struct gfs2_glock *gl)
+static struct gfs2_inode *gfs2_grab_existing_inode(struct gfs2_glock *gl)
+{
+	struct gfs2_inode *ip;
+
+	spin_lock(&gl->gl_lockref.lock);
+	ip = gl->gl_object;
+	if (ip && !igrab(&ip->i_inode))
+		ip = NULL;
+	spin_unlock(&gl->gl_lockref.lock);
+	if (ip) {
+		wait_on_inode(&ip->i_inode);
+		if (is_bad_inode(&ip->i_inode)) {
+			iput(&ip->i_inode);
+			ip = NULL;
+		}
+	}
+	return ip;
+}
+
+static void gfs2_try_evict(struct gfs2_glock *gl)
 {
 	struct gfs2_inode *ip;
-	bool evicted = false;
 
 	/*
 	 * If there is contention on the iopen glock and we have an inode, try
-	 * to grab and release the inode so that it can be evicted.  This will
-	 * allow the remote node to go ahead and delete the inode without us
-	 * having to do it, which will avoid rgrp glock thrashing.
+	 * to grab and release the inode so that it can be evicted.  The
+	 * GIF_DEFER_DELETE flag indicates to gfs2_evict_inode() that the inode
+	 * should not be deleted locally.  This will allow the remote node to
+	 * go ahead and delete the inode without us having to do it, which will
+	 * avoid rgrp glock thrashing.
 	 *
 	 * The remote node is likely still holding the corresponding inode
 	 * glock, so it will run before we get to verify that the delete has
-	 * happened below.
+	 * happened below.  (Verification is triggered by the call to
+	 * gfs2_queue_verify_delete() in gfs2_evict_inode().)
 	 */
-	spin_lock(&gl->gl_lockref.lock);
-	ip = gl->gl_object;
-	if (ip && !igrab(&ip->i_inode))
-		ip = NULL;
-	spin_unlock(&gl->gl_lockref.lock);
+	ip = gfs2_grab_existing_inode(gl);
 	if (ip) {
-		gl->gl_no_formal_ino = ip->i_no_formal_ino;
-		set_bit(GIF_DEFERRED_DELETE, &ip->i_flags);
+		set_bit(GLF_DEFER_DELETE, &gl->gl_flags);
 		d_prune_aliases(&ip->i_inode);
 		iput(&ip->i_inode);
+		clear_bit(GLF_DEFER_DELETE, &gl->gl_flags);
 
 		/* If the inode was evicted, gl->gl_object will now be NULL. */
-		spin_lock(&gl->gl_lockref.lock);
-		ip = gl->gl_object;
-		if (ip) {
-			clear_bit(GIF_DEFERRED_DELETE, &ip->i_flags);
-			if (!igrab(&ip->i_inode))
-				ip = NULL;
-		}
-		spin_unlock(&gl->gl_lockref.lock);
+		ip = gfs2_grab_existing_inode(gl);
 		if (ip) {
 			gfs2_glock_poke(ip->i_gl);
 			iput(&ip->i_inode);
 		}
-		evicted = !ip;
 	}
-	return evicted;
 }
 
 bool gfs2_queue_try_to_evict(struct gfs2_glock *gl)
@@ -1009,18 +1021,18 @@ bool gfs2_queue_try_to_evict(struct gfs2_glock *gl)
 
 	if (test_and_set_bit(GLF_TRY_TO_EVICT, &gl->gl_flags))
 		return false;
-	return queue_delayed_work(sdp->sd_delete_wq,
-				  &gl->gl_delete, 0);
+	return !mod_delayed_work(sdp->sd_delete_wq, &gl->gl_delete, 0);
 }
 
-static bool gfs2_queue_verify_evict(struct gfs2_glock *gl)
+bool gfs2_queue_verify_delete(struct gfs2_glock *gl, bool later)
 {
 	struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
+	unsigned long delay;
 
-	if (test_and_set_bit(GLF_VERIFY_EVICT, &gl->gl_flags))
+	if (test_and_set_bit(GLF_VERIFY_DELETE, &gl->gl_flags))
 		return false;
-	return queue_delayed_work(sdp->sd_delete_wq,
-				  &gl->gl_delete, 5 * HZ);
+	delay = later ? HZ + get_random_long() % (HZ * 9) : 0;
+	return queue_delayed_work(sdp->sd_delete_wq, &gl->gl_delete, delay);
 }
 
 static void delete_work_func(struct work_struct *work)
@@ -1028,43 +1040,21 @@ static void delete_work_func(struct work_struct *work)
 	struct delayed_work *dwork = to_delayed_work(work);
 	struct gfs2_glock *gl = container_of(dwork, struct gfs2_glock, gl_delete);
 	struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
-	struct inode *inode;
-	u64 no_addr = gl->gl_name.ln_number;
+	bool verify_delete = test_and_clear_bit(GLF_VERIFY_DELETE, &gl->gl_flags);
 
-	if (test_and_clear_bit(GLF_TRY_TO_EVICT, &gl->gl_flags)) {
-		/*
-		 * If we can evict the inode, give the remote node trying to
-		 * delete the inode some time before verifying that the delete
-		 * has happened.  Otherwise, if we cause contention on the inode glock
-		 * immediately, the remote node will think that we still have
-		 * the inode in use, and so it will give up waiting.
-		 *
-		 * If we can't evict the inode, signal to the remote node that
-		 * the inode is still in use.  We'll later try to delete the
-		 * inode locally in gfs2_evict_inode.
-		 *
-		 * FIXME: We only need to verify that the remote node has
-		 * deleted the inode because nodes before this remote delete
-		 * rework won't cooperate.  At a later time, when we no longer
-		 * care about compatibility with such nodes, we can skip this
-		 * step entirely.
-		 */
-		if (gfs2_try_evict(gl)) {
-			if (test_bit(SDF_KILL, &sdp->sd_flags))
-				goto out;
-			if (gfs2_queue_verify_evict(gl))
-				return;
-		}
-		goto out;
-	}
+	if (test_and_clear_bit(GLF_TRY_TO_EVICT, &gl->gl_flags))
+		gfs2_try_evict(gl);
+
+	if (verify_delete) {
+		u64 no_addr = gl->gl_name.ln_number;
+		struct inode *inode;
 
-	if (test_and_clear_bit(GLF_VERIFY_EVICT, &gl->gl_flags)) {
 		inode = gfs2_lookup_by_inum(sdp, no_addr, gl->gl_no_formal_ino,
 					    GFS2_BLKST_UNLINKED);
 		if (IS_ERR(inode)) {
 			if (PTR_ERR(inode) == -EAGAIN &&
 			    !test_bit(SDF_KILL, &sdp->sd_flags) &&
-			    gfs2_queue_verify_evict(gl))
+			    gfs2_queue_verify_delete(gl, true))
 				return;
 		} else {
 			d_prune_aliases(inode);
@@ -1072,7 +1062,6 @@ static void delete_work_func(struct work_struct *work)
 		}
 	}
 
-out:
 	gfs2_glock_put(gl);
 }
 
@@ -1101,7 +1090,7 @@ static void glock_work_func(struct work_struct *work)
 
 		if (!delay) {
 			clear_bit(GLF_PENDING_DEMOTE, &gl->gl_flags);
-			gfs2_set_demote(gl);
+			gfs2_set_demote(GLF_DEMOTE, gl);
 		}
 	}
 	run_queue(gl, 0);
@@ -1177,7 +1166,6 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 		   const struct gfs2_glock_operations *glops, int create,
 		   struct gfs2_glock **glp)
 {
-	struct super_block *s = sdp->sd_vfs;
 	struct lm_lockname name = { .ln_number = number,
 				    .ln_type = glops->go_type,
 				    .ln_sbd = sdp };
@@ -1218,8 +1206,8 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 	if (glops->go_instantiate)
 		gl->gl_flags |= BIT(GLF_INSTANTIATE_NEEDED);
 	gl->gl_name = name;
+	lockref_init(&gl->gl_lockref);
 	lockdep_set_subclass(&gl->gl_lockref.lock, glops->go_subclass);
-	gl->gl_lockref.count = 1;
 	gl->gl_state = LM_ST_UNLOCKED;
 	gl->gl_target = LM_ST_UNLOCKED;
 	gl->gl_demote_state = LM_ST_EXCLUSIVE;
@@ -1240,7 +1228,7 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 	mapping = gfs2_glock2aspace(gl);
 	if (mapping) {
                 mapping->a_ops = &gfs2_meta_aops;
-		mapping->host = s->s_bdev->bd_mapping->host;
+		mapping->host = sdp->sd_inode;
 		mapping->flags = 0;
 		mapping_set_gfp_mask(mapping, GFP_NOFS);
 		mapping->i_private_data = NULL;
@@ -1443,10 +1431,7 @@ out:
 static void request_demote(struct gfs2_glock *gl, unsigned int state,
 			   unsigned long delay, bool remote)
 {
-	if (delay)
-		set_bit(GLF_PENDING_DEMOTE, &gl->gl_flags);
-	else
-		gfs2_set_demote(gl);
+	gfs2_set_demote(delay ? GLF_PENDING_DEMOTE : GLF_DEMOTE, gl);
 	if (gl->gl_demote_state == LM_ST_EXCLUSIVE) {
 		gl->gl_demote_state = state;
 		gl->gl_demote_time = jiffies;
@@ -1482,9 +1467,7 @@ static inline bool pid_is_meaningful(const struct gfs2_holder *gh)
 {
         if (!(gh->gh_flags & GL_NOPID))
                 return true;
-        if (gh->gh_state == LM_ST_UNLOCKED)
-                return true;
-        return false;
+	return !test_bit(HIF_HOLDER, &gh->gh_iflags);
 }
 
 /**
@@ -1503,7 +1486,6 @@ __acquires(&gl->gl_lockref.lock)
 {
 	struct gfs2_glock *gl = gh->gh_gl;
 	struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
-	struct list_head *insert_pt = NULL;
 	struct gfs2_holder *gh2;
 	int try_futile = 0;
 
@@ -1539,21 +1521,11 @@ fail:
 			gfs2_holder_wake(gh);
 			return;
 		}
-		if (test_bit(HIF_HOLDER, &gh2->gh_iflags))
-			continue;
 	}
 	trace_gfs2_glock_queue(gh, 1);
 	gfs2_glstats_inc(gl, GFS2_LKS_QCOUNT);
 	gfs2_sbstats_inc(gl, GFS2_LKS_QCOUNT);
-	if (likely(insert_pt == NULL)) {
-		list_add_tail(&gh->gh_list, &gl->gl_holders);
-		return;
-	}
-	list_add_tail(&gh->gh_list, insert_pt);
-	spin_unlock(&gl->gl_lockref.lock);
-	if (sdp->sd_lockstruct.ls_ops->lm_cancel)
-		sdp->sd_lockstruct.ls_ops->lm_cancel(gl);
-	spin_lock(&gl->gl_lockref.lock);
+	list_add_tail(&gh->gh_list, &gl->gl_holders);
 	return;
 
 trap_recursive:
@@ -1636,12 +1608,6 @@ int gfs2_glock_poll(struct gfs2_holder *gh)
 	return test_bit(HIF_WAIT, &gh->gh_iflags) ? 0 : 1;
 }
 
-static inline bool needs_demote(struct gfs2_glock *gl)
-{
-	return (test_bit(GLF_DEMOTE, &gl->gl_flags) ||
-		test_bit(GLF_PENDING_DEMOTE, &gl->gl_flags));
-}
-
 static void __gfs2_glock_dq(struct gfs2_holder *gh)
 {
 	struct gfs2_glock *gl = gh->gh_gl;
@@ -1650,8 +1616,8 @@ static void __gfs2_glock_dq(struct gfs2_holder *gh)
 
 	/*
 	 * This holder should not be cached, so mark it for demote.
-	 * Note: this should be done before the check for needs_demote
-	 * below.
+	 * Note: this should be done before the glock_needs_demote
+	 * check below.
 	 */
 	if (gh->gh_flags & GL_NOCACHE)
 		request_demote(gl, LM_ST_UNLOCKED, 0, false);
@@ -1664,7 +1630,7 @@ static void __gfs2_glock_dq(struct gfs2_holder *gh)
 	 * If there hasn't been a demote request we are done.
 	 * (Let the remaining holders, if any, keep holding it.)
 	 */
-	if (!needs_demote(gl)) {
+	if (!glock_needs_demote(gl)) {
 		if (list_empty(&gl->gl_holders))
 			fast_path = 1;
 	}
@@ -1699,11 +1665,19 @@ void gfs2_glock_dq(struct gfs2_holder *gh)
 	}
 
 	if (list_is_first(&gh->gh_list, &gl->gl_holders) &&
-	    !test_bit(HIF_HOLDER, &gh->gh_iflags)) {
+	    !test_bit(HIF_HOLDER, &gh->gh_iflags) &&
+	    test_bit(GLF_LOCK, &gl->gl_flags) &&
+	    !test_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags) &&
+	    !test_bit(GLF_CANCELING, &gl->gl_flags)) {
+		set_bit(GLF_CANCELING, &gl->gl_flags);
 		spin_unlock(&gl->gl_lockref.lock);
 		gl->gl_name.ln_sbd->sd_lockstruct.ls_ops->lm_cancel(gl);
 		wait_on_bit(&gh->gh_iflags, HIF_WAIT, TASK_UNINTERRUPTIBLE);
 		spin_lock(&gl->gl_lockref.lock);
+		clear_bit(GLF_CANCELING, &gl->gl_flags);
+		clear_bit(GLF_LOCK, &gl->gl_flags);
+		if (!gfs2_holder_queued(gh))
+			goto out;
 	}
 
 	/*
@@ -1949,6 +1923,7 @@ void gfs2_glock_complete(struct gfs2_glock *gl, int ret)
 	struct lm_lockstruct *ls = &gl->gl_name.ln_sbd->sd_lockstruct;
 
 	spin_lock(&gl->gl_lockref.lock);
+	clear_bit(GLF_PENDING_REPLY, &gl->gl_flags);
 	gl->gl_reply = ret;
 
 	if (unlikely(test_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags))) {
@@ -2118,7 +2093,7 @@ static void glock_hash_walk(glock_examiner examiner, const struct gfs2_sbd *sdp)
 void gfs2_cancel_delete_work(struct gfs2_glock *gl)
 {
 	clear_bit(GLF_TRY_TO_EVICT, &gl->gl_flags);
-	clear_bit(GLF_VERIFY_EVICT, &gl->gl_flags);
+	clear_bit(GLF_VERIFY_DELETE, &gl->gl_flags);
 	if (cancel_delayed_work(&gl->gl_delete))
 		gfs2_glock_put(gl);
 }
@@ -2349,6 +2324,8 @@ static const char *gflags2str(char *buf, const struct gfs2_glock *gl)
 		*p++ = 'f';
 	if (test_bit(GLF_INVALIDATE_IN_PROGRESS, gflags))
 		*p++ = 'i';
+	if (test_bit(GLF_PENDING_REPLY, gflags))
+		*p++ = 'R';
 	if (test_bit(GLF_HAVE_REPLY, gflags))
 		*p++ = 'r';
 	if (test_bit(GLF_INITIAL, gflags))
@@ -2371,8 +2348,12 @@ static const char *gflags2str(char *buf, const struct gfs2_glock *gl)
 		*p++ = 'N';
 	if (test_bit(GLF_TRY_TO_EVICT, gflags))
 		*p++ = 'e';
-	if (test_bit(GLF_VERIFY_EVICT, gflags))
+	if (test_bit(GLF_VERIFY_DELETE, gflags))
 		*p++ = 'E';
+	if (test_bit(GLF_DEFER_DELETE, gflags))
+		*p++ = 's';
+	if (test_bit(GLF_CANCELING, gflags))
+		*p++ = 'C';
 	*p = 0;
 	return buf;
 }
@@ -2768,25 +2749,18 @@ static struct file *gfs2_glockfd_next_file(struct gfs2_glockfd_iter *i)
 		i->file = NULL;
 	}
 
-	rcu_read_lock();
 	for(;; i->fd++) {
-		struct inode *inode;
-
-		i->file = task_lookup_next_fdget_rcu(i->task, &i->fd);
+		i->file = fget_task_next(i->task, &i->fd);
 		if (!i->file) {
 			i->fd = 0;
 			break;
 		}
 
-		inode = file_inode(i->file);
-		if (inode->i_sb == i->sb)
+		if (file_inode(i->file)->i_sb == i->sb)
 			break;
 
-		rcu_read_unlock();
 		fput(i->file);
-		rcu_read_lock();
 	}
-	rcu_read_unlock();
 	return i->file;
 }