[Cluster-devel] GFS2: Pre-pull patch posting (merge window)

[Cluster-devel] GFS2: Pre-pull patch posting (merge window)

[Steven Whitehouse]

Hi,

Not a lot new this time... the addition of a cache for the directory hash table
improve directory read/lookup speed, automatic adjustment of the glock hold
time improves performance for some contention corner cases. S_NOSEC support
is another performance related change, plus a nice clean up from Eric
Sandeen,

Steve.

[Cluster-devel] [PATCH 1/4] GFS2: Cache dir hash table in a contiguous buffer

[Steven Whitehouse]

This patch adds a cache for the hash table to the directory code
in order to help simplify the way in which the hash table is
accessed. This is intended to be a first step towards introducing
some performance improvements in the directory code.

There are two follow ups that I'm hoping to see fairly shortly. One
is to simplify the hash table reading code now that we always read the
complete hash table, whether we want one entry or all of them. The
other is to introduce readahead on the heads of the hash chains
which are referred to from the table.

The hash table is a maximum of 128k in size, so it is not worth trying
to read it in small chunks.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 091ee47..1cc2f8e 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -339,6 +339,67 @@ fail:
 	return (copied) ? copied : error;
 }
 
+/**
+ * gfs2_dir_get_hash_table - Get pointer to the dir hash table
+ * @ip: The inode in question
+ *
+ * Returns: The hash table or an error
+ */
+
+static __be64 *gfs2_dir_get_hash_table(struct gfs2_inode *ip)
+{
+	struct inode *inode = &ip->i_inode;
+	int ret;
+	u32 hsize;
+	__be64 *hc;
+
+	BUG_ON(!(ip->i_diskflags & GFS2_DIF_EXHASH));
+
+	hc = ip->i_hash_cache;
+	if (hc)
+		return hc;
+
+	hsize = 1 << ip->i_depth;
+	hsize *= sizeof(__be64);
+	if (hsize != i_size_read(&ip->i_inode)) {
+		gfs2_consist_inode(ip);
+		return ERR_PTR(-EIO);
+	}
+
+	hc = kmalloc(hsize, GFP_NOFS);
+	ret = -ENOMEM;
+	if (hc == NULL)
+		return ERR_PTR(-ENOMEM);
+
+	ret = gfs2_dir_read_data(ip, (char *)hc, 0, hsize, 1);
+	if (ret < 0) {
+		kfree(hc);
+		return ERR_PTR(ret);
+	}
+
+	spin_lock(&inode->i_lock);
+	if (ip->i_hash_cache)
+		kfree(hc);
+	else
+		ip->i_hash_cache = hc;
+	spin_unlock(&inode->i_lock);
+
+	return ip->i_hash_cache;
+}
+
+/**
+ * gfs2_dir_hash_inval - Invalidate dir hash
+ * @ip: The directory inode
+ *
+ * Must be called with an exclusive glock, or during glock invalidation.
+ */
+void gfs2_dir_hash_inval(struct gfs2_inode *ip)
+{
+	__be64 *hc = ip->i_hash_cache;
+	ip->i_hash_cache = NULL;
+	kfree(hc);
+}
+
 static inline int gfs2_dirent_sentinel(const struct gfs2_dirent *dent)
 {
 	return dent->de_inum.no_addr == 0 || dent->de_inum.no_formal_ino == 0;
@@ -686,17 +747,12 @@ static int get_leaf(struct gfs2_inode *dip, u64 leaf_no,
 static int get_leaf_nr(struct gfs2_inode *dip, u32 index,
 		       u64 *leaf_out)
 {
-	__be64 leaf_no;
-	int error;
-
-	error = gfs2_dir_read_data(dip, (char *)&leaf_no,
-				    index * sizeof(__be64),
-				    sizeof(__be64), 0);
-	if (error != sizeof(u64))
-		return (error < 0) ? error : -EIO;
-
-	*leaf_out = be64_to_cpu(leaf_no);
+	__be64 *hash;
 
+	hash = gfs2_dir_get_hash_table(dip);
+	if (IS_ERR(hash))
+		return PTR_ERR(hash);
+	*leaf_out = be64_to_cpu(*(hash + index));
 	return 0;
 }
 
@@ -966,6 +1022,8 @@ static int dir_split_leaf(struct inode *inode, const struct qstr *name)
 	for (x = 0; x < half_len; x++)
 		lp[x] = cpu_to_be64(bn);
 
+	gfs2_dir_hash_inval(dip);
+
 	error = gfs2_dir_write_data(dip, (char *)lp, start * sizeof(u64),
 				    half_len * sizeof(u64));
 	if (error != half_len * sizeof(u64)) {
@@ -1052,70 +1110,54 @@ fail_brelse:
 
 static int dir_double_exhash(struct gfs2_inode *dip)
 {
-	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
 	struct buffer_head *dibh;
 	u32 hsize;
-	u64 *buf;
-	u64 *from, *to;
-	u64 block;
-	u64 disksize = i_size_read(&dip->i_inode);
+	u32 hsize_bytes;
+	__be64 *hc;
+	__be64 *hc2, *h;
 	int x;
 	int error = 0;
 
 	hsize = 1 << dip->i_depth;
-	if (hsize * sizeof(u64) != disksize) {
-		gfs2_consist_inode(dip);
-		return -EIO;
-	}
+	hsize_bytes = hsize * sizeof(__be64);
 
-	/*  Allocate both the "from" and "to" buffers in one big chunk  */
+	hc = gfs2_dir_get_hash_table(dip);
+	if (IS_ERR(hc))
+		return PTR_ERR(hc);
 
-	buf = kcalloc(3, sdp->sd_hash_bsize, GFP_NOFS);
-	if (!buf)
+	h = hc2 = kmalloc(hsize_bytes * 2, GFP_NOFS);
+	if (!hc2)
 		return -ENOMEM;
 
-	for (block = disksize >> sdp->sd_hash_bsize_shift; block--;) {
-		error = gfs2_dir_read_data(dip, (char *)buf,
-					    block * sdp->sd_hash_bsize,
-					    sdp->sd_hash_bsize, 1);
-		if (error != sdp->sd_hash_bsize) {
-			if (error >= 0)
-				error = -EIO;
-			goto fail;
-		}
-
-		from = buf;
-		to = (u64 *)((char *)buf + sdp->sd_hash_bsize);
-
-		for (x = sdp->sd_hash_ptrs; x--; from++) {
-			*to++ = *from;	/*  No endianess worries  */
-			*to++ = *from;
-		}
+	error = gfs2_meta_inode_buffer(dip, &dibh);
+	if (error)
+		goto out_kfree;
 
-		error = gfs2_dir_write_data(dip,
-					     (char *)buf + sdp->sd_hash_bsize,
-					     block * sdp->sd_sb.sb_bsize,
-					     sdp->sd_sb.sb_bsize);
-		if (error != sdp->sd_sb.sb_bsize) {
-			if (error >= 0)
-				error = -EIO;
-			goto fail;
-		}
+	for (x = 0; x < hsize; x++) {
+		*h++ = *hc;
+		*h++ = *hc;
+		hc++;
 	}
 
-	kfree(buf);
-
-	error = gfs2_meta_inode_buffer(dip, &dibh);
-	if (!gfs2_assert_withdraw(sdp, !error)) {
-		dip->i_depth++;
-		gfs2_dinode_out(dip, dibh->b_data);
-		brelse(dibh);
-	}
+	error = gfs2_dir_write_data(dip, (char *)hc2, 0, hsize_bytes * 2);
+	if (error != (hsize_bytes * 2))
+		goto fail;
 
-	return error;
+	gfs2_dir_hash_inval(dip);
+	dip->i_hash_cache = hc2;
+	dip->i_depth++;
+	gfs2_dinode_out(dip, dibh->b_data);
+	brelse(dibh);
+	return 0;
 
 fail:
-	kfree(buf);
+	/* Replace original hash table & size */
+	gfs2_dir_write_data(dip, (char *)hc, 0, hsize_bytes);
+	i_size_write(&dip->i_inode, hsize_bytes);
+	gfs2_dinode_out(dip, dibh->b_data);
+	brelse(dibh);
+out_kfree:
+	kfree(hc2);
 	return error;
 }
 
@@ -1348,6 +1390,7 @@ out:
 	return error;
 }
 
+
 /**
  * dir_e_read - Reads the entries from a directory into a filldir buffer
  * @dip: dinode pointer
@@ -1362,9 +1405,7 @@ static int dir_e_read(struct inode *inode, u64 *offset, void *opaque,
 		      filldir_t filldir)
 {
 	struct gfs2_inode *dip = GFS2_I(inode);
-	struct gfs2_sbd *sdp = GFS2_SB(inode);
 	u32 hsize, len = 0;
-	u32 ht_offset, lp_offset, ht_offset_cur = -1;
 	u32 hash, index;
 	__be64 *lp;
 	int copied = 0;
@@ -1372,37 +1413,17 @@ static int dir_e_read(struct inode *inode, u64 *offset, void *opaque,
 	unsigned depth = 0;
 
 	hsize = 1 << dip->i_depth;
-	if (hsize * sizeof(u64) != i_size_read(inode)) {
-		gfs2_consist_inode(dip);
-		return -EIO;
-	}
-
 	hash = gfs2_dir_offset2hash(*offset);
 	index = hash >> (32 - dip->i_depth);
 
-	lp = kmalloc(sdp->sd_hash_bsize, GFP_NOFS);
-	if (!lp)
-		return -ENOMEM;
+	lp = gfs2_dir_get_hash_table(dip);
+	if (IS_ERR(lp))
+		return PTR_ERR(lp);
 
 	while (index < hsize) {
-		lp_offset = index & (sdp->sd_hash_ptrs - 1);
-		ht_offset = index - lp_offset;
-
-		if (ht_offset_cur != ht_offset) {
-			error = gfs2_dir_read_data(dip, (char *)lp,
-						ht_offset * sizeof(__be64),
-						sdp->sd_hash_bsize, 1);
-			if (error != sdp->sd_hash_bsize) {
-				if (error >= 0)
-					error = -EIO;
-				goto out;
-			}
-			ht_offset_cur = ht_offset;
-		}
-
 		error = gfs2_dir_read_leaf(inode, offset, opaque, filldir,
 					   &copied, &depth,
-					   be64_to_cpu(lp[lp_offset]));
+					   be64_to_cpu(lp[index]));
 		if (error)
 			break;
 
@@ -1410,8 +1431,6 @@ static int dir_e_read(struct inode *inode, u64 *offset, void *opaque,
 		index = (index & ~(len - 1)) + len;
 	}
 
-out:
-	kfree(lp);
 	if (error > 0)
 		error = 0;
 	return error;
@@ -1914,43 +1933,22 @@ out:
 
 int gfs2_dir_exhash_dealloc(struct gfs2_inode *dip)
 {
-	struct gfs2_sbd *sdp = GFS2_SB(&dip->i_inode);
 	struct buffer_head *bh;
 	struct gfs2_leaf *leaf;
 	u32 hsize, len;
-	u32 ht_offset, lp_offset, ht_offset_cur = -1;
 	u32 index = 0, next_index;
 	__be64 *lp;
 	u64 leaf_no;
 	int error = 0, last;
 
 	hsize = 1 << dip->i_depth;
-	if (hsize * sizeof(u64) != i_size_read(&dip->i_inode)) {
-		gfs2_consist_inode(dip);
-		return -EIO;
-	}
 
-	lp = kmalloc(sdp->sd_hash_bsize, GFP_NOFS);
-	if (!lp)
-		return -ENOMEM;
+	lp = gfs2_dir_get_hash_table(dip);
+	if (IS_ERR(lp))
+		return PTR_ERR(lp);
 
 	while (index < hsize) {
-		lp_offset = index & (sdp->sd_hash_ptrs - 1);
-		ht_offset = index - lp_offset;
-
-		if (ht_offset_cur != ht_offset) {
-			error = gfs2_dir_read_data(dip, (char *)lp,
-						ht_offset * sizeof(__be64),
-						sdp->sd_hash_bsize, 1);
-			if (error != sdp->sd_hash_bsize) {
-				if (error >= 0)
-					error = -EIO;
-				goto out;
-			}
-			ht_offset_cur = ht_offset;
-		}
-
-		leaf_no = be64_to_cpu(lp[lp_offset]);
+		leaf_no = be64_to_cpu(lp[index]);
 		if (leaf_no) {
 			error = get_leaf(dip, leaf_no, &bh);
 			if (error)
@@ -1976,7 +1974,6 @@ int gfs2_dir_exhash_dealloc(struct gfs2_inode *dip)
 	}
 
 out:
-	kfree(lp);
 
 	return error;
 }
diff --git a/fs/gfs2/dir.h b/fs/gfs2/dir.h
index e686af1..ff5772f 100644
--- a/fs/gfs2/dir.h
+++ b/fs/gfs2/dir.h
@@ -35,6 +35,7 @@ extern int gfs2_diradd_alloc_required(struct inode *dir,
 				      const struct qstr *filename);
 extern int gfs2_dir_get_new_buffer(struct gfs2_inode *ip, u64 block,
 				   struct buffer_head **bhp);
+extern void gfs2_dir_hash_inval(struct gfs2_inode *ip);
 
 static inline u32 gfs2_disk_hash(const char *data, int len)
 {
diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c
index 2cca293..95788ae 100644
--- a/fs/gfs2/glops.c
+++ b/fs/gfs2/glops.c
@@ -26,6 +26,7 @@
 #include "rgrp.h"
 #include "util.h"
 #include "trans.h"
+#include "dir.h"
 
 /**
  * __gfs2_ail_flush - remove all buffers for a given lock from the AIL
@@ -218,6 +219,7 @@ static void inode_go_inval(struct gfs2_glock *gl, int flags)
 		if (ip) {
 			set_bit(GIF_INVALID, &ip->i_flags);
 			forget_all_cached_acls(&ip->i_inode);
+			gfs2_dir_hash_inval(ip);
 		}
 	}
 
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 81206e7..24cd55f 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -285,6 +285,7 @@ struct gfs2_inode {
 	u64 i_goal;	/* goal block for allocations */
 	struct rw_semaphore i_rw_mutex;
 	struct list_head i_trunc_list;
+	__be64 *i_hash_cache;
 	u32 i_entries;
 	u32 i_diskflags;
 	u8 i_height;
diff --git a/fs/gfs2/main.c b/fs/gfs2/main.c
index c2b34cd..29e1ace 100644
--- a/fs/gfs2/main.c
+++ b/fs/gfs2/main.c
@@ -41,6 +41,7 @@ static void gfs2_init_inode_once(void *foo)
 	init_rwsem(&ip->i_rw_mutex);
 	INIT_LIST_HEAD(&ip->i_trunc_list);
 	ip->i_alloc = NULL;
+	ip->i_hash_cache = NULL;
 }
 
 static void gfs2_init_glock_once(void *foo)
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index fb0edf7..b7beadd 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -1533,7 +1533,7 @@ out:
 	/* Case 3 starts here */
 	truncate_inode_pages(&inode->i_data, 0);
 	end_writeback(inode);
-
+	gfs2_dir_hash_inval(ip);
 	ip->i_gl->gl_object = NULL;
 	gfs2_glock_add_to_lru(ip->i_gl);
 	gfs2_glock_put(ip->i_gl);
-- 
1.7.4

[Cluster-devel] [PATCH 2/4] GFS2: Automatically adjust glock min hold time

[Steven Whitehouse]

From: Bob Peterson <rpeterso@redhat.com>

This patch is a performance improvement for GFS2 in a clustered
environment. It makes the glock hold time self-adjusting.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 1c1336e..88e8a23 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -409,6 +409,10 @@ static void state_change(struct gfs2_glock *gl, unsigned int new_state)
 	if (held1 && held2 && list_empty(&gl->gl_holders))
 		clear_bit(GLF_QUEUED, &gl->gl_flags);
 
+	if (new_state != gl->gl_target)
+		/* shorten our minimum hold time */
+		gl->gl_hold_time = max(gl->gl_hold_time - GL_GLOCK_HOLD_DECR,
+				       GL_GLOCK_MIN_HOLD);
 	gl->gl_state = new_state;
 	gl->gl_tchange = jiffies;
 }
@@ -668,7 +672,7 @@ static void glock_work_func(struct work_struct *work)
 	    gl->gl_demote_state != LM_ST_EXCLUSIVE) {
 		unsigned long holdtime, now = jiffies;
 
-		holdtime = gl->gl_tchange + gl->gl_ops->go_min_hold_time;
+		holdtime = gl->gl_tchange + gl->gl_hold_time;
 		if (time_before(now, holdtime))
 			delay = holdtime - now;
 
@@ -679,9 +683,14 @@ static void glock_work_func(struct work_struct *work)
 	}
 	run_queue(gl, 0);
 	spin_unlock(&gl->gl_spin);
-	if (!delay ||
-	    queue_delayed_work(glock_workqueue, &gl->gl_work, delay) == 0)
+	if (!delay)
 		gfs2_glock_put(gl);
+	else {
+		if (gl->gl_name.ln_type != LM_TYPE_INODE)
+			delay = 0;
+		if (queue_delayed_work(glock_workqueue, &gl->gl_work, delay) == 0)
+			gfs2_glock_put(gl);
+	}
 	if (drop_ref)
 		gfs2_glock_put(gl);
 }
@@ -743,6 +752,7 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 	gl->gl_tchange = jiffies;
 	gl->gl_object = NULL;
 	gl->gl_sbd = sdp;
+	gl->gl_hold_time = GL_GLOCK_DFT_HOLD;
 	INIT_DELAYED_WORK(&gl->gl_work, glock_work_func);
 	INIT_WORK(&gl->gl_delete, delete_work_func);
 
@@ -855,8 +865,15 @@ static int gfs2_glock_demote_wait(void *word)
 
 static void wait_on_holder(struct gfs2_holder *gh)
 {
+	unsigned long time1 = jiffies;
+
 	might_sleep();
 	wait_on_bit(&gh->gh_iflags, HIF_WAIT, gfs2_glock_holder_wait, TASK_UNINTERRUPTIBLE);
+	if (time_after(jiffies, time1 + HZ)) /* have we waited > a second? */
+		/* Lengthen the minimum hold time. */
+		gh->gh_gl->gl_hold_time = min(gh->gh_gl->gl_hold_time +
+					      GL_GLOCK_HOLD_INCR,
+					      GL_GLOCK_MAX_HOLD);
 }
 
 static void wait_on_demote(struct gfs2_glock *gl)
@@ -1093,8 +1110,9 @@ void gfs2_glock_dq(struct gfs2_holder *gh)
 
 	gfs2_glock_hold(gl);
 	if (test_bit(GLF_PENDING_DEMOTE, &gl->gl_flags) &&
-	    !test_bit(GLF_DEMOTE, &gl->gl_flags))
-		delay = gl->gl_ops->go_min_hold_time;
+	    !test_bit(GLF_DEMOTE, &gl->gl_flags) &&
+	    gl->gl_name.ln_type == LM_TYPE_INODE)
+		delay = gl->gl_hold_time;
 	if (queue_delayed_work(glock_workqueue, &gl->gl_work, delay) == 0)
 		gfs2_glock_put(gl);
 }
@@ -1273,12 +1291,13 @@ void gfs2_glock_cb(struct gfs2_glock *gl, unsigned int state)
 	unsigned long now = jiffies;
 
 	gfs2_glock_hold(gl);
-	holdtime = gl->gl_tchange + gl->gl_ops->go_min_hold_time;
-	if (test_bit(GLF_QUEUED, &gl->gl_flags)) {
+	holdtime = gl->gl_tchange + gl->gl_hold_time;
+	if (test_bit(GLF_QUEUED, &gl->gl_flags) &&
+	    gl->gl_name.ln_type == LM_TYPE_INODE) {
 		if (time_before(now, holdtime))
 			delay = holdtime - now;
 		if (test_bit(GLF_REPLY_PENDING, &gl->gl_flags))
-			delay = gl->gl_ops->go_min_hold_time;
+			delay = gl->gl_hold_time;
 	}
 
 	spin_lock(&gl->gl_spin);
@@ -1667,7 +1686,7 @@ static int __dump_glock(struct seq_file *seq, const struct gfs2_glock *gl)
 	dtime *= 1000000/HZ; /* demote time in uSec */
 	if (!test_bit(GLF_DEMOTE, &gl->gl_flags))
 		dtime = 0;
-	gfs2_print_dbg(seq, "G:  s:%s n:%u/%llx f:%s t:%s d:%s/%llu a:%d v:%d r:%d\n",
+	gfs2_print_dbg(seq, "G:  s:%s n:%u/%llx f:%s t:%s d:%s/%llu a:%d v:%d r:%d m:%ld\n",
 		  state2str(gl->gl_state),
 		  gl->gl_name.ln_type,
 		  (unsigned long long)gl->gl_name.ln_number,
@@ -1676,7 +1695,7 @@ static int __dump_glock(struct seq_file *seq, const struct gfs2_glock *gl)
 		  state2str(gl->gl_demote_state), dtime,
 		  atomic_read(&gl->gl_ail_count),
 		  atomic_read(&gl->gl_revokes),
-		  atomic_read(&gl->gl_ref));
+		  atomic_read(&gl->gl_ref), gl->gl_hold_time);
 
 	list_for_each_entry(gh, &gl->gl_holders, gh_list) {
 		error = dump_holder(seq, gh);
diff --git a/fs/gfs2/glock.h b/fs/gfs2/glock.h
index 6b2f757..6670711 100644
--- a/fs/gfs2/glock.h
+++ b/fs/gfs2/glock.h
@@ -113,6 +113,12 @@ enum {
 
 #define GLR_TRYFAILED		13
 
+#define GL_GLOCK_MAX_HOLD        (long)(HZ / 5)
+#define GL_GLOCK_DFT_HOLD        (long)(HZ / 5)
+#define GL_GLOCK_MIN_HOLD        (long)(10)
+#define GL_GLOCK_HOLD_INCR       (long)(HZ / 20)
+#define GL_GLOCK_HOLD_DECR       (long)(HZ / 40)
+
 struct lm_lockops {
 	const char *lm_proto_name;
 	int (*lm_mount) (struct gfs2_sbd *sdp, const char *fsname);
diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c
index 95788ae..a4222c3 100644
--- a/fs/gfs2/glops.c
+++ b/fs/gfs2/glops.c
@@ -551,7 +551,6 @@ const struct gfs2_glock_operations gfs2_inode_glops = {
 	.go_lock = inode_go_lock,
 	.go_dump = inode_go_dump,
 	.go_type = LM_TYPE_INODE,
-	.go_min_hold_time = HZ / 5,
 	.go_flags = GLOF_ASPACE,
 };
 
@@ -562,7 +561,6 @@ const struct gfs2_glock_operations gfs2_rgrp_glops = {
 	.go_unlock = rgrp_go_unlock,
 	.go_dump = gfs2_rgrp_dump,
 	.go_type = LM_TYPE_RGRP,
-	.go_min_hold_time = HZ / 5,
 	.go_flags = GLOF_ASPACE,
 };
 
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 24cd55f..892ac37 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -163,7 +163,6 @@ struct gfs2_glock_operations {
 	int (*go_dump)(struct seq_file *seq, const struct gfs2_glock *gl);
 	void (*go_callback) (struct gfs2_glock *gl);
 	const int go_type;
-	const unsigned long go_min_hold_time;
 	const unsigned long go_flags;
 #define GLOF_ASPACE 1
 };
@@ -221,6 +220,7 @@ struct gfs2_glock {
 
 	unsigned int gl_hash;
 	unsigned long gl_demote_time; /* time of first demote request */
+	long gl_hold_time;
 	struct list_head gl_holders;
 
 	const struct gfs2_glock_operations *gl_ops;
-- 
1.7.4

[Cluster-devel] [PATCH 3/4] GFS2: Add S_NOSEC support

[Steven Whitehouse]

This adds S_NOSEC support to GFS2. We set/reset the flag either when
a user calls setattr or when we have just regained the glock
from another node. The flag is only set if there are no xattrs
on the inode and there is no suid bit set.

Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Cc: Al Viro <viro@ZenIV.linux.org.uk>

diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index a9f5cbe..bc2590e 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -174,7 +174,9 @@ void gfs2_set_inode_flags(struct inode *inode)
 	struct gfs2_inode *ip = GFS2_I(inode);
 	unsigned int flags = inode->i_flags;
 
-	flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
+	flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|S_NOSEC);
+	if ((ip->i_eattr == 0) && !is_sxid(inode->i_mode))
+		inode->i_flags |= S_NOSEC;
 	if (ip->i_diskflags & GFS2_DIF_IMMUTABLE)
 		flags |= S_IMMUTABLE;
 	if (ip->i_diskflags & GFS2_DIF_APPENDONLY)
diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c
index a4222c3..da21eca 100644
--- a/fs/gfs2/glops.c
+++ b/fs/gfs2/glops.c
@@ -318,6 +318,8 @@ static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
 	ip->i_generation = be64_to_cpu(str->di_generation);
 
 	ip->i_diskflags = be32_to_cpu(str->di_flags);
+	ip->i_eattr = be64_to_cpu(str->di_eattr);
+	/* i_diskflags and i_eattr must be set before gfs2_set_inode_flags() */
 	gfs2_set_inode_flags(&ip->i_inode);
 	height = be16_to_cpu(str->di_height);
 	if (unlikely(height > GFS2_MAX_META_HEIGHT))
@@ -330,7 +332,6 @@ static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
 	ip->i_depth = (u8)depth;
 	ip->i_entries = be32_to_cpu(str->di_entries);
 
-	ip->i_eattr = be64_to_cpu(str->di_eattr);
 	if (S_ISREG(ip->i_inode.i_mode))
 		gfs2_set_aops(&ip->i_inode);
 
diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c
index 2a77071..516516e 100644
--- a/fs/gfs2/ops_fstype.c
+++ b/fs/gfs2/ops_fstype.c
@@ -1094,6 +1094,7 @@ static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent
 	if (sdp->sd_args.ar_nobarrier)
 		set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
 
+	sb->s_flags |= MS_NOSEC;
 	sb->s_magic = GFS2_MAGIC;
 	sb->s_op = &gfs2_super_ops;
 	sb->s_d_op = &gfs2_dops;
-- 
1.7.4

[Cluster-devel] [PATCH 4/4] GFS2: combine duplicated block freeing routines

[Steven Whitehouse]

From: Eric Sandeen <sandeen@redhat.com>

__gfs2_free_data and __gfs2_free_meta are almost identical, and
can be trivially combined.

[This is as per Eric's original patch minus gfs2_free_data() which had
 no callers left and plus the conversion of the bmap.c calls to these
 functions. All in all, a nice clean up]

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index e65493a..42e477f 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -854,11 +854,7 @@ static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh,
 			blen++;
 		else {
 			if (bstart) {
-				if (metadata)
-					__gfs2_free_meta(ip, bstart, blen);
-				else
-					__gfs2_free_data(ip, bstart, blen);
-
+				__gfs2_free_blocks(ip, bstart, blen, metadata);
 				btotal += blen;
 			}
 
@@ -870,11 +866,7 @@ static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh,
 		gfs2_add_inode_blocks(&ip->i_inode, -1);
 	}
 	if (bstart) {
-		if (metadata)
-			__gfs2_free_meta(ip, bstart, blen);
-		else
-			__gfs2_free_data(ip, bstart, blen);
-
+		__gfs2_free_blocks(ip, bstart, blen, metadata);
 		btotal += blen;
 	}
 
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 9b780df..7f8af1e 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -1607,14 +1607,15 @@ rgrp_error:
 }
 
 /**
- * gfs2_free_data - free a contiguous run of data block(s)
+ * __gfs2_free_blocks - free a contiguous run of block(s)
  * @ip: the inode these blocks are being freed from
  * @bstart: first block of a run of contiguous blocks
  * @blen: the length of the block run
+ * @meta: 1 if the blocks represent metadata
  *
  */
 
-void __gfs2_free_data(struct gfs2_inode *ip, u64 bstart, u32 blen)
+void __gfs2_free_blocks(struct gfs2_inode *ip, u64 bstart, u32 blen, int meta)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct gfs2_rgrpd *rgd;
@@ -1631,54 +1632,11 @@ void __gfs2_free_data(struct gfs2_inode *ip, u64 bstart, u32 blen)
 	gfs2_trans_add_rg(rgd);
 
 	/* Directories keep their data in the metadata address space */
-	if (ip->i_depth)
+	if (meta || ip->i_depth)
 		gfs2_meta_wipe(ip, bstart, blen);
 }
 
 /**
- * gfs2_free_data - free a contiguous run of data block(s)
- * @ip: the inode these blocks are being freed from
- * @bstart: first block of a run of contiguous blocks
- * @blen: the length of the block run
- *
- */
-
-void gfs2_free_data(struct gfs2_inode *ip, u64 bstart, u32 blen)
-{
-	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
-
-	__gfs2_free_data(ip, bstart, blen);
-	gfs2_statfs_change(sdp, 0, +blen, 0);
-	gfs2_quota_change(ip, -(s64)blen, ip->i_inode.i_uid, ip->i_inode.i_gid);
-}
-
-/**
- * gfs2_free_meta - free a contiguous run of data block(s)
- * @ip: the inode these blocks are being freed from
- * @bstart: first block of a run of contiguous blocks
- * @blen: the length of the block run
- *
- */
-
-void __gfs2_free_meta(struct gfs2_inode *ip, u64 bstart, u32 blen)
-{
-	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
-	struct gfs2_rgrpd *rgd;
-
-	rgd = rgblk_free(sdp, bstart, blen, GFS2_BLKST_FREE);
-	if (!rgd)
-		return;
-	trace_gfs2_block_alloc(ip, bstart, blen, GFS2_BLKST_FREE);
-	rgd->rd_free += blen;
-
-	gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1);
-	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
-
-	gfs2_trans_add_rg(rgd);
-	gfs2_meta_wipe(ip, bstart, blen);
-}
-
-/**
  * gfs2_free_meta - free a contiguous run of data block(s)
  * @ip: the inode these blocks are being freed from
  * @bstart: first block of a run of contiguous blocks
@@ -1690,7 +1648,7 @@ void gfs2_free_meta(struct gfs2_inode *ip, u64 bstart, u32 blen)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 
-	__gfs2_free_meta(ip, bstart, blen);
+	__gfs2_free_blocks(ip, bstart, blen, 1);
 	gfs2_statfs_change(sdp, 0, +blen, 0);
 	gfs2_quota_change(ip, -(s64)blen, ip->i_inode.i_uid, ip->i_inode.i_gid);
 }
diff --git a/fs/gfs2/rgrp.h b/fs/gfs2/rgrp.h
index a80e303..d253f9a 100644
--- a/fs/gfs2/rgrp.h
+++ b/fs/gfs2/rgrp.h
@@ -52,9 +52,7 @@ extern int gfs2_ri_update(struct gfs2_inode *ip);
 extern int gfs2_alloc_block(struct gfs2_inode *ip, u64 *bn, unsigned int *n);
 extern int gfs2_alloc_di(struct gfs2_inode *ip, u64 *bn, u64 *generation);
 
-extern void __gfs2_free_data(struct gfs2_inode *ip, u64 bstart, u32 blen);
-extern void gfs2_free_data(struct gfs2_inode *ip, u64 bstart, u32 blen);
-extern void __gfs2_free_meta(struct gfs2_inode *ip, u64 bstart, u32 blen);
+extern void __gfs2_free_blocks(struct gfs2_inode *ip, u64 bstart, u32 blen, int meta);
 extern void gfs2_free_meta(struct gfs2_inode *ip, u64 bstart, u32 blen);
 extern void gfs2_free_di(struct gfs2_rgrpd *rgd, struct gfs2_inode *ip);
 extern void gfs2_unlink_di(struct inode *inode);
-- 
1.7.4