Skip to contentSkip to conversation

GFS2: Pre-pull patch posting (-nmw)

PATCH New Steven Whitehouse 1 participant 16 messages Dec 17, 2010 GFS2 file system

lore
Share & tools
lore
#0 lore raw

Hi,

This is a bit early this time, but I'm not expecting to add any more features to this patch set before the merge window. In fact it is mostly clean up and fixes anyway. So I'm pushing it out now to let everybody see it ahead of time.

The only user visible new feature this time is the addition of support for updating quota usage values through the quotactl interface. This addition means that all of GFS2's quota operations will be accessible via the generic quota-tools userland package very shortly. This is a precursor to being able to drop the userland gfs2_quota tool and use the generic quota tools for all GFS2 quota operations at some future date.

I'm planning to keep the glock scalability patch (previously posted on cluster-devel) until the following merge window, since that is a fairly large patch, and it should be ready to go into my -nmw tree as the first patch of the new -nmw tree with a bit of luck.

Other than that I don't think there is much else noteworthy this time,

Steve.

#1 lore raw

PATCH 1/15 GFS2: Change two WQ_RESCUERs into WQ_MEM_RECLAIM

The WQ_RESCUER flag should only be used internally to the workqueue implementation.

fs/gfs2/glock.c +4 -3
index f92c177..5a56568 100644
@@ -1782,12 +1782,13 @@ int __init gfs2_glock_init(void)
 	}
 #endif
 
-	glock_workqueue = alloc_workqueue("glock_workqueue", WQ_RESCUER |
+	glock_workqueue = alloc_workqueue("glock_workqueue", WQ_MEM_RECLAIM |
 					  WQ_HIGHPRI | WQ_FREEZEABLE, 0);
 	if (IS_ERR(glock_workqueue))
 		return PTR_ERR(glock_workqueue);
-	gfs2_delete_workqueue = alloc_workqueue("delete_workqueue", WQ_RESCUER |
-						WQ_FREEZEABLE, 0);
+	gfs2_delete_workqueue = alloc_workqueue("delete_workqueue",
+						WQ_MEM_RECLAIM | WQ_FREEZEABLE,
+						0);
 	if (IS_ERR(gfs2_delete_workqueue)) {
 		destroy_workqueue(glock_workqueue);
 		return PTR_ERR(gfs2_delete_workqueue);
Signature · 1 line
1.7.3.3
#2 lore raw

PATCH 2/15 GFS2: fs/gfs2/glock.c: Convert sprintf_symbol to %pS

In reply to Steven Whitehouse

From: Joe Perches <[email protected]>

fs/gfs2/glock.c +7 -8
index 5a56568..13155f6 100644
@@ -1626,18 +1626,17 @@ static const char *hflags2str(char *buf, unsigned flags, unsigned long iflags)
 static int dump_holder(struct seq_file *seq, const struct gfs2_holder *gh)
 {
 	struct task_struct *gh_owner = NULL;
-	char buffer[KSYM_SYMBOL_LEN];
 	char flags_buf[32];
 
-	sprint_symbol(buffer, gh->gh_ip);
 	if (gh->gh_owner_pid)
 		gh_owner = pid_task(gh->gh_owner_pid, PIDTYPE_PID);
-	gfs2_print_dbg(seq, " H: s:%s f:%s e:%d p:%ld [%s] %s\n",
-		  state2str(gh->gh_state),
-		  hflags2str(flags_buf, gh->gh_flags, gh->gh_iflags),
-		  gh->gh_error, 
-		  gh->gh_owner_pid ? (long)pid_nr(gh->gh_owner_pid) : -1,
-		  gh_owner ? gh_owner->comm : "(ended)", buffer);
+	gfs2_print_dbg(seq, " H: s:%s f:%s e:%d p:%ld [%s] %pS\n",
+		       state2str(gh->gh_state),
+		       hflags2str(flags_buf, gh->gh_flags, gh->gh_iflags),
+		       gh->gh_error,
+		       gh->gh_owner_pid ? (long)pid_nr(gh->gh_owner_pid) : -1,
+		       gh_owner ? gh_owner->comm : "(ended)",
+		       (void *)gh->gh_ip);
 	return 0;
 }
 
Signature · 1 line
1.7.3.3
#3 lore raw

PATCH 3/15 GFS2: Remove unreachable calls to vmtruncate

In reply to Steven Whitehouse
fs/gfs2/inode.c +0 -9
index e1213f7..14e682d 100644
@@ -916,17 +916,8 @@ static int __gfs2_setattr_simple(struct gfs2_inode *ip, struct iattr *attr)
 	if (error)
 		return error;
 
-	if ((attr->ia_valid & ATTR_SIZE) &&
-	    attr->ia_size != i_size_read(inode)) {
-		error = vmtruncate(inode, attr->ia_size);
-		if (error)
-			return error;
-	}
-
 	setattr_copy(inode, attr);
 	mark_inode_dirty(inode);
-
-	gfs2_assert_warn(GFS2_SB(inode), !error);
 	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
 	gfs2_dinode_out(ip, dibh->b_data);
 	brelse(dibh);
fs/gfs2/ops_inode.c +0 -9
index 12cbea7..457ecd1 100644
@@ -1104,17 +1104,8 @@ static int setattr_chown(struct inode *inode, struct iattr *attr)
 	if (error)
 		goto out_end_trans;
 
-	if ((attr->ia_valid & ATTR_SIZE) &&
-	    attr->ia_size != i_size_read(inode)) {
-		int error;
-
-		error = vmtruncate(inode, attr->ia_size);
-		gfs2_assert_warn(sdp, !error);
-	}
-
 	setattr_copy(inode, attr);
 	mark_inode_dirty(inode);
-
 	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
 	gfs2_dinode_out(ip, dibh->b_data);
 	brelse(dibh);
fs/gfs2/xattr.c +0 -9
index 30b58f0..ecfd40d 100644
@@ -1325,17 +1325,8 @@ int gfs2_xattr_acl_chmod(struct gfs2_inode *ip, struct iattr *attr, char *data)
 	if (error)
 		goto out_trans_end;
 
-	if ((attr->ia_valid & ATTR_SIZE) &&
-	    attr->ia_size != i_size_read(inode)) {
-		int error;
-
-		error = vmtruncate(inode, attr->ia_size);
-		gfs2_assert_warn(GFS2_SB(inode), !error);
-	}
-
 	setattr_copy(inode, attr);
 	mark_inode_dirty(inode);
-
 	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
 	gfs2_dinode_out(ip, dibh->b_data);
 	brelse(dibh);
Signature · 1 line
1.7.3.3
#4 lore raw

PATCH 4/15 GFS2: Clean up duplicated setattr code

In reply to Steven Whitehouse

While preparing the last patch I noticed that the gfs2_setattr_simple code had been duplicated into two other places. This patch updates those to call gfs2_setattr_simple rather than open coding it.

fs/gfs2/ops_inode.c +1 -8
index 457ecd1..1db6b73 100644
@@ -1069,7 +1069,6 @@ static int setattr_chown(struct inode *inode, struct iattr *attr)
 {
 	struct gfs2_inode *ip = GFS2_I(inode);
 	struct gfs2_sbd *sdp = GFS2_SB(inode);
-	struct buffer_head *dibh;
 	u32 ouid, ogid, nuid, ngid;
 	int error;
 
@@ -1100,16 +1099,10 @@ static int setattr_chown(struct inode *inode, struct iattr *attr)
 	if (error)
 		goto out_gunlock_q;
 
-	error = gfs2_meta_inode_buffer(ip, &dibh);
+	error = gfs2_setattr_simple(ip, attr);
 	if (error)
 		goto out_end_trans;
 
-	setattr_copy(inode, attr);
-	mark_inode_dirty(inode);
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
-	gfs2_dinode_out(ip, dibh->b_data);
-	brelse(dibh);
-
 	if (ouid != NO_QUOTA_CHANGE || ogid != NO_QUOTA_CHANGE) {
 		u64 blocks = gfs2_get_inode_blocks(&ip->i_inode);
 		gfs2_quota_change(ip, -blocks, ouid, ogid);
fs/gfs2/xattr.c +1 -13
index ecfd40d..439b61c 100644
@@ -1296,10 +1296,8 @@ fail:
 
 int gfs2_xattr_acl_chmod(struct gfs2_inode *ip, struct iattr *attr, char *data)
 {
-	struct inode *inode = &ip->i_inode;
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct gfs2_ea_location el;
-	struct buffer_head *dibh;
 	int error;
 
 	error = gfs2_ea_find(ip, GFS2_EATYPE_SYS, GFS2_POSIX_ACL_ACCESS, &el);
@@ -1321,17 +1319,7 @@ int gfs2_xattr_acl_chmod(struct gfs2_inode *ip, struct iattr *attr, char *data)
 	if (error)
 		return error;
 
-	error = gfs2_meta_inode_buffer(ip, &dibh);
-	if (error)
-		goto out_trans_end;
-
-	setattr_copy(inode, attr);
-	mark_inode_dirty(inode);
-	gfs2_trans_add_bh(ip->i_gl, dibh, 1);
-	gfs2_dinode_out(ip, dibh->b_data);
-	brelse(dibh);
-
-out_trans_end:
+	error = gfs2_setattr_simple(ip, attr);
 	gfs2_trans_end(sdp);
 	return error;
 }
Signature · 1 line
1.7.3.3
#5 lore raw

PATCH 5/15 GFS2: fs/gfs2/glock.c: Use printf extension %pV

In reply to Steven Whitehouse

From: Joe Perches <[email protected]>

Using %pV reduces the number of printk calls and
eliminates any possible message interleaving from
other printk calls.

fs/gfs2/glock.c +7 -2
index 13155f6..a9b53a4 100644
@@ -951,17 +951,22 @@ int gfs2_glock_wait(struct gfs2_holder *gh)
 
 void gfs2_print_dbg(struct seq_file *seq, const char *fmt, ...)
 {
+	struct va_format vaf;
 	va_list args;
 
 	va_start(args, fmt);
+
 	if (seq) {
 		struct gfs2_glock_iter *gi = seq->private;
 		vsprintf(gi->string, fmt, args);
 		seq_printf(seq, gi->string);
 	} else {
-		printk(KERN_ERR " ");
-		vprintk(fmt, args);
+		vaf.fmt = fmt;
+		vaf.va = &args;
+
+		printk(KERN_ERR " %pV", &vaf);
 	}
+
 	va_end(args);
 }
 
Signature · 1 line
1.7.3.3
#6 lore raw

PATCH 6/15 GFS2: fs/gfs2/glock.h: Add __attribute__((format(printf,2,3)) to gfs2_print_dbg

In reply to Steven Whitehouse

From: Joe Perches <[email protected]>

Functions that use printf formatting, especially those that use %pV, should have their uses of printf format and arguments checked by the compiler.

fs/gfs2/glock.h +2 -0
index db1c26d..a12d117 100644
@@ -212,6 +212,8 @@ int gfs2_glock_nq_num(struct gfs2_sbd *sdp,
 int gfs2_glock_nq_m(unsigned int num_gh, struct gfs2_holder *ghs);
 void gfs2_glock_dq_m(unsigned int num_gh, struct gfs2_holder *ghs);
 void gfs2_glock_dq_uninit_m(unsigned int num_gh, struct gfs2_holder *ghs);
+
+__attribute__ ((format(printf, 2, 3)))
 void gfs2_print_dbg(struct seq_file *seq, const char *fmt, ...);
 
 /**
Signature · 1 line
1.7.3.3
#7 lore raw

PATCH 7/15 GFS2: Allow gfs2 to update quota usage values through the quotactl interface

In reply to Steven Whitehouse

From: Abhijith Das <[email protected]>

With this patch the gfs2_set_dqblk() function will be able to update the quota usage block count (FS_DQ_BCOUNT) in addition to the already supported FS_DQ_BHARD (limit) and FS_DQ_BSOFT (warn) fields of the dquot structure.

fs/gfs2/quota.c +11 -2
index f606baf..a689901 100644
@@ -666,6 +666,10 @@ static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
 			qp->qu_limit = cpu_to_be64(fdq->d_blk_hardlimit >> sdp->sd_fsb2bb_shift);
 			qd->qd_qb.qb_limit = qp->qu_limit;
 		}
+		if (fdq->d_fieldmask & FS_DQ_BCOUNT) {
+			qp->qu_value = cpu_to_be64(fdq->d_bcount >> sdp->sd_fsb2bb_shift);
+			qd->qd_qb.qb_value = qp->qu_value;
+		}
 	}
 
 	/* Write the quota into the quota file on disk */
@@ -1509,7 +1513,7 @@ out:
 }
 
 /* GFS2 only supports a subset of the XFS fields */
-#define GFS2_FIELDMASK (FS_DQ_BSOFT|FS_DQ_BHARD)
+#define GFS2_FIELDMASK (FS_DQ_BSOFT|FS_DQ_BHARD|FS_DQ_BCOUNT)
 
 static int gfs2_set_dqblk(struct super_block *sb, int type, qid_t id,
 			  struct fs_disk_quota *fdq)
@@ -1569,9 +1573,15 @@ static int gfs2_set_dqblk(struct super_block *sb, int type, qid_t id,
 	if ((fdq->d_fieldmask & FS_DQ_BSOFT) &&
 	    ((fdq->d_blk_softlimit >> sdp->sd_fsb2bb_shift) == be64_to_cpu(qd->qd_qb.qb_warn)))
 		fdq->d_fieldmask ^= FS_DQ_BSOFT;
+
 	if ((fdq->d_fieldmask & FS_DQ_BHARD) &&
 	    ((fdq->d_blk_hardlimit >> sdp->sd_fsb2bb_shift) == be64_to_cpu(qd->qd_qb.qb_limit)))
 		fdq->d_fieldmask ^= FS_DQ_BHARD;
+
+	if ((fdq->d_fieldmask & FS_DQ_BCOUNT) &&
+	    ((fdq->d_bcount >> sdp->sd_fsb2bb_shift) == be64_to_cpu(qd->qd_qb.qb_value)))
+		fdq->d_fieldmask ^= FS_DQ_BCOUNT;
+
 	if (fdq->d_fieldmask == 0)
 		goto out_i;
 
@@ -1620,4 +1630,3 @@ const struct quotactl_ops gfs2_quotactl_ops = {
 	.get_dqblk	= gfs2_get_dqblk,
 	.set_dqblk	= gfs2_set_dqblk,
 };
-
Signature · 1 line
1.7.3.3
#8 lore raw

PATCH 8/15 GFS2: Clean up of gdlm_lock function

In reply to Steven Whitehouse

The DLM never returns -EAGAIN in response to dlm_lock(), and even if it did, the test in gdlm_lock() was wrong anyway. Once that test is removed, it is possible to greatly simplify this code by simply using a "normal" error return code (0 for success).

We then no longer need the LM_OUT_ASYNC return code which can be removed.

fs/gfs2/glock.c +9 -22
index a9b53a4..2dd1d72 100644
@@ -541,21 +541,6 @@ out_locked:
 	spin_unlock(&gl->gl_spin);
 }
 
-static unsigned int gfs2_lm_lock(struct gfs2_sbd *sdp, void *lock,
-				 unsigned int req_state,
-				 unsigned int flags)
-{
-	int ret = LM_OUT_ERROR;
-
-	if (!sdp->sd_lockstruct.ls_ops->lm_lock)
-		return req_state == LM_ST_UNLOCKED ? 0 : req_state;
-
-	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
-		ret = sdp->sd_lockstruct.ls_ops->lm_lock(lock,
-							 req_state, flags);
-	return ret;
-}
-
 /**
  * do_xmote - Calls the DLM to change the state of a lock
  * @gl: The lock state
@@ -575,8 +560,8 @@ __acquires(&gl->gl_spin)
 
 	lck_flags &= (LM_FLAG_TRY | LM_FLAG_TRY_1CB | LM_FLAG_NOEXP |
 		      LM_FLAG_PRIORITY);
-	BUG_ON(gl->gl_state == target);
-	BUG_ON(gl->gl_state == gl->gl_target);
+	GLOCK_BUG_ON(gl, gl->gl_state == target);
+	GLOCK_BUG_ON(gl, gl->gl_state == gl->gl_target);
 	if ((target == LM_ST_UNLOCKED || target == LM_ST_DEFERRED) &&
 	    glops->go_inval) {
 		set_bit(GLF_INVALIDATE_IN_PROGRESS, &gl->gl_flags);
@@ -594,15 +579,17 @@ __acquires(&gl->gl_spin)
 	    gl->gl_state == LM_ST_DEFERRED) &&
 	    !(lck_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB)))
 		lck_flags |= LM_FLAG_TRY_1CB;
-	ret = gfs2_lm_lock(sdp, gl, target, lck_flags);
 
-	if (!(ret & LM_OUT_ASYNC)) {
-		finish_xmote(gl, ret);
+	if (sdp->sd_lockstruct.ls_ops->lm_lock)	{
+		/* lock_dlm */
+		ret = sdp->sd_lockstruct.ls_ops->lm_lock(gl, target, lck_flags);
+		GLOCK_BUG_ON(gl, ret);
+	} else { /* lock_nolock */
+		finish_xmote(gl, target);
 		if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
 			gfs2_glock_put(gl);
-	} else {
-		GLOCK_BUG_ON(gl, ret != LM_OUT_ASYNC);
 	}
+
 	spin_lock(&gl->gl_spin);
 }
 
fs/gfs2/glock.h +4 -8
index a12d117..ba6865c 100644
@@ -91,7 +91,7 @@ enum {
 #define GL_NOCACHE		0x00000400
   
 /*
- * lm_lock() and lm_async_cb return flags
+ * lm_async_cb return flags
  *
  * LM_OUT_ST_MASK
  * Masks the lower two bits of lock state in the returned value.
@@ -99,15 +99,11 @@ enum {
  * LM_OUT_CANCELED
  * The lock request was canceled.
  *
- * LM_OUT_ASYNC
- * The result of the request will be returned in an LM_CB_ASYNC callback.
- *
  */
 
 #define LM_OUT_ST_MASK		0x00000003
 #define LM_OUT_CANCELED		0x00000008
-#define LM_OUT_ASYNC		0x00000080
-#define LM_OUT_ERROR		0x00000100
+#define LM_OUT_ERROR		0x00000004
 
 /*
  * lm_recovery_done() messages
@@ -124,8 +120,8 @@ struct lm_lockops {
  	void (*lm_unmount) (struct gfs2_sbd *sdp);
 	void (*lm_withdraw) (struct gfs2_sbd *sdp);
 	void (*lm_put_lock) (struct kmem_cache *cachep, struct gfs2_glock *gl);
-	unsigned int (*lm_lock) (struct gfs2_glock *gl,
-				 unsigned int req_state, unsigned int flags);
+	int (*lm_lock) (struct gfs2_glock *gl, unsigned int req_state,
+			unsigned int flags);
 	void (*lm_cancel) (struct gfs2_glock *gl);
 	const match_table_t *lm_tokens;
 };
fs/gfs2/lock_dlm.c +4 -10
index 1c09425..f40ce34 100644
@@ -146,11 +146,10 @@ static u32 make_flags(const u32 lkid, const unsigned int gfs_flags,
 	return lkf;
 }
 
-static unsigned int gdlm_lock(struct gfs2_glock *gl,
-			      unsigned int req_state, unsigned int flags)
+static int gdlm_lock(struct gfs2_glock *gl, unsigned int req_state,
+		     unsigned int flags)
 {
 	struct lm_lockstruct *ls = &gl->gl_sbd->sd_lockstruct;
-	int error;
 	int req;
 	u32 lkf;
 
@@ -162,13 +161,8 @@ static unsigned int gdlm_lock(struct gfs2_glock *gl,
 	 * Submit the actual lock request.
 	 */
 
-	error = dlm_lock(ls->ls_dlm, req, &gl->gl_lksb, lkf, gl->gl_strname,
-			 GDLM_STRNAME_BYTES - 1, 0, gdlm_ast, gl, gdlm_bast);
-	if (error == -EAGAIN)
-		return 0;
-	if (error)
-		return LM_OUT_ERROR;
-	return LM_OUT_ASYNC;
+	return dlm_lock(ls->ls_dlm, req, &gl->gl_lksb, lkf, gl->gl_strname,
+			GDLM_STRNAME_BYTES - 1, 0, gdlm_ast, gl, gdlm_bast);
 }
 
 static void gdlm_put_lock(struct kmem_cache *cachep, struct gfs2_glock *gl)
Signature · 1 line
1.7.3.3
#9 lore raw

PATCH 9/15 GFS2: Remove duplicate #defines from glock.h

In reply to Steven Whitehouse

There are a number of duplicated #defines in glock.h plus one which is unused. This removes the extra definitions.

fs/gfs2/glock.h +0 -14
index ba6865c..691851c 100644
@@ -87,7 +87,6 @@ enum {
 #define GL_ASYNC		0x00000040
 #define GL_EXACT		0x00000080
 #define GL_SKIP			0x00000100
-#define GL_ATIME		0x00000200
 #define GL_NOCACHE		0x00000400
   
 /*
@@ -126,19 +125,6 @@ struct lm_lockops {
 	const match_table_t *lm_tokens;
 };
 
-#define LM_FLAG_TRY		0x00000001
-#define LM_FLAG_TRY_1CB		0x00000002
-#define LM_FLAG_NOEXP		0x00000004
-#define LM_FLAG_ANY		0x00000008
-#define LM_FLAG_PRIORITY	0x00000010
-
-#define GL_ASYNC		0x00000040
-#define GL_EXACT		0x00000080
-#define GL_SKIP			0x00000100
-#define GL_NOCACHE		0x00000400
-
-#define GLR_TRYFAILED		13
-
 extern struct workqueue_struct *gfs2_delete_workqueue;
 static inline struct gfs2_holder *gfs2_glock_is_locked_by_me(struct gfs2_glock *gl)
 {
Signature · 1 line
1.7.3.3
#10 lore raw

PATCH 10/15 GFS2: reread rindex when necessary to grow rindex

In reply to Steven Whitehouse

From: Benjamin Marzinski <[email protected]>

When GFS2 grew the filesystem, it was never rereading the rindex file during the grow. This is necessary for large grows when the filesystem is almost full, and GFS2 needs to use some of the space allocated earlier in the grow to complete it. Now, if GFS2 fails to reserve the necessary space and the rindex file is not uptodate, it rereads it. Also, the only difference between gfs2_ri_update() and gfs2_ri_update_special() was that gfs2_ri_update_special() didn't clear out the existing resource groups, since you knew that it was only called when there were no resource groups. Attempting to clear out the resource groups when there are none takes almost no time, and rarely happens, so I simply removed gfs2_ri_update_special().

fs/gfs2/rgrp.c +11 -42
index 33c8407..6375b38 100644
@@ -614,46 +614,6 @@ static int gfs2_ri_update(struct gfs2_inode *ip)
 }
 
 /**
- * gfs2_ri_update_special - Pull in a new resource index from the disk
- *
- * This is a special version that's safe to call from gfs2_inplace_reserve_i.
- * In this case we know that we don't have any resource groups in memory yet.
- *
- * @ip: pointer to the rindex inode
- *
- * Returns: 0 on successful update, error code otherwise
- */
-static int gfs2_ri_update_special(struct gfs2_inode *ip)
-{
-	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
-	struct inode *inode = &ip->i_inode;
-	struct file_ra_state ra_state;
-	struct gfs2_rgrpd *rgd;
-	unsigned int max_data = 0;
-	int error;
-
-	file_ra_state_init(&ra_state, inode->i_mapping);
-	for (sdp->sd_rgrps = 0;; sdp->sd_rgrps++) {
-		/* Ignore partials */
-		if ((sdp->sd_rgrps + 1) * sizeof(struct gfs2_rindex) >
-		    i_size_read(inode))
-			break;
-		error = read_rindex_entry(ip, &ra_state);
-		if (error) {
-			clear_rgrpdi(sdp);
-			return error;
-		}
-	}
-	list_for_each_entry(rgd, &sdp->sd_rindex_list, rd_list)
-		if (rgd->rd_data > max_data)
-			max_data = rgd->rd_data;
-	sdp->sd_max_rg_data = max_data;
-
-	sdp->sd_rindex_uptodate = 1;
-	return 0;
-}
-
-/**
  * gfs2_rindex_hold - Grab a lock on the rindex
  * @sdp: The GFS2 superblock
  * @ri_gh: the glock holder
@@ -1226,16 +1186,25 @@ int gfs2_inplace_reserve_i(struct gfs2_inode *ip, int hold_rindex,
 			error = gfs2_rindex_hold(sdp, &al->al_ri_gh);
 		else if (!sdp->sd_rgrps) /* We may not have the rindex read
 					    in, so: */
-			error = gfs2_ri_update_special(ip);
+			error = gfs2_ri_update(ip);
 		if (error)
 			return error;
 	}
 
+try_again:
 	do {
 		error = get_local_rgrp(ip, &last_unlinked);
 		/* If there is no space, flushing the log may release some */
-		if (error)
+		if (error) {
+			if (ip == GFS2_I(sdp->sd_rindex) &&
+			    !sdp->sd_rindex_uptodate) {
+				error = gfs2_ri_update(ip);
+				if (error)
+					return error;
+				goto try_again;
+			}
 			gfs2_log_flush(sdp, NULL);
+		}
 	} while (error && tries++ < 3);
 
 	if (error) {
Signature · 1 line
1.7.3.3
#11 lore raw

PATCH 11/15 GFS2: fix recursive locking during rindex truncates

In reply to Steven Whitehouse

From: Benjamin Marzinski <[email protected]>

When you truncate the rindex file, you need to avoid calling gfs2_rindex_hold, since you already hold it. However, if you haven't already read in the resource groups, you need to do that.

fs/gfs2/bmap.c +7 -2
index 5476c06..ae7d205 100644
@@ -780,7 +780,11 @@ static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh,
 	if (metadata)
 		revokes = (height) ? sdp->sd_inptrs : sdp->sd_diptrs;
 
-	error = gfs2_rindex_hold(sdp, &ip->i_alloc->al_ri_gh);
+	if (ip != GFS2_I(sdp->sd_rindex))
+		error = gfs2_rindex_hold(sdp, &ip->i_alloc->al_ri_gh);
+	else if (!sdp->sd_rgrps)
+		error = gfs2_ri_update(ip);
+
 	if (error)
 		return error;
 
@@ -879,7 +883,8 @@ out_rg_gunlock:
 out_rlist:
 	gfs2_rlist_free(&rlist);
 out:
-	gfs2_glock_dq_uninit(&ip->i_alloc->al_ri_gh);
+	if (ip != GFS2_I(sdp->sd_rindex))
+		gfs2_glock_dq_uninit(&ip->i_alloc->al_ri_gh);
 	return error;
 }
 
fs/gfs2/rgrp.c +1 -1
index 6375b38..25dbe5c 100644
@@ -583,7 +583,7 @@ static int read_rindex_entry(struct gfs2_inode *ip,
  * Returns: 0 on successful update, error code otherwise
  */
 
-static int gfs2_ri_update(struct gfs2_inode *ip)
+int gfs2_ri_update(struct gfs2_inode *ip)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct inode *inode = &ip->i_inode;
fs/gfs2/rgrp.h +1 -0
index 0e35c04..50c2bb0 100644
@@ -48,6 +48,7 @@ extern int gfs2_inplace_reserve_i(struct gfs2_inode *ip, int hold_rindex,
 
 extern void gfs2_inplace_release(struct gfs2_inode *ip);
 
+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);
 
Signature · 1 line
1.7.3.3
#12 lore raw

PATCH 12/15 GFS2: Fix uninitialised error value in previous patch

In reply to Steven Whitehouse
fs/gfs2/bmap.c +1 -1
index ae7d205..3c4039d 100644
@@ -763,7 +763,7 @@ static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh,
 	int metadata;
 	unsigned int revokes = 0;
 	int x;
-	int error;
+	int error = 0;
 
 	if (!*top)
 		sm->sm_first = 0;
Signature · 1 line
1.7.3.3
#13 lore raw

PATCH 13/15 GFS2: Merge glock state fields into a bitfield

In reply to Steven Whitehouse

We can only merge the fields into a bitfield if the locking rules for them are the same. In this case gl_spin covers all of the fields (write side) but a couple of them are used with GLF_LOCK as the read side lock, which should be ok since we know that the field in question won't be changing at the time.

The gl_req setting has to be done earlier (in glock.c) in order to place it under gl_spin. The gl_reply setting also has to be brought under gl_spin in order to comply with the new rules.

This saves 4*sizeof(unsigned int) per glock.

fs/gfs2/glock.c +7 -2
index 2dd1d72..08a8beb 100644
@@ -567,6 +567,7 @@ __acquires(&gl->gl_spin)
 		set_bit(GLF_INVALIDATE_IN_PROGRESS, &gl->gl_flags);
 		do_error(gl, 0); /* Fail queued try locks */
 	}
+	gl->gl_req = target;
 	spin_unlock(&gl->gl_spin);
 	if (glops->go_xmote_th)
 		glops->go_xmote_th(gl);
@@ -1353,24 +1354,28 @@ static int gfs2_should_freeze(const struct gfs2_glock *gl)
  * @gl: Pointer to the glock
  * @ret: The return value from the dlm
  *
+ * The gl_reply field is under the gl_spin lock so that it is ok
+ * to use a bitfield shared with other glock state fields.
  */
 
 void gfs2_glock_complete(struct gfs2_glock *gl, int ret)
 {
 	struct lm_lockstruct *ls = &gl->gl_sbd->sd_lockstruct;
 
+	spin_lock(&gl->gl_spin);
 	gl->gl_reply = ret;
 
 	if (unlikely(test_bit(DFL_BLOCK_LOCKS, &ls->ls_flags))) {
-		spin_lock(&gl->gl_spin);
 		if (gfs2_should_freeze(gl)) {
 			set_bit(GLF_FROZEN, &gl->gl_flags);
 			spin_unlock(&gl->gl_spin);
 			return;
 		}
-		spin_unlock(&gl->gl_spin);
 	}
+
+	spin_unlock(&gl->gl_spin);
 	set_bit(GLF_REPLY_PENDING, &gl->gl_flags);
+	smp_wmb();
 	gfs2_glock_hold(gl);
 	if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0)
 		gfs2_glock_put(gl);
fs/gfs2/incore.h +7 -5
index 764fbb4..8d3d2b4 100644
@@ -207,12 +207,14 @@ struct gfs2_glock {
 
 	spinlock_t gl_spin;
 
-	unsigned int gl_state;
-	unsigned int gl_target;
-	unsigned int gl_reply;
+	/* State fields protected by gl_spin */
+	unsigned int gl_state:2,	/* Current state */
+		     gl_target:2,	/* Target state */
+		     gl_demote_state:2,	/* State requested by remote node */
+		     gl_req:2,		/* State in last dlm request */
+		     gl_reply:8;	/* Last reply from the dlm */
+
 	unsigned int gl_hash;
-	unsigned int gl_req;
-	unsigned int gl_demote_state; /* state requested by remote node */
 	unsigned long gl_demote_time; /* time of first demote request */
 	struct list_head gl_holders;
 
fs/gfs2/lock_dlm.c +0 -1
index f40ce34..6e493ae 100644
@@ -153,7 +153,6 @@ static int gdlm_lock(struct gfs2_glock *gl, unsigned int req_state,
 	int req;
 	u32 lkf;
 
-	gl->gl_req = req_state;
 	req = make_mode(req_state);
 	lkf = make_flags(gl->gl_lksb.sb_lkid, flags, req);
 
Signature · 1 line
1.7.3.3
#14 lore raw

PATCH 14/15 GFS2: fsck.gfs2 reported statfs error after gfs2_grow

In reply to Steven Whitehouse

From: Bob Peterson <[email protected]>

When you do gfs2_grow it failed to take the very last rgrp into account when adding up the new free space due to an off-by-one error. It was not reading the last rgrp from the rindex because of a check for "<=" that should have been "<". Therefore, fsck.gfs2 was finding (and fixing) an error with the system statfs file.

fs/gfs2/rgrp.c +1 -1
index 25dbe5c..7293ea2 100644
@@ -500,7 +500,7 @@ u64 gfs2_ri_total(struct gfs2_sbd *sdp)
 	for (rgrps = 0;; rgrps++) {
 		loff_t pos = rgrps * sizeof(struct gfs2_rindex);
 
-		if (pos + sizeof(struct gfs2_rindex) >= i_size_read(inode))
+		if (pos + sizeof(struct gfs2_rindex) > i_size_read(inode))
 			break;
 		error = gfs2_internal_read(ip, &ra_state, buf, &pos,
 					   sizeof(struct gfs2_rindex));
Signature · 1 line
1.7.3.3
#15 lore raw

PATCH 15/15 GFS2: Don't flush delete workqueue when releasing the transaction lock

In reply to Steven Whitehouse

There is no requirement to flush the delete workqueue before a gfs2 filesystem is suspended. The workqueue's work will just be suspended along with the rest of the tasks on the filesystem.

The resolves a deadlock situation where the transaction lock's demotion code was trying to flush the delete workqueue while at the same time, the workqueue was waiting for the transaction lock.

The delete workqueue is flushed by gfs2_make_fs_ro() already, so that umount/remount are correctly protected anyway.

fs/gfs2/glops.c +0 -1
index 0d149dc..263561b 100644
@@ -325,7 +325,6 @@ static void trans_go_sync(struct gfs2_glock *gl)
 
 	if (gl->gl_state != LM_ST_UNLOCKED &&
 	    test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags)) {
-		flush_workqueue(gfs2_delete_workqueue);
 		gfs2_meta_syncfs(sdp);
 		gfs2_log_shutdown(sdp);
 	}
Signature · 1 line
1.7.3.3