Merge tag 'gfs2-merge-window' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2

Pull GFS2 updates from Bob Peterson:
"Here is a list of patches we've accumulated for GFS2 for the current
upstream merge window.

Most of the patches fix GFS2 quotas, which were not properly enforced.
There's another that adds me as a GFS2 co-maintainer, and a couple
patches that fix a kernel panic doing splice_write on GFS2 as well as
a few correctness patches"

* tag 'gfs2-merge-window' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2:
gfs2: fix quota refresh race in do_glock()
gfs2: incorrect check for debugfs returns
gfs2: allow fallocate to max out quotas/fs efficiently
gfs2: allow quota_check and inplace_reserve to return available blocks
gfs2: perform quota checks against allocation parameters
GFS2: Move gfs2_file_splice_write outside of #ifdef
GFS2: Allocate reservation during splice_write
GFS2: gfs2_set_acl(): Cache "no acl" as well
Add myself (Bob Peterson) as a maintainer of GFS2

MAINTAINERS

···
 
 GFS2 FILE SYSTEM
 M:	Steven Whitehouse <[e-mail]>
+M:	Bob Peterson <[e-mail]>
 L:	[e-mail]
 W:	http://sources.redhat.com/cluster/
-T:	git git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-3.0-fixes.git
-T:	git git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-3.0-nmw.git
+T:	git git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2.git
 S:	Supported
 F:	Documentation/filesystems/gfs2*.txt
 F:	fs/gfs2/

fs/gfs2/acl.c

···
 	error = __gfs2_xattr_set(inode, name, data, len, 0, GFS2_EATYPE_SYS);
 	if (error)
 		goto out;
-
-	if (acl)
-		set_cached_acl(inode, type, acl);
-	else
-		forget_cached_acl(inode, type);
+	set_cached_acl(inode, type, acl);
 out:
 	kfree(data);
 	return error;

fs/gfs2/aops.c

···
 
 	if (alloc_required) {
 		struct gfs2_alloc_parms ap = { .aflags = 0, };
-		error = gfs2_quota_lock_check(ip);
+		requested = data_blocks + ind_blocks;
+		ap.target = requested;
+		error = gfs2_quota_lock_check(ip, &ap);
 		if (error)
 			goto out_unlock;
 
-		requested = data_blocks + ind_blocks;
-		ap.target = requested;
 		error = gfs2_inplace_reserve(ip, &ap);
 		if (error)
 			goto out_qunlock;

fs/gfs2/bmap.c

···
 
 	if (gfs2_is_stuffed(ip) &&
 	    (size > (sdp->sd_sb.sb_bsize - sizeof(struct gfs2_dinode)))) {
-		error = gfs2_quota_lock_check(ip);
+		error = gfs2_quota_lock_check(ip, &ap);
 		if (error)
 			return error;
 

fs/gfs2/file.c

···
 	if (ret)
 		goto out_unlock;
 
-	ret = gfs2_quota_lock_check(ip);
-	if (ret)
-		goto out_unlock;
 	gfs2_write_calc_reserv(ip, PAGE_CACHE_SIZE, &data_blocks, &ind_blocks);
 	ap.target = data_blocks + ind_blocks;
+	ret = gfs2_quota_lock_check(ip, &ap);
+	if (ret)
+		goto out_unlock;
 	ret = gfs2_inplace_reserve(ip, &ap);
 	if (ret)
 		goto out_quota_unlock;
···
 	brelse(dibh);
 	return error;
 }
-
-static void calc_max_reserv(struct gfs2_inode *ip, loff_t max, loff_t *len,
-			    unsigned int *data_blocks, unsigned int *ind_blocks)
+/**
+ * calc_max_reserv() - Reverse of write_calc_reserv. Given a number of
+ *                     blocks, determine how many bytes can be written.
+ * @ip:          The inode in question.
+ * @len:         Max cap of bytes. What we return in *len must be <= this.
+ * @data_blocks: Compute and return the number of data blocks needed
+ * @ind_blocks:  Compute and return the number of indirect blocks needed
+ * @max_blocks:  The total blocks available to work with.
+ *
+ * Returns: void, but @len, @data_blocks and @ind_blocks are filled in.
+ */
+static void calc_max_reserv(struct gfs2_inode *ip, loff_t *len,
+			    unsigned int *data_blocks, unsigned int *ind_blocks,
+			    unsigned int max_blocks)
 {
+	loff_t max = *len;
 	const struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
-	unsigned int max_blocks = ip->i_rgd->rd_free_clone;
 	unsigned int tmp, max_data = max_blocks - 3 * (sdp->sd_max_height - 1);
 
 	for (tmp = max_data; tmp > sdp->sd_diptrs;) {
 		tmp = DIV_ROUND_UP(tmp, sdp->sd_inptrs);
 		max_data -= tmp;
 	}
-	/* This calculation isn't the exact reverse of gfs2_write_calc_reserve,
-	   so it might end up with fewer data blocks */
-	if (max_data <= *data_blocks)
-		return;
+
 	*data_blocks = max_data;
 	*ind_blocks = max_blocks - max_data;
 	*len = ((loff_t)max_data - 3) << sdp->sd_sb.sb_bsize_shift;
···
 	struct gfs2_inode *ip = GFS2_I(inode);
 	struct gfs2_alloc_parms ap = { .aflags = 0, };
 	unsigned int data_blocks = 0, ind_blocks = 0, rblocks;
-	loff_t bytes, max_bytes;
+	loff_t bytes, max_bytes, max_blks = UINT_MAX;
 	int error;
 	const loff_t pos = offset;
 	const loff_t count = len;
···
 
 	gfs2_size_hint(file, offset, len);
 
+	gfs2_write_calc_reserv(ip, PAGE_SIZE, &data_blocks, &ind_blocks);
+	ap.min_target = data_blocks + ind_blocks;
+
 	while (len > 0) {
 		if (len < bytes)
 			bytes = len;
···
 			offset += bytes;
 			continue;
 		}
-		error = gfs2_quota_lock_check(ip);
+
+		/* We need to determine how many bytes we can actually
+		 * fallocate without exceeding quota or going over the
+		 * end of the fs. We start off optimistically by assuming
+		 * we can write max_bytes */
+		max_bytes = (len > max_chunk_size) ? max_chunk_size : len;
+
+		/* Since max_bytes is most likely a theoretical max, we
+		 * calculate a more realistic 'bytes' to serve as a good
+		 * starting point for the number of bytes we may be able
+		 * to write */
+		gfs2_write_calc_reserv(ip, bytes, &data_blocks, &ind_blocks);
+		ap.target = data_blocks + ind_blocks;
+
+		error = gfs2_quota_lock_check(ip, &ap);
 		if (error)
 			return error;
-retry:
-		gfs2_write_calc_reserv(ip, bytes, &data_blocks, &ind_blocks);
+		/* ap.allowed tells us how many blocks quota will allow
+		 * us to write. Check if this reduces max_blks */
+		if (ap.allowed && ap.allowed < max_blks)
+			max_blks = ap.allowed;
 
-		ap.target = data_blocks + ind_blocks;
 		error = gfs2_inplace_reserve(ip, &ap);
-		if (error) {
-			if (error == -ENOSPC && bytes > sdp->sd_sb.sb_bsize) {
-				bytes >>= 1;
-				bytes &= bsize_mask;
-				if (bytes == 0)
-					bytes = sdp->sd_sb.sb_bsize;
-				goto retry;
-			}
+		if (error)
 			goto out_qunlock;
-		}
-		max_bytes = bytes;
-		calc_max_reserv(ip, (len > max_chunk_size)? max_chunk_size: len,
-				&max_bytes, &data_blocks, &ind_blocks);
+
+		/* check if the selected rgrp limits our max_blks further */
+		if (ap.allowed && ap.allowed < max_blks)
+			max_blks = ap.allowed;
+
+		/* Almost done. Calculate bytes that can be written using
+		 * max_blks. We also recompute max_bytes, data_blocks and
+		 * ind_blocks */
+		calc_max_reserv(ip, &max_bytes, &data_blocks,
+				&ind_blocks, max_blks);
 
 		rblocks = RES_DINODE + ind_blocks + RES_STATFS + RES_QUOTA +
 			  RES_RG_HDR + gfs2_rg_blocks(ip, data_blocks + ind_blocks);
···
 	gfs2_holder_uninit(&gh);
 	mutex_unlock(&inode->i_mutex);
 	return ret;
+}
+
+static ssize_t gfs2_file_splice_write(struct pipe_inode_info *pipe,
+				      struct file *out, loff_t *ppos,
+				      size_t len, unsigned int flags)
+{
+	int error;
+	struct gfs2_inode *ip = GFS2_I(out->f_mapping->host);
+
+	error = gfs2_rs_alloc(ip);
+	if (error)
+		return (ssize_t)error;
+
+	gfs2_size_hint(out, *ppos, len);
+
+	return iter_file_splice_write(pipe, out, ppos, len, flags);
 }
 
 #ifdef CONFIG_GFS2_FS_LOCKING_DLM
···
 	.lock		= gfs2_lock,
 	.flock		= gfs2_flock,
 	.splice_read	= generic_file_splice_read,
-	.splice_write	= iter_file_splice_write,
+	.splice_write	= gfs2_file_splice_write,
 	.setlease	= simple_nosetlease,
 	.fallocate	= gfs2_fallocate,
 };
···
 	.release	= gfs2_release,
 	.fsync		= gfs2_fsync,
 	.splice_read	= generic_file_splice_read,
-	.splice_write	= iter_file_splice_write,
+	.splice_write	= gfs2_file_splice_write,
 	.setlease	= generic_setlease,
 	.fallocate	= gfs2_fallocate,
 };

fs/gfs2/glock.c

···
 
 int gfs2_create_debugfs_file(struct gfs2_sbd *sdp)
 {
-	sdp->debugfs_dir = debugfs_create_dir(sdp->sd_table_name, gfs2_root);
-	if (!sdp->debugfs_dir)
-		return -ENOMEM;
-	sdp->debugfs_dentry_glocks = debugfs_create_file("glocks",
-							 S_IFREG | S_IRUGO,
-							 sdp->debugfs_dir, sdp,
-							 &gfs2_glocks_fops);
-	if (!sdp->debugfs_dentry_glocks)
-		goto fail;
+	struct dentry *dent;
 
-	sdp->debugfs_dentry_glstats = debugfs_create_file("glstats",
-							S_IFREG | S_IRUGO,
-							sdp->debugfs_dir, sdp,
-							&gfs2_glstats_fops);
-	if (!sdp->debugfs_dentry_glstats)
+	dent = debugfs_create_dir(sdp->sd_table_name, gfs2_root);
+	if (IS_ERR_OR_NULL(dent))
 		goto fail;
+	sdp->debugfs_dir = dent;
 
-	sdp->debugfs_dentry_sbstats = debugfs_create_file("sbstats",
-							S_IFREG | S_IRUGO,
-							sdp->debugfs_dir, sdp,
-							&gfs2_sbstats_fops);
-	if (!sdp->debugfs_dentry_sbstats)
+	dent = debugfs_create_file("glocks",
+				   S_IFREG | S_IRUGO,
+				   sdp->debugfs_dir, sdp,
+				   &gfs2_glocks_fops);
+	if (IS_ERR_OR_NULL(dent))
 		goto fail;
+	sdp->debugfs_dentry_glocks = dent;
+
+	dent = debugfs_create_file("glstats",
+				   S_IFREG | S_IRUGO,
+				   sdp->debugfs_dir, sdp,
+				   &gfs2_glstats_fops);
+	if (IS_ERR_OR_NULL(dent))
+		goto fail;
+	sdp->debugfs_dentry_glstats = dent;
+
+	dent = debugfs_create_file("sbstats",
+				   S_IFREG | S_IRUGO,
+				   sdp->debugfs_dir, sdp,
+				   &gfs2_sbstats_fops);
+	if (IS_ERR_OR_NULL(dent))
+		goto fail;
+	sdp->debugfs_dentry_sbstats = dent;
 
 	return 0;
 fail:
 	gfs2_delete_debugfs_file(sdp);
-	return -ENOMEM;
+	return dent ? PTR_ERR(dent) : -ENOMEM;
 }
 
 void gfs2_delete_debugfs_file(struct gfs2_sbd *sdp)
···
 int gfs2_register_debugfs(void)
 {
 	gfs2_root = debugfs_create_dir("gfs2", NULL);
+	if (IS_ERR(gfs2_root))
+		return PTR_ERR(gfs2_root);
 	return gfs2_root ? 0 : -ENOMEM;
 }
 

fs/gfs2/incore.h

···
  * to the allocation code.
  */
 struct gfs2_alloc_parms {
-	u32 target;
+	u64 target;
+	u32 min_target;
 	u32 aflags;
+	u64 allowed;
 };
 
 enum {

fs/gfs2/inode.c

···
 	struct gfs2_alloc_parms ap = { .target = *dblocks, .aflags = flags, };
 	int error;
 
-	error = gfs2_quota_lock_check(ip);
+	error = gfs2_quota_lock_check(ip, &ap);
 	if (error)
 		goto out;
 
···
 	int error;
 
 	if (da->nr_blocks) {
-		error = gfs2_quota_lock_check(dip);
+		error = gfs2_quota_lock_check(dip, &ap);
 		if (error)
 			goto fail_quota_locks;
 
···
 
 	if (da.nr_blocks) {
 		struct gfs2_alloc_parms ap = { .target = da.nr_blocks, };
-		error = gfs2_quota_lock_check(dip);
+		error = gfs2_quota_lock_check(dip, &ap);
 		if (error)
 			goto out_gunlock;
 
···
 
 	if (da.nr_blocks) {
 		struct gfs2_alloc_parms ap = { .target = da.nr_blocks, };
-		error = gfs2_quota_lock_check(ndip);
+		error = gfs2_quota_lock_check(ndip, &ap);
 		if (error)
 			goto out_gunlock;
 
···
 	kuid_t ouid, nuid;
 	kgid_t ogid, ngid;
 	int error;
+	struct gfs2_alloc_parms ap;
 
 	ouid = inode->i_uid;
 	ogid = inode->i_gid;
···
 	if (error)
 		goto out;
 
+	ap.target = gfs2_get_inode_blocks(&ip->i_inode);
+
 	if (!uid_eq(ouid, NO_UID_QUOTA_CHANGE) ||
 	    !gid_eq(ogid, NO_GID_QUOTA_CHANGE)) {
-		error = gfs2_quota_check(ip, nuid, ngid);
+		error = gfs2_quota_check(ip, nuid, ngid, &ap);
 		if (error)
 			goto out_gunlock_q;
 	}
···
 
 	if (!uid_eq(ouid, NO_UID_QUOTA_CHANGE) ||
 	    !gid_eq(ogid, NO_GID_QUOTA_CHANGE)) {
-		u64 blocks = gfs2_get_inode_blocks(&ip->i_inode);
-		gfs2_quota_change(ip, -blocks, ouid, ogid);
-		gfs2_quota_change(ip, blocks, nuid, ngid);
+		gfs2_quota_change(ip, -ap.target, ouid, ogid);
+		gfs2_quota_change(ip, ap.target, nuid, ngid);
 	}
 
 out_end_trans:

fs/gfs2/quota.c

···
 	if (error)
 		return error;
 
+	if (test_and_clear_bit(QDF_REFRESH, &qd->qd_flags))
+		force_refresh = FORCE;
+
 	qd->qd_qb = *(struct gfs2_quota_lvb *)qd->qd_gl->gl_lksb.sb_lvbptr;
 
 	if (force_refresh || qd->qd_qb.qb_magic != cpu_to_be32(GFS2_MAGIC)) {
···
 	     sizeof(struct gfs2_quota_data *), sort_qd, NULL);
 
 	for (x = 0; x < ip->i_res->rs_qa_qd_num; x++) {
-		int force = NO_FORCE;
 		qd = ip->i_res->rs_qa_qd[x];
-		if (test_and_clear_bit(QDF_REFRESH, &qd->qd_flags))
-			force = FORCE;
-		error = do_glock(qd, force, &ip->i_res->rs_qa_qd_ghs[x]);
+		error = do_glock(qd, NO_FORCE, &ip->i_res->rs_qa_qd_ghs[x]);
 		if (error)
 			break;
 	}
···
 	return 0;
 }
 
-int gfs2_quota_check(struct gfs2_inode *ip, kuid_t uid, kgid_t gid)
+/**
+ * gfs2_quota_check - check if allocating new blocks will exceed quota
+ * @ip:  The inode for which this check is being performed
+ * @uid: The uid to check against
+ * @gid: The gid to check against
+ * @ap:  The allocation parameters. ap->target contains the requested
+ *       blocks. ap->min_target, if set, contains the minimum blks
+ *       requested.
+ *
+ * Returns: 0 on success.
+ *                  min_req = ap->min_target ? ap->min_target : ap->target;
+ *                  quota must allow atleast min_req blks for success and
+ *                  ap->allowed is set to the number of blocks allowed
+ *
+ *          -EDQUOT otherwise, quota violation. ap->allowed is set to number
+ *                  of blocks available.
+ */
+int gfs2_quota_check(struct gfs2_inode *ip, kuid_t uid, kgid_t gid,
+		     struct gfs2_alloc_parms *ap)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct gfs2_quota_data *qd;
-	s64 value;
+	s64 value, warn, limit;
 	unsigned int x;
 	int error = 0;
 
+	ap->allowed = UINT_MAX; /* Assume we are permitted a whole lot */
 	if (!test_bit(GIF_QD_LOCKED, &ip->i_flags))
 		return 0;
 
···
 		      qid_eq(qd->qd_id, make_kqid_gid(gid))))
 			continue;
 
+		warn = (s64)be64_to_cpu(qd->qd_qb.qb_warn);
+		limit = (s64)be64_to_cpu(qd->qd_qb.qb_limit);
 		value = (s64)be64_to_cpu(qd->qd_qb.qb_value);
 		spin_lock(&qd_lock);
 		value += qd->qd_change;
 		spin_unlock(&qd_lock);
 
-		if (be64_to_cpu(qd->qd_qb.qb_limit) && (s64)be64_to_cpu(qd->qd_qb.qb_limit) < value) {
-			print_message(qd, "exceeded");
-			quota_send_warning(qd->qd_id,
-					   sdp->sd_vfs->s_dev, QUOTA_NL_BHARDWARN);
-
-			error = -EDQUOT;
-			break;
-		} else if (be64_to_cpu(qd->qd_qb.qb_warn) &&
-			   (s64)be64_to_cpu(qd->qd_qb.qb_warn) < value &&
+		if (limit > 0 && (limit - value) < ap->allowed)
+			ap->allowed = limit - value;
+		/* If we can't meet the target */
+		if (limit && limit < (value + (s64)ap->target)) {
+			/* If no min_target specified or we don't meet
+			 * min_target, return -EDQUOT */
+			if (!ap->min_target || ap->min_target > ap->allowed) {
+				print_message(qd, "exceeded");
+				quota_send_warning(qd->qd_id,
+						   sdp->sd_vfs->s_dev,
+						   QUOTA_NL_BHARDWARN);
+				error = -EDQUOT;
+				break;
+			}
+		} else if (warn && warn < value &&
 			   time_after_eq(jiffies, qd->qd_last_warn +
-					 gfs2_tune_get(sdp,
-						gt_quota_warn_period) * HZ)) {
+					 gfs2_tune_get(sdp, gt_quota_warn_period)
+					 * HZ)) {
 			quota_send_warning(qd->qd_id,
 					   sdp->sd_vfs->s_dev, QUOTA_NL_BSOFTWARN);
 			error = print_message(qd, "warning");
 			qd->qd_last_warn = jiffies;
 		}
 	}
-
 	return error;
 }
 

fs/gfs2/quota.h

···
 extern int gfs2_quota_lock(struct gfs2_inode *ip, kuid_t uid, kgid_t gid);
 extern void gfs2_quota_unlock(struct gfs2_inode *ip);
 
-extern int gfs2_quota_check(struct gfs2_inode *ip, kuid_t uid, kgid_t gid);
+extern int gfs2_quota_check(struct gfs2_inode *ip, kuid_t uid, kgid_t gid,
+			    struct gfs2_alloc_parms *ap);
 extern void gfs2_quota_change(struct gfs2_inode *ip, s64 change,
 			      kuid_t uid, kgid_t gid);
 
···
 
 extern void gfs2_wake_up_statfs(struct gfs2_sbd *sdp);
 
-static inline int gfs2_quota_lock_check(struct gfs2_inode *ip)
+static inline int gfs2_quota_lock_check(struct gfs2_inode *ip,
+					struct gfs2_alloc_parms *ap)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	int ret;
···
 		return ret;
 	if (sdp->sd_args.ar_quota != GFS2_QUOTA_ON)
 		return 0;
-	ret = gfs2_quota_check(ip, ip->i_inode.i_uid, ip->i_inode.i_gid);
+	ret = gfs2_quota_check(ip, ip->i_inode.i_uid, ip->i_inode.i_gid, ap);
 	if (ret)
 		gfs2_quota_unlock(ip);
 	return ret;

fs/gfs2/rgrp.c

···
  * @ip: the inode to reserve space for
  * @ap: the allocation parameters
  *
- * Returns: errno
+ * We try our best to find an rgrp that has at least ap->target blocks
+ * available. After a couple of passes (loops == 2), the prospects of finding
+ * such an rgrp diminish. At this stage, we return the first rgrp that has
+ * atleast ap->min_target blocks available. Either way, we set ap->allowed to
+ * the number of blocks available in the chosen rgrp.
+ *
+ * Returns: 0 on success,
+ *          -ENOMEM if a suitable rgrp can't be found
+ *          errno otherwise
  */
 
-int gfs2_inplace_reserve(struct gfs2_inode *ip, const struct gfs2_alloc_parms *ap)
+int gfs2_inplace_reserve(struct gfs2_inode *ip, struct gfs2_alloc_parms *ap)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct gfs2_rgrpd *begin = NULL;
···
 		/* Skip unuseable resource groups */
 		if ((rs->rs_rbm.rgd->rd_flags & (GFS2_RGF_NOALLOC |
 						 GFS2_RDF_ERROR)) ||
-		    (ap->target > rs->rs_rbm.rgd->rd_extfail_pt))
+		    (loops == 0 && ap->target > rs->rs_rbm.rgd->rd_extfail_pt))
 			goto skip_rgrp;
 
 		if (sdp->sd_args.ar_rgrplvb)
···
 			goto check_rgrp;
 
 		/* If rgrp has enough free space, use it */
-		if (rs->rs_rbm.rgd->rd_free_clone >= ap->target) {
+		if (rs->rs_rbm.rgd->rd_free_clone >= ap->target ||
+		    (loops == 2 && ap->min_target &&
+		     rs->rs_rbm.rgd->rd_free_clone >= ap->min_target)) {
 			ip->i_rgd = rs->rs_rbm.rgd;
+			ap->allowed = ip->i_rgd->rd_free_clone;
 			return 0;
 		}
-
 check_rgrp:
 		/* Check for unlinked inodes which can be reclaimed */
 		if (rs->rs_rbm.rgd->rd_flags & GFS2_RDF_CHECK)

fs/gfs2/rgrp.h

···
 extern struct gfs2_alloc *gfs2_alloc_get(struct gfs2_inode *ip);
 
 #define GFS2_AF_ORLOV 1
-extern int gfs2_inplace_reserve(struct gfs2_inode *ip, const struct gfs2_alloc_parms *ap);
+extern int gfs2_inplace_reserve(struct gfs2_inode *ip,
+				struct gfs2_alloc_parms *ap);
 extern void gfs2_inplace_release(struct gfs2_inode *ip);
 
 extern int gfs2_alloc_blocks(struct gfs2_inode *ip, u64 *bn, unsigned int *n,

fs/gfs2/xattr.c

···
 	if (error)
 		return error;
 
-	error = gfs2_quota_lock_check(ip);
+	error = gfs2_quota_lock_check(ip, &ap);
 	if (error)
 		return error;