quota-deadlock-on-pagelock-core.patch

lustre 1.6.5 source code

 #include <asm/byteorder.h>
 
 MODULE_AUTHOR("Jan Kara");
@@ -41,23 +40,14 @@ static void v1_mem2disk_dqblk(struct v1_
 static int v1_read_dqblk(struct dquot *dquot)
 {
 	int type = dquot->dq_type;
-	struct file *filp;
-	mm_segment_t fs;
-	loff_t offset;
 	struct v1_disk_dqblk dqblk;
 
-	filp = sb_dqopt(dquot->dq_sb)->files[type];
-	if (filp == (struct file *)NULL)
+	if (!sb_dqopt(dquot->dq_sb)->files[type])
 		return -EINVAL;
 
-	/* Now we are sure filp is valid */
-	offset = v1_dqoff(dquot->dq_id);
 	/* Set structure to 0s in case read fails/is after end of file */
 	memset(&dqblk, 0, sizeof(struct v1_disk_dqblk));
-	fs = get_fs();
-	set_fs(KERNEL_DS);
-	filp->f_op->read(filp, (char *)&dqblk, sizeof(struct v1_disk_dqblk), &offset);
-	set_fs(fs);
+	dquot->dq_sb->s_op->quota_read(dquot->dq_sb, type, (char *)&dqblk, sizeof(struct v1_disk_dqblk), v1_dqoff(dquot->dq_id));
 
 	v1_disk2mem_dqblk(&dquot->dq_dqb, &dqblk);
 	if (dquot->dq_dqb.dqb_bhardlimit == 0 && dquot->dq_dqb.dqb_bsoftlimit == 0 &&
@@ -71,26 +61,18 @@ static int v1_read_dqblk(struct dquot *d
 static int v1_commit_dqblk(struct dquot *dquot)
 {
 	short type = dquot->dq_type;
-	struct file *filp;
-	mm_segment_t fs;
-	loff_t offset;
 	ssize_t ret;
 	struct v1_disk_dqblk dqblk;
 
-	filp = sb_dqopt(dquot->dq_sb)->files[type];
-	offset = v1_dqoff(dquot->dq_id);
-	fs = get_fs();
-	set_fs(KERNEL_DS);
-
 	v1_mem2disk_dqblk(&dqblk, &dquot->dq_dqb);
 	if (dquot->dq_id == 0) {
 		dqblk.dqb_btime = sb_dqopt(dquot->dq_sb)->info[type].dqi_bgrace;
 		dqblk.dqb_itime = sb_dqopt(dquot->dq_sb)->info[type].dqi_igrace;
 	}
 	ret = 0;
-	if (filp)
-		ret = filp->f_op->write(filp, (char *)&dqblk,
-					sizeof(struct v1_disk_dqblk), &offset);
+	if (sb_dqopt(dquot->dq_sb)->files[type])
+		ret = dquot->dq_sb->s_op->quota_write(dquot->dq_sb, type, (char *)&dqblk,
+					sizeof(struct v1_disk_dqblk), v1_dqoff(dquot->dq_id));
 	if (ret != sizeof(struct v1_disk_dqblk)) {
 		printk(KERN_WARNING "VFS: dquota write failed on dev %s\n",
 			dquot->dq_sb->s_id);
@@ -101,7 +83,6 @@ static int v1_commit_dqblk(struct dquot 
 	ret = 0;
 
 out:
-	set_fs(fs);
 	dqstats.writes++;
 
 	return ret;
@@ -121,14 +102,11 @@ struct v2_disk_dqheader {
 
 static int v1_check_quota_file(struct super_block *sb, int type)
 {
-	struct file *f = sb_dqopt(sb)->files[type];
-	struct inode *inode = f->f_dentry->d_inode;
+	struct inode *inode = sb_dqopt(sb)->files[type];
 	ulong blocks;
 	size_t off; 
 	struct v2_disk_dqheader dqhead;
-	mm_segment_t fs;
 	ssize_t size;
-	loff_t offset = 0;
 	loff_t isize;
 	static const uint quota_magics[] = V2_INITQMAGICS;
 
@@ -140,10 +118,7 @@ static int v1_check_quota_file(struct su
 	if ((blocks % sizeof(struct v1_disk_dqblk) * BLOCK_SIZE + off) % sizeof(struct v1_disk_dqblk))
 		return 0;
 	/* Doublecheck whether we didn't get file with new format - with old quotactl() this could happen */
-	fs = get_fs();
-	set_fs(KERNEL_DS);
-	size = f->f_op->read(f, (char *)&dqhead, sizeof(struct v2_disk_dqheader), &offset);
-	set_fs(fs);
+	size = sb->s_op->quota_read(sb, type, (char *)&dqhead, sizeof(struct v2_disk_dqheader), 0);
 	if (size != sizeof(struct v2_disk_dqheader))
 		return 1;	/* Probably not new format */
 	if (le32_to_cpu(dqhead.dqh_magic) != quota_magics[type])
@@ -155,16 +130,10 @@ static int v1_check_quota_file(struct su
 static int v1_read_file_info(struct super_block *sb, int type)
 {
 	struct quota_info *dqopt = sb_dqopt(sb);
-	mm_segment_t fs;
-	loff_t offset;
-	struct file *filp = dqopt->files[type];
 	struct v1_disk_dqblk dqblk;
 	int ret;
 
-	offset = v1_dqoff(0);
-	fs = get_fs();
-	set_fs(KERNEL_DS);
-	if ((ret = filp->f_op->read(filp, (char *)&dqblk, sizeof(struct v1_disk_dqblk), &offset)) != sizeof(struct v1_disk_dqblk)) {
+	if ((ret = sb->s_op->quota_read(sb, type, (char *)&dqblk, sizeof(struct v1_disk_dqblk), v1_dqoff(0))) != sizeof(struct v1_disk_dqblk)) {
 		if (ret >= 0)
 			ret = -EIO;
 		goto out;
@@ -173,38 +142,31 @@ static int v1_read_file_info(struct supe
 	dqopt->info[type].dqi_igrace = dqblk.dqb_itime ? dqblk.dqb_itime : MAX_IQ_TIME;
 	dqopt->info[type].dqi_bgrace = dqblk.dqb_btime ? dqblk.dqb_btime : MAX_DQ_TIME;
 out:
-	set_fs(fs);
 	return ret;
 }
 
 static int v1_write_file_info(struct super_block *sb, int type)
 {
 	struct quota_info *dqopt = sb_dqopt(sb);
-	mm_segment_t fs;
-	struct file *filp = dqopt->files[type];
 	struct v1_disk_dqblk dqblk;
-	loff_t offset;
 	int ret;
 
 	dqopt->info[type].dqi_flags &= ~DQF_INFO_DIRTY;
-	offset = v1_dqoff(0);
-	fs = get_fs();
-	set_fs(KERNEL_DS);
-	if ((ret = filp->f_op->read(filp, (char *)&dqblk, sizeof(struct v1_disk_dqblk), &offset)) != sizeof(struct v1_disk_dqblk)) {
+	if ((ret = sb->s_op->quota_read(sb, type, (char *)&dqblk,
+	    sizeof(struct v1_disk_dqblk), v1_dqoff(0))) != sizeof(struct v1_disk_dqblk)) {
 		if (ret >= 0)
 			ret = -EIO;
 		goto out;
 	}
 	dqblk.dqb_itime = dqopt->info[type].dqi_igrace;
 	dqblk.dqb_btime = dqopt->info[type].dqi_bgrace;
-	offset = v1_dqoff(0);
-	ret = filp->f_op->write(filp, (char *)&dqblk, sizeof(struct v1_disk_dqblk), &offset);
+	ret = sb->s_op->quota_write(sb, type, (char *)&dqblk,
+	      sizeof(struct v1_disk_dqblk), v1_dqoff(0));
 	if (ret == sizeof(struct v1_disk_dqblk))
 		ret = 0;
 	else if (ret > 0)
 		ret = -EIO;
 out:
-	set_fs(fs);
 	return ret;
 }
 
diff -rup RH_2_6_9_55.orig/fs/quota_v2.c RH_2_6_9_55/fs/quota_v2.c
--- RH_2_6_9_55.orig/fs/quota_v2.c
+++ RH_2_6_9_55/fs/quota_v2.c
@@ -13,7 +13,6 @@
 #include <linux/slab.h>
 
 #include <asm/byteorder.h>
-#include <asm/uaccess.h>
 
 MODULE_AUTHOR("Jan Kara");
 MODULE_DESCRIPTION("Quota format v2 support");
@@ -30,19 +29,15 @@ typedef char *dqbuf_t;
 static int v2_check_quota_file(struct super_block *sb, int type)
 {
 	struct v2_disk_dqheader dqhead;
-	struct file *f = sb_dqopt(sb)->files[type];
-	mm_segment_t fs;
 	ssize_t size;
-	loff_t offset = 0;
 	static const uint quota_magics[] = V2_INITQMAGICS;
 	static const uint quota_versions[] = V2_INITQVERSIONS;
  
-	fs = get_fs();
-	set_fs(KERNEL_DS);
-	size = f->f_op->read(f, (char *)&dqhead, sizeof(struct v2_disk_dqheader), &offset);
-	set_fs(fs);
-	if (size != sizeof(struct v2_disk_dqheader))
+	size = sb->s_op->quota_read(sb, type, (char *)&dqhead, sizeof(struct v2_disk_dqheader), 0);
+	if (size != sizeof(struct v2_disk_dqheader)) {
+		printk("failed read\n");
 		return 0;
+	}
 	if (le32_to_cpu(dqhead.dqh_magic) != quota_magics[type] ||
 	    le32_to_cpu(dqhead.dqh_version) != quota_versions[type])
 		return 0;
@@ -52,20 +47,15 @@ static int v2_check_quota_file(struct su
 /* Read information header from quota file */
 static int v2_read_file_info(struct super_block *sb, int type)
 {
-	mm_segment_t fs;
 	struct v2_disk_dqinfo dinfo;
 	struct mem_dqinfo *info = sb_dqopt(sb)->info+type;
-	struct file *f = sb_dqopt(sb)->files[type];
 	ssize_t size;
-	loff_t offset = V2_DQINFOOFF;
 
-	fs = get_fs();
-	set_fs(KERNEL_DS);
-	size = f->f_op->read(f, (char *)&dinfo, sizeof(struct v2_disk_dqinfo), &offset);
-	set_fs(fs);
+	size = sb->s_op->quota_read(sb, type, (char *)&dinfo,
+	       sizeof(struct v2_disk_dqinfo), V2_DQINFOOFF);
 	if (size != sizeof(struct v2_disk_dqinfo)) {
 		printk(KERN_WARNING "Can't read info structure on device %s.\n",
-			f->f_dentry->d_sb->s_id);
+			sb->s_id);
 		return -1;
 	}
 	info->dqi_bgrace = le32_to_cpu(dinfo.dqi_bgrace);
@@ -80,12 +70,9 @@ static int v2_read_file_info(struct supe
 /* Write information header to quota file */
 static int v2_write_file_info(struct super_block *sb, int type)
 {
-	mm_segment_t fs;
 	struct v2_disk_dqinfo dinfo;
 	struct mem_dqinfo *info = sb_dqopt(sb)->info+type;
-	struct file *f = sb_dqopt(sb)->files[type];
 	ssize_t size;
-	loff_t offset = V2_DQINFOOFF;
 
 	spin_lock(&dq_data_lock);
 	info->dqi_flags &= ~DQF_INFO_DIRTY;
@@ -96,13 +83,11 @@ static int v2_write_file_info(struct sup
 	dinfo.dqi_blocks = cpu_to_le32(info->u.v2_i.dqi_blocks);
 	dinfo.dqi_free_blk = cpu_to_le32(info->u.v2_i.dqi_free_blk);
 	dinfo.dqi_free_entry = cpu_to_le32(info->u.v2_i.dqi_free_entry);
-	fs = get_fs();
-	set_fs(KERNEL_DS);
-	size = f->f_op->write(f, (char *)&dinfo, sizeof(struct v2_disk_dqinfo), &offset);
-	set_fs(fs);
+	size = sb->s_op->quota_write(sb, type, (char *)&dinfo,
+	       sizeof(struct v2_disk_dqinfo), V2_DQINFOOFF);
 	if (size != sizeof(struct v2_disk_dqinfo)) {
 		printk(KERN_WARNING "Can't write info structure on device %s.\n",
-			f->f_dentry->d_sb->s_id);
+			sb->s_id);
 		return -1;
 	}
 	return 0;
@@ -146,39 +131,24 @@ static inline void freedqbuf(dqbuf_t buf
 	kfree(buf);
 }
 
-static ssize_t read_blk(struct file *filp, uint blk, dqbuf_t buf)
+static inline ssize_t read_blk(struct super_block *sb, int type, uint blk, dqbuf_t buf)
 {
-	mm_segment_t fs;
-	ssize_t ret;
-	loff_t offset = blk<<V2_DQBLKSIZE_BITS;
-
 	memset(buf, 0, V2_DQBLKSIZE);
-	fs = get_fs();
-	set_fs(KERNEL_DS);
-	ret = filp->f_op->read(filp, (char *)buf, V2_DQBLKSIZE, &offset);
-	set_fs(fs);
-	return ret;
+	return sb->s_op->quota_read(sb, type, (char *)buf,
+	       V2_DQBLKSIZE, blk << V2_DQBLKSIZE_BITS);
 }
 
-static ssize_t write_blk(struct file *filp, uint blk, dqbuf_t buf)
+static inline ssize_t write_blk(struct super_block *sb, int type, uint blk, dqbuf_t buf)
 {
-	mm_segment_t fs;
-	ssize_t ret;
-	loff_t offset = blk<<V2_DQBLKSIZE_BITS;
-
-	fs = get_fs();
-	set_fs(KERNEL_DS);
-	ret = filp->f_op->write(filp, (char *)buf, V2_DQBLKSIZE, &offset);
-	set_fs(fs);
-	return ret;
-
+	return sb->s_op->quota_write(sb, type, (char *)buf,
+	       V2_DQBLKSIZE, blk << V2_DQBLKSIZE_BITS);
 }
 
 /* Remove empty block from list and return it */
-static int get_free_dqblk(struct file *filp, int type)
+static int get_free_dqblk(struct super_block *sb, int type)
 {
 	dqbuf_t buf = getdqbuf();
-	struct mem_dqinfo *info = sb_dqinfo(filp->f_dentry->d_sb, type);
+	struct mem_dqinfo *info = sb_dqinfo(sb, type);
 	struct v2_disk_dqdbheader *dh = (struct v2_disk_dqdbheader *)buf;
 	int ret, blk;
 
@@ -186,17 +156,18 @@ static int get_free_dqblk(struct file *f
 		return -ENOMEM;
 	if (info->u.v2_i.dqi_free_blk) {
 		blk = info->u.v2_i.dqi_free_blk;
-		if ((ret = read_blk(filp, blk, buf)) < 0)
+		if ((ret = read_blk(sb, type, blk, buf)) < 0)
 			goto out_buf;
 		info->u.v2_i.dqi_free_blk = le32_to_cpu(dh->dqdh_next_free);
 	}
 	else {
 		memset(buf, 0, V2_DQBLKSIZE);
-		if ((ret = write_blk(filp, info->u.v2_i.dqi_blocks, buf)) < 0)	/* Assure block allocation... */
+		/* Assure block allocation... */
+		if ((ret = write_blk(sb, type, info->u.v2_i.dqi_blocks, buf)) < 0)
 			goto out_buf;
 		blk = info->u.v2_i.dqi_blocks++;
 	}
-	mark_info_dirty(filp->f_dentry->d_sb, type);
+	mark_info_dirty(sb, type);
 	ret = blk;
 out_buf:
 	freedqbuf(buf);
@@ -204,9 +175,9 @@ out_buf:
 }
 
 /* Insert empty block to the list */
-static int put_free_dqblk(struct file *filp, int type, dqbuf_t buf, uint blk)
+static int put_free_dqblk(struct super_block *sb, int type, dqbuf_t buf, uint blk)
 {
-	struct mem_dqinfo *info = sb_dqinfo(filp->f_dentry->d_sb, type);
+	struct mem_dqinfo *info = sb_dqinfo(sb, type);
 	struct v2_disk_dqdbheader *dh = (struct v2_disk_dqdbheader *)buf;
 	int err;
 
@@ -214,17 +185,18 @@ static int put_free_dqblk(struct file *f
 	dh->dqdh_prev_free = cpu_to_le32(0);
 	dh->dqdh_entries = cpu_to_le16(0);
 	info->u.v2_i.dqi_free_blk = blk;
-	mark_info_dirty(filp->f_dentry->d_sb, type);
-	if ((err = write_blk(filp, blk, buf)) < 0)	/* Some strange block. We had better leave it... */
+	mark_info_dirty(sb, type);
+	/* Some strange block. We had better leave it... */
+	if ((err = write_blk(sb, type, blk, buf)) < 0)
 		return err;
 	return 0;
 }
 
 /* Remove given block from the list of blocks with free entries */
-static int remove_free_dqentry(struct file *filp, int type, dqbuf_t buf, uint blk)
+static int remove_free_dqentry(struct super_block *sb, int type, dqbuf_t buf, uint blk)
 {
 	dqbuf_t tmpbuf = getdqbuf();
-	struct mem_dqinfo *info = sb_dqinfo(filp->f_dentry->d_sb, type);
+	struct mem_dqinfo *info = sb_dqinfo(sb, type);
 	struct v2_disk_dqdbheader *dh = (struct v2_disk_dqdbheader *)buf;
 	uint nextblk = le32_to_cpu(dh->dqdh_next_free), prevblk = le32_to_cpu(dh->dqdh_prev_free);
 	int err;
@@ -232,26 +204,27 @@ static int remove_free_dqentry(struct fi
 	if (!tmpbuf)
 		return -ENOMEM;
 	if (nextblk) {
-		if ((err = read_blk(filp, nextblk, tmpbuf)) < 0)
+		if ((err = read_blk(sb, type, nextblk, tmpbuf)) < 0)
 			goto out_buf;
 		((struct v2_disk_dqdbheader *)tmpbuf)->dqdh_prev_free = dh->dqdh_prev_free;
-		if ((err = write_blk(filp, nextblk, tmpbuf)) < 0)
+		if ((err = write_blk(sb, type, nextblk, tmpbuf)) < 0)
 			goto out_buf;
 	}
 	if (prevblk) {
-		if ((err = read_blk(filp, prevblk, tmpbuf)) < 0)
+		if ((err = read_blk(sb, type, prevblk, tmpbuf)) < 0)
 			goto out_buf;
 		((struct v2_disk_dqdbheader *)tmpbuf)->dqdh_next_free = dh->dqdh_next_free;
-		if ((err = write_blk(filp, prevblk, tmpbuf)) < 0)
+		if ((err = write_blk(sb, type, prevblk, tmpbuf)) < 0)
 			goto out_buf;
 	}
 	else {
 		info->u.v2_i.dqi_free_entry = nextblk;
-		mark_info_dirty(filp->f_dentry->d_sb, type);
+		mark_info_dirty(sb, type);
 	}
 	freedqbuf(tmpbuf);
 	dh->dqdh_next_free = dh->dqdh_prev_free = cpu_to_le32(0);
-	if (write_blk(filp, blk, buf) < 0)	/* No matter whether write succeeds block is out of list */
+	/* No matter whether write succeeds block is out of list */
+	if (write_blk(sb, type, blk, buf) < 0)
 		printk(KERN_ERR "VFS: Can't write block (%u) with free entries.\n", blk);
 	return 0;
 out_buf:
@@ -260,10 +233,10 @@ out_buf:
 }
 
 /* Insert given block to the beginning of list with free entries */
-static int insert_free_dqentry(struct file *filp, int type, dqbuf_t buf, uint blk)
+static int insert_free_dqentry(struct super_block *sb, int type, dqbuf_t buf, uint blk)
 {
 	dqbuf_t tmpbuf = getdqbuf();
-	struct mem_dqinfo *info = sb_dqinfo(filp->f_dentry->d_sb, type);
+	struct mem_dqinfo *info = sb_dqinfo(sb, type);
 	struct v2_disk_dqdbheader *dh = (struct v2_disk_dqdbheader *)buf;
 	int err;
 
@@ -271,18 +244,18 @@ static int insert_free_dqentry(struct fi
 		return -ENOMEM;
 	dh->dqdh_next_free = cpu_to_le32(info->u.v2_i.dqi_free_entry);
 	dh->dqdh_prev_free = cpu_to_le32(0);
-	if ((err = write_blk(filp, blk, buf)) < 0)
+	if ((err = write_blk(sb, type, blk, buf)) < 0)
 		goto out_buf;
 	if (info->u.v2_i.dqi_free_entry) {
-		if ((err = read_blk(filp, info->u.v2_i.dqi_free_entry, tmpbuf)) < 0)
+		if ((err = read_blk(sb, type, info->u.v2_i.dqi_free_entry, tmpbuf)) < 0)
 			goto out_buf;
 		((struct v2_disk_dqdbheader *)tmpbuf)->dqdh_prev_free = cpu_to_le32(blk);
-		if ((err = write_blk(filp, info->u.v2_i.dqi_free_entry, tmpbuf)) < 0)
+		if ((err = write_blk(sb, type, info->u.v2_i.dqi_free_entry, tmpbuf)) < 0)
 			goto out_buf;
 	}
 	freedqbuf(tmpbuf);
 	info->u.v2_i.dqi_free_entry = blk;
-	mark_info_dirty(filp->f_dentry->d_sb, type);
+	mark_info_dirty(sb, type);
 	return 0;
 out_buf:
 	freedqbuf(tmpbuf);
@@ -292,8 +265,8 @@ out_buf: