patch-2.1.101 linux/fs/dquot.c

Next file: linux/fs/ext2/balloc.c
Previous file: linux/fs/buffer.c
Back to the patch index
Back to the overall index

diff -u --recursive --new-file v2.1.100/linux/fs/dquot.c linux/fs/dquot.c
@@ -9,33 +9,22 @@
  * have a consistent diskquota tracking system. The ideas of both
  * user and group quotas are based on the Melbourne quota system as
  * used on BSD derived systems. The internal implementation is 
- * based on the LINUX inode-subsystem with added complexity of the
- * diskquota system. This implementation is not based on any BSD
- * kernel sourcecode.
+ * based on one of the several variants of the LINUX inode-subsystem
+ * with added complexity of the diskquota system.
  * 
- * Version: $Id: dquot.c,v 1.11 1997/01/06 06:53:02 davem Exp $
+ * Version: $Id: dquot.c,v 6.3 1996/11/17 18:35:34 mvw Exp mvw $
  * 
- * Author:  Marco van Wieringen <mvw@mcs.ow.nl> <mvw@tnix.net>
- * 
- * Fixes:   Dmitry Gorodchanin <pgmdsg@ibi.com>, 11 Feb 96
- *	    removed race conditions in dqput(), dqget() and iput(). 
- *          Andi Kleen removed all verify_area() calls, 31 Dec 96  
- *          Nick Kralevich <nickkral@cal.alumni.berkeley.edu>, 21 Jul 97
- *          Fixed a condition where user and group quotas could get mixed up.
+ * Author:	Marco van Wieringen <mvw@planets.elm.net>
  *
- *          Chris Rankin <rankinc@bellsouth.net>, 31 Dec 97, 2-4 Jan 98
- *          Fixed kernel API so that the user can get the quota for any
- *          group s/he belongs to. Also return useful error codes when
- *          turning quotas off, and fixed sync_dquot() so that all devices
- *          are synced when dev==NODEV. 
- *
- * (C) Copyright 1994, 1995 Marco van Wieringen 
+ * Fixes:   Dmitry Gorodchanin <pgmdsg@ibi.com>, 11 Feb 96
  *
+ * (C) Copyright 1994 - 1997 Marco van Wieringen 
  */
 
 #include <linux/errno.h>
 #include <linux/kernel.h>
 #include <linux/sched.h>
+
 #include <linux/types.h>
 #include <linux/string.h>
 #include <linux/fcntl.h>
@@ -46,129 +35,153 @@
 #include <linux/mount.h>
 #include <linux/smp.h>
 #include <linux/smp_lock.h>
+#include <linux/init.h>
+#include <linux/slab.h>
 
 #include <asm/uaccess.h>
 
-#define __DQUOT_VERSION__	"dquot_5.6.0"
+#define __DQUOT_VERSION__	"dquot_6.4.0"
+
+int nr_dquots = 0, nr_free_dquots = 0;
+int max_dquots = NR_DQUOTS;
 
 static char quotamessage[MAX_QUOTA_MESSAGE];
 static char *quotatypes[] = INITQFNAMES;
 
-static int nr_dquots = 0, nr_free_dquots = 0;
-static struct dquot *hash_table[NR_DQHASH];
-static struct dquot *first_dquot;
+static kmem_cache_t *dquot_cachep;
+
+static struct dquot *dquot_hash[NR_DQHASH];
+static struct free_dquot_queue {
+	struct dquot *head;
+	struct dquot **last;
+} free_dquots = { NULL, &free_dquots.head };
+static struct dquot *inuse_list = NULL;
+static int dquot_updating[NR_DQHASH];
+
 static struct dqstats dqstats;
+static struct wait_queue *dquot_wait = (struct wait_queue *)NULL,
+                         *update_wait = (struct wait_queue *)NULL;
 
-static struct wait_queue *dquot_wait = (struct wait_queue *)NULL;
+static inline char is_enabled(struct vfsmount *vfsmnt, short type)
+{
+	switch (type) {
+		case USRQUOTA:
+			return((vfsmnt->mnt_dquot.flags & DQUOT_USR_ENABLED) != 0);
+		case GRPQUOTA:
+			return((vfsmnt->mnt_dquot.flags & DQUOT_GRP_ENABLED) != 0);
+	}
+	return(0);
+}
 
-extern void add_dquot_ref(kdev_t dev, short type);
-extern void reset_dquot_ptrs(kdev_t dev, short type);
+static inline char sb_has_quota_enabled(struct super_block *sb, short type)
+{
+	struct vfsmount *vfsmnt;
 
-#ifndef min
-#define min(a,b) ((a) < (b)) ? (a) : (b)
-#endif
+	return((vfsmnt = lookup_vfsmnt(sb->s_dev)) != (struct vfsmount *)NULL && is_enabled(vfsmnt, type));
+}
 
-/*
- * Functions for management of the hashlist.
- */
-static inline int const hashfn(kdev_t dev, unsigned int id, short type)
+static inline char dev_has_quota_enabled(kdev_t dev, short type)
 {
-	return((HASHDEV(dev) ^ id) * (MAXQUOTAS - type)) % NR_DQHASH;
+	struct vfsmount *vfsmnt;
+
+	return((vfsmnt = lookup_vfsmnt(dev)) != (struct vfsmount *)NULL && is_enabled(vfsmnt, type));
 }
 
-static inline struct dquot **const hash(kdev_t dev, unsigned int id, short type)
+static inline int const hashfn(kdev_t dev, unsigned int id, short type)
 {
-	return(hash_table + hashfn(dev, id, type));
+	return((HASHDEV(dev) ^ id) * (MAXQUOTAS - type)) % NR_DQHASH;
 }
 
-static inline int has_quota_enabled(kdev_t dev, short type)
+static inline void insert_dquot_hash(struct dquot *dquot)
 {
-	struct vfsmount *vfsmnt;
+	struct dquot **htable;
 
-	return((vfsmnt = lookup_vfsmnt(dev)) != (struct vfsmount *)NULL &&
-	       (vfsmnt->mnt_quotas[type] != (struct file *)NULL));
+	htable = &dquot_hash[hashfn(dquot->dq_dev, dquot->dq_id, dquot->dq_type)];
+	if ((dquot->dq_hash_next = *htable) != NULL)
+		(*htable)->dq_hash_pprev = &dquot->dq_hash_next;
+	*htable = dquot;
+	dquot->dq_hash_pprev = htable;
 }
 
-static void insert_dquot_free(struct dquot *dquot)
+static inline void hash_dquot(struct dquot *dquot)
 {
-	dquot->dq_next = first_dquot;
-	dquot->dq_prev = first_dquot->dq_prev;
-	dquot->dq_next->dq_prev = dquot;
-	dquot->dq_prev->dq_next = dquot;
-	first_dquot = dquot;
+	insert_dquot_hash(dquot);
 }
 
-static void remove_dquot_free(struct dquot *dquot)
+static inline void unhash_dquot(struct dquot *dquot)
 {
-	if (first_dquot == dquot)
-		first_dquot = first_dquot->dq_next;
-	if (dquot->dq_next)
-		dquot->dq_next->dq_prev = dquot->dq_prev;
-	if (dquot->dq_prev)
-		dquot->dq_prev->dq_next = dquot->dq_next;
-	dquot->dq_next = dquot->dq_prev = NODQUOT;
+	if (dquot->dq_hash_pprev) {
+		if (dquot->dq_hash_next)
+			dquot->dq_hash_next->dq_hash_pprev = dquot->dq_hash_pprev;
+		*(dquot->dq_hash_pprev) = dquot->dq_hash_next;
+		dquot->dq_hash_pprev = NULL;
+	}
 }
 
-static void insert_dquot_hash(struct dquot *dquot)
+static inline struct dquot *find_dquot(unsigned int hashent, kdev_t dev, unsigned int id, short type)
 {
-	struct dquot **hash_ent;
+	struct dquot *dquot;
 
-	hash_ent = hash(dquot->dq_dev, dquot->dq_id, dquot->dq_type);
-	dquot->dq_hash_next = *hash_ent;
-	dquot->dq_hash_prev = NODQUOT;
-	if (dquot->dq_hash_next)
-		dquot->dq_hash_next->dq_hash_prev = dquot;
-	*hash_ent = dquot;
+	for (dquot = dquot_hash[hashent]; dquot; dquot = dquot->dq_hash_next)
+		if (dquot->dq_dev == dev && dquot->dq_id == id && dquot->dq_type == type)
+			break;
+	return dquot;
 }
 
-static void remove_dquot_hash(struct dquot *dquot)
+static inline void put_dquot_head(struct dquot *dquot)
 {
-	struct dquot **hash_ent;
+	if ((dquot->dq_next = free_dquots.head) != NULL)
+		free_dquots.head->dq_pprev = &dquot->dq_next;
+	else
+		free_dquots.last = &dquot->dq_next;
+	free_dquots.head = dquot;
+	dquot->dq_pprev = &free_dquots.head;
+	nr_free_dquots++;
+}
 
-	hash_ent = hash(dquot->dq_dev, dquot->dq_id, dquot->dq_type);
-	if (*hash_ent == dquot)
-		*hash_ent = dquot->dq_hash_next;
-	if (dquot->dq_hash_next)
-		dquot->dq_hash_next->dq_hash_prev = dquot->dq_hash_prev;
-	if (dquot->dq_hash_prev)
-		dquot->dq_hash_prev->dq_hash_next = dquot->dq_hash_next;
-	dquot->dq_hash_prev = dquot->dq_hash_next = NODQUOT;
+static inline void put_dquot_last(struct dquot *dquot)
+{
+	dquot->dq_next = NULL;
+	dquot->dq_pprev = free_dquots.last;
+	*free_dquots.last = dquot;
+	free_dquots.last = &dquot->dq_next;
+	nr_free_dquots++;
 }
 
-static void put_last_free(struct dquot *dquot)
+static inline void remove_free_dquot(struct dquot *dquot)
 {
-	remove_dquot_free(dquot);
-	dquot->dq_prev = first_dquot->dq_prev;
-	dquot->dq_prev->dq_next = dquot;
-	dquot->dq_next = first_dquot;
-	dquot->dq_next->dq_prev = dquot;
+	if (dquot->dq_pprev) {
+		if (dquot->dq_next)
+			dquot->dq_next->dq_pprev = dquot->dq_pprev;
+		else
+			free_dquots.last = dquot->dq_pprev;
+		*dquot->dq_pprev = dquot->dq_next;
+		dquot->dq_pprev = NULL;
+		nr_free_dquots--;
+	}
 }
 
-static void grow_dquots(void)
+static inline void put_inuse(struct dquot *dquot)
 {
-	struct dquot *dquot;
-	int cnt;
+	if ((dquot->dq_next = inuse_list) != NULL)
+		inuse_list->dq_pprev = &dquot->dq_next;
+	inuse_list = dquot;
+	dquot->dq_pprev = &inuse_list;
+}
 
-	if (!(dquot = (struct dquot*) get_free_page(GFP_KERNEL)))
-		return;
-	dqstats.pages_allocated++;
-	cnt = PAGE_SIZE / sizeof(struct dquot);
-	nr_dquots += cnt;
-	nr_free_dquots += cnt;
-	if (!first_dquot) {
-		dquot->dq_next = dquot->dq_prev = first_dquot = dquot++;
-		cnt--;
+static inline void remove_inuse(struct dquot *dquot)
+{
+	if (dquot->dq_pprev) {
+		if (dquot->dq_next)
+			dquot->dq_next->dq_pprev = dquot->dq_pprev;
+		*dquot->dq_pprev = dquot->dq_next;
+		dquot->dq_pprev = NULL;
 	}
-	for (; cnt; cnt--)
-		insert_dquot_free(dquot++);
 }
 
-/*
- * Functions for locking and waiting on dquots.
- */
 static void __wait_on_dquot(struct dquot *dquot)
 {
-	struct wait_queue wait = {current, NULL};
+	struct wait_queue wait = { current, NULL };
 
 	add_wait_queue(&dquot->dq_wait, &wait);
 repeat:
@@ -202,39 +215,22 @@
 		wake_up(&dquot->dq_wait);
 	}
 }
-/*
- * Note that we don't want to disturb any wait-queues when we discard
- * an dquot.
- *
- * FIXME: As soon as we have a nice solution for the inode problem we
- *		  can also fix this one. I.e. the volatile part.
- */
-static void clear_dquot(struct dquot * dquot)
-{
-	struct wait_queue *wait;
-
-	wait_on_dquot(dquot);
-	remove_dquot_hash(dquot);
-	remove_dquot_free(dquot);
-	wait = ((volatile struct dquot *) dquot)->dq_wait;
-	if (dquot->dq_count)
-		nr_free_dquots++;
-	memset(dquot, 0, sizeof(*dquot));
-	((volatile struct dquot *) dquot)->dq_wait = wait;
-	insert_dquot_free(dquot);
-}
 
 static void write_dquot(struct dquot *dquot)
 {
-	short type = dquot->dq_type;
-	struct file *filp = dquot->dq_mnt->mnt_quotas[type];
+	short type;
+	struct file *filp;
 	mm_segment_t fs;
 	loff_t offset;
 
+	type = dquot->dq_type;
+	filp = dquot->dq_mnt->mnt_dquot.files[type];
+
 	if (!(dquot->dq_flags & DQ_MOD) || (filp == (struct file *)NULL))
 		return;
+
 	lock_dquot(dquot);
-	down(&dquot->dq_mnt->mnt_sem);
+	down(&dquot->dq_mnt->mnt_dquot.semaphore);
 	offset = dqoff(dquot->dq_id);
 	fs = get_fs();
 	set_fs(KERNEL_DS);
@@ -242,29 +238,35 @@
 	if (filp->f_op->write(filp, (char *)&dquot->dq_dqb, sizeof(struct dqblk), &offset) == sizeof(struct dqblk))
 		dquot->dq_flags &= ~DQ_MOD;
 
-	up(&dquot->dq_mnt->mnt_sem);
+	up(&dquot->dq_mnt->mnt_dquot.semaphore);
 	set_fs(fs);
+
 	unlock_dquot(dquot);
 	dqstats.writes++;
 }
 
 static void read_dquot(struct dquot *dquot)
 {
-	short type = dquot->dq_type;
-	struct file *filp = dquot->dq_mnt->mnt_quotas[type];
+	short type;
+	struct file *filp;
 	mm_segment_t fs;
 	loff_t offset;
 
+	type = dquot->dq_type;
+	filp = dquot->dq_mnt->mnt_dquot.files[type];
+
 	if (filp == (struct file *)NULL)
 		return;
+
 	lock_dquot(dquot);
-	down(&dquot->dq_mnt->mnt_sem);
+	down(&dquot->dq_mnt->mnt_dquot.semaphore);
 	offset = dqoff(dquot->dq_id);
 	fs = get_fs();
 	set_fs(KERNEL_DS);
 	filp->f_op->read(filp, (char *)&dquot->dq_dqb, sizeof(struct dqblk), &offset);
-	up(&dquot->dq_mnt->mnt_sem);
+	up(&dquot->dq_mnt->mnt_dquot.semaphore);
 	set_fs(fs);
+
 	if (dquot->dq_bhardlimit == 0 && dquot->dq_bsoftlimit == 0 &&
 	    dquot->dq_ihardlimit == 0 && dquot->dq_isoftlimit == 0)
 		dquot->dq_flags |= DQ_FAKE;
@@ -272,46 +274,292 @@
 	dqstats.reads++;
 }
 
+void clear_dquot(struct dquot *dquot)
+{
+        struct wait_queue *wait;
+
+        /* So we don't disappear. */
+        dquot->dq_count++;
+
+        wait_on_dquot(dquot);
+
+        if (--dquot->dq_count > 0)
+                remove_inuse(dquot);
+        else
+                remove_free_dquot(dquot);
+        unhash_dquot(dquot);
+        wait = dquot->dq_wait;
+        memset(dquot, 0, sizeof(*dquot)); barrier();
+        dquot->dq_wait = wait;
+        put_dquot_head(dquot);
+}
+
+void invalidate_dquots(kdev_t dev, short type)
+{
+	struct dquot *dquot, *next = NULL;
+	int pass = 0;
+
+	dquot = free_dquots.head;
+repeat:
+	while (dquot) {
+		next = dquot->dq_next;
+		if (dquot->dq_dev != dev || dquot->dq_type != type)
+			goto next;
+		clear_dquot(dquot);
+	next:
+		dquot = next;
+	}
+
+	if (pass == 0) {
+		dquot = inuse_list;
+		pass = 1;
+		goto repeat;
+	}
+}
+
 int sync_dquots(kdev_t dev, short type)
 {
-	struct dquot *dquot = first_dquot;
-	int i;
+	struct dquot *dquot, *next;
+	int pass = 0;
 
-	dqstats.syncs++;
-	for (i = 0; i < nr_dquots * 2; i++, dquot = dquot->dq_next) {
-		if ((dev != NODEV && dquot->dq_dev != dev) || dquot->dq_count == 0 )
-			continue;
-		if (type != -1 && dquot->dq_type != type)
-			continue;
+	dquot = free_dquots.head;
+repeat:
+	while (dquot) {
+		next = dquot->dq_next;
+		if ((dev && dquot->dq_dev != dev) ||
+                    (type != -1 && dquot->dq_type != type))
+			goto next;
 		wait_on_dquot(dquot);
 		if (dquot->dq_flags & DQ_MOD)
 			write_dquot(dquot);
+	next:
+		dquot = next;
+	}
+
+	if (pass == 0) {
+		dquot = inuse_list;
+		pass = 1;
+		goto repeat;
 	}
+	dqstats.syncs++;
 	return(0);
 }
 
-/*
- * Trash the cache for a certain type on a device.
- */
-void invalidate_dquots(kdev_t dev, short type)
+void dqput(struct dquot *dquot)
 {
-	struct dquot *dquot, *next;
-	int cnt;
+	if (!dquot)
+		return;
 
-	next = first_dquot;
-	for (cnt = nr_dquots ; cnt > 0 ; cnt--) {
-		dquot = next;
-		next = dquot->dq_next;
-		if (dquot->dq_dev != dev || dquot->dq_type != type)
+	/*
+	 * If the dq_mnt pointer isn't initialized this entry needs no
+	 * checking and doesn't need to be written. It just an empty
+	 * dquot that is put back on to the freelist.
+	 */
+	if (dquot->dq_mnt != (struct vfsmount *)NULL) {
+		dqstats.drops++;
+		wait_on_dquot(dquot);
+
+		if (!dquot->dq_count) {
+			printk("VFS: dqput: trying to free free dquot\n");
+			printk("VFS: device %s, dquot of %s %d\n", kdevname(dquot->dq_dev),
+			       quotatypes[dquot->dq_type], dquot->dq_id);
+			return;
+		}
+we_slept:
+		if (dquot->dq_count > 1) {
+			dquot->dq_count--;
+			return;
+		} else {
+			wake_up(&dquot_wait);
+
+			if (dquot->dq_flags & DQ_MOD) {
+				write_dquot(dquot);
+				wait_on_dquot(dquot);
+				goto we_slept;
+			}
+		}
+	}
+
+	if (--dquot->dq_count == 0) {
+		remove_inuse(dquot);
+		put_dquot_last(dquot); /* Place at end of LRU free queue */
+	}
+
+	return;
+}
+
+static void grow_dquots(void)
+{
+	struct dquot *dquot;
+	int cnt = 32;
+
+	while (cnt > 0) {
+		dquot = kmem_cache_alloc(dquot_cachep, SLAB_KERNEL);
+		if(!dquot)
+			return;
+
+		nr_dquots++;
+		memset((caddr_t)dquot, 0, sizeof(struct dquot));
+		put_dquot_head(dquot);
+		cnt--;
+	}
+}
+
+static struct dquot *find_best_candidate_weighted(struct dquot *dquot)
+{
+	int limit, myscore;
+	unsigned long bestscore;
+	struct dquot *best = NULL;
+
+	if (dquot) {
+		bestscore = 2147483647;
+		limit = nr_free_dquots >> 2;
+		do {
+			if (!((dquot->dq_flags & DQ_LOCKED) || (dquot->dq_flags & DQ_MOD))) {
+				myscore = dquot->dq_referenced;
+				if (myscore < bestscore) {
+					bestscore = myscore;
+					best = dquot;
+				}
+			}
+			dquot = dquot->dq_next;
+		} while (dquot && --limit);
+	}
+	return best;
+}
+
+static inline struct dquot *find_best_free(struct dquot *dquot)
+{
+	int limit;
+
+	if (dquot) {
+		limit = nr_free_dquots >> 5;
+		do {
+			if (dquot->dq_referenced == 0)
+				return dquot;
+			dquot = dquot->dq_next;
+		} while (dquot && --limit);
+	}
+	return NULL;
+}
+
+struct dquot *get_empty_dquot(void)
+{
+	struct dquot *dquot;
+
+repeat:
+	dquot = find_best_free(free_dquots.head);
+	if (!dquot)
+		goto pressure;
+got_it:
+	dquot->dq_count++;
+	wait_on_dquot(dquot);
+	unhash_dquot(dquot);
+	remove_free_dquot(dquot);
+
+	memset(dquot, 0, sizeof(*dquot));
+	dquot->dq_count = 1;
+
+	put_inuse(dquot);
+	return dquot;
+pressure:
+	if (nr_dquots < max_dquots) {
+		grow_dquots();
+		goto repeat;
+	}
+
+	dquot = find_best_candidate_weighted(free_dquots.head);
+	if (!dquot) {
+		printk("VFS: No free dquots, contact mvw@planets.elm.net\n");
+		sleep_on(&dquot_wait);
+		goto repeat;
+	}
+	if (dquot->dq_flags & DQ_LOCKED) {
+		wait_on_dquot(dquot);
+		goto repeat;
+	} else if (dquot->dq_flags & DQ_MOD) {
+		write_dquot(dquot);
+		goto repeat;
+	}
+	goto got_it;
+}
+
+struct dquot *dqget(kdev_t dev, unsigned int id, short type)
+{
+	unsigned int hashent = hashfn(dev, id, type);
+	struct dquot *dquot, *empty = NULL;
+	struct vfsmount *vfsmnt;
+
+        if ((vfsmnt = lookup_vfsmnt(dev)) == (struct vfsmount *)NULL || is_enabled(vfsmnt, type) == 0)
+                return(NODQUOT);
+
+we_slept:
+	if ((dquot = find_dquot(hashent, dev, id, type)) == NULL) {
+		if (empty == NULL) {
+			dquot_updating[hashent]++;
+			empty = get_empty_dquot();
+			if (!--dquot_updating[hashent])
+				wake_up(&update_wait);
+			goto we_slept;
+		}
+		dquot = empty;
+        	dquot->dq_id = id;
+        	dquot->dq_type = type;
+        	dquot->dq_dev = dev;
+        	dquot->dq_mnt = vfsmnt;
+        	read_dquot(dquot);
+		hash_dquot(dquot);
+	} else {
+		if (!dquot->dq_count++) {
+			remove_free_dquot(dquot);
+			put_inuse(dquot);
+		} else
+			dqstats.cache_hits++;
+		wait_on_dquot(dquot);
+		if (empty)
+			dqput(empty);
+	}
+
+	while (dquot_updating[hashent])
+		sleep_on(&update_wait);
+
+	dquot->dq_referenced++;
+	dqstats.lookups++;
+
+	return dquot;
+}
+
+static void add_dquot_ref(kdev_t dev, short type)
+{
+	struct file *filp;
+	struct inode *inode;
+
+	for (filp = inuse_filps; filp; filp = filp->f_next) {
+		inode = filp->f_dentry->d_inode;
+		if (!inode || inode->i_dev != dev)
 			continue;
-		if (dquot->dq_flags & DQ_LOCKED) {
-			printk("VFS: dquot busy on removed device %s\n", kdevname(dev));
+		if (filp->f_mode & FMODE_WRITE && inode->i_sb && inode->i_sb->dq_op) {
+			inode->i_sb->dq_op->initialize(inode, type);
+			inode->i_flags |= S_QUOTA;
+		}
+	}
+}
+
+static void reset_dquot_ptrs(kdev_t dev, short type)
+{
+	struct file *filp;
+	struct inode *inode;
+
+	for (filp = inuse_filps; filp; filp = filp->f_next) {
+		inode = filp->f_dentry->d_inode;
+		if (!inode || inode->i_dev != dev)
 			continue;
+		if (IS_QUOTAINIT(inode)) {
+			if (inode->i_sb && inode->i_sb->dq_op)
+				inode->i_sb->dq_op->drop(inode);
+			inode->i_dquot[type] = NODQUOT;
+			inode->i_flags &= ~S_QUOTA;
 		}
-		if (dquot->dq_flags & DQ_MOD)
-			write_dquot(dquot);
-		dqstats.drops++;
-		clear_dquot(dquot);
 	}
 }
 
@@ -359,229 +607,108 @@
 	unlock_dquot(dquot);
 }
 
-static inline int need_print_warning(short type, struct dquot *dquot)
+static inline char need_print_warning(short type, uid_t initiator, struct dquot *dquot)
 {
 	switch (type) {
 		case USRQUOTA:
-			return(current->fsuid == dquot->dq_id);
+			return(initiator == dquot->dq_id);
 		case GRPQUOTA:
-			return(current->fsgid == dquot->dq_id);
+			return(initiator == dquot->dq_id);
 	}
 	return(0);
 }
 
-static int check_idq(struct dquot *dquot, short type, u_long short inodes)
+static inline char ignore_hardlimit(struct dquot *dquot, uid_t initiator)
+{
+	return(initiator == 0 && dquot->dq_mnt->mnt_dquot.rsquash[dquot->dq_type] == 0);
+}
+
+static int check_idq(struct dquot *dquot, short type, u_long short inodes, uid_t initiator, struct tty_struct *tty)
 {
 	if (inodes <= 0 || dquot->dq_flags & DQ_FAKE)
 		return(QUOTA_OK);
+
 	if (dquot->dq_ihardlimit &&
-	   (dquot->dq_curinodes + inodes) > dquot->dq_ihardlimit && 
-	    !capable(CAP_SYS_RESOURCE)) {
+	   (dquot->dq_curinodes + inodes) > dquot->dq_ihardlimit &&
+            !ignore_hardlimit(dquot, initiator)) {
 		if ((dquot->dq_flags & DQ_INODES) == 0 &&
-                     need_print_warning(type, dquot)) {
-			sprintf(quotamessage, "%s: write failed, %s file limit reached\r\n",
+                     need_print_warning(type, initiator, dquot)) {
+			sprintf(quotamessage, "%s: write failed, %s file limit reached\n",
 			        dquot->dq_mnt->mnt_dirname, quotatypes[type]);
-			tty_write_message(current->tty, quotamessage);
+			tty_write_message(tty, quotamessage);
 			dquot->dq_flags |= DQ_INODES;
 		}
 		return(NO_QUOTA);
 	}
+
 	if (dquot->dq_isoftlimit &&
 	   (dquot->dq_curinodes + inodes) > dquot->dq_isoftlimit &&
-	    dquot->dq_itime && CURRENT_TIME >= dquot->dq_itime && 
-	    !capable(CAP_SYS_RESOURCE)) {
-                if (need_print_warning(type, dquot)) {
-			sprintf(quotamessage, "%s: warning, %s file quota exceeded too long.\r\n",
+	    dquot->dq_itime && CURRENT_TIME >= dquot->dq_itime &&
+            !ignore_hardlimit(dquot, initiator)) {
+                if (need_print_warning(type, initiator, dquot)) {
+			sprintf(quotamessage, "%s: warning, %s file quota exceeded too long.\n",
 		        	dquot->dq_mnt->mnt_dirname, quotatypes[type]);
-			tty_write_message(current->tty, quotamessage);
+			tty_write_message(tty, quotamessage);
 		}
 		return(NO_QUOTA);
 	}
+
 	if (dquot->dq_isoftlimit &&
 	   (dquot->dq_curinodes + inodes) > dquot->dq_isoftlimit &&
-	    dquot->dq_itime == 0 && 
-	    !capable(CAP_SYS_RESOURCE)) {
-                if (need_print_warning(type, dquot)) {
-			sprintf(quotamessage, "%s: warning, %s file quota exceeded\r\n",
+	    dquot->dq_itime == 0) {
+                if (need_print_warning(type, initiator, dquot)) {
+			sprintf(quotamessage, "%s: warning, %s file quota exceeded\n",
 		        	dquot->dq_mnt->mnt_dirname, quotatypes[type]);
-			tty_write_message(current->tty, quotamessage);
+			tty_write_message(tty, quotamessage);
 		}
-		dquot->dq_itime = CURRENT_TIME + dquot->dq_mnt->mnt_iexp[type];
+		dquot->dq_itime = CURRENT_TIME + dquot->dq_mnt->mnt_dquot.inode_expire[type];
 	}
+
 	return(QUOTA_OK);
 }
 
-static int check_bdq(struct dquot *dquot, short type, u_long blocks)
+static int check_bdq(struct dquot *dquot, short type, u_long blocks, uid_t initiator, struct tty_struct *tty, char warn)
 {
 	if (blocks <= 0 || dquot->dq_flags & DQ_FAKE)
 		return(QUOTA_OK);
+
 	if (dquot->dq_bhardlimit &&
-	   (dquot->dq_curblocks + blocks) > dquot->dq_bhardlimit && 
-	    !capable(CAP_SYS_RESOURCE)) {
-		if ((dquot->dq_flags & DQ_BLKS) == 0 &&
-                     need_print_warning(type, dquot)) {
-			sprintf(quotamessage, "%s: write failed, %s disk limit reached.\r\n",
+	   (dquot->dq_curblocks + blocks) > dquot->dq_bhardlimit &&
+            !ignore_hardlimit(dquot, initiator)) {
+		if (warn && (dquot->dq_flags & DQ_BLKS) == 0 &&
+                     need_print_warning(type, initiator, dquot)) {
+			sprintf(quotamessage, "%s: write failed, %s disk limit reached.\n",
 			        dquot->dq_mnt->mnt_dirname, quotatypes[type]);
-			tty_write_message(current->tty, quotamessage);
+			tty_write_message(tty, quotamessage);
 			dquot->dq_flags |= DQ_BLKS;
 		}
 		return(NO_QUOTA);
 	}
+
 	if (dquot->dq_bsoftlimit &&
 	   (dquot->dq_curblocks + blocks) > dquot->dq_bsoftlimit &&
-	    dquot->dq_btime && CURRENT_TIME >= dquot->dq_btime && 
-	    !capable(CAP_SYS_RESOURCE)) {
-                if (need_print_warning(type, dquot)) {
-			sprintf(quotamessage, "%s: write failed, %s disk quota exceeded too long.\r\n",
+	    dquot->dq_btime && CURRENT_TIME >= dquot->dq_btime &&
+            !ignore_hardlimit(dquot, initiator)) {
+                if (warn && need_print_warning(type, initiator, dquot)) {
+			sprintf(quotamessage, "%s: write failed, %s disk quota exceeded too long.\n",
 		        	dquot->dq_mnt->mnt_dirname, quotatypes[type]);
-			tty_write_message(current->tty, quotamessage);
+			tty_write_message(tty, quotamessage);
 		}
 		return(NO_QUOTA);
 	}
+
 	if (dquot->dq_bsoftlimit &&
 	   (dquot->dq_curblocks + blocks) > dquot->dq_bsoftlimit &&
-	    dquot->dq_btime == 0 && 
-	    !capable(CAP_SYS_RESOURCE)) {
-                if (need_print_warning(type, dquot)) {
-			sprintf(quotamessage, "%s: warning, %s disk quota exceeded\r\n",
+	    dquot->dq_btime == 0) {
+                if (warn && need_print_warning(type, initiator, dquot)) {
+			sprintf(quotamessage, "%s: warning, %s disk quota exceeded\n",
 		        	dquot->dq_mnt->mnt_dirname, quotatypes[type]);
-			tty_write_message(current->tty, quotamessage);
-		}
-		dquot->dq_btime = CURRENT_TIME + dquot->dq_mnt->mnt_bexp[type];
-	}
-	return(QUOTA_OK);
-}
-
-static void dqput(struct dquot *dquot)
-{
-	if (!dquot)
-		return;
-	/*
-	 * If the dq_mnt pointer isn't initialized this entry needs no
-	 * checking and doesn't need to be written. It just an empty
-	 * dquot that is put back into the freelist.
-	 */
-	if (dquot->dq_mnt != (struct vfsmount *)NULL) {
-		dqstats.drops++;
-		wait_on_dquot(dquot);
-		if (!dquot->dq_count) {
-			printk("VFS: dqput: trying to free free dquot\n");
-			printk("VFS: device %s, dquot of %s %d\n", kdevname(dquot->dq_dev),
-			       quotatypes[dquot->dq_type], dquot->dq_id);
-			return;
+			tty_write_message(tty, quotamessage);
 		}
-repeat:
-		if (dquot->dq_count > 1) {
-			dquot->dq_count--;
-			return;
-		}
-		wake_up(&dquot_wait);
-		if (dquot->dq_flags & DQ_MOD) {
-			write_dquot(dquot);	/* we can sleep - so do again */
-			wait_on_dquot(dquot);
-			goto repeat;
-		}
-	}
-	if (dquot->dq_count) {
-		dquot->dq_count--;
-		nr_free_dquots++;
+		dquot->dq_btime = CURRENT_TIME + dquot->dq_mnt->mnt_dquot.block_expire[type];
 	}
-	return;
-}
-
-static struct dquot *get_empty_dquot(void)
-{
-	struct dquot *dquot, *best;
-	int cnt;
-
-	if (nr_dquots < NR_DQUOTS && nr_free_dquots < (nr_dquots >> 2))
-		grow_dquots();
 
-repeat:
-	dquot = first_dquot;
-	best = NODQUOT;
-	for (cnt = 0; cnt < nr_dquots; dquot = dquot->dq_next, cnt++) {
-		if (!dquot->dq_count) {
-			if (!best)
-				best = dquot;
-			if (!(dquot->dq_flags & DQ_MOD) && !(dquot->dq_flags & DQ_LOCKED)) {
-				best = dquot;
-				break;
-			}
-		}
-	}
-	if (!best || best->dq_flags & DQ_MOD || best->dq_flags & DQ_LOCKED)
-		if (nr_dquots < NR_DQUOTS) {
-			grow_dquots();
-			goto repeat;
-		}
-	dquot = best;
-	if (!dquot) {
-		printk("VFS: No free dquots - contact mvw@mcs.ow.org\n");
-		sleep_on(&dquot_wait);
-		goto repeat;
-	}
-	if (dquot->dq_flags & DQ_LOCKED) {
-		wait_on_dquot(dquot);
-		goto repeat;
-	}
-	if (dquot->dq_flags & DQ_MOD) {
-		write_dquot(dquot);
-		goto repeat;
-	}
-	if (dquot->dq_count)
-		goto repeat;
-	clear_dquot(dquot);
-	dquot->dq_count = 1;
-	nr_free_dquots--;
-	if (nr_free_dquots < 0) {
-		printk ("VFS: get_empty_dquot: bad free dquot count.\n");
-		nr_free_dquots = 0;
-	}
-	return(dquot);
-}
-
-static struct dquot *dqget(kdev_t dev, unsigned int id, short type)
-{
-	struct dquot *dquot, *empty;
-	struct vfsmount *vfsmnt;
-
-	if ((vfsmnt = lookup_vfsmnt(dev)) == (struct vfsmount *)NULL ||
-	    (vfsmnt->mnt_quotas[type] == (struct file *)0))
-		return(NODQUOT);
-	dqstats.lookups++;
-	empty = get_empty_dquot();
-repeat:
-	dquot = *(hash(dev, id, type));
-	while (dquot) {
-		if (dquot->dq_dev != dev || dquot->dq_id != id ||
-		    dquot->dq_type != type) {
-			dquot = dquot->dq_hash_next;
-			continue;
-		}
-		wait_on_dquot(dquot);
-		if (dquot->dq_dev != dev || dquot->dq_id != id ||
-		    dquot->dq_type != type)
-			goto repeat;
-		if (!dquot->dq_count)
-			nr_free_dquots--;
-		dquot->dq_count++;
-		if (empty)
-			dqput(empty);
-		dqstats.cache_hits++;
-		return(dquot);
-	}
-	if (!empty)
-		return(NODQUOT);
-	dquot = empty;
-	dquot->dq_id = id;
-	dquot->dq_type = type;
-	dquot->dq_dev = dev;
-	dquot->dq_mnt = vfsmnt;
-	put_last_free(dquot);
-	insert_dquot_hash(dquot);
-	read_dquot(dquot);
-	return(dquot);
+	return(QUOTA_OK);
 }
 
 /*
@@ -590,54 +717,57 @@
  */ 
 static int set_dqblk(kdev_t dev, int id, short type, int flags, struct dqblk *dqblk)
 {
+	int error;
 	struct dquot *dquot;
 	struct dqblk dq_dqblk;
 
 	if (dqblk == (struct dqblk *)NULL)
 		return(-EFAULT);
 
-	if (flags & QUOTA_SYSCALL) {
-		if (copy_from_user(&dq_dqblk, dqblk, sizeof(struct dqblk)))
-			return -EFAULT;	
-	} else {
-		memcpy(&dq_dqblk, dqblk, sizeof(struct dqblk));
-	}
+	if (flags & QUOTA_SYSCALL)
+		if ((error = copy_from_user((caddr_t)&dq_dqblk, (caddr_t)dqblk, sizeof(struct dqblk))) != 0)
+			return(error);
+	else
+		memcpy((caddr_t)&dq_dqblk, (caddr_t)dqblk, sizeof(struct dqblk));
+
 	if ((dquot = dqget(dev, id, type)) != NODQUOT) {
 		lock_dquot(dquot);
-		if (id > 0 && (flags & (SET_QUOTA|SET_QLIMIT))) {
+
+		if (id > 0 && ((flags & SET_QUOTA) || (flags & SET_QLIMIT))) {
 			dquot->dq_bhardlimit = dq_dqblk.dqb_bhardlimit;
 			dquot->dq_bsoftlimit = dq_dqblk.dqb_bsoftlimit;
 			dquot->dq_ihardlimit = dq_dqblk.dqb_ihardlimit;
 			dquot->dq_isoftlimit = dq_dqblk.dqb_isoftlimit;
 		}
-		if (flags & (SET_QUOTA|SET_USE)) {
+
+		if ((flags & SET_QUOTA) || (flags & SET_USE)) {
 			if (dquot->dq_isoftlimit &&
 			    dquot->dq_curinodes < dquot->dq_isoftlimit &&
 			    dq_dqblk.dqb_curinodes >= dquot->dq_isoftlimit)
-				dquot->dq_itime = CURRENT_TIME + dquot->dq_mnt->mnt_iexp[type];
+				dquot->dq_itime = CURRENT_TIME + dquot->dq_mnt->mnt_dquot.inode_expire[type];
 			dquot->dq_curinodes = dq_dqblk.dqb_curinodes;
 			if (dquot->dq_curinodes < dquot->dq_isoftlimit)
 				dquot->dq_flags &= ~DQ_INODES;
 			if (dquot->dq_bsoftlimit &&
 			    dquot->dq_curblocks < dquot->dq_bsoftlimit &&
 			    dq_dqblk.dqb_curblocks >= dquot->dq_bsoftlimit)
-				dquot->dq_btime = CURRENT_TIME + dquot->dq_mnt->mnt_bexp[type];
+				dquot->dq_btime = CURRENT_TIME + dquot->dq_mnt->mnt_dquot.block_expire[type];
 			dquot->dq_curblocks = dq_dqblk.dqb_curblocks;
 			if (dquot->dq_curblocks < dquot->dq_bsoftlimit)
 				dquot->dq_flags &= ~DQ_BLKS;
 		}
+
 		if (id == 0) {
-			/* 
-			 * Change in expiretimes, change them in dq_mnt.
-			 */
-			dquot->dq_mnt->mnt_bexp[type] = dquot->dq_btime = dq_dqblk.dqb_btime;
-			dquot->dq_mnt->mnt_iexp[type] = dquot->dq_itime = dq_dqblk.dqb_itime;
+			dquot->dq_mnt->mnt_dquot.block_expire[type] = dquot->dq_btime = dq_dqblk.dqb_btime;
+			dquot->dq_mnt->mnt_dquot.inode_expire[type] = dquot->dq_itime = dq_dqblk.dqb_itime;
 		}
+
 		if (dq_dqblk.dqb_bhardlimit == 0 && dq_dqblk.dqb_bsoftlimit == 0 &&
 		    dq_dqblk.dqb_ihardlimit == 0 && dq_dqblk.dqb_isoftlimit == 0)
 			dquot->dq_flags |= DQ_FAKE;
 		else
 			dquot->dq_flags &= ~DQ_FAKE;
+
 		dquot->dq_flags |= DQ_MOD;
 		unlock_dquot(dquot);
 		dqput(dquot);
@@ -650,14 +780,14 @@
 	struct dquot *dquot;
 	int error;
 
-	if (has_quota_enabled(dev, type)) {
+	if (dev_has_quota_enabled(dev, type)) {
 		if (dqblk == (struct dqblk *)NULL)
 			return(-EFAULT);
 
 		if ((dquot = dqget(dev, id, type)) != NODQUOT) {
-			error = copy_to_user(dqblk, (char *)&dquot->dq_dqb, sizeof(struct dqblk));
+			error = copy_to_user((caddr_t)dqblk, (caddr_t)&dquot->dq_dqb, sizeof(struct dqblk));
 			dqput(dquot);
-			return error ? -EFAULT : 0;
+			return(error);
 		}
 	}
 	return(-ESRCH);
@@ -667,25 +797,67 @@
 {
 	dqstats.allocated_dquots = nr_dquots;
 	dqstats.free_dquots = nr_free_dquots;
-	return copy_to_user(addr, (caddr_t)&dqstats, sizeof(struct dqstats)) 
-		? -EFAULT : 0; 
+	return(copy_to_user(addr, (caddr_t)&dqstats, sizeof(struct dqstats)));
+}
+
+static int quota_root_squash(kdev_t dev, short type, int *addr)
+{
+	struct vfsmount *vfsmnt;
+	int new_value, error;
+
+	if ((vfsmnt = lookup_vfsmnt(dev)) == (struct vfsmount *)NULL)
+		return(-ENODEV);
+
+	if ((error = copy_from_user((caddr_t)&new_value, (caddr_t)addr, sizeof(int))) != 0)
+		return(error);
+
+	vfsmnt->mnt_dquot.rsquash[type] = new_value;
+	return(0);
 }
 
 /*
- * Initialize pointer in an inode to the right dquots.
+ * This is a simple algorithm that calculates the size of a file in blocks.
+ * This is only used on filesystems that do not have an i_blocks count.
+ */
+static u_long isize_to_blocks(size_t isize, size_t blksize)
+{
+	u_long blocks;
+	u_long indirect;
+
+	if (!blksize)
+		blksize = BLOCK_SIZE;
+	blocks = (isize / blksize) + ((isize % blksize) ? 1 : 0);
+	if (blocks > 10) {
+		indirect = ((blocks - 11) >> 8) + 1; /* single indirect blocks */
+		if (blocks > (10 + 256)) {
+			indirect += ((blocks - 267) >> 16) + 1; /* double indirect blocks */
+			if (blocks > (10 + 256 + (256 << 8)))
+				indirect++; /* triple indirect blocks */
+		}
+		blocks += indirect;
+	}
+	return(blocks);
+}
+
+/*
+ * Externally referenced functions through dquot_operations in inode.
  */
 void dquot_initialize(struct inode *inode, short type)
 {
+	struct dquot *dquot;
 	unsigned int id = 0;
 	short cnt;
-	struct dquot *tmp;
 
-	if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) {
+	if (S_ISREG(inode->i_mode) ||
+            S_ISDIR(inode->i_mode) ||
+            S_ISLNK(inode->i_mode)) {
 		for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
 			if (type != -1 && cnt != type)
 				continue;
-			if (!has_quota_enabled(inode->i_dev, cnt))
+
+			if (!sb_has_quota_enabled(inode->i_sb, cnt))
 				continue;
+
 			if (inode->i_dquot[cnt] == NODQUOT) {
 				switch (cnt) {
 					case USRQUOTA:
@@ -695,17 +867,13 @@
 						id = inode->i_gid;
 						break;
 				}
-
-				tmp = dqget(inode->i_dev, id, cnt);
-				/* We may sleep in dqget(), so check it again.
-				 * 	Dmitry Gorodchanin 02/11/96
-				 */
+				dquot = dqget(inode->i_dev, id, cnt);
 				if (inode->i_dquot[cnt] != NODQUOT) {
-					dqput(tmp);
+					dqput(dquot);
 					continue;
 				} 
-				inode->i_dquot[cnt] = tmp;
-				inode->i_flags |= S_WRITE;
+				inode->i_dquot[cnt] = dquot;
+				inode->i_flags |= S_QUOTA;
 			}
 		}
 	}
@@ -713,82 +881,58 @@
 
 void dquot_drop(struct inode *inode)
 {
+	struct dquot *dquot;
 	short cnt;
-	struct dquot * tmp;
 
+	inode->i_flags &= ~S_QUOTA;
 	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
 		if (inode->i_dquot[cnt] == NODQUOT)
 			continue;
-		/* We can sleep at dqput(). So we must do it this way.
-		 * 	Dmitry Gorodchanin 02/11/96
-		 */
-		tmp = inode->i_dquot[cnt];
+		dquot = inode->i_dquot[cnt];
 		inode->i_dquot[cnt] = NODQUOT;
-		dqput(tmp);
-	}
-	inode->i_flags &= ~S_WRITE;
-}
-
-/*
- * This is a simple algorithm that calculates the size of a file in blocks.
- * This is only used on filesystems that do not have an i_blocks count.
- */
-static u_long isize_to_blocks(size_t isize, size_t blksize)
-{
-	u_long blocks;
-	u_long indirect;
-
-	if (!blksize)
-		blksize = BLOCK_SIZE;
-	blocks = (isize / blksize) + ((isize % blksize) ? 1 : 0);
-	if (blocks > 10) {
-		indirect = ((blocks - 11) >> 8) + 1; /* single indirect blocks */
-		if (blocks > (10 + 256)) {
-			indirect += ((blocks - 267) >> 16) + 1; /* double indirect blocks */
-			if (blocks > (10 + 256 + (256 << 8)))
-				indirect++; /* triple indirect blocks */
-		}
-		blocks += indirect;
+		dqput(dquot);
 	}
-	return(blocks);
 }
 
-/*
- * Externally referenced functions through dquot_operations.
- */
-int dquot_alloc_block(const struct inode *inode, unsigned long number)
+int dquot_alloc_block(const struct inode *inode, unsigned long number, uid_t initiator, char warn)
 {
 	unsigned short cnt;
+	struct tty_struct *tty = current->tty;
 
 	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
 		if (inode->i_dquot[cnt] == NODQUOT)
 			continue;
-		if (check_bdq(inode->i_dquot[cnt], cnt, number))
+		if (check_bdq(inode->i_dquot[cnt], cnt, number, initiator, tty, warn))
 			return(NO_QUOTA);
 	}
+
 	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
 		if (inode->i_dquot[cnt] == NODQUOT)
 			continue;
 		dquot_incr_blocks(inode->i_dquot[cnt], number);
 	}
+
 	return(QUOTA_OK);
 }
 
-int dquot_alloc_inode(const struct inode *inode, unsigned long number)
+int dquot_alloc_inode(const struct inode *inode, unsigned long number, uid_t initiator)
 {
 	unsigned short cnt;
+	struct tty_struct *tty = current->tty;
 
 	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
 		if (inode->i_dquot[cnt] == NODQUOT)
 			continue;
-		if (check_idq(inode->i_dquot[cnt], cnt, number))
+		if (check_idq(inode->i_dquot[cnt], cnt, number, initiator, tty))
 			return(NO_QUOTA);
 	}
+
 	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
 		if (inode->i_dquot[cnt] == NODQUOT)
 			continue;
 		dquot_incr_inodes(inode->i_dquot[cnt], number);
 	}
+
 	return(QUOTA_OK);
 }
 
@@ -817,11 +961,12 @@
 /*
  * Transfer the number of inode and blocks from one diskquota to an other.
  */
-int dquot_transfer(struct inode *inode, struct iattr *iattr, char direction)
+int dquot_transfer(struct inode *inode, struct iattr *iattr, char direction, uid_t initiator)
 {
 	unsigned long blocks;
 	struct dquot *transfer_from[MAXQUOTAS];
 	struct dquot *transfer_to[MAXQUOTAS];
+	struct tty_struct *tty = current->tty;
 	short cnt, disc;
 
 	/*
@@ -840,7 +985,7 @@
 		transfer_from[cnt] = NODQUOT;
 		transfer_to[cnt] = NODQUOT;
 
-		if (!has_quota_enabled(inode->i_dev, cnt))
+		if (!sb_has_quota_enabled(inode->i_sb, cnt))
 			continue;
 
 		switch (cnt) {
@@ -858,8 +1003,8 @@
 				break;
 		}
 
-		if (check_idq(transfer_to[cnt], cnt, 1) == NO_QUOTA ||
-		    check_bdq(transfer_to[cnt], cnt, blocks) == NO_QUOTA) {
+		if (check_idq(transfer_to[cnt], cnt, 1, initiator, tty) == NO_QUOTA ||
+		    check_bdq(transfer_to[cnt], cnt, blocks, initiator, tty, 0) == NO_QUOTA) {
 			for (disc = 0; disc <= cnt; disc++) {
 				dqput(transfer_from[disc]);
 				dqput(transfer_to[disc]);
@@ -883,10 +1028,12 @@
 			dquot_decr_inodes(transfer_from[cnt], 1);
 			dquot_decr_blocks(transfer_from[cnt], blocks);
 		}
+
 		if (transfer_to[cnt] != NODQUOT) {
 			dquot_incr_inodes(transfer_to[cnt], 1);
 			dquot_incr_blocks(transfer_to[cnt], blocks);
 		}
+
 		if (inode->i_dquot[cnt] != NODQUOT) {
 			dqput(transfer_from[cnt]);
 			dqput(inode->i_dquot[cnt]);
@@ -896,16 +1043,24 @@
 			dqput(transfer_to[cnt]);
 		}
 	}
+
 	return(QUOTA_OK);
 }
 
-void dquot_init(void)
+
+__initfunc(void dquot_init_hash(void))
 {
-	printk(KERN_NOTICE "VFS: Diskquotas version %s initialized\r\n",
-	       __DQUOT_VERSION__);
-	memset(hash_table, 0, sizeof(hash_table));
+	printk(KERN_NOTICE "VFS: Diskquotas version %s initialized\n", __DQUOT_VERSION__);
+
+	dquot_cachep = kmem_cache_create("dquot", sizeof(struct dquot),
+					 sizeof(unsigned long) * 4,
+					 SLAB_HWCACHE_ALIGN, NULL, NULL);
+
+	if (!dquot_cachep)
+		panic("Cannot create dquot SLAB cache\n");
+
+	memset(dquot_hash, 0, sizeof(dquot_hash));
 	memset((caddr_t)&dqstats, 0, sizeof(dqstats));
-	first_dquot = NODQUOT;
 }
 
 /*
@@ -921,6 +1076,30 @@
 	dquot_transfer
 };
 
+static inline void set_enable_flags(struct vfsmount *vfsmnt, short type)
+{
+	switch (type) {
+		case USRQUOTA:
+			vfsmnt->mnt_dquot.flags |= DQUOT_USR_ENABLED;
+			break;
+		case GRPQUOTA:
+			vfsmnt->mnt_dquot.flags |= DQUOT_GRP_ENABLED;
+			break;
+	}
+}
+
+static inline void reset_enable_flags(struct vfsmount *vfsmnt, short type)
+{
+	switch (type) {
+		case USRQUOTA:
+			vfsmnt->mnt_dquot.flags &= ~DQUOT_USR_ENABLED;
+			break;
+		case GRPQUOTA:
+			vfsmnt->mnt_dquot.flags &= ~DQUOT_GRP_ENABLED;
+			break;
+	}
+}
+
 /*
  * Turn quota off on a device. type == -1 ==> quotaoff for all types (umount)
  */
@@ -929,25 +1108,28 @@
 	struct vfsmount *vfsmnt;
 	short cnt;
 
-	if ( !(vfsmnt = lookup_vfsmnt(dev)) )
-		return -ENODEV;
-
 	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
 		if (type != -1 && cnt != type)
 			continue;
-		if (vfsmnt->mnt_quotas[cnt] == (struct file *)NULL)
-		{
-			if(type == -1)
-				continue;
-			return -ESRCH;
-		}
+
+		if ((vfsmnt = lookup_vfsmnt(dev)) == (struct vfsmount *)NULL ||
+                    is_enabled(vfsmnt, cnt) == 0 ||
+                    vfsmnt->mnt_sb == (struct super_block *)NULL) 
+			continue;
+
 		vfsmnt->mnt_sb->dq_op = (struct dquot_operations *)NULL;
+
 		reset_dquot_ptrs(dev, cnt);
 		invalidate_dquots(dev, cnt);
-		fput(vfsmnt->mnt_quotas[cnt]);
-		vfsmnt->mnt_quotas[cnt] = (struct file *)NULL;
-		vfsmnt->mnt_iexp[cnt] = vfsmnt->mnt_bexp[cnt] = (time_t)NULL;
+
+		fput(vfsmnt->mnt_dquot.files[cnt]);
+
+		reset_enable_flags(vfsmnt, cnt);
+		vfsmnt->mnt_dquot.files[cnt] = (struct file *)NULL;
+		vfsmnt->mnt_dquot.inode_expire[cnt] = 0;
+		vfsmnt->mnt_dquot.block_expire[cnt] = 0;
 	}
+
 	return(0);
 }
 
@@ -962,11 +1144,11 @@
 	int error;
 
 	vfsmnt = lookup_vfsmnt(dev);
-	if (vfsmnt == NULL)
+	if (vfsmnt == (struct vfsmount *)NULL)
 		return -ENODEV;
 
-	if (vfsmnt->mnt_quotas[type] != NULL)
-		return -EBUSY;
+	if (is_enabled(vfsmnt, type))
+		return(-EBUSY);
 
 	tmp = getname(path);
 	error = PTR_ERR(tmp);
@@ -986,8 +1168,13 @@
 		return -EACCES;
 	}
 
+	if (inode->i_size == 0 || (inode->i_size % sizeof(struct dqblk)) != 0) {
+		iput(inode);
+		return(-EINVAL);
+	}
+
 	filp = get_empty_filp();
-	if (filp != NULL) {
+	if (filp != (struct file *)NULL) {
 		filp->f_mode = (O_RDWR + 1) & O_ACCMODE;
 		filp->f_flags = O_RDWR;
 		filp->f_dentry = dentry;
@@ -1000,14 +1187,18 @@
 				if (filp->f_op && filp->f_op->open)
 					error = filp->f_op->open(inode, filp);
 				if (!error) {
-					vfsmnt->mnt_quotas[type] = filp;
+					set_enable_flags(vfsmnt, type);
+					vfsmnt->mnt_dquot.files[type] = filp;
+
 					dquot = dqget(dev, 0, type);
-					vfsmnt->mnt_iexp[type] = (dquot) ? dquot->dq_itime : MAX_IQ_TIME;
-					vfsmnt->mnt_bexp[type] = (dquot) ? dquot->dq_btime : MAX_DQ_TIME;
+					vfsmnt->mnt_dquot.inode_expire[type] = (dquot) ? dquot->dq_itime : MAX_IQ_TIME;
+					vfsmnt->mnt_dquot.block_expire[type] = (dquot) ? dquot->dq_btime : MAX_DQ_TIME;
 					dqput(dquot);
+
 					vfsmnt->mnt_sb->dq_op = &dquot_operations;
 					add_dquot_ref(dev, type);
-					return 0;
+
+					return(0);
 				}
 				put_write_access(inode);
 			}
@@ -1016,8 +1207,10 @@
 		put_filp(filp);
 	} else
 		error = -EMFILE;
+
 	dput(dentry);
-	return error;
+
+	return(error);
 }
 
 /*
@@ -1045,17 +1238,17 @@
 			break;
 		case Q_GETQUOTA:
 			if (((type == USRQUOTA && current->uid != id) ||
-			     (type == GRPQUOTA && in_group_p(id))) && 
-			    !capable(CAP_SYS_ADMIN))
+			     (type == GRPQUOTA && current->gid != id)) &&
+			    !capable(CAP_SYS_RESOURCE))
 				goto out;
 			break;
 		default:
-			if (!capable(CAP_SYS_ADMIN))
+			if (!capable(CAP_SYS_RESOURCE))
 				goto out;
 	}
 
 	ret = -EINVAL;
-	dev = NODEV;
+	dev = 0;
 	if (special != NULL || (cmds != Q_SYNC && cmds != Q_GETSTATS)) {
 		mode_t mode;
 		struct dentry * dentry;
@@ -1098,12 +1291,17 @@
 			goto out;
 		case Q_GETSTATS:
 			ret = get_stats(addr);
+			goto out;
+		case Q_RSQUASH:
+			ret = quota_root_squash(dev, type, (int *) addr);
+			goto out;
 		default:
 			goto out;
 	}
 
 	flags |= QUOTA_SYSCALL;
-	if (has_quota_enabled(dev, type))
+
+	if (dev_has_quota_enabled(dev, type))
 		ret = set_dqblk(dev, id, type, flags, (struct dqblk *) addr);
 	else
 		ret = -ESRCH;

FUNET's LINUX-ADM group, linux-adm@nic.funet.fi
TCL-scripts by Sam Shen, slshen@lbl.gov