quota.c 38.5 KB
Newer Older
D
David Teigland 已提交
1 2
/*
 * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
3
 * Copyright (C) 2004-2007 Red Hat, Inc.  All rights reserved.
D
David Teigland 已提交
4 5 6
 *
 * This copyrighted material is made available to anyone wishing to use,
 * modify, copy, or redistribute it subject to the terms and conditions
7
 * of the GNU General Public License version 2.
D
David Teigland 已提交
8 9 10 11 12 13 14 15 16 17
 */

/*
 * Quota change tags are associated with each transaction that allocates or
 * deallocates space.  Those changes are accumulated locally to each node (in a
 * per-node file) and then are periodically synced to the quota file.  This
 * avoids the bottleneck of constantly touching the quota file, but introduces
 * fuzziness in the current usage value of IDs that are being used on different
 * nodes in the cluster simultaneously.  So, it is possible for a user on
 * multiple nodes to overrun their quota, but that overrun is controlable.
18
 * Since quota tags are part of transactions, there is no need for a quota check
D
David Teigland 已提交
19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
 * program to be run on node crashes or anything like that.
 *
 * There are couple of knobs that let the administrator manage the quota
 * fuzziness.  "quota_quantum" sets the maximum time a quota change can be
 * sitting on one node before being synced to the quota file.  (The default is
 * 60 seconds.)  Another knob, "quota_scale" controls how quickly the frequency
 * of quota file syncs increases as the user moves closer to their limit.  The
 * more frequent the syncs, the more accurate the quota enforcement, but that
 * means that there is more contention between the nodes for the quota file.
 * The default value is one.  This sets the maximum theoretical quota overrun
 * (with infinite node with infinite bandwidth) to twice the user's limit.  (In
 * practice, the maximum overrun you see should be much less.)  A "quota_scale"
 * number greater than one makes quota syncs more frequent and reduces the
 * maximum overrun.  Numbers less than one (but greater than zero) make quota
 * syncs less frequent.
 *
 * GFS quotas also use per-ID Lock Value Blocks (LVBs) to cache the contents of
 * the quota file, so it is not being constantly read.
 */

#include <linux/sched.h>
#include <linux/slab.h>
41
#include <linux/mm.h>
D
David Teigland 已提交
42 43 44 45
#include <linux/spinlock.h>
#include <linux/completion.h>
#include <linux/buffer_head.h>
#include <linux/sort.h>
46
#include <linux/fs.h>
47
#include <linux/bio.h>
48
#include <linux/gfs2_ondisk.h>
49 50
#include <linux/kthread.h>
#include <linux/freezer.h>
S
Steven Whitehouse 已提交
51
#include <linux/quota.h>
52
#include <linux/dqblk_xfs.h>
D
David Teigland 已提交
53 54

#include "gfs2.h"
55
#include "incore.h"
D
David Teigland 已提交
56 57 58 59 60 61 62 63 64
#include "bmap.h"
#include "glock.h"
#include "glops.h"
#include "log.h"
#include "meta_io.h"
#include "quota.h"
#include "rgrp.h"
#include "super.h"
#include "trans.h"
65
#include "inode.h"
66
#include "util.h"
D
David Teigland 已提交
67 68 69 70

#define QUOTA_USER 1
#define QUOTA_GROUP 0

71 72 73 74 75 76
struct gfs2_quota_change_host {
	u64 qc_change;
	u32 qc_flags; /* GFS2_QCF_... */
	u32 qc_id;
};

77 78
static LIST_HEAD(qd_lru_list);
static atomic_t qd_lru_count = ATOMIC_INIT(0);
X
Xu Gang 已提交
79
static DEFINE_SPINLOCK(qd_lru_lock);
80

81
int gfs2_shrink_qd_memory(struct shrinker *shrink, struct shrink_control *sc)
82 83 84
{
	struct gfs2_quota_data *qd;
	struct gfs2_sbd *sdp;
85
	int nr_to_scan = sc->nr_to_scan;
86

87
	if (nr_to_scan == 0)
88 89
		goto out;

90
	if (!(sc->gfp_mask & __GFP_FS))
91 92 93
		return -1;

	spin_lock(&qd_lru_lock);
94
	while (nr_to_scan && !list_empty(&qd_lru_list)) {
95 96 97 98 99 100 101 102 103 104 105
		qd = list_entry(qd_lru_list.next,
				struct gfs2_quota_data, qd_reclaim);
		sdp = qd->qd_gl->gl_sbd;

		/* Free from the filesystem-specific list */
		list_del(&qd->qd_list);

		gfs2_assert_warn(sdp, !qd->qd_change);
		gfs2_assert_warn(sdp, !qd->qd_slot_count);
		gfs2_assert_warn(sdp, !qd->qd_bh_count);

106
		gfs2_glock_put(qd->qd_gl);
107 108 109 110 111 112 113 114
		atomic_dec(&sdp->sd_quota_count);

		/* Delete it from the common reclaim list */
		list_del_init(&qd->qd_reclaim);
		atomic_dec(&qd_lru_count);
		spin_unlock(&qd_lru_lock);
		kmem_cache_free(gfs2_quotad_cachep, qd);
		spin_lock(&qd_lru_lock);
115
		nr_to_scan--;
116 117 118 119 120 121 122
	}
	spin_unlock(&qd_lru_lock);

out:
	return (atomic_read(&qd_lru_count) * sysctl_vfs_cache_pressure) / 100;
}

123
static u64 qd2offset(struct gfs2_quota_data *qd)
D
David Teigland 已提交
124
{
125
	u64 offset;
D
David Teigland 已提交
126

127
	offset = 2 * (u64)qd->qd_id + !test_bit(QDF_USER, &qd->qd_flags);
D
David Teigland 已提交
128 129 130 131 132
	offset *= sizeof(struct gfs2_quota);

	return offset;
}

133
static int qd_alloc(struct gfs2_sbd *sdp, int user, u32 id,
D
David Teigland 已提交
134 135 136 137 138
		    struct gfs2_quota_data **qdp)
{
	struct gfs2_quota_data *qd;
	int error;

139
	qd = kmem_cache_zalloc(gfs2_quotad_cachep, GFP_NOFS);
D
David Teigland 已提交
140 141 142
	if (!qd)
		return -ENOMEM;

143
	atomic_set(&qd->qd_count, 1);
D
David Teigland 已提交
144 145 146 147
	qd->qd_id = id;
	if (user)
		set_bit(QDF_USER, &qd->qd_flags);
	qd->qd_slot = -1;
148
	INIT_LIST_HEAD(&qd->qd_reclaim);
D
David Teigland 已提交
149

150
	error = gfs2_glock_get(sdp, 2 * (u64)id + !user,
D
David Teigland 已提交
151 152 153 154 155 156 157 158
			      &gfs2_quota_glops, CREATE, &qd->qd_gl);
	if (error)
		goto fail;

	*qdp = qd;

	return 0;

159
fail:
160
	kmem_cache_free(gfs2_quotad_cachep, qd);
D
David Teigland 已提交
161 162 163
	return error;
}

164
static int qd_get(struct gfs2_sbd *sdp, int user, u32 id,
D
David Teigland 已提交
165 166 167 168 169 170 171 172 173
		  struct gfs2_quota_data **qdp)
{
	struct gfs2_quota_data *qd = NULL, *new_qd = NULL;
	int error, found;

	*qdp = NULL;

	for (;;) {
		found = 0;
174
		spin_lock(&qd_lru_lock);
D
David Teigland 已提交
175 176 177
		list_for_each_entry(qd, &sdp->sd_quota_list, qd_list) {
			if (qd->qd_id == id &&
			    !test_bit(QDF_USER, &qd->qd_flags) == !user) {
178 179 180 181 182 183 184
				if (!atomic_read(&qd->qd_count) &&
				    !list_empty(&qd->qd_reclaim)) {
					/* Remove it from reclaim list */
					list_del_init(&qd->qd_reclaim);
					atomic_dec(&qd_lru_count);
				}
				atomic_inc(&qd->qd_count);
D
David Teigland 已提交
185 186 187 188 189 190 191 192 193 194 195 196 197 198 199
				found = 1;
				break;
			}
		}

		if (!found)
			qd = NULL;

		if (!qd && new_qd) {
			qd = new_qd;
			list_add(&qd->qd_list, &sdp->sd_quota_list);
			atomic_inc(&sdp->sd_quota_count);
			new_qd = NULL;
		}

200
		spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
201

202
		if (qd) {
D
David Teigland 已提交
203
			if (new_qd) {
204
				gfs2_glock_put(new_qd->qd_gl);
205
				kmem_cache_free(gfs2_quotad_cachep, new_qd);
D
David Teigland 已提交
206 207 208 209 210 211 212 213 214 215 216 217 218 219
			}
			*qdp = qd;
			return 0;
		}

		error = qd_alloc(sdp, user, id, &new_qd);
		if (error)
			return error;
	}
}

static void qd_hold(struct gfs2_quota_data *qd)
{
	struct gfs2_sbd *sdp = qd->qd_gl->gl_sbd;
220 221
	gfs2_assert(sdp, atomic_read(&qd->qd_count));
	atomic_inc(&qd->qd_count);
D
David Teigland 已提交
222 223 224 225
}

static void qd_put(struct gfs2_quota_data *qd)
{
226 227 228 229 230 231
	if (atomic_dec_and_lock(&qd->qd_count, &qd_lru_lock)) {
		/* Add to the reclaim list */
		list_add_tail(&qd->qd_reclaim, &qd_lru_list);
		atomic_inc(&qd_lru_count);
		spin_unlock(&qd_lru_lock);
	}
D
David Teigland 已提交
232 233 234 235 236 237 238 239
}

static int slot_get(struct gfs2_quota_data *qd)
{
	struct gfs2_sbd *sdp = qd->qd_gl->gl_sbd;
	unsigned int c, o = 0, b;
	unsigned char byte = 0;

240
	spin_lock(&qd_lru_lock);
D
David Teigland 已提交
241 242

	if (qd->qd_slot_count++) {
243
		spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
244 245 246 247 248 249 250 251 252 253 254 255
		return 0;
	}

	for (c = 0; c < sdp->sd_quota_chunks; c++)
		for (o = 0; o < PAGE_SIZE; o++) {
			byte = sdp->sd_quota_bitmap[c][o];
			if (byte != 0xFF)
				goto found;
		}

	goto fail;

256
found:
D
David Teigland 已提交
257 258 259 260 261 262 263 264 265 266
	for (b = 0; b < 8; b++)
		if (!(byte & (1 << b)))
			break;
	qd->qd_slot = c * (8 * PAGE_SIZE) + o * 8 + b;

	if (qd->qd_slot >= sdp->sd_quota_slots)
		goto fail;

	sdp->sd_quota_bitmap[c][o] |= 1 << b;

267
	spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
268 269 270

	return 0;

271
fail:
D
David Teigland 已提交
272
	qd->qd_slot_count--;
273
	spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
274 275 276 277 278 279 280
	return -ENOSPC;
}

static void slot_hold(struct gfs2_quota_data *qd)
{
	struct gfs2_sbd *sdp = qd->qd_gl->gl_sbd;

281
	spin_lock(&qd_lru_lock);
D
David Teigland 已提交
282 283
	gfs2_assert(sdp, qd->qd_slot_count);
	qd->qd_slot_count++;
284
	spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
285 286 287 288 289 290
}

static void slot_put(struct gfs2_quota_data *qd)
{
	struct gfs2_sbd *sdp = qd->qd_gl->gl_sbd;

291
	spin_lock(&qd_lru_lock);
D
David Teigland 已提交
292 293 294 295 296
	gfs2_assert(sdp, qd->qd_slot_count);
	if (!--qd->qd_slot_count) {
		gfs2_icbit_munge(sdp, sdp->sd_quota_bitmap, qd->qd_slot, 0);
		qd->qd_slot = -1;
	}
297
	spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
298 299 300 301 302
}

static int bh_get(struct gfs2_quota_data *qd)
{
	struct gfs2_sbd *sdp = qd->qd_gl->gl_sbd;
303
	struct gfs2_inode *ip = GFS2_I(sdp->sd_qc_inode);
D
David Teigland 已提交
304 305 306
	unsigned int block, offset;
	struct buffer_head *bh;
	int error;
307
	struct buffer_head bh_map = { .b_state = 0, .b_blocknr = 0 };
D
David Teigland 已提交
308

309
	mutex_lock(&sdp->sd_quota_mutex);
D
David Teigland 已提交
310 311

	if (qd->qd_bh_count++) {
312
		mutex_unlock(&sdp->sd_quota_mutex);
D
David Teigland 已提交
313 314 315 316
		return 0;
	}

	block = qd->qd_slot / sdp->sd_qc_per_block;
317
	offset = qd->qd_slot % sdp->sd_qc_per_block;
D
David Teigland 已提交
318

319
	bh_map.b_size = 1 << ip->i_inode.i_blkbits;
320
	error = gfs2_block_map(&ip->i_inode, block, &bh_map, 0);
D
David Teigland 已提交
321 322
	if (error)
		goto fail;
S
Steven Whitehouse 已提交
323
	error = gfs2_meta_read(ip->i_gl, bh_map.b_blocknr, DIO_WAIT, &bh);
D
David Teigland 已提交
324 325 326 327 328 329 330 331 332 333 334
	if (error)
		goto fail;
	error = -EIO;
	if (gfs2_metatype_check(sdp, bh, GFS2_METATYPE_QC))
		goto fail_brelse;

	qd->qd_bh = bh;
	qd->qd_bh_qc = (struct gfs2_quota_change *)
		(bh->b_data + sizeof(struct gfs2_meta_header) +
		 offset * sizeof(struct gfs2_quota_change));

J
Josef Whiter 已提交
335
	mutex_unlock(&sdp->sd_quota_mutex);
D
David Teigland 已提交
336 337 338

	return 0;

339
fail_brelse:
D
David Teigland 已提交
340
	brelse(bh);
341
fail:
D
David Teigland 已提交
342
	qd->qd_bh_count--;
343
	mutex_unlock(&sdp->sd_quota_mutex);
D
David Teigland 已提交
344 345 346 347 348 349 350
	return error;
}

static void bh_put(struct gfs2_quota_data *qd)
{
	struct gfs2_sbd *sdp = qd->qd_gl->gl_sbd;

351
	mutex_lock(&sdp->sd_quota_mutex);
D
David Teigland 已提交
352 353 354 355 356 357
	gfs2_assert(sdp, qd->qd_bh_count);
	if (!--qd->qd_bh_count) {
		brelse(qd->qd_bh);
		qd->qd_bh = NULL;
		qd->qd_bh_qc = NULL;
	}
358
	mutex_unlock(&sdp->sd_quota_mutex);
D
David Teigland 已提交
359 360 361 362 363 364 365 366 367 368 369 370 371
}

static int qd_fish(struct gfs2_sbd *sdp, struct gfs2_quota_data **qdp)
{
	struct gfs2_quota_data *qd = NULL;
	int error;
	int found = 0;

	*qdp = NULL;

	if (sdp->sd_vfs->s_flags & MS_RDONLY)
		return 0;

372
	spin_lock(&qd_lru_lock);
D
David Teigland 已提交
373 374 375 376 377 378 379 380 381 382

	list_for_each_entry(qd, &sdp->sd_quota_list, qd_list) {
		if (test_bit(QDF_LOCKED, &qd->qd_flags) ||
		    !test_bit(QDF_CHANGE, &qd->qd_flags) ||
		    qd->qd_sync_gen >= sdp->sd_quota_sync_gen)
			continue;

		list_move_tail(&qd->qd_list, &sdp->sd_quota_list);

		set_bit(QDF_LOCKED, &qd->qd_flags);
383 384
		gfs2_assert_warn(sdp, atomic_read(&qd->qd_count));
		atomic_inc(&qd->qd_count);
D
David Teigland 已提交
385 386 387 388 389 390 391 392 393 394 395
		qd->qd_change_sync = qd->qd_change;
		gfs2_assert_warn(sdp, qd->qd_slot_count);
		qd->qd_slot_count++;
		found = 1;

		break;
	}

	if (!found)
		qd = NULL;

396
	spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420

	if (qd) {
		gfs2_assert_warn(sdp, qd->qd_change_sync);
		error = bh_get(qd);
		if (error) {
			clear_bit(QDF_LOCKED, &qd->qd_flags);
			slot_put(qd);
			qd_put(qd);
			return error;
		}
	}

	*qdp = qd;

	return 0;
}

static int qd_trylock(struct gfs2_quota_data *qd)
{
	struct gfs2_sbd *sdp = qd->qd_gl->gl_sbd;

	if (sdp->sd_vfs->s_flags & MS_RDONLY)
		return 0;

421
	spin_lock(&qd_lru_lock);
D
David Teigland 已提交
422 423 424

	if (test_bit(QDF_LOCKED, &qd->qd_flags) ||
	    !test_bit(QDF_CHANGE, &qd->qd_flags)) {
425
		spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
426 427 428 429 430 431
		return 0;
	}

	list_move_tail(&qd->qd_list, &sdp->sd_quota_list);

	set_bit(QDF_LOCKED, &qd->qd_flags);
432 433
	gfs2_assert_warn(sdp, atomic_read(&qd->qd_count));
	atomic_inc(&qd->qd_count);
D
David Teigland 已提交
434 435 436 437
	qd->qd_change_sync = qd->qd_change;
	gfs2_assert_warn(sdp, qd->qd_slot_count);
	qd->qd_slot_count++;

438
	spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
439 440 441 442 443 444 445 446 447 448 449 450 451 452

	gfs2_assert_warn(sdp, qd->qd_change_sync);
	if (bh_get(qd)) {
		clear_bit(QDF_LOCKED, &qd->qd_flags);
		slot_put(qd);
		qd_put(qd);
		return 0;
	}

	return 1;
}

static void qd_unlock(struct gfs2_quota_data *qd)
{
453 454
	gfs2_assert_warn(qd->qd_gl->gl_sbd,
			 test_bit(QDF_LOCKED, &qd->qd_flags));
D
David Teigland 已提交
455 456 457 458 459 460
	clear_bit(QDF_LOCKED, &qd->qd_flags);
	bh_put(qd);
	slot_put(qd);
	qd_put(qd);
}

461
static int qdsb_get(struct gfs2_sbd *sdp, int user, u32 id,
D
David Teigland 已提交
462 463 464 465
		    struct gfs2_quota_data **qdp)
{
	int error;

466
	error = qd_get(sdp, user, id, qdp);
D
David Teigland 已提交
467 468 469 470 471 472 473 474 475 476 477 478 479
	if (error)
		return error;

	error = slot_get(*qdp);
	if (error)
		goto fail;

	error = bh_get(*qdp);
	if (error)
		goto fail_slot;

	return 0;

480
fail_slot:
D
David Teigland 已提交
481
	slot_put(*qdp);
482
fail:
D
David Teigland 已提交
483 484 485 486 487 488 489 490 491 492 493
	qd_put(*qdp);
	return error;
}

static void qdsb_put(struct gfs2_quota_data *qd)
{
	bh_put(qd);
	slot_put(qd);
	qd_put(qd);
}

494
int gfs2_quota_hold(struct gfs2_inode *ip, u32 uid, u32 gid)
D
David Teigland 已提交
495
{
496
	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
497
	struct gfs2_quota_data **qd;
D
David Teigland 已提交
498 499
	int error;

500 501 502 503 504
	if (ip->i_res == NULL) {
		error = gfs2_rs_alloc(ip);
		if (error)
			return error;
	}
505 506 507 508

	qd = ip->i_res->rs_qa_qd;

	if (gfs2_assert_warn(sdp, !ip->i_res->rs_qa_qd_num) ||
D
David Teigland 已提交
509 510 511 512 513 514
	    gfs2_assert_warn(sdp, !test_bit(GIF_QD_LOCKED, &ip->i_flags)))
		return -EIO;

	if (sdp->sd_args.ar_quota == GFS2_QUOTA_OFF)
		return 0;

515
	error = qdsb_get(sdp, QUOTA_USER, ip->i_inode.i_uid, qd);
D
David Teigland 已提交
516 517
	if (error)
		goto out;
518
	ip->i_res->rs_qa_qd_num++;
D
David Teigland 已提交
519 520
	qd++;

521
	error = qdsb_get(sdp, QUOTA_GROUP, ip->i_inode.i_gid, qd);
D
David Teigland 已提交
522 523
	if (error)
		goto out;
524
	ip->i_res->rs_qa_qd_num++;
D
David Teigland 已提交
525 526
	qd++;

527
	if (uid != NO_QUOTA_CHANGE && uid != ip->i_inode.i_uid) {
528
		error = qdsb_get(sdp, QUOTA_USER, uid, qd);
D
David Teigland 已提交
529 530
		if (error)
			goto out;
531
		ip->i_res->rs_qa_qd_num++;
D
David Teigland 已提交
532 533 534
		qd++;
	}

535
	if (gid != NO_QUOTA_CHANGE && gid != ip->i_inode.i_gid) {
536
		error = qdsb_get(sdp, QUOTA_GROUP, gid, qd);
D
David Teigland 已提交
537 538
		if (error)
			goto out;
539
		ip->i_res->rs_qa_qd_num++;
D
David Teigland 已提交
540 541 542
		qd++;
	}

543
out:
D
David Teigland 已提交
544 545 546 547 548 549 550
	if (error)
		gfs2_quota_unhold(ip);
	return error;
}

void gfs2_quota_unhold(struct gfs2_inode *ip)
{
551
	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
D
David Teigland 已提交
552 553
	unsigned int x;

554 555
	if (ip->i_res == NULL)
		return;
D
David Teigland 已提交
556 557
	gfs2_assert_warn(sdp, !test_bit(GIF_QD_LOCKED, &ip->i_flags));

558 559 560
	for (x = 0; x < ip->i_res->rs_qa_qd_num; x++) {
		qdsb_put(ip->i_res->rs_qa_qd[x]);
		ip->i_res->rs_qa_qd[x] = NULL;
D
David Teigland 已提交
561
	}
562
	ip->i_res->rs_qa_qd_num = 0;
D
David Teigland 已提交
563 564 565 566
}

static int sort_qd(const void *a, const void *b)
{
567 568
	const struct gfs2_quota_data *qd_a = *(const struct gfs2_quota_data **)a;
	const struct gfs2_quota_data *qd_b = *(const struct gfs2_quota_data **)b;
D
David Teigland 已提交
569 570 571 572

	if (!test_bit(QDF_USER, &qd_a->qd_flags) !=
	    !test_bit(QDF_USER, &qd_b->qd_flags)) {
		if (test_bit(QDF_USER, &qd_a->qd_flags))
573
			return -1;
D
David Teigland 已提交
574
		else
575
			return 1;
D
David Teigland 已提交
576
	}
577 578 579 580
	if (qd_a->qd_id < qd_b->qd_id)
		return -1;
	if (qd_a->qd_id > qd_b->qd_id)
		return 1;
D
David Teigland 已提交
581

582
	return 0;
D
David Teigland 已提交
583 584
}

585
static void do_qc(struct gfs2_quota_data *qd, s64 change)
D
David Teigland 已提交
586 587
{
	struct gfs2_sbd *sdp = qd->qd_gl->gl_sbd;
588
	struct gfs2_inode *ip = GFS2_I(sdp->sd_qc_inode);
D
David Teigland 已提交
589
	struct gfs2_quota_change *qc = qd->qd_bh_qc;
590
	s64 x;
D
David Teigland 已提交
591

592
	mutex_lock(&sdp->sd_quota_mutex);
593
	gfs2_trans_add_bh(ip->i_gl, qd->qd_bh, 1);
D
David Teigland 已提交
594 595 596 597 598 599 600 601 602

	if (!test_bit(QDF_CHANGE, &qd->qd_flags)) {
		qc->qc_change = 0;
		qc->qc_flags = 0;
		if (test_bit(QDF_USER, &qd->qd_flags))
			qc->qc_flags = cpu_to_be32(GFS2_QCF_USER);
		qc->qc_id = cpu_to_be32(qd->qd_id);
	}

A
Al Viro 已提交
603
	x = be64_to_cpu(qc->qc_change) + change;
D
David Teigland 已提交
604 605
	qc->qc_change = cpu_to_be64(x);

606
	spin_lock(&qd_lru_lock);
D
David Teigland 已提交
607
	qd->qd_change = x;
608
	spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
609 610 611 612 613 614 615 616 617 618 619 620

	if (!x) {
		gfs2_assert_warn(sdp, test_bit(QDF_CHANGE, &qd->qd_flags));
		clear_bit(QDF_CHANGE, &qd->qd_flags);
		qc->qc_flags = 0;
		qc->qc_id = 0;
		slot_put(qd);
		qd_put(qd);
	} else if (!test_and_set_bit(QDF_CHANGE, &qd->qd_flags)) {
		qd_hold(qd);
		slot_hold(qd);
	}
621

622
	mutex_unlock(&sdp->sd_quota_mutex);
D
David Teigland 已提交
623 624
}

625
/**
626 627 628
 * gfs2_adjust_quota - adjust record of current block usage
 * @ip: The quota inode
 * @loc: Offset of the entry in the quota file
S
Steven Whitehouse 已提交
629
 * @change: The amount of usage change to record
630
 * @qd: The quota data
S
Steven Whitehouse 已提交
631
 * @fdq: The updated limits to record
632 633 634
 *
 * This function was mostly borrowed from gfs2_block_truncate_page which was
 * in turn mostly borrowed from ext3
635 636
 *
 * Returns: 0 or -ve on error
637
 */
638

639
static int gfs2_adjust_quota(struct gfs2_inode *ip, loff_t loc,
S
Steven Whitehouse 已提交
640 641
			     s64 change, struct gfs2_quota_data *qd,
			     struct fs_disk_quota *fdq)
642
{
643
	struct inode *inode = &ip->i_inode;
644
	struct gfs2_sbd *sdp = GFS2_SB(inode);
645 646
	struct address_space *mapping = inode->i_mapping;
	unsigned long index = loc >> PAGE_CACHE_SHIFT;
647
	unsigned offset = loc & (PAGE_CACHE_SIZE - 1);
648
	unsigned blocksize, iblock, pos;
S
Steven Whitehouse 已提交
649
	struct buffer_head *bh;
650
	struct page *page;
651 652 653
	void *kaddr, *ptr;
	struct gfs2_quota q, *qp;
	int err, nbytes;
S
Steven Whitehouse 已提交
654
	u64 size;
655

S
Steven Whitehouse 已提交
656 657 658 659 660
	if (gfs2_is_stuffed(ip)) {
		err = gfs2_unstuff_dinode(ip, NULL);
		if (err)
			return err;
	}
661 662

	memset(&q, 0, sizeof(struct gfs2_quota));
663
	err = gfs2_internal_read(ip, (char *)&q, &loc, sizeof(q));
664 665 666 667 668 669 670 671 672 673 674
	if (err < 0)
		return err;

	err = -EIO;
	qp = &q;
	qp->qu_value = be64_to_cpu(qp->qu_value);
	qp->qu_value += change;
	qp->qu_value = cpu_to_be64(qp->qu_value);
	qd->qd_qb.qb_value = qp->qu_value;
	if (fdq) {
		if (fdq->d_fieldmask & FS_DQ_BSOFT) {
675
			qp->qu_warn = cpu_to_be64(fdq->d_blk_softlimit >> sdp->sd_fsb2bb_shift);
676 677 678
			qd->qd_qb.qb_warn = qp->qu_warn;
		}
		if (fdq->d_fieldmask & FS_DQ_BHARD) {
679
			qp->qu_limit = cpu_to_be64(fdq->d_blk_hardlimit >> sdp->sd_fsb2bb_shift);
680 681
			qd->qd_qb.qb_limit = qp->qu_limit;
		}
682 683 684 685
		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;
		}
686 687 688 689 690 691
	}

	/* Write the quota into the quota file on disk */
	ptr = qp;
	nbytes = sizeof(struct gfs2_quota);
get_a_page:
692
	page = find_or_create_page(mapping, index, GFP_NOFS);
693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710
	if (!page)
		return -ENOMEM;

	blocksize = inode->i_sb->s_blocksize;
	iblock = index << (PAGE_CACHE_SHIFT - inode->i_sb->s_blocksize_bits);

	if (!page_has_buffers(page))
		create_empty_buffers(page, blocksize, 0);

	bh = page_buffers(page);
	pos = blocksize;
	while (offset >= pos) {
		bh = bh->b_this_page;
		iblock++;
		pos += blocksize;
	}

	if (!buffer_mapped(bh)) {
711
		gfs2_block_map(inode, iblock, bh, 1);
712
		if (!buffer_mapped(bh))
713 714
			goto unlock_out;
		/* If it's a newly allocated disk block for quota, zero it */
A
Abhijith Das 已提交
715 716
		if (buffer_new(bh))
			zero_user(page, pos - blocksize, bh->b_size);
717 718 719 720 721 722
	}

	if (PageUptodate(page))
		set_buffer_uptodate(bh);

	if (!buffer_uptodate(bh)) {
S
Steven Whitehouse 已提交
723
		ll_rw_block(READ | REQ_META, 1, &bh);
724 725
		wait_on_buffer(bh);
		if (!buffer_uptodate(bh))
726
			goto unlock_out;
727 728 729 730
	}

	gfs2_trans_add_bh(ip->i_gl, bh, 0);

731
	kaddr = kmap_atomic(page);
732 733 734
	if (offset + sizeof(struct gfs2_quota) > PAGE_CACHE_SIZE)
		nbytes = PAGE_CACHE_SIZE - offset;
	memcpy(kaddr + offset, ptr, nbytes);
735
	flush_dcache_page(page);
736
	kunmap_atomic(kaddr);
737 738 739 740 741
	unlock_page(page);
	page_cache_release(page);

	/* If quota straddles page boundary, we need to update the rest of the
	 * quota at the beginning of the next page */
A
Abhijith Das 已提交
742
	if ((offset + sizeof(struct gfs2_quota)) > PAGE_CACHE_SIZE) {
743 744 745 746 747 748
		ptr = ptr + nbytes;
		nbytes = sizeof(struct gfs2_quota) - nbytes;
		offset = 0;
		index++;
		goto get_a_page;
	}
S
Steven Whitehouse 已提交
749 750

	size = loc + sizeof(struct gfs2_quota);
S
Steven Whitehouse 已提交
751
	if (size > inode->i_size)
S
Steven Whitehouse 已提交
752 753 754
		i_size_write(inode, size);
	inode->i_mtime = inode->i_atime = CURRENT_TIME;
	mark_inode_dirty(inode);
755
	return 0;
S
Steven Whitehouse 已提交
756

757
unlock_out:
758 759 760 761 762
	unlock_page(page);
	page_cache_release(page);
	return err;
}

D
David Teigland 已提交
763 764 765
static int do_sync(unsigned int num_qd, struct gfs2_quota_data **qda)
{
	struct gfs2_sbd *sdp = (*qda)->qd_gl->gl_sbd;
766
	struct gfs2_inode *ip = GFS2_I(sdp->sd_quota_inode);
D
David Teigland 已提交
767 768 769 770
	unsigned int data_blocks, ind_blocks;
	struct gfs2_holder *ghs, i_gh;
	unsigned int qx, x;
	struct gfs2_quota_data *qd;
771
	unsigned reserved;
772
	loff_t offset;
773
	unsigned int nalloc = 0, blocks;
D
David Teigland 已提交
774 775
	int error;

776 777 778 779
	error = gfs2_rs_alloc(ip);
	if (error)
		return error;

D
David Teigland 已提交
780 781 782
	gfs2_write_calc_reserv(ip, sizeof(struct gfs2_quota),
			      &data_blocks, &ind_blocks);

J
Josef Bacik 已提交
783
	ghs = kcalloc(num_qd, sizeof(struct gfs2_holder), GFP_NOFS);
D
David Teigland 已提交
784 785 786 787
	if (!ghs)
		return -ENOMEM;

	sort(qda, num_qd, sizeof(struct gfs2_quota_data *), sort_qd, NULL);
J
Jan Kara 已提交
788
	mutex_lock(&ip->i_inode.i_mutex);
D
David Teigland 已提交
789
	for (qx = 0; qx < num_qd; qx++) {
790
		error = gfs2_glock_nq_init(qda[qx]->qd_gl, LM_ST_EXCLUSIVE,
D
David Teigland 已提交
791 792 793 794 795 796 797 798 799 800 801
					   GL_NOCACHE, &ghs[qx]);
		if (error)
			goto out;
	}

	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &i_gh);
	if (error)
		goto out;

	for (x = 0; x < num_qd; x++) {
		offset = qd2offset(qda[x]);
802 803
		if (gfs2_write_alloc_required(ip, offset,
					      sizeof(struct gfs2_quota)))
D
David Teigland 已提交
804 805 806
			nalloc++;
	}

807 808 809 810 811 812
	/* 
	 * 1 blk for unstuffing inode if stuffed. We add this extra
	 * block to the reservation unconditionally. If the inode
	 * doesn't need unstuffing, the block will be released to the 
	 * rgrp since it won't be allocated during the transaction
	 */
813 814 815 816
	/* +3 in the end for unstuffing block, inode size update block
	 * and another block in case quota straddles page boundary and 
	 * two blocks need to be updated instead of 1 */
	blocks = num_qd * data_blocks + RES_DINODE + num_qd + 3;
D
David Teigland 已提交
817

818
	reserved = 1 + (nalloc * (data_blocks + ind_blocks));
S
Steven Whitehouse 已提交
819
	error = gfs2_inplace_reserve(ip, reserved, 0);
820 821
	if (error)
		goto out_alloc;
D
David Teigland 已提交
822

823
	if (nalloc)
824
		blocks += gfs2_rg_blocks(ip, reserved) + nalloc * ind_blocks + RES_STATFS;
825 826 827 828

	error = gfs2_trans_begin(sdp, blocks, 0);
	if (error)
		goto out_ipres;
D
David Teigland 已提交
829 830 831 832

	for (x = 0; x < num_qd; x++) {
		qd = qda[x];
		offset = qd2offset(qd);
S
Steven Whitehouse 已提交
833
		error = gfs2_adjust_quota(ip, offset, qd->qd_change_sync, qd, NULL);
834
		if (error)
D
David Teigland 已提交
835 836 837
			goto out_end_trans;

		do_qc(qd, -qd->qd_change_sync);
838
		set_bit(QDF_REFRESH, &qd->qd_flags);
D
David Teigland 已提交
839 840 841 842
	}

	error = 0;

843
out_end_trans:
D
David Teigland 已提交
844
	gfs2_trans_end(sdp);
845
out_ipres:
846
	gfs2_inplace_release(ip);
847
out_alloc:
D
David Teigland 已提交
848
	gfs2_glock_dq_uninit(&i_gh);
849
out:
D
David Teigland 已提交
850 851
	while (qx--)
		gfs2_glock_dq_uninit(&ghs[qx]);
S
Steven Whitehouse 已提交
852
	mutex_unlock(&ip->i_inode.i_mutex);
D
David Teigland 已提交
853
	kfree(ghs);
854
	gfs2_log_flush(ip->i_gl->gl_sbd, ip->i_gl);
D
David Teigland 已提交
855 856 857
	return error;
}

S
Steven Whitehouse 已提交
858 859 860 861 862 863 864 865 866 867
static int update_qd(struct gfs2_sbd *sdp, struct gfs2_quota_data *qd)
{
	struct gfs2_inode *ip = GFS2_I(sdp->sd_quota_inode);
	struct gfs2_quota q;
	struct gfs2_quota_lvb *qlvb;
	loff_t pos;
	int error;

	memset(&q, 0, sizeof(struct gfs2_quota));
	pos = qd2offset(qd);
868
	error = gfs2_internal_read(ip, (char *)&q, &pos, sizeof(q));
S
Steven Whitehouse 已提交
869 870 871 872 873 874 875 876 877 878 879 880 881 882
	if (error < 0)
		return error;

	qlvb = (struct gfs2_quota_lvb *)qd->qd_gl->gl_lvb;
	qlvb->qb_magic = cpu_to_be32(GFS2_MAGIC);
	qlvb->__pad = 0;
	qlvb->qb_limit = q.qu_limit;
	qlvb->qb_warn = q.qu_warn;
	qlvb->qb_value = q.qu_value;
	qd->qd_qb = *qlvb;

	return 0;
}

D
David Teigland 已提交
883 884 885 886
static int do_glock(struct gfs2_quota_data *qd, int force_refresh,
		    struct gfs2_holder *q_gh)
{
	struct gfs2_sbd *sdp = qd->qd_gl->gl_sbd;
887
	struct gfs2_inode *ip = GFS2_I(sdp->sd_quota_inode);
D
David Teigland 已提交
888 889 890
	struct gfs2_holder i_gh;
	int error;

891
restart:
D
David Teigland 已提交
892 893 894 895
	error = gfs2_glock_nq_init(qd->qd_gl, LM_ST_SHARED, 0, q_gh);
	if (error)
		return error;

896
	qd->qd_qb = *(struct gfs2_quota_lvb *)qd->qd_gl->gl_lvb;
D
David Teigland 已提交
897

898
	if (force_refresh || qd->qd_qb.qb_magic != cpu_to_be32(GFS2_MAGIC)) {
D
David Teigland 已提交
899
		gfs2_glock_dq_uninit(q_gh);
900 901
		error = gfs2_glock_nq_init(qd->qd_gl, LM_ST_EXCLUSIVE,
					   GL_NOCACHE, q_gh);
D
David Teigland 已提交
902 903 904
		if (error)
			return error;

905
		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, 0, &i_gh);
D
David Teigland 已提交
906 907 908
		if (error)
			goto fail;

S
Steven Whitehouse 已提交
909 910
		error = update_qd(sdp, qd);
		if (error)
911
			goto fail_gunlock;
D
David Teigland 已提交
912

S
Steven Whitehouse 已提交
913
		gfs2_glock_dq_uninit(&i_gh);
914 915 916
		gfs2_glock_dq_uninit(q_gh);
		force_refresh = 0;
		goto restart;
D
David Teigland 已提交
917 918 919 920
	}

	return 0;

921
fail_gunlock:
D
David Teigland 已提交
922
	gfs2_glock_dq_uninit(&i_gh);
923
fail:
D
David Teigland 已提交
924 925 926 927
	gfs2_glock_dq_uninit(q_gh);
	return error;
}

928
int gfs2_quota_lock(struct gfs2_inode *ip, u32 uid, u32 gid)
D
David Teigland 已提交
929
{
930
	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
931
	struct gfs2_quota_data *qd;
D
David Teigland 已提交
932 933 934
	unsigned int x;
	int error = 0;

S
Steven Whitehouse 已提交
935 936 937
	error = gfs2_quota_hold(ip, uid, gid);
	if (error)
		return error;
D
David Teigland 已提交
938 939 940 941 942

	if (capable(CAP_SYS_RESOURCE) ||
	    sdp->sd_args.ar_quota != GFS2_QUOTA_ON)
		return 0;

943 944
	sort(ip->i_res->rs_qa_qd, ip->i_res->rs_qa_qd_num,
	     sizeof(struct gfs2_quota_data *), sort_qd, NULL);
D
David Teigland 已提交
945

946
	for (x = 0; x < ip->i_res->rs_qa_qd_num; x++) {
947
		int force = NO_FORCE;
948
		qd = ip->i_res->rs_qa_qd[x];
949 950
		if (test_and_clear_bit(QDF_REFRESH, &qd->qd_flags))
			force = FORCE;
951
		error = do_glock(qd, force, &ip->i_res->rs_qa_qd_ghs[x]);
D
David Teigland 已提交
952 953 954 955 956 957 958 959
		if (error)
			break;
	}

	if (!error)
		set_bit(GIF_QD_LOCKED, &ip->i_flags);
	else {
		while (x--)
960
			gfs2_glock_dq_uninit(&ip->i_res->rs_qa_qd_ghs[x]);
D
David Teigland 已提交
961 962 963 964 965 966 967 968 969 970
		gfs2_quota_unhold(ip);
	}

	return error;
}

static int need_sync(struct gfs2_quota_data *qd)
{
	struct gfs2_sbd *sdp = qd->qd_gl->gl_sbd;
	struct gfs2_tune *gt = &sdp->sd_tune;
971
	s64 value;
D
David Teigland 已提交
972 973 974 975 976 977
	unsigned int num, den;
	int do_sync = 1;

	if (!qd->qd_qb.qb_limit)
		return 0;

978
	spin_lock(&qd_lru_lock);
D
David Teigland 已提交
979
	value = qd->qd_change;
980
	spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
981 982 983 984 985 986 987 988

	spin_lock(&gt->gt_spin);
	num = gt->gt_quota_scale_num;
	den = gt->gt_quota_scale_den;
	spin_unlock(&gt->gt_spin);

	if (value < 0)
		do_sync = 0;
989 990
	else if ((s64)be64_to_cpu(qd->qd_qb.qb_value) >=
		 (s64)be64_to_cpu(qd->qd_qb.qb_limit))
D
David Teigland 已提交
991 992 993
		do_sync = 0;
	else {
		value *= gfs2_jindex_size(sdp) * num;
994
		value = div_s64(value, den);
995
		value += (s64)be64_to_cpu(qd->qd_qb.qb_value);
996
		if (value < (s64)be64_to_cpu(qd->qd_qb.qb_limit))
D
David Teigland 已提交
997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
			do_sync = 0;
	}

	return do_sync;
}

void gfs2_quota_unlock(struct gfs2_inode *ip)
{
	struct gfs2_quota_data *qda[4];
	unsigned int count = 0;
	unsigned int x;

	if (!test_and_clear_bit(GIF_QD_LOCKED, &ip->i_flags))
		goto out;

1012
	for (x = 0; x < ip->i_res->rs_qa_qd_num; x++) {
D
David Teigland 已提交
1013 1014 1015
		struct gfs2_quota_data *qd;
		int sync;

1016
		qd = ip->i_res->rs_qa_qd[x];
D
David Teigland 已提交
1017 1018
		sync = need_sync(qd);

1019
		gfs2_glock_dq_uninit(&ip->i_res->rs_qa_qd_ghs[x]);
D
David Teigland 已提交
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030

		if (sync && qd_trylock(qd))
			qda[count++] = qd;
	}

	if (count) {
		do_sync(count, qda);
		for (x = 0; x < count; x++)
			qd_unlock(qda[x]);
	}

1031
out:
D
David Teigland 已提交
1032 1033 1034 1035 1036 1037 1038 1039 1040
	gfs2_quota_unhold(ip);
}

#define MAX_LINE 256

static int print_message(struct gfs2_quota_data *qd, char *type)
{
	struct gfs2_sbd *sdp = qd->qd_gl->gl_sbd;

S
Steven Whitehouse 已提交
1041
	printk(KERN_INFO "GFS2: fsid=%s: quota %s for %s %u\n",
1042 1043 1044
	       sdp->sd_fsname, type,
	       (test_bit(QDF_USER, &qd->qd_flags)) ? "user" : "group",
	       qd->qd_id);
D
David Teigland 已提交
1045 1046 1047 1048

	return 0;
}

1049
int gfs2_quota_check(struct gfs2_inode *ip, u32 uid, u32 gid)
D
David Teigland 已提交
1050
{
1051
	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
D
David Teigland 已提交
1052
	struct gfs2_quota_data *qd;
1053
	s64 value;
D
David Teigland 已提交
1054 1055 1056 1057 1058 1059 1060 1061 1062
	unsigned int x;
	int error = 0;

	if (!test_bit(GIF_QD_LOCKED, &ip->i_flags))
		return 0;

        if (sdp->sd_args.ar_quota != GFS2_QUOTA_ON)
                return 0;

1063 1064
	for (x = 0; x < ip->i_res->rs_qa_qd_num; x++) {
		qd = ip->i_res->rs_qa_qd[x];
D
David Teigland 已提交
1065 1066 1067 1068 1069

		if (!((qd->qd_id == uid && test_bit(QDF_USER, &qd->qd_flags)) ||
		      (qd->qd_id == gid && !test_bit(QDF_USER, &qd->qd_flags))))
			continue;

1070
		value = (s64)be64_to_cpu(qd->qd_qb.qb_value);
1071
		spin_lock(&qd_lru_lock);
D
David Teigland 已提交
1072
		value += qd->qd_change;
1073
		spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
1074

1075
		if (be64_to_cpu(qd->qd_qb.qb_limit) && (s64)be64_to_cpu(qd->qd_qb.qb_limit) < value) {
D
David Teigland 已提交
1076
			print_message(qd, "exceeded");
1077 1078 1079 1080
			quota_send_warning(make_kqid(&init_user_ns,
						     test_bit(QDF_USER, &qd->qd_flags) ?
						     USRQUOTA : GRPQUOTA,
						     qd->qd_id),
S
Steven Whitehouse 已提交
1081 1082
					   sdp->sd_vfs->s_dev, QUOTA_NL_BHARDWARN);

D
David Teigland 已提交
1083 1084
			error = -EDQUOT;
			break;
1085
		} else if (be64_to_cpu(qd->qd_qb.qb_warn) &&
1086
			   (s64)be64_to_cpu(qd->qd_qb.qb_warn) < value &&
D
David Teigland 已提交
1087
			   time_after_eq(jiffies, qd->qd_last_warn +
1088 1089
					 gfs2_tune_get(sdp,
						gt_quota_warn_period) * HZ)) {
1090 1091 1092 1093
			quota_send_warning(make_kqid(&init_user_ns,
						     test_bit(QDF_USER, &qd->qd_flags) ?
						     USRQUOTA : GRPQUOTA,
						     qd->qd_id),
S
Steven Whitehouse 已提交
1094
					   sdp->sd_vfs->s_dev, QUOTA_NL_BSOFTWARN);
D
David Teigland 已提交
1095 1096 1097 1098 1099 1100 1101 1102
			error = print_message(qd, "warning");
			qd->qd_last_warn = jiffies;
		}
	}

	return error;
}

1103 1104
void gfs2_quota_change(struct gfs2_inode *ip, s64 change,
		       u32 uid, u32 gid)
D
David Teigland 已提交
1105 1106 1107 1108
{
	struct gfs2_quota_data *qd;
	unsigned int x;

1109
	if (gfs2_assert_warn(GFS2_SB(&ip->i_inode), change))
D
David Teigland 已提交
1110
		return;
1111
	if (ip->i_diskflags & GFS2_DIF_SYSTEM)
D
David Teigland 已提交
1112 1113
		return;

1114 1115
	for (x = 0; x < ip->i_res->rs_qa_qd_num; x++) {
		qd = ip->i_res->rs_qa_qd[x];
D
David Teigland 已提交
1116 1117 1118 1119 1120 1121 1122 1123

		if ((qd->qd_id == uid && test_bit(QDF_USER, &qd->qd_flags)) ||
		    (qd->qd_id == gid && !test_bit(QDF_USER, &qd->qd_flags))) {
			do_qc(qd, change);
		}
	}
}

1124
int gfs2_quota_sync(struct super_block *sb, int type)
D
David Teigland 已提交
1125
{
1126
	struct gfs2_sbd *sdp = sb->s_fs_info;
D
David Teigland 已提交
1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167
	struct gfs2_quota_data **qda;
	unsigned int max_qd = gfs2_tune_get(sdp, gt_quota_simul_sync);
	unsigned int num_qd;
	unsigned int x;
	int error = 0;

	sdp->sd_quota_sync_gen++;

	qda = kcalloc(max_qd, sizeof(struct gfs2_quota_data *), GFP_KERNEL);
	if (!qda)
		return -ENOMEM;

	do {
		num_qd = 0;

		for (;;) {
			error = qd_fish(sdp, qda + num_qd);
			if (error || !qda[num_qd])
				break;
			if (++num_qd == max_qd)
				break;
		}

		if (num_qd) {
			if (!error)
				error = do_sync(num_qd, qda);
			if (!error)
				for (x = 0; x < num_qd; x++)
					qda[x]->qd_sync_gen =
						sdp->sd_quota_sync_gen;

			for (x = 0; x < num_qd; x++)
				qd_unlock(qda[x]);
		}
	} while (!error && num_qd == max_qd);

	kfree(qda);

	return error;
}

1168 1169
static int gfs2_quota_sync_timeo(struct super_block *sb, int type)
{
1170
	return gfs2_quota_sync(sb, type);
1171 1172
}

1173
int gfs2_quota_refresh(struct gfs2_sbd *sdp, int user, u32 id)
D
David Teigland 已提交
1174 1175 1176 1177 1178
{
	struct gfs2_quota_data *qd;
	struct gfs2_holder q_gh;
	int error;

1179
	error = qd_get(sdp, user, id, &qd);
D
David Teigland 已提交
1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190
	if (error)
		return error;

	error = do_glock(qd, FORCE, &q_gh);
	if (!error)
		gfs2_glock_dq_uninit(&q_gh);

	qd_put(qd);
	return error;
}

1191 1192 1193 1194 1195 1196 1197 1198 1199
static void gfs2_quota_change_in(struct gfs2_quota_change_host *qc, const void *buf)
{
	const struct gfs2_quota_change *str = buf;

	qc->qc_change = be64_to_cpu(str->qc_change);
	qc->qc_flags = be32_to_cpu(str->qc_flags);
	qc->qc_id = be32_to_cpu(str->qc_id);
}

D
David Teigland 已提交
1200 1201
int gfs2_quota_init(struct gfs2_sbd *sdp)
{
1202
	struct gfs2_inode *ip = GFS2_I(sdp->sd_qc_inode);
S
Steven Whitehouse 已提交
1203 1204
	u64 size = i_size_read(sdp->sd_qc_inode);
	unsigned int blocks = size >> sdp->sd_sb.sb_bsize_shift;
D
David Teigland 已提交
1205 1206
	unsigned int x, slot = 0;
	unsigned int found = 0;
1207 1208
	u64 dblock;
	u32 extlen = 0;
D
David Teigland 已提交
1209 1210
	int error;

S
Steven Whitehouse 已提交
1211
	if (gfs2_check_internal_file_size(sdp->sd_qc_inode, 1, 64 << 20))
1212
		return -EIO;
S
Steven Whitehouse 已提交
1213

D
David Teigland 已提交
1214
	sdp->sd_quota_slots = blocks * sdp->sd_qc_per_block;
1215
	sdp->sd_quota_chunks = DIV_ROUND_UP(sdp->sd_quota_slots, 8 * PAGE_SIZE);
D
David Teigland 已提交
1216 1217 1218 1219

	error = -ENOMEM;

	sdp->sd_quota_bitmap = kcalloc(sdp->sd_quota_chunks,
J
Josef Bacik 已提交
1220
				       sizeof(unsigned char *), GFP_NOFS);
D
David Teigland 已提交
1221 1222 1223 1224
	if (!sdp->sd_quota_bitmap)
		return error;

	for (x = 0; x < sdp->sd_quota_chunks; x++) {
J
Josef Bacik 已提交
1225
		sdp->sd_quota_bitmap[x] = kzalloc(PAGE_SIZE, GFP_NOFS);
D
David Teigland 已提交
1226 1227 1228 1229 1230 1231 1232 1233 1234 1235
		if (!sdp->sd_quota_bitmap[x])
			goto fail;
	}

	for (x = 0; x < blocks; x++) {
		struct buffer_head *bh;
		unsigned int y;

		if (!extlen) {
			int new = 0;
1236
			error = gfs2_extent_map(&ip->i_inode, x, &new, &dblock, &extlen);
D
David Teigland 已提交
1237 1238 1239 1240
			if (error)
				goto fail;
		}
		error = -EIO;
S
Steven Whitehouse 已提交
1241 1242 1243
		bh = gfs2_meta_ra(ip->i_gl, dblock, extlen);
		if (!bh)
			goto fail;
D
David Teigland 已提交
1244 1245 1246 1247 1248
		if (gfs2_metatype_check(sdp, bh, GFS2_METATYPE_QC)) {
			brelse(bh);
			goto fail;
		}

S
Steven Whitehouse 已提交
1249
		for (y = 0; y < sdp->sd_qc_per_block && slot < sdp->sd_quota_slots;
D
David Teigland 已提交
1250
		     y++, slot++) {
1251
			struct gfs2_quota_change_host qc;
D
David Teigland 已提交
1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271
			struct gfs2_quota_data *qd;

			gfs2_quota_change_in(&qc, bh->b_data +
					  sizeof(struct gfs2_meta_header) +
					  y * sizeof(struct gfs2_quota_change));
			if (!qc.qc_change)
				continue;

			error = qd_alloc(sdp, (qc.qc_flags & GFS2_QCF_USER),
					 qc.qc_id, &qd);
			if (error) {
				brelse(bh);
				goto fail;
			}

			set_bit(QDF_CHANGE, &qd->qd_flags);
			qd->qd_change = qc.qc_change;
			qd->qd_slot = slot;
			qd->qd_slot_count = 1;

1272
			spin_lock(&qd_lru_lock);
D
David Teigland 已提交
1273 1274 1275
			gfs2_icbit_munge(sdp, sdp->sd_quota_bitmap, slot, 1);
			list_add(&qd->qd_list, &sdp->sd_quota_list);
			atomic_inc(&sdp->sd_quota_count);
1276
			spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290

			found++;
		}

		brelse(bh);
		dblock++;
		extlen--;
	}

	if (found)
		fs_info(sdp, "found %u quota changes\n", found);

	return 0;

1291
fail:
D
David Teigland 已提交
1292 1293 1294 1295 1296 1297 1298 1299 1300 1301
	gfs2_quota_cleanup(sdp);
	return error;
}

void gfs2_quota_cleanup(struct gfs2_sbd *sdp)
{
	struct list_head *head = &sdp->sd_quota_list;
	struct gfs2_quota_data *qd;
	unsigned int x;

1302
	spin_lock(&qd_lru_lock);
D
David Teigland 已提交
1303 1304 1305
	while (!list_empty(head)) {
		qd = list_entry(head->prev, struct gfs2_quota_data, qd_list);

1306 1307 1308 1309 1310
		if (atomic_read(&qd->qd_count) > 1 ||
		    (atomic_read(&qd->qd_count) &&
		     !test_bit(QDF_CHANGE, &qd->qd_flags))) {
			list_move(&qd->qd_list, head);
			spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
1311
			schedule();
1312
			spin_lock(&qd_lru_lock);
D
David Teigland 已提交
1313 1314 1315 1316
			continue;
		}

		list_del(&qd->qd_list);
1317 1318 1319 1320 1321
		/* Also remove if this qd exists in the reclaim list */
		if (!list_empty(&qd->qd_reclaim)) {
			list_del_init(&qd->qd_reclaim);
			atomic_dec(&qd_lru_count);
		}
D
David Teigland 已提交
1322
		atomic_dec(&sdp->sd_quota_count);
1323
		spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
1324

1325
		if (!atomic_read(&qd->qd_count)) {
D
David Teigland 已提交
1326 1327 1328 1329 1330 1331
			gfs2_assert_warn(sdp, !qd->qd_change);
			gfs2_assert_warn(sdp, !qd->qd_slot_count);
		} else
			gfs2_assert_warn(sdp, qd->qd_slot_count == 1);
		gfs2_assert_warn(sdp, !qd->qd_bh_count);

1332
		gfs2_glock_put(qd->qd_gl);
1333
		kmem_cache_free(gfs2_quotad_cachep, qd);
D
David Teigland 已提交
1334

1335
		spin_lock(&qd_lru_lock);
D
David Teigland 已提交
1336
	}
1337
	spin_unlock(&qd_lru_lock);
D
David Teigland 已提交
1338 1339 1340 1341 1342 1343 1344 1345 1346 1347

	gfs2_assert_warn(sdp, !atomic_read(&sdp->sd_quota_count));

	if (sdp->sd_quota_bitmap) {
		for (x = 0; x < sdp->sd_quota_chunks; x++)
			kfree(sdp->sd_quota_bitmap[x]);
		kfree(sdp->sd_quota_bitmap);
	}
}

1348 1349 1350 1351 1352 1353 1354 1355 1356
static void quotad_error(struct gfs2_sbd *sdp, const char *msg, int error)
{
	if (error == 0 || error == -EROFS)
		return;
	if (!test_bit(SDF_SHUTDOWN, &sdp->sd_flags))
		fs_err(sdp, "gfs2_quotad: %s error %d\n", msg, error);
}

static void quotad_check_timeo(struct gfs2_sbd *sdp, const char *msg,
1357
			       int (*fxn)(struct super_block *sb, int type),
1358 1359 1360 1361
			       unsigned long t, unsigned long *timeo,
			       unsigned int *new_timeo)
{
	if (t >= *timeo) {
1362
		int error = fxn(sdp->sd_vfs, 0);
1363 1364 1365 1366 1367 1368 1369
		quotad_error(sdp, msg, error);
		*timeo = gfs2_tune_get_i(&sdp->sd_tune, new_timeo) * HZ;
	} else {
		*timeo -= t;
	}
}

1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388
static void quotad_check_trunc_list(struct gfs2_sbd *sdp)
{
	struct gfs2_inode *ip;

	while(1) {
		ip = NULL;
		spin_lock(&sdp->sd_trunc_lock);
		if (!list_empty(&sdp->sd_trunc_list)) {
			ip = list_entry(sdp->sd_trunc_list.next,
					struct gfs2_inode, i_trunc_list);
			list_del_init(&ip->i_trunc_list);
		}
		spin_unlock(&sdp->sd_trunc_lock);
		if (ip == NULL)
			return;
		gfs2_glock_finish_truncate(ip);
	}
}

1389 1390 1391 1392 1393 1394 1395 1396
void gfs2_wake_up_statfs(struct gfs2_sbd *sdp) {
	if (!sdp->sd_statfs_force_sync) {
		sdp->sd_statfs_force_sync = 1;
		wake_up(&sdp->sd_quota_wait);
	}
}


1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410
/**
 * gfs2_quotad - Write cached quota changes into the quota file
 * @sdp: Pointer to GFS2 superblock
 *
 */

int gfs2_quotad(void *data)
{
	struct gfs2_sbd *sdp = data;
	struct gfs2_tune *tune = &sdp->sd_tune;
	unsigned long statfs_timeo = 0;
	unsigned long quotad_timeo = 0;
	unsigned long t = 0;
	DEFINE_WAIT(wait);
1411
	int empty;
1412 1413 1414 1415

	while (!kthread_should_stop()) {

		/* Update the master statfs file */
1416 1417 1418 1419 1420 1421 1422 1423 1424
		if (sdp->sd_statfs_force_sync) {
			int error = gfs2_statfs_sync(sdp->sd_vfs, 0);
			quotad_error(sdp, "statfs", error);
			statfs_timeo = gfs2_tune_get(sdp, gt_statfs_quantum) * HZ;
		}
		else
			quotad_check_timeo(sdp, "statfs", gfs2_statfs_sync, t,
				   	   &statfs_timeo,
					   &tune->gt_statfs_quantum);
1425 1426

		/* Update quota file */
1427
		quotad_check_timeo(sdp, "sync", gfs2_quota_sync_timeo, t,
1428 1429
				   &quotad_timeo, &tune->gt_quota_quantum);

1430 1431 1432
		/* Check for & recover partially truncated inodes */
		quotad_check_trunc_list(sdp);

1433 1434
		try_to_freeze();

1435 1436
		t = min(quotad_timeo, statfs_timeo);

1437
		prepare_to_wait(&sdp->sd_quota_wait, &wait, TASK_INTERRUPTIBLE);
1438 1439 1440
		spin_lock(&sdp->sd_trunc_lock);
		empty = list_empty(&sdp->sd_trunc_list);
		spin_unlock(&sdp->sd_trunc_lock);
1441
		if (empty && !sdp->sd_statfs_force_sync)
1442 1443 1444
			t -= schedule_timeout(t);
		else
			t = 0;
1445 1446 1447 1448 1449 1450
		finish_wait(&sdp->sd_quota_wait, &wait);
	}

	return 0;
}

1451 1452 1453 1454 1455 1456 1457
static int gfs2_quota_get_xstate(struct super_block *sb,
				 struct fs_quota_stat *fqs)
{
	struct gfs2_sbd *sdp = sb->s_fs_info;

	memset(fqs, 0, sizeof(struct fs_quota_stat));
	fqs->qs_version = FS_QSTAT_VERSION;
1458 1459 1460

	switch (sdp->sd_args.ar_quota) {
	case GFS2_QUOTA_ON:
1461
		fqs->qs_flags |= (FS_QUOTA_UDQ_ENFD | FS_QUOTA_GDQ_ENFD);
1462 1463
		/*FALLTHRU*/
	case GFS2_QUOTA_ACCOUNT:
1464
		fqs->qs_flags |= (FS_QUOTA_UDQ_ACCT | FS_QUOTA_GDQ_ACCT);
1465 1466 1467 1468 1469
		break;
	case GFS2_QUOTA_OFF:
		break;
	}

1470 1471 1472 1473 1474 1475 1476 1477 1478 1479
	if (sdp->sd_quota_inode) {
		fqs->qs_uquota.qfs_ino = GFS2_I(sdp->sd_quota_inode)->i_no_addr;
		fqs->qs_uquota.qfs_nblks = sdp->sd_quota_inode->i_blocks;
	}
	fqs->qs_uquota.qfs_nextents = 1; /* unsupported */
	fqs->qs_gquota = fqs->qs_uquota; /* its the same inode in both cases */
	fqs->qs_incoredqs = atomic_read(&qd_lru_count);
	return 0;
}

E
Eric W. Biederman 已提交
1480
static int gfs2_get_dqblk(struct super_block *sb, struct kqid qid,
C
Christoph Hellwig 已提交
1481
			  struct fs_disk_quota *fdq)
S
Steven Whitehouse 已提交
1482 1483 1484 1485 1486 1487
{
	struct gfs2_sbd *sdp = sb->s_fs_info;
	struct gfs2_quota_lvb *qlvb;
	struct gfs2_quota_data *qd;
	struct gfs2_holder q_gh;
	int error;
E
Eric W. Biederman 已提交
1488
	int type;
S
Steven Whitehouse 已提交
1489 1490 1491 1492 1493 1494

	memset(fdq, 0, sizeof(struct fs_disk_quota));

	if (sdp->sd_args.ar_quota == GFS2_QUOTA_OFF)
		return -ESRCH; /* Crazy XFS error code */

E
Eric W. Biederman 已提交
1495
	if (qid.type == USRQUOTA)
S
Steven Whitehouse 已提交
1496
		type = QUOTA_USER;
E
Eric W. Biederman 已提交
1497
	else if (qid.type == GRPQUOTA)
S
Steven Whitehouse 已提交
1498 1499 1500 1501
		type = QUOTA_GROUP;
	else
		return -EINVAL;

E
Eric W. Biederman 已提交
1502
	error = qd_get(sdp, type, from_kqid(&init_user_ns, qid), &qd);
S
Steven Whitehouse 已提交
1503 1504 1505 1506 1507 1508 1509 1510
	if (error)
		return error;
	error = do_glock(qd, FORCE, &q_gh);
	if (error)
		goto out;

	qlvb = (struct gfs2_quota_lvb *)qd->qd_gl->gl_lvb;
	fdq->d_version = FS_DQUOT_VERSION;
1511
	fdq->d_flags = (type == QUOTA_USER) ? FS_USER_QUOTA : FS_GROUP_QUOTA;
E
Eric W. Biederman 已提交
1512
	fdq->d_id = from_kqid(&init_user_ns, qid);
1513 1514 1515
	fdq->d_blk_hardlimit = be64_to_cpu(qlvb->qb_limit) << sdp->sd_fsb2bb_shift;
	fdq->d_blk_softlimit = be64_to_cpu(qlvb->qb_warn) << sdp->sd_fsb2bb_shift;
	fdq->d_bcount = be64_to_cpu(qlvb->qb_value) << sdp->sd_fsb2bb_shift;
S
Steven Whitehouse 已提交
1516 1517 1518 1519 1520 1521 1522

	gfs2_glock_dq_uninit(&q_gh);
out:
	qd_put(qd);
	return error;
}

S
Steven Whitehouse 已提交
1523
/* GFS2 only supports a subset of the XFS fields */
1524
#define GFS2_FIELDMASK (FS_DQ_BSOFT|FS_DQ_BHARD|FS_DQ_BCOUNT)
S
Steven Whitehouse 已提交
1525

E
Eric W. Biederman 已提交
1526
static int gfs2_set_dqblk(struct super_block *sb, struct kqid qid,
C
Christoph Hellwig 已提交
1527
			  struct fs_disk_quota *fdq)
S
Steven Whitehouse 已提交
1528 1529 1530 1531 1532 1533 1534 1535 1536 1537
{
	struct gfs2_sbd *sdp = sb->s_fs_info;
	struct gfs2_inode *ip = GFS2_I(sdp->sd_quota_inode);
	struct gfs2_quota_data *qd;
	struct gfs2_holder q_gh, i_gh;
	unsigned int data_blocks, ind_blocks;
	unsigned int blocks = 0;
	int alloc_required;
	loff_t offset;
	int error;
E
Eric W. Biederman 已提交
1538
	int type;
S
Steven Whitehouse 已提交
1539 1540 1541 1542

	if (sdp->sd_args.ar_quota == GFS2_QUOTA_OFF)
		return -ESRCH; /* Crazy XFS error code */

E
Eric W. Biederman 已提交
1543
	switch(qid.type) {
S
Steven Whitehouse 已提交
1544 1545
	case USRQUOTA:
		type = QUOTA_USER;
1546
		if (fdq->d_flags != FS_USER_QUOTA)
S
Steven Whitehouse 已提交
1547 1548 1549 1550
			return -EINVAL;
		break;
	case GRPQUOTA:
		type = QUOTA_GROUP;
1551
		if (fdq->d_flags != FS_GROUP_QUOTA)
S
Steven Whitehouse 已提交
1552 1553 1554 1555 1556 1557 1558 1559
			return -EINVAL;
		break;
	default:
		return -EINVAL;
	}

	if (fdq->d_fieldmask & ~GFS2_FIELDMASK)
		return -EINVAL;
E
Eric W. Biederman 已提交
1560
	if (fdq->d_id != from_kqid(&init_user_ns, qid))
S
Steven Whitehouse 已提交
1561 1562
		return -EINVAL;

E
Eric W. Biederman 已提交
1563
	error = qd_get(sdp, type, from_kqid(&init_user_ns, qid), &qd);
S
Steven Whitehouse 已提交
1564 1565 1566
	if (error)
		return error;

1567 1568 1569 1570
	error = gfs2_rs_alloc(ip);
	if (error)
		goto out_put;

S
Steven Whitehouse 已提交
1571 1572 1573
	mutex_lock(&ip->i_inode.i_mutex);
	error = gfs2_glock_nq_init(qd->qd_gl, LM_ST_EXCLUSIVE, 0, &q_gh);
	if (error)
1574
		goto out_unlockput;
S
Steven Whitehouse 已提交
1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585
	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &i_gh);
	if (error)
		goto out_q;

	/* Check for existing entry, if none then alloc new blocks */
	error = update_qd(sdp, qd);
	if (error)
		goto out_i;

	/* If nothing has changed, this is a no-op */
	if ((fdq->d_fieldmask & FS_DQ_BSOFT) &&
1586
	    ((fdq->d_blk_softlimit >> sdp->sd_fsb2bb_shift) == be64_to_cpu(qd->qd_qb.qb_warn)))
S
Steven Whitehouse 已提交
1587
		fdq->d_fieldmask ^= FS_DQ_BSOFT;
1588

S
Steven Whitehouse 已提交
1589
	if ((fdq->d_fieldmask & FS_DQ_BHARD) &&
1590
	    ((fdq->d_blk_hardlimit >> sdp->sd_fsb2bb_shift) == be64_to_cpu(qd->qd_qb.qb_limit)))
S
Steven Whitehouse 已提交
1591
		fdq->d_fieldmask ^= FS_DQ_BHARD;
1592 1593 1594 1595 1596

	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;

S
Steven Whitehouse 已提交
1597 1598 1599 1600
	if (fdq->d_fieldmask == 0)
		goto out_i;

	offset = qd2offset(qd);
1601
	alloc_required = gfs2_write_alloc_required(ip, offset, sizeof(struct gfs2_quota));
A
Abhijith Das 已提交
1602 1603
	if (gfs2_is_stuffed(ip))
		alloc_required = 1;
S
Steven Whitehouse 已提交
1604 1605 1606
	if (alloc_required) {
		gfs2_write_calc_reserv(ip, sizeof(struct gfs2_quota),
				       &data_blocks, &ind_blocks);
1607
		blocks = 1 + data_blocks + ind_blocks;
S
Steven Whitehouse 已提交
1608
		error = gfs2_inplace_reserve(ip, blocks, 0);
S
Steven Whitehouse 已提交
1609
		if (error)
1610
			goto out_i;
1611
		blocks += gfs2_rg_blocks(ip, blocks);
S
Steven Whitehouse 已提交
1612 1613
	}

A
Abhijith Das 已提交
1614 1615 1616
	/* Some quotas span block boundaries and can update two blocks,
	   adding an extra block to the transaction to handle such quotas */
	error = gfs2_trans_begin(sdp, blocks + RES_DINODE + 2, 0);
S
Steven Whitehouse 已提交
1617 1618 1619 1620 1621 1622 1623 1624
	if (error)
		goto out_release;

	/* Apply changes */
	error = gfs2_adjust_quota(ip, offset, 0, qd, fdq);

	gfs2_trans_end(sdp);
out_release:
1625
	if (alloc_required)
S
Steven Whitehouse 已提交
1626 1627 1628 1629 1630
		gfs2_inplace_release(ip);
out_i:
	gfs2_glock_dq_uninit(&i_gh);
out_q:
	gfs2_glock_dq_uninit(&q_gh);
1631
out_unlockput:
S
Steven Whitehouse 已提交
1632
	mutex_unlock(&ip->i_inode.i_mutex);
1633
out_put:
S
Steven Whitehouse 已提交
1634 1635 1636 1637
	qd_put(qd);
	return error;
}

1638 1639
const struct quotactl_ops gfs2_quotactl_ops = {
	.quota_sync     = gfs2_quota_sync,
1640
	.get_xstate     = gfs2_quota_get_xstate,
C
Christoph Hellwig 已提交
1641
	.get_dqblk	= gfs2_get_dqblk,
C
Christoph Hellwig 已提交
1642
	.set_dqblk	= gfs2_set_dqblk,
1643
};