mtdswap.c 34.3 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52
/*
 * Swap block device support for MTDs
 * Turns an MTD device into a swap device with block wear leveling
 *
 * Copyright © 2007,2011 Nokia Corporation. All rights reserved.
 *
 * Authors: Jarkko Lavinen <jarkko.lavinen@nokia.com>
 *
 * Based on Richard Purdie's earlier implementation in 2007. Background
 * support and lock-less operation written by Adrian Hunter.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * version 2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
 * 02110-1301 USA
 */

#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/mtd/mtd.h>
#include <linux/mtd/blktrans.h>
#include <linux/rbtree.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/vmalloc.h>
#include <linux/genhd.h>
#include <linux/swap.h>
#include <linux/debugfs.h>
#include <linux/seq_file.h>
#include <linux/device.h>
#include <linux/math64.h>

#define MTDSWAP_PREFIX "mtdswap"

/*
 * The number of free eraseblocks when GC should stop
 */
#define CLEAN_BLOCK_THRESHOLD	20

/*
 * Number of free eraseblocks below which GC can also collect low frag
 * blocks.
 */
53
#define LOW_FRAG_GC_THRESHOLD	5
54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88

/*
 * Wear level cost amortization. We want to do wear leveling on the background
 * without disturbing gc too much. This is made by defining max GC frequency.
 * Frequency value 6 means 1/6 of the GC passes will pick an erase block based
 * on the biggest wear difference rather than the biggest dirtiness.
 *
 * The lower freq2 should be chosen so that it makes sure the maximum erase
 * difference will decrease even if a malicious application is deliberately
 * trying to make erase differences large.
 */
#define MAX_ERASE_DIFF		4000
#define COLLECT_NONDIRTY_BASE	MAX_ERASE_DIFF
#define COLLECT_NONDIRTY_FREQ1	6
#define COLLECT_NONDIRTY_FREQ2	4

#define PAGE_UNDEF		UINT_MAX
#define BLOCK_UNDEF		UINT_MAX
#define BLOCK_ERROR		(UINT_MAX - 1)
#define BLOCK_MAX		(UINT_MAX - 2)

#define EBLOCK_BAD		(1 << 0)
#define EBLOCK_NOMAGIC		(1 << 1)
#define EBLOCK_BITFLIP		(1 << 2)
#define EBLOCK_FAILED		(1 << 3)
#define EBLOCK_READERR		(1 << 4)
#define EBLOCK_IDX_SHIFT	5

struct swap_eb {
	struct rb_node rb;
	struct rb_root *root;

	unsigned int flags;
	unsigned int active_count;
	unsigned int erase_count;
Brian Norris's avatar
Brian Norris committed
89
	unsigned int pad;		/* speeds up pointer decrement */
90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145
};

#define MTDSWAP_ECNT_MIN(rbroot) (rb_entry(rb_first(rbroot), struct swap_eb, \
				rb)->erase_count)
#define MTDSWAP_ECNT_MAX(rbroot) (rb_entry(rb_last(rbroot), struct swap_eb, \
				rb)->erase_count)

struct mtdswap_tree {
	struct rb_root root;
	unsigned int count;
};

enum {
	MTDSWAP_CLEAN,
	MTDSWAP_USED,
	MTDSWAP_LOWFRAG,
	MTDSWAP_HIFRAG,
	MTDSWAP_DIRTY,
	MTDSWAP_BITFLIP,
	MTDSWAP_FAILING,
	MTDSWAP_TREE_CNT,
};

struct mtdswap_dev {
	struct mtd_blktrans_dev *mbd_dev;
	struct mtd_info *mtd;
	struct device *dev;

	unsigned int *page_data;
	unsigned int *revmap;

	unsigned int eblks;
	unsigned int spare_eblks;
	unsigned int pages_per_eblk;
	unsigned int max_erase_count;
	struct swap_eb *eb_data;

	struct mtdswap_tree trees[MTDSWAP_TREE_CNT];

	unsigned long long sect_read_count;
	unsigned long long sect_write_count;
	unsigned long long mtd_write_count;
	unsigned long long mtd_read_count;
	unsigned long long discard_count;
	unsigned long long discard_page_count;

	unsigned int curr_write_pos;
	struct swap_eb *curr_write;

	char *page_buf;
	char *oob_buf;
};

struct mtdswap_oobdata {
	__le16 magic;
	__le32 count;
Brian Norris's avatar
Brian Norris committed
146
} __packed;
147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274

#define MTDSWAP_MAGIC_CLEAN	0x2095
#define MTDSWAP_MAGIC_DIRTY	(MTDSWAP_MAGIC_CLEAN + 1)
#define MTDSWAP_TYPE_CLEAN	0
#define MTDSWAP_TYPE_DIRTY	1
#define MTDSWAP_OOBSIZE		sizeof(struct mtdswap_oobdata)

#define MTDSWAP_ERASE_RETRIES	3 /* Before marking erase block bad */
#define MTDSWAP_IO_RETRIES	3

enum {
	MTDSWAP_SCANNED_CLEAN,
	MTDSWAP_SCANNED_DIRTY,
	MTDSWAP_SCANNED_BITFLIP,
	MTDSWAP_SCANNED_BAD,
};

/*
 * In the worst case mtdswap_writesect() has allocated the last clean
 * page from the current block and is then pre-empted by the GC
 * thread. The thread can consume a full erase block when moving a
 * block.
 */
#define MIN_SPARE_EBLOCKS	2
#define MIN_ERASE_BLOCKS	(MIN_SPARE_EBLOCKS + 1)

#define TREE_ROOT(d, name) (&d->trees[MTDSWAP_ ## name].root)
#define TREE_EMPTY(d, name) (TREE_ROOT(d, name)->rb_node == NULL)
#define TREE_NONEMPTY(d, name) (!TREE_EMPTY(d, name))
#define TREE_COUNT(d, name) (d->trees[MTDSWAP_ ## name].count)

#define MTDSWAP_MBD_TO_MTDSWAP(dev) ((struct mtdswap_dev *)dev->priv)

static char partitions[128] = "";
module_param_string(partitions, partitions, sizeof(partitions), 0444);
MODULE_PARM_DESC(partitions, "MTD partition numbers to use as swap "
		"partitions=\"1,3,5\"");

static unsigned int spare_eblocks = 10;
module_param(spare_eblocks, uint, 0444);
MODULE_PARM_DESC(spare_eblocks, "Percentage of spare erase blocks for "
		"garbage collection (default 10%)");

static bool header; /* false */
module_param(header, bool, 0444);
MODULE_PARM_DESC(header,
		"Include builtin swap header (default 0, without header)");

static int mtdswap_gc(struct mtdswap_dev *d, unsigned int background);

static loff_t mtdswap_eb_offset(struct mtdswap_dev *d, struct swap_eb *eb)
{
	return (loff_t)(eb - d->eb_data) * d->mtd->erasesize;
}

static void mtdswap_eb_detach(struct mtdswap_dev *d, struct swap_eb *eb)
{
	unsigned int oldidx;
	struct mtdswap_tree *tp;

	if (eb->root) {
		tp = container_of(eb->root, struct mtdswap_tree, root);
		oldidx = tp - &d->trees[0];

		d->trees[oldidx].count--;
		rb_erase(&eb->rb, eb->root);
	}
}

static void __mtdswap_rb_add(struct rb_root *root, struct swap_eb *eb)
{
	struct rb_node **p, *parent = NULL;
	struct swap_eb *cur;

	p = &root->rb_node;
	while (*p) {
		parent = *p;
		cur = rb_entry(parent, struct swap_eb, rb);
		if (eb->erase_count > cur->erase_count)
			p = &(*p)->rb_right;
		else
			p = &(*p)->rb_left;
	}

	rb_link_node(&eb->rb, parent, p);
	rb_insert_color(&eb->rb, root);
}

static void mtdswap_rb_add(struct mtdswap_dev *d, struct swap_eb *eb, int idx)
{
	struct rb_root *root;

	if (eb->root == &d->trees[idx].root)
		return;

	mtdswap_eb_detach(d, eb);
	root = &d->trees[idx].root;
	__mtdswap_rb_add(root, eb);
	eb->root = root;
	d->trees[idx].count++;
}

static struct rb_node *mtdswap_rb_index(struct rb_root *root, unsigned int idx)
{
	struct rb_node *p;
	unsigned int i;

	p = rb_first(root);
	i = 0;
	while (i < idx && p) {
		p = rb_next(p);
		i++;
	}

	return p;
}

static int mtdswap_handle_badblock(struct mtdswap_dev *d, struct swap_eb *eb)
{
	int ret;
	loff_t offset;

	d->spare_eblks--;
	eb->flags |= EBLOCK_BAD;
	mtdswap_eb_detach(d, eb);
	eb->root = NULL;

	/* badblocks not supported */
275
	if (!mtd_can_have_bb(d->mtd))
276 277 278 279
		return 1;

	offset = mtdswap_eb_offset(d, eb);
	dev_warn(d->dev, "Marking bad block at %08llx\n", offset);
280
	ret = mtd_block_markbad(d->mtd, offset);
281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312

	if (ret) {
		dev_warn(d->dev, "Mark block bad failed for block at %08llx "
			"error %d\n", offset, ret);
		return ret;
	}

	return 1;

}

static int mtdswap_handle_write_error(struct mtdswap_dev *d, struct swap_eb *eb)
{
	unsigned int marked = eb->flags & EBLOCK_FAILED;
	struct swap_eb *curr_write = d->curr_write;

	eb->flags |= EBLOCK_FAILED;
	if (curr_write == eb) {
		d->curr_write = NULL;

		if (!marked && d->curr_write_pos != 0) {
			mtdswap_rb_add(d, eb, MTDSWAP_FAILING);
			return 0;
		}
	}

	return mtdswap_handle_badblock(d, eb);
}

static int mtdswap_read_oob(struct mtdswap_dev *d, loff_t from,
			struct mtd_oob_ops *ops)
{
313
	int ret = mtd_read_oob(d->mtd, from, ops);
314

315
	if (mtd_is_bitflip(ret))
316 317 318 319 320 321 322 323 324 325
		return ret;

	if (ret) {
		dev_warn(d->dev, "Read OOB failed %d for block at %08llx\n",
			ret, from);
		return ret;
	}

	if (ops->oobretlen < ops->ooblen) {
		dev_warn(d->dev, "Read OOB return short read (%zd bytes not "
326
			"%zd) for block at %08llx\n",
327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343
			ops->oobretlen, ops->ooblen, from);
		return -EIO;
	}

	return 0;
}

static int mtdswap_read_markers(struct mtdswap_dev *d, struct swap_eb *eb)
{
	struct mtdswap_oobdata *data, *data2;
	int ret;
	loff_t offset;
	struct mtd_oob_ops ops;

	offset = mtdswap_eb_offset(d, eb);

	/* Check first if the block is bad. */
344
	if (mtd_can_have_bb(d->mtd) && mtd_block_isbad(d->mtd, offset))
345 346
		return MTDSWAP_SCANNED_BAD;

347
	ops.ooblen = 2 * d->mtd->oobavail;
348 349 350
	ops.oobbuf = d->oob_buf;
	ops.ooboffs = 0;
	ops.datbuf = NULL;
351
	ops.mode = MTD_OPS_AUTO_OOB;
352 353 354

	ret = mtdswap_read_oob(d, offset, &ops);

355
	if (ret && !mtd_is_bitflip(ret))
356 357 358 359
		return ret;

	data = (struct mtdswap_oobdata *)d->oob_buf;
	data2 = (struct mtdswap_oobdata *)
360
		(d->oob_buf + d->mtd->oobavail);
361 362 363

	if (le16_to_cpu(data->magic) == MTDSWAP_MAGIC_CLEAN) {
		eb->erase_count = le32_to_cpu(data->count);
364
		if (mtd_is_bitflip(ret))
365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389
			ret = MTDSWAP_SCANNED_BITFLIP;
		else {
			if (le16_to_cpu(data2->magic) == MTDSWAP_MAGIC_DIRTY)
				ret = MTDSWAP_SCANNED_DIRTY;
			else
				ret = MTDSWAP_SCANNED_CLEAN;
		}
	} else {
		eb->flags |= EBLOCK_NOMAGIC;
		ret = MTDSWAP_SCANNED_DIRTY;
	}

	return ret;
}

static int mtdswap_write_marker(struct mtdswap_dev *d, struct swap_eb *eb,
				u16 marker)
{
	struct mtdswap_oobdata n;
	int ret;
	loff_t offset;
	struct mtd_oob_ops ops;

	ops.ooboffs = 0;
	ops.oobbuf = (uint8_t *)&n;
390
	ops.mode = MTD_OPS_AUTO_OOB;
391 392 393 394 395 396 397 398 399 400 401 402 403
	ops.datbuf = NULL;

	if (marker == MTDSWAP_TYPE_CLEAN) {
		n.magic = cpu_to_le16(MTDSWAP_MAGIC_CLEAN);
		n.count = cpu_to_le32(eb->erase_count);
		ops.ooblen = MTDSWAP_OOBSIZE;
		offset = mtdswap_eb_offset(d, eb);
	} else {
		n.magic = cpu_to_le16(MTDSWAP_MAGIC_DIRTY);
		ops.ooblen = sizeof(n.magic);
		offset = mtdswap_eb_offset(d, eb) + d->mtd->writesize;
	}

404
	ret = mtd_write_oob(d->mtd, offset, &ops);
405 406 407 408

	if (ret) {
		dev_warn(d->dev, "Write OOB failed for block at %08llx "
			"error %d\n", offset, ret);
409
		if (ret == -EIO || mtd_is_eccerr(ret))
410 411 412 413 414 415
			mtdswap_handle_write_error(d, eb);
		return ret;
	}

	if (ops.oobretlen != ops.ooblen) {
		dev_warn(d->dev, "Short OOB write for block at %08llx: "
416
			"%zd not %zd\n",
417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554
			offset, ops.oobretlen, ops.ooblen);
		return ret;
	}

	return 0;
}

/*
 * Are there any erase blocks without MAGIC_CLEAN header, presumably
 * because power was cut off after erase but before header write? We
 * need to guestimate the erase count.
 */
static void mtdswap_check_counts(struct mtdswap_dev *d)
{
	struct rb_root hist_root = RB_ROOT;
	struct rb_node *medrb;
	struct swap_eb *eb;
	unsigned int i, cnt, median;

	cnt = 0;
	for (i = 0; i < d->eblks; i++) {
		eb = d->eb_data + i;

		if (eb->flags & (EBLOCK_NOMAGIC | EBLOCK_BAD | EBLOCK_READERR))
			continue;

		__mtdswap_rb_add(&hist_root, eb);
		cnt++;
	}

	if (cnt == 0)
		return;

	medrb = mtdswap_rb_index(&hist_root, cnt / 2);
	median = rb_entry(medrb, struct swap_eb, rb)->erase_count;

	d->max_erase_count = MTDSWAP_ECNT_MAX(&hist_root);

	for (i = 0; i < d->eblks; i++) {
		eb = d->eb_data + i;

		if (eb->flags & (EBLOCK_NOMAGIC | EBLOCK_READERR))
			eb->erase_count = median;

		if (eb->flags & (EBLOCK_NOMAGIC | EBLOCK_BAD | EBLOCK_READERR))
			continue;

		rb_erase(&eb->rb, &hist_root);
	}
}

static void mtdswap_scan_eblks(struct mtdswap_dev *d)
{
	int status;
	unsigned int i, idx;
	struct swap_eb *eb;

	for (i = 0; i < d->eblks; i++) {
		eb = d->eb_data + i;

		status = mtdswap_read_markers(d, eb);
		if (status < 0)
			eb->flags |= EBLOCK_READERR;
		else if (status == MTDSWAP_SCANNED_BAD) {
			eb->flags |= EBLOCK_BAD;
			continue;
		}

		switch (status) {
		case MTDSWAP_SCANNED_CLEAN:
			idx = MTDSWAP_CLEAN;
			break;
		case MTDSWAP_SCANNED_DIRTY:
		case MTDSWAP_SCANNED_BITFLIP:
			idx = MTDSWAP_DIRTY;
			break;
		default:
			idx = MTDSWAP_FAILING;
		}

		eb->flags |= (idx << EBLOCK_IDX_SHIFT);
	}

	mtdswap_check_counts(d);

	for (i = 0; i < d->eblks; i++) {
		eb = d->eb_data + i;

		if (eb->flags & EBLOCK_BAD)
			continue;

		idx = eb->flags >> EBLOCK_IDX_SHIFT;
		mtdswap_rb_add(d, eb, idx);
	}
}

/*
 * Place eblk into a tree corresponding to its number of active blocks
 * it contains.
 */
static void mtdswap_store_eb(struct mtdswap_dev *d, struct swap_eb *eb)
{
	unsigned int weight = eb->active_count;
	unsigned int maxweight = d->pages_per_eblk;

	if (eb == d->curr_write)
		return;

	if (eb->flags & EBLOCK_BITFLIP)
		mtdswap_rb_add(d, eb, MTDSWAP_BITFLIP);
	else if (eb->flags & (EBLOCK_READERR | EBLOCK_FAILED))
		mtdswap_rb_add(d, eb, MTDSWAP_FAILING);
	if (weight == maxweight)
		mtdswap_rb_add(d, eb, MTDSWAP_USED);
	else if (weight == 0)
		mtdswap_rb_add(d, eb, MTDSWAP_DIRTY);
	else if (weight > (maxweight/2))
		mtdswap_rb_add(d, eb, MTDSWAP_LOWFRAG);
	else
		mtdswap_rb_add(d, eb, MTDSWAP_HIFRAG);
}

static int mtdswap_erase_block(struct mtdswap_dev *d, struct swap_eb *eb)
{
	struct mtd_info *mtd = d->mtd;
	struct erase_info erase;
	unsigned int retries = 0;
	int ret;

	eb->erase_count++;
	if (eb->erase_count > d->max_erase_count)
		d->max_erase_count = eb->erase_count;

retry:
	memset(&erase, 0, sizeof(struct erase_info));
	erase.addr	= mtdswap_eb_offset(d, eb);
	erase.len	= mtd->erasesize;

555
	ret = mtd_erase(mtd, &erase);
556
	if (ret) {
557
		if (retries++ < MTDSWAP_ERASE_RETRIES) {
558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594
			dev_warn(d->dev,
				"erase of erase block %#llx on %s failed",
				erase.addr, mtd->name);
			yield();
			goto retry;
		}

		dev_err(d->dev, "Cannot erase erase block %#llx on %s\n",
			erase.addr, mtd->name);

		mtdswap_handle_badblock(d, eb);
		return -EIO;
	}

	return 0;
}

static int mtdswap_map_free_block(struct mtdswap_dev *d, unsigned int page,
				unsigned int *block)
{
	int ret;
	struct swap_eb *old_eb = d->curr_write;
	struct rb_root *clean_root;
	struct swap_eb *eb;

	if (old_eb == NULL || d->curr_write_pos >= d->pages_per_eblk) {
		do {
			if (TREE_EMPTY(d, CLEAN))
				return -ENOSPC;

			clean_root = TREE_ROOT(d, CLEAN);
			eb = rb_entry(rb_first(clean_root), struct swap_eb, rb);
			rb_erase(&eb->rb, clean_root);
			eb->root = NULL;
			TREE_COUNT(d, CLEAN)--;

			ret = mtdswap_write_marker(d, eb, MTDSWAP_TYPE_DIRTY);
595
		} while (ret == -EIO || mtd_is_eccerr(ret));
596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644

		if (ret)
			return ret;

		d->curr_write_pos = 0;
		d->curr_write = eb;
		if (old_eb)
			mtdswap_store_eb(d, old_eb);
	}

	*block = (d->curr_write - d->eb_data) * d->pages_per_eblk +
		d->curr_write_pos;

	d->curr_write->active_count++;
	d->revmap[*block] = page;
	d->curr_write_pos++;

	return 0;
}

static unsigned int mtdswap_free_page_cnt(struct mtdswap_dev *d)
{
	return TREE_COUNT(d, CLEAN) * d->pages_per_eblk +
		d->pages_per_eblk - d->curr_write_pos;
}

static unsigned int mtdswap_enough_free_pages(struct mtdswap_dev *d)
{
	return mtdswap_free_page_cnt(d) > d->pages_per_eblk;
}

static int mtdswap_write_block(struct mtdswap_dev *d, char *buf,
			unsigned int page, unsigned int *bp, int gc_context)
{
	struct mtd_info *mtd = d->mtd;
	struct swap_eb *eb;
	size_t retlen;
	loff_t writepos;
	int ret;

retry:
	if (!gc_context)
		while (!mtdswap_enough_free_pages(d))
			if (mtdswap_gc(d, 0) > 0)
				return -ENOSPC;

	ret = mtdswap_map_free_block(d, page, bp);
	eb = d->eb_data + (*bp / d->pages_per_eblk);

645
	if (ret == -EIO || mtd_is_eccerr(ret)) {
646 647 648 649 650 651 652 653 654 655
		d->curr_write = NULL;
		eb->active_count--;
		d->revmap[*bp] = PAGE_UNDEF;
		goto retry;
	}

	if (ret < 0)
		return ret;

	writepos = (loff_t)*bp << PAGE_SHIFT;
656
	ret =  mtd_write(mtd, writepos, PAGE_SIZE, &retlen, buf);
657
	if (ret == -EIO || mtd_is_eccerr(ret)) {
658 659 660 661 662 663 664 665
		d->curr_write_pos--;
		eb->active_count--;
		d->revmap[*bp] = PAGE_UNDEF;
		mtdswap_handle_write_error(d, eb);
		goto retry;
	}

	if (ret < 0) {
666
		dev_err(d->dev, "Write to MTD device failed: %d (%zd written)",
667 668 669 670 671
			ret, retlen);
		goto err;
	}

	if (retlen != PAGE_SIZE) {
672
		dev_err(d->dev, "Short write to MTD device: %zd written",
673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702
			retlen);
		ret = -EIO;
		goto err;
	}

	return ret;

err:
	d->curr_write_pos--;
	eb->active_count--;
	d->revmap[*bp] = PAGE_UNDEF;

	return ret;
}

static int mtdswap_move_block(struct mtdswap_dev *d, unsigned int oldblock,
		unsigned int *newblock)
{
	struct mtd_info *mtd = d->mtd;
	struct swap_eb *eb, *oldeb;
	int ret;
	size_t retlen;
	unsigned int page, retries;
	loff_t readpos;

	page = d->revmap[oldblock];
	readpos = (loff_t) oldblock << PAGE_SHIFT;
	retries = 0;

retry:
703
	ret = mtd_read(mtd, readpos, PAGE_SIZE, &retlen, d->page_buf);
704

705
	if (ret < 0 && !mtd_is_bitflip(ret)) {
706 707 708 709 710 711 712 713 714 715 716 717 718
		oldeb = d->eb_data + oldblock / d->pages_per_eblk;
		oldeb->flags |= EBLOCK_READERR;

		dev_err(d->dev, "Read Error: %d (block %u)\n", ret,
			oldblock);
		retries++;
		if (retries < MTDSWAP_IO_RETRIES)
			goto retry;

		goto read_error;
	}

	if (retlen != PAGE_SIZE) {
719
		dev_err(d->dev, "Short read: %zd (block %u)\n", retlen,
720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773
		       oldblock);
		ret = -EIO;
		goto read_error;
	}

	ret = mtdswap_write_block(d, d->page_buf, page, newblock, 1);
	if (ret < 0) {
		d->page_data[page] = BLOCK_ERROR;
		dev_err(d->dev, "Write error: %d\n", ret);
		return ret;
	}

	eb = d->eb_data + *newblock / d->pages_per_eblk;
	d->page_data[page] = *newblock;
	d->revmap[oldblock] = PAGE_UNDEF;
	eb = d->eb_data + oldblock / d->pages_per_eblk;
	eb->active_count--;

	return 0;

read_error:
	d->page_data[page] = BLOCK_ERROR;
	d->revmap[oldblock] = PAGE_UNDEF;
	return ret;
}

static int mtdswap_gc_eblock(struct mtdswap_dev *d, struct swap_eb *eb)
{
	unsigned int i, block, eblk_base, newblock;
	int ret, errcode;

	errcode = 0;
	eblk_base = (eb - d->eb_data) * d->pages_per_eblk;

	for (i = 0; i < d->pages_per_eblk; i++) {
		if (d->spare_eblks < MIN_SPARE_EBLOCKS)
			return -ENOSPC;

		block = eblk_base + i;
		if (d->revmap[block] == PAGE_UNDEF)
			continue;

		ret = mtdswap_move_block(d, block, &newblock);
		if (ret < 0 && !errcode)
			errcode = ret;
	}

	return errcode;
}

static int __mtdswap_choose_gc_tree(struct mtdswap_dev *d)
{
	int idx, stopat;

774
	if (TREE_COUNT(d, CLEAN) < LOW_FRAG_GC_THRESHOLD)
775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816
		stopat = MTDSWAP_LOWFRAG;
	else
		stopat = MTDSWAP_HIFRAG;

	for (idx = MTDSWAP_BITFLIP; idx >= stopat; idx--)
		if (d->trees[idx].root.rb_node != NULL)
			return idx;

	return -1;
}

static int mtdswap_wlfreq(unsigned int maxdiff)
{
	unsigned int h, x, y, dist, base;

	/*
	 * Calculate linear ramp down from f1 to f2 when maxdiff goes from
	 * MAX_ERASE_DIFF to MAX_ERASE_DIFF + COLLECT_NONDIRTY_BASE.  Similar
	 * to triangle with height f1 - f1 and width COLLECT_NONDIRTY_BASE.
	 */

	dist = maxdiff - MAX_ERASE_DIFF;
	if (dist > COLLECT_NONDIRTY_BASE)
		dist = COLLECT_NONDIRTY_BASE;

	/*
	 * Modelling the slop as right angular triangle with base
	 * COLLECT_NONDIRTY_BASE and height freq1 - freq2. The ratio y/x is
	 * equal to the ratio h/base.
	 */
	h = COLLECT_NONDIRTY_FREQ1 - COLLECT_NONDIRTY_FREQ2;
	base = COLLECT_NONDIRTY_BASE;

	x = dist - base;
	y = (x * h + base / 2) / base;

	return COLLECT_NONDIRTY_FREQ2 + y;
}

static int mtdswap_choose_wl_tree(struct mtdswap_dev *d)
{
	static unsigned int pick_cnt;
817
	unsigned int i, idx = -1, wear, max;
818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897
	struct rb_root *root;

	max = 0;
	for (i = 0; i <= MTDSWAP_DIRTY; i++) {
		root = &d->trees[i].root;
		if (root->rb_node == NULL)
			continue;

		wear = d->max_erase_count - MTDSWAP_ECNT_MIN(root);
		if (wear > max) {
			max = wear;
			idx = i;
		}
	}

	if (max > MAX_ERASE_DIFF && pick_cnt >= mtdswap_wlfreq(max) - 1) {
		pick_cnt = 0;
		return idx;
	}

	pick_cnt++;
	return -1;
}

static int mtdswap_choose_gc_tree(struct mtdswap_dev *d,
				unsigned int background)
{
	int idx;

	if (TREE_NONEMPTY(d, FAILING) &&
		(background || (TREE_EMPTY(d, CLEAN) && TREE_EMPTY(d, DIRTY))))
		return MTDSWAP_FAILING;

	idx = mtdswap_choose_wl_tree(d);
	if (idx >= MTDSWAP_CLEAN)
		return idx;

	return __mtdswap_choose_gc_tree(d);
}

static struct swap_eb *mtdswap_pick_gc_eblk(struct mtdswap_dev *d,
					unsigned int background)
{
	struct rb_root *rp = NULL;
	struct swap_eb *eb = NULL;
	int idx;

	if (background && TREE_COUNT(d, CLEAN) > CLEAN_BLOCK_THRESHOLD &&
		TREE_EMPTY(d, DIRTY) && TREE_EMPTY(d, FAILING))
		return NULL;

	idx = mtdswap_choose_gc_tree(d, background);
	if (idx < 0)
		return NULL;

	rp = &d->trees[idx].root;
	eb = rb_entry(rb_first(rp), struct swap_eb, rb);

	rb_erase(&eb->rb, rp);
	eb->root = NULL;
	d->trees[idx].count--;
	return eb;
}

static unsigned int mtdswap_test_patt(unsigned int i)
{
	return i % 2 ? 0x55555555 : 0xAAAAAAAA;
}

static unsigned int mtdswap_eblk_passes(struct mtdswap_dev *d,
					struct swap_eb *eb)
{
	struct mtd_info *mtd = d->mtd;
	unsigned int test, i, j, patt, mtd_pages;
	loff_t base, pos;
	unsigned int *p1 = (unsigned int *)d->page_buf;
	unsigned char *p2 = (unsigned char *)d->oob_buf;
	struct mtd_oob_ops ops;
	int ret;

898
	ops.mode = MTD_OPS_AUTO_OOB;
899
	ops.len = mtd->writesize;
900
	ops.ooblen = mtd->oobavail;
901 902 903 904 905 906 907 908 909 910 911
	ops.ooboffs = 0;
	ops.datbuf = d->page_buf;
	ops.oobbuf = d->oob_buf;
	base = mtdswap_eb_offset(d, eb);
	mtd_pages = d->pages_per_eblk * PAGE_SIZE / mtd->writesize;

	for (test = 0; test < 2; test++) {
		pos = base;
		for (i = 0; i < mtd_pages; i++) {
			patt = mtdswap_test_patt(test + i);
			memset(d->page_buf, patt, mtd->writesize);
912
			memset(d->oob_buf, patt, mtd->oobavail);
913
			ret = mtd_write_oob(mtd, pos, &ops);
914 915 916 917 918 919 920 921
			if (ret)
				goto error;

			pos += mtd->writesize;
		}

		pos = base;
		for (i = 0; i < mtd_pages; i++) {
922
			ret = mtd_read_oob(mtd, pos, &ops);
923 924 925 926 927 928 929 930
			if (ret)
				goto error;

			patt = mtdswap_test_patt(test + i);
			for (j = 0; j < mtd->writesize/sizeof(int); j++)
				if (p1[j] != patt)
					goto error;

931
			for (j = 0; j < mtd->oobavail; j++)
932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982
				if (p2[j] != (unsigned char)patt)
					goto error;

			pos += mtd->writesize;
		}

		ret = mtdswap_erase_block(d, eb);
		if (ret)
			goto error;
	}

	eb->flags &= ~EBLOCK_READERR;
	return 1;

error:
	mtdswap_handle_badblock(d, eb);
	return 0;
}

static int mtdswap_gc(struct mtdswap_dev *d, unsigned int background)
{
	struct swap_eb *eb;
	int ret;

	if (d->spare_eblks < MIN_SPARE_EBLOCKS)
		return 1;

	eb = mtdswap_pick_gc_eblk(d, background);
	if (!eb)
		return 1;

	ret = mtdswap_gc_eblock(d, eb);
	if (ret == -ENOSPC)
		return 1;

	if (eb->flags & EBLOCK_FAILED) {
		mtdswap_handle_badblock(d, eb);
		return 0;
	}

	eb->flags &= ~EBLOCK_BITFLIP;
	ret = mtdswap_erase_block(d, eb);
	if ((eb->flags & EBLOCK_READERR) &&
		(ret || !mtdswap_eblk_passes(d, eb)))
		return 0;

	if (ret == 0)
		ret = mtdswap_write_marker(d, eb, MTDSWAP_TYPE_CLEAN);

	if (ret == 0)
		mtdswap_rb_add(d, eb, MTDSWAP_CLEAN);
983
	else if (ret != -EIO && !mtd_is_eccerr(ret))
984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013
		mtdswap_rb_add(d, eb, MTDSWAP_DIRTY);

	return 0;
}

static void mtdswap_background(struct mtd_blktrans_dev *dev)
{
	struct mtdswap_dev *d = MTDSWAP_MBD_TO_MTDSWAP(dev);
	int ret;

	while (1) {
		ret = mtdswap_gc(d, 1);
		if (ret || mtd_blktrans_cease_background(dev))
			return;
	}
}

static void mtdswap_cleanup(struct mtdswap_dev *d)
{
	vfree(d->eb_data);
	vfree(d->revmap);
	vfree(d->page_data);
	kfree(d->oob_buf);
	kfree(d->page_buf);
}

static int mtdswap_flush(struct mtd_blktrans_dev *dev)
{
	struct mtdswap_dev *d = MTDSWAP_MBD_TO_MTDSWAP(dev);

1014
	mtd_sync(d->mtd);
1015 1016 1017 1018 1019 1020 1021 1022 1023 1024
	return 0;
}

static unsigned int mtdswap_badblocks(struct mtd_info *mtd, uint64_t size)
{
	loff_t offset;
	unsigned int badcnt;

	badcnt = 0;

1025
	if (mtd_can_have_bb(mtd))
1026
		for (offset = 0; offset < size; offset += mtd->erasesize)
1027
			if (mtd_block_isbad(mtd, offset))
1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126
				badcnt++;

	return badcnt;
}

static int mtdswap_writesect(struct mtd_blktrans_dev *dev,
			unsigned long page, char *buf)
{
	struct mtdswap_dev *d = MTDSWAP_MBD_TO_MTDSWAP(dev);
	unsigned int newblock, mapped;
	struct swap_eb *eb;
	int ret;

	d->sect_write_count++;

	if (d->spare_eblks < MIN_SPARE_EBLOCKS)
		return -ENOSPC;

	if (header) {
		/* Ignore writes to the header page */
		if (unlikely(page == 0))
			return 0;

		page--;
	}

	mapped = d->page_data[page];
	if (mapped <= BLOCK_MAX) {
		eb = d->eb_data + (mapped / d->pages_per_eblk);
		eb->active_count--;
		mtdswap_store_eb(d, eb);
		d->page_data[page] = BLOCK_UNDEF;
		d->revmap[mapped] = PAGE_UNDEF;
	}

	ret = mtdswap_write_block(d, buf, page, &newblock, 0);
	d->mtd_write_count++;

	if (ret < 0)
		return ret;

	eb = d->eb_data + (newblock / d->pages_per_eblk);
	d->page_data[page] = newblock;

	return 0;
}

/* Provide a dummy swap header for the kernel */
static int mtdswap_auto_header(struct mtdswap_dev *d, char *buf)
{
	union swap_header *hd = (union swap_header *)(buf);

	memset(buf, 0, PAGE_SIZE - 10);

	hd->info.version = 1;
	hd->info.last_page = d->mbd_dev->size - 1;
	hd->info.nr_badpages = 0;

	memcpy(buf + PAGE_SIZE - 10, "SWAPSPACE2", 10);

	return 0;
}

static int mtdswap_readsect(struct mtd_blktrans_dev *dev,
			unsigned long page, char *buf)
{
	struct mtdswap_dev *d = MTDSWAP_MBD_TO_MTDSWAP(dev);
	struct mtd_info *mtd = d->mtd;
	unsigned int realblock, retries;
	loff_t readpos;
	struct swap_eb *eb;
	size_t retlen;
	int ret;

	d->sect_read_count++;

	if (header) {
		if (unlikely(page == 0))
			return mtdswap_auto_header(d, buf);

		page--;
	}

	realblock = d->page_data[page];
	if (realblock > BLOCK_MAX) {
		memset(buf, 0x0, PAGE_SIZE);
		if (realblock == BLOCK_UNDEF)
			return 0;
		else
			return -EIO;
	}

	eb = d->eb_data + (realblock / d->pages_per_eblk);
	BUG_ON(d->revmap[realblock] == PAGE_UNDEF);

	readpos = (loff_t)realblock << PAGE_SHIFT;
	retries = 0;

retry:
1127
	ret = mtd_read(mtd, readpos, PAGE_SIZE, &retlen, buf);
1128 1129

	d->mtd_read_count++;
1130
	if (mtd_is_bitflip(ret)) {
1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147
		eb->flags |= EBLOCK_BITFLIP;
		mtdswap_rb_add(d, eb, MTDSWAP_BITFLIP);
		ret = 0;
	}

	if (ret < 0) {
		dev_err(d->dev, "Read error %d\n", ret);
		eb->flags |= EBLOCK_READERR;
		mtdswap_rb_add(d, eb, MTDSWAP_FAILING);
		retries++;
		if (retries < MTDSWAP_IO_RETRIES)
			goto retry;

		return ret;
	}

	if (retlen != PAGE_SIZE) {
1148
		dev_err(d->dev, "Short read %zd\n", retlen);
1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191
		return -EIO;
	}

	return 0;
}

static int mtdswap_discard(struct mtd_blktrans_dev *dev, unsigned long first,
			unsigned nr_pages)
{
	struct mtdswap_dev *d = MTDSWAP_MBD_TO_MTDSWAP(dev);
	unsigned long page;
	struct swap_eb *eb;
	unsigned int mapped;

	d->discard_count++;

	for (page = first; page < first + nr_pages; page++) {
		mapped = d->page_data[page];
		if (mapped <= BLOCK_MAX) {
			eb = d->eb_data + (mapped / d->pages_per_eblk);
			eb->active_count--;
			mtdswap_store_eb(d, eb);
			d->page_data[page] = BLOCK_UNDEF;
			d->revmap[mapped] = PAGE_UNDEF;
			d->discard_page_count++;
		} else if (mapped == BLOCK_ERROR) {
			d->page_data[page] = BLOCK_UNDEF;
			d->discard_page_count++;
		}
	}

	return 0;
}

static int mtdswap_show(struct seq_file *s, void *data)
{
	struct mtdswap_dev *d = (struct mtdswap_dev *) s->private;
	unsigned long sum;
	unsigned int count[MTDSWAP_TREE_CNT];
	unsigned int min[MTDSWAP_TREE_CNT];
	unsigned int max[MTDSWAP_TREE_CNT];
	unsigned int i, cw = 0, cwp = 0, cwecount = 0, bb_cnt, mapped, pages;
	uint64_t use_size;
1192 1193 1194
	static const char * const name[] = {
		"clean", "used", "low", "high", "dirty", "bitflip", "failing"
	};
1195 1196 1197 1198 1199 1200 1201 1202

	mutex_lock(&d->mbd_dev->lock);

	for (i = 0; i < MTDSWAP_TREE_CNT; i++) {
		struct rb_root *root = &d->trees[i].root;

		if (root->rb_node) {
			count[i] = d->trees[i].count;
1203 1204
			min[i] = MTDSWAP_ECNT_MIN(root);
			max[i] = MTDSWAP_ECNT_MAX(root);
1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252
		} else
			count[i] = 0;
	}

	if (d->curr_write) {
		cw = 1;
		cwp = d->curr_write_pos;
		cwecount = d->curr_write->erase_count;
	}

	sum = 0;
	for (i = 0; i < d->eblks; i++)
		sum += d->eb_data[i].erase_count;

	use_size = (uint64_t)d->eblks * d->mtd->erasesize;
	bb_cnt = mtdswap_badblocks(d->mtd, use_size);

	mapped = 0;
	pages = d->mbd_dev->size;
	for (i = 0; i < pages; i++)
		if (d->page_data[i] != BLOCK_UNDEF)
			mapped++;

	mutex_unlock(&d->mbd_dev->lock);

	for (i = 0; i < MTDSWAP_TREE_CNT; i++) {
		if (!count[i])
			continue;

		if (min[i] != max[i])
			seq_printf(s, "%s:\t%5d erase blocks, erased min %d, "
				"max %d times\n",
				name[i], count[i], min[i], max[i]);
		else
			seq_printf(s, "%s:\t%5d erase blocks, all erased %d "
				"times\n", name[i], count[i], min[i]);
	}

	if (bb_cnt)
		seq_printf(s, "bad:\t%5u erase blocks\n", bb_cnt);

	if (cw)
		seq_printf(s, "current erase block: %u pages used, %u free, "
			"erased %u times\n",
			cwp, d->pages_per_eblk - cwp, cwecount);

	seq_printf(s, "total erasures: %lu\n", sum);

1253
	seq_puts(s, "\n");
1254 1255 1256 1257 1258 1259 1260 1261

	seq_printf(s, "mtdswap_readsect count: %llu\n", d->sect_read_count);
	seq_printf(s, "mtdswap_writesect count: %llu\n", d->sect_write_count);
	seq_printf(s, "mtdswap_discard count: %llu\n", d->discard_count);
	seq_printf(s, "mtd read count: %llu\n", d->mtd_read_count);
	seq_printf(s, "mtd write count: %llu\n", d->mtd_write_count);
	seq_printf(s, "discarded pages count: %llu\n", d->discard_page_count);

1262
	seq_puts(s, "\n");
1263
	seq_printf(s, "total pages: %u\n", pages);
1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282
	seq_printf(s, "pages mapped: %u\n", mapped);

	return 0;
}

static int mtdswap_open(struct inode *inode, struct file *file)
{
	return single_open(file, mtdswap_show, inode->i_private);
}

static const struct file_operations mtdswap_fops = {
	.open		= mtdswap_open,
	.read		= seq_read,
	.llseek		= seq_lseek,
	.release	= single_release,
};

static int mtdswap_add_debugfs(struct mtdswap_dev *d)
{
1283
	struct dentry *root = d->mtd->dbg.dfs_dir;
1284 1285
	struct dentry *dent;

1286
	if (!IS_ENABLED(CONFIG_DEBUG_FS))
1287 1288
		return 0;

1289
	if (IS_ERR_OR_NULL(root))
1290 1291
		return -1;

1292
	dent = debugfs_create_file("mtdswap_stats", S_IRUSR, root, d,
1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319
				&mtdswap_fops);
	if (!dent) {
		dev_err(d->dev, "debugfs_create_file failed\n");
		return -1;
	}

	return 0;
}

static int mtdswap_init(struct mtdswap_dev *d, unsigned int eblocks,
			unsigned int spare_cnt)
{
	struct mtd_info *mtd = d->mbd_dev->mtd;
	unsigned int i, eblk_bytes, pages, blocks;
	int ret = -ENOMEM;

	d->mtd = mtd;
	d->eblks = eblocks;
	d->spare_eblks = spare_cnt;
	d->pages_per_eblk = mtd->erasesize >> PAGE_SHIFT;

	pages = d->mbd_dev->size;
	blocks = eblocks * d->pages_per_eblk;

	for (i = 0; i < MTDSWAP_TREE_CNT; i++)
		d->trees[i].root = RB_ROOT;

1320
	d->page_data = vmalloc(array_size(pages, sizeof(int)));
1321 1322 1323
	if (!d->page_data)
		goto page_data_fail;

1324
	d->revmap = vmalloc(array_size(blocks, sizeof(int)));
1325 1326 1327 1328
	if (!d->revmap)
		goto revmap_fail;

	eblk_bytes = sizeof(struct swap_eb)*d->eblks;
1329
	d->eb_data = vzalloc(eblk_bytes);
1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342
	if (!d->eb_data)
		goto eb_data_fail;

	for (i = 0; i < pages; i++)
		d->page_data[i] = BLOCK_UNDEF;

	for (i = 0; i < blocks; i++)
		d->revmap[i] = PAGE_UNDEF;

	d->page_buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
	if (!d->page_buf)
		goto page_buf_fail;

1343
	d->oob_buf = kmalloc_array(2, mtd->oobavail, GFP_KERNEL);
1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379
	if (!d->oob_buf)
		goto oob_buf_fail;

	mtdswap_scan_eblks(d);

	return 0;

oob_buf_fail:
	kfree(d->page_buf);
page_buf_fail:
	vfree(d->eb_data);
eb_data_fail:
	vfree(d->revmap);
revmap_fail:
	vfree(d->page_data);
page_data_fail:
	printk(KERN_ERR "%s: init failed (%d)\n", MTDSWAP_PREFIX, ret);
	return ret;
}

static void mtdswap_add_mtd(struct mtd_blktrans_ops *tr, struct mtd_info *mtd)
{
	struct mtdswap_dev *d;
	struct mtd_blktrans_dev *mbd_dev;
	char *parts;
	char *this_opt;
	unsigned long part;
	unsigned int eblocks, eavailable, bad_blocks, spare_cnt;
	uint64_t swap_size, use_size, size_limit;
	int ret;

	parts = &partitions[0];
	if (!*parts)
		return;

	while ((this_opt = strsep(&parts, ",")) != NULL) {
1380
		if (kstrtoul(this_opt, 0, &part) < 0)
1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401
			return;

		if (mtd->index == part)
			break;
	}

	if (mtd->index != part)
		return;

	if (mtd->erasesize < PAGE_SIZE || mtd->erasesize % PAGE_SIZE) {
		printk(KERN_ERR "%s: Erase size %u not multiple of PAGE_SIZE "
			"%lu\n", MTDSWAP_PREFIX, mtd->erasesize, PAGE_SIZE);
		return;
	}

	if (PAGE_SIZE % mtd->writesize || mtd->writesize > PAGE_SIZE) {
		printk(KERN_ERR "%s: PAGE_SIZE %lu not multiple of write size"
			" %u\n", MTDSWAP_PREFIX, PAGE_SIZE, mtd->writesize);
		return;
	}

1402
	if (!mtd->oobsize || mtd->oobavail < MTDSWAP_OOBSIZE) {
1403
		printk(KERN_ERR "%s: Not enough free bytes in OOB, "
1404
			"%d available, %zu needed.\n",
1405
			MTDSWAP_PREFIX, mtd->oobavail, MTDSWAP_OOBSIZE);
1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421
		return;
	}

	if (spare_eblocks > 100)
		spare_eblocks = 100;

	use_size = mtd->size;
	size_limit = (uint64_t) BLOCK_MAX * PAGE_SIZE;

	if (mtd->size > size_limit) {
		printk(KERN_WARNING "%s: Device too large. Limiting size to "
			"%llu bytes\n", MTDSWAP_PREFIX, size_limit);
		use_size = size_limit;
	}

	eblocks = mtd_div_by_eb(use_size, mtd);
1422
	use_size = (uint64_t)eblocks * mtd->erasesize;
1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535
	bad_blocks = mtdswap_badblocks(mtd, use_size);
	eavailable = eblocks - bad_blocks;

	if (eavailable < MIN_ERASE_BLOCKS) {
		printk(KERN_ERR "%s: Not enough erase blocks. %u available, "
			"%d needed\n", MTDSWAP_PREFIX, eavailable,
			MIN_ERASE_BLOCKS);
		return;
	}

	spare_cnt = div_u64((uint64_t)eavailable * spare_eblocks, 100);

	if (spare_cnt < MIN_SPARE_EBLOCKS)
		spare_cnt = MIN_SPARE_EBLOCKS;

	if (spare_cnt > eavailable - 1)
		spare_cnt = eavailable - 1;

	swap_size = (uint64_t)(eavailable - spare_cnt) * mtd->erasesize +
		(header ? PAGE_SIZE : 0);

	printk(KERN_INFO "%s: Enabling MTD swap on device %lu, size %llu KB, "
		"%u spare, %u bad blocks\n",
		MTDSWAP_PREFIX, part, swap_size / 1024, spare_cnt, bad_blocks);

	d = kzalloc(sizeof(struct mtdswap_dev), GFP_KERNEL);
	if (!d)
		return;

	mbd_dev = kzalloc(sizeof(struct mtd_blktrans_dev), GFP_KERNEL);
	if (!mbd_dev) {
		kfree(d);
		return;
	}

	d->mbd_dev = mbd_dev;
	mbd_dev->priv = d;

	mbd_dev->mtd = mtd;
	mbd_dev->devnum = mtd->index;
	mbd_dev->size = swap_size >> PAGE_SHIFT;
	mbd_dev->tr = tr;

	if (!(mtd->flags & MTD_WRITEABLE))
		mbd_dev->readonly = 1;

	if (mtdswap_init(d, eblocks, spare_cnt) < 0)
		goto init_failed;

	if (add_mtd_blktrans_dev(mbd_dev) < 0)
		goto cleanup;

	d->dev = disk_to_dev(mbd_dev->disk);

	ret = mtdswap_add_debugfs(d);
	if (ret < 0)
		goto debugfs_failed;

	return;

debugfs_failed:
	del_mtd_blktrans_dev(mbd_dev);

cleanup:
	mtdswap_cleanup(d);

init_failed:
	kfree(mbd_dev);
	kfree(d);
}

static void mtdswap_remove_dev(struct mtd_blktrans_dev *dev)
{
	struct mtdswap_dev *d = MTDSWAP_MBD_TO_MTDSWAP(dev);

	del_mtd_blktrans_dev(dev);
	mtdswap_cleanup(d);
	kfree(d);
}

static struct mtd_blktrans_ops mtdswap_ops = {
	.name		= "mtdswap",
	.major		= 0,
	.part_bits	= 0,
	.blksize	= PAGE_SIZE,
	.flush		= mtdswap_flush,
	.readsect	= mtdswap_readsect,
	.writesect	= mtdswap_writesect,
	.discard	= mtdswap_discard,
	.background	= mtdswap_background,
	.add_mtd	= mtdswap_add_mtd,
	.remove_dev	= mtdswap_remove_dev,
	.owner		= THIS_MODULE,
};

static int __init mtdswap_modinit(void)
{
	return register_mtd_blktrans(&mtdswap_ops);
}

static void __exit mtdswap_modexit(void)
{
	deregister_mtd_blktrans(&mtdswap_ops);
}

module_init(mtdswap_modinit);
module_exit(mtdswap_modexit);


MODULE_LICENSE("GPL");
MODULE_AUTHOR("Jarkko Lavinen <jarkko.lavinen@nokia.com>");
MODULE_DESCRIPTION("Block device access to an MTD suitable for using as "
		"swap space");