gennvm.c 10.2 KB
Newer Older
M
Matias Bjørling 已提交
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 53 54 55 56 57 58 59 60 61 62
/*
 * Copyright (C) 2015 Matias Bjorling <m@bjorling.me>
 *
 * 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; see the file COPYING.  If not, write to
 * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139,
 * USA.
 *
 * Implementation of a generic nvm manager for Open-Channel SSDs.
 */

#include "gennvm.h"

static void gennvm_blocks_free(struct nvm_dev *dev)
{
	struct gen_nvm *gn = dev->mp;
	struct gen_lun *lun;
	int i;

	gennvm_for_each_lun(gn, lun, i) {
		if (!lun->vlun.blocks)
			break;
		vfree(lun->vlun.blocks);
	}
}

static void gennvm_luns_free(struct nvm_dev *dev)
{
	struct gen_nvm *gn = dev->mp;

	kfree(gn->luns);
}

static int gennvm_luns_init(struct nvm_dev *dev, struct gen_nvm *gn)
{
	struct gen_lun *lun;
	int i;

	gn->luns = kcalloc(dev->nr_luns, sizeof(struct gen_lun), GFP_KERNEL);
	if (!gn->luns)
		return -ENOMEM;

	gennvm_for_each_lun(gn, lun, i) {
		spin_lock_init(&lun->vlun.lock);
		INIT_LIST_HEAD(&lun->free_list);
		INIT_LIST_HEAD(&lun->used_list);
		INIT_LIST_HEAD(&lun->bb_list);

		lun->reserved_blocks = 2; /* for GC only */
		lun->vlun.id = i;
		lun->vlun.lun_id = i % dev->luns_per_chnl;
		lun->vlun.chnl_id = i / dev->luns_per_chnl;
		lun->vlun.nr_free_blocks = dev->blks_per_lun;
63 64
		lun->vlun.nr_inuse_blocks = 0;
		lun->vlun.nr_bad_blocks = 0;
M
Matias Bjørling 已提交
65 66 67 68
	}
	return 0;
}

69
static int gennvm_block_bb(struct ppa_addr ppa, int nr_blocks, u8 *blks,
M
Matias Bjørling 已提交
70 71 72
								void *private)
{
	struct gen_nvm *gn = private;
73 74
	struct nvm_dev *dev = gn->dev;
	struct gen_lun *lun;
M
Matias Bjørling 已提交
75 76 77
	struct nvm_block *blk;
	int i;

78
	lun = &gn->luns[(dev->luns_per_chnl * ppa.g.ch) + ppa.g.lun];
79 80 81 82

	for (i = 0; i < nr_blocks; i++) {
		if (blks[i] == 0)
			continue;
M
Matias Bjørling 已提交
83 84 85 86 87 88 89 90

		blk = &lun->vlun.blocks[i];
		if (!blk) {
			pr_err("gennvm: BB data is out of bounds.\n");
			return -EINVAL;
		}

		list_move_tail(&blk->list, &lun->bb_list);
91
		lun->vlun.nr_bad_blocks++;
92
		lun->vlun.nr_free_blocks--;
M
Matias Bjørling 已提交
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
	}

	return 0;
}

static int gennvm_block_map(u64 slba, u32 nlb, __le64 *entries, void *private)
{
	struct nvm_dev *dev = private;
	struct gen_nvm *gn = dev->mp;
	sector_t max_pages = dev->total_pages * (dev->sec_size >> 9);
	u64 elba = slba + nlb;
	struct gen_lun *lun;
	struct nvm_block *blk;
	u64 i;
	int lun_id;

	if (unlikely(elba > dev->total_pages)) {
		pr_err("gennvm: L2P data from device is out of bounds!\n");
		return -EINVAL;
	}

	for (i = 0; i < nlb; i++) {
		u64 pba = le64_to_cpu(entries[i]);

		if (unlikely(pba >= max_pages && pba != U64_MAX)) {
			pr_err("gennvm: L2P data entry is out of bounds!\n");
			return -EINVAL;
		}

		/* Address zero is a special one. The first page on a disk is
		 * protected. It often holds internal device boot
		 * information.
		 */
		if (!pba)
			continue;

		/* resolve block from physical address */
		lun_id = div_u64(pba, dev->sec_per_lun);
		lun = &gn->luns[lun_id];

		/* Calculate block offset into lun */
		pba = pba - (dev->sec_per_lun * lun_id);
		blk = &lun->vlun.blocks[div_u64(pba, dev->sec_per_blk)];

		if (!blk->type) {
			/* at this point, we don't know anything about the
			 * block. It's up to the FTL on top to re-etablish the
			 * block state
			 */
			list_move_tail(&blk->list, &lun->used_list);
			blk->type = 1;
			lun->vlun.nr_free_blocks--;
145
			lun->vlun.nr_inuse_blocks++;
M
Matias Bjørling 已提交
146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173
		}
	}

	return 0;
}

static int gennvm_blocks_init(struct nvm_dev *dev, struct gen_nvm *gn)
{
	struct gen_lun *lun;
	struct nvm_block *block;
	sector_t lun_iter, blk_iter, cur_block_id = 0;
	int ret;

	gennvm_for_each_lun(gn, lun, lun_iter) {
		lun->vlun.blocks = vzalloc(sizeof(struct nvm_block) *
							dev->blks_per_lun);
		if (!lun->vlun.blocks)
			return -ENOMEM;

		for (blk_iter = 0; blk_iter < dev->blks_per_lun; blk_iter++) {
			block = &lun->vlun.blocks[blk_iter];

			INIT_LIST_HEAD(&block->list);

			block->lun = &lun->vlun;
			block->id = cur_block_id++;

			/* First block is reserved for device */
174 175
			if (unlikely(lun_iter == 0 && blk_iter == 0)) {
				lun->vlun.nr_free_blocks--;
M
Matias Bjørling 已提交
176
				continue;
177
			}
M
Matias Bjørling 已提交
178 179 180 181 182

			list_add_tail(&block->list, &lun->free_list);
		}

		if (dev->ops->get_bb_tbl) {
183 184 185 186 187
			struct ppa_addr ppa;

			ppa.ppa = 0;
			ppa.g.ch = lun->vlun.chnl_id;
			ppa.g.lun = lun->vlun.id;
188
			ppa = generic_to_dev_addr(dev, ppa);
189

190
			ret = dev->ops->get_bb_tbl(dev, ppa,
191 192
						dev->blks_per_lun,
						gennvm_block_bb, gn);
M
Matias Bjørling 已提交
193 194 195 196 197 198
			if (ret)
				pr_err("gennvm: could not read BB table\n");
		}
	}

	if (dev->ops->get_l2p_tbl) {
199
		ret = dev->ops->get_l2p_tbl(dev, 0, dev->total_pages,
M
Matias Bjørling 已提交
200 201 202 203 204 205 206 207 208 209
							gennvm_block_map, dev);
		if (ret) {
			pr_err("gennvm: could not read L2P table.\n");
			pr_warn("gennvm: default block initialization");
		}
	}

	return 0;
}

210 211 212 213 214 215 216 217
static void gennvm_free(struct nvm_dev *dev)
{
	gennvm_blocks_free(dev);
	gennvm_luns_free(dev);
	kfree(dev->mp);
	dev->mp = NULL;
}

M
Matias Bjørling 已提交
218 219 220 221 222
static int gennvm_register(struct nvm_dev *dev)
{
	struct gen_nvm *gn;
	int ret;

223 224 225
	if (!try_module_get(THIS_MODULE))
		return -ENODEV;

M
Matias Bjørling 已提交
226 227 228 229
	gn = kzalloc(sizeof(struct gen_nvm), GFP_KERNEL);
	if (!gn)
		return -ENOMEM;

230
	gn->dev = dev;
M
Matias Bjørling 已提交
231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247
	gn->nr_luns = dev->nr_luns;
	dev->mp = gn;

	ret = gennvm_luns_init(dev, gn);
	if (ret) {
		pr_err("gennvm: could not initialize luns\n");
		goto err;
	}

	ret = gennvm_blocks_init(dev, gn);
	if (ret) {
		pr_err("gennvm: could not initialize blocks\n");
		goto err;
	}

	return 1;
err:
248
	gennvm_free(dev);
249
	module_put(THIS_MODULE);
M
Matias Bjørling 已提交
250 251 252 253 254
	return ret;
}

static void gennvm_unregister(struct nvm_dev *dev)
{
255
	gennvm_free(dev);
256
	module_put(THIS_MODULE);
M
Matias Bjørling 已提交
257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273
}

static struct nvm_block *gennvm_get_blk(struct nvm_dev *dev,
				struct nvm_lun *vlun, unsigned long flags)
{
	struct gen_lun *lun = container_of(vlun, struct gen_lun, vlun);
	struct nvm_block *blk = NULL;
	int is_gc = flags & NVM_IOTYPE_GC;

	spin_lock(&vlun->lock);

	if (list_empty(&lun->free_list)) {
		pr_err_ratelimited("gennvm: lun %u have no free pages available",
								lun->vlun.id);
		goto out;
	}

274
	if (!is_gc && lun->vlun.nr_free_blocks < lun->reserved_blocks)
M
Matias Bjørling 已提交
275 276 277 278 279 280 281
		goto out;

	blk = list_first_entry(&lun->free_list, struct nvm_block, list);
	list_move_tail(&blk->list, &lun->used_list);
	blk->type = 1;

	lun->vlun.nr_free_blocks--;
282
	lun->vlun.nr_inuse_blocks++;
M
Matias Bjørling 已提交
283 284

out:
285
	spin_unlock(&vlun->lock);
M
Matias Bjørling 已提交
286 287 288 289 290 291 292 293 294 295 296 297 298 299
	return blk;
}

static void gennvm_put_blk(struct nvm_dev *dev, struct nvm_block *blk)
{
	struct nvm_lun *vlun = blk->lun;
	struct gen_lun *lun = container_of(vlun, struct gen_lun, vlun);

	spin_lock(&vlun->lock);

	switch (blk->type) {
	case 1:
		list_move_tail(&blk->list, &lun->free_list);
		lun->vlun.nr_free_blocks++;
300
		lun->vlun.nr_inuse_blocks--;
M
Matias Bjørling 已提交
301 302 303 304
		blk->type = 0;
		break;
	case 2:
		list_move_tail(&blk->list, &lun->bb_list);
305 306
		lun->vlun.nr_bad_blocks++;
		lun->vlun.nr_inuse_blocks--;
M
Matias Bjørling 已提交
307 308 309 310 311 312
		break;
	default:
		WARN_ON_ONCE(1);
		pr_err("gennvm: erroneous block type (%lu -> %u)\n",
							blk->id, blk->type);
		list_move_tail(&blk->list, &lun->bb_list);
313 314
		lun->vlun.nr_bad_blocks++;
		lun->vlun.nr_inuse_blocks--;
M
Matias Bjørling 已提交
315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349
	}

	spin_unlock(&vlun->lock);
}

static void gennvm_blk_set_type(struct nvm_dev *dev, struct ppa_addr *ppa,
								int type)
{
	struct gen_nvm *gn = dev->mp;
	struct gen_lun *lun;
	struct nvm_block *blk;

	if (unlikely(ppa->g.ch > dev->nr_chnls ||
					ppa->g.lun > dev->luns_per_chnl ||
					ppa->g.blk > dev->blks_per_lun)) {
		WARN_ON_ONCE(1);
		pr_err("gennvm: ppa broken (ch: %u > %u lun: %u > %u blk: %u > %u",
				ppa->g.ch, dev->nr_chnls,
				ppa->g.lun, dev->luns_per_chnl,
				ppa->g.blk, dev->blks_per_lun);
		return;
	}

	lun = &gn->luns[ppa->g.lun * ppa->g.ch];
	blk = &lun->vlun.blocks[ppa->g.blk];

	/* will be moved to bb list on put_blk from target */
	blk->type = type;
}

/* mark block bad. It is expected the target recover from the error. */
static void gennvm_mark_blk_bad(struct nvm_dev *dev, struct nvm_rq *rqd)
{
	int i;

350
	if (!dev->ops->set_bb_tbl)
M
Matias Bjørling 已提交
351 352
		return;

353
	if (dev->ops->set_bb_tbl(dev, rqd, 1))
M
Matias Bjørling 已提交
354 355
		return;

356
	nvm_addr_to_generic_mode(dev, rqd);
M
Matias Bjørling 已提交
357 358 359 360 361 362 363 364 365

	/* look up blocks and mark them as bad */
	if (rqd->nr_pages > 1)
		for (i = 0; i < rqd->nr_pages; i++)
			gennvm_blk_set_type(dev, &rqd->ppa_list[i], 2);
	else
		gennvm_blk_set_type(dev, &rqd->ppa_addr, 2);
}

366
static void gennvm_end_io(struct nvm_rq *rqd)
M
Matias Bjørling 已提交
367 368 369
{
	struct nvm_tgt_instance *ins = rqd->ins;

370
	switch (rqd->error) {
M
Matias Bjørling 已提交
371 372 373 374 375 376 377
	case NVM_RSP_SUCCESS:
	case NVM_RSP_ERR_EMPTYPAGE:
		break;
	case NVM_RSP_ERR_FAILWRITE:
		gennvm_mark_blk_bad(rqd->dev, rqd);
	}

378
	ins->tt->end_io(rqd);
379
}
M
Matias Bjørling 已提交
380

381 382 383 384 385 386 387 388 389 390 391
static int gennvm_submit_io(struct nvm_dev *dev, struct nvm_rq *rqd)
{
	if (!dev->ops->submit_io)
		return -ENODEV;

	/* Convert address space */
	nvm_generic_to_addr_mode(dev, rqd);

	rqd->dev = dev;
	rqd->end_io = gennvm_end_io;
	return dev->ops->submit_io(dev, rqd);
M
Matias Bjørling 已提交
392 393 394 395 396
}

static int gennvm_erase_blk(struct nvm_dev *dev, struct nvm_block *blk,
							unsigned long flags)
{
397
	struct ppa_addr addr = block_to_ppa(dev, blk);
M
Matias Bjørling 已提交
398

399
	return nvm_erase_ppa(dev, &addr, 1);
M
Matias Bjørling 已提交
400 401 402 403 404 405 406 407 408
}

static struct nvm_lun *gennvm_get_lun(struct nvm_dev *dev, int lunid)
{
	struct gen_nvm *gn = dev->mp;

	return &gn->luns[lunid].vlun;
}

409
static void gennvm_lun_info_print(struct nvm_dev *dev)
M
Matias Bjørling 已提交
410 411 412 413 414
{
	struct gen_nvm *gn = dev->mp;
	struct gen_lun *lun;
	unsigned int i;

415 416 417 418 419 420 421 422 423 424 425 426

	gennvm_for_each_lun(gn, lun, i) {
		spin_lock(&lun->vlun.lock);

		pr_info("%s: lun%8u\t%u\t%u\t%u\n",
				dev->name, i,
				lun->vlun.nr_free_blocks,
				lun->vlun.nr_inuse_blocks,
				lun->vlun.nr_bad_blocks);

		spin_unlock(&lun->vlun.lock);
	}
M
Matias Bjørling 已提交
427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442
}

static struct nvmm_type gennvm = {
	.name		= "gennvm",
	.version	= {0, 1, 0},

	.register_mgr	= gennvm_register,
	.unregister_mgr	= gennvm_unregister,

	.get_blk	= gennvm_get_blk,
	.put_blk	= gennvm_put_blk,

	.submit_io	= gennvm_submit_io,
	.erase_blk	= gennvm_erase_blk,

	.get_lun	= gennvm_get_lun,
443
	.lun_info_print = gennvm_lun_info_print,
M
Matias Bjørling 已提交
444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
};

static int __init gennvm_module_init(void)
{
	return nvm_register_mgr(&gennvm);
}

static void gennvm_module_exit(void)
{
	nvm_unregister_mgr(&gennvm);
}

module_init(gennvm_module_init);
module_exit(gennvm_module_exit);
MODULE_LICENSE("GPL v2");
MODULE_DESCRIPTION("Generic media manager for Open-Channel SSDs");