summaryrefslogtreecommitdiffstats
path: root/drivers/lightnvm/pblk-write.c
blob: aef6fd7c4a0cbae0859398fcd07ad9854d4a1f77 (plain)
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
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
89
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
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
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
275
276
277
278
279
280
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
313
314
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
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
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
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
/*
 * Copyright (C) 2016 CNEX Labs
 * Initial release: Javier Gonzalez <javier@cnexlabs.com>
 *                  Matias Bjorling <matias@cnexlabs.com>
 *
 * 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.
 *
 * pblk-write.c - pblk's write path from write buffer to media
 */

#include "pblk.h"

static void pblk_sync_line(struct pblk *pblk, struct pblk_line *line)
{
#ifdef CONFIG_NVM_DEBUG
	atomic_long_inc(&pblk->sync_writes);
#endif

	/* Counter protected by rb sync lock */
	line->left_ssecs--;
	if (!line->left_ssecs)
		pblk_line_run_ws(pblk, line, NULL, pblk_line_close_ws);
}

static unsigned long pblk_end_w_bio(struct pblk *pblk, struct nvm_rq *rqd,
				    struct pblk_c_ctx *c_ctx)
{
	struct nvm_tgt_dev *dev = pblk->dev;
	struct bio *original_bio;
	unsigned long ret;
	int i;

	for (i = 0; i < c_ctx->nr_valid; i++) {
		struct pblk_w_ctx *w_ctx;
		struct ppa_addr p;
		struct pblk_line *line;

		w_ctx = pblk_rb_w_ctx(&pblk->rwb, c_ctx->sentry + i);

		p = rqd->ppa_list[i];
		line = &pblk->lines[pblk_dev_ppa_to_line(p)];
		pblk_sync_line(pblk, line);

		while ((original_bio = bio_list_pop(&w_ctx->bios)))
			bio_endio(original_bio);
	}

#ifdef CONFIG_NVM_DEBUG
	atomic_long_add(c_ctx->nr_valid, &pblk->compl_writes);
#endif

	ret = pblk_rb_sync_advance(&pblk->rwb, c_ctx->nr_valid);

	if (rqd->meta_list)
		nvm_dev_dma_free(dev->parent, rqd->meta_list,
							rqd->dma_meta_list);

	bio_put(rqd->bio);
	pblk_free_rqd(pblk, rqd, WRITE);

	return ret;
}

static unsigned long pblk_end_queued_w_bio(struct pblk *pblk,
					   struct nvm_rq *rqd,
					   struct pblk_c_ctx *c_ctx)
{
	list_del(&c_ctx->list);
	return pblk_end_w_bio(pblk, rqd, c_ctx);
}

static void pblk_complete_write(struct pblk *pblk, struct nvm_rq *rqd,
				struct pblk_c_ctx *c_ctx)
{
	struct pblk_c_ctx *c, *r;
	unsigned long flags;
	unsigned long pos;

#ifdef CONFIG_NVM_DEBUG
	atomic_long_sub(c_ctx->nr_valid, &pblk->inflight_writes);
#endif

	pblk_up_rq(pblk, rqd->ppa_list, rqd->nr_ppas, c_ctx->lun_bitmap);

	pos = pblk_rb_sync_init(&pblk->rwb, &flags);
	if (pos == c_ctx->sentry) {
		pos = pblk_end_w_bio(pblk, rqd, c_ctx);

retry:
		list_for_each_entry_safe(c, r, &pblk->compl_list, list) {
			rqd = nvm_rq_from_c_ctx(c);
			if (c->sentry == pos) {
				pos = pblk_end_queued_w_bio(pblk, rqd, c);
				goto retry;
			}
		}
	} else {
		WARN_ON(nvm_rq_from_c_ctx(c_ctx) != rqd);
		list_add_tail(&c_ctx->list, &pblk->compl_list);
	}
	pblk_rb_sync_end(&pblk->rwb, &flags);
}

/* When a write fails, we are not sure whether the block has grown bad or a page
 * range is more susceptible to write errors. If a high number of pages fail, we
 * assume that the block is bad and we mark it accordingly. In all cases, we
 * remap and resubmit the failed entries as fast as possible; if a flush is
 * waiting on a completion, the whole stack would stall otherwise.
 */
static void pblk_end_w_fail(struct pblk *pblk, struct nvm_rq *rqd)
{
	void *comp_bits = &rqd->ppa_status;
	struct pblk_c_ctx *c_ctx = nvm_rq_to_pdu(rqd);
	struct pblk_rec_ctx *recovery;
	struct ppa_addr *ppa_list = rqd->ppa_list;
	int nr_ppas = rqd->nr_ppas;
	unsigned int c_entries;
	int bit, ret;

	if (unlikely(nr_ppas == 1))
		ppa_list = &rqd->ppa_addr;

	recovery = mempool_alloc(pblk->rec_pool, GFP_ATOMIC);
	if (!recovery) {
		pr_err("pblk: could not allocate recovery context\n");
		return;
	}
	INIT_LIST_HEAD(&recovery->failed);

	bit = -1;
	while ((bit = find_next_bit(comp_bits, nr_ppas, bit + 1)) < nr_ppas) {
		struct pblk_rb_entry *entry;
		struct ppa_addr ppa;

		/* Logic error */
		if (bit > c_ctx->nr_valid) {
			WARN_ONCE(1, "pblk: corrupted write request\n");
			mempool_free(recovery, pblk->rec_pool);
			goto out;
		}

		ppa = ppa_list[bit];
		entry = pblk_rb_sync_scan_entry(&pblk->rwb, &ppa);
		if (!entry) {
			pr_err("pblk: could not scan entry on write failure\n");
			mempool_free(recovery, pblk->rec_pool);
			goto out;
		}

		/* The list is filled first and emptied afterwards. No need for
		 * protecting it with a lock
		 */
		list_add_tail(&entry->index, &recovery->failed);
	}

	c_entries = find_first_bit(comp_bits, nr_ppas);
	ret = pblk_recov_setup_rq(pblk, c_ctx, recovery, comp_bits, c_entries);
	if (ret) {
		pr_err("pblk: could not recover from write failure\n");
		mempool_free(recovery, pblk->rec_pool);
		goto out;
	}

	INIT_WORK(&recovery->ws_rec, pblk_submit_rec);
	queue_work(pblk->kw_wq, &recovery->ws_rec);

out:
	pblk_complete_write(pblk, rqd, c_ctx);
}

static void pblk_end_io_write(struct nvm_rq *rqd)
{
	struct pblk *pblk = rqd->private;
	struct pblk_c_ctx *c_ctx = nvm_rq_to_pdu(rqd);

	if (rqd->error) {
		pblk_log_write_err(pblk, rqd);
		return pblk_end_w_fail(pblk, rqd);
	}
#ifdef CONFIG_NVM_DEBUG
	else
		WARN_ONCE(rqd->bio->bi_error, "pblk: corrupted write error\n");
#endif

	pblk_complete_write(pblk, rqd, c_ctx);
}

static int pblk_alloc_w_rq(struct pblk *pblk, struct nvm_rq *rqd,
			   unsigned int nr_secs)
{
	struct nvm_tgt_dev *dev = pblk->dev;

	/* Setup write request */
	rqd->opcode = NVM_OP_PWRITE;
	rqd->nr_ppas = nr_secs;
	rqd->flags = pblk_set_progr_mode(pblk, WRITE);
	rqd->private = pblk;
	rqd->end_io = pblk_end_io_write;

	rqd->meta_list = nvm_dev_dma_alloc(dev->parent, GFP_KERNEL,
							&rqd->dma_meta_list);
	if (!rqd->meta_list)
		return -ENOMEM;

	if (unlikely(nr_secs == 1))
		return 0;

	rqd->ppa_list = rqd->meta_list + pblk_dma_meta_size;
	rqd->dma_ppa_list = rqd->dma_meta_list + pblk_dma_meta_size;

	return 0;
}

static int pblk_setup_w_rq(struct pblk *pblk, struct nvm_rq *rqd,
			   struct pblk_c_ctx *c_ctx)
{
	struct pblk_line_meta *lm = &pblk->lm;
	struct pblk_line *e_line = pblk_line_get_data_next(pblk);
	struct ppa_addr erase_ppa;
	unsigned int valid = c_ctx->nr_valid;
	unsigned int padded = c_ctx->nr_padded;
	unsigned int nr_secs = valid + padded;
	unsigned long *lun_bitmap;
	int ret = 0;

	lun_bitmap = kzalloc(lm->lun_bitmap_len, GFP_KERNEL);
	if (!lun_bitmap) {
		ret = -ENOMEM;
		goto out;
	}
	c_ctx->lun_bitmap = lun_bitmap;

	ret = pblk_alloc_w_rq(pblk, rqd, nr_secs);
	if (ret) {
		kfree(lun_bitmap);
		goto out;
	}

	ppa_set_empty(&erase_ppa);
	if (likely(!e_line || !atomic_read(&e_line->left_eblks)))
		pblk_map_rq(pblk, rqd, c_ctx->sentry, lun_bitmap, valid, 0);
	else
		pblk_map_erase_rq(pblk, rqd, c_ctx->sentry, lun_bitmap,
							valid, &erase_ppa);

out:
	if (unlikely(e_line && !ppa_empty(erase_ppa))) {
		if (pblk_blk_erase_async(pblk, erase_ppa)) {
			struct nvm_tgt_dev *dev = pblk->dev;
			struct nvm_geo *geo = &dev->geo;
			int bit;

			atomic_inc(&e_line->left_eblks);
			bit = erase_ppa.g.lun * geo->nr_chnls + erase_ppa.g.ch;
			WARN_ON(!test_and_clear_bit(bit, e_line->erase_bitmap));
			up(&pblk->erase_sem);
		}
	}

	return ret;
}

int pblk_setup_w_rec_rq(struct pblk *pblk, struct nvm_rq *rqd,
			struct pblk_c_ctx *c_ctx)
{
	struct pblk_line_meta *lm = &pblk->lm;
	unsigned long *lun_bitmap;
	int ret;

	lun_bitmap = kzalloc(lm->lun_bitmap_len, GFP_KERNEL);
	if (!lun_bitmap)
		return -ENOMEM;

	c_ctx->lun_bitmap = lun_bitmap;

	ret = pblk_alloc_w_rq(pblk, rqd, rqd->nr_ppas);
	if (ret)
		return ret;

	pblk_map_rq(pblk, rqd, c_ctx->sentry, lun_bitmap, c_ctx->nr_valid, 0);

	rqd->ppa_status = (u64)0;
	rqd->flags = pblk_set_progr_mode(pblk, WRITE);

	return ret;
}

static int pblk_calc_secs_to_sync(struct pblk *pblk, unsigned int secs_avail,
				  unsigned int secs_to_flush)
{
	int secs_to_sync;

	secs_to_sync = pblk_calc_secs(pblk, secs_avail, secs_to_flush);

#ifdef CONFIG_NVM_DEBUG
	if ((!secs_to_sync && secs_to_flush)
			|| (secs_to_sync < 0)
			|| (secs_to_sync > secs_avail && !secs_to_flush)) {
		pr_err("pblk: bad sector calculation (a:%d,s:%d,f:%d)\n",
				secs_avail, secs_to_sync, secs_to_flush);
	}
#endif

	return secs_to_sync;
}

static int pblk_submit_write(struct pblk *pblk)
{
	struct bio *bio;
	struct nvm_rq *rqd;
	struct pblk_c_ctx *c_ctx;
	unsigned int pgs_read;
	unsigned int secs_avail, secs_to_sync, secs_to_com;
	unsigned int secs_to_flush;
	unsigned long pos;
	int err;

	/* If there are no sectors in the cache, flushes (bios without data)
	 * will be cleared on the cache threads
	 */
	secs_avail = pblk_rb_read_count(&pblk->rwb);
	if (!secs_avail)
		return 1;

	secs_to_flush = pblk_rb_sync_point_count(&pblk->rwb);
	if (!secs_to_flush && secs_avail < pblk->min_write_pgs)
		return 1;

	rqd = pblk_alloc_rqd(pblk, WRITE);
	if (IS_ERR(rqd)) {
		pr_err("pblk: cannot allocate write req.\n");
		return 1;
	}
	c_ctx = nvm_rq_to_pdu(rqd);

	bio = bio_alloc(GFP_KERNEL, pblk->max_write_pgs);
	if (!bio) {
		pr_err("pblk: cannot allocate write bio\n");
		goto fail_free_rqd;
	}
	bio->bi_iter.bi_sector = 0; /* internal bio */
	bio_set_op_attrs(bio, REQ_OP_WRITE, 0);
	rqd->bio = bio;

	secs_to_sync = pblk_calc_secs_to_sync(pblk, secs_avail, secs_to_flush);
	if (secs_to_sync > pblk->max_write_pgs) {
		pr_err("pblk: bad buffer sync calculation\n");
		goto fail_put_bio;
	}

	secs_to_com = (secs_to_sync > secs_avail) ? secs_avail : secs_to_sync;
	pos = pblk_rb_read_commit(&pblk->rwb, secs_to_com);

	pgs_read = pblk_rb_read_to_bio(&pblk->rwb, bio, c_ctx, pos,
						secs_to_sync, secs_avail);
	if (!pgs_read) {
		pr_err("pblk: corrupted write bio\n");
		goto fail_put_bio;
	}

	if (c_ctx->nr_padded)
		if (pblk_bio_add_pages(pblk, bio, GFP_KERNEL, c_ctx->nr_padded))
			goto fail_put_bio;

	/* Assign lbas to ppas and populate request structure */
	err = pblk_setup_w_rq(pblk, rqd, c_ctx);
	if (err) {
		pr_err("pblk: could not setup write request\n");
		goto fail_free_bio;
	}

	err = pblk_submit_io(pblk, rqd);
	if (err) {
		pr_err("pblk: I/O submission failed: %d\n", err);
		goto fail_free_bio;
	}

#ifdef CONFIG_NVM_DEBUG
	atomic_long_add(secs_to_sync, &pblk->sub_writes);
#endif

	return 0;

fail_free_bio:
	if (c_ctx->nr_padded)
		pblk_bio_free_pages(pblk, bio, secs_to_sync, c_ctx->nr_padded);
fail_put_bio:
	bio_put(bio);
fail_free_rqd:
	pblk_free_rqd(pblk, rqd, WRITE);

	return 1;
}

int pblk_write_ts(void *data)
{
	struct pblk *pblk = data;

	while (!kthread_should_stop()) {
		if (!pblk_submit_write(pblk))
			continue;
		set_current_state(TASK_INTERRUPTIBLE);
		io_schedule();
	}

	return 0;
}