1// SPDX-License-Identifier: GPL-2.0
  2/*
  3 * Copyright (C) 2016 CNEX Labs
  4 * Initial release: Javier Gonzalez <javier@cnexlabs.com>
  5 *                  Matias Bjorling <matias@cnexlabs.com>
  6 *
  7 * This program is free software; you can redistribute it and/or
  8 * modify it under the terms of the GNU General Public License version
  9 * 2 as published by the Free Software Foundation.
 10 *
 11 * This program is distributed in the hope that it will be useful, but
 12 * WITHOUT ANY WARRANTY; without even the implied warranty of
 13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 14 * General Public License for more details.
 15 *
 16 * pblk-read.c - pblk's read path
 17 */
 18
 19#include "pblk.h"
 20
 21/*
 22 * There is no guarantee that the value read from cache has not been updated and
 23 * resides at another location in the cache. We guarantee though that if the
 24 * value is read from the cache, it belongs to the mapped lba. In order to
 25 * guarantee and order between writes and reads are ordered, a flush must be
 26 * issued.
 27 */
 28static int pblk_read_from_cache(struct pblk *pblk, struct bio *bio,
 29				sector_t lba, struct ppa_addr ppa)
 30{
 31#ifdef CONFIG_NVM_PBLK_DEBUG
 32	/* Callers must ensure that the ppa points to a cache address */
 33	BUG_ON(pblk_ppa_empty(ppa));
 34	BUG_ON(!pblk_addr_in_cache(ppa));
 35#endif
 36
 37	return pblk_rb_copy_to_bio(&pblk->rwb, bio, lba, ppa);
 38}
 39
 40static int pblk_read_ppalist_rq(struct pblk *pblk, struct nvm_rq *rqd,
 41				 struct bio *bio, sector_t blba,
 42				 bool *from_cache)
 43{
 44	void *meta_list = rqd->meta_list;
 45	int nr_secs, i;
 46
 47retry:
 48	nr_secs = pblk_lookup_l2p_seq(pblk, rqd->ppa_list, blba, rqd->nr_ppas,
 49					from_cache);
 50
 51	if (!*from_cache)
 52		goto end;
 53
 54	for (i = 0; i < nr_secs; i++) {
 55		struct pblk_sec_meta *meta = pblk_get_meta(pblk, meta_list, i);
 56		sector_t lba = blba + i;
 57
 58		if (pblk_ppa_empty(rqd->ppa_list[i])) {
 59			__le64 addr_empty = cpu_to_le64(ADDR_EMPTY);
 60
 61			meta->lba = addr_empty;
 62		} else if (pblk_addr_in_cache(rqd->ppa_list[i])) {
 63			/*
 64			 * Try to read from write buffer. The address is later
 65			 * checked on the write buffer to prevent retrieving
 66			 * overwritten data.
 67			 */
 68			if (!pblk_read_from_cache(pblk, bio, lba,
 69							rqd->ppa_list[i])) {
 70				if (i == 0) {
 71					/*
 72					 * We didn't call with bio_advance()
 73					 * yet, so we can just retry.
 74					 */
 75					goto retry;
 76				} else {
 77					/*
 78					 * We already call bio_advance()
 79					 * so we cannot retry and we need
 80					 * to quit that function in order
 81					 * to allow caller to handle the bio
 82					 * splitting in the current sector
 83					 * position.
 84					 */
 85					nr_secs = i;
 86					goto end;
 87				}
 88			}
 89			meta->lba = cpu_to_le64(lba);
 90#ifdef CONFIG_NVM_PBLK_DEBUG
 91			atomic_long_inc(&pblk->cache_reads);
 92#endif
 93		}
 94		bio_advance(bio, PBLK_EXPOSED_PAGE_SIZE);
 95	}
 96
 97end:
 98	if (pblk_io_aligned(pblk, nr_secs))
 99		rqd->is_seq = 1;
100
101#ifdef CONFIG_NVM_PBLK_DEBUG
102	atomic_long_add(nr_secs, &pblk->inflight_reads);
103#endif
104
105	return nr_secs;
106}
107
108
109static void pblk_read_check_seq(struct pblk *pblk, struct nvm_rq *rqd,
110				sector_t blba)
111{
112	void *meta_list = rqd->meta_list;
113	int nr_lbas = rqd->nr_ppas;
114	int i;
115
116	if (!pblk_is_oob_meta_supported(pblk))
117		return;
118
119	for (i = 0; i < nr_lbas; i++) {
120		struct pblk_sec_meta *meta = pblk_get_meta(pblk, meta_list, i);
121		u64 lba = le64_to_cpu(meta->lba);
122
123		if (lba == ADDR_EMPTY)
124			continue;
125
126		if (lba != blba + i) {
127#ifdef CONFIG_NVM_PBLK_DEBUG
128			struct ppa_addr *ppa_list = nvm_rq_to_ppa_list(rqd);
129
130			print_ppa(pblk, &ppa_list[i], "seq", i);
131#endif
132			pblk_err(pblk, "corrupted read LBA (%llu/%llu)\n",
133							lba, (u64)blba + i);
134			WARN_ON(1);
135		}
136	}
137}
138
139/*
140 * There can be holes in the lba list.
141 */
142static void pblk_read_check_rand(struct pblk *pblk, struct nvm_rq *rqd,
143				 u64 *lba_list, int nr_lbas)
144{
145	void *meta_lba_list = rqd->meta_list;
146	int i, j;
147
148	if (!pblk_is_oob_meta_supported(pblk))
149		return;
150
151	for (i = 0, j = 0; i < nr_lbas; i++) {
152		struct pblk_sec_meta *meta = pblk_get_meta(pblk,
153							   meta_lba_list, j);
154		u64 lba = lba_list[i];
155		u64 meta_lba;
156
157		if (lba == ADDR_EMPTY)
158			continue;
159
160		meta_lba = le64_to_cpu(meta->lba);
161
162		if (lba != meta_lba) {
163#ifdef CONFIG_NVM_PBLK_DEBUG
164			struct ppa_addr *ppa_list = nvm_rq_to_ppa_list(rqd);
165
166			print_ppa(pblk, &ppa_list[j], "rnd", j);
167#endif
168			pblk_err(pblk, "corrupted read LBA (%llu/%llu)\n",
169							meta_lba, lba);
170			WARN_ON(1);
171		}
172
173		j++;
174	}
175
176	WARN_ONCE(j != rqd->nr_ppas, "pblk: corrupted random request\n");
177}
178
179static void pblk_end_user_read(struct bio *bio, int error)
180{
181	if (error && error != NVM_RSP_WARN_HIGHECC)
182		bio_io_error(bio);
183	else
184		bio_endio(bio);
185}
186
187static void __pblk_end_io_read(struct pblk *pblk, struct nvm_rq *rqd,
188			       bool put_line)
189{
190	struct nvm_tgt_dev *dev = pblk->dev;
191	struct pblk_g_ctx *r_ctx = nvm_rq_to_pdu(rqd);
192	struct bio *int_bio = rqd->bio;
193	unsigned long start_time = r_ctx->start_time;
194
195	generic_end_io_acct(dev->q, REQ_OP_READ, &pblk->disk->part0, start_time);
196
197	if (rqd->error)
198		pblk_log_read_err(pblk, rqd);
199
200	pblk_read_check_seq(pblk, rqd, r_ctx->lba);
201	bio_put(int_bio);
202
203	if (put_line)
204		pblk_rq_to_line_put(pblk, rqd);
205
206#ifdef CONFIG_NVM_PBLK_DEBUG
207	atomic_long_add(rqd->nr_ppas, &pblk->sync_reads);
208	atomic_long_sub(rqd->nr_ppas, &pblk->inflight_reads);
209#endif
210
211	pblk_free_rqd(pblk, rqd, PBLK_READ);
212	atomic_dec(&pblk->inflight_io);
213}
214
215static void pblk_end_io_read(struct nvm_rq *rqd)
216{
217	struct pblk *pblk = rqd->private;
218	struct pblk_g_ctx *r_ctx = nvm_rq_to_pdu(rqd);
219	struct bio *bio = (struct bio *)r_ctx->private;
220
221	pblk_end_user_read(bio, rqd->error);
222	__pblk_end_io_read(pblk, rqd, true);
223}
224
225static void pblk_read_rq(struct pblk *pblk, struct nvm_rq *rqd, struct bio *bio,
226			 sector_t lba, bool *from_cache)
227{
228	struct pblk_sec_meta *meta = pblk_get_meta(pblk, rqd->meta_list, 0);
229	struct ppa_addr ppa;
230
231	pblk_lookup_l2p_seq(pblk, &ppa, lba, 1, from_cache);
232
233#ifdef CONFIG_NVM_PBLK_DEBUG
234	atomic_long_inc(&pblk->inflight_reads);
235#endif
236
237retry:
238	if (pblk_ppa_empty(ppa)) {
239		__le64 addr_empty = cpu_to_le64(ADDR_EMPTY);
240
241		meta->lba = addr_empty;
242		return;
243	}
244
245	/* Try to read from write buffer. The address is later checked on the
246	 * write buffer to prevent retrieving overwritten data.
247	 */
248	if (pblk_addr_in_cache(ppa)) {
249		if (!pblk_read_from_cache(pblk, bio, lba, ppa)) {
250			pblk_lookup_l2p_seq(pblk, &ppa, lba, 1, from_cache);
251			goto retry;
252		}
253
254		meta->lba = cpu_to_le64(lba);
255
256#ifdef CONFIG_NVM_PBLK_DEBUG
257		atomic_long_inc(&pblk->cache_reads);
258#endif
259	} else {
260		rqd->ppa_addr = ppa;
261	}
262}
263
264void pblk_submit_read(struct pblk *pblk, struct bio *bio)
265{
266	struct nvm_tgt_dev *dev = pblk->dev;
267	struct request_queue *q = dev->q;
268	sector_t blba = pblk_get_lba(bio);
269	unsigned int nr_secs = pblk_get_secs(bio);
270	bool from_cache;
271	struct pblk_g_ctx *r_ctx;
272	struct nvm_rq *rqd;
273	struct bio *int_bio, *split_bio;
274
275	generic_start_io_acct(q, REQ_OP_READ, bio_sectors(bio),
276			      &pblk->disk->part0);
277
278	rqd = pblk_alloc_rqd(pblk, PBLK_READ);
279
280	rqd->opcode = NVM_OP_PREAD;
281	rqd->nr_ppas = nr_secs;
282	rqd->private = pblk;
283	rqd->end_io = pblk_end_io_read;
284
285	r_ctx = nvm_rq_to_pdu(rqd);
286	r_ctx->start_time = jiffies;
287	r_ctx->lba = blba;
288
289	if (pblk_alloc_rqd_meta(pblk, rqd)) {
290		bio_io_error(bio);
291		pblk_free_rqd(pblk, rqd, PBLK_READ);
292		return;
293	}
294
295	/* Clone read bio to deal internally with:
296	 * -read errors when reading from drive
297	 * -bio_advance() calls during cache reads
298	 */
299	int_bio = bio_clone_fast(bio, GFP_KERNEL, &pblk_bio_set);
300
301	if (nr_secs > 1)
302		nr_secs = pblk_read_ppalist_rq(pblk, rqd, int_bio, blba,
303						&from_cache);
304	else
305		pblk_read_rq(pblk, rqd, int_bio, blba, &from_cache);
306
307split_retry:
308	r_ctx->private = bio; /* original bio */
309	rqd->bio = int_bio; /* internal bio */
310
311	if (from_cache && nr_secs == rqd->nr_ppas) {
312		/* All data was read from cache, we can complete the IO. */
313		pblk_end_user_read(bio, 0);
314		atomic_inc(&pblk->inflight_io);
315		__pblk_end_io_read(pblk, rqd, false);
316	} else if (nr_secs != rqd->nr_ppas) {
317		/* The read bio request could be partially filled by the write
318		 * buffer, but there are some holes that need to be read from
319		 * the drive. In order to handle this, we will use block layer
320		 * mechanism to split this request in to smaller ones and make
321		 * a chain of it.
322		 */
323		split_bio = bio_split(bio, nr_secs * NR_PHY_IN_LOG, GFP_KERNEL,
324					&pblk_bio_set);
325		bio_chain(split_bio, bio);
326		generic_make_request(bio);
327
328		/* New bio contains first N sectors of the previous one, so
329		 * we can continue to use existing rqd, but we need to shrink
330		 * the number of PPAs in it. New bio is also guaranteed that
331		 * it contains only either data from cache or from drive, newer
332		 * mix of them.
333		 */
334		bio = split_bio;
335		rqd->nr_ppas = nr_secs;
336		if (rqd->nr_ppas == 1)
337			rqd->ppa_addr = rqd->ppa_list[0];
338
339		/* Recreate int_bio - existing might have some needed internal
340		 * fields modified already.
341		 */
342		bio_put(int_bio);
343		int_bio = bio_clone_fast(bio, GFP_KERNEL, &pblk_bio_set);
344		goto split_retry;
345	} else if (pblk_submit_io(pblk, rqd, NULL)) {
346		/* Submitting IO to drive failed, let's report an error */
347		rqd->error = -ENODEV;
348		pblk_end_io_read(rqd);
349	}
350}
351
352static int read_ppalist_rq_gc(struct pblk *pblk, struct nvm_rq *rqd,
353			      struct pblk_line *line, u64 *lba_list,
354			      u64 *paddr_list_gc, unsigned int nr_secs)
355{
356	struct ppa_addr ppa_list_l2p[NVM_MAX_VLBA];
357	struct ppa_addr ppa_gc;
358	int valid_secs = 0;
359	int i;
360
361	pblk_lookup_l2p_rand(pblk, ppa_list_l2p, lba_list, nr_secs);
362
363	for (i = 0; i < nr_secs; i++) {
364		if (lba_list[i] == ADDR_EMPTY)
365			continue;
366
367		ppa_gc = addr_to_gen_ppa(pblk, paddr_list_gc[i], line->id);
368		if (!pblk_ppa_comp(ppa_list_l2p[i], ppa_gc)) {
369			paddr_list_gc[i] = lba_list[i] = ADDR_EMPTY;
370			continue;
371		}
372
373		rqd->ppa_list[valid_secs++] = ppa_list_l2p[i];
374	}
375
376#ifdef CONFIG_NVM_PBLK_DEBUG
377	atomic_long_add(valid_secs, &pblk->inflight_reads);
378#endif
379
380	return valid_secs;
381}
382
383static int read_rq_gc(struct pblk *pblk, struct nvm_rq *rqd,
384		      struct pblk_line *line, sector_t lba,
385		      u64 paddr_gc)
386{
387	struct ppa_addr ppa_l2p, ppa_gc;
388	int valid_secs = 0;
389
390	if (lba == ADDR_EMPTY)
391		goto out;
392
393	/* logic error: lba out-of-bounds */
394	if (lba >= pblk->capacity) {
395		WARN(1, "pblk: read lba out of bounds\n");
396		goto out;
397	}
398
399	spin_lock(&pblk->trans_lock);
400	ppa_l2p = pblk_trans_map_get(pblk, lba);
401	spin_unlock(&pblk->trans_lock);
402
403	ppa_gc = addr_to_gen_ppa(pblk, paddr_gc, line->id);
404	if (!pblk_ppa_comp(ppa_l2p, ppa_gc))
405		goto out;
406
407	rqd->ppa_addr = ppa_l2p;
408	valid_secs = 1;
409
410#ifdef CONFIG_NVM_PBLK_DEBUG
411	atomic_long_inc(&pblk->inflight_reads);
412#endif
413
414out:
415	return valid_secs;
416}
417
418int pblk_submit_read_gc(struct pblk *pblk, struct pblk_gc_rq *gc_rq)
419{
420	struct nvm_rq rqd;
421	int ret = NVM_IO_OK;
422
423	memset(&rqd, 0, sizeof(struct nvm_rq));
424
425	ret = pblk_alloc_rqd_meta(pblk, &rqd);
426	if (ret)
427		return ret;
428
429	if (gc_rq->nr_secs > 1) {
430		gc_rq->secs_to_gc = read_ppalist_rq_gc(pblk, &rqd, gc_rq->line,
431							gc_rq->lba_list,
432							gc_rq->paddr_list,
433							gc_rq->nr_secs);
434		if (gc_rq->secs_to_gc == 1)
435			rqd.ppa_addr = rqd.ppa_list[0];
436	} else {
437		gc_rq->secs_to_gc = read_rq_gc(pblk, &rqd, gc_rq->line,
438							gc_rq->lba_list[0],
439							gc_rq->paddr_list[0]);
440	}
441
442	if (!(gc_rq->secs_to_gc))
443		goto out;
444
445	rqd.opcode = NVM_OP_PREAD;
446	rqd.nr_ppas = gc_rq->secs_to_gc;
447
448	if (pblk_submit_io_sync(pblk, &rqd, gc_rq->data)) {
449		ret = -EIO;
450		goto err_free_dma;
451	}
452
453	pblk_read_check_rand(pblk, &rqd, gc_rq->lba_list, gc_rq->nr_secs);
454
455	atomic_dec(&pblk->inflight_io);
456
457	if (rqd.error) {
458		atomic_long_inc(&pblk->read_failed_gc);
459#ifdef CONFIG_NVM_PBLK_DEBUG
460		pblk_print_failed_rqd(pblk, &rqd, rqd.error);
461#endif
462	}
463
464#ifdef CONFIG_NVM_PBLK_DEBUG
465	atomic_long_add(gc_rq->secs_to_gc, &pblk->sync_reads);
466	atomic_long_add(gc_rq->secs_to_gc, &pblk->recov_gc_reads);
467	atomic_long_sub(gc_rq->secs_to_gc, &pblk->inflight_reads);
468#endif
469
470out:
471	pblk_free_rqd_meta(pblk, &rqd);
472	return ret;
473
474err_free_dma:
475	pblk_free_rqd_meta(pblk, &rqd);
476	return ret;
477}