1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3 * Copyright (C) 2015 IT University of Copenhagen (rrpc.h)
4 * Copyright (C) 2016 CNEX Labs
5 * Initial release: Matias Bjorling <matias@cnexlabs.com>
6 * Write buffering: Javier Gonzalez <javier@cnexlabs.com>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License version
10 * 2 as published by the Free Software Foundation.
11 *
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License for more details.
16 *
17 * Implementation of a Physical Block-device target for Open-channel SSDs.
18 *
19 */
20
21 #ifndef PBLK_H_
22 #define PBLK_H_
23
24 #include <linux/blkdev.h>
25 #include <linux/blk-mq.h>
26 #include <linux/bio.h>
27 #include <linux/module.h>
28 #include <linux/kthread.h>
29 #include <linux/vmalloc.h>
30 #include <linux/crc32.h>
31 #include <linux/uuid.h>
32
33 #include <linux/lightnvm.h>
34
35 /* Run only GC if less than 1/X blocks are free */
36 #define GC_LIMIT_INVERSE 5
37 #define GC_TIME_MSECS 1000
38
39 #define PBLK_SECTOR (512)
40 #define PBLK_EXPOSED_PAGE_SIZE (4096)
41
42 #define PBLK_NR_CLOSE_JOBS (4)
43
44 #define PBLK_CACHE_NAME_LEN (DISK_NAME_LEN + 16)
45
46 /* Max 512 LUNs per device */
47 #define PBLK_MAX_LUNS_BITMAP (4)
48
49 #define NR_PHY_IN_LOG (PBLK_EXPOSED_PAGE_SIZE / PBLK_SECTOR)
50
51 /* Static pool sizes */
52 #define PBLK_GEN_WS_POOL_SIZE (2)
53
54 #define PBLK_DEFAULT_OP (11)
55
56 enum {
57 PBLK_READ = READ,
58 PBLK_WRITE = WRITE,/* Write from write buffer */
59 PBLK_WRITE_INT, /* Internal write - no write buffer */
60 PBLK_READ_RECOV, /* Recovery read - errors allowed */
61 PBLK_ERASE,
62 };
63
64 enum {
65 /* IO Types */
66 PBLK_IOTYPE_USER = 1 << 0,
67 PBLK_IOTYPE_GC = 1 << 1,
68
69 /* Write buffer flags */
70 PBLK_FLUSH_ENTRY = 1 << 2,
71 PBLK_WRITTEN_DATA = 1 << 3,
72 PBLK_SUBMITTED_ENTRY = 1 << 4,
73 PBLK_WRITABLE_ENTRY = 1 << 5,
74 };
75
76 enum {
77 PBLK_BLK_ST_OPEN = 0x1,
78 PBLK_BLK_ST_CLOSED = 0x2,
79 };
80
81 enum {
82 PBLK_CHUNK_RESET_START,
83 PBLK_CHUNK_RESET_DONE,
84 PBLK_CHUNK_RESET_FAILED,
85 };
86
87 struct pblk_sec_meta {
88 u64 reserved;
89 __le64 lba;
90 };
91
92 /* The number of GC lists and the rate-limiter states go together. This way the
93 * rate-limiter can dictate how much GC is needed based on resource utilization.
94 */
95 #define PBLK_GC_NR_LISTS 4
96
97 enum {
98 PBLK_RL_OFF = 0,
99 PBLK_RL_WERR = 1,
100 PBLK_RL_HIGH = 2,
101 PBLK_RL_MID = 3,
102 PBLK_RL_LOW = 4
103 };
104
105 #define pblk_dma_ppa_size (sizeof(u64) * NVM_MAX_VLBA)
106
107 /* write buffer completion context */
108 struct pblk_c_ctx {
109 struct list_head list; /* Head for out-of-order completion */
110
111 unsigned long *lun_bitmap; /* Luns used on current request */
112 unsigned int sentry;
113 unsigned int nr_valid;
114 unsigned int nr_padded;
115 };
116
117 /* read context */
118 struct pblk_g_ctx {
119 void *private;
120 unsigned long start_time;
121 u64 lba;
122 };
123
124 /* Pad context */
125 struct pblk_pad_rq {
126 struct pblk *pblk;
127 struct completion wait;
128 struct kref ref;
129 };
130
131 /* Recovery context */
132 struct pblk_rec_ctx {
133 struct pblk *pblk;
134 struct nvm_rq *rqd;
135 struct work_struct ws_rec;
136 };
137
138 /* Write context */
139 struct pblk_w_ctx {
140 struct bio_list bios; /* Original bios - used for completion
141 * in REQ_FUA, REQ_FLUSH case
142 */
143 u64 lba; /* Logic addr. associated with entry */
144 struct ppa_addr ppa; /* Physic addr. associated with entry */
145 int flags; /* Write context flags */
146 };
147
148 struct pblk_rb_entry {
149 struct ppa_addr cacheline; /* Cacheline for this entry */
150 void *data; /* Pointer to data on this entry */
151 struct pblk_w_ctx w_ctx; /* Context for this entry */
152 struct list_head index; /* List head to enable indexes */
153 };
154
155 #define EMPTY_ENTRY (~0U)
156
157 struct pblk_rb_pages {
158 struct page *pages;
159 int order;
160 struct list_head list;
161 };
162
163 struct pblk_rb {
164 struct pblk_rb_entry *entries; /* Ring buffer entries */
165 unsigned int mem; /* Write offset - points to next
166 * writable entry in memory
167 */
168 unsigned int subm; /* Read offset - points to last entry
169 * that has been submitted to the media
170 * to be persisted
171 */
172 unsigned int sync; /* Synced - backpointer that signals
173 * the last submitted entry that has
174 * been successfully persisted to media
175 */
176 unsigned int flush_point; /* Sync point - last entry that must be
177 * flushed to the media. Used with
178 * REQ_FLUSH and REQ_FUA
179 */
180 unsigned int l2p_update; /* l2p update point - next entry for
181 * which l2p mapping will be updated to
182 * contain a device ppa address (instead
183 * of a cacheline
184 */
185 unsigned int nr_entries; /* Number of entries in write buffer -
186 * must be a power of two
187 */
188 unsigned int seg_size; /* Size of the data segments being
189 * stored on each entry. Typically this
190 * will be 4KB
191 */
192
193 unsigned int back_thres; /* Threshold that shall be maintained by
194 * the backpointer in order to respect
195 * geo->mw_cunits on a per chunk basis
196 */
197
198 struct list_head pages; /* List of data pages */
199
200 spinlock_t w_lock; /* Write lock */
201 spinlock_t s_lock; /* Sync lock */
202
203 #ifdef CONFIG_NVM_PBLK_DEBUG
204 atomic_t inflight_flush_point; /* Not served REQ_FLUSH | REQ_FUA */
205 #endif
206 };
207
208 #define PBLK_RECOVERY_SECTORS 16
209
210 struct pblk_lun {
211 struct ppa_addr bppa;
212 struct semaphore wr_sem;
213 };
214
215 struct pblk_gc_rq {
216 struct pblk_line *line;
217 void *data;
218 u64 paddr_list[NVM_MAX_VLBA];
219 u64 lba_list[NVM_MAX_VLBA];
220 int nr_secs;
221 int secs_to_gc;
222 struct list_head list;
223 };
224
225 struct pblk_gc {
226 /* These states are not protected by a lock since (i) they are in the
227 * fast path, and (ii) they are not critical.
228 */
229 int gc_active;
230 int gc_enabled;
231 int gc_forced;
232
233 struct task_struct *gc_ts;
234 struct task_struct *gc_writer_ts;
235 struct task_struct *gc_reader_ts;
236
237 struct workqueue_struct *gc_line_reader_wq;
238 struct workqueue_struct *gc_reader_wq;
239
240 struct timer_list gc_timer;
241
242 struct semaphore gc_sem;
243 atomic_t read_inflight_gc; /* Number of lines with inflight GC reads */
244 atomic_t pipeline_gc; /* Number of lines in the GC pipeline -
245 * started reads to finished writes
246 */
247 int w_entries;
248
249 struct list_head w_list;
250 struct list_head r_list;
251
252 spinlock_t lock;
253 spinlock_t w_lock;
254 spinlock_t r_lock;
255 };
256
257 struct pblk_rl {
258 unsigned int high; /* Upper threshold for rate limiter (free run -
259 * user I/O rate limiter
260 */
261 unsigned int high_pw; /* High rounded up as a power of 2 */
262
263 #define PBLK_USER_HIGH_THRS 8 /* Begin write limit at 12% available blks */
264 #define PBLK_USER_LOW_THRS 10 /* Aggressive GC at 10% available blocks */
265
266 int rb_windows_pw; /* Number of rate windows in the write buffer
267 * given as a power-of-2. This guarantees that
268 * when user I/O is being rate limited, there
269 * will be reserved enough space for the GC to
270 * place its payload. A window is of
271 * pblk->max_write_pgs size, which in NVMe is
272 * 64, i.e., 256kb.
273 */
274 int rb_budget; /* Total number of entries available for I/O */
275 int rb_user_max; /* Max buffer entries available for user I/O */
276 int rb_gc_max; /* Max buffer entries available for GC I/O */
277 int rb_gc_rsv; /* Reserved buffer entries for GC I/O */
278 int rb_state; /* Rate-limiter current state */
279 int rb_max_io; /* Maximum size for an I/O giving the config */
280
281 atomic_t rb_user_cnt; /* User I/O buffer counter */
282 atomic_t rb_gc_cnt; /* GC I/O buffer counter */
283 atomic_t rb_space; /* Space limit in case of reaching capacity */
284
285 int rsv_blocks; /* Reserved blocks for GC */
286
287 int rb_user_active;
288 int rb_gc_active;
289
290 atomic_t werr_lines; /* Number of write error lines that needs gc */
291
292 struct timer_list u_timer;
293
294 unsigned long total_blocks;
295
296 atomic_t free_blocks; /* Total number of free blocks (+ OP) */
297 atomic_t free_user_blocks; /* Number of user free blocks (no OP) */
298 };
299
300 #define PBLK_LINE_EMPTY (~0U)
301
302 enum {
303 /* Line Types */
304 PBLK_LINETYPE_FREE = 0,
305 PBLK_LINETYPE_LOG = 1,
306 PBLK_LINETYPE_DATA = 2,
307
308 /* Line state */
309 PBLK_LINESTATE_NEW = 9,
310 PBLK_LINESTATE_FREE = 10,
311 PBLK_LINESTATE_OPEN = 11,
312 PBLK_LINESTATE_CLOSED = 12,
313 PBLK_LINESTATE_GC = 13,
314 PBLK_LINESTATE_BAD = 14,
315 PBLK_LINESTATE_CORRUPT = 15,
316
317 /* GC group */
318 PBLK_LINEGC_NONE = 20,
319 PBLK_LINEGC_EMPTY = 21,
320 PBLK_LINEGC_LOW = 22,
321 PBLK_LINEGC_MID = 23,
322 PBLK_LINEGC_HIGH = 24,
323 PBLK_LINEGC_FULL = 25,
324 PBLK_LINEGC_WERR = 26
325 };
326
327 #define PBLK_MAGIC 0x70626c6b /*pblk*/
328
329 /* emeta/smeta persistent storage format versions:
330 * Changes in major version requires offline migration.
331 * Changes in minor version are handled automatically during
332 * recovery.
333 */
334
335 #define SMETA_VERSION_MAJOR (0)
336 #define SMETA_VERSION_MINOR (1)
337
338 #define EMETA_VERSION_MAJOR (0)
339 #define EMETA_VERSION_MINOR (2)
340
341 struct line_header {
342 __le32 crc;
343 __le32 identifier; /* pblk identifier */
344 __u8 uuid[16]; /* instance uuid */
345 __le16 type; /* line type */
346 __u8 version_major; /* version major */
347 __u8 version_minor; /* version minor */
348 __le32 id; /* line id for current line */
349 };
350
351 struct line_smeta {
352 struct line_header header;
353
354 __le32 crc; /* Full structure including struct crc */
355 /* Previous line metadata */
356 __le32 prev_id; /* Line id for previous line */
357
358 /* Current line metadata */
359 __le64 seq_nr; /* Sequence number for current line */
360
361 /* Active writers */
362 __le32 window_wr_lun; /* Number of parallel LUNs to write */
363
364 __le32 rsvd[2];
365
366 __le64 lun_bitmap[];
367 };
368
369
370 /*
371 * Metadata layout in media:
372 * First sector:
373 * 1. struct line_emeta
374 * 2. bad block bitmap (u64 * window_wr_lun)
375 * 3. write amplification counters
376 * Mid sectors (start at lbas_sector):
377 * 3. nr_lbas (u64) forming lba list
378 * Last sectors (start at vsc_sector):
379 * 4. u32 valid sector count (vsc) for all lines (~0U: free line)
380 */
381 struct line_emeta {
382 struct line_header header;
383
384 __le32 crc; /* Full structure including struct crc */
385
386 /* Previous line metadata */
387 __le32 prev_id; /* Line id for prev line */
388
389 /* Current line metadata */
390 __le64 seq_nr; /* Sequence number for current line */
391
392 /* Active writers */
393 __le32 window_wr_lun; /* Number of parallel LUNs to write */
394
395 /* Bookkeeping for recovery */
396 __le32 next_id; /* Line id for next line */
397 __le64 nr_lbas; /* Number of lbas mapped in line */
398 __le64 nr_valid_lbas; /* Number of valid lbas mapped in line */
399 __le64 bb_bitmap[]; /* Updated bad block bitmap for line */
400 };
401
402
403 /* Write amplification counters stored on media */
404 struct wa_counters {
405 __le64 user; /* Number of user written sectors */
406 __le64 gc; /* Number of sectors written by GC*/
407 __le64 pad; /* Number of padded sectors */
408 };
409
410 struct pblk_emeta {
411 struct line_emeta *buf; /* emeta buffer in media format */
412 int mem; /* Write offset - points to next
413 * writable entry in memory
414 */
415 atomic_t sync; /* Synced - backpointer that signals the
416 * last entry that has been successfully
417 * persisted to media
418 */
419 unsigned int nr_entries; /* Number of emeta entries */
420 };
421
422 struct pblk_smeta {
423 struct line_smeta *buf; /* smeta buffer in persistent format */
424 };
425
426 struct pblk_w_err_gc {
427 int has_write_err;
428 int has_gc_err;
429 __le64 *lba_list;
430 };
431
432 struct pblk_line {
433 struct pblk *pblk;
434 unsigned int id; /* Line number corresponds to the
435 * block line
436 */
437 unsigned int seq_nr; /* Unique line sequence number */
438
439 int state; /* PBLK_LINESTATE_X */
440 int type; /* PBLK_LINETYPE_X */
441 int gc_group; /* PBLK_LINEGC_X */
442 struct list_head list; /* Free, GC lists */
443
444 unsigned long *lun_bitmap; /* Bitmap for LUNs mapped in line */
445
446 struct nvm_chk_meta *chks; /* Chunks forming line */
447
448 struct pblk_smeta *smeta; /* Start metadata */
449 struct pblk_emeta *emeta; /* End medatada */
450
451 int meta_line; /* Metadata line id */
452 int meta_distance; /* Distance between data and metadata */
453
454 u64 emeta_ssec; /* Sector where emeta starts */
455
456 unsigned int sec_in_line; /* Number of usable secs in line */
457
458 atomic_t blk_in_line; /* Number of good blocks in line */
459 unsigned long *blk_bitmap; /* Bitmap for valid/invalid blocks */
460 unsigned long *erase_bitmap; /* Bitmap for erased blocks */
461
462 unsigned long *map_bitmap; /* Bitmap for mapped sectors in line */
463 unsigned long *invalid_bitmap; /* Bitmap for invalid sectors in line */
464
465 atomic_t left_eblks; /* Blocks left for erasing */
466 atomic_t left_seblks; /* Blocks left for sync erasing */
467
468 int left_msecs; /* Sectors left for mapping */
469 unsigned int cur_sec; /* Sector map pointer */
470 unsigned int nr_valid_lbas; /* Number of valid lbas in line */
471
472 __le32 *vsc; /* Valid sector count in line */
473
474 struct kref ref; /* Write buffer L2P references */
475 atomic_t sec_to_update; /* Outstanding L2P updates to ppa */
476
477 struct pblk_w_err_gc *w_err_gc; /* Write error gc recovery metadata */
478
479 spinlock_t lock; /* Necessary for invalid_bitmap only */
480 };
481
482 #define PBLK_DATA_LINES 4
483
484 enum {
485 PBLK_EMETA_TYPE_HEADER = 1, /* struct line_emeta first sector */
486 PBLK_EMETA_TYPE_LLBA = 2, /* lba list - type: __le64 */
487 PBLK_EMETA_TYPE_VSC = 3, /* vsc list - type: __le32 */
488 };
489
490 struct pblk_line_mgmt {
491 int nr_lines; /* Total number of full lines */
492 int nr_free_lines; /* Number of full lines in free list */
493
494 /* Free lists - use free_lock */
495 struct list_head free_list; /* Full lines ready to use */
496 struct list_head corrupt_list; /* Full lines corrupted */
497 struct list_head bad_list; /* Full lines bad */
498
499 /* GC lists - use gc_lock */
500 struct list_head *gc_lists[PBLK_GC_NR_LISTS];
501 struct list_head gc_high_list; /* Full lines ready to GC, high isc */
502 struct list_head gc_mid_list; /* Full lines ready to GC, mid isc */
503 struct list_head gc_low_list; /* Full lines ready to GC, low isc */
504
505 struct list_head gc_werr_list; /* Write err recovery list */
506
507 struct list_head gc_full_list; /* Full lines ready to GC, no valid */
508 struct list_head gc_empty_list; /* Full lines close, all valid */
509
510 struct pblk_line *log_line; /* Current FTL log line */
511 struct pblk_line *data_line; /* Current data line */
512 struct pblk_line *log_next; /* Next FTL log line */
513 struct pblk_line *data_next; /* Next data line */
514
515 struct list_head emeta_list; /* Lines queued to schedule emeta */
516
517 __le32 *vsc_list; /* Valid sector counts for all lines */
518
519 /* Pre-allocated metadata for data lines */
520 struct pblk_smeta *sline_meta[PBLK_DATA_LINES];
521 struct pblk_emeta *eline_meta[PBLK_DATA_LINES];
522 unsigned long meta_bitmap;
523
524 /* Cache and mempool for map/invalid bitmaps */
525 struct kmem_cache *bitmap_cache;
526 mempool_t *bitmap_pool;
527
528 /* Helpers for fast bitmap calculations */
529 unsigned long *bb_template;
530 unsigned long *bb_aux;
531
532 unsigned long d_seq_nr; /* Data line unique sequence number */
533 unsigned long l_seq_nr; /* Log line unique sequence number */
534
535 spinlock_t free_lock;
536 spinlock_t close_lock;
537 spinlock_t gc_lock;
538 };
539
540 struct pblk_line_meta {
541 unsigned int smeta_len; /* Total length for smeta */
542 unsigned int smeta_sec; /* Sectors needed for smeta */
543
544 unsigned int emeta_len[4]; /* Lengths for emeta:
545 * [0]: Total
546 * [1]: struct line_emeta +
547 * bb_bitmap + struct wa_counters
548 * [2]: L2P portion
549 * [3]: vsc
550 */
551 unsigned int emeta_sec[4]; /* Sectors needed for emeta. Same layout
552 * as emeta_len
553 */
554
555 unsigned int emeta_bb; /* Boundary for bb that affects emeta */
556
557 unsigned int vsc_list_len; /* Length for vsc list */
558 unsigned int sec_bitmap_len; /* Length for sector bitmap in line */
559 unsigned int blk_bitmap_len; /* Length for block bitmap in line */
560 unsigned int lun_bitmap_len; /* Length for lun bitmap in line */
561
562 unsigned int blk_per_line; /* Number of blocks in a full line */
563 unsigned int sec_per_line; /* Number of sectors in a line */
564 unsigned int dsec_per_line; /* Number of data sectors in a line */
565 unsigned int min_blk_line; /* Min. number of good blocks in line */
566
567 unsigned int mid_thrs; /* Threshold for GC mid list */
568 unsigned int high_thrs; /* Threshold for GC high list */
569
570 unsigned int meta_distance; /* Distance between data and metadata */
571 };
572
573 enum {
574 PBLK_STATE_RUNNING = 0,
575 PBLK_STATE_STOPPING = 1,
576 PBLK_STATE_RECOVERING = 2,
577 PBLK_STATE_STOPPED = 3,
578 };
579
580 /* Internal format to support not power-of-2 device formats */
581 struct pblk_addrf {
582 /* gen to dev */
583 int sec_stripe;
584 int ch_stripe;
585 int lun_stripe;
586
587 /* dev to gen */
588 int sec_lun_stripe;
589 int sec_ws_stripe;
590 };
591
592 struct pblk {
593 struct nvm_tgt_dev *dev;
594 struct gendisk *disk;
595
596 struct kobject kobj;
597
598 struct pblk_lun *luns;
599
600 struct pblk_line *lines; /* Line array */
601 struct pblk_line_mgmt l_mg; /* Line management */
602 struct pblk_line_meta lm; /* Line metadata */
603
604 struct nvm_addrf addrf; /* Aligned address format */
605 struct pblk_addrf uaddrf; /* Unaligned address format */
606 int addrf_len;
607
608 struct pblk_rb rwb;
609
610 int state; /* pblk line state */
611
612 int min_write_pgs; /* Minimum amount of pages required by controller */
613 int min_write_pgs_data; /* Minimum amount of payload pages */
614 int max_write_pgs; /* Maximum amount of pages supported by controller */
615 int oob_meta_size; /* Size of OOB sector metadata */
616
617 sector_t capacity; /* Device capacity when bad blocks are subtracted */
618
619 int op; /* Percentage of device used for over-provisioning */
620 int op_blks; /* Number of blocks used for over-provisioning */
621
622 /* pblk provisioning values. Used by rate limiter */
623 struct pblk_rl rl;
624
625 int sec_per_write;
626
627 guid_t instance_uuid;
628
629 /* Persistent write amplification counters, 4kb sector I/Os */
630 atomic64_t user_wa; /* Sectors written by user */
631 atomic64_t gc_wa; /* Sectors written by GC */
632 atomic64_t pad_wa; /* Padded sectors written */
633
634 /* Reset values for delta write amplification measurements */
635 u64 user_rst_wa;
636 u64 gc_rst_wa;
637 u64 pad_rst_wa;
638
639 /* Counters used for calculating padding distribution */
640 atomic64_t *pad_dist; /* Padding distribution buckets */
641 u64 nr_flush_rst; /* Flushes reset value for pad dist.*/
642 atomic64_t nr_flush; /* Number of flush/fua I/O */
643
644 #ifdef CONFIG_NVM_PBLK_DEBUG
645 /* Non-persistent debug counters, 4kb sector I/Os */
646 atomic_long_t inflight_writes; /* Inflight writes (user and gc) */
647 atomic_long_t padded_writes; /* Sectors padded due to flush/fua */
648 atomic_long_t padded_wb; /* Sectors padded in write buffer */
649 atomic_long_t req_writes; /* Sectors stored on write buffer */
650 atomic_long_t sub_writes; /* Sectors submitted from buffer */
651 atomic_long_t sync_writes; /* Sectors synced to media */
652 atomic_long_t inflight_reads; /* Inflight sector read requests */
653 atomic_long_t cache_reads; /* Read requests that hit the cache */
654 atomic_long_t sync_reads; /* Completed sector read requests */
655 atomic_long_t recov_writes; /* Sectors submitted from recovery */
656 atomic_long_t recov_gc_writes; /* Sectors submitted from write GC */
657 atomic_long_t recov_gc_reads; /* Sectors submitted from read GC */
658 #endif
659
660 spinlock_t lock;
661
662 atomic_long_t read_failed;
663 atomic_long_t read_empty;
664 atomic_long_t read_high_ecc;
665 atomic_long_t read_failed_gc;
666 atomic_long_t write_failed;
667 atomic_long_t erase_failed;
668
669 atomic_t inflight_io; /* General inflight I/O counter */
670
671 struct task_struct *writer_ts;
672
673 /* Simple translation map of logical addresses to physical addresses.
674 * The logical addresses is known by the host system, while the physical
675 * addresses are used when writing to the disk block device.
676 */
677 unsigned char *trans_map;
678 spinlock_t trans_lock;
679
680 struct list_head compl_list;
681
682 spinlock_t resubmit_lock; /* Resubmit list lock */
683 struct list_head resubmit_list; /* Resubmit list for failed writes*/
684
685 mempool_t page_bio_pool;
686 mempool_t gen_ws_pool;
687 mempool_t rec_pool;
688 mempool_t r_rq_pool;
689 mempool_t w_rq_pool;
690 mempool_t e_rq_pool;
691
692 struct workqueue_struct *close_wq;
693 struct workqueue_struct *bb_wq;
694 struct workqueue_struct *r_end_wq;
695
696 struct timer_list wtimer;
697
698 struct pblk_gc gc;
699 };
700
701 struct pblk_line_ws {
702 struct pblk *pblk;
703 struct pblk_line *line;
704 void *priv;
705 struct work_struct ws;
706 };
707
708 #define pblk_g_rq_size (sizeof(struct nvm_rq) + sizeof(struct pblk_g_ctx))
709 #define pblk_w_rq_size (sizeof(struct nvm_rq) + sizeof(struct pblk_c_ctx))
710
711 #define pblk_err(pblk, fmt, ...) \
712 pr_err("pblk %s: " fmt, pblk->disk->disk_name, ##__VA_ARGS__)
713 #define pblk_info(pblk, fmt, ...) \
714 pr_info("pblk %s: " fmt, pblk->disk->disk_name, ##__VA_ARGS__)
715 #define pblk_warn(pblk, fmt, ...) \
716 pr_warn("pblk %s: " fmt, pblk->disk->disk_name, ##__VA_ARGS__)
717 #define pblk_debug(pblk, fmt, ...) \
718 pr_debug("pblk %s: " fmt, pblk->disk->disk_name, ##__VA_ARGS__)
719
720 /*
721 * pblk ring buffer operations
722 */
723 int pblk_rb_init(struct pblk_rb *rb, unsigned int size, unsigned int threshold,
724 unsigned int seg_sz);
725 int pblk_rb_may_write_user(struct pblk_rb *rb, struct bio *bio,
726 unsigned int nr_entries, unsigned int *pos);
727 int pblk_rb_may_write_gc(struct pblk_rb *rb, unsigned int nr_entries,
728 unsigned int *pos);
729 void pblk_rb_write_entry_user(struct pblk_rb *rb, void *data,
730 struct pblk_w_ctx w_ctx, unsigned int pos);
731 void pblk_rb_write_entry_gc(struct pblk_rb *rb, void *data,
732 struct pblk_w_ctx w_ctx, struct pblk_line *line,
733 u64 paddr, unsigned int pos);
734 struct pblk_w_ctx *pblk_rb_w_ctx(struct pblk_rb *rb, unsigned int pos);
735 void pblk_rb_flush(struct pblk_rb *rb);
736
737 void pblk_rb_sync_l2p(struct pblk_rb *rb);
738 unsigned int pblk_rb_read_to_bio(struct pblk_rb *rb, struct nvm_rq *rqd,
739 unsigned int pos, unsigned int nr_entries,
740 unsigned int count);
741 int pblk_rb_copy_to_bio(struct pblk_rb *rb, struct bio *bio, sector_t lba,
742 struct ppa_addr ppa);
743 unsigned int pblk_rb_read_commit(struct pblk_rb *rb, unsigned int entries);
744
745 unsigned int pblk_rb_sync_init(struct pblk_rb *rb, unsigned long *flags);
746 unsigned int pblk_rb_sync_advance(struct pblk_rb *rb, unsigned int nr_entries);
747 unsigned int pblk_rb_ptr_wrap(struct pblk_rb *rb, unsigned int p,
748 unsigned int nr_entries);
749 void pblk_rb_sync_end(struct pblk_rb *rb, unsigned long *flags);
750 unsigned int pblk_rb_flush_point_count(struct pblk_rb *rb);
751
752 unsigned int pblk_rb_read_count(struct pblk_rb *rb);
753 unsigned int pblk_rb_sync_count(struct pblk_rb *rb);
754 unsigned int pblk_rb_wrap_pos(struct pblk_rb *rb, unsigned int pos);
755
756 int pblk_rb_tear_down_check(struct pblk_rb *rb);
757 int pblk_rb_pos_oob(struct pblk_rb *rb, u64 pos);
758 void pblk_rb_free(struct pblk_rb *rb);
759 ssize_t pblk_rb_sysfs(struct pblk_rb *rb, char *buf);
760
761 /*
762 * pblk core
763 */
764 struct nvm_rq *pblk_alloc_rqd(struct pblk *pblk, int type);
765 void pblk_free_rqd(struct pblk *pblk, struct nvm_rq *rqd, int type);
766 int pblk_alloc_rqd_meta(struct pblk *pblk, struct nvm_rq *rqd);
767 void pblk_free_rqd_meta(struct pblk *pblk, struct nvm_rq *rqd);
768 void pblk_set_sec_per_write(struct pblk *pblk, int sec_per_write);
769 int pblk_setup_w_rec_rq(struct pblk *pblk, struct nvm_rq *rqd,
770 struct pblk_c_ctx *c_ctx);
771 void pblk_discard(struct pblk *pblk, struct bio *bio);
772 struct nvm_chk_meta *pblk_get_chunk_meta(struct pblk *pblk);
773 struct nvm_chk_meta *pblk_chunk_get_off(struct pblk *pblk,
774 struct nvm_chk_meta *lp,
775 struct ppa_addr ppa);
776 void pblk_log_write_err(struct pblk *pblk, struct nvm_rq *rqd);
777 void pblk_log_read_err(struct pblk *pblk, struct nvm_rq *rqd);
778 int pblk_submit_io(struct pblk *pblk, struct nvm_rq *rqd, void *buf);
779 int pblk_submit_io_sync(struct pblk *pblk, struct nvm_rq *rqd, void *buf);
780 int pblk_submit_meta_io(struct pblk *pblk, struct pblk_line *meta_line);
781 void pblk_check_chunk_state_update(struct pblk *pblk, struct nvm_rq *rqd);
782 struct pblk_line *pblk_line_get(struct pblk *pblk);
783 struct pblk_line *pblk_line_get_first_data(struct pblk *pblk);
784 struct pblk_line *pblk_line_replace_data(struct pblk *pblk);
785 void pblk_ppa_to_line_put(struct pblk *pblk, struct ppa_addr ppa);
786 void pblk_rq_to_line_put(struct pblk *pblk, struct nvm_rq *rqd);
787 int pblk_line_recov_alloc(struct pblk *pblk, struct pblk_line *line);
788 void pblk_line_recov_close(struct pblk *pblk, struct pblk_line *line);
789 struct pblk_line *pblk_line_get_data(struct pblk *pblk);
790 struct pblk_line *pblk_line_get_erase(struct pblk *pblk);
791 int pblk_line_erase(struct pblk *pblk, struct pblk_line *line);
792 int pblk_line_is_full(struct pblk_line *line);
793 void pblk_line_free(struct pblk_line *line);
794 void pblk_line_close_meta(struct pblk *pblk, struct pblk_line *line);
795 void pblk_line_close(struct pblk *pblk, struct pblk_line *line);
796 void pblk_line_close_ws(struct work_struct *work);
797 void pblk_pipeline_stop(struct pblk *pblk);
798 void __pblk_pipeline_stop(struct pblk *pblk);
799 void __pblk_pipeline_flush(struct pblk *pblk);
800 void pblk_gen_run_ws(struct pblk *pblk, struct pblk_line *line, void *priv,
801 void (*work)(struct work_struct *), gfp_t gfp_mask,
802 struct workqueue_struct *wq);
803 u64 pblk_line_smeta_start(struct pblk *pblk, struct pblk_line *line);
804 int pblk_line_smeta_read(struct pblk *pblk, struct pblk_line *line);
805 int pblk_line_emeta_read(struct pblk *pblk, struct pblk_line *line,
806 void *emeta_buf);
807 int pblk_blk_erase_async(struct pblk *pblk, struct ppa_addr erase_ppa);
808 void pblk_line_put(struct kref *ref);
809 void pblk_line_put_wq(struct kref *ref);
810 struct list_head *pblk_line_gc_list(struct pblk *pblk, struct pblk_line *line);
811 u64 pblk_lookup_page(struct pblk *pblk, struct pblk_line *line);
812 void pblk_dealloc_page(struct pblk *pblk, struct pblk_line *line, int nr_secs);
813 u64 pblk_alloc_page(struct pblk *pblk, struct pblk_line *line, int nr_secs);
814 u64 __pblk_alloc_page(struct pblk *pblk, struct pblk_line *line, int nr_secs);
815 int pblk_calc_secs(struct pblk *pblk, unsigned long secs_avail,
816 unsigned long secs_to_flush, bool skip_meta);
817 void pblk_down_rq(struct pblk *pblk, struct ppa_addr ppa,
818 unsigned long *lun_bitmap);
819 void pblk_down_chunk(struct pblk *pblk, struct ppa_addr ppa);
820 void pblk_up_chunk(struct pblk *pblk, struct ppa_addr ppa);
821 void pblk_up_rq(struct pblk *pblk, unsigned long *lun_bitmap);
822 int pblk_bio_add_pages(struct pblk *pblk, struct bio *bio, gfp_t flags,
823 int nr_pages);
824 void pblk_bio_free_pages(struct pblk *pblk, struct bio *bio, int off,
825 int nr_pages);
826 void pblk_map_invalidate(struct pblk *pblk, struct ppa_addr ppa);
827 void __pblk_map_invalidate(struct pblk *pblk, struct pblk_line *line,
828 u64 paddr);
829 void pblk_update_map(struct pblk *pblk, sector_t lba, struct ppa_addr ppa);
830 void pblk_update_map_cache(struct pblk *pblk, sector_t lba,
831 struct ppa_addr ppa);
832 void pblk_update_map_dev(struct pblk *pblk, sector_t lba,
833 struct ppa_addr ppa, struct ppa_addr entry_line);
834 int pblk_update_map_gc(struct pblk *pblk, sector_t lba, struct ppa_addr ppa,
835 struct pblk_line *gc_line, u64 paddr);
836 void pblk_lookup_l2p_rand(struct pblk *pblk, struct ppa_addr *ppas,
837 u64 *lba_list, int nr_secs);
838 int pblk_lookup_l2p_seq(struct pblk *pblk, struct ppa_addr *ppas,
839 sector_t blba, int nr_secs, bool *from_cache);
840 void *pblk_get_meta_for_writes(struct pblk *pblk, struct nvm_rq *rqd);
841 void pblk_get_packed_meta(struct pblk *pblk, struct nvm_rq *rqd);
842
843 /*
844 * pblk user I/O write path
845 */
846 void pblk_write_to_cache(struct pblk *pblk, struct bio *bio,
847 unsigned long flags);
848 int pblk_write_gc_to_cache(struct pblk *pblk, struct pblk_gc_rq *gc_rq);
849
850 /*
851 * pblk map
852 */
853 int pblk_map_erase_rq(struct pblk *pblk, struct nvm_rq *rqd,
854 unsigned int sentry, unsigned long *lun_bitmap,
855 unsigned int valid_secs, struct ppa_addr *erase_ppa);
856 int pblk_map_rq(struct pblk *pblk, struct nvm_rq *rqd, unsigned int sentry,
857 unsigned long *lun_bitmap, unsigned int valid_secs,
858 unsigned int off);
859
860 /*
861 * pblk write thread
862 */
863 int pblk_write_ts(void *data);
864 void pblk_write_timer_fn(struct timer_list *t);
865 void pblk_write_should_kick(struct pblk *pblk);
866 void pblk_write_kick(struct pblk *pblk);
867
868 /*
869 * pblk read path
870 */
871 extern struct bio_set pblk_bio_set;
872 void pblk_submit_read(struct pblk *pblk, struct bio *bio);
873 int pblk_submit_read_gc(struct pblk *pblk, struct pblk_gc_rq *gc_rq);
874 /*
875 * pblk recovery
876 */
877 struct pblk_line *pblk_recov_l2p(struct pblk *pblk);
878 int pblk_recov_pad(struct pblk *pblk);
879 int pblk_recov_check_emeta(struct pblk *pblk, struct line_emeta *emeta);
880
881 /*
882 * pblk gc
883 */
884 #define PBLK_GC_MAX_READERS 8 /* Max number of outstanding GC reader jobs */
885 #define PBLK_GC_RQ_QD 128 /* Queue depth for inflight GC requests */
886 #define PBLK_GC_L_QD 4 /* Queue depth for inflight GC lines */
887
888 int pblk_gc_init(struct pblk *pblk);
889 void pblk_gc_exit(struct pblk *pblk, bool graceful);
890 void pblk_gc_should_start(struct pblk *pblk);
891 void pblk_gc_should_stop(struct pblk *pblk);
892 void pblk_gc_should_kick(struct pblk *pblk);
893 void pblk_gc_free_full_lines(struct pblk *pblk);
894 void pblk_gc_sysfs_state_show(struct pblk *pblk, int *gc_enabled,
895 int *gc_active);
896 int pblk_gc_sysfs_force(struct pblk *pblk, int force);
897 void pblk_put_line_back(struct pblk *pblk, struct pblk_line *line);
898
899 /*
900 * pblk rate limiter
901 */
902 void pblk_rl_init(struct pblk_rl *rl, int budget, int threshold);
903 void pblk_rl_free(struct pblk_rl *rl);
904 void pblk_rl_update_rates(struct pblk_rl *rl);
905 int pblk_rl_high_thrs(struct pblk_rl *rl);
906 unsigned long pblk_rl_nr_free_blks(struct pblk_rl *rl);
907 unsigned long pblk_rl_nr_user_free_blks(struct pblk_rl *rl);
908 int pblk_rl_user_may_insert(struct pblk_rl *rl, int nr_entries);
909 void pblk_rl_inserted(struct pblk_rl *rl, int nr_entries);
910 void pblk_rl_user_in(struct pblk_rl *rl, int nr_entries);
911 int pblk_rl_gc_may_insert(struct pblk_rl *rl, int nr_entries);
912 void pblk_rl_gc_in(struct pblk_rl *rl, int nr_entries);
913 void pblk_rl_out(struct pblk_rl *rl, int nr_user, int nr_gc);
914 int pblk_rl_max_io(struct pblk_rl *rl);
915 void pblk_rl_free_lines_inc(struct pblk_rl *rl, struct pblk_line *line);
916 void pblk_rl_free_lines_dec(struct pblk_rl *rl, struct pblk_line *line,
917 bool used);
918 int pblk_rl_is_limit(struct pblk_rl *rl);
919
920 void pblk_rl_werr_line_in(struct pblk_rl *rl);
921 void pblk_rl_werr_line_out(struct pblk_rl *rl);
922
923 /*
924 * pblk sysfs
925 */
926 int pblk_sysfs_init(struct gendisk *tdisk);
927 void pblk_sysfs_exit(struct gendisk *tdisk);
928
nvm_rq_from_c_ctx(void * c_ctx)929 static inline struct nvm_rq *nvm_rq_from_c_ctx(void *c_ctx)
930 {
931 return c_ctx - sizeof(struct nvm_rq);
932 }
933
emeta_to_bb(struct line_emeta * emeta)934 static inline void *emeta_to_bb(struct line_emeta *emeta)
935 {
936 return emeta->bb_bitmap;
937 }
938
emeta_to_wa(struct pblk_line_meta * lm,struct line_emeta * emeta)939 static inline void *emeta_to_wa(struct pblk_line_meta *lm,
940 struct line_emeta *emeta)
941 {
942 return emeta->bb_bitmap + lm->blk_bitmap_len;
943 }
944
emeta_to_lbas(struct pblk * pblk,struct line_emeta * emeta)945 static inline void *emeta_to_lbas(struct pblk *pblk, struct line_emeta *emeta)
946 {
947 return ((void *)emeta + pblk->lm.emeta_len[1]);
948 }
949
emeta_to_vsc(struct pblk * pblk,struct line_emeta * emeta)950 static inline void *emeta_to_vsc(struct pblk *pblk, struct line_emeta *emeta)
951 {
952 return (emeta_to_lbas(pblk, emeta) + pblk->lm.emeta_len[2]);
953 }
954
pblk_line_vsc(struct pblk_line * line)955 static inline int pblk_line_vsc(struct pblk_line *line)
956 {
957 return le32_to_cpu(*line->vsc);
958 }
959
pblk_ppa_to_line_id(struct ppa_addr p)960 static inline int pblk_ppa_to_line_id(struct ppa_addr p)
961 {
962 return p.a.blk;
963 }
964
pblk_ppa_to_line(struct pblk * pblk,struct ppa_addr p)965 static inline struct pblk_line *pblk_ppa_to_line(struct pblk *pblk,
966 struct ppa_addr p)
967 {
968 return &pblk->lines[pblk_ppa_to_line_id(p)];
969 }
970
pblk_ppa_to_pos(struct nvm_geo * geo,struct ppa_addr p)971 static inline int pblk_ppa_to_pos(struct nvm_geo *geo, struct ppa_addr p)
972 {
973 return p.a.lun * geo->num_ch + p.a.ch;
974 }
975
addr_to_gen_ppa(struct pblk * pblk,u64 paddr,u64 line_id)976 static inline struct ppa_addr addr_to_gen_ppa(struct pblk *pblk, u64 paddr,
977 u64 line_id)
978 {
979 struct nvm_tgt_dev *dev = pblk->dev;
980 struct nvm_geo *geo = &dev->geo;
981 struct ppa_addr ppa;
982
983 if (geo->version == NVM_OCSSD_SPEC_12) {
984 struct nvm_addrf_12 *ppaf = (struct nvm_addrf_12 *)&pblk->addrf;
985
986 ppa.ppa = 0;
987 ppa.g.blk = line_id;
988 ppa.g.pg = (paddr & ppaf->pg_mask) >> ppaf->pg_offset;
989 ppa.g.lun = (paddr & ppaf->lun_mask) >> ppaf->lun_offset;
990 ppa.g.ch = (paddr & ppaf->ch_mask) >> ppaf->ch_offset;
991 ppa.g.pl = (paddr & ppaf->pln_mask) >> ppaf->pln_offset;
992 ppa.g.sec = (paddr & ppaf->sec_mask) >> ppaf->sec_offset;
993 } else {
994 struct pblk_addrf *uaddrf = &pblk->uaddrf;
995 int secs, chnls, luns;
996
997 ppa.ppa = 0;
998
999 ppa.m.chk = line_id;
1000
1001 paddr = div_u64_rem(paddr, uaddrf->sec_stripe, &secs);
1002 ppa.m.sec = secs;
1003
1004 paddr = div_u64_rem(paddr, uaddrf->ch_stripe, &chnls);
1005 ppa.m.grp = chnls;
1006
1007 paddr = div_u64_rem(paddr, uaddrf->lun_stripe, &luns);
1008 ppa.m.pu = luns;
1009
1010 ppa.m.sec += uaddrf->sec_stripe * paddr;
1011 }
1012
1013 return ppa;
1014 }
1015
pblk_dev_ppa_to_chunk(struct pblk * pblk,struct ppa_addr p)1016 static inline struct nvm_chk_meta *pblk_dev_ppa_to_chunk(struct pblk *pblk,
1017 struct ppa_addr p)
1018 {
1019 struct nvm_tgt_dev *dev = pblk->dev;
1020 struct nvm_geo *geo = &dev->geo;
1021 struct pblk_line *line = pblk_ppa_to_line(pblk, p);
1022 int pos = pblk_ppa_to_pos(geo, p);
1023
1024 return &line->chks[pos];
1025 }
1026
pblk_dev_ppa_to_chunk_addr(struct pblk * pblk,struct ppa_addr p)1027 static inline u64 pblk_dev_ppa_to_chunk_addr(struct pblk *pblk,
1028 struct ppa_addr p)
1029 {
1030 struct nvm_tgt_dev *dev = pblk->dev;
1031
1032 return dev_to_chunk_addr(dev->parent, &pblk->addrf, p);
1033 }
1034
pblk_dev_ppa_to_line_addr(struct pblk * pblk,struct ppa_addr p)1035 static inline u64 pblk_dev_ppa_to_line_addr(struct pblk *pblk,
1036 struct ppa_addr p)
1037 {
1038 struct nvm_tgt_dev *dev = pblk->dev;
1039 struct nvm_geo *geo = &dev->geo;
1040 u64 paddr;
1041
1042 if (geo->version == NVM_OCSSD_SPEC_12) {
1043 struct nvm_addrf_12 *ppaf = (struct nvm_addrf_12 *)&pblk->addrf;
1044
1045 paddr = (u64)p.g.ch << ppaf->ch_offset;
1046 paddr |= (u64)p.g.lun << ppaf->lun_offset;
1047 paddr |= (u64)p.g.pg << ppaf->pg_offset;
1048 paddr |= (u64)p.g.pl << ppaf->pln_offset;
1049 paddr |= (u64)p.g.sec << ppaf->sec_offset;
1050 } else {
1051 struct pblk_addrf *uaddrf = &pblk->uaddrf;
1052 u64 secs = p.m.sec;
1053 int sec_stripe;
1054
1055 paddr = (u64)p.m.grp * uaddrf->sec_stripe;
1056 paddr += (u64)p.m.pu * uaddrf->sec_lun_stripe;
1057
1058 secs = div_u64_rem(secs, uaddrf->sec_stripe, &sec_stripe);
1059 paddr += secs * uaddrf->sec_ws_stripe;
1060 paddr += sec_stripe;
1061 }
1062
1063 return paddr;
1064 }
1065
pblk_ppa32_to_ppa64(struct pblk * pblk,u32 ppa32)1066 static inline struct ppa_addr pblk_ppa32_to_ppa64(struct pblk *pblk, u32 ppa32)
1067 {
1068 struct nvm_tgt_dev *dev = pblk->dev;
1069
1070 return nvm_ppa32_to_ppa64(dev->parent, &pblk->addrf, ppa32);
1071 }
1072
pblk_ppa64_to_ppa32(struct pblk * pblk,struct ppa_addr ppa64)1073 static inline u32 pblk_ppa64_to_ppa32(struct pblk *pblk, struct ppa_addr ppa64)
1074 {
1075 struct nvm_tgt_dev *dev = pblk->dev;
1076
1077 return nvm_ppa64_to_ppa32(dev->parent, &pblk->addrf, ppa64);
1078 }
1079
pblk_trans_map_get(struct pblk * pblk,sector_t lba)1080 static inline struct ppa_addr pblk_trans_map_get(struct pblk *pblk,
1081 sector_t lba)
1082 {
1083 struct ppa_addr ppa;
1084
1085 if (pblk->addrf_len < 32) {
1086 u32 *map = (u32 *)pblk->trans_map;
1087
1088 ppa = pblk_ppa32_to_ppa64(pblk, map[lba]);
1089 } else {
1090 struct ppa_addr *map = (struct ppa_addr *)pblk->trans_map;
1091
1092 ppa = map[lba];
1093 }
1094
1095 return ppa;
1096 }
1097
pblk_trans_map_set(struct pblk * pblk,sector_t lba,struct ppa_addr ppa)1098 static inline void pblk_trans_map_set(struct pblk *pblk, sector_t lba,
1099 struct ppa_addr ppa)
1100 {
1101 if (pblk->addrf_len < 32) {
1102 u32 *map = (u32 *)pblk->trans_map;
1103
1104 map[lba] = pblk_ppa64_to_ppa32(pblk, ppa);
1105 } else {
1106 u64 *map = (u64 *)pblk->trans_map;
1107
1108 map[lba] = ppa.ppa;
1109 }
1110 }
1111
pblk_ppa_empty(struct ppa_addr ppa_addr)1112 static inline int pblk_ppa_empty(struct ppa_addr ppa_addr)
1113 {
1114 return (ppa_addr.ppa == ADDR_EMPTY);
1115 }
1116
pblk_ppa_set_empty(struct ppa_addr * ppa_addr)1117 static inline void pblk_ppa_set_empty(struct ppa_addr *ppa_addr)
1118 {
1119 ppa_addr->ppa = ADDR_EMPTY;
1120 }
1121
pblk_ppa_comp(struct ppa_addr lppa,struct ppa_addr rppa)1122 static inline bool pblk_ppa_comp(struct ppa_addr lppa, struct ppa_addr rppa)
1123 {
1124 return (lppa.ppa == rppa.ppa);
1125 }
1126
pblk_addr_in_cache(struct ppa_addr ppa)1127 static inline int pblk_addr_in_cache(struct ppa_addr ppa)
1128 {
1129 return (ppa.ppa != ADDR_EMPTY && ppa.c.is_cached);
1130 }
1131
pblk_addr_to_cacheline(struct ppa_addr ppa)1132 static inline int pblk_addr_to_cacheline(struct ppa_addr ppa)
1133 {
1134 return ppa.c.line;
1135 }
1136
pblk_cacheline_to_addr(int addr)1137 static inline struct ppa_addr pblk_cacheline_to_addr(int addr)
1138 {
1139 struct ppa_addr p;
1140
1141 p.c.line = addr;
1142 p.c.is_cached = 1;
1143
1144 return p;
1145 }
1146
pblk_calc_meta_header_crc(struct pblk * pblk,struct line_header * header)1147 static inline u32 pblk_calc_meta_header_crc(struct pblk *pblk,
1148 struct line_header *header)
1149 {
1150 u32 crc = ~(u32)0;
1151
1152 crc = crc32_le(crc, (unsigned char *)header + sizeof(crc),
1153 sizeof(struct line_header) - sizeof(crc));
1154
1155 return crc;
1156 }
1157
pblk_calc_smeta_crc(struct pblk * pblk,struct line_smeta * smeta)1158 static inline u32 pblk_calc_smeta_crc(struct pblk *pblk,
1159 struct line_smeta *smeta)
1160 {
1161 struct pblk_line_meta *lm = &pblk->lm;
1162 u32 crc = ~(u32)0;
1163
1164 crc = crc32_le(crc, (unsigned char *)smeta +
1165 sizeof(struct line_header) + sizeof(crc),
1166 lm->smeta_len -
1167 sizeof(struct line_header) - sizeof(crc));
1168
1169 return crc;
1170 }
1171
pblk_calc_emeta_crc(struct pblk * pblk,struct line_emeta * emeta)1172 static inline u32 pblk_calc_emeta_crc(struct pblk *pblk,
1173 struct line_emeta *emeta)
1174 {
1175 struct pblk_line_meta *lm = &pblk->lm;
1176 u32 crc = ~(u32)0;
1177
1178 crc = crc32_le(crc, (unsigned char *)emeta +
1179 sizeof(struct line_header) + sizeof(crc),
1180 lm->emeta_len[0] -
1181 sizeof(struct line_header) - sizeof(crc));
1182
1183 return crc;
1184 }
1185
pblk_io_aligned(struct pblk * pblk,int nr_secs)1186 static inline int pblk_io_aligned(struct pblk *pblk, int nr_secs)
1187 {
1188 return !(nr_secs % pblk->min_write_pgs);
1189 }
1190
1191 #ifdef CONFIG_NVM_PBLK_DEBUG
print_ppa(struct pblk * pblk,struct ppa_addr * p,char * msg,int error)1192 static inline void print_ppa(struct pblk *pblk, struct ppa_addr *p,
1193 char *msg, int error)
1194 {
1195 struct nvm_geo *geo = &pblk->dev->geo;
1196
1197 if (p->c.is_cached) {
1198 pblk_err(pblk, "ppa: (%s: %x) cache line: %llu\n",
1199 msg, error, (u64)p->c.line);
1200 } else if (geo->version == NVM_OCSSD_SPEC_12) {
1201 pblk_err(pblk, "ppa: (%s: %x):ch:%d,lun:%d,blk:%d,pg:%d,pl:%d,sec:%d\n",
1202 msg, error,
1203 p->g.ch, p->g.lun, p->g.blk,
1204 p->g.pg, p->g.pl, p->g.sec);
1205 } else {
1206 pblk_err(pblk, "ppa: (%s: %x):ch:%d,lun:%d,chk:%d,sec:%d\n",
1207 msg, error,
1208 p->m.grp, p->m.pu, p->m.chk, p->m.sec);
1209 }
1210 }
1211
pblk_print_failed_rqd(struct pblk * pblk,struct nvm_rq * rqd,int error)1212 static inline void pblk_print_failed_rqd(struct pblk *pblk, struct nvm_rq *rqd,
1213 int error)
1214 {
1215 int bit = -1;
1216
1217 if (rqd->nr_ppas == 1) {
1218 print_ppa(pblk, &rqd->ppa_addr, "rqd", error);
1219 return;
1220 }
1221
1222 while ((bit = find_next_bit((void *)&rqd->ppa_status, rqd->nr_ppas,
1223 bit + 1)) < rqd->nr_ppas) {
1224 print_ppa(pblk, &rqd->ppa_list[bit], "rqd", error);
1225 }
1226
1227 pblk_err(pblk, "error:%d, ppa_status:%llx\n", error, rqd->ppa_status);
1228 }
1229
pblk_boundary_ppa_checks(struct nvm_tgt_dev * tgt_dev,struct ppa_addr * ppas,int nr_ppas)1230 static inline int pblk_boundary_ppa_checks(struct nvm_tgt_dev *tgt_dev,
1231 struct ppa_addr *ppas, int nr_ppas)
1232 {
1233 struct nvm_geo *geo = &tgt_dev->geo;
1234 struct ppa_addr *ppa;
1235 int i;
1236
1237 for (i = 0; i < nr_ppas; i++) {
1238 ppa = &ppas[i];
1239
1240 if (geo->version == NVM_OCSSD_SPEC_12) {
1241 if (!ppa->c.is_cached &&
1242 ppa->g.ch < geo->num_ch &&
1243 ppa->g.lun < geo->num_lun &&
1244 ppa->g.pl < geo->num_pln &&
1245 ppa->g.blk < geo->num_chk &&
1246 ppa->g.pg < geo->num_pg &&
1247 ppa->g.sec < geo->ws_min)
1248 continue;
1249 } else {
1250 if (!ppa->c.is_cached &&
1251 ppa->m.grp < geo->num_ch &&
1252 ppa->m.pu < geo->num_lun &&
1253 ppa->m.chk < geo->num_chk &&
1254 ppa->m.sec < geo->clba)
1255 continue;
1256 }
1257
1258 print_ppa(tgt_dev->q->queuedata, ppa, "boundary", i);
1259
1260 return 1;
1261 }
1262 return 0;
1263 }
1264
pblk_check_io(struct pblk * pblk,struct nvm_rq * rqd)1265 static inline int pblk_check_io(struct pblk *pblk, struct nvm_rq *rqd)
1266 {
1267 struct nvm_tgt_dev *dev = pblk->dev;
1268 struct ppa_addr *ppa_list = nvm_rq_to_ppa_list(rqd);
1269
1270 if (pblk_boundary_ppa_checks(dev, ppa_list, rqd->nr_ppas)) {
1271 WARN_ON(1);
1272 return -EINVAL;
1273 }
1274
1275 if (rqd->opcode == NVM_OP_PWRITE) {
1276 struct pblk_line *line;
1277 int i;
1278
1279 for (i = 0; i < rqd->nr_ppas; i++) {
1280 line = pblk_ppa_to_line(pblk, ppa_list[i]);
1281
1282 spin_lock(&line->lock);
1283 if (line->state != PBLK_LINESTATE_OPEN) {
1284 pblk_err(pblk, "bad ppa: line:%d,state:%d\n",
1285 line->id, line->state);
1286 WARN_ON(1);
1287 spin_unlock(&line->lock);
1288 return -EINVAL;
1289 }
1290 spin_unlock(&line->lock);
1291 }
1292 }
1293
1294 return 0;
1295 }
1296 #endif
1297
pblk_boundary_paddr_checks(struct pblk * pblk,u64 paddr)1298 static inline int pblk_boundary_paddr_checks(struct pblk *pblk, u64 paddr)
1299 {
1300 struct pblk_line_meta *lm = &pblk->lm;
1301
1302 if (paddr > lm->sec_per_line)
1303 return 1;
1304
1305 return 0;
1306 }
1307
pblk_get_bi_idx(struct bio * bio)1308 static inline unsigned int pblk_get_bi_idx(struct bio *bio)
1309 {
1310 return bio->bi_iter.bi_idx;
1311 }
1312
pblk_get_lba(struct bio * bio)1313 static inline sector_t pblk_get_lba(struct bio *bio)
1314 {
1315 return bio->bi_iter.bi_sector / NR_PHY_IN_LOG;
1316 }
1317
pblk_get_secs(struct bio * bio)1318 static inline unsigned int pblk_get_secs(struct bio *bio)
1319 {
1320 return bio->bi_iter.bi_size / PBLK_EXPOSED_PAGE_SIZE;
1321 }
1322
pblk_disk_name(struct pblk * pblk)1323 static inline char *pblk_disk_name(struct pblk *pblk)
1324 {
1325 struct gendisk *disk = pblk->disk;
1326
1327 return disk->disk_name;
1328 }
1329
pblk_get_min_chks(struct pblk * pblk)1330 static inline unsigned int pblk_get_min_chks(struct pblk *pblk)
1331 {
1332 struct pblk_line_meta *lm = &pblk->lm;
1333 /* In a worst-case scenario every line will have OP invalid sectors.
1334 * We will then need a minimum of 1/OP lines to free up a single line
1335 */
1336
1337 return DIV_ROUND_UP(100, pblk->op) * lm->blk_per_line;
1338 }
1339
pblk_get_meta(struct pblk * pblk,void * meta,int index)1340 static inline struct pblk_sec_meta *pblk_get_meta(struct pblk *pblk,
1341 void *meta, int index)
1342 {
1343 return meta +
1344 max_t(int, sizeof(struct pblk_sec_meta), pblk->oob_meta_size)
1345 * index;
1346 }
1347
pblk_dma_meta_size(struct pblk * pblk)1348 static inline int pblk_dma_meta_size(struct pblk *pblk)
1349 {
1350 return max_t(int, sizeof(struct pblk_sec_meta), pblk->oob_meta_size)
1351 * NVM_MAX_VLBA;
1352 }
1353
pblk_is_oob_meta_supported(struct pblk * pblk)1354 static inline int pblk_is_oob_meta_supported(struct pblk *pblk)
1355 {
1356 return pblk->oob_meta_size >= sizeof(struct pblk_sec_meta);
1357 }
1358 #endif /* PBLK_H_ */
1359