?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
perf tools: Don't clone maps from parent when synthesizing forks
[linux/fpc-iii.git] / drivers / lightnvm / pblk.h
blob02bb2e98f8a9c9dfbfdf09b061b5825fd4bb9f9c
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 #define PBLK_COMMAND_TIMEOUT_MS 30000
47
48 /* Max 512 LUNs per device */
49 #define PBLK_MAX_LUNS_BITMAP (4)
50
51 #define NR_PHY_IN_LOG (PBLK_EXPOSED_PAGE_SIZE / PBLK_SECTOR)
52
53 /* Static pool sizes */
54 #define PBLK_GEN_WS_POOL_SIZE (2)
55
56 #define PBLK_DEFAULT_OP (11)
57
58 enum {
59 PBLK_READ = READ,
60 PBLK_WRITE = WRITE,/* Write from write buffer */
61 PBLK_WRITE_INT, /* Internal write - no write buffer */
62 PBLK_READ_RECOV, /* Recovery read - errors allowed */
63 PBLK_ERASE,
64 };
65
66 enum {
67 /* IO Types */
68 PBLK_IOTYPE_USER = 1 << 0,
69 PBLK_IOTYPE_GC = 1 << 1,
70
71 /* Write buffer flags */
72 PBLK_FLUSH_ENTRY = 1 << 2,
73 PBLK_WRITTEN_DATA = 1 << 3,
74 PBLK_SUBMITTED_ENTRY = 1 << 4,
75 PBLK_WRITABLE_ENTRY = 1 << 5,
76 };
77
78 enum {
79 PBLK_BLK_ST_OPEN = 0x1,
80 PBLK_BLK_ST_CLOSED = 0x2,
81 };
82
83 enum {
84 PBLK_CHUNK_RESET_START,
85 PBLK_CHUNK_RESET_DONE,
86 PBLK_CHUNK_RESET_FAILED,
87 };
88
89 struct pblk_sec_meta {
90 u64 reserved;
91 __le64 lba;
92 };
93
94 /* The number of GC lists and the rate-limiter states go together. This way the
95 * rate-limiter can dictate how much GC is needed based on resource utilization.
96 */
97 #define PBLK_GC_NR_LISTS 4
98
99 enum {
100 PBLK_RL_OFF = 0,
101 PBLK_RL_WERR = 1,
102 PBLK_RL_HIGH = 2,
103 PBLK_RL_MID = 3,
104 PBLK_RL_LOW = 4
105 };
106
107 #define pblk_dma_meta_size (sizeof(struct pblk_sec_meta) * NVM_MAX_VLBA)
108 #define pblk_dma_ppa_size (sizeof(u64) * NVM_MAX_VLBA)
109
110 /* write buffer completion context */
111 struct pblk_c_ctx {
112 struct list_head list; /* Head for out-of-order completion */
113
114 unsigned long *lun_bitmap; /* Luns used on current request */
115 unsigned int sentry;
116 unsigned int nr_valid;
117 unsigned int nr_padded;
118 };
119
120 /* read context */
121 struct pblk_g_ctx {
122 void *private;
123 unsigned long start_time;
124 u64 lba;
125 };
126
127 /* partial read context */
128 struct pblk_pr_ctx {
129 struct bio *orig_bio;
130 DECLARE_BITMAP(bitmap, NVM_MAX_VLBA);
131 unsigned int orig_nr_secs;
132 unsigned int bio_init_idx;
133 void *ppa_ptr;
134 dma_addr_t dma_ppa_list;
135 };
136
137 /* Pad context */
138 struct pblk_pad_rq {
139 struct pblk *pblk;
140 struct completion wait;
141 struct kref ref;
142 };
143
144 /* Recovery context */
145 struct pblk_rec_ctx {
146 struct pblk *pblk;
147 struct nvm_rq *rqd;
148 struct work_struct ws_rec;
149 };
150
151 /* Write context */
152 struct pblk_w_ctx {
153 struct bio_list bios; /* Original bios - used for completion
154 * in REQ_FUA, REQ_FLUSH case
155 */
156 u64 lba; /* Logic addr. associated with entry */
157 struct ppa_addr ppa; /* Physic addr. associated with entry */
158 int flags; /* Write context flags */
159 };
160
161 struct pblk_rb_entry {
162 struct ppa_addr cacheline; /* Cacheline for this entry */
163 void *data; /* Pointer to data on this entry */
164 struct pblk_w_ctx w_ctx; /* Context for this entry */
165 struct list_head index; /* List head to enable indexes */
166 };
167
168 #define EMPTY_ENTRY (~0U)
169
170 struct pblk_rb_pages {
171 struct page *pages;
172 int order;
173 struct list_head list;
174 };
175
176 struct pblk_rb {
177 struct pblk_rb_entry *entries; /* Ring buffer entries */
178 unsigned int mem; /* Write offset - points to next
179 * writable entry in memory
180 */
181 unsigned int subm; /* Read offset - points to last entry
182 * that has been submitted to the media
183 * to be persisted
184 */
185 unsigned int sync; /* Synced - backpointer that signals
186 * the last submitted entry that has
187 * been successfully persisted to media
188 */
189 unsigned int flush_point; /* Sync point - last entry that must be
190 * flushed to the media. Used with
191 * REQ_FLUSH and REQ_FUA
192 */
193 unsigned int l2p_update; /* l2p update point - next entry for
194 * which l2p mapping will be updated to
195 * contain a device ppa address (instead
196 * of a cacheline
197 */
198 unsigned int nr_entries; /* Number of entries in write buffer -
199 * must be a power of two
200 */
201 unsigned int seg_size; /* Size of the data segments being
202 * stored on each entry. Typically this
203 * will be 4KB
204 */
205
206 unsigned int back_thres; /* Threshold that shall be maintained by
207 * the backpointer in order to respect
208 * geo->mw_cunits on a per chunk basis
209 */
210
211 struct list_head pages; /* List of data pages */
212
213 spinlock_t w_lock; /* Write lock */
214 spinlock_t s_lock; /* Sync lock */
215
216 #ifdef CONFIG_NVM_PBLK_DEBUG
217 atomic_t inflight_flush_point; /* Not served REQ_FLUSH | REQ_FUA */
218 #endif
219 };
220
221 #define PBLK_RECOVERY_SECTORS 16
222
223 struct pblk_lun {
224 struct ppa_addr bppa;
225 struct semaphore wr_sem;
226 };
227
228 struct pblk_gc_rq {
229 struct pblk_line *line;
230 void *data;
231 u64 paddr_list[NVM_MAX_VLBA];
232 u64 lba_list[NVM_MAX_VLBA];
233 int nr_secs;
234 int secs_to_gc;
235 struct list_head list;
236 };
237
238 struct pblk_gc {
239 /* These states are not protected by a lock since (i) they are in the
240 * fast path, and (ii) they are not critical.
241 */
242 int gc_active;
243 int gc_enabled;
244 int gc_forced;
245
246 struct task_struct *gc_ts;
247 struct task_struct *gc_writer_ts;
248 struct task_struct *gc_reader_ts;
249
250 struct workqueue_struct *gc_line_reader_wq;
251 struct workqueue_struct *gc_reader_wq;
252
253 struct timer_list gc_timer;
254
255 struct semaphore gc_sem;
256 atomic_t read_inflight_gc; /* Number of lines with inflight GC reads */
257 atomic_t pipeline_gc; /* Number of lines in the GC pipeline -
258 * started reads to finished writes
259 */
260 int w_entries;
261
262 struct list_head w_list;
263 struct list_head r_list;
264
265 spinlock_t lock;
266 spinlock_t w_lock;
267 spinlock_t r_lock;
268 };
269
270 struct pblk_rl {
271 unsigned int high; /* Upper threshold for rate limiter (free run -
272 * user I/O rate limiter
273 */
274 unsigned int high_pw; /* High rounded up as a power of 2 */
275
276 #define PBLK_USER_HIGH_THRS 8 /* Begin write limit at 12% available blks */
277 #define PBLK_USER_LOW_THRS 10 /* Aggressive GC at 10% available blocks */
278
279 int rb_windows_pw; /* Number of rate windows in the write buffer
280 * given as a power-of-2. This guarantees that
281 * when user I/O is being rate limited, there
282 * will be reserved enough space for the GC to
283 * place its payload. A window is of
284 * pblk->max_write_pgs size, which in NVMe is
285 * 64, i.e., 256kb.
286 */
287 int rb_budget; /* Total number of entries available for I/O */
288 int rb_user_max; /* Max buffer entries available for user I/O */
289 int rb_gc_max; /* Max buffer entries available for GC I/O */
290 int rb_gc_rsv; /* Reserved buffer entries for GC I/O */
291 int rb_state; /* Rate-limiter current state */
292 int rb_max_io; /* Maximum size for an I/O giving the config */
293
294 atomic_t rb_user_cnt; /* User I/O buffer counter */
295 atomic_t rb_gc_cnt; /* GC I/O buffer counter */
296 atomic_t rb_space; /* Space limit in case of reaching capacity */
297
298 int rsv_blocks; /* Reserved blocks for GC */
299
300 int rb_user_active;
301 int rb_gc_active;
302
303 atomic_t werr_lines; /* Number of write error lines that needs gc */
304
305 struct timer_list u_timer;
306
307 unsigned long long nr_secs;
308 unsigned long total_blocks;
309
310 atomic_t free_blocks; /* Total number of free blocks (+ OP) */
311 atomic_t free_user_blocks; /* Number of user free blocks (no OP) */
312 };
313
314 #define PBLK_LINE_EMPTY (~0U)
315
316 enum {
317 /* Line Types */
318 PBLK_LINETYPE_FREE = 0,
319 PBLK_LINETYPE_LOG = 1,
320 PBLK_LINETYPE_DATA = 2,
321
322 /* Line state */
323 PBLK_LINESTATE_NEW = 9,
324 PBLK_LINESTATE_FREE = 10,
325 PBLK_LINESTATE_OPEN = 11,
326 PBLK_LINESTATE_CLOSED = 12,
327 PBLK_LINESTATE_GC = 13,
328 PBLK_LINESTATE_BAD = 14,
329 PBLK_LINESTATE_CORRUPT = 15,
330
331 /* GC group */
332 PBLK_LINEGC_NONE = 20,
333 PBLK_LINEGC_EMPTY = 21,
334 PBLK_LINEGC_LOW = 22,
335 PBLK_LINEGC_MID = 23,
336 PBLK_LINEGC_HIGH = 24,
337 PBLK_LINEGC_FULL = 25,
338 PBLK_LINEGC_WERR = 26
339 };
340
341 #define PBLK_MAGIC 0x70626c6b /*pblk*/
342
343 /* emeta/smeta persistent storage format versions:
344 * Changes in major version requires offline migration.
345 * Changes in minor version are handled automatically during
346 * recovery.
347 */
348
349 #define SMETA_VERSION_MAJOR (0)
350 #define SMETA_VERSION_MINOR (1)
351
352 #define EMETA_VERSION_MAJOR (0)
353 #define EMETA_VERSION_MINOR (2)
354
355 struct line_header {
356 __le32 crc;
357 __le32 identifier; /* pblk identifier */
358 __u8 uuid[16]; /* instance uuid */
359 __le16 type; /* line type */
360 __u8 version_major; /* version major */
361 __u8 version_minor; /* version minor */
362 __le32 id; /* line id for current line */
363 };
364
365 struct line_smeta {
366 struct line_header header;
367
368 __le32 crc; /* Full structure including struct crc */
369 /* Previous line metadata */
370 __le32 prev_id; /* Line id for previous line */
371
372 /* Current line metadata */
373 __le64 seq_nr; /* Sequence number for current line */
374
375 /* Active writers */
376 __le32 window_wr_lun; /* Number of parallel LUNs to write */
377
378 __le32 rsvd[2];
379
380 __le64 lun_bitmap[];
381 };
382
383
384 /*
385 * Metadata layout in media:
386 * First sector:
387 * 1. struct line_emeta
388 * 2. bad block bitmap (u64 * window_wr_lun)
389 * 3. write amplification counters
390 * Mid sectors (start at lbas_sector):
391 * 3. nr_lbas (u64) forming lba list
392 * Last sectors (start at vsc_sector):
393 * 4. u32 valid sector count (vsc) for all lines (~0U: free line)
394 */
395 struct line_emeta {
396 struct line_header header;
397
398 __le32 crc; /* Full structure including struct crc */
399
400 /* Previous line metadata */
401 __le32 prev_id; /* Line id for prev line */
402
403 /* Current line metadata */
404 __le64 seq_nr; /* Sequence number for current line */
405
406 /* Active writers */
407 __le32 window_wr_lun; /* Number of parallel LUNs to write */
408
409 /* Bookkeeping for recovery */
410 __le32 next_id; /* Line id for next line */
411 __le64 nr_lbas; /* Number of lbas mapped in line */
412 __le64 nr_valid_lbas; /* Number of valid lbas mapped in line */
413 __le64 bb_bitmap[]; /* Updated bad block bitmap for line */
414 };
415
416
417 /* Write amplification counters stored on media */
418 struct wa_counters {
419 __le64 user; /* Number of user written sectors */
420 __le64 gc; /* Number of sectors written by GC*/
421 __le64 pad; /* Number of padded sectors */
422 };
423
424 struct pblk_emeta {
425 struct line_emeta *buf; /* emeta buffer in media format */
426 int mem; /* Write offset - points to next
427 * writable entry in memory
428 */
429 atomic_t sync; /* Synced - backpointer that signals the
430 * last entry that has been successfully
431 * persisted to media
432 */
433 unsigned int nr_entries; /* Number of emeta entries */
434 };
435
436 struct pblk_smeta {
437 struct line_smeta *buf; /* smeta buffer in persistent format */
438 };
439
440 struct pblk_w_err_gc {
441 int has_write_err;
442 __le64 *lba_list;
443 };
444
445 struct pblk_line {
446 struct pblk *pblk;
447 unsigned int id; /* Line number corresponds to the
448 * block line
449 */
450 unsigned int seq_nr; /* Unique line sequence number */
451
452 int state; /* PBLK_LINESTATE_X */
453 int type; /* PBLK_LINETYPE_X */
454 int gc_group; /* PBLK_LINEGC_X */
455 struct list_head list; /* Free, GC lists */
456
457 unsigned long *lun_bitmap; /* Bitmap for LUNs mapped in line */
458
459 struct nvm_chk_meta *chks; /* Chunks forming line */
460
461 struct pblk_smeta *smeta; /* Start metadata */
462 struct pblk_emeta *emeta; /* End medatada */
463
464 int meta_line; /* Metadata line id */
465 int meta_distance; /* Distance between data and metadata */
466
467 u64 smeta_ssec; /* Sector where smeta starts */
468 u64 emeta_ssec; /* Sector where emeta starts */
469
470 unsigned int sec_in_line; /* Number of usable secs in line */
471
472 atomic_t blk_in_line; /* Number of good blocks in line */
473 unsigned long *blk_bitmap; /* Bitmap for valid/invalid blocks */
474 unsigned long *erase_bitmap; /* Bitmap for erased blocks */
475
476 unsigned long *map_bitmap; /* Bitmap for mapped sectors in line */
477 unsigned long *invalid_bitmap; /* Bitmap for invalid sectors in line */
478
479 atomic_t left_eblks; /* Blocks left for erasing */
480 atomic_t left_seblks; /* Blocks left for sync erasing */
481
482 int left_msecs; /* Sectors left for mapping */
483 unsigned int cur_sec; /* Sector map pointer */
484 unsigned int nr_valid_lbas; /* Number of valid lbas in line */
485
486 __le32 *vsc; /* Valid sector count in line */
487
488 struct kref ref; /* Write buffer L2P references */
489
490 struct pblk_w_err_gc *w_err_gc; /* Write error gc recovery metadata */
491
492 spinlock_t lock; /* Necessary for invalid_bitmap only */
493 };
494
495 #define PBLK_DATA_LINES 4
496
497 enum {
498 PBLK_KMALLOC_META = 1,
499 PBLK_VMALLOC_META = 2,
500 };
501
502 enum {
503 PBLK_EMETA_TYPE_HEADER = 1, /* struct line_emeta first sector */
504 PBLK_EMETA_TYPE_LLBA = 2, /* lba list - type: __le64 */
505 PBLK_EMETA_TYPE_VSC = 3, /* vsc list - type: __le32 */
506 };
507
508 struct pblk_line_mgmt {
509 int nr_lines; /* Total number of full lines */
510 int nr_free_lines; /* Number of full lines in free list */
511
512 /* Free lists - use free_lock */
513 struct list_head free_list; /* Full lines ready to use */
514 struct list_head corrupt_list; /* Full lines corrupted */
515 struct list_head bad_list; /* Full lines bad */
516
517 /* GC lists - use gc_lock */
518 struct list_head *gc_lists[PBLK_GC_NR_LISTS];
519 struct list_head gc_high_list; /* Full lines ready to GC, high isc */
520 struct list_head gc_mid_list; /* Full lines ready to GC, mid isc */
521 struct list_head gc_low_list; /* Full lines ready to GC, low isc */
522
523 struct list_head gc_werr_list; /* Write err recovery list */
524
525 struct list_head gc_full_list; /* Full lines ready to GC, no valid */
526 struct list_head gc_empty_list; /* Full lines close, all valid */
527
528 struct pblk_line *log_line; /* Current FTL log line */
529 struct pblk_line *data_line; /* Current data line */
530 struct pblk_line *log_next; /* Next FTL log line */
531 struct pblk_line *data_next; /* Next data line */
532
533 struct list_head emeta_list; /* Lines queued to schedule emeta */
534
535 __le32 *vsc_list; /* Valid sector counts for all lines */
536
537 /* Metadata allocation type: VMALLOC | KMALLOC */
538 int emeta_alloc_type;
539
540 /* Pre-allocated metadata for data lines */
541 struct pblk_smeta *sline_meta[PBLK_DATA_LINES];
542 struct pblk_emeta *eline_meta[PBLK_DATA_LINES];
543 unsigned long meta_bitmap;
544
545 /* Cache and mempool for map/invalid bitmaps */
546 struct kmem_cache *bitmap_cache;
547 mempool_t *bitmap_pool;
548
549 /* Helpers for fast bitmap calculations */
550 unsigned long *bb_template;
551 unsigned long *bb_aux;
552
553 unsigned long d_seq_nr; /* Data line unique sequence number */
554 unsigned long l_seq_nr; /* Log line unique sequence number */
555
556 spinlock_t free_lock;
557 spinlock_t close_lock;
558 spinlock_t gc_lock;
559 };
560
561 struct pblk_line_meta {
562 unsigned int smeta_len; /* Total length for smeta */
563 unsigned int smeta_sec; /* Sectors needed for smeta */
564
565 unsigned int emeta_len[4]; /* Lengths for emeta:
566 * [0]: Total
567 * [1]: struct line_emeta +
568 * bb_bitmap + struct wa_counters
569 * [2]: L2P portion
570 * [3]: vsc
571 */
572 unsigned int emeta_sec[4]; /* Sectors needed for emeta. Same layout
573 * as emeta_len
574 */
575
576 unsigned int emeta_bb; /* Boundary for bb that affects emeta */
577
578 unsigned int vsc_list_len; /* Length for vsc list */
579 unsigned int sec_bitmap_len; /* Length for sector bitmap in line */
580 unsigned int blk_bitmap_len; /* Length for block bitmap in line */
581 unsigned int lun_bitmap_len; /* Length for lun bitmap in line */
582
583 unsigned int blk_per_line; /* Number of blocks in a full line */
584 unsigned int sec_per_line; /* Number of sectors in a line */
585 unsigned int dsec_per_line; /* Number of data sectors in a line */
586 unsigned int min_blk_line; /* Min. number of good blocks in line */
587
588 unsigned int mid_thrs; /* Threshold for GC mid list */
589 unsigned int high_thrs; /* Threshold for GC high list */
590
591 unsigned int meta_distance; /* Distance between data and metadata */
592 };
593
594 enum {
595 PBLK_STATE_RUNNING = 0,
596 PBLK_STATE_STOPPING = 1,
597 PBLK_STATE_RECOVERING = 2,
598 PBLK_STATE_STOPPED = 3,
599 };
600
601 /* Internal format to support not power-of-2 device formats */
602 struct pblk_addrf {
603 /* gen to dev */
604 int sec_stripe;
605 int ch_stripe;
606 int lun_stripe;
607
608 /* dev to gen */
609 int sec_lun_stripe;
610 int sec_ws_stripe;
611 };
612
613 struct pblk {
614 struct nvm_tgt_dev *dev;
615 struct gendisk *disk;
616
617 struct kobject kobj;
618
619 struct pblk_lun *luns;
620
621 struct pblk_line *lines; /* Line array */
622 struct pblk_line_mgmt l_mg; /* Line management */
623 struct pblk_line_meta lm; /* Line metadata */
624
625 struct nvm_addrf addrf; /* Aligned address format */
626 struct pblk_addrf uaddrf; /* Unaligned address format */
627 int addrf_len;
628
629 struct pblk_rb rwb;
630
631 int state; /* pblk line state */
632
633 int min_write_pgs; /* Minimum amount of pages required by controller */
634 int max_write_pgs; /* Maximum amount of pages supported by controller */
635
636 sector_t capacity; /* Device capacity when bad blocks are subtracted */
637
638 int op; /* Percentage of device used for over-provisioning */
639 int op_blks; /* Number of blocks used for over-provisioning */
640
641 /* pblk provisioning values. Used by rate limiter */
642 struct pblk_rl rl;
643
644 int sec_per_write;
645
646 unsigned char instance_uuid[16];
647
648 /* Persistent write amplification counters, 4kb sector I/Os */
649 atomic64_t user_wa; /* Sectors written by user */
650 atomic64_t gc_wa; /* Sectors written by GC */
651 atomic64_t pad_wa; /* Padded sectors written */
652
653 /* Reset values for delta write amplification measurements */
654 u64 user_rst_wa;
655 u64 gc_rst_wa;
656 u64 pad_rst_wa;
657
658 /* Counters used for calculating padding distribution */
659 atomic64_t *pad_dist; /* Padding distribution buckets */
660 u64 nr_flush_rst; /* Flushes reset value for pad dist.*/
661 atomic64_t nr_flush; /* Number of flush/fua I/O */
662
663 #ifdef CONFIG_NVM_PBLK_DEBUG
664 /* Non-persistent debug counters, 4kb sector I/Os */
665 atomic_long_t inflight_writes; /* Inflight writes (user and gc) */
666 atomic_long_t padded_writes; /* Sectors padded due to flush/fua */
667 atomic_long_t padded_wb; /* Sectors padded in write buffer */
668 atomic_long_t req_writes; /* Sectors stored on write buffer */
669 atomic_long_t sub_writes; /* Sectors submitted from buffer */
670 atomic_long_t sync_writes; /* Sectors synced to media */
671 atomic_long_t inflight_reads; /* Inflight sector read requests */
672 atomic_long_t cache_reads; /* Read requests that hit the cache */
673 atomic_long_t sync_reads; /* Completed sector read requests */
674 atomic_long_t recov_writes; /* Sectors submitted from recovery */
675 atomic_long_t recov_gc_writes; /* Sectors submitted from write GC */
676 atomic_long_t recov_gc_reads; /* Sectors submitted from read GC */
677 #endif
678
679 spinlock_t lock;
680
681 atomic_long_t read_failed;
682 atomic_long_t read_empty;
683 atomic_long_t read_high_ecc;
684 atomic_long_t read_failed_gc;
685 atomic_long_t write_failed;
686 atomic_long_t erase_failed;
687
688 atomic_t inflight_io; /* General inflight I/O counter */
689
690 struct task_struct *writer_ts;
691
692 /* Simple translation map of logical addresses to physical addresses.
693 * The logical addresses is known by the host system, while the physical
694 * addresses are used when writing to the disk block device.
695 */
696 unsigned char *trans_map;
697 spinlock_t trans_lock;
698
699 struct list_head compl_list;
700
701 spinlock_t resubmit_lock; /* Resubmit list lock */
702 struct list_head resubmit_list; /* Resubmit list for failed writes*/
703
704 mempool_t page_bio_pool;
705 mempool_t gen_ws_pool;
706 mempool_t rec_pool;
707 mempool_t r_rq_pool;
708 mempool_t w_rq_pool;
709 mempool_t e_rq_pool;
710
711 struct workqueue_struct *close_wq;
712 struct workqueue_struct *bb_wq;
713 struct workqueue_struct *r_end_wq;
714
715 struct timer_list wtimer;
716
717 struct pblk_gc gc;
718 };
719
720 struct pblk_line_ws {
721 struct pblk *pblk;
722 struct pblk_line *line;
723 void *priv;
724 struct work_struct ws;
725 };
726
727 #define pblk_g_rq_size (sizeof(struct nvm_rq) + sizeof(struct pblk_g_ctx))
728 #define pblk_w_rq_size (sizeof(struct nvm_rq) + sizeof(struct pblk_c_ctx))
729
730 #define pblk_err(pblk, fmt, ...) \
731 pr_err("pblk %s: " fmt, pblk->disk->disk_name, ##__VA_ARGS__)
732 #define pblk_info(pblk, fmt, ...) \
733 pr_info("pblk %s: " fmt, pblk->disk->disk_name, ##__VA_ARGS__)
734 #define pblk_warn(pblk, fmt, ...) \
735 pr_warn("pblk %s: " fmt, pblk->disk->disk_name, ##__VA_ARGS__)
736 #define pblk_debug(pblk, fmt, ...) \
737 pr_debug("pblk %s: " fmt, pblk->disk->disk_name, ##__VA_ARGS__)
738
739 /*
740 * pblk ring buffer operations
741 */
742 int pblk_rb_init(struct pblk_rb *rb, unsigned int size, unsigned int threshold,
743 unsigned int seg_sz);
744 int pblk_rb_may_write_user(struct pblk_rb *rb, struct bio *bio,
745 unsigned int nr_entries, unsigned int *pos);
746 int pblk_rb_may_write_gc(struct pblk_rb *rb, unsigned int nr_entries,
747 unsigned int *pos);
748 void pblk_rb_write_entry_user(struct pblk_rb *rb, void *data,
749 struct pblk_w_ctx w_ctx, unsigned int pos);
750 void pblk_rb_write_entry_gc(struct pblk_rb *rb, void *data,
751 struct pblk_w_ctx w_ctx, struct pblk_line *line,
752 u64 paddr, unsigned int pos);
753 struct pblk_w_ctx *pblk_rb_w_ctx(struct pblk_rb *rb, unsigned int pos);
754 void pblk_rb_flush(struct pblk_rb *rb);
755
756 void pblk_rb_sync_l2p(struct pblk_rb *rb);
757 unsigned int pblk_rb_read_to_bio(struct pblk_rb *rb, struct nvm_rq *rqd,
758 unsigned int pos, unsigned int nr_entries,
759 unsigned int count);
760 int pblk_rb_copy_to_bio(struct pblk_rb *rb, struct bio *bio, sector_t lba,
761 struct ppa_addr ppa, int bio_iter, bool advanced_bio);
762 unsigned int pblk_rb_read_commit(struct pblk_rb *rb, unsigned int entries);
763
764 unsigned int pblk_rb_sync_init(struct pblk_rb *rb, unsigned long *flags);
765 unsigned int pblk_rb_sync_advance(struct pblk_rb *rb, unsigned int nr_entries);
766 unsigned int pblk_rb_ptr_wrap(struct pblk_rb *rb, unsigned int p,
767 unsigned int nr_entries);
768 void pblk_rb_sync_end(struct pblk_rb *rb, unsigned long *flags);
769 unsigned int pblk_rb_flush_point_count(struct pblk_rb *rb);
770
771 unsigned int pblk_rb_read_count(struct pblk_rb *rb);
772 unsigned int pblk_rb_sync_count(struct pblk_rb *rb);
773 unsigned int pblk_rb_wrap_pos(struct pblk_rb *rb, unsigned int pos);
774
775 int pblk_rb_tear_down_check(struct pblk_rb *rb);
776 int pblk_rb_pos_oob(struct pblk_rb *rb, u64 pos);
777 void pblk_rb_free(struct pblk_rb *rb);
778 ssize_t pblk_rb_sysfs(struct pblk_rb *rb, char *buf);
779
780 /*
781 * pblk core
782 */
783 struct nvm_rq *pblk_alloc_rqd(struct pblk *pblk, int type);
784 void pblk_free_rqd(struct pblk *pblk, struct nvm_rq *rqd, int type);
785 int pblk_alloc_rqd_meta(struct pblk *pblk, struct nvm_rq *rqd);
786 void pblk_free_rqd_meta(struct pblk *pblk, struct nvm_rq *rqd);
787 void pblk_set_sec_per_write(struct pblk *pblk, int sec_per_write);
788 int pblk_setup_w_rec_rq(struct pblk *pblk, struct nvm_rq *rqd,
789 struct pblk_c_ctx *c_ctx);
790 void pblk_discard(struct pblk *pblk, struct bio *bio);
791 struct nvm_chk_meta *pblk_get_chunk_meta(struct pblk *pblk);
792 struct nvm_chk_meta *pblk_chunk_get_off(struct pblk *pblk,
793 struct nvm_chk_meta *lp,
794 struct ppa_addr ppa);
795 void pblk_log_write_err(struct pblk *pblk, struct nvm_rq *rqd);
796 void pblk_log_read_err(struct pblk *pblk, struct nvm_rq *rqd);
797 int pblk_submit_io(struct pblk *pblk, struct nvm_rq *rqd);
798 int pblk_submit_io_sync(struct pblk *pblk, struct nvm_rq *rqd);
799 int pblk_submit_io_sync_sem(struct pblk *pblk, struct nvm_rq *rqd);
800 int pblk_submit_meta_io(struct pblk *pblk, struct pblk_line *meta_line);
801 void pblk_check_chunk_state_update(struct pblk *pblk, struct nvm_rq *rqd);
802 struct bio *pblk_bio_map_addr(struct pblk *pblk, void *data,
803 unsigned int nr_secs, unsigned int len,
804 int alloc_type, gfp_t gfp_mask);
805 struct pblk_line *pblk_line_get(struct pblk *pblk);
806 struct pblk_line *pblk_line_get_first_data(struct pblk *pblk);
807 struct pblk_line *pblk_line_replace_data(struct pblk *pblk);
808 void pblk_ppa_to_line_put(struct pblk *pblk, struct ppa_addr ppa);
809 void pblk_rq_to_line_put(struct pblk *pblk, struct nvm_rq *rqd);
810 int pblk_line_recov_alloc(struct pblk *pblk, struct pblk_line *line);
811 void pblk_line_recov_close(struct pblk *pblk, struct pblk_line *line);
812 struct pblk_line *pblk_line_get_data(struct pblk *pblk);
813 struct pblk_line *pblk_line_get_erase(struct pblk *pblk);
814 int pblk_line_erase(struct pblk *pblk, struct pblk_line *line);
815 int pblk_line_is_full(struct pblk_line *line);
816 void pblk_line_free(struct pblk_line *line);
817 void pblk_line_close_meta(struct pblk *pblk, struct pblk_line *line);
818 void pblk_line_close(struct pblk *pblk, struct pblk_line *line);
819 void pblk_line_close_ws(struct work_struct *work);
820 void pblk_pipeline_stop(struct pblk *pblk);
821 void __pblk_pipeline_stop(struct pblk *pblk);
822 void __pblk_pipeline_flush(struct pblk *pblk);
823 void pblk_gen_run_ws(struct pblk *pblk, struct pblk_line *line, void *priv,
824 void (*work)(struct work_struct *), gfp_t gfp_mask,
825 struct workqueue_struct *wq);
826 u64 pblk_line_smeta_start(struct pblk *pblk, struct pblk_line *line);
827 int pblk_line_smeta_read(struct pblk *pblk, struct pblk_line *line);
828 int pblk_line_emeta_read(struct pblk *pblk, struct pblk_line *line,
829 void *emeta_buf);
830 int pblk_blk_erase_async(struct pblk *pblk, struct ppa_addr erase_ppa);
831 void pblk_line_put(struct kref *ref);
832 void pblk_line_put_wq(struct kref *ref);
833 struct list_head *pblk_line_gc_list(struct pblk *pblk, struct pblk_line *line);
834 u64 pblk_lookup_page(struct pblk *pblk, struct pblk_line *line);
835 void pblk_dealloc_page(struct pblk *pblk, struct pblk_line *line, int nr_secs);
836 u64 pblk_alloc_page(struct pblk *pblk, struct pblk_line *line, int nr_secs);
837 u64 __pblk_alloc_page(struct pblk *pblk, struct pblk_line *line, int nr_secs);
838 int pblk_calc_secs(struct pblk *pblk, unsigned long secs_avail,
839 unsigned long secs_to_flush);
840 void pblk_down_rq(struct pblk *pblk, struct ppa_addr ppa,
841 unsigned long *lun_bitmap);
842 void pblk_down_chunk(struct pblk *pblk, struct ppa_addr ppa);
843 void pblk_up_chunk(struct pblk *pblk, struct ppa_addr ppa);
844 void pblk_up_rq(struct pblk *pblk, unsigned long *lun_bitmap);
845 int pblk_bio_add_pages(struct pblk *pblk, struct bio *bio, gfp_t flags,
846 int nr_pages);
847 void pblk_bio_free_pages(struct pblk *pblk, struct bio *bio, int off,
848 int nr_pages);
849 void pblk_map_invalidate(struct pblk *pblk, struct ppa_addr ppa);
850 void __pblk_map_invalidate(struct pblk *pblk, struct pblk_line *line,
851 u64 paddr);
852 void pblk_update_map(struct pblk *pblk, sector_t lba, struct ppa_addr ppa);
853 void pblk_update_map_cache(struct pblk *pblk, sector_t lba,
854 struct ppa_addr ppa);
855 void pblk_update_map_dev(struct pblk *pblk, sector_t lba,
856 struct ppa_addr ppa, struct ppa_addr entry_line);
857 int pblk_update_map_gc(struct pblk *pblk, sector_t lba, struct ppa_addr ppa,
858 struct pblk_line *gc_line, u64 paddr);
859 void pblk_lookup_l2p_rand(struct pblk *pblk, struct ppa_addr *ppas,
860 u64 *lba_list, int nr_secs);
861 void pblk_lookup_l2p_seq(struct pblk *pblk, struct ppa_addr *ppas,
862 sector_t blba, int nr_secs);
863
864 /*
865 * pblk user I/O write path
866 */
867 int pblk_write_to_cache(struct pblk *pblk, struct bio *bio,
868 unsigned long flags);
869 int pblk_write_gc_to_cache(struct pblk *pblk, struct pblk_gc_rq *gc_rq);
870
871 /*
872 * pblk map
873 */
874 void pblk_map_erase_rq(struct pblk *pblk, struct nvm_rq *rqd,
875 unsigned int sentry, unsigned long *lun_bitmap,
876 unsigned int valid_secs, struct ppa_addr *erase_ppa);
877 void pblk_map_rq(struct pblk *pblk, struct nvm_rq *rqd, unsigned int sentry,
878 unsigned long *lun_bitmap, unsigned int valid_secs,
879 unsigned int off);
880
881 /*
882 * pblk write thread
883 */
884 int pblk_write_ts(void *data);
885 void pblk_write_timer_fn(struct timer_list *t);
886 void pblk_write_should_kick(struct pblk *pblk);
887 void pblk_write_kick(struct pblk *pblk);
888
889 /*
890 * pblk read path
891 */
892 extern struct bio_set pblk_bio_set;
893 int pblk_submit_read(struct pblk *pblk, struct bio *bio);
894 int pblk_submit_read_gc(struct pblk *pblk, struct pblk_gc_rq *gc_rq);
895 /*
896 * pblk recovery
897 */
898 struct pblk_line *pblk_recov_l2p(struct pblk *pblk);
899 int pblk_recov_pad(struct pblk *pblk);
900 int pblk_recov_check_emeta(struct pblk *pblk, struct line_emeta *emeta);
901
902 /*
903 * pblk gc
904 */
905 #define PBLK_GC_MAX_READERS 8 /* Max number of outstanding GC reader jobs */
906 #define PBLK_GC_RQ_QD 128 /* Queue depth for inflight GC requests */
907 #define PBLK_GC_L_QD 4 /* Queue depth for inflight GC lines */
908 #define PBLK_GC_RSV_LINE 1 /* Reserved lines for GC */
909
910 int pblk_gc_init(struct pblk *pblk);
911 void pblk_gc_exit(struct pblk *pblk, bool graceful);
912 void pblk_gc_should_start(struct pblk *pblk);
913 void pblk_gc_should_stop(struct pblk *pblk);
914 void pblk_gc_should_kick(struct pblk *pblk);
915 void pblk_gc_free_full_lines(struct pblk *pblk);
916 void pblk_gc_sysfs_state_show(struct pblk *pblk, int *gc_enabled,
917 int *gc_active);
918 int pblk_gc_sysfs_force(struct pblk *pblk, int force);
919
920 /*
921 * pblk rate limiter
922 */
923 void pblk_rl_init(struct pblk_rl *rl, int budget);
924 void pblk_rl_free(struct pblk_rl *rl);
925 void pblk_rl_update_rates(struct pblk_rl *rl);
926 int pblk_rl_high_thrs(struct pblk_rl *rl);
927 unsigned long pblk_rl_nr_free_blks(struct pblk_rl *rl);
928 unsigned long pblk_rl_nr_user_free_blks(struct pblk_rl *rl);
929 int pblk_rl_user_may_insert(struct pblk_rl *rl, int nr_entries);
930 void pblk_rl_inserted(struct pblk_rl *rl, int nr_entries);
931 void pblk_rl_user_in(struct pblk_rl *rl, int nr_entries);
932 int pblk_rl_gc_may_insert(struct pblk_rl *rl, int nr_entries);
933 void pblk_rl_gc_in(struct pblk_rl *rl, int nr_entries);
934 void pblk_rl_out(struct pblk_rl *rl, int nr_user, int nr_gc);
935 int pblk_rl_max_io(struct pblk_rl *rl);
936 void pblk_rl_free_lines_inc(struct pblk_rl *rl, struct pblk_line *line);
937 void pblk_rl_free_lines_dec(struct pblk_rl *rl, struct pblk_line *line,
938 bool used);
939 int pblk_rl_is_limit(struct pblk_rl *rl);
940
941 void pblk_rl_werr_line_in(struct pblk_rl *rl);
942 void pblk_rl_werr_line_out(struct pblk_rl *rl);
943
944 /*
945 * pblk sysfs
946 */
947 int pblk_sysfs_init(struct gendisk *tdisk);
948 void pblk_sysfs_exit(struct gendisk *tdisk);
949
950 static inline void *pblk_malloc(size_t size, int type, gfp_t flags)
951 {
952 if (type == PBLK_KMALLOC_META)
953 return kmalloc(size, flags);
954 return vmalloc(size);
955 }
956
957 static inline void pblk_mfree(void *ptr, int type)
958 {
959 if (type == PBLK_KMALLOC_META)
960 kfree(ptr);
961 else
962 vfree(ptr);
963 }
964
965 static inline struct nvm_rq *nvm_rq_from_c_ctx(void *c_ctx)
966 {
967 return c_ctx - sizeof(struct nvm_rq);
968 }
969
970 static inline void *emeta_to_bb(struct line_emeta *emeta)
971 {
972 return emeta->bb_bitmap;
973 }
974
975 static inline void *emeta_to_wa(struct pblk_line_meta *lm,
976 struct line_emeta *emeta)
977 {
978 return emeta->bb_bitmap + lm->blk_bitmap_len;
979 }
980
981 static inline void *emeta_to_lbas(struct pblk *pblk, struct line_emeta *emeta)
982 {
983 return ((void *)emeta + pblk->lm.emeta_len[1]);
984 }
985
986 static inline void *emeta_to_vsc(struct pblk *pblk, struct line_emeta *emeta)
987 {
988 return (emeta_to_lbas(pblk, emeta) + pblk->lm.emeta_len[2]);
989 }
990
991 static inline int pblk_line_vsc(struct pblk_line *line)
992 {
993 return le32_to_cpu(*line->vsc);
994 }
995
996 static inline int pblk_ppa_to_line_id(struct ppa_addr p)
997 {
998 return p.a.blk;
999 }
1000
1001 static inline struct pblk_line *pblk_ppa_to_line(struct pblk *pblk,
1002 struct ppa_addr p)
1003 {
1004 return &pblk->lines[pblk_ppa_to_line_id(p)];
1005 }
1006
1007 static inline int pblk_ppa_to_pos(struct nvm_geo *geo, struct ppa_addr p)
1008 {
1009 return p.a.lun * geo->num_ch + p.a.ch;
1010 }
1011
1012 static inline struct ppa_addr addr_to_gen_ppa(struct pblk *pblk, u64 paddr,
1013 u64 line_id)
1014 {
1015 struct nvm_tgt_dev *dev = pblk->dev;
1016 struct nvm_geo *geo = &dev->geo;
1017 struct ppa_addr ppa;
1018
1019 if (geo->version == NVM_OCSSD_SPEC_12) {
1020 struct nvm_addrf_12 *ppaf = (struct nvm_addrf_12 *)&pblk->addrf;
1021
1022 ppa.ppa = 0;
1023 ppa.g.blk = line_id;
1024 ppa.g.pg = (paddr & ppaf->pg_mask) >> ppaf->pg_offset;
1025 ppa.g.lun = (paddr & ppaf->lun_mask) >> ppaf->lun_offset;
1026 ppa.g.ch = (paddr & ppaf->ch_mask) >> ppaf->ch_offset;
1027 ppa.g.pl = (paddr & ppaf->pln_mask) >> ppaf->pln_offset;
1028 ppa.g.sec = (paddr & ppaf->sec_mask) >> ppaf->sec_offset;
1029 } else {
1030 struct pblk_addrf *uaddrf = &pblk->uaddrf;
1031 int secs, chnls, luns;
1032
1033 ppa.ppa = 0;
1034
1035 ppa.m.chk = line_id;
1036
1037 paddr = div_u64_rem(paddr, uaddrf->sec_stripe, &secs);
1038 ppa.m.sec = secs;
1039
1040 paddr = div_u64_rem(paddr, uaddrf->ch_stripe, &chnls);
1041 ppa.m.grp = chnls;
1042
1043 paddr = div_u64_rem(paddr, uaddrf->lun_stripe, &luns);
1044 ppa.m.pu = luns;
1045
1046 ppa.m.sec += uaddrf->sec_stripe * paddr;
1047 }
1048
1049 return ppa;
1050 }
1051
1052 static inline struct nvm_chk_meta *pblk_dev_ppa_to_chunk(struct pblk *pblk,
1053 struct ppa_addr p)
1054 {
1055 struct nvm_tgt_dev *dev = pblk->dev;
1056 struct nvm_geo *geo = &dev->geo;
1057 struct pblk_line *line = pblk_ppa_to_line(pblk, p);
1058 int pos = pblk_ppa_to_pos(geo, p);
1059
1060 return &line->chks[pos];
1061 }
1062
1063 static inline u64 pblk_dev_ppa_to_chunk_addr(struct pblk *pblk,
1064 struct ppa_addr p)
1065 {
1066 struct nvm_tgt_dev *dev = pblk->dev;
1067
1068 return dev_to_chunk_addr(dev->parent, &pblk->addrf, p);
1069 }
1070
1071 static inline u64 pblk_dev_ppa_to_line_addr(struct pblk *pblk,
1072 struct ppa_addr p)
1073 {
1074 struct nvm_tgt_dev *dev = pblk->dev;
1075 struct nvm_geo *geo = &dev->geo;
1076 u64 paddr;
1077
1078 if (geo->version == NVM_OCSSD_SPEC_12) {
1079 struct nvm_addrf_12 *ppaf = (struct nvm_addrf_12 *)&pblk->addrf;
1080
1081 paddr = (u64)p.g.ch << ppaf->ch_offset;
1082 paddr |= (u64)p.g.lun << ppaf->lun_offset;
1083 paddr |= (u64)p.g.pg << ppaf->pg_offset;
1084 paddr |= (u64)p.g.pl << ppaf->pln_offset;
1085 paddr |= (u64)p.g.sec << ppaf->sec_offset;
1086 } else {
1087 struct pblk_addrf *uaddrf = &pblk->uaddrf;
1088 u64 secs = p.m.sec;
1089 int sec_stripe;
1090
1091 paddr = (u64)p.m.grp * uaddrf->sec_stripe;
1092 paddr += (u64)p.m.pu * uaddrf->sec_lun_stripe;
1093
1094 secs = div_u64_rem(secs, uaddrf->sec_stripe, &sec_stripe);
1095 paddr += secs * uaddrf->sec_ws_stripe;
1096 paddr += sec_stripe;
1097 }
1098
1099 return paddr;
1100 }
1101
1102 static inline struct ppa_addr pblk_ppa32_to_ppa64(struct pblk *pblk, u32 ppa32)
1103 {
1104 struct nvm_tgt_dev *dev = pblk->dev;
1105
1106 return nvm_ppa32_to_ppa64(dev->parent, &pblk->addrf, ppa32);
1107 }
1108
1109 static inline u32 pblk_ppa64_to_ppa32(struct pblk *pblk, struct ppa_addr ppa64)
1110 {
1111 struct nvm_tgt_dev *dev = pblk->dev;
1112
1113 return nvm_ppa64_to_ppa32(dev->parent, &pblk->addrf, ppa64);
1114 }
1115
1116 static inline struct ppa_addr pblk_trans_map_get(struct pblk *pblk,
1117 sector_t lba)
1118 {
1119 struct ppa_addr ppa;
1120
1121 if (pblk->addrf_len < 32) {
1122 u32 *map = (u32 *)pblk->trans_map;
1123
1124 ppa = pblk_ppa32_to_ppa64(pblk, map[lba]);
1125 } else {
1126 struct ppa_addr *map = (struct ppa_addr *)pblk->trans_map;
1127
1128 ppa = map[lba];
1129 }
1130
1131 return ppa;
1132 }
1133
1134 static inline void pblk_trans_map_set(struct pblk *pblk, sector_t lba,
1135 struct ppa_addr ppa)
1136 {
1137 if (pblk->addrf_len < 32) {
1138 u32 *map = (u32 *)pblk->trans_map;
1139
1140 map[lba] = pblk_ppa64_to_ppa32(pblk, ppa);
1141 } else {
1142 u64 *map = (u64 *)pblk->trans_map;
1143
1144 map[lba] = ppa.ppa;
1145 }
1146 }
1147
1148 static inline int pblk_ppa_empty(struct ppa_addr ppa_addr)
1149 {
1150 return (ppa_addr.ppa == ADDR_EMPTY);
1151 }
1152
1153 static inline void pblk_ppa_set_empty(struct ppa_addr *ppa_addr)
1154 {
1155 ppa_addr->ppa = ADDR_EMPTY;
1156 }
1157
1158 static inline bool pblk_ppa_comp(struct ppa_addr lppa, struct ppa_addr rppa)
1159 {
1160 return (lppa.ppa == rppa.ppa);
1161 }
1162
1163 static inline int pblk_addr_in_cache(struct ppa_addr ppa)
1164 {
1165 return (ppa.ppa != ADDR_EMPTY && ppa.c.is_cached);
1166 }
1167
1168 static inline int pblk_addr_to_cacheline(struct ppa_addr ppa)
1169 {
1170 return ppa.c.line;
1171 }
1172
1173 static inline struct ppa_addr pblk_cacheline_to_addr(int addr)
1174 {
1175 struct ppa_addr p;
1176
1177 p.c.line = addr;
1178 p.c.is_cached = 1;
1179
1180 return p;
1181 }
1182
1183 static inline u32 pblk_calc_meta_header_crc(struct pblk *pblk,
1184 struct line_header *header)
1185 {
1186 u32 crc = ~(u32)0;
1187
1188 crc = crc32_le(crc, (unsigned char *)header + sizeof(crc),
1189 sizeof(struct line_header) - sizeof(crc));
1190
1191 return crc;
1192 }
1193
1194 static inline u32 pblk_calc_smeta_crc(struct pblk *pblk,
1195 struct line_smeta *smeta)
1196 {
1197 struct pblk_line_meta *lm = &pblk->lm;
1198 u32 crc = ~(u32)0;
1199
1200 crc = crc32_le(crc, (unsigned char *)smeta +
1201 sizeof(struct line_header) + sizeof(crc),
1202 lm->smeta_len -
1203 sizeof(struct line_header) - sizeof(crc));
1204
1205 return crc;
1206 }
1207
1208 static inline u32 pblk_calc_emeta_crc(struct pblk *pblk,
1209 struct line_emeta *emeta)
1210 {
1211 struct pblk_line_meta *lm = &pblk->lm;
1212 u32 crc = ~(u32)0;
1213
1214 crc = crc32_le(crc, (unsigned char *)emeta +
1215 sizeof(struct line_header) + sizeof(crc),
1216 lm->emeta_len[0] -
1217 sizeof(struct line_header) - sizeof(crc));
1218
1219 return crc;
1220 }
1221
1222 static inline int pblk_io_aligned(struct pblk *pblk, int nr_secs)
1223 {
1224 return !(nr_secs % pblk->min_write_pgs);
1225 }
1226
1227 #ifdef CONFIG_NVM_PBLK_DEBUG
1228 static inline void print_ppa(struct pblk *pblk, struct ppa_addr *p,
1229 char *msg, int error)
1230 {
1231 struct nvm_geo *geo = &pblk->dev->geo;
1232
1233 if (p->c.is_cached) {
1234 pblk_err(pblk, "ppa: (%s: %x) cache line: %llu\n",
1235 msg, error, (u64)p->c.line);
1236 } else if (geo->version == NVM_OCSSD_SPEC_12) {
1237 pblk_err(pblk, "ppa: (%s: %x):ch:%d,lun:%d,blk:%d,pg:%d,pl:%d,sec:%d\n",
1238 msg, error,
1239 p->g.ch, p->g.lun, p->g.blk,
1240 p->g.pg, p->g.pl, p->g.sec);
1241 } else {
1242 pblk_err(pblk, "ppa: (%s: %x):ch:%d,lun:%d,chk:%d,sec:%d\n",
1243 msg, error,
1244 p->m.grp, p->m.pu, p->m.chk, p->m.sec);
1245 }
1246 }
1247
1248 static inline void pblk_print_failed_rqd(struct pblk *pblk, struct nvm_rq *rqd,
1249 int error)
1250 {
1251 int bit = -1;
1252
1253 if (rqd->nr_ppas == 1) {
1254 print_ppa(pblk, &rqd->ppa_addr, "rqd", error);
1255 return;
1256 }
1257
1258 while ((bit = find_next_bit((void *)&rqd->ppa_status, rqd->nr_ppas,
1259 bit + 1)) < rqd->nr_ppas) {
1260 print_ppa(pblk, &rqd->ppa_list[bit], "rqd", error);
1261 }
1262
1263 pblk_err(pblk, "error:%d, ppa_status:%llx\n", error, rqd->ppa_status);
1264 }
1265
1266 static inline int pblk_boundary_ppa_checks(struct nvm_tgt_dev *tgt_dev,
1267 struct ppa_addr *ppas, int nr_ppas)
1268 {
1269 struct nvm_geo *geo = &tgt_dev->geo;
1270 struct ppa_addr *ppa;
1271 int i;
1272
1273 for (i = 0; i < nr_ppas; i++) {
1274 ppa = &ppas[i];
1275
1276 if (geo->version == NVM_OCSSD_SPEC_12) {
1277 if (!ppa->c.is_cached &&
1278 ppa->g.ch < geo->num_ch &&
1279 ppa->g.lun < geo->num_lun &&
1280 ppa->g.pl < geo->num_pln &&
1281 ppa->g.blk < geo->num_chk &&
1282 ppa->g.pg < geo->num_pg &&
1283 ppa->g.sec < geo->ws_min)
1284 continue;
1285 } else {
1286 if (!ppa->c.is_cached &&
1287 ppa->m.grp < geo->num_ch &&
1288 ppa->m.pu < geo->num_lun &&
1289 ppa->m.chk < geo->num_chk &&
1290 ppa->m.sec < geo->clba)
1291 continue;
1292 }
1293
1294 print_ppa(tgt_dev->q->queuedata, ppa, "boundary", i);
1295
1296 return 1;
1297 }
1298 return 0;
1299 }
1300
1301 static inline int pblk_check_io(struct pblk *pblk, struct nvm_rq *rqd)
1302 {
1303 struct nvm_tgt_dev *dev = pblk->dev;
1304 struct ppa_addr *ppa_list = nvm_rq_to_ppa_list(rqd);
1305
1306 if (pblk_boundary_ppa_checks(dev, ppa_list, rqd->nr_ppas)) {
1307 WARN_ON(1);
1308 return -EINVAL;
1309 }
1310
1311 if (rqd->opcode == NVM_OP_PWRITE) {
1312 struct pblk_line *line;
1313 int i;
1314
1315 for (i = 0; i < rqd->nr_ppas; i++) {
1316 line = pblk_ppa_to_line(pblk, ppa_list[i]);
1317
1318 spin_lock(&line->lock);
1319 if (line->state != PBLK_LINESTATE_OPEN) {
1320 pblk_err(pblk, "bad ppa: line:%d,state:%d\n",
1321 line->id, line->state);
1322 WARN_ON(1);
1323 spin_unlock(&line->lock);
1324 return -EINVAL;
1325 }
1326 spin_unlock(&line->lock);
1327 }
1328 }
1329
1330 return 0;
1331 }
1332 #endif
1333
1334 static inline int pblk_boundary_paddr_checks(struct pblk *pblk, u64 paddr)
1335 {
1336 struct pblk_line_meta *lm = &pblk->lm;
1337
1338 if (paddr > lm->sec_per_line)
1339 return 1;
1340
1341 return 0;
1342 }
1343
1344 static inline unsigned int pblk_get_bi_idx(struct bio *bio)
1345 {
1346 return bio->bi_iter.bi_idx;
1347 }
1348
1349 static inline sector_t pblk_get_lba(struct bio *bio)
1350 {
1351 return bio->bi_iter.bi_sector / NR_PHY_IN_LOG;
1352 }
1353
1354 static inline unsigned int pblk_get_secs(struct bio *bio)
1355 {
1356 return bio->bi_iter.bi_size / PBLK_EXPOSED_PAGE_SIZE;
1357 }
1358
1359 static inline void pblk_setup_uuid(struct pblk *pblk)
1360 {
1361 uuid_le uuid;
1362
1363 uuid_le_gen(&uuid);
1364 memcpy(pblk->instance_uuid, uuid.b, 16);
1365 }
1366
1367 static inline char *pblk_disk_name(struct pblk *pblk)
1368 {
1369 struct gendisk *disk = pblk->disk;
1370
1371 return disk->disk_name;
1372 }
1373 #endif /* PBLK_H_ */
Linux with added FPC-III support