2 * Copyright (C) 2015 IT University of Copenhagen (rrpc.h)
3 * Copyright (C) 2016 CNEX Labs
4 * Initial release: Matias Bjorling <matias@cnexlabs.com>
5 * Write buffering: Javier Gonzalez <javier@cnexlabs.com>
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.
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.
16 * Implementation of a Physical Block-device target for Open-channel SSDs.
23 #include <linux/blkdev.h>
24 #include <linux/blk-mq.h>
25 #include <linux/bio.h>
26 #include <linux/module.h>
27 #include <linux/kthread.h>
28 #include <linux/vmalloc.h>
29 #include <linux/crc32.h>
30 #include <linux/uuid.h>
32 #include <linux/lightnvm.h>
34 /* Run only GC if less than 1/X blocks are free */
35 #define GC_LIMIT_INVERSE 5
36 #define GC_TIME_MSECS 1000
38 #define PBLK_SECTOR (512)
39 #define PBLK_EXPOSED_PAGE_SIZE (4096)
40 #define PBLK_MAX_REQ_ADDRS (64)
41 #define PBLK_MAX_REQ_ADDRS_PW (6)
43 #define PBLK_CACHE_NAME_LEN (DISK_NAME_LEN + 16)
45 #define PBLK_COMMAND_TIMEOUT_MS 30000
47 /* Max 512 LUNs per device */
48 #define PBLK_MAX_LUNS_BITMAP (4)
50 #define NR_PHY_IN_LOG (PBLK_EXPOSED_PAGE_SIZE / PBLK_SECTOR)
52 #define pblk_for_each_lun(pblk, rlun, i) \
53 for ((i) = 0, rlun = &(pblk)->luns[0]; \
54 (i) < (pblk)->nr_luns; (i)++, rlun = &(pblk)->luns[(i)])
56 #define ERASE 2 /* READ = 0, WRITE = 1 */
60 PBLK_IOTYPE_USER = 1 << 0,
61 PBLK_IOTYPE_GC = 1 << 1,
63 /* Write buffer flags */
64 PBLK_FLUSH_ENTRY = 1 << 2,
65 PBLK_WRITTEN_DATA = 1 << 3,
66 PBLK_SUBMITTED_ENTRY = 1 << 4,
67 PBLK_WRITABLE_ENTRY = 1 << 5,
71 PBLK_BLK_ST_OPEN = 0x1,
72 PBLK_BLK_ST_CLOSED = 0x2,
75 /* The number of GC lists and the rate-limiter states go together. This way the
76 * rate-limiter can dictate how much GC is needed based on resource utilization.
78 #define PBLK_NR_GC_LISTS 3
79 #define PBLK_MAX_GC_JOBS 32
87 struct pblk_sec_meta {
92 #define pblk_dma_meta_size (sizeof(struct pblk_sec_meta) * PBLK_MAX_REQ_ADDRS)
94 /* write buffer completion context */
96 struct list_head list; /* Head for out-of-order completion */
98 unsigned long *lun_bitmap; /* Luns used on current request */
100 unsigned int nr_valid;
101 unsigned int nr_padded;
104 /* generic context */
109 /* Recovery context */
110 struct pblk_rec_ctx {
113 struct list_head failed;
114 struct work_struct ws_rec;
119 struct bio_list bios; /* Original bios - used for completion
120 * in REQ_FUA, REQ_FLUSH case
122 u64 lba; /* Logic addr. associated with entry */
123 struct ppa_addr ppa; /* Physic addr. associated with entry */
124 int flags; /* Write context flags */
127 struct pblk_rb_entry {
128 struct ppa_addr cacheline; /* Cacheline for this entry */
129 void *data; /* Pointer to data on this entry */
130 struct pblk_w_ctx w_ctx; /* Context for this entry */
131 struct list_head index; /* List head to enable indexes */
134 #define EMPTY_ENTRY (~0U)
136 struct pblk_rb_pages {
139 struct list_head list;
143 struct pblk_rb_entry *entries; /* Ring buffer entries */
144 unsigned int mem; /* Write offset - points to next
145 * writable entry in memory
147 unsigned int subm; /* Read offset - points to last entry
148 * that has been submitted to the media
151 unsigned int sync; /* Synced - backpointer that signals
152 * the last submitted entry that has
153 * been successfully persisted to media
155 unsigned int sync_point; /* Sync point - last entry that must be
156 * flushed to the media. Used with
157 * REQ_FLUSH and REQ_FUA
159 unsigned int l2p_update; /* l2p update point - next entry for
160 * which l2p mapping will be updated to
161 * contain a device ppa address (instead
164 unsigned int nr_entries; /* Number of entries in write buffer -
165 * must be a power of two
167 unsigned int seg_size; /* Size of the data segments being
168 * stored on each entry. Typically this
172 struct list_head pages; /* List of data pages */
174 spinlock_t w_lock; /* Write lock */
175 spinlock_t s_lock; /* Sync lock */
177 #ifdef CONFIG_NVM_DEBUG
178 atomic_t inflight_sync_point; /* Not served REQ_FLUSH | REQ_FUA */
182 #define PBLK_RECOVERY_SECTORS 16
185 struct ppa_addr bppa;
187 u8 *bb_list; /* Bad block list for LUN. Only used on
188 * bring up. Bad blocks are managed
189 * within lines on run-time.
192 struct semaphore wr_sem;
196 struct pblk_line *line;
201 struct list_head list;
209 atomic_t inflight_gc;
211 struct task_struct *gc_ts;
212 struct task_struct *gc_writer_ts;
213 struct workqueue_struct *gc_reader_wq;
214 struct timer_list gc_timer;
217 struct list_head w_list;
224 unsigned int high; /* Upper threshold for rate limiter (free run -
225 * user I/O rate limiter
227 unsigned int low; /* Lower threshold for rate limiter (user I/O
228 * rate limiter - stall)
230 unsigned int high_pw; /* High rounded up as a power of 2 */
232 #define PBLK_USER_HIGH_THRS 2 /* Begin write limit at 50 percent
235 #define PBLK_USER_LOW_THRS 20 /* Aggressive GC at 5% available blocks */
237 int rb_windows_pw; /* Number of rate windows in the write buffer
238 * given as a power-of-2. This guarantees that
239 * when user I/O is being rate limited, there
240 * will be reserved enough space for the GC to
241 * place its payload. A window is of
242 * pblk->max_write_pgs size, which in NVMe is
245 int rb_budget; /* Total number of entries available for I/O */
246 int rb_user_max; /* Max buffer entries available for user I/O */
247 atomic_t rb_user_cnt; /* User I/O buffer counter */
248 int rb_gc_max; /* Max buffer entries available for GC I/O */
249 int rb_gc_rsv; /* Reserved buffer entries for GC I/O */
250 int rb_state; /* Rate-limiter current state */
251 atomic_t rb_gc_cnt; /* GC I/O buffer counter */
254 struct timer_list u_timer;
256 unsigned long long nr_secs;
257 unsigned long total_blocks;
258 atomic_t free_blocks;
261 #define PBLK_LINE_EMPTY (~0U)
265 PBLK_LINETYPE_FREE = 0,
266 PBLK_LINETYPE_LOG = 1,
267 PBLK_LINETYPE_DATA = 2,
270 PBLK_LINESTATE_FREE = 10,
271 PBLK_LINESTATE_OPEN = 11,
272 PBLK_LINESTATE_CLOSED = 12,
273 PBLK_LINESTATE_GC = 13,
274 PBLK_LINESTATE_BAD = 14,
275 PBLK_LINESTATE_CORRUPT = 15,
278 PBLK_LINEGC_NONE = 20,
279 PBLK_LINEGC_EMPTY = 21,
280 PBLK_LINEGC_LOW = 22,
281 PBLK_LINEGC_MID = 23,
282 PBLK_LINEGC_HIGH = 24,
283 PBLK_LINEGC_FULL = 25,
286 #define PBLK_MAGIC 0x70626c6b /*pblk*/
290 __le32 identifier; /* pblk identifier */
291 __u8 uuid[16]; /* instance uuid */
292 __le16 type; /* line type */
293 __le16 version; /* type version */
294 __le32 id; /* line id for current line */
298 struct line_header header;
300 __le32 crc; /* Full structure including struct crc */
301 /* Previous line metadata */
302 __le32 prev_id; /* Line id for previous line */
304 /* Current line metadata */
305 __le64 seq_nr; /* Sequence number for current line */
308 __le32 window_wr_lun; /* Number of parallel LUNs to write */
316 * Metadata layout in media:
318 * 1. struct line_emeta
319 * 2. bad block bitmap (u64 * window_wr_lun)
320 * Mid sectors (start at lbas_sector):
321 * 3. nr_lbas (u64) forming lba list
322 * Last sectors (start at vsc_sector):
323 * 4. u32 valid sector count (vsc) for all lines (~0U: free line)
326 struct line_header header;
328 __le32 crc; /* Full structure including struct crc */
330 /* Previous line metadata */
331 __le32 prev_id; /* Line id for prev line */
333 /* Current line metadata */
334 __le64 seq_nr; /* Sequence number for current line */
337 __le32 window_wr_lun; /* Number of parallel LUNs to write */
339 /* Bookkeeping for recovery */
340 __le32 next_id; /* Line id for next line */
341 __le64 nr_lbas; /* Number of lbas mapped in line */
342 __le64 nr_valid_lbas; /* Number of valid lbas mapped in line */
343 __le64 bb_bitmap[]; /* Updated bad block bitmap for line */
347 struct line_emeta *buf; /* emeta buffer in media format */
348 int mem; /* Write offset - points to next
349 * writable entry in memory
351 atomic_t sync; /* Synced - backpointer that signals the
352 * last entry that has been successfully
355 unsigned int nr_entries; /* Number of emeta entries */
359 struct line_smeta *buf; /* smeta buffer in persistent format */
364 unsigned int id; /* Line number corresponds to the
367 unsigned int seq_nr; /* Unique line sequence number */
369 int state; /* PBLK_LINESTATE_X */
370 int type; /* PBLK_LINETYPE_X */
371 int gc_group; /* PBLK_LINEGC_X */
372 struct list_head list; /* Free, GC lists */
374 unsigned long *lun_bitmap; /* Bitmap for LUNs mapped in line */
376 struct pblk_smeta *smeta; /* Start metadata */
377 struct pblk_emeta *emeta; /* End medatada */
379 int meta_line; /* Metadata line id */
380 int meta_distance; /* Distance between data and metadata */
382 u64 smeta_ssec; /* Sector where smeta starts */
383 u64 emeta_ssec; /* Sector where emeta starts */
385 unsigned int sec_in_line; /* Number of usable secs in line */
387 atomic_t blk_in_line; /* Number of good blocks in line */
388 unsigned long *blk_bitmap; /* Bitmap for valid/invalid blocks */
389 unsigned long *erase_bitmap; /* Bitmap for erased blocks */
391 unsigned long *map_bitmap; /* Bitmap for mapped sectors in line */
392 unsigned long *invalid_bitmap; /* Bitmap for invalid sectors in line */
394 atomic_t left_eblks; /* Blocks left for erasing */
395 atomic_t left_seblks; /* Blocks left for sync erasing */
397 int left_msecs; /* Sectors left for mapping */
398 unsigned int cur_sec; /* Sector map pointer */
399 unsigned int nr_valid_lbas; /* Number of valid lbas in line */
401 __le32 *vsc; /* Valid sector count in line */
403 struct kref ref; /* Write buffer L2P references */
405 spinlock_t lock; /* Necessary for invalid_bitmap only */
408 #define PBLK_DATA_LINES 4
411 PBLK_KMALLOC_META = 1,
412 PBLK_VMALLOC_META = 2,
416 PBLK_EMETA_TYPE_HEADER = 1, /* struct line_emeta first sector */
417 PBLK_EMETA_TYPE_LLBA = 2, /* lba list - type: __le64 */
418 PBLK_EMETA_TYPE_VSC = 3, /* vsc list - type: __le32 */
421 struct pblk_line_mgmt {
422 int nr_lines; /* Total number of full lines */
423 int nr_free_lines; /* Number of full lines in free list */
425 /* Free lists - use free_lock */
426 struct list_head free_list; /* Full lines ready to use */
427 struct list_head corrupt_list; /* Full lines corrupted */
428 struct list_head bad_list; /* Full lines bad */
430 /* GC lists - use gc_lock */
431 struct list_head *gc_lists[PBLK_NR_GC_LISTS];
432 struct list_head gc_high_list; /* Full lines ready to GC, high isc */
433 struct list_head gc_mid_list; /* Full lines ready to GC, mid isc */
434 struct list_head gc_low_list; /* Full lines ready to GC, low isc */
436 struct list_head gc_full_list; /* Full lines ready to GC, no valid */
437 struct list_head gc_empty_list; /* Full lines close, all valid */
439 struct pblk_line *log_line; /* Current FTL log line */
440 struct pblk_line *data_line; /* Current data line */
441 struct pblk_line *log_next; /* Next FTL log line */
442 struct pblk_line *data_next; /* Next data line */
444 struct list_head emeta_list; /* Lines queued to schedule emeta */
446 __le32 *vsc_list; /* Valid sector counts for all lines */
448 /* Metadata allocation type: VMALLOC | KMALLOC */
449 int smeta_alloc_type;
450 int emeta_alloc_type;
452 /* Pre-allocated metadata for data lines */
453 struct pblk_smeta *sline_meta[PBLK_DATA_LINES];
454 struct pblk_emeta *eline_meta[PBLK_DATA_LINES];
455 unsigned long meta_bitmap;
457 /* Helpers for fast bitmap calculations */
458 unsigned long *bb_template;
459 unsigned long *bb_aux;
461 unsigned long d_seq_nr; /* Data line unique sequence number */
462 unsigned long l_seq_nr; /* Log line unique sequence number */
464 spinlock_t free_lock;
465 spinlock_t close_lock;
469 struct pblk_line_meta {
470 unsigned int smeta_len; /* Total length for smeta */
471 unsigned int smeta_sec; /* Sectors needed for smeta */
473 unsigned int emeta_len[4]; /* Lengths for emeta:
475 * [1]: struct line_emeta length
476 * [2]: L2P portion length
477 * [3]: vsc list length
479 unsigned int emeta_sec[4]; /* Sectors needed for emeta. Same layout
483 unsigned int emeta_bb; /* Boundary for bb that affects emeta */
485 unsigned int vsc_list_len; /* Length for vsc list */
486 unsigned int sec_bitmap_len; /* Length for sector bitmap in line */
487 unsigned int blk_bitmap_len; /* Length for block bitmap in line */
488 unsigned int lun_bitmap_len; /* Length for lun bitmap in line */
490 unsigned int blk_per_line; /* Number of blocks in a full line */
491 unsigned int sec_per_line; /* Number of sectors in a line */
492 unsigned int dsec_per_line; /* Number of data sectors in a line */
493 unsigned int min_blk_line; /* Min. number of good blocks in line */
495 unsigned int mid_thrs; /* Threshold for GC mid list */
496 unsigned int high_thrs; /* Threshold for GC high list */
498 unsigned int meta_distance; /* Distance between data and metadata */
501 struct pblk_addr_format {
517 struct nvm_tgt_dev *dev;
518 struct gendisk *disk;
522 struct pblk_lun *luns;
524 struct pblk_line *lines; /* Line array */
525 struct pblk_line_mgmt l_mg; /* Line management */
526 struct pblk_line_meta lm; /* Line metadata */
529 struct pblk_addr_format ppaf;
533 int min_write_pgs; /* Minimum amount of pages required by controller */
534 int max_write_pgs; /* Maximum amount of pages supported by controller */
535 int pgs_in_buffer; /* Number of pages that need to be held in buffer to
536 * guarantee successful reads.
539 sector_t capacity; /* Device capacity when bad blocks are subtracted */
540 int over_pct; /* Percentage of device used for over-provisioning */
542 /* pblk provisioning values. Used by rate limiter */
547 unsigned char instance_uuid[16];
548 #ifdef CONFIG_NVM_DEBUG
549 /* All debug counters apply to 4kb sector I/Os */
550 atomic_long_t inflight_writes; /* Inflight writes (user and gc) */
551 atomic_long_t padded_writes; /* Sectors padded due to flush/fua */
552 atomic_long_t padded_wb; /* Sectors padded in write buffer */
553 atomic_long_t nr_flush; /* Number of flush/fua I/O */
554 atomic_long_t req_writes; /* Sectors stored on write buffer */
555 atomic_long_t sub_writes; /* Sectors submitted from buffer */
556 atomic_long_t sync_writes; /* Sectors synced to media */
557 atomic_long_t inflight_reads; /* Inflight sector read requests */
558 atomic_long_t cache_reads; /* Read requests that hit the cache */
559 atomic_long_t sync_reads; /* Completed sector read requests */
560 atomic_long_t recov_writes; /* Sectors submitted from recovery */
561 atomic_long_t recov_gc_writes; /* Sectors submitted from write GC */
562 atomic_long_t recov_gc_reads; /* Sectors submitted from read GC */
567 atomic_long_t read_failed;
568 atomic_long_t read_empty;
569 atomic_long_t read_high_ecc;
570 atomic_long_t read_failed_gc;
571 atomic_long_t write_failed;
572 atomic_long_t erase_failed;
574 struct task_struct *writer_ts;
576 /* Simple translation map of logical addresses to physical addresses.
577 * The logical addresses is known by the host system, while the physical
578 * addresses are used when writing to the disk block device.
580 unsigned char *trans_map;
581 spinlock_t trans_lock;
583 struct list_head compl_list;
585 mempool_t *page_pool;
586 mempool_t *line_ws_pool;
588 mempool_t *g_rq_pool;
589 mempool_t *w_rq_pool;
590 mempool_t *line_meta_pool;
592 struct workqueue_struct *kw_wq;
593 struct timer_list wtimer;
598 struct pblk_line_ws {
600 struct pblk_line *line;
602 struct work_struct ws;
605 #define pblk_g_rq_size (sizeof(struct nvm_rq) + sizeof(struct pblk_g_ctx))
606 #define pblk_w_rq_size (sizeof(struct nvm_rq) + sizeof(struct pblk_c_ctx))
609 * pblk ring buffer operations
611 int pblk_rb_init(struct pblk_rb *rb, struct pblk_rb_entry *rb_entry_base,
612 unsigned int power_size, unsigned int power_seg_sz);
613 unsigned int pblk_rb_calculate_size(unsigned int nr_entries);
614 void *pblk_rb_entries_ref(struct pblk_rb *rb);
615 int pblk_rb_may_write_user(struct pblk_rb *rb, struct bio *bio,
616 unsigned int nr_entries, unsigned int *pos);
617 int pblk_rb_may_write_gc(struct pblk_rb *rb, unsigned int nr_entries,
619 void pblk_rb_write_entry_user(struct pblk_rb *rb, void *data,
620 struct pblk_w_ctx w_ctx, unsigned int pos);
621 void pblk_rb_write_entry_gc(struct pblk_rb *rb, void *data,
622 struct pblk_w_ctx w_ctx, struct pblk_line *gc_line,
624 struct pblk_w_ctx *pblk_rb_w_ctx(struct pblk_rb *rb, unsigned int pos);
626 void pblk_rb_sync_l2p(struct pblk_rb *rb);
627 unsigned int pblk_rb_read_to_bio(struct pblk_rb *rb, struct nvm_rq *rqd,
628 struct bio *bio, unsigned int pos,
629 unsigned int nr_entries, unsigned int count);
630 unsigned int pblk_rb_read_to_bio_list(struct pblk_rb *rb, struct bio *bio,
631 struct list_head *list,
633 int pblk_rb_copy_to_bio(struct pblk_rb *rb, struct bio *bio, sector_t lba,
634 u64 pos, int bio_iter);
635 unsigned int pblk_rb_read_commit(struct pblk_rb *rb, unsigned int entries);
637 unsigned int pblk_rb_sync_init(struct pblk_rb *rb, unsigned long *flags);
638 unsigned int pblk_rb_sync_advance(struct pblk_rb *rb, unsigned int nr_entries);
639 struct pblk_rb_entry *pblk_rb_sync_scan_entry(struct pblk_rb *rb,
640 struct ppa_addr *ppa);
641 void pblk_rb_sync_end(struct pblk_rb *rb, unsigned long *flags);
642 unsigned int pblk_rb_sync_point_count(struct pblk_rb *rb);
644 unsigned int pblk_rb_read_count(struct pblk_rb *rb);
645 unsigned int pblk_rb_wrap_pos(struct pblk_rb *rb, unsigned int pos);
647 int pblk_rb_tear_down_check(struct pblk_rb *rb);
648 int pblk_rb_pos_oob(struct pblk_rb *rb, u64 pos);
649 void pblk_rb_data_free(struct pblk_rb *rb);
650 ssize_t pblk_rb_sysfs(struct pblk_rb *rb, char *buf);
655 struct nvm_rq *pblk_alloc_rqd(struct pblk *pblk, int rw);
656 void pblk_set_sec_per_write(struct pblk *pblk, int sec_per_write);
657 int pblk_setup_w_rec_rq(struct pblk *pblk, struct nvm_rq *rqd,
658 struct pblk_c_ctx *c_ctx);
659 void pblk_free_rqd(struct pblk *pblk, struct nvm_rq *rqd, int rw);
660 void pblk_flush_writer(struct pblk *pblk);
661 struct ppa_addr pblk_get_lba_map(struct pblk *pblk, sector_t lba);
662 void pblk_discard(struct pblk *pblk, struct bio *bio);
663 void pblk_log_write_err(struct pblk *pblk, struct nvm_rq *rqd);
664 void pblk_log_read_err(struct pblk *pblk, struct nvm_rq *rqd);
665 int pblk_submit_io(struct pblk *pblk, struct nvm_rq *rqd);
666 int pblk_submit_meta_io(struct pblk *pblk, struct pblk_line *meta_line);
667 struct bio *pblk_bio_map_addr(struct pblk *pblk, void *data,
668 unsigned int nr_secs, unsigned int len,
670 struct pblk_line *pblk_line_get(struct pblk *pblk);
671 struct pblk_line *pblk_line_get_first_data(struct pblk *pblk);
672 struct pblk_line *pblk_line_replace_data(struct pblk *pblk);
673 int pblk_line_recov_alloc(struct pblk *pblk, struct pblk_line *line);
674 void pblk_line_recov_close(struct pblk *pblk, struct pblk_line *line);
675 struct pblk_line *pblk_line_get_data(struct pblk *pblk);
676 struct pblk_line *pblk_line_get_erase(struct pblk *pblk);
677 int pblk_line_erase(struct pblk *pblk, struct pblk_line *line);
678 int pblk_line_is_full(struct pblk_line *line);
679 void pblk_line_free(struct pblk *pblk, struct pblk_line *line);
680 void pblk_line_close_meta(struct pblk *pblk, struct pblk_line *line);
681 void pblk_line_close(struct pblk *pblk, struct pblk_line *line);
682 void pblk_line_close_ws(struct work_struct *work);
683 void pblk_line_mark_bb(struct work_struct *work);
684 void pblk_line_run_ws(struct pblk *pblk, struct pblk_line *line, void *priv,
685 void (*work)(struct work_struct *));
686 u64 pblk_line_smeta_start(struct pblk *pblk, struct pblk_line *line);
687 int pblk_line_read_smeta(struct pblk *pblk, struct pblk_line *line);
688 int pblk_line_read_emeta(struct pblk *pblk, struct pblk_line *line,
690 int pblk_blk_erase_async(struct pblk *pblk, struct ppa_addr erase_ppa);
691 void pblk_line_put(struct kref *ref);
692 struct list_head *pblk_line_gc_list(struct pblk *pblk, struct pblk_line *line);
693 u64 pblk_lookup_page(struct pblk *pblk, struct pblk_line *line);
694 void pblk_dealloc_page(struct pblk *pblk, struct pblk_line *line, int nr_secs);
695 u64 pblk_alloc_page(struct pblk *pblk, struct pblk_line *line, int nr_secs);
696 u64 __pblk_alloc_page(struct pblk *pblk, struct pblk_line *line, int nr_secs);
697 int pblk_calc_secs(struct pblk *pblk, unsigned long secs_avail,
698 unsigned long secs_to_flush);
699 void pblk_down_rq(struct pblk *pblk, struct ppa_addr *ppa_list, int nr_ppas,
700 unsigned long *lun_bitmap);
701 void pblk_up_rq(struct pblk *pblk, struct ppa_addr *ppa_list, int nr_ppas,
702 unsigned long *lun_bitmap);
703 void pblk_end_bio_sync(struct bio *bio);
704 void pblk_end_io_sync(struct nvm_rq *rqd);
705 int pblk_bio_add_pages(struct pblk *pblk, struct bio *bio, gfp_t flags,
707 void pblk_bio_free_pages(struct pblk *pblk, struct bio *bio, int off,
709 void pblk_map_invalidate(struct pblk *pblk, struct ppa_addr ppa);
710 void __pblk_map_invalidate(struct pblk *pblk, struct pblk_line *line,
712 void pblk_update_map(struct pblk *pblk, sector_t lba, struct ppa_addr ppa);
713 void pblk_update_map_cache(struct pblk *pblk, sector_t lba,
714 struct ppa_addr ppa);
715 void pblk_update_map_dev(struct pblk *pblk, sector_t lba,
716 struct ppa_addr ppa, struct ppa_addr entry_line);
717 int pblk_update_map_gc(struct pblk *pblk, sector_t lba, struct ppa_addr ppa,
718 struct pblk_line *gc_line);
719 void pblk_lookup_l2p_rand(struct pblk *pblk, struct ppa_addr *ppas,
720 u64 *lba_list, int nr_secs);
721 void pblk_lookup_l2p_seq(struct pblk *pblk, struct ppa_addr *ppas,
722 sector_t blba, int nr_secs);
725 * pblk user I/O write path
727 int pblk_write_to_cache(struct pblk *pblk, struct bio *bio,
728 unsigned long flags);
729 int pblk_write_gc_to_cache(struct pblk *pblk, void *data, u64 *lba_list,
730 unsigned int nr_entries, unsigned int nr_rec_entries,
731 struct pblk_line *gc_line, unsigned long flags);
736 void pblk_map_erase_rq(struct pblk *pblk, struct nvm_rq *rqd,
737 unsigned int sentry, unsigned long *lun_bitmap,
738 unsigned int valid_secs, struct ppa_addr *erase_ppa);
739 void pblk_map_rq(struct pblk *pblk, struct nvm_rq *rqd, unsigned int sentry,
740 unsigned long *lun_bitmap, unsigned int valid_secs,
746 int pblk_write_ts(void *data);
747 void pblk_write_timer_fn(unsigned long data);
748 void pblk_write_should_kick(struct pblk *pblk);
753 extern struct bio_set *pblk_bio_set;
754 int pblk_submit_read(struct pblk *pblk, struct bio *bio);
755 int pblk_submit_read_gc(struct pblk *pblk, u64 *lba_list, void *data,
756 unsigned int nr_secs, unsigned int *secs_to_gc,
757 struct pblk_line *line);
761 void pblk_submit_rec(struct work_struct *work);
762 struct pblk_line *pblk_recov_l2p(struct pblk *pblk);
763 void pblk_recov_pad(struct pblk *pblk);
764 __le64 *pblk_recov_get_lba_list(struct pblk *pblk, struct line_emeta *emeta);
765 int pblk_recov_setup_rq(struct pblk *pblk, struct pblk_c_ctx *c_ctx,
766 struct pblk_rec_ctx *recovery, u64 *comp_bits,
772 #define PBLK_GC_TRIES 3
774 int pblk_gc_init(struct pblk *pblk);
775 void pblk_gc_exit(struct pblk *pblk);
776 void pblk_gc_should_start(struct pblk *pblk);
777 void pblk_gc_should_stop(struct pblk *pblk);
778 int pblk_gc_status(struct pblk *pblk);
779 void pblk_gc_sysfs_state_show(struct pblk *pblk, int *gc_enabled,
781 void pblk_gc_sysfs_force(struct pblk *pblk, int force);
786 void pblk_rl_init(struct pblk_rl *rl, int budget);
787 void pblk_rl_free(struct pblk_rl *rl);
788 int pblk_rl_gc_thrs(struct pblk_rl *rl);
789 unsigned long pblk_rl_nr_free_blks(struct pblk_rl *rl);
790 int pblk_rl_user_may_insert(struct pblk_rl *rl, int nr_entries);
791 void pblk_rl_user_in(struct pblk_rl *rl, int nr_entries);
792 int pblk_rl_gc_may_insert(struct pblk_rl *rl, int nr_entries);
793 void pblk_rl_gc_in(struct pblk_rl *rl, int nr_entries);
794 void pblk_rl_out(struct pblk_rl *rl, int nr_user, int nr_gc);
795 void pblk_rl_set_gc_rsc(struct pblk_rl *rl, int rsv);
796 int pblk_rl_sysfs_rate_show(struct pblk_rl *rl);
797 void pblk_rl_free_lines_inc(struct pblk_rl *rl, struct pblk_line *line);
798 void pblk_rl_free_lines_dec(struct pblk_rl *rl, struct pblk_line *line);
803 int pblk_sysfs_init(struct gendisk *tdisk);
804 void pblk_sysfs_exit(struct gendisk *tdisk);
806 static inline void *pblk_malloc(size_t size, int type, gfp_t flags)
808 if (type == PBLK_KMALLOC_META)
809 return kmalloc(size, flags);
810 return vmalloc(size);
813 static inline void pblk_mfree(void *ptr, int type)
815 if (type == PBLK_KMALLOC_META)
821 static inline struct nvm_rq *nvm_rq_from_c_ctx(void *c_ctx)
823 return c_ctx - sizeof(struct nvm_rq);
826 static inline void *emeta_to_bb(struct line_emeta *emeta)
828 return emeta->bb_bitmap;
831 static inline void *emeta_to_lbas(struct pblk *pblk, struct line_emeta *emeta)
833 return ((void *)emeta + pblk->lm.emeta_len[1]);
836 static inline void *emeta_to_vsc(struct pblk *pblk, struct line_emeta *emeta)
838 return (emeta_to_lbas(pblk, emeta) + pblk->lm.emeta_len[2]);
841 #define NVM_MEM_PAGE_WRITE (8)
843 static inline int pblk_pad_distance(struct pblk *pblk)
845 struct nvm_tgt_dev *dev = pblk->dev;
846 struct nvm_geo *geo = &dev->geo;
848 return NVM_MEM_PAGE_WRITE * geo->nr_luns * geo->sec_per_pl;
851 static inline int pblk_dev_ppa_to_line(struct ppa_addr p)
856 static inline int pblk_tgt_ppa_to_line(struct ppa_addr p)
861 static inline int pblk_ppa_to_pos(struct nvm_geo *geo, struct ppa_addr p)
863 return p.g.lun * geo->nr_chnls + p.g.ch;
866 /* A block within a line corresponds to the lun */
867 static inline int pblk_dev_ppa_to_pos(struct nvm_geo *geo, struct ppa_addr p)
869 return p.g.lun * geo->nr_chnls + p.g.ch;
872 static inline struct ppa_addr pblk_ppa32_to_ppa64(struct pblk *pblk, u32 ppa32)
874 struct ppa_addr ppa64;
879 ppa64.ppa = ADDR_EMPTY;
880 } else if (ppa32 & (1U << 31)) {
881 ppa64.c.line = ppa32 & ((~0U) >> 1);
882 ppa64.c.is_cached = 1;
884 ppa64.g.blk = (ppa32 & pblk->ppaf.blk_mask) >>
885 pblk->ppaf.blk_offset;
886 ppa64.g.pg = (ppa32 & pblk->ppaf.pg_mask) >>
887 pblk->ppaf.pg_offset;
888 ppa64.g.lun = (ppa32 & pblk->ppaf.lun_mask) >>
889 pblk->ppaf.lun_offset;
890 ppa64.g.ch = (ppa32 & pblk->ppaf.ch_mask) >>
891 pblk->ppaf.ch_offset;
892 ppa64.g.pl = (ppa32 & pblk->ppaf.pln_mask) >>
893 pblk->ppaf.pln_offset;
894 ppa64.g.sec = (ppa32 & pblk->ppaf.sec_mask) >>
895 pblk->ppaf.sec_offset;
901 static inline struct ppa_addr pblk_trans_map_get(struct pblk *pblk,
906 if (pblk->ppaf_bitsize < 32) {
907 u32 *map = (u32 *)pblk->trans_map;
909 ppa = pblk_ppa32_to_ppa64(pblk, map[lba]);
911 struct ppa_addr *map = (struct ppa_addr *)pblk->trans_map;
919 static inline u32 pblk_ppa64_to_ppa32(struct pblk *pblk, struct ppa_addr ppa64)
923 if (ppa64.ppa == ADDR_EMPTY) {
925 } else if (ppa64.c.is_cached) {
926 ppa32 |= ppa64.c.line;
929 ppa32 |= ppa64.g.blk << pblk->ppaf.blk_offset;
930 ppa32 |= ppa64.g.pg << pblk->ppaf.pg_offset;
931 ppa32 |= ppa64.g.lun << pblk->ppaf.lun_offset;
932 ppa32 |= ppa64.g.ch << pblk->ppaf.ch_offset;
933 ppa32 |= ppa64.g.pl << pblk->ppaf.pln_offset;
934 ppa32 |= ppa64.g.sec << pblk->ppaf.sec_offset;
940 static inline void pblk_trans_map_set(struct pblk *pblk, sector_t lba,
943 if (pblk->ppaf_bitsize < 32) {
944 u32 *map = (u32 *)pblk->trans_map;
946 map[lba] = pblk_ppa64_to_ppa32(pblk, ppa);
948 u64 *map = (u64 *)pblk->trans_map;
954 static inline u64 pblk_dev_ppa_to_line_addr(struct pblk *pblk,
960 paddr |= (u64)p.g.pg << pblk->ppaf.pg_offset;
961 paddr |= (u64)p.g.lun << pblk->ppaf.lun_offset;
962 paddr |= (u64)p.g.ch << pblk->ppaf.ch_offset;
963 paddr |= (u64)p.g.pl << pblk->ppaf.pln_offset;
964 paddr |= (u64)p.g.sec << pblk->ppaf.sec_offset;
969 static inline int pblk_ppa_empty(struct ppa_addr ppa_addr)
971 return (ppa_addr.ppa == ADDR_EMPTY);
974 static inline void pblk_ppa_set_empty(struct ppa_addr *ppa_addr)
976 ppa_addr->ppa = ADDR_EMPTY;
979 static inline int pblk_addr_in_cache(struct ppa_addr ppa)
981 return (ppa.ppa != ADDR_EMPTY && ppa.c.is_cached);
984 static inline int pblk_addr_to_cacheline(struct ppa_addr ppa)
989 static inline struct ppa_addr pblk_cacheline_to_addr(int addr)
999 static inline struct ppa_addr addr_to_gen_ppa(struct pblk *pblk, u64 paddr,
1002 struct ppa_addr ppa;
1005 ppa.g.blk = line_id;
1006 ppa.g.pg = (paddr & pblk->ppaf.pg_mask) >> pblk->ppaf.pg_offset;
1007 ppa.g.lun = (paddr & pblk->ppaf.lun_mask) >> pblk->ppaf.lun_offset;
1008 ppa.g.ch = (paddr & pblk->ppaf.ch_mask) >> pblk->ppaf.ch_offset;
1009 ppa.g.pl = (paddr & pblk->ppaf.pln_mask) >> pblk->ppaf.pln_offset;
1010 ppa.g.sec = (paddr & pblk->ppaf.sec_mask) >> pblk->ppaf.sec_offset;
1015 static inline struct ppa_addr addr_to_pblk_ppa(struct pblk *pblk, u64 paddr,
1018 struct ppa_addr ppa;
1020 ppa = addr_to_gen_ppa(pblk, paddr, line_id);
1025 static inline u32 pblk_calc_meta_header_crc(struct pblk *pblk,
1026 struct line_header *header)
1030 crc = crc32_le(crc, (unsigned char *)header + sizeof(crc),
1031 sizeof(struct line_header) - sizeof(crc));
1036 static inline u32 pblk_calc_smeta_crc(struct pblk *pblk,
1037 struct line_smeta *smeta)
1039 struct pblk_line_meta *lm = &pblk->lm;
1042 crc = crc32_le(crc, (unsigned char *)smeta +
1043 sizeof(struct line_header) + sizeof(crc),
1045 sizeof(struct line_header) - sizeof(crc));
1050 static inline u32 pblk_calc_emeta_crc(struct pblk *pblk,
1051 struct line_emeta *emeta)
1053 struct pblk_line_meta *lm = &pblk->lm;
1056 crc = crc32_le(crc, (unsigned char *)emeta +
1057 sizeof(struct line_header) + sizeof(crc),
1059 sizeof(struct line_header) - sizeof(crc));
1064 static inline int pblk_set_progr_mode(struct pblk *pblk, int type)
1066 struct nvm_tgt_dev *dev = pblk->dev;
1067 struct nvm_geo *geo = &dev->geo;
1070 flags = geo->plane_mode >> 1;
1073 flags |= NVM_IO_SCRAMBLE_ENABLE;
1079 PBLK_READ_RANDOM = 0,
1080 PBLK_READ_SEQUENTIAL = 1,
1083 static inline int pblk_set_read_mode(struct pblk *pblk, int type)
1085 struct nvm_tgt_dev *dev = pblk->dev;
1086 struct nvm_geo *geo = &dev->geo;
1089 flags = NVM_IO_SUSPEND | NVM_IO_SCRAMBLE_ENABLE;
1090 if (type == PBLK_READ_SEQUENTIAL)
1091 flags |= geo->plane_mode >> 1;
1096 static inline int pblk_io_aligned(struct pblk *pblk, int nr_secs)
1098 return !(nr_secs % pblk->min_write_pgs);
1101 #ifdef CONFIG_NVM_DEBUG
1102 static inline void print_ppa(struct ppa_addr *p, char *msg, int error)
1104 if (p->c.is_cached) {
1105 pr_err("ppa: (%s: %x) cache line: %llu\n",
1106 msg, error, (u64)p->c.line);
1108 pr_err("ppa: (%s: %x):ch:%d,lun:%d,blk:%d,pg:%d,pl:%d,sec:%d\n",
1110 p->g.ch, p->g.lun, p->g.blk,
1111 p->g.pg, p->g.pl, p->g.sec);
1115 static inline void pblk_print_failed_rqd(struct pblk *pblk, struct nvm_rq *rqd,
1120 if (rqd->nr_ppas == 1) {
1121 print_ppa(&rqd->ppa_addr, "rqd", error);
1125 while ((bit = find_next_bit((void *)&rqd->ppa_status, rqd->nr_ppas,
1126 bit + 1)) < rqd->nr_ppas) {
1127 print_ppa(&rqd->ppa_list[bit], "rqd", error);
1130 pr_err("error:%d, ppa_status:%llx\n", error, rqd->ppa_status);
1134 static inline int pblk_boundary_ppa_checks(struct nvm_tgt_dev *tgt_dev,
1135 struct ppa_addr *ppas, int nr_ppas)
1137 struct nvm_geo *geo = &tgt_dev->geo;
1138 struct ppa_addr *ppa;
1141 for (i = 0; i < nr_ppas; i++) {
1144 if (!ppa->c.is_cached &&
1145 ppa->g.ch < geo->nr_chnls &&
1146 ppa->g.lun < geo->luns_per_chnl &&
1147 ppa->g.pl < geo->nr_planes &&
1148 ppa->g.blk < geo->blks_per_lun &&
1149 ppa->g.pg < geo->pgs_per_blk &&
1150 ppa->g.sec < geo->sec_per_pg)
1153 #ifdef CONFIG_NVM_DEBUG
1154 print_ppa(ppa, "boundary", i);
1161 static inline int pblk_boundary_paddr_checks(struct pblk *pblk, u64 paddr)
1163 struct pblk_line_meta *lm = &pblk->lm;
1165 if (paddr > lm->sec_per_line)
1171 static inline unsigned int pblk_get_bi_idx(struct bio *bio)
1173 return bio->bi_iter.bi_idx;
1176 static inline sector_t pblk_get_lba(struct bio *bio)
1178 return bio->bi_iter.bi_sector / NR_PHY_IN_LOG;
1181 static inline unsigned int pblk_get_secs(struct bio *bio)
1183 return bio->bi_iter.bi_size / PBLK_EXPOSED_PAGE_SIZE;
1186 static inline sector_t pblk_get_sector(sector_t lba)
1188 return lba * NR_PHY_IN_LOG;
1191 static inline void pblk_setup_uuid(struct pblk *pblk)
1196 memcpy(pblk->instance_uuid, uuid.b, 16);
1198 #endif /* PBLK_H_ */