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 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
;
106 struct bio
*orig_bio
;
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_NR_LUN_BITMAP 2
262 #define PBLK_LINE_NR_SEC_BITMAP 2
263 #define PBLK_LINE_EMPTY (~0U)
267 PBLK_LINETYPE_FREE
= 0,
268 PBLK_LINETYPE_LOG
= 1,
269 PBLK_LINETYPE_DATA
= 2,
272 PBLK_LINESTATE_FREE
= 10,
273 PBLK_LINESTATE_OPEN
= 11,
274 PBLK_LINESTATE_CLOSED
= 12,
275 PBLK_LINESTATE_GC
= 13,
276 PBLK_LINESTATE_BAD
= 14,
277 PBLK_LINESTATE_CORRUPT
= 15,
280 PBLK_LINEGC_NONE
= 20,
281 PBLK_LINEGC_EMPTY
= 21,
282 PBLK_LINEGC_LOW
= 22,
283 PBLK_LINEGC_MID
= 23,
284 PBLK_LINEGC_HIGH
= 24,
285 PBLK_LINEGC_FULL
= 25,
288 #define PBLK_MAGIC 0x70626c6b /*pblk*/
292 __le32 identifier
; /* pblk identifier */
293 __u8 uuid
[16]; /* instance uuid */
294 __le16 type
; /* line type */
295 __le16 version
; /* type version */
296 __le32 id
; /* line id for current line */
300 struct line_header header
;
302 __le32 crc
; /* Full structure including struct crc */
303 /* Previous line metadata */
304 __le32 prev_id
; /* Line id for previous line */
306 /* Current line metadata */
307 __le64 seq_nr
; /* Sequence number for current line */
310 __le32 window_wr_lun
; /* Number of parallel LUNs to write */
317 * 1. struct pblk_emeta
318 * 2. nr_lbas u64 forming lba list
319 * 3. nr_lines (all) u32 valid sector count (vsc) (~0U: non-alloc line)
320 * 4. nr_luns bits (u64 format) forming line bad block bitmap
322 * 3. and 4. will be part of FTL log
325 struct line_header header
;
327 __le32 crc
; /* Full structure including struct crc */
329 /* Previous line metadata */
330 __le32 prev_id
; /* Line id for prev line */
332 /* Current line metadata */
333 __le64 seq_nr
; /* Sequence number for current line */
336 __le32 window_wr_lun
; /* Number of parallel LUNs to write */
338 /* Bookkeeping for recovery */
339 __le32 next_id
; /* Line id for next line */
340 __le64 nr_lbas
; /* Number of lbas mapped in line */
341 __le64 nr_valid_lbas
; /* Number of valid lbas mapped in line */
346 unsigned int id
; /* Line number corresponds to the
349 unsigned int seq_nr
; /* Unique line sequence number */
351 int state
; /* PBLK_LINESTATE_X */
352 int type
; /* PBLK_LINETYPE_X */
353 int gc_group
; /* PBLK_LINEGC_X */
354 struct list_head list
; /* Free, GC lists */
356 unsigned long *lun_bitmap
; /* Bitmap for LUNs mapped in line */
358 struct line_smeta
*smeta
; /* Start metadata */
359 struct line_emeta
*emeta
; /* End metadata */
360 int meta_line
; /* Metadata line id */
361 u64 smeta_ssec
; /* Sector where smeta starts */
362 u64 emeta_ssec
; /* Sector where emeta starts */
364 unsigned int sec_in_line
; /* Number of usable secs in line */
366 unsigned int blk_in_line
; /* Number of good blocks in line */
367 unsigned long *blk_bitmap
; /* Bitmap for valid/invalid blocks */
368 unsigned long *erase_bitmap
; /* Bitmap for erased blocks */
370 unsigned long *map_bitmap
; /* Bitmap for mapped sectors in line */
371 unsigned long *invalid_bitmap
; /* Bitmap for invalid sectors in line */
373 int left_eblks
; /* Blocks left for erasing */
374 atomic_t left_seblks
; /* Blocks left for sync erasing */
376 int left_msecs
; /* Sectors left for mapping */
377 int left_ssecs
; /* Sectors left to sync */
378 unsigned int cur_sec
; /* Sector map pointer */
379 unsigned int vsc
; /* Valid sector count in line */
381 struct kref ref
; /* Write buffer L2P references */
383 spinlock_t lock
; /* Necessary for invalid_bitmap only */
386 #define PBLK_DATA_LINES 2
389 PBLK_KMALLOC_META
= 1,
390 PBLK_VMALLOC_META
= 2,
393 struct pblk_line_metadata
{
397 struct pblk_line_mgmt
{
398 int nr_lines
; /* Total number of full lines */
399 int nr_free_lines
; /* Number of full lines in free list */
401 /* Free lists - use free_lock */
402 struct list_head free_list
; /* Full lines ready to use */
403 struct list_head corrupt_list
; /* Full lines corrupted */
404 struct list_head bad_list
; /* Full lines bad */
406 /* GC lists - use gc_lock */
407 struct list_head
*gc_lists
[PBLK_NR_GC_LISTS
];
408 struct list_head gc_high_list
; /* Full lines ready to GC, high isc */
409 struct list_head gc_mid_list
; /* Full lines ready to GC, mid isc */
410 struct list_head gc_low_list
; /* Full lines ready to GC, low isc */
412 struct list_head gc_full_list
; /* Full lines ready to GC, no valid */
413 struct list_head gc_empty_list
; /* Full lines close, all valid */
415 struct pblk_line
*log_line
; /* Current FTL log line */
416 struct pblk_line
*data_line
; /* Current data line */
417 struct pblk_line
*log_next
; /* Next FTL log line */
418 struct pblk_line
*data_next
; /* Next data line */
420 /* Metadata allocation type: VMALLOC | KMALLOC */
421 int smeta_alloc_type
;
422 int emeta_alloc_type
;
424 /* Pre-allocated metadata for data lines */
425 struct pblk_line_metadata sline_meta
[PBLK_DATA_LINES
];
426 struct pblk_line_metadata eline_meta
[PBLK_DATA_LINES
];
427 unsigned long meta_bitmap
;
429 /* Helpers for fast bitmap calculations */
430 unsigned long *bb_template
;
431 unsigned long *bb_aux
;
433 unsigned long d_seq_nr
; /* Data line unique sequence number */
434 unsigned long l_seq_nr
; /* Log line unique sequence number */
436 spinlock_t free_lock
;
440 struct pblk_line_meta
{
441 unsigned int smeta_len
; /* Total length for smeta */
442 unsigned int smeta_sec
; /* Sectors needed for smeta*/
443 unsigned int emeta_len
; /* Total length for emeta */
444 unsigned int emeta_sec
; /* Sectors needed for emeta*/
445 unsigned int emeta_bb
; /* Boundary for bb that affects emeta */
446 unsigned int sec_bitmap_len
; /* Length for sector bitmap in line */
447 unsigned int blk_bitmap_len
; /* Length for block bitmap in line */
448 unsigned int lun_bitmap_len
; /* Length for lun bitmap in line */
450 unsigned int blk_per_line
; /* Number of blocks in a full line */
451 unsigned int sec_per_line
; /* Number of sectors in a line */
452 unsigned int min_blk_line
; /* Min. number of good blocks in line */
454 unsigned int mid_thrs
; /* Threshold for GC mid list */
455 unsigned int high_thrs
; /* Threshold for GC high list */
458 struct pblk_addr_format
{
474 struct nvm_tgt_dev
*dev
;
475 struct gendisk
*disk
;
479 struct pblk_lun
*luns
;
481 struct pblk_line
*lines
; /* Line array */
482 struct pblk_line_mgmt l_mg
; /* Line management */
483 struct pblk_line_meta lm
; /* Line metadata */
486 struct pblk_addr_format ppaf
;
490 int min_write_pgs
; /* Minimum amount of pages required by controller */
491 int max_write_pgs
; /* Maximum amount of pages supported by controller */
492 int pgs_in_buffer
; /* Number of pages that need to be held in buffer to
493 * guarantee successful reads.
496 sector_t capacity
; /* Device capacity when bad blocks are subtracted */
497 int over_pct
; /* Percentage of device used for over-provisioning */
499 /* pblk provisioning values. Used by rate limiter */
502 struct semaphore erase_sem
;
504 unsigned char instance_uuid
[16];
505 #ifdef CONFIG_NVM_DEBUG
506 /* All debug counters apply to 4kb sector I/Os */
507 atomic_long_t inflight_writes
; /* Inflight writes (user and gc) */
508 atomic_long_t padded_writes
; /* Sectors padded due to flush/fua */
509 atomic_long_t padded_wb
; /* Sectors padded in write buffer */
510 atomic_long_t nr_flush
; /* Number of flush/fua I/O */
511 atomic_long_t req_writes
; /* Sectors stored on write buffer */
512 atomic_long_t sub_writes
; /* Sectors submitted from buffer */
513 atomic_long_t sync_writes
; /* Sectors synced to media */
514 atomic_long_t compl_writes
; /* Sectors completed in write bio */
515 atomic_long_t inflight_reads
; /* Inflight sector read requests */
516 atomic_long_t sync_reads
; /* Completed sector read requests */
517 atomic_long_t recov_writes
; /* Sectors submitted from recovery */
518 atomic_long_t recov_gc_writes
; /* Sectors submitted from write GC */
519 atomic_long_t recov_gc_reads
; /* Sectors submitted from read GC */
524 atomic_long_t read_failed
;
525 atomic_long_t read_empty
;
526 atomic_long_t read_high_ecc
;
527 atomic_long_t read_failed_gc
;
528 atomic_long_t write_failed
;
529 atomic_long_t erase_failed
;
531 struct task_struct
*writer_ts
;
533 /* Simple translation map of logical addresses to physical addresses.
534 * The logical addresses is known by the host system, while the physical
535 * addresses are used when writing to the disk block device.
537 unsigned char *trans_map
;
538 spinlock_t trans_lock
;
540 struct list_head compl_list
;
542 mempool_t
*page_pool
;
543 mempool_t
*line_ws_pool
;
545 mempool_t
*r_rq_pool
;
546 mempool_t
*w_rq_pool
;
547 mempool_t
*line_meta_pool
;
549 struct workqueue_struct
*kw_wq
;
550 struct timer_list wtimer
;
555 struct pblk_line_ws
{
557 struct pblk_line
*line
;
559 struct work_struct ws
;
562 #define pblk_r_rq_size (sizeof(struct nvm_rq) + sizeof(struct pblk_r_ctx))
563 #define pblk_w_rq_size (sizeof(struct nvm_rq) + sizeof(struct pblk_c_ctx))
566 * pblk ring buffer operations
568 int pblk_rb_init(struct pblk_rb
*rb
, struct pblk_rb_entry
*rb_entry_base
,
569 unsigned int power_size
, unsigned int power_seg_sz
);
570 unsigned int pblk_rb_calculate_size(unsigned int nr_entries
);
571 void *pblk_rb_entries_ref(struct pblk_rb
*rb
);
572 int pblk_rb_may_write_user(struct pblk_rb
*rb
, struct bio
*bio
,
573 unsigned int nr_entries
, unsigned int *pos
);
574 int pblk_rb_may_write_gc(struct pblk_rb
*rb
, unsigned int nr_entries
,
576 void pblk_rb_write_entry_user(struct pblk_rb
*rb
, void *data
,
577 struct pblk_w_ctx w_ctx
, unsigned int pos
);
578 void pblk_rb_write_entry_gc(struct pblk_rb
*rb
, void *data
,
579 struct pblk_w_ctx w_ctx
, struct pblk_line
*gc_line
,
581 struct pblk_w_ctx
*pblk_rb_w_ctx(struct pblk_rb
*rb
, unsigned int pos
);
583 void pblk_rb_sync_l2p(struct pblk_rb
*rb
);
584 unsigned int pblk_rb_read_to_bio(struct pblk_rb
*rb
, struct bio
*bio
,
585 struct pblk_c_ctx
*c_ctx
,
587 unsigned int nr_entries
,
589 unsigned int pblk_rb_read_to_bio_list(struct pblk_rb
*rb
, struct bio
*bio
,
590 struct list_head
*list
,
592 int pblk_rb_copy_to_bio(struct pblk_rb
*rb
, struct bio
*bio
, sector_t lba
,
593 u64 pos
, int bio_iter
);
594 unsigned int pblk_rb_read_commit(struct pblk_rb
*rb
, unsigned int entries
);
596 unsigned int pblk_rb_sync_init(struct pblk_rb
*rb
, unsigned long *flags
);
597 unsigned int pblk_rb_sync_advance(struct pblk_rb
*rb
, unsigned int nr_entries
);
598 struct pblk_rb_entry
*pblk_rb_sync_scan_entry(struct pblk_rb
*rb
,
599 struct ppa_addr
*ppa
);
600 void pblk_rb_sync_end(struct pblk_rb
*rb
, unsigned long *flags
);
601 unsigned int pblk_rb_sync_point_count(struct pblk_rb
*rb
);
603 unsigned int pblk_rb_read_count(struct pblk_rb
*rb
);
604 unsigned int pblk_rb_wrap_pos(struct pblk_rb
*rb
, unsigned int pos
);
606 int pblk_rb_tear_down_check(struct pblk_rb
*rb
);
607 int pblk_rb_pos_oob(struct pblk_rb
*rb
, u64 pos
);
608 void pblk_rb_data_free(struct pblk_rb
*rb
);
609 ssize_t
pblk_rb_sysfs(struct pblk_rb
*rb
, char *buf
);
614 struct nvm_rq
*pblk_alloc_rqd(struct pblk
*pblk
, int rw
);
615 int pblk_setup_w_rec_rq(struct pblk
*pblk
, struct nvm_rq
*rqd
,
616 struct pblk_c_ctx
*c_ctx
);
617 void pblk_free_rqd(struct pblk
*pblk
, struct nvm_rq
*rqd
, int rw
);
618 void pblk_flush_writer(struct pblk
*pblk
);
619 struct ppa_addr
pblk_get_lba_map(struct pblk
*pblk
, sector_t lba
);
620 void pblk_discard(struct pblk
*pblk
, struct bio
*bio
);
621 void pblk_log_write_err(struct pblk
*pblk
, struct nvm_rq
*rqd
);
622 void pblk_log_read_err(struct pblk
*pblk
, struct nvm_rq
*rqd
);
623 int pblk_submit_io(struct pblk
*pblk
, struct nvm_rq
*rqd
);
624 struct bio
*pblk_bio_map_addr(struct pblk
*pblk
, void *data
,
625 unsigned int nr_secs
, unsigned int len
,
627 struct pblk_line
*pblk_line_get(struct pblk
*pblk
);
628 struct pblk_line
*pblk_line_get_first_data(struct pblk
*pblk
);
629 struct pblk_line
*pblk_line_replace_data(struct pblk
*pblk
);
630 int pblk_line_recov_alloc(struct pblk
*pblk
, struct pblk_line
*line
);
631 void pblk_line_recov_close(struct pblk
*pblk
, struct pblk_line
*line
);
632 struct pblk_line
*pblk_line_get_data(struct pblk
*pblk
);
633 struct pblk_line
*pblk_line_get_data_next(struct pblk
*pblk
);
634 int pblk_line_erase(struct pblk
*pblk
, struct pblk_line
*line
);
635 int pblk_line_is_full(struct pblk_line
*line
);
636 void pblk_line_free(struct pblk
*pblk
, struct pblk_line
*line
);
637 void pblk_line_close_ws(struct work_struct
*work
);
638 void pblk_line_close(struct pblk
*pblk
, struct pblk_line
*line
);
639 void pblk_line_mark_bb(struct work_struct
*work
);
640 void pblk_line_run_ws(struct pblk
*pblk
, struct pblk_line
*line
, void *priv
,
641 void (*work
)(struct work_struct
*));
642 u64
pblk_line_smeta_start(struct pblk
*pblk
, struct pblk_line
*line
);
643 int pblk_line_read_smeta(struct pblk
*pblk
, struct pblk_line
*line
);
644 int pblk_line_read_emeta(struct pblk
*pblk
, struct pblk_line
*line
);
645 int pblk_blk_erase_async(struct pblk
*pblk
, struct ppa_addr erase_ppa
);
646 void pblk_line_put(struct kref
*ref
);
647 struct list_head
*pblk_line_gc_list(struct pblk
*pblk
, struct pblk_line
*line
);
648 u64
pblk_alloc_page(struct pblk
*pblk
, struct pblk_line
*line
, int nr_secs
);
649 int pblk_calc_secs(struct pblk
*pblk
, unsigned long secs_avail
,
650 unsigned long secs_to_flush
);
651 void pblk_down_rq(struct pblk
*pblk
, struct ppa_addr
*ppa_list
, int nr_ppas
,
652 unsigned long *lun_bitmap
);
653 void pblk_up_rq(struct pblk
*pblk
, struct ppa_addr
*ppa_list
, int nr_ppas
,
654 unsigned long *lun_bitmap
);
655 void pblk_end_bio_sync(struct bio
*bio
);
656 void pblk_end_io_sync(struct nvm_rq
*rqd
);
657 int pblk_bio_add_pages(struct pblk
*pblk
, struct bio
*bio
, gfp_t flags
,
659 void pblk_map_pad_invalidate(struct pblk
*pblk
, struct pblk_line
*line
,
661 void pblk_bio_free_pages(struct pblk
*pblk
, struct bio
*bio
, int off
,
663 void pblk_map_invalidate(struct pblk
*pblk
, struct ppa_addr ppa
);
664 void pblk_update_map(struct pblk
*pblk
, sector_t lba
, struct ppa_addr ppa
);
665 void pblk_update_map_cache(struct pblk
*pblk
, sector_t lba
,
666 struct ppa_addr ppa
);
667 void pblk_update_map_dev(struct pblk
*pblk
, sector_t lba
,
668 struct ppa_addr ppa
, struct ppa_addr entry_line
);
669 int pblk_update_map_gc(struct pblk
*pblk
, sector_t lba
, struct ppa_addr ppa
,
670 struct pblk_line
*gc_line
);
671 void pblk_lookup_l2p_rand(struct pblk
*pblk
, struct ppa_addr
*ppas
,
672 u64
*lba_list
, int nr_secs
);
673 void pblk_lookup_l2p_seq(struct pblk
*pblk
, struct ppa_addr
*ppas
,
674 sector_t blba
, int nr_secs
);
677 * pblk user I/O write path
679 int pblk_write_to_cache(struct pblk
*pblk
, struct bio
*bio
,
680 unsigned long flags
);
681 int pblk_write_gc_to_cache(struct pblk
*pblk
, void *data
, u64
*lba_list
,
682 unsigned int nr_entries
, unsigned int nr_rec_entries
,
683 struct pblk_line
*gc_line
, unsigned long flags
);
688 void pblk_map_erase_rq(struct pblk
*pblk
, struct nvm_rq
*rqd
,
689 unsigned int sentry
, unsigned long *lun_bitmap
,
690 unsigned int valid_secs
, struct ppa_addr
*erase_ppa
);
691 void pblk_map_rq(struct pblk
*pblk
, struct nvm_rq
*rqd
, unsigned int sentry
,
692 unsigned long *lun_bitmap
, unsigned int valid_secs
,
698 int pblk_write_ts(void *data
);
699 void pblk_write_timer_fn(unsigned long data
);
700 void pblk_write_should_kick(struct pblk
*pblk
);
705 int pblk_submit_read(struct pblk
*pblk
, struct bio
*bio
);
706 int pblk_submit_read_gc(struct pblk
*pblk
, u64
*lba_list
, void *data
,
707 unsigned int nr_secs
, unsigned int *secs_to_gc
,
708 struct pblk_line
*line
);
712 void pblk_submit_rec(struct work_struct
*work
);
713 struct pblk_line
*pblk_recov_l2p(struct pblk
*pblk
);
714 void pblk_recov_pad(struct pblk
*pblk
);
715 __le64
*pblk_recov_get_lba_list(struct pblk
*pblk
, struct line_emeta
*emeta
);
716 int pblk_recov_setup_rq(struct pblk
*pblk
, struct pblk_c_ctx
*c_ctx
,
717 struct pblk_rec_ctx
*recovery
, u64
*comp_bits
,
723 #define PBLK_GC_TRIES 3
725 int pblk_gc_init(struct pblk
*pblk
);
726 void pblk_gc_exit(struct pblk
*pblk
);
727 void pblk_gc_should_start(struct pblk
*pblk
);
728 void pblk_gc_should_stop(struct pblk
*pblk
);
729 int pblk_gc_status(struct pblk
*pblk
);
730 void pblk_gc_sysfs_state_show(struct pblk
*pblk
, int *gc_enabled
,
732 void pblk_gc_sysfs_force(struct pblk
*pblk
, int force
);
737 void pblk_rl_init(struct pblk_rl
*rl
, int budget
);
738 void pblk_rl_free(struct pblk_rl
*rl
);
739 int pblk_rl_gc_thrs(struct pblk_rl
*rl
);
740 unsigned long pblk_rl_nr_free_blks(struct pblk_rl
*rl
);
741 int pblk_rl_user_may_insert(struct pblk_rl
*rl
, int nr_entries
);
742 void pblk_rl_user_in(struct pblk_rl
*rl
, int nr_entries
);
743 int pblk_rl_gc_may_insert(struct pblk_rl
*rl
, int nr_entries
);
744 void pblk_rl_gc_in(struct pblk_rl
*rl
, int nr_entries
);
745 void pblk_rl_out(struct pblk_rl
*rl
, int nr_user
, int nr_gc
);
746 void pblk_rl_set_gc_rsc(struct pblk_rl
*rl
, int rsv
);
747 int pblk_rl_sysfs_rate_show(struct pblk_rl
*rl
);
748 void pblk_rl_free_lines_inc(struct pblk_rl
*rl
, struct pblk_line
*line
);
749 void pblk_rl_free_lines_dec(struct pblk_rl
*rl
, struct pblk_line
*line
);
754 int pblk_sysfs_init(struct gendisk
*tdisk
);
755 void pblk_sysfs_exit(struct gendisk
*tdisk
);
757 static inline void *pblk_malloc(size_t size
, int type
, gfp_t flags
)
759 if (type
== PBLK_KMALLOC_META
)
760 return kmalloc(size
, flags
);
761 return vmalloc(size
);
764 static inline void pblk_mfree(void *ptr
, int type
)
766 if (type
== PBLK_KMALLOC_META
)
772 static inline struct nvm_rq
*nvm_rq_from_c_ctx(void *c_ctx
)
774 return c_ctx
- sizeof(struct nvm_rq
);
777 static inline void *pblk_line_emeta_to_lbas(struct line_emeta
*emeta
)
782 #define NVM_MEM_PAGE_WRITE (8)
784 static inline int pblk_pad_distance(struct pblk
*pblk
)
786 struct nvm_tgt_dev
*dev
= pblk
->dev
;
787 struct nvm_geo
*geo
= &dev
->geo
;
789 return NVM_MEM_PAGE_WRITE
* geo
->nr_luns
* geo
->sec_per_pl
;
792 static inline int pblk_dev_ppa_to_line(struct ppa_addr p
)
797 static inline int pblk_tgt_ppa_to_line(struct ppa_addr p
)
802 static inline int pblk_ppa_to_pos(struct nvm_geo
*geo
, struct ppa_addr p
)
804 return p
.g
.lun
* geo
->nr_chnls
+ p
.g
.ch
;
807 /* A block within a line corresponds to the lun */
808 static inline int pblk_dev_ppa_to_pos(struct nvm_geo
*geo
, struct ppa_addr p
)
810 return p
.g
.lun
* geo
->nr_chnls
+ p
.g
.ch
;
813 static inline struct ppa_addr
pblk_ppa32_to_ppa64(struct pblk
*pblk
, u32 ppa32
)
815 struct ppa_addr ppa64
;
820 ppa64
.ppa
= ADDR_EMPTY
;
821 } else if (ppa32
& (1U << 31)) {
822 ppa64
.c
.line
= ppa32
& ((~0U) >> 1);
823 ppa64
.c
.is_cached
= 1;
825 ppa64
.g
.blk
= (ppa32
& pblk
->ppaf
.blk_mask
) >>
826 pblk
->ppaf
.blk_offset
;
827 ppa64
.g
.pg
= (ppa32
& pblk
->ppaf
.pg_mask
) >>
828 pblk
->ppaf
.pg_offset
;
829 ppa64
.g
.lun
= (ppa32
& pblk
->ppaf
.lun_mask
) >>
830 pblk
->ppaf
.lun_offset
;
831 ppa64
.g
.ch
= (ppa32
& pblk
->ppaf
.ch_mask
) >>
832 pblk
->ppaf
.ch_offset
;
833 ppa64
.g
.pl
= (ppa32
& pblk
->ppaf
.pln_mask
) >>
834 pblk
->ppaf
.pln_offset
;
835 ppa64
.g
.sec
= (ppa32
& pblk
->ppaf
.sec_mask
) >>
836 pblk
->ppaf
.sec_offset
;
842 static inline struct ppa_addr
pblk_trans_map_get(struct pblk
*pblk
,
847 if (pblk
->ppaf_bitsize
< 32) {
848 u32
*map
= (u32
*)pblk
->trans_map
;
850 ppa
= pblk_ppa32_to_ppa64(pblk
, map
[lba
]);
852 struct ppa_addr
*map
= (struct ppa_addr
*)pblk
->trans_map
;
860 static inline u32
pblk_ppa64_to_ppa32(struct pblk
*pblk
, struct ppa_addr ppa64
)
864 if (ppa64
.ppa
== ADDR_EMPTY
) {
866 } else if (ppa64
.c
.is_cached
) {
867 ppa32
|= ppa64
.c
.line
;
870 ppa32
|= ppa64
.g
.blk
<< pblk
->ppaf
.blk_offset
;
871 ppa32
|= ppa64
.g
.pg
<< pblk
->ppaf
.pg_offset
;
872 ppa32
|= ppa64
.g
.lun
<< pblk
->ppaf
.lun_offset
;
873 ppa32
|= ppa64
.g
.ch
<< pblk
->ppaf
.ch_offset
;
874 ppa32
|= ppa64
.g
.pl
<< pblk
->ppaf
.pln_offset
;
875 ppa32
|= ppa64
.g
.sec
<< pblk
->ppaf
.sec_offset
;
881 static inline void pblk_trans_map_set(struct pblk
*pblk
, sector_t lba
,
884 if (pblk
->ppaf_bitsize
< 32) {
885 u32
*map
= (u32
*)pblk
->trans_map
;
887 map
[lba
] = pblk_ppa64_to_ppa32(pblk
, ppa
);
889 u64
*map
= (u64
*)pblk
->trans_map
;
895 static inline u64
pblk_dev_ppa_to_line_addr(struct pblk
*pblk
,
901 paddr
|= (u64
)p
.g
.pg
<< pblk
->ppaf
.pg_offset
;
902 paddr
|= (u64
)p
.g
.lun
<< pblk
->ppaf
.lun_offset
;
903 paddr
|= (u64
)p
.g
.ch
<< pblk
->ppaf
.ch_offset
;
904 paddr
|= (u64
)p
.g
.pl
<< pblk
->ppaf
.pln_offset
;
905 paddr
|= (u64
)p
.g
.sec
<< pblk
->ppaf
.sec_offset
;
910 static inline int pblk_ppa_empty(struct ppa_addr ppa_addr
)
912 return (ppa_addr
.ppa
== ADDR_EMPTY
);
915 static inline void pblk_ppa_set_empty(struct ppa_addr
*ppa_addr
)
917 ppa_addr
->ppa
= ADDR_EMPTY
;
920 static inline int pblk_addr_in_cache(struct ppa_addr ppa
)
922 return (ppa
.ppa
!= ADDR_EMPTY
&& ppa
.c
.is_cached
);
925 static inline int pblk_addr_to_cacheline(struct ppa_addr ppa
)
930 static inline struct ppa_addr
pblk_cacheline_to_addr(int addr
)
940 static inline struct ppa_addr
addr_to_gen_ppa(struct pblk
*pblk
, u64 paddr
,
947 ppa
.g
.pg
= (paddr
& pblk
->ppaf
.pg_mask
) >> pblk
->ppaf
.pg_offset
;
948 ppa
.g
.lun
= (paddr
& pblk
->ppaf
.lun_mask
) >> pblk
->ppaf
.lun_offset
;
949 ppa
.g
.ch
= (paddr
& pblk
->ppaf
.ch_mask
) >> pblk
->ppaf
.ch_offset
;
950 ppa
.g
.pl
= (paddr
& pblk
->ppaf
.pln_mask
) >> pblk
->ppaf
.pln_offset
;
951 ppa
.g
.sec
= (paddr
& pblk
->ppaf
.sec_mask
) >> pblk
->ppaf
.sec_offset
;
956 static inline struct ppa_addr
addr_to_pblk_ppa(struct pblk
*pblk
, u64 paddr
,
961 ppa
= addr_to_gen_ppa(pblk
, paddr
, line_id
);
966 static inline u32
pblk_calc_meta_header_crc(struct pblk
*pblk
,
967 struct line_smeta
*smeta
)
971 crc
= crc32_le(crc
, (unsigned char *)smeta
+ sizeof(crc
),
972 sizeof(struct line_header
) - sizeof(crc
));
977 static inline u32
pblk_calc_smeta_crc(struct pblk
*pblk
,
978 struct line_smeta
*smeta
)
980 struct pblk_line_meta
*lm
= &pblk
->lm
;
983 crc
= crc32_le(crc
, (unsigned char *)smeta
+
984 sizeof(struct line_header
) + sizeof(crc
),
986 sizeof(struct line_header
) - sizeof(crc
));
991 static inline u32
pblk_calc_emeta_crc(struct pblk
*pblk
,
992 struct line_emeta
*emeta
)
994 struct pblk_line_meta
*lm
= &pblk
->lm
;
997 crc
= crc32_le(crc
, (unsigned char *)emeta
+
998 sizeof(struct line_header
) + sizeof(crc
),
1000 sizeof(struct line_header
) - sizeof(crc
));
1005 static inline int pblk_set_progr_mode(struct pblk
*pblk
, int type
)
1007 struct nvm_tgt_dev
*dev
= pblk
->dev
;
1008 struct nvm_geo
*geo
= &dev
->geo
;
1011 flags
= geo
->plane_mode
>> 1;
1014 flags
|= NVM_IO_SCRAMBLE_ENABLE
;
1019 static inline int pblk_set_read_mode(struct pblk
*pblk
)
1021 return NVM_IO_SNGL_ACCESS
| NVM_IO_SUSPEND
| NVM_IO_SCRAMBLE_ENABLE
;
1024 #ifdef CONFIG_NVM_DEBUG
1025 static inline void print_ppa(struct ppa_addr
*p
, char *msg
, int error
)
1027 if (p
->c
.is_cached
) {
1028 pr_err("ppa: (%s: %x) cache line: %llu\n",
1029 msg
, error
, (u64
)p
->c
.line
);
1031 pr_err("ppa: (%s: %x):ch:%d,lun:%d,blk:%d,pg:%d,pl:%d,sec:%d\n",
1033 p
->g
.ch
, p
->g
.lun
, p
->g
.blk
,
1034 p
->g
.pg
, p
->g
.pl
, p
->g
.sec
);
1038 static inline void pblk_print_failed_rqd(struct pblk
*pblk
, struct nvm_rq
*rqd
,
1043 if (rqd
->nr_ppas
== 1) {
1044 print_ppa(&rqd
->ppa_addr
, "rqd", error
);
1048 while ((bit
= find_next_bit((void *)&rqd
->ppa_status
, rqd
->nr_ppas
,
1049 bit
+ 1)) < rqd
->nr_ppas
) {
1050 print_ppa(&rqd
->ppa_list
[bit
], "rqd", error
);
1053 pr_err("error:%d, ppa_status:%llx\n", error
, rqd
->ppa_status
);
1057 static inline int pblk_boundary_ppa_checks(struct nvm_tgt_dev
*tgt_dev
,
1058 struct ppa_addr
*ppas
, int nr_ppas
)
1060 struct nvm_geo
*geo
= &tgt_dev
->geo
;
1061 struct ppa_addr
*ppa
;
1064 for (i
= 0; i
< nr_ppas
; i
++) {
1067 if (!ppa
->c
.is_cached
&&
1068 ppa
->g
.ch
< geo
->nr_chnls
&&
1069 ppa
->g
.lun
< geo
->luns_per_chnl
&&
1070 ppa
->g
.pl
< geo
->nr_planes
&&
1071 ppa
->g
.blk
< geo
->blks_per_lun
&&
1072 ppa
->g
.pg
< geo
->pgs_per_blk
&&
1073 ppa
->g
.sec
< geo
->sec_per_pg
)
1076 #ifdef CONFIG_NVM_DEBUG
1077 print_ppa(ppa
, "boundary", i
);
1084 static inline int pblk_boundary_paddr_checks(struct pblk
*pblk
, u64 paddr
)
1086 struct pblk_line_meta
*lm
= &pblk
->lm
;
1088 if (paddr
> lm
->sec_per_line
)
1094 static inline unsigned int pblk_get_bi_idx(struct bio
*bio
)
1096 return bio
->bi_iter
.bi_idx
;
1099 static inline sector_t
pblk_get_lba(struct bio
*bio
)
1101 return bio
->bi_iter
.bi_sector
/ NR_PHY_IN_LOG
;
1104 static inline unsigned int pblk_get_secs(struct bio
*bio
)
1106 return bio
->bi_iter
.bi_size
/ PBLK_EXPOSED_PAGE_SIZE
;
1109 static inline sector_t
pblk_get_sector(sector_t lba
)
1111 return lba
* NR_PHY_IN_LOG
;
1114 static inline void pblk_setup_uuid(struct pblk
*pblk
)
1119 memcpy(pblk
->instance_uuid
, uuid
.b
, 16);
1121 #endif /* PBLK_H_ */