]> git.sesse.net Git - bcachefs-tools-debian/blob - libbcachefs/sysfs.c
Update bcachefs sources to 8c94740b1bf8 bcachefs: Add missing vaidation for jset_entr...
[bcachefs-tools-debian] / libbcachefs / sysfs.c
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * bcache sysfs interfaces
4  *
5  * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
6  * Copyright 2012 Google, Inc.
7  */
8
9 #ifndef NO_BCACHEFS_SYSFS
10
11 #include "bcachefs.h"
12 #include "alloc_background.h"
13 #include "alloc_foreground.h"
14 #include "sysfs.h"
15 #include "btree_cache.h"
16 #include "btree_io.h"
17 #include "btree_iter.h"
18 #include "btree_key_cache.h"
19 #include "btree_update.h"
20 #include "btree_update_interior.h"
21 #include "btree_gc.h"
22 #include "buckets.h"
23 #include "clock.h"
24 #include "disk_groups.h"
25 #include "ec.h"
26 #include "inode.h"
27 #include "journal.h"
28 #include "keylist.h"
29 #include "move.h"
30 #include "movinggc.h"
31 #include "nocow_locking.h"
32 #include "opts.h"
33 #include "rebalance.h"
34 #include "replicas.h"
35 #include "super-io.h"
36 #include "tests.h"
37
38 #include <linux/blkdev.h>
39 #include <linux/sort.h>
40 #include <linux/sched/clock.h>
41
42 #include "util.h"
43
44 #define SYSFS_OPS(type)                                                 \
45 const struct sysfs_ops type ## _sysfs_ops = {                           \
46         .show   = type ## _show,                                        \
47         .store  = type ## _store                                        \
48 }
49
50 #define SHOW(fn)                                                        \
51 static ssize_t fn ## _to_text(struct printbuf *,                        \
52                               struct kobject *, struct attribute *);    \
53                                                                         \
54 static ssize_t fn ## _show(struct kobject *kobj, struct attribute *attr,\
55                            char *buf)                                   \
56 {                                                                       \
57         struct printbuf out = PRINTBUF;                                 \
58         ssize_t ret = fn ## _to_text(&out, kobj, attr);                 \
59                                                                         \
60         if (out.pos && out.buf[out.pos - 1] != '\n')                    \
61                 prt_newline(&out);                                      \
62                                                                         \
63         if (!ret && out.allocation_failure)                             \
64                 ret = -ENOMEM;                                          \
65                                                                         \
66         if (!ret) {                                                     \
67                 ret = min_t(size_t, out.pos, PAGE_SIZE - 1);            \
68                 memcpy(buf, out.buf, ret);                              \
69         }                                                               \
70         printbuf_exit(&out);                                            \
71         return bch2_err_class(ret);                                     \
72 }                                                                       \
73                                                                         \
74 static ssize_t fn ## _to_text(struct printbuf *out, struct kobject *kobj,\
75                               struct attribute *attr)
76
77 #define STORE(fn)                                                       \
78 static ssize_t fn ## _store_inner(struct kobject *, struct attribute *,\
79                             const char *, size_t);                      \
80                                                                         \
81 static ssize_t fn ## _store(struct kobject *kobj, struct attribute *attr,\
82                             const char *buf, size_t size)               \
83 {                                                                       \
84         return bch2_err_class(fn##_store_inner(kobj, attr, buf, size)); \
85 }                                                                       \
86                                                                         \
87 static ssize_t fn ## _store_inner(struct kobject *kobj, struct attribute *attr,\
88                                   const char *buf, size_t size)
89
90 #define __sysfs_attribute(_name, _mode)                                 \
91         static struct attribute sysfs_##_name =                         \
92                 { .name = #_name, .mode = _mode }
93
94 #define write_attribute(n)      __sysfs_attribute(n, 0200)
95 #define read_attribute(n)       __sysfs_attribute(n, 0444)
96 #define rw_attribute(n)         __sysfs_attribute(n, 0644)
97
98 #define sysfs_printf(file, fmt, ...)                                    \
99 do {                                                                    \
100         if (attr == &sysfs_ ## file)                                    \
101                 prt_printf(out, fmt "\n", __VA_ARGS__);                 \
102 } while (0)
103
104 #define sysfs_print(file, var)                                          \
105 do {                                                                    \
106         if (attr == &sysfs_ ## file)                                    \
107                 snprint(out, var);                                      \
108 } while (0)
109
110 #define sysfs_hprint(file, val)                                         \
111 do {                                                                    \
112         if (attr == &sysfs_ ## file)                                    \
113                 prt_human_readable_s64(out, val);                       \
114 } while (0)
115
116 #define sysfs_strtoul(file, var)                                        \
117 do {                                                                    \
118         if (attr == &sysfs_ ## file)                                    \
119                 return strtoul_safe(buf, var) ?: (ssize_t) size;        \
120 } while (0)
121
122 #define sysfs_strtoul_clamp(file, var, min, max)                        \
123 do {                                                                    \
124         if (attr == &sysfs_ ## file)                                    \
125                 return strtoul_safe_clamp(buf, var, min, max)           \
126                         ?: (ssize_t) size;                              \
127 } while (0)
128
129 #define strtoul_or_return(cp)                                           \
130 ({                                                                      \
131         unsigned long _v;                                               \
132         int _r = kstrtoul(cp, 10, &_v);                                 \
133         if (_r)                                                         \
134                 return _r;                                              \
135         _v;                                                             \
136 })
137
138 write_attribute(trigger_gc);
139 write_attribute(trigger_discards);
140 write_attribute(trigger_invalidates);
141 write_attribute(prune_cache);
142 write_attribute(btree_wakeup);
143 rw_attribute(btree_gc_periodic);
144 rw_attribute(gc_gens_pos);
145
146 read_attribute(uuid);
147 read_attribute(minor);
148 read_attribute(bucket_size);
149 read_attribute(first_bucket);
150 read_attribute(nbuckets);
151 rw_attribute(durability);
152 read_attribute(io_done);
153 read_attribute(io_errors);
154 write_attribute(io_errors_reset);
155
156 read_attribute(io_latency_read);
157 read_attribute(io_latency_write);
158 read_attribute(io_latency_stats_read);
159 read_attribute(io_latency_stats_write);
160 read_attribute(congested);
161
162 read_attribute(btree_write_stats);
163
164 read_attribute(btree_cache_size);
165 read_attribute(compression_stats);
166 read_attribute(journal_debug);
167 read_attribute(btree_updates);
168 read_attribute(btree_cache);
169 read_attribute(btree_key_cache);
170 read_attribute(stripes_heap);
171 read_attribute(open_buckets);
172 read_attribute(open_buckets_partial);
173 read_attribute(write_points);
174 read_attribute(nocow_lock_table);
175
176 #ifdef BCH_WRITE_REF_DEBUG
177 read_attribute(write_refs);
178
179 static const char * const bch2_write_refs[] = {
180 #define x(n)    #n,
181         BCH_WRITE_REFS()
182 #undef x
183         NULL
184 };
185
186 static void bch2_write_refs_to_text(struct printbuf *out, struct bch_fs *c)
187 {
188         bch2_printbuf_tabstop_push(out, 24);
189
190         for (unsigned i = 0; i < ARRAY_SIZE(c->writes); i++) {
191                 prt_str(out, bch2_write_refs[i]);
192                 prt_tab(out);
193                 prt_printf(out, "%li", atomic_long_read(&c->writes[i]));
194                 prt_newline(out);
195         }
196 }
197 #endif
198
199 read_attribute(internal_uuid);
200 read_attribute(disk_groups);
201
202 read_attribute(has_data);
203 read_attribute(alloc_debug);
204
205 #define x(t, n, ...) read_attribute(t);
206 BCH_PERSISTENT_COUNTERS()
207 #undef x
208
209 rw_attribute(discard);
210 rw_attribute(label);
211
212 rw_attribute(copy_gc_enabled);
213 read_attribute(copy_gc_wait);
214
215 rw_attribute(rebalance_enabled);
216 sysfs_pd_controller_attribute(rebalance);
217 read_attribute(rebalance_status);
218 rw_attribute(promote_whole_extents);
219
220 read_attribute(new_stripes);
221
222 read_attribute(io_timers_read);
223 read_attribute(io_timers_write);
224
225 read_attribute(moving_ctxts);
226
227 #ifdef CONFIG_BCACHEFS_TESTS
228 write_attribute(perf_test);
229 #endif /* CONFIG_BCACHEFS_TESTS */
230
231 #define x(_name)                                                \
232         static struct attribute sysfs_time_stat_##_name =               \
233                 { .name = #_name, .mode = 0444 };
234         BCH_TIME_STATS()
235 #undef x
236
237 static struct attribute sysfs_state_rw = {
238         .name = "state",
239         .mode =  0444,
240 };
241
242 static size_t bch2_btree_cache_size(struct bch_fs *c)
243 {
244         size_t ret = 0;
245         struct btree *b;
246
247         mutex_lock(&c->btree_cache.lock);
248         list_for_each_entry(b, &c->btree_cache.live, list)
249                 ret += btree_bytes(c);
250
251         mutex_unlock(&c->btree_cache.lock);
252         return ret;
253 }
254
255 static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c)
256 {
257         struct btree_trans *trans;
258         struct btree_iter iter;
259         struct bkey_s_c k;
260         enum btree_id id;
261         struct compression_type_stats {
262                 u64             nr_extents;
263                 u64             sectors_compressed;
264                 u64             sectors_uncompressed;
265         } s[BCH_COMPRESSION_TYPE_NR];
266         u64 compressed_incompressible = 0;
267         int ret = 0;
268
269         memset(s, 0, sizeof(s));
270
271         if (!test_bit(BCH_FS_STARTED, &c->flags))
272                 return -EPERM;
273
274         trans = bch2_trans_get(c);
275
276         for (id = 0; id < BTREE_ID_NR; id++) {
277                 if (!btree_type_has_ptrs(id))
278                         continue;
279
280                 for_each_btree_key(trans, iter, id, POS_MIN,
281                                    BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
282                         struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
283                         struct bch_extent_crc_unpacked crc;
284                         const union bch_extent_entry *entry;
285                         bool compressed = false, incompressible = false;
286
287                         bkey_for_each_crc(k.k, ptrs, crc, entry) {
288                                 incompressible  |= crc.compression_type == BCH_COMPRESSION_TYPE_incompressible;
289                                 compressed      |= crc_is_compressed(crc);
290
291                                 if (crc_is_compressed(crc)) {
292                                         s[crc.compression_type].nr_extents++;
293                                         s[crc.compression_type].sectors_compressed += crc.compressed_size;
294                                         s[crc.compression_type].sectors_uncompressed += crc.uncompressed_size;
295                                 }
296                         }
297
298                         compressed_incompressible += compressed && incompressible;
299
300                         if (!compressed) {
301                                 unsigned t = incompressible ? BCH_COMPRESSION_TYPE_incompressible : 0;
302
303                                 s[t].nr_extents++;
304                                 s[t].sectors_compressed += k.k->size;
305                                 s[t].sectors_uncompressed += k.k->size;
306                         }
307                 }
308                 bch2_trans_iter_exit(trans, &iter);
309         }
310
311         bch2_trans_put(trans);
312
313         if (ret)
314                 return ret;
315
316         prt_str(out, "type");
317         printbuf_tabstop_push(out, 12);
318         prt_tab(out);
319
320         prt_str(out, "compressed");
321         printbuf_tabstop_push(out, 16);
322         prt_tab_rjust(out);
323
324         prt_str(out, "uncompressed");
325         printbuf_tabstop_push(out, 16);
326         prt_tab_rjust(out);
327
328         prt_str(out, "average extent size");
329         printbuf_tabstop_push(out, 24);
330         prt_tab_rjust(out);
331         prt_newline(out);
332
333         for (unsigned i = 0; i < ARRAY_SIZE(s); i++) {
334                 prt_str(out, bch2_compression_types[i]);
335                 prt_tab(out);
336
337                 prt_human_readable_u64(out, s[i].sectors_compressed << 9);
338                 prt_tab_rjust(out);
339
340                 prt_human_readable_u64(out, s[i].sectors_uncompressed << 9);
341                 prt_tab_rjust(out);
342
343                 prt_human_readable_u64(out, s[i].nr_extents
344                                        ? div_u64(s[i].sectors_uncompressed << 9, s[i].nr_extents)
345                                        : 0);
346                 prt_tab_rjust(out);
347                 prt_newline(out);
348         }
349
350         if (compressed_incompressible) {
351                 prt_printf(out, "%llu compressed & incompressible extents", compressed_incompressible);
352                 prt_newline(out);
353         }
354
355         return 0;
356 }
357
358 static void bch2_gc_gens_pos_to_text(struct printbuf *out, struct bch_fs *c)
359 {
360         prt_printf(out, "%s: ", bch2_btree_id_str(c->gc_gens_btree));
361         bch2_bpos_to_text(out, c->gc_gens_pos);
362         prt_printf(out, "\n");
363 }
364
365 static void bch2_btree_wakeup_all(struct bch_fs *c)
366 {
367         struct btree_trans *trans;
368
369         seqmutex_lock(&c->btree_trans_lock);
370         list_for_each_entry(trans, &c->btree_trans_list, list) {
371                 struct btree_bkey_cached_common *b = READ_ONCE(trans->locking);
372
373                 if (b)
374                         six_lock_wakeup_all(&b->lock);
375
376         }
377         seqmutex_unlock(&c->btree_trans_lock);
378 }
379
380 SHOW(bch2_fs)
381 {
382         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
383
384         sysfs_print(minor,                      c->minor);
385         sysfs_printf(internal_uuid, "%pU",      c->sb.uuid.b);
386
387         sysfs_hprint(btree_cache_size,          bch2_btree_cache_size(c));
388
389         if (attr == &sysfs_btree_write_stats)
390                 bch2_btree_write_stats_to_text(out, c);
391
392         sysfs_printf(btree_gc_periodic, "%u",   (int) c->btree_gc_periodic);
393
394         if (attr == &sysfs_gc_gens_pos)
395                 bch2_gc_gens_pos_to_text(out, c);
396
397         sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
398
399         sysfs_printf(rebalance_enabled,         "%i", c->rebalance.enabled);
400         sysfs_pd_controller_show(rebalance,     &c->rebalance.pd); /* XXX */
401
402         if (attr == &sysfs_copy_gc_wait)
403                 bch2_copygc_wait_to_text(out, c);
404
405         if (attr == &sysfs_rebalance_status)
406                 bch2_rebalance_status_to_text(out, c);
407
408         sysfs_print(promote_whole_extents,      c->promote_whole_extents);
409
410         /* Debugging: */
411
412         if (attr == &sysfs_journal_debug)
413                 bch2_journal_debug_to_text(out, &c->journal);
414
415         if (attr == &sysfs_btree_updates)
416                 bch2_btree_updates_to_text(out, c);
417
418         if (attr == &sysfs_btree_cache)
419                 bch2_btree_cache_to_text(out, &c->btree_cache);
420
421         if (attr == &sysfs_btree_key_cache)
422                 bch2_btree_key_cache_to_text(out, &c->btree_key_cache);
423
424         if (attr == &sysfs_stripes_heap)
425                 bch2_stripes_heap_to_text(out, c);
426
427         if (attr == &sysfs_open_buckets)
428                 bch2_open_buckets_to_text(out, c);
429
430         if (attr == &sysfs_open_buckets_partial)
431                 bch2_open_buckets_partial_to_text(out, c);
432
433         if (attr == &sysfs_write_points)
434                 bch2_write_points_to_text(out, c);
435
436         if (attr == &sysfs_compression_stats)
437                 bch2_compression_stats_to_text(out, c);
438
439         if (attr == &sysfs_new_stripes)
440                 bch2_new_stripes_to_text(out, c);
441
442         if (attr == &sysfs_io_timers_read)
443                 bch2_io_timers_to_text(out, &c->io_clock[READ]);
444
445         if (attr == &sysfs_io_timers_write)
446                 bch2_io_timers_to_text(out, &c->io_clock[WRITE]);
447
448         if (attr == &sysfs_moving_ctxts)
449                 bch2_fs_moving_ctxts_to_text(out, c);
450
451 #ifdef BCH_WRITE_REF_DEBUG
452         if (attr == &sysfs_write_refs)
453                 bch2_write_refs_to_text(out, c);
454 #endif
455
456         if (attr == &sysfs_nocow_lock_table)
457                 bch2_nocow_locks_to_text(out, &c->nocow_locks);
458
459         if (attr == &sysfs_disk_groups)
460                 bch2_disk_groups_to_text(out, c);
461
462         return 0;
463 }
464
465 STORE(bch2_fs)
466 {
467         struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
468
469         if (attr == &sysfs_btree_gc_periodic) {
470                 ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
471                         ?: (ssize_t) size;
472
473                 wake_up_process(c->gc_thread);
474                 return ret;
475         }
476
477         if (attr == &sysfs_copy_gc_enabled) {
478                 ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
479                         ?: (ssize_t) size;
480
481                 if (c->copygc_thread)
482                         wake_up_process(c->copygc_thread);
483                 return ret;
484         }
485
486         if (attr == &sysfs_rebalance_enabled) {
487                 ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
488                         ?: (ssize_t) size;
489
490                 rebalance_wakeup(c);
491                 return ret;
492         }
493
494         sysfs_pd_controller_store(rebalance,    &c->rebalance.pd);
495
496         sysfs_strtoul(promote_whole_extents,    c->promote_whole_extents);
497
498         /* Debugging: */
499
500         if (!test_bit(BCH_FS_STARTED, &c->flags))
501                 return -EPERM;
502
503         /* Debugging: */
504
505         if (!test_bit(BCH_FS_RW, &c->flags))
506                 return -EROFS;
507
508         if (attr == &sysfs_prune_cache) {
509                 struct shrink_control sc;
510
511                 sc.gfp_mask = GFP_KERNEL;
512                 sc.nr_to_scan = strtoul_or_return(buf);
513                 c->btree_cache.shrink->scan_objects(c->btree_cache.shrink, &sc);
514         }
515
516         if (attr == &sysfs_btree_wakeup)
517                 bch2_btree_wakeup_all(c);
518
519         if (attr == &sysfs_trigger_gc) {
520                 /*
521                  * Full gc is currently incompatible with btree key cache:
522                  */
523 #if 0
524                 down_read(&c->state_lock);
525                 bch2_gc(c, false, false);
526                 up_read(&c->state_lock);
527 #else
528                 bch2_gc_gens(c);
529 #endif
530         }
531
532         if (attr == &sysfs_trigger_discards)
533                 bch2_do_discards(c);
534
535         if (attr == &sysfs_trigger_invalidates)
536                 bch2_do_invalidates(c);
537
538 #ifdef CONFIG_BCACHEFS_TESTS
539         if (attr == &sysfs_perf_test) {
540                 char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
541                 char *test              = strsep(&p, " \t\n");
542                 char *nr_str            = strsep(&p, " \t\n");
543                 char *threads_str       = strsep(&p, " \t\n");
544                 unsigned threads;
545                 u64 nr;
546                 int ret = -EINVAL;
547
548                 if (threads_str &&
549                     !(ret = kstrtouint(threads_str, 10, &threads)) &&
550                     !(ret = bch2_strtoull_h(nr_str, &nr)))
551                         ret = bch2_btree_perf_test(c, test, nr, threads);
552                 kfree(tmp);
553
554                 if (ret)
555                         size = ret;
556         }
557 #endif
558         return size;
559 }
560 SYSFS_OPS(bch2_fs);
561
562 struct attribute *bch2_fs_files[] = {
563         &sysfs_minor,
564         &sysfs_btree_cache_size,
565         &sysfs_btree_write_stats,
566
567         &sysfs_promote_whole_extents,
568
569         &sysfs_compression_stats,
570
571 #ifdef CONFIG_BCACHEFS_TESTS
572         &sysfs_perf_test,
573 #endif
574         NULL
575 };
576
577 /* counters dir */
578
579 SHOW(bch2_fs_counters)
580 {
581         struct bch_fs *c = container_of(kobj, struct bch_fs, counters_kobj);
582         u64 counter = 0;
583         u64 counter_since_mount = 0;
584
585         printbuf_tabstop_push(out, 32);
586
587         #define x(t, ...) \
588                 if (attr == &sysfs_##t) {                                       \
589                         counter             = percpu_u64_get(&c->counters[BCH_COUNTER_##t]);\
590                         counter_since_mount = counter - c->counters_on_mount[BCH_COUNTER_##t];\
591                         prt_printf(out, "since mount:");                                \
592                         prt_tab(out);                                           \
593                         prt_human_readable_u64(out, counter_since_mount);       \
594                         prt_newline(out);                                       \
595                                                                                 \
596                         prt_printf(out, "since filesystem creation:");          \
597                         prt_tab(out);                                           \
598                         prt_human_readable_u64(out, counter);                   \
599                         prt_newline(out);                                       \
600                 }
601         BCH_PERSISTENT_COUNTERS()
602         #undef x
603         return 0;
604 }
605
606 STORE(bch2_fs_counters) {
607         return 0;
608 }
609
610 SYSFS_OPS(bch2_fs_counters);
611
612 struct attribute *bch2_fs_counters_files[] = {
613 #define x(t, ...) \
614         &sysfs_##t,
615         BCH_PERSISTENT_COUNTERS()
616 #undef x
617         NULL
618 };
619 /* internal dir - just a wrapper */
620
621 SHOW(bch2_fs_internal)
622 {
623         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
624
625         return bch2_fs_to_text(out, &c->kobj, attr);
626 }
627
628 STORE(bch2_fs_internal)
629 {
630         struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
631
632         return bch2_fs_store(&c->kobj, attr, buf, size);
633 }
634 SYSFS_OPS(bch2_fs_internal);
635
636 struct attribute *bch2_fs_internal_files[] = {
637         &sysfs_journal_debug,
638         &sysfs_btree_updates,
639         &sysfs_btree_cache,
640         &sysfs_btree_key_cache,
641         &sysfs_new_stripes,
642         &sysfs_stripes_heap,
643         &sysfs_open_buckets,
644         &sysfs_open_buckets_partial,
645         &sysfs_write_points,
646 #ifdef BCH_WRITE_REF_DEBUG
647         &sysfs_write_refs,
648 #endif
649         &sysfs_nocow_lock_table,
650         &sysfs_io_timers_read,
651         &sysfs_io_timers_write,
652
653         &sysfs_trigger_gc,
654         &sysfs_trigger_discards,
655         &sysfs_trigger_invalidates,
656         &sysfs_prune_cache,
657         &sysfs_btree_wakeup,
658
659         &sysfs_gc_gens_pos,
660
661         &sysfs_copy_gc_enabled,
662         &sysfs_copy_gc_wait,
663
664         &sysfs_rebalance_enabled,
665         &sysfs_rebalance_status,
666         sysfs_pd_controller_files(rebalance),
667
668         &sysfs_moving_ctxts,
669
670         &sysfs_internal_uuid,
671
672         &sysfs_disk_groups,
673         NULL
674 };
675
676 /* options */
677
678 SHOW(bch2_fs_opts_dir)
679 {
680         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
681         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
682         int id = opt - bch2_opt_table;
683         u64 v = bch2_opt_get_by_id(&c->opts, id);
684
685         bch2_opt_to_text(out, c, c->disk_sb.sb, opt, v, OPT_SHOW_FULL_LIST);
686         prt_char(out, '\n');
687
688         return 0;
689 }
690
691 STORE(bch2_fs_opts_dir)
692 {
693         struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
694         const struct bch_option *opt = container_of(attr, struct bch_option, attr);
695         int ret, id = opt - bch2_opt_table;
696         char *tmp;
697         u64 v;
698
699         /*
700          * We don't need to take c->writes for correctness, but it eliminates an
701          * unsightly error message in the dmesg log when we're RO:
702          */
703         if (unlikely(!bch2_write_ref_tryget(c, BCH_WRITE_REF_sysfs)))
704                 return -EROFS;
705
706         tmp = kstrdup(buf, GFP_KERNEL);
707         if (!tmp) {
708                 ret = -ENOMEM;
709                 goto err;
710         }
711
712         ret = bch2_opt_parse(c, opt, strim(tmp), &v, NULL);
713         kfree(tmp);
714
715         if (ret < 0)
716                 goto err;
717
718         ret = bch2_opt_check_may_set(c, id, v);
719         if (ret < 0)
720                 goto err;
721
722         bch2_opt_set_sb(c, opt, v);
723         bch2_opt_set_by_id(&c->opts, id, v);
724
725         if ((id == Opt_background_target ||
726              id == Opt_background_compression) && v)
727                 bch2_set_rebalance_needs_scan(c, 0);
728
729         ret = size;
730 err:
731         bch2_write_ref_put(c, BCH_WRITE_REF_sysfs);
732         return ret;
733 }
734 SYSFS_OPS(bch2_fs_opts_dir);
735
736 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
737
738 int bch2_opts_create_sysfs_files(struct kobject *kobj)
739 {
740         const struct bch_option *i;
741         int ret;
742
743         for (i = bch2_opt_table;
744              i < bch2_opt_table + bch2_opts_nr;
745              i++) {
746                 if (!(i->flags & OPT_FS))
747                         continue;
748
749                 ret = sysfs_create_file(kobj, &i->attr);
750                 if (ret)
751                         return ret;
752         }
753
754         return 0;
755 }
756
757 /* time stats */
758
759 SHOW(bch2_fs_time_stats)
760 {
761         struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
762
763 #define x(name)                                                         \
764         if (attr == &sysfs_time_stat_##name)                            \
765                 bch2_time_stats_to_text(out, &c->times[BCH_TIME_##name]);
766         BCH_TIME_STATS()
767 #undef x
768
769         return 0;
770 }
771
772 STORE(bch2_fs_time_stats)
773 {
774         return size;
775 }
776 SYSFS_OPS(bch2_fs_time_stats);
777
778 struct attribute *bch2_fs_time_stats_files[] = {
779 #define x(name)                                         \
780         &sysfs_time_stat_##name,
781         BCH_TIME_STATS()
782 #undef x
783         NULL
784 };
785
786 static void dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
787 {
788         struct bch_fs *c = ca->fs;
789         struct bch_dev_usage stats = bch2_dev_usage_read(ca);
790         unsigned i, nr[BCH_DATA_NR];
791
792         memset(nr, 0, sizeof(nr));
793
794         for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
795                 nr[c->open_buckets[i].data_type]++;
796
797         printbuf_tabstop_push(out, 8);
798         printbuf_tabstop_push(out, 16);
799         printbuf_tabstop_push(out, 16);
800         printbuf_tabstop_push(out, 16);
801         printbuf_tabstop_push(out, 16);
802
803         bch2_dev_usage_to_text(out, &stats);
804
805         prt_newline(out);
806
807         prt_printf(out, "reserves:");
808         prt_newline(out);
809         for (i = 0; i < BCH_WATERMARK_NR; i++) {
810                 prt_str(out, bch2_watermarks[i]);
811                 prt_tab(out);
812                 prt_u64(out, bch2_dev_buckets_reserved(ca, i));
813                 prt_tab_rjust(out);
814                 prt_newline(out);
815         }
816
817         prt_newline(out);
818
819         printbuf_tabstops_reset(out);
820         printbuf_tabstop_push(out, 24);
821
822         prt_str(out, "freelist_wait");
823         prt_tab(out);
824         prt_str(out, c->freelist_wait.list.first ? "waiting" : "empty");
825         prt_newline(out);
826
827         prt_str(out, "open buckets allocated");
828         prt_tab(out);
829         prt_u64(out, OPEN_BUCKETS_COUNT - c->open_buckets_nr_free);
830         prt_newline(out);
831
832         prt_str(out, "open buckets this dev");
833         prt_tab(out);
834         prt_u64(out, ca->nr_open_buckets);
835         prt_newline(out);
836
837         prt_str(out, "open buckets total");
838         prt_tab(out);
839         prt_u64(out, OPEN_BUCKETS_COUNT);
840         prt_newline(out);
841
842         prt_str(out, "open_buckets_wait");
843         prt_tab(out);
844         prt_str(out, c->open_buckets_wait.list.first ? "waiting" : "empty");
845         prt_newline(out);
846
847         prt_str(out, "open_buckets_btree");
848         prt_tab(out);
849         prt_u64(out, nr[BCH_DATA_btree]);
850         prt_newline(out);
851
852         prt_str(out, "open_buckets_user");
853         prt_tab(out);
854         prt_u64(out, nr[BCH_DATA_user]);
855         prt_newline(out);
856
857         prt_str(out, "buckets_to_invalidate");
858         prt_tab(out);
859         prt_u64(out, should_invalidate_buckets(ca, stats));
860         prt_newline(out);
861
862         prt_str(out, "btree reserve cache");
863         prt_tab(out);
864         prt_u64(out, c->btree_reserve_cache_nr);
865         prt_newline(out);
866 }
867
868 static const char * const bch2_rw[] = {
869         "read",
870         "write",
871         NULL
872 };
873
874 static void dev_io_done_to_text(struct printbuf *out, struct bch_dev *ca)
875 {
876         int rw, i;
877
878         for (rw = 0; rw < 2; rw++) {
879                 prt_printf(out, "%s:\n", bch2_rw[rw]);
880
881                 for (i = 1; i < BCH_DATA_NR; i++)
882                         prt_printf(out, "%-12s:%12llu\n",
883                                bch2_data_types[i],
884                                percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
885         }
886 }
887
888 SHOW(bch2_dev)
889 {
890         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
891         struct bch_fs *c = ca->fs;
892
893         sysfs_printf(uuid,              "%pU\n", ca->uuid.b);
894
895         sysfs_print(bucket_size,        bucket_bytes(ca));
896         sysfs_print(first_bucket,       ca->mi.first_bucket);
897         sysfs_print(nbuckets,           ca->mi.nbuckets);
898         sysfs_print(durability,         ca->mi.durability);
899         sysfs_print(discard,            ca->mi.discard);
900
901         if (attr == &sysfs_label) {
902                 if (ca->mi.group)
903                         bch2_disk_path_to_text(out, c, ca->mi.group - 1);
904                 prt_char(out, '\n');
905         }
906
907         if (attr == &sysfs_has_data) {
908                 prt_bitflags(out, bch2_data_types, bch2_dev_has_data(c, ca));
909                 prt_char(out, '\n');
910         }
911
912         if (attr == &sysfs_state_rw) {
913                 prt_string_option(out, bch2_member_states, ca->mi.state);
914                 prt_char(out, '\n');
915         }
916
917         if (attr == &sysfs_io_done)
918                 dev_io_done_to_text(out, ca);
919
920         if (attr == &sysfs_io_errors)
921                 bch2_dev_io_errors_to_text(out, ca);
922
923         sysfs_print(io_latency_read,            atomic64_read(&ca->cur_latency[READ]));
924         sysfs_print(io_latency_write,           atomic64_read(&ca->cur_latency[WRITE]));
925
926         if (attr == &sysfs_io_latency_stats_read)
927                 bch2_time_stats_to_text(out, &ca->io_latency[READ]);
928
929         if (attr == &sysfs_io_latency_stats_write)
930                 bch2_time_stats_to_text(out, &ca->io_latency[WRITE]);
931
932         sysfs_printf(congested,                 "%u%%",
933                      clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
934                      * 100 / CONGESTED_MAX);
935
936         if (attr == &sysfs_alloc_debug)
937                 dev_alloc_debug_to_text(out, ca);
938
939         return 0;
940 }
941
942 STORE(bch2_dev)
943 {
944         struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
945         struct bch_fs *c = ca->fs;
946         struct bch_member *mi;
947
948         if (attr == &sysfs_discard) {
949                 bool v = strtoul_or_return(buf);
950
951                 mutex_lock(&c->sb_lock);
952                 mi = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
953
954                 if (v != BCH_MEMBER_DISCARD(mi)) {
955                         SET_BCH_MEMBER_DISCARD(mi, v);
956                         bch2_write_super(c);
957                 }
958                 mutex_unlock(&c->sb_lock);
959         }
960
961         if (attr == &sysfs_durability) {
962                 u64 v = strtoul_or_return(buf);
963
964                 mutex_lock(&c->sb_lock);
965                 mi = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
966
967                 if (v + 1 != BCH_MEMBER_DURABILITY(mi)) {
968                         SET_BCH_MEMBER_DURABILITY(mi, v + 1);
969                         bch2_write_super(c);
970                 }
971                 mutex_unlock(&c->sb_lock);
972         }
973
974         if (attr == &sysfs_label) {
975                 char *tmp;
976                 int ret;
977
978                 tmp = kstrdup(buf, GFP_KERNEL);
979                 if (!tmp)
980                         return -ENOMEM;
981
982                 ret = bch2_dev_group_set(c, ca, strim(tmp));
983                 kfree(tmp);
984                 if (ret)
985                         return ret;
986         }
987
988         if (attr == &sysfs_io_errors_reset)
989                 bch2_dev_errors_reset(ca);
990
991         return size;
992 }
993 SYSFS_OPS(bch2_dev);
994
995 struct attribute *bch2_dev_files[] = {
996         &sysfs_uuid,
997         &sysfs_bucket_size,
998         &sysfs_first_bucket,
999         &sysfs_nbuckets,
1000         &sysfs_durability,
1001
1002         /* settings: */
1003         &sysfs_discard,
1004         &sysfs_state_rw,
1005         &sysfs_label,
1006
1007         &sysfs_has_data,
1008         &sysfs_io_done,
1009         &sysfs_io_errors,
1010         &sysfs_io_errors_reset,
1011
1012         &sysfs_io_latency_read,
1013         &sysfs_io_latency_write,
1014         &sysfs_io_latency_stats_read,
1015         &sysfs_io_latency_stats_write,
1016         &sysfs_congested,
1017
1018         /* debug: */
1019         &sysfs_alloc_debug,
1020         NULL
1021 };
1022
1023 #endif  /* _BCACHEFS_SYSFS_H_ */