xref: /linux/fs/bcachefs/sysfs.c (revision 5e2cb28dd7e182dfa641550dfa225913509ad45d)
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(iodone);
153 
154 read_attribute(io_latency_read);
155 read_attribute(io_latency_write);
156 read_attribute(io_latency_stats_read);
157 read_attribute(io_latency_stats_write);
158 read_attribute(congested);
159 
160 read_attribute(btree_write_stats);
161 
162 read_attribute(btree_cache_size);
163 read_attribute(compression_stats);
164 read_attribute(journal_debug);
165 read_attribute(btree_updates);
166 read_attribute(btree_cache);
167 read_attribute(btree_key_cache);
168 read_attribute(stripes_heap);
169 read_attribute(open_buckets);
170 read_attribute(open_buckets_partial);
171 read_attribute(write_points);
172 read_attribute(nocow_lock_table);
173 
174 #ifdef BCH_WRITE_REF_DEBUG
175 read_attribute(write_refs);
176 
177 static const char * const bch2_write_refs[] = {
178 #define x(n)	#n,
179 	BCH_WRITE_REFS()
180 #undef x
181 	NULL
182 };
183 
184 static void bch2_write_refs_to_text(struct printbuf *out, struct bch_fs *c)
185 {
186 	bch2_printbuf_tabstop_push(out, 24);
187 
188 	for (unsigned i = 0; i < ARRAY_SIZE(c->writes); i++) {
189 		prt_str(out, bch2_write_refs[i]);
190 		prt_tab(out);
191 		prt_printf(out, "%li", atomic_long_read(&c->writes[i]));
192 		prt_newline(out);
193 	}
194 }
195 #endif
196 
197 read_attribute(internal_uuid);
198 read_attribute(disk_groups);
199 
200 read_attribute(has_data);
201 read_attribute(alloc_debug);
202 
203 #define x(t, n, ...) read_attribute(t);
204 BCH_PERSISTENT_COUNTERS()
205 #undef x
206 
207 rw_attribute(discard);
208 rw_attribute(label);
209 
210 rw_attribute(copy_gc_enabled);
211 read_attribute(copy_gc_wait);
212 
213 rw_attribute(rebalance_enabled);
214 sysfs_pd_controller_attribute(rebalance);
215 read_attribute(rebalance_work);
216 rw_attribute(promote_whole_extents);
217 
218 read_attribute(new_stripes);
219 
220 read_attribute(io_timers_read);
221 read_attribute(io_timers_write);
222 
223 read_attribute(moving_ctxts);
224 
225 #ifdef CONFIG_BCACHEFS_TESTS
226 write_attribute(perf_test);
227 #endif /* CONFIG_BCACHEFS_TESTS */
228 
229 #define x(_name)						\
230 	static struct attribute sysfs_time_stat_##_name =		\
231 		{ .name = #_name, .mode = 0444 };
232 	BCH_TIME_STATS()
233 #undef x
234 
235 static struct attribute sysfs_state_rw = {
236 	.name = "state",
237 	.mode =  0444,
238 };
239 
240 static size_t bch2_btree_cache_size(struct bch_fs *c)
241 {
242 	size_t ret = 0;
243 	struct btree *b;
244 
245 	mutex_lock(&c->btree_cache.lock);
246 	list_for_each_entry(b, &c->btree_cache.live, list)
247 		ret += btree_bytes(c);
248 
249 	mutex_unlock(&c->btree_cache.lock);
250 	return ret;
251 }
252 
253 static int bch2_compression_stats_to_text(struct printbuf *out, struct bch_fs *c)
254 {
255 	struct btree_trans *trans;
256 	struct btree_iter iter;
257 	struct bkey_s_c k;
258 	enum btree_id id;
259 	u64 nr_uncompressed_extents = 0,
260 	    nr_compressed_extents = 0,
261 	    nr_incompressible_extents = 0,
262 	    uncompressed_sectors = 0,
263 	    incompressible_sectors = 0,
264 	    compressed_sectors_compressed = 0,
265 	    compressed_sectors_uncompressed = 0;
266 	int ret = 0;
267 
268 	if (!test_bit(BCH_FS_STARTED, &c->flags))
269 		return -EPERM;
270 
271 	trans = bch2_trans_get(c);
272 
273 	for (id = 0; id < BTREE_ID_NR; id++) {
274 		if (!btree_type_has_ptrs(id))
275 			continue;
276 
277 		for_each_btree_key(trans, iter, id, POS_MIN,
278 				   BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
279 			struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
280 			const union bch_extent_entry *entry;
281 			struct extent_ptr_decoded p;
282 			bool compressed = false, uncompressed = false, incompressible = false;
283 
284 			bkey_for_each_ptr_decode(k.k, ptrs, p, entry) {
285 				switch (p.crc.compression_type) {
286 				case BCH_COMPRESSION_TYPE_none:
287 					uncompressed = true;
288 					uncompressed_sectors += k.k->size;
289 					break;
290 				case BCH_COMPRESSION_TYPE_incompressible:
291 					incompressible = true;
292 					incompressible_sectors += k.k->size;
293 					break;
294 				default:
295 					compressed_sectors_compressed +=
296 						p.crc.compressed_size;
297 					compressed_sectors_uncompressed +=
298 						p.crc.uncompressed_size;
299 					compressed = true;
300 					break;
301 				}
302 			}
303 
304 			if (incompressible)
305 				nr_incompressible_extents++;
306 			else if (uncompressed)
307 				nr_uncompressed_extents++;
308 			else if (compressed)
309 				nr_compressed_extents++;
310 		}
311 		bch2_trans_iter_exit(trans, &iter);
312 	}
313 
314 	bch2_trans_put(trans);
315 
316 	if (ret)
317 		return ret;
318 
319 	prt_printf(out, "uncompressed:\n");
320 	prt_printf(out, "	nr extents:		%llu\n", nr_uncompressed_extents);
321 	prt_printf(out, "	size:			");
322 	prt_human_readable_u64(out, uncompressed_sectors << 9);
323 	prt_printf(out, "\n");
324 
325 	prt_printf(out, "compressed:\n");
326 	prt_printf(out, "	nr extents:		%llu\n", nr_compressed_extents);
327 	prt_printf(out, "	compressed size:	");
328 	prt_human_readable_u64(out, compressed_sectors_compressed << 9);
329 	prt_printf(out, "\n");
330 	prt_printf(out, "	uncompressed size:	");
331 	prt_human_readable_u64(out, compressed_sectors_uncompressed << 9);
332 	prt_printf(out, "\n");
333 
334 	prt_printf(out, "incompressible:\n");
335 	prt_printf(out, "	nr extents:		%llu\n", nr_incompressible_extents);
336 	prt_printf(out, "	size:			");
337 	prt_human_readable_u64(out, incompressible_sectors << 9);
338 	prt_printf(out, "\n");
339 	return 0;
340 }
341 
342 static void bch2_gc_gens_pos_to_text(struct printbuf *out, struct bch_fs *c)
343 {
344 	prt_printf(out, "%s: ", bch2_btree_ids[c->gc_gens_btree]);
345 	bch2_bpos_to_text(out, c->gc_gens_pos);
346 	prt_printf(out, "\n");
347 }
348 
349 static void bch2_btree_wakeup_all(struct bch_fs *c)
350 {
351 	struct btree_trans *trans;
352 
353 	seqmutex_lock(&c->btree_trans_lock);
354 	list_for_each_entry(trans, &c->btree_trans_list, list) {
355 		struct btree_bkey_cached_common *b = READ_ONCE(trans->locking);
356 
357 		if (b)
358 			six_lock_wakeup_all(&b->lock);
359 
360 	}
361 	seqmutex_unlock(&c->btree_trans_lock);
362 }
363 
364 SHOW(bch2_fs)
365 {
366 	struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
367 
368 	sysfs_print(minor,			c->minor);
369 	sysfs_printf(internal_uuid, "%pU",	c->sb.uuid.b);
370 
371 	sysfs_hprint(btree_cache_size,		bch2_btree_cache_size(c));
372 
373 	if (attr == &sysfs_btree_write_stats)
374 		bch2_btree_write_stats_to_text(out, c);
375 
376 	sysfs_printf(btree_gc_periodic, "%u",	(int) c->btree_gc_periodic);
377 
378 	if (attr == &sysfs_gc_gens_pos)
379 		bch2_gc_gens_pos_to_text(out, c);
380 
381 	sysfs_printf(copy_gc_enabled, "%i", c->copy_gc_enabled);
382 
383 	sysfs_printf(rebalance_enabled,		"%i", c->rebalance.enabled);
384 	sysfs_pd_controller_show(rebalance,	&c->rebalance.pd); /* XXX */
385 
386 	if (attr == &sysfs_copy_gc_wait)
387 		bch2_copygc_wait_to_text(out, c);
388 
389 	if (attr == &sysfs_rebalance_work)
390 		bch2_rebalance_work_to_text(out, c);
391 
392 	sysfs_print(promote_whole_extents,	c->promote_whole_extents);
393 
394 	/* Debugging: */
395 
396 	if (attr == &sysfs_journal_debug)
397 		bch2_journal_debug_to_text(out, &c->journal);
398 
399 	if (attr == &sysfs_btree_updates)
400 		bch2_btree_updates_to_text(out, c);
401 
402 	if (attr == &sysfs_btree_cache)
403 		bch2_btree_cache_to_text(out, c);
404 
405 	if (attr == &sysfs_btree_key_cache)
406 		bch2_btree_key_cache_to_text(out, &c->btree_key_cache);
407 
408 	if (attr == &sysfs_stripes_heap)
409 		bch2_stripes_heap_to_text(out, c);
410 
411 	if (attr == &sysfs_open_buckets)
412 		bch2_open_buckets_to_text(out, c);
413 
414 	if (attr == &sysfs_open_buckets_partial)
415 		bch2_open_buckets_partial_to_text(out, c);
416 
417 	if (attr == &sysfs_write_points)
418 		bch2_write_points_to_text(out, c);
419 
420 	if (attr == &sysfs_compression_stats)
421 		bch2_compression_stats_to_text(out, c);
422 
423 	if (attr == &sysfs_new_stripes)
424 		bch2_new_stripes_to_text(out, c);
425 
426 	if (attr == &sysfs_io_timers_read)
427 		bch2_io_timers_to_text(out, &c->io_clock[READ]);
428 
429 	if (attr == &sysfs_io_timers_write)
430 		bch2_io_timers_to_text(out, &c->io_clock[WRITE]);
431 
432 	if (attr == &sysfs_moving_ctxts)
433 		bch2_fs_moving_ctxts_to_text(out, c);
434 
435 #ifdef BCH_WRITE_REF_DEBUG
436 	if (attr == &sysfs_write_refs)
437 		bch2_write_refs_to_text(out, c);
438 #endif
439 
440 	if (attr == &sysfs_nocow_lock_table)
441 		bch2_nocow_locks_to_text(out, &c->nocow_locks);
442 
443 	if (attr == &sysfs_disk_groups)
444 		bch2_disk_groups_to_text(out, c);
445 
446 	return 0;
447 }
448 
449 STORE(bch2_fs)
450 {
451 	struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
452 
453 	if (attr == &sysfs_btree_gc_periodic) {
454 		ssize_t ret = strtoul_safe(buf, c->btree_gc_periodic)
455 			?: (ssize_t) size;
456 
457 		wake_up_process(c->gc_thread);
458 		return ret;
459 	}
460 
461 	if (attr == &sysfs_copy_gc_enabled) {
462 		ssize_t ret = strtoul_safe(buf, c->copy_gc_enabled)
463 			?: (ssize_t) size;
464 
465 		if (c->copygc_thread)
466 			wake_up_process(c->copygc_thread);
467 		return ret;
468 	}
469 
470 	if (attr == &sysfs_rebalance_enabled) {
471 		ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
472 			?: (ssize_t) size;
473 
474 		rebalance_wakeup(c);
475 		return ret;
476 	}
477 
478 	sysfs_pd_controller_store(rebalance,	&c->rebalance.pd);
479 
480 	sysfs_strtoul(promote_whole_extents,	c->promote_whole_extents);
481 
482 	/* Debugging: */
483 
484 	if (!test_bit(BCH_FS_STARTED, &c->flags))
485 		return -EPERM;
486 
487 	/* Debugging: */
488 
489 	if (!test_bit(BCH_FS_RW, &c->flags))
490 		return -EROFS;
491 
492 	if (attr == &sysfs_prune_cache) {
493 		struct shrink_control sc;
494 
495 		sc.gfp_mask = GFP_KERNEL;
496 		sc.nr_to_scan = strtoul_or_return(buf);
497 		c->btree_cache.shrink->scan_objects(c->btree_cache.shrink, &sc);
498 	}
499 
500 	if (attr == &sysfs_btree_wakeup)
501 		bch2_btree_wakeup_all(c);
502 
503 	if (attr == &sysfs_trigger_gc) {
504 		/*
505 		 * Full gc is currently incompatible with btree key cache:
506 		 */
507 #if 0
508 		down_read(&c->state_lock);
509 		bch2_gc(c, false, false);
510 		up_read(&c->state_lock);
511 #else
512 		bch2_gc_gens(c);
513 #endif
514 	}
515 
516 	if (attr == &sysfs_trigger_discards)
517 		bch2_do_discards(c);
518 
519 	if (attr == &sysfs_trigger_invalidates)
520 		bch2_do_invalidates(c);
521 
522 #ifdef CONFIG_BCACHEFS_TESTS
523 	if (attr == &sysfs_perf_test) {
524 		char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
525 		char *test		= strsep(&p, " \t\n");
526 		char *nr_str		= strsep(&p, " \t\n");
527 		char *threads_str	= strsep(&p, " \t\n");
528 		unsigned threads;
529 		u64 nr;
530 		int ret = -EINVAL;
531 
532 		if (threads_str &&
533 		    !(ret = kstrtouint(threads_str, 10, &threads)) &&
534 		    !(ret = bch2_strtoull_h(nr_str, &nr)))
535 			ret = bch2_btree_perf_test(c, test, nr, threads);
536 		kfree(tmp);
537 
538 		if (ret)
539 			size = ret;
540 	}
541 #endif
542 	return size;
543 }
544 SYSFS_OPS(bch2_fs);
545 
546 struct attribute *bch2_fs_files[] = {
547 	&sysfs_minor,
548 	&sysfs_btree_cache_size,
549 	&sysfs_btree_write_stats,
550 
551 	&sysfs_promote_whole_extents,
552 
553 	&sysfs_compression_stats,
554 
555 #ifdef CONFIG_BCACHEFS_TESTS
556 	&sysfs_perf_test,
557 #endif
558 	NULL
559 };
560 
561 /* counters dir */
562 
563 SHOW(bch2_fs_counters)
564 {
565 	struct bch_fs *c = container_of(kobj, struct bch_fs, counters_kobj);
566 	u64 counter = 0;
567 	u64 counter_since_mount = 0;
568 
569 	printbuf_tabstop_push(out, 32);
570 
571 	#define x(t, ...) \
572 		if (attr == &sysfs_##t) {					\
573 			counter             = percpu_u64_get(&c->counters[BCH_COUNTER_##t]);\
574 			counter_since_mount = counter - c->counters_on_mount[BCH_COUNTER_##t];\
575 			prt_printf(out, "since mount:");				\
576 			prt_tab(out);						\
577 			prt_human_readable_u64(out, counter_since_mount);	\
578 			prt_newline(out);					\
579 										\
580 			prt_printf(out, "since filesystem creation:");		\
581 			prt_tab(out);						\
582 			prt_human_readable_u64(out, counter);			\
583 			prt_newline(out);					\
584 		}
585 	BCH_PERSISTENT_COUNTERS()
586 	#undef x
587 	return 0;
588 }
589 
590 STORE(bch2_fs_counters) {
591 	return 0;
592 }
593 
594 SYSFS_OPS(bch2_fs_counters);
595 
596 struct attribute *bch2_fs_counters_files[] = {
597 #define x(t, ...) \
598 	&sysfs_##t,
599 	BCH_PERSISTENT_COUNTERS()
600 #undef x
601 	NULL
602 };
603 /* internal dir - just a wrapper */
604 
605 SHOW(bch2_fs_internal)
606 {
607 	struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
608 
609 	return bch2_fs_to_text(out, &c->kobj, attr);
610 }
611 
612 STORE(bch2_fs_internal)
613 {
614 	struct bch_fs *c = container_of(kobj, struct bch_fs, internal);
615 
616 	return bch2_fs_store(&c->kobj, attr, buf, size);
617 }
618 SYSFS_OPS(bch2_fs_internal);
619 
620 struct attribute *bch2_fs_internal_files[] = {
621 	&sysfs_journal_debug,
622 	&sysfs_btree_updates,
623 	&sysfs_btree_cache,
624 	&sysfs_btree_key_cache,
625 	&sysfs_new_stripes,
626 	&sysfs_stripes_heap,
627 	&sysfs_open_buckets,
628 	&sysfs_open_buckets_partial,
629 	&sysfs_write_points,
630 #ifdef BCH_WRITE_REF_DEBUG
631 	&sysfs_write_refs,
632 #endif
633 	&sysfs_nocow_lock_table,
634 	&sysfs_io_timers_read,
635 	&sysfs_io_timers_write,
636 
637 	&sysfs_trigger_gc,
638 	&sysfs_trigger_discards,
639 	&sysfs_trigger_invalidates,
640 	&sysfs_prune_cache,
641 	&sysfs_btree_wakeup,
642 
643 	&sysfs_gc_gens_pos,
644 
645 	&sysfs_copy_gc_enabled,
646 	&sysfs_copy_gc_wait,
647 
648 	&sysfs_rebalance_enabled,
649 	&sysfs_rebalance_work,
650 	sysfs_pd_controller_files(rebalance),
651 
652 	&sysfs_moving_ctxts,
653 
654 	&sysfs_internal_uuid,
655 
656 	&sysfs_disk_groups,
657 	NULL
658 };
659 
660 /* options */
661 
662 SHOW(bch2_fs_opts_dir)
663 {
664 	struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
665 	const struct bch_option *opt = container_of(attr, struct bch_option, attr);
666 	int id = opt - bch2_opt_table;
667 	u64 v = bch2_opt_get_by_id(&c->opts, id);
668 
669 	bch2_opt_to_text(out, c, c->disk_sb.sb, opt, v, OPT_SHOW_FULL_LIST);
670 	prt_char(out, '\n');
671 
672 	return 0;
673 }
674 
675 STORE(bch2_fs_opts_dir)
676 {
677 	struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
678 	const struct bch_option *opt = container_of(attr, struct bch_option, attr);
679 	int ret, id = opt - bch2_opt_table;
680 	char *tmp;
681 	u64 v;
682 
683 	/*
684 	 * We don't need to take c->writes for correctness, but it eliminates an
685 	 * unsightly error message in the dmesg log when we're RO:
686 	 */
687 	if (unlikely(!bch2_write_ref_tryget(c, BCH_WRITE_REF_sysfs)))
688 		return -EROFS;
689 
690 	tmp = kstrdup(buf, GFP_KERNEL);
691 	if (!tmp) {
692 		ret = -ENOMEM;
693 		goto err;
694 	}
695 
696 	ret = bch2_opt_parse(c, opt, strim(tmp), &v, NULL);
697 	kfree(tmp);
698 
699 	if (ret < 0)
700 		goto err;
701 
702 	ret = bch2_opt_check_may_set(c, id, v);
703 	if (ret < 0)
704 		goto err;
705 
706 	bch2_opt_set_sb(c, opt, v);
707 	bch2_opt_set_by_id(&c->opts, id, v);
708 
709 	if ((id == Opt_background_target ||
710 	     id == Opt_background_compression) && v) {
711 		bch2_rebalance_add_work(c, S64_MAX);
712 		rebalance_wakeup(c);
713 	}
714 
715 	ret = size;
716 err:
717 	bch2_write_ref_put(c, BCH_WRITE_REF_sysfs);
718 	return ret;
719 }
720 SYSFS_OPS(bch2_fs_opts_dir);
721 
722 struct attribute *bch2_fs_opts_dir_files[] = { NULL };
723 
724 int bch2_opts_create_sysfs_files(struct kobject *kobj)
725 {
726 	const struct bch_option *i;
727 	int ret;
728 
729 	for (i = bch2_opt_table;
730 	     i < bch2_opt_table + bch2_opts_nr;
731 	     i++) {
732 		if (!(i->flags & OPT_FS))
733 			continue;
734 
735 		ret = sysfs_create_file(kobj, &i->attr);
736 		if (ret)
737 			return ret;
738 	}
739 
740 	return 0;
741 }
742 
743 /* time stats */
744 
745 SHOW(bch2_fs_time_stats)
746 {
747 	struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
748 
749 #define x(name)								\
750 	if (attr == &sysfs_time_stat_##name)				\
751 		bch2_time_stats_to_text(out, &c->times[BCH_TIME_##name]);
752 	BCH_TIME_STATS()
753 #undef x
754 
755 	return 0;
756 }
757 
758 STORE(bch2_fs_time_stats)
759 {
760 	return size;
761 }
762 SYSFS_OPS(bch2_fs_time_stats);
763 
764 struct attribute *bch2_fs_time_stats_files[] = {
765 #define x(name)						\
766 	&sysfs_time_stat_##name,
767 	BCH_TIME_STATS()
768 #undef x
769 	NULL
770 };
771 
772 static void dev_alloc_debug_to_text(struct printbuf *out, struct bch_dev *ca)
773 {
774 	struct bch_fs *c = ca->fs;
775 	struct bch_dev_usage stats = bch2_dev_usage_read(ca);
776 	unsigned i, nr[BCH_DATA_NR];
777 
778 	memset(nr, 0, sizeof(nr));
779 
780 	for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
781 		nr[c->open_buckets[i].data_type]++;
782 
783 	printbuf_tabstop_push(out, 8);
784 	printbuf_tabstop_push(out, 16);
785 	printbuf_tabstop_push(out, 16);
786 	printbuf_tabstop_push(out, 16);
787 	printbuf_tabstop_push(out, 16);
788 
789 	prt_tab(out);
790 	prt_str(out, "buckets");
791 	prt_tab_rjust(out);
792 	prt_str(out, "sectors");
793 	prt_tab_rjust(out);
794 	prt_str(out, "fragmented");
795 	prt_tab_rjust(out);
796 	prt_newline(out);
797 
798 	for (i = 0; i < BCH_DATA_NR; i++) {
799 		prt_str(out, bch2_data_types[i]);
800 		prt_tab(out);
801 		prt_u64(out, stats.d[i].buckets);
802 		prt_tab_rjust(out);
803 		prt_u64(out, stats.d[i].sectors);
804 		prt_tab_rjust(out);
805 		prt_u64(out, stats.d[i].fragmented);
806 		prt_tab_rjust(out);
807 		prt_newline(out);
808 	}
809 
810 	prt_str(out, "ec");
811 	prt_tab(out);
812 	prt_u64(out, stats.buckets_ec);
813 	prt_tab_rjust(out);
814 	prt_newline(out);
815 
816 	prt_newline(out);
817 
818 	prt_printf(out, "reserves:");
819 	prt_newline(out);
820 	for (i = 0; i < BCH_WATERMARK_NR; i++) {
821 		prt_str(out, bch2_watermarks[i]);
822 		prt_tab(out);
823 		prt_u64(out, bch2_dev_buckets_reserved(ca, i));
824 		prt_tab_rjust(out);
825 		prt_newline(out);
826 	}
827 
828 	prt_newline(out);
829 
830 	printbuf_tabstops_reset(out);
831 	printbuf_tabstop_push(out, 24);
832 
833 	prt_str(out, "freelist_wait");
834 	prt_tab(out);
835 	prt_str(out, c->freelist_wait.list.first ? "waiting" : "empty");
836 	prt_newline(out);
837 
838 	prt_str(out, "open buckets allocated");
839 	prt_tab(out);
840 	prt_u64(out, OPEN_BUCKETS_COUNT - c->open_buckets_nr_free);
841 	prt_newline(out);
842 
843 	prt_str(out, "open buckets this dev");
844 	prt_tab(out);
845 	prt_u64(out, ca->nr_open_buckets);
846 	prt_newline(out);
847 
848 	prt_str(out, "open buckets total");
849 	prt_tab(out);
850 	prt_u64(out, OPEN_BUCKETS_COUNT);
851 	prt_newline(out);
852 
853 	prt_str(out, "open_buckets_wait");
854 	prt_tab(out);
855 	prt_str(out, c->open_buckets_wait.list.first ? "waiting" : "empty");
856 	prt_newline(out);
857 
858 	prt_str(out, "open_buckets_btree");
859 	prt_tab(out);
860 	prt_u64(out, nr[BCH_DATA_btree]);
861 	prt_newline(out);
862 
863 	prt_str(out, "open_buckets_user");
864 	prt_tab(out);
865 	prt_u64(out, nr[BCH_DATA_user]);
866 	prt_newline(out);
867 
868 	prt_str(out, "buckets_to_invalidate");
869 	prt_tab(out);
870 	prt_u64(out, should_invalidate_buckets(ca, stats));
871 	prt_newline(out);
872 
873 	prt_str(out, "btree reserve cache");
874 	prt_tab(out);
875 	prt_u64(out, c->btree_reserve_cache_nr);
876 	prt_newline(out);
877 }
878 
879 static const char * const bch2_rw[] = {
880 	"read",
881 	"write",
882 	NULL
883 };
884 
885 static void dev_iodone_to_text(struct printbuf *out, struct bch_dev *ca)
886 {
887 	int rw, i;
888 
889 	for (rw = 0; rw < 2; rw++) {
890 		prt_printf(out, "%s:\n", bch2_rw[rw]);
891 
892 		for (i = 1; i < BCH_DATA_NR; i++)
893 			prt_printf(out, "%-12s:%12llu\n",
894 			       bch2_data_types[i],
895 			       percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
896 	}
897 }
898 
899 SHOW(bch2_dev)
900 {
901 	struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
902 	struct bch_fs *c = ca->fs;
903 
904 	sysfs_printf(uuid,		"%pU\n", ca->uuid.b);
905 
906 	sysfs_print(bucket_size,	bucket_bytes(ca));
907 	sysfs_print(first_bucket,	ca->mi.first_bucket);
908 	sysfs_print(nbuckets,		ca->mi.nbuckets);
909 	sysfs_print(durability,		ca->mi.durability);
910 	sysfs_print(discard,		ca->mi.discard);
911 
912 	if (attr == &sysfs_label) {
913 		if (ca->mi.group) {
914 			mutex_lock(&c->sb_lock);
915 			bch2_disk_path_to_text(out, c->disk_sb.sb,
916 					       ca->mi.group - 1);
917 			mutex_unlock(&c->sb_lock);
918 		}
919 
920 		prt_char(out, '\n');
921 	}
922 
923 	if (attr == &sysfs_has_data) {
924 		prt_bitflags(out, bch2_data_types, bch2_dev_has_data(c, ca));
925 		prt_char(out, '\n');
926 	}
927 
928 	if (attr == &sysfs_state_rw) {
929 		prt_string_option(out, bch2_member_states, ca->mi.state);
930 		prt_char(out, '\n');
931 	}
932 
933 	if (attr == &sysfs_iodone)
934 		dev_iodone_to_text(out, ca);
935 
936 	sysfs_print(io_latency_read,		atomic64_read(&ca->cur_latency[READ]));
937 	sysfs_print(io_latency_write,		atomic64_read(&ca->cur_latency[WRITE]));
938 
939 	if (attr == &sysfs_io_latency_stats_read)
940 		bch2_time_stats_to_text(out, &ca->io_latency[READ]);
941 
942 	if (attr == &sysfs_io_latency_stats_write)
943 		bch2_time_stats_to_text(out, &ca->io_latency[WRITE]);
944 
945 	sysfs_printf(congested,			"%u%%",
946 		     clamp(atomic_read(&ca->congested), 0, CONGESTED_MAX)
947 		     * 100 / CONGESTED_MAX);
948 
949 	if (attr == &sysfs_alloc_debug)
950 		dev_alloc_debug_to_text(out, ca);
951 
952 	return 0;
953 }
954 
955 STORE(bch2_dev)
956 {
957 	struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
958 	struct bch_fs *c = ca->fs;
959 	struct bch_member *mi;
960 
961 	if (attr == &sysfs_discard) {
962 		bool 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 != BCH_MEMBER_DISCARD(mi)) {
968 			SET_BCH_MEMBER_DISCARD(mi, v);
969 			bch2_write_super(c);
970 		}
971 		mutex_unlock(&c->sb_lock);
972 	}
973 
974 	if (attr == &sysfs_durability) {
975 		u64 v = strtoul_or_return(buf);
976 
977 		mutex_lock(&c->sb_lock);
978 		mi = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
979 
980 		if (v + 1 != BCH_MEMBER_DURABILITY(mi)) {
981 			SET_BCH_MEMBER_DURABILITY(mi, v + 1);
982 			bch2_write_super(c);
983 		}
984 		mutex_unlock(&c->sb_lock);
985 	}
986 
987 	if (attr == &sysfs_label) {
988 		char *tmp;
989 		int ret;
990 
991 		tmp = kstrdup(buf, GFP_KERNEL);
992 		if (!tmp)
993 			return -ENOMEM;
994 
995 		ret = bch2_dev_group_set(c, ca, strim(tmp));
996 		kfree(tmp);
997 		if (ret)
998 			return ret;
999 	}
1000 
1001 	return size;
1002 }
1003 SYSFS_OPS(bch2_dev);
1004 
1005 struct attribute *bch2_dev_files[] = {
1006 	&sysfs_uuid,
1007 	&sysfs_bucket_size,
1008 	&sysfs_first_bucket,
1009 	&sysfs_nbuckets,
1010 	&sysfs_durability,
1011 
1012 	/* settings: */
1013 	&sysfs_discard,
1014 	&sysfs_state_rw,
1015 	&sysfs_label,
1016 
1017 	&sysfs_has_data,
1018 	&sysfs_iodone,
1019 
1020 	&sysfs_io_latency_read,
1021 	&sysfs_io_latency_write,
1022 	&sysfs_io_latency_stats_read,
1023 	&sysfs_io_latency_stats_write,
1024 	&sysfs_congested,
1025 
1026 	/* debug: */
1027 	&sysfs_alloc_debug,
1028 	NULL
1029 };
1030 
1031 #endif  /* _BCACHEFS_SYSFS_H_ */
1032