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#include "bcache.h"
  10#include "sysfs.h"
  11#include "btree.h"
  12#include "request.h"
  13#include "writeback.h"
  14#include "features.h"
  15
  16#include <linux/blkdev.h>
  17#include <linux/sort.h>
  18#include <linux/sched/clock.h>
  19
  20extern bool bcache_is_reboot;
  21
  22/* Default is 0 ("writethrough") */
  23static const char * const bch_cache_modes[] = {
  24	"writethrough",
  25	"writeback",
  26	"writearound",
  27	"none",
  28	NULL
  29};
  30
  31static const char * const bch_reada_cache_policies[] = {
  32	"all",
  33	"meta-only",
  34	NULL
  35};
  36
  37/* Default is 0 ("auto") */
  38static const char * const bch_stop_on_failure_modes[] = {
  39	"auto",
  40	"always",
  41	NULL
  42};
  43
  44static const char * const cache_replacement_policies[] = {
  45	"lru",
  46	"fifo",
  47	"random",
  48	NULL
  49};
  50
  51static const char * const error_actions[] = {
  52	"unregister",
  53	"panic",
  54	NULL
  55};
  56
  57write_attribute(attach);
  58write_attribute(detach);
  59write_attribute(unregister);
  60write_attribute(stop);
  61write_attribute(clear_stats);
  62write_attribute(trigger_gc);
  63write_attribute(prune_cache);
  64write_attribute(flash_vol_create);
  65
  66read_attribute(bucket_size);
  67read_attribute(block_size);
  68read_attribute(nbuckets);
  69read_attribute(tree_depth);
  70read_attribute(root_usage_percent);
  71read_attribute(priority_stats);
  72read_attribute(btree_cache_size);
  73read_attribute(btree_cache_max_chain);
  74read_attribute(cache_available_percent);
  75read_attribute(written);
  76read_attribute(btree_written);
  77read_attribute(metadata_written);
  78read_attribute(active_journal_entries);
  79read_attribute(backing_dev_name);
  80read_attribute(backing_dev_uuid);
  81
  82sysfs_time_stats_attribute(btree_gc,	sec, ms);
  83sysfs_time_stats_attribute(btree_split, sec, us);
  84sysfs_time_stats_attribute(btree_sort,	ms,  us);
  85sysfs_time_stats_attribute(btree_read,	ms,  us);
  86
  87read_attribute(btree_nodes);
  88read_attribute(btree_used_percent);
  89read_attribute(average_key_size);
  90read_attribute(dirty_data);
  91read_attribute(bset_tree_stats);
  92read_attribute(feature_compat);
  93read_attribute(feature_ro_compat);
  94read_attribute(feature_incompat);
  95
  96read_attribute(state);
  97read_attribute(cache_read_races);
  98read_attribute(reclaim);
  99read_attribute(reclaimed_journal_buckets);
 100read_attribute(flush_write);
 101read_attribute(writeback_keys_done);
 102read_attribute(writeback_keys_failed);
 103read_attribute(io_errors);
 104read_attribute(congested);
 105read_attribute(cutoff_writeback);
 106read_attribute(cutoff_writeback_sync);
 107rw_attribute(congested_read_threshold_us);
 108rw_attribute(congested_write_threshold_us);
 109
 110rw_attribute(sequential_cutoff);
 111rw_attribute(data_csum);
 112rw_attribute(cache_mode);
 113rw_attribute(readahead_cache_policy);
 114rw_attribute(stop_when_cache_set_failed);
 115rw_attribute(writeback_metadata);
 116rw_attribute(writeback_running);
 117rw_attribute(writeback_percent);
 118rw_attribute(writeback_delay);
 119rw_attribute(writeback_rate);
 120rw_attribute(writeback_consider_fragment);
 121
 122rw_attribute(writeback_rate_update_seconds);
 123rw_attribute(writeback_rate_i_term_inverse);
 124rw_attribute(writeback_rate_p_term_inverse);
 125rw_attribute(writeback_rate_fp_term_low);
 126rw_attribute(writeback_rate_fp_term_mid);
 127rw_attribute(writeback_rate_fp_term_high);
 128rw_attribute(writeback_rate_minimum);
 129read_attribute(writeback_rate_debug);
 130
 131read_attribute(stripe_size);
 132read_attribute(partial_stripes_expensive);
 133
 134rw_attribute(synchronous);
 135rw_attribute(journal_delay_ms);
 136rw_attribute(io_disable);
 137rw_attribute(running);
 138rw_attribute(label);
 139rw_attribute(errors);
 140rw_attribute(io_error_limit);
 141rw_attribute(io_error_halflife);
 142rw_attribute(verify);
 143rw_attribute(bypass_torture_test);
 144rw_attribute(key_merging_disabled);
 145rw_attribute(gc_always_rewrite);
 146rw_attribute(expensive_debug_checks);
 147rw_attribute(cache_replacement_policy);
 148rw_attribute(btree_shrinker_disabled);
 149rw_attribute(copy_gc_enabled);
 150rw_attribute(idle_max_writeback_rate);
 151rw_attribute(gc_after_writeback);
 152rw_attribute(size);
 153
 154static ssize_t bch_snprint_string_list(char *buf,
 155				       size_t size,
 156				       const char * const list[],
 157				       size_t selected)
 158{
 159	char *out = buf;
 160	size_t i;
 161
 162	for (i = 0; list[i]; i++)
 163		out += scnprintf(out, buf + size - out,
 164				i == selected ? "[%s] " : "%s ", list[i]);
 165
 166	out[-1] = '\n';
 167	return out - buf;
 168}
 169
 170SHOW(__bch_cached_dev)
 171{
 172	struct cached_dev *dc = container_of(kobj, struct cached_dev,
 173					     disk.kobj);
 174	char const *states[] = { "no cache", "clean", "dirty", "inconsistent" };
 175	int wb = dc->writeback_running;
 176
 177#define var(stat)		(dc->stat)
 178
 179	if (attr == &sysfs_cache_mode)
 180		return bch_snprint_string_list(buf, PAGE_SIZE,
 181					       bch_cache_modes,
 182					       BDEV_CACHE_MODE(&dc->sb));
 183
 184	if (attr == &sysfs_readahead_cache_policy)
 185		return bch_snprint_string_list(buf, PAGE_SIZE,
 186					      bch_reada_cache_policies,
 187					      dc->cache_readahead_policy);
 188
 189	if (attr == &sysfs_stop_when_cache_set_failed)
 190		return bch_snprint_string_list(buf, PAGE_SIZE,
 191					       bch_stop_on_failure_modes,
 192					       dc->stop_when_cache_set_failed);
 193
 194
 195	sysfs_printf(data_csum,		"%i", dc->disk.data_csum);
 196	var_printf(verify,		"%i");
 197	var_printf(bypass_torture_test,	"%i");
 198	var_printf(writeback_metadata,	"%i");
 199	var_printf(writeback_running,	"%i");
 200	var_printf(writeback_consider_fragment,	"%i");
 201	var_print(writeback_delay);
 202	var_print(writeback_percent);
 203	sysfs_hprint(writeback_rate,
 204		     wb ? atomic_long_read(&dc->writeback_rate.rate) << 9 : 0);
 205	sysfs_printf(io_errors,		"%i", atomic_read(&dc->io_errors));
 206	sysfs_printf(io_error_limit,	"%i", dc->error_limit);
 207	sysfs_printf(io_disable,	"%i", dc->io_disable);
 208	var_print(writeback_rate_update_seconds);
 209	var_print(writeback_rate_i_term_inverse);
 210	var_print(writeback_rate_p_term_inverse);
 211	var_print(writeback_rate_fp_term_low);
 212	var_print(writeback_rate_fp_term_mid);
 213	var_print(writeback_rate_fp_term_high);
 214	var_print(writeback_rate_minimum);
 215
 216	if (attr == &sysfs_writeback_rate_debug) {
 217		char rate[20];
 218		char dirty[20];
 219		char target[20];
 220		char proportional[20];
 221		char integral[20];
 222		char change[20];
 223		s64 next_io;
 224
 225		/*
 226		 * Except for dirty and target, other values should
 227		 * be 0 if writeback is not running.
 228		 */
 229		bch_hprint(rate,
 230			   wb ? atomic_long_read(&dc->writeback_rate.rate) << 9
 231			      : 0);
 232		bch_hprint(dirty, bcache_dev_sectors_dirty(&dc->disk) << 9);
 233		bch_hprint(target, dc->writeback_rate_target << 9);
 234		bch_hprint(proportional,
 235			   wb ? dc->writeback_rate_proportional << 9 : 0);
 236		bch_hprint(integral,
 237			   wb ? dc->writeback_rate_integral_scaled << 9 : 0);
 238		bch_hprint(change, wb ? dc->writeback_rate_change << 9 : 0);
 239		next_io = wb ? div64_s64(dc->writeback_rate.next-local_clock(),
 240					 NSEC_PER_MSEC) : 0;
 241
 242		return sprintf(buf,
 243			       "rate:\t\t%s/sec\n"
 244			       "dirty:\t\t%s\n"
 245			       "target:\t\t%s\n"
 246			       "proportional:\t%s\n"
 247			       "integral:\t%s\n"
 248			       "change:\t\t%s/sec\n"
 249			       "next io:\t%llims\n",
 250			       rate, dirty, target, proportional,
 251			       integral, change, next_io);
 252	}
 253
 254	sysfs_hprint(dirty_data,
 255		     bcache_dev_sectors_dirty(&dc->disk) << 9);
 256
 257	sysfs_hprint(stripe_size,	 ((uint64_t)dc->disk.stripe_size) << 9);
 258	var_printf(partial_stripes_expensive,	"%u");
 259
 260	var_hprint(sequential_cutoff);
 261
 262	sysfs_print(running,		atomic_read(&dc->running));
 263	sysfs_print(state,		states[BDEV_STATE(&dc->sb)]);
 264
 265	if (attr == &sysfs_label) {
 266		memcpy(buf, dc->sb.label, SB_LABEL_SIZE);
 267		buf[SB_LABEL_SIZE + 1] = '\0';
 268		strcat(buf, "\n");
 269		return strlen(buf);
 270	}
 271
 272	if (attr == &sysfs_backing_dev_name) {
 273		snprintf(buf, BDEVNAME_SIZE + 1, "%pg", dc->bdev);
 274		strcat(buf, "\n");
 275		return strlen(buf);
 276	}
 277
 278	if (attr == &sysfs_backing_dev_uuid) {
 279		/* convert binary uuid into 36-byte string plus '\0' */
 280		snprintf(buf, 36+1, "%pU", dc->sb.uuid);
 281		strcat(buf, "\n");
 282		return strlen(buf);
 283	}
 284
 285#undef var
 286	return 0;
 287}
 288SHOW_LOCKED(bch_cached_dev)
 289
 290STORE(__cached_dev)
 291{
 292	struct cached_dev *dc = container_of(kobj, struct cached_dev,
 293					     disk.kobj);
 294	ssize_t v;
 295	struct cache_set *c;
 296	struct kobj_uevent_env *env;
 297
 298	/* no user space access if system is rebooting */
 299	if (bcache_is_reboot)
 300		return -EBUSY;
 301
 302#define d_strtoul(var)		sysfs_strtoul(var, dc->var)
 303#define d_strtoul_nonzero(var)	sysfs_strtoul_clamp(var, dc->var, 1, INT_MAX)
 304#define d_strtoi_h(var)		sysfs_hatoi(var, dc->var)
 305
 306	sysfs_strtoul(data_csum,	dc->disk.data_csum);
 307	d_strtoul(verify);
 308	sysfs_strtoul_bool(bypass_torture_test, dc->bypass_torture_test);
 309	sysfs_strtoul_bool(writeback_metadata, dc->writeback_metadata);
 310	sysfs_strtoul_bool(writeback_running, dc->writeback_running);
 311	sysfs_strtoul_bool(writeback_consider_fragment, dc->writeback_consider_fragment);
 312	sysfs_strtoul_clamp(writeback_delay, dc->writeback_delay, 0, UINT_MAX);
 313
 314	sysfs_strtoul_clamp(writeback_percent, dc->writeback_percent,
 315			    0, bch_cutoff_writeback);
 316
 317	if (attr == &sysfs_writeback_rate) {
 318		ssize_t ret;
 319		long int v = atomic_long_read(&dc->writeback_rate.rate);
 320
 321		ret = strtoul_safe_clamp(buf, v, 1, INT_MAX);
 322
 323		if (!ret) {
 324			atomic_long_set(&dc->writeback_rate.rate, v);
 325			ret = size;
 326		}
 327
 328		return ret;
 329	}
 330
 331	sysfs_strtoul_clamp(writeback_rate_update_seconds,
 332			    dc->writeback_rate_update_seconds,
 333			    1, WRITEBACK_RATE_UPDATE_SECS_MAX);
 334	sysfs_strtoul_clamp(writeback_rate_i_term_inverse,
 335			    dc->writeback_rate_i_term_inverse,
 336			    1, UINT_MAX);
 337	sysfs_strtoul_clamp(writeback_rate_p_term_inverse,
 338			    dc->writeback_rate_p_term_inverse,
 339			    1, UINT_MAX);
 340	sysfs_strtoul_clamp(writeback_rate_fp_term_low,
 341			    dc->writeback_rate_fp_term_low,
 342			    1, dc->writeback_rate_fp_term_mid - 1);
 343	sysfs_strtoul_clamp(writeback_rate_fp_term_mid,
 344			    dc->writeback_rate_fp_term_mid,
 345			    dc->writeback_rate_fp_term_low + 1,
 346			    dc->writeback_rate_fp_term_high - 1);
 347	sysfs_strtoul_clamp(writeback_rate_fp_term_high,
 348			    dc->writeback_rate_fp_term_high,
 349			    dc->writeback_rate_fp_term_mid + 1, UINT_MAX);
 350	sysfs_strtoul_clamp(writeback_rate_minimum,
 351			    dc->writeback_rate_minimum,
 352			    1, UINT_MAX);
 353
 354	sysfs_strtoul_clamp(io_error_limit, dc->error_limit, 0, INT_MAX);
 355
 356	if (attr == &sysfs_io_disable) {
 357		int v = strtoul_or_return(buf);
 358
 359		dc->io_disable = v ? 1 : 0;
 360	}
 361
 362	sysfs_strtoul_clamp(sequential_cutoff,
 363			    dc->sequential_cutoff,
 364			    0, UINT_MAX);
 365
 366	if (attr == &sysfs_clear_stats)
 367		bch_cache_accounting_clear(&dc->accounting);
 368
 369	if (attr == &sysfs_running &&
 370	    strtoul_or_return(buf)) {
 371		v = bch_cached_dev_run(dc);
 372		if (v)
 373			return v;
 374	}
 375
 376	if (attr == &sysfs_cache_mode) {
 377		v = __sysfs_match_string(bch_cache_modes, -1, buf);
 378		if (v < 0)
 379			return v;
 380
 381		if ((unsigned int) v != BDEV_CACHE_MODE(&dc->sb)) {
 382			SET_BDEV_CACHE_MODE(&dc->sb, v);
 383			bch_write_bdev_super(dc, NULL);
 384		}
 385	}
 386
 387	if (attr == &sysfs_readahead_cache_policy) {
 388		v = __sysfs_match_string(bch_reada_cache_policies, -1, buf);
 389		if (v < 0)
 390			return v;
 391
 392		if ((unsigned int) v != dc->cache_readahead_policy)
 393			dc->cache_readahead_policy = v;
 394	}
 395
 396	if (attr == &sysfs_stop_when_cache_set_failed) {
 397		v = __sysfs_match_string(bch_stop_on_failure_modes, -1, buf);
 398		if (v < 0)
 399			return v;
 400
 401		dc->stop_when_cache_set_failed = v;
 402	}
 403
 404	if (attr == &sysfs_label) {
 405		if (size > SB_LABEL_SIZE)
 406			return -EINVAL;
 407		memcpy(dc->sb.label, buf, size);
 408		if (size < SB_LABEL_SIZE)
 409			dc->sb.label[size] = '\0';
 410		if (size && dc->sb.label[size - 1] == '\n')
 411			dc->sb.label[size - 1] = '\0';
 412		bch_write_bdev_super(dc, NULL);
 413		if (dc->disk.c) {
 414			memcpy(dc->disk.c->uuids[dc->disk.id].label,
 415			       buf, SB_LABEL_SIZE);
 416			bch_uuid_write(dc->disk.c);
 417		}
 418		env = kzalloc(sizeof(struct kobj_uevent_env), GFP_KERNEL);
 419		if (!env)
 420			return -ENOMEM;
 421		add_uevent_var(env, "DRIVER=bcache");
 422		add_uevent_var(env, "CACHED_UUID=%pU", dc->sb.uuid);
 423		add_uevent_var(env, "CACHED_LABEL=%s", buf);
 424		kobject_uevent_env(&disk_to_dev(dc->disk.disk)->kobj,
 425				   KOBJ_CHANGE,
 426				   env->envp);
 427		kfree(env);
 428	}
 429
 430	if (attr == &sysfs_attach) {
 431		uint8_t		set_uuid[16];
 432
 433		if (bch_parse_uuid(buf, set_uuid) < 16)
 434			return -EINVAL;
 435
 436		v = -ENOENT;
 437		list_for_each_entry(c, &bch_cache_sets, list) {
 438			v = bch_cached_dev_attach(dc, c, set_uuid);
 439			if (!v)
 440				return size;
 441		}
 442		if (v == -ENOENT)
 443			pr_err("Can't attach %s: cache set not found\n", buf);
 444		return v;
 445	}
 446
 447	if (attr == &sysfs_detach && dc->disk.c)
 448		bch_cached_dev_detach(dc);
 449
 450	if (attr == &sysfs_stop)
 451		bcache_device_stop(&dc->disk);
 452
 453	return size;
 454}
 455
 456STORE(bch_cached_dev)
 457{
 458	struct cached_dev *dc = container_of(kobj, struct cached_dev,
 459					     disk.kobj);
 460
 461	/* no user space access if system is rebooting */
 462	if (bcache_is_reboot)
 463		return -EBUSY;
 464
 465	mutex_lock(&bch_register_lock);
 466	size = __cached_dev_store(kobj, attr, buf, size);
 467
 468	if (attr == &sysfs_writeback_running) {
 469		/* dc->writeback_running changed in __cached_dev_store() */
 470		if (IS_ERR_OR_NULL(dc->writeback_thread)) {
 471			/*
 472			 * reject setting it to 1 via sysfs if writeback
 473			 * kthread is not created yet.
 474			 */
 475			if (dc->writeback_running) {
 476				dc->writeback_running = false;
 477				pr_err("%s: failed to run non-existent writeback thread\n",
 478						dc->disk.disk->disk_name);
 479			}
 480		} else
 481			/*
 482			 * writeback kthread will check if dc->writeback_running
 483			 * is true or false.
 484			 */
 485			bch_writeback_queue(dc);
 486	}
 487
 488	/*
 489	 * Only set BCACHE_DEV_WB_RUNNING when cached device attached to
 490	 * a cache set, otherwise it doesn't make sense.
 491	 */
 492	if (attr == &sysfs_writeback_percent)
 493		if ((dc->disk.c != NULL) &&
 494		    (!test_and_set_bit(BCACHE_DEV_WB_RUNNING, &dc->disk.flags)))
 495			schedule_delayed_work(&dc->writeback_rate_update,
 496				      dc->writeback_rate_update_seconds * HZ);
 497
 498	mutex_unlock(&bch_register_lock);
 499	return size;
 500}
 501
 502static struct attribute *bch_cached_dev_attrs[] = {
 503	&sysfs_attach,
 504	&sysfs_detach,
 505	&sysfs_stop,
 506#if 0
 507	&sysfs_data_csum,
 508#endif
 509	&sysfs_cache_mode,
 510	&sysfs_readahead_cache_policy,
 511	&sysfs_stop_when_cache_set_failed,
 512	&sysfs_writeback_metadata,
 513	&sysfs_writeback_running,
 514	&sysfs_writeback_delay,
 515	&sysfs_writeback_percent,
 516	&sysfs_writeback_rate,
 517	&sysfs_writeback_consider_fragment,
 518	&sysfs_writeback_rate_update_seconds,
 519	&sysfs_writeback_rate_i_term_inverse,
 520	&sysfs_writeback_rate_p_term_inverse,
 521	&sysfs_writeback_rate_fp_term_low,
 522	&sysfs_writeback_rate_fp_term_mid,
 523	&sysfs_writeback_rate_fp_term_high,
 524	&sysfs_writeback_rate_minimum,
 525	&sysfs_writeback_rate_debug,
 526	&sysfs_io_errors,
 527	&sysfs_io_error_limit,
 528	&sysfs_io_disable,
 529	&sysfs_dirty_data,
 530	&sysfs_stripe_size,
 531	&sysfs_partial_stripes_expensive,
 532	&sysfs_sequential_cutoff,
 533	&sysfs_clear_stats,
 534	&sysfs_running,
 535	&sysfs_state,
 536	&sysfs_label,
 537#ifdef CONFIG_BCACHE_DEBUG
 538	&sysfs_verify,
 539	&sysfs_bypass_torture_test,
 540#endif
 541	&sysfs_backing_dev_name,
 542	&sysfs_backing_dev_uuid,
 543	NULL
 544};
 545ATTRIBUTE_GROUPS(bch_cached_dev);
 546KTYPE(bch_cached_dev);
 547
 548SHOW(bch_flash_dev)
 549{
 550	struct bcache_device *d = container_of(kobj, struct bcache_device,
 551					       kobj);
 552	struct uuid_entry *u = &d->c->uuids[d->id];
 553
 554	sysfs_printf(data_csum,	"%i", d->data_csum);
 555	sysfs_hprint(size,	u->sectors << 9);
 556
 557	if (attr == &sysfs_label) {
 558		memcpy(buf, u->label, SB_LABEL_SIZE);
 559		buf[SB_LABEL_SIZE + 1] = '\0';
 560		strcat(buf, "\n");
 561		return strlen(buf);
 562	}
 563
 564	return 0;
 565}
 566
 567STORE(__bch_flash_dev)
 568{
 569	struct bcache_device *d = container_of(kobj, struct bcache_device,
 570					       kobj);
 571	struct uuid_entry *u = &d->c->uuids[d->id];
 572
 573	/* no user space access if system is rebooting */
 574	if (bcache_is_reboot)
 575		return -EBUSY;
 576
 577	sysfs_strtoul(data_csum,	d->data_csum);
 578
 579	if (attr == &sysfs_size) {
 580		uint64_t v;
 581
 582		strtoi_h_or_return(buf, v);
 583
 584		u->sectors = v >> 9;
 585		bch_uuid_write(d->c);
 586		set_capacity(d->disk, u->sectors);
 587	}
 588
 589	if (attr == &sysfs_label) {
 590		memcpy(u->label, buf, SB_LABEL_SIZE);
 591		bch_uuid_write(d->c);
 592	}
 593
 594	if (attr == &sysfs_unregister) {
 595		set_bit(BCACHE_DEV_DETACHING, &d->flags);
 596		bcache_device_stop(d);
 597	}
 598
 599	return size;
 600}
 601STORE_LOCKED(bch_flash_dev)
 602
 603static struct attribute *bch_flash_dev_attrs[] = {
 604	&sysfs_unregister,
 605#if 0
 606	&sysfs_data_csum,
 607#endif
 608	&sysfs_label,
 609	&sysfs_size,
 610	NULL
 611};
 612ATTRIBUTE_GROUPS(bch_flash_dev);
 613KTYPE(bch_flash_dev);
 614
 615struct bset_stats_op {
 616	struct btree_op op;
 617	size_t nodes;
 618	struct bset_stats stats;
 619};
 620
 621static int bch_btree_bset_stats(struct btree_op *b_op, struct btree *b)
 622{
 623	struct bset_stats_op *op = container_of(b_op, struct bset_stats_op, op);
 624
 625	op->nodes++;
 626	bch_btree_keys_stats(&b->keys, &op->stats);
 627
 628	return MAP_CONTINUE;
 629}
 630
 631static int bch_bset_print_stats(struct cache_set *c, char *buf)
 632{
 633	struct bset_stats_op op;
 634	int ret;
 635
 636	memset(&op, 0, sizeof(op));
 637	bch_btree_op_init(&op.op, -1);
 638
 639	ret = bch_btree_map_nodes(&op.op, c, &ZERO_KEY, bch_btree_bset_stats);
 640	if (ret < 0)
 641		return ret;
 642
 643	return snprintf(buf, PAGE_SIZE,
 644			"btree nodes:		%zu\n"
 645			"written sets:		%zu\n"
 646			"unwritten sets:		%zu\n"
 647			"written key bytes:	%zu\n"
 648			"unwritten key bytes:	%zu\n"
 649			"floats:			%zu\n"
 650			"failed:			%zu\n",
 651			op.nodes,
 652			op.stats.sets_written, op.stats.sets_unwritten,
 653			op.stats.bytes_written, op.stats.bytes_unwritten,
 654			op.stats.floats, op.stats.failed);
 655}
 656
 657static unsigned int bch_root_usage(struct cache_set *c)
 658{
 659	unsigned int bytes = 0;
 660	struct bkey *k;
 661	struct btree *b;
 662	struct btree_iter_stack iter;
 663
 664	goto lock_root;
 665
 666	do {
 667		rw_unlock(false, b);
 668lock_root:
 669		b = c->root;
 670		rw_lock(false, b, b->level);
 671	} while (b != c->root);
 672
 673	for_each_key_filter(&b->keys, k, &iter, bch_ptr_bad)
 674		bytes += bkey_bytes(k);
 675
 676	rw_unlock(false, b);
 677
 678	return (bytes * 100) / btree_bytes(c);
 679}
 680
 681static size_t bch_cache_size(struct cache_set *c)
 682{
 683	size_t ret = 0;
 684	struct btree *b;
 685
 686	mutex_lock(&c->bucket_lock);
 687	list_for_each_entry(b, &c->btree_cache, list)
 688		ret += 1 << (b->keys.page_order + PAGE_SHIFT);
 689
 690	mutex_unlock(&c->bucket_lock);
 691	return ret;
 692}
 693
 694static unsigned int bch_cache_max_chain(struct cache_set *c)
 695{
 696	unsigned int ret = 0;
 697	struct hlist_head *h;
 698
 699	mutex_lock(&c->bucket_lock);
 700
 701	for (h = c->bucket_hash;
 702	     h < c->bucket_hash + (1 << BUCKET_HASH_BITS);
 703	     h++) {
 704		ret = max(ret, hlist_count_nodes(h));
 705	}
 706
 707	mutex_unlock(&c->bucket_lock);
 708	return ret;
 709}
 710
 711static unsigned int bch_btree_used(struct cache_set *c)
 712{
 713	return div64_u64(c->gc_stats.key_bytes * 100,
 714			 (c->gc_stats.nodes ?: 1) * btree_bytes(c));
 715}
 716
 717static unsigned int bch_average_key_size(struct cache_set *c)
 718{
 719	return c->gc_stats.nkeys
 720		? div64_u64(c->gc_stats.data, c->gc_stats.nkeys)
 721		: 0;
 722}
 723
 724SHOW(__bch_cache_set)
 725{
 726	struct cache_set *c = container_of(kobj, struct cache_set, kobj);
 727
 728	sysfs_print(synchronous,		CACHE_SYNC(&c->cache->sb));
 729	sysfs_print(journal_delay_ms,		c->journal_delay_ms);
 730	sysfs_hprint(bucket_size,		bucket_bytes(c->cache));
 731	sysfs_hprint(block_size,		block_bytes(c->cache));
 732	sysfs_print(tree_depth,			c->root->level);
 733	sysfs_print(root_usage_percent,		bch_root_usage(c));
 734
 735	sysfs_hprint(btree_cache_size,		bch_cache_size(c));
 736	sysfs_print(btree_cache_max_chain,	bch_cache_max_chain(c));
 737	sysfs_print(cache_available_percent,	100 - c->gc_stats.in_use);
 738
 739	sysfs_print_time_stats(&c->btree_gc_time,	btree_gc, sec, ms);
 740	sysfs_print_time_stats(&c->btree_split_time,	btree_split, sec, us);
 741	sysfs_print_time_stats(&c->sort.time,		btree_sort, ms, us);
 742	sysfs_print_time_stats(&c->btree_read_time,	btree_read, ms, us);
 743
 744	sysfs_print(btree_used_percent,	bch_btree_used(c));
 745	sysfs_print(btree_nodes,	c->gc_stats.nodes);
 746	sysfs_hprint(average_key_size,	bch_average_key_size(c));
 747
 748	sysfs_print(cache_read_races,
 749		    atomic_long_read(&c->cache_read_races));
 750
 751	sysfs_print(reclaim,
 752		    atomic_long_read(&c->reclaim));
 753
 754	sysfs_print(reclaimed_journal_buckets,
 755		    atomic_long_read(&c->reclaimed_journal_buckets));
 756
 757	sysfs_print(flush_write,
 758		    atomic_long_read(&c->flush_write));
 759
 760	sysfs_print(writeback_keys_done,
 761		    atomic_long_read(&c->writeback_keys_done));
 762	sysfs_print(writeback_keys_failed,
 763		    atomic_long_read(&c->writeback_keys_failed));
 764
 765	if (attr == &sysfs_errors)
 766		return bch_snprint_string_list(buf, PAGE_SIZE, error_actions,
 767					       c->on_error);
 768
 769	/* See count_io_errors for why 88 */
 770	sysfs_print(io_error_halflife,	c->error_decay * 88);
 771	sysfs_print(io_error_limit,	c->error_limit);
 772
 773	sysfs_hprint(congested,
 774		     ((uint64_t) bch_get_congested(c)) << 9);
 775	sysfs_print(congested_read_threshold_us,
 776		    c->congested_read_threshold_us);
 777	sysfs_print(congested_write_threshold_us,
 778		    c->congested_write_threshold_us);
 779
 780	sysfs_print(cutoff_writeback, bch_cutoff_writeback);
 781	sysfs_print(cutoff_writeback_sync, bch_cutoff_writeback_sync);
 782
 783	sysfs_print(active_journal_entries,	fifo_used(&c->journal.pin));
 784	sysfs_printf(verify,			"%i", c->verify);
 785	sysfs_printf(key_merging_disabled,	"%i", c->key_merging_disabled);
 786	sysfs_printf(expensive_debug_checks,
 787		     "%i", c->expensive_debug_checks);
 788	sysfs_printf(gc_always_rewrite,		"%i", c->gc_always_rewrite);
 789	sysfs_printf(btree_shrinker_disabled,	"%i", c->shrinker_disabled);
 790	sysfs_printf(copy_gc_enabled,		"%i", c->copy_gc_enabled);
 791	sysfs_printf(idle_max_writeback_rate,	"%i",
 792		     c->idle_max_writeback_rate_enabled);
 793	sysfs_printf(gc_after_writeback,	"%i", c->gc_after_writeback);
 794	sysfs_printf(io_disable,		"%i",
 795		     test_bit(CACHE_SET_IO_DISABLE, &c->flags));
 796
 797	if (attr == &sysfs_bset_tree_stats)
 798		return bch_bset_print_stats(c, buf);
 799
 800	if (attr == &sysfs_feature_compat)
 801		return bch_print_cache_set_feature_compat(c, buf, PAGE_SIZE);
 802	if (attr == &sysfs_feature_ro_compat)
 803		return bch_print_cache_set_feature_ro_compat(c, buf, PAGE_SIZE);
 804	if (attr == &sysfs_feature_incompat)
 805		return bch_print_cache_set_feature_incompat(c, buf, PAGE_SIZE);
 806
 807	return 0;
 808}
 809SHOW_LOCKED(bch_cache_set)
 810
 811STORE(__bch_cache_set)
 812{
 813	struct cache_set *c = container_of(kobj, struct cache_set, kobj);
 814	ssize_t v;
 815
 816	/* no user space access if system is rebooting */
 817	if (bcache_is_reboot)
 818		return -EBUSY;
 819
 820	if (attr == &sysfs_unregister)
 821		bch_cache_set_unregister(c);
 822
 823	if (attr == &sysfs_stop)
 824		bch_cache_set_stop(c);
 825
 826	if (attr == &sysfs_synchronous) {
 827		bool sync = strtoul_or_return(buf);
 828
 829		if (sync != CACHE_SYNC(&c->cache->sb)) {
 830			SET_CACHE_SYNC(&c->cache->sb, sync);
 831			bcache_write_super(c);
 832		}
 833	}
 834
 835	if (attr == &sysfs_flash_vol_create) {
 836		int r;
 837		uint64_t v;
 838
 839		strtoi_h_or_return(buf, v);
 840
 841		r = bch_flash_dev_create(c, v);
 842		if (r)
 843			return r;
 844	}
 845
 846	if (attr == &sysfs_clear_stats) {
 847		atomic_long_set(&c->writeback_keys_done,	0);
 848		atomic_long_set(&c->writeback_keys_failed,	0);
 849
 850		memset(&c->gc_stats, 0, sizeof(struct gc_stat));
 851		bch_cache_accounting_clear(&c->accounting);
 852	}
 853
 854	if (attr == &sysfs_trigger_gc)
 855		force_wake_up_gc(c);
 856
 857	if (attr == &sysfs_prune_cache) {
 858		struct shrink_control sc;
 859
 860		sc.gfp_mask = GFP_KERNEL;
 861		sc.nr_to_scan = strtoul_or_return(buf);
 862		if (c->shrink)
 863			c->shrink->scan_objects(c->shrink, &sc);
 864	}
 865
 866	sysfs_strtoul_clamp(congested_read_threshold_us,
 867			    c->congested_read_threshold_us,
 868			    0, UINT_MAX);
 869	sysfs_strtoul_clamp(congested_write_threshold_us,
 870			    c->congested_write_threshold_us,
 871			    0, UINT_MAX);
 872
 873	if (attr == &sysfs_errors) {
 874		v = __sysfs_match_string(error_actions, -1, buf);
 875		if (v < 0)
 876			return v;
 877
 878		c->on_error = v;
 879	}
 880
 881	sysfs_strtoul_clamp(io_error_limit, c->error_limit, 0, UINT_MAX);
 882
 883	/* See count_io_errors() for why 88 */
 884	if (attr == &sysfs_io_error_halflife) {
 885		unsigned long v = 0;
 886		ssize_t ret;
 887
 888		ret = strtoul_safe_clamp(buf, v, 0, UINT_MAX);
 889		if (!ret) {
 890			c->error_decay = v / 88;
 891			return size;
 892		}
 893		return ret;
 894	}
 895
 896	if (attr == &sysfs_io_disable) {
 897		v = strtoul_or_return(buf);
 898		if (v) {
 899			if (test_and_set_bit(CACHE_SET_IO_DISABLE,
 900					     &c->flags))
 901				pr_warn("CACHE_SET_IO_DISABLE already set\n");
 902		} else {
 903			if (!test_and_clear_bit(CACHE_SET_IO_DISABLE,
 904						&c->flags))
 905				pr_warn("CACHE_SET_IO_DISABLE already cleared\n");
 906		}
 907	}
 908
 909	sysfs_strtoul_clamp(journal_delay_ms,
 910			    c->journal_delay_ms,
 911			    0, USHRT_MAX);
 912	sysfs_strtoul_bool(verify,		c->verify);
 913	sysfs_strtoul_bool(key_merging_disabled, c->key_merging_disabled);
 914	sysfs_strtoul(expensive_debug_checks,	c->expensive_debug_checks);
 915	sysfs_strtoul_bool(gc_always_rewrite,	c->gc_always_rewrite);
 916	sysfs_strtoul_bool(btree_shrinker_disabled, c->shrinker_disabled);
 917	sysfs_strtoul_bool(copy_gc_enabled,	c->copy_gc_enabled);
 918	sysfs_strtoul_bool(idle_max_writeback_rate,
 919			   c->idle_max_writeback_rate_enabled);
 920
 921	/*
 922	 * write gc_after_writeback here may overwrite an already set
 923	 * BCH_DO_AUTO_GC, it doesn't matter because this flag will be
 924	 * set in next chance.
 925	 */
 926	sysfs_strtoul_clamp(gc_after_writeback, c->gc_after_writeback, 0, 1);
 927
 928	return size;
 929}
 930STORE_LOCKED(bch_cache_set)
 931
 932SHOW(bch_cache_set_internal)
 933{
 934	struct cache_set *c = container_of(kobj, struct cache_set, internal);
 935
 936	return bch_cache_set_show(&c->kobj, attr, buf);
 937}
 938
 939STORE(bch_cache_set_internal)
 940{
 941	struct cache_set *c = container_of(kobj, struct cache_set, internal);
 942
 943	/* no user space access if system is rebooting */
 944	if (bcache_is_reboot)
 945		return -EBUSY;
 946
 947	return bch_cache_set_store(&c->kobj, attr, buf, size);
 948}
 949
 950static void bch_cache_set_internal_release(struct kobject *k)
 951{
 952}
 953
 954static struct attribute *bch_cache_set_attrs[] = {
 955	&sysfs_unregister,
 956	&sysfs_stop,
 957	&sysfs_synchronous,
 958	&sysfs_journal_delay_ms,
 959	&sysfs_flash_vol_create,
 960
 961	&sysfs_bucket_size,
 962	&sysfs_block_size,
 963	&sysfs_tree_depth,
 964	&sysfs_root_usage_percent,
 965	&sysfs_btree_cache_size,
 966	&sysfs_cache_available_percent,
 967
 968	&sysfs_average_key_size,
 969
 970	&sysfs_errors,
 971	&sysfs_io_error_limit,
 972	&sysfs_io_error_halflife,
 973	&sysfs_congested,
 974	&sysfs_congested_read_threshold_us,
 975	&sysfs_congested_write_threshold_us,
 976	&sysfs_clear_stats,
 977	NULL
 978};
 979ATTRIBUTE_GROUPS(bch_cache_set);
 980KTYPE(bch_cache_set);
 981
 982static struct attribute *bch_cache_set_internal_attrs[] = {
 983	&sysfs_active_journal_entries,
 984
 985	sysfs_time_stats_attribute_list(btree_gc, sec, ms)
 986	sysfs_time_stats_attribute_list(btree_split, sec, us)
 987	sysfs_time_stats_attribute_list(btree_sort, ms, us)
 988	sysfs_time_stats_attribute_list(btree_read, ms, us)
 989
 990	&sysfs_btree_nodes,
 991	&sysfs_btree_used_percent,
 992	&sysfs_btree_cache_max_chain,
 993
 994	&sysfs_bset_tree_stats,
 995	&sysfs_cache_read_races,
 996	&sysfs_reclaim,
 997	&sysfs_reclaimed_journal_buckets,
 998	&sysfs_flush_write,
 999	&sysfs_writeback_keys_done,
1000	&sysfs_writeback_keys_failed,
1001
1002	&sysfs_trigger_gc,
1003	&sysfs_prune_cache,
1004#ifdef CONFIG_BCACHE_DEBUG
1005	&sysfs_verify,
1006	&sysfs_key_merging_disabled,
1007	&sysfs_expensive_debug_checks,
1008#endif
1009	&sysfs_gc_always_rewrite,
1010	&sysfs_btree_shrinker_disabled,
1011	&sysfs_copy_gc_enabled,
1012	&sysfs_idle_max_writeback_rate,
1013	&sysfs_gc_after_writeback,
1014	&sysfs_io_disable,
1015	&sysfs_cutoff_writeback,
1016	&sysfs_cutoff_writeback_sync,
1017	&sysfs_feature_compat,
1018	&sysfs_feature_ro_compat,
1019	&sysfs_feature_incompat,
1020	NULL
1021};
1022ATTRIBUTE_GROUPS(bch_cache_set_internal);
1023KTYPE(bch_cache_set_internal);
1024
1025static int __bch_cache_cmp(const void *l, const void *r)
1026{
1027	cond_resched();
1028	return *((uint16_t *)r) - *((uint16_t *)l);
1029}
1030
1031SHOW(__bch_cache)
1032{
1033	struct cache *ca = container_of(kobj, struct cache, kobj);
1034
1035	sysfs_hprint(bucket_size,	bucket_bytes(ca));
1036	sysfs_hprint(block_size,	block_bytes(ca));
1037	sysfs_print(nbuckets,		ca->sb.nbuckets);
1038	sysfs_hprint(written, atomic_long_read(&ca->sectors_written) << 9);
1039	sysfs_hprint(btree_written,
1040		     atomic_long_read(&ca->btree_sectors_written) << 9);
1041	sysfs_hprint(metadata_written,
1042		     (atomic_long_read(&ca->meta_sectors_written) +
1043		      atomic_long_read(&ca->btree_sectors_written)) << 9);
1044
1045	sysfs_print(io_errors,
1046		    atomic_read(&ca->io_errors) >> IO_ERROR_SHIFT);
1047
1048	if (attr == &sysfs_cache_replacement_policy)
1049		return bch_snprint_string_list(buf, PAGE_SIZE,
1050					       cache_replacement_policies,
1051					       CACHE_REPLACEMENT(&ca->sb));
1052
1053	if (attr == &sysfs_priority_stats) {
1054		struct bucket *b;
1055		size_t n = ca->sb.nbuckets, i;
1056		size_t unused = 0, available = 0, dirty = 0, meta = 0;
1057		uint64_t sum = 0;
1058		/* Compute 31 quantiles */
1059		uint16_t q[31], *p, *cached;
1060		ssize_t ret;
1061
1062		cached = p = vmalloc(array_size(sizeof(uint16_t),
1063						ca->sb.nbuckets));
1064		if (!p)
1065			return -ENOMEM;
1066
1067		mutex_lock(&ca->set->bucket_lock);
1068		for_each_bucket(b, ca) {
1069			if (!GC_SECTORS_USED(b))
1070				unused++;
1071			if (GC_MARK(b) == GC_MARK_RECLAIMABLE)
1072				available++;
1073			if (GC_MARK(b) == GC_MARK_DIRTY)
1074				dirty++;
1075			if (GC_MARK(b) == GC_MARK_METADATA)
1076				meta++;
1077		}
1078
1079		for (i = ca->sb.first_bucket; i < n; i++)
1080			p[i] = ca->buckets[i].prio;
1081		mutex_unlock(&ca->set->bucket_lock);
1082
1083		sort(p, n, sizeof(uint16_t), __bch_cache_cmp, NULL);
1084
1085		while (n &&
1086		       !cached[n - 1])
1087			--n;
1088
1089		while (cached < p + n &&
1090		       *cached == BTREE_PRIO) {
1091			cached++;
1092			n--;
1093		}
1094
1095		for (i = 0; i < n; i++)
1096			sum += INITIAL_PRIO - cached[i];
1097
1098		if (n)
1099			sum = div64_u64(sum, n);
1100
1101		for (i = 0; i < ARRAY_SIZE(q); i++)
1102			q[i] = INITIAL_PRIO - cached[n * (i + 1) /
1103				(ARRAY_SIZE(q) + 1)];
1104
1105		vfree(p);
1106
1107		ret = sysfs_emit(buf,
1108				 "Unused:		%zu%%\n"
1109				 "Clean:		%zu%%\n"
1110				 "Dirty:		%zu%%\n"
1111				 "Metadata:	%zu%%\n"
1112				 "Average:	%llu\n"
1113				 "Sectors per Q:	%zu\n"
1114				 "Quantiles:	[",
1115				 unused * 100 / (size_t) ca->sb.nbuckets,
1116				 available * 100 / (size_t) ca->sb.nbuckets,
1117				 dirty * 100 / (size_t) ca->sb.nbuckets,
1118				 meta * 100 / (size_t) ca->sb.nbuckets, sum,
1119				 n * ca->sb.bucket_size / (ARRAY_SIZE(q) + 1));
1120
1121		for (i = 0; i < ARRAY_SIZE(q); i++)
1122			ret += sysfs_emit_at(buf, ret, "%u ", q[i]);
1123		ret--;
1124
1125		ret += sysfs_emit_at(buf, ret, "]\n");
1126
1127		return ret;
1128	}
1129
1130	return 0;
1131}
1132SHOW_LOCKED(bch_cache)
1133
1134STORE(__bch_cache)
1135{
1136	struct cache *ca = container_of(kobj, struct cache, kobj);
1137	ssize_t v;
1138
1139	/* no user space access if system is rebooting */
1140	if (bcache_is_reboot)
1141		return -EBUSY;
1142
1143	if (attr == &sysfs_cache_replacement_policy) {
1144		v = __sysfs_match_string(cache_replacement_policies, -1, buf);
1145		if (v < 0)
1146			return v;
1147
1148		if ((unsigned int) v != CACHE_REPLACEMENT(&ca->sb)) {
1149			mutex_lock(&ca->set->bucket_lock);
1150			SET_CACHE_REPLACEMENT(&ca->sb, v);
1151			mutex_unlock(&ca->set->bucket_lock);
1152
1153			bcache_write_super(ca->set);
1154		}
1155	}
1156
1157	if (attr == &sysfs_clear_stats) {
1158		atomic_long_set(&ca->sectors_written, 0);
1159		atomic_long_set(&ca->btree_sectors_written, 0);
1160		atomic_long_set(&ca->meta_sectors_written, 0);
1161		atomic_set(&ca->io_count, 0);
1162		atomic_set(&ca->io_errors, 0);
1163	}
1164
1165	return size;
1166}
1167STORE_LOCKED(bch_cache)
1168
1169static struct attribute *bch_cache_attrs[] = {
1170	&sysfs_bucket_size,
1171	&sysfs_block_size,
1172	&sysfs_nbuckets,
1173	&sysfs_priority_stats,
1174	&sysfs_written,
1175	&sysfs_btree_written,
1176	&sysfs_metadata_written,
1177	&sysfs_io_errors,
1178	&sysfs_clear_stats,
1179	&sysfs_cache_replacement_policy,
1180	NULL
1181};
1182ATTRIBUTE_GROUPS(bch_cache);
1183KTYPE(bch_cache);