+// SPDX-License-Identifier: GPL-2.0
/*
* bcache sysfs interfaces
*
#ifndef NO_BCACHEFS_SYSFS
#include "bcachefs.h"
-#include "alloc.h"
-#include "compress.h"
+#include "alloc_background.h"
#include "sysfs.h"
#include "btree_cache.h"
#include "btree_io.h"
#include "btree_update_interior.h"
#include "btree_gc.h"
#include "buckets.h"
+#include "clock.h"
#include "disk_groups.h"
+#include "ec.h"
#include "inode.h"
#include "journal.h"
#include "keylist.h"
#include "move.h"
#include "opts.h"
+#include "rebalance.h"
#include "replicas.h"
#include "super-io.h"
-#include "tier.h"
+#include "tests.h"
#include <linux/blkdev.h>
#include <linux/sort.h>
#define sysfs_hprint(file, val) \
do { \
if (attr == &sysfs_ ## file) { \
- ssize_t ret = bch2_hprint(buf, val); \
- strcat(buf, "\n"); \
- return ret + 1; \
+ struct printbuf out = _PBUF(buf, PAGE_SIZE); \
+ bch2_hprint(&out, val); \
+ pr_buf(&out, "\n"); \
+ return out.pos - buf; \
} \
} while (0)
read_attribute(journal_pins);
read_attribute(btree_updates);
read_attribute(dirty_btree_nodes);
+read_attribute(btree_transactions);
read_attribute(internal_uuid);
sysfs_pd_controller_attribute(copy_gc);
rw_attribute(rebalance_enabled);
-rw_attribute(rebalance_percent);
sysfs_pd_controller_attribute(rebalance);
+read_attribute(rebalance_work);
rw_attribute(promote_whole_extents);
+read_attribute(new_stripes);
+
rw_attribute(pd_controllers_update_seconds);
read_attribute(meta_replicas_have);
read_attribute(data_replicas_have);
+read_attribute(io_timers_read);
+read_attribute(io_timers_write);
+
+#ifdef CONFIG_BCACHEFS_TESTS
+write_attribute(perf_test);
+#endif /* CONFIG_BCACHEFS_TESTS */
+
#define BCH_DEBUG_PARAM(name, description) \
rw_attribute(name);
BCH_DEBUG_PARAMS()
#undef BCH_DEBUG_PARAM
-#define BCH_TIME_STAT(_name) \
+#define x(_name) \
static struct attribute sysfs_time_stat_##_name = \
{ .name = #_name, .mode = S_IRUGO };
BCH_TIME_STATS()
-#undef BCH_TIME_STAT
+#undef x
static struct attribute sysfs_state_rw = {
.name = "state",
static ssize_t show_fs_alloc_debug(struct bch_fs *c, char *buf)
{
- struct bch_fs_usage stats = bch2_fs_usage_read(c);
+ struct printbuf out = _PBUF(buf, PAGE_SIZE);
+ struct bch_fs_usage *fs_usage = bch2_fs_usage_read(c);
- return scnprintf(buf, PAGE_SIZE,
- "capacity:\t\t%llu\n"
- "1 replicas:\n"
- "\tmeta:\t\t%llu\n"
- "\tdirty:\t\t%llu\n"
- "\treserved:\t%llu\n"
- "2 replicas:\n"
- "\tmeta:\t\t%llu\n"
- "\tdirty:\t\t%llu\n"
- "\treserved:\t%llu\n"
- "3 replicas:\n"
- "\tmeta:\t\t%llu\n"
- "\tdirty:\t\t%llu\n"
- "\treserved:\t%llu\n"
- "4 replicas:\n"
- "\tmeta:\t\t%llu\n"
- "\tdirty:\t\t%llu\n"
- "\treserved:\t%llu\n"
- "online reserved:\t%llu\n",
- c->capacity,
- stats.s[0].data[S_META],
- stats.s[0].data[S_DIRTY],
- stats.s[0].persistent_reserved,
- stats.s[1].data[S_META],
- stats.s[1].data[S_DIRTY],
- stats.s[1].persistent_reserved,
- stats.s[2].data[S_META],
- stats.s[2].data[S_DIRTY],
- stats.s[2].persistent_reserved,
- stats.s[3].data[S_META],
- stats.s[3].data[S_DIRTY],
- stats.s[3].persistent_reserved,
- stats.online_reserved);
+ if (!fs_usage)
+ return -ENOMEM;
+
+ bch2_fs_usage_to_text(&out, c, fs_usage);
+
+ percpu_up_read(&c->mark_lock);
+
+ kfree(fs_usage);
+
+ return out.pos - buf;
}
static ssize_t bch2_compression_stats(struct bch_fs *c, char *buf)
{
- struct btree_iter iter;
+ struct btree_trans trans;
+ struct btree_iter *iter;
struct bkey_s_c k;
u64 nr_uncompressed_extents = 0, uncompressed_sectors = 0,
nr_compressed_extents = 0,
compressed_sectors_compressed = 0,
compressed_sectors_uncompressed = 0;
+ int ret;
- if (!bch2_fs_running(c))
+ if (!test_bit(BCH_FS_STARTED, &c->flags))
return -EPERM;
- for_each_btree_key(&iter, c, BTREE_ID_EXTENTS, POS_MIN, 0, k)
- if (k.k->type == BCH_EXTENT) {
+ bch2_trans_init(&trans, c, 0, 0);
+
+ for_each_btree_key(&trans, iter, BTREE_ID_EXTENTS, POS_MIN, 0, k, ret)
+ if (k.k->type == KEY_TYPE_extent) {
struct bkey_s_c_extent e = bkey_s_c_to_extent(k);
- const struct bch_extent_ptr *ptr;
- struct bch_extent_crc_unpacked crc;
+ const union bch_extent_entry *entry;
+ struct extent_ptr_decoded p;
- extent_for_each_ptr_crc(e, ptr, crc) {
- if (crc.compression_type == BCH_COMPRESSION_NONE) {
+ extent_for_each_ptr_decode(e, p, entry) {
+ if (!crc_is_compressed(p.crc)) {
nr_uncompressed_extents++;
uncompressed_sectors += e.k->size;
} else {
nr_compressed_extents++;
compressed_sectors_compressed +=
- crc.compressed_size;
+ p.crc.compressed_size;
compressed_sectors_uncompressed +=
- crc.uncompressed_size;
+ p.crc.uncompressed_size;
}
/* only looking at the first ptr */
break;
}
}
- bch2_btree_iter_unlock(&iter);
+
+ ret = bch2_trans_exit(&trans) ?: ret;
+ if (ret)
+ return ret;
return scnprintf(buf, PAGE_SIZE,
"uncompressed data:\n"
compressed_sectors_uncompressed << 9);
}
+static ssize_t bch2_new_stripes(struct bch_fs *c, char *buf)
+{
+ char *out = buf, *end = buf + PAGE_SIZE;
+ struct ec_stripe_head *h;
+ struct ec_stripe_new *s;
+
+ mutex_lock(&c->ec_new_stripe_lock);
+ list_for_each_entry(h, &c->ec_new_stripe_list, list) {
+ out += scnprintf(out, end - out,
+ "target %u algo %u redundancy %u:\n",
+ h->target, h->algo, h->redundancy);
+
+ if (h->s)
+ out += scnprintf(out, end - out,
+ "\tpending: blocks %u allocated %u\n",
+ h->s->blocks.nr,
+ bitmap_weight(h->s->blocks_allocated,
+ h->s->blocks.nr));
+
+ mutex_lock(&h->lock);
+ list_for_each_entry(s, &h->stripes, list)
+ out += scnprintf(out, end - out,
+ "\tin flight: blocks %u allocated %u pin %u\n",
+ s->blocks.nr,
+ bitmap_weight(s->blocks_allocated,
+ s->blocks.nr),
+ atomic_read(&s->pin));
+ mutex_unlock(&h->lock);
+
+ }
+ mutex_unlock(&c->ec_new_stripe_lock);
+
+ return out - buf;
+}
+
SHOW(bch2_fs)
{
struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
sysfs_print(pd_controllers_update_seconds,
c->pd_controllers_update_seconds);
- sysfs_printf(rebalance_enabled, "%i", c->rebalance_enabled);
- sysfs_print(rebalance_percent, c->rebalance_percent);
- sysfs_pd_controller_show(rebalance, &c->rebalance_pd); /* XXX */
+ sysfs_printf(rebalance_enabled, "%i", c->rebalance.enabled);
+ sysfs_pd_controller_show(rebalance, &c->rebalance.pd); /* XXX */
+
+ if (attr == &sysfs_rebalance_work)
+ return bch2_rebalance_work_show(c, buf);
sysfs_print(promote_whole_extents, c->promote_whole_extents);
- sysfs_printf(meta_replicas_have, "%u", bch2_replicas_online(c, true));
- sysfs_printf(data_replicas_have, "%u", bch2_replicas_online(c, false));
+ sysfs_printf(meta_replicas_have, "%i", bch2_replicas_online(c, true));
+ sysfs_printf(data_replicas_have, "%i", bch2_replicas_online(c, false));
/* Debugging: */
if (attr == &sysfs_dirty_btree_nodes)
return bch2_dirty_btree_nodes_print(c, buf);
+ if (attr == &sysfs_btree_transactions) {
+ struct printbuf out = _PBUF(buf, PAGE_SIZE);
+
+ bch2_btree_trans_to_text(&out, c);
+ return out.pos - buf;
+ }
if (attr == &sysfs_compression_stats)
return bch2_compression_stats(c, buf);
+ if (attr == &sysfs_new_stripes)
+ return bch2_new_stripes(c, buf);
+
+ if (attr == &sysfs_io_timers_read)
+ return bch2_io_timers_show(&c->io_clock[READ], buf);
+ if (attr == &sysfs_io_timers_write)
+ return bch2_io_timers_show(&c->io_clock[WRITE], buf);
+
#define BCH_DEBUG_PARAM(name, description) sysfs_print(name, c->name);
BCH_DEBUG_PARAMS()
#undef BCH_DEBUG_PARAM
return 0;
}
-STORE(__bch2_fs)
+STORE(bch2_fs)
{
struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
}
if (attr == &sysfs_rebalance_enabled) {
- ssize_t ret = strtoul_safe(buf, c->rebalance_enabled)
+ ssize_t ret = strtoul_safe(buf, c->rebalance.enabled)
?: (ssize_t) size;
rebalance_wakeup(c);
sysfs_strtoul(pd_controllers_update_seconds,
c->pd_controllers_update_seconds);
-
- sysfs_strtoul(rebalance_percent, c->rebalance_percent);
- sysfs_pd_controller_store(rebalance, &c->rebalance_pd);
+ sysfs_pd_controller_store(rebalance, &c->rebalance.pd);
sysfs_strtoul(promote_whole_extents, c->promote_whole_extents);
BCH_DEBUG_PARAMS()
#undef BCH_DEBUG_PARAM
- if (!bch2_fs_running(c))
+ if (!test_bit(BCH_FS_STARTED, &c->flags))
return -EPERM;
/* Debugging: */
if (attr == &sysfs_trigger_btree_coalesce)
bch2_coalesce(c);
- if (attr == &sysfs_trigger_gc)
- bch2_gc(c);
+ if (attr == &sysfs_trigger_gc) {
+ /*
+ * Full gc is currently incompatible with btree key cache:
+ */
+#if 0
+ down_read(&c->state_lock);
+ bch2_gc(c, NULL, false, false);
+ up_read(&c->state_lock);
+#else
+ bch2_gc_gens(c);
+#endif
+ }
if (attr == &sysfs_prune_cache) {
struct shrink_control sc;
c->btree_cache.shrink.scan_objects(&c->btree_cache.shrink, &sc);
}
- return size;
-}
-
-STORE(bch2_fs)
-{
- struct bch_fs *c = container_of(kobj, struct bch_fs, kobj);
-
- mutex_lock(&c->state_lock);
- size = __bch2_fs_store(kobj, attr, buf, size);
- mutex_unlock(&c->state_lock);
-
+#ifdef CONFIG_BCACHEFS_TESTS
+ if (attr == &sysfs_perf_test) {
+ char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
+ char *test = strsep(&p, " \t\n");
+ char *nr_str = strsep(&p, " \t\n");
+ char *threads_str = strsep(&p, " \t\n");
+ unsigned threads;
+ u64 nr;
+ int ret = -EINVAL;
+
+ if (threads_str &&
+ !(ret = kstrtouint(threads_str, 10, &threads)) &&
+ !(ret = bch2_strtoull_h(nr_str, &nr)))
+ bch2_btree_perf_test(c, test, nr, threads);
+ else
+ size = ret;
+ kfree(tmp);
+ }
+#endif
return size;
}
SYSFS_OPS(bch2_fs);
&sysfs_journal_write_delay_ms,
&sysfs_journal_reclaim_delay_ms,
- &sysfs_rebalance_percent,
&sysfs_promote_whole_extents,
&sysfs_compression_stats,
+
+#ifdef CONFIG_BCACHEFS_TESTS
+ &sysfs_perf_test,
+#endif
NULL
};
&sysfs_journal_pins,
&sysfs_btree_updates,
&sysfs_dirty_btree_nodes,
+ &sysfs_btree_transactions,
&sysfs_read_realloc_races,
&sysfs_extent_migrate_done,
&sysfs_prune_cache,
&sysfs_copy_gc_enabled,
+
&sysfs_rebalance_enabled,
+ &sysfs_rebalance_work,
sysfs_pd_controller_files(rebalance),
+
+ &sysfs_new_stripes,
+
+ &sysfs_io_timers_read,
+ &sysfs_io_timers_write,
+
&sysfs_internal_uuid,
#define BCH_DEBUG_PARAM(name, description) &sysfs_##name,
SHOW(bch2_fs_opts_dir)
{
- char *out = buf, *end = buf + PAGE_SIZE;
+ struct printbuf out = _PBUF(buf, PAGE_SIZE);
struct bch_fs *c = container_of(kobj, struct bch_fs, opts_dir);
const struct bch_option *opt = container_of(attr, struct bch_option, attr);
int id = opt - bch2_opt_table;
u64 v = bch2_opt_get_by_id(&c->opts, id);
- out += bch2_opt_to_text(c, out, end - out, opt, v, OPT_SHOW_FULL_LIST);
- out += scnprintf(out, end - out, "\n");
+ bch2_opt_to_text(&out, c, opt, v, OPT_SHOW_FULL_LIST);
+ pr_buf(&out, "\n");
- return out - buf;
+ return out.pos - buf;
}
STORE(bch2_fs_opts_dir)
if (ret < 0)
return ret;
- if (id == Opt_compression ||
- id == Opt_background_compression) {
- int ret = bch2_check_set_has_compressed_data(c, v);
- if (ret) {
- mutex_unlock(&c->sb_lock);
- return ret;
- }
- }
+ ret = bch2_opt_check_may_set(c, id, v);
+ if (ret < 0)
+ return ret;
if (opt->set_sb != SET_NO_SB_OPT) {
mutex_lock(&c->sb_lock);
for (i = bch2_opt_table;
i < bch2_opt_table + bch2_opts_nr;
i++) {
- if (i->mode == OPT_INTERNAL)
+ if (!(i->mode & (OPT_FORMAT|OPT_MOUNT|OPT_RUNTIME)))
continue;
ret = sysfs_create_file(kobj, &i->attr);
{
struct bch_fs *c = container_of(kobj, struct bch_fs, time_stats);
-#define BCH_TIME_STAT(name) \
+#define x(name) \
if (attr == &sysfs_time_stat_##name) \
- return bch2_time_stats_print(&c->name##_time, buf, PAGE_SIZE);
+ return bch2_time_stats_print(&c->times[BCH_TIME_##name],\
+ buf, PAGE_SIZE);
BCH_TIME_STATS()
-#undef BCH_TIME_STAT
+#undef x
return 0;
}
SYSFS_OPS(bch2_fs_time_stats);
struct attribute *bch2_fs_time_stats_files[] = {
-#define BCH_TIME_STAT(name) \
+#define x(name) \
&sysfs_time_stat_##name,
BCH_TIME_STATS()
-#undef BCH_TIME_STAT
+#undef x
NULL
};
static int unsigned_cmp(const void *_l, const void *_r)
{
- unsigned l = *((unsigned *) _l);
- unsigned r = *((unsigned *) _r);
+ const unsigned *l = _l;
+ const unsigned *r = _r;
- return (l > r) - (l < r);
+ return cmp_int(*l, *r);
}
static ssize_t show_quantiles(struct bch_fs *c, struct bch_dev *ca,
static ssize_t show_reserve_stats(struct bch_dev *ca, char *buf)
{
+ struct printbuf out = _PBUF(buf, PAGE_SIZE);
enum alloc_reserve i;
- ssize_t ret;
- spin_lock(&ca->freelist_lock);
+ spin_lock(&ca->fs->freelist_lock);
- ret = scnprintf(buf, PAGE_SIZE,
- "free_inc:\t%zu\t%zu\n",
- fifo_used(&ca->free_inc),
- ca->free_inc.size);
+ pr_buf(&out, "free_inc:\t%zu\t%zu\n",
+ fifo_used(&ca->free_inc),
+ ca->free_inc.size);
for (i = 0; i < RESERVE_NR; i++)
- ret += scnprintf(buf + ret, PAGE_SIZE - ret,
- "free[%u]:\t%zu\t%zu\n", i,
- fifo_used(&ca->free[i]),
- ca->free[i].size);
+ pr_buf(&out, "free[%u]:\t%zu\t%zu\n", i,
+ fifo_used(&ca->free[i]),
+ ca->free[i].size);
- spin_unlock(&ca->freelist_lock);
+ spin_unlock(&ca->fs->freelist_lock);
- return ret;
+ return out.pos - buf;
}
static ssize_t show_dev_alloc_debug(struct bch_dev *ca, char *buf)
{
struct bch_fs *c = ca->fs;
struct bch_dev_usage stats = bch2_dev_usage_read(c, ca);
+ unsigned i, nr[BCH_DATA_NR];
+
+ memset(nr, 0, sizeof(nr));
+
+ for (i = 0; i < ARRAY_SIZE(c->open_buckets); i++)
+ nr[c->open_buckets[i].type]++;
return scnprintf(buf, PAGE_SIZE,
"free_inc: %zu/%zu\n"
" meta: %llu\n"
" user: %llu\n"
" cached: %llu\n"
- " available: %llu\n"
+ " erasure coded: %llu\n"
+ " available: %lli\n"
"sectors:\n"
" sb: %llu\n"
" journal: %llu\n"
" meta: %llu\n"
" user: %llu\n"
" cached: %llu\n"
+ " fragmented: %llu\n"
+ " copygc threshold: %llu\n"
"freelist_wait: %s\n"
"open buckets: %u/%u (reserved %u)\n"
- "open_buckets_wait: %s\n",
+ "open_buckets_wait: %s\n"
+ "open_buckets_btree: %u\n"
+ "open_buckets_user: %u\n"
+ "btree reserve cache: %u\n",
fifo_used(&ca->free_inc), ca->free_inc.size,
fifo_used(&ca->free[RESERVE_BTREE]), ca->free[RESERVE_BTREE].size,
fifo_used(&ca->free[RESERVE_MOVINGGC]), ca->free[RESERVE_MOVINGGC].size,
stats.buckets[BCH_DATA_BTREE],
stats.buckets[BCH_DATA_USER],
stats.buckets[BCH_DATA_CACHED],
- __dev_buckets_available(ca, stats),
+ stats.buckets_ec,
+ ca->mi.nbuckets - ca->mi.first_bucket - stats.buckets_unavailable,
stats.sectors[BCH_DATA_SB],
stats.sectors[BCH_DATA_JOURNAL],
stats.sectors[BCH_DATA_BTREE],
stats.sectors[BCH_DATA_USER],
stats.sectors[BCH_DATA_CACHED],
+ stats.sectors_fragmented,
+ ca->copygc_threshold,
c->freelist_wait.list.first ? "waiting" : "empty",
- c->open_buckets_nr_free, OPEN_BUCKETS_COUNT, BTREE_NODE_RESERVE,
- c->open_buckets_wait.list.first ? "waiting" : "empty");
+ c->open_buckets_nr_free, OPEN_BUCKETS_COUNT,
+ BTREE_NODE_OPEN_BUCKET_RESERVE,
+ c->open_buckets_wait.list.first ? "waiting" : "empty",
+ nr[BCH_DATA_BTREE],
+ nr[BCH_DATA_USER],
+ c->btree_reserve_cache_nr);
}
static const char * const bch2_rw[] = {
static ssize_t show_dev_iodone(struct bch_dev *ca, char *buf)
{
- char *out = buf, *end = buf + PAGE_SIZE;
- int rw, i, cpu;
+ struct printbuf out = _PBUF(buf, PAGE_SIZE);
+ int rw, i;
for (rw = 0; rw < 2; rw++) {
- out += scnprintf(out, end - out, "%s:\n", bch2_rw[rw]);
-
- for (i = 1; i < BCH_DATA_NR; i++) {
- u64 n = 0;
-
- for_each_possible_cpu(cpu)
- n += per_cpu_ptr(ca->io_done, cpu)->sectors[rw][i];
+ pr_buf(&out, "%s:\n", bch2_rw[rw]);
- out += scnprintf(out, end - out, "%-12s:%12llu\n",
- bch2_data_types[i], n << 9);
- }
+ for (i = 1; i < BCH_DATA_NR; i++)
+ pr_buf(&out, "%-12s:%12llu\n",
+ bch2_data_types[i],
+ percpu_u64_get(&ca->io_done->sectors[rw][i]) << 9);
}
- return out - buf;
+ return out.pos - buf;
}
SHOW(bch2_dev)
{
struct bch_dev *ca = container_of(kobj, struct bch_dev, kobj);
struct bch_fs *c = ca->fs;
- char *out = buf, *end = buf + PAGE_SIZE;
+ struct printbuf out = _PBUF(buf, PAGE_SIZE);
sysfs_printf(uuid, "%pU\n", ca->uuid.b);
if (attr == &sysfs_label) {
if (ca->mi.group) {
mutex_lock(&c->sb_lock);
- out += bch2_disk_path_print(&c->disk_sb, out, end - out,
- ca->mi.group - 1);
+ bch2_disk_path_to_text(&out, &c->disk_sb,
+ ca->mi.group - 1);
mutex_unlock(&c->sb_lock);
- } else {
- out += scnprintf(out, end - out, "none");
}
- out += scnprintf(out, end - out, "\n");
- return out - buf;
+ pr_buf(&out, "\n");
+ return out.pos - buf;
}
if (attr == &sysfs_has_data) {
- out += bch2_scnprint_flag_list(out, end - out,
- bch2_data_types,
- bch2_dev_has_data(c, ca));
- out += scnprintf(out, end - out, "\n");
- return out - buf;
+ bch2_flags_to_text(&out, bch2_data_types,
+ bch2_dev_has_data(c, ca));
+ pr_buf(&out, "\n");
+ return out.pos - buf;
}
sysfs_pd_controller_show(copy_gc, &ca->copygc_pd);
if (attr == &sysfs_cache_replacement_policy) {
- out += bch2_scnprint_string_list(out, end - out,
- bch2_cache_replacement_policies,
- ca->mi.replacement);
- out += scnprintf(out, end - out, "\n");
- return out - buf;
+ bch2_string_opt_to_text(&out,
+ bch2_cache_replacement_policies,
+ ca->mi.replacement);
+ pr_buf(&out, "\n");
+ return out.pos - buf;
}
if (attr == &sysfs_state_rw) {
- out += bch2_scnprint_string_list(out, end - out,
- bch2_dev_state,
- ca->mi.state);
- out += scnprintf(out, end - out, "\n");
- return out - buf;
+ bch2_string_opt_to_text(&out, bch2_dev_state,
+ ca->mi.state);
+ pr_buf(&out, "\n");
+ return out.pos - buf;
}
if (attr == &sysfs_iodone)
}
if (attr == &sysfs_cache_replacement_policy) {
- ssize_t v = bch2_read_string_list(buf, bch2_cache_replacement_policies);
+ ssize_t v = __sysfs_match_string(bch2_cache_replacement_policies, -1, buf);
if (v < 0)
return v;
projects / bcachefs-tools-debian / blobdiff