linux/fs/bcachefs/move.c
Kent Overstreet 7ef2a73a58 bcachefs: Fix check for if extent update is allocating
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2023-10-22 17:08:14 -04:00

808 lines
20 KiB
C

// SPDX-License-Identifier: GPL-2.0
#include "bcachefs.h"
#include "alloc_foreground.h"
#include "btree_gc.h"
#include "btree_update.h"
#include "btree_update_interior.h"
#include "buckets.h"
#include "disk_groups.h"
#include "inode.h"
#include "io.h"
#include "journal_reclaim.h"
#include "keylist.h"
#include "move.h"
#include "replicas.h"
#include "super-io.h"
#include "trace.h"
#include <linux/ioprio.h>
#include <linux/kthread.h>
#define SECTORS_IN_FLIGHT_PER_DEVICE 2048
struct moving_io {
struct list_head list;
struct closure cl;
bool read_completed;
unsigned read_sectors;
unsigned write_sectors;
struct bch_read_bio rbio;
struct migrate_write write;
/* Must be last since it is variable size */
struct bio_vec bi_inline_vecs[0];
};
struct moving_context {
/* Closure for waiting on all reads and writes to complete */
struct closure cl;
struct bch_move_stats *stats;
struct list_head reads;
/* in flight sectors: */
atomic_t read_sectors;
atomic_t write_sectors;
wait_queue_head_t wait;
};
static int bch2_migrate_index_update(struct bch_write_op *op)
{
struct bch_fs *c = op->c;
struct migrate_write *m =
container_of(op, struct migrate_write, op);
struct keylist *keys = &op->insert_keys;
struct btree_iter iter;
int ret = 0;
bch2_btree_iter_init(&iter, c, BTREE_ID_EXTENTS,
bkey_start_pos(&bch2_keylist_front(keys)->k),
BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
while (1) {
struct bkey_s_c k = bch2_btree_iter_peek_slot(&iter);
struct bkey_i_extent *insert, *new =
bkey_i_to_extent(bch2_keylist_front(keys));
BKEY_PADDED(k) _new, _insert;
const union bch_extent_entry *entry;
struct extent_ptr_decoded p;
bool did_work = false;
int nr;
if (btree_iter_err(k)) {
ret = bch2_btree_iter_unlock(&iter);
break;
}
if (bversion_cmp(k.k->version, new->k.version) ||
!bkey_extent_is_data(k.k) ||
!bch2_extent_matches_ptr(c, bkey_s_c_to_extent(k),
m->ptr, m->offset))
goto nomatch;
if (m->data_cmd == DATA_REWRITE &&
!bch2_extent_has_device(bkey_s_c_to_extent(k),
m->data_opts.rewrite_dev))
goto nomatch;
bkey_reassemble(&_insert.k, k);
insert = bkey_i_to_extent(&_insert.k);
bkey_copy(&_new.k, bch2_keylist_front(keys));
new = bkey_i_to_extent(&_new.k);
bch2_cut_front(iter.pos, &insert->k_i);
bch2_cut_back(new->k.p, &insert->k);
bch2_cut_back(insert->k.p, &new->k);
if (m->data_cmd == DATA_REWRITE)
bch2_bkey_drop_device(extent_i_to_s(insert).s,
m->data_opts.rewrite_dev);
extent_for_each_ptr_decode(extent_i_to_s(new), p, entry) {
if (bch2_extent_has_device(extent_i_to_s_c(insert), p.ptr.dev)) {
/*
* raced with another move op? extent already
* has a pointer to the device we just wrote
* data to
*/
continue;
}
bch2_extent_ptr_decoded_append(insert, &p);
did_work = true;
}
if (!did_work)
goto nomatch;
bch2_extent_narrow_crcs(insert,
(struct bch_extent_crc_unpacked) { 0 });
bch2_extent_normalize(c, extent_i_to_s(insert).s);
bch2_extent_mark_replicas_cached(c, extent_i_to_s(insert),
op->opts.background_target,
op->opts.data_replicas);
/*
* If we're not fully overwriting @k, and it's compressed, we
* need a reservation for all the pointers in @insert
*/
nr = bch2_bkey_nr_dirty_ptrs(bkey_i_to_s_c(&insert->k_i)) -
m->nr_ptrs_reserved;
if (insert->k.size < k.k->size &&
bch2_extent_is_compressed(k) &&
nr > 0) {
/*
* can't call bch2_disk_reservation_add() with btree
* locks held, at least not without a song and dance
*/
bch2_btree_iter_unlock(&iter);
ret = bch2_disk_reservation_add(c, &op->res,
keylist_sectors(keys) * nr, 0);
if (ret)
goto out;
m->nr_ptrs_reserved += nr;
goto next;
}
ret = bch2_btree_insert_at(c, &op->res,
op_journal_seq(op),
BTREE_INSERT_ATOMIC|
BTREE_INSERT_NOFAIL|
BTREE_INSERT_USE_RESERVE|
m->data_opts.btree_insert_flags,
BTREE_INSERT_ENTRY(&iter, &insert->k_i));
if (!ret)
atomic_long_inc(&c->extent_migrate_done);
if (ret == -EINTR)
ret = 0;
if (ret)
break;
next:
while (bkey_cmp(iter.pos, bch2_keylist_front(keys)->k.p) >= 0) {
bch2_keylist_pop_front(keys);
if (bch2_keylist_empty(keys))
goto out;
}
bch2_cut_front(iter.pos, bch2_keylist_front(keys));
continue;
nomatch:
if (m->ctxt)
atomic64_add(k.k->p.offset - iter.pos.offset,
&m->ctxt->stats->sectors_raced);
atomic_long_inc(&c->extent_migrate_raced);
trace_move_race(&new->k);
bch2_btree_iter_next_slot(&iter);
goto next;
}
out:
bch2_btree_iter_unlock(&iter);
return ret;
}
void bch2_migrate_read_done(struct migrate_write *m, struct bch_read_bio *rbio)
{
/* write bio must own pages: */
BUG_ON(!m->op.wbio.bio.bi_vcnt);
m->ptr = rbio->pick.ptr;
m->offset = rbio->pos.offset - rbio->pick.crc.offset;
m->op.devs_have = rbio->devs_have;
m->op.pos = rbio->pos;
m->op.version = rbio->version;
m->op.crc = rbio->pick.crc;
m->op.wbio.bio.bi_iter.bi_size = m->op.crc.compressed_size << 9;
if (bch2_csum_type_is_encryption(m->op.crc.csum_type)) {
m->op.nonce = m->op.crc.nonce + m->op.crc.offset;
m->op.csum_type = m->op.crc.csum_type;
}
if (m->data_cmd == DATA_REWRITE)
bch2_dev_list_drop_dev(&m->op.devs_have, m->data_opts.rewrite_dev);
}
int bch2_migrate_write_init(struct bch_fs *c, struct migrate_write *m,
struct write_point_specifier wp,
struct bch_io_opts io_opts,
enum data_cmd data_cmd,
struct data_opts data_opts,
struct bkey_s_c k)
{
int ret;
m->data_cmd = data_cmd;
m->data_opts = data_opts;
m->nr_ptrs_reserved = 0;
bch2_write_op_init(&m->op, c, io_opts);
m->op.compression_type =
bch2_compression_opt_to_type[io_opts.background_compression ?:
io_opts.compression];
m->op.target = data_opts.target,
m->op.write_point = wp;
if (m->data_opts.btree_insert_flags & BTREE_INSERT_USE_RESERVE)
m->op.alloc_reserve = RESERVE_MOVINGGC;
m->op.flags |= BCH_WRITE_ONLY_SPECIFIED_DEVS|
BCH_WRITE_PAGES_STABLE|
BCH_WRITE_PAGES_OWNED|
BCH_WRITE_DATA_ENCODED;
m->op.nr_replicas = 1;
m->op.nr_replicas_required = 1;
m->op.index_update_fn = bch2_migrate_index_update;
switch (data_cmd) {
case DATA_ADD_REPLICAS: {
/*
* DATA_ADD_REPLICAS is used for moving data to a different
* device in the background, and due to compression the new copy
* might take up more space than the old copy:
*/
#if 0
int nr = (int) io_opts.data_replicas -
bch2_bkey_nr_dirty_ptrs(k);
#endif
int nr = (int) io_opts.data_replicas;
if (nr > 0) {
m->op.nr_replicas = m->nr_ptrs_reserved = nr;
ret = bch2_disk_reservation_get(c, &m->op.res,
k.k->size, m->op.nr_replicas, 0);
if (ret)
return ret;
}
break;
}
case DATA_REWRITE: {
const union bch_extent_entry *entry;
struct extent_ptr_decoded p;
unsigned compressed_sectors = 0;
extent_for_each_ptr_decode(bkey_s_c_to_extent(k), p, entry)
if (!p.ptr.cached &&
p.crc.compression_type != BCH_COMPRESSION_NONE &&
bch2_dev_in_target(c, p.ptr.dev, data_opts.target))
compressed_sectors += p.crc.compressed_size;
if (compressed_sectors) {
ret = bch2_disk_reservation_add(c, &m->op.res,
compressed_sectors,
BCH_DISK_RESERVATION_NOFAIL);
if (ret)
return ret;
}
break;
}
case DATA_PROMOTE:
m->op.flags |= BCH_WRITE_ALLOC_NOWAIT;
m->op.flags |= BCH_WRITE_CACHED;
break;
default:
BUG();
}
return 0;
}
static void move_free(struct closure *cl)
{
struct moving_io *io = container_of(cl, struct moving_io, cl);
struct moving_context *ctxt = io->write.ctxt;
struct bvec_iter_all iter;
struct bio_vec *bv;
bch2_disk_reservation_put(io->write.op.c, &io->write.op.res);
bio_for_each_segment_all(bv, &io->write.op.wbio.bio, iter)
if (bv->bv_page)
__free_page(bv->bv_page);
wake_up(&ctxt->wait);
kfree(io);
}
static void move_write_done(struct closure *cl)
{
struct moving_io *io = container_of(cl, struct moving_io, cl);
atomic_sub(io->write_sectors, &io->write.ctxt->write_sectors);
closure_return_with_destructor(cl, move_free);
}
static void move_write(struct closure *cl)
{
struct moving_io *io = container_of(cl, struct moving_io, cl);
if (unlikely(io->rbio.bio.bi_status || io->rbio.hole)) {
closure_return_with_destructor(cl, move_free);
return;
}
bch2_migrate_read_done(&io->write, &io->rbio);
atomic_add(io->write_sectors, &io->write.ctxt->write_sectors);
closure_call(&io->write.op.cl, bch2_write, NULL, cl);
continue_at(cl, move_write_done, NULL);
}
static inline struct moving_io *next_pending_write(struct moving_context *ctxt)
{
struct moving_io *io =
list_first_entry_or_null(&ctxt->reads, struct moving_io, list);
return io && io->read_completed ? io : NULL;
}
static void move_read_endio(struct bio *bio)
{
struct moving_io *io = container_of(bio, struct moving_io, rbio.bio);
struct moving_context *ctxt = io->write.ctxt;
atomic_sub(io->read_sectors, &ctxt->read_sectors);
io->read_completed = true;
if (next_pending_write(ctxt))
wake_up(&ctxt->wait);
closure_put(&ctxt->cl);
}
static void do_pending_writes(struct moving_context *ctxt)
{
struct moving_io *io;
while ((io = next_pending_write(ctxt))) {
list_del(&io->list);
closure_call(&io->cl, move_write, NULL, &ctxt->cl);
}
}
#define move_ctxt_wait_event(_ctxt, _cond) \
do { \
do_pending_writes(_ctxt); \
\
if (_cond) \
break; \
__wait_event((_ctxt)->wait, \
next_pending_write(_ctxt) || (_cond)); \
} while (1)
static void bch2_move_ctxt_wait_for_io(struct moving_context *ctxt)
{
unsigned sectors_pending = atomic_read(&ctxt->write_sectors);
move_ctxt_wait_event(ctxt,
!atomic_read(&ctxt->write_sectors) ||
atomic_read(&ctxt->write_sectors) != sectors_pending);
}
static int bch2_move_extent(struct bch_fs *c,
struct moving_context *ctxt,
struct write_point_specifier wp,
struct bch_io_opts io_opts,
struct bkey_s_c_extent e,
enum data_cmd data_cmd,
struct data_opts data_opts)
{
struct moving_io *io;
const union bch_extent_entry *entry;
struct extent_ptr_decoded p;
unsigned sectors = e.k->size, pages;
int ret = -ENOMEM;
move_ctxt_wait_event(ctxt,
atomic_read(&ctxt->write_sectors) <
SECTORS_IN_FLIGHT_PER_DEVICE);
move_ctxt_wait_event(ctxt,
atomic_read(&ctxt->read_sectors) <
SECTORS_IN_FLIGHT_PER_DEVICE);
/* write path might have to decompress data: */
extent_for_each_ptr_decode(e, p, entry)
sectors = max_t(unsigned, sectors, p.crc.uncompressed_size);
pages = DIV_ROUND_UP(sectors, PAGE_SECTORS);
io = kzalloc(sizeof(struct moving_io) +
sizeof(struct bio_vec) * pages, GFP_KERNEL);
if (!io)
goto err;
io->write.ctxt = ctxt;
io->read_sectors = e.k->size;
io->write_sectors = e.k->size;
bio_init(&io->write.op.wbio.bio, NULL, io->bi_inline_vecs, pages, 0);
bio_set_prio(&io->write.op.wbio.bio,
IOPRIO_PRIO_VALUE(IOPRIO_CLASS_IDLE, 0));
if (bch2_bio_alloc_pages(&io->write.op.wbio.bio, sectors << 9,
GFP_KERNEL))
goto err_free;
io->rbio.opts = io_opts;
bio_init(&io->rbio.bio, NULL, io->bi_inline_vecs, pages, 0);
io->rbio.bio.bi_vcnt = pages;
bio_set_prio(&io->rbio.bio, IOPRIO_PRIO_VALUE(IOPRIO_CLASS_IDLE, 0));
io->rbio.bio.bi_iter.bi_size = sectors << 9;
io->rbio.bio.bi_opf = REQ_OP_READ;
io->rbio.bio.bi_iter.bi_sector = bkey_start_offset(e.k);
io->rbio.bio.bi_end_io = move_read_endio;
ret = bch2_migrate_write_init(c, &io->write, wp, io_opts,
data_cmd, data_opts, e.s_c);
if (ret)
goto err_free_pages;
atomic64_inc(&ctxt->stats->keys_moved);
atomic64_add(e.k->size, &ctxt->stats->sectors_moved);
trace_move_extent(e.k);
atomic_add(io->read_sectors, &ctxt->read_sectors);
list_add_tail(&io->list, &ctxt->reads);
/*
* dropped by move_read_endio() - guards against use after free of
* ctxt when doing wakeup
*/
closure_get(&ctxt->cl);
bch2_read_extent(c, &io->rbio, e.s_c,
BCH_READ_NODECODE|
BCH_READ_LAST_FRAGMENT);
return 0;
err_free_pages:
bio_free_pages(&io->write.op.wbio.bio);
err_free:
kfree(io);
err:
trace_move_alloc_fail(e.k);
return ret;
}
int bch2_move_data(struct bch_fs *c,
struct bch_ratelimit *rate,
struct write_point_specifier wp,
struct bpos start,
struct bpos end,
move_pred_fn pred, void *arg,
struct bch_move_stats *stats)
{
bool kthread = (current->flags & PF_KTHREAD) != 0;
struct moving_context ctxt = { .stats = stats };
struct bch_io_opts io_opts = bch2_opts_to_inode_opts(c->opts);
BKEY_PADDED(k) tmp;
struct bkey_s_c k;
struct data_opts data_opts;
enum data_cmd data_cmd;
u64 delay, cur_inum = U64_MAX;
int ret = 0, ret2;
closure_init_stack(&ctxt.cl);
INIT_LIST_HEAD(&ctxt.reads);
init_waitqueue_head(&ctxt.wait);
stats->data_type = BCH_DATA_USER;
bch2_btree_iter_init(&stats->iter, c, BTREE_ID_EXTENTS, start,
BTREE_ITER_PREFETCH);
if (rate)
bch2_ratelimit_reset(rate);
while (1) {
do {
delay = rate ? bch2_ratelimit_delay(rate) : 0;
if (delay) {
bch2_btree_iter_unlock(&stats->iter);
set_current_state(TASK_INTERRUPTIBLE);
}
if (kthread && (ret = kthread_should_stop())) {
__set_current_state(TASK_RUNNING);
goto out;
}
if (delay)
schedule_timeout(delay);
if (unlikely(freezing(current))) {
bch2_btree_iter_unlock(&stats->iter);
move_ctxt_wait_event(&ctxt, list_empty(&ctxt.reads));
try_to_freeze();
}
} while (delay);
peek:
k = bch2_btree_iter_peek(&stats->iter);
if (!k.k)
break;
ret = btree_iter_err(k);
if (ret)
break;
if (bkey_cmp(bkey_start_pos(k.k), end) >= 0)
break;
if (!bkey_extent_is_data(k.k))
goto next_nondata;
if (cur_inum != k.k->p.inode) {
struct bch_inode_unpacked inode;
/* don't hold btree locks while looking up inode: */
bch2_btree_iter_unlock(&stats->iter);
io_opts = bch2_opts_to_inode_opts(c->opts);
if (!bch2_inode_find_by_inum(c, k.k->p.inode, &inode))
bch2_io_opts_apply(&io_opts, bch2_inode_opts_get(&inode));
cur_inum = k.k->p.inode;
goto peek;
}
switch ((data_cmd = pred(c, arg, k, &io_opts, &data_opts))) {
case DATA_SKIP:
goto next;
case DATA_SCRUB:
BUG();
case DATA_ADD_REPLICAS:
case DATA_REWRITE:
case DATA_PROMOTE:
break;
default:
BUG();
}
/* unlock before doing IO: */
bkey_reassemble(&tmp.k, k);
k = bkey_i_to_s_c(&tmp.k);
bch2_btree_iter_unlock(&stats->iter);
ret2 = bch2_move_extent(c, &ctxt, wp, io_opts,
bkey_s_c_to_extent(k),
data_cmd, data_opts);
if (ret2) {
if (ret2 == -ENOMEM) {
/* memory allocation failure, wait for some IO to finish */
bch2_move_ctxt_wait_for_io(&ctxt);
continue;
}
/* XXX signal failure */
goto next;
}
if (rate)
bch2_ratelimit_increment(rate, k.k->size);
next:
atomic64_add(k.k->size * bch2_bkey_nr_dirty_ptrs(k),
&stats->sectors_seen);
next_nondata:
bch2_btree_iter_next(&stats->iter);
bch2_btree_iter_cond_resched(&stats->iter);
}
out:
bch2_btree_iter_unlock(&stats->iter);
move_ctxt_wait_event(&ctxt, list_empty(&ctxt.reads));
closure_sync(&ctxt.cl);
EBUG_ON(atomic_read(&ctxt.write_sectors));
trace_move_data(c,
atomic64_read(&stats->sectors_moved),
atomic64_read(&stats->keys_moved));
return ret;
}
static int bch2_gc_data_replicas(struct bch_fs *c)
{
struct btree_iter iter;
struct bkey_s_c k;
int ret;
mutex_lock(&c->replicas_gc_lock);
bch2_replicas_gc_start(c, (1 << BCH_DATA_USER)|(1 << BCH_DATA_CACHED));
for_each_btree_key(&iter, c, BTREE_ID_EXTENTS, POS_MIN,
BTREE_ITER_PREFETCH, k) {
ret = bch2_mark_bkey_replicas(c, k);
if (ret)
break;
}
ret = bch2_btree_iter_unlock(&iter) ?: ret;
bch2_replicas_gc_end(c, ret);
mutex_unlock(&c->replicas_gc_lock);
return ret;
}
static int bch2_gc_btree_replicas(struct bch_fs *c)
{
struct btree_iter iter;
struct btree *b;
unsigned id;
int ret = 0;
mutex_lock(&c->replicas_gc_lock);
bch2_replicas_gc_start(c, 1 << BCH_DATA_BTREE);
for (id = 0; id < BTREE_ID_NR; id++) {
for_each_btree_node(&iter, c, id, POS_MIN, BTREE_ITER_PREFETCH, b) {
ret = bch2_mark_bkey_replicas(c, bkey_i_to_s_c(&b->key));
bch2_btree_iter_cond_resched(&iter);
}
ret = bch2_btree_iter_unlock(&iter) ?: ret;
}
bch2_replicas_gc_end(c, ret);
mutex_unlock(&c->replicas_gc_lock);
return ret;
}
static int bch2_move_btree(struct bch_fs *c,
move_pred_fn pred,
void *arg,
struct bch_move_stats *stats)
{
struct bch_io_opts io_opts = bch2_opts_to_inode_opts(c->opts);
struct btree *b;
unsigned id;
struct data_opts data_opts;
enum data_cmd cmd;
int ret = 0;
stats->data_type = BCH_DATA_BTREE;
for (id = 0; id < BTREE_ID_NR; id++) {
for_each_btree_node(&stats->iter, c, id, POS_MIN, BTREE_ITER_PREFETCH, b) {
switch ((cmd = pred(c, arg,
bkey_i_to_s_c(&b->key),
&io_opts, &data_opts))) {
case DATA_SKIP:
goto next;
case DATA_SCRUB:
BUG();
case DATA_ADD_REPLICAS:
case DATA_REWRITE:
break;
default:
BUG();
}
ret = bch2_btree_node_rewrite(c, &stats->iter,
b->data->keys.seq, 0) ?: ret;
next:
bch2_btree_iter_cond_resched(&stats->iter);
}
ret = bch2_btree_iter_unlock(&stats->iter) ?: ret;
}
return ret;
}
#if 0
static enum data_cmd scrub_pred(struct bch_fs *c, void *arg,
struct bkey_s_c k,
struct bch_io_opts *io_opts,
struct data_opts *data_opts)
{
return DATA_SCRUB;
}
#endif
static enum data_cmd rereplicate_pred(struct bch_fs *c, void *arg,
struct bkey_s_c k,
struct bch_io_opts *io_opts,
struct data_opts *data_opts)
{
unsigned nr_good = bch2_bkey_durability(c, k);
unsigned replicas = 0;
switch (k.k->type) {
case KEY_TYPE_btree_ptr:
replicas = c->opts.metadata_replicas;
break;
case KEY_TYPE_extent:
replicas = io_opts->data_replicas;
break;
}
if (!nr_good || nr_good >= replicas)
return DATA_SKIP;
data_opts->target = 0;
data_opts->btree_insert_flags = 0;
return DATA_ADD_REPLICAS;
}
static enum data_cmd migrate_pred(struct bch_fs *c, void *arg,
struct bkey_s_c k,
struct bch_io_opts *io_opts,
struct data_opts *data_opts)
{
struct bch_ioctl_data *op = arg;
if (!bch2_bkey_has_device(k, op->migrate.dev))
return DATA_SKIP;
data_opts->target = 0;
data_opts->btree_insert_flags = 0;
data_opts->rewrite_dev = op->migrate.dev;
return DATA_REWRITE;
}
int bch2_data_job(struct bch_fs *c,
struct bch_move_stats *stats,
struct bch_ioctl_data op)
{
int ret = 0;
switch (op.op) {
case BCH_DATA_OP_REREPLICATE:
stats->data_type = BCH_DATA_JOURNAL;
ret = bch2_journal_flush_device_pins(&c->journal, -1);
ret = bch2_move_btree(c, rereplicate_pred, c, stats) ?: ret;
while (1) {
closure_wait_event(&c->btree_interior_update_wait,
!bch2_btree_interior_updates_nr_pending(c) ||
c->btree_roots_dirty);
if (!bch2_btree_interior_updates_nr_pending(c))
break;
bch2_journal_meta(&c->journal);
}
ret = bch2_gc_btree_replicas(c) ?: ret;
ret = bch2_move_data(c, NULL,
writepoint_hashed((unsigned long) current),
op.start,
op.end,
rereplicate_pred, c, stats) ?: ret;
ret = bch2_gc_data_replicas(c) ?: ret;
break;
case BCH_DATA_OP_MIGRATE:
if (op.migrate.dev >= c->sb.nr_devices)
return -EINVAL;
stats->data_type = BCH_DATA_JOURNAL;
ret = bch2_journal_flush_device_pins(&c->journal, op.migrate.dev);
ret = bch2_move_btree(c, migrate_pred, &op, stats) ?: ret;
ret = bch2_gc_btree_replicas(c) ?: ret;
ret = bch2_move_data(c, NULL,
writepoint_hashed((unsigned long) current),
op.start,
op.end,
migrate_pred, &op, stats) ?: ret;
ret = bch2_gc_data_replicas(c) ?: ret;
break;
default:
ret = -EINVAL;
}
return ret;
}