linux/fs/bcachefs/reflink.c

415 lines
11 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0
#include "bcachefs.h"
#include "bkey_buf.h"
#include "btree_update.h"
#include "buckets.h"
#include "extents.h"
#include "inode.h"
#include "io_misc.h"
#include "io_write.h"
bcachefs: rebalance_work This adds a new btree, rebalance_work, to eliminate scanning required for finding extents that need work done on them in the background - i.e. for the background_target and background_compression options. rebalance_work is a bitset btree, where a KEY_TYPE_set corresponds to an extent in the extents or reflink btree at the same pos. A new extent field is added, bch_extent_rebalance, which indicates that this extent has work that needs to be done in the background - and which options to use. This allows per-inode options to be propagated to indirect extents - at least in some circumstances. In this patch, changing IO options on a file will not propagate the new options to indirect extents pointed to by that file. Updating (setting/clearing) the rebalance_work btree is done by the extent trigger, which looks at the bch_extent_rebalance field. Scanning is still requrired after changing IO path options - either just for a given inode, or for the whole filesystem. We indicate that scanning is required by adding a KEY_TYPE_cookie key to the rebalance_work btree: the cookie counter is so that we can detect that scanning is still required when an option has been flipped mid-way through an existing scan. Future possible work: - Propagate options to indirect extents when being changed - Add other IO path options - nr_replicas, ec, to rebalance_work so they can be applied in the background when they change - Add a counter, for bcachefs fs usage output, showing the pending amount of rebalance work: we'll probably want to do this after the disk space accounting rewrite (moving it to a new btree) Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2023-10-21 01:33:14 +08:00
#include "rebalance.h"
#include "reflink.h"
#include "subvolume.h"
#include "super-io.h"
#include <linux/sched/signal.h>
static inline unsigned bkey_type_to_indirect(const struct bkey *k)
{
switch (k->type) {
case KEY_TYPE_extent:
return KEY_TYPE_reflink_v;
case KEY_TYPE_inline_data:
return KEY_TYPE_indirect_inline_data;
default:
return 0;
}
}
/* reflink pointers */
int bch2_reflink_p_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
struct printbuf *err)
{
struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k);
if (c->sb.version >= bcachefs_metadata_version_reflink_p_fix &&
le64_to_cpu(p.v->idx) < le32_to_cpu(p.v->front_pad)) {
prt_printf(err, "idx < front_pad (%llu < %u)",
le64_to_cpu(p.v->idx), le32_to_cpu(p.v->front_pad));
return -EINVAL;
}
return 0;
}
void bch2_reflink_p_to_text(struct printbuf *out, struct bch_fs *c,
struct bkey_s_c k)
{
struct bkey_s_c_reflink_p p = bkey_s_c_to_reflink_p(k);
prt_printf(out, "idx %llu front_pad %u back_pad %u",
le64_to_cpu(p.v->idx),
le32_to_cpu(p.v->front_pad),
le32_to_cpu(p.v->back_pad));
}
bool bch2_reflink_p_merge(struct bch_fs *c, struct bkey_s _l, struct bkey_s_c _r)
{
struct bkey_s_reflink_p l = bkey_s_to_reflink_p(_l);
struct bkey_s_c_reflink_p r = bkey_s_c_to_reflink_p(_r);
/*
* Disabled for now, the triggers code needs to be reworked for merging
* of reflink pointers to work:
*/
return false;
if (le64_to_cpu(l.v->idx) + l.k->size != le64_to_cpu(r.v->idx))
return false;
bch2_key_resize(l.k, l.k->size + r.k->size);
return true;
}
/* indirect extents */
int bch2_reflink_v_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
struct printbuf *err)
{
return bch2_bkey_ptrs_invalid(c, k, flags, err);
}
void bch2_reflink_v_to_text(struct printbuf *out, struct bch_fs *c,
struct bkey_s_c k)
{
struct bkey_s_c_reflink_v r = bkey_s_c_to_reflink_v(k);
prt_printf(out, "refcount: %llu ", le64_to_cpu(r.v->refcount));
bch2_bkey_ptrs_to_text(out, c, k);
}
#if 0
Currently disabled, needs to be debugged:
bool bch2_reflink_v_merge(struct bch_fs *c, struct bkey_s _l, struct bkey_s_c _r)
{
struct bkey_s_reflink_v l = bkey_s_to_reflink_v(_l);
struct bkey_s_c_reflink_v r = bkey_s_c_to_reflink_v(_r);
return l.v->refcount == r.v->refcount && bch2_extent_merge(c, _l, _r);
}
#endif
static inline void check_indirect_extent_deleting(struct bkey_i *new, unsigned *flags)
{
if ((*flags & BTREE_TRIGGER_INSERT) && !*bkey_refcount(new)) {
new->k.type = KEY_TYPE_deleted;
new->k.size = 0;
set_bkey_val_u64s(&new->k, 0);;
*flags &= ~BTREE_TRIGGER_INSERT;
}
}
int bch2_trans_mark_reflink_v(struct btree_trans *trans,
enum btree_id btree_id, unsigned level,
struct bkey_s_c old, struct bkey_i *new,
unsigned flags)
{
check_indirect_extent_deleting(new, &flags);
if (old.k->type == KEY_TYPE_reflink_v &&
new->k.type == KEY_TYPE_reflink_v &&
old.k->u64s == new->k.u64s &&
!memcmp(bkey_s_c_to_reflink_v(old).v->start,
bkey_i_to_reflink_v(new)->v.start,
bkey_val_bytes(&new->k) - 8))
return 0;
return bch2_trans_mark_extent(trans, btree_id, level, old, new, flags);
}
/* indirect inline data */
int bch2_indirect_inline_data_invalid(struct bch_fs *c, struct bkey_s_c k,
enum bkey_invalid_flags flags,
struct printbuf *err)
{
return 0;
}
void bch2_indirect_inline_data_to_text(struct printbuf *out,
struct bch_fs *c, struct bkey_s_c k)
{
struct bkey_s_c_indirect_inline_data d = bkey_s_c_to_indirect_inline_data(k);
unsigned datalen = bkey_inline_data_bytes(k.k);
prt_printf(out, "refcount %llu datalen %u: %*phN",
le64_to_cpu(d.v->refcount), datalen,
min(datalen, 32U), d.v->data);
}
int bch2_trans_mark_indirect_inline_data(struct btree_trans *trans,
enum btree_id btree_id, unsigned level,
struct bkey_s_c old, struct bkey_i *new,
unsigned flags)
{
check_indirect_extent_deleting(new, &flags);
return 0;
}
static int bch2_make_extent_indirect(struct btree_trans *trans,
struct btree_iter *extent_iter,
struct bkey_i *orig)
{
struct bch_fs *c = trans->c;
struct btree_iter reflink_iter = { NULL };
struct bkey_s_c k;
struct bkey_i *r_v;
struct bkey_i_reflink_p *r_p;
__le64 *refcount;
int ret;
if (orig->k.type == KEY_TYPE_inline_data)
bch2_check_set_feature(c, BCH_FEATURE_reflink_inline_data);
bch2_trans_iter_init(trans, &reflink_iter, BTREE_ID_reflink, POS_MAX,
BTREE_ITER_INTENT);
k = bch2_btree_iter_peek_prev(&reflink_iter);
ret = bkey_err(k);
if (ret)
goto err;
r_v = bch2_trans_kmalloc(trans, sizeof(__le64) + bkey_bytes(&orig->k));
ret = PTR_ERR_OR_ZERO(r_v);
if (ret)
goto err;
bkey_init(&r_v->k);
r_v->k.type = bkey_type_to_indirect(&orig->k);
r_v->k.p = reflink_iter.pos;
bch2_key_resize(&r_v->k, orig->k.size);
r_v->k.version = orig->k.version;
set_bkey_val_bytes(&r_v->k, sizeof(__le64) + bkey_val_bytes(&orig->k));
refcount = bkey_refcount(r_v);
*refcount = 0;
memcpy(refcount + 1, &orig->v, bkey_val_bytes(&orig->k));
ret = bch2_trans_update(trans, &reflink_iter, r_v, 0);
if (ret)
goto err;
/*
* orig is in a bkey_buf which statically allocates 5 64s for the val,
* so we know it will be big enough:
*/
orig->k.type = KEY_TYPE_reflink_p;
r_p = bkey_i_to_reflink_p(orig);
set_bkey_val_bytes(&r_p->k, sizeof(r_p->v));
/* FORTIFY_SOURCE is broken here, and doesn't provide unsafe_memset() */
#if !defined(__NO_FORTIFY) && defined(__OPTIMIZE__) && defined(CONFIG_FORTIFY_SOURCE)
__underlying_memset(&r_p->v, 0, sizeof(r_p->v));
#else
memset(&r_p->v, 0, sizeof(r_p->v));
#endif
r_p->v.idx = cpu_to_le64(bkey_start_offset(&r_v->k));
ret = bch2_trans_update(trans, extent_iter, &r_p->k_i,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
err:
bch2_trans_iter_exit(trans, &reflink_iter);
return ret;
}
static struct bkey_s_c get_next_src(struct btree_iter *iter, struct bpos end)
{
struct bkey_s_c k;
int ret;
for_each_btree_key_upto_continue_norestart(*iter, end, 0, k, ret) {
if (bkey_extent_is_unwritten(k))
continue;
if (bkey_extent_is_data(k.k))
return k;
}
if (bkey_ge(iter->pos, end))
bch2_btree_iter_set_pos(iter, end);
return ret ? bkey_s_c_err(ret) : bkey_s_c_null;
}
s64 bch2_remap_range(struct bch_fs *c,
subvol_inum dst_inum, u64 dst_offset,
subvol_inum src_inum, u64 src_offset,
u64 remap_sectors,
u64 new_i_size, s64 *i_sectors_delta)
{
struct btree_trans *trans;
struct btree_iter dst_iter, src_iter;
struct bkey_s_c src_k;
struct bkey_buf new_dst, new_src;
struct bpos dst_start = POS(dst_inum.inum, dst_offset);
struct bpos src_start = POS(src_inum.inum, src_offset);
struct bpos dst_end = dst_start, src_end = src_start;
bcachefs: rebalance_work This adds a new btree, rebalance_work, to eliminate scanning required for finding extents that need work done on them in the background - i.e. for the background_target and background_compression options. rebalance_work is a bitset btree, where a KEY_TYPE_set corresponds to an extent in the extents or reflink btree at the same pos. A new extent field is added, bch_extent_rebalance, which indicates that this extent has work that needs to be done in the background - and which options to use. This allows per-inode options to be propagated to indirect extents - at least in some circumstances. In this patch, changing IO options on a file will not propagate the new options to indirect extents pointed to by that file. Updating (setting/clearing) the rebalance_work btree is done by the extent trigger, which looks at the bch_extent_rebalance field. Scanning is still requrired after changing IO path options - either just for a given inode, or for the whole filesystem. We indicate that scanning is required by adding a KEY_TYPE_cookie key to the rebalance_work btree: the cookie counter is so that we can detect that scanning is still required when an option has been flipped mid-way through an existing scan. Future possible work: - Propagate options to indirect extents when being changed - Add other IO path options - nr_replicas, ec, to rebalance_work so they can be applied in the background when they change - Add a counter, for bcachefs fs usage output, showing the pending amount of rebalance work: we'll probably want to do this after the disk space accounting rewrite (moving it to a new btree) Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2023-10-21 01:33:14 +08:00
struct bch_io_opts opts;
struct bpos src_want;
bcachefs: rebalance_work This adds a new btree, rebalance_work, to eliminate scanning required for finding extents that need work done on them in the background - i.e. for the background_target and background_compression options. rebalance_work is a bitset btree, where a KEY_TYPE_set corresponds to an extent in the extents or reflink btree at the same pos. A new extent field is added, bch_extent_rebalance, which indicates that this extent has work that needs to be done in the background - and which options to use. This allows per-inode options to be propagated to indirect extents - at least in some circumstances. In this patch, changing IO options on a file will not propagate the new options to indirect extents pointed to by that file. Updating (setting/clearing) the rebalance_work btree is done by the extent trigger, which looks at the bch_extent_rebalance field. Scanning is still requrired after changing IO path options - either just for a given inode, or for the whole filesystem. We indicate that scanning is required by adding a KEY_TYPE_cookie key to the rebalance_work btree: the cookie counter is so that we can detect that scanning is still required when an option has been flipped mid-way through an existing scan. Future possible work: - Propagate options to indirect extents when being changed - Add other IO path options - nr_replicas, ec, to rebalance_work so they can be applied in the background when they change - Add a counter, for bcachefs fs usage output, showing the pending amount of rebalance work: we'll probably want to do this after the disk space accounting rewrite (moving it to a new btree) Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2023-10-21 01:33:14 +08:00
u64 dst_done = 0;
u32 dst_snapshot, src_snapshot;
int ret = 0, ret2 = 0;
if (!bch2_write_ref_tryget(c, BCH_WRITE_REF_reflink))
return -BCH_ERR_erofs_no_writes;
bch2_check_set_feature(c, BCH_FEATURE_reflink);
dst_end.offset += remap_sectors;
src_end.offset += remap_sectors;
bch2_bkey_buf_init(&new_dst);
bch2_bkey_buf_init(&new_src);
trans = bch2_trans_get(c);
bcachefs: rebalance_work This adds a new btree, rebalance_work, to eliminate scanning required for finding extents that need work done on them in the background - i.e. for the background_target and background_compression options. rebalance_work is a bitset btree, where a KEY_TYPE_set corresponds to an extent in the extents or reflink btree at the same pos. A new extent field is added, bch_extent_rebalance, which indicates that this extent has work that needs to be done in the background - and which options to use. This allows per-inode options to be propagated to indirect extents - at least in some circumstances. In this patch, changing IO options on a file will not propagate the new options to indirect extents pointed to by that file. Updating (setting/clearing) the rebalance_work btree is done by the extent trigger, which looks at the bch_extent_rebalance field. Scanning is still requrired after changing IO path options - either just for a given inode, or for the whole filesystem. We indicate that scanning is required by adding a KEY_TYPE_cookie key to the rebalance_work btree: the cookie counter is so that we can detect that scanning is still required when an option has been flipped mid-way through an existing scan. Future possible work: - Propagate options to indirect extents when being changed - Add other IO path options - nr_replicas, ec, to rebalance_work so they can be applied in the background when they change - Add a counter, for bcachefs fs usage output, showing the pending amount of rebalance work: we'll probably want to do this after the disk space accounting rewrite (moving it to a new btree) Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2023-10-21 01:33:14 +08:00
ret = bch2_inum_opts_get(trans, src_inum, &opts);
if (ret)
goto err;
bch2_trans_iter_init(trans, &src_iter, BTREE_ID_extents, src_start,
BTREE_ITER_INTENT);
bch2_trans_iter_init(trans, &dst_iter, BTREE_ID_extents, dst_start,
BTREE_ITER_INTENT);
while ((ret == 0 ||
bch2_err_matches(ret, BCH_ERR_transaction_restart)) &&
bkey_lt(dst_iter.pos, dst_end)) {
struct disk_reservation disk_res = { 0 };
bch2_trans_begin(trans);
if (fatal_signal_pending(current)) {
ret = -EINTR;
break;
}
ret = bch2_subvolume_get_snapshot(trans, src_inum.subvol,
&src_snapshot);
if (ret)
continue;
bch2_btree_iter_set_snapshot(&src_iter, src_snapshot);
ret = bch2_subvolume_get_snapshot(trans, dst_inum.subvol,
&dst_snapshot);
if (ret)
continue;
bch2_btree_iter_set_snapshot(&dst_iter, dst_snapshot);
dst_done = dst_iter.pos.offset - dst_start.offset;
src_want = POS(src_start.inode, src_start.offset + dst_done);
bch2_btree_iter_set_pos(&src_iter, src_want);
src_k = get_next_src(&src_iter, src_end);
ret = bkey_err(src_k);
if (ret)
continue;
if (bkey_lt(src_want, src_iter.pos)) {
ret = bch2_fpunch_at(trans, &dst_iter, dst_inum,
min(dst_end.offset,
dst_iter.pos.offset +
src_iter.pos.offset - src_want.offset),
i_sectors_delta);
continue;
}
if (src_k.k->type != KEY_TYPE_reflink_p) {
bch2_btree_iter_set_pos_to_extent_start(&src_iter);
bch2_bkey_buf_reassemble(&new_src, c, src_k);
src_k = bkey_i_to_s_c(new_src.k);
ret = bch2_make_extent_indirect(trans, &src_iter,
new_src.k);
if (ret)
continue;
BUG_ON(src_k.k->type != KEY_TYPE_reflink_p);
}
if (src_k.k->type == KEY_TYPE_reflink_p) {
struct bkey_s_c_reflink_p src_p =
bkey_s_c_to_reflink_p(src_k);
struct bkey_i_reflink_p *dst_p =
bkey_reflink_p_init(new_dst.k);
u64 offset = le64_to_cpu(src_p.v->idx) +
(src_want.offset -
bkey_start_offset(src_k.k));
dst_p->v.idx = cpu_to_le64(offset);
} else {
BUG();
}
new_dst.k->k.p = dst_iter.pos;
bch2_key_resize(&new_dst.k->k,
min(src_k.k->p.offset - src_want.offset,
dst_end.offset - dst_iter.pos.offset));
bcachefs: rebalance_work This adds a new btree, rebalance_work, to eliminate scanning required for finding extents that need work done on them in the background - i.e. for the background_target and background_compression options. rebalance_work is a bitset btree, where a KEY_TYPE_set corresponds to an extent in the extents or reflink btree at the same pos. A new extent field is added, bch_extent_rebalance, which indicates that this extent has work that needs to be done in the background - and which options to use. This allows per-inode options to be propagated to indirect extents - at least in some circumstances. In this patch, changing IO options on a file will not propagate the new options to indirect extents pointed to by that file. Updating (setting/clearing) the rebalance_work btree is done by the extent trigger, which looks at the bch_extent_rebalance field. Scanning is still requrired after changing IO path options - either just for a given inode, or for the whole filesystem. We indicate that scanning is required by adding a KEY_TYPE_cookie key to the rebalance_work btree: the cookie counter is so that we can detect that scanning is still required when an option has been flipped mid-way through an existing scan. Future possible work: - Propagate options to indirect extents when being changed - Add other IO path options - nr_replicas, ec, to rebalance_work so they can be applied in the background when they change - Add a counter, for bcachefs fs usage output, showing the pending amount of rebalance work: we'll probably want to do this after the disk space accounting rewrite (moving it to a new btree) Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2023-10-21 01:33:14 +08:00
ret = bch2_bkey_set_needs_rebalance(c, new_dst.k,
opts.background_target,
opts.background_compression) ?:
bch2_extent_update(trans, dst_inum, &dst_iter,
new_dst.k, &disk_res,
new_i_size, i_sectors_delta,
true);
bch2_disk_reservation_put(c, &disk_res);
}
bch2_trans_iter_exit(trans, &dst_iter);
bch2_trans_iter_exit(trans, &src_iter);
BUG_ON(!ret && !bkey_eq(dst_iter.pos, dst_end));
BUG_ON(bkey_gt(dst_iter.pos, dst_end));
dst_done = dst_iter.pos.offset - dst_start.offset;
new_i_size = min(dst_iter.pos.offset << 9, new_i_size);
do {
struct bch_inode_unpacked inode_u;
struct btree_iter inode_iter = { NULL };
bch2_trans_begin(trans);
ret2 = bch2_inode_peek(trans, &inode_iter, &inode_u,
dst_inum, BTREE_ITER_INTENT);
if (!ret2 &&
inode_u.bi_size < new_i_size) {
inode_u.bi_size = new_i_size;
ret2 = bch2_inode_write(trans, &inode_iter, &inode_u) ?:
bch2_trans_commit(trans, NULL, NULL,
BTREE_INSERT_NOFAIL);
}
bch2_trans_iter_exit(trans, &inode_iter);
} while (bch2_err_matches(ret2, BCH_ERR_transaction_restart));
bcachefs: rebalance_work This adds a new btree, rebalance_work, to eliminate scanning required for finding extents that need work done on them in the background - i.e. for the background_target and background_compression options. rebalance_work is a bitset btree, where a KEY_TYPE_set corresponds to an extent in the extents or reflink btree at the same pos. A new extent field is added, bch_extent_rebalance, which indicates that this extent has work that needs to be done in the background - and which options to use. This allows per-inode options to be propagated to indirect extents - at least in some circumstances. In this patch, changing IO options on a file will not propagate the new options to indirect extents pointed to by that file. Updating (setting/clearing) the rebalance_work btree is done by the extent trigger, which looks at the bch_extent_rebalance field. Scanning is still requrired after changing IO path options - either just for a given inode, or for the whole filesystem. We indicate that scanning is required by adding a KEY_TYPE_cookie key to the rebalance_work btree: the cookie counter is so that we can detect that scanning is still required when an option has been flipped mid-way through an existing scan. Future possible work: - Propagate options to indirect extents when being changed - Add other IO path options - nr_replicas, ec, to rebalance_work so they can be applied in the background when they change - Add a counter, for bcachefs fs usage output, showing the pending amount of rebalance work: we'll probably want to do this after the disk space accounting rewrite (moving it to a new btree) Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
2023-10-21 01:33:14 +08:00
err:
bch2_trans_put(trans);
bch2_bkey_buf_exit(&new_src, c);
bch2_bkey_buf_exit(&new_dst, c);
bch2_write_ref_put(c, BCH_WRITE_REF_reflink);
return dst_done ?: ret ?: ret2;
}