+// SPDX-License-Identifier: GPL-2.0
#include "bcachefs.h"
+#include "btree_key_cache.h"
#include "bkey_methods.h"
#include "btree_update.h"
+#include "buckets.h"
#include "error.h"
#include "extents.h"
+#include "extent_update.h"
#include "inode.h"
-#include "io.h"
-#include "keylist.h"
+#include "str_hash.h"
+#include "subvolume.h"
+#include "varint.h"
#include <linux/random.h>
};
static const u8 byte_table[8] = { 1, 2, 3, 4, 6, 8, 10, 13 };
-static const u8 bits_table[8] = {
- 1 * 8 - 1,
- 2 * 8 - 2,
- 3 * 8 - 3,
- 4 * 8 - 4,
- 6 * 8 - 5,
- 8 * 8 - 6,
- 10 * 8 - 7,
- 13 * 8 - 8,
-};
-
-static int inode_encode_field(u8 *out, u8 *end, u64 hi, u64 lo)
-{
- __be64 in[2] = { cpu_to_be64(hi), cpu_to_be64(lo), };
- unsigned shift, bytes, bits = likely(!hi)
- ? fls64(lo)
- : fls64(hi) + 64;
-
- for (shift = 1; shift <= 8; shift++)
- if (bits < bits_table[shift - 1])
- goto got_shift;
-
- BUG();
-got_shift:
- bytes = byte_table[shift - 1];
-
- BUG_ON(out + bytes > end);
-
- memcpy(out, (u8 *) in + 16 - bytes, bytes);
- *out |= (1 << 8) >> shift;
-
- return bytes;
-}
static int inode_decode_field(const u8 *in, const u8 *end,
u64 out[2], unsigned *out_bits)
return bytes;
}
-void bch2_inode_pack(struct bkey_inode_buf *packed,
+void bch2_inode_pack(struct bch_fs *c,
+ struct bkey_inode_buf *packed,
const struct bch_inode_unpacked *inode)
{
- u8 *out = packed->inode.v.fields;
+ struct bkey_i_inode_v2 *k = &packed->inode;
+ u8 *out = k->v.fields;
u8 *end = (void *) &packed[1];
u8 *last_nonzero_field = out;
unsigned nr_fields = 0, last_nonzero_fieldnr = 0;
+ unsigned bytes;
+ int ret;
- bkey_inode_init(&packed->inode.k_i);
- packed->inode.k.p.inode = inode->bi_inum;
+ bkey_inode_v2_init(&packed->inode.k_i);
+ packed->inode.k.p.offset = inode->bi_inum;
+ packed->inode.v.bi_journal_seq = cpu_to_le64(inode->bi_journal_seq);
packed->inode.v.bi_hash_seed = inode->bi_hash_seed;
- packed->inode.v.bi_flags = cpu_to_le32(inode->bi_flags);
+ packed->inode.v.bi_flags = cpu_to_le64(inode->bi_flags);
+ packed->inode.v.bi_flags = cpu_to_le64(inode->bi_flags);
packed->inode.v.bi_mode = cpu_to_le16(inode->bi_mode);
-#define x(_name, _bits) \
- out += inode_encode_field(out, end, 0, inode->_name); \
+#define x(_name, _bits) \
nr_fields++; \
\
if (inode->_name) { \
+ ret = bch2_varint_encode_fast(out, inode->_name); \
+ out += ret; \
+ \
+ if (_bits > 64) \
+ *out++ = 0; \
+ \
last_nonzero_field = out; \
last_nonzero_fieldnr = nr_fields; \
+ } else { \
+ *out++ = 0; \
+ \
+ if (_bits > 64) \
+ *out++ = 0; \
}
BCH_INODE_FIELDS()
#undef x
+ BUG_ON(out > end);
out = last_nonzero_field;
nr_fields = last_nonzero_fieldnr;
- set_bkey_val_bytes(&packed->inode.k, out - (u8 *) &packed->inode.v);
- memset(out, 0,
- (u8 *) &packed->inode.v +
- bkey_val_bytes(&packed->inode.k) - out);
+ bytes = out - (u8 *) &packed->inode.v;
+ set_bkey_val_bytes(&packed->inode.k, bytes);
+ memset_u64s_tail(&packed->inode.v, 0, bytes);
- SET_INODE_NR_FIELDS(&packed->inode.v, nr_fields);
+ SET_INODEv2_NR_FIELDS(&k->v, nr_fields);
if (IS_ENABLED(CONFIG_BCACHEFS_DEBUG)) {
struct bch_inode_unpacked unpacked;
- int ret = bch2_inode_unpack(inode_i_to_s_c(&packed->inode),
+ int ret = bch2_inode_unpack(bkey_i_to_s_c(&packed->inode.k_i),
&unpacked);
BUG_ON(ret);
BUG_ON(unpacked.bi_inum != inode->bi_inum);
BUG_ON(unpacked.bi_hash_seed != inode->bi_hash_seed);
BUG_ON(unpacked.bi_mode != inode->bi_mode);
-#define x(_name, _bits) BUG_ON(unpacked._name != inode->_name);
+#define x(_name, _bits) if (unpacked._name != inode->_name) \
+ panic("unpacked %llu should be %llu", \
+ (u64) unpacked._name, (u64) inode->_name);
BCH_INODE_FIELDS()
#undef x
}
}
-int bch2_inode_unpack(struct bkey_s_c_inode inode,
- struct bch_inode_unpacked *unpacked)
+static noinline int bch2_inode_unpack_v1(struct bkey_s_c_inode inode,
+ struct bch_inode_unpacked *unpacked)
{
const u8 *in = inode.v->fields;
- const u8 *end = (void *) inode.v + bkey_val_bytes(inode.k);
+ const u8 *end = bkey_val_end(inode);
u64 field[2];
unsigned fieldnr = 0, field_bits;
int ret;
- unpacked->bi_inum = inode.k->p.inode;
- unpacked->bi_hash_seed = inode.v->bi_hash_seed;
- unpacked->bi_flags = le32_to_cpu(inode.v->bi_flags);
- unpacked->bi_mode = le16_to_cpu(inode.v->bi_mode);
-
#define x(_name, _bits) \
if (fieldnr++ == INODE_NR_FIELDS(inode.v)) { \
- memset(&unpacked->_name, 0, \
- sizeof(*unpacked) - \
- offsetof(struct bch_inode_unpacked, _name)); \
+ unsigned offset = offsetof(struct bch_inode_unpacked, _name);\
+ memset((void *) unpacked + offset, 0, \
+ sizeof(*unpacked) - offset); \
return 0; \
} \
\
#undef x
/* XXX: signal if there were more fields than expected? */
+ return 0;
+}
+
+static int bch2_inode_unpack_v2(struct bch_inode_unpacked *unpacked,
+ const u8 *in, const u8 *end,
+ unsigned nr_fields)
+{
+ unsigned fieldnr = 0;
+ int ret;
+ u64 v[2];
+
+#define x(_name, _bits) \
+ if (fieldnr < nr_fields) { \
+ ret = bch2_varint_decode_fast(in, end, &v[0]); \
+ if (ret < 0) \
+ return ret; \
+ in += ret; \
+ \
+ if (_bits > 64) { \
+ ret = bch2_varint_decode_fast(in, end, &v[1]); \
+ if (ret < 0) \
+ return ret; \
+ in += ret; \
+ } else { \
+ v[1] = 0; \
+ } \
+ } else { \
+ v[0] = v[1] = 0; \
+ } \
+ \
+ unpacked->_name = v[0]; \
+ if (v[1] || v[0] != unpacked->_name) \
+ return -1; \
+ fieldnr++;
+ BCH_INODE_FIELDS()
+#undef x
+
+ /* XXX: signal if there were more fields than expected? */
return 0;
}
-const char *bch2_inode_invalid(const struct bch_fs *c, struct bkey_s_c k)
+int bch2_inode_unpack(struct bkey_s_c k,
+ struct bch_inode_unpacked *unpacked)
{
+ switch (k.k->type) {
+ case KEY_TYPE_inode: {
struct bkey_s_c_inode inode = bkey_s_c_to_inode(k);
- struct bch_inode_unpacked unpacked;
- if (k.k->p.offset)
- return "nonzero offset";
+ unpacked->bi_inum = inode.k->p.offset;
+ unpacked->bi_journal_seq= 0;
+ unpacked->bi_hash_seed = inode.v->bi_hash_seed;
+ unpacked->bi_flags = le32_to_cpu(inode.v->bi_flags);
+ unpacked->bi_mode = le16_to_cpu(inode.v->bi_mode);
+
+ if (INODE_NEW_VARINT(inode.v)) {
+ return bch2_inode_unpack_v2(unpacked, inode.v->fields,
+ bkey_val_end(inode),
+ INODE_NR_FIELDS(inode.v));
+ } else {
+ return bch2_inode_unpack_v1(inode, unpacked);
+ }
+ break;
+ }
+ case KEY_TYPE_inode_v2: {
+ struct bkey_s_c_inode_v2 inode = bkey_s_c_to_inode_v2(k);
+
+ unpacked->bi_inum = inode.k->p.offset;
+ unpacked->bi_journal_seq= le64_to_cpu(inode.v->bi_journal_seq);
+ unpacked->bi_hash_seed = inode.v->bi_hash_seed;
+ unpacked->bi_flags = le64_to_cpu(inode.v->bi_flags);
+ unpacked->bi_mode = le16_to_cpu(inode.v->bi_mode);
+
+ return bch2_inode_unpack_v2(unpacked, inode.v->fields,
+ bkey_val_end(inode),
+ INODEv2_NR_FIELDS(inode.v));
+ }
+ default:
+ BUG();
+ }
+}
+
+int bch2_inode_peek(struct btree_trans *trans,
+ struct btree_iter *iter,
+ struct bch_inode_unpacked *inode,
+ subvol_inum inum, unsigned flags)
+{
+ struct bkey_s_c k;
+ u32 snapshot;
+ int ret;
+
+ ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
+ if (ret)
+ return ret;
+
+ bch2_trans_iter_init(trans, iter, BTREE_ID_inodes,
+ SPOS(0, inum.inum, snapshot),
+ flags|BTREE_ITER_CACHED);
+ k = bch2_btree_iter_peek_slot(iter);
+ ret = bkey_err(k);
+ if (ret)
+ goto err;
+
+ ret = bkey_is_inode(k.k) ? 0 : -ENOENT;
+ if (ret)
+ goto err;
+
+ ret = bch2_inode_unpack(k, inode);
+ if (ret)
+ goto err;
+
+ return 0;
+err:
+ bch2_trans_iter_exit(trans, iter);
+ return ret;
+}
+
+int bch2_inode_write(struct btree_trans *trans,
+ struct btree_iter *iter,
+ struct bch_inode_unpacked *inode)
+{
+ struct bkey_inode_buf *inode_p;
+
+ inode_p = bch2_trans_kmalloc(trans, sizeof(*inode_p));
+ if (IS_ERR(inode_p))
+ return PTR_ERR(inode_p);
+
+ bch2_inode_pack(trans->c, inode_p, inode);
+ inode_p->inode.k.p.snapshot = iter->snapshot;
+ return bch2_trans_update(trans, iter, &inode_p->inode.k_i, 0);
+}
- if (bkey_val_bytes(k.k) < sizeof(struct bch_inode))
- return "incorrect value size";
+static int __bch2_inode_invalid(struct bkey_s_c k, struct printbuf *err)
+{
+ struct bch_inode_unpacked unpacked;
- if (k.k->p.inode < BLOCKDEV_INODE_MAX)
- return "fs inode in blockdev range";
+ if (k.k->p.inode) {
+ pr_buf(err, "nonzero k.p.inode");
+ return -EINVAL;
+ }
- if (INODE_STR_HASH(inode.v) >= BCH_STR_HASH_NR)
- return "invalid str hash type";
+ if (k.k->p.offset < BLOCKDEV_INODE_MAX) {
+ pr_buf(err, "fs inode in blockdev range");
+ return -EINVAL;
+ }
- if (bch2_inode_unpack(inode, &unpacked))
- return "invalid variable length fields";
+ if (bch2_inode_unpack(k, &unpacked)){
+ pr_buf(err, "invalid variable length fields");
+ return -EINVAL;
+ }
- if (unpacked.bi_data_checksum >= BCH_CSUM_OPT_NR + 1)
- return "invalid data checksum type";
+ if (unpacked.bi_data_checksum >= BCH_CSUM_OPT_NR + 1) {
+ pr_buf(err, "invalid data checksum type (%u >= %u",
+ unpacked.bi_data_checksum, BCH_CSUM_OPT_NR + 1);
+ return -EINVAL;
+ }
- if (unpacked.bi_compression >= BCH_COMPRESSION_OPT_NR + 1)
- return "invalid data checksum type";
+ if (unpacked.bi_compression >= BCH_COMPRESSION_OPT_NR + 1) {
+ pr_buf(err, "invalid data checksum type (%u >= %u)",
+ unpacked.bi_compression, BCH_COMPRESSION_OPT_NR + 1);
+ return -EINVAL;
+ }
if ((unpacked.bi_flags & BCH_INODE_UNLINKED) &&
- unpacked.bi_nlink != 0)
- return "flagged as unlinked but bi_nlink != 0";
+ unpacked.bi_nlink != 0) {
+ pr_buf(err, "flagged as unlinked but bi_nlink != 0");
+ return -EINVAL;
+ }
+
+ if (unpacked.bi_subvol && !S_ISDIR(unpacked.bi_mode)) {
+ pr_buf(err, "subvolume root but not a directory");
+ return -EINVAL;
+ }
- return NULL;
+ return 0;
}
-void bch2_inode_to_text(struct printbuf *out, struct bch_fs *c,
- struct bkey_s_c k)
+int bch2_inode_invalid(const struct bch_fs *c, struct bkey_s_c k,
+ int rw, struct printbuf *err)
{
struct bkey_s_c_inode inode = bkey_s_c_to_inode(k);
- struct bch_inode_unpacked unpacked;
- if (bch2_inode_unpack(inode, &unpacked)) {
- pr_buf(out, "(unpack error)");
- return;
+ if (bkey_val_bytes(k.k) < sizeof(*inode.v)) {
+ pr_buf(err, "incorrect value size (%zu < %zu)",
+ bkey_val_bytes(k.k), sizeof(*inode.v));
+ return -EINVAL;
+ }
+
+ if (INODE_STR_HASH(inode.v) >= BCH_STR_HASH_NR) {
+ pr_buf(err, "invalid str hash type (%llu >= %u)",
+ INODE_STR_HASH(inode.v), BCH_STR_HASH_NR);
+ return -EINVAL;
+ }
+
+ return __bch2_inode_invalid(k, err);
+}
+
+int bch2_inode_v2_invalid(const struct bch_fs *c, struct bkey_s_c k,
+ int rw, struct printbuf *err)
+{
+ struct bkey_s_c_inode_v2 inode = bkey_s_c_to_inode_v2(k);
+
+ if (bkey_val_bytes(k.k) < sizeof(*inode.v)) {
+ pr_buf(err, "incorrect value size (%zu < %zu)",
+ bkey_val_bytes(k.k), sizeof(*inode.v));
+ return -EINVAL;
+ }
+
+ if (INODEv2_STR_HASH(inode.v) >= BCH_STR_HASH_NR) {
+ pr_buf(err, "invalid str hash type (%llu >= %u)",
+ INODEv2_STR_HASH(inode.v), BCH_STR_HASH_NR);
+ return -EINVAL;
}
+ return __bch2_inode_invalid(k, err);
+}
+
+static void __bch2_inode_unpacked_to_text(struct printbuf *out, struct bch_inode_unpacked *inode)
+{
+ pr_buf(out, "mode %o flags %x journal_seq %llu",
+ inode->bi_mode, inode->bi_flags,
+ inode->bi_journal_seq);
+
#define x(_name, _bits) \
- pr_buf(out, #_name ": %llu ", (u64) unpacked._name);
+ pr_buf(out, " "#_name " %llu", (u64) inode->_name);
BCH_INODE_FIELDS()
#undef x
}
-const char *bch2_inode_generation_invalid(const struct bch_fs *c,
- struct bkey_s_c k)
+void bch2_inode_unpacked_to_text(struct printbuf *out, struct bch_inode_unpacked *inode)
{
- if (k.k->p.offset)
- return "nonzero offset";
+ pr_buf(out, "inum: %llu ", inode->bi_inum);
+ __bch2_inode_unpacked_to_text(out, inode);
+}
- if (bkey_val_bytes(k.k) != sizeof(struct bch_inode_generation))
- return "incorrect value size";
+void bch2_inode_to_text(struct printbuf *out, struct bch_fs *c,
+ struct bkey_s_c k)
+{
+ struct bch_inode_unpacked inode;
- return NULL;
+ if (bch2_inode_unpack(k, &inode)) {
+ pr_buf(out, "(unpack error)");
+ return;
+ }
+
+ __bch2_inode_unpacked_to_text(out, &inode);
+}
+
+int bch2_inode_generation_invalid(const struct bch_fs *c, struct bkey_s_c k,
+ int rw, struct printbuf *err)
+{
+ if (k.k->p.inode) {
+ pr_buf(err, "nonzero k.p.inode");
+ return -EINVAL;
+ }
+
+ if (bkey_val_bytes(k.k) != sizeof(struct bch_inode_generation)) {
+ pr_buf(err, "incorrect value size (%zu != %zu)",
+ bkey_val_bytes(k.k), sizeof(struct bch_inode_generation));
+ return -EINVAL;
+ }
+
+ return 0;
}
void bch2_inode_generation_to_text(struct printbuf *out, struct bch_fs *c,
struct bkey_s_c k)
{
+ struct bkey_s_c_inode_generation gen = bkey_s_c_to_inode_generation(k);
+
+ pr_buf(out, "generation: %u", le32_to_cpu(gen.v->bi_generation));
}
-void bch2_inode_init(struct bch_fs *c, struct bch_inode_unpacked *inode_u,
- uid_t uid, gid_t gid, umode_t mode, dev_t rdev,
- struct bch_inode_unpacked *parent)
+void bch2_inode_init_early(struct bch_fs *c,
+ struct bch_inode_unpacked *inode_u)
{
- s64 now = timespec_to_bch2_time(c,
- timespec64_trunc(current_kernel_time64(),
- c->sb.time_precision));
+ enum bch_str_hash_type str_hash =
+ bch2_str_hash_opt_to_type(c, c->opts.str_hash);
memset(inode_u, 0, sizeof(*inode_u));
/* ick */
- inode_u->bi_flags |= c->opts.str_hash << INODE_STR_HASH_OFFSET;
+ inode_u->bi_flags |= str_hash << INODE_STR_HASH_OFFSET;
get_random_bytes(&inode_u->bi_hash_seed,
sizeof(inode_u->bi_hash_seed));
+}
+void bch2_inode_init_late(struct bch_inode_unpacked *inode_u, u64 now,
+ uid_t uid, gid_t gid, umode_t mode, dev_t rdev,
+ struct bch_inode_unpacked *parent)
+{
inode_u->bi_mode = mode;
inode_u->bi_uid = uid;
inode_u->bi_gid = gid;
inode_u->bi_ctime = now;
inode_u->bi_otime = now;
+ if (parent && parent->bi_mode & S_ISGID) {
+ inode_u->bi_gid = parent->bi_gid;
+ if (S_ISDIR(mode))
+ inode_u->bi_mode |= S_ISGID;
+ }
+
if (parent) {
#define x(_name, ...) inode_u->bi_##_name = parent->bi_##_name;
BCH_INODE_OPTS()
}
}
+void bch2_inode_init(struct bch_fs *c, struct bch_inode_unpacked *inode_u,
+ uid_t uid, gid_t gid, umode_t mode, dev_t rdev,
+ struct bch_inode_unpacked *parent)
+{
+ bch2_inode_init_early(c, inode_u);
+ bch2_inode_init_late(inode_u, bch2_current_time(c),
+ uid, gid, mode, rdev, parent);
+}
+
static inline u32 bkey_generation(struct bkey_s_c k)
{
switch (k.k->type) {
case KEY_TYPE_inode:
+ case KEY_TYPE_inode_v2:
BUG();
case KEY_TYPE_inode_generation:
return le32_to_cpu(bkey_s_c_to_inode_generation(k).v->bi_generation);
}
}
-int __bch2_inode_create(struct btree_trans *trans,
- struct bch_inode_unpacked *inode_u,
- u64 min, u64 max, u64 *hint)
+/*
+ * This just finds an empty slot:
+ */
+int bch2_inode_create(struct btree_trans *trans,
+ struct btree_iter *iter,
+ struct bch_inode_unpacked *inode_u,
+ u32 snapshot, u64 cpu)
{
struct bch_fs *c = trans->c;
- struct bkey_inode_buf *inode_p;
- struct btree_iter *iter;
- u64 start;
- int ret;
-
- if (!max)
- max = ULLONG_MAX;
-
- if (c->opts.inodes_32bit)
- max = min_t(u64, max, U32_MAX);
+ struct bkey_s_c k;
+ u64 min, max, start, pos, *hint;
+ int ret = 0;
+ unsigned bits = (c->opts.inodes_32bit ? 31 : 63);
+
+ if (c->opts.shard_inode_numbers) {
+ bits -= c->inode_shard_bits;
+
+ min = (cpu << bits);
+ max = (cpu << bits) | ~(ULLONG_MAX << bits);
+
+ min = max_t(u64, min, BLOCKDEV_INODE_MAX);
+ hint = c->unused_inode_hints + cpu;
+ } else {
+ min = BLOCKDEV_INODE_MAX;
+ max = ~(ULLONG_MAX << bits);
+ hint = c->unused_inode_hints;
+ }
start = READ_ONCE(*hint);
if (start >= max || start < min)
start = min;
- inode_p = bch2_trans_kmalloc(trans, sizeof(*inode_p));
- if (IS_ERR(inode_p))
- return PTR_ERR(inode_p);
-
- iter = bch2_trans_get_iter(trans,
- BTREE_ID_INODES, POS(start, 0),
- BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
- if (IS_ERR(iter))
- return PTR_ERR(iter);
+ pos = start;
+ bch2_trans_iter_init(trans, iter, BTREE_ID_inodes, POS(0, pos),
+ BTREE_ITER_ALL_SNAPSHOTS|
+ BTREE_ITER_INTENT);
again:
- while (1) {
- struct bkey_s_c k = bch2_btree_iter_peek_slot(iter);
+ while ((k = bch2_btree_iter_peek(iter)).k &&
+ !(ret = bkey_err(k)) &&
+ bkey_cmp(k.k->p, POS(0, max)) < 0) {
+ while (pos < iter->pos.offset) {
+ if (!bch2_btree_key_cache_find(c, BTREE_ID_inodes, POS(0, pos)))
+ goto found_slot;
+
+ pos++;
+ }
- ret = btree_iter_err(k);
- if (ret)
- return ret;
+ if (k.k->p.snapshot == snapshot &&
+ !bkey_is_inode(k.k) &&
+ !bch2_btree_key_cache_find(c, BTREE_ID_inodes, SPOS(0, pos, snapshot))) {
+ bch2_btree_iter_advance(iter);
+ continue;
+ }
- switch (k.k->type) {
- case KEY_TYPE_inode:
- /* slot used */
- if (iter->pos.inode >= max)
- goto out;
+ /*
+ * We don't need to iterate over keys in every snapshot once
+ * we've found just one:
+ */
+ pos = iter->pos.offset + 1;
+ bch2_btree_iter_set_pos(iter, POS(0, pos));
+ }
- bch2_btree_iter_next_slot(iter);
- break;
+ while (!ret && pos < max) {
+ if (!bch2_btree_key_cache_find(c, BTREE_ID_inodes, POS(0, pos)))
+ goto found_slot;
- default:
- *hint = k.k->p.inode;
- inode_u->bi_inum = k.k->p.inode;
- inode_u->bi_generation = bkey_generation(k);
+ pos++;
+ }
- bch2_inode_pack(inode_p, inode_u);
- bch2_trans_update(trans,
- BTREE_INSERT_ENTRY(iter, &inode_p->inode.k_i));
- return 0;
- }
+ if (!ret && start == min)
+ ret = -ENOSPC;
+
+ if (ret) {
+ bch2_trans_iter_exit(trans, iter);
+ return ret;
}
-out:
- if (start != min) {
- /* Retry from start */
- start = min;
- bch2_btree_iter_set_pos(iter, POS(start, 0));
- goto again;
+
+ /* Retry from start */
+ pos = start = min;
+ bch2_btree_iter_set_pos(iter, POS(0, pos));
+ goto again;
+found_slot:
+ bch2_btree_iter_set_pos(iter, SPOS(0, pos, snapshot));
+ k = bch2_btree_iter_peek_slot(iter);
+ ret = bkey_err(k);
+ if (ret) {
+ bch2_trans_iter_exit(trans, iter);
+ return ret;
}
- return -ENOSPC;
+ /* We may have raced while the iterator wasn't pointing at pos: */
+ if (bkey_is_inode(k.k) ||
+ bch2_btree_key_cache_find(c, BTREE_ID_inodes, k.k->p))
+ goto again;
+
+ *hint = k.k->p.offset;
+ inode_u->bi_inum = k.k->p.offset;
+ inode_u->bi_generation = bkey_generation(k);
+ return 0;
}
-int bch2_inode_create(struct bch_fs *c, struct bch_inode_unpacked *inode_u,
- u64 min, u64 max, u64 *hint)
+static int bch2_inode_delete_keys(struct btree_trans *trans,
+ subvol_inum inum, enum btree_id id)
{
- return bch2_trans_do(c, NULL, BTREE_INSERT_ATOMIC,
- __bch2_inode_create(&trans, inode_u, min, max, hint));
+ struct btree_iter iter;
+ struct bkey_s_c k;
+ struct bkey_i delete;
+ u32 snapshot;
+ int ret = 0;
+
+ /*
+ * We're never going to be deleting extents, no need to use an extent
+ * iterator:
+ */
+ bch2_trans_iter_init(trans, &iter, id, POS(inum.inum, 0),
+ BTREE_ITER_NOT_EXTENTS|
+ BTREE_ITER_INTENT);
+
+ while (1) {
+ bch2_trans_begin(trans);
+
+ ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
+ if (ret)
+ goto err;
+
+ bch2_btree_iter_set_snapshot(&iter, snapshot);
+
+ k = bch2_btree_iter_peek_upto(&iter, POS(inum.inum, U64_MAX));
+ ret = bkey_err(k);
+ if (ret)
+ goto err;
+
+ if (!k.k)
+ break;
+
+ bkey_init(&delete.k);
+ delete.k.p = iter.pos;
+
+ ret = bch2_trans_update(trans, &iter, &delete, 0) ?:
+ bch2_trans_commit(trans, NULL, NULL,
+ BTREE_INSERT_NOFAIL);
+err:
+ if (ret && ret != -EINTR)
+ break;
+ }
+
+ bch2_trans_iter_exit(trans, &iter);
+ return ret;
}
-int bch2_inode_rm(struct bch_fs *c, u64 inode_nr)
+int bch2_inode_rm(struct bch_fs *c, subvol_inum inum)
{
struct btree_trans trans;
- struct btree_iter *iter;
+ struct btree_iter iter = { NULL };
struct bkey_i_inode_generation delete;
- struct bpos start = POS(inode_nr, 0);
- struct bpos end = POS(inode_nr + 1, 0);
+ struct bch_inode_unpacked inode_u;
+ struct bkey_s_c k;
+ u32 snapshot;
int ret;
+ bch2_trans_init(&trans, c, 0, 1024);
+
/*
* If this was a directory, there shouldn't be any real dirents left -
* but there could be whiteouts (from hash collisions) that we should
* XXX: the dirent could ideally would delete whiteouts when they're no
* longer needed
*/
- ret = bch2_btree_delete_range(c, BTREE_ID_EXTENTS,
- start, end, NULL) ?:
- bch2_btree_delete_range(c, BTREE_ID_XATTRS,
- start, end, NULL) ?:
- bch2_btree_delete_range(c, BTREE_ID_DIRENTS,
- start, end, NULL);
+ ret = bch2_inode_delete_keys(&trans, inum, BTREE_ID_extents) ?:
+ bch2_inode_delete_keys(&trans, inum, BTREE_ID_xattrs) ?:
+ bch2_inode_delete_keys(&trans, inum, BTREE_ID_dirents);
if (ret)
- return ret;
-
- bch2_trans_init(&trans, c);
-
- iter = bch2_trans_get_iter(&trans, BTREE_ID_INODES, POS(inode_nr, 0),
- BTREE_ITER_SLOTS|BTREE_ITER_INTENT);
- do {
- struct bkey_s_c k = bch2_btree_iter_peek_slot(iter);
- u32 bi_generation = 0;
+ goto err;
+retry:
+ bch2_trans_begin(&trans);
- ret = btree_iter_err(k);
- if (ret)
- break;
+ ret = bch2_subvolume_get_snapshot(&trans, inum.subvol, &snapshot);
+ if (ret)
+ goto err;
- bch2_fs_inconsistent_on(k.k->type != KEY_TYPE_inode, c,
- "inode %llu not found when deleting",
- inode_nr);
+ bch2_trans_iter_init(&trans, &iter, BTREE_ID_inodes,
+ SPOS(0, inum.inum, snapshot),
+ BTREE_ITER_INTENT|BTREE_ITER_CACHED);
+ k = bch2_btree_iter_peek_slot(&iter);
- switch (k.k->type) {
- case KEY_TYPE_inode: {
- struct bch_inode_unpacked inode_u;
+ ret = bkey_err(k);
+ if (ret)
+ goto err;
+
+ if (!bkey_is_inode(k.k)) {
+ bch2_fs_inconsistent(trans.c,
+ "inode %llu not found when deleting",
+ inum.inum);
+ ret = -EIO;
+ goto err;
+ }
- if (!bch2_inode_unpack(bkey_s_c_to_inode(k), &inode_u))
- bi_generation = inode_u.bi_generation + 1;
- break;
- }
- case KEY_TYPE_inode_generation: {
- struct bkey_s_c_inode_generation g =
- bkey_s_c_to_inode_generation(k);
- bi_generation = le32_to_cpu(g.v->bi_generation);
- break;
- }
- }
+ bch2_inode_unpack(k, &inode_u);
- if (!bi_generation) {
- bkey_init(&delete.k);
- delete.k.p.inode = inode_nr;
- } else {
- bkey_inode_generation_init(&delete.k_i);
- delete.k.p.inode = inode_nr;
- delete.v.bi_generation = cpu_to_le32(bi_generation);
- }
+ /* Subvolume root? */
+ BUG_ON(inode_u.bi_subvol);
- bch2_trans_update(&trans,
- BTREE_INSERT_ENTRY(iter, &delete.k_i));
+ bkey_inode_generation_init(&delete.k_i);
+ delete.k.p = iter.pos;
+ delete.v.bi_generation = cpu_to_le32(inode_u.bi_generation + 1);
- ret = bch2_trans_commit(&trans, NULL, NULL,
- BTREE_INSERT_ATOMIC|
- BTREE_INSERT_NOFAIL);
- } while (ret == -EINTR);
+ ret = bch2_trans_update(&trans, &iter, &delete.k_i, 0) ?:
+ bch2_trans_commit(&trans, NULL, NULL,
+ BTREE_INSERT_NOFAIL);
+err:
+ bch2_trans_iter_exit(&trans, &iter);
+ if (ret == -EINTR)
+ goto retry;
bch2_trans_exit(&trans);
return ret;
}
-int bch2_inode_find_by_inum(struct bch_fs *c, u64 inode_nr,
- struct bch_inode_unpacked *inode)
+int bch2_inode_find_by_inum_trans(struct btree_trans *trans,
+ subvol_inum inum,
+ struct bch_inode_unpacked *inode)
{
struct btree_iter iter;
- struct bkey_s_c k;
- int ret = -ENOENT;
-
- for_each_btree_key(&iter, c, BTREE_ID_INODES,
- POS(inode_nr, 0),
- BTREE_ITER_SLOTS, k) {
- switch (k.k->type) {
- case KEY_TYPE_inode:
- ret = bch2_inode_unpack(bkey_s_c_to_inode(k), inode);
- break;
- default:
- /* hole, not found */
- break;
- }
-
- break;
- }
+ int ret;
- return bch2_btree_iter_unlock(&iter) ?: ret;
+ ret = bch2_inode_peek(trans, &iter, inode, inum, 0);
+ if (!ret)
+ bch2_trans_iter_exit(trans, &iter);
+ return ret;
}
-#ifdef CONFIG_BCACHEFS_DEBUG
-void bch2_inode_pack_test(void)
+int bch2_inode_find_by_inum(struct bch_fs *c, subvol_inum inum,
+ struct bch_inode_unpacked *inode)
{
- struct bch_inode_unpacked *u, test_inodes[] = {
- {
- .bi_atime = U64_MAX,
- .bi_ctime = U64_MAX,
- .bi_mtime = U64_MAX,
- .bi_otime = U64_MAX,
- .bi_size = U64_MAX,
- .bi_sectors = U64_MAX,
- .bi_uid = U32_MAX,
- .bi_gid = U32_MAX,
- .bi_nlink = U32_MAX,
- .bi_generation = U32_MAX,
- .bi_dev = U32_MAX,
- },
- };
-
- for (u = test_inodes;
- u < test_inodes + ARRAY_SIZE(test_inodes);
- u++) {
- struct bkey_inode_buf p;
-
- bch2_inode_pack(&p, u);
- }
+ return bch2_trans_do(c, NULL, NULL, 0,
+ bch2_inode_find_by_inum_trans(&trans, inum, inode));
}
-#endif