// SPDX-License-Identifier: GPL-2.0
#include "bcachefs.h"
+#include "backpointers.h"
#include "bkey_methods.h"
#include "btree_types.h"
#include "alloc_background.h"
#include "error.h"
#include "extents.h"
#include "inode.h"
+#include "io_misc.h"
+#include "lru.h"
#include "quota.h"
#include "reflink.h"
+#include "snapshot.h"
+#include "subvolume.h"
#include "xattr.h"
const char * const bch2_bkey_types[] = {
NULL
};
-static const char *deleted_key_invalid(const struct bch_fs *c,
- struct bkey_s_c k)
+static int deleted_key_invalid(const struct bch_fs *c, struct bkey_s_c k,
+ enum bkey_invalid_flags flags, struct printbuf *err)
{
- return NULL;
+ return 0;
}
-#define bch2_bkey_ops_deleted (struct bkey_ops) { \
+#define bch2_bkey_ops_deleted ((struct bkey_ops) { \
.key_invalid = deleted_key_invalid, \
-}
+})
-#define bch2_bkey_ops_discard (struct bkey_ops) { \
+#define bch2_bkey_ops_whiteout ((struct bkey_ops) { \
.key_invalid = deleted_key_invalid, \
-}
+})
-static const char *empty_val_key_invalid(const struct bch_fs *c, struct bkey_s_c k)
+static int empty_val_key_invalid(const struct bch_fs *c, struct bkey_s_c k,
+ enum bkey_invalid_flags flags, struct printbuf *err)
{
- if (bkey_val_bytes(k.k))
- return "value size should be zero";
+ if (bkey_val_bytes(k.k)) {
+ prt_printf(err, "incorrect value size (%zu != 0)",
+ bkey_val_bytes(k.k));
+ return -BCH_ERR_invalid_bkey;
+ }
- return NULL;
+ return 0;
}
-#define bch2_bkey_ops_error (struct bkey_ops) { \
+#define bch2_bkey_ops_error ((struct bkey_ops) { \
.key_invalid = empty_val_key_invalid, \
-}
+})
-static const char *key_type_cookie_invalid(const struct bch_fs *c,
- struct bkey_s_c k)
+static int key_type_cookie_invalid(const struct bch_fs *c, struct bkey_s_c k,
+ enum bkey_invalid_flags flags, struct printbuf *err)
{
- if (bkey_val_bytes(k.k) != sizeof(struct bch_cookie))
- return "incorrect value size";
-
- return NULL;
+ return 0;
}
-#define bch2_bkey_ops_cookie (struct bkey_ops) { \
- .key_invalid = key_type_cookie_invalid, \
-}
+#define bch2_bkey_ops_cookie ((struct bkey_ops) { \
+ .key_invalid = key_type_cookie_invalid, \
+ .min_val_size = 8, \
+})
-#define bch2_bkey_ops_hash_whiteout (struct bkey_ops) { \
+#define bch2_bkey_ops_hash_whiteout ((struct bkey_ops) {\
.key_invalid = empty_val_key_invalid, \
-}
+})
-static const char *key_type_inline_data_invalid(const struct bch_fs *c,
- struct bkey_s_c k)
+static int key_type_inline_data_invalid(const struct bch_fs *c, struct bkey_s_c k,
+ enum bkey_invalid_flags flags, struct printbuf *err)
{
- return NULL;
+ return 0;
}
static void key_type_inline_data_to_text(struct printbuf *out, struct bch_fs *c,
struct bkey_s_c_inline_data d = bkey_s_c_to_inline_data(k);
unsigned datalen = bkey_inline_data_bytes(k.k);
- pr_buf(out, "datalen %u: %*phN",
+ prt_printf(out, "datalen %u: %*phN",
datalen, min(datalen, 32U), d.v->data);
}
-#define bch2_bkey_ops_inline_data (struct bkey_ops) { \
+#define bch2_bkey_ops_inline_data ((struct bkey_ops) { \
.key_invalid = key_type_inline_data_invalid, \
.val_to_text = key_type_inline_data_to_text, \
+})
+
+static int key_type_set_invalid(const struct bch_fs *c, struct bkey_s_c k,
+ enum bkey_invalid_flags flags, struct printbuf *err)
+{
+ if (bkey_val_bytes(k.k)) {
+ prt_printf(err, "incorrect value size (%zu != %zu)",
+ bkey_val_bytes(k.k), sizeof(struct bch_cookie));
+ return -BCH_ERR_invalid_bkey;
+ }
+
+ return 0;
+}
+
+static bool key_type_set_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r)
+{
+ bch2_key_resize(l.k, l.k->size + r.k->size);
+ return true;
}
+#define bch2_bkey_ops_set ((struct bkey_ops) { \
+ .key_invalid = key_type_set_invalid, \
+ .key_merge = key_type_set_merge, \
+})
+
const struct bkey_ops bch2_bkey_ops[] = {
#define x(name, nr) [KEY_TYPE_##name] = bch2_bkey_ops_##name,
BCH_BKEY_TYPES()
#undef x
};
-const char *bch2_bkey_val_invalid(struct bch_fs *c, struct bkey_s_c k)
+const struct bkey_ops bch2_bkey_null_ops = {
+ .min_val_size = U8_MAX,
+};
+
+int bch2_bkey_val_invalid(struct bch_fs *c, struct bkey_s_c k,
+ enum bkey_invalid_flags flags,
+ struct printbuf *err)
{
- if (k.k->type >= KEY_TYPE_MAX)
- return "invalid type";
+ const struct bkey_ops *ops = bch2_bkey_type_ops(k.k->type);
+
+ if (bkey_val_bytes(k.k) < ops->min_val_size) {
+ prt_printf(err, "bad val size (%zu < %u)",
+ bkey_val_bytes(k.k), ops->min_val_size);
+ return -BCH_ERR_invalid_bkey;
+ }
- return bch2_bkey_ops[k.k->type].key_invalid(c, k);
+ if (!ops->key_invalid)
+ return 0;
+
+ return ops->key_invalid(c, k, flags, err);
}
-static unsigned bch2_key_types_allowed[] = {
- [BKEY_TYPE_extents] =
- (1U << KEY_TYPE_error)|
- (1U << KEY_TYPE_cookie)|
- (1U << KEY_TYPE_extent)|
- (1U << KEY_TYPE_reservation)|
- (1U << KEY_TYPE_reflink_p)|
- (1U << KEY_TYPE_inline_data),
- [BKEY_TYPE_inodes] =
- (1U << KEY_TYPE_inode)|
- (1U << KEY_TYPE_inode_generation),
- [BKEY_TYPE_dirents] =
- (1U << KEY_TYPE_hash_whiteout)|
- (1U << KEY_TYPE_dirent),
- [BKEY_TYPE_xattrs] =
- (1U << KEY_TYPE_cookie)|
- (1U << KEY_TYPE_hash_whiteout)|
- (1U << KEY_TYPE_xattr),
- [BKEY_TYPE_alloc] =
- (1U << KEY_TYPE_alloc)|
- (1U << KEY_TYPE_alloc_v2),
- [BKEY_TYPE_quotas] =
- (1U << KEY_TYPE_quota),
- [BKEY_TYPE_stripes] =
- (1U << KEY_TYPE_stripe),
- [BKEY_TYPE_reflink] =
- (1U << KEY_TYPE_reflink_v)|
- (1U << KEY_TYPE_indirect_inline_data),
+static u64 bch2_key_types_allowed[] = {
+#define x(name, nr, flags, keys) [BKEY_TYPE_##name] = BIT_ULL(KEY_TYPE_deleted)|keys,
+ BCH_BTREE_IDS()
+#undef x
[BKEY_TYPE_btree] =
- (1U << KEY_TYPE_btree_ptr)|
- (1U << KEY_TYPE_btree_ptr_v2),
+ BIT_ULL(KEY_TYPE_deleted)|
+ BIT_ULL(KEY_TYPE_btree_ptr)|
+ BIT_ULL(KEY_TYPE_btree_ptr_v2),
};
-const char *__bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
- enum btree_node_type type)
+int __bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
+ enum btree_node_type type,
+ enum bkey_invalid_flags flags,
+ struct printbuf *err)
{
- unsigned key_types_allowed = (1U << KEY_TYPE_deleted)|
- bch2_key_types_allowed[type] ;
-
- if (k.k->u64s < BKEY_U64s)
- return "u64s too small";
-
- if (!(key_types_allowed & (1U << k.k->type)))
- return "invalid key type for this btree";
+ if (k.k->u64s < BKEY_U64s) {
+ prt_printf(err, "u64s too small (%u < %zu)", k.k->u64s, BKEY_U64s);
+ return -BCH_ERR_invalid_bkey;
+ }
- if (type == BKEY_TYPE_btree &&
- bkey_val_u64s(k.k) > BKEY_BTREE_PTR_VAL_U64s_MAX)
- return "value too big";
+ if (flags & BKEY_INVALID_COMMIT &&
+ !(bch2_key_types_allowed[type] & BIT_ULL(k.k->type))) {
+ prt_printf(err, "invalid key type for btree %s (%s)",
+ bch2_btree_ids[type], bch2_bkey_types[k.k->type]);
+ return -BCH_ERR_invalid_bkey;
+ }
- if (btree_node_type_is_extents(type)) {
- if ((k.k->size == 0) != bkey_deleted(k.k))
- return "bad size field";
+ if (btree_node_type_is_extents(type) && !bkey_whiteout(k.k)) {
+ if (k.k->size == 0) {
+ prt_printf(err, "size == 0");
+ return -BCH_ERR_invalid_bkey;
+ }
- if (k.k->size > k.k->p.offset)
- return "size greater than offset";
+ if (k.k->size > k.k->p.offset) {
+ prt_printf(err, "size greater than offset (%u > %llu)",
+ k.k->size, k.k->p.offset);
+ return -BCH_ERR_invalid_bkey;
+ }
} else {
- if (k.k->size)
- return "nonzero size field";
+ if (k.k->size) {
+ prt_printf(err, "size != 0");
+ return -BCH_ERR_invalid_bkey;
+ }
}
- if (type != BKEY_TYPE_btree &&
- !btree_type_has_snapshots(type) &&
- k.k->p.snapshot)
- return "nonzero snapshot";
-
- if (type != BKEY_TYPE_btree &&
- btree_type_has_snapshots(type) &&
- k.k->p.snapshot != U32_MAX)
- return "invalid snapshot field";
+ if (type != BKEY_TYPE_btree) {
+ if (!btree_type_has_snapshots((enum btree_id) type) &&
+ k.k->p.snapshot) {
+ prt_printf(err, "nonzero snapshot");
+ return -BCH_ERR_invalid_bkey;
+ }
- if (type != BKEY_TYPE_btree &&
- !bkey_cmp(k.k->p, POS_MAX))
- return "POS_MAX key";
+ if (btree_type_has_snapshots((enum btree_id) type) &&
+ !k.k->p.snapshot) {
+ prt_printf(err, "snapshot == 0");
+ return -BCH_ERR_invalid_bkey;
+ }
- return NULL;
-}
+ if (bkey_eq(k.k->p, POS_MAX)) {
+ prt_printf(err, "key at POS_MAX");
+ return -BCH_ERR_invalid_bkey;
+ }
+ }
-const char *bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
- enum btree_node_type type)
-{
- return __bch2_bkey_invalid(c, k, type) ?:
- bch2_bkey_val_invalid(c, k);
+ return 0;
}
-const char *bch2_bkey_in_btree_node(struct btree *b, struct bkey_s_c k)
+int bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
+ enum btree_node_type type,
+ enum bkey_invalid_flags flags,
+ struct printbuf *err)
{
- if (bpos_cmp(k.k->p, b->data->min_key) < 0)
- return "key before start of btree node";
-
- if (bpos_cmp(k.k->p, b->data->max_key) > 0)
- return "key past end of btree node";
-
- return NULL;
+ return __bch2_bkey_invalid(c, k, type, flags, err) ?:
+ bch2_bkey_val_invalid(c, k, flags, err);
}
-void bch2_bkey_debugcheck(struct bch_fs *c, struct btree *b, struct bkey_s_c k)
+int bch2_bkey_in_btree_node(struct btree *b, struct bkey_s_c k,
+ struct printbuf *err)
{
- const char *invalid;
-
- BUG_ON(!k.k->u64s);
-
- invalid = bch2_bkey_invalid(c, k, btree_node_type(b)) ?:
- bch2_bkey_in_btree_node(b, k);
- if (invalid) {
- char buf[160];
+ if (bpos_lt(k.k->p, b->data->min_key)) {
+ prt_printf(err, "key before start of btree node");
+ return -BCH_ERR_invalid_bkey;
+ }
- bch2_bkey_val_to_text(&PBUF(buf), c, k);
- bch2_fs_inconsistent(c, "invalid bkey %s: %s", buf, invalid);
+ if (bpos_gt(k.k->p, b->data->max_key)) {
+ prt_printf(err, "key past end of btree node");
+ return -BCH_ERR_invalid_bkey;
}
+
+ return 0;
}
void bch2_bpos_to_text(struct printbuf *out, struct bpos pos)
{
- if (!bpos_cmp(pos, POS_MIN))
- pr_buf(out, "POS_MIN");
- else if (!bpos_cmp(pos, POS_MAX))
- pr_buf(out, "POS_MAX");
+ if (bpos_eq(pos, POS_MIN))
+ prt_printf(out, "POS_MIN");
+ else if (bpos_eq(pos, POS_MAX))
+ prt_printf(out, "POS_MAX");
+ else if (bpos_eq(pos, SPOS_MAX))
+ prt_printf(out, "SPOS_MAX");
else {
if (pos.inode == U64_MAX)
- pr_buf(out, "U64_MAX");
+ prt_printf(out, "U64_MAX");
else
- pr_buf(out, "%llu", pos.inode);
- pr_buf(out, ":");
+ prt_printf(out, "%llu", pos.inode);
+ prt_printf(out, ":");
if (pos.offset == U64_MAX)
- pr_buf(out, "U64_MAX");
+ prt_printf(out, "U64_MAX");
else
- pr_buf(out, "%llu", pos.offset);
- pr_buf(out, ":");
+ prt_printf(out, "%llu", pos.offset);
+ prt_printf(out, ":");
if (pos.snapshot == U32_MAX)
- pr_buf(out, "U32_MAX");
+ prt_printf(out, "U32_MAX");
else
- pr_buf(out, "%u", pos.snapshot);
+ prt_printf(out, "%u", pos.snapshot);
}
}
void bch2_bkey_to_text(struct printbuf *out, const struct bkey *k)
{
if (k) {
- pr_buf(out, "u64s %u type ", k->u64s);
+ prt_printf(out, "u64s %u type ", k->u64s);
if (k->type < KEY_TYPE_MAX)
- pr_buf(out, "%s ", bch2_bkey_types[k->type]);
+ prt_printf(out, "%s ", bch2_bkey_types[k->type]);
else
- pr_buf(out, "%u ", k->type);
+ prt_printf(out, "%u ", k->type);
bch2_bpos_to_text(out, k->p);
- pr_buf(out, " len %u ver %llu", k->size, k->version.lo);
+ prt_printf(out, " len %u ver %llu", k->size, k->version.lo);
} else {
- pr_buf(out, "(null)");
+ prt_printf(out, "(null)");
}
}
void bch2_val_to_text(struct printbuf *out, struct bch_fs *c,
struct bkey_s_c k)
{
- if (k.k->type < KEY_TYPE_MAX) {
- const struct bkey_ops *ops = &bch2_bkey_ops[k.k->type];
+ const struct bkey_ops *ops = bch2_bkey_type_ops(k.k->type);
- if (likely(ops->val_to_text))
- ops->val_to_text(out, c, k);
- } else {
- pr_buf(out, "(invalid type %u)", k.k->type);
- }
+ if (likely(ops->val_to_text))
+ ops->val_to_text(out, c, k);
}
void bch2_bkey_val_to_text(struct printbuf *out, struct bch_fs *c,
bch2_bkey_to_text(out, k.k);
if (bkey_val_bytes(k.k)) {
- pr_buf(out, ": ");
+ prt_printf(out, ": ");
bch2_val_to_text(out, c, k);
}
}
void bch2_bkey_swab_val(struct bkey_s k)
{
- const struct bkey_ops *ops = &bch2_bkey_ops[k.k->type];
+ const struct bkey_ops *ops = bch2_bkey_type_ops(k.k->type);
if (ops->swab)
ops->swab(k);
bool bch2_bkey_normalize(struct bch_fs *c, struct bkey_s k)
{
- const struct bkey_ops *ops = &bch2_bkey_ops[k.k->type];
+ const struct bkey_ops *ops = bch2_bkey_type_ops(k.k->type);
return ops->key_normalize
? ops->key_normalize(c, k)
bool bch2_bkey_merge(struct bch_fs *c, struct bkey_s l, struct bkey_s_c r)
{
- const struct bkey_ops *ops = &bch2_bkey_ops[l.k->type];
+ const struct bkey_ops *ops = bch2_bkey_type_ops(l.k->type);
- return bch2_bkey_maybe_mergable(l.k, r.k) && ops->key_merge(c, l, r);
+ return ops->key_merge &&
+ bch2_bkey_maybe_mergable(l.k, r.k) &&
+ (u64) l.k->size + r.k->size <= KEY_SIZE_MAX &&
+ !bch2_key_merging_disabled &&
+ ops->key_merge(c, l, r);
}
static const struct old_bkey_type {
{
const struct bkey_ops *ops;
struct bkey uk;
- struct bkey_s u;
unsigned nr_compat = 5;
int i;
btree_id == BTREE_ID_inodes) {
if (!bkey_packed(k)) {
struct bkey_i *u = packed_to_bkey(k);
+
swap(u->k.p.inode, u->k.p.offset);
} else if (f->bits_per_field[BKEY_FIELD_INODE] &&
f->bits_per_field[BKEY_FIELD_OFFSET]) {
u->k.p.snapshot = write
? 0 : U32_MAX;
} else {
- u64 min_packed = f->field_offset[BKEY_FIELD_SNAPSHOT];
+ u64 min_packed = le64_to_cpu(f->field_offset[BKEY_FIELD_SNAPSHOT]);
u64 max_packed = min_packed +
~(~0ULL << f->bits_per_field[BKEY_FIELD_SNAPSHOT]);
}
break;
- case 4:
+ case 4: {
+ struct bkey_s u;
+
if (!bkey_packed(k)) {
u = bkey_i_to_s(packed_to_bkey(k));
} else {
if (big_endian != CPU_BIG_ENDIAN)
bch2_bkey_swab_val(u);
- ops = &bch2_bkey_ops[k->type];
+ ops = bch2_bkey_type_ops(k->type);
if (ops->compat)
ops->compat(btree_id, version, big_endian, write, u);
break;
+ }
default:
BUG();
}