int
dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
void *tag, dmu_buf_t **dbp, int flags)
{
dnode_t *dn;
uint64_t blkid;
dmu_buf_impl_t *db;
int err;
int db_flags = DB_RF_CANFAIL;
if (flags & DMU_READ_NO_PREFETCH)
db_flags |= DB_RF_NOPREFETCH;
err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
blkid = dbuf_whichblock(dn, offset);
rw_enter(&dn->dn_struct_rwlock, RW_READER);
db = dbuf_hold(dn, blkid, tag);
rw_exit(&dn->dn_struct_rwlock);
if (db == NULL) {
err = EIO;
} else {
err = dbuf_read(db, NULL, db_flags);
if (err) {
dbuf_rele(db, tag);
db = NULL;
}
}
dnode_rele(dn, FTAG);
*dbp = &db->db; /* NULL db plus first field offset is NULL */
return (err);
}
int
dmu_buf_hold_noread(objset_t *os, uint64_t object, uint64_t offset,
void *tag, dmu_buf_t **dbp)
{
dnode_t *dn;
uint64_t blkid;
dmu_buf_impl_t *db;
int err;
err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
blkid = dbuf_whichblock(dn, offset);
rw_enter(&dn->dn_struct_rwlock, RW_READER);
db = dbuf_hold(dn, blkid, tag);
rw_exit(&dn->dn_struct_rwlock);
dnode_rele(dn, FTAG);
if (db == NULL) {
*dbp = NULL;
return (SET_ERROR(EIO));
}
*dbp = &db->db;
return (err);
}
int
dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
void *tag, dmu_buf_t **dbp)
{
dnode_t *dn;
uint64_t blkid;
dmu_buf_impl_t *db;
int err;
err = dnode_hold(os->os, object, FTAG, &dn);
if (err)
return (err);
blkid = dbuf_whichblock(dn, offset);
rw_enter(&dn->dn_struct_rwlock, RW_READER);
db = dbuf_hold(dn, blkid, tag);
rw_exit(&dn->dn_struct_rwlock);
if (db == NULL) {
err = EIO;
} else {
err = dbuf_read(db, NULL, DB_RF_CANFAIL);
if (err) {
dbuf_rele(db, tag);
db = NULL;
}
}
dnode_rele(dn, FTAG);
*dbp = &db->db;
return (err);
}
/*
* returns ENOENT, EIO, or 0.
*/
int
dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **dbp)
{
dnode_t *dn;
dmu_buf_impl_t *db;
int error;
error = dnode_hold(os->os, object, FTAG, &dn);
if (error)
return (error);
rw_enter(&dn->dn_struct_rwlock, RW_READER);
if (dn->dn_bonus == NULL) {
rw_exit(&dn->dn_struct_rwlock);
rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
if (dn->dn_bonus == NULL)
dbuf_create_bonus(dn);
}
db = dn->dn_bonus;
rw_exit(&dn->dn_struct_rwlock);
/* as long as the bonus buf is held, the dnode will be held */
if (refcount_add(&db->db_holds, tag) == 1)
VERIFY(dnode_add_ref(dn, db));
dnode_rele(dn, FTAG);
VERIFY(0 == dbuf_read(db, NULL, DB_RF_MUST_SUCCEED));
*dbp = &db->db;
return (0);
}
int
dmu_rm_spill(objset_t *os, uint64_t object, dmu_tx_t *tx)
{
dnode_t *dn;
int error;
error = dnode_hold(os, object, FTAG, &dn);
dbuf_rm_spill(dn, tx);
rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
dnode_rm_spill(dn, tx);
rw_exit(&dn->dn_struct_rwlock);
dnode_rele(dn, FTAG);
return (error);
}
static dmu_tx_hold_t *
dmu_tx_hold_object_impl(dmu_tx_t *tx, objset_t *os, uint64_t object,
enum dmu_tx_hold_type type, uint64_t arg1, uint64_t arg2)
{
dmu_tx_hold_t *txh;
dnode_t *dn = NULL;
int err;
if (object != DMU_NEW_OBJECT) {
err = dnode_hold(os, object, tx, &dn);
if (err) {
tx->tx_err = err;
return (NULL);
}
if (err == 0 && tx->tx_txg != 0) {
mutex_enter(&dn->dn_mtx);
/*
* dn->dn_assigned_txg == tx->tx_txg doesn't pose a
* problem, but there's no way for it to happen (for
* now, at least).
*/
ASSERT(dn->dn_assigned_txg == 0);
dn->dn_assigned_txg = tx->tx_txg;
(void) refcount_add(&dn->dn_tx_holds, tx);
mutex_exit(&dn->dn_mtx);
}
}
txh = kmem_zalloc(sizeof (dmu_tx_hold_t), KM_SLEEP);
txh->txh_tx = tx;
txh->txh_dnode = dn;
refcount_create(&txh->txh_space_towrite);
refcount_create(&txh->txh_space_tofree);
refcount_create(&txh->txh_space_tooverwrite);
refcount_create(&txh->txh_space_tounref);
refcount_create(&txh->txh_memory_tohold);
refcount_create(&txh->txh_fudge);
#ifdef ZFS_DEBUG
txh->txh_type = type;
txh->txh_arg1 = arg1;
txh->txh_arg2 = arg2;
#endif
list_insert_tail(&tx->tx_holds, txh);
return (txh);
}
static dmu_tx_hold_t *
dmu_tx_hold_object_impl(dmu_tx_t *tx, objset_t *os, uint64_t object,
enum dmu_tx_hold_type type, uint64_t arg1, uint64_t arg2)
{
dnode_t *dn = NULL;
dmu_tx_hold_t *txh;
int err;
if (object != DMU_NEW_OBJECT) {
err = dnode_hold(os, object, FTAG, &dn);
if (err != 0) {
tx->tx_err = err;
return (NULL);
}
}
txh = dmu_tx_hold_dnode_impl(tx, dn, type, arg1, arg2);
if (dn != NULL)
dnode_rele(dn, FTAG);
return (txh);
}
/*
* returns ENOENT, EIO, or 0.
*/
int
dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **dbp)
{
dnode_t *dn;
dmu_buf_impl_t *db;
int error;
error = dnode_hold(os, object, FTAG, &dn);
if (error)
return (error);
rw_enter(&dn->dn_struct_rwlock, RW_READER);
if (dn->dn_bonus == NULL) {
rw_exit(&dn->dn_struct_rwlock);
rw_enter(&dn->dn_struct_rwlock, RW_WRITER);
if (dn->dn_bonus == NULL)
dbuf_create_bonus(dn);
}
db = dn->dn_bonus;
/* as long as the bonus buf is held, the dnode will be held */
if (refcount_add(&db->db_holds, tag) == 1) {
VERIFY(dnode_add_ref(dn, db));
atomic_inc_32(&dn->dn_dbufs_count);
}
/*
* Wait to drop dn_struct_rwlock until after adding the bonus dbuf's
* hold and incrementing the dbuf count to ensure that dnode_move() sees
* a dnode hold for every dbuf.
*/
rw_exit(&dn->dn_struct_rwlock);
dnode_rele(dn, FTAG);
VERIFY(0 == dbuf_read(db, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH));
*dbp = &db->db;
return (0);
}
/*
* Calculate the real range based on the type, level, and range given.
*/
static int
calculate_range(const char *dataset, err_type_t type, int level, char *range,
zinject_record_t *record)
{
objset_t *os = NULL;
dnode_t *dn = NULL;
int err;
int ret = -1;
/*
* Determine the numeric range from the string.
*/
if (range == NULL) {
/*
* If range is unspecified, set the range to [0,-1], which
* indicates that the whole object should be treated as an
* error.
*/
record->zi_start = 0;
record->zi_end = -1ULL;
} else {
char *end;
/* XXX add support for suffixes */
record->zi_start = strtoull(range, &end, 10);
if (*end == '\0')
record->zi_end = record->zi_start + 1;
else if (*end == ',')
record->zi_end = strtoull(end + 1, &end, 10);
if (*end != '\0') {
(void) fprintf(stderr, "invalid range '%s': must be "
"a numeric range of the form 'start[,end]'\n",
range);
goto out;
}
}
switch (type) {
case TYPE_DATA:
break;
case TYPE_DNODE:
/*
* If this is a request to inject faults into the dnode, then we
* must translate the current (objset,object) pair into an
* offset within the metadnode for the objset. Specifying any
* kind of range with type 'dnode' is illegal.
*/
if (range != NULL) {
(void) fprintf(stderr, "range cannot be specified when "
"type is 'dnode'\n");
goto out;
}
record->zi_start = record->zi_object * sizeof (dnode_phys_t);
record->zi_end = record->zi_start + sizeof (dnode_phys_t);
record->zi_object = 0;
break;
}
/*
* Get the dnode associated with object, so we can calculate the block
* size.
*/
if ((err = dmu_objset_own(dataset, DMU_OST_ANY,
B_TRUE, FTAG, &os)) != 0) {
(void) fprintf(stderr, "cannot open dataset '%s': %s\n",
dataset, strerror(err));
goto out;
}
if (record->zi_object == 0) {
dn = DMU_META_DNODE(os);
} else {
err = dnode_hold(os, record->zi_object, FTAG, &dn);
if (err != 0) {
(void) fprintf(stderr, "failed to hold dnode "
"for object %llu\n",
(u_longlong_t)record->zi_object);
goto out;
}
}
ziprintf("data shift: %d\n", (int)dn->dn_datablkshift);
ziprintf(" ind shift: %d\n", (int)dn->dn_indblkshift);
/*
* Translate range into block IDs.
*/
if (record->zi_start != 0 || record->zi_end != -1ULL) {
record->zi_start >>= dn->dn_datablkshift;
record->zi_end >>= dn->dn_datablkshift;
}