static void
dmu_tx_count_twig(dmu_tx_hold_t *txh, dnode_t *dn, dmu_buf_impl_t *db,
int level, uint64_t blkid, boolean_t freeable, uint64_t *history)
{
objset_t *os = dn->dn_objset;
dsl_dataset_t *ds = os->os_dsl_dataset;
int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
dmu_buf_impl_t *parent = NULL;
blkptr_t *bp = NULL;
uint64_t space;
if (level >= dn->dn_nlevels || history[level] == blkid)
return;
history[level] = blkid;
space = (level == 0) ? dn->dn_datablksz : (1ULL << dn->dn_indblkshift);
if (db == NULL || db == dn->dn_dbuf) {
ASSERT(level != 0);
db = NULL;
} else {
ASSERT(DB_DNODE(db) == dn);
ASSERT(db->db_level == level);
ASSERT(db->db.db_size == space);
ASSERT(db->db_blkid == blkid);
bp = db->db_blkptr;
parent = db->db_parent;
}
freeable = (bp && (freeable ||
dsl_dataset_block_freeable(ds, bp, bp->blk_birth)));
if (freeable) {
(void) refcount_add_many(&txh->txh_space_tooverwrite,
space, FTAG);
} else {
(void) refcount_add_many(&txh->txh_space_towrite,
space, FTAG);
}
if (bp) {
(void) refcount_add_many(&txh->txh_space_tounref,
bp_get_dsize(os->os_spa, bp), FTAG);
}
dmu_tx_count_twig(txh, dn, parent, level + 1,
blkid >> epbs, freeable, history);
}
int64_t
refcount_add(refcount_t *rc, void *holder)
{
return (refcount_add_many(rc, 1, holder));
}
/* ARGSUSED */
static void
dmu_tx_count_write(dmu_tx_hold_t *txh, uint64_t off, uint64_t len)
{
dnode_t *dn = txh->txh_dnode;
uint64_t start, end, i;
int min_bs, max_bs, min_ibs, max_ibs, epbs, bits;
int err = 0;
if (len == 0)
return;
min_bs = SPA_MINBLOCKSHIFT;
max_bs = highbit64(txh->txh_tx->tx_objset->os_recordsize) - 1;
min_ibs = DN_MIN_INDBLKSHIFT;
max_ibs = DN_MAX_INDBLKSHIFT;
if (dn) {
uint64_t history[DN_MAX_LEVELS];
int nlvls = dn->dn_nlevels;
int delta;
/*
* For i/o error checking, read the first and last level-0
* blocks (if they are not aligned), and all the level-1 blocks.
*/
if (dn->dn_maxblkid == 0) {
delta = dn->dn_datablksz;
start = (off < dn->dn_datablksz) ? 0 : 1;
end = (off+len <= dn->dn_datablksz) ? 0 : 1;
if (start == 0 && (off > 0 || len < dn->dn_datablksz)) {
err = dmu_tx_check_ioerr(NULL, dn, 0, 0);
if (err)
goto out;
delta -= off;
}
} else {
zio_t *zio = zio_root(dn->dn_objset->os_spa,
NULL, NULL, ZIO_FLAG_CANFAIL);
/* first level-0 block */
start = off >> dn->dn_datablkshift;
if (P2PHASE(off, dn->dn_datablksz) ||
len < dn->dn_datablksz) {
err = dmu_tx_check_ioerr(zio, dn, 0, start);
if (err)
goto out;
}
/* last level-0 block */
end = (off+len-1) >> dn->dn_datablkshift;
if (end != start && end <= dn->dn_maxblkid &&
P2PHASE(off+len, dn->dn_datablksz)) {
err = dmu_tx_check_ioerr(zio, dn, 0, end);
if (err)
goto out;
}
/* level-1 blocks */
if (nlvls > 1) {
int shft = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
for (i = (start>>shft)+1; i < end>>shft; i++) {
err = dmu_tx_check_ioerr(zio, dn, 1, i);
if (err)
goto out;
}
}
err = zio_wait(zio);
if (err)
goto out;
delta = P2NPHASE(off, dn->dn_datablksz);
}
min_ibs = max_ibs = dn->dn_indblkshift;
if (dn->dn_maxblkid > 0) {
/*
* The blocksize can't change,
* so we can make a more precise estimate.
*/
ASSERT(dn->dn_datablkshift != 0);
min_bs = max_bs = dn->dn_datablkshift;
} else {
/*
* The blocksize can increase up to the recordsize,
* or if it is already more than the recordsize,
* up to the next power of 2.
*/
min_bs = highbit64(dn->dn_datablksz - 1);
max_bs = MAX(max_bs, highbit64(dn->dn_datablksz - 1));
}
/*
* If this write is not off the end of the file
* we need to account for overwrites/unref.
*/
if (start <= dn->dn_maxblkid) {
for (int l = 0; l < DN_MAX_LEVELS; l++)
history[l] = -1ULL;
}
while (start <= dn->dn_maxblkid) {
dmu_buf_impl_t *db;
rw_enter(&dn->dn_struct_rwlock, RW_READER);
err = dbuf_hold_impl(dn, 0, start,
FALSE, FALSE, FTAG, &db);
rw_exit(&dn->dn_struct_rwlock);
if (err) {
txh->txh_tx->tx_err = err;
return;
}
dmu_tx_count_twig(txh, dn, db, 0, start, B_FALSE,
history);
dbuf_rele(db, FTAG);
if (++start > end) {
/*
* Account for new indirects appearing
* before this IO gets assigned into a txg.
*/
bits = 64 - min_bs;
epbs = min_ibs - SPA_BLKPTRSHIFT;
for (bits -= epbs * (nlvls - 1);
bits >= 0; bits -= epbs) {
(void) refcount_add_many(
&txh->txh_fudge,
1ULL << max_ibs, FTAG);
}
goto out;
}
off += delta;
if (len >= delta)
len -= delta;
delta = dn->dn_datablksz;
}
}
/* ARGSUSED */
static void
dmu_tx_count_write(dmu_tx_hold_t *txh, uint64_t off, uint64_t len)
{
dnode_t *dn = txh->txh_dnode;
int err = 0;
if (len == 0)
return;
(void) refcount_add_many(&txh->txh_space_towrite, len, FTAG);
if (refcount_count(&txh->txh_space_towrite) > 2 * DMU_MAX_ACCESS)
err = SET_ERROR(EFBIG);
if (dn == NULL)
return;
/*
* For i/o error checking, read the blocks that will be needed
* to perform the write: the first and last level-0 blocks (if
* they are not aligned, i.e. if they are partial-block writes),
* and all the level-1 blocks.
*/
if (dn->dn_maxblkid == 0) {
if (off < dn->dn_datablksz &&
(off > 0 || len < dn->dn_datablksz)) {
err = dmu_tx_check_ioerr(NULL, dn, 0, 0);
if (err != 0) {
txh->txh_tx->tx_err = err;
}
}
} else {
zio_t *zio = zio_root(dn->dn_objset->os_spa,
NULL, NULL, ZIO_FLAG_CANFAIL);
/* first level-0 block */
uint64_t start = off >> dn->dn_datablkshift;
if (P2PHASE(off, dn->dn_datablksz) || len < dn->dn_datablksz) {
err = dmu_tx_check_ioerr(zio, dn, 0, start);
if (err != 0) {
txh->txh_tx->tx_err = err;
}
}
/* last level-0 block */
uint64_t end = (off + len - 1) >> dn->dn_datablkshift;
if (end != start && end <= dn->dn_maxblkid &&
P2PHASE(off + len, dn->dn_datablksz)) {
err = dmu_tx_check_ioerr(zio, dn, 0, end);
if (err != 0) {
txh->txh_tx->tx_err = err;
}
}
/* level-1 blocks */
if (dn->dn_nlevels > 1) {
int shft = dn->dn_indblkshift - SPA_BLKPTRSHIFT;
for (uint64_t i = (start >> shft) + 1;
i < end >> shft; i++) {
err = dmu_tx_check_ioerr(zio, dn, 1, i);
if (err != 0) {
txh->txh_tx->tx_err = err;
}
}
}
err = zio_wait(zio);
if (err != 0) {
txh->txh_tx->tx_err = err;
}
}
