void
zfs_sa_symlink(znode_t *zp, char *link, int len, dmu_tx_t *tx)
{
dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
if (ZFS_OLD_ZNODE_PHYS_SIZE + len <= dmu_bonus_max()) {
VERIFY(dmu_set_bonus(db,
len + ZFS_OLD_ZNODE_PHYS_SIZE, tx) == 0);
if (len) {
bcopy(link, (caddr_t)db->db_data +
ZFS_OLD_ZNODE_PHYS_SIZE, len);
}
} else {
dmu_buf_t *dbp;
zfs_grow_blocksize(zp, len, tx);
VERIFY(0 == dmu_buf_hold(zp->z_zfsvfs->z_os,
zp->z_id, 0, FTAG, &dbp, DMU_READ_NO_PREFETCH));
dmu_buf_will_dirty(dbp, tx);
ASSERT3U(len, <=, dbp->db_size);
bcopy(link, dbp->db_data, len);
dmu_buf_rele(dbp, FTAG);
}
}
void
fzap_upgrade(zap_t *zap, dmu_tx_t *tx, zap_flags_t flags)
{
dmu_buf_t *db;
zap_leaf_t *l;
int i;
zap_phys_t *zp;
ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
zap->zap_ismicro = FALSE;
(void) dmu_buf_update_user(zap->zap_dbuf, zap, zap,
&zap->zap_f.zap_phys, zap_evict);
mutex_init(&zap->zap_f.zap_num_entries_mtx, 0, 0, 0);
zap->zap_f.zap_block_shift = highbit(zap->zap_dbuf->db_size) - 1;
zp = zap->zap_f.zap_phys;
/*
* explicitly zero it since it might be coming from an
* initialized microzap
*/
bzero(zap->zap_dbuf->db_data, zap->zap_dbuf->db_size);
zp->zap_block_type = ZBT_HEADER;
zp->zap_magic = ZAP_MAGIC;
zp->zap_ptrtbl.zt_shift = ZAP_EMBEDDED_PTRTBL_SHIFT(zap);
zp->zap_freeblk = 2; /* block 1 will be the first leaf */
zp->zap_num_leafs = 1;
zp->zap_num_entries = 0;
zp->zap_salt = zap->zap_salt;
zp->zap_normflags = zap->zap_normflags;
zp->zap_flags = flags;
/* block 1 will be the first leaf */
for (i = 0; i < (1<<zp->zap_ptrtbl.zt_shift); i++)
ZAP_EMBEDDED_PTRTBL_ENT(zap, i) = 1;
/*
* set up block 1 - the first leaf
*/
VERIFY(0 == dmu_buf_hold(zap->zap_objset, zap->zap_object,
1<<FZAP_BLOCK_SHIFT(zap), FTAG, &db, DMU_READ_NO_PREFETCH));
dmu_buf_will_dirty(db, tx);
l = kmem_zalloc(sizeof (zap_leaf_t), KM_PUSHPAGE);
l->l_dbuf = db;
l->l_phys = db->db_data;
zap_leaf_init(l, zp->zap_normflags != 0);
kmem_free(l, sizeof (zap_leaf_t));
dmu_buf_rele(db, FTAG);
}
static int
zap_table_store(zap_t *zap, zap_table_phys_t *tbl, uint64_t idx, uint64_t val,
dmu_tx_t *tx)
{
int err;
uint64_t blk, off;
int bs = FZAP_BLOCK_SHIFT(zap);
dmu_buf_t *db;
ASSERT(RW_LOCK_HELD(&zap->zap_rwlock));
ASSERT(tbl->zt_blk != 0);
dprintf("storing %llx at index %llx\n", val, idx);
blk = idx >> (bs-3);
off = idx & ((1<<(bs-3))-1);
err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
(tbl->zt_blk + blk) << bs, FTAG, &db, DMU_READ_NO_PREFETCH);
if (err)
return (err);
dmu_buf_will_dirty(db, tx);
if (tbl->zt_nextblk != 0) {
uint64_t idx2 = idx * 2;
uint64_t blk2 = idx2 >> (bs-3);
uint64_t off2 = idx2 & ((1<<(bs-3))-1);
dmu_buf_t *db2;
err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
(tbl->zt_nextblk + blk2) << bs, FTAG, &db2,
DMU_READ_NO_PREFETCH);
if (err) {
dmu_buf_rele(db, FTAG);
return (err);
}
dmu_buf_will_dirty(db2, tx);
((uint64_t *)db2->db_data)[off2] = val;
((uint64_t *)db2->db_data)[off2+1] = val;
dmu_buf_rele(db2, FTAG);
}
/*
* Get data to generate a TX_WRITE intent log record.
*/
static int
zvol_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
{
zvol_state_t *zv = arg;
objset_t *os = zv->zv_objset;
uint64_t offset = lr->lr_offset;
uint64_t size = lr->lr_length;
dmu_buf_t *db;
zgd_t *zgd;
int error;
ASSERT(zio != NULL);
ASSERT(size != 0);
zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_PUSHPAGE);
zgd->zgd_zilog = zv->zv_zilog;
zgd->zgd_rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_READER);
/*
* Write records come in two flavors: immediate and indirect.
* For small writes it's cheaper to store the data with the
* log record (immediate); for large writes it's cheaper to
* sync the data and get a pointer to it (indirect) so that
* we don't have to write the data twice.
*/
if (buf != NULL) { /* immediate write */
error = dmu_read(os, ZVOL_OBJ, offset, size, buf,
DMU_READ_NO_PREFETCH);
} else {
size = zv->zv_volblocksize;
offset = P2ALIGN_TYPED(offset, size, uint64_t);
error = dmu_buf_hold(os, ZVOL_OBJ, offset, zgd, &db,
DMU_READ_NO_PREFETCH);
if (error == 0) {
zgd->zgd_db = db;
zgd->zgd_bp = &lr->lr_blkptr;
ASSERT(db != NULL);
ASSERT(db->db_offset == offset);
ASSERT(db->db_size == size);
error = dmu_sync(zio, lr->lr_common.lrc_txg,
zvol_get_done, zgd);
if (error == 0)
return (0);
}
}
zvol_get_done(zgd, error);
return (error);
}
static int
bplist_cache(bplist_t *bpl, uint64_t blkid)
{
int err = 0;
if (bpl->bpl_cached_dbuf == NULL ||
bpl->bpl_cached_dbuf->db_offset != (blkid << bpl->bpl_blockshift)) {
if (bpl->bpl_cached_dbuf != NULL)
dmu_buf_rele(bpl->bpl_cached_dbuf, bpl);
err = dmu_buf_hold(bpl->bpl_mos,
bpl->bpl_object, blkid << bpl->bpl_blockshift,
bpl, &bpl->bpl_cached_dbuf);
ASSERT(err || bpl->bpl_cached_dbuf->db_size ==
1ULL << bpl->bpl_blockshift);
}
return (err);
}
int
zfs_sa_readlink(znode_t *zp, uio_t *uio)
{
dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
size_t bufsz;
int error;
bufsz = zp->z_size;
if (bufsz + ZFS_OLD_ZNODE_PHYS_SIZE <= db->db_size) {
error = uiomove((caddr_t)db->db_data +
ZFS_OLD_ZNODE_PHYS_SIZE,
MIN((size_t)bufsz, uio_resid(uio)), UIO_READ, uio);
} else {
dmu_buf_t *dbp;
if ((error = dmu_buf_hold(zp->z_zfsvfs->z_os, zp->z_id,
0, FTAG, &dbp, DMU_READ_NO_PREFETCH)) == 0) {
error = uiomove(dbp->db_data,
MIN((size_t)bufsz, uio_resid(uio)), UIO_READ, uio);
dmu_buf_rele(dbp, FTAG);
}
}
return (error);
}
static int
zap_table_grow(zap_t *zap, zap_table_phys_t *tbl,
void (*transfer_func)(const uint64_t *src, uint64_t *dst, int n),
dmu_tx_t *tx)
{
uint64_t b, newblk;
dmu_buf_t *db_old, *db_new;
int err;
int bs = FZAP_BLOCK_SHIFT(zap);
int hepb = 1<<(bs-4);
/* hepb = half the number of entries in a block */
ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
ASSERT(tbl->zt_blk != 0);
ASSERT(tbl->zt_numblks > 0);
if (tbl->zt_nextblk != 0) {
newblk = tbl->zt_nextblk;
} else {
newblk = zap_allocate_blocks(zap, tbl->zt_numblks * 2);
tbl->zt_nextblk = newblk;
ASSERT0(tbl->zt_blks_copied);
dmu_prefetch(zap->zap_objset, zap->zap_object,
tbl->zt_blk << bs, tbl->zt_numblks << bs);
}
/*
* Copy the ptrtbl from the old to new location.
*/
b = tbl->zt_blks_copied;
err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
(tbl->zt_blk + b) << bs, FTAG, &db_old, DMU_READ_NO_PREFETCH);
if (err)
return (err);
/* first half of entries in old[b] go to new[2*b+0] */
VERIFY(0 == dmu_buf_hold(zap->zap_objset, zap->zap_object,
(newblk + 2*b+0) << bs, FTAG, &db_new, DMU_READ_NO_PREFETCH));
dmu_buf_will_dirty(db_new, tx);
transfer_func(db_old->db_data, db_new->db_data, hepb);
dmu_buf_rele(db_new, FTAG);
/* second half of entries in old[b] go to new[2*b+1] */
VERIFY(0 == dmu_buf_hold(zap->zap_objset, zap->zap_object,
(newblk + 2*b+1) << bs, FTAG, &db_new, DMU_READ_NO_PREFETCH));
dmu_buf_will_dirty(db_new, tx);
transfer_func((uint64_t *)db_old->db_data + hepb,
db_new->db_data, hepb);
dmu_buf_rele(db_new, FTAG);
dmu_buf_rele(db_old, FTAG);
tbl->zt_blks_copied++;
dprintf("copied block %llu of %llu\n",
tbl->zt_blks_copied, tbl->zt_numblks);
if (tbl->zt_blks_copied == tbl->zt_numblks) {
(void) dmu_free_range(zap->zap_objset, zap->zap_object,
tbl->zt_blk << bs, tbl->zt_numblks << bs, tx);
tbl->zt_blk = newblk;
tbl->zt_numblks *= 2;
tbl->zt_shift++;
tbl->zt_nextblk = 0;
tbl->zt_blks_copied = 0;
dprintf("finished; numblocks now %llu (%lluk entries)\n",
tbl->zt_numblks, 1<<(tbl->zt_shift-10));
}
return (0);
}
int
zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, offset_t off, ssize_t len,
char *addr)
{
dmu_object_info_t doi;
ssize_t nbytes;
itx_t *itx;
lr_write_t *lr;
objset_t *os;
dmu_buf_t *db;
uint64_t txg;
uint64_t boff;
int error;
uint32_t blocksize;
/* handle common case */
if (len <= zvol_immediate_write_sz) {
itx = zvol_immediate_itx(off, len, addr);
(void) zil_itx_assign(zv->zv_zilog, itx, tx);
return (0);
}
txg = dmu_tx_get_txg(tx);
os = zv->zv_objset;
/*
* We need to dmu_sync() each block in the range.
* For this we need the blocksize.
*/
error = dmu_object_info(os, ZVOL_OBJ, &doi);
if (error)
return (error);
blocksize = doi.doi_data_block_size;
/*
* We need to immediate write or dmu_sync() each block in the range.
*/
while (len) {
nbytes = MIN(len, blocksize - P2PHASE(off, blocksize));
if (nbytes <= zvol_immediate_write_sz) {
itx = zvol_immediate_itx(off, nbytes, addr);
} else {
boff = P2ALIGN_TYPED(off, blocksize, uint64_t);
itx = zil_itx_create(TX_WRITE, sizeof (*lr));
lr = (lr_write_t *)&itx->itx_lr;
lr->lr_foid = ZVOL_OBJ;
lr->lr_offset = off;
lr->lr_length = nbytes;
lr->lr_blkoff = off - boff;
BP_ZERO(&lr->lr_blkptr);
/* XXX - we should do these IOs in parallel */
VERIFY(0 == dmu_buf_hold(os, ZVOL_OBJ, boff,
FTAG, &db));
ASSERT(boff == db->db_offset);
error = dmu_sync(NULL, db, &lr->lr_blkptr,
txg, NULL, NULL);
dmu_buf_rele(db, FTAG);
if (error) {
kmem_free(itx, offsetof(itx_t, itx_lr));
return (error);
}
itx->itx_wr_state = WR_COPIED;
}
(void) zil_itx_assign(zv->zv_zilog, itx, tx);
len -= nbytes;
off += nbytes;
}
return (0);
}
int
zap_lockdir(objset_t *os, uint64_t obj, dmu_tx_t *tx,
krw_t lti, int fatreader, zap_t **zapp)
{
zap_t *zap;
dmu_buf_t *db;
krw_t lt;
int err;
*zapp = NULL;
err = dmu_buf_hold(os, obj, 0, NULL, &db);
if (err)
return (err);
#ifdef ZFS_DEBUG
{
dmu_object_info_t doi;
dmu_object_info_from_db(db, &doi);
ASSERT(dmu_ot[doi.doi_type].ot_byteswap == zap_byteswap);
}
#endif
zap = dmu_buf_get_user(db);
if (zap == NULL)
zap = mzap_open(os, obj, db);
/*
* We're checking zap_ismicro without the lock held, in order to
* tell what type of lock we want. Once we have some sort of
* lock, see if it really is the right type. In practice this
* can only be different if it was upgraded from micro to fat,
* and micro wanted WRITER but fat only needs READER.
*/
lt = (!zap->zap_ismicro && fatreader) ? RW_READER : lti;
rw_enter(&zap->zap_rwlock, lt);
if (lt != ((!zap->zap_ismicro && fatreader) ? RW_READER : lti)) {
/* it was upgraded, now we only need reader */
ASSERT(lt == RW_WRITER);
ASSERT(RW_READER ==
(!zap->zap_ismicro && fatreader) ? RW_READER : lti);
rw_downgrade(&zap->zap_rwlock);
lt = RW_READER;
}
zap->zap_objset = os;
if (lt == RW_WRITER)
dmu_buf_will_dirty(db, tx);
ASSERT3P(zap->zap_dbuf, ==, db);
ASSERT(!zap->zap_ismicro ||
zap->zap_m.zap_num_entries <= zap->zap_m.zap_num_chunks);
if (zap->zap_ismicro && tx &&
zap->zap_m.zap_num_entries == zap->zap_m.zap_num_chunks) {
uint64_t newsz = db->db_size + SPA_MINBLOCKSIZE;
if (newsz > MZAP_MAX_BLKSZ) {
dprintf("upgrading obj %llu: num_entries=%u\n",
obj, zap->zap_m.zap_num_entries);
mzap_upgrade(zap, tx);
*zapp = zap;
return (0);
}
err = dmu_object_set_blocksize(os, obj, newsz, 0, tx);
ASSERT3U(err, ==, 0);
zap->zap_m.zap_num_chunks =
db->db_size / MZAP_ENT_LEN - 1;
}
