static int
spa_history_write(spa_t *spa, void *buf, uint64_t len, spa_history_phys_t *shpp,
dmu_tx_t *tx)
{
uint64_t firstwrite, phys_eof;
objset_t *mos = spa->spa_meta_objset;
int err;
ASSERT(MUTEX_HELD(&spa->spa_history_lock));
/* see if we need to reset logical BOF */
while (shpp->sh_phys_max_off - shpp->sh_pool_create_len -
(shpp->sh_eof - shpp->sh_bof) <= len) {
if ((err = spa_history_advance_bof(spa, shpp)) != 0) {
return (err);
}
}
phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp);
firstwrite = MIN(len, shpp->sh_phys_max_off - phys_eof);
shpp->sh_eof += len;
dmu_write(mos, spa->spa_history, phys_eof, firstwrite, buf, tx);
len -= firstwrite;
if (len > 0) {
/* write out the rest at the beginning of physical file */
dmu_write(mos, spa->spa_history, shpp->sh_pool_create_len,
len, (char *)buf + firstwrite, tx);
}
return (0);
}
/*
* Replay a TX_WRITE ZIL transaction that didn't get committed
* after a system failure
*/
static int
zvol_replay_write(zvol_state_t *zv, lr_write_t *lr, boolean_t byteswap)
{
objset_t *os = zv->zv_objset;
char *data = (char *)(lr + 1); /* data follows lr_write_t */
uint64_t off = lr->lr_offset;
uint64_t len = lr->lr_length;
dmu_tx_t *tx;
int error;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, ZVOL_OBJ, off, len);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
} else {
dmu_write(os, ZVOL_OBJ, off, len, data, tx);
dmu_tx_commit(tx);
}
return (SET_ERROR(error));
}
void
vdev_indirect_births_add_entry(vdev_indirect_births_t *vib,
uint64_t max_offset, uint64_t txg, dmu_tx_t *tx)
{
vdev_indirect_birth_entry_phys_t vibe;
uint64_t old_size;
uint64_t new_size;
vdev_indirect_birth_entry_phys_t *new_entries;
ASSERT(dmu_tx_is_syncing(tx));
ASSERT(dsl_pool_sync_context(dmu_tx_pool(tx)));
ASSERT(vdev_indirect_births_verify(vib));
dmu_buf_will_dirty(vib->vib_dbuf, tx);
vibe.vibe_offset = max_offset;
vibe.vibe_phys_birth_txg = txg;
old_size = vdev_indirect_births_size_impl(vib);
dmu_write(vib->vib_objset, vib->vib_object, old_size, sizeof (vibe),
&vibe, tx);
vib->vib_phys->vib_count++;
new_size = vdev_indirect_births_size_impl(vib);
new_entries = kmem_alloc(new_size, KM_SLEEP);
if (old_size > 0) {
bcopy(vib->vib_entries, new_entries, old_size);
kmem_free(vib->vib_entries, old_size);
}
new_entries[vib->vib_phys->vib_count - 1] = vibe;
vib->vib_entries = new_entries;
}
static ssize_t osd_write(const struct lu_env *env, struct dt_object *dt,
const struct lu_buf *buf, loff_t *pos,
struct thandle *th, int ignore_quota)
{
struct osd_object *obj = osd_dt_obj(dt);
struct osd_device *osd = osd_obj2dev(obj);
struct osd_thandle *oh;
uint64_t offset = *pos;
int rc;
ENTRY;
LASSERT(dt_object_exists(dt));
LASSERT(obj->oo_db);
LASSERT(th != NULL);
oh = container_of0(th, struct osd_thandle, ot_super);
record_start_io(osd, WRITE, 0);
dmu_write(osd->od_os, obj->oo_db->db_object, offset,
(uint64_t)buf->lb_len, buf->lb_buf, oh->ot_tx);
write_lock(&obj->oo_attr_lock);
if (obj->oo_attr.la_size < offset + buf->lb_len) {
obj->oo_attr.la_size = offset + buf->lb_len;
write_unlock(&obj->oo_attr_lock);
/* osd_object_sa_update() will be copying directly from oo_attr
* into dbuf. any update within a single txg will copy the
* most actual */
rc = osd_object_sa_update(obj, SA_ZPL_SIZE(osd),
&obj->oo_attr.la_size, 8, oh);
if (unlikely(rc))
GOTO(out, rc);
} else {
write_unlock(&obj->oo_attr_lock);
}
*pos += buf->lb_len;
rc = buf->lb_len;
out:
record_end_io(osd, WRITE, 0, buf->lb_len,
buf->lb_len >> PAGE_CACHE_SHIFT);
RETURN(rc);
}
static int
zpios_dmu_write(run_args_t *run_args, objset_t *os, uint64_t object,
uint64_t offset, uint64_t size, const void *buf)
{
struct dmu_tx *tx;
int rc, how = TXG_WAIT;
// int flags = 0;
if (run_args->flags & DMU_WRITE_NOWAIT)
how = TXG_NOWAIT;
while (1) {
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, object, offset, size);
rc = dmu_tx_assign(tx, how);
if (rc) {
if (rc == ERESTART && how == TXG_NOWAIT) {
dmu_tx_wait(tx);
dmu_tx_abort(tx);
continue;
}
zpios_print(run_args->file,
"Error in dmu_tx_assign(), %d", rc);
dmu_tx_abort(tx);
return (rc);
}
break;
}
// if (run_args->flags & DMU_WRITE_ZC)
// flags |= DMU_WRITE_ZEROCOPY;
dmu_write(os, object, offset, size, buf, tx);
dmu_tx_commit(tx);
return (0);
}
static
#endif
int
zfs_fuid_find_by_domain(zfsvfs_t *zfsvfs, const char *domain, char **retdomain,
dmu_tx_t *tx)
{
fuid_domain_t searchnode, *findnode;
avl_index_t loc;
/*
* If the dummy "nobody" domain then return an index of 0
* to cause the created FUID to be a standard POSIX id
* for the user nobody.
*/
if (domain[0] == '\0') {
*retdomain = "";
return (0);
}
searchnode.f_ksid = ksid_lookupdomain(domain);
if (retdomain) {
*retdomain = searchnode.f_ksid->kd_name;
}
if (!zfsvfs->z_fuid_loaded)
zfs_fuid_init(zfsvfs, tx);
rw_enter(&zfsvfs->z_fuid_lock, RW_READER);
findnode = avl_find(&zfsvfs->z_fuid_domain, &searchnode, &loc);
rw_exit(&zfsvfs->z_fuid_lock);
if (findnode) {
ksiddomain_rele(searchnode.f_ksid);
return (findnode->f_idx);
} else {
fuid_domain_t *domnode;
nvlist_t *nvp;
nvlist_t **fuids;
uint64_t retidx;
size_t nvsize = 0;
char *packed;
dmu_buf_t *db;
int i = 0;
domnode = kmem_alloc(sizeof (fuid_domain_t), KM_SLEEP);
domnode->f_ksid = searchnode.f_ksid;
rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
retidx = domnode->f_idx = avl_numnodes(&zfsvfs->z_fuid_idx) + 1;
avl_add(&zfsvfs->z_fuid_domain, domnode);
avl_add(&zfsvfs->z_fuid_idx, domnode);
/*
* Now resync the on-disk nvlist.
*/
VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
domnode = avl_first(&zfsvfs->z_fuid_domain);
fuids = kmem_alloc(retidx * sizeof (void *), KM_SLEEP);
while (domnode) {
VERIFY(nvlist_alloc(&fuids[i],
NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX,
domnode->f_idx) == 0);
VERIFY(nvlist_add_uint64(fuids[i],
FUID_OFFSET, 0) == 0);
VERIFY(nvlist_add_string(fuids[i++], FUID_DOMAIN,
domnode->f_ksid->kd_name) == 0);
domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode);
}
VERIFY(nvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY,
fuids, retidx) == 0);
for (i = 0; i != retidx; i++)
nvlist_free(fuids[i]);
kmem_free(fuids, retidx * sizeof (void *));
VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0);
packed = kmem_alloc(nvsize, KM_SLEEP);
VERIFY(nvlist_pack(nvp, &packed, &nvsize,
NV_ENCODE_XDR, KM_SLEEP) == 0);
nvlist_free(nvp);
zfsvfs->z_fuid_size = nvsize;
dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0,
zfsvfs->z_fuid_size, packed, tx);
kmem_free(packed, zfsvfs->z_fuid_size);
VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj,
FTAG, &db));
dmu_buf_will_dirty(db, tx);
*(uint64_t *)db->db_data = zfsvfs->z_fuid_size;
dmu_buf_rele(db, FTAG);
rw_exit(&zfsvfs->z_fuid_lock);
return (retidx);
}
}
int
__osd_xattr_set(const struct lu_env *env, struct osd_object *obj,
const struct lu_buf *buf, const char *name, int fl,
struct osd_thandle *oh)
{
struct osd_device *osd = osd_obj2dev(obj);
dmu_buf_t *xa_zap_db = NULL;
dmu_buf_t *xa_data_db = NULL;
uint64_t xa_data_obj;
sa_handle_t *sa_hdl = NULL;
dmu_tx_t *tx = oh->ot_tx;
uint64_t size;
int rc;
LASSERT(obj->oo_sa_hdl);
if (obj->oo_xattr == ZFS_NO_OBJECT) {
struct lu_attr *la = &osd_oti_get(env)->oti_la;
la->la_valid = LA_MODE;
la->la_mode = S_IFDIR | S_IRUGO | S_IWUSR | S_IXUGO;
rc = __osd_zap_create(env, osd, &xa_zap_db, tx, la,
obj->oo_db->db_object, 0);
if (rc)
return rc;
obj->oo_xattr = xa_zap_db->db_object;
rc = osd_object_sa_update(obj, SA_ZPL_XATTR(osd),
&obj->oo_xattr, 8, oh);
if (rc)
goto out;
}
rc = -zap_lookup(osd->od_os, obj->oo_xattr, name, sizeof(uint64_t), 1,
&xa_data_obj);
if (rc == 0) {
if (fl & LU_XATTR_CREATE) {
rc = -EEXIST;
goto out;
}
/*
* Entry already exists.
* We'll truncate the existing object.
*/
rc = __osd_obj2dbuf(env, osd->od_os, xa_data_obj,
&xa_data_db);
if (rc)
goto out;
rc = -sa_handle_get(osd->od_os, xa_data_obj, NULL,
SA_HDL_PRIVATE, &sa_hdl);
if (rc)
goto out;
rc = -sa_lookup(sa_hdl, SA_ZPL_SIZE(osd), &size, 8);
if (rc)
goto out_sa;
rc = -dmu_free_range(osd->od_os, xa_data_db->db_object,
0, DMU_OBJECT_END, tx);
if (rc)
goto out_sa;
} else if (rc == -ENOENT) {
struct lu_attr *la = &osd_oti_get(env)->oti_la;
/*
* Entry doesn't exist, we need to create a new one and a new
* object to store the value.
*/
if (fl & LU_XATTR_REPLACE) {
/* should be ENOATTR according to the
* man, but that is undefined here */
rc = -ENODATA;
goto out;
}
la->la_valid = LA_MODE;
la->la_mode = S_IFREG | S_IRUGO | S_IWUSR;
rc = __osd_object_create(env, obj, &xa_data_db, tx, la,
obj->oo_xattr);
if (rc)
goto out;
xa_data_obj = xa_data_db->db_object;
rc = -sa_handle_get(osd->od_os, xa_data_obj, NULL,
SA_HDL_PRIVATE, &sa_hdl);
if (rc)
goto out;
rc = -zap_add(osd->od_os, obj->oo_xattr, name, sizeof(uint64_t),
1, &xa_data_obj, tx);
if (rc)
goto out_sa;
} else {
/* There was an error looking up the xattr name */
goto out;
}
/* Finally write the xattr value */
dmu_write(osd->od_os, xa_data_obj, 0, buf->lb_len, buf->lb_buf, tx);
size = buf->lb_len;
rc = -sa_update(sa_hdl, SA_ZPL_SIZE(osd), &size, 8, tx);
out_sa:
sa_handle_destroy(sa_hdl);
out:
if (xa_data_db != NULL)
dmu_buf_rele(xa_data_db, FTAG);
if (xa_zap_db != NULL)
dmu_buf_rele(xa_zap_db, FTAG);
return rc;
}
/*
* sync out AVL trees to persistent storage.
*/
void
zfs_fuid_sync(zfsvfs_t *zfsvfs, dmu_tx_t *tx)
{
#ifdef HAVE_ZPL
nvlist_t *nvp;
nvlist_t **fuids;
size_t nvsize = 0;
char *packed;
dmu_buf_t *db;
fuid_domain_t *domnode;
int numnodes;
int i;
if (!zfsvfs->z_fuid_dirty) {
return;
}
rw_enter(&zfsvfs->z_fuid_lock, RW_WRITER);
/*
* First see if table needs to be created?
*/
if (zfsvfs->z_fuid_obj == 0) {
zfsvfs->z_fuid_obj = dmu_object_alloc(zfsvfs->z_os,
DMU_OT_FUID, 1 << 14, DMU_OT_FUID_SIZE,
sizeof (uint64_t), tx);
VERIFY(zap_add(zfsvfs->z_os, MASTER_NODE_OBJ,
ZFS_FUID_TABLES, sizeof (uint64_t), 1,
&zfsvfs->z_fuid_obj, tx) == 0);
}
VERIFY(nvlist_alloc(&nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0);
numnodes = avl_numnodes(&zfsvfs->z_fuid_idx);
fuids = kmem_alloc(numnodes * sizeof (void *), KM_SLEEP);
for (i = 0, domnode = avl_first(&zfsvfs->z_fuid_domain); domnode; i++,
domnode = AVL_NEXT(&zfsvfs->z_fuid_domain, domnode)) {
VERIFY(nvlist_alloc(&fuids[i], NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_uint64(fuids[i], FUID_IDX,
domnode->f_idx) == 0);
VERIFY(nvlist_add_uint64(fuids[i], FUID_OFFSET, 0) == 0);
VERIFY(nvlist_add_string(fuids[i], FUID_DOMAIN,
domnode->f_ksid->kd_name) == 0);
}
VERIFY(nvlist_add_nvlist_array(nvp, FUID_NVP_ARRAY,
fuids, numnodes) == 0);
for (i = 0; i != numnodes; i++)
nvlist_free(fuids[i]);
kmem_free(fuids, numnodes * sizeof (void *));
VERIFY(nvlist_size(nvp, &nvsize, NV_ENCODE_XDR) == 0);
packed = kmem_alloc(nvsize, KM_SLEEP);
VERIFY(nvlist_pack(nvp, &packed, &nvsize,
NV_ENCODE_XDR, KM_SLEEP) == 0);
nvlist_free(nvp);
zfsvfs->z_fuid_size = nvsize;
dmu_write(zfsvfs->z_os, zfsvfs->z_fuid_obj, 0,
zfsvfs->z_fuid_size, packed, tx);
kmem_free(packed, zfsvfs->z_fuid_size);
VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, zfsvfs->z_fuid_obj,
FTAG, &db));
dmu_buf_will_dirty(db, tx);
*(uint64_t *)db->db_data = zfsvfs->z_fuid_size;
dmu_buf_rele(db, FTAG);
zfsvfs->z_fuid_dirty = B_FALSE;
rw_exit(&zfsvfs->z_fuid_lock);
#endif /* HAVE_ZPL */
}
int
zvol_strategy(buf_t *bp)
{
zvol_state_t *zv = ddi_get_soft_state(zvol_state, getminor(bp->b_edev));
uint64_t off, volsize;
size_t size, resid;
char *addr;
objset_t *os;
int error = 0;
int sync;
int reading;
int txg_sync_needed = B_FALSE;
if (zv == NULL) {
bioerror(bp, ENXIO);
biodone(bp);
return (0);
}
if (getminor(bp->b_edev) == 0) {
bioerror(bp, EINVAL);
biodone(bp);
return (0);
}
if (zv->zv_readonly && !(bp->b_flags & B_READ)) {
bioerror(bp, EROFS);
biodone(bp);
return (0);
}
off = ldbtob(bp->b_blkno);
volsize = zv->zv_volsize;
os = zv->zv_objset;
ASSERT(os != NULL);
sync = !(bp->b_flags & B_ASYNC) && !(zil_disable);
bp_mapin(bp);
addr = bp->b_un.b_addr;
resid = bp->b_bcount;
/*
* There must be no buffer changes when doing a dmu_sync() because
* we can't change the data whilst calculating the checksum.
* A better approach than a per zvol rwlock would be to lock ranges.
*/
reading = bp->b_flags & B_READ;
if (reading || resid <= zvol_immediate_write_sz)
rw_enter(&zv->zv_dslock, RW_READER);
else
rw_enter(&zv->zv_dslock, RW_WRITER);
while (resid != 0 && off < volsize) {
size = MIN(resid, 1UL << 20); /* cap at 1MB per tx */
if (size > volsize - off) /* don't write past the end */
size = volsize - off;
if (reading) {
error = dmu_read(os, ZVOL_OBJ, off, size, addr);
} else {
dmu_tx_t *tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, ZVOL_OBJ, off, size);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
} else {
dmu_write(os, ZVOL_OBJ, off, size, addr, tx);
if (sync) {
/* use the ZIL to commit this write */
if (zvol_log_write(zv, tx, off, size,
addr) != 0) {
txg_sync_needed = B_TRUE;
}
}
dmu_tx_commit(tx);
}
}
if (error)
break;
off += size;
addr += size;
resid -= size;
}
rw_exit(&zv->zv_dslock);
if ((bp->b_resid = resid) == bp->b_bcount)
bioerror(bp, off > volsize ? EINVAL : error);
biodone(bp);
if (sync) {
if (txg_sync_needed)
txg_wait_synced(dmu_objset_pool(os), 0);
else
zil_commit(zv->zv_zilog, UINT64_MAX, 0);
}
return (0);
}
