*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
+ * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
+ * Copyright (c) 2013, Joyent, Inc. All rights reserved.
+ * Copyright (c) 2016, Nexenta Systems, Inc. All rights reserved.
+ * Copyright (c) 2015 by Chunwei Chen. All rights reserved.
*/
#include <sys/dmu.h>
#include <sys/zfs_ioctl.h>
#include <sys/zap.h>
#include <sys/zio_checksum.h>
+#include <sys/zio_compress.h>
#include <sys/sa.h>
+#include <sys/zfeature.h>
+#include <sys/abd.h>
+#include <sys/trace_dmu.h>
+#include <sys/zfs_rlock.h>
#ifdef _KERNEL
#include <sys/vmsystm.h>
#include <sys/zfs_znode.h>
#endif
+/*
+ * Enable/disable nopwrite feature.
+ */
+int zfs_nopwrite_enabled = 1;
+
+/*
+ * Tunable to control percentage of dirtied blocks from frees in one TXG.
+ * After this threshold is crossed, additional dirty blocks from frees
+ * wait until the next TXG.
+ * A value of zero will disable this throttle.
+ */
+unsigned long zfs_per_txg_dirty_frees_percent = 30;
+
+/*
+ * Enable/disable forcing txg sync when dirty in dmu_offset_next.
+ */
+int zfs_dmu_offset_next_sync = 0;
+
const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = {
- { byteswap_uint8_array, TRUE, "unallocated" },
- { zap_byteswap, TRUE, "object directory" },
- { byteswap_uint64_array, TRUE, "object array" },
- { byteswap_uint8_array, TRUE, "packed nvlist" },
- { byteswap_uint64_array, TRUE, "packed nvlist size" },
- { byteswap_uint64_array, TRUE, "bpobj" },
- { byteswap_uint64_array, TRUE, "bpobj header" },
- { byteswap_uint64_array, TRUE, "SPA space map header" },
- { byteswap_uint64_array, TRUE, "SPA space map" },
- { byteswap_uint64_array, TRUE, "ZIL intent log" },
- { dnode_buf_byteswap, TRUE, "DMU dnode" },
- { dmu_objset_byteswap, TRUE, "DMU objset" },
- { byteswap_uint64_array, TRUE, "DSL directory" },
- { zap_byteswap, TRUE, "DSL directory child map"},
- { zap_byteswap, TRUE, "DSL dataset snap map" },
- { zap_byteswap, TRUE, "DSL props" },
- { byteswap_uint64_array, TRUE, "DSL dataset" },
- { zfs_znode_byteswap, TRUE, "ZFS znode" },
- { zfs_oldacl_byteswap, TRUE, "ZFS V0 ACL" },
- { byteswap_uint8_array, FALSE, "ZFS plain file" },
- { zap_byteswap, TRUE, "ZFS directory" },
- { zap_byteswap, TRUE, "ZFS master node" },
- { zap_byteswap, TRUE, "ZFS delete queue" },
- { byteswap_uint8_array, FALSE, "zvol object" },
- { zap_byteswap, TRUE, "zvol prop" },
- { byteswap_uint8_array, FALSE, "other uint8[]" },
- { byteswap_uint64_array, FALSE, "other uint64[]" },
- { zap_byteswap, TRUE, "other ZAP" },
- { zap_byteswap, TRUE, "persistent error log" },
- { byteswap_uint8_array, TRUE, "SPA history" },
- { byteswap_uint64_array, TRUE, "SPA history offsets" },
- { zap_byteswap, TRUE, "Pool properties" },
- { zap_byteswap, TRUE, "DSL permissions" },
- { zfs_acl_byteswap, TRUE, "ZFS ACL" },
- { byteswap_uint8_array, TRUE, "ZFS SYSACL" },
- { byteswap_uint8_array, TRUE, "FUID table" },
- { byteswap_uint64_array, TRUE, "FUID table size" },
- { zap_byteswap, TRUE, "DSL dataset next clones"},
- { zap_byteswap, TRUE, "scan work queue" },
- { zap_byteswap, TRUE, "ZFS user/group used" },
- { zap_byteswap, TRUE, "ZFS user/group quota" },
- { zap_byteswap, TRUE, "snapshot refcount tags"},
- { zap_byteswap, TRUE, "DDT ZAP algorithm" },
- { zap_byteswap, TRUE, "DDT statistics" },
- { byteswap_uint8_array, TRUE, "System attributes" },
- { zap_byteswap, TRUE, "SA master node" },
- { zap_byteswap, TRUE, "SA attr registration" },
- { zap_byteswap, TRUE, "SA attr layouts" },
- { zap_byteswap, TRUE, "scan translations" },
- { byteswap_uint8_array, FALSE, "deduplicated block" },
- { zap_byteswap, TRUE, "DSL deadlist map" },
- { byteswap_uint64_array, TRUE, "DSL deadlist map hdr" },
- { zap_byteswap, TRUE, "DSL dir clones" },
- { byteswap_uint64_array, TRUE, "bpobj subobj" },
+ { DMU_BSWAP_UINT8, TRUE, FALSE, "unallocated" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "object directory" },
+ { DMU_BSWAP_UINT64, TRUE, FALSE, "object array" },
+ { DMU_BSWAP_UINT8, TRUE, FALSE, "packed nvlist" },
+ { DMU_BSWAP_UINT64, TRUE, FALSE, "packed nvlist size" },
+ { DMU_BSWAP_UINT64, TRUE, FALSE, "bpobj" },
+ { DMU_BSWAP_UINT64, TRUE, FALSE, "bpobj header" },
+ { DMU_BSWAP_UINT64, TRUE, FALSE, "SPA space map header" },
+ { DMU_BSWAP_UINT64, TRUE, FALSE, "SPA space map" },
+ { DMU_BSWAP_UINT64, TRUE, TRUE, "ZIL intent log" },
+ { DMU_BSWAP_DNODE, TRUE, TRUE, "DMU dnode" },
+ { DMU_BSWAP_OBJSET, TRUE, FALSE, "DMU objset" },
+ { DMU_BSWAP_UINT64, TRUE, FALSE, "DSL directory" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL directory child map"},
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL dataset snap map" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL props" },
+ { DMU_BSWAP_UINT64, TRUE, FALSE, "DSL dataset" },
+ { DMU_BSWAP_ZNODE, TRUE, FALSE, "ZFS znode" },
+ { DMU_BSWAP_OLDACL, TRUE, TRUE, "ZFS V0 ACL" },
+ { DMU_BSWAP_UINT8, FALSE, TRUE, "ZFS plain file" },
+ { DMU_BSWAP_ZAP, TRUE, TRUE, "ZFS directory" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "ZFS master node" },
+ { DMU_BSWAP_ZAP, TRUE, TRUE, "ZFS delete queue" },
+ { DMU_BSWAP_UINT8, FALSE, TRUE, "zvol object" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "zvol prop" },
+ { DMU_BSWAP_UINT8, FALSE, TRUE, "other uint8[]" },
+ { DMU_BSWAP_UINT64, FALSE, TRUE, "other uint64[]" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "other ZAP" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "persistent error log" },
+ { DMU_BSWAP_UINT8, TRUE, FALSE, "SPA history" },
+ { DMU_BSWAP_UINT64, TRUE, FALSE, "SPA history offsets" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "Pool properties" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL permissions" },
+ { DMU_BSWAP_ACL, TRUE, TRUE, "ZFS ACL" },
+ { DMU_BSWAP_UINT8, TRUE, TRUE, "ZFS SYSACL" },
+ { DMU_BSWAP_UINT8, TRUE, TRUE, "FUID table" },
+ { DMU_BSWAP_UINT64, TRUE, FALSE, "FUID table size" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL dataset next clones"},
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "scan work queue" },
+ { DMU_BSWAP_ZAP, TRUE, TRUE, "ZFS user/group used" },
+ { DMU_BSWAP_ZAP, TRUE, TRUE, "ZFS user/group quota" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "snapshot refcount tags"},
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "DDT ZAP algorithm" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "DDT statistics" },
+ { DMU_BSWAP_UINT8, TRUE, TRUE, "System attributes" },
+ { DMU_BSWAP_ZAP, TRUE, TRUE, "SA master node" },
+ { DMU_BSWAP_ZAP, TRUE, TRUE, "SA attr registration" },
+ { DMU_BSWAP_ZAP, TRUE, TRUE, "SA attr layouts" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "scan translations" },
+ { DMU_BSWAP_UINT8, FALSE, TRUE, "deduplicated block" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL deadlist map" },
+ { DMU_BSWAP_UINT64, TRUE, FALSE, "DSL deadlist map hdr" },
+ { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL dir clones" },
+ { DMU_BSWAP_UINT64, TRUE, FALSE, "bpobj subobj" }
+};
+
+const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS] = {
+ { byteswap_uint8_array, "uint8" },
+ { byteswap_uint16_array, "uint16" },
+ { byteswap_uint32_array, "uint32" },
+ { byteswap_uint64_array, "uint64" },
+ { zap_byteswap, "zap" },
+ { dnode_buf_byteswap, "dnode" },
+ { dmu_objset_byteswap, "objset" },
+ { zfs_znode_byteswap, "znode" },
+ { zfs_oldacl_byteswap, "oldacl" },
+ { zfs_acl_byteswap, "acl" }
};
int
-dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
- void *tag, dmu_buf_t **dbp, int flags)
+dmu_buf_hold_noread_by_dnode(dnode_t *dn, uint64_t offset,
+ void *tag, dmu_buf_t **dbp)
+{
+ uint64_t blkid;
+ dmu_buf_impl_t *db;
+
+ blkid = dbuf_whichblock(dn, 0, offset);
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ db = dbuf_hold(dn, blkid, tag);
+ rw_exit(&dn->dn_struct_rwlock);
+
+ if (db == NULL) {
+ *dbp = NULL;
+ return (SET_ERROR(EIO));
+ }
+
+ *dbp = &db->db;
+ return (0);
+}
+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;
- 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);
+ blkid = dbuf_whichblock(dn, 0, 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) {
- err = EIO;
- } else {
+ *dbp = NULL;
+ return (SET_ERROR(EIO));
+ }
+
+ *dbp = &db->db;
+ return (err);
+}
+
+int
+dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset,
+ void *tag, dmu_buf_t **dbp, int flags)
+{
+ int err;
+ int db_flags = DB_RF_CANFAIL;
+
+ if (flags & DMU_READ_NO_PREFETCH)
+ db_flags |= DB_RF_NOPREFETCH;
+ if (flags & DMU_READ_NO_DECRYPT)
+ db_flags |= DB_RF_NO_DECRYPT;
+
+ err = dmu_buf_hold_noread_by_dnode(dn, offset, tag, dbp);
+ if (err == 0) {
+ dmu_buf_impl_t *db = (dmu_buf_impl_t *)(*dbp);
err = dbuf_read(db, NULL, db_flags);
- if (err) {
+ if (err != 0) {
dbuf_rele(db, tag);
- db = NULL;
+ *dbp = NULL;
+ }
+ }
+
+ return (err);
+}
+
+int
+dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset,
+ void *tag, dmu_buf_t **dbp, int flags)
+{
+ int err;
+ int db_flags = DB_RF_CANFAIL;
+
+ if (flags & DMU_READ_NO_PREFETCH)
+ db_flags |= DB_RF_NOPREFETCH;
+ if (flags & DMU_READ_NO_DECRYPT)
+ db_flags |= DB_RF_NO_DECRYPT;
+
+ err = dmu_buf_hold_noread(os, object, offset, tag, dbp);
+ if (err == 0) {
+ dmu_buf_impl_t *db = (dmu_buf_impl_t *)(*dbp);
+ err = dbuf_read(db, NULL, db_flags);
+ if (err != 0) {
+ dbuf_rele(db, tag);
+ *dbp = NULL;
}
}
- dnode_rele(dn, FTAG);
- *dbp = &db->db; /* NULL db plus first field offset is NULL */
return (err);
}
int
dmu_bonus_max(void)
{
- return (DN_MAX_BONUSLEN);
+ return (DN_OLD_MAX_BONUSLEN);
}
int
dn = DB_DNODE(db);
if (dn->dn_bonus != db) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
} else if (newsize < 0 || newsize > db_fake->db_size) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
} else {
dnode_setbonuslen(dn, newsize, tx);
error = 0;
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
- if (type > DMU_OT_NUMTYPES) {
- error = EINVAL;
+ if (!DMU_OT_IS_VALID(type)) {
+ error = SET_ERROR(EINVAL);
} else if (dn->dn_bonus != db) {
- error = EINVAL;
+ error = SET_ERROR(EINVAL);
} else {
dnode_setbonus_type(dn, type, tx);
error = 0;
* returns ENOENT, EIO, or 0.
*/
int
-dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **dbp)
+dmu_bonus_hold_impl(objset_t *os, uint64_t object, void *tag, uint32_t flags,
+ dmu_buf_t **dbp)
{
dnode_t *dn;
dmu_buf_impl_t *db;
int error;
+ uint32_t db_flags = DB_RF_MUST_SUCCEED;
+
+ if (flags & DMU_READ_NO_PREFETCH)
+ db_flags |= DB_RF_NOPREFETCH;
+ if (flags & DMU_READ_NO_DECRYPT)
+ db_flags |= DB_RF_NO_DECRYPT;
error = dnode_hold(os, object, FTAG, &dn);
if (error)
/* 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));
- (void) atomic_inc_32_nv(&dn->dn_dbufs_count);
+ atomic_inc_32(&dn->dn_dbufs_count);
}
/*
dnode_rele(dn, FTAG);
- VERIFY(0 == dbuf_read(db, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH));
+ error = dbuf_read(db, NULL, db_flags);
+ if (error) {
+ dnode_evict_bonus(dn);
+ dbuf_rele(db, tag);
+ *dbp = NULL;
+ return (error);
+ }
*dbp = &db->db;
return (0);
}
+int
+dmu_bonus_hold(objset_t *os, uint64_t obj, void *tag, dmu_buf_t **dbp)
+{
+ return (dmu_bonus_hold_impl(os, obj, tag, DMU_READ_NO_PREFETCH, dbp));
+}
+
/*
* returns ENOENT, EIO, or 0.
*
if ((flags & DB_RF_HAVESTRUCT) == 0)
rw_exit(&dn->dn_struct_rwlock);
- ASSERT(db != NULL);
+ if (db == NULL) {
+ *dbp = NULL;
+ return (SET_ERROR(EIO));
+ }
err = dbuf_read(db, NULL, flags);
if (err == 0)
*dbp = &db->db;
- else
+ else {
dbuf_rele(db, tag);
+ *dbp = NULL;
+ }
return (err);
}
dn = DB_DNODE(db);
if (spa_version(dn->dn_objset->os_spa) < SPA_VERSION_SA) {
- err = EINVAL;
+ err = SET_ERROR(EINVAL);
} else {
rw_enter(&dn->dn_struct_rwlock, RW_READER);
if (!dn->dn_have_spill) {
- err = ENOENT;
+ err = SET_ERROR(ENOENT);
} else {
err = dmu_spill_hold_by_dnode(dn,
DB_RF_HAVESTRUCT | DB_RF_CANFAIL, tag, dbp);
*/
static int
dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, uint64_t length,
- int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp, uint32_t flags)
+ boolean_t read, void *tag, int *numbufsp, dmu_buf_t ***dbpp, uint32_t flags)
{
- dsl_pool_t *dp = NULL;
dmu_buf_t **dbp;
uint64_t blkid, nblks, i;
uint32_t dbuf_flags;
int err;
zio_t *zio;
- hrtime_t start = 0;
ASSERT(length <= DMU_MAX_ACCESS);
- dbuf_flags = DB_RF_CANFAIL | DB_RF_NEVERWAIT | DB_RF_HAVESTRUCT;
- if (flags & DMU_READ_NO_PREFETCH || length > zfetch_array_rd_sz)
- dbuf_flags |= DB_RF_NOPREFETCH;
+ /*
+ * Note: We directly notify the prefetch code of this read, so that
+ * we can tell it about the multi-block read. dbuf_read() only knows
+ * about the one block it is accessing.
+ */
+ dbuf_flags = DB_RF_CANFAIL | DB_RF_NEVERWAIT | DB_RF_HAVESTRUCT |
+ DB_RF_NOPREFETCH;
rw_enter(&dn->dn_struct_rwlock, RW_READER);
if (dn->dn_datablkshift) {
int blkshift = dn->dn_datablkshift;
- nblks = (P2ROUNDUP(offset+length, 1ULL<<blkshift) -
- P2ALIGN(offset, 1ULL<<blkshift)) >> blkshift;
+ nblks = (P2ROUNDUP(offset + length, 1ULL << blkshift) -
+ P2ALIGN(offset, 1ULL << blkshift)) >> blkshift;
} else {
if (offset + length > dn->dn_datablksz) {
zfs_panic_recover("zfs: accessing past end of object "
(longlong_t)dn->dn_object, dn->dn_datablksz,
(longlong_t)offset, (longlong_t)length);
rw_exit(&dn->dn_struct_rwlock);
- return (EIO);
+ return (SET_ERROR(EIO));
}
nblks = 1;
}
- dbp = kmem_zalloc(sizeof (dmu_buf_t *) * nblks, KM_PUSHPAGE | KM_NODEBUG);
+ dbp = kmem_zalloc(sizeof (dmu_buf_t *) * nblks, KM_SLEEP);
- if (dn->dn_objset->os_dsl_dataset)
- dp = dn->dn_objset->os_dsl_dataset->ds_dir->dd_pool;
- if (dp && dsl_pool_sync_context(dp))
- start = gethrtime();
zio = zio_root(dn->dn_objset->os_spa, NULL, NULL, ZIO_FLAG_CANFAIL);
- blkid = dbuf_whichblock(dn, offset);
+ blkid = dbuf_whichblock(dn, 0, offset);
for (i = 0; i < nblks; i++) {
- dmu_buf_impl_t *db = dbuf_hold(dn, blkid+i, tag);
+ dmu_buf_impl_t *db = dbuf_hold(dn, blkid + i, tag);
if (db == NULL) {
rw_exit(&dn->dn_struct_rwlock);
dmu_buf_rele_array(dbp, nblks, tag);
zio_nowait(zio);
- return (EIO);
+ return (SET_ERROR(EIO));
}
+
/* initiate async i/o */
- if (read) {
+ if (read)
(void) dbuf_read(db, zio, dbuf_flags);
- }
dbp[i] = &db->db;
}
+
+ if ((flags & DMU_READ_NO_PREFETCH) == 0 &&
+ DNODE_META_IS_CACHEABLE(dn) && length <= zfetch_array_rd_sz) {
+ dmu_zfetch(&dn->dn_zfetch, blkid, nblks,
+ read && DNODE_IS_CACHEABLE(dn));
+ }
rw_exit(&dn->dn_struct_rwlock);
/* wait for async i/o */
err = zio_wait(zio);
- /* track read overhead when we are in sync context */
- if (dp && dsl_pool_sync_context(dp))
- dp->dp_read_overhead += gethrtime() - start;
if (err) {
dmu_buf_rele_array(dbp, nblks, tag);
return (err);
db->db_state == DB_FILL)
cv_wait(&db->db_changed, &db->db_mtx);
if (db->db_state == DB_UNCACHED)
- err = EIO;
+ err = SET_ERROR(EIO);
mutex_exit(&db->db_mtx);
if (err) {
dmu_buf_rele_array(dbp, nblks, tag);
int
dmu_buf_hold_array_by_bonus(dmu_buf_t *db_fake, uint64_t offset,
- uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp)
+ uint64_t length, boolean_t read, void *tag, int *numbufsp,
+ dmu_buf_t ***dbpp)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
dnode_t *dn;
kmem_free(dbp, sizeof (dmu_buf_t *) * numbufs);
}
+/*
+ * Issue prefetch i/os for the given blocks. If level is greater than 0, the
+ * indirect blocks prefeteched will be those that point to the blocks containing
+ * the data starting at offset, and continuing to offset + len.
+ *
+ * Note that if the indirect blocks above the blocks being prefetched are not
+ * in cache, they will be asychronously read in.
+ */
void
-dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset, uint64_t len)
+dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset,
+ uint64_t len, zio_priority_t pri)
{
dnode_t *dn;
uint64_t blkid;
- int nblks, i, err;
-
- if (zfs_prefetch_disable)
- return;
+ int nblks, err;
if (len == 0) { /* they're interested in the bonus buffer */
dn = DMU_META_DNODE(os);
return;
rw_enter(&dn->dn_struct_rwlock, RW_READER);
- blkid = dbuf_whichblock(dn, object * sizeof (dnode_phys_t));
- dbuf_prefetch(dn, blkid);
+ blkid = dbuf_whichblock(dn, level,
+ object * sizeof (dnode_phys_t));
+ dbuf_prefetch(dn, level, blkid, pri, 0);
rw_exit(&dn->dn_struct_rwlock);
return;
}
return;
rw_enter(&dn->dn_struct_rwlock, RW_READER);
- if (dn->dn_datablkshift) {
- int blkshift = dn->dn_datablkshift;
- nblks = (P2ROUNDUP(offset+len, 1<<blkshift) -
- P2ALIGN(offset, 1<<blkshift)) >> blkshift;
+ /*
+ * offset + len - 1 is the last byte we want to prefetch for, and offset
+ * is the first. Then dbuf_whichblk(dn, level, off + len - 1) is the
+ * last block we want to prefetch, and dbuf_whichblock(dn, level,
+ * offset) is the first. Then the number we need to prefetch is the
+ * last - first + 1.
+ */
+ if (level > 0 || dn->dn_datablkshift != 0) {
+ nblks = dbuf_whichblock(dn, level, offset + len - 1) -
+ dbuf_whichblock(dn, level, offset) + 1;
} else {
nblks = (offset < dn->dn_datablksz);
}
if (nblks != 0) {
- blkid = dbuf_whichblock(dn, offset);
- for (i = 0; i < nblks; i++)
- dbuf_prefetch(dn, blkid+i);
+ blkid = dbuf_whichblock(dn, level, offset);
+ for (int i = 0; i < nblks; i++)
+ dbuf_prefetch(dn, level, blkid + i, pri, 0);
}
rw_exit(&dn->dn_struct_rwlock);
* the end so that the file gets shorter over time (if we crashes in the
* middle, this will leave us in a better state). We find allocated file
* data by simply searching the allocated level 1 indirects.
+ *
+ * On input, *start should be the first offset that does not need to be
+ * freed (e.g. "offset + length"). On return, *start will be the first
+ * offset that should be freed.
*/
static int
-get_next_chunk(dnode_t *dn, uint64_t *start, uint64_t limit)
+get_next_chunk(dnode_t *dn, uint64_t *start, uint64_t minimum)
{
- uint64_t len = *start - limit;
- uint64_t blkcnt = 0;
- uint64_t maxblks = DMU_MAX_ACCESS / (1ULL << (dn->dn_indblkshift + 1));
+ uint64_t maxblks = DMU_MAX_ACCESS >> (dn->dn_indblkshift + 1);
+ /* bytes of data covered by a level-1 indirect block */
uint64_t iblkrange =
dn->dn_datablksz * EPB(dn->dn_indblkshift, SPA_BLKPTRSHIFT);
- ASSERT(limit <= *start);
+ ASSERT3U(minimum, <=, *start);
- if (len <= iblkrange * maxblks) {
- *start = limit;
+ if (*start - minimum <= iblkrange * maxblks) {
+ *start = minimum;
return (0);
}
ASSERT(ISP2(iblkrange));
- while (*start > limit && blkcnt < maxblks) {
+ for (uint64_t blks = 0; *start > minimum && blks < maxblks; blks++) {
int err;
- /* find next allocated L1 indirect */
+ /*
+ * dnode_next_offset(BACKWARDS) will find an allocated L1
+ * indirect block at or before the input offset. We must
+ * decrement *start so that it is at the end of the region
+ * to search.
+ */
+ (*start)--;
err = dnode_next_offset(dn,
DNODE_FIND_BACKWARDS, start, 2, 1, 0);
- /* if there are no more, then we are done */
+ /* if there are no indirect blocks before start, we are done */
if (err == ESRCH) {
- *start = limit;
- return (0);
- } else if (err) {
+ *start = minimum;
+ break;
+ } else if (err != 0) {
return (err);
}
- blkcnt += 1;
- /* reset offset to end of "next" block back */
+ /* set start to the beginning of this L1 indirect */
*start = P2ALIGN(*start, iblkrange);
- if (*start <= limit)
- *start = limit;
- else
- *start -= 1;
}
+ if (*start < minimum)
+ *start = minimum;
return (0);
}
+/*
+ * If this objset is of type OST_ZFS return true if vfs's unmounted flag is set,
+ * otherwise return false.
+ * Used below in dmu_free_long_range_impl() to enable abort when unmounting
+ */
+/*ARGSUSED*/
+static boolean_t
+dmu_objset_zfs_unmounting(objset_t *os)
+{
+#ifdef _KERNEL
+ if (dmu_objset_type(os) == DMU_OST_ZFS)
+ return (zfs_get_vfs_flag_unmounted(os));
+#endif
+ return (B_FALSE);
+}
+
static int
dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset,
- uint64_t length, boolean_t free_dnode)
+ uint64_t length, boolean_t raw)
{
- dmu_tx_t *tx;
- uint64_t object_size, start, end, len;
- boolean_t trunc = (length == DMU_OBJECT_END);
- int align, err;
-
- align = 1 << dn->dn_datablkshift;
- ASSERT(align > 0);
- object_size = align == 1 ? dn->dn_datablksz :
- (dn->dn_maxblkid + 1) << dn->dn_datablkshift;
-
- end = offset + length;
- if (trunc || end > object_size)
- end = object_size;
- if (end <= offset)
+ uint64_t object_size;
+ int err;
+ uint64_t dirty_frees_threshold;
+ dsl_pool_t *dp = dmu_objset_pool(os);
+
+ if (dn == NULL)
+ return (SET_ERROR(EINVAL));
+
+ object_size = (dn->dn_maxblkid + 1) * dn->dn_datablksz;
+ if (offset >= object_size)
return (0);
- length = end - offset;
- while (length) {
- start = end;
- /* assert(offset <= start) */
- err = get_next_chunk(dn, &start, offset);
+ if (zfs_per_txg_dirty_frees_percent <= 100)
+ dirty_frees_threshold =
+ zfs_per_txg_dirty_frees_percent * zfs_dirty_data_max / 100;
+ else
+ dirty_frees_threshold = zfs_dirty_data_max / 4;
+
+ if (length == DMU_OBJECT_END || offset + length > object_size)
+ length = object_size - offset;
+
+ while (length != 0) {
+ uint64_t chunk_end, chunk_begin, chunk_len;
+ uint64_t long_free_dirty_all_txgs = 0;
+ dmu_tx_t *tx;
+
+ if (dmu_objset_zfs_unmounting(dn->dn_objset))
+ return (SET_ERROR(EINTR));
+
+ chunk_end = chunk_begin = offset + length;
+
+ /* move chunk_begin backwards to the beginning of this chunk */
+ err = get_next_chunk(dn, &chunk_begin, offset);
if (err)
return (err);
- len = trunc ? DMU_OBJECT_END : end - start;
+ ASSERT3U(chunk_begin, >=, offset);
+ ASSERT3U(chunk_begin, <=, chunk_end);
+
+ chunk_len = chunk_end - chunk_begin;
+
+ mutex_enter(&dp->dp_lock);
+ for (int t = 0; t < TXG_SIZE; t++) {
+ long_free_dirty_all_txgs +=
+ dp->dp_long_free_dirty_pertxg[t];
+ }
+ mutex_exit(&dp->dp_lock);
+
+ /*
+ * To avoid filling up a TXG with just frees wait for
+ * the next TXG to open before freeing more chunks if
+ * we have reached the threshold of frees
+ */
+ if (dirty_frees_threshold != 0 &&
+ long_free_dirty_all_txgs >= dirty_frees_threshold) {
+ txg_wait_open(dp, 0);
+ continue;
+ }
tx = dmu_tx_create(os);
- dmu_tx_hold_free(tx, dn->dn_object, start, len);
+ dmu_tx_hold_free(tx, dn->dn_object, chunk_begin, chunk_len);
+
+ /*
+ * Mark this transaction as typically resulting in a net
+ * reduction in space used.
+ */
+ dmu_tx_mark_netfree(tx);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err) {
dmu_tx_abort(tx);
return (err);
}
- dnode_free_range(dn, start, trunc ? -1 : len, tx);
-
- if (start == 0 && free_dnode) {
- ASSERT(trunc);
- dnode_free(dn, tx);
+ mutex_enter(&dp->dp_lock);
+ dp->dp_long_free_dirty_pertxg[dmu_tx_get_txg(tx) & TXG_MASK] +=
+ chunk_len;
+ mutex_exit(&dp->dp_lock);
+ DTRACE_PROBE3(free__long__range,
+ uint64_t, long_free_dirty_all_txgs, uint64_t, chunk_len,
+ uint64_t, dmu_tx_get_txg(tx));
+ dnode_free_range(dn, chunk_begin, chunk_len, tx);
+
+ /* if this is a raw free, mark the dirty record as such */
+ if (raw) {
+ dbuf_dirty_record_t *dr = dn->dn_dbuf->db_last_dirty;
+
+ while (dr != NULL && dr->dr_txg > tx->tx_txg)
+ dr = dr->dr_next;
+ if (dr != NULL && dr->dr_txg == tx->tx_txg)
+ dr->dt.dl.dr_raw = B_TRUE;
}
- length -= end - start;
-
dmu_tx_commit(tx);
- end = start;
+
+ length -= chunk_len;
}
return (0);
}
err = dnode_hold(os, object, FTAG, &dn);
if (err != 0)
return (err);
- err = dmu_free_long_range_impl(os, dn, offset, length, FALSE);
+ err = dmu_free_long_range_impl(os, dn, offset, length, B_FALSE);
+
+ /*
+ * It is important to zero out the maxblkid when freeing the entire
+ * file, so that (a) subsequent calls to dmu_free_long_range_impl()
+ * will take the fast path, and (b) dnode_reallocate() can verify
+ * that the entire file has been freed.
+ */
+ if (err == 0 && offset == 0 && length == DMU_OBJECT_END)
+ dn->dn_maxblkid = 0;
+
dnode_rele(dn, FTAG);
return (err);
}
+/*
+ * This function is equivalent to dmu_free_long_range(), but also
+ * marks the new dirty record as a raw write.
+ */
int
-dmu_free_object(objset_t *os, uint64_t object)
+dmu_free_long_range_raw(objset_t *os, uint64_t object,
+ uint64_t offset, uint64_t length)
{
dnode_t *dn;
+ int err;
+
+ err = dnode_hold(os, object, FTAG, &dn);
+ if (err != 0)
+ return (err);
+ err = dmu_free_long_range_impl(os, dn, offset, length, B_TRUE);
+
+ /*
+ * It is important to zero out the maxblkid when freeing the entire
+ * file, so that (a) subsequent calls to dmu_free_long_range_impl()
+ * will take the fast path, and (b) dnode_reallocate() can verify
+ * that the entire file has been freed.
+ */
+ if (err == 0 && offset == 0 && length == DMU_OBJECT_END)
+ dn->dn_maxblkid = 0;
+
+ dnode_rele(dn, FTAG);
+ return (err);
+}
+
+static int
+dmu_free_long_object_impl(objset_t *os, uint64_t object, boolean_t raw)
+{
dmu_tx_t *tx;
int err;
- err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED,
- FTAG, &dn);
+ err = dmu_free_long_range(os, object, 0, DMU_OBJECT_END);
if (err != 0)
return (err);
- if (dn->dn_nlevels == 1) {
- tx = dmu_tx_create(os);
- dmu_tx_hold_bonus(tx, object);
- dmu_tx_hold_free(tx, dn->dn_object, 0, DMU_OBJECT_END);
- err = dmu_tx_assign(tx, TXG_WAIT);
- if (err == 0) {
- dnode_free_range(dn, 0, DMU_OBJECT_END, tx);
- dnode_free(dn, tx);
- dmu_tx_commit(tx);
- } else {
- dmu_tx_abort(tx);
- }
+
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_bonus(tx, object);
+ dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END);
+ dmu_tx_mark_netfree(tx);
+ err = dmu_tx_assign(tx, TXG_WAIT);
+ if (err == 0) {
+ if (raw)
+ err = dmu_object_dirty_raw(os, object, tx);
+ if (err == 0)
+ err = dmu_object_free(os, object, tx);
+
+ dmu_tx_commit(tx);
} else {
- err = dmu_free_long_range_impl(os, dn, 0, DMU_OBJECT_END, TRUE);
+ dmu_tx_abort(tx);
}
- dnode_rele(dn, FTAG);
+
return (err);
}
+int
+dmu_free_long_object(objset_t *os, uint64_t object)
+{
+ return (dmu_free_long_object_impl(os, object, B_FALSE));
+}
+
+int
+dmu_free_long_object_raw(objset_t *os, uint64_t object)
+{
+ return (dmu_free_long_object_impl(os, object, B_TRUE));
+}
+
+
int
dmu_free_range(objset_t *os, uint64_t object, uint64_t offset,
uint64_t size, dmu_tx_t *tx)
if (err)
return (err);
ASSERT(offset < UINT64_MAX);
- ASSERT(size == -1ULL || size <= UINT64_MAX - offset);
+ ASSERT(size == DMU_OBJECT_END || size <= UINT64_MAX - offset);
dnode_free_range(dn, offset, size, tx);
dnode_rele(dn, FTAG);
return (0);
}
-int
-dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
+static int
+dmu_read_impl(dnode_t *dn, uint64_t offset, uint64_t size,
void *buf, uint32_t flags)
{
- dnode_t *dn;
dmu_buf_t **dbp;
- int numbufs, err;
-
- err = dnode_hold(os, object, FTAG, &dn);
- if (err)
- return (err);
+ int numbufs, err = 0;
/*
* Deal with odd block sizes, where there can't be data past the first
* handle that here as well.
*/
if (dn->dn_maxblkid == 0) {
- int newsz = offset > dn->dn_datablksz ? 0 :
+ uint64_t newsz = offset > dn->dn_datablksz ? 0 :
MIN(size, dn->dn_datablksz - offset);
bzero((char *)buf + newsz, size - newsz);
size = newsz;
break;
for (i = 0; i < numbufs; i++) {
- int tocpy;
- int bufoff;
+ uint64_t tocpy;
+ int64_t bufoff;
dmu_buf_t *db = dbp[i];
ASSERT(size > 0);
bufoff = offset - db->db_offset;
- tocpy = (int)MIN(db->db_size - bufoff, size);
+ tocpy = MIN(db->db_size - bufoff, size);
- bcopy((char *)db->db_data + bufoff, buf, tocpy);
+ (void) memcpy(buf, (char *)db->db_data + bufoff, tocpy);
offset += tocpy;
size -= tocpy;
}
dmu_buf_rele_array(dbp, numbufs, FTAG);
}
- dnode_rele(dn, FTAG);
return (err);
}
-void
-dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
- const void *buf, dmu_tx_t *tx)
+int
+dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
+ void *buf, uint32_t flags)
{
- dmu_buf_t **dbp;
- int numbufs, i;
+ dnode_t *dn;
+ int err;
- if (size == 0)
- return;
+ err = dnode_hold(os, object, FTAG, &dn);
+ if (err != 0)
+ return (err);
- VERIFY(0 == dmu_buf_hold_array(os, object, offset, size,
- FALSE, FTAG, &numbufs, &dbp));
+ err = dmu_read_impl(dn, offset, size, buf, flags);
+ dnode_rele(dn, FTAG);
+ return (err);
+}
+
+int
+dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf,
+ uint32_t flags)
+{
+ return (dmu_read_impl(dn, offset, size, buf, flags));
+}
+
+static void
+dmu_write_impl(dmu_buf_t **dbp, int numbufs, uint64_t offset, uint64_t size,
+ const void *buf, dmu_tx_t *tx)
+{
+ int i;
for (i = 0; i < numbufs; i++) {
- int tocpy;
- int bufoff;
+ uint64_t tocpy;
+ int64_t bufoff;
dmu_buf_t *db = dbp[i];
ASSERT(size > 0);
bufoff = offset - db->db_offset;
- tocpy = (int)MIN(db->db_size - bufoff, size);
+ tocpy = MIN(db->db_size - bufoff, size);
ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);
size -= tocpy;
buf = (char *)buf + tocpy;
}
+}
+
+void
+dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size,
+ const void *buf, dmu_tx_t *tx)
+{
+ dmu_buf_t **dbp;
+ int numbufs;
+
+ if (size == 0)
+ return;
+
+ VERIFY0(dmu_buf_hold_array(os, object, offset, size,
+ FALSE, FTAG, &numbufs, &dbp));
+ dmu_write_impl(dbp, numbufs, offset, size, buf, tx);
+ dmu_buf_rele_array(dbp, numbufs, FTAG);
+}
+
+void
+dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size,
+ const void *buf, dmu_tx_t *tx)
+{
+ dmu_buf_t **dbp;
+ int numbufs;
+
+ if (size == 0)
+ return;
+
+ VERIFY0(dmu_buf_hold_array_by_dnode(dn, offset, size,
+ FALSE, FTAG, &numbufs, &dbp, DMU_READ_PREFETCH));
+ dmu_write_impl(dbp, numbufs, offset, size, buf, tx);
dmu_buf_rele_array(dbp, numbufs, FTAG);
}
dmu_buf_rele_array(dbp, numbufs, FTAG);
}
+void
+dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset,
+ void *data, uint8_t etype, uint8_t comp, int uncompressed_size,
+ int compressed_size, int byteorder, dmu_tx_t *tx)
+{
+ dmu_buf_t *db;
+
+ ASSERT3U(etype, <, NUM_BP_EMBEDDED_TYPES);
+ ASSERT3U(comp, <, ZIO_COMPRESS_FUNCTIONS);
+ VERIFY0(dmu_buf_hold_noread(os, object, offset,
+ FTAG, &db));
+
+ dmu_buf_write_embedded(db,
+ data, (bp_embedded_type_t)etype, (enum zio_compress)comp,
+ uncompressed_size, compressed_size, byteorder, tx);
+
+ dmu_buf_rele(db, FTAG);
+}
+
/*
* DMU support for xuio
*/
{ "write_buf_nocopy", KSTAT_DATA_UINT64 }
};
-#define XUIOSTAT_INCR(stat, val) \
- atomic_add_64(&xuio_stats.stat.value.ui64, (val))
-#define XUIOSTAT_BUMP(stat) XUIOSTAT_INCR(stat, 1)
+#define XUIOSTAT_INCR(stat, val) \
+ atomic_add_64(&xuio_stats.stat.value.ui64, (val))
+#define XUIOSTAT_BUMP(stat) XUIOSTAT_INCR(stat, 1)
+#ifdef HAVE_UIO_ZEROCOPY
int
dmu_xuio_init(xuio_t *xuio, int nblk)
{
uio_t *uio = &xuio->xu_uio;
uio->uio_iovcnt = nblk;
- uio->uio_iov = kmem_zalloc(nblk * sizeof (iovec_t), KM_PUSHPAGE);
+ uio->uio_iov = kmem_zalloc(nblk * sizeof (iovec_t), KM_SLEEP);
- priv = kmem_zalloc(sizeof (dmu_xuio_t), KM_PUSHPAGE);
+ priv = kmem_zalloc(sizeof (dmu_xuio_t), KM_SLEEP);
priv->cnt = nblk;
- priv->bufs = kmem_zalloc(nblk * sizeof (arc_buf_t *), KM_PUSHPAGE);
- priv->iovp = uio->uio_iov;
+ priv->bufs = kmem_zalloc(nblk * sizeof (arc_buf_t *), KM_SLEEP);
+ priv->iovp = (iovec_t *)uio->uio_iov;
XUIO_XUZC_PRIV(xuio) = priv;
if (XUIO_XUZC_RW(xuio) == UIO_READ)
int i = priv->next++;
ASSERT(i < priv->cnt);
- ASSERT(off + n <= arc_buf_size(abuf));
- iov = uio->uio_iov + i;
+ ASSERT(off + n <= arc_buf_lsize(abuf));
+ iov = (iovec_t *)uio->uio_iov + i;
iov->iov_base = (char *)abuf->b_data + off;
iov->iov_len = n;
priv->bufs[i] = abuf;
ASSERT(i < priv->cnt);
priv->bufs[i] = NULL;
}
+#endif /* HAVE_UIO_ZEROCOPY */
static void
xuio_stat_init(void)
}
void
-xuio_stat_wbuf_copied()
+xuio_stat_wbuf_copied(void)
{
XUIOSTAT_BUMP(xuiostat_wbuf_copied);
}
void
-xuio_stat_wbuf_nocopy()
+xuio_stat_wbuf_nocopy(void)
{
XUIOSTAT_BUMP(xuiostat_wbuf_nocopy);
}
#ifdef _KERNEL
-
-/*
- * Copy up to size bytes between arg_buf and req based on the data direction
- * described by the req. If an entire req's data cannot be transfered the
- * req's is updated such that it's current index and bv offsets correctly
- * reference any residual data which could not be copied. The return value
- * is the number of bytes successfully copied to arg_buf.
- */
-static int
-dmu_req_copy(void *arg_buf, int size, int *offset, struct request *req)
+int
+dmu_read_uio_dnode(dnode_t *dn, uio_t *uio, uint64_t size)
{
- struct bio_vec *bv;
- struct req_iterator iter;
- char *bv_buf;
- int tocpy;
-
- *offset = 0;
- rq_for_each_segment(bv, req, iter) {
-
- /* Fully consumed the passed arg_buf */
- ASSERT3S(*offset, <=, size);
- if (size == *offset)
- break;
-
- /* Skip fully consumed bv's */
- if (bv->bv_len == 0)
- continue;
-
- tocpy = MIN(bv->bv_len, size - *offset);
- ASSERT3S(tocpy, >=, 0);
-
- bv_buf = page_address(bv->bv_page) + bv->bv_offset;
- ASSERT3P(bv_buf, !=, NULL);
-
- if (rq_data_dir(req) == WRITE)
- memcpy(arg_buf + *offset, bv_buf, tocpy);
- else
- memcpy(bv_buf, arg_buf + *offset, tocpy);
-
- *offset += tocpy;
- bv->bv_offset += tocpy;
- bv->bv_len -= tocpy;
- }
-
- return 0;
-}
-
-static void
-dmu_bio_put(struct bio *bio)
-{
- struct bio *bio_next;
-
- while (bio) {
- bio_next = bio->bi_next;
- bio_put(bio);
- bio = bio_next;
- }
-}
-
-static int
-dmu_bio_clone(struct bio *bio, struct bio **bio_copy)
-{
- struct bio *bio_root = NULL;
- struct bio *bio_last = NULL;
- struct bio *bio_new;
-
- if (bio == NULL)
- return EINVAL;
-
- while (bio) {
- bio_new = bio_clone(bio, GFP_NOIO);
- if (bio_new == NULL) {
- dmu_bio_put(bio_root);
- return ENOMEM;
- }
-
- if (bio_last) {
- bio_last->bi_next = bio_new;
- bio_last = bio_new;
- } else {
- bio_root = bio_new;
- bio_last = bio_new;
- }
-
- bio = bio->bi_next;
- }
-
- *bio_copy = bio_root;
-
- return 0;
-}
-
-int
-dmu_read_req(objset_t *os, uint64_t object, struct request *req)
-{
- uint64_t size = blk_rq_bytes(req);
- uint64_t offset = blk_rq_pos(req) << 9;
- struct bio *bio_saved = req->bio;
dmu_buf_t **dbp;
int numbufs, i, err;
+#ifdef HAVE_UIO_ZEROCOPY
+ xuio_t *xuio = NULL;
+#endif
/*
* NB: we could do this block-at-a-time, but it's nice
* to be reading in parallel.
*/
- err = dmu_buf_hold_array(os, object, offset, size, TRUE, FTAG,
- &numbufs, &dbp);
+ err = dmu_buf_hold_array_by_dnode(dn, uio->uio_loffset, size,
+ TRUE, FTAG, &numbufs, &dbp, 0);
if (err)
return (err);
- /*
- * Clone the bio list so the bv->bv_offset and bv->bv_len members
- * can be safely modified. The original bio list is relinked in to
- * the request when the function exits. This is required because
- * some file systems blindly assume that these values will remain
- * constant between bio_submit() and the IO completion callback.
- */
- err = dmu_bio_clone(bio_saved, &req->bio);
- if (err)
- goto error;
-
for (i = 0; i < numbufs; i++) {
- int tocpy, didcpy, bufoff;
+ uint64_t tocpy;
+ int64_t bufoff;
dmu_buf_t *db = dbp[i];
- bufoff = offset - db->db_offset;
- ASSERT3S(bufoff, >=, 0);
-
- tocpy = (int)MIN(db->db_size - bufoff, size);
- if (tocpy == 0)
- break;
+ ASSERT(size > 0);
- err = dmu_req_copy(db->db_data + bufoff, tocpy, &didcpy, req);
+ bufoff = uio->uio_loffset - db->db_offset;
+ tocpy = MIN(db->db_size - bufoff, size);
- if (didcpy < tocpy)
- err = EIO;
+#ifdef HAVE_UIO_ZEROCOPY
+ if (xuio) {
+ dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db;
+ arc_buf_t *dbuf_abuf = dbi->db_buf;
+ arc_buf_t *abuf = dbuf_loan_arcbuf(dbi);
+ err = dmu_xuio_add(xuio, abuf, bufoff, tocpy);
+ if (!err) {
+ uio->uio_resid -= tocpy;
+ uio->uio_loffset += tocpy;
+ }
+ if (abuf == dbuf_abuf)
+ XUIOSTAT_BUMP(xuiostat_rbuf_nocopy);
+ else
+ XUIOSTAT_BUMP(xuiostat_rbuf_copied);
+ } else
+#endif
+ err = uiomove((char *)db->db_data + bufoff, tocpy,
+ UIO_READ, uio);
if (err)
break;
size -= tocpy;
- offset += didcpy;
- err = 0;
}
-
- dmu_bio_put(req->bio);
- req->bio = bio_saved;
-error:
dmu_buf_rele_array(dbp, numbufs, FTAG);
return (err);
}
+/*
+ * Read 'size' bytes into the uio buffer.
+ * From object zdb->db_object.
+ * Starting at offset uio->uio_loffset.
+ *
+ * If the caller already has a dbuf in the target object
+ * (e.g. its bonus buffer), this routine is faster than dmu_read_uio(),
+ * because we don't have to find the dnode_t for the object.
+ */
int
-dmu_write_req(objset_t *os, uint64_t object, struct request *req, dmu_tx_t *tx)
+dmu_read_uio_dbuf(dmu_buf_t *zdb, uio_t *uio, uint64_t size)
{
- uint64_t size = blk_rq_bytes(req);
- uint64_t offset = blk_rq_pos(req) << 9;
- struct bio *bio_saved = req->bio;
- dmu_buf_t **dbp;
- int numbufs;
- int err = 0;
- int i;
+ dmu_buf_impl_t *db = (dmu_buf_impl_t *)zdb;
+ dnode_t *dn;
+ int err;
if (size == 0)
return (0);
- err = dmu_buf_hold_array(os, object, offset, size, FALSE, FTAG,
- &numbufs, &dbp);
- if (err)
- return (err);
-
- /*
- * Clone the bio list so the bv->bv_offset and bv->bv_len members
- * can be safely modified. The original bio list is relinked in to
- * the request when the function exits. This is required because
- * some file systems blindly assume that these values will remain
- * constant between bio_submit() and the IO completion callback.
- */
- err = dmu_bio_clone(bio_saved, &req->bio);
- if (err)
- goto error;
-
- for (i = 0; i < numbufs; i++) {
- int tocpy, didcpy, bufoff;
- dmu_buf_t *db = dbp[i];
-
- bufoff = offset - db->db_offset;
- ASSERT3S(bufoff, >=, 0);
-
- tocpy = (int)MIN(db->db_size - bufoff, size);
- if (tocpy == 0)
- break;
-
- ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);
-
- if (tocpy == db->db_size)
- dmu_buf_will_fill(db, tx);
- else
- dmu_buf_will_dirty(db, tx);
-
- err = dmu_req_copy(db->db_data + bufoff, tocpy, &didcpy, req);
-
- if (tocpy == db->db_size)
- dmu_buf_fill_done(db, tx);
-
- if (didcpy < tocpy)
- err = EIO;
-
- if (err)
- break;
-
- size -= tocpy;
- offset += didcpy;
- err = 0;
- }
-
- dmu_bio_put(req->bio);
- req->bio = bio_saved;
-error:
- dmu_buf_rele_array(dbp, numbufs, FTAG);
+ DB_DNODE_ENTER(db);
+ dn = DB_DNODE(db);
+ err = dmu_read_uio_dnode(dn, uio, size);
+ DB_DNODE_EXIT(db);
return (err);
}
+/*
+ * Read 'size' bytes into the uio buffer.
+ * From the specified object
+ * Starting at offset uio->uio_loffset.
+ */
int
dmu_read_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size)
{
- dmu_buf_t **dbp;
- int numbufs, i, err;
- xuio_t *xuio = NULL;
+ dnode_t *dn;
+ int err;
- /*
- * NB: we could do this block-at-a-time, but it's nice
- * to be reading in parallel.
- */
- err = dmu_buf_hold_array(os, object, uio->uio_loffset, size, TRUE, FTAG,
- &numbufs, &dbp);
+ if (size == 0)
+ return (0);
+
+ err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
- for (i = 0; i < numbufs; i++) {
- int tocpy;
- int bufoff;
- dmu_buf_t *db = dbp[i];
+ err = dmu_read_uio_dnode(dn, uio, size);
- ASSERT(size > 0);
-
- bufoff = uio->uio_loffset - db->db_offset;
- tocpy = (int)MIN(db->db_size - bufoff, size);
-
- if (xuio) {
- dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db;
- arc_buf_t *dbuf_abuf = dbi->db_buf;
- arc_buf_t *abuf = dbuf_loan_arcbuf(dbi);
- err = dmu_xuio_add(xuio, abuf, bufoff, tocpy);
- if (!err) {
- uio->uio_resid -= tocpy;
- uio->uio_loffset += tocpy;
- }
-
- if (abuf == dbuf_abuf)
- XUIOSTAT_BUMP(xuiostat_rbuf_nocopy);
- else
- XUIOSTAT_BUMP(xuiostat_rbuf_copied);
- } else {
- err = uiomove((char *)db->db_data + bufoff, tocpy,
- UIO_READ, uio);
- }
- if (err)
- break;
-
- size -= tocpy;
- }
- dmu_buf_rele_array(dbp, numbufs, FTAG);
+ dnode_rele(dn, FTAG);
return (err);
}
-static int
+int
dmu_write_uio_dnode(dnode_t *dn, uio_t *uio, uint64_t size, dmu_tx_t *tx)
{
dmu_buf_t **dbp;
return (err);
for (i = 0; i < numbufs; i++) {
- int tocpy;
- int bufoff;
+ uint64_t tocpy;
+ int64_t bufoff;
dmu_buf_t *db = dbp[i];
ASSERT(size > 0);
bufoff = uio->uio_loffset - db->db_offset;
- tocpy = (int)MIN(db->db_size - bufoff, size);
+ tocpy = MIN(db->db_size - bufoff, size);
ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size);
return (err);
}
+/*
+ * Write 'size' bytes from the uio buffer.
+ * To object zdb->db_object.
+ * Starting at offset uio->uio_loffset.
+ *
+ * If the caller already has a dbuf in the target object
+ * (e.g. its bonus buffer), this routine is faster than dmu_write_uio(),
+ * because we don't have to find the dnode_t for the object.
+ */
int
dmu_write_uio_dbuf(dmu_buf_t *zdb, uio_t *uio, uint64_t size,
dmu_tx_t *tx)
return (err);
}
+/*
+ * Write 'size' bytes from the uio buffer.
+ * To the specified object.
+ * Starting at offset uio->uio_loffset.
+ */
int
dmu_write_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size,
dmu_tx_t *tx)
dmu_request_arcbuf(dmu_buf_t *handle, int size)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)handle;
- spa_t *spa;
- DB_GET_SPA(&spa, db);
- return (arc_loan_buf(spa, size));
+ return (arc_loan_buf(db->db_objset->os_spa, B_FALSE, size));
}
/*
dmu_return_arcbuf(arc_buf_t *buf)
{
arc_return_buf(buf, FTAG);
- VERIFY(arc_buf_remove_ref(buf, FTAG) == 1);
+ arc_buf_destroy(buf, FTAG);
+}
+
+void
+dmu_convert_to_raw(dmu_buf_t *handle, boolean_t byteorder, const uint8_t *salt,
+ const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx)
+{
+ dmu_object_type_t type;
+ dmu_buf_impl_t *db = (dmu_buf_impl_t *)handle;
+ uint64_t dsobj = dmu_objset_id(db->db_objset);
+
+ ASSERT3P(db->db_buf, !=, NULL);
+ ASSERT3U(dsobj, !=, 0);
+
+ dmu_buf_will_change_crypt_params(handle, tx);
+
+ DB_DNODE_ENTER(db);
+ type = DB_DNODE(db)->dn_type;
+ DB_DNODE_EXIT(db);
+
+ /*
+ * This technically violates the assumption the dmu code makes
+ * that dnode blocks are only released in syncing context.
+ */
+ (void) arc_release(db->db_buf, db);
+ arc_convert_to_raw(db->db_buf, dsobj, byteorder, type, salt, iv, mac);
+}
+
+void
+dmu_copy_from_buf(objset_t *os, uint64_t object, uint64_t offset,
+ dmu_buf_t *handle, dmu_tx_t *tx)
+{
+ dmu_buf_t *dst_handle;
+ dmu_buf_impl_t *dstdb;
+ dmu_buf_impl_t *srcdb = (dmu_buf_impl_t *)handle;
+ arc_buf_t *abuf;
+ uint64_t datalen;
+ boolean_t byteorder;
+ uint8_t salt[ZIO_DATA_SALT_LEN];
+ uint8_t iv[ZIO_DATA_IV_LEN];
+ uint8_t mac[ZIO_DATA_MAC_LEN];
+
+ ASSERT3P(srcdb->db_buf, !=, NULL);
+
+ /* hold the db that we want to write to */
+ VERIFY0(dmu_buf_hold(os, object, offset, FTAG, &dst_handle,
+ DMU_READ_NO_DECRYPT));
+ dstdb = (dmu_buf_impl_t *)dst_handle;
+ datalen = arc_buf_size(srcdb->db_buf);
+
+ /* allocated an arc buffer that matches the type of srcdb->db_buf */
+ if (arc_is_encrypted(srcdb->db_buf)) {
+ arc_get_raw_params(srcdb->db_buf, &byteorder, salt, iv, mac);
+ abuf = arc_loan_raw_buf(os->os_spa, dmu_objset_id(os),
+ byteorder, salt, iv, mac, DB_DNODE(dstdb)->dn_type,
+ datalen, arc_buf_lsize(srcdb->db_buf),
+ arc_get_compression(srcdb->db_buf));
+ } else {
+ /* we won't get a compressed db back from dmu_buf_hold() */
+ ASSERT3U(arc_get_compression(srcdb->db_buf),
+ ==, ZIO_COMPRESS_OFF);
+ abuf = arc_loan_buf(os->os_spa,
+ DMU_OT_IS_METADATA(DB_DNODE(dstdb)->dn_type), datalen);
+ }
+
+ ASSERT3U(datalen, ==, arc_buf_size(abuf));
+
+ /* copy the data to the new buffer and assign it to the dstdb */
+ bcopy(srcdb->db_buf->b_data, abuf->b_data, datalen);
+ dbuf_assign_arcbuf(dstdb, abuf, tx);
+ dmu_buf_rele(dst_handle, FTAG);
}
/*
* dmu_write().
*/
void
-dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, arc_buf_t *buf,
+dmu_assign_arcbuf_by_dnode(dnode_t *dn, uint64_t offset, arc_buf_t *buf,
dmu_tx_t *tx)
{
- dmu_buf_impl_t *dbuf = (dmu_buf_impl_t *)handle;
- dnode_t *dn;
dmu_buf_impl_t *db;
- uint32_t blksz = (uint32_t)arc_buf_size(buf);
+ objset_t *os = dn->dn_objset;
+ uint64_t object = dn->dn_object;
+ uint32_t blksz = (uint32_t)arc_buf_lsize(buf);
uint64_t blkid;
- DB_DNODE_ENTER(dbuf);
- dn = DB_DNODE(dbuf);
rw_enter(&dn->dn_struct_rwlock, RW_READER);
- blkid = dbuf_whichblock(dn, offset);
+ blkid = dbuf_whichblock(dn, 0, offset);
VERIFY((db = dbuf_hold(dn, blkid, FTAG)) != NULL);
rw_exit(&dn->dn_struct_rwlock);
- DB_DNODE_EXIT(dbuf);
+ /*
+ * We can only assign if the offset is aligned, the arc buf is the
+ * same size as the dbuf, and the dbuf is not metadata.
+ */
if (offset == db->db.db_offset && blksz == db->db.db_size) {
dbuf_assign_arcbuf(db, buf, tx);
dbuf_rele(db, FTAG);
} else {
- objset_t *os;
- uint64_t object;
-
- DB_DNODE_ENTER(dbuf);
- dn = DB_DNODE(dbuf);
- os = dn->dn_objset;
- object = dn->dn_object;
- DB_DNODE_EXIT(dbuf);
+ /* compressed bufs must always be assignable to their dbuf */
+ ASSERT3U(arc_get_compression(buf), ==, ZIO_COMPRESS_OFF);
+ ASSERT(!(buf->b_flags & ARC_BUF_FLAG_COMPRESSED));
dbuf_rele(db, FTAG);
dmu_write(os, object, offset, blksz, buf->b_data, tx);
}
}
+void
+dmu_assign_arcbuf_by_dbuf(dmu_buf_t *handle, uint64_t offset, arc_buf_t *buf,
+ dmu_tx_t *tx)
+{
+ dmu_buf_impl_t *dbuf = (dmu_buf_impl_t *)handle;
+
+ DB_DNODE_ENTER(dbuf);
+ dmu_assign_arcbuf_by_dnode(DB_DNODE(dbuf), offset, buf, tx);
+ DB_DNODE_EXIT(dbuf);
+}
+
typedef struct {
dbuf_dirty_record_t *dsa_dr;
dmu_sync_cb_t *dsa_done;
* block size still needs to be known for replay.
*/
BP_SET_LSIZE(bp, db->db_size);
- } else {
+ } else if (!BP_IS_EMBEDDED(bp)) {
ASSERT(BP_GET_LEVEL(bp) == 0);
- bp->blk_fill = 1;
+ BP_SET_FILL(bp, 1);
}
}
}
mutex_enter(&db->db_mtx);
ASSERT(dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC);
if (zio->io_error == 0) {
+ dr->dt.dl.dr_nopwrite = !!(zio->io_flags & ZIO_FLAG_NOPWRITE);
+ if (dr->dt.dl.dr_nopwrite) {
+ blkptr_t *bp = zio->io_bp;
+ blkptr_t *bp_orig = &zio->io_bp_orig;
+ uint8_t chksum = BP_GET_CHECKSUM(bp_orig);
+
+ ASSERT(BP_EQUAL(bp, bp_orig));
+ VERIFY(BP_EQUAL(bp, db->db_blkptr));
+ ASSERT(zio->io_prop.zp_compress != ZIO_COMPRESS_OFF);
+ VERIFY(zio_checksum_table[chksum].ci_flags &
+ ZCHECKSUM_FLAG_NOPWRITE);
+ }
dr->dt.dl.dr_overridden_by = *zio->io_bp;
dr->dt.dl.dr_override_state = DR_OVERRIDDEN;
dr->dt.dl.dr_copies = zio->io_prop.zp_copies;
- if (BP_IS_HOLE(&dr->dt.dl.dr_overridden_by))
+
+ /*
+ * Old style holes are filled with all zeros, whereas
+ * new-style holes maintain their lsize, type, level,
+ * and birth time (see zio_write_compress). While we
+ * need to reset the BP_SET_LSIZE() call that happened
+ * in dmu_sync_ready for old style holes, we do *not*
+ * want to wipe out the information contained in new
+ * style holes. Thus, only zero out the block pointer if
+ * it's an old style hole.
+ */
+ if (BP_IS_HOLE(&dr->dt.dl.dr_overridden_by) &&
+ dr->dt.dl.dr_overridden_by.blk_birth == 0)
BP_ZERO(&dr->dt.dl.dr_overridden_by);
} else {
dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN;
{
blkptr_t *bp = zio->io_bp;
dmu_sync_arg_t *dsa = zio->io_private;
+ ASSERTV(blkptr_t *bp_orig = &zio->io_bp_orig);
if (zio->io_error == 0 && !BP_IS_HOLE(bp)) {
+ ASSERT(!(zio->io_flags & ZIO_FLAG_NOPWRITE));
+ ASSERT(BP_IS_HOLE(bp_orig) || !BP_EQUAL(bp, bp_orig));
ASSERT(zio->io_bp->blk_birth == zio->io_txg);
ASSERT(zio->io_txg > spa_syncing_txg(zio->io_spa));
zio_free(zio->io_spa, zio->io_txg, zio->io_bp);
dsa->dsa_done(dsa->dsa_zgd, zio->io_error);
+ abd_put(zio->io_abd);
kmem_free(dsa, sizeof (*dsa));
}
static int
dmu_sync_late_arrival(zio_t *pio, objset_t *os, dmu_sync_cb_t *done, zgd_t *zgd,
- zio_prop_t *zp, zbookmark_t *zb)
+ zio_prop_t *zp, zbookmark_phys_t *zb)
{
dmu_sync_arg_t *dsa;
dmu_tx_t *tx;
dmu_tx_hold_space(tx, zgd->zgd_db->db_size);
if (dmu_tx_assign(tx, TXG_WAIT) != 0) {
dmu_tx_abort(tx);
- return (EIO); /* Make zl_get_data do txg_waited_synced() */
+ /* Make zl_get_data do txg_waited_synced() */
+ return (SET_ERROR(EIO));
}
- dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_PUSHPAGE);
+ dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP);
dsa->dsa_dr = NULL;
dsa->dsa_done = done;
dsa->dsa_zgd = zgd;
dsa->dsa_tx = tx;
+ /*
+ * Since we are currently syncing this txg, it's nontrivial to
+ * determine what BP to nopwrite against, so we disable nopwrite.
+ *
+ * When syncing, the db_blkptr is initially the BP of the previous
+ * txg. We can not nopwrite against it because it will be changed
+ * (this is similar to the non-late-arrival case where the dbuf is
+ * dirty in a future txg).
+ *
+ * Then dbuf_write_ready() sets bp_blkptr to the location we will write.
+ * We can not nopwrite against it because although the BP will not
+ * (typically) be changed, the data has not yet been persisted to this
+ * location.
+ *
+ * Finally, when dbuf_write_done() is called, it is theoretically
+ * possible to always nopwrite, because the data that was written in
+ * this txg is the same data that we are trying to write. However we
+ * would need to check that this dbuf is not dirty in any future
+ * txg's (as we do in the normal dmu_sync() path). For simplicity, we
+ * don't nopwrite in this case.
+ */
+ zp->zp_nopwrite = B_FALSE;
+
zio_nowait(zio_write(pio, os->os_spa, dmu_tx_get_txg(tx), zgd->zgd_bp,
- zgd->zgd_db->db_data, zgd->zgd_db->db_size, zp,
- dmu_sync_late_arrival_ready, dmu_sync_late_arrival_done, dsa,
- ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL | ZIO_FLAG_FASTWRITE, zb));
+ abd_get_from_buf(zgd->zgd_db->db_data, zgd->zgd_db->db_size),
+ zgd->zgd_db->db_size, zgd->zgd_db->db_size, zp,
+ dmu_sync_late_arrival_ready, NULL, NULL, dmu_sync_late_arrival_done,
+ dsa, ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, zb));
return (0);
}
*
* Return values:
*
- * EEXIST: this txg has already been synced, so there's nothing to to.
+ * EEXIST: this txg has already been synced, so there's nothing to do.
* The caller should not log the write.
*
* ENOENT: the block was dbuf_free_range()'d, so there's nothing to do.
int
dmu_sync(zio_t *pio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd)
{
- blkptr_t *bp = zgd->zgd_bp;
dmu_buf_impl_t *db = (dmu_buf_impl_t *)zgd->zgd_db;
objset_t *os = db->db_objset;
dsl_dataset_t *ds = os->os_dsl_dataset;
dbuf_dirty_record_t *dr;
dmu_sync_arg_t *dsa;
- zbookmark_t zb;
+ zbookmark_phys_t zb;
zio_prop_t zp;
dnode_t *dn;
ASSERT(pio != NULL);
- ASSERT(BP_IS_HOLE(bp));
ASSERT(txg != 0);
+ /* dbuf is within the locked range */
+ ASSERT3U(db->db.db_offset, >=, zgd->zgd_rl->r_off);
+ ASSERT3U(db->db.db_offset + db->db.db_size, <=,
+ zgd->zgd_rl->r_off + zgd->zgd_rl->r_len);
+
SET_BOOKMARK(&zb, ds->ds_object,
db->db.db_object, db->db_level, db->db_blkid);
* This txg has already synced. There's nothing to do.
*/
mutex_exit(&db->db_mtx);
- return (EEXIST);
+ return (SET_ERROR(EEXIST));
}
if (txg <= spa_syncing_txg(os->os_spa)) {
* There's no need to log writes to freed blocks, so we're done.
*/
mutex_exit(&db->db_mtx);
- return (ENOENT);
+ return (SET_ERROR(ENOENT));
}
+ ASSERT(dr->dr_next == NULL || dr->dr_next->dr_txg < txg);
+
+ if (db->db_blkptr != NULL) {
+ /*
+ * We need to fill in zgd_bp with the current blkptr so that
+ * the nopwrite code can check if we're writing the same
+ * data that's already on disk. We can only nopwrite if we
+ * are sure that after making the copy, db_blkptr will not
+ * change until our i/o completes. We ensure this by
+ * holding the db_mtx, and only allowing nopwrite if the
+ * block is not already dirty (see below). This is verified
+ * by dmu_sync_done(), which VERIFYs that the db_blkptr has
+ * not changed.
+ */
+ *zgd->zgd_bp = *db->db_blkptr;
+ }
+
+ /*
+ * Assume the on-disk data is X, the current syncing data (in
+ * txg - 1) is Y, and the current in-memory data is Z (currently
+ * in dmu_sync).
+ *
+ * We usually want to perform a nopwrite if X and Z are the
+ * same. However, if Y is different (i.e. the BP is going to
+ * change before this write takes effect), then a nopwrite will
+ * be incorrect - we would override with X, which could have
+ * been freed when Y was written.
+ *
+ * (Note that this is not a concern when we are nop-writing from
+ * syncing context, because X and Y must be identical, because
+ * all previous txgs have been synced.)
+ *
+ * Therefore, we disable nopwrite if the current BP could change
+ * before this TXG. There are two ways it could change: by
+ * being dirty (dr_next is non-NULL), or by being freed
+ * (dnode_block_freed()). This behavior is verified by
+ * zio_done(), which VERIFYs that the override BP is identical
+ * to the on-disk BP.
+ */
+ DB_DNODE_ENTER(db);
+ dn = DB_DNODE(db);
+ if (dr->dr_next != NULL || dnode_block_freed(dn, db->db_blkid))
+ zp.zp_nopwrite = B_FALSE;
+ DB_DNODE_EXIT(db);
+
ASSERT(dr->dr_txg == txg);
if (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC ||
dr->dt.dl.dr_override_state == DR_OVERRIDDEN) {
* have been dirtied since, or we would have cleared the state.
*/
mutex_exit(&db->db_mtx);
- return (EALREADY);
+ return (SET_ERROR(EALREADY));
}
ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN);
dr->dt.dl.dr_override_state = DR_IN_DMU_SYNC;
mutex_exit(&db->db_mtx);
- dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_PUSHPAGE);
+ dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP);
dsa->dsa_dr = dr;
dsa->dsa_done = done;
dsa->dsa_zgd = zgd;
dsa->dsa_tx = NULL;
zio_nowait(arc_write(pio, os->os_spa, txg,
- bp, dr->dt.dl.dr_data, DBUF_IS_L2CACHEABLE(db), &zp,
- dmu_sync_ready, dmu_sync_done, dsa,
- ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL | ZIO_FLAG_FASTWRITE, &zb));
+ zgd->zgd_bp, dr->dt.dl.dr_data, DBUF_IS_L2CACHEABLE(db),
+ &zp, dmu_sync_ready, NULL, NULL, dmu_sync_done, dsa,
+ ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, &zb));
return (0);
}
+int
+dmu_object_set_nlevels(objset_t *os, uint64_t object, int nlevels, dmu_tx_t *tx)
+{
+ dnode_t *dn;
+ int err;
+
+ err = dnode_hold(os, object, FTAG, &dn);
+ if (err)
+ return (err);
+ err = dnode_set_nlevels(dn, nlevels, tx);
+ dnode_rele(dn, FTAG);
+ return (err);
+}
+
int
dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, int ibs,
- dmu_tx_t *tx)
+ dmu_tx_t *tx)
{
dnode_t *dn;
int err;
void
dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum,
- dmu_tx_t *tx)
+ dmu_tx_t *tx)
{
dnode_t *dn;
- /* XXX assumes dnode_hold will not get an i/o error */
- (void) dnode_hold(os, object, FTAG, &dn);
- ASSERT(checksum < ZIO_CHECKSUM_FUNCTIONS);
+ /*
+ * Send streams include each object's checksum function. This
+ * check ensures that the receiving system can understand the
+ * checksum function transmitted.
+ */
+ ASSERT3U(checksum, <, ZIO_CHECKSUM_LEGACY_FUNCTIONS);
+
+ VERIFY0(dnode_hold(os, object, FTAG, &dn));
+ ASSERT3U(checksum, <, ZIO_CHECKSUM_FUNCTIONS);
dn->dn_checksum = checksum;
dnode_setdirty(dn, tx);
dnode_rele(dn, FTAG);
void
dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
- dmu_tx_t *tx)
+ dmu_tx_t *tx)
{
dnode_t *dn;
- /* XXX assumes dnode_hold will not get an i/o error */
- (void) dnode_hold(os, object, FTAG, &dn);
- ASSERT(compress < ZIO_COMPRESS_FUNCTIONS);
+ /*
+ * Send streams include each object's compression function. This
+ * check ensures that the receiving system can understand the
+ * compression function transmitted.
+ */
+ ASSERT3U(compress, <, ZIO_COMPRESS_LEGACY_FUNCTIONS);
+
+ VERIFY0(dnode_hold(os, object, FTAG, &dn));
dn->dn_compress = compress;
dnode_setdirty(dn, tx);
dnode_rele(dn, FTAG);
}
+/*
+ * Dirty an object and set the dirty record's raw flag. This is used
+ * when writing raw data to an object that will not effect the
+ * encryption parameters, specifically during raw receives.
+ */
+int
+dmu_object_dirty_raw(objset_t *os, uint64_t object, dmu_tx_t *tx)
+{
+ dnode_t *dn;
+ int err;
+
+ err = dnode_hold(os, object, FTAG, &dn);
+ if (err)
+ return (err);
+ dmu_buf_will_change_crypt_params((dmu_buf_t *)dn->dn_dbuf, tx);
+ dnode_rele(dn, FTAG);
+ return (err);
+}
+
int zfs_mdcomp_disable = 0;
+/*
+ * When the "redundant_metadata" property is set to "most", only indirect
+ * blocks of this level and higher will have an additional ditto block.
+ */
+int zfs_redundant_metadata_most_ditto_level = 2;
+
void
dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, zio_prop_t *zp)
{
dmu_object_type_t type = dn ? dn->dn_type : DMU_OT_OBJSET;
- boolean_t ismd = (level > 0 || dmu_ot[type].ot_metadata ||
+ boolean_t ismd = (level > 0 || DMU_OT_IS_METADATA(type) ||
(wp & WP_SPILL));
enum zio_checksum checksum = os->os_checksum;
enum zio_compress compress = os->os_compress;
enum zio_checksum dedup_checksum = os->os_dedup_checksum;
- boolean_t dedup;
+ boolean_t dedup = B_FALSE;
+ boolean_t nopwrite = B_FALSE;
boolean_t dedup_verify = os->os_dedup_verify;
+ boolean_t encrypt = B_FALSE;
int copies = os->os_copies;
/*
- * Determine checksum setting.
+ * We maintain different write policies for each of the following
+ * types of data:
+ * 1. metadata
+ * 2. preallocated blocks (i.e. level-0 blocks of a dump device)
+ * 3. all other level 0 blocks
*/
if (ismd) {
+ if (zfs_mdcomp_disable) {
+ compress = ZIO_COMPRESS_EMPTY;
+ } else {
+ /*
+ * XXX -- we should design a compression algorithm
+ * that specializes in arrays of bps.
+ */
+ compress = zio_compress_select(os->os_spa,
+ ZIO_COMPRESS_ON, ZIO_COMPRESS_ON);
+ }
+
/*
* Metadata always gets checksummed. If the data
* checksum is multi-bit correctable, and it's not a
* as well. Otherwise, the metadata checksum defaults
* to fletcher4.
*/
- if (zio_checksum_table[checksum].ci_correctable < 1 ||
- zio_checksum_table[checksum].ci_eck)
+ if (!(zio_checksum_table[checksum].ci_flags &
+ ZCHECKSUM_FLAG_METADATA) ||
+ (zio_checksum_table[checksum].ci_flags &
+ ZCHECKSUM_FLAG_EMBEDDED))
checksum = ZIO_CHECKSUM_FLETCHER_4;
- } else {
- checksum = zio_checksum_select(dn->dn_checksum, checksum);
- }
- /*
- * Determine compression setting.
- */
- if (ismd) {
+ if (os->os_redundant_metadata == ZFS_REDUNDANT_METADATA_ALL ||
+ (os->os_redundant_metadata ==
+ ZFS_REDUNDANT_METADATA_MOST &&
+ (level >= zfs_redundant_metadata_most_ditto_level ||
+ DMU_OT_IS_METADATA(type) || (wp & WP_SPILL))))
+ copies++;
+ } else if (wp & WP_NOFILL) {
+ ASSERT(level == 0);
+
/*
- * XXX -- we should design a compression algorithm
- * that specializes in arrays of bps.
+ * If we're writing preallocated blocks, we aren't actually
+ * writing them so don't set any policy properties. These
+ * blocks are currently only used by an external subsystem
+ * outside of zfs (i.e. dump) and not written by the zio
+ * pipeline.
*/
- compress = zfs_mdcomp_disable ? ZIO_COMPRESS_EMPTY :
- ZIO_COMPRESS_LZJB;
+ compress = ZIO_COMPRESS_OFF;
+ checksum = ZIO_CHECKSUM_OFF;
} else {
- compress = zio_compress_select(dn->dn_compress, compress);
+ compress = zio_compress_select(os->os_spa, dn->dn_compress,
+ compress);
+
+ checksum = (dedup_checksum == ZIO_CHECKSUM_OFF) ?
+ zio_checksum_select(dn->dn_checksum, checksum) :
+ dedup_checksum;
+
+ /*
+ * Determine dedup setting. If we are in dmu_sync(),
+ * we won't actually dedup now because that's all
+ * done in syncing context; but we do want to use the
+ * dedup checkum. If the checksum is not strong
+ * enough to ensure unique signatures, force
+ * dedup_verify.
+ */
+ if (dedup_checksum != ZIO_CHECKSUM_OFF) {
+ dedup = (wp & WP_DMU_SYNC) ? B_FALSE : B_TRUE;
+ if (!(zio_checksum_table[checksum].ci_flags &
+ ZCHECKSUM_FLAG_DEDUP))
+ dedup_verify = B_TRUE;
+ }
+
+ /*
+ * Enable nopwrite if we have secure enough checksum
+ * algorithm (see comment in zio_nop_write) and
+ * compression is enabled. We don't enable nopwrite if
+ * dedup is enabled as the two features are mutually
+ * exclusive.
+ */
+ nopwrite = (!dedup && (zio_checksum_table[checksum].ci_flags &
+ ZCHECKSUM_FLAG_NOPWRITE) &&
+ compress != ZIO_COMPRESS_OFF && zfs_nopwrite_enabled);
}
/*
- * Determine dedup setting. If we are in dmu_sync(), we won't
- * actually dedup now because that's all done in syncing context;
- * but we do want to use the dedup checkum. If the checksum is not
- * strong enough to ensure unique signatures, force dedup_verify.
+ * All objects in an encrypted objset are protected from modification
+ * via a MAC. Encrypted objects store their IV and salt in the last DVA
+ * in the bp, so we cannot use all copies. Encrypted objects are also
+ * not subject to nopwrite since writing the same data will still
+ * result in a new ciphertext. Only encrypted blocks can be dedup'd
+ * to avoid ambiguity in the dedup code since the DDT does not store
+ * object types.
*/
- dedup = (!ismd && dedup_checksum != ZIO_CHECKSUM_OFF);
- if (dedup) {
- checksum = dedup_checksum;
- if (!zio_checksum_table[checksum].ci_dedup)
- dedup_verify = 1;
- }
+ if (os->os_encrypted && (wp & WP_NOFILL) == 0) {
+ encrypt = B_TRUE;
- if (wp & WP_DMU_SYNC)
- dedup = 0;
+ if (DMU_OT_IS_ENCRYPTED(type)) {
+ copies = MIN(copies, SPA_DVAS_PER_BP - 1);
+ nopwrite = B_FALSE;
+ } else {
+ dedup = B_FALSE;
+ }
- if (wp & WP_NOFILL) {
- ASSERT(!ismd && level == 0);
- checksum = ZIO_CHECKSUM_OFF;
- compress = ZIO_COMPRESS_OFF;
- dedup = B_FALSE;
+ if (type == DMU_OT_DNODE || type == DMU_OT_OBJSET)
+ compress = ZIO_COMPRESS_EMPTY;
}
- zp->zp_checksum = checksum;
zp->zp_compress = compress;
+ zp->zp_checksum = checksum;
zp->zp_type = (wp & WP_SPILL) ? dn->dn_bonustype : type;
zp->zp_level = level;
- zp->zp_copies = MIN(copies + ismd, spa_max_replication(os->os_spa));
+ zp->zp_copies = MIN(copies, spa_max_replication(os->os_spa));
zp->zp_dedup = dedup;
zp->zp_dedup_verify = dedup && dedup_verify;
+ zp->zp_nopwrite = nopwrite;
+ zp->zp_encrypt = encrypt;
+ zp->zp_byteorder = ZFS_HOST_BYTEORDER;
+ bzero(zp->zp_salt, ZIO_DATA_SALT_LEN);
+ bzero(zp->zp_iv, ZIO_DATA_IV_LEN);
+ bzero(zp->zp_mac, ZIO_DATA_MAC_LEN);
+
+ ASSERT3U(zp->zp_compress, !=, ZIO_COMPRESS_INHERIT);
}
+/*
+ * This function is only called from zfs_holey_common() for zpl_llseek()
+ * in order to determine the location of holes. In order to accurately
+ * report holes all dirty data must be synced to disk. This causes extremely
+ * poor performance when seeking for holes in a dirty file. As a compromise,
+ * only provide hole data when the dnode is clean. When a dnode is dirty
+ * report the dnode as having no holes which is always a safe thing to do.
+ */
int
dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, uint64_t *off)
{
dnode_t *dn;
int i, err;
+ boolean_t clean = B_TRUE;
err = dnode_hold(os, object, FTAG, &dn);
if (err)
return (err);
+
/*
- * Sync any current changes before
- * we go trundling through the block pointers.
+ * Check if dnode is dirty
*/
- for (i = 0; i < TXG_SIZE; i++) {
- if (list_link_active(&dn->dn_dirty_link[i]))
- break;
+ if (dn->dn_dirtyctx != DN_UNDIRTIED) {
+ for (i = 0; i < TXG_SIZE; i++) {
+ if (!list_is_empty(&dn->dn_dirty_records[i])) {
+ clean = B_FALSE;
+ break;
+ }
+ }
}
- if (i != TXG_SIZE) {
+
+ /*
+ * If compatibility option is on, sync any current changes before
+ * we go trundling through the block pointers.
+ */
+ if (!clean && zfs_dmu_offset_next_sync) {
+ clean = B_TRUE;
dnode_rele(dn, FTAG);
txg_wait_synced(dmu_objset_pool(os), 0);
err = dnode_hold(os, object, FTAG, &dn);
return (err);
}
- err = dnode_next_offset(dn, (hole ? DNODE_FIND_HOLE : 0), off, 1, 1, 0);
+ if (clean)
+ err = dnode_next_offset(dn,
+ (hole ? DNODE_FIND_HOLE : 0), off, 1, 1, 0);
+ else
+ err = SET_ERROR(EBUSY);
+
dnode_rele(dn, FTAG);
return (err);
}
void
-dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi)
+__dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi)
{
- dnode_phys_t *dnp;
- int i;
-
- rw_enter(&dn->dn_struct_rwlock, RW_READER);
- mutex_enter(&dn->dn_mtx);
-
- dnp = dn->dn_phys;
+ dnode_phys_t *dnp = dn->dn_phys;
doi->doi_data_block_size = dn->dn_datablksz;
doi->doi_metadata_block_size = dn->dn_indblkshift ?
doi->doi_type = dn->dn_type;
doi->doi_bonus_type = dn->dn_bonustype;
doi->doi_bonus_size = dn->dn_bonuslen;
+ doi->doi_dnodesize = dn->dn_num_slots << DNODE_SHIFT;
doi->doi_indirection = dn->dn_nlevels;
doi->doi_checksum = dn->dn_checksum;
doi->doi_compress = dn->dn_compress;
+ doi->doi_nblkptr = dn->dn_nblkptr;
doi->doi_physical_blocks_512 = (DN_USED_BYTES(dnp) + 256) >> 9;
- doi->doi_max_offset = (dnp->dn_maxblkid + 1) * dn->dn_datablksz;
+ doi->doi_max_offset = (dn->dn_maxblkid + 1) * dn->dn_datablksz;
doi->doi_fill_count = 0;
- for (i = 0; i < dnp->dn_nblkptr; i++)
- doi->doi_fill_count += dnp->dn_blkptr[i].blk_fill;
+ for (int i = 0; i < dnp->dn_nblkptr; i++)
+ doi->doi_fill_count += BP_GET_FILL(&dnp->dn_blkptr[i]);
+}
+
+void
+dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi)
+{
+ rw_enter(&dn->dn_struct_rwlock, RW_READER);
+ mutex_enter(&dn->dn_mtx);
+
+ __dmu_object_info_from_dnode(dn, doi);
mutex_exit(&dn->dn_mtx);
rw_exit(&dn->dn_struct_rwlock);
dn = DB_DNODE(db);
*blksize = dn->dn_datablksz;
- /* add 1 for dnode space */
+ /* add in number of slots used for the dnode itself */
*nblk512 = ((DN_USED_BYTES(dn->dn_phys) + SPA_MINBLOCKSIZE/2) >>
- SPA_MINBLOCKSHIFT) + 1;
+ SPA_MINBLOCKSHIFT) + dn->dn_num_slots;
+ DB_DNODE_EXIT(db);
+}
+
+void
+dmu_object_dnsize_from_db(dmu_buf_t *db_fake, int *dnsize)
+{
+ dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake;
+ dnode_t *dn;
+
+ DB_DNODE_ENTER(db);
+ dn = DB_DNODE(db);
+ *dnsize = dn->dn_num_slots << DNODE_SHIFT;
DB_DNODE_EXIT(db);
}
void
dmu_init(void)
{
+ abd_init();
zfs_dbgmsg_init();
sa_cache_init();
xuio_stat_init();
dmu_objset_init();
dnode_init();
- dbuf_init();
zfetch_init();
dmu_tx_init();
- arc_init();
l2arc_init();
+ arc_init();
+ dbuf_init();
}
void
dmu_fini(void)
{
+ arc_fini(); /* arc depends on l2arc, so arc must go first */
l2arc_fini();
- arc_fini();
dmu_tx_fini();
zfetch_fini();
dbuf_fini();
xuio_stat_fini();
sa_cache_fini();
zfs_dbgmsg_fini();
+ abd_fini();
}
#if defined(_KERNEL) && defined(HAVE_SPL)
EXPORT_SYMBOL(dmu_bonus_hold);
EXPORT_SYMBOL(dmu_buf_hold_array_by_bonus);
EXPORT_SYMBOL(dmu_buf_rele_array);
+EXPORT_SYMBOL(dmu_prefetch);
EXPORT_SYMBOL(dmu_free_range);
+EXPORT_SYMBOL(dmu_free_long_range);
+EXPORT_SYMBOL(dmu_free_long_range_raw);
+EXPORT_SYMBOL(dmu_free_long_object);
+EXPORT_SYMBOL(dmu_free_long_object_raw);
EXPORT_SYMBOL(dmu_read);
+EXPORT_SYMBOL(dmu_read_by_dnode);
EXPORT_SYMBOL(dmu_write);
+EXPORT_SYMBOL(dmu_write_by_dnode);
+EXPORT_SYMBOL(dmu_prealloc);
EXPORT_SYMBOL(dmu_object_info);
EXPORT_SYMBOL(dmu_object_info_from_dnode);
EXPORT_SYMBOL(dmu_object_info_from_db);
EXPORT_SYMBOL(dmu_object_size_from_db);
+EXPORT_SYMBOL(dmu_object_dnsize_from_db);
+EXPORT_SYMBOL(dmu_object_set_nlevels);
EXPORT_SYMBOL(dmu_object_set_blocksize);
EXPORT_SYMBOL(dmu_object_set_checksum);
EXPORT_SYMBOL(dmu_object_set_compress);
+EXPORT_SYMBOL(dmu_write_policy);
+EXPORT_SYMBOL(dmu_sync);
EXPORT_SYMBOL(dmu_request_arcbuf);
EXPORT_SYMBOL(dmu_return_arcbuf);
-EXPORT_SYMBOL(dmu_assign_arcbuf);
+EXPORT_SYMBOL(dmu_assign_arcbuf_by_dnode);
+EXPORT_SYMBOL(dmu_assign_arcbuf_by_dbuf);
EXPORT_SYMBOL(dmu_buf_hold);
EXPORT_SYMBOL(dmu_ot);
+/* BEGIN CSTYLED */
module_param(zfs_mdcomp_disable, int, 0644);
MODULE_PARM_DESC(zfs_mdcomp_disable, "Disable meta data compression");
+
+module_param(zfs_nopwrite_enabled, int, 0644);
+MODULE_PARM_DESC(zfs_nopwrite_enabled, "Enable NOP writes");
+
+module_param(zfs_per_txg_dirty_frees_percent, ulong, 0644);
+MODULE_PARM_DESC(zfs_per_txg_dirty_frees_percent,
+ "percentage of dirtied blocks from frees in one TXG");
+
+module_param(zfs_dmu_offset_next_sync, int, 0644);
+MODULE_PARM_DESC(zfs_dmu_offset_next_sync,
+ "Enable forcing txg sync to find holes");
+
+/* END CSTYLED */
+
#endif
projects / mirror_zfs.git / blobdiff