(void) close(fd);
}
+static uint64_t num_large_blocks;
+
/*ARGSUSED*/
static int
dump_one_dir(const char *dsname, void *arg)
(void) printf("Could not open %s, error %d\n", dsname, error);
return (0);
}
+ if (dmu_objset_ds(os)->ds_large_blocks)
+ num_large_blocks++;
dump_dir(os);
dmu_objset_disown(os, FTAG);
fuid_table_destroy();
/*
* Block statistics.
*/
-#define PSIZE_HISTO_SIZE (SPA_MAXBLOCKSIZE / SPA_MINBLOCKSIZE + 1)
+#define PSIZE_HISTO_SIZE (SPA_OLD_MAXBLOCKSIZE / SPA_MINBLOCKSIZE + 2)
typedef struct zdb_blkstats {
uint64_t zb_asize;
uint64_t zb_lsize;
zb->zb_lsize += BP_GET_LSIZE(bp);
zb->zb_psize += BP_GET_PSIZE(bp);
zb->zb_count++;
- zb->zb_psize_histogram[BP_GET_PSIZE(bp) >> SPA_MINBLOCKSHIFT]++;
+
+ /*
+ * The histogram is only big enough to record blocks up to
+ * SPA_OLD_MAXBLOCKSIZE; larger blocks go into the last,
+ * "other", bucket.
+ */
+ int idx = BP_GET_PSIZE(bp) >> SPA_MINBLOCKSHIFT;
+ idx = MIN(idx, SPA_OLD_MAXBLOCKSIZE / SPA_MINBLOCKSIZE + 1);
+ zb->zb_psize_histogram[idx]++;
zb->zb_gangs += BP_COUNT_GANG(bp);
dump_metaslab_groups(spa);
if (dump_opt['d'] || dump_opt['i']) {
+ uint64_t refcount;
dump_dir(dp->dp_meta_objset);
if (dump_opt['d'] >= 3) {
dump_bpobj(&spa->spa_deferred_bpobj,
}
(void) dmu_objset_find(spa_name(spa), dump_one_dir,
NULL, DS_FIND_SNAPSHOTS | DS_FIND_CHILDREN);
+
+ (void) feature_get_refcount(spa,
+ &spa_feature_table[SPA_FEATURE_LARGE_BLOCKS], &refcount);
+ if (num_large_blocks != refcount) {
+ (void) printf("large_blocks feature refcount mismatch: "
+ "expected %lld != actual %lld\n",
+ (longlong_t)num_large_blocks,
+ (longlong_t)refcount);
+ rc = 2;
+ } else {
+ (void) printf("Verified large_blocks feature refcount "
+ "is correct (%llu)\n", (longlong_t)refcount);
+ }
}
- if (dump_opt['b'] || dump_opt['c'])
+ if (rc == 0 && (dump_opt['b'] || dump_opt['c']))
rc = dump_block_stats(spa);
if (rc == 0)
case HELP_ROLLBACK:
return (gettext("\trollback [-rRf] <snapshot>\n"));
case HELP_SEND:
- return (gettext("\tsend [-DnPpRrve] [-[iI] snapshot] "
+ return (gettext("\tsend [-DnPpRvLe] [-[iI] snapshot] "
"<snapshot>\n"
- "\tsend [-e] [-i snapshot|bookmark] "
+ "\tsend [-Le] [-i snapshot|bookmark] "
"<filesystem|volume|snapshot>\n"));
case HELP_SET:
return (gettext("\tset <property=value> "
boolean_t extraverbose = B_FALSE;
/* check options */
- while ((c = getopt(argc, argv, ":i:I:RDpvnPe")) != -1) {
+ while ((c = getopt(argc, argv, ":i:I:RDpvnPLe")) != -1) {
switch (c) {
case 'i':
if (fromname)
case 'n':
flags.dryrun = B_TRUE;
break;
+ case 'L':
+ flags.largeblock = B_TRUE;
+ break;
case 'e':
flags.embed_data = B_TRUE;
break;
if (zhp == NULL)
return (1);
+ if (flags.largeblock)
+ lzc_flags |= LZC_SEND_FLAG_LARGE_BLOCK;
if (flags.embed_data)
lzc_flags |= LZC_SEND_FLAG_EMBED_DATA;
FILE *send_stream = 0;
boolean_t do_byteswap = B_FALSE;
boolean_t do_cksum = B_TRUE;
-#define INITIAL_BUFLEN (1<<20)
static void
usage(void)
exit(1);
}
+static void *
+safe_malloc(size_t size)
+{
+ void *rv = malloc(size);
+ if (rv == NULL) {
+ (void) fprintf(stderr, "ERROR; failed to allocate %u bytes\n",
+ (unsigned)size);
+ abort();
+ }
+ return (rv);
+}
+
/*
* ssread - send stream read.
*
int
main(int argc, char *argv[])
{
- char *buf = malloc(INITIAL_BUFLEN);
+ char *buf = safe_malloc(SPA_MAXBLOCKSIZE);
uint64_t drr_record_count[DRR_NUMTYPES] = { 0 };
uint64_t total_records = 0;
dmu_replay_record_t thedrr;
nvlist_t *nv;
int sz = drr->drr_payloadlen;
- if (sz > INITIAL_BUFLEN) {
+ if (sz > SPA_MAXBLOCKSIZE) {
free(buf);
- buf = malloc(sz);
+ buf = safe_malloc(sz);
}
(void) ssread(buf, sz, &zc);
if (ferror(send_stream))
static int
ztest_random_blocksize(void)
{
- // Choose a block size >= the ashift.
- uint64_t block_shift =
- ztest_random(SPA_MAXBLOCKSHIFT - ztest_spa_get_ashift() + 1);
+ /*
+ * Choose a block size >= the ashift.
+ * If the SPA supports new MAXBLOCKSIZE, test up to 1MB blocks.
+ */
+ int maxbs = SPA_OLD_MAXBLOCKSHIFT;
+ if (spa_maxblocksize(ztest_spa) == SPA_MAXBLOCKSIZE)
+ maxbs = 20;
+ uint64_t block_shift = ztest_random(maxbs - ztest_spa_get_ashift() + 1);
return (1 << (SPA_MINBLOCKSHIFT + block_shift));
}
char *path0;
char *pathrand;
size_t fsize;
- int bshift = SPA_MAXBLOCKSHIFT + 2; /* don't scrog all labels */
+ int bshift = SPA_OLD_MAXBLOCKSHIFT + 2; /* don't scrog all labels */
int iters = 1000;
int maxfaults;
int mirror_save;
/* show progress (ie. -v) */
boolean_t progress;
+ /* large blocks (>128K) are permitted */
+ boolean_t largeblock;
+
/* WRITE_EMBEDDED records of type DATA are permitted */
boolean_t embed_data;
} sendflags_t;
int lzc_get_holds(const char *, nvlist_t **);
enum lzc_send_flags {
- LZC_SEND_FLAG_EMBED_DATA = 1 << 0
+ LZC_SEND_FLAG_EMBED_DATA = 1 << 0,
+ LZC_SEND_FLAG_LARGE_BLOCK = 1 << 1
};
int lzc_send(const char *, const char *, int, enum lzc_send_flags);
* The maximum number of bytes that can be accessed as part of one
* operation, including metadata.
*/
-#define DMU_MAX_ACCESS (10<<20) /* 10MB */
+#define DMU_MAX_ACCESS (64 * 1024 * 1024) /* 64MB */
#define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */
#define DMU_USERUSED_OBJECT (-1ULL)
void xuio_stat_wbuf_nocopy(void);
extern int zfs_prefetch_disable;
+extern int zfs_max_recordsize;
/*
* Asynchronously try to read in the data.
zfs_cache_type_t os_secondary_cache;
zfs_sync_type_t os_sync;
zfs_redundant_metadata_type_t os_redundant_metadata;
+ int os_recordsize;
/* no lock needed: */
struct dmu_tx *os_synctx; /* XXX sketchy */
struct drr_begin;
struct avl_tree;
-int dmu_send(const char *tosnap, const char *fromsnap, boolean_t embedok,
+int dmu_send(const char *tosnap, const char *fromsnap,
+ boolean_t embedok, boolean_t large_block_ok,
int outfd, struct vnode *vp, offset_t *off);
int dmu_send_estimate(struct dsl_dataset *ds, struct dsl_dataset *fromds,
uint64_t *sizep);
int dmu_send_obj(const char *pool, uint64_t tosnap, uint64_t fromsnap,
- boolean_t embedok, int outfd, vnode_t *vp, offset_t *off);
+ boolean_t embedok, boolean_t large_block_ok,
+ int outfd, struct vnode *vp, offset_t *off);
typedef struct dmu_recv_cookie {
struct dsl_dataset *drc_ds;
*/
#define DS_FIELD_BOOKMARK_NAMES "com.delphix:bookmarks"
+/*
+ * This field is present (with value=0) if this dataset may contain large
+ * blocks (>128KB). If it is present, then this dataset
+ * is counted in the refcount of the SPA_FEATURE_LARGE_BLOCKS feature.
+ */
+#define DS_FIELD_LARGE_BLOCKS "org.open-zfs:large_blocks"
+
/*
* DS_FLAG_CI_DATASET is set if the dataset contains a file system whose
* name lookups should be performed case-insensitively.
/* only used in syncing context, only valid for non-snapshots: */
struct dsl_dataset *ds_prev;
uint64_t ds_bookmarks; /* DMU_OTN_ZAP_METADATA */
+ boolean_t ds_large_blocks;
+ boolean_t ds_need_large_blocks;
/* has internal locking: */
dsl_deadlist_t ds_deadlist;
int dsl_dataset_space_wouldfree(dsl_dataset_t *firstsnap, dsl_dataset_t *last,
uint64_t *usedp, uint64_t *compp, uint64_t *uncompp);
boolean_t dsl_dataset_is_dirty(dsl_dataset_t *ds);
+int dsl_dataset_activate_large_blocks(const char *dsname);
+void dsl_dataset_activate_large_blocks_sync_impl(uint64_t dsobj, dmu_tx_t *tx);
int dsl_dsobj_to_dsname(char *pname, uint64_t obj, char *buf);
ZPOOL_PROP_FREEING,
ZPOOL_PROP_FRAGMENTATION,
ZPOOL_PROP_LEAKED,
+ ZPOOL_PROP_MAXBLOCKSIZE,
ZPOOL_PROP_TNAME,
ZPOOL_NUM_PROPS
} zpool_prop_t;
_NOTE(CONSTCOND) } while (0)
/*
- * We currently support nine block sizes, from 512 bytes to 128K.
- * We could go higher, but the benefits are near-zero and the cost
- * of COWing a giant block to modify one byte would become excessive.
+ * We currently support block sizes from 512 bytes to 16MB.
+ * The benefits of larger blocks, and thus larger IO, need to be weighed
+ * against the cost of COWing a giant block to modify one byte, and the
+ * large latency of reading or writing a large block.
+ *
+ * Note that although blocks up to 16MB are supported, the recordsize
+ * property can not be set larger than zfs_max_recordsize (default 1MB).
+ * See the comment near zfs_max_recordsize in dsl_dataset.c for details.
+ *
+ * Note that although the LSIZE field of the blkptr_t can store sizes up
+ * to 32MB, the dnode's dn_datablkszsec can only store sizes up to
+ * 32MB - 512 bytes. Therefore, we limit SPA_MAXBLOCKSIZE to 16MB.
*/
#define SPA_MINBLOCKSHIFT 9
-#define SPA_MAXBLOCKSHIFT 17
+#define SPA_OLD_MAXBLOCKSHIFT 17
+#define SPA_MAXBLOCKSHIFT 24
#define SPA_MINBLOCKSIZE (1ULL << SPA_MINBLOCKSHIFT)
+#define SPA_OLD_MAXBLOCKSIZE (1ULL << SPA_OLD_MAXBLOCKSHIFT)
#define SPA_MAXBLOCKSIZE (1ULL << SPA_MAXBLOCKSHIFT)
-#define SPA_BLOCKSIZES (SPA_MAXBLOCKSHIFT - SPA_MINBLOCKSHIFT + 1)
-
/*
* Size of block to hold the configuration data (a packed nvlist)
*/
extern boolean_t spa_is_root(spa_t *spa);
extern boolean_t spa_writeable(spa_t *spa);
extern boolean_t spa_has_pending_synctask(spa_t *spa);
+extern int spa_maxblocksize(spa_t *spa);
extern void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp);
extern int spa_mode(spa_t *spa);
#define MZAP_ENT_LEN 64
#define MZAP_NAME_LEN (MZAP_ENT_LEN - 8 - 4 - 2)
-#define MZAP_MAX_BLKSHIFT SPA_MAXBLOCKSHIFT
-#define MZAP_MAX_BLKSZ (1 << MZAP_MAX_BLKSHIFT)
+#define MZAP_MAX_BLKSZ SPA_OLD_MAXBLOCKSIZE
#define ZAP_NEED_CD (-1U)
/* flags #3 - #15 are reserved for incompatible closed-source implementations */
#define DMU_BACKUP_FEATURE_EMBED_DATA (1<<16)
#define DMU_BACKUP_FEATURE_EMBED_DATA_LZ4 (1<<17)
+/* flag #18 is reserved for a Delphix feature */
+#define DMU_BACKUP_FEATURE_LARGE_BLOCKS (1<<19)
/*
* Mask of all supported backup features
*/
#define DMU_BACKUP_FEATURE_MASK (DMU_BACKUP_FEATURE_DEDUP | \
DMU_BACKUP_FEATURE_DEDUPPROPS | DMU_BACKUP_FEATURE_SA_SPILL | \
- DMU_BACKUP_FEATURE_EMBED_DATA | DMU_BACKUP_FEATURE_EMBED_DATA_LZ4)
+ DMU_BACKUP_FEATURE_EMBED_DATA | DMU_BACKUP_FEATURE_EMBED_DATA_LZ4 | \
+ DMU_BACKUP_FEATURE_LARGE_BLOCKS)
/* Are all features in the given flag word currently supported? */
#define DMU_STREAM_SUPPORTED(x) (!((x) & ~DMU_BACKUP_FEATURE_MASK))
#ifdef _KERNEL
#define DXATTR_MAX_ENTRY_SIZE (32768)
-#define DXATTR_MAX_SA_SIZE (SPA_MAXBLOCKSIZE >> 1)
+#define DXATTR_MAX_SA_SIZE (SPA_OLD_MAXBLOCKSIZE >> 1)
int zfs_sa_readlink(struct znode *, uio_t *);
void zfs_sa_symlink(struct znode *, char *link, int len, dmu_tx_t *);
#define ZFS_SHARES_DIR "SHARES"
#define ZFS_SA_ATTRS "SA_ATTRS"
-#define ZFS_MAX_BLOCKSIZE (SPA_MAXBLOCKSIZE)
-
/*
* Path component length
*
} zil_chain_t;
#define ZIL_MIN_BLKSZ 4096ULL
-#define ZIL_MAX_BLKSZ SPA_MAXBLOCKSIZE
/*
* The words of a log block checksum.
avl_node_t zn_node;
} zil_bp_node_t;
-#define ZIL_MAX_LOG_DATA (SPA_MAXBLOCKSIZE - sizeof (zil_chain_t) - \
+#define ZIL_MAX_LOG_DATA (SPA_OLD_MAXBLOCKSIZE - sizeof (zil_chain_t) - \
sizeof (lr_write_t))
#ifdef __cplusplus
SPA_FEATURE_EMBEDDED_DATA,
SPA_FEATURE_BOOKMARKS,
SPA_FEATURE_FS_SS_LIMIT,
+ SPA_FEATURE_LARGE_BLOCKS,
SPA_FEATURES
} spa_feature_t;
break;
}
- case ZFS_PROP_RECORDSIZE:
case ZFS_PROP_VOLBLOCKSIZE:
- /* must be power of two within SPA_{MIN,MAX}BLOCKSIZE */
+ case ZFS_PROP_RECORDSIZE:
+ {
+ int maxbs = SPA_MAXBLOCKSIZE;
+ if (zhp != NULL) {
+ maxbs = zpool_get_prop_int(zhp->zpool_hdl,
+ ZPOOL_PROP_MAXBLOCKSIZE, NULL);
+ }
+ /*
+ * The value must be a power of two between
+ * SPA_MINBLOCKSIZE and maxbs.
+ */
if (intval < SPA_MINBLOCKSIZE ||
- intval > SPA_MAXBLOCKSIZE || !ISP2(intval)) {
+ intval > maxbs || !ISP2(intval)) {
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
- "'%s' must be power of 2 from %u "
- "to %uk"), propname,
- (uint_t)SPA_MINBLOCKSIZE,
- (uint_t)SPA_MAXBLOCKSIZE >> 10);
+ "'%s' must be power of 2 from 512B "
+ "to %uKB"), propname, maxbs >> 10);
(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
goto error;
}
break;
-
+ }
case ZFS_PROP_MLSLABEL:
{
#ifdef HAVE_MLSLABEL
break;
case ERANGE:
- if (prop == ZFS_PROP_COMPRESSION) {
+ if (prop == ZFS_PROP_COMPRESSION ||
+ prop == ZFS_PROP_RECORDSIZE) {
(void) zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"property setting is not allowed on "
"bootable datasets"));
case EDOM:
zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
"volume block size must be power of 2 from "
- "%u to %uk"),
- (uint_t)SPA_MINBLOCKSIZE,
- (uint_t)SPA_MAXBLOCKSIZE >> 10);
+ "512B to %uKB"), zfs_max_recordsize >> 10);
return (zfs_error(hdl, EZFS_BADPROP, errbuf));
cksummer(void *arg)
{
dedup_arg_t *dda = arg;
- char *buf = malloc(1<<20);
+ char *buf = zfs_alloc(dda->dedup_hdl, SPA_MAXBLOCKSIZE);
dmu_replay_record_t thedrr;
dmu_replay_record_t *drr = &thedrr;
struct drr_begin *drrb = &thedrr.drr_u.drr_begin;
DMU_COMPOUNDSTREAM && drr->drr_payloadlen != 0) {
int sz = drr->drr_payloadlen;
- if (sz > 1<<20) {
- free(buf);
- buf = malloc(sz);
+ if (sz > SPA_MAXBLOCKSIZE) {
+ buf = zfs_realloc(dda->dedup_hdl, buf,
+ SPA_MAXBLOCKSIZE, sz);
}
(void) ssread(buf, sz, ofp);
if (ferror(stdin))
char prevsnap[ZFS_MAXNAMELEN];
uint64_t prevsnap_obj;
boolean_t seenfrom, seento, replicate, doall, fromorigin;
- boolean_t verbose, dryrun, parsable, progress, embed_data;
+ boolean_t verbose, dryrun, parsable, progress, embed_data, large_block;
int outfd;
boolean_t err;
nvlist_t *fss;
}
enum lzc_send_flags flags = 0;
+ if (sdd->large_block)
+ flags |= LZC_SEND_FLAG_LARGE_BLOCK;
if (sdd->embed_data)
flags |= LZC_SEND_FLAG_EMBED_DATA;
sdd.parsable = flags->parsable;
sdd.progress = flags->progress;
sdd.dryrun = flags->dryrun;
+ sdd.large_block = flags->largeblock;
sdd.embed_data = flags->embed_data;
sdd.filter_cb = filter_func;
sdd.filter_cb_arg = cb_arg;
recv_skip(libzfs_handle_t *hdl, int fd, boolean_t byteswap)
{
dmu_replay_record_t *drr;
- void *buf = malloc(1<<20);
+ void *buf = zfs_alloc(hdl, SPA_MAXBLOCKSIZE);
char errbuf[1024];
(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
*
* "fd" is the file descriptor to write the send stream to.
*
+ * If "flags" contains LZC_SEND_FLAG_LARGE_BLOCK, the stream is permitted
+ * to contain DRR_WRITE records with drr_length > 128K, and DRR_OBJECT
+ * records with drr_blksz > 128K.
+ *
* If "flags" contains LZC_SEND_FLAG_EMBED_DATA, the stream is permitted
* to contain DRR_WRITE_EMBEDDED records with drr_etype==BP_EMBEDDED_TYPE_DATA,
* which the receiving system must support (as indicated by support
fnvlist_add_int32(args, "fd", fd);
if (from != NULL)
fnvlist_add_string(args, "fromsnap", from);
+ if (flags & LZC_SEND_FLAG_LARGE_BLOCK)
+ fnvlist_add_boolean(args, "largeblockok");
if (flags & LZC_SEND_FLAG_EMBED_DATA)
fnvlist_add_boolean(args, "embedok");
err = lzc_ioctl(ZFS_IOC_SEND_NEW, snapname, args, NULL);
Default value: \fB32,768\fR.
.RE
+.sp
+.ne 2
+.na
+\fBzfs_max_recordsize\fR (int)
+.ad
+.RS 12n
+We currently support block sizes from 512 bytes to 16MB. The benefits of
+larger blocks, and thus larger IO, need to be weighed against the cost of
+COWing a giant block to modify one byte. Additionally, very large blocks
+can have an impact on i/o latency, and also potentially on the memory
+allocator. Therefore, we do not allow the recordsize to be set larger than
+zfs_max_recordsize (default 1MB). Larger blocks can be created by changing
+this tunable, and pools with larger blocks can always be imported and used,
+regardless of this setting.
+.sp
+Default value: \fB1,048,576\fR.
+.RE
+
.sp
.ne 2
.na
.RE
+.sp
+.ne 2
+.na
+\fB\fBlarge_blocks\fR\fR
+.ad
+.RS 4n
+.TS
+l l .
+GUID org.open-zfs:large_block
+READ\-ONLY COMPATIBLE no
+DEPENDENCIES extensible_dataset
+.TE
+
+The \fBlarge_block\fR feature allows the record size on a dataset to be
+set larger than 128KB.
+
+This feature becomes \fBactive\fR once a \fBrecordsize\fR property has been
+set larger than 128KB, and will return to being \fBenabled\fR once all
+filesystems that have ever had their recordsize larger than 128KB are destroyed.
+.RE
+
.SH "SEE ALSO"
\fBzpool\fR(8)
.LP
.nf
-\fBzfs\fR \fBsend\fR [\fB-DnPpRve\fR] [\fB-\fR[\fBiI\fR] \fIsnapshot\fR] \fIsnapshot\fR
+\fBzfs\fR \fBsend\fR [\fB-DnPpRveL\fR] [\fB-\fR[\fBiI\fR] \fIsnapshot\fR] \fIsnapshot\fR
.fi
.LP
.nf
-\fBzfs\fR \fBsend\fR [\fB-e\fR] [\fB-i \fIsnapshot\fR|\fIbookmark\fR]\fR \fIfilesystem\fR|\fIvolume\fR|\fIsnapshot\fR
+\fBzfs\fR \fBsend\fR [\fB-eL\fR] [\fB-i \fIsnapshot\fR|\fIbookmark\fR]\fR \fIfilesystem\fR|\fIvolume\fR|\fIsnapshot\fR
.fi
.LP
.sp
.ne 2
.na
-\fBzfs send\fR [\fB-DnPpRve\fR] [\fB-\fR[\fBiI\fR] \fIsnapshot\fR] \fIsnapshot\fR
+\fBzfs send\fR [\fB-DnPpRveL\fR] [\fB-\fR[\fBiI\fR] \fIsnapshot\fR] \fIsnapshot\fR
.ad
.sp .6
.RS 4n
Generate a deduplicated stream. Blocks which would have been sent multiple times in the send stream will only be sent once. The receiving system must also support this feature to receive a deduplicated stream. This flag can be used regardless of the dataset's dedup property, but performance will be much better if the filesystem uses a dedup-capable checksum (eg. sha256).
.RE
+.sp
+.ne 2
+.mk
+.na
+\fB\fB-L\fR\fR
+.ad
+.sp .6
+.RS 4n
+Generate a stream which may contain blocks larger than 128KB. This flag
+has no effect if the \fBlarge_blocks\fR pool feature is disabled, or if
+the \fRrecordsize\fR property of this filesystem has never been set above
+128KB. The receiving system must have the \fBlarge_blocks\fR pool feature
+enabled as well. See \fBzpool-features\fR(5) for details on ZFS feature
+flags and the \fBlarge_blocks\fR feature.
+.RE
+
.sp
.ne 2
.mk
.sp
.ne 2
.na
-\fBzfs send\fR [\fB-e\fR] [\fB-i\fR \fIsnapshot\fR|\fIbookmark\fR] \fIfilesystem\fR|\fIvolume\fR|\fIsnapshot\fR
+\fBzfs send\fR [\fB-eL\fR] [\fB-i\fR \fIsnapshot\fR|\fIbookmark\fR] \fIfilesystem\fR|\fIvolume\fR|\fIsnapshot\fR
.ad
.sp .6
.RS 4n
or the origin's origin, etc.
.RE
+.sp
+.ne 2
+.mk
+.na
+\fB\fB-L\fR\fR
+.ad
+.sp .6
+.RS 4n
+Generate a stream which may contain blocks larger than 128KB. This flag
+has no effect if the \fBlarge_blocks\fR pool feature is disabled, or if
+the \fRrecordsize\fR property of this filesystem has never been set above
+128KB. The receiving system must have the \fBlarge_blocks\fR pool feature
+enabled as well. See \fBzpool-features\fR(5) for details on ZFS feature
+flags and the \fBlarge_blocks\fR feature.
+.RE
+
.sp
.ne 2
.mk
Discard the first element of the sent snapshot's file system name, using the remaining elements to determine the name of the target file system for the new snapshot as described in the paragraph above.
.RE
-
.sp
.ne 2
.na
/* inherit number properties */
zprop_register_number(ZFS_PROP_RECORDSIZE, "recordsize",
- SPA_MAXBLOCKSIZE, PROP_INHERIT,
- ZFS_TYPE_FILESYSTEM, "512 to 128k, power of 2", "RECSIZE");
+ SPA_OLD_MAXBLOCKSIZE, PROP_INHERIT,
+ ZFS_TYPE_FILESYSTEM, "512 to 1M, power of 2", "RECSIZE");
/* hidden properties */
zprop_register_hidden(ZFS_PROP_CREATETXG, "createtxg", PROP_TYPE_NUMBER,
/* hidden properties */
zprop_register_hidden(ZPOOL_PROP_NAME, "name", PROP_TYPE_STRING,
PROP_READONLY, ZFS_TYPE_POOL, "NAME");
+ zprop_register_hidden(ZPOOL_PROP_MAXBLOCKSIZE, "maxblocksize",
+ PROP_TYPE_NUMBER, PROP_READONLY, ZFS_TYPE_POOL, "MAXBLOCKSIZE");
zprop_register_hidden(ZPOOL_PROP_TNAME, "tname", PROP_TYPE_STRING,
PROP_ONETIME, ZFS_TYPE_POOL, "TNAME");
}
arc_buf_hdr_t *hdr;
arc_buf_t *buf;
- VERIFY3U(size, <=, SPA_MAXBLOCKSIZE);
+ VERIFY3U(size, <=, spa_maxblocksize(spa));
hdr = kmem_cache_alloc(hdr_cache, KM_PUSHPAGE);
ASSERT(BUF_EMPTY(hdr));
hdr->b_size = size;
* Gracefully handle a damaged logical block size as a
* checksum error by passing a dummy zio to the done callback.
*/
- if (size > SPA_MAXBLOCKSIZE) {
+ if (size > spa_maxblocksize(spa)) {
if (done) {
rzio = zio_null(pio, spa, NULL,
NULL, NULL, zio_flags);
if (!spa_feature_is_active(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
ASSERT0(dp->dp_empty_bpobj);
dp->dp_empty_bpobj =
- bpobj_alloc(os, SPA_MAXBLOCKSIZE, tx);
+ bpobj_alloc(os, SPA_OLD_MAXBLOCKSIZE, tx);
VERIFY(zap_add(os,
DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_EMPTY_BPOBJ, sizeof (uint64_t), 1,
dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
if (bpo->bpo_phys->bpo_subobjs == 0) {
bpo->bpo_phys->bpo_subobjs = dmu_object_alloc(bpo->bpo_os,
- DMU_OT_BPOBJ_SUBOBJ, SPA_MAXBLOCKSIZE, DMU_OT_NONE, 0, tx);
+ DMU_OT_BPOBJ_SUBOBJ, SPA_OLD_MAXBLOCKSIZE,
+ DMU_OT_NONE, 0, tx);
}
ASSERT0(dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, &doi));
bptree_phys_t *bt;
obj = dmu_object_alloc(os, DMU_OTN_UINT64_METADATA,
- SPA_MAXBLOCKSIZE, DMU_OTN_UINT64_METADATA,
+ SPA_OLD_MAXBLOCKSIZE, DMU_OTN_UINT64_METADATA,
sizeof (bptree_phys_t), tx);
/*
return (SET_ERROR(ENOTSUP));
if (blksz == 0)
blksz = SPA_MINBLOCKSIZE;
- if (blksz > SPA_MAXBLOCKSIZE)
- blksz = SPA_MAXBLOCKSIZE;
- else
- blksz = P2ROUNDUP(blksz, SPA_MINBLOCKSIZE);
+ ASSERT3U(blksz, <=, spa_maxblocksize(dmu_objset_spa(db->db_objset)));
+ blksz = P2ROUNDUP(blksz, SPA_MINBLOCKSIZE);
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
zil_set_logbias(os->os_zil, newval);
}
+static void
+recordsize_changed_cb(void *arg, uint64_t newval)
+{
+ objset_t *os = arg;
+
+ os->os_recordsize = newval;
+}
+
void
dmu_objset_byteswap(void *buf, size_t size)
{
ZFS_PROP_REDUNDANT_METADATA),
redundant_metadata_changed_cb, os);
}
+ if (err == 0) {
+ err = dsl_prop_register(ds,
+ zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
+ recordsize_changed_cb, os);
+ }
}
if (err != 0) {
VERIFY(arc_buf_remove_ref(os->os_phys_buf,
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_REDUNDANT_METADATA),
redundant_metadata_changed_cb, os));
+ VERIFY0(dsl_prop_unregister(ds,
+ zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
+ recordsize_changed_cb, os));
}
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE),
drrw->drr_offset = offset;
drrw->drr_length = blksz;
drrw->drr_toguid = dsp->dsa_toguid;
- if (BP_IS_EMBEDDED(bp)) {
+ if (bp == NULL || BP_IS_EMBEDDED(bp)) {
/*
- * There's no pre-computed checksum of embedded BP's, so
- * (like fletcher4-checkummed blocks) userland will have
- * to compute a dedup-capable checksum itself.
+ * There's no pre-computed checksum for partial-block
+ * writes or embedded BP's, so (like
+ * fletcher4-checkummed blocks) userland will have to
+ * compute a dedup-capable checksum itself.
*/
drrw->drr_checksumtype = ZIO_CHECKSUM_OFF;
} else {
drro->drr_compress = dnp->dn_compress;
drro->drr_toguid = dsp->dsa_toguid;
+ if (!(dsp->dsa_featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) &&
+ drro->drr_blksz > SPA_OLD_MAXBLOCKSIZE)
+ drro->drr_blksz = SPA_OLD_MAXBLOCKSIZE;
+
if (dump_bytes(dsp, dsp->dsa_drr, sizeof (dmu_replay_record_t)) != 0)
return (SET_ERROR(EINTR));
zb->zb_blkid * blksz, blksz, bp);
} else { /* it's a level-0 block of a regular object */
uint32_t aflags = ARC_WAIT;
+ uint64_t offset;
arc_buf_t *abuf;
int blksz = BP_GET_LSIZE(bp);
}
}
- err = dump_write(dsp, type, zb->zb_object, zb->zb_blkid * blksz,
- blksz, bp, abuf->b_data);
+ offset = zb->zb_blkid * blksz;
+
+ if (!(dsp->dsa_featureflags &
+ DMU_BACKUP_FEATURE_LARGE_BLOCKS) &&
+ blksz > SPA_OLD_MAXBLOCKSIZE) {
+ char *buf = abuf->b_data;
+ while (blksz > 0 && err == 0) {
+ int n = MIN(blksz, SPA_OLD_MAXBLOCKSIZE);
+ err = dump_write(dsp, type, zb->zb_object,
+ offset, n, NULL, buf);
+ offset += n;
+ buf += n;
+ blksz -= n;
+ }
+ } else {
+ err = dump_write(dsp, type, zb->zb_object,
+ offset, blksz, bp, abuf->b_data);
+ }
(void) arc_buf_remove_ref(abuf, &abuf);
}
static int
dmu_send_impl(void *tag, dsl_pool_t *dp, dsl_dataset_t *ds,
zfs_bookmark_phys_t *fromzb, boolean_t is_clone, boolean_t embedok,
- int outfd, vnode_t *vp, offset_t *off)
+ boolean_t large_block_ok, int outfd, vnode_t *vp, offset_t *off)
{
objset_t *os;
dmu_replay_record_t *drr;
}
#endif
+ if (large_block_ok && ds->ds_large_blocks)
+ featureflags |= DMU_BACKUP_FEATURE_LARGE_BLOCKS;
if (embedok &&
spa_feature_is_active(dp->dp_spa, SPA_FEATURE_EMBEDDED_DATA)) {
featureflags |= DMU_BACKUP_FEATURE_EMBED_DATA;
int
dmu_send_obj(const char *pool, uint64_t tosnap, uint64_t fromsnap,
- boolean_t embedok, int outfd, vnode_t *vp, offset_t *off)
+ boolean_t embedok, boolean_t large_block_ok,
+ int outfd, vnode_t *vp, offset_t *off)
{
dsl_pool_t *dp;
dsl_dataset_t *ds;
zb.zbm_guid = dsl_dataset_phys(fromds)->ds_guid;
is_clone = (fromds->ds_dir != ds->ds_dir);
dsl_dataset_rele(fromds, FTAG);
- err = dmu_send_impl(FTAG, dp, ds, &zb, is_clone, embedok,
- outfd, vp, off);
+ err = dmu_send_impl(FTAG, dp, ds, &zb, is_clone,
+ embedok, large_block_ok, outfd, vp, off);
} else {
- err = dmu_send_impl(FTAG, dp, ds, NULL, B_FALSE, embedok,
- outfd, vp, off);
+ err = dmu_send_impl(FTAG, dp, ds, NULL, B_FALSE,
+ embedok, large_block_ok, outfd, vp, off);
}
dsl_dataset_rele(ds, FTAG);
return (err);
}
int
-dmu_send(const char *tosnap, const char *fromsnap, boolean_t embedok,
+dmu_send(const char *tosnap, const char *fromsnap,
+ boolean_t embedok, boolean_t large_block_ok,
int outfd, vnode_t *vp, offset_t *off)
{
dsl_pool_t *dp;
dsl_pool_rele(dp, FTAG);
return (err);
}
- err = dmu_send_impl(FTAG, dp, ds, &zb, is_clone, embedok,
- outfd, vp, off);
+ err = dmu_send_impl(FTAG, dp, ds, &zb, is_clone,
+ embedok, large_block_ok, outfd, vp, off);
} else {
- err = dmu_send_impl(FTAG, dp, ds, NULL, B_FALSE, embedok,
- outfd, vp, off);
+ err = dmu_send_impl(FTAG, dp, ds, NULL, B_FALSE,
+ embedok, large_block_ok, outfd, vp, off);
}
if (owned)
dsl_dataset_disown(ds, FTAG);
!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LZ4_COMPRESS))
return (SET_ERROR(ENOTSUP));
+ /*
+ * The receiving code doesn't know how to translate large blocks
+ * to smaller ones, so the pool must have the LARGE_BLOCKS
+ * feature enabled if the stream has LARGE_BLOCKS.
+ */
+ if ((featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) &&
+ !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LARGE_BLOCKS))
+ return (SET_ERROR(ENOTSUP));
+
error = dsl_dataset_hold(dp, tofs, FTAG, &ds);
if (error == 0) {
/* target fs already exists; recv into temp clone */
}
VERIFY0(dsl_dataset_own_obj(dp, dsobj, dmu_recv_tag, &newds));
+ if ((DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
+ DMU_BACKUP_FEATURE_LARGE_BLOCKS) &&
+ !newds->ds_large_blocks) {
+ dsl_dataset_activate_large_blocks_sync_impl(dsobj, tx);
+ newds->ds_large_blocks = B_TRUE;
+ }
+
dmu_buf_will_dirty(newds->ds_dbuf, tx);
dsl_dataset_phys(newds)->ds_flags |= DS_FLAG_INCONSISTENT;
/* some things will require 8-byte alignment, so everything must */
ASSERT0(len % 8);
+ ASSERT3U(len, <=, ra->bufsize);
while (done < len) {
ssize_t resid;
drro->drr_compress >= ZIO_COMPRESS_FUNCTIONS ||
P2PHASE(drro->drr_blksz, SPA_MINBLOCKSIZE) ||
drro->drr_blksz < SPA_MINBLOCKSIZE ||
- drro->drr_blksz > SPA_MAXBLOCKSIZE ||
+ drro->drr_blksz > spa_maxblocksize(dmu_objset_spa(os)) ||
drro->drr_bonuslen > DN_MAX_BONUSLEN) {
return (SET_ERROR(EINVAL));
}
int err;
if (drrs->drr_length < SPA_MINBLOCKSIZE ||
- drrs->drr_length > SPA_MAXBLOCKSIZE)
+ drrs->drr_length > spa_maxblocksize(dmu_objset_spa(os)))
return (SET_ERROR(EINVAL));
data = restore_read(ra, drrs->drr_length, NULL);
ra.cksum = drc->drc_cksum;
ra.vp = vp;
ra.voff = *voffp;
- ra.bufsize = 1<<20;
+ ra.bufsize = SPA_MAXBLOCKSIZE;
ra.buf = vmem_alloc(ra.bufsize, KM_SLEEP);
/* these were verified in dmu_recv_begin */
return;
min_bs = SPA_MINBLOCKSHIFT;
- max_bs = SPA_MAXBLOCKSHIFT;
+ max_bs = highbit64(txh->txh_tx->tx_objset->os_recordsize) - 1;
min_ibs = DN_MIN_INDBLKSHIFT;
max_ibs = DN_MAX_INDBLKSHIFT;
*/
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));
}
/*
bp = &dn->dn_phys->dn_blkptr[0];
if (dsl_dataset_block_freeable(dn->dn_objset->os_dsl_dataset,
bp, bp->blk_birth))
- txh->txh_space_tooverwrite += SPA_MAXBLOCKSIZE;
+ txh->txh_space_tooverwrite += MZAP_MAX_BLKSZ;
else
- txh->txh_space_towrite += SPA_MAXBLOCKSIZE;
+ txh->txh_space_towrite += MZAP_MAX_BLKSZ;
if (!BP_IS_HOLE(bp))
- txh->txh_space_tounref += SPA_MAXBLOCKSIZE;
+ txh->txh_space_tounref += MZAP_MAX_BLKSZ;
return;
}
/* If blkptr doesn't exist then add space to towrite */
if (!(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) {
- txh->txh_space_towrite += SPA_MAXBLOCKSIZE;
+ txh->txh_space_towrite += SPA_OLD_MAXBLOCKSIZE;
} else {
blkptr_t *bp;
bp = &dn->dn_phys->dn_spill;
if (dsl_dataset_block_freeable(dn->dn_objset->os_dsl_dataset,
bp, bp->blk_birth))
- txh->txh_space_tooverwrite += SPA_MAXBLOCKSIZE;
+ txh->txh_space_tooverwrite += SPA_OLD_MAXBLOCKSIZE;
else
- txh->txh_space_towrite += SPA_MAXBLOCKSIZE;
+ txh->txh_space_towrite += SPA_OLD_MAXBLOCKSIZE;
if (!BP_IS_HOLE(bp))
- txh->txh_space_tounref += SPA_MAXBLOCKSIZE;
+ txh->txh_space_tounref += SPA_OLD_MAXBLOCKSIZE;
}
}
{
int i;
+ ASSERT3U(blocksize, <=,
+ spa_maxblocksize(dmu_objset_spa(dn->dn_objset)));
if (blocksize == 0)
blocksize = 1 << zfs_default_bs;
- else if (blocksize > SPA_MAXBLOCKSIZE)
- blocksize = SPA_MAXBLOCKSIZE;
else
blocksize = P2ROUNDUP(blocksize, SPA_MINBLOCKSIZE);
int nblkptr;
ASSERT3U(blocksize, >=, SPA_MINBLOCKSIZE);
- ASSERT3U(blocksize, <=, SPA_MAXBLOCKSIZE);
+ ASSERT3U(blocksize, <=,
+ spa_maxblocksize(dmu_objset_spa(dn->dn_objset)));
ASSERT0(blocksize % SPA_MINBLOCKSIZE);
ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT || dmu_tx_private_ok(tx));
ASSERT(tx->tx_txg != 0);
dmu_buf_impl_t *db;
int err;
+ ASSERT3U(size, <=, spa_maxblocksize(dmu_objset_spa(dn->dn_objset)));
if (size == 0)
size = SPA_MINBLOCKSIZE;
- if (size > SPA_MAXBLOCKSIZE)
- size = SPA_MAXBLOCKSIZE;
else
size = P2ROUNDUP(size, SPA_MINBLOCKSIZE);
#include <sys/dsl_userhold.h>
#include <sys/dsl_bookmark.h>
+/*
+ * The SPA supports block sizes up to 16MB. However, very large blocks
+ * can have an impact on i/o latency (e.g. tying up a spinning disk for
+ * ~300ms), and also potentially on the memory allocator. Therefore,
+ * we do not allow the recordsize to be set larger than zfs_max_recordsize
+ * (default 1MB). Larger blocks can be created by changing this tunable,
+ * and pools with larger blocks can always be imported and used, regardless
+ * of this setting.
+ */
+int zfs_max_recordsize = 1 * 1024 * 1024;
+
#define SWITCH64(x, y) \
{ \
uint64_t __tmp = (x); \
#define DS_REF_MAX (1ULL << 62)
-#define DSL_DEADLIST_BLOCKSIZE SPA_MAXBLOCKSIZE
-
extern inline dsl_dataset_phys_t *dsl_dataset_phys(dsl_dataset_t *ds);
/*
dsl_dataset_phys(ds)->ds_compressed_bytes += compressed;
dsl_dataset_phys(ds)->ds_uncompressed_bytes += uncompressed;
dsl_dataset_phys(ds)->ds_unique_bytes += used;
+ if (BP_GET_LSIZE(bp) > SPA_OLD_MAXBLOCKSIZE)
+ ds->ds_need_large_blocks = B_TRUE;
mutex_exit(&ds->ds_lock);
dsl_dir_diduse_space(ds->ds_dir, DD_USED_HEAD, delta,
compressed, uncompressed, tx);
list_create(&ds->ds_sendstreams, sizeof (dmu_sendarg_t),
offsetof(dmu_sendarg_t, dsa_link));
+ if (doi.doi_type == DMU_OTN_ZAP_METADATA) {
+ err = zap_contains(mos, dsobj, DS_FIELD_LARGE_BLOCKS);
+ if (err == 0)
+ ds->ds_large_blocks = B_TRUE;
+ else
+ ASSERT3U(err, ==, ENOENT);
+ }
+
if (err == 0) {
err = dsl_dir_hold_obj(dp,
dsl_dataset_phys(ds)->ds_dir_obj, NULL, ds,
dsphys->ds_flags |= dsl_dataset_phys(origin)->ds_flags &
(DS_FLAG_INCONSISTENT | DS_FLAG_CI_DATASET);
+ if (origin->ds_large_blocks)
+ dsl_dataset_activate_large_blocks_sync_impl(dsobj, tx);
+
dmu_buf_will_dirty(origin->ds_dbuf, tx);
dsl_dataset_phys(origin)->ds_num_children++;
dsphys->ds_bp = dsl_dataset_phys(ds)->ds_bp;
dmu_buf_rele(dbuf, FTAG);
+ if (ds->ds_large_blocks)
+ dsl_dataset_activate_large_blocks_sync_impl(dsobj, tx);
+
ASSERT3U(ds->ds_prev != 0, ==,
dsl_dataset_phys(ds)->ds_prev_snap_obj != 0);
if (ds->ds_prev) {
dsl_dataset_phys(ds)->ds_fsid_guid = ds->ds_fsid_guid;
dmu_objset_sync(ds->ds_objset, zio, tx);
+
+ if (ds->ds_need_large_blocks && !ds->ds_large_blocks) {
+ dsl_dataset_activate_large_blocks_sync_impl(ds->ds_object, tx);
+ ds->ds_large_blocks = B_TRUE;
+ }
}
static void
return (err);
}
+static int
+dsl_dataset_activate_large_blocks_check(void *arg, dmu_tx_t *tx)
+{
+ const char *dsname = arg;
+ dsl_dataset_t *ds;
+ dsl_pool_t *dp = dmu_tx_pool(tx);
+ int error = 0;
+
+ if (!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LARGE_BLOCKS))
+ return (SET_ERROR(ENOTSUP));
+
+ ASSERT(spa_feature_is_enabled(dp->dp_spa,
+ SPA_FEATURE_EXTENSIBLE_DATASET));
+
+ error = dsl_dataset_hold(dp, dsname, FTAG, &ds);
+ if (error != 0)
+ return (error);
+
+ if (ds->ds_large_blocks)
+ error = EALREADY;
+ dsl_dataset_rele(ds, FTAG);
+
+ return (error);
+}
+
+void
+dsl_dataset_activate_large_blocks_sync_impl(uint64_t dsobj, dmu_tx_t *tx)
+{
+ spa_t *spa = dmu_tx_pool(tx)->dp_spa;
+ objset_t *mos = dmu_tx_pool(tx)->dp_meta_objset;
+ uint64_t zero = 0;
+
+ spa_feature_incr(spa, SPA_FEATURE_LARGE_BLOCKS, tx);
+ dmu_object_zapify(mos, dsobj, DMU_OT_DSL_DATASET, tx);
+
+ VERIFY0(zap_add(mos, dsobj, DS_FIELD_LARGE_BLOCKS,
+ sizeof (zero), 1, &zero, tx));
+}
+
+static void
+dsl_dataset_activate_large_blocks_sync(void *arg, dmu_tx_t *tx)
+{
+ const char *dsname = arg;
+ dsl_dataset_t *ds;
+
+ VERIFY0(dsl_dataset_hold(dmu_tx_pool(tx), dsname, FTAG, &ds));
+
+ dsl_dataset_activate_large_blocks_sync_impl(ds->ds_object, tx);
+ ASSERT(!ds->ds_large_blocks);
+ ds->ds_large_blocks = B_TRUE;
+ dsl_dataset_rele(ds, FTAG);
+}
+
+int
+dsl_dataset_activate_large_blocks(const char *dsname)
+{
+ int error;
+
+ error = dsl_sync_task(dsname,
+ dsl_dataset_activate_large_blocks_check,
+ dsl_dataset_activate_large_blocks_sync, (void *)dsname,
+ 1, ZFS_SPACE_CHECK_RESERVED);
+
+ /*
+ * EALREADY indicates that this dataset already supports large blocks.
+ */
+ if (error == EALREADY)
+ error = 0;
+ return (error);
+}
+
/*
* Return TRUE if 'earlier' is an earlier snapshot in 'later's timeline.
* For example, they could both be snapshots of the same filesystem, and
}
#if defined(_KERNEL) && defined(HAVE_SPL)
+#if defined(_LP64)
+module_param(zfs_max_recordsize, int, 0644);
+MODULE_PARM_DESC(zfs_max_recordsize, "Max allowed record size");
+#else
+/* Limited to 1M on 32-bit platforms due to lack of virtual address space */
+module_param(zfs_max_recordsize, int, 0444);
+MODULE_PARM_DESC(zfs_max_recordsize, "Max allowed record size");
+#endif
+
EXPORT_SYMBOL(dsl_dataset_hold);
EXPORT_SYMBOL(dsl_dataset_hold_obj);
EXPORT_SYMBOL(dsl_dataset_own);
dsl_deadlist_alloc(objset_t *os, dmu_tx_t *tx)
{
if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_DEADLISTS)
- return (bpobj_alloc(os, SPA_MAXBLOCKSIZE, tx));
+ return (bpobj_alloc(os, SPA_OLD_MAXBLOCKSIZE, tx));
return (zap_create(os, DMU_OT_DEADLIST, DMU_OT_DEADLIST_HDR,
sizeof (dsl_deadlist_phys_t), tx));
}
{
if (dle->dle_bpobj.bpo_object ==
dmu_objset_pool(dl->dl_os)->dp_empty_bpobj) {
- uint64_t obj = bpobj_alloc(dl->dl_os, SPA_MAXBLOCKSIZE, tx);
+ uint64_t obj = bpobj_alloc(dl->dl_os, SPA_OLD_MAXBLOCKSIZE, tx);
bpobj_close(&dle->dle_bpobj);
bpobj_decr_empty(dl->dl_os, tx);
VERIFY3U(0, ==, bpobj_open(&dle->dle_bpobj, dl->dl_os, obj));
dle = kmem_alloc(sizeof (*dle), KM_SLEEP);
dle->dle_mintxg = mintxg;
- obj = bpobj_alloc_empty(dl->dl_os, SPA_MAXBLOCKSIZE, tx);
+ obj = bpobj_alloc_empty(dl->dl_os, SPA_OLD_MAXBLOCKSIZE, tx);
VERIFY3U(0, ==, bpobj_open(&dle->dle_bpobj, dl->dl_os, obj));
avl_add(&dl->dl_tree, dle);
if (dle->dle_mintxg >= maxtxg)
break;
- obj = bpobj_alloc_empty(dl->dl_os, SPA_MAXBLOCKSIZE, tx);
+ obj = bpobj_alloc_empty(dl->dl_os, SPA_OLD_MAXBLOCKSIZE, tx);
VERIFY3U(0, ==, zap_add_int_key(dl->dl_os, newobj,
dle->dle_mintxg, obj, tx));
}
obj = ds->ds_object;
+ if (ds->ds_large_blocks) {
+ ASSERT0(zap_contains(mos, obj, DS_FIELD_LARGE_BLOCKS));
+ spa_feature_decr(dp->dp_spa, SPA_FEATURE_LARGE_BLOCKS, tx);
+ }
if (dsl_dataset_phys(ds)->ds_prev_snap_obj != 0) {
ASSERT3P(ds->ds_prev, ==, NULL);
VERIFY0(dsl_dataset_hold_obj(dp,
ASSERT0(ds->ds_reserved);
}
+ if (ds->ds_large_blocks)
+ spa_feature_decr(dp->dp_spa, SPA_FEATURE_LARGE_BLOCKS, tx);
+
dsl_scan_ds_destroyed(ds, tx);
obj = ds->ds_object;
FREE_DIR_NAME, &dp->dp_free_dir));
/* create and open the free_bplist */
- obj = bpobj_alloc(dp->dp_meta_objset, SPA_MAXBLOCKSIZE, tx);
+ obj = bpobj_alloc(dp->dp_meta_objset, SPA_OLD_MAXBLOCKSIZE, tx);
VERIFY(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj, tx) == 0);
VERIFY0(bpobj_open(&dp->dp_free_bpobj,
* subobj support. So call dmu_object_alloc() directly.
*/
obj = dmu_object_alloc(dp->dp_meta_objset, DMU_OT_BPOBJ,
- SPA_MAXBLOCKSIZE, DMU_OT_BPOBJ_HDR, sizeof (bpobj_phys_t), tx);
+ SPA_OLD_MAXBLOCKSIZE, DMU_OT_BPOBJ_HDR, sizeof (bpobj_phys_t), tx);
VERIFY0(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj, tx));
VERIFY0(bpobj_open(&dp->dp_free_bpobj, dp->dp_meta_objset, obj));
if (size == 0) {
blocksize = SPA_MINBLOCKSIZE;
- } else if (size > SPA_MAXBLOCKSIZE) {
+ } else if (size > SPA_OLD_MAXBLOCKSIZE) {
ASSERT(0);
return (SET_ERROR(EFBIG));
} else {
hdrsize = sa_find_sizes(sa, attr_desc, attr_count, hdl->sa_bonus,
SA_BONUS, &spill_idx, &used, &spilling);
- if (used > SPA_MAXBLOCKSIZE)
+ if (used > SPA_OLD_MAXBLOCKSIZE)
return (SET_ERROR(EFBIG));
VERIFY(0 == dmu_set_bonus(hdl->sa_bonus, spilling ?
attr_count - spill_idx, hdl->sa_spill, SA_SPILL, &i,
&spill_used, &dummy);
- if (spill_used > SPA_MAXBLOCKSIZE)
+ if (spill_used > SPA_OLD_MAXBLOCKSIZE)
return (SET_ERROR(EFBIG));
if (BUF_SPACE_NEEDED(spill_used, spillhdrsize) >
spa_prop_add_list(*nvp, ZPOOL_PROP_ALTROOT, spa->spa_root,
0, ZPROP_SRC_LOCAL);
+ if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL,
+ MIN(zfs_max_recordsize, SPA_MAXBLOCKSIZE), ZPROP_SRC_NONE);
+ } else {
+ spa_prop_add_list(*nvp, ZPOOL_PROP_MAXBLOCKSIZE, NULL,
+ SPA_OLD_MAXBLOCKSIZE, ZPROP_SRC_NONE);
+ }
+
if ((dp = list_head(&spa->spa_config_list)) != NULL) {
if (dp->scd_path == NULL) {
spa_prop_add_list(*nvp, ZPOOL_PROP_CACHEFILE,
if (!error) {
objset_t *os;
- uint64_t compress;
+ uint64_t propval;
if (strval == NULL || strval[0] == '\0') {
objnum = zpool_prop_default_numeric(
if (error)
break;
- /* Must be ZPL and not gzip compressed. */
+ /*
+ * Must be ZPL, and its property settings
+ * must be supported by GRUB (compression
+ * is not gzip, and large blocks are not used).
+ */
if (dmu_objset_type(os) != DMU_OST_ZFS) {
error = SET_ERROR(ENOTSUP);
} else if ((error =
dsl_prop_get_int_ds(dmu_objset_ds(os),
zfs_prop_to_name(ZFS_PROP_COMPRESSION),
- &compress)) == 0 &&
- !BOOTFS_COMPRESS_VALID(compress)) {
+ &propval)) == 0 &&
+ !BOOTFS_COMPRESS_VALID(propval)) {
+ error = SET_ERROR(ENOTSUP);
+ } else if ((error =
+ dsl_prop_get_int_ds(dmu_objset_ds(os),
+ zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
+ &propval)) == 0 &&
+ propval > SPA_OLD_MAXBLOCKSIZE) {
error = SET_ERROR(ENOTSUP);
} else {
objnum = dmu_objset_id(os);
ASSERT(spa->spa_history == 0);
spa->spa_history = dmu_object_alloc(mos, DMU_OT_SPA_HISTORY,
- SPA_MAXBLOCKSIZE, DMU_OT_SPA_HISTORY_OFFSETS,
+ SPA_OLD_MAXBLOCKSIZE, DMU_OT_SPA_HISTORY_OFFSETS,
sizeof (spa_history_phys_t), tx);
VERIFY(zap_add(mos, DMU_POOL_DIRECTORY_OBJECT,
return (spa->spa_debug);
}
+int
+spa_maxblocksize(spa_t *spa)
+{
+ if (spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS))
+ return (SPA_MAXBLOCKSIZE);
+ else
+ return (SPA_OLD_MAXBLOCKSIZE);
+}
+
#if defined(_KERNEL) && defined(HAVE_SPL)
/* Namespace manipulation */
EXPORT_SYMBOL(spa_lookup);
EXPORT_SYMBOL(spa_bootfs);
EXPORT_SYMBOL(spa_delegation);
EXPORT_SYMBOL(spa_meta_objset);
+EXPORT_SYMBOL(spa_maxblocksize);
/* Miscellaneous support routines */
EXPORT_SYMBOL(spa_rename);
/*
* Compute the raidz-deflation ratio. Note, we hard-code
- * in 128k (1 << 17) because it is the current "typical" blocksize.
- * Even if SPA_MAXBLOCKSIZE changes, this algorithm must never change,
- * or we will inconsistently account for existing bp's.
+ * in 128k (1 << 17) because it is the "typical" blocksize.
+ * Even though SPA_MAXBLOCKSIZE changed, this algorithm can not change,
+ * otherwise it would inconsistently account for existing bp's.
*/
vd->vdev_deflate_ratio = (1 << 17) /
(vdev_psize_to_asize(vd, 1 << 17) >> SPA_MINBLOCKSHIFT);
goto retry;
}
- dr->dr_bio[i] = bio_alloc(GFP_NOIO,
- bio_nr_pages(bio_ptr, bio_size));
/* bio_alloc() with __GFP_WAIT never returns NULL */
+ dr->dr_bio[i] = bio_alloc(GFP_NOIO,
+ MIN(bio_nr_pages(bio_ptr, bio_size), BIO_MAX_PAGES));
if (unlikely(dr->dr_bio[i] == NULL)) {
vdev_disk_dio_free(dr);
return (ENOMEM);
* we include spans of optional I/Os to aid aggregation at the disk even when
* they aren't able to help us aggregate at this level.
*/
-int zfs_vdev_aggregation_limit = SPA_MAXBLOCKSIZE;
+int zfs_vdev_aggregation_limit = SPA_OLD_MAXBLOCKSIZE;
int zfs_vdev_read_gap_limit = 32 << 10;
int zfs_vdev_write_gap_limit = 4 << 10;
#include <sys/zap_leaf.h>
#include <sys/avl.h>
#include <sys/arc.h>
+#include <sys/dmu_objset.h>
#ifdef _KERNEL
#include <sys/sunddi.h>
uint64_t obj = dmu_object_alloc(os, ot, 0, bonustype, bonuslen, tx);
ASSERT(leaf_blockshift >= SPA_MINBLOCKSHIFT &&
- leaf_blockshift <= SPA_MAXBLOCKSHIFT &&
+ leaf_blockshift <= SPA_OLD_MAXBLOCKSHIFT &&
indirect_blockshift >= SPA_MINBLOCKSHIFT &&
- indirect_blockshift <= SPA_MAXBLOCKSHIFT);
+ indirect_blockshift <= SPA_OLD_MAXBLOCKSHIFT);
VERIFY(dmu_object_set_blocksize(os, obj,
1ULL << leaf_blockshift, indirect_blockshift, tx) == 0);
zap_t *zap;
int err = 0;
-
/*
* Since, we don't have a name, we cannot figure out which blocks will
* be affected in this operation. So, account for the worst case :
* large microzap results in a promotion to fatzap.
*/
if (name == NULL) {
- *towrite += (3 + (add ? 4 : 0)) * SPA_MAXBLOCKSIZE;
+ *towrite += (3 + (add ? 4 : 0)) * SPA_OLD_MAXBLOCKSIZE;
return (err);
}
/*
* We treat this case as similar to (name == NULL)
*/
- *towrite += (3 + (add ? 4 : 0)) * SPA_MAXBLOCKSIZE;
+ *towrite += (3 + (add ? 4 : 0)) * SPA_OLD_MAXBLOCKSIZE;
}
} else {
/*
* ptrtbl blocks
*/
if (dmu_buf_freeable(zap->zap_dbuf))
- *tooverwrite += SPA_MAXBLOCKSIZE;
+ *tooverwrite += MZAP_MAX_BLKSZ;
else
- *towrite += SPA_MAXBLOCKSIZE;
+ *towrite += MZAP_MAX_BLKSZ;
if (add) {
- *towrite += 4 * SPA_MAXBLOCKSIZE;
+ *towrite += 4 * MZAP_MAX_BLKSZ;
}
}
{
return ((c >= 'a' && c <= 'z') ||
(c >= '0' && c <= '9') ||
- c == (after_colon ? '_' : '.'));
+ (after_colon && c == '_') ||
+ (!after_colon && (c == '.' || c == '-')));
}
/*
"com.delphix:embedded_data", "embedded_data",
"Blocks which compress very well use even less space.",
B_FALSE, B_TRUE, B_TRUE, NULL);
+
+ {
+ static const spa_feature_t large_blocks_deps[] = {
+ SPA_FEATURE_EXTENSIBLE_DATASET,
+ SPA_FEATURE_NONE
+ };
+ zfeature_register(SPA_FEATURE_LARGE_BLOCKS,
+ "org.open-zfs:large_blocks", "large_blocks",
+ "Support for blocks larger than 128KB.", B_FALSE, B_FALSE, B_FALSE,
+ large_blocks_deps);
+ }
}
const char *propname = nvpair_name(pair);
zfs_prop_t prop = zfs_name_to_prop(propname);
uint64_t intval;
- int err;
+ int err = -1;
if (prop == ZPROP_INVAL) {
if (zfs_prop_userquota(propname))
* the SPA supports it. We ignore any errors here since
* we'll catch them later.
*/
- if (nvpair_type(pair) == DATA_TYPE_UINT64 &&
- nvpair_value_uint64(pair, &intval) == 0) {
+ if (nvpair_value_uint64(pair, &intval) == 0) {
if (intval >= ZIO_COMPRESS_GZIP_1 &&
intval <= ZIO_COMPRESS_GZIP_9 &&
zfs_earlier_version(dsname,
return (SET_ERROR(ENOTSUP));
break;
+ case ZFS_PROP_RECORDSIZE:
+ /* Record sizes above 128k need the feature to be enabled */
+ if (nvpair_value_uint64(pair, &intval) == 0 &&
+ intval > SPA_OLD_MAXBLOCKSIZE) {
+ spa_t *spa;
+
+ /*
+ * If this is a bootable dataset then
+ * the we don't allow large (>128K) blocks,
+ * because GRUB doesn't support them.
+ */
+ if (zfs_is_bootfs(dsname) &&
+ intval > SPA_OLD_MAXBLOCKSIZE) {
+ return (SET_ERROR(EDOM));
+ }
+
+ /*
+ * We don't allow setting the property above 1MB,
+ * unless the tunable has been changed.
+ */
+ if (intval > zfs_max_recordsize ||
+ intval > SPA_MAXBLOCKSIZE)
+ return (SET_ERROR(EDOM));
+
+ if ((err = spa_open(dsname, &spa, FTAG)) != 0)
+ return (err);
+
+ if (!spa_feature_is_enabled(spa,
+ SPA_FEATURE_LARGE_BLOCKS)) {
+ spa_close(spa, FTAG);
+ return (SET_ERROR(ENOTSUP));
+ }
+ spa_close(spa, FTAG);
+ }
+ break;
+
case ZFS_PROP_SHARESMB:
if (zpl_earlier_version(dsname, ZPL_VERSION_FUID))
return (SET_ERROR(ENOTSUP));
* zc_fromobj objsetid of incremental fromsnap (may be zero)
* zc_guid if set, estimate size of stream only. zc_cookie is ignored.
* output size in zc_objset_type.
- * zc_flags if =1, WRITE_EMBEDDED records are permitted
+ * zc_flags lzc_send_flags
*
* outputs:
* zc_objset_type estimated size, if zc_guid is set
offset_t off;
boolean_t estimate = (zc->zc_guid != 0);
boolean_t embedok = (zc->zc_flags & 0x1);
+ boolean_t large_block_ok = (zc->zc_flags & 0x2);
if (zc->zc_obj != 0) {
dsl_pool_t *dp;
off = fp->f_offset;
error = dmu_send_obj(zc->zc_name, zc->zc_sendobj,
- zc->zc_fromobj, embedok, zc->zc_cookie, fp->f_vnode, &off);
+ zc->zc_fromobj, embedok, large_block_ok,
+ zc->zc_cookie, fp->f_vnode, &off);
if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
fp->f_offset = off;
* innvl: {
* "fd" -> file descriptor to write stream to (int32)
* (optional) "fromsnap" -> full snap name to send an incremental from
+ * (optional) "largeblockok" -> (value ignored)
+ * indicates that blocks > 128KB are permitted
* (optional) "embedok" -> (value ignored)
* presence indicates DRR_WRITE_EMBEDDED records are permitted
* }
char *fromname = NULL;
int fd;
file_t *fp;
+ boolean_t largeblockok;
boolean_t embedok;
error = nvlist_lookup_int32(innvl, "fd", &fd);
(void) nvlist_lookup_string(innvl, "fromsnap", &fromname);
+ largeblockok = nvlist_exists(innvl, "largeblockok");
embedok = nvlist_exists(innvl, "embedok");
if ((fp = getf(fd)) == NULL)
return (SET_ERROR(EBADF));
off = fp->f_offset;
- error = dmu_send(snapname, fromname, embedok, fd, fp->f_vnode, &off);
+ error = dmu_send(snapname, fromname, embedok, largeblockok,
+ fd, fp->f_vnode, &off);
if (VOP_SEEK(fp->f_vnode, fp->f_offset, &off, NULL) == 0)
fp->f_offset = off;
* If the write would overflow the largest block then split it.
*/
if (write_state != WR_INDIRECT && resid > ZIL_MAX_LOG_DATA)
- len = SPA_MAXBLOCKSIZE >> 1;
+ len = SPA_OLD_MAXBLOCKSIZE >> 1;
else
len = resid;
blksz_changed_cb(void *arg, uint64_t newval)
{
zfs_sb_t *zsb = arg;
-
- if (newval < SPA_MINBLOCKSIZE ||
- newval > SPA_MAXBLOCKSIZE || !ISP2(newval))
- newval = SPA_MAXBLOCKSIZE;
+ ASSERT3U(newval, <=, spa_maxblocksize(dmu_objset_spa(zsb->z_os)));
+ ASSERT3U(newval, >=, SPA_MINBLOCKSIZE);
+ ASSERT(ISP2(newval));
zsb->z_max_blksz = newval;
}
*/
zsb->z_sb = NULL;
zsb->z_parent = zsb;
- zsb->z_max_blksz = SPA_MAXBLOCKSIZE;
+ zsb->z_max_blksz = SPA_OLD_MAXBLOCKSIZE;
zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
zsb->z_os = os;
uint64_t new_blksz;
if (zp->z_blksz > max_blksz) {
+ /*
+ * File's blocksize is already larger than the
+ * "recordsize" property. Only let it grow to
+ * the next power of 2.
+ */
ASSERT(!ISP2(zp->z_blksz));
- new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE);
+ new_blksz = MIN(end_size,
+ 1 << highbit64(zp->z_blksz));
} else {
new_blksz = MIN(end_size, max_blksz);
}
#endif /* _KERNEL */
#include <sys/dmu.h>
+#include <sys/dmu_objset.h>
#include <sys/refcount.h>
#include <sys/stat.h>
#include <sys/zap.h>
* We are growing the file past the current block size.
*/
if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) {
+ /*
+ * File's blocksize is already larger than the
+ * "recordsize" property. Only let it grow to
+ * the next power of 2.
+ */
ASSERT(!ISP2(zp->z_blksz));
- newblksz = MIN(end, SPA_MAXBLOCKSIZE);
+ newblksz = MIN(end, 1 << highbit64(zp->z_blksz));
} else {
newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz);
}
sizeof (cksum)) || BP_IS_HOLE(&zilc->zc_next_blk)) {
error = SET_ERROR(ECKSUM);
} else {
+ ASSERT3U(len, <=, SPA_OLD_MAXBLOCKSIZE);
bcopy(lr, dst, len);
*end = (char *)dst + len;
*nbp = zilc->zc_next_blk;
(zilc->zc_nused > (size - sizeof (*zilc)))) {
error = SET_ERROR(ECKSUM);
} else {
+ ASSERT3U(zilc->zc_nused, <=,
+ SPA_OLD_MAXBLOCKSIZE);
bcopy(lr, dst, zilc->zc_nused);
*end = (char *)dst + zilc->zc_nused;
*nbp = zilc->zc_next_blk;
* If the log has been claimed, stop if we encounter a sequence
* number greater than the highest claimed sequence number.
*/
- lrbuf = zio_buf_alloc(SPA_MAXBLOCKSIZE);
+ lrbuf = zio_buf_alloc(SPA_OLD_MAXBLOCKSIZE);
zil_bp_tree_init(zilog);
for (blk = zh->zh_log; !BP_IS_HOLE(&blk); blk = next_blk) {
(max_blk_seq == claim_blk_seq && max_lr_seq == claim_lr_seq));
zil_bp_tree_fini(zilog);
- zio_buf_free(lrbuf, SPA_MAXBLOCKSIZE);
+ zio_buf_free(lrbuf, SPA_OLD_MAXBLOCKSIZE);
return (error);
}
*
* These must be a multiple of 4KB. Note only the amount used (again
* aligned to 4KB) actually gets written. However, we can't always just
- * allocate SPA_MAXBLOCKSIZE as the slog space could be exhausted.
+ * allocate SPA_OLD_MAXBLOCKSIZE as the slog space could be exhausted.
*/
uint64_t zil_block_buckets[] = {
4096, /* non TX_WRITE */
continue;
zil_blksz = zil_block_buckets[i];
if (zil_blksz == UINT64_MAX)
- zil_blksz = SPA_MAXBLOCKSIZE;
+ zil_blksz = SPA_OLD_MAXBLOCKSIZE;
zilog->zl_prev_blks[zilog->zl_prev_rotor] = zil_blksz;
for (i = 0; i < ZIL_PREV_BLKS; i++)
zil_blksz = MAX(zil_blksz, zilog->zl_prev_blks[i]);
* Copyright (c) 2011 Nexenta Systems, Inc. All rights reserved.
*/
+#include <sys/sysmacros.h>
#include <sys/zfs_context.h>
#include <sys/fm/fs/zfs.h>
#include <sys/spa.h>
/*
* For small buffers, we want a cache for each multiple of
- * SPA_MINBLOCKSIZE. For medium-size buffers, we want a cache
- * for each quarter-power of 2. For large buffers, we want
- * a cache for each multiple of PAGESIZE.
+ * SPA_MINBLOCKSIZE. For larger buffers, we want a cache
+ * for each quarter-power of 2.
*/
for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) {
size_t size = (c + 1) << SPA_MINBLOCKSHIFT;
size_t align = 0;
size_t cflags = (size > zio_buf_debug_limit) ? KMC_NODEBUG : 0;
- while (p2 & (p2 - 1))
+#ifdef _ILP32
+ /*
+ * Cache size limited to 1M on 32-bit platforms until ARC
+ * buffers no longer require virtual address space.
+ */
+ if (size > zfs_max_recordsize)
+ break;
+#endif
+
+ while (!ISP2(p2))
p2 &= p2 - 1;
#ifndef _KERNEL
#endif
if (size <= 4 * SPA_MINBLOCKSIZE) {
align = SPA_MINBLOCKSIZE;
- } else if (IS_P2ALIGNED(size, PAGESIZE)) {
- align = PAGESIZE;
} else if (IS_P2ALIGNED(size, p2 >> 2)) {
- align = p2 >> 2;
+ align = MIN(p2 >> 2, PAGESIZE);
}
if (align != 0) {
kmem_cache_t *last_data_cache = NULL;
for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) {
+#ifdef _ILP32
+ /*
+ * Cache size limited to 1M on 32-bit platforms until ARC
+ * buffers no longer require virtual address space.
+ */
+ if (((c + 1) << SPA_MINBLOCKSHIFT) > zfs_max_recordsize)
+ break;
+#endif
if (zio_buf_cache[c] != last_cache) {
last_cache = zio_buf_cache[c];
kmem_cache_destroy(zio_buf_cache[c]);
projects / mirror_zfs.git / commitdiff