/*
* Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2012, 2016 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/zio_checksum.h>
#include <sys/zio_compress.h>
#include <sys/dsl_scan.h>
+#include <sys/abd.h>
+
+static kmem_cache_t *ddt_cache;
+static kmem_cache_t *ddt_entry_cache;
/*
* Enable/disable prefetching of dedup-ed blocks which are going to be freed.
*/
-int zfs_dedup_prefetch = 1;
+int zfs_dedup_prefetch = 0;
static const ddt_ops_t *ddt_ops[DDT_TYPES] = {
&ddt_zap_ops,
spa_t *spa = ddt->ddt_spa;
objset_t *os = ddt->ddt_os;
uint64_t *objectp = &ddt->ddt_object[type][class];
- boolean_t prehash = zio_checksum_table[ddt->ddt_checksum].ci_dedup;
+ boolean_t prehash = zio_checksum_table[ddt->ddt_checksum].ci_flags &
+ ZCHECKSUM_FLAG_DEDUP;
char name[DDT_NAMELEN];
ddt_object_name(ddt, type, class, name);
spa_t *spa = ddt->ddt_spa;
objset_t *os = ddt->ddt_os;
uint64_t *objectp = &ddt->ddt_object[type][class];
+ uint64_t count;
char name[DDT_NAMELEN];
ddt_object_name(ddt, type, class, name);
ASSERT(*objectp != 0);
- ASSERT(ddt_object_count(ddt, type, class) == 0);
ASSERT(ddt_histogram_empty(&ddt->ddt_histogram[type][class]));
+ VERIFY(ddt_object_count(ddt, type, class, &count) == 0 && count == 0);
VERIFY(zap_remove(os, DMU_POOL_DIRECTORY_OBJECT, name, tx) == 0);
VERIFY(zap_remove(os, spa->spa_ddt_stat_object, name, tx) == 0);
VERIFY(ddt_ops[type]->ddt_op_destroy(os, *objectp, tx) == 0);
{
ddt_object_t *ddo = &ddt->ddt_object_stats[type][class];
dmu_object_info_t doi;
+ uint64_t count;
char name[DDT_NAMELEN];
int error;
error = zap_lookup(ddt->ddt_os, DMU_POOL_DIRECTORY_OBJECT, name,
sizeof (uint64_t), 1, &ddt->ddt_object[type][class]);
-
- if (error)
+ if (error != 0)
return (error);
error = zap_lookup(ddt->ddt_os, ddt->ddt_spa->spa_ddt_stat_object, name,
sizeof (uint64_t), sizeof (ddt_histogram_t) / sizeof (uint64_t),
&ddt->ddt_histogram[type][class]);
+ if (error != 0)
+ return (error);
/*
* Seed the cached statistics.
*/
- VERIFY(ddt_object_info(ddt, type, class, &doi) == 0);
+ error = ddt_object_info(ddt, type, class, &doi);
+ if (error)
+ return (error);
- ddo->ddo_count = ddt_object_count(ddt, type, class);
+ error = ddt_object_count(ddt, type, class, &count);
+ if (error)
+ return (error);
+
+ ddo->ddo_count = count;
ddo->ddo_dspace = doi.doi_physical_blocks_512 << 9;
ddo->ddo_mspace = doi.doi_fill_count * doi.doi_data_block_size;
- ASSERT(error == 0);
- return (error);
+ return (0);
}
static void
{
ddt_object_t *ddo = &ddt->ddt_object_stats[type][class];
dmu_object_info_t doi;
+ uint64_t count;
char name[DDT_NAMELEN];
ddt_object_name(ddt, type, class, name);
* Cache DDT statistics; this is the only time they'll change.
*/
VERIFY(ddt_object_info(ddt, type, class, &doi) == 0);
+ VERIFY(ddt_object_count(ddt, type, class, &count) == 0);
- ddo->ddo_count = ddt_object_count(ddt, type, class);
+ ddo->ddo_count = count;
ddo->ddo_dspace = doi.doi_physical_blocks_512 << 9;
ddo->ddo_mspace = doi.doi_fill_count * doi.doi_data_block_size;
}
ddt_entry_t *dde)
{
if (!ddt_object_exists(ddt, type, class))
- return (ENOENT);
+ return (SET_ERROR(ENOENT));
return (ddt_ops[type]->ddt_op_lookup(ddt->ddt_os,
ddt->ddt_object[type][class], dde));
ddt->ddt_object[type][class], dde, walk));
}
-uint64_t
-ddt_object_count(ddt_t *ddt, enum ddt_type type, enum ddt_class class)
+int
+ddt_object_count(ddt_t *ddt, enum ddt_type type, enum ddt_class class,
+ uint64_t *count)
{
ASSERT(ddt_object_exists(ddt, type, class));
return (ddt_ops[type]->ddt_op_count(ddt->ddt_os,
- ddt->ddt_object[type][class]));
+ ddt->ddt_object[type][class], count));
}
int
dmu_object_info_t *doi)
{
if (!ddt_object_exists(ddt, type, class))
- return (ENOENT);
+ return (SET_ERROR(ENOENT));
return (dmu_object_info(ddt->ddt_os, ddt->ddt_object[type][class],
doi));
BP_SET_BIRTH(bp, txg, ddp->ddp_phys_birth);
}
+/*
+ * The bp created via this function may be used for repairs and scrub, but it
+ * will be missing the salt / IV required to do a full decrypting read.
+ */
void
ddt_bp_create(enum zio_checksum checksum,
const ddt_key_t *ddk, const ddt_phys_t *ddp, blkptr_t *bp)
ddt_bp_fill(ddp, bp, ddp->ddp_phys_birth);
bp->blk_cksum = ddk->ddk_cksum;
- bp->blk_fill = 1;
BP_SET_LSIZE(bp, DDK_GET_LSIZE(ddk));
BP_SET_PSIZE(bp, DDK_GET_PSIZE(ddk));
BP_SET_COMPRESS(bp, DDK_GET_COMPRESS(ddk));
+ BP_SET_CRYPT(bp, DDK_GET_CRYPT(ddk));
+ BP_SET_FILL(bp, 1);
BP_SET_CHECKSUM(bp, checksum);
BP_SET_TYPE(bp, DMU_OT_DEDUP);
BP_SET_LEVEL(bp, 0);
- BP_SET_DEDUP(bp, 0);
+ BP_SET_DEDUP(bp, 1);
BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER);
}
ddk->ddk_cksum = bp->blk_cksum;
ddk->ddk_prop = 0;
+ ASSERT(BP_IS_ENCRYPTED(bp) || !BP_USES_CRYPT(bp));
+
DDK_SET_LSIZE(ddk, BP_GET_LSIZE(bp));
DDK_SET_PSIZE(ddk, BP_GET_PSIZE(bp));
DDK_SET_COMPRESS(ddk, BP_GET_COMPRESS(bp));
+ DDK_SET_CRYPT(ddk, BP_USES_CRYPT(bp));
}
void
void
ddt_phys_decref(ddt_phys_t *ddp)
{
- ASSERT((int64_t)ddp->ddp_refcnt > 0);
- ddp->ddp_refcnt--;
+ if (ddp) {
+ ASSERT(ddp->ddp_refcnt > 0);
+ ddp->ddp_refcnt--;
+ }
}
void
blkptr_t blk;
ddt_bp_create(ddt->ddt_checksum, ddk, ddp, &blk);
+
+ /*
+ * We clear the dedup bit so that zio_free() will actually free the
+ * space, rather than just decrementing the refcount in the DDT.
+ */
+ BP_SET_DEDUP(&blk, 0);
+
ddt_phys_clear(ddp);
zio_free(ddt->ddt_spa, txg, &blk);
}
if (ddp->ddp_phys_birth == 0)
continue;
- for (int d = 0; d < SPA_DVAS_PER_BP; d++)
+ for (int d = 0; d < DDE_GET_NDVAS(dde); d++)
dsize += dva_get_dsize_sync(spa, &ddp->ddp_dva[d]);
dds->dds_blocks += 1;
ddt_stat_generate(ddt, dde, &dds);
- bucket = highbit(dds.dds_ref_blocks) - 1;
+ bucket = highbit64(dds.dds_ref_blocks) - 1;
ASSERT(bucket >= 0);
ddh = &ddt->ddt_histogram[dde->dde_type][dde->dde_class];
uint64_t
ddt_get_dedup_dspace(spa_t *spa)
{
- ddt_stat_t dds_total = { 0 };
+ ddt_stat_t dds_total;
+
+ if (spa->spa_dedup_dspace != ~0ULL)
+ return (spa->spa_dedup_dspace);
+ bzero(&dds_total, sizeof (ddt_stat_t));
+
+ /* Calculate and cache the stats */
ddt_get_dedup_stats(spa, &dds_total);
- return (dds_total.dds_ref_dsize - dds_total.dds_dsize);
+ spa->spa_dedup_dspace = dds_total.dds_ref_dsize - dds_total.dds_dsize;
+ return (spa->spa_dedup_dspace);
}
uint64_t
uint64_t ditto = spa->spa_dedup_ditto;
int total_copies = 0;
int desired_copies = 0;
+ int copies_needed = 0;
for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) {
ddt_phys_t *ddp = &dde->dde_phys[p];
if (total_refcnt >= ditto * ditto)
desired_copies++;
- return (MAX(desired_copies, total_copies) - total_copies);
+ copies_needed = MAX(desired_copies, total_copies) - total_copies;
+
+ /* encrypted blocks store their IV in DVA[2] */
+ if (DDK_GET_CRYPT(&dde->dde_key))
+ copies_needed = MIN(copies_needed, SPA_DVAS_PER_BP - 1);
+
+ return (copies_needed);
}
int
dva_t *dva = ddp->ddp_dva;
int copies = 0 - DVA_GET_GANG(dva);
- for (int d = 0; d < SPA_DVAS_PER_BP; d++, dva++)
+ for (int d = 0; d < DDE_GET_NDVAS(dde); d++, dva++)
if (DVA_IS_VALID(dva))
copies++;
bcopy(src, dst, s_len);
}
- *version = (ZFS_HOST_BYTEORDER & DDT_COMPRESS_BYTEORDER_MASK) | cpfunc;
+ *version = cpfunc;
+ /* CONSTCOND */
+ if (ZFS_HOST_BYTEORDER)
+ *version |= DDT_COMPRESS_BYTEORDER_MASK;
return (c_len + 1);
}
else
bcopy(src, dst, d_len);
- if ((version ^ ZFS_HOST_BYTEORDER) & DDT_COMPRESS_BYTEORDER_MASK)
+ if (((version & DDT_COMPRESS_BYTEORDER_MASK) != 0) !=
+ (ZFS_HOST_BYTEORDER != 0))
byteswap_uint64_array(dst, d_len);
}
mutex_exit(&ddt->ddt_lock);
}
+void
+ddt_init(void)
+{
+ ddt_cache = kmem_cache_create("ddt_cache",
+ sizeof (ddt_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
+ ddt_entry_cache = kmem_cache_create("ddt_entry_cache",
+ sizeof (ddt_entry_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
+}
+
+void
+ddt_fini(void)
+{
+ kmem_cache_destroy(ddt_entry_cache);
+ kmem_cache_destroy(ddt_cache);
+}
+
static ddt_entry_t *
ddt_alloc(const ddt_key_t *ddk)
{
ddt_entry_t *dde;
- dde = kmem_zalloc(sizeof (ddt_entry_t), KM_SLEEP);
+ dde = kmem_cache_alloc(ddt_entry_cache, KM_SLEEP);
+ bzero(dde, sizeof (ddt_entry_t));
cv_init(&dde->dde_cv, NULL, CV_DEFAULT, NULL);
dde->dde_key = *ddk;
for (int p = 0; p < DDT_PHYS_TYPES; p++)
ASSERT(dde->dde_lead_zio[p] == NULL);
- if (dde->dde_repair_data != NULL)
- zio_buf_free(dde->dde_repair_data,
- DDK_GET_PSIZE(&dde->dde_key));
+ if (dde->dde_repair_abd != NULL)
+ abd_free(dde->dde_repair_abd);
cv_destroy(&dde->dde_cv);
- kmem_free(dde, sizeof (*dde));
+ kmem_cache_free(ddt_entry_cache, dde);
}
void
for (type = 0; type < DDT_TYPES; type++) {
for (class = 0; class < DDT_CLASSES; class++) {
error = ddt_object_lookup(ddt, type, class, dde);
- if (error != ENOENT)
+ if (error != ENOENT) {
+ ASSERT0(error);
break;
+ }
}
if (error != ENOENT)
break;
}
- ASSERT(error == 0 || error == ENOENT);
-
ddt_enter(ddt);
ASSERT(dde->dde_loaded == B_FALSE);
}
}
+/*
+ * Opaque struct used for ddt_key comparison
+ */
+#define DDT_KEY_CMP_LEN (sizeof (ddt_key_t) / sizeof (uint16_t))
+
+typedef struct ddt_key_cmp {
+ uint16_t u16[DDT_KEY_CMP_LEN];
+} ddt_key_cmp_t;
+
int
ddt_entry_compare(const void *x1, const void *x2)
{
const ddt_entry_t *dde1 = x1;
const ddt_entry_t *dde2 = x2;
- const uint64_t *u1 = (const uint64_t *)&dde1->dde_key;
- const uint64_t *u2 = (const uint64_t *)&dde2->dde_key;
-
- for (int i = 0; i < DDT_KEY_WORDS; i++) {
- if (u1[i] < u2[i])
- return (-1);
- if (u1[i] > u2[i])
- return (1);
+ const ddt_key_cmp_t *k1 = (const ddt_key_cmp_t *)&dde1->dde_key;
+ const ddt_key_cmp_t *k2 = (const ddt_key_cmp_t *)&dde2->dde_key;
+ int32_t cmp = 0;
+
+ for (int i = 0; i < DDT_KEY_CMP_LEN; i++) {
+ cmp = (int32_t)k1->u16[i] - (int32_t)k2->u16[i];
+ if (likely(cmp))
+ break;
}
- return (0);
+ return (AVL_ISIGN(cmp));
}
static ddt_t *
{
ddt_t *ddt;
- ddt = kmem_zalloc(sizeof (*ddt), KM_SLEEP);
+ ddt = kmem_cache_alloc(ddt_cache, KM_SLEEP);
+ bzero(ddt, sizeof (ddt_t));
mutex_init(&ddt->ddt_lock, NULL, MUTEX_DEFAULT, NULL);
avl_create(&ddt->ddt_tree, ddt_entry_compare,
avl_destroy(&ddt->ddt_tree);
avl_destroy(&ddt->ddt_repair_tree);
mutex_destroy(&ddt->ddt_lock);
- kmem_free(ddt, sizeof (*ddt));
+ kmem_cache_free(ddt_cache, ddt);
}
void
*/
bcopy(ddt->ddt_histogram, &ddt->ddt_histogram_cache,
sizeof (ddt->ddt_histogram));
+ spa->spa_dedup_dspace = ~0ULL;
}
return (0);
ddt_class_contains(spa_t *spa, enum ddt_class max_class, const blkptr_t *bp)
{
ddt_t *ddt;
- ddt_entry_t dde;
+ ddt_entry_t *dde;
if (!BP_GET_DEDUP(bp))
return (B_FALSE);
return (B_TRUE);
ddt = spa->spa_ddt[BP_GET_CHECKSUM(bp)];
+ dde = kmem_cache_alloc(ddt_entry_cache, KM_SLEEP);
- ddt_key_fill(&dde.dde_key, bp);
+ ddt_key_fill(&(dde->dde_key), bp);
- for (enum ddt_type type = 0; type < DDT_TYPES; type++)
- for (enum ddt_class class = 0; class <= max_class; class++)
- if (ddt_object_lookup(ddt, type, class, &dde) == 0)
+ for (enum ddt_type type = 0; type < DDT_TYPES; type++) {
+ for (enum ddt_class class = 0; class <= max_class; class++) {
+ if (ddt_object_lookup(ddt, type, class, dde) == 0) {
+ kmem_cache_free(ddt_entry_cache, dde);
return (B_TRUE);
+ }
+ }
+ }
+ kmem_cache_free(ddt_entry_cache, dde);
return (B_FALSE);
}
ddt_enter(ddt);
- if (dde->dde_repair_data != NULL && spa_writeable(ddt->ddt_spa) &&
+ if (dde->dde_repair_abd != NULL && spa_writeable(ddt->ddt_spa) &&
avl_find(&ddt->ddt_repair_tree, dde, &where) == NULL)
avl_insert(&ddt->ddt_repair_tree, dde, where);
else
continue;
ddt_bp_create(ddt->ddt_checksum, ddk, ddp, &blk);
zio_nowait(zio_rewrite(zio, zio->io_spa, 0, &blk,
- rdde->dde_repair_data, DDK_GET_PSIZE(rddk), NULL, NULL,
+ rdde->dde_repair_abd, DDK_GET_PSIZE(rddk), NULL, NULL,
ZIO_PRIORITY_SYNC_WRITE, ZIO_DDT_CHILD_FLAGS(zio), NULL));
}
for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) {
ASSERT(dde->dde_lead_zio[p] == NULL);
- ASSERT((int64_t)ddp->ddp_refcnt >= 0);
if (ddp->ddp_phys_birth == 0) {
ASSERT(ddp->ddp_refcnt == 0);
continue;
ASSERT(spa->spa_uberblock.ub_version >= SPA_VERSION_DEDUP);
if (spa->spa_ddt_stat_object == 0) {
- spa->spa_ddt_stat_object = zap_create(ddt->ddt_os,
- DMU_OT_DDT_STATS, DMU_OT_NONE, 0, tx);
- VERIFY(zap_add(ddt->ddt_os, DMU_POOL_DIRECTORY_OBJECT,
- DMU_POOL_DDT_STATS, sizeof (uint64_t), 1,
- &spa->spa_ddt_stat_object, tx) == 0);
+ spa->spa_ddt_stat_object = zap_create_link(ddt->ddt_os,
+ DMU_OT_DDT_STATS, DMU_POOL_DIRECTORY_OBJECT,
+ DMU_POOL_DDT_STATS, tx);
}
while ((dde = avl_destroy_nodes(&ddt->ddt_tree, &cookie)) != NULL) {
}
for (enum ddt_type type = 0; type < DDT_TYPES; type++) {
- uint64_t count = 0;
+ uint64_t add, count = 0;
for (enum ddt_class class = 0; class < DDT_CLASSES; class++) {
if (ddt_object_exists(ddt, type, class)) {
ddt_object_sync(ddt, type, class, tx);
- count += ddt_object_count(ddt, type, class);
+ VERIFY(ddt_object_count(ddt, type, class,
+ &add) == 0);
+ count += add;
}
}
for (enum ddt_class class = 0; class < DDT_CLASSES; class++) {
bcopy(ddt->ddt_histogram, &ddt->ddt_histogram_cache,
sizeof (ddt->ddt_histogram));
+ spa->spa_dedup_dspace = ~0ULL;
}
void
ddt_sync(spa_t *spa, uint64_t txg)
{
+ dsl_scan_t *scn = spa->spa_dsl_pool->dp_scan;
dmu_tx_t *tx;
- zio_t *rio = zio_root(spa, NULL, NULL,
- ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE);
+ zio_t *rio;
ASSERT(spa_syncing_txg(spa) == txg);
tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
+ rio = zio_root(spa, NULL, NULL,
+ ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_SELF_HEAL);
+
+ /*
+ * This function may cause an immediate scan of ddt blocks (see
+ * the comment above dsl_scan_ddt() for details). We set the
+ * scan's root zio here so that we can wait for any scan IOs in
+ * addition to the regular ddt IOs.
+ */
+ ASSERT3P(scn->scn_zio_root, ==, NULL);
+ scn->scn_zio_root = rio;
+
for (enum zio_checksum c = 0; c < ZIO_CHECKSUM_FUNCTIONS; c++) {
ddt_t *ddt = spa->spa_ddt[c];
if (ddt == NULL)
}
(void) zio_wait(rio);
+ scn->scn_zio_root = NULL;
dmu_tx_commit(tx);
}
ddb->ddb_type = 0;
} while (++ddb->ddb_class < DDT_CLASSES);
- return (ENOENT);
+ return (SET_ERROR(ENOENT));
}
+
+#if defined(_KERNEL)
+module_param(zfs_dedup_prefetch, int, 0644);
+MODULE_PARM_DESC(zfs_dedup_prefetch, "Enable prefetching dedup-ed blks");
+#endif
projects / mirror_zfs.git / blobdiff