* CDDL HEADER END
*/
/*
- * Copyright 2009 Sun Microsystems, Inc. All rights reserved.
- * Use is subject to license terms.
+ * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2012 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
#include <sys/spa.h>
#include <sys/zio.h>
#include <sys/dmu_impl.h>
+#include <sys/sa.h>
+#include <sys/sa_impl.h>
#include <sys/callb.h>
-#define SET_BOOKMARK(zb, objset, object, level, blkid) \
-{ \
- (zb)->zb_objset = objset; \
- (zb)->zb_object = object; \
- (zb)->zb_level = level; \
- (zb)->zb_blkid = blkid; \
-}
+int zfs_pd_blks_max = 100;
-struct prefetch_data {
+typedef struct prefetch_data {
kmutex_t pd_mtx;
kcondvar_t pd_cv;
int pd_blks_max;
int pd_flags;
boolean_t pd_cancel;
boolean_t pd_exited;
-};
+} prefetch_data_t;
-struct traverse_data {
+typedef struct traverse_data {
spa_t *td_spa;
uint64_t td_objset;
blkptr_t *td_rootbp;
uint64_t td_min_txg;
+ zbookmark_t *td_resume;
int td_flags;
- struct prefetch_data *td_pfd;
+ prefetch_data_t *td_pfd;
blkptr_cb_t *td_func;
void *td_arg;
-};
+} traverse_data_t;
-static int traverse_dnode(struct traverse_data *td, const dnode_phys_t *dnp,
+static int traverse_dnode(traverse_data_t *td, const dnode_phys_t *dnp,
+ arc_buf_t *buf, uint64_t objset, uint64_t object);
+static void prefetch_dnode_metadata(traverse_data_t *td, const dnode_phys_t *,
arc_buf_t *buf, uint64_t objset, uint64_t object);
-/* ARGSUSED */
-static void
+static int
traverse_zil_block(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t claim_txg)
{
- struct traverse_data *td = arg;
+ traverse_data_t *td = arg;
zbookmark_t zb;
if (bp->blk_birth == 0)
- return;
+ return (0);
if (claim_txg == 0 && bp->blk_birth >= spa_first_txg(td->td_spa))
- return;
+ return (0);
+
+ SET_BOOKMARK(&zb, td->td_objset, ZB_ZIL_OBJECT, ZB_ZIL_LEVEL,
+ bp->blk_cksum.zc_word[ZIL_ZC_SEQ]);
- zb.zb_objset = td->td_objset;
- zb.zb_object = 0;
- zb.zb_level = -1;
- zb.zb_blkid = bp->blk_cksum.zc_word[ZIL_ZC_SEQ];
- VERIFY(0 == td->td_func(td->td_spa, bp, &zb, NULL, td->td_arg));
+ (void) td->td_func(td->td_spa, zilog, bp, NULL, &zb, NULL, td->td_arg);
+
+ return (0);
}
-/* ARGSUSED */
-static void
+static int
traverse_zil_record(zilog_t *zilog, lr_t *lrc, void *arg, uint64_t claim_txg)
{
- struct traverse_data *td = arg;
+ traverse_data_t *td = arg;
if (lrc->lrc_txtype == TX_WRITE) {
lr_write_t *lr = (lr_write_t *)lrc;
zbookmark_t zb;
if (bp->blk_birth == 0)
- return;
+ return (0);
if (claim_txg == 0 || bp->blk_birth < claim_txg)
- return;
+ return (0);
- zb.zb_objset = td->td_objset;
- zb.zb_object = lr->lr_foid;
- zb.zb_level = BP_GET_LEVEL(bp);
- zb.zb_blkid = lr->lr_offset / BP_GET_LSIZE(bp);
- VERIFY(0 == td->td_func(td->td_spa, bp, &zb, NULL, td->td_arg));
+ SET_BOOKMARK(&zb, td->td_objset, lr->lr_foid,
+ ZB_ZIL_LEVEL, lr->lr_offset / BP_GET_LSIZE(bp));
+
+ (void) td->td_func(td->td_spa, zilog, bp, NULL, &zb, NULL,
+ td->td_arg);
}
+ return (0);
}
static void
-traverse_zil(struct traverse_data *td, zil_header_t *zh)
+traverse_zil(traverse_data_t *td, zil_header_t *zh)
{
uint64_t claim_txg = zh->zh_claim_txg;
zilog_t *zilog;
/*
* We only want to visit blocks that have been claimed but not yet
- * replayed (or, in read-only mode, blocks that *would* be claimed).
+ * replayed; plus, in read-only mode, blocks that are already stable.
*/
if (claim_txg == 0 && spa_writeable(td->td_spa))
return;
zil_free(zilog);
}
+typedef enum resume_skip {
+ RESUME_SKIP_ALL,
+ RESUME_SKIP_NONE,
+ RESUME_SKIP_CHILDREN
+} resume_skip_t;
+
+/*
+ * Returns RESUME_SKIP_ALL if td indicates that we are resuming a traversal and
+ * the block indicated by zb does not need to be visited at all. Returns
+ * RESUME_SKIP_CHILDREN if we are resuming a post traversal and we reach the
+ * resume point. This indicates that this block should be visited but not its
+ * children (since they must have been visited in a previous traversal).
+ * Otherwise returns RESUME_SKIP_NONE.
+ */
+static resume_skip_t
+resume_skip_check(traverse_data_t *td, const dnode_phys_t *dnp,
+ const zbookmark_t *zb)
+{
+ if (td->td_resume != NULL && !ZB_IS_ZERO(td->td_resume)) {
+ /*
+ * If we already visited this bp & everything below,
+ * don't bother doing it again.
+ */
+ if (zbookmark_is_before(dnp, zb, td->td_resume))
+ return (RESUME_SKIP_ALL);
+
+ /*
+ * If we found the block we're trying to resume from, zero
+ * the bookmark out to indicate that we have resumed.
+ */
+ ASSERT3U(zb->zb_object, <=, td->td_resume->zb_object);
+ if (bcmp(zb, td->td_resume, sizeof (*zb)) == 0) {
+ bzero(td->td_resume, sizeof (*zb));
+ if (td->td_flags & TRAVERSE_POST)
+ return (RESUME_SKIP_CHILDREN);
+ }
+ }
+ return (RESUME_SKIP_NONE);
+}
+
+static void
+traverse_pause(traverse_data_t *td, const zbookmark_t *zb)
+{
+ ASSERT(td->td_resume != NULL);
+ ASSERT0(zb->zb_level);
+ bcopy(zb, td->td_resume, sizeof (*td->td_resume));
+}
+
+static void
+traverse_prefetch_metadata(traverse_data_t *td,
+ arc_buf_t *pbuf, const blkptr_t *bp, const zbookmark_t *zb)
+{
+ uint32_t flags = ARC_NOWAIT | ARC_PREFETCH;
+
+ if (!(td->td_flags & TRAVERSE_PREFETCH_METADATA))
+ return;
+ /*
+ * If we are in the process of resuming, don't prefetch, because
+ * some children will not be needed (and in fact may have already
+ * been freed).
+ */
+ if (td->td_resume != NULL && !ZB_IS_ZERO(td->td_resume))
+ return;
+ if (BP_IS_HOLE(bp) || bp->blk_birth <= td->td_min_txg)
+ return;
+ if (BP_GET_LEVEL(bp) == 0 && BP_GET_TYPE(bp) != DMU_OT_DNODE)
+ return;
+
+ (void) arc_read(NULL, td->td_spa, bp,
+ pbuf, NULL, NULL, ZIO_PRIORITY_ASYNC_READ,
+ ZIO_FLAG_CANFAIL, &flags, zb);
+}
+
static int
-traverse_visitbp(struct traverse_data *td, const dnode_phys_t *dnp,
- arc_buf_t *pbuf, blkptr_t *bp, const zbookmark_t *zb)
+traverse_visitbp(traverse_data_t *td, const dnode_phys_t *dnp,
+ arc_buf_t *pbuf, const blkptr_t *bp, const zbookmark_t *zb)
{
zbookmark_t czb;
- int err = 0;
+ int err = 0, lasterr = 0;
arc_buf_t *buf = NULL;
- struct prefetch_data *pd = td->td_pfd;
+ prefetch_data_t *pd = td->td_pfd;
+ boolean_t hard = td->td_flags & TRAVERSE_HARD;
+ boolean_t pause = B_FALSE;
- if (bp->blk_birth == 0) {
- err = td->td_func(td->td_spa, NULL, zb, dnp, td->td_arg);
+ switch (resume_skip_check(td, dnp, zb)) {
+ case RESUME_SKIP_ALL:
+ return (0);
+ case RESUME_SKIP_CHILDREN:
+ goto post;
+ case RESUME_SKIP_NONE:
+ break;
+ default:
+ ASSERT(0);
+ }
+
+ if (BP_IS_HOLE(bp)) {
+ err = td->td_func(td->td_spa, NULL, NULL, pbuf, zb, dnp,
+ td->td_arg);
return (err);
}
}
if (td->td_flags & TRAVERSE_PRE) {
- err = td->td_func(td->td_spa, bp, zb, dnp, td->td_arg);
- if (err)
- return (err);
+ err = td->td_func(td->td_spa, NULL, bp, pbuf, zb, dnp,
+ td->td_arg);
+ if (err == TRAVERSE_VISIT_NO_CHILDREN)
+ return (0);
+ if (err == ERESTART)
+ pause = B_TRUE; /* handle pausing at a common point */
+ if (err != 0)
+ goto post;
}
if (BP_GET_LEVEL(bp) > 0) {
blkptr_t *cbp;
int epb = BP_GET_LSIZE(bp) >> SPA_BLKPTRSHIFT;
- err = arc_read(NULL, td->td_spa, bp, pbuf,
+ err = dsl_read(NULL, td->td_spa, bp, pbuf,
arc_getbuf_func, &buf,
ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
if (err)
return (err);
+ cbp = buf->b_data;
+
+ for (i = 0; i < epb; i++) {
+ SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
+ zb->zb_level - 1,
+ zb->zb_blkid * epb + i);
+ traverse_prefetch_metadata(td, buf, &cbp[i], &czb);
+ }
/* recursively visitbp() blocks below this */
- cbp = buf->b_data;
- for (i = 0; i < epb; i++, cbp++) {
+ for (i = 0; i < epb; i++) {
SET_BOOKMARK(&czb, zb->zb_objset, zb->zb_object,
zb->zb_level - 1,
zb->zb_blkid * epb + i);
- err = traverse_visitbp(td, dnp, buf, cbp, &czb);
- if (err)
- break;
+ err = traverse_visitbp(td, dnp, buf, &cbp[i], &czb);
+ if (err) {
+ if (!hard)
+ break;
+ lasterr = err;
+ }
}
} else if (BP_GET_TYPE(bp) == DMU_OT_DNODE) {
uint32_t flags = ARC_WAIT;
int i;
int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT;
- err = arc_read(NULL, td->td_spa, bp, pbuf,
+ err = dsl_read(NULL, td->td_spa, bp, pbuf,
arc_getbuf_func, &buf,
ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
if (err)
return (err);
+ dnp = buf->b_data;
+
+ for (i = 0; i < epb; i++) {
+ prefetch_dnode_metadata(td, &dnp[i], buf, zb->zb_objset,
+ zb->zb_blkid * epb + i);
+ }
/* recursively visitbp() blocks below this */
- dnp = buf->b_data;
- for (i = 0; i < epb && err == 0; i++, dnp++) {
- err = traverse_dnode(td, dnp, buf, zb->zb_objset,
+ for (i = 0; i < epb; i++) {
+ err = traverse_dnode(td, &dnp[i], buf, zb->zb_objset,
zb->zb_blkid * epb + i);
- if (err)
- break;
+ if (err) {
+ if (!hard)
+ break;
+ lasterr = err;
+ }
}
} else if (BP_GET_TYPE(bp) == DMU_OT_OBJSET) {
uint32_t flags = ARC_WAIT;
objset_phys_t *osp;
dnode_phys_t *dnp;
- err = arc_read_nolock(NULL, td->td_spa, bp,
+ err = dsl_read_nolock(NULL, td->td_spa, bp,
arc_getbuf_func, &buf,
ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL, &flags, zb);
if (err)
return (err);
osp = buf->b_data;
- traverse_zil(td, &osp->os_zil_header);
-
dnp = &osp->os_meta_dnode;
- err = traverse_dnode(td, dnp, buf, zb->zb_objset, 0);
+ prefetch_dnode_metadata(td, dnp, buf, zb->zb_objset,
+ DMU_META_DNODE_OBJECT);
+ if (arc_buf_size(buf) >= sizeof (objset_phys_t)) {
+ prefetch_dnode_metadata(td, &osp->os_userused_dnode,
+ buf, zb->zb_objset, DMU_USERUSED_OBJECT);
+ prefetch_dnode_metadata(td, &osp->os_groupused_dnode,
+ buf, zb->zb_objset, DMU_USERUSED_OBJECT);
+ }
+
+ err = traverse_dnode(td, dnp, buf, zb->zb_objset,
+ DMU_META_DNODE_OBJECT);
+ if (err && hard) {
+ lasterr = err;
+ err = 0;
+ }
if (err == 0 && arc_buf_size(buf) >= sizeof (objset_phys_t)) {
dnp = &osp->os_userused_dnode;
err = traverse_dnode(td, dnp, buf, zb->zb_objset,
DMU_USERUSED_OBJECT);
}
+ if (err && hard) {
+ lasterr = err;
+ err = 0;
+ }
if (err == 0 && arc_buf_size(buf) >= sizeof (objset_phys_t)) {
dnp = &osp->os_groupused_dnode;
err = traverse_dnode(td, dnp, buf, zb->zb_objset,
if (buf)
(void) arc_buf_remove_ref(buf, &buf);
- if (err == 0 && (td->td_flags & TRAVERSE_POST))
- err = td->td_func(td->td_spa, bp, zb, dnp, td->td_arg);
+post:
+ if (err == 0 && lasterr == 0 && (td->td_flags & TRAVERSE_POST)) {
+ err = td->td_func(td->td_spa, NULL, bp, pbuf, zb, dnp,
+ td->td_arg);
+ if (err == ERESTART)
+ pause = B_TRUE;
+ }
- return (err);
+ if (pause && td->td_resume != NULL) {
+ ASSERT3U(err, ==, ERESTART);
+ ASSERT(!hard);
+ traverse_pause(td, zb);
+ }
+
+ return (err != 0 ? err : lasterr);
+}
+
+static void
+prefetch_dnode_metadata(traverse_data_t *td, const dnode_phys_t *dnp,
+ arc_buf_t *buf, uint64_t objset, uint64_t object)
+{
+ int j;
+ zbookmark_t czb;
+
+ for (j = 0; j < dnp->dn_nblkptr; j++) {
+ SET_BOOKMARK(&czb, objset, object, dnp->dn_nlevels - 1, j);
+ traverse_prefetch_metadata(td, buf, &dnp->dn_blkptr[j], &czb);
+ }
+
+ if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
+ SET_BOOKMARK(&czb, objset, object, 0, DMU_SPILL_BLKID);
+ traverse_prefetch_metadata(td, buf, &dnp->dn_spill, &czb);
+ }
}
static int
-traverse_dnode(struct traverse_data *td, const dnode_phys_t *dnp,
+traverse_dnode(traverse_data_t *td, const dnode_phys_t *dnp,
arc_buf_t *buf, uint64_t objset, uint64_t object)
{
- int j, err = 0;
+ int j, err = 0, lasterr = 0;
zbookmark_t czb;
+ boolean_t hard = (td->td_flags & TRAVERSE_HARD);
for (j = 0; j < dnp->dn_nblkptr; j++) {
SET_BOOKMARK(&czb, objset, object, dnp->dn_nlevels - 1, j);
- err = traverse_visitbp(td, dnp, buf,
- (blkptr_t *)&dnp->dn_blkptr[j], &czb);
- if (err)
- break;
+ err = traverse_visitbp(td, dnp, buf, &dnp->dn_blkptr[j], &czb);
+ if (err) {
+ if (!hard)
+ break;
+ lasterr = err;
+ }
}
- return (err);
+
+ if (dnp->dn_flags & DNODE_FLAG_SPILL_BLKPTR) {
+ SET_BOOKMARK(&czb, objset, object, 0, DMU_SPILL_BLKID);
+ err = traverse_visitbp(td, dnp, buf, &dnp->dn_spill, &czb);
+ if (err) {
+ if (!hard)
+ return (err);
+ lasterr = err;
+ }
+ }
+ return (err != 0 ? err : lasterr);
}
/* ARGSUSED */
static int
-traverse_prefetcher(spa_t *spa, blkptr_t *bp, const zbookmark_t *zb,
- const dnode_phys_t *dnp, void *arg)
+traverse_prefetcher(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
+ arc_buf_t *pbuf, const zbookmark_t *zb, const dnode_phys_t *dnp,
+ void *arg)
{
- struct prefetch_data *pfd = arg;
+ prefetch_data_t *pfd = arg;
uint32_t aflags = ARC_NOWAIT | ARC_PREFETCH;
ASSERT(pfd->pd_blks_fetched >= 0);
return (EINTR);
if (bp == NULL || !((pfd->pd_flags & TRAVERSE_PREFETCH_DATA) ||
- BP_GET_TYPE(bp) == DMU_OT_DNODE || BP_GET_LEVEL(bp) > 0))
+ BP_GET_TYPE(bp) == DMU_OT_DNODE || BP_GET_LEVEL(bp) > 0) ||
+ BP_GET_TYPE(bp) == DMU_OT_INTENT_LOG)
return (0);
mutex_enter(&pfd->pd_mtx);
cv_broadcast(&pfd->pd_cv);
mutex_exit(&pfd->pd_mtx);
- (void) arc_read_nolock(NULL, spa, bp, NULL, NULL,
+ (void) dsl_read(NULL, spa, bp, pbuf, NULL, NULL,
ZIO_PRIORITY_ASYNC_READ,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE,
&aflags, zb);
static void
traverse_prefetch_thread(void *arg)
{
- struct traverse_data *td_main = arg;
- struct traverse_data td = *td_main;
+ traverse_data_t *td_main = arg;
+ traverse_data_t td = *td_main;
zbookmark_t czb;
td.td_func = traverse_prefetcher;
td.td_arg = td_main->td_pfd;
td.td_pfd = NULL;
- SET_BOOKMARK(&czb, td.td_objset, 0, -1, 0);
+ SET_BOOKMARK(&czb, td.td_objset,
+ ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
(void) traverse_visitbp(&td, NULL, NULL, td.td_rootbp, &czb);
mutex_enter(&td_main->td_pfd->pd_mtx);
* in syncing context).
*/
static int
-traverse_impl(spa_t *spa, uint64_t objset, blkptr_t *rootbp,
- uint64_t txg_start, int flags, blkptr_cb_t func, void *arg)
+traverse_impl(spa_t *spa, dsl_dataset_t *ds, uint64_t objset, blkptr_t *rootbp,
+ uint64_t txg_start, zbookmark_t *resume, int flags,
+ blkptr_cb_t func, void *arg)
{
- struct traverse_data td;
- struct prefetch_data pd = { 0 };
- zbookmark_t czb;
+ traverse_data_t *td;
+ prefetch_data_t *pd;
+ zbookmark_t *czb;
int err;
- td.td_spa = spa;
- td.td_objset = objset;
- td.td_rootbp = rootbp;
- td.td_min_txg = txg_start;
- td.td_func = func;
- td.td_arg = arg;
- td.td_pfd = &pd;
- td.td_flags = flags;
-
- pd.pd_blks_max = 100;
- pd.pd_flags = flags;
- mutex_init(&pd.pd_mtx, NULL, MUTEX_DEFAULT, NULL);
- cv_init(&pd.pd_cv, NULL, CV_DEFAULT, NULL);
-
- if (!(flags & TRAVERSE_PREFETCH) ||
+ ASSERT(ds == NULL || objset == ds->ds_object);
+ ASSERT(!(flags & TRAVERSE_PRE) || !(flags & TRAVERSE_POST));
+
+ /*
+ * The data prefetching mechanism (the prefetch thread) is incompatible
+ * with resuming from a bookmark.
+ */
+ ASSERT(resume == NULL || !(flags & TRAVERSE_PREFETCH_DATA));
+
+ td = kmem_alloc(sizeof(traverse_data_t), KM_PUSHPAGE);
+ pd = kmem_zalloc(sizeof(prefetch_data_t), KM_PUSHPAGE);
+ czb = kmem_alloc(sizeof(zbookmark_t), KM_PUSHPAGE);
+
+ td->td_spa = spa;
+ td->td_objset = objset;
+ td->td_rootbp = rootbp;
+ td->td_min_txg = txg_start;
+ td->td_resume = resume;
+ td->td_func = func;
+ td->td_arg = arg;
+ td->td_pfd = pd;
+ td->td_flags = flags;
+
+ pd->pd_blks_max = zfs_pd_blks_max;
+ pd->pd_flags = flags;
+ mutex_init(&pd->pd_mtx, NULL, MUTEX_DEFAULT, NULL);
+ cv_init(&pd->pd_cv, NULL, CV_DEFAULT, NULL);
+
+ /* See comment on ZIL traversal in dsl_scan_visitds. */
+ if (ds != NULL && !dsl_dataset_is_snapshot(ds)) {
+ objset_t *os;
+
+ err = dmu_objset_from_ds(ds, &os);
+ if (err)
+ return (err);
+
+ traverse_zil(td, &os->os_zil_header);
+ }
+
+ if (!(flags & TRAVERSE_PREFETCH_DATA) ||
0 == taskq_dispatch(system_taskq, traverse_prefetch_thread,
- &td, TQ_NOQUEUE))
- pd.pd_exited = B_TRUE;
+ td, TQ_NOQUEUE))
+ pd->pd_exited = B_TRUE;
+
+ SET_BOOKMARK(czb, td->td_objset,
+ ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
+ err = traverse_visitbp(td, NULL, NULL, rootbp, czb);
- SET_BOOKMARK(&czb, objset, 0, -1, 0);
- err = traverse_visitbp(&td, NULL, NULL, rootbp, &czb);
+ mutex_enter(&pd->pd_mtx);
+ pd->pd_cancel = B_TRUE;
+ cv_broadcast(&pd->pd_cv);
+ while (!pd->pd_exited)
+ cv_wait(&pd->pd_cv, &pd->pd_mtx);
+ mutex_exit(&pd->pd_mtx);
- mutex_enter(&pd.pd_mtx);
- pd.pd_cancel = B_TRUE;
- cv_broadcast(&pd.pd_cv);
- while (!pd.pd_exited)
- cv_wait(&pd.pd_cv, &pd.pd_mtx);
- mutex_exit(&pd.pd_mtx);
+ mutex_destroy(&pd->pd_mtx);
+ cv_destroy(&pd->pd_cv);
- mutex_destroy(&pd.pd_mtx);
- cv_destroy(&pd.pd_cv);
+ kmem_free(czb, sizeof(zbookmark_t));
+ kmem_free(pd, sizeof(struct prefetch_data));
+ kmem_free(td, sizeof(struct traverse_data));
return (err);
}
traverse_dataset(dsl_dataset_t *ds, uint64_t txg_start, int flags,
blkptr_cb_t func, void *arg)
{
- return (traverse_impl(ds->ds_dir->dd_pool->dp_spa, ds->ds_object,
- &ds->ds_phys->ds_bp, txg_start, flags, func, arg));
+ return (traverse_impl(ds->ds_dir->dd_pool->dp_spa, ds, ds->ds_object,
+ &ds->ds_phys->ds_bp, txg_start, NULL, flags, func, arg));
+}
+
+int
+traverse_dataset_destroyed(spa_t *spa, blkptr_t *blkptr,
+ uint64_t txg_start, zbookmark_t *resume, int flags,
+ blkptr_cb_t func, void *arg)
+{
+ return (traverse_impl(spa, NULL, ZB_DESTROYED_OBJSET,
+ blkptr, txg_start, resume, flags, func, arg));
}
/*
* NB: pool must not be changing on-disk (eg, from zdb or sync context).
*/
int
-traverse_pool(spa_t *spa, blkptr_cb_t func, void *arg)
+traverse_pool(spa_t *spa, uint64_t txg_start, int flags,
+ blkptr_cb_t func, void *arg)
{
- int err;
+ int err, lasterr = 0;
uint64_t obj;
dsl_pool_t *dp = spa_get_dsl(spa);
objset_t *mos = dp->dp_meta_objset;
+ boolean_t hard = (flags & TRAVERSE_HARD);
/* visit the MOS */
- err = traverse_impl(spa, 0, spa_get_rootblkptr(spa),
- 0, TRAVERSE_PRE, func, arg);
+ err = traverse_impl(spa, NULL, 0, spa_get_rootblkptr(spa),
+ txg_start, NULL, flags, func, arg);
if (err)
return (err);
/* visit each dataset */
- for (obj = 1; err == 0; err = dmu_object_next(mos, &obj, FALSE, 0)) {
+ for (obj = 1; err == 0 || (err != ESRCH && hard);
+ err = dmu_object_next(mos, &obj, FALSE, txg_start)) {
dmu_object_info_t doi;
err = dmu_object_info(mos, obj, &doi);
- if (err)
- return (err);
+ if (err) {
+ if (!hard)
+ return (err);
+ lasterr = err;
+ continue;
+ }
if (doi.doi_type == DMU_OT_DSL_DATASET) {
dsl_dataset_t *ds;
+ uint64_t txg = txg_start;
+
rw_enter(&dp->dp_config_rwlock, RW_READER);
err = dsl_dataset_hold_obj(dp, obj, FTAG, &ds);
rw_exit(&dp->dp_config_rwlock);
- if (err)
- return (err);
- err = traverse_dataset(ds,
- ds->ds_phys->ds_prev_snap_txg, TRAVERSE_PRE,
- func, arg);
+ if (err) {
+ if (!hard)
+ return (err);
+ lasterr = err;
+ continue;
+ }
+ if (ds->ds_phys->ds_prev_snap_txg > txg)
+ txg = ds->ds_phys->ds_prev_snap_txg;
+ err = traverse_dataset(ds, txg, flags, func, arg);
dsl_dataset_rele(ds, FTAG);
- if (err)
- return (err);
+ if (err) {
+ if (!hard)
+ return (err);
+ lasterr = err;
+ }
}
}
if (err == ESRCH)
err = 0;
- return (err);
+ return (err != 0 ? err : lasterr);
}
+
+#if defined(_KERNEL) && defined(HAVE_SPL)
+EXPORT_SYMBOL(traverse_dataset);
+EXPORT_SYMBOL(traverse_pool);
+
+module_param(zfs_pd_blks_max, int, 0644);
+MODULE_PARM_DESC(zfs_pd_blks_max, "Max number of blocks to prefetch");
+#endif
git://git.camperquake.de / zfs.git / blobdiff