Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Raidz time dependent geometry (misc improvements) #15

Merged
merged 3 commits into from
Nov 3, 2020
Merged

Raidz time dependent geometry (misc improvements) #15

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
862f908
Fix reflow txgs tree creation in case if expansion is progress
fuporovvStack Sep 9, 2020
5bb5bed
Restore vdev_raidz_io_verify()
fuporovvStack Sep 9, 2020
43d4d74
Improve raidz expansion test
fuporovvStack Sep 9, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
115 changes: 73 additions & 42 deletions module/zfs/vdev_raidz.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1921,7 +1921,30 @@ vdev_raidz_close(vdev_t *vd)
}

static uint64_t
vdev_raidz_asize(vdev_t *vd, uint64_t psize)
vdev_raidz_get_logical_width(vdev_raidz_t *vdrz, zio_t *zio)
{
#if 1
reflow_node_t lookup = {
.re_txg = BP_PHYSICAL_BIRTH(zio->io_bp)
};
avl_index_t where;

reflow_node_t *re = avl_find(&vdrz->vd_expand_txgs, &lookup, &where);
if (re != NULL)
return (re->re_logical_width);

re = avl_nearest(&vdrz->vd_expand_txgs, where, AVL_BEFORE);
if (re == NULL)
return (vdrz->vd_original_width);

return (re->re_logical_width);
#else
return (vdrz->vd_original_width);
#endif
}

static uint64_t
vdev_raidz_asize_impl(zio_t *zio, vdev_t *vd, uint64_t psize)
{
vdev_raidz_t *vdrz = vd->vdev_tsd;
uint64_t asize;
Expand All @@ -1930,6 +1953,9 @@ vdev_raidz_asize(vdev_t *vd, uint64_t psize)
uint64_t cols = vdrz->vd_original_width;
uint64_t nparity = vdrz->vd_nparity;

if (zio != NULL)
cols = vdev_raidz_get_logical_width(vdrz, zio);

asize = ((psize - 1) >> ashift) + 1;
asize += nparity * ((asize + cols - nparity - 1) / (cols - nparity));
asize = roundup(asize, nparity + 1) << ashift;
Expand All @@ -1943,6 +1969,12 @@ vdev_raidz_asize(vdev_t *vd, uint64_t psize)
return (asize);
}

static uint64_t
vdev_raidz_asize(vdev_t *vd, uint64_t psize)
{
return (vdev_raidz_asize_impl(NULL, vd, psize));
}

static void
vdev_raidz_child_done(zio_t *zio)
{
Expand All @@ -1961,9 +1993,8 @@ vdev_raidz_shadow_child_done(zio_t *zio)
rc->rc_shadow_error = zio->io_error;
}

#if 0
static void
vdev_raidz_io_verify(zio_t *zio, raidz_row_t *rr, int col)
vdev_raidz_io_verify(zio_t *zio, raidz_map_t *rm, raidz_row_t *rr, int col)
{
#ifdef ZFS_DEBUG
vdev_t *vd = zio->io_vd;
Expand All @@ -1972,29 +2003,34 @@ vdev_raidz_io_verify(zio_t *zio, raidz_row_t *rr, int col)
range_seg64_t logical_rs, physical_rs;
logical_rs.rs_start = zio->io_offset;
logical_rs.rs_end = logical_rs.rs_start +
vdev_raidz_asize(zio->io_vd, zio->io_size);
vdev_raidz_asize_impl(zio, zio->io_vd, zio->io_size);

raidz_col_t *rc = &rr->rr_col[col];
vdev_t *cvd = vd->vdev_child[rc->rc_devidx];

vdev_xlate(cvd, &logical_rs, &physical_rs);
ASSERT3U(rc->rc_offset, ==, physical_rs.rs_start);
if (rm->rm_nrows == 1)
ASSERT3U(rc->rc_offset, ==, physical_rs.rs_start);
else
ASSERT3U(rc->rc_offset, >=, physical_rs.rs_start);

ASSERT3U(rc->rc_offset, <, physical_rs.rs_end);

/*
* It would be nice to assert that rs_end is equal
* to rc_offset + rc_size but there might be an
* optional I/O at the end that is not accounted in
* rc_size.
*/
if (physical_rs.rs_end > rc->rc_offset + rc->rc_size) {
ASSERT3U(physical_rs.rs_end, ==, rc->rc_offset +
rc->rc_size + (1 << tvd->vdev_ashift));
if (rm->rm_nrows == 1)
ASSERT3U(physical_rs.rs_end, ==, rc->rc_offset +
rc->rc_size + (1 << tvd->vdev_ashift));
} else {
ASSERT3U(physical_rs.rs_end, ==, rc->rc_offset + rc->rc_size);
}
#endif
}
#endif

static void
vdev_raidz_io_start_write(zio_t *zio, raidz_row_t *rr)
Expand All @@ -2010,6 +2046,11 @@ vdev_raidz_io_start_write(zio_t *zio, raidz_row_t *rr)
continue;
vdev_t *cvd = vd->vdev_child[rc->rc_devidx];

/*
* Verify physical to logical translation.
*/
vdev_raidz_io_verify(zio, rm, rr, c);

zio_nowait(zio_vdev_child_io(zio, NULL, cvd,
rc->rc_offset, rc->rc_abd, rc->rc_size,
zio->io_type, zio->io_priority, 0,
Expand Down Expand Up @@ -2092,29 +2133,6 @@ vdev_raidz_io_start_read(zio_t *zio, raidz_row_t *rr, boolean_t forceparity)
}
}

static uint64_t
vdev_raidz_get_logical_width(vdev_raidz_t *vdrz, zio_t *zio)
{
#if 1
reflow_node_t lookup = {
.re_txg = BP_PHYSICAL_BIRTH(zio->io_bp)
};
avl_index_t where;

reflow_node_t *re = avl_find(&vdrz->vd_expand_txgs, &lookup, &where);
if (re != NULL)
return (re->re_logical_width);

re = avl_nearest(&vdrz->vd_expand_txgs, where, AVL_BEFORE);
if (re == NULL)
return (vdrz->vd_original_width);

return (re->re_logical_width);
#else
return (vdrz->vd_original_width);
#endif
}

/*
* Start an IO operation on a RAIDZ VDev
*
Expand Down Expand Up @@ -3027,8 +3045,12 @@ vdev_raidz_xlate(vdev_t *cvd, const range_seg64_t *in, range_seg64_t *res)
vdev_t *raidvd = cvd->vdev_parent;
ASSERT(raidvd->vdev_ops == &vdev_raidz_ops);

/* XXX deal with different logical and physical widths */
uint64_t width = raidvd->vdev_children;
vdev_raidz_t *vdrz = raidvd->vdev_tsd;
uint64_t children = vdrz->vd_physical_width;
if (in->rs_start > vdrz->vn_vre.vre_offset_phys)
children--;

uint64_t width = children;
uint64_t tgt_col = cvd->vdev_id;
uint64_t ashift = raidvd->vdev_top->vdev_ashift;

Expand Down Expand Up @@ -3580,6 +3602,7 @@ vdev_raidz_get_tsd(spa_t *spa, nvlist_t *nv)
uint64_t nparity, *txgs;
uint_t txgs_size;
vdev_raidz_t *vdrz = kmem_zalloc(sizeof (*vdrz), KM_SLEEP);
boolean_t reflow_in_progress = B_FALSE;

vdrz->vn_vre.vre_vdev_id = -1;
vdrz->vn_vre.vre_offset = UINT64_MAX;
Expand All @@ -3602,6 +3625,18 @@ vdev_raidz_get_tsd(spa_t *spa, nvlist_t *nv)
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
&vdrz->vn_vre.vre_vdev_id);

if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_RAIDZ_EXPAND_OFFSET,
&vdrz->vn_vre.vre_offset_phys) == 0) {
vdrz->vn_vre.vre_offset = vdrz->vn_vre.vre_offset_phys;
if (vdrz->vn_vre.vre_offset != UINT64_MAX)
reflow_in_progress = B_TRUE;

/*
* vdev_load() will set spa_raidz_expand.
*/
}


avl_create(&vdrz->vd_expand_txgs, vedv_raidz_reflow_compare,
sizeof (reflow_node_t), offsetof(reflow_node_t, re_link));

Expand All @@ -3612,22 +3647,18 @@ vdev_raidz_get_tsd(spa_t *spa, nvlist_t *nv)
reflow_node_t *re = kmem_zalloc(sizeof (*re), KM_SLEEP);
re->re_txg = txgs[txgs_size - i - 1];
re->re_logical_width = vdrz->vd_physical_width - i;

if (reflow_in_progress)
re->re_logical_width--;

avl_add(&vdrz->vd_expand_txgs, re);
}

vdrz->vd_original_width = vdrz->vd_physical_width - txgs_size;
}
if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_RAIDZ_EXPAND_OFFSET,
&vdrz->vn_vre.vre_offset_phys) == 0) {
vdrz->vn_vre.vre_offset = vdrz->vn_vre.vre_offset_phys;
if (reflow_in_progress)
vdrz->vd_original_width--;

/*
* vdev_load() will set spa_raidz_expand.
*/
}


if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
&nparity) == 0) {
if (nparity == 0 || nparity > VDEV_RAIDZ_MAXPARITY)
Expand Down
9 changes: 7 additions & 2 deletions tests/zfs-tests/tests/functional/raidz/raidz_expand_002_pos.ksh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -97,12 +97,15 @@ for nparity in 1 2 3; do
typeset pool_size=$(get_pool_prop size $pool)

for disk in ${disks[$(($nparity+1))+1..$devs]}; do
log_must dd if=/dev/urandom of=/${pool}/FILE-$RANDOM bs=1M \
count=128

log_must zpool attach $pool ${raid}-0 $disk

wait_expand_completion

log_must zpool export $pool
log_must zpool import $opts -d $dir $pool
# Wait some time for pool size increase
sleep 5

typeset disk_attached=$(get_disklist $pool | grep $disk)
if [[ -z $disk_attached ]]; then
Expand All @@ -114,6 +117,8 @@ for nparity in 1 2 3; do
log_fail "pool $pool not expanded"
fi

verify_pool $pool

pool_size=$expand_size
done

Expand Down