c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2025-01-13 19:50:25 +03:00
Code Issues Packages Projects Releases Wiki Activity
715c996d3b
mirror_zfs/lib/libzfs/libzfs_iter.c
Paul Dagnelie ca5777793e Reduce loaded range tree memory usage 
This patch implements a new tree structure for ZFS, and uses it to 
store range trees more efficiently.

The new structure is approximately a B-tree, though there are some 
small differences from the usual characterizations. The tree has core 
nodes and leaf nodes; each contain data elements, which the elements 
in the core nodes acting as separators between its children. The 
difference between core and leaf nodes is that the core nodes have an 
array of children, while leaf nodes don't. Every node in the tree may 
be only partially full; in most cases, they are all at least 50% full 
(in terms of element count) except for the root node, which can be 
less full. Underfull nodes will steal from their neighbors or merge to 
remain full enough, while overfull nodes will split in two. The data 
elements are contained in tree-controlled buffers; they are copied 
into these on insertion, and overwritten on deletion. This means that 
the elements are not independently allocated, which reduces overhead, 
but also means they can't be shared between trees (and also that 
pointers to them are only valid until a side-effectful tree operation 
occurs). The overhead varies based on how dense the tree is, but is 
usually on the order of about 50% of the element size; the per-node 
overheads are very small, and so don't make a significant difference. 
The trees can accept arbitrary records; they accept a size and a 
comparator to allow them to be used for a variety of purposes.

The new trees replace the AVL trees used in the range trees today. 
Currently, the range_seg_t structure contains three 8 byte integers 
of payload and two 24 byte avl_tree_node_ts to handle its storage in 
both an offset-sorted tree and a size-sorted tree (total size: 64 
bytes). In the new model, the range seg structures are usually two 4 
byte integers, but a separate one needs to exist for the size-sorted 
and offset-sorted tree. Between the raw size, the 50% overhead, and 
the double storage, the new btrees are expected to use 8*1.5*2 = 24 
bytes per record, or 33.3% as much memory as the AVL trees (this is 
for the purposes of storing metaslab range trees; for other purposes, 
like scrubs, they use ~50% as much memory).

We reduced the size of the payload in the range segments by teaching 
range trees about starting offsets and shifts; since metaslabs have a 
fixed starting offset, and they all operate in terms of disk sectors, 
we can store the ranges using 4-byte integers as long as the size of 
the metaslab divided by the sector size is less than 2^32. For 512-byte
sectors, this is a 2^41 (or 2TB) metaslab, which with the default
settings corresponds to a 256PB disk. 4k sector disks can handle 
metaslabs up to 2^46 bytes, or 2^63 byte disks. Since we do not 
anticipate disks of this size in the near future, there should be 
almost no cases where metaslabs need 64-byte integers to store their 
ranges. We do still have the capability to store 64-byte integer ranges 
to account for cases where we are storing per-vdev (or per-dnode) trees, 
which could reasonably go above the limits discussed. We also do not 
store fill information in the compact version of the node, since it 
is only used for sorted scrub.

We also optimized the metaslab loading process in various other ways
to offset some inefficiencies in the btree model. While individual
operations (find, insert, remove_from) are faster for the btree than 
they are for the avl tree, remove usually requires a find operation, 
while in the AVL tree model the element itself suffices. Some clever 
changes actually caused an overall speedup in metaslab loading; we use 
approximately 40% less cpu to load metaslabs in our tests on Illumos.

Another memory and performance optimization was achieved by changing 
what is stored in the size-sorted trees. When a disk is heavily 
fragmented, the df algorithm used by default in ZFS will almost always 
find a number of small regions in its initial cursor-based search; it 
will usually only fall back to the size-sorted tree to find larger 
regions. If we increase the size of the cursor-based search slightly, 
and don't store segments that are smaller than a tunable size floor 
in the size-sorted tree, we can further cut memory usage down to 
below 20% of what the AVL trees store. This also results in further 
reductions in CPU time spent loading metaslabs.

The 16KiB size floor was chosen because it results in substantial memory 
usage reduction while not usually resulting in situations where we can't 
find an appropriate chunk with the cursor and are forced to use an 
oversized chunk from the size-sorted tree. In addition, even if we do 
have to use an oversized chunk from the size-sorted tree, the chunk 
would be too small to use for ZIL allocations, so it isn't as big of a 
loss as it might otherwise be. And often, more small allocations will 
follow the initial one, and the cursor search will now find the 
remainder of the chunk we didn't use all of and use it for subsequent 
allocations. Practical testing has shown little or no change in 
fragmentation as a result of this change.

If the size-sorted tree becomes empty while the offset sorted one still 
has entries, it will load all the entries from the offset sorted tree 
and disregard the size floor until it is unloaded again. This operation 
occurs rarely with the default setting, only on incredibly thoroughly 
fragmented pools.

There are some other small changes to zdb to teach it to handle btrees, 
but nothing major.
                                           
Reviewed-by: George Wilson <gwilson@delphix.com>
Reviewed-by: Matt Ahrens <matt@delphix.com>
Reviewed by: Sebastien Roy seb@delphix.com
Reviewed-by: Igor Kozhukhov <igor@dilos.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Paul Dagnelie <pcd@delphix.com>
Closes #9181
2019-10-09 10:36:03 -07:00

601 lines
14 KiB
C
Raw Blame History

/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013, 2019 by Delphix. All rights reserved.
* Copyright 2014 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2019 Datto Inc.
*/
#include <stdio.h>
#include <stdlib.h>
#include <strings.h>
#include <unistd.h>
#include <stddef.h>
#include <libintl.h>
#include <libzfs.h>
#include <libzutil.h>
#include <sys/mntent.h>
#include "libzfs_impl.h"
int
zfs_iter_clones(zfs_handle_t *zhp, zfs_iter_f func, void *data)
{
nvlist_t *nvl = zfs_get_clones_nvl(zhp);
nvpair_t *pair;
if (nvl == NULL)
return (0);
for (pair = nvlist_next_nvpair(nvl, NULL); pair != NULL;
pair = nvlist_next_nvpair(nvl, pair)) {
zfs_handle_t *clone = zfs_open(zhp->zfs_hdl, nvpair_name(pair),
ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME);
if (clone != NULL) {
int err = func(clone, data);
if (err != 0)
return (err);
}
}
return (0);
}
static int
zfs_do_list_ioctl(zfs_handle_t *zhp, int arg, zfs_cmd_t *zc)
{
int rc;
uint64_t orig_cookie;
orig_cookie = zc->zc_cookie;
top:
(void) strlcpy(zc->zc_name, zhp->zfs_name, sizeof (zc->zc_name));
rc = ioctl(zhp->zfs_hdl->libzfs_fd, arg, zc);
if (rc == -1) {
switch (errno) {
case ENOMEM:
/* expand nvlist memory and try again */
if (zcmd_expand_dst_nvlist(zhp->zfs_hdl, zc) != 0) {
zcmd_free_nvlists(zc);
return (-1);
}
zc->zc_cookie = orig_cookie;
goto top;
/*
* An errno value of ESRCH indicates normal completion.
* If ENOENT is returned, then the underlying dataset
* has been removed since we obtained the handle.
*/
case ESRCH:
case ENOENT:
rc = 1;
break;
default:
rc = zfs_standard_error(zhp->zfs_hdl, errno,
dgettext(TEXT_DOMAIN,
"cannot iterate filesystems"));
break;
}
}
return (rc);
}
/*
* Iterate over all child filesystems
*/
int
zfs_iter_filesystems(zfs_handle_t *zhp, zfs_iter_f func, void *data)
{
zfs_cmd_t zc = {"\0"};
zfs_handle_t *nzhp;
int ret;
if (zhp->zfs_type != ZFS_TYPE_FILESYSTEM)
return (0);
if (zcmd_alloc_dst_nvlist(zhp->zfs_hdl, &zc, 0) != 0)
return (-1);
while ((ret = zfs_do_list_ioctl(zhp, ZFS_IOC_DATASET_LIST_NEXT,
&zc)) == 0) {
/*
* Silently ignore errors, as the only plausible explanation is
* that the pool has since been removed.
*/
if ((nzhp = make_dataset_handle_zc(zhp->zfs_hdl,
&zc)) == NULL) {
continue;
}
if ((ret = func(nzhp, data)) != 0) {
zcmd_free_nvlists(&zc);
return (ret);
}
}
zcmd_free_nvlists(&zc);
return ((ret < 0) ? ret : 0);
}
/*
* Iterate over all snapshots
*/
int
zfs_iter_snapshots(zfs_handle_t *zhp, boolean_t simple, zfs_iter_f func,
void *data, uint64_t min_txg, uint64_t max_txg)
{
zfs_cmd_t zc = {"\0"};
zfs_handle_t *nzhp;
int ret;
nvlist_t *range_nvl = NULL;
if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT ||
zhp->zfs_type == ZFS_TYPE_BOOKMARK)
return (0);
zc.zc_simple = simple;
if (zcmd_alloc_dst_nvlist(zhp->zfs_hdl, &zc, 0) != 0)
return (-1);
if (min_txg != 0) {
range_nvl = fnvlist_alloc();
fnvlist_add_uint64(range_nvl, SNAP_ITER_MIN_TXG, min_txg);
}
if (max_txg != 0) {
if (range_nvl == NULL)
range_nvl = fnvlist_alloc();
fnvlist_add_uint64(range_nvl, SNAP_ITER_MAX_TXG, max_txg);
}
if (range_nvl != NULL &&
zcmd_write_src_nvlist(zhp->zfs_hdl, &zc, range_nvl) != 0) {
zcmd_free_nvlists(&zc);
fnvlist_free(range_nvl);
return (-1);
}
while ((ret = zfs_do_list_ioctl(zhp, ZFS_IOC_SNAPSHOT_LIST_NEXT,
&zc)) == 0) {
if (simple)
nzhp = make_dataset_simple_handle_zc(zhp, &zc);
else
nzhp = make_dataset_handle_zc(zhp->zfs_hdl, &zc);
if (nzhp == NULL)
continue;
if ((ret = func(nzhp, data)) != 0) {
zcmd_free_nvlists(&zc);
fnvlist_free(range_nvl);
return (ret);
}
}
zcmd_free_nvlists(&zc);
fnvlist_free(range_nvl);
return ((ret < 0) ? ret : 0);
}
/*
* Iterate over all bookmarks
*/
int
zfs_iter_bookmarks(zfs_handle_t *zhp, zfs_iter_f func, void *data)
{
zfs_handle_t *nzhp;
nvlist_t *props = NULL;
nvlist_t *bmarks = NULL;
int err;
nvpair_t *pair;
if ((zfs_get_type(zhp) & (ZFS_TYPE_SNAPSHOT | ZFS_TYPE_BOOKMARK)) != 0)
return (0);
/* Setup the requested properties nvlist. */
props = fnvlist_alloc();
for (zfs_prop_t p = 0; p < ZFS_NUM_PROPS; p++) {
if (zfs_prop_valid_for_type(p, ZFS_TYPE_BOOKMARK, B_FALSE)) {
fnvlist_add_boolean(props, zfs_prop_to_name(p));
}
}
fnvlist_add_boolean(props, "redact_complete");
if ((err = lzc_get_bookmarks(zhp->zfs_name, props, &bmarks)) != 0)
goto out;
for (pair = nvlist_next_nvpair(bmarks, NULL);
pair != NULL; pair = nvlist_next_nvpair(bmarks, pair)) {
char name[ZFS_MAX_DATASET_NAME_LEN];
char *bmark_name;
nvlist_t *bmark_props;
bmark_name = nvpair_name(pair);
bmark_props = fnvpair_value_nvlist(pair);
if (snprintf(name, sizeof (name), "%s#%s", zhp->zfs_name,
bmark_name) >= sizeof (name)) {
err = EINVAL;
goto out;
}
nzhp = make_bookmark_handle(zhp, name, bmark_props);
if (nzhp == NULL)
continue;
if ((err = func(nzhp, data)) != 0)
goto out;
}
out:
fnvlist_free(props);
fnvlist_free(bmarks);
return (err);
}
/*
* Routines for dealing with the sorted snapshot functionality
*/
typedef struct zfs_node {
zfs_handle_t *zn_handle;
avl_node_t zn_avlnode;
} zfs_node_t;
static int
zfs_sort_snaps(zfs_handle_t *zhp, void *data)
{
avl_tree_t *avl = data;
zfs_node_t *node;
zfs_node_t search;
search.zn_handle = zhp;
node = avl_find(avl, &search, NULL);
if (node) {
/*
* If this snapshot was renamed while we were creating the
* AVL tree, it's possible that we already inserted it under
* its old name. Remove the old handle before adding the new
* one.
*/
zfs_close(node->zn_handle);
avl_remove(avl, node);
free(node);
}
node = zfs_alloc(zhp->zfs_hdl, sizeof (zfs_node_t));
node->zn_handle = zhp;
avl_add(avl, node);
return (0);
}
static int
zfs_snapshot_compare(const void *larg, const void *rarg)
{
zfs_handle_t *l = ((zfs_node_t *)larg)->zn_handle;
zfs_handle_t *r = ((zfs_node_t *)rarg)->zn_handle;
uint64_t lcreate, rcreate;
/*
* Sort them according to creation time. We use the hidden
* CREATETXG property to get an absolute ordering of snapshots.
*/
lcreate = zfs_prop_get_int(l, ZFS_PROP_CREATETXG);
rcreate = zfs_prop_get_int(r, ZFS_PROP_CREATETXG);
return (TREE_CMP(lcreate, rcreate));
}
int
zfs_iter_snapshots_sorted(zfs_handle_t *zhp, zfs_iter_f callback, void *data,
uint64_t min_txg, uint64_t max_txg)
{
int ret = 0;
zfs_node_t *node;
avl_tree_t avl;
void *cookie = NULL;
avl_create(&avl, zfs_snapshot_compare,
sizeof (zfs_node_t), offsetof(zfs_node_t, zn_avlnode));
ret = zfs_iter_snapshots(zhp, B_FALSE, zfs_sort_snaps, &avl, min_txg,
max_txg);
for (node = avl_first(&avl); node != NULL; node = AVL_NEXT(&avl, node))
ret |= callback(node->zn_handle, data);
while ((node = avl_destroy_nodes(&avl, &cookie)) != NULL)
free(node);
avl_destroy(&avl);
return (ret);
}
typedef struct {
char *ssa_first;
char *ssa_last;
boolean_t ssa_seenfirst;
boolean_t ssa_seenlast;
zfs_iter_f ssa_func;
void *ssa_arg;
} snapspec_arg_t;
static int
snapspec_cb(zfs_handle_t *zhp, void *arg)
{
snapspec_arg_t *ssa = arg;
const char *shortsnapname;
int err = 0;
if (ssa->ssa_seenlast)
return (0);
shortsnapname = strchr(zfs_get_name(zhp), '@') + 1;
if (!ssa->ssa_seenfirst && strcmp(shortsnapname, ssa->ssa_first) == 0)
ssa->ssa_seenfirst = B_TRUE;
if (strcmp(shortsnapname, ssa->ssa_last) == 0)
ssa->ssa_seenlast = B_TRUE;
if (ssa->ssa_seenfirst) {
err = ssa->ssa_func(zhp, ssa->ssa_arg);
} else {
zfs_close(zhp);
}
return (err);
}
/*
* spec is a string like "A,B%C,D"
*
* <snaps>, where <snaps> can be:
* <snap> (single snapshot)
* <snap>%<snap> (range of snapshots, inclusive)
* %<snap> (range of snapshots, starting with earliest)
* <snap>% (range of snapshots, ending with last)
* % (all snapshots)
* <snaps>[,...] (comma separated list of the above)
*
* If a snapshot can not be opened, continue trying to open the others, but
* return ENOENT at the end.
*/
int
zfs_iter_snapspec(zfs_handle_t *fs_zhp, const char *spec_orig,
zfs_iter_f func, void *arg)
{
char *buf, *comma_separated, *cp;
int err = 0;
int ret = 0;
buf = zfs_strdup(fs_zhp->zfs_hdl, spec_orig);
cp = buf;
while ((comma_separated = strsep(&cp, ",")) != NULL) {
char *pct = strchr(comma_separated, '%');
if (pct != NULL) {
snapspec_arg_t ssa = { 0 };
ssa.ssa_func = func;
ssa.ssa_arg = arg;
if (pct == comma_separated)
ssa.ssa_seenfirst = B_TRUE;
else
ssa.ssa_first = comma_separated;
*pct = '\0';
ssa.ssa_last = pct + 1;
/*
* If there is a lastname specified, make sure it
* exists.
*/
if (ssa.ssa_last[0] != '\0') {
char snapname[ZFS_MAX_DATASET_NAME_LEN];
(void) snprintf(snapname, sizeof (snapname),
"%s@%s", zfs_get_name(fs_zhp),
ssa.ssa_last);
if (!zfs_dataset_exists(fs_zhp->zfs_hdl,
snapname, ZFS_TYPE_SNAPSHOT)) {
ret = ENOENT;
continue;
}
}
err = zfs_iter_snapshots_sorted(fs_zhp,
snapspec_cb, &ssa, 0, 0);
if (ret == 0)
ret = err;
if (ret == 0 && (!ssa.ssa_seenfirst ||
(ssa.ssa_last[0] != '\0' && !ssa.ssa_seenlast))) {
ret = ENOENT;
}
} else {
char snapname[ZFS_MAX_DATASET_NAME_LEN];
zfs_handle_t *snap_zhp;
(void) snprintf(snapname, sizeof (snapname), "%s@%s",
zfs_get_name(fs_zhp), comma_separated);
snap_zhp = make_dataset_handle(fs_zhp->zfs_hdl,
snapname);
if (snap_zhp == NULL) {
ret = ENOENT;
continue;
}
err = func(snap_zhp, arg);
if (ret == 0)
ret = err;
}
}
free(buf);
return (ret);
}
/*
* Iterate over all children, snapshots and filesystems
* Process snapshots before filesystems because they are nearer the input
* handle: this is extremely important when used with zfs_iter_f functions
* looking for data, following the logic that we would like to find it as soon
* and as close as possible.
*/
int
zfs_iter_children(zfs_handle_t *zhp, zfs_iter_f func, void *data)
{
int ret;
if ((ret = zfs_iter_snapshots(zhp, B_FALSE, func, data, 0, 0)) != 0)
return (ret);
return (zfs_iter_filesystems(zhp, func, data));
}
typedef struct iter_stack_frame {
struct iter_stack_frame *next;
zfs_handle_t *zhp;
} iter_stack_frame_t;
typedef struct iter_dependents_arg {
boolean_t first;
boolean_t allowrecursion;
iter_stack_frame_t *stack;
zfs_iter_f func;
void *data;
} iter_dependents_arg_t;
static int
iter_dependents_cb(zfs_handle_t *zhp, void *arg)
{
iter_dependents_arg_t *ida = arg;
int err = 0;
boolean_t first = ida->first;
ida->first = B_FALSE;
if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) {
err = zfs_iter_clones(zhp, iter_dependents_cb, ida);
} else if (zhp->zfs_type != ZFS_TYPE_BOOKMARK) {
iter_stack_frame_t isf;
iter_stack_frame_t *f;
/*
* check if there is a cycle by seeing if this fs is already
* on the stack.
*/
for (f = ida->stack; f != NULL; f = f->next) {
if (f->zhp->zfs_dmustats.dds_guid ==
zhp->zfs_dmustats.dds_guid) {
if (ida->allowrecursion) {
zfs_close(zhp);
return (0);
} else {
zfs_error_aux(zhp->zfs_hdl,
dgettext(TEXT_DOMAIN,
"recursive dependency at '%s'"),
zfs_get_name(zhp));
err = zfs_error(zhp->zfs_hdl,
EZFS_RECURSIVE,
dgettext(TEXT_DOMAIN,
"cannot determine dependent "
"datasets"));
zfs_close(zhp);
return (err);
}
}
}
isf.zhp = zhp;
isf.next = ida->stack;
ida->stack = &isf;
err = zfs_iter_filesystems(zhp, iter_dependents_cb, ida);
if (err == 0)
err = zfs_iter_snapshots(zhp, B_FALSE,
iter_dependents_cb, ida, 0, 0);
ida->stack = isf.next;
}
if (!first && err == 0)
err = ida->func(zhp, ida->data);
else
zfs_close(zhp);
return (err);
}
int
zfs_iter_dependents(zfs_handle_t *zhp, boolean_t allowrecursion,
zfs_iter_f func, void *data)
{
iter_dependents_arg_t ida;
ida.allowrecursion = allowrecursion;
ida.stack = NULL;
ida.func = func;
ida.data = data;
ida.first = B_TRUE;
return (iter_dependents_cb(zfs_handle_dup(zhp), &ida));
}
/*
* Iterate over mounted children of the specified dataset
*/
int
zfs_iter_mounted(zfs_handle_t *zhp, zfs_iter_f func, void *data)
{
char mnt_prop[ZFS_MAXPROPLEN];
struct mnttab entry;
zfs_handle_t *mtab_zhp;
size_t namelen = strlen(zhp->zfs_name);
FILE *mnttab;
int err = 0;
if ((mnttab = fopen(MNTTAB, "r")) == NULL)
return (ENOENT);
while (err == 0 && getmntent(mnttab, &entry) == 0) {
/* Ignore non-ZFS entries */
if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0)
continue;
/* Ignore datasets not within the provided dataset */
if (strncmp(entry.mnt_special, zhp->zfs_name, namelen) != 0 ||
(entry.mnt_special[namelen] != '/' &&
entry.mnt_special[namelen] != '@'))
continue;
if ((mtab_zhp = zfs_open(zhp->zfs_hdl, entry.mnt_special,
ZFS_TYPE_FILESYSTEM)) == NULL)
continue;
/* Ignore legacy mounts as they are user managed */
verify(zfs_prop_get(mtab_zhp, ZFS_PROP_MOUNTPOINT, mnt_prop,
sizeof (mnt_prop), NULL, NULL, 0, B_FALSE) == 0);
if (strcmp(mnt_prop, "legacy") == 0) {
zfs_close(mtab_zhp);
continue;
}
err = func(mtab_zhp, data);
}
fclose(mnttab);
return (err);
}
Reference in New Issue View Git Blame Copy Permalink