doc/rados/operations/placement-groups: added bias + profile

Added documentations on the autoscale profile and bias

Signed-off-by: Kamoltat <ksirivad@redhat.com>

pybind/mgr/pg_autoscaler: Added PROFILE to autoscale-status

autoscaler now shows the PROFILE status of each pool
in `ceph osd pool autoscale-status`

Signed-off-by: Kamoltat <ksirivad@redhat.com>

Merge pull request #42550 from dang/wip-dang-zipper-writer

Zipper Writer API
Reviewed-by: cbodley@redhat.com

Merge pull request #31454 from soumyakoduri/dbstore

rgw/Zipper: DB Backend store

Reviewed-by: dang@redhat.com
Reviewed-by: amaredia@redhat.com

RGW - Zipper - Proper Writer API

With the implementation of DBStore, it was determined that the API used
for writing in Zipper was too tied to RADOS.  Implement a clean writing
API named Writer.

Signed-off-by: Daniel Gryniewicz <dang@redhat.com>

Merge pull request #42266 from dang/wip-dang-zipper-raw_obj

Wip dang zipper raw obj

Reviewed-by: Soumya Koduri <skoduri@redhat.com>

Merge pull request #42308 from jtlayton/wip-51644

osd: don't assert on zero-length OP_ZERO request

Reviewed-by: Ilya Dryomov <idryomov@gmail.com>

Merge pull request #42523 from mgfritch/cephadm-fsid-validate

cephadm: validate `fsid` command arg

Reviewed-by: Sage Weil <sage@redhat.com>
Reviewed-by: Adam King <adking@redhat.com>

Merge pull request #42528 from liewegas/fix-51816

mon/LogMonitor: fix crash when cluster log file is not writeable

Reviewed-by: Neha Ojha <nojha@redhat.com>

Merge pull request #42538 from dsavineau/issue_51902

cephadm: don't use ctx.fsid for clean_cgroup

Reviewed-by: Adam King <adking@redhat.com>

Merge pull request #42558 from tchaikov/wip-crimson-cleanup

crimson/os: cleanups for building with Clang

Reviewed-by: Yingxin Cheng <yingxin.cheng@intel.com>

crimson/os: do not capture unused variable

Signed-off-by: Kefu Chai <kchai@redhat.com>

crimson/os: reference this explicitly

to silence false alarm from Clang that `this` is not used.

Signed-off-by: Kefu Chai <kchai@redhat.com>

crimson/os: do not capture labels

structured binding does not define variables, so we cannot capture them
without defining variables in capture list.

in this change, instead of using a map<> for defining labels, just
create labels on the fly.

Signed-off-by: Kefu Chai <kchai@redhat.com>

Merge pull request #42556 from tchaikov/wip-fair-mutex

common: add ceph::fair_mutex

Reviewed-by: Xiubo Li <xiubli@redhat.com>

Merge pull request #42539 from cyx1231st/wip-seastore-cache-metrics-2

crimson/os/seastore/cache: refine metrics

Reviewed-by: Samuel Just <sjust@redhat.com>

common: add ceph::fair_mutex

a mutex which enqueues and wakes up the waiters in FIFO order, to
ensure the fairness of the mutex.

Signed-off-by: Kefu Chai <kchai@redhat.com>

crimson/os/seastore: reassign extent_types_t values and remove extent_type_to_index()

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

crimson/os/seastore/cache: misc cleanup to metrics

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

crimson/os/seastore/cache: remove derived metrics

Only keep the basic metrics to minimize the total number of metrics.

Derived metrics can be numerous according to different needs and can be
confusing with labels.

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

crimson/os/seastore/cache: remove counter labels

Do not label metrics by counter type which could be confusing.

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

crimson/os/seastore/cache: cleanup, replace unordered_map by array

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

Merge pull request #40965 from rokj/patch-3

doc: mention copying keyrings and adjust node names in manual deployment example

Reviewed-by: Ilya Dryomov <idryomov@gmail.com>

rgw/dbstore: Fix library link issues

Now that rgw_common is no more linked with rgw_a library (commit#7b61667),
dbstore (rgw_sal_dbstore) should be linked directly to rgw_common.

Signed-off-by: Soumya Koduri <skoduri@redhat.com>

Merge pull request #42432 from tchaikov/wip-mon-crush-cleanup

mon: let CrushWrapper::get_validated_type_id() return an optional<>

Reviewed-by: Radoslaw Zarzynski <rzarzyns@redhat.com>

Merge pull request #42524 from guits/cv_wait_destroy_tests

ceph-volume/tests: retry when destroying osd

Merge pull request #42515 from rhcs-dashboard/decouple-unit-tests-from-build-dir

mgr/dashboard: backend unit tests: decouple from build dir

Reviewed-by: Waad Alkhoury <walkhour@redhat.com>
Reviewed-by: Alfonso Martínez <almartin@redhat.com>
Reviewed-by: Avan Thakkar <athakkar@redhat.com>

Merge pull request #42516 from tchaikov/wip-win32-snappy

win32_deps_build.sh: bump snappy version to 1.1.9

Reviewed-by: Nathan Cutler <ncutler@suse.com>

doc: adding missing command. changed node naming.

Signed-off-by: Rok Jaklič <rokj@rasca.net>

crimson/os/seastore/cache: cleanup, rename to get_by_src()

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

crimson/os/seastore: measure cache hit ratio by src

Remove excessive amount of cache hit/access metrics by extent type.

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

crimson/os/seastore: measure committed efforts by extent

In order to cross-check the writes at segment manager level, and
evaluate the write amplification from each sub-component.

Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>

win32_deps_build.sh: only clone the tip of required tag

no need to clone the whole repo, just clone the tip of the specified
tag. this saves the bandwidth, disk IO and precious time.

Signed-off-by: Kefu Chai <kchai@redhat.com>

win32_deps_build.sh: bump snappy version to 1.1.9

in snappy, the commit of 26102a0c66175bc39edbf484c994a21902e986dc
fixes the SNAPPY_VERSION generation. and this commit was included by
v1.1.8 and v1.1.9.

also, in v1.1.9, a change was introduced, where the function signature
was changed, and more importantly, this change is not backward
compatible:

<   bool GetUncompressedLength(Source* source, uint32_t* result);
---
>   bool GetUncompressedLength(Source* source, uint32* result);

see also, https://tracker.ceph.com/issues/50934

so we check SNAPPY_VERSION to tell if we should use `uint32_t` or
`uint32`.

in this change, snappy version used to build win32 client is bumped
to the latest stable version, v1.1.9, to include the fix of
SNAPPY_VERSION. this paves the road to fix of https://tracker.ceph.com/issues/50934

Signed-off-by: Kefu Chai <kchai@redhat.com>

compression/snappy: use uint32_t to be compatible with 1.1.9

The snappy project made the following change in snappy.h between version 1.1.8
and 1.1.9:

<   bool GetUncompressedLength(Source* source, uint32_t* result);
---
>   bool GetUncompressedLength(Source* source, uint32* result);

This causes Ceph to FTBFS with snappy 1.1.9.

Thanks to Chris Denice for bringing this to our attention via Redmine.

Fixes: https://tracker.ceph.com/issues/50934
Signed-off-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #42442 from badone/wip-insights-reports-non-persistent-storage

Don't persist report data

Reviewed-by: Josh Durgin <jdurgin@redhat.com>
Reviewed-by: Neha Ojha <nojha@redhat.com>

cephadm: don't use ctx.fsid for clean_cgroup

The clean_cgroup method assumes that the ctx.fsid is set while this is
true for the bootstrap command, it isn't set for adopt or deploy commands
(and maybe others).

This ends up to the adopt command to fails:

Traceback (most recent call last):
  File "/sbin/cephadm", line 8301, in <module>
    main()
  File "/sbin/cephadm", line 8289, in main
    r = ctx.func(ctx)
  File "/sbin/cephadm", line 1764, in _default_image
    return func(ctx)
  File "/sbin/cephadm", line 5091, in command_adopt
    command_adopt_ceph(ctx, daemon_type, daemon_id, fsid)
  File "/sbin/cephadm", line 5299, in command_adopt_ceph
    osd_fsid=osd_fsid)
  File "/sbin/cephadm", line 2884, in deploy_daemon_units
    clean_cgroup(ctx, unit_name)
  File "/sbin/cephadm", line 2724, in clean_cgroup
    if not ctx.fsid:
  File "/sbin/cephadm", line 155, in __getattr__
    return super().__getattribute__(name)
AttributeError: 'CephadmContext' object has no attribute 'fsid'

Since we already have the fsid value in deploy_daemon_units (which calls
clean_cgroup) then we can pass the fsid value directly.

This fixes a regression introduced by 1fee255

Fixes: https://tracker.ceph.com/issues/51902
Signed-off-by: Dimitri Savineau <dsavinea@redhat.com>

Merge pull request #42527 from ceph/ljflores-patch-3

doc/mgr/telemetry: fix formatting problem

Reviewed-by: Neha Ojha <nojha@redhat.com>

Merge PR #42349 into master

* refs/pull/42349/head:
mon/MDSMonitor: propose if FSMap struct_v is too old
mon/MDSMonitor: give a proper error message if FSMap struct_v is too old
mds/FSMap: use DECODE_OLDEST to gate FSMap version
qa: add tests for fs dump of epoch and trimming
qa: add file system support for dumping epoch
mon/MDSMonitor: return mon_mds_force_trim_to even if equal to current epoch
mon: add debugging for trimming methods
mon: fix debug spacing
qa: add nofs upgrade suite

Reviewed-by: Kefu Chai <kchai@redhat.com>
Reviewed-by: Neha Ojha <nojha@redhat.com>
Reviewed-by: Ramana Raja <rraja@redhat.com>

Merge PR #42199 into master

* refs/pull/42199/head:
mds: add debugging when rejecting mksnap with EPERM

Reviewed-by: Milind Changire <mchangir@redhat.com>

Merge PR #41025 into master

* refs/pull/41025/head:
qa: wait pgs to be clean before using the pools
qa: ignore PG_RECOVERY_FULL and PG_DEGRADED for mds-full
qa: wait more time since there have many more pgs than before
qa: do not multiple the full ratio twice
qa: do not raise for kclient for _fsync test
qa: use the pg autoscale mode to calcuate the pg_num
qa: set the object_size to 1M
qa: move the is_full() to parent class

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge PR #38388 into master

* refs/pull/38388/head:
mds: check rejoin_ack_gather before enter rejoin_gather_finish

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Sidharth Anupkrishnan <sanupkri@redhat.com>

Merge pull request #42453 from sebastian-philipp/githubmap-rh

.githubmap: Update Sebastian Wagner's mapping

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: David Galloway <dgallowa@redhat.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>

Merge pull request #42501 from ybwang0211/doc-cap

doc/man: add missing right parenthesis in manpage.

Reviewed-by: Kefu Chai <kchai@redhat.com>

Merge pull request #42495 from hjwsm1989/wip-51842

crush: cancel upmaps with up set size != pool size

Reviewed-by: xie xingguo <xie.xingguo@zte.com.cn>

Merge pull request #42511 from adk3798/shutil-copy-exception

cephadm: don't fail hard on SameFileError during shutil.copy

Reviewed-by: Michael Fritch <mfritch@suse.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>

Merge pull request #40337 from ideepika/wip-bugzilla-1857447

mon/PGMap: remove DIRTY field in `ceph df detail` when cache tiering  is not in use

Reviewed-by: Neha Ojha <nojha@redhat.com>

Merge pull request #42508 from cybozu/kv-rocksdbstore-enrich-debug-message

kv/RocksDBStore: enrich debug message

Reviewed-by: Kefu Chai <kchai@redhat.com>

Merge pull request #42500 from ybwang0211/doc-list-get-attr

tools/rados: improve the usage message of {get,set}omapaheader

Reviewed-by: Kefu Chai <kchai@redhat.com>

RGW - Zipper - MultipartUpload

Create a MultipartUpload object in the Zipper API.

Signed-off-by: Daniel Gryniewicz <dang@redhat.com>

Remove get_raw_obj() from Object

Signed-off-by: Daniel Gryniewicz <dang@redhat.com>

rgw - Configure ceph.conf for orphan-list

Signed-off-by: Daniel Gryniewicz <dang@redhat.com>

mon/LogMonitor: fix crash when cluster log file is not writeable

If we are in this block, then p == channel_fds.end() and p->first is not
valid.

Also, no need to populate channel_fds with an fd of -1.

Fixes: https://tracker.ceph.com/issues/51816
Signed-off-by: Sage Weil <sage@newdream.net>

doc/mgr/telemetry: fix formatting problem

There was strange bolding and bullet point placement due to a missing new line in the perf description.

Signed-off-by: Laura Flores <lflores@redhat.com>

Merge pull request #42502 from tchaikov/wip-bloomfilter-cleanups

include/intarith, common/bloom_filter: add popcount() and cleanups

Reviewed-by: Sage Weil <sage@redhat.com>

ceph-volume/tests: retry when destroying osd

Sometimes, it can happen that the osds being destroyed in those tests
are not yet marked as 'down' for some reason. Let's add some retries on
those tasks to avoid CI failures.

Fixes: https://tracker.ceph.com/issues/51903
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>

mon/MDSMonitor: propose if FSMap struct_v is too old

To flush older versions which may still be an empty MDSMap (for clusters
that have never used CephFS), we need to force a proposal so older
versions of the struct are trimmed.

This is the main fix of this branch. We removed code which processed old
encodings of the MDSMap in the mon store via 60bc524. That broke old
ceph clusters which never used CephFS (see cited ticket below).  This is
because the initial epoch is an empty MDSMap (back in Infernalis/Hammer)
that is never updated. So, the fix here is to just do proposals
periodically until all of the old structs are automatically trimmed by
the mons.

Fixes: 60bc524827bac072658203e56b1fa3dede9641c5
Fixes: https://tracker.ceph.com/issues/51673
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>

mon/MDSMonitor: give a proper error message if FSMap struct_v is too old

Fixes: https://tracker.ceph.com/issues/51673
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>

mds/FSMap: use DECODE_OLDEST to gate FSMap version

This throws a proper exception which can be handled.

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>

qa: add tests for fs dump of epoch and trimming

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>

qa: add file system support for dumping epoch

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>

mon/MDSMonitor: return mon_mds_force_trim_to even if equal to current epoch

The PaxosService code already excludes the value returned by
PaxosService::get_trim_to as the upper bound of the range of epochs to
trim. Without this fix, you need to set mon_mds_force_trim_to to one
greater than the epoch you want to trim _and_ force the current epoch to
be one greater than that; the net result being that you can only force
trimming up to 2 epochs behind the current epoch.

This change is helpful for resolving issue 51673, but not strictly
necessary.

Related-to: https://tracker.ceph.com/issues/51673
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>

mon: add debugging for trimming methods

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>

mon: fix debug spacing

Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>

qa: add nofs upgrade suite

This adds an upgrade suite to ensure that a Ceph cluster without a
CephFS file system does not blow up on upgrade (in particular, that the
MDSMonitor does not trip). This was developed to potentially reproduce
tracker 51673 but the actual cause for that issue was an old encoding
for the MDSMap which was obsoleted in Pacific. You must create a cluster
older than the FSMap (~Hammer or Infernalis) to reproduce. In any case,
this upgrade suite may be useful in the future so let's keep it!

Related-to: https://tracker.ceph.com/issues/51673
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>

cephadm: validate fsid during ceph-volume command

Signed-off-by: Michael Fritch <mfritch@suse.com>

cephadm: validate fsid during shell command

Signed-off-by: Michael Fritch <mfritch@suse.com>

test: add a test for zero-length OP_ZERO request

This triggered an assertion before. Verify that it's a no-op now.

Signed-off-by: Jeff Layton <jlayton@redhat.com>

mgr/dashboard: backend unit tests: decouple from build dir

After https://github.com/ceph/ceph/pull/42255 some unit tests got coupled
to the frontend folder of the build dir.
Backend unit tests do not need a build folder: they only test our business logic.

Signed-off-by: Alfonso Martínez <almartin@redhat.com>

Merge pull request #42169 from Matan-B/lua-nats-example

rgw: Add an example for lua scripting

Merge pull request #42350 from TRYTOBE8TME/wip-rgw-notification

rgw/notifications: Change in multipart upload notification behavior

rgw/dbstore: Add DBStore base class def to namespace rgw::store

Signed-off-by: Soumya Koduri <skoduri@redhat.com>

rgw/dbstore: Use dpp for logging

Aligning with other Ceph RGW modules, use dpp for logging.
A default dpp is defined for DBStore to be used for unit-tests.

Signed-off-by: Soumya Koduri <skoduri@redhat.com>

rgw/dbstore: Use CEPH RGW logging in ceph context

When running dbstore as RGW backend, use the logging mechanism
and config values provided by default by RGW.
For other tests, set logfile and loglevel to the input values provided.

Signed-off-by: Soumya Koduri <skoduri@redhat.com>

rgw/dbstore: DB backend for RGW

As part of Zipper, adding support for DB backend (using SQLite) for Stackable RGW.

The base class methods implemented are generic which could be extended to any backend database, not just SQLite.

More details on design/implementation can be found at -
https://docs.google.com/document/d/1xCoHT5DCujqbe1pnEfYpSaSqiBcW8z87CgQCC5LivOo/edit#

Current status:

[Done]
- User related APIs
- Bucket APIs
- Testcases (using Gtest) to test the DBStore APIs

[Not handled in the first pass]
- Quota
- Usage Stats
- Swift Users
- Multiple Zones/Zonegroups

[WIP]
- Object APIs

Signed-off-by: Soumya Koduri <skoduri@redhat.com>

common/bloom_filter: move bloom_filter::density() to .cc

for faster compilation, also, bloom_filter::density()
is not sitting in the critical paths, so no need to make it an
inline function.

Signed-off-by: Kefu Chai <kchai@redhat.com>

Merge pull request #42389 from asannes/ingress-fix

mgr/cephadm: ingress: fix typo in spec.virtual_interface_networks reference

Reviewed-by: Sage Weil <sage@redhat.com>

Merge pull request #42510 from rzarzynski/wip-crimson-use-random-nonce

crimson/osd: fix misdirecting msgs when an OSD flips at Sepia.

Reviewed-by: Samuel Just <sjust@redhat.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>

doc/man: add missing right parenthesis in manpage.

Signed-off-by: wangyingbin <ybwang0211@163.com>

crush: cancel upmaps with up set size > pool size

Fixes: https://tracker.ceph.com/issues/51842
Signed-off-by: huangjun <huangjun@xsky.com>

Merge pull request #42322 from ljflores/wip-lflores-telemetry-docs

doc/mgr/telemetry: update Telemetry Module docs to include perf channel

Reviewed-by: Josh Durgin <jdurgin@redhat.com>
Reviewed-by: Zac Dover <zac.dover@gmail.com>

Merge pull request #42492 from zdover23/wip-doc-rados-top-level-section-2021-07-26

doc/rados: rewrite index.rst

Reviewed-by: Josh Durgin <jdurgin@redhat.com>

osd: don't assert on zero-length OP_ZERO request

I was able to crash the OSD by sending it a CEPH_OSD_OP_ZERO request
that had an op.extent.length that was 0. Convert that to a no-op
instead.

Fixes: https://tracker.ceph.com/issues/51644
Signed-off-by: Jeff Layton <jlayton@redhat.com>

Merge pull request #42473 from tchaikov/wip-cmake-rgw-cleanup

cmake: make rgw_common a static library

Reviewed-by: Casey Bodley <cbodley@redhat.com>

cephadm: don't fail hard on SameFileError during shutil.copy

Fixes: https://tracker.ceph.com/issues/51829
Signed-off-by: Adam King <adking@redhat.com>

Merge PR #42406 into master

* refs/pull/42406/head:
mgr/nfs: remove unused 'realm' arg for 'nfs export create rgw'
doc/mgr/rook: update title
doc/mgr/nfs: reference customizing ingress
doc/mgr/nfs: add section for manual ganesha config; reframe
doc/mgr/nfs: document ingress in more detail
doc/mgr/nfs: typo
doc/mgr/nfs: add note about incomplete ingress
qa/suites/orch/cephadm: add rgw nfs export test
mgr/cephadm: ingress: tolerate no daemons
mgr/nfs: add --squash option to 'nfs export create rgw ...'
mgr/nfs: use bucket owner creds for rgw bucket export
mgr/cephadm: use new CEPH_IMAGE_TYPES for all daemons using ceph container image
qa/tasks/python: simple task to run python code
doc/mgr/nfs: revisions
mgr/nfs/export: nicer exceptions on cap update

Reviewed-by: Varsha Rao <varao@redhat.com>

crimson/osd: make the get_nonce() static.

Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>

crimson/osd: print OSD::nonce on start-up.

Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>

crimson/osd: fix misdirecting msgs when an OSD flips at Sepia.

For description please refer to:
  * https://gist.github.com/rzarzynski/6cbc20fb28d7ac096ecc4b68500454bf#gistcomment-3830762,
  * https://gist.github.com/rzarzynski/6cbc20fb28d7ac096ecc4b68500454bf#gistcomment-3830939.

Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>

Merge pull request #42476 from tchaikov/wip-crimson-more-aliens

doc, vstart.sh: adapt to the generalize AlienStore support

Reviewed-by: Radoslaw Zarzynski <rzarzyns@redhat.com>

Merge pull request #42222 from flipkart-incubator/aws_sse_s3

rgw: Introduce BucketEncryption APIs to support SSE-S3 feature

Reviewed-by: Casey Bodley <cbodley@redhat.com>

doc/dev/crimson: document cyanstore and alienstore

Signed-off-by: Kefu Chai <kchai@redhat.com>

mgr/nfs: remove unused 'realm' arg for 'nfs export create rgw'

This argument is not used.  The rgw exports currently only work on the
default realm.

Signed-off-by: Sage Weil <sage@newdream.net>

doc/mgr/rook: update title

Signed-off-by: Sage Weil <sage@newdream.net>

doc/mgr/nfs: reference customizing ingress

Link to the cephadm docs on modifying the service directly.

Signed-off-by: Sage Weil <sage@newdream.net>

doc/mgr/nfs: add section for manual ganesha config; reframe

This documentation is incomplete because this mode of operation is not
tested/validated.

Signed-off-by: Sage Weil <sage@newdream.net>

kv/RocksDBStore: enrich debug message

It's better to print why ListColumnFamilies() failed.

Signed-off-by: Toshikuni Fukaya <toshiq2@gmail.com>
Signed-off-by: Satoru Takeuchi <satoru.takeuchi@gmail.com>

tools/rados: improve the usage message of {get,set}omapaheader

Signed-off-by: wangyingbin <ybwang0211@163.com>

common/bloom_filter: use popcount() to implement density()

instead of counting the bits using a loop, it'd be more efficient to do
this using the SIMD instruction.

Signed-off-by: Kefu Chai <kchai@redhat.com>

common/bloom_filter: s/bits_per_char/CHAR_BIT/

use macro defined by standard libary instead of inventing our own.

Signed-off-by: Kefu Chai <kchai@redhat.com>