msg/async/rdma: Use RDMA resources only after fork

Thanks to previous patch [1], no need to access RDMA resources before
the fork. Initialize Infiniband class only before a connection is
established or a listener is created. [1] is making sure that the call
to RDMAWorker::listen() is postponed till after the fork.

[1] - 7393db45644d ("msg/async: Postpone bind if network stack is not ready")

While backporting from master branch into the stable branch had to pull
some missing stuff from master:
- Make Infiniband ctor an empty shell. Actual initialization is done
  through a call to init()
- RDMADispatcher polling thread is started only after Infiniband::init()
  is called

(cherry picked from commit 972c7416deae2dd3a763643be6c9334d4edd1c17)

Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>
Change-Id: Iab1c450937713e6c4b83daf03c903e2fe5562ba2

msg/async: Postpone bind if network stack is not ready

RDMAStack shouldn't access hardware from the parent process.
The only reason to do so, is because bind is called before the fork.
After this patch the bind is postponed until the NetworkStack reports
that it is ready to bind.
For NetworkStack types will always return true, except the RDMAStack
which will return true only after the fork (after
AsyncMessenger::ready() is called).

This patch is based on a patch by Haomai Wang <haomai@xsky.com>

Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>
(cherry picked from commit a5b87e2fb83adec2b0fe060d3b22acfb1b9db1ed)

Change-Id: I5f8be4f93a5fe0f2a7ef3b29d755be52b0790c06

Revert "msg/async/rdma: Introduce Device.{cc,h}"

This reverts commit d930a3db13d4a5f9ab408faae7cdc237d20466f2.

Change-Id: Ib55ff961af125be5800ea90b2cce6588e3122f4d
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Move resource handling to Device"

This reverts commit 9cba3b36ef90a1f3d5d474f0fefe48e9c7ec6bad.

Change-Id: I5953a9380ecd5a07e8ea107cf1fc35d0772e8f61
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Move async event handling to Device"

This reverts commit 4b10ba7578b5d378edf715711bbf222dfe41b91d.

Change-Id: I9a294890c1cf01461356d87491bfa5e954b42444
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Initialize device on first connect"

This reverts commit 94eddb1ec53205e3e1b21033a9b841c4a72d7a77.

Change-Id: Ia16384eaed1295af6a4352e03308bc13bfb2f6d2
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Extract sockets stuff from RDMAStack.h"

This reverts commit 1890e924ca417c78c023f0a798ece8c3ea759413.

Change-Id: I66a33e55206b3f8f8c85d1f1ee1a2e85e3c2a3ad
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Introduce RDMAConnMgr"

This reverts commit 91bb13dceddb433966e6f775716e53ec5f523f5e.

Change-Id: I91dbe3c02651402f0e97e89151ec6a5154afdca3
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Debug prints for ibv_*"

This reverts commit 7d1475f6274531988854cf62f6056f9cfd13ff29.

Change-Id: Ieae3ea07f4e40d4558332e15fe6fbfcd22e0553e
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "Revert "msg/async/rdma: Debug prints for ibv_*""

This reverts commit 7e0f0748ef59842e329aa845baf153e812349450.

Change-Id: Iaccdd1739647a82aeafeff3ab46ed9f4d3fe0782
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Debug prints for ibv_*"

This reverts commit 42078045bc935122f939b71020bf3c66d7cef051.

Change-Id: If3a7b236636af0492bbe3576d25042b7a4dbf762
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Device::last_poll_dev must be positive"

This reverts commit fb7787dd58988fcf0e36f0b4155d707fd77b6a71.

Change-Id: I42d47cd683bd69ac2ebac1a6379a1ee0f2c1d543
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Fix small memory leaks detected by valgrind"

This reverts commit 23fd3efaafd4b4034bded605c88b57932b06646f.

Change-Id: Id04d88f4d08174632271c6eb224212155e0a4a38
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Make port number an attribute of the Connection not of the Device"

This reverts commit af2f1ecd2700ef3fad7e7d16d61f8ee86794b3ae.

Change-Id: I40f1da9da613dfb90aa8f09073c70ebb30b59aa1
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Make poll_blocking() poll for async events in addition to tx/rx"

This reverts commit 2386cea1eb31bec9ec49eafa0135eff82bbe696c.

Change-Id: I93a23f216cd4ce55051b125e9dd7edb3e297aaad
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Pass specific ConnMgr info in constructor"

This reverts commit ce4677612fdabab6e762cf0cd21914be438ce8e1.

Change-Id: I5098c109e7be17402ca512c25ba0c250ad610cea
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: get_device() by ibv_context"

This reverts commit f47025f83711d9af4a52470327fbd7411aafdafe.

Change-Id: Icc3667da1828868bc5001d1d8e8b2ec14e5f6c8e
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async: Postpone bind if network stack is not ready"

This reverts commit a5b87e2fb83adec2b0fe060d3b22acfb1b9db1ed.

Change-Id: Ia1bc0535636de27d599cfb2dc7be2a555e0a2cd7
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Revert "msg/async/rdma: Use RDMA resources only after fork"

This reverts commit 972c7416deae2dd3a763643be6c9334d4edd1c17.

Change-Id: I4a5627ce02ef8e055c2088db41ac4897dd9a3425
Issue: 995322
Signed-off-by: Amir Vadai <amir@vadai.me>

Merge pull request #15195 from theanalyst/doc/12.0.3-typo-fix

doc: fixing an error in 12.0.3 release notes

Reviewed-by: Nathan Cutler <ncutler@suse.com>
Reviewed-by: Loic Dachary <loic@dachary.org>

Merge pull request #15186 from adamemerson/wip-clangtastic-mr-clocks

rgw: Remove pessimizing move
common: Forward-declare container I/O overloads

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

Merge pull request #12861 from theanalyst/wip-absolute-urls

rgw: pass authentication domain to civetweb

Reviewed-by: Yehuda Sadeh <yehuda@redhat.com>

Merge pull request #15150 from ovh/bp-doc-perf-histograms

doc: document perf historgrams

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

doc: document perf historgrams

Documentation for perf histograms schema and dump.

Signed-off-by: Piotr Dałek <piotr.dalek@corp.ovh.com>

rgw: Remove pessimizing move

Do not std::move values from returned functions.

Signed-off-by: Adam C. Emerson <aemerson@redhat.com>

common: Forward-declare container I/O overloads

This will allow strict ADL to find the right overload in the case of a
container of containers and unbreak compilation with Clang.

Signed-off-by: Adam C. Emerson <aemerson@redhat.com>

Merge pull request #15162 from dillaman/wip-19970

librbd: reduce potential of erroneous blacklisting on image close

Reviewed-by: Mykola Golub <mgolub@mirantis.com>

doc: fixing an error in 12.0.3 release notes

Thanks @smithfarm for pointing this out

Signed-off-by: Abhishek Lekshmanan <abhishek@suse.com>

Merge pull request #15191 from Linbing/modify-note-word-error

mon: Fixed typo in @post of _active()

Note word error of _active(), take iif with if

Signed-off-by: Linbing <hawkerous@gmail.com>

Merge pull request #15181 from tchaikov/wip-cxx1z-size

common: add ceph::size()

Reviewed-By: Adam C. Emerson <aemerson@redhat.com>

Merge pull request #14946 from dmick/wip-mgr-dashboard

mgr: dashboard GUI module

Merge pull request #14994 from theanalyst/f/tier-fix

rgw_rados: create sync module instances only if run_sync_thread is set

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

test: add mgr/ and smoke test for mgr dashboard

Signed-off-by: Dan Mick <dan.mick@redhat.com>

rgw_rados: create sync module instances only if run_sync_thread is set

This allows for radosgw-admin to run even if there was a
misconfiguration in setting tier types which would otherwise make it
difficult to revert an invalid tier type setting. Since sync modules are
necessary only for rgw and not for the admin itself, this allows for
invalid tier configurations to be corrected as rgw-admin wouldn't fail

Fixes: http://tracker.ceph.com/issues/19830
Signed-off-by: Abhishek Lekshmanan <abhishek@suse.com>

Merge pull request #15153 from zhangsw/optimize-torrent

rgw: optimize generating torrent file. Object data won't stay in memory  now.

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

Merge pull request #15001 from chardan/jfw-wip-halflife_atomic_t-kamakuras

rgw: migrate atomic_t to std::atomic<>

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

Merge pull request #14827 from liuyuhong66/rgw-swift-auth-improve

rgw/rgw_swift_auth.cc: using string::back() instead as the C++11 recommend

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

Merge pull request #15025 from mikulely/jiaying-fix-lc-list

rgw: fix lc list failure when shards not be all created

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

Merge pull request #12300 from idealguo/rename-s3_code

rgw: rename s3_code to err_code for swift

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

Merge pull request #15045 from shashalu/fix-slo

rgw: fix X-Object-Meta-Static-Large-Object in SLO download

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

Merge pull request #14800 from theanalyst/doc/release-notes/sse-c-correction

doc: release-notes clarify about rgw encryption

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

Merge pull request #14922 from ukernel/wip-kcephfs-whitelist

qa: update log whitelists for kcephfs suite

Reviewed-by: John Spray <john.spray@redhat.com>

Merge pull request #14941 from wjwithagen/wip-wjw-freebsd-doc-manual-install

doc: add FreeBSD manual install

Merge pull request #14988 from ShiqiCooperation/master

test/unittest_bluefs: check whether mounted success

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

Merge pull request #14688 from cbodley/wip-rgw-multi-suite

qa/rgw: add multisite suite to configure and run multisite tests

Reviewed-by: Orit Wasserman <owasserm@redhat.com>

Merge pull request #15071 from cbodley/wip-qa-dnsmasq

qa: add task for dnsmasq configuration

Reviewed-by: Vasu Kulkarni <vasu@redhat.com>
Reviewed-by: Abhishek Lekshmanan <abhishek.lekshmanan@gmail.com>

Merge pull request #15090 from theanalyst/release/v12.0.3/notes

doc: release notes for v12.0.3 (dev)

Merge pull request #15133 from cbodley/wip-doc-rgw-remove-fcgi

doc/rgw: remove fastcgi page and sample configs

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

Merge pull request #15144 from drunkard/master

doc: typo fixes on hyperlink/words

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

Merge pull request #15161 from dzafman/wip-test-fix

test: Test fix for SnapSet change

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

common: add ceph::size()

backport from C++17

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

Merge pull request #15147 from bassam/pr-speedup-ccache

cmake: improved build speed by 5x when using ccache

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

Merge pull request #15158 from liewegas/wip-mgr-dump

mon/MgrMonitor: add 'mgr dump [epoch]' command

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

Merge pull request #15159 from LiumxNL/fix-ping-monitor

pybind/rados: avoid call free() on invalid pointer

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

Merge pull request #15164 from tchaikov/wip-check-gperf-before-using-it

cmake: check the existence of gperf before using it

Reviewed-By: Adam C. Emerson <aemerson@redhat.com>

librbd: block RPC requests before unregistering watch during close

The watch should be released after exclusive lock is released to
avoid a potential race condition which would result in the
client being blacklisted by a peer.

Fixes: http://tracker.ceph.com/issues/19970
Signed-off-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #15089 from zhsj/add-mirror

doc: add new cn ceph mirror to doc and mirroring

Reviewed-by: Wido den Hollander <wido@42on.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>

Merge pull request #15173 from joscollin/wip-test-crush-warning

test/crush: silence warnings from -Walloc-size-larger-than= and -Wstringop-overflow=

Merge pull request #15171 from ztczll/master

test/objectstore: Check apply_transaction() return values

Merge pull request #15034 from liewegas/wip-bluestore-sync-write

os/bluestore/KernelDevice: fix sync write vs flush

Reviewed-by: Varada Kari <varada.kari@gmail.com>

os/bluestore/KernelDevice: fix flush() vs sync write()

We need to set the atomic flag so that a flush will actually flush.

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

librbd: introduced support for blocking watch notifications

Signed-off-by: Jason Dillaman <dillaman@redhat.com>

librbd: recheck current lock owner before blacklist

Signed-off-by: Jason Dillaman <dillaman@redhat.com>

test/crush: silence warnings from -Walloc-size-larger-than= and -Wstringop-overflow=

The following warnings appear during make:

/home/jcollin/workspace/ceph/src/test/crush/CrushWrapper.cc: In member function ‘virtual void CrushWrapper_choose_args_compat_Test::TestBody()’:
/home/jcollin/workspace/ceph/src/test/crush/CrushWrapper.cc:1057:20: warning: argument 1 value ‘18446744073709550912’ exceeds maximum object size 9223372036854775807 [-Walloc-size-larger-than=]
   crush_choose_arg choose_args[c.get_max_buckets()];
                    ^~~~~~~~~~~

In function ‘void* memset(void*, int, size_t)’,
    inlined from ‘virtual void CrushWrapper_choose_args_compat_Test::TestBody()’ at /home/jcollin/workspace/ceph/src/test/crush/CrushWrapper.cc:1058:9:
/usr/include/bits/string3.h:90:70: warning: ‘void* __builtin_memset(void*, int, long unsigned int)’: specified size 18446744073709550912 exceeds maximum object size 9223372036854775807 [-Wstringop-overflow=]
   return __builtin___memset_chk (__dest, __ch, __len, __bos0 (__dest));

Signed-off-by: Jos Collin <jcollin@redhat.com>

Check apply_transaction return values

Signed-off-by: zhanglei <243290414@qq.com>

doc: v12.0.3 release notes

Signed-off-by: Abhishek Lekshmanan <abhishek@suse.com>

doc: PendingReleaseNotes move the 12.0.2 section of release notes

So that we can track whatever diff goes in after

Signed-off-by: Abhishek Lekshmanan <abhishek@suse.com>

ceph.spec.in, debian/control: build/runtime dependency for cherrypy

Signed-off-by: Dan Mick <dan.mick@redhat.com>

mgr: change config_prefix to be 'mgr/<id>/' and separate handle/key by '/'

dm_crypt config-key usage has established '/' as the separator; keep
that as a convention for now

Signed-off-by: Dan Mick <dan.mick@redhat.com>

mgr/PyModules: move constructor definition to .cc

Otherwise, the incomplete ServeThread type causes problems
for others who include PyModules.h

Signed-off-by: Dan Mick <dan.mick@redhat.com>

pybind/mgr/dashboard/module.py: allow custom server addr, port

Signed-off-by: Dan Mick <dan.mick@redhat.com>

mgr/PyModules.cc,PyState.cc: improve config access debug logging

Signed-off-by: Dan Mick <dan.mick@redhat.com>

client/SyntheticClient.cc: filer is now a std::unique_ptr

Signed-off-by: Dan Mick <dan.mick@redhat.com>

ceph_syn.cc, client/SyntheticClient.cc: Client -> StandaloneClient

Signed-off-by: Dan Mick <dan.mick@redhat.com>

ServeThread: add 'running' flag, don't notify non-running threads

Signed-off-by: Dan Mick <dan.mick@redhat.com>

pybind/mgr: web dashboard

A read only display of some of the cluster's state, plus
some cephfs/rbd/osd pages that demonstrate how to go
read state remotely from daemons.

Signed-off-by: John Spray <john.spray@redhat.com>

vstart: enable gui plugin

Signed-off-by: John Spray <john.spray@redhat.com>

osd: wire up "histogram dump" to `tell`

...so that mgr modules can get at it.

Signed-off-by: John Spray <john.spray@redhat.com>

Merge pull request #15148 from zealoussnow/wip-spec

build: move bash_completion.d/ceph to ceph-common

Reviewed-by: Nathan Cutler <ncutler@suse.com>
Reviewed-by: Ken Dreyer <kdreyer@redhat.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>

cmake: check the existence of gperf before using it

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

Merge pull request #15160 from bassam/pr-system-boost

cmake: fix boost components for WITH_SYSTEM_BOOST

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

Merge pull request #15107 from guihecheng/rgw_file-cleanup

rgw: use pre-defined calls to replace raw flag operation

doc: hyperlink fixes

Signed-off-by: Drunkard Zhang <gongfan193@gmail.com>

doc: typo fixes

Signed-off-by: Drunkard Zhang <gongfan193@gmail.com>

rgw_file: use pre-defined calls to replace raw flag operation

Signed-off-by: Gui Hecheng <guihecheng@cmss.chinamobile.com>

doc: remove duplicated section in doc/radosgw/adminops

Signed-off-by: Drunkard Zhang <gongfan193@gmail.com>

Merge pull request #15053 from dillaman/wip-librbd-blkin

blkin: librbd trace hooks

Reviewed-by: Josh Durgin <jdurgin@redhat.com>
Reviewed-by: Mykola Golub <mgolub@mirantis.com>

blkin: rbd object map update traces

Signed-off-by: Jason Dillaman <dillaman@redhat.com>

blkin: initial trace hooks for IO path

To collect blkin traced for RBD, set the "rbd blkin trace all"
configuration parameter to true.

Signed-off-by: Jason Dillaman <dillaman@redhat.com>

blkin: add config variable rbd_blkin_trace_all

Signed-off-by: Jason Dillaman <dillaman@redhat.com>

blkin: improved objecter trace events

Signed-off-by: Jason Dillaman <dillaman@redhat.com>

blkin: librados should start new child trace

Signed-off-by: Jason Dillaman <dillaman@redhat.com>

blkin: improve trace visibility into object cacher

Signed-off-by: Jason Dillaman <dillaman@redhat.com>

blkin: pass trace information to object cacher

Signed-off-by: Victor Araujo <ve.ar91@gmail.com>
Signed-off-by: Jason Dillaman <dillaman@redhat.com>

common/zipkin_trace: added missing include for encoding

Signed-off-by: Jason Dillaman <dillaman@redhat.com>

test/librados_test_stub: add new aio_operate overloads for tracing

Signed-off-by: Jason Dillaman <dillaman@redhat.com>

test: Test fix for SnapSet change

Caused by: a7b59221691954aedfac88695df8c89e36e10739

Signed-off-by: David Zafman <dzafman@redhat.com>

Merge pull request #10736 from oliveiradan/SebastienHan_config_diff_get_fr

common/ceph_context: 'config diff get' option added

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

Merge pull request #14914 from neha-ojha/wip-mon-crushtype-health-warn

mon: add crush type down health warnings

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