os/bluestore: fix SharedBlob unregistration

We use the SharedBlobSet remove() in three cases:

- from SharedBlob::put(), we try to remove ourselves from the set, but
  have to deal with a racing lookup, so the removal is conditional on
  nref still being 0.
- from split_cache(), we move the SharedBlob to another collection
- from make_blob_unshared(), we remove the entry when we clear the sbid.

The problem is that the condtiional remove() (for the first case) was being
used for all three cases, and in the second two cases nref is always != 0,
so it doesn't actually happen.  This can lead to a crash during cache
shutdown.

Fix by making two variants: remove() that is unconditional, and
try_remove() that is conditional.

Set the sb->coll pointer after because remove() asserts the parent matches
where we are unregistering.

Fixes: http://tracker.ceph.com/issues/22039
Signed-off-by: Sage Weil <sage@redhat.com>
(cherry picked from commit 78a465b90b1c5187cfec9f25eb58b3bf617ca39b)

Merge pull request #18957 from tchaikov/wip-pr-18808-luminous

luminous: qa: do not wait for down/out osd for pg convergence

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

qa: do not wait for down/out osd for pg convergence

that osd is not invovlved in the PG state changes.

Signed-off-by: Kefu Chai <kchai@redhat.com>
(cherry picked from commit 7f549af459b40c5bb51a24a3a443ed7a7116915e)

Merge pull request #18908 from smithfarm/wip-luminous-fixup

qa/suites/fs/basic_functional/clusters: more osds

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Abhishek Lekshmanan <abhishek.lekshmanan@gmail.com>

Merge pull request #18910 from ceph/backport-18882

luminous: ceph-volume: add functional tests for simple, rearrange lvm tests

Reviewed-by: Alfredo Deza <adeza@redhat.com>

ceph-volume tests.functional update lvm/tox.ini paths after moving it

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 7d787849556788961155534039886aedfcdb2a88)

ceph-volume tests.functional move top level tox.ini to lvm/tox.ini

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 9b14b96d3bf6c8326cfc91e657161ffab3b8c2b4)

ceph-volume tests.functional move xenial to lvm/xenial

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 00d576d266b63080129e58482bd6fba62534ee97)

ceph-volume tests.functional lvm move xenial/bluestore to xenial/bluestore/create

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit d5ec7d1c58141f4593ed9d1e4501c9f3be6d9755)

ceph-volume tests.functional lvm move xenial/create to xenial/filestore/create

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit c504b8781b3a31475280a032cd75c9e1eaa6d8c9)

ceph-volume tests.functional move centos7 to lvm/centos7

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit dc1da3fd0aeafe59a82343fb16fe1fcea66ae3eb)

ceph-volume tests.functional lvm move bluestore to bluestore/create

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit aec1a80a3e1e61720f494980711808bfee783a03)

ceph-volume tests.functional lvm move create to filestore/create

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit d6905f3f620d4c733835e21a275d4127cdcb769b)

ceph-volume tests.functional simple xenial bluestore scenario

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 8f0fdc975951a2a8af31bbaa3abd2b5d807a0add)

ceph-volume tests.functional simple xenial filestore scenario

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 7452e1b6f9b82c2ef62bae19c632e81039638ba5)

ceph-volume tests.functional simple tox.ini update for xenial scnearios

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 76ed6a8f2ecab7c58def0453ffed9d8a997e0d5b)

ceph-volume tests.functional simple tox.ini addition for bluestore

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 02154e445b87a8fdcabbd699a2532833764140b5)

ceph-volume tests.functional add initial files for simple bluestore activate

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 3320f28e799e1d77d1ea69be9ced603bb69cc5fa)

ceph-volume tests.functional add initial files for simple filestore activate

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 9d18bae52ad77be0e64ef105c33a019446b6f1cf)

ceph-volume tests.functional create a specific test.yml for simple activate

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 37bde0d3b37a87c456438b0edc0cb9e2dcd39120)

ceph-volume tests.functional create a separate tox.ini environ for
`simple`

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 89757ad12e6938fe2bf997e0f21a5ffac082e30b)

Merge pull request #18907 from ceph/luminous-wip-bz1498200

luminous: ceph-volume: support GPT and other deployed OSDs

Reviewed-by: Andrew Schoen <aschoen@redhat.com>

qa/suites/fs/basic_functional/clusters: more osds

Signed-off-by: Sage Weil <sage@redhat.com>
(cherry picked from commit 998122c324047c96c16d60d7fbb3a21f4dd63c56)

doc/ceph-volume include the top level systemd.rst in the toctree

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit e2f281c115c15e820588b478214c3600ff97470f)

ceph-volume simple.activate add --file in help docstring as an example

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit ff60592897b63581ebce372b3f6ffa0f10d2fe5a)

doc/ceph-volume activate has a --file option for json files

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 29913951b71d68b3f3f9b9fa076f309c0252da36)

ceph-volume simple.activate enable the osd when activating

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 366fb8c43af54e9c92f4e95d9a0582c1fc378dbc)

ceph-volume systemd.systemctl always force symlink when masking ceph-disk

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit cb470622fd4f13fafe9cb2f38fb25fd1fbd8d89c)

ceph-volume simple.activate use the contents of the JSON file, fallback to CLI args

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit c954c913395557bcd89f25b0496c4428e6109494)

ceph-volume simple.activate only check for missing id and uuid if file is not passed in

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 826c2709dd5ab0fbf6eb419a735d15e718a7712c)

ceph-volume simple.activate allow to pass a file to activate

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit bd3498f3eabf07637cb936e6c036d1d8d75415b0)

ceph-volume tests.util add tests for OSDPath validator

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit d7fc26ab4e2cebe485647510119ea300d47db184)

ceph-volume systemd.systemctl masking ceph-disk needs to be done by directly symlinking to dev/null

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit cda47c247b929401a7dce3f1ef70371ea80eeceb)

ceph-volume systemd.systemctl masking needs to use systemctl mask, not disable

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit bad2f62b34243addc854abf691854b886aecdc19)

ceph-volume simple.scan if mounted device cannot be detected, raise an error

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 6b2fc4c0ba63e4bd51774ea09cc6542738128621)

ceph-volume simple.activate remove str formatting from logging

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 230faf04a185c0b95c8ad497ddd5501dc70c1409)

ceph-volume util.arg_validators always convert the path to an absolute path

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit b5acf4d720b5395ae4ccbf007c94c99610e40089)

ceph-volume util.arg_validators require superuser privs to check paths

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 376dad74159532f901d0044f95330fc02f93bd25)

ceph-volume simple.scan skip directories to scan

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 6bc1e9ee0b8d7bc475f218ae3dd8c6c7fe5a983a)

ceph-volume tests.devices.simple add activate CLI unit tests

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit cde0d517563b537319d06c758b93e421cdabe59d)

ceph-volume tests.devices add help menu unit test

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit dcc7c72196a327f9501045f025761b228b5379df)

doc/ceph-volume systemd describe the interaction with 'simple' activation

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 7071eeaeb5d77c6ab6e7a24f1b630cb0cf8a8e7c)

doc/ceph-volume index initial description of simple

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit f9ee51f9feadb6d0f91b76a2b2ba8a24f1523a77)

doc/ceph-volume lvm add back a system.rst with just the lvm portion of it

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit f872cd29b93e02b77172f7b382c4746af9ba845d)

doc/ceph-volume make systemd.rst generic enough to be related to other sub-commands

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit cf4e32a0dc3ab4f55fe7857b51afbc63a638ec66)

doc/ceph-volume move lvm/systemd.rst to top level, making it generic

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 0672dd608d4601a936d703ea5ba6edc14a161d8e)

doc/ceph-volume simple.activate initial take on documenting activation

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 80c796bd1c43125f5678710b494b03c486b3fce6)

doc/ceph-volume lvm update systemd anchor to reflect lvm parent

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 53420a4756a5e4fa6b0f2aa74d6ad1e2919f1954)

doc/ceph-volume lvm.activate update systemd anchor to reflect lvm parent

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 7fc0e7ebcae06f28f6ed08c0db14075ada34e8a7)

doc/ceph-volume lvm.systemd update anchor to reflect lvm parent

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 01a8b7e6a45f6fd85bf53d3cab15acc81d0bae1b)

doc/ceph-volume document the scan functionality

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit b30a3eaf17a20a1273724b2cfcaf0a7b36eb08f6)

doc/ceph-volume update the index to include the simple sub command and files

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit d7f85c64e4b409ea95e30ccb5a1388278843c478)

ceph-volume tests.simple add tests for trigger parsing from systemd

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 0e2620234e82abf85274b3c08b39cdf76430ce48)

ceph-volume tests.simple create etc_path scan tests

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 90ebc4e1ffa7b3273787d708bb2c05180530b5d1)

ceph-volume tests.simple add checks for get_contents in scan

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 647bfe0883d2bb05b4ff7f99894075df3e73d2d5)

ceph-volume tests.util add binary verification tests

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 827aa57557561355dd4068e26801b5f1ed62f295)

ceph-volume tests create a fixture for temporary files

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 23da3445718c01f71e2a5748fb4fc1c5fdc85ca5)

ceph-volume systemd.systemctl helpers to mask ceph-disk

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 3da23fa4caea8c9dc91da3eee5b5acc1860bf242)

ceph-volume devices include simple at the module import level

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 027a09b3ea6bac416ff632a69a53393badd51708)

ceph-volume main add simple to the sub-commands

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 94579c77ffe4d5ba2d457ad3981a2b9ed232a90f)

ceph-volume util add an OSDPath argument validator

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit a826fdf4a2ed9ff704fb823d05d3cb706ba89e38)

ceph-volume simple.system add a helper to check if a file is a binary or not

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 7bf8676045b4f966b60cdadf4d5af1c89fca69ae)

ceph-volume util.system add a context manager for temporary mounting/unmounting

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 6394cdd41943e3cda6867ecc59ab835cb2151707)

ceph-volume simple.scan initial take on directory/device scanning

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 0e54b9be621cc0969a3bbc972d1870080ccdae77)

ceph-volume simple.trigger enable systemd triggering for simple setups

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit cbc0c0f775e96a5b0ab5d5bc141acebe4cc48373)

ceph-volume simple.activate initial take on activation

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 5a97374605408d18e9be29bcc051ae7b913d1675)

ceph-volume simple add __init__ for the module

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit 285e5bdd28a42d98418ca90ad0cdec6bd87849db)

ceph-volume simple create the initial, main module

Signed-off-by: Alfredo Deza <adeza@redhat.com>
(cherry picked from commit f050502425a1993ae8edfdb5785b24129cffeeba)

Merge pull request #18801 from linuxbox2/luminous-rgw-readdir2

rgw_file: implement variant offset readdir processing -- pullup 2 of 2

Reviewed-By: Abhishek Lekshmanan <abhishek@suse.com>

Merge pull request #18734 from shinobu-x/wip-21648-luminous

luminous: mgr[zabbix] float division by zero

Reviewed-by: Wido den Hollander <wido@widodh.nl>

Merge pull request #18851 from tchaikov/wip-18759-pr-luminous

mon/LogMonitor: "log last" should return up to n entries

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

Merge pull request #18741 from shinobu-x/wip-22019-luminous

luminous: tests: "ceph osd create" is not idempotent

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

mon/LogMonitor: "log last" should return up to n entries

limit the # of returned entries to "num", and backoff the start iterator
by one if it reaches the rend().

Signed-off-by: Kefu Chai <kchai@redhat.com>
(cherry picked from commit de14103f5201da69b360d9b01e249896d59376b3)

rgw_file: implement variant offset readdir processing

Introduce new rgw_readdir2(...), which in which continues
from an arbitrary dirent name, which presumably has been
seen in a prior partial enumeration.

Add single-file unit test for READDIR cases, librgw_file_marker.cc.

Signed-off-by: Matt Benjamin <mbenjamin@redhat.com>
(cherry picked from commit a2c34c597371809bd8fd40ee1fc6b8e6a5145609)
Signed-off-by: Matt Benjamin <mbenjamin@redhat.com>

Merge pull request #18800 from linuxbox2/luminous-rgw_file-fsid-mount

Luminous rgw file fsid mount -- pullup 1 of 2

Reviewed-By: Abhishek Lekshmanan <abhishek@suse.com>

Merge pull request #18767 from smithfarm/wip-22020-luminous

luminous: multisite: race between sync of bucket and bucket instance metadata

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

Merge pull request #18794 from theanalyst/wip-18709

luminous: rgw: Stale bucket index entry remains after object deletion

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

Merge pull request #18726 from shinobu-x/wip-luminous-22035

luminous: Spurious ceph-mgr failovers during mon elections

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

Merge pull request #18738 from shinobu-x/wip-21547-luminous

luminous: ceph-mgr gets process called "exe" after respawn

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

Merge pull request #18727 from shinobu-x/wip-luminous-22034

luminous: key mismatch for mgr after upgrade from jewel to luminous(dev)

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

Merge pull request #18723 from shinobu-x/wip-luminous-22023

luminous: osd: make stat_bytes and stat_bytes_used counters PRIO_USEFUL

Merge pull request #18737 from shinobu-x/wip-21549-luminous

luminous: the dashboard uses absolute links for filesystems and clients

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

Merge pull request #18736 from shinobu-x/wip-21638-luminous

luminous: dashboard OSD list has servers and osds in arbitrary order

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

Merge pull request #18728 from shinobu-x/wip-luminous-22032

luminous: dashboard barfs on nulls where it expects numbers

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

Merge pull request #18699 from shinobu-x/wip-luminous-21538

luminous: upmap does not respect osd reweights

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

Merge pull request #18719 from tchaikov/wip-pr-18373-luminous

luminous: osd,mon: add max-pg-per-osd limit

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

doc: fix the desc of "osd max pg per osd hard ratio"

Signed-off-by: Kefu Chai <kchai@redhat.com>
(cherry picked from commit d940f6220346ad21dfc4f351c57389e8430c0257)

mon: change "mon_pg_warn_min_per_osd" to uint64_t

* this silences a warning from -Wsign-compare.
* also switch all its user to the new-style option

Signed-off-by: Kefu Chai <kchai@redhat.com>
(cherry picked from commit c3cd0c13a1585388df0f17ab53c170e39c711615)

Merge pull request #18702 from liewegas/wip-pr-18670-luminous

luminous: qa/tasks/thrashosds: set min_in default to 4

Merge pull request #18596 from smithfarm/wip-21955-luminous

luminous: tests: add EC data pool to testing

Reviewed-By: Patrick Donelly <pdonelly@redhat.com>

rgw_file: introduce rgw_mount2 with a bucket name parameter

This explicitly allow mount a specified bucket in librgw.
Originally, mounting a bucket is implemented like a sub-directory
mount in nfs-ganesha FSAL_RGW with lookup_path.
With this change, we allow root_fh to points exactly to the root
of a mounted fs instance, which is a bucket or "/" rather than
always let root_fh points to "/".

Signed-off-by: Gui Hecheng <guihecheng@cmss.chinamobile.com>
(cherry picked from commit 860716e17e840df11a7e1d8b6ca0c1ee694b038b)

rgw_file: new fsid for fs instance

We use an in-memory fs_inst_counter for fsid currently, but it
inevitably cause nfs automount problem across a nfs-ganesha crash.
So here we relate the fsid to the s3 uid with a hash.
With this, fsid remains consistent across all fs instances and
across crashes.

We should be able to upgrade from old fsid to new fsid with
update_fh().

Signed-off-by: Gui Hecheng <guihecheng@cmss.chinamobile.com>
(cherry picked from commit 178b547e83a2c0eeb024bbe4ca2ff8d8e7d50572)

Merge pull request #18628 from batrick/i21953

luminous: mds: sanitize mdsmap of removed pools

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

Merge pull request #18626 from dillaman/wip-21968

luminous: cls/journal: fixed possible infinite loop in expire_tags

Reviewed-By: Mykola Golub <trociny@gmail.com>

Merge pull request #18688 from liewegas/wip-balancer-luminous

luminous: mgr/balancer: backport crush-compat work from master

Reviewed-By: John Spray <jspray@redhat.com>

rgw: fix BZ 1500904, Stale bucket index entry remains after object deletion

We have a race condition:

 1. RGW client #1: requests an object be deleted.
 2. RGW client #1: sends a prepare op to bucket index OSD #1.
 3. OSD #1:        prepares the op, adding pending ops to the bucket dir entry
 4. RGW client #2: sends a list bucket to OSD #1
 5. RGW client #2: sees that there are pending operations on bucket
                   dir entry, and calls check_disk_state
 6. RGW client #2: check_disk_state sees that the object still exists, so it
                   sends CEPH_RGW_UPDATE to bucket index OSD (#1)
 7. RGW client #1: sends a delete object to object OSD (#2)
 8. OSD #2:        deletes the object
 9. RGW client #2: sends a complete op to bucket index OSD (#1)
10. OSD #1:        completes the op
11. OSD #1:        receives the CEPH_RGW_UPDATE and updates the bucket index
                   entry, thereby **RECREATING** it

Solution implemented:

At step #5 the object's dir entry exists. If we get to beginning of
step #11 and the object's dir entry no longer exists, we know that the
dir entry was just actively being modified, and ignore the
CEPH_RGW_UPDATE operation, thereby NOT recreating it.

Signed-off-by: J. Eric Ivancich <ivancich@redhat.com>
(cherry picked from commit b33f529e79b74314a2030231e1308ee225717743)

Merge pull request #18569 from kmroz/wip-21939-luminous

luminous: list bucket which enable versioning get wrong result when user marker

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

Merge pull request #18591 from cbodley/wip-21938

luminous: rgw: fix error handling in ListBucketIndexesCR
(approved in irc & standups)

Reviewed-By: Abhishek Lekshmanan <abhishek@suse.com>

Merge pull request #18539 from linuxbox2/luminous-baixuyeyu-refcnt

luminous: baixuyeyu refcnt

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

Merge pull request #18516 from ceph/wip-bp-pr18193-luminous

qa/rgw: ignore errors from 'pool application enable'

Reviewed-By: Nathan Cutler <ncutler@suse.com>
Reviewed-By: Casey Bodley <cbodley@redhat.com>