ceph-volume: allow removable devices but exclude USB

Changed the logic to exclude USB devices to pass hot swap devices that were
passed through before the bug.

Fixes: https://tracker.ceph.com/issues/57907
Fixes: https://tracker.ceph.com/issues/58189
Fixes: https://tracker.ceph.com/issues/58306
Fixes: https://tracker.ceph.com/issues/58591
Signed-off-by: Kim Minjong <make.dirty.code@gmail.com>
(cherry picked from commit bd5e1a83495e31e457827f564c56fba23f4da8c9)

ceph-volume: fix a bug in _check_generic_reject_reasons

The types of removable and ro are wrong. Here, both filters are not
working at all. Changed this from integer to string and corrected the test
data.

Delete redundant logic in get_block_devs_sysfs. Given the name of the
function, I think it is correct to judge from _check_generic_reject_reasons,
and in fact it was before v17.2.4.

Fixes: https://tracker.ceph.com/issues/58591
Signed-off-by: Kim Minjong <make.dirty.code@gmail.com>
(cherry picked from commit a78e660728c6c0442cdbfa65db776b5856aee933)

ceph-volume: drop is_locked_raw_device()

This functions works for what it is supposed to do:

check if a device is busy.

That being said, this induces a race condition in `get_devices()`

Indeed, it does:

1/ `os.open()` with `(os.O_RDWR | os.O_EXCL)`
2/ `os.close()`

The second call has an effect: it triggers a udev event which causes
systemd-udevd to re-process the device. This seems to be a question of
millisecond but because of this, /sys (sysfs) isn't fully populated as
expected. Given that get_devices() collects a lot of details from sysfs
in a loop, some of these details can be missed.

ceph-volume overall doesn't make decisions based on `is_locked_raw_device()`
This detail is used only for reporting (inventory).
For this reason, dropping this function seems reasonnable.
As a compromise, we can check if the device has partitions and/or a FileSystem
on it.

Signed-off-by: Guillaume Abrioux <gabrioux@ibm.com>
(cherry picked from commit 2422ad867dff9d526d7e8be543178c897991097f)

ceph-volume/ceph_volume/util: ceph-volume multi-actuator drive attribute

As storage capacities grow, multi-actuator technology introduced by Seagate addresses the downward pressure on performance that comes with growing drive capacities and areal densities. Having multiple actuators enables drives to maintain the performance needs of customers with data-intensive applications. However, this innovation requires storage stack changes because a single hard drive is now represented by two or more independent actuators (independent_access_ranges) that transfer data concurrently and are represented by a single LBA address space.

This code addition to the `ceph-volume` command assists Ceph administrators in identifying drives with multiple actuators by utilizing the Linux kernel's multi-actuator support introduced in version 5.16. Dual-actuator hard drives are gaining market share and becoming more important to Ceph deployments on large discs.

The code has been tested with Seagate Osprey Exos 2X18 with Mach.2 technology.

$ sudo ceph-volume inventory /dev/sdd

====== Device report /dev/sdd ======

     path                      /dev/sdd
     ceph device               None
     lsm data                  {}
     available                 False
     rejected reasons          Has GPT headers
     device id                 ST18000NM0092-3CX103_MVV00H3J
     removable                 0
     ro                        0
     vendor                    ATA
     model                     ST18000NM0092-3C
     sas address
     rotational                1
     actuators                 2
     scheduler mode            mq-deadline
     human readable size       16.37 TB

Copyright (c) 2022 Seagate Technology LLC and/or its Affiliates

Signed-off-by: Michael English <michael.english@seagate.com>
(cherry picked from commit e34237990631bb817f885a02b6a4abff79125035)

ceph-volume/tests: add a unit test

In order to test that get_devices() returns the correct
number of actuators when such a device is being processed.

Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
(cherry picked from commit ea900373b524c321478fcc8f550090709dba2ecb)

Merge pull request #54652 from k0ste/wip-58829-pacific

pacific: mgr/dashboard: fix constraints.txt

Reviewed-by: Ernesto Puerta <epuertat@redhat.com>
Reviewed-by: Nizamudeen A <nia@redhat.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>

Merge pull request #54474 from k0ste/wip-63441-pacific

pacific: Fixing example of BlueStore resharding.

Reviewed-by: Laura Flores <lflores@redhat.com>
Reviewed-by: Joshua Baergen <jbaergen@digitalocean.com>

Merge pull request #53567 from pdvian/wip-59178-pacific

pacific: blk/kernel: Add O_EXCL for block devices

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

Merge pull request #52055 from trociny/wip-61433-pacific

pacific: rgw: multisite data log flag not used

Reviewed-by: Shilpa Jagannath <smanjara@redhat.com>

Merge pull request #52051 from trociny/wip-59610-pacific

pacific: rgw/sts: AssumeRole no longer writes to user metadata

Reviewed-by: Shilpa Jagannath <smanjara@redhat.com>

Merge pull request #51602 from alimaredia/wip-58584-pacific

pacific: rgw: set keys from from master zone on admin api user create

Reviewed-by: Jiffin Tony Thottan <jthottan@redhat.com>

Merge pull request #51600 from alimaredia/wip-61175-pacific

pacific: rgw/swift: check position of first slash in slo manifest files

Reviewed-by: Shilpa Jagannath <smanjara@redhat.com>

Merge pull request #51445 from k0ste/wip-59729-pacific

pacific: RGW: Solving the issue of not populating etag in Multipart upload result

Reviewed-by: Shilpa Jagannath <smanjara@redhat.com>

Merge pull request #51256 from k0ste/wip-55063-pacific

pacific: radosgw-admin: fix segfault on pipe modify without source/dest zone specified

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

Merge pull request #50540 from k0ste/wip-55149-pacific

pacific: rgw: Update "CEPH_RGW_DIR_SUGGEST_LOG_OP" for remove entries

Reviewed-by: Shilpa Jagannath <smanjara@redhat.com>

Merge pull request #49526 from dang/wip-58234-pacific

pacific: rgw: Fix truncated ListBuckets response.

Reviewed-by: Shilpa Jagannath <smanjara@redhat.com>

Merge pull request #46106 from BenoitKnecht/wip-55500-pacific

pacific: rgw: Avoid segfault when OPA authz is enabled

Reviewed-by: Shilpa Jagannath <smanjara@redhat.com>

Merge pull request #44471 from pritha-srivastava/wip-53648-pacific

pacific: rgw/sts: fix read_obj_policy permission evaluation

Reviewed-by: Shilpa Jagannath <smanjara@redhat.com>

Merge pull request #44464 from pritha-srivastava/wip-52778-pacific

pacific: rgw/sts: code to fetch certs using .well-known/openid-configuration URL

Reviewed-by: Shilpa Jagannath <smanjara@redhat.com>

Merge pull request #44463 from pritha-srivastava/wip-52785-pacific

pacific: rgw/sts: fixes getsessiontoken authenticated with LDAP,

Reviewed-by: Shilpa Jagannath <smanjara@redhat.com>

Merge pull request #53693 from pdvian/wip-62996-pacific

pacific: osd/OpRequest: Add detail description for delayed op in osd log file

Merge pull request #54434 from ifed01/wip-ifed-no-alloc-lba-align-pac

pacific: os/bluestore: get rid off resulting lba alignment in allocators

Reviewed-by: Adam Kupczyk <akupczyk@redhat.com>

Merge pull request #53464 from k0ste/wip-56649-pacific

pacific: pybind/mgr/autoscaler: Donot show NEW PG_NUM value if autoscaler is not on

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

Merge pull request #52203 from trociny/wip-61822-pacific

pacific: tools/osdmaptool: fix possible segfaults when there are down osds

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

Merge pull request #51262 from k0ste/wip-52841-pacific

pacific: osd: fix shard-threads cannot wakeup bug

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

Merge pull request #50552 from rzarzynski/wip-59085-pacific

pacific: mon: add proxy to cache tier options

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

Merge pull request #50194 from trociny/wip-58805-pacific

pacific: mgr: don't dump global config holding gil

Reviewed-by: Tim Serong <tserong@suse.com>
Reviewed-by: John Mulligan <jmulligan@redhat.com>

mgr/dashboard: fix constraints.txt

Replaces strict version matching with compatible version syntax.

Reference: https://peps.python.org/pep-0440/#version-specifiers

Fixes: https://tracker.ceph.com/issues/58827
Signed-off-by: Ernesto Puerta <epuertat@redhat.com>
(cherry picked from commit d5f8927a5cd4bfd9870f02fea1c66b32191d1fba)

Merge pull request #54192 from cfsnyder/wip-62621-pacific

pacific: mgr/dashboard: add 'omit_usage' query param to dashboard api 'get rbd' endpoint

Reviewed-by: Nizamudeen A <nia@redhat.com>

Merge pull request #54410 from mihalicyn/wip-63478-pacific

pacific: MClientRequest: properly handle ceph_mds_request_head_legacy for ext_num_retry, ext_num_fwd, owner_uid, owner_gid

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #54318 from batrick/wip-63414-pacific

pacific: mon: fix mds metadata lost in one case.

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #54245 from vshankar/wip-63283-pacific

pacific: client: always refresh mds feature bits on session open

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53716 from leonid-s-usov/bp/msg-wait/pacific

pacific: msg/AsyncMessenger: re-evaluate the stop condition when woken up in 'wait()'

Reviewed-by: Rishabh Dave <ridave@redhat.com>
Reviewed-by: Ilya Dryomov <idryomov@redhat.com>

Merge pull request #53634 from vshankar/wip-62584-pacific

pacific: mds: blocklist clients with "bloated" session metadata

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53552 from batrick/wip-62906-pacific

pacific: mds,qa: some balancer debug messages (<=5) not printed when debug_mds is >=5

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #54604 from lxbsz/wip-63586-p

pacific: osd: log the number of extents for sparse read

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

osd: add more debug logs for sparse read

This will be very important to get to know what exactly has happened
when client get a very large number of extents.

URL: https://tracker.ceph.com/issues/63586
Signed-off-by: Xiubo Li <xiubli@redhat.com>
(cherry picked from commit 1cf5ecb93fb73a5ed4a82935f635f98efccf7de6)

Merge pull request #54542 from aaSharma14/wip-63570-pacific

pacific: mgr/dashboard: Consider null values as zero in grafana panels

Reviewed-by: Nizamudeen A <nia@redhat.com>

mgr/dashboard: Consider null values as zero in grafana panels

After upgrading from RHCS4 to RHCS5..some of the grafana charts broke.
This is because in RHCS5 we do not generate the metrics if its value is
zero as a result the null value from that metric breaks the grafana
charts or graphs. This PR is to fix the above mentioned issue.

Fixes: https://tracker.ceph.com/issues/63088
Signed-off-by: Aashish Sharma <aasharma@redhat.com>
(cherry picked from commit 6f3f58cb8e4ce100cc7186858465b4d11d5c2c49)

Merge pull request #54010 from k0ste/wip-53152-pacific

pacific: radosgw-admin: allow 'bi purge' to delete index if entrypoint doesn't exist

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

Merge pull request #54160 from cbodley/wip-61351-pacific

pacific: rgw/s3: ListObjectsV2 returns correct object owners

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

Merge pull request #53764 from cbodley/wip-63043-pacific

pacific: rgw/keystone: EC2Engine uses reject() for ERR_SIGNATURE_NO_MATCH

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

Merge pull request #53443 from trociny/wip-61872-pacific

pacific: rgw: pick http_date in case of http_x_amz_date absence

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

Merge pull request #52996 from k0ste/wip-59026-pacific

pacific: rgw: use unique_ptr for flat_map emplace in BucketTrimWatche

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

Merge pull request #52605 from cbodley/wip-62138-pacific

pacific: rgw: s3 object lock avoids overflow in retention date

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

Merge pull request #54016 from k0ste/wip-57199-pacific

pacific: rgw: 'bucket check' deletes index of multipart meta when its pending_map is nonempty

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

Merge pull request #54014 from k0ste/wip-59361-pacific

pacific: rgw: fix rgw cache invalidation after unregister_watch() error

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

Merge pull request #53775 from cbodley/wip-63055-pacific

pacific: rgw: improve buffer list utilization in the chunkupload scenario

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

Merge pull request #53772 from cbodley/wip-63052-pacific

pacific: rgw: fix SignatureDoesNotMatch when extra headers start with 'x-amz'

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

Merge pull request #53769 from cbodley/wip-63049-pacific

pacific: rgw: s3website doesn't prefetch for web_dir() check

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

Merge pull request #53759 from cbodley/wip-63058-pacific

pacific: rgw: fix unwatch crash at radosgw startup

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

Merge pull request #53593 from trociny/wip-58478-pacific

pacific: rgw: fix FP error when calculating enteries per bi shard

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

Merge pull request #53474 from k0ste/wip-55701-pacific

pacific: radosgw-admin: don't crash on --placement-id without --storage-class

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

Merge pull request #53472 from k0ste/wip-57635-pacific

pacific: rgw: Drain async_processor request queue during shutdown

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

Merge pull request #53439 from k0ste/wip-62823-pacific

pacific: RadosGW API: incorrect bucket quota in response to HEAD /{bucket}/?usage

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

Merge pull request #53410 from trociny/wip-62308-pacific

pacific: rgw/sync-policy: Correct "sync status" & "sync group" commands

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

Merge pull request #53400 from trociny/wip-62751-pacific

pacific: rgw: fix 2 null versionID after convert_plain_entry_to_versioned

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

Merge pull request #53376 from jzhu116-bloomberg/wip-59692-pacific

pacific: rgw/notification: remove non x-amz-meta-* attributes from bucket notifications

Reviewed-by: Yuval Lifshitz <ylifshit@redhat.com>

Merge pull request #53356 from k0ste/wip-53658-pacific

pacific: rgw: fix UploadPartCopy error code when src object not exist and src bucket not exist

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

Merge pull request #52936 from k0ste/wip-58902-pacific

pacific: rgw: Fix Browser POST content-length-range min value

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

Merge pull request #52797 from cbodley/wip-62300-pacific

pacific: rgw: retry metadata cache notifications with INVALIDATE_OBJ

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

Merge pull request #52729 from theanalyst/wip-58817

pacific: rgw: swift : check for valid key in POST forms

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

Merge pull request #52113 from cbodley/wip-61728-pacific

pacific: rgw/beast: add max_header_size option with 16k default, up from 4k

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

Merge pull request #54167 from cbodley/wip-58238-pacific

pacific: rgw: beast frontend checks for local_endpoint() errors

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

Merge pull request #54120 from dparmar18/wip-63269-pacific

pacific: mds: report clients laggy due laggy OSDs only after checking any OSD is laggy

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53916 from kotreshhr/wip-63164-pacific

pacific: pybind/mgr/volumes: log mutex locks to help debug deadlocks

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53576 from mchangir/wip-57157-pacific

pacific: doc/cephfs: note regarding start time time zone

Reviewed-by: Rishabh Dave <ridave@redhat.com>
Reviewed-by: Anthony D'Atri <anthony.datri@gmail.com>

Merge pull request #53556 from batrick/wip-62731-pacific

pacific: mds: add event for batching getattr/lookup

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53555 from batrick/wip-62897-pacific

pacific: qa: lengthen shutdown timeout for thrashed MDS

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53550 from batrick/wip-62902-pacific

pacific: mds: log message when exiting due to asok command

Reviewed-by: Rishabh Dave <ridave@redhat.com>
Reviewed-by: Christopher Hoffman <choffman@redhat.com>

Merge pull request #53495 from lxbsz/wip-62859

pacific: mds: fix deadlock between unlinking and linkmerge

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53486 from batrick/wip-62854-pacific

pacific: qa: ignore expected cluster warning from damage tests

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53453 from joscollin/wip-62834-pacific

pacific: cephfs-top: include the missing fields in --dump output

Reviewed-by: Rishabh Dave <ridave@redhat.com>

os/bluestore: proper init sequent in BlueFS::mount()

Requires in pacific backport only due to the lack of
https://github.com/ceph/ceph/pull/41557

Signed-off-by: Igor Fedotov <igor.fedotov@croit.io>

Merge PR #52852 into pacific

* refs/pull/52852/head:
mds: remove calculating caps after adding revokes back
test/libcephfs: add test case for revoking caps
client: issue a cap release immediately if no cap exists
mds: add the revoking caps back to _revokes list
mds: move confirm_receipt() to Capability.cc

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

doc/rados/configuration/bluestore-config-ref: Fix lowcase typo

Letters "L P" got lowcased to "l p".
Reverted that.

Signed-off-by: Adam Kupczyk <akupczyk@ibm.com>
(cherry picked from commit 59a7f01b5c3cc48aa1e15ee44204d3552ffd593a)

qa: typo fix when checking for perf counter - s/md_thresh_evicted/mdthresh_evicted

Signed-off-by: Venky Shankar <vshankar@redhat.com>
(cherry picked from commit 5856a1e6b72dc77d3b4d9109385481b924e4fe36)

qa: lower mds_session_metadata_threshold for tests

... and increase the number of files that are created so as to
hit the threshold with a high probability.

Fixes: http://tracker.ceph.com/issues/62873
Signed-off-by: Venky Shankar <vshankar@redhat.com>
(cherry picked from commit 92200d9d10d30ed35f640fd1120be65577a3ce64)

PendingReleaseNotes: add a note for `mds_session_metadata_threshold` mds config

Signed-off-by: Venky Shankar <vshankar@redhat.com>
(cherry picked from commit ac3ab1a203ad65c7db07bc2c02acb5620800fb25)

test: add test to verify that a buggy client is blocklisted

... when its session metadata is bloated due to buildup of
`completed_requests`.

Signed-off-by: Venky Shankar <vshankar@redhat.com>
(cherry picked from commit 84df4b3d0c9e767a74cf5af80e8138239992df2c)

mds: add perf counter to track number of sessions evicted due to metadata threshold being exceeded

Signed-off-by: Venky Shankar <vshankar@redhat.com>
(cherry picked from commit 59dd587ddb3bc95a95d0fb8715511d194181ec6f)

mds: blocklist clients with "bloated" session metadata

Buggy clients (or maybe a MDS bug) causes a huge buildup of
`completed_requests` metadata in its session information.
This could cause the MDS to go read-only when its flushing
session metadata to the journal since the bloated metadata
causes the ODSOp payload to exceed the maximum write size.

Blocklist such clients so as to allow the MDS to continue
servicing requests.

Fixes: http://tracker.ceph.com/issues/61947
Signed-off-by: Venky Shankar <vshankar@redhat.com>
(cherry picked from commit bc6814d72a9fbec9c41ed75aee2314666cfca34b)

 Conflicts:
src/common/options/mds.yaml.in
src/mds/MDSRank.cc

Pacific uses old-style config file (src/common/options.cc), so
adjust to that.

Merge pull request #54268 from ronen-fr/wip-63372-pacific

pacific: osd: fix use-after-move in build_incremental_map_msg()

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

Merge pull request #53662 from lxbsz/wip-62523

pacific: ceph: allow xlock state to be LOCK_PREXLOCK when putting it

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53645 from vshankar/wip-61803-pacific

pacific: cephfs-journal-tool: disambiguate usage of all keyword (in tool help).

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53640 from vshankar/wip-62949-pacific

pacific: cephfs-mirror: do not run concurrent C_RestartMirroring context

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53362 from k0ste/wip-57110-pacific

pacific: mds: replacing bootstrap session only if handle client session message

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53270 from mchangir/wip-59001-pacific

pacific: cephfs_mirror: correctly set top level dir permissions

Reviewed-by: Rishabh Dave <ridave@redhat.com>

Merge pull request #53169 from leonid-s-usov/bp/cap-throttle-event/pacific

pacific: mds/Server: mark a cap acquisition throttle event in the request

Reviewed-by: Rishabh Dave <ridave@redhat.com>

os/bluestore: get rid off resulting lba alignment in allocators

Fixes: https://tracker.ceph.com/issues/62815
Signed-off-by: Igor Fedotov <igor.fedotov@croit.io>
(cherry picked from commit 5a98cdfdf932300b6e467e33f692c20a2f67f728)

 Conflicts:
src/os/bluestore/BtreeAllocator.cc
 * missong BtreeAllocator.cc

os/bluestore: adjust and validate bluefs_shared_alloc_size

Make sure it's in-sync (meaning it's higher or equal and properly aligned)
with bluestore_min_alloc_size into account

Signed-off-by: Igor Fedotov <igor.fedotov@croit.io>
(cherry picked from commit 33a87bda7f65a9c6d2f881cc0a69e1d98f21d14a)

os/bluestore: dump bluefs alloc unit sizes with perf counters dump

Signed-off-by: Igor Fedotov <igor.fedotov@croit.io>
(cherry picked from commit d17986306d23cfbf410434aaf789546707dd7dc9)

 Conflicts:
src/os/bluestore/BlueFS.cc
 * missed backports

client: always refresh mds feature bits on session open

Fixes: http://tracker.ceph.com/issues/63188
Signed-off-by: Venky Shankar <vshankar@redhat.com>
(cherry picked from commit 167fdd22e38d9febef8d7e0969403261314e3c98)

 Conflicts:
src/client/Client.cc

In pacific, the client would close the mds session if it has missing
features - adjust the code due to that. Also adjust for `session`
variable in Client::handle_client_session() as in pacific its of
type MetaSession* rather than MetaSesssionRef.

Merge pull request #54294 from ajarr/wip-63385-pacific

pacific: qa/suites/rbd: add test to check rbd_support module recovery

Reviewed-by: Mykola Golub <mgolub@suse.com>
Reviewed-by: Ilya Dryomov <idryomov@redhat.com>

Merge pull request #54293 from ajarr/wip-63382-pacific

pacific: mgr/rbd_support: fix recursive locking on CreateSnapshotRequests lock

Reviewed-by: Ilya Dryomov <idryomov@redhat.com>
Reviewed-by: Mykola Golub <mgolub@suse.com>

Merge pull request #54256 from pkalever/wip-63351-pacific

pacific: rbd-nbd: fix stuck with disable request

Reviewed-by: Mykola Golub <mgolub@suse.com>
Reviewed-by: Ilya Dryomov <idryomov@redhat.com>

MClientRequest: handle ext_num_retry and ext_num_fwd from ceph_mds_request_head_legacy

When a client is too old and uses struct ceph_mds_request_head_legacy we must
fill new ext_num_retry and ext_num_fwd fields from an old num_retry and num_fwd.

Fixes: https://github.com/ceph/ceph/pull/45669
Fixes: https://tracker.ceph.com/issues/63288
Fixes: commit cbd7e3040208 ("ceph_fs.h: add 32 bits extended num_retry and num_fwd support")
Signed-off-by: Alexander Mikhalitsyn <aleksandr.mikhalitsyn@canonical.com>
(cherry picked from commit 43f32a46aa9095b19525357ba7ca215e842b4f77)

mgr/rbd_support: remove CreateSnapshotRequests __del__()

There is no need for CreateSnapshotRequests.__del__() that calls
CreateSnapshotRequests.wait_for_pending().
MirrorSnapshotScheduleHandler.shutdown() already calls
CreateSnapshotRequests.wait_for_pending().

Signed-off-by: Ramana Raja <rraja@redhat.com>
(cherry picked from commit fed1e87685a698876cf167b3681327e5b0066ee6)

Conflicts:
       src/pybind/mgr/rbd_support/mirror_snapshot_schedule.py
 - Above conflict was due to commit e4a16e2
   ("mgr/rbd_support: add type annotation") not in pacific

mgr/rbd_support: fix recursive locking on CreateSnapshotRequests lock

The MirrorSnapshotScheduleHandler's run thread issues asynchronous
create snapshot requests using a CreateSnapshotRequests instance. When
the thread invokes a CreateSnapshotRequests instance's get_ioctx(),
the instance's class variable lock is acquired. With the class
variable lock held, the garbage collection of a CreateSnapshotRequests
instance may race in the thread. The thread would then call
CreateSnapshotRequests __del__() that tries to acquire the class
variable lock that the thread already holds. Fix this
recursive deadlock by converting the CreateSnapshotRequests lock from
a class variable to an instance variable. There is no need to share
the lock across CreateSnapshotRequests instances.

Also convert MirrorSnapshotScheduleHandler, PerfHandler and
TrashPurgeScheduleHandler class variables to instance variables
that don't need to be shared across the instances.

Fixes: https://tracker.ceph.com/issues/62994
Signed-off-by: Ramana Raja <rraja@redhat.com>
Co-Authored-By: Ilya Dryomov <idryomov@gmail.com>
(cherry picked from commit 4452bc22d1c6c8499cf55d6e39090adf7ae1dcbf)

 Conflicts:
src/pybind/mgr/rbd_support/mirror_snapshot_schedule.py
src/pybind/mgr/rbd_support/perf.py
src/pybind/mgr/rbd_support/trash_purge_schedule.py
 - Above conflicts were due to commit e4a16e2
   ("mgr/rbd_support: add type annotation") not in pacific

Merge pull request #53970 from Matan-B/wip-63179-pacific

pacific: osd/OSD: introduce reset_purged_snaps_last

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