mgr/smb: update fs.py to request all caps for the cephfs volume

Update the authorizer class in fs.py to request all caps (*) for our
volume. This is necessary in order to make use of the fscrypt feature
from smbd.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

cephadm: add pid file directory option to cfg watch sidecar

This will enable the config watch sidecar to signal processes
with a SIGHUP to tell them to reload configuration when config
watch has detected a configuration change. Currently only used
by keybridge.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

cephadm: add pidfile option to smb keybridge sidecar

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/smb: fix unit test assertion

The unit test was looking for the wrong string (but the right
condition). This fixes the string to check.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

doc: add documentation for keybridge and fscrypt options

Add docs for the keybrige configuration and cephfs fscrypt options
added to the smb mgr module resource definitions.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/smb: add some keybridge related unit test cases

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/smb: add support for generating keybridge configuration

Add support for generating the sambacc configuration section for
keybridge. Add support for configuring smb shares for keybridge access.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/smb: add cross-check validation for keybridge scopes

Validate that scope names are not re-used, etc. Check on things that
can't be done in single object validation.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/smb: add new cephfs parameter for getting fscrypt keys

Add a new field to the cephfs configuration section for shares. This
section selects the keybridge scope and key name to use when acquiring
the key to use for fscrypt.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/smb: add keybridge configuration to cluster resource

Add keybridge service configuration classes and parameters to the
resources module. This supports enabling the keybridge, setting up
scopes for the keybridge and it's access control.

A helper class is added that parses and helps validate the scope names.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/smb: add enums that will be used for configuring keybridge

Add a pair of enum types that will be used for configuring the
keybridge. The scope type identifies what kind of scope is being
used. The peer policy can be used to allow a dev or other user
more access to the keybridge api for development purposes.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/smb: add raw data methods to MemConfigStore

Add the set_data/get_data methods to the MemConfigStore so that future
test updates will not fail to save tls credential objects.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/smb: fix a resource error unpacking str instead of list

Add special handling for the case where a string is passed instead of a
list. Without this fix a string will be converted into a list of single
letter items, something pretty much no one ever wants. Raise an
exception instead.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

cephadm: add keybridge sidecar to smb daemon module

The keybridge uses the sambacc configuration but can also be passed
CLI options. Since cephadm writes the cert files, cephadm must also
pass the file names to use to the container args.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

mgr/cephadm: enable setting up SSL/TLS files for keybridge sidecar

Signed-off-by: John Mulligan <jmulligan@redhat.com>

python-common/deployment: add keybridge feature to smb service spec

The keybridge sidecar is enabled by the keybridge feature flag.
This sidecar will be used to help fetch keys over various protocols
for the ceph module to use to set up fs encryption.

Signed-off-by: John Mulligan <jmulligan@redhat.com>

Merge pull request #67314 from aainscow/block_non_4k_chunks

mon: Deny EC optimizations (fast EC) for non-4k-aligned chunk-sizes.

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Bill Scales <bill_scales@uk.ibm.com>

Merge pull request #67723 from perezjosibm/wip-perezjos-tracker75202

qa/workunits: add crimson/test_crimson_dump_metrics.sh

Merge pull request #67936 from shraddhaag/wip-shraddhaag-fix-crimson-qa

qa/distros/crimson-supported-all-distro: reinstate centos

Reviewed-by: Matan Breizman <mbreizma@redhat.com>

Merge pull request #57395 from rzarzynski/wip-wip-crimson-ec-with-recovery

crimson/osd: bring support for EC, including recovery

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

qa/distros/crimson-supported-all-distro: reinstate centos

Signed-off-by: Shraddha Agrawal <shraddha.agrawal000@gmail.com>

Revert "crimson/osd, test: turn the EC tests on for crimson"

This reverts commit 36fc26b2b803fcf425de674a578c4aba20a5d9ea.

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

crimson/osd: fixes after crimson::os::with_store<>()

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

crimson/osd: Avoid pwlc spanning intervals

This is a crimson port of `dc0a195b937eb175cf663d48a856f486435e987f`.

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

crimson/osd: improve !is_omap() logging

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

crimon/osd: drop present_shards from PGLog handler's partial_write()

In correspondence with classical's 880a17e39626d99a0b6cc8259523daa83c72802c.

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

crimson/osd: improve debugs in ECTransaction::Generate::Generate()

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

crimon/osd: implement PG::PGLogEntryHandler::partial_write() for FastEC

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

crimson/osd: port changes from classial FastEC

This commit ports changes from c116b8615d68a3926dc78a4965cc0a28ff85d4f2.

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

tests: LibRadosIoEC allows for overwrites / unalligned writes on crimson

1. crimson-osd does support solely the FastEC flavor of EC pools;
   SlowEC isn't now and won't be implemented.
2. FastEC imposes the EC overwrites support.
3. When run against crimson's EC implementation, the tests shall
   not enforce e.g. ENOTSUPP on overwrite.

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

crimson/osd: implement handling of RADOS' extended attirbutes in FastEC

Support for `attr_cache` in replicated backend is planned for later.

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

crimson/osd: drop unused variable from PGBackend::getxattr()

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

crimson/osd: drop unneccessary variant of PGBackend::getxattr()

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

crimson/osd: don't decode EC only attrs are needed

This a crimson coutnerpart of 417fb71c9b5628726d3217909ba1b6d3e7bf251a.

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

rmw_pipeline in RMWPipeline::Op

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

fixup: crimson/osd: settle ECRecoveryBackend -- local MOSDPGPush

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

fixup: crimson/osd: settle RMWPipeline and RMWPipeline within ECBackend

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

fixup: crimson/osd: wire MOSDECSubOpWriteReply up with ECBackend

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

fixup: crimson/osd: route EC messages over PG

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

crimson/osd: fix unused variable warnings pg.cc and seastore

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

ECCommon: fixups for crimson's FastEC support

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

fixup: crimson/osd: bring ECRecoveryBackend, phase 0.1

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

fixup: crimson/osd: translate for os::Transaction into ECTransaction

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

fixup: crimson/osd: switch return type of PG::get_dpp to unify with ECListener

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

crimson/osd: switch return type of PG::get_dpp to unify with ECListener

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

crimson/osd: settle ECExtentCache within ShardService

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

crimson/osd: PGBackend::_read() takes object_size for the sake of FastEC

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

fixup: fixup: crimson/osd: add support for MOSDECSubOpRead

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

fixup: crimson/osd: wire MOSDECSubOpWriteReply up with ECBackend

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

fixup: crimson/osd: implement ECBackend::handle_rep_read_reply()

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

fixup: crimson/osd: translate for os::Transaction into ECTransaction

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

fixup: crimson/osd: implement ECBackend::objects_read_and_reconstruc

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

fixup: crimson/osd: settle RMWPipeline and RMWPipeline within ECBacken

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

fixup: crimson/osd: settle RMWPipeline and RMWPipeline within ECBackend

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

crimson/osd: settle RMWPipeline and RMWPipeline within ECBackend

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

crimson/osd: tackle the get_write_plan() problem of ECBackend

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

fixup: crimson/osd: settle ECRecoveryBackend -- misc small bits of ECListener

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

fixup: crimson/osd: bring ECRecoveryBackend, phase 0.1

Per ecda6780c93db66865533a27dab5b0beb052f5fa.

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

crimson/osd, test: turn the EC tests on for crimson

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

Revert "crimson/.../object_context: drop recovery_read_marker"

This reverts commit c1b7435d6c3a598ee13dd57ba8b83014c4cca893.

EC support needs `wait_for_recovery()`.

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

osd: ECCommon respects crimson non-null obc for new objects

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

crimson/osd: bring get_temp_recovery_object() to PGRecovery for EC

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

crimson/osd: settle ECRecoveryBackend -- misc small bits of ECListener

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

crimson/osd: settle ECRecoveryBackend -- local MOSDPGPush

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

crimson/osd: settle ECRecoveryBackend -- bring maybe_load_obc()

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

crimson/osd: make objctx loading agnostic about source of attrs

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

crimson/errorator: helpers for tl::expected

`ErrorHelper` is supposed to facilitate transformation of `std::error_code`
(and similar) into an immediately failed errorated `future` with extra error
checking.

The main idea is to let something like:

```cpp
my_ertr_t::future<my_value_t> foo() {
  // ...
  // `maybe_decoded` here is an instance of `tl::expected` carrying `std::error_code`
  // in its error part
  if (!maybe_decoded) {
    // `std::error_code` can convey errors `my_ertr_t` does not allow – in that case,
    // it would be helpful to get at least a run-time assert.
    return ErrorHelper<my_ertr_t>::from_error<my_value_t>(maybe_decoded.error());
  }
  // ...
}
```

Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>
(cherry picked from commit a7ffc18eb40b9145fca5f064c4b3d097bffeb8d0)

crimson/*: rename errorator-loop.h into errorator-utils.h

Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>
(cherry picked from commit 682e4ff4322a895a0bdb6c9cc0d00a516499baad)

crimson/osd: delegate md loading outside of ObjectContextLoader::load_obc()

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

crimson/osd: dissect metadata decoding from PGBackend::load_metadata()

This loosens the coupling between loading and decoding making
the latter reusable with metadata coming from other source than
local object store.

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

crimson/osd: make OSD::osd_stat compliant with multi-core

The EC recovery backend needs `inc_osd_stat_repaired()`.

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

crimson/osd: add support for reads over EC pool

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

crimson/osd: fix compiler warnings on copy elision being prevented

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

crimson/osd: on the cloning path bump obj version closer to its usage

My understanding at the time of writing is this change is a pure
refactoring.

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

crimson/osd: fix std::moving in ShardServices

```
/home/rzarzynski/ceph2/src/crimson/osd/shard_services.cc:595:9:   required from here
/home/rzarzynski/ceph2/src/crimson/osd/shard_services.cc:596:10: warning: moving a temporary object prevents copy elision [-Wpessimizing-move]
  596 |     auto [pool, name, ec_profile] = std::move(std::get<0>(ret).get0());
      |          ^~~~~~~~~~~~~~~~~~~~~~~~
/home/rzarzynski/ceph2/src/crimson/osd/shard_services.cc:596:10: note: remove ‘std::move’ call
/home/rzarzynski/ceph2/src/crimson/osd/shard_services.cc:597:10: warning: moving a temporary object prevents copy elision [-Wpessimizing-move]
  597 |     auto coll = std::move(std::get<1>(ret).get0());
      |          ^~~~
/home/rzarzynski/ceph2/src/crimson/osd/shard_services.cc:597:10: note: remove ‘std::move’ call
```

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

crimson/osd: implement ECRecoveryBackend::commit_txn_send_replies()

This bypasses the crimson-msgr for local ops.

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

crimson/osd: implement ECRecoveryBackend::recover_object()

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

crimson/osd: wire up MOSDPGPush and MOSDPGPushReply in ECRecoveryBackend

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

crimson/osd: ECReplicatedBackend embraces ECCommon::RecoveryBackend

The idea is to compose the latter into the former (thus the private
inheritance) to easily build a class that fulfills the crimon iface
called `crimson::ReplicatedBackend`.

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

crimson/osd: settle ECCommon::RecoveryBackend within crimson-osd

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

fixup: crimson/osd: compile and link with ECTransaction.cc

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

fixup: crimson/osd: add support for error handling around MOSDECSubOpRead

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

Revert "crimson: clarify RecoveryBackend::handle_backfill_op"

This reverts commit ecda6780c93db66865533a27dab5b0beb052f5fa.

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

crimson/osd: wire MOSDPGPush{,Reply} with ECRecoveryBackend, phase 2

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

crimson/osd: wire MOSDPGPush{,Reply} with ECRecoveryBackend

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

crimson/osd: bring factory to RecoveryBackend

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

crimson/osd: bring ECRecoveryBackend, phase 0.1

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

crimson/osd: conceptually move delete handling out of ReplicatedRecoveryBackend

This logic is shareable between ReplicatedRecoveryBackend and
coming ECRecoveryBackend.

The switch happens currently only structural level; shuffling
betweeen source files will happen later.

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

crimson/osd: bring ECRecoveryBackend, phase 0

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

crimson/osd: drop the green thread for ECBackend::submit_transaction()

Signed-off-by: Radosław Zarzyński <rzarzyns@redhat.com>

crimson/osd: ECBackend::handle_sub_write() handles the end of local write

Signed-off-by: Radosław Zarzyński <rzarzyns@redhat.com>

crimson/osd: improve debugs in ECBackend::handle_rep_write_reply

Signed-off-by: Radosław Zarzyński <rzarzyns@redhat.com>

crimson/osd: handle_rep_write_reply() takes ECSubWriteReply

Taking this ECSubWriteReply instead of MOSDECSubOpWriteReply
makes this method easier to reuse for handling local writes.

Signed-off-by: Radosław Zarzyński <rzarzyns@redhat.com>

crimson/osd: extend PG::should_send_op() and move it to .cc

Signed-off-by: Radosław Zarzyński <rzarzyns@redhat.com>

crimson/osd: implement ECBackend::objects_read_and_reconstruct

Signed-off-by: Radosław Zarzyński <rzarzyns@redhat.com>

crimson/osd: implement submit_transaction() of ECBackend

Signed-off-by: Radosław Zarzyński <rzarzyns@redhat.com>

crimson/osd: PCBackend::_submit_transaction() takes ObjectContext

For the sake of ECBackend.

Signed-off-by: Radosław Zarzyński <rzarzyns@redhat.com>

crimson/osd: bring attr cache to ObjectContext

Signed-off-by: Radosław Zarzyński <rzarzyns@redhat.com>

crimson/osd: compile and link with ECTransaction.cc

Signed-off-by: Radosław Zarzyński <rzarzyns@redhat.com>

crimson/osd: drop the mock of ECTransaction-for-crimson from ECCommon

Signed-off-by: Radosław Zarzyński <rzarzyns@redhat.com>