* added timeout options to adhoc and console
* added test
* fix typosesz
* fix conflict
* task_timeout
* fix timeout option, added extra vars to console
* actually use right cli switch .. DUH!
* added timeout to include 'valid' but ignored keys
* fix default
* fixes per review
* Add DarwinStrategy class and integration tests
macOS has three seprate hostname params that need to be set. One of those params, LocalHostName, has more stringent requirements than the other two, which accept special characters and spaces. Create a method to scrub the hostname to ensure it works well with the system requirements.
* Update documentation
* Account for virtualization type returned on Azure Pipelines
* Do not be dependent on order of self.name_types
Use the scrubbed name when the name type is LocalHostName
A bug existed in systemd 245 that did not properly handle unknown kernel
capabilities gracefully. This resulted in incomplete output when querying
for the service status. It is possible to get service status by other means.
This PR works around this issue by getting service status using other commands
in the event of a failure due to this bug.
Change:
- Cryptography 3.2 drops support for OpenSSL 1.0.2. Some of our CI
infrastructure still uses this version (FreeBSD, namely). For now,
just add a constraint to use old cryptography.
Test Plan:
- CI
Signed-off-by: Rick Elrod <rick@elrod.me>
* GitHub is removing the underlying API used to implement the `login` command. Since the general consensus seems to be that relatively nobody currently uses this command (in favor of explicit token passing), support was simply removed for interactive login. If a future need arises, this command should be reimplemented via OAuth Device Auth Grants.
* login or role login commands now produce a fatal error with a descriptive message
* updated 2.10 and 2.11 porting guide entries
* remove dead code/config, update messages and porting guides
* Fix ansible-test docker container detection.
* Attach test containers to the correct network.
* Do not assume `localhost` for accesing Docker.
* Look for containers on current network.
* Always map /var/run/docker.sock into containers.
This fixes issues when using a remote Docker host.
* Support container IP lookup from networks list.
* Fix container network attachment.
* Remove redundant container detection messages.
* Limit DOCKER_HOST parsing to TCP.
* Restore docker socket existence check.
The check is skipped if the docker hostname is not localhost.
* Correct changelog entry.
* Fix the reversed order of return values
AnsibleModule.run_command returns a tuple of return code, stdout and stderr.
The module main function of the user module expects user.create_user to
return a tuple of return code, stdout and stderr.
Fix the locations where stdout and stderr got reversed.
* Fix another missed occurance
* Add changelog
* Fix further reversed order of return values occurrences
Followup to 72c3ba62c8e2606a6a35d51075fba29ce5faf34e:
Fix further reversed order of return values occurrences.
Change:
- containerd is now recognized as container tech
Test Plan:
- Tested with Docker inside Github Actions in a private repo:
"ansible_virtualization_role": "guest",
"ansible_virtualization_tech_guest": [
"containerd",
"container",
"VirtualPC"
],
"ansible_virtualization_tech_host": [],
"ansible_virtualization_type": "containerd",
Tickets:
- Fixes#66304 because this is what docker containers show up as
in Github Actions.
Signed-off-by: Rick Elrod <rick@elrod.me>
* Ensure we call action_loader.get with collection_list. Fixes#72170
* Add tests and changelog
* Remove grep, do assertion in playbook. ci_complete
* Skip old jinja2 versions
* ci_complete
* dedupe
make collections whitelist follow normal flow
* fixes missing set_options call and adhoc and stdout processing rules
* avoid dupes
* fixed to handle redirects
* also updated tests with new and more accurate skip message
* fix callback tests for envs with cowsay installed
* lots MOAR comments on why the code is as it is, some todos to refactor in future
Co-authored-by: Matt Davis <nitzmahone@users.noreply.github.com>
Change:
- When a plugin defines `type: str` on a parameter, treat more kinds of
input as a string instead of whatever it is parsed as.
Test Plan:
- New unit tests
- CI
Signed-off-by: Rick Elrod <rick@elrod.me>
* Add support for GSSAPI/Kerberos to urls.py
* Test out changes with the latest test container
* Get remote hosts working
* Fix up httptester_krb5_password reader
* Fix tests for opensuse and macOS
* Hopefully last lot of testing changes
* Dont do CBT on macOS
* Fixes from review
Change:
- No longer fail due to old Fedora/RHEL and our failure to try to cast
gids to integers before trying to pass them to getgrgid() before
trying to use them.
- Add tests for user/mode for various unarchive formats.
Test Plan:
- New integration tests, ran against centos6 container
Tickets:
- Fixes#71903
Change:
- Make KubeVirt be identified as such, and with a guest role.
- This is a breaking change, since playbooks could be relying on the
incorrect behavior previously.
Test Plan:
- Issue author
Tickets:
- Fixes#72001
Signed-off-by: Rick Elrod <rick@elrod.me>
Change:
- podman > 2 && < 2.2 does not support "images --format {{json .}}"
- podman also now outputs images JSON differently than docker
- Work around both of the above.
Test Plan:
- Tested with podman 2.0.6 in Fedora 31.
Signed-off-by: Rick Elrod <rick@elrod.me>
Co-authored-by: Sviatoslav Sydorenko <wk.cvs.github@sydorenko.org.ua>
Change:
- Use `sysctl -n` for openbsd uptime information
- Allow `get_sysctl()` to account for multi-line sysctl settings
- Add unit tests for `get_sysctl()`
Test Plan:
- New unit tests
Tickets:
- Fixes#71968
- Refs #72025
- Refs #72067
Signed-off-by: Rick Elrod <rick@elrod.me>
Co-authored-by: Brian Coca <brian.coca+git@gmail.com>
Co-authored-by: Brian Coca <bcoca@ansible.com>
The luseradd / lusermod commands do not support the -e option. Set
the expiry time in this case via lchage after the user was
created / modified.
Fixes: #71942
In Python3 math.floor returns an integer whereas Python2 returns a float.
Hence always convert the result of math.floor to an int to ensure that
lexpires is an integer.
Move local expires tests in a separate file and import the tasks to the
main.yml to keep main.yml smaller.
b6b238a fixed the SLES4SAP detection, which was at this time ok.
Sadly Suse changed with SLES 15 the /etc/os-release file, so the above
change will no longer work.
This commit updates the SLES4SAP detection regarding
https://www.suse.com/support/kb/doc/?id=000019341.
The symlink realpath is matched with endswith, because in SLES 12+ the
link target is SLES_SAP.prod, but in SLES 11 the link target is
SUSE_SLES_SAP.prod.
* Validate removal versions.
* Validate that removal collection versions and version_added collection versions conform to semver spec.
* Validate removal version numbers in meta/runtime.yml.
* Stricter validation for isodates (f.ex. YYYY-M-D is not allowed).
* Improve error reporting.
* Validate removal collection versions.
Co-authored-by: Andrew Klychkov <aaklychkov@mail.ru>
* Let get_file_attributes() work without `lsattr -v`
Change:
- module_utils's get_file_attributes() expects `lsattr -v` to work, but
in some cases, it may not.
- The function now takes an optional include_version bool parameter,
which removes this expectation.
- Places where we call get_file_attributes() without using the 'version'
it returns, we now call it with include_version=False.
Test Plan:
- New unit tests
Signed-off-by: Rick Elrod <rick@elrod.me>
* Reorder test/sanity/ignore.txt to better see changes
* Remove extra empty line causing pass fail
* Apply suggestions made by Akasurde
* Minor fix in package_facts and remove two unnecessary ignores
* Fix subversion based on a suggestion made by felixfontein
* Apply suggestions made by felixfontein and Andersson007
* Fix subversion.py as suggested by felixfontein
* Minor reformatting in yum_repository description
* Reformat changelog
* Add key to apt_key, add deprecated changelog
* Add PR url to changelog
* Ignore paramater-type-not-in-doc in favour of adding key back to apt_key
* Fix apt_key
* Remove undocumented-paramater from apt_key ignore
* Ignore doc-choices-do-not-match-spec in package_facts
* Fix package_facts
* Fix filter option in setup module
Change:
- In some cases (always with free strategy, sometimes with linear), the
default callback would not show the task banner for include_tasks.
- This only affects the include_tasks task itself, not the tasks in the
included file.
Test Plan:
- Updated default callback tests
Tickets:
- Fixes#71277
Signed-off-by: Rick Elrod <rick@elrod.me>
Change:
- Previously when `security: true` and `bugfix: true` were both given,
only security updates would get applied. Filters now accumulate so
that both get applied in this case.
Test Plan:
- New integration tests for both check_mode and not. These tests make
use of a contrived yum repository which is stored in S3.
Tickets:
- Fixes#70854
Signed-off-by: Rick Elrod <rick@elrod.me>
Co-authored-by: Matt Martz <matt@sivel.net>
Co-authored-by: Matt Martz <matt@sivel.net>
* fix coverage output from synthetic packages
* synthetic packages (eg, implicit collection packages without `__init__.py`) were always created at runtime with empty string source, which was compiled to a code object and exec'd during the package load. When run with code coverage, it created a bogus coverage entry (since the `__synthetic__`-suffixed `__file__` entry didn't exist on disk).
* modified collection loader `get_code` to preserve the distinction between `None` (eg synthetic package) and empty string (eg empty `__init__.py`) values from `get_source`, and to return `None` when the source is `None`. This allows the package loader to skip `exec`ing things that truly have no source file on disk, thus not creating bogus coverage entries, while preserving behavior and coverage reporting for empty package inits that actually exist.
* add unit test
On some systems (tested with official Debian Buster-based Python 3.6
docker image), setting the LC_ALL environment variable to en_US.UTF-8
will trigger Python into switching its preferred encoding to ASCII.
If any python process tries to read a non-ASCII file in this scenario,
it will terminate with an error.
And this is exactly what happens to pytest when it tries to load its
configuration that ansible-test supplies because the configuration
contains an em dash.
In order to bypass this issue, we replaced the em dash with a regular
dash and things started working again.
An alternative solution would be to replace the en_US.UTF-8 locale with
something safer, but unfortunately, the en_US.UTF-8 is probably as safe
as it gets.
The setup module can now filter out multiple pattern by providing a list
to the filter parameter instead of just a string. Single string sill
works. Previous behaviour remains.
(cherry picked from commit b5c36dac483fdd74d6c570d77cc8f3e396720366)
* Use InventoryManager for inventory_hostnames lookup. Fixes#17268
* Add test for using patterns
* s/it's/its/
* Allow bypassing parse_sources in InventoryManager
* Remove unneeded method call
Change:
- By default the dnf API does not gpg-verify packages. This is a feature
that is executed in its CLI code. It never made it into Ansible's
usage of the API, so packages were previously not verified.
- This fixes CVE-2020-14365.
Test Plan:
- New integration tests
Signed-off-by: Rick Elrod <rick@elrod.me>
* Validate salt when using crypt. Respect salt_size in password lookup. Repair salt for bcrypt. Fixes#71107. Fixes#53750. Fixes#36129.
* Handle algorithms we don't know about, and make sure to return the salt
* Account for old passlib
* Add tests for salt constraints
* Add changelog fragment
* Add test for #36129
* added configurable dir for tree callback
- allows usage in playbooks
- also made errors more specific if bad dir is supplied
- ensure we expand paths
* docfix
Change:
- Now sends meta tasks to the task start callback
- Lets callback plugins opt-in to receiving implicit tasks
Test Plan:
- New integration tests
Tickets:
- Indirectly fixes#71007 by allowing custom callbacks with this data
Signed-off-by: Rick Elrod <rick@elrod.me>
skip_missing parameter in subelements lookup plugin is accepted from
inside the dictionary.
Fixes: #38182
Signed-off-by: Abhijeet Kasurde <akasurde@redhat.com>
* check run state of current block only
* Add changelog and test
* Add test for issue 29047
* Fix for both tests
* blerg
* Change test messages
* fix tests
* Add multi-level block in rescue test case
* Add recursive rescue check and multi-level test
* Should probably run the new test
* ci_complete
* Merge new tests
* ci_complete
* Don't do conflict check on sdist and egg_info. Fixes#71279
* Add changelog
* adjust changelog PR link
Co-authored-by: Rick Elrod <rick@elrod.me>
* add warning about bdist_wheel
Co-authored-by: Rick Elrod <rick@elrod.me>
This would be a partial solution for #69364 in that the SHASUMS file can be downloaded and gpg verified but then used from the downloaded location to verify the get_url's file.
* Make checksum url parsing more explicit
Use urlsplit to test if the checksum string has a (currently tested and) supported url scheme.
* Fix whitespace
* Changelog fragment
* Added tests
* Fix typo in test setup
The message generated by systemctl has been updated in 9321e23c40, which requires a corresponding change in the systemd module.
In addition, this fixes the module when the SYSTEMD_OFFLINE environment variable is set.
* POC for supporting callback events that come from the worker
* linting fixes. ci_complete
* fix up units. ci_complete
* Try moving the sentinel put higher. ci_complete
* safeguards. ci_complete
* Move queue killing to terminate
* LINTING. ci_complete
* Subclass Queue, to add helper send_callback method
* Just use _final_q instead of adding another queue and thread
* Revert a few changes
* Add helper for inserting a TaskResult into the _final_q
* Add changelog fragment
* Address rebase issue
* ci_complete
* Add test to assert async poll callback from fork
* Don't use full path
* ci_complete
* Use _results_lock as a context manager
* Add new generic lock decorator, and use it with send_callback
* Revert "Change default file permissions so they are not world readable (#70221)"
This reverts commit 5260527c4a.
* Revert "Fix warning for new default permissions when mode is not specified (#70976)"
This reverts commit dc79528cc6.
* Ensure -k is set to delegated hosts without a pass
* Fix up some broken tests
* Update task_executor.py
one possible fix, the other is updating winrm to normalize on 'password' like the other connection plugins
* Add alias for winrm and fix incorrect assumption
* Make sure aliases are used for keyword options
* Conditionally run test if sshpass is present, fix sanity
Co-authored-by: Brian Coca <bcoca@users.noreply.github.com>
Change:
- Allow systems to declare multiple virt techs. For example if a system
is both a docker container, but virtualized on KVM, show both. If a
system is set up to run virtualbox and KVM VMs, show both.
- This is done by introducing new facts keys:
- virtualization_tech_guest
- virtualization_tech_host
- Backwards compatibility is preserved by keeping track of the previous
return-points and refusing to update those keys after we would have
returned, but now returning them at the end, so that the new keys can
accumulate their data.
Test Plan:
- Local
- CI
Tickets:
- Refs #66304
- Refs #17151
- Refs #17058
- Probably others
Signed-off-by: Rick Elrod <rick@elrod.me>
* Fix tty_ify bugs and refactor
* Move tty_ify() and supporting attributes to the DocCLI class as that's
the only thing using it.
* Add unittest for the code.
* Fix a bug where the substitution macros can be detected when they are
a part of another word.
* Add support for L(), R(), and HORIZONTALLINE which were added to the
website docs many years ago.
* Update test/units/cli/test_doc.py
Co-authored-by: Matt Clay <matt@mystile.com>
Co-authored-by: Matt Clay <matt@mystile.com>
* add jinja2 global to reserved names
also allow expansion by additional context provided from caller
fixes#41955
Co-authored-by: Abhijeet Kasurde <akasurde@redhat.com>
Change:
- Use `chmod +a` in the fallback chain to allow MacOS to use ACLs to
allow an unprivileged user to become an unprivileged user.
Test Plan:
- CI, new tests
Tickets:
- Fixes#70648
Signed-off-by: Rick Elrod <rick@elrod.me>
Change:
- Add a parameter `read_whole_file` which allows for reading the whole
file when doing a `contains` regex search.
- This allows for (for example) matching a pattern at the very end of
a file.
Test Plan:
- New integration tests
Tickets:
- Fixes#63378
Signed-off-by: Rick Elrod <rick@elrod.me>
Change:
- Previously CachePluginAdjudicator#flush only removed entries from the
cache backend that it knew about by using them earlier. Now it calls
the underlying plugin's flush() method.
Test Plan:
- New unit tests
Tickets:
- Fixes#68770
Signed-off-by: Rick Elrod <rick@elrod.me>
* Strip spaces in module names in explicit actions
Change:
- When an action is called like "action: copy foo=bar", strip spaces
around the action name.
- This allows "action: copy foo=bar" to work as expected.
Test Plan:
- New integration tests
Tickets:
- Fixes#62136
Signed-off-by: Rick Elrod <rick@elrod.me>
Add integration tests for unarchiving as unprivileged user
Break tasks into separate files for easier reading and maintenance
Create a user by specifying a default group of 'staff' for macOS.
The user module does not actually remove the user directory on macOS,
so explicitly remove it.
Put the removal tasks in an always block to ensure they always run
Co-authored-by: Philip Douglass <philip.douglass@amadeus.com>
Co-authored-by: Sam Doran <sdoran@redhat.com>
The iso8601_micro and iso8601 facts incorrectly called now.utcnow(), resulting
in a new timestamp at the time it was called, not a conversion of the previously
stored timestamp.
Correct this by capturing the UTC timestamp once then calculating the local
time using the UTC offset of the current system.
* Use time.time() for getting the current time
* Convert from that stored epoch timestamp to local and UTC times
* Used existing timestamp for epoch time
* Add unit tests that validate the formate of the return value rather than an exact value since mocking time and timezone is non-trivial
Follow up to #70221
Related to #67794
CVE-2020-1736
When set_mode_if_different() is called with mode of 'None', ensure we issue
a warning about the change in default permissions.
Add integration tests to ensure the warning works properly.
* Fix tests
- actually use custom module 🤦♂️
- verify file permission on created files
- use remote_tmp_dir so we're ready for split controller
- improve test module so we can skip the call to set_fs_attributes_if_different()
- fix tests for CentOS 6
Change:
- Allow older git to verify tags again
- Enable verification tests everywhere, even if most of them only work
on newer git. Some of them work on older git and they test the --raw
parameter.
Test Plan:
- Re-enabled subset of git tests
Tickets:
- Fixes#64469
Signed-off-by: Rick Elrod <rick@elrod.me>
Change:
- Previously, we only showed that something would have changed, not what
would have changed. This allows us to show what will chang as well.
Test Plan:
- Local RHEL8 VM
- New integration tests
Tickets:
- Fixes#66132
Signed-off-by: Rick Elrod <rick@elrod.me>
* Also validate top-level version_added.
* Fix error code.
* Produce same version_added validation error in schema than in code (and stop returning it twice).
* Return correct error codes for invalid version_added for options and return values.
* Add changelog.
* Fix forgotten closing braket.
* Accept 'historical' for some top-level version_added.
* dnf: Add nobest option
* dnf: Fix indent, add nobest specifically to dnf not yum
* Add changelog for dnf: add nobest option
* dnf: Add nobest to yumdnf module argument_spec
* dnf: remove nobest from module paramaters in yumdnf.py
* dnf: Add test for nobest option
* dnf: Cleanup packages in nobest test at last
* dnf: Cleanup manually added repos in nobest test at last
* dnf: Remove dnf-plugins-core as well in nobest test
* dnf: Change nobest release version to 2.11
* Change changelog number according to change in PR number
* Change changelog number according to change in PR number
Change:
- On Linux, there are situations where a host might be a KVM host but
not have the kernel module enabled (it might be compiled in instead).
In these cases, /dev/kvm will still exist, and rather than reporting
NA, we should report that the host is a KVM host.
Test Plan:
- Local
Signed-off-by: Rick Elrod <rick@elrod.me>
* Change default file permissions so they are not world readable
CVE-2020-1736
Set the default permissions for files we create with atomic_move() to 0o0660. Track
which files we create that did not exist and warn if the module supports 'mode'
and it was not specified and the module did not call set_mode_if_different(). This allows the user to take action and specify a mode rather than using the defaults.
A code audit is needed to find all instances of modules that call atomic_move()
but do not call set_mode_if_different(). The findings need to be documented in
a changelog since we are not warning. Warning in those instances would be frustrating
to the user since they have no way to change the module code.
- use a set for storing list of created files
- just check the argument spac and params rather than using another property
- improve the warning message to include the default permissions
* ensure hostvars are available on delegation
* also inventory_hostname must point to current host and not delegated one
* fix get_connection since it was still mixing original host vars and delegated ones
* also return connection vars for delegation and non delegation alike
* add test to ensure we have expected usage when directly assigning for non delegated host
Some platform such as ESXi does not implement EpollSelector,
which is selected by DefaultSelector. Use SelectSelector which is
based upon 'select' implementation. This works perfectly with
a platform like VMware ESXi.
Fixes: #70238
Signed-off-by: Abhijeet Kasurde <akasurde@redhat.com>
* Encode/Decode files in UTF-8
* Use helper function in ansible
* Add an integration test
* Use emoji in test data.
* add changelog
* Also support non-ascii chars in filepath and add tests about this.
* Also use non-ascii chars in replaced text and ensure not to break cron syntax.
* rename self.existing to self.n_existing
* rename crontab.existing to crontab.n_existing
Change:
- Refactoring to make it harder to get wrong and easier to read.
- Generalize become_unprivileged tests and fix some that never worked
but also never failed.
Test Plan:
- CI, new units/integration tests
Signed-off-by: Rick Elrod <rick@elrod.me>
* Fix missing quoting for remote_tmp in second mkdir of shell module. Issue #69577
* adding changelog
* fixing typo in changelog entry
* adding test case
Adding test case written by bmillemayhias.
* using $HOME instead of ~
* fixing commit measage
* Update 69578-shell-remote_tmp-quoting.yaml
Co-authored-by: Brian Kohles <me@briankohles.com>
debconf module exposes sensitive information to logs, console.
Add a note to user about using no_log=True to hide such
information from console.
Fixes: #32386
Signed-off-by: Abhijeet Kasurde <akasurde@redhat.com>
A couple of years ago Slackware -current began using a plus (“+”) at the end of the distribution version string to indicate a future version work-in-progress.
Rearrange distribution_files unit tests to easily support more tests
- add conftest with common fixtures
- use parametrize for testing multiple scenarios
* Add changelog
* Add unit tests for Slackware distribution parsing
* Use correct fixtures for Slackware
Data comes from /etc/slackware-version
Co-authored-by: Sam Doran <sdoran@redhat.com>
Co-authored-by: <Eduard Rozenberg <eduardr@pobox.com>>
A recent updated to psutil, which is a dependency of ansible-runner, fails
to install on older versions of pip.
Commit with the breaking change:
135628639b
pipe lookup plugin uses Popen with shell=True intentionally.
This is considered a security issue if user input is not validated.
Updated docs to reflect this information for the user. Also, added
Bandit B602 documentation link for further reading.
Fixes: #70159
Signed-off-by: Abhijeet Kasurde <akasurde@redhat.com>
Change:
- Add integration tests for various cases
- Fix wrong use of "its" in an exception thrown in varnames when it
throws an AnsibleError, given a term of the wrong type.
Test Plan:
- new tests, CI
Tickets:
- Fixes#70546
Signed-off-by: Rick Elrod <rick@elrod.me>
- ensure we preserve the typeerror part of the exception so loop defereed error handling
can postpone those caused by undefined variables until the when check is done.
- fix tests to comply with the 'new normal'
- human_to_bytes and others can issue TypeError not only on 'non string'
but also bad string that is not convertable.
Co-authored-by: Sloane Hertel <shertel@redhat.com>
Co-authored-by: Sloane Hertel <shertel@redhat.com>
Change:
- Use parse_kv() for parsing in the csvfile lookup plugin. This allows
us to handle multi-word search keys and filenames. Previously, the
plugin split on space and so none of these things worked as expected.
- Add integration tests for csvfile, testing a plethora of weird cases.
Test Plan:
- New integration tests, CI
Tickets:
- Fixes#70545
Signed-off-by: Rick Elrod <rick@elrod.me>
Change:
- Followup to #70504. We need to not pass the 'warn' parameter from the
action plugin either, unless it's True. Otherwise, even though it
defaults to false, we always show the deprecation.
Test Plan:
- Local
Signed-off-by: Rick Elrod <rick@elrod.me>
* speed up modify_module
* Remove debugging
* ci_complete
* Simplify generic_visit, alias to visit, eliminate some attr lookups
* ci_complete
* Add changelog fragment
* Fix ansible-test error in community.aws
* Add changelog entry for fix
* Change check from None to string_types
* Update changelogs/fragments/70507-validate-null-author.yaml
clarify wording "or a list of strings"
Co-authored-by: Felix Fontein <felix@fontein.de>
* Update test/lib/ansible_test/_data/sanity/validate-modules/validate_modules/schema.py
clarify wording - single string or not specified valid
Co-authored-by: Felix Fontein <felix@fontein.de>
* Do not fail but return None when given outside list
Co-authored-by: Felix Fontein <felix@fontein.de>
Change:
- The command warnings feature which suggests that users use modules
instead of certain commands is now deprecated. Its `warn` paramater
and `COMMAND_WARNINGS` configuration options are also deprecated.
Their use will become an error in version 2.13.
Test Plan:
- CI
Signed-off-by: Rick Elrod <rick@elrod.me>
Co-authored-by: Sam Doran <sdoran@redhat.com>
Change:
Our handling of NetBSD virtualization facts led to facts that were just
plain incorrect. One example is reporting Xen even when the system is
running on something completely different (like KVM).
As stated by the reporter of #69352, NetBSD has a better sysctl setting
to use for this information, machdep.hypervisor.
This PR does the following:
- Try to use machdep.hypervisor sysctl value if the other sysctl values
we check don't end up with enough information to be useful
- Only look for /dev/xencons and assume Xen if nothing else works
(Really this should probably return 'unknown' since the file exists on
non-Xen systems and is not very useful).
- Add a few more patterns (Xen matches and also Hyper-V) to
VirtualSysctlDetectionMixin#detect_virt_product.
This change is slightly breaking:
- If the first two attempts at using sysctl worked before,
(machdep.dmi.system-product and machdep.dmi.system-vendor), they will
continue to work.
- For cases when those values didn't work, previously the existence of
/dev/xencons was checked, and if found, we reported 'xen' (even on
non-Xen systems when the file existed). After this PR, we try the
machdep.hypervisor sysctl key before still falling back to
/dev/xencons. This means that in some cases, we might go from
(wrongly) saying "xen" to giving a more accurate value such as "kvm"
or "Hyper-V".
Test Plan:
- Tested with local NetBSD VM and got 'kvm' instead of 'xen' back.
Tickets:
- Fixes#69352
Signed-off-by: Rick Elrod <rick@elrod.me>
Brian Coca