Import LXD changes from stable-5.0 #988

This was risky because if another snap (like mdl) moved to a core20 base, we'd be removing core20 just to download it again. Signed-off-by: Simon Deziel <simon.deziel@canonical.com> (cherry picked from commit 3491b6e2e11d8fde67c58878a3a23fe75944da9f)

This is to reclaim some space. Signed-off-by: Simon Deziel <simon.deziel@canonical.com> (cherry picked from commit 875480ca306b421442bb7ebbce45ff2634b2cea8)

Signed-off-by: Simon Deziel <simon.deziel@canonical.com> (cherry picked from commit b7656135800b9e3d3d27b209700f52e6d2dd2ebf)

Signed-off-by: Simon Deziel <simon.deziel@canonical.com> (cherry picked from commit 117e76b35d1f2f560299e6836797cf1fa3d395b5)

Signed-off-by: Simon Deziel <simon.deziel@canonical.com> (cherry picked from commit d5c5c7d7799c23553bc5eed1db1effd0f90ed5d9)

…ptor when probing for Direct I/O support Fixes #12808 Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com> (cherry picked from commit d0b991dca948d400b31f724a5d4ea6b274bf6763)

ovn-controller.log no longer reports the protocol on the `severity` field, instead including it with the `direction` field. Previously, the log line would include: severity=info: tcp Whereas now, it simply reports the log level: severity=info The protocol is now reported as with the direction: direction=to-lport: icmp Signed-off-by: Max Asnaashari <max.asnaashari@canonical.com> (cherry picked from commit f4d093fc0e00d7f68483eccdeea84c0534eeb223)

Fixes #12874 Signed-off-by: Wesley Hershberger <wesley.hershberger@canonical.com> (cherry picked from commit f0d0128ed636065080607170969244033b2d4616)

…874) Signed-off-by: Wesley Hershberger <wesley.hershberger@canonical.com> (cherry picked from commit 14da90f72592a7c3cf32dae85d0cfa245cc92d46)

Resolves #13038 Signed-off-by: Wesley Hershberger <wesley.hershberger@canonical.com> (cherry picked from commit f76c8bcf52ade8a18aa47c9ddf5ea9cc072d4376)

Signed-off-by: Simon Deziel <simon.deziel@canonical.com>

…hat blocksize is in bytes Signed-off-by: Simon Deziel <simon.deziel@canonical.com>

…or clarity Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

Signed-off-by: ggouzi <gaetan.gouzi@canonical.com>

Signed-off-by: Din Music <din.music@canonical.com>

…d signals Allow dnsmasq to be restarted / killed and remove the peer restriction on the signal rule for the archive profile to ensure it can be sent by a lxd which is confined under a profile with the unconfined flag. Signed-off-by: Alex Murray <alex.murray@canonical.com>

Starting with rsync 3.1.3 using a negative window will take nanoseconds into account. Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>

…resh Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>

…efreshes based on nanoseconds Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>

Signed-off-by: ggouzi <gaetan.gouzi@canonical.com>

Pyspelling has a new option to run jobs in parallel. This should make the spelling check considerably faster. Signed-off-by: Ruth Fuchss <ruth.fuchss@canonical.com>

Signed-off-by: Simon Deziel <simon.deziel@canonical.com>

… on block volumes Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>

…on block volumes Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>

Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>

…ocstring Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>

The previous error message didn't exist anymore. A good example of why not to use hard-coded error messages :) Signed-off-by: Max Asnaashari <max.asnaashari@canonical.com>

…field. Signed-off-by: Mark Laing <mark.laing@canonical.com>

Signed-off-by: Mark Laing <mark.laing@canonical.com>

…es twice This was added in https://github.com/lxc/incus/pull/6369 but is redundant since the same check is also happening afterwards when creating the actual DB record Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>

Incus generate cannot create update method for associative tables if the reference table has more than one "natural key" or if it doesn't have a field called 'Name' or 'Fingerprint'. In the former case it will fail silently (generating incorrect methods). In the latter case it will panic. This commit returns a helpful error message instead. Signed-off-by: Mark Laing <mark.laing@canonical.com>

…essages Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>

Don't panic if the given volume name doesn't contain a project Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>

Signed-off-by: Simon Deziel <simon.deziel@canonical.com>

…nalImportFromBackup Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…vice type for error in NewDeviceType Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…ce and volume DB records have been generated in CreateInstanceFromBackup This ensures any device modifications are applied to the on-disk backup.yaml file. Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

Signed-off-by: Simon Deziel <simon.deziel@canonical.com>

…nit:config disk device is removed Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…umeProject This was causing the unvalidated request project name to be used directly in storage layer sometimes. https://github.com/lxc/incus/security/code-scanning/147 Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…definition in doCustomVolumeRefresh Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…definition in doVolumeCreateOrCopy Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…VolumeRefresh And convert to effective project name. Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…t in doVolumeCreateOrCopy And convert to effective project name. Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

Signed-off-by: Mark Laing <mark.laing@canonical.com>

Additionally, capitalise error messages. Signed-off-by: Mark Laing <mark.laing@canonical.com>

…migrationSourceWs.DoStorage This fixes https://github.com/lxc/incus/security/code-scanning/147 Although wasn't an actual security problem because the DB check would fail the request if the volume specified didn't exist rather than passing unverified input to the storage layer. Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

… int64 to int in inRoutingTable Fixes https://github.com/lxc/incus/security/code-scanning/32 by using arch dependent int size in ParseInt() so its compatible with net.CIDRMask. Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…t.Sprintf by passing base to ParseInt Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

SyncResponse's `success` field can hint that SyncResponse can potentially handle errors. However, when an error is passed in like so: ``` err = ceph.ServicePlacementHandler(interfaces.CephState{State: s}, payload) if err != nil { return response.SyncResponse(false, err) } ``` But the syncResponse does not set the `error` field of the response, so the result has no metadata or error: ``` { "type": "sync", "status": "Failure", "status_code": 400, "operation": "", "error_code": 0, "error": "", "metadata": {} } ``` It seems `SyncResponse(false, ...)` is never used in the Incus codebase for any reason, so I think it's safe to just handle this type of response like an error. So I've changed `Render` to just fallback to a SmartError, which will also try to maintain the http information from the underlying error. Signed-off-by: Max Asnaashari <max.asnaashari@canonical.com>

…g in CreateVolumeFromCopy Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…ive channels together in CreateVolumeFromCopy Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…eceiver fails in CreateVolumeFromCopy As sender has by definition already ended. Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…e errors in CreateVolumeFromCopy Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…btrfs receive fails in RefreshVolume Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…ine errors in RefreshVolume Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…lume with BTRFS driver For consistency. Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

…eaders Signed-off-by: Max Asnaashari <max.asnaashari@canonical.com>

…t in BackupCustomVolume Fixes https://github.com/lxc/incus/security/code-scanning/142 Although not an actual security issue because the VolumeDBGet was already validating the input was a valid volume name. Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>

… discovery times out Signed-off-by: Julian Pelizäus <julian.pelizaeus@canonical.com>

Signed-off-by: Stéphane Graber <stgraber@stgraber.org>

Signed-off-by: Mark Laing <mark.laing@canonical.com>

…kSizeBytes Need to be able to check the block size associatd with a volume for zfs (see next commit) Signed-off-by: Wesley Hershberger <wesley.hershberger@canonical.com>

Fixes #13420 Signed-off-by: Wesley Hershberger <wesley.hershberger@canonical.com>

Signed-off-by: Wesley Hershberger <wesley.hershberger@canonical.com>

This test will be added to lxd-ci in https://github.com/lxc/incus-ci/pull/175 It doesn't fit well anywhere in this suite as it requires a VM image. Signed-off-by: Wesley Hershberger <wesley.hershberger@canonical.com>

Signed-off-by: Wesley Hershberger <wesley.hershberger@canonical.com>

…activating a volume Instead of giving a udev rules half a second to be applied, wait for the device up to 30 seconds to appear. This fixes the issue where zvol device cannot be located if the rules are not applied in time. Signed-off-by: Din Music <din.music@canonical.com>

Signed-off-by: Din Music <din.music@canonical.com>

Signed-off-by: Stéphane Graber <stgraber@stgraber.org>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Import LXD changes from stable-5.0 #988

Import LXD changes from stable-5.0 #988

Commits on Jul 12, 2024