summaryrefslogtreecommitdiff
AgeCommit message (Collapse)Author
2024-03-06Merge branch 'nvme-initramfs'Olivier Gayot
https://code.launchpad.net/~curtin-dev/curtin/+git/curtin/+merge/461452
2024-03-04Check to ensure ischroot exists before ChrootableTarget bind mountMitchell Augustin
Some systems (such as centos7) do not include /usr/bin/ischroot, which will cause the bind mount to it in ChrootableTarget to fail. Adding a check for this ensures that the file exists before mounting and aborts the mount if it does not.
2024-03-01Bind ischroot->true in ChrootableTarget and --mount-proc in unshareMitchell Augustin
Some programs that are installed by default in Ubuntu rely on checking /proc/$pid to determine if certain processes are still alive. Curtin's current default behavior for "curtin in-target" is to run children inside a separate PID namespace without also isolating /proc, meaning checks for /proc/$pid usually wait on the wrong process entirely. One such example is in any DKMS package, as DKMS checks /proc/$pid. (I have submitted a similar patch to them to address this as well: https://github.com/dell/dkms/pull/403). Adding --mount-proc to unshare --pid args resolves that issue since /proc is already mounted in util.py. However, it introduces a new issue with postinstall scripts that rely on ischroot to determine whether to restart systemd, since ischroot behavior is undefined in pid namespaces. Bind-mounting /usr/bin/ischroot to /usr/bin/true resolves this issue and is always correct since the symlink is added only within ChrootableTargets.
2024-02-29util: human2bytes actually return intDan Bungert
2024-02-29curthooks: nvmeotcp: check if network in initramfs is neededOlivier Gayot
To decide if we should bring up the network in the initramfs and then run `nvme connect-all` commands, we now look for mounpoints that have the `_netdev` option. Those are the ones that require the network to be up before mounting. If any of those mounpoints corresponds to something essential for booting (e.g., /, /usr, /var or alike), then we decide that the network is required in the initramfs. Signed-off-by: Olivier Gayot <[email protected]>
2024-02-29curthooks: nvmeotcp: bring up the network in the initramfsOlivier Gayot
If the network is required in the initramfs for NVMe over TCP, we now extract the netplan configuration, generate a set of `ip` commands (plus dhcpcd if needed) and execute those in the initramfs. This is a bit of a hack because the netplan configuration supplied by subiquity is provided using write_files directives. Also, only trivial network configurations are supported at the moment. Signed-off-by: Olivier Gayot <[email protected]>
2024-02-29curthooks: nvmeotcp: run nvme connect-all commands in the initramfsOlivier Gayot
If the network is required in the initramfs for NVMe over TCP, we now execute a set of `nvme connect-all` commands as required by the NVMe over TCP configuration. Signed-off-by: Olivier Gayot <[email protected]>
2024-02-28zpool: create keystore encrypted setupsDan Bungert
Add support for the 'luks_keystore' style encrypted ZFS install. Native ZFS encryption is supported by a small dm_crypt dataset containing the real key. This is a functional transcription of the way that Ubiquity has implemented encrypted guided ZFS.
2024-02-28curthooks: nvmeotcp: add required bins and kmods to initramfsOlivier Gayot
When network is required in the initramfs for NVMe over TCP, we need to the nvme-tcp kernel module and the nvme CLI available in the initramfs. This is now done using an initramfs-tools hook. Signed-off-by: Olivier Gayot <[email protected]>
2024-02-28curthooks: nvmeotcp: only use nvme-stas wording when relevantOlivier Gayot
When configuring NVMe over TCP, we can only rely solely on nvme-stas if none of the essential filesystems required for booting are on remote storage. In other scenarios, use the wording "NVMe over TCP". Signed-off-by: Olivier Gayot <[email protected]>
2024-02-27dm_crypt: tests related to keyfile removalDan Bungert
2024-02-27dm_crypt: clearer name for keyfile removalDan Bungert
2024-02-27dm_crypt: create and use cryptoswap as PLAINDan Bungert
Previously, cryptoswap was acting strangely, in that at install time it was considered to be a LUKS2 device and durign actual system use it would be considered a PLAIN device, in the cryptsetup(8) meanings of these terms. This caused problems when attempting to reference the device, as we would do get_volume_spec(), the returned /dev/disk/by-id/dm-uuid-... value would be wrong - it would change from dm-uuid-CRYPT-LUKS2-... to dm-uuid-CRYPT-PLAIN-... As the real runtime behavior is to be a PLAIN device for cryptoswap, stop creating it as a LUKS2 device and just use it as a PLAIN device, which doesn't involve a "create" step at all, you just open it as PLAIN directly.
2024-02-27dm_crypt: split create and open operationsDan Bungert
2024-02-28block_meta: improve get_volume_uuidDan Bungert
blkid can return 3 different types of UUIDs, and they aren't interchangable. Clarify which one we're using. Also, leverage the existing blkid() wrapper to do so.
2024-02-23raise exception on bcache validation errorsAlexsander Silva de Souza
returning the exception is probably a typo
2024-02-22pycodestyle: make use of isinstance() instead of type()Florent 'Skia' Jacquet
Should fix curtin-style-tip job.
2024-02-22Disable a flaky test to reach forever green CI.Florent 'Skia' Jacquet
2024-02-08Merge branch 'nvme-o-tcp-storageconfig'Olivier Gayot
https://code.launchpad.net/~ogayot/curtin/+git/curtin/+merge/458446
2024-02-08block: fix kname kernel refDan Bungert
2024-02-08nvme: configure nvme-stas when necessaryOlivier Gayot
When a NVMe controller is present in the storage config, we now configure nvme-stas so that the remote drives can be discovered and made available using nvme-tcp at boot. Signed-off-by: Olivier Gayot <[email protected]>
2024-02-08nvme: add NVMe controller info for NVMe disks, as an optional fieldOlivier Gayot
In the storage config, disks now support the optional "nvme_controller" property. This property holds the identifier of an existing "nvme_controller" object (which is new). a nvme_controller object has the following properties: * id: a unique identifier that can be referenced in a disk. * type: hardcoded to "nvme_controller" * transport: usually "pcie", but we will look for the value "tcp" * tcp_port: only if transport is set to "tcp" - will indicate the port to connect to as an integer. * tcp_addr: only if transport is set to "tcp" - will indicate the IP address (could be IPv4 or IPv6) to connect to. The presence of a "nvme_controller" object in the config will trigger installation of nvme-cli and nvme-stas during curthooks, even if the controller specifies transport = "pcie" or if the controller is actually unused for the installation. For now, we will recommend not having any "nvme_controller" objects in the config if we don't install to a drive connected using NVMe over TCP. Signed-off-by: Olivier Gayot <[email protected]>
2024-01-18block_meta: dm_crypt extensions for cryptswapDan Bungert
2024-01-11unittest/block_meta: make dm_crypt tests easierDan Bungert
Refactor TestDmCryptHandler a bit so that tests can inherit from DmCryptCommon and use setUpStorageConfig to provide a custom storage config.
2024-01-09vmtests: +cryptsetupDan Bungert
2024-01-03doc/apt_source: fix list markupKenyon Ralph
2023-12-13Add missing build dependencies when building curtin as debsOlivier Gayot
When building debian packages for curtin, the test suite runs and fails because of a few missing dependencies. Add what's missing so we can have a green build. Signed-off-by: Olivier Gayot <[email protected]>
2023-11-16vmtests: skip all failing tests to get the jobs green againFlorent 'Skia' Jacquet
2023-11-10tools: add usage to jenkins-runnerFlorent 'Skia' Jacquet
2023-10-26curtooks/install-grub: workaround the efivarfs statfs kernel regressionOlivier Gayot
On systems affected by the statfs kernel bug for efivarfs, we now export the LIBEFIVAR_OPS=efivarfs variable to work around the issue. The variable instructs libefivar (and therefore efibootmgr) to access EFI variables through /sys/firmware/efi/efivars without leaning on statfs. LP: #2040190
2023-10-17Add support for Oracle LinuxStamatis Katsaounis
2023-10-05install: do not log merged configDan Bungert
2023-09-26Revert "vmtests: add mantic after jammy"Dan Bungert
This reverts commit abebab1e1e30a16d4be891a051060e81254ff0c5.
2023-09-21block_meta: also add recovery key if zkey is usedOlivier Gayot
When zkey is properly setup, we do not invoke cryptsetup luksFormat ourselves. Instead we lean on zkey to invoke cryptsetup luksFormat for us. zkey seems to have no native support for invoking cryptsetup luksAddKey, so we need to manually call it if we want to add a recovery key in a second slot. Signed-off-by: Olivier Gayot <[email protected]>
2023-09-20vmtests: make loading error verboseDan Bungert
2023-09-20vmtests: +python3-debian depDan Bungert
2023-09-20vmtests: add mantic after jammyDan Bungert
2023-09-20vmtests: add jammy after focalChad Smith
(cherry picked from commit 7c7c29b008128ee5476cfa82b24762c18a0d2fcc)
2023-09-20vmtests: remove hirsuteChad Smith
(cherry picked from commit 5acd262e37cd3f2ae2b6134bf568806bd2736b0e)
2023-09-20vmtests: remove impishChad Smith
(cherry picked from commit 48e5bc483e15d56b456fe629c8a9544cf61d8d64)
2023-09-19apt: disable default deb822 migrationNick Rosbrook
Turn off migration to deb822 by default, and update appropriate docs/test to reflect this. When we are actually ready for this, the logic in want_deb822() can be restored.
2023-09-13storage_config: handle partitions on 4k diskDan Bungert
In LP: #2035034, a user reported that the layout of existing partitions was showing as 1/8th the correct size. Update storage_config to no longer hard-code a 512 byte sector and read the value instead.
2023-09-13tests/data: 4k sector diskDan Bungert
2023-09-14extract: log source informationDan Bungert
If you try hard enough and follow mount commands backwards you can surely get this same information, but when analyzing what the extract is actually doing it's nice to just be able to see what the image stack is.
2023-09-08apt_source: do not produce a deb822 source stanza with zero suiteOlivier Gayot
When no suite is set for a given repository, we end up with a deb822 styled source such as:  Type: deb  Suites:  URIs: http://archive.ubuntu.com/ubuntu This is invalid and apt/apt-get will raise the following error: E: Malformed entry 2 in sources file /etc/apt/sources.list.d/ubuntu.sources (Suite) Signed-off-by: Olivier Gayot <[email protected]>
2023-09-01docs: link to sources.list manpage for apt deb822Dan Bungert
2023-09-01swap: add loggingDan Bungert
The suggested swap size code just sort of returns 0 sometimes and figuring out why is a complicated analysis of several variables. Provide diagnostic info on those values, and the reason for the decision.
2023-08-31distro: remove even more code from apt_updateMichael Hudson-Doyle
The existing code tries to avoid redundant calls to 'apt update' but it has flaws and redundant calls to 'apt update' should be fairly quick anyway.
2023-08-30distro: remove disabling of deb-src lines in apt_updateMichael Hudson-Doyle
The current code has the side-effect of ignoring deb822 .sources files which is a problem that could be fixed but really it's hard to see why this code is still useful (we do not expect to deploy images with deb-src enabled!) so just deleting it makes more sense.
2023-08-30ensure grub-pc is installed on non-UEFI amd64 installsMichael Hudson-Doyle
I was sure curtin did this already but LP: #2033496 says not.
generated by cgit v1.2.3 (git 2.25.1) at 2025-10-22 17:24:18 +0000