On 2026-11-05 10:28 +03:00, Pascal Hambourg wrote:
> On 04/11/2026 at 19:26, Alper Nebi Yasak wrote:
>> On 2026-11-04 17:02 +03:00, Pascal Hambourg wrote:
>>>
>>>> [2] https://salsa.debian.org/installer-team/partman-cros/-/merge_requests/3
>>>
>>> All other automatic partitioning recipes are gathered in partman-auto
>>> and recipes for current architectures are built by a script for easier
>>> maintenance and better consistency. Introducing recipes in another
>>> package would make maintenance more difficult.
>>
>> I agree. I think I had a libdebian-installer MR that would add amd64/cros
>> and arm64/cros to archdetect, which partman-auto would want?
> 
> I guess yes, it would allow to ship recipes for this subarch. But it may 
> affect other d-i packages which use archdetect. What is the current 
> reported subarchitecture on ChromeOS platforms ?

They are not EFI, so it defaults to "amd64/generic" or "arm64/generic" IIRC.

> The MR adds recipes for armhf, are there any ChromeOS machines with this 
> architecture ?

Yes, but we can't boot d-i through this boot method on those due to size
limits (8/16/32 MiB depending on board) on the bootable image size.

We can sometimes chainload e.g. U-Boot from these partitions but I don't 
know how to detect the board as running ChromeOS bootloader in those cases, 
so not a concern for now.

>>> Regarding the recipes themselves,
>>> - Do any ChromeOS machines require unpartitioned reserved space for the
>>> boot loader ?
>>
>> Nope.
> 
> Then it can be removed. This reserved partition was added mostly for 
> arm64 platforms using Rockchip SoCs.
> 
>> Also, the bootloader can only boot from GPT, and we don't need a /boot
>> partition even if rootfs is encrypted (but it doesn't hurt).
> 
> So the encrypted /boot check would need to be adjusted in partman-crypto.
> 
>>>> [3] https://salsa.debian.org/installer-team/partman-cros/-/merge_requests/2
>>>
>>> Never clearing the flag seems a bit extreme. What about using the same
>>> logic as in partman-efi to preserve unused existing ESPs instead ?
>>
>> I'm not sure the analysis is correct there, I have to test things again.
>> I remember installing to a USB disk without messing up ChromeOS on the
>> eMMC. Maybe they chose "Do not use" on existing partitions expecting it
>> to be like the ESP case? IIRC I thought it makes no sense to mark a CrOS
>> partition as "Do not use" except to indicate it should stop being one.
> 
> OK, but partman-auto marks all partitions "do not use" before starting 
> automatic partitioning.

Ah. In that case switching to the ESP logic makes sense.

>> One difference with ESPs is that we don't choose which partition will be
>> used here. depthcharge-tools(-installer) will look at the disks that
>> hold /boot and rootfs (as they are the only disks we can assume will be
>> present during boot) and rotate between them itself.
> I see. But I think we still need to be able to clear the flag if the 
> partition is re-used for anything else, or did not originally have the 
> flag set. Otherwise IIUC the partition may be overwritten by 
> depthcharge-tools.

Yes and yes.

(And I should make depthcharge-tools check if the targeted partition is 
mounted or has a filesystem...)

Reply via email to