Re: [PATCH] expand-on-first-boot: Add support for devicemapper

[Tobias Schaffner <tobias.schaffner@siemens.com>]

Switching to systemd-repart would also fix a rare race condition. 
Expand-on-first-boot can cause systemd to not be able to mount its mount 
units if it makes the kernel reload the partition table in a bad moment.

Tobias

On 4/21/22 21:05, Henning Schild wrote:
> Am Thu, 21 Apr 2022 19:43:47 +0200
> schrieb Henning Schild <henning.schild@siemens.com>:
> 
>> I learned yesterday about systemd-repart
>> https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.freedesktop.org%2Fsoftware%2Fsystemd%2Fman%2Fsystemd-repart.service.html&amp;data=05%7C01%7Ctobias.schaffner%40siemens.com%7C045ffc25cea442b91a7808da23c9e432%7C38ae3bcd95794fd4addab42e1495d55a%7C1%7C0%7C637861648059484524%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&amp;sdata=qlzBS8YjhJBvZE7LnofOlT6uE76Nrsu45xZ25E9ocj4%3D&amp;reserved=0
> 
> This currently needs >= bookworm to even start playing with systemd >=
> 250.1-2
> 
> https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fbugs.debian.org%2Fcgi-bin%2Fbugreport.cgi%3Fbug%3D976959&amp;data=05%7C01%7Ctobias.schaffner%40siemens.com%7C045ffc25cea442b91a7808da23c9e432%7C38ae3bcd95794fd4addab42e1495d55a%7C1%7C0%7C637861648059484524%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&amp;sdata=Zlq1hF7%2BbE5FS3gayX6jhaVzykvJNrwc1lhITORUZ%2BU%3D&amp;reserved=0
> 
> So maybe a little too early for isar, but still something we could look
> into.
> 
> systemd-growfs is also available in older versions, but that just deals
> with the filesystem, not with the partition table.
> 
> Henning
> 
>> We should look at this and how mkosi puts/uses it for debian.
>>
>> Not a review, just a generic remark that there might be more powerful
>> and ready to use tools out there. At some point i will play with this
>> and see if it can replace expand-on-first-boot, i bet we will soon
>> hear the first btrfs stories ... and maybe luks.
>>
>> I know people do that with isar and their own funny scripting which
>> can not be found upstream.
>>
>> Am Thu, 21 Apr 2022 19:10:44 +0200
>> schrieb Jan Kiszka <jan.kiszka@siemens.com>:
>>
>>> From: Jan Kiszka <jan.kiszka@siemens.com>
>>>
>>> When the rootfs is inside a devicemapper volume, e.g. dm-verity, the
>>> script so far fails to identify the boot device. This adds at least
>>> basic support for that case (single level, single slave). Once there
>>> is a good example for more complex setups, this could be taken into
>>> account as well.
>>>
>>> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
>>> ---
>>>   .../expand-on-first-boot/files/expand-last-partition.sh    | 7
>>> ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
>>>
>>> diff --git
>>> a/meta/recipes-support/expand-on-first-boot/files/expand-last-partition.sh
>>> b/meta/recipes-support/expand-on-first-boot/files/expand-last-partition.sh
>>> index 7175dfd3..bb371e9f 100755 ---
>>> a/meta/recipes-support/expand-on-first-boot/files/expand-last-partition.sh
>>> +++
>>> b/meta/recipes-support/expand-on-first-boot/files/expand-last-partition.sh
>>> @@ -10,8 +10,13 @@ set -e ROOT_DEV="$(findmnt / -o source -n)"
>>> -BOOT_DEV="$(echo "${ROOT_DEV}" | sed 's/p\?[0-9]*$//')"
>>> +ROOT_DEV_NAME=${ROOT_DEV##*/}
>>> +ROOT_DEV_SLAVE=$(ls -d /sys/block/${ROOT_DEV_NAME}/slaves/*
>>> 2>/dev/null | head -1) +if [ -n ${ROOT_DEV_SLAVE} ]; then
>>> +	ROOT_DEV=/dev/${ROOT_DEV_SLAVE##*/}
>>> +fi
>>>   
>>> +BOOT_DEV="$(echo "${ROOT_DEV}" | sed 's/p\?[0-9]*$//')"
>>
>> Please add a wks and switch at least one CI target. In case wic can
>> not generate that dm-verity ... NACK, or provide the wic plugin as
>> well.
>>
>> Henning
>>
>>>   if [ "${ROOT_DEV}" = "${BOOT_DEV}" ]; then
>>>   	echo "Boot device equals root device - no partitioning
>>> found" >&2 exit 1
>>
>