Partition sizing

Partition sizing

[Demi Marie Obenour]

[-- Attachment #1.1.1: Type: text/plain, Size: 968 bytes --]

What size should the A and B partitions be?  I can’t think of
any reasonable way to size them that is also future-proof.

Android had the same problem, and they solved it by using
device mapper.  Unfortunately, I doubt LVM meets our security
goals, which means a whole new userspace implementation would
be required.  That would not be fun, especially since a bug
would require a reinstall to correct.

Other image-based Linux distros generally don't have this
problem because they use filesystem or ostree snapshots,
rather than separate partitions.  Chromium OS has a whole
team of paid developers, so I think they can deal with
constraints a bit better than we can :).

One horrifying worst-case option is to add a file on the
user data partition, create a loop device based on it,
and then add it to the dm-verity table.  I really don't
want to do that if there is any other option, though.
-- 
Sincerely,
Demi Marie Obenour (she/her/hers)

[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 7253 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: Partition sizing

[Alyssa Ross]

[-- Attachment #1: Type: text/plain, Size: 1562 bytes --]

Demi Marie Obenour <demiobenour@gmail.com> writes:

> What size should the A and B partitions be?  I can’t think of
> any reasonable way to size them that is also future-proof.

I think we need to just do our best estimate, erring on the side of them
being larger, and I don't think we'll even be in a position to estimate
that until we've done work on minimizing dependencies etc., but
obviously we will need to come up with an answer to this before we start
encouraging people to install an image.

If we pick too big, will we be able to shrink them later for new
installs?  Will it be possible for space-constrained users to manually
partition with smaller sizes?

> Android had the same problem, and they solved it by using
> device mapper.  Unfortunately, I doubt LVM meets our security
> goals, which means a whole new userspace implementation would
> be required.  That would not be fun, especially since a bug
> would require a reinstall to correct.
>
> Other image-based Linux distros generally don't have this
> problem because they use filesystem or ostree snapshots,
> rather than separate partitions.  Chromium OS has a whole
> team of paid developers, so I think they can deal with
> constraints a bit better than we can :).
>
> One horrifying worst-case option is to add a file on the
> user data partition, create a loop device based on it,
> and then add it to the dm-verity table.  I really don't
> want to do that if there is any other option, though.
> -- 
> Sincerely,
> Demi Marie Obenour (she/her/hers)

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 227 bytes --]

Re: Partition sizing

[Demi Marie Obenour]

[-- Attachment #1.1.1: Type: text/plain, Size: 1268 bytes --]

On 9/17/25 07:23, Alyssa Ross wrote:
> Demi Marie Obenour <demiobenour@gmail.com> writes:
> 
>> What size should the A and B partitions be?  I can’t think of
>> any reasonable way to size them that is also future-proof.
> 
> I think we need to just do our best estimate, erring on the side of them
> being larger, and I don't think we'll even be in a position to estimate
> that until we've done work on minimizing dependencies etc., but
> obviously we will need to come up with an answer to this before we start
> encouraging people to install an image.
> 
> If we pick too big, will we be able to shrink them later for new
> installs?

Yes, easily.

> Will it be possible for space-constrained users to manually
> partition with smaller sizes?
For new installs, yes if the installer UI allows it.  For
existing installs, maybe, but it would be fragile at best without
some sort of device-mapper trick like Android uses.  It would be
more robust to reinstall.

In the future, it would be much better to adopt Android's update
system, which solves all of these problems and provides features
like update compression.  AOSP isn't open contribution, though,
so this is likely to be a fork.
-- 
Sincerely,
Demi Marie Obenour (she/her/hers)

[-- Attachment #1.1.2: OpenPGP public key --]
[-- Type: application/pgp-keys, Size: 7253 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]