2016-11-29 21:30:00

by Richard Weinberger

[permalink] [raw]
Subject: Idea behind EXT4_IOC_GET_ENCRYPTION_PWSALT?

Hi!

As the subject states, I'm a bit confused wrt. EXT4_IOC_GET_ENCRYPTION_PWSALT.
Will common fscrypt userspace depend on it?

IIUC you want to store some salt to seed the user password.
But what if /home/rw and /home/dags have different keys? Is it okay
to use the same salt for both keys?
Or is it an ad-hoc solution to allow an encrypted filesystem root
because you need a way to store the seed on the same filesystem but not as
file since all files are encrypted?
Wouldn't a xattr stored in the filesystem root directory also do it?

Long story short, I'm not sure whether we really need this ioctl command
in UBIFS and what the designated semantics are. :-)

Thanks,
//richard


2016-11-30 00:44:53

by Joe Richey

[permalink] [raw]
Subject: Re: Idea behind EXT4_IOC_GET_ENCRYPTION_PWSALT?

Richard,

Our current design for fscrypt (tentative name for the userspace
filesystem encryption manager) does not use the global filesystem salt
(EXT4_IOC_GET_ENCRYPTION_PWSALT), we are planning on having a
different salt for each password used in the system. We are using
planning on using Argon2id as the password stretching algorithm, so
we'll have costs for memory, time, and parallelism stored for each
password as well as a salt.

I can't speak to other uses of the ioctl.

Joe


On Tue, Nov 29, 2016 at 1:30 PM, Richard Weinberger <[email protected]> wrote:
> Hi!
>
> As the subject states, I'm a bit confused wrt. EXT4_IOC_GET_ENCRYPTION_PWSALT.
> Will common fscrypt userspace depend on it?
>
> IIUC you want to store some salt to seed the user password.
> But what if /home/rw and /home/dags have different keys? Is it okay
> to use the same salt for both keys?
> Or is it an ad-hoc solution to allow an encrypted filesystem root
> because you need a way to store the seed on the same filesystem but not as
> file since all files are encrypted?
> Wouldn't a xattr stored in the filesystem root directory also do it?
>
> Long story short, I'm not sure whether we really need this ioctl command
> in UBIFS and what the designated semantics are. :-)
>
> Thanks,
> //richard
> --
> To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html


Attachments:
smime.p7s (4.74 kB)
S/MIME Cryptographic Signature

2016-11-30 08:29:40

by Richard Weinberger

[permalink] [raw]
Subject: Re: Idea behind EXT4_IOC_GET_ENCRYPTION_PWSALT?

Joe,

On 30.11.2016 01:44, Joe Richey wrote:
> Richard,
>
> Our current design for fscrypt (tentative name for the userspace
> filesystem encryption manager) does not use the global filesystem salt
> (EXT4_IOC_GET_ENCRYPTION_PWSALT), we are planning on having a
> different salt for each password used in the system. We are using
> planning on using Argon2id as the password stretching algorithm, so
> we'll have costs for memory, time, and parallelism stored for each
> password as well as a salt.


Makes sense.
So, for now I'll not implement this ioctl command in UBIFS.

Thanks,
//richard