linux-kernel - Re: [PATCH] Add new open(2) flag

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 19 Apr 2023 06:15:29 +0500
From:   Ameer Hamza <ahamza@...ystems.com>
To:     Christian Brauner <brauner@...nel.org>
Cc:     viro@...iv.linux.org.uk, jlayton@...nel.org,
        chuck.lever@...cle.com, arnd@...db.de, guoren@...nel.org,
        palmer@...osinc.com, f.fainelli@...il.com, slark_xiao@....com,
        linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-arch@...r.kernel.org, awalker@...ystems.com
Subject: Re: [PATCH] Add new open(2) flag - O_EMPTY_PATH

On Fri, Jan 06, 2023 at 02:06:51PM +0100, Christian Brauner wrote:
> On Wed, Dec 28, 2022 at 09:02:49PM +0500, Ameer Hamza wrote:
> > This patch adds a new flag O_EMPTY_PATH that allows openat and open
> > system calls to open a file referenced by fd if the path is empty,
> > and it is very similar to the FreeBSD O_EMPTY_PATH flag. This can be
> > beneficial in some cases since it would avoid having to grant /proc
> > access to things like samba containers for reopening files to change
> > flags in a race-free way.
> > 
> > Signed-off-by: Ameer Hamza <ahamza@...ystems.com>
> > ---
> 
> In general this isn't a bad idea and Aleksa and I proposed this as part
> of the openat2() patchset (see [1]).
> 
> However, the reason we didn't do this right away was that we concluded
> that it shouldn't be simply adding a flag. Reopening file descriptors
> through procfs is indeed very useful and is often required. But it's
> also been an endless source of subtle bugs and security holes as it
> allows reopening file descriptors with more permissions than the
> original file descriptor had.
> 
> The same lax behavior should not be encoded into O_EMPTYPATH. Ideally we
> would teach O_EMPTYPATH to adhere to magic link modes by default. This
> would be tied to the idea of upgrade mask in openat2() (cf. [2]). They
> allow a caller to specify the permissions that a file descriptor may be
> reopened with at the time the fd is opened.
> 
> [1]: https://lore.kernel.org/lkml/20190930183316.10190-4-cyphar@cyphar.com/
> [2]: https://lore.kernel.org/all/20220526130355.fo6gzbst455fxywy@senku/Kk

Thank you for the detailed explanation and sorry for getting back late
at it. It seems like a pre-requisite for O_EMPTYPATH is to make it safe
and that depends on a patchset that Aleksa was working on. It would be
helpful to know the current status of that effort and if we could expect
it in the near future.

The repo[1] that was mentioned here[2] seems to be private. I am wondering
if there's a way to look at the patch somehow.

[1]: https://github.com/cyphar/linux/tree/magiclink/main
[2]: https://lore.kernel.org/all/20220526130952.z5efngrnh7xtli32@senku/