linux-kernel - Re: [GIT PULL] SafeSetID LSM changes for 5.4

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]

Message-ID: <CAHk-=wh_CHD9fQOyF6D2q3hVdAhFOmR8vNzcq5ZPcxKW3Nc+2Q@mail.gmail.com>
Date:   Mon, 23 Sep 2019 12:28:19 -0700
From:   Linus Torvalds <torvalds@...ux-foundation.org>
To:     Micah Morton <mortonm@...omium.org>, Jann Horn <jannh@...gle.com>,
        Bart Van Assche <bart.vanassche@....com>,
        "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
Cc:     Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        linux-security-module <linux-security-module@...r.kernel.org>
Subject: Re: [GIT PULL] SafeSetID LSM changes for 5.4

On Mon, Sep 23, 2019 at 12:01 PM Linus Torvalds
<torvalds@...ux-foundation.org> wrote:
>
> Anyway, this bug would likely had been avoided if rcu_swap_protected()
> just returned the old pointer instead of changing the argument.

Also, I have to say that the fact that I got the fundamentally buggy
commit  in a pull request during the 5.3 merge window, and merged it
on July 16, but then get the pull request for the fix two months
later, after 5.3 has been released, makes me very unhappy with the
state of safesetid.

The pull request itself was clearly never tested. That's a big problem.

And *nobody* used it at all or tested it at all during the whole
release process. That's another big problem.

Should we just remove safesetid again? It's not really maintained, and
it's apparently not used.  It was merged in March (with the first
commit in January), and here we are at end of September and this
happens.

So yes, syntactically I'll blame the bad RCU interfaces for why the
bug happened.

But the fact that the code didn't _work_ and was never tested by
anybody for two months, that's not the fault of the RCU code.

                  Linus