linux-kernel - Re: [kernel-hardening] Re: [PATCH v9 1/4] syscalls: Verify address limit before returning to user-mode

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, 10 May 2017 09:08:41 +0100
From:   Al Viro <viro@...IV.linux.org.uk>
To:     Arnd Bergmann <arnd@...db.de>
Cc:     Andy Lutomirski <luto@...nel.org>,
        Christoph Hellwig <hch@...radead.org>,
        Ingo Molnar <mingo@...nel.org>, Greg KH <greg@...ah.com>,
        Thomas Garnier <thgarnie@...gle.com>,
        Martin Schwidefsky <schwidefsky@...ibm.com>,
        Heiko Carstens <heiko.carstens@...ibm.com>,
        Dave Hansen <dave.hansen@...el.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        David Howells <dhowells@...hat.com>,
        René Nyffenegger <mail@...enyffenegger.ch>,
        Andrew Morton <akpm@...ux-foundation.org>,
        "Paul E . McKenney" <paulmck@...ux.vnet.ibm.com>,
        "Eric W . Biederman" <ebiederm@...ssion.com>,
        Oleg Nesterov <oleg@...hat.com>,
        Pavel Tikhomirov <ptikhomirov@...tuozzo.com>,
        Ingo Molnar <mingo@...hat.com>,
        "H . Peter Anvin" <hpa@...or.com>,
        Paolo Bonzini <pbonzini@...hat.com>,
        Rik van Riel <riel@...hat.com>,
        Kees Cook <keescook@...omium.org>,
        Josh Poimboeuf <jpoimboe@...hat.com>,
        Borislav Petkov <bp@...en8.de>,
        Brian Gerst <brgerst@...il.com>,
        "Kirill A . Shutemov" <kirill.shutemov@...ux.intel.com>,
        Christian Borntraeger <borntraeger@...ibm.com>,
        Russell King <linux@...linux.org.uk>,
        Will Deacon <will.deacon@....com>,
        Catalin Marinas <catalin.marinas@....com>,
        Mark Rutland <mark.rutland@....com>,
        James Morse <james.morse@....com>,
        linux-s390 <linux-s390@...r.kernel.org>,
        LKML <linux-kernel@...r.kernel.org>,
        Linux API <linux-api@...r.kernel.org>,
        the arch/x86 maintainers <x86@...nel.org>,
        "linux-arm-kernel@...ts.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>,
        Kernel Hardening <kernel-hardening@...ts.openwall.com>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Peter Zijlstra <a.p.zijlstra@...llo.nl>
Subject: Re: [kernel-hardening] Re: [PATCH v9 1/4] syscalls: Verify address
 limit before returning to user-mode

On Wed, May 10, 2017 at 09:37:04AM +0200, Arnd Bergmann wrote:

> > How about trying to remove all of them?  If we could actually get rid
> > of all of them, we could drop the arch support, and we'd get faster,
> > simpler, shorter uaccess code throughout the kernel.

BTW, not all get_user() under KERNEL_DS are plain loads.  There is an
exception - probe_kernel_read().

> > The ones in kernel/compat.c are generally garbage.  They should be
> > using compat_alloc_user_space().  Ditto for kernel/power/user.c.
> 
> compat_alloc_user_space() has some problems too, it adds
> complexity to a rarely-tested code path and can add some noticeable
> overhead in cases where user space access is slow because of
> extra checks.
> 
> It's clearly better than set_fs(), but the way I prefer to convert the
> code is to avoid both and instead move compat handlers next to
> the native code, and splitting out the common code between native
> and compat mode into a helper that takes a regular kernel pointer.
> 
> I think that's what both Al has done in the past on compat_ioctl()
> and select() and what Christoph does in his latest series, but
> it seems worth pointing out for others that decide to help out here.

Folks, reducing the amount of places where we play with set_fs() is certainly
a good thing.  Getting rid of them completely is something entirely different;
I have tried to plot out patch series in this direction many times during the
last 5 years or so, but it's not going to be easy.  Tomorrow I can start
posting my notes in that direction (and there are tons of those, unfortunately
mixed with git grep results, highly unprintable personal comments, etc.);
just let me grab some sleep first...

BTW, slow userland access is not just due to extra checks; access_ok(),
in particular, is pretty much noise.  The real PITA comes from the things
like STAC/CLAC on recent x86.  Or hardware overhead of cross-address-space
block copy insn (e.g. on s390, where it's optimized for multi-cacheline
blocks).  Or things like uml, where it's a matter of walking the page
tables for each sodding __get_user().  It's not always just a matter of
address space limit...