linux-kernel - Re: [PATCH 1/3] x86/entry: Clear extra registers beyond syscall arguments for 64bit kernels

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:   Mon, 5 Feb 2018 20:25:16 +0000
From:   Andy Lutomirski <luto@...nel.org>
To:     Ingo Molnar <mingo@...nel.org>
Cc:     Brian Gerst <brgerst@...il.com>, Andy Lutomirski <luto@...nel.org>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Dan Williams <dan.j.williams@...el.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Andi Kleen <ak@...ux.intel.com>,
        "the arch/x86 maintainers" <x86@...nel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Ingo Molnar <mingo@...hat.com>,
        "H. Peter Anvin" <hpa@...or.com>
Subject: Re: [PATCH 1/3] x86/entry: Clear extra registers beyond syscall
 arguments for 64bit kernels

On Mon, Feb 5, 2018 at 7:48 PM, Ingo Molnar <mingo@...nel.org> wrote:
>
> * Brian Gerst <brgerst@...il.com> wrote:
>
>> On Mon, Feb 5, 2018 at 1:29 PM, Ingo Molnar <mingo@...nel.org> wrote:
>> >
>> > * Andy Lutomirski <luto@...nel.org> wrote:
>> >
>> >> [...] Clearing R10 is mostly useless in the syscall path because we'll just
>> >> unconditionally reload it in do_syscall_64().
>> >
>> > AFAICS do_syscall_64() doesn't touch R10 at all. So how does it reload R10?
>> >
>> > In fact do_syscall_64() as a C function does not touch R10, R11, R12, R13, R14,
>> > R15 - it passes their values through.
>> >
>> > What am I missing?
>>
>> The syscall ABI uses R10 for the 4th argument instead of RCX, because
>> RCX gets clobbered by the SYSCALL instruction for RIP.
>
> But we only reload the syscall-entry value of R10 it into RCX (4th C function
> argument):
>
>                 regs->ax = sys_call_table[nr](
>                         regs->di, regs->si, regs->dx,
>                         regs->r10, regs->r8, regs->r9);
>
> while RCX is a clobbered register, so in practice, while it will be briefly
> present in do_syscall_64() and the high level syscall functions, the value in RCX
> will be cleared from RCX in the overwhelming majority of cases.
>
> But the real R10 will survive much longer, because it's only used in a very small
> minority of the C functions!

Yes, indeed, brain fart on my part.