linux-kernel - Re: [PATCH v1 1/8] x86/entry/clearregs: Remove partial stack frame in fast system call

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:   Thu, 11 Jan 2018 10:44:40 -0800
From:   Andi Kleen <andi@...stfloor.org>
To:     Andi Kleen <andi@...stfloor.org>
Cc:     Andy Lutomirski <luto@...capital.net>,
        Brian Gerst <brgerst@...il.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        the arch/x86 maintainers <x86@...nel.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        David Woodhouse <dwmw@...zon.co.uk>,
        Paul Turner <pjt@...gle.com>,
        Andy Lutomirski <luto@...nel.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Tom Lendacky <thomas.lendacky@....com>,
        Tim Chen <tim.c.chen@...ux.intel.com>,
        Greg Kroah-Hartman <gregkh@...ux-foundation.org>,
        Dave Hansen <dave.hansen@...el.com>,
        Jiri Kosina <jikos@...nel.org>, Andi Kleen <ak@...ux.intel.com>
Subject: Re: [PATCH v1 1/8] x86/entry/clearregs: Remove partial stack frame
 in fast system call

> Well it's kernel hardening to guard against possible future speculation
> attacks. Linus discussed it here for example:
> 
> https://www.mail-archive.com/linux-kernel@vger.kernel.org/msg1580667.html
> 
> For the on/off switch I can add a CONFIG to enable it, even though
> it seems somewhat silly.

I did some micro benchmarking now, sampling different real system
calls.

For the entry code (entry to call) I get on average 62 cycles 
for the old code, vs 78 cycles with clear regs and full
stack frame saving on Skylake.

So it's roughly ~20 cycles difference, if we include the restore.

I would conclude 20 cycles are not significant for a syscall,
so there's not a lot of motivation to add a switch
for less security.

-Andi