[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1516011195.22147.132.camel@infradead.org>
Date: Mon, 15 Jan 2018 11:13:16 +0100
From: David Woodhouse <dwmw2@...radead.org>
To: Andi Kleen <ak@...ux.intel.com>, pjt@...gle.com,
linux-kernel@...r.kernel.org, tim.c.chen@...ux.intel.com,
riel@...hat.com, jikos@...nel.org, jpoimboe@...hat.com,
luto@...capital.net, torvalds@...ux-foundation.org,
tglx@...utronix.de, hpa@...or.com, peterz@...radead.org,
dave.hansen@...el.com, mingo@...nel.org, keescook@...gle.com,
gregkh@...ux-foundation.org, arjan@...ux.intel.com
Cc: linux-tip-commits@...r.kernel.org
Subject: Re: [tip:x86/pti] x86/retpoline: Fill RSB on context switch for
affected CPUs
On Sun, 2018-01-14 at 16:05 -0800, Andi Kleen wrote:
> > + if ((!boot_cpu_has(X86_FEATURE_PTI) &&
> > + !boot_cpu_has(X86_FEATURE_SMEP)) || is_skylake_era()) {
> > + setup_force_cpu_cap(X86_FEATURE_RSB_CTXSW);
> > + pr_info("Filling RSB on context switch\n");
> > + }
>
> Missing an option to turn this off.
Deliberately so. You can already boot with 'spectre_v2=off' to turn off
the mitigations. We are not intending to permit all the bullshit micro-
management of IBRS=3/IBPB=2/RSB=π nonsense.
If you choose retpoline, you get the RSB stuffing which is appropriate
along with that. With IBRS, you get the RSB stuffing which is
appropriate with that. You don't get command line or sysfs tunables to
mess it. You *do* have the source code, if you really want to make
changes. Don't.
Download attachment "smime.p7s" of type "application/x-pkcs7-signature" (5213 bytes)
Powered by blists - more mailing lists