Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Feedback-ID: ieff94742:Fastmail
User-Agent: Cyrus-JMAP/3.7.0-alpha0-811-gb808317eab-fm-20220801.001-gb808317e
Mime-Version: 1.0
Message-Id: <c2f42cfd-82d4-4ab7-92b9-a173d6cdd29d@www.fastmail.com>
In-Reply-To: <01768c70-a012-0d08-cfa5-a7a87ff0a19c@intel.com>
References: <20220805173009.3128098-1-ira.weiny@intel.com>
 <20220805173009.3128098-6-ira.weiny@intel.com>
 <5d62c1d0-7425-d5bb-ecb5-1dc3b4d7d245@intel.com> <Yu4t0jkpIY4grgxL@gmail.com>
 <Yu4wON0MRGH7h5Jv@zn.tnic> <Yu+Nwbn4CZUmyD14@gmail.com>
 <Yu+VV8xFTYRh0D0a@zn.tnic> <YvAaXet3sBg3mRDe@iweiny-desk3>
 <YvDtfKRyMGenRMU5@gmail.com> <01768c70-a012-0d08-cfa5-a7a87ff0a19c@intel.com>
Date:   Tue, 09 Aug 2022 14:19:19 -0700
From:   "Andy Lutomirski" <luto@kernel.org>
To:     "Dave Hansen" <dave.hansen@intel.com>,
        "Ingo Molnar" <mingo@kernel.org>, "Ira Weiny" <ira.weiny@intel.com>
Cc:     "Borislav Petkov" <bp@alien8.de>,
        "Rik van Riel" <riel@surriel.com>,
        "the arch/x86 maintainers" <x86@kernel.org>,
        "Linux Kernel Mailing List" <linux-kernel@vger.kernel.org>,
        kernel-team@fb.com
Subject: Re: [RFC PATCH 5/5] x86/entry: Store CPU info on exception entry
Content-Type: text/plain;charset=utf-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk


On Mon, Aug 8, 2022, at 9:16 AM, Dave Hansen wrote:
> On 8/8/22 04:03, Ingo Molnar wrote:
>>> Again, I don't believe this is too much overhead but I don't want pe=
ople=20
>>> to say it was not discussed.
>> Is it necessary to do this, what are the alternatives, can this overh=
ead be=20
>> avoided?
>
> One thing Andy mentioned is that we _could_ get it down to two instruc=
tions:
>
> 	rdgsbase $reg
> 	push $reg
>
> This could be hidden in:
>
> 	PUSH_PTREGS_AUXILIARY
>
> where, today, it would only net add a single instruction.  But, if we
> ever add more stuff to PUSH_PTREGS_AUXILIARY, it would move back to
> needing two instructions since we'd need both the:
>
> 	subq $PTREGS_AUX_SIZE, %rsp
>
> and something to write gsbase to the stack.
>
> That doesn't get us the smp_processor_id() directly, but we can derive
> it later on from the gsbase value.
>
> The downside is that we're doing it in assembly.  We'd also have
> something additional which is a bit uglier and that reads memory on
> !X86_FEATURE_FSGSBASE, probably:
>=09
> 	movq    PER_CPU_VAR(cpu_number), %reg
> 	push %reg

Nah, I believe the same value that RDGSBASE reads is already in percpu m=
emory as 'per_cpu_offset', so the alternative can just read that and the=
 code that uses it doesn=E2=80=99t need to care about the alternative.

>
> Which would require some different code to decode what was there:
>
> int read_exception_cpu_number(ext_pt_regs *e)
> {
> 	if (cpu_feature_enabled(X86_FEATURE_FSGSBASE))
> 		return gsbase_to_cpu_number(e->ext_cpu_nr);
> 	else
> 		return e->ext_cpu_nr;
> }
>
> I'm thinking that the whole racy smp_processor_id() thing wasn't so bad
> in the first place.