Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Content-Type: text/plain; charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 10.3 \(3273\))
Subject: Re: [PATCH RFC v2 5/6] x86: Use global pages when PTI is disabled
From:   Nadav Amit <nadav.amit@gmail.com>
In-Reply-To: <3f69237a-ee32-3d06-dacd-a7f7897f6251@linux.intel.com>
Date:   Thu, 15 Feb 2018 12:45:11 -0800
Cc:     Andy Lutomirski <luto@kernel.org>, Ingo Molnar <mingo@redhat.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Peter Zijlstra <peterz@infradead.org>,
        Willy Tarreau <w@1wt.eu>, X86 ML <x86@kernel.org>,
        LKML <linux-kernel@vger.kernel.org>
Content-Transfer-Encoding: quoted-printable
Message-Id: <7B00EFC7-7BF2-4A17-981E-01EC69AEADA8@gmail.com>
References: <20180215163602.61162-1-namit@vmware.com>
 <20180215163602.61162-6-namit@vmware.com>
 <CALCETrX+F5apPenXGBf9oAc+aJieQYDA81zigPk6q6q6pa_i5A@mail.gmail.com>
 <3f69237a-ee32-3d06-dacd-a7f7897f6251@linux.intel.com>
To:     Dave Hansen <dave.hansen@linux.intel.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Dave Hansen <dave.hansen@linux.intel.com> wrote:

> On 02/15/2018 11:53 AM, Andy Lutomirski wrote:
>>> --- a/arch/x86/include/asm/tlbflush.h
>>> +++ b/arch/x86/include/asm/tlbflush.h
>>> @@ -319,6 +319,12 @@ static inline void set_cpu_pti_disable(unsigned =
short disable)
>>>        WARN_ON_ONCE(preemptible());
>>>=20
>>>        pti_update_user_cs64(cpu_pti_disable(), disable);
>>> +       if (__supported_pte_mask & _PAGE_GLOBAL) {
>>> +               if (disable)
>>> +                       cr4_set_bits(X86_CR4_PGE);
>>> +               else
>>> +                       cr4_clear_bits(X86_CR4_PGE);
>>> +       }
>> This will be *extremely* slow, and I don't see the point at all.  =
What
>> are you accomplishing here?
>=20
> It won't be slow if you always run compat processes, I guess.
>=20
> But mixing these in here will eat a big chunk of the benefit of having
> global pages (or PCIDs for that matter) in the first place.

These are all good points. The idea was to allow workloads like Apache, =
that
spawn multiple processes that frequently perform context-switches to =
incur
TLB misses on kernel pages.=20

The double-flushing was not intentional - I missed it. Anyhow, based on =
your
comments, and because I don=E2=80=99t see an easy way to make the global
cpu_entry_area (your recent patches) to work with this patch, I think I =
will
drop this patch.