All of lore.kernel.org
 help / color / mirror / Atom feed
From: Dave Hansen <dave.hansen@intel.com>
To: Tim Chen <tim.c.chen@linux.intel.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Andy Lutomirski <luto@kernel.org>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Greg KH <gregkh@linuxfoundation.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>,
	Andi Kleen <ak@linux.intel.com>,
	Arjan Van De Ven <arjan.van.de.ven@intel.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 3/7] x86/enter: Use IBRS on syscall and interrupts
Date: Thu, 4 Jan 2018 12:45:14 -0800	[thread overview]
Message-ID: <e2b435e0-7ecc-1ed4-4937-230fed226866@intel.com> (raw)
In-Reply-To: <0c525c4c6c817e9c42c7ed583d86dc591a86efde.1515086770.git.tim.c.chen@linux.intel.com>

On 01/04/2018 09:56 AM, Tim Chen wrote:
> If NMI runs when exiting kernel between IBRS_DISABLE and
> SWAPGS, the NMI would have turned on IBRS bit 0 and then it would have
> left enabled when exiting the NMI. IBRS bit 0 would then be left
> enabled in userland until the next enter kernel.
> 
> That is a minor inefficiency only, but we can eliminate it by saving
> the MSR when entering the NMI in save_paranoid and restoring it when
> exiting the NMI.

Can I suggest and alternate description for the NMI case?  This is
long-winded, but it should keep me from having to think through it yet
again. :)

"
The normal interrupt code uses the 'error_entry' path which uses the
Code Segment (CS) of the instruction that was interrupted to tell
whether it interrupted the kernel or userspace and thus has to switch
IBRS, or leave it alone.

The NMI code is different.  It uses 'paranoid_entry' because it can
interrupt the kernel while it is running with a userspace IBRS (and %GS
and CR3) value, but has a kernel CS.  If we used the same approach as
the normal interrupt code, we might do the following;

	SYSENTER_entry
<-------------- NMI HERE
	IBRS=1
		do_something()
	IBRS=0
	SYSRET

The NMI code might notice that we are running in the kernel and decide
that it is OK to skip the IBRS=1.  This would leave it running
unprotected with IBRS=0, which is bad.

However, if we unconditionally set IBRS=1, in the NMI, we might get the
following case:

	SYSENTER_entry
	IBRS=1
		do_something()
	IBRS=0
<-------------- NMI HERE (set IBRS=1)
	SYSRET

and we would return to userspace with IBRS=1.  Userspace would run
slowly until we entered and exited the kernel again.  (This is the case
Tim is alluding to in the patch description).

Instead of those two approaches, we chose a third one where we simply
save the IBRS value in a scratch register (%r13) and then restore that
value, verbatim.  This is what PTI does with CR3 and it works beautifully.
"

  parent reply	other threads:[~2018-01-04 20:45 UTC|newest]

Thread overview: 116+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-01-04 17:56 [PATCH 0/7] IBRS patch series Tim Chen
2018-01-04 17:56 ` [PATCH 1/7] x86/feature: Detect the x86 feature to control Speculation Tim Chen
2018-01-04 19:58   ` Greg KH
2018-01-04 20:47     ` Tim Chen
2018-01-05 11:14   ` David Woodhouse
2018-01-05 15:14     ` Tom Lendacky
2018-01-05 17:07       ` Tim Chen
2018-01-05 13:09   ` Thomas Gleixner
2018-01-05 13:44     ` Andrea Arcangeli
2018-01-05 13:51       ` Thomas Gleixner
2018-01-04 17:56 ` [PATCH 2/7] x86/enter: MACROS to set/clear IBRS Tim Chen
2018-01-04 22:16   ` Peter Zijlstra
2018-01-04 22:21     ` Tim Chen
2018-01-04 22:23       ` Dave Hansen
2018-01-05  4:54         ` Andy Lutomirski
2018-01-05  5:05           ` Dave Hansen
2018-01-05 13:19       ` Thomas Gleixner
2018-01-04 17:56 ` [PATCH 3/7] x86/enter: Use IBRS on syscall and interrupts Tim Chen
2018-01-04 20:00   ` Greg KH
2018-01-04 20:26     ` Tim Chen
2018-01-04 20:45   ` Dave Hansen [this message]
2018-01-04 22:33   ` Peter Zijlstra
2018-01-04 23:12     ` Andrea Arcangeli
2018-01-05  0:08     ` Dave Hansen
2018-01-05  4:51       ` Andy Lutomirski
2018-01-05  5:11         ` Dave Hansen
2018-01-05 12:01           ` Alan Cox
2018-01-05 13:35   ` Thomas Gleixner
2018-01-04 17:56 ` [PATCH 4/7] x86/idle: Disable IBRS entering idle and enable it on wakeup Tim Chen
2018-01-04 22:47   ` Peter Zijlstra
2018-01-04 23:00     ` Andrea Arcangeli
2018-01-04 23:22     ` Tim Chen
2018-01-04 23:42       ` Andrea Arcangeli
2018-01-04 23:45         ` Thomas Gleixner
2018-01-05  0:03           ` Andrea Arcangeli
2018-01-08  8:24       ` Peter Zijlstra
2018-01-04 17:56 ` [PATCH 5/7] x86: Use IBRS for firmware update path Tim Chen
2018-01-04 18:48   ` Alan Cox
2018-01-04 20:05   ` Greg KH
2018-01-04 20:08     ` Woodhouse, David
2018-01-05 16:08       ` gregkh
2018-01-05 16:37         ` Andrea Arcangeli
2018-01-04 20:21     ` Andrew Cooper
2018-01-04 20:48     ` Andrea Arcangeli
2018-01-04 20:51   ` Yves-Alexis Perez
2018-01-04 21:13     ` Tim Chen
2018-01-04 22:51   ` Peter Zijlstra
2018-01-05 13:40   ` Thomas Gleixner
2018-01-04 17:56 ` [PATCH 6/7] x86/spec_ctrl: Add sysctl knobs to enable/disable SPEC_CTRL feature Tim Chen
2018-01-04 18:33   ` Borislav Petkov
2018-01-04 18:36     ` Dave Hansen
2018-01-04 18:52       ` Borislav Petkov
2018-01-04 18:57         ` Andrea Arcangeli
2018-01-04 18:59         ` Dave Hansen
2018-01-04 19:06           ` Borislav Petkov
2018-01-05 13:48       ` Thomas Gleixner
2018-01-04 18:38     ` Andrea Arcangeli
2018-01-04 18:54       ` Dave Hansen
2018-01-04 18:56         ` Borislav Petkov
2018-01-04 18:55       ` Borislav Petkov
2018-01-04 18:34   ` Andrea Arcangeli
2018-01-04 19:02     ` Tim Chen
2018-01-04 18:50   ` Alan Cox
2018-01-04 20:16   ` Greg KH
2018-01-04 20:58     ` Tim Chen
2018-01-04 22:54   ` Peter Zijlstra
2018-01-04 23:26     ` Tim Chen
2018-01-04 23:51       ` Andrea Arcangeli
2018-01-04 23:59         ` Borislav Petkov
2018-01-05  0:07           ` Thomas Gleixner
2018-01-05 11:16   ` David Woodhouse
2018-01-06  1:29     ` Tim Chen
2018-01-04 17:56 ` [PATCH 7/7] x86/microcode: Recheck IBRS features on microcode reload Tim Chen
2018-01-04 18:28   ` Borislav Petkov
2018-01-04 18:34     ` Andrea Arcangeli
2018-01-04 18:50       ` Borislav Petkov
2018-01-04 19:10         ` Tim Chen
2018-01-05 13:32         ` Greg KH
2018-01-05 13:37           ` Borislav Petkov
2018-01-05 13:47             ` Greg KH
2018-01-05 15:28           ` Andrea Arcangeli
2018-01-04 19:00 ` [PATCH 0/7] IBRS patch series Linus Torvalds
2018-01-04 19:19   ` David Woodhouse
2018-01-04 19:33     ` Linus Torvalds
2018-01-04 19:39       ` David Woodhouse
2018-01-04 19:40       ` Andrew Cooper
2018-01-04 19:46         ` David Woodhouse
2018-01-04 21:22       ` Van De Ven, Arjan
2018-01-05 11:32         ` Paolo Bonzini
2018-01-05 12:09           ` Paul Turner
2018-01-05 14:45           ` Van De Ven, Arjan
2018-01-05 14:43         ` Andrea Arcangeli
2018-01-05 14:52           ` Van De Ven, Arjan
2018-01-05 15:03             ` Andrea Arcangeli
2018-01-05 14:54           ` Thomas Gleixner
2018-01-05 11:52       ` Paul Turner
2018-01-05 14:28         ` David Woodhouse
2018-01-05 14:42           ` Van De Ven, Arjan
2018-01-05 15:38             ` David Woodhouse
2018-01-05 16:05               ` Andrea Arcangeli
2018-01-05 16:37                 ` David Woodhouse
2018-01-05 16:42                   ` Andrea Arcangeli
2018-01-05 16:44                     ` Van De Ven, Arjan
2018-01-05 16:46                     ` David Woodhouse
2018-01-05  5:25   ` Florian Weimer
2018-01-05 11:05     ` David Woodhouse
2018-01-04 19:05 ` Justin Forbes
2018-01-04 19:10   ` Tim Chen
2018-01-04 21:01     ` Yves-Alexis Perez
2018-01-05 13:28       ` Greg KH
2018-01-05 13:47         ` Yves-Alexis Perez
2018-01-05 14:01           ` Greg KH
2018-01-05 14:26             ` Paolo Bonzini
2018-01-05 14:54               ` Yves-Alexis Perez
2018-01-05 12:27 [PATCH 3/7] x86/enter: Use IBRS on syscall and interrupts Dr. David Alan Gilbert
2018-01-05 18:00 ` Dave Hansen

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=e2b435e0-7ecc-1ed4-4937-230fed226866@intel.com \
    --to=dave.hansen@intel.com \
    --cc=aarcange@redhat.com \
    --cc=ak@linux.intel.com \
    --cc=arjan.van.de.ven@intel.com \
    --cc=gregkh@linuxfoundation.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=luto@kernel.org \
    --cc=tglx@linutronix.de \
    --cc=tim.c.chen@linux.intel.com \
    --cc=torvalds@linux-foundation.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.