Netdev List
 help / color / mirror / Atom feed
From: Alexei Starovoitov <alexei.starovoitov@gmail.com>
To: Alan Cox <gnomes@lxorguk.ukuu.org.uk>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
	Dan Williams <dan.j.williams@intel.com>,
	linux-kernel@vger.kernel.org, linux-arch@vger.kernel.org,
	Andi Kleen <ak@linux.intel.com>, Arnd Bergmann <arnd@arndb.de>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Peter Zijlstra <peterz@infradead.org>,
	netdev@vger.kernel.org, Ingo Molnar <mingo@redhat.com>,
	"H. Peter Anvin" <hpa@zytor.com>,
	Thomas Gleixner <tglx@linutronix.de>
Subject: Re: [PATCH 06/18] x86, barrier: stop speculation for failed access_ok
Date: Sat, 6 Jan 2018 10:13:33 -0800	[thread overview]
Message-ID: <20180106181331.mmrqwwbu2jcjj2si@ast-mbp> (raw)
In-Reply-To: <20180106123242.77f4d860@alans-desktop>

On Sat, Jan 06, 2018 at 12:32:42PM +0000, Alan Cox wrote:
> On Fri, 5 Jan 2018 18:52:07 -0800
> Linus Torvalds <torvalds@linux-foundation.org> wrote:
> 
> > On Fri, Jan 5, 2018 at 5:10 PM, Dan Williams <dan.j.williams@intel.com> wrote:
> > > From: Andi Kleen <ak@linux.intel.com>
> > >
> > > When access_ok fails we should always stop speculating.
> > > Add the required barriers to the x86 access_ok macro.  
> > 
> > Honestly, this seems completely bogus.
> 
> Also for x86-64 if we are trusting that an AND with a constant won't get
> speculated into something else surely we can just and the address with ~(1
> << 63) before copying from/to user space ? The user will then just
> speculatively steal their own memory.

+1

Any type of straight line code can address variant 1.
Like changing:
  array[index]
into
  array[index & mask]
works even when 'mask' is a variable.
To proceed with speculative load from array cpu has to speculatively
load 'mask' from memory and speculatively do '&' alu.
If attacker cannot influence 'mask' the speculative value of it
will bound 'index & mask' value to be within array limits.

I think "lets sprinkle lfence everywhere" approach is going to
cause serious performance degradation. Yet people pushing for lfence
didn't present any numbers.
Last time lfence was removed from the networking drivers via dma_rmb()
packet-per-second metric jumped 10-30%. lfence forces all outstanding loads
to complete. If any prior load is waiting on L3 or memory,
lfence will cause 100+ ns stall and overall kernel performance will tank.
If kernel adopts this "lfence everywhere" approach it will be
the end of the kernel as we know it. All high performance operations
will move into user space. Networking and IO will be first.
Since it will takes years to design new cpus and even longer
to upgrade all servers the industry will have no choice,
but to move as much logic as possible from the kernel.

kpti already made crossing user/kernel boundary slower, but
kernel itself is still fast. If kernel will have lfence everywhere
the kernel itself will be slow.

In that sense retpolining the kernel is not as horrible as it sounds,
since both user space and kernel has to be retpolined.

  parent reply	other threads:[~2018-01-06 18:13 UTC|newest]

Thread overview: 158+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-01-06  1:09 [PATCH 00/18] prevent bounds-check bypass via speculative execution Dan Williams
2018-01-06  1:09 ` [PATCH 01/18] asm-generic/barrier: add generic nospec helpers Dan Williams
2018-01-06  2:55   ` Linus Torvalds
2018-01-06  5:23     ` Dan Williams
2018-01-06 17:08       ` Mark Rutland
2018-01-06  1:10 ` [PATCH 02/18] Documentation: document " Dan Williams
2018-01-08 16:29   ` Jonathan Corbet
2018-01-08 17:09     ` Mark Rutland
2018-01-08 21:19       ` Jonathan Corbet
2018-01-06  1:10 ` [PATCH 03/18] arm64: implement nospec_ptr() Dan Williams
2018-01-06  1:10 ` [PATCH 04/18] arm: " Dan Williams
2018-01-10  2:04   ` Laura Abbott
2018-01-10  7:40     ` Hanjun Guo
2018-01-10 17:24       ` Laura Abbott
2018-01-06  1:10 ` [PATCH 05/18] x86: implement nospec_barrier() Dan Williams
2018-01-06  1:10 ` [PATCH 06/18] x86, barrier: stop speculation for failed access_ok Dan Williams
2018-01-06  2:52   ` Linus Torvalds
2018-01-06  3:09     ` Linus Torvalds
2018-01-06 23:31       ` Dan Williams
2018-01-07  1:20         ` Linus Torvalds
2018-01-08 21:09           ` Dan Williams
2018-01-08 23:44             ` Linus Torvalds
2018-01-08 23:53               ` Dan Williams
2018-01-06  5:47     ` Dan Williams
2018-01-06 12:32     ` Alan Cox
2018-01-06 17:56       ` Linus Torvalds
2018-01-06 18:13       ` Alexei Starovoitov [this message]
2018-01-06 18:29         ` Dan Williams
2018-01-06 18:39           ` Alexei Starovoitov
2018-01-06 18:54             ` Dan Williams
2018-01-06 19:25               ` Alexei Starovoitov
2018-01-06 19:36                 ` Dan Williams
2018-01-06 19:41                 ` Thomas Gleixner
2018-01-08 10:02                   ` Andrea Arcangeli
2018-01-06 18:38         ` Alan Cox
2018-01-06 18:51           ` Alexei Starovoitov
2018-01-06 19:55             ` Alan Cox
2018-01-06 20:09               ` Alexei Starovoitov
2018-01-06 20:22                 ` Alan Cox
2018-01-06 21:17                   ` Alexei Starovoitov
2018-01-06 21:21                     ` Thomas Gleixner
2018-01-06 23:05                     ` Alan Cox
2018-01-07  3:38                       ` Alexei Starovoitov
2018-01-07  6:33                         ` Willy Tarreau
2018-01-07 19:47                           ` Linus Torvalds
2018-01-07 20:12                             ` Willy Tarreau
2018-01-07 20:17                               ` Linus Torvalds
2018-01-07 20:56                                 ` Thomas Gleixner
2018-01-08  2:23                                   ` David Miller
2018-01-08  7:38                                     ` Greg KH
2018-01-07 22:15                                 ` Willy Tarreau
2018-01-07 20:15                             ` Dan Williams
2018-01-08  2:24                               ` Alexei Starovoitov
2018-01-08  9:51                                 ` Peter Zijlstra
2018-01-08 18:21                                   ` Ingo Molnar
2018-01-08 12:00                             ` David Laight
2018-01-08 12:12                               ` Alan Cox
2018-01-08 12:33                                 ` David Laight
2018-01-07 10:08                         ` Thomas Gleixner
2018-01-08  2:09                           ` Alexei Starovoitov
2018-01-07 13:59                         ` Alan Cox
2018-01-08  2:57                           ` Alexei Starovoitov
2018-01-08  9:57                             ` Peter Zijlstra
2018-01-06 20:42           ` Willy Tarreau
2018-01-07  1:36             ` David Miller
2018-01-07 17:19               ` James Bottomley
2018-01-07 18:31                 ` Thomas Gleixner
2018-01-08  2:04                   ` David Miller
2018-01-07 19:24                 ` Alan Cox
2018-01-09 21:41     ` Josh Poimboeuf
2018-01-09 21:47       ` Dan Williams
2018-01-09 21:49         ` Josh Poimboeuf
2018-01-09 21:59           ` Dan Williams
2018-01-09 22:23             ` Josh Poimboeuf
2018-01-09 22:35               ` Dan Williams
2018-01-06  1:10 ` [PATCH 07/18] [media] uvcvideo: prevent bounds-check bypass via speculative execution Dan Williams
2018-01-06  9:09   ` Greg KH
2018-01-06  9:40     ` Greg KH
2018-01-06 17:41       ` Dan Williams
2018-01-07  9:09         ` Greg KH
2018-01-07 19:37           ` Dan Williams
2018-01-09  8:40       ` Laurent Pinchart
2018-01-09 10:04         ` Greg KH
2018-01-09 14:26           ` Laurent Pinchart
2018-01-09 14:47             ` Greg KH
2018-01-08 11:23   ` Laurent Pinchart
2018-01-09  2:11     ` Dan Williams
2018-01-06  1:10 ` [PATCH 08/18] carl9170: " Dan Williams
2018-01-06 10:01   ` Sergei Shtylyov
2018-01-06 14:23   ` Christian Lamparter
2018-01-06 15:06     ` Alan Cox
2018-01-06 16:38       ` Christian Lamparter
2018-01-06 16:34     ` Dan Williams
2018-01-06  1:10 ` [PATCH 09/18] p54: " Dan Williams
2018-01-06 10:01   ` Sergei Shtylyov
2018-01-06  1:10 ` [PATCH 10/18] qla2xxx: " Dan Williams
2018-01-06  9:03   ` Greg KH
2018-01-06  9:42     ` Greg KH
2018-01-11 22:15     ` Dan Williams
2018-01-12  7:27       ` Greg KH
2018-01-12 15:25         ` James Bottomley
2018-01-06  1:10 ` [PATCH 11/18] cw1200: " Dan Williams
2018-01-06  1:10 ` [PATCH 12/18] Thermal/int340x: " Dan Williams
2018-01-06  1:53   ` Srinivas Pandruvada
2018-01-06  1:57     ` Dan Williams
2018-01-06 17:24       ` Srinivas Pandruvada
2018-01-06 10:03   ` Sergei Shtylyov
2018-01-06  1:11 ` [PATCH 13/18] ipv6: " Dan Williams
2018-01-06 10:04   ` Sergei Shtylyov
2018-01-06 14:48   ` Stephen Hemminger
2018-01-06 18:05     ` Dan Williams
2018-01-06  1:11 ` [PATCH 14/18] ipv4: " Dan Williams
2018-01-06  9:00   ` Greg KH
2018-01-06  9:01   ` Greg KH
2018-01-06 12:23     ` Alan Cox
2018-01-06 15:14       ` Greg KH
2018-01-06 16:29         ` Dan Williams
2018-01-06 18:10           ` Dan Williams
2018-01-06 10:04   ` Sergei Shtylyov
2018-01-06  1:11 ` [PATCH 15/18] vfs, fdtable: " Dan Williams
2018-01-06 10:05   ` Sergei Shtylyov
2018-01-06  1:11 ` [PATCH 16/18] net: mpls: " Dan Williams
2018-01-06 10:06   ` Sergei Shtylyov
2018-01-09  3:11   ` Eric W. Biederman
2018-01-09  3:42     ` Dan Williams
2018-01-09  4:13       ` Linus Torvalds
2018-01-09  4:21         ` Linus Torvalds
2018-01-10  0:48         ` Dan Williams
2018-01-10  1:33           ` Dan Williams
2018-01-10  1:57           ` Alexei Starovoitov
2018-01-10  2:22             ` Dan Williams
2018-01-10  3:07               ` Alexei Starovoitov
2018-01-10  3:27           ` Linus Torvalds
2018-01-09 16:17       ` Eric W. Biederman
2018-01-09 18:01         ` Dan Williams
2018-01-10  0:54           ` Eric W. Biederman
2018-01-10  1:31             ` Dan Williams
2018-01-06  1:11 ` [PATCH 17/18] udf: " Dan Williams
2018-01-08 10:20   ` Jan Kara
2018-01-06  1:11 ` [PATCH 18/18] userns: " Dan Williams
2018-01-06  2:22 ` [PATCH 00/18] " Eric W. Biederman
     [not found]   ` <87y3lbpvzp.fsf-aS9lmoZGLiVWk0Htik3J/w@public.gmane.org>
2018-01-06  6:30     ` Dan Williams
2018-01-08 10:08       ` Peter Zijlstra
2018-01-08 11:43         ` Alan Cox
2018-01-08 11:55           ` Peter Zijlstra
2018-01-08 18:33           ` Ingo Molnar
2018-01-08 16:20       ` Bart Van Assche
2018-01-06 18:56 ` Florian Fainelli
2018-01-06 18:59   ` Arjan van de Ven
2018-01-06 19:37 ` Dan Williams
2018-01-06 20:07   ` Dan Williams
2018-01-09 19:34 ` Jiri Kosina
2018-01-09 19:44   ` Dan Williams
2018-01-09 20:55     ` Josh Poimboeuf
2018-01-11  9:54       ` Jiri Kosina
2018-01-11 15:58         ` Dan Williams
2018-01-11 16:34           ` Daniel Borkmann
  -- strict thread matches above, loose matches on Subject: below --
2018-01-09  0:12 [PATCH 06/18] x86, barrier: stop speculation for failed access_ok Linus Torvalds

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=20180106181331.mmrqwwbu2jcjj2si@ast-mbp \
    --to=alexei.starovoitov@gmail.com \
    --cc=ak@linux.intel.com \
    --cc=arnd@arndb.de \
    --cc=dan.j.williams@intel.com \
    --cc=gnomes@lxorguk.ukuu.org.uk \
    --cc=gregkh@linuxfoundation.org \
    --cc=hpa@zytor.com \
    --cc=linux-arch@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@redhat.com \
    --cc=netdev@vger.kernel.org \
    --cc=peterz@infradead.org \
    --cc=tglx@linutronix.de \
    --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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox