[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v3 2/7] xen/nospec: Use always_inline to fix code gen for evaluate_nospec


  • To: Norbert Manthey <nmanthey@xxxxxxxxx>, Jan Beulich <jbeulich@xxxxxxxx>
  • From: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
  • Date: Tue, 29 Oct 2019 13:46:41 +0000
  • Authentication-results: esa2.hc3370-68.iphmx.com; dkim=none (message not signed) header.i=none; spf=None smtp.pra=andrew.cooper3@xxxxxxxxxx; spf=Pass smtp.mailfrom=Andrew.Cooper3@xxxxxxxxxx; spf=None smtp.helo=postmaster@xxxxxxxxxxxxxxx
  • Autocrypt: addr=andrew.cooper3@xxxxxxxxxx; prefer-encrypt=mutual; keydata= mQINBFLhNn8BEADVhE+Hb8i0GV6mihnnr/uiQQdPF8kUoFzCOPXkf7jQ5sLYeJa0cQi6Penp VtiFYznTairnVsN5J+ujSTIb+OlMSJUWV4opS7WVNnxHbFTPYZVQ3erv7NKc2iVizCRZ2Kxn srM1oPXWRic8BIAdYOKOloF2300SL/bIpeD+x7h3w9B/qez7nOin5NzkxgFoaUeIal12pXSR Q354FKFoy6Vh96gc4VRqte3jw8mPuJQpfws+Pb+swvSf/i1q1+1I4jsRQQh2m6OTADHIqg2E ofTYAEh7R5HfPx0EXoEDMdRjOeKn8+vvkAwhviWXTHlG3R1QkbE5M/oywnZ83udJmi+lxjJ5 YhQ5IzomvJ16H0Bq+TLyVLO/VRksp1VR9HxCzItLNCS8PdpYYz5TC204ViycobYU65WMpzWe LFAGn8jSS25XIpqv0Y9k87dLbctKKA14Ifw2kq5OIVu2FuX+3i446JOa2vpCI9GcjCzi3oHV e00bzYiHMIl0FICrNJU0Kjho8pdo0m2uxkn6SYEpogAy9pnatUlO+erL4LqFUO7GXSdBRbw5 gNt25XTLdSFuZtMxkY3tq8MFss5QnjhehCVPEpE6y9ZjI4XB8ad1G4oBHVGK5LMsvg22PfMJ ISWFSHoF/B5+lHkCKWkFxZ0gZn33ju5n6/FOdEx4B8cMJt+cWwARAQABtClBbmRyZXcgQ29v cGVyIDxhbmRyZXcuY29vcGVyM0BjaXRyaXguY29tPokCOgQTAQgAJAIbAwULCQgHAwUVCgkI CwUWAgMBAAIeAQIXgAUCWKD95wIZAQAKCRBlw/kGpdefoHbdD/9AIoR3k6fKl+RFiFpyAhvO 59ttDFI7nIAnlYngev2XUR3acFElJATHSDO0ju+hqWqAb8kVijXLops0gOfqt3VPZq9cuHlh IMDquatGLzAadfFx2eQYIYT+FYuMoPZy/aTUazmJIDVxP7L383grjIkn+7tAv+qeDfE+txL4 SAm1UHNvmdfgL2/lcmL3xRh7sub3nJilM93RWX1Pe5LBSDXO45uzCGEdst6uSlzYR/MEr+5Z JQQ32JV64zwvf/aKaagSQSQMYNX9JFgfZ3TKWC1KJQbX5ssoX/5hNLqxMcZV3TN7kU8I3kjK mPec9+1nECOjjJSO/h4P0sBZyIUGfguwzhEeGf4sMCuSEM4xjCnwiBwftR17sr0spYcOpqET ZGcAmyYcNjy6CYadNCnfR40vhhWuCfNCBzWnUW0lFoo12wb0YnzoOLjvfD6OL3JjIUJNOmJy RCsJ5IA/Iz33RhSVRmROu+TztwuThClw63g7+hoyewv7BemKyuU6FTVhjjW+XUWmS/FzknSi dAG+insr0746cTPpSkGl3KAXeWDGJzve7/SBBfyznWCMGaf8E2P1oOdIZRxHgWj0zNr1+ooF /PzgLPiCI4OMUttTlEKChgbUTQ+5o0P080JojqfXwbPAyumbaYcQNiH1/xYbJdOFSiBv9rpt TQTBLzDKXok86LkCDQRS4TZ/ARAAkgqudHsp+hd82UVkvgnlqZjzz2vyrYfz7bkPtXaGb9H4 Rfo7mQsEQavEBdWWjbga6eMnDqtu+FC+qeTGYebToxEyp2lKDSoAsvt8w82tIlP/EbmRbDVn 7bhjBlfRcFjVYw8uVDPptT0TV47vpoCVkTwcyb6OltJrvg/QzV9f07DJswuda1JH3/qvYu0p vjPnYvCq4NsqY2XSdAJ02HrdYPFtNyPEntu1n1KK+gJrstjtw7KsZ4ygXYrsm/oCBiVW/OgU g/XIlGErkrxe4vQvJyVwg6YH653YTX5hLLUEL1NS4TCo47RP+wi6y+TnuAL36UtK/uFyEuPy wwrDVcC4cIFhYSfsO0BumEI65yu7a8aHbGfq2lW251UcoU48Z27ZUUZd2Dr6O/n8poQHbaTd 6bJJSjzGGHZVbRP9UQ3lkmkmc0+XCHmj5WhwNNYjgbbmML7y0fsJT5RgvefAIFfHBg7fTY/i kBEimoUsTEQz+N4hbKwo1hULfVxDJStE4sbPhjbsPCrlXf6W9CxSyQ0qmZ2bXsLQYRj2xqd1 bpA+1o1j2N4/au1R/uSiUFjewJdT/LX1EklKDcQwpk06Af/N7VZtSfEJeRV04unbsKVXWZAk uAJyDDKN99ziC0Wz5kcPyVD1HNf8bgaqGDzrv3TfYjwqayRFcMf7xJaL9xXedMcAEQEAAYkC HwQYAQgACQUCUuE2fwIbDAAKCRBlw/kGpdefoG4XEACD1Qf/er8EA7g23HMxYWd3FXHThrVQ HgiGdk5Yh632vjOm9L4sd/GCEACVQKjsu98e8o3ysitFlznEns5EAAXEbITrgKWXDDUWGYxd pnjj2u+GkVdsOAGk0kxczX6s+VRBhpbBI2PWnOsRJgU2n10PZ3mZD4Xu9kU2IXYmuW+e5KCA vTArRUdCrAtIa1k01sPipPPw6dfxx2e5asy21YOytzxuWFfJTGnVxZZSCyLUO83sh6OZhJkk b9rxL9wPmpN/t2IPaEKoAc0FTQZS36wAMOXkBh24PQ9gaLJvfPKpNzGD8XWR5HHF0NLIJhgg 4ZlEXQ2fVp3XrtocHqhu4UZR4koCijgB8sB7Tb0GCpwK+C4UePdFLfhKyRdSXuvY3AHJd4CP 4JzW0Bzq/WXY3XMOzUTYApGQpnUpdOmuQSfpV9MQO+/jo7r6yPbxT7CwRS5dcQPzUiuHLK9i nvjREdh84qycnx0/6dDroYhp0DFv4udxuAvt1h4wGwTPRQZerSm4xaYegEFusyhbZrI0U9tJ B8WrhBLXDiYlyJT6zOV2yZFuW47VrLsjYnHwn27hmxTC/7tvG3euCklmkn9Sl9IAKFu29RSo d5bD8kMSCYsTqtTfT6W4A3qHGvIDta3ptLYpIAOD2sY3GYq2nf3Bbzx81wZK14JdDDHUX2Rs 6+ahAA==
  • Cc: Juergen Gross <jgross@xxxxxxxx>, Stefano Stabellini <sstabellini@xxxxxxxxxx>, Julien Grall <julien@xxxxxxx>, Wei Liu <wl@xxxxxxx>, Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>, George Dunlap <George.Dunlap@xxxxxxxxxxxxx>, Ian Jackson <ian.jackson@xxxxxxxxxx>, Xen-devel <xen-devel@xxxxxxxxxxxxxxxxxxxx>
  • Delivery-date: Tue, 29 Oct 2019 13:46:57 +0000
  • Ironport-sdr: v/fZ3+UOGLhbbBLIZNzk+u1yaSlqzbQUBV53ZeY99v7bWdqvTxBii4tr8qmhnEbqfaO1lywZgj 14XIkB02SL31vJ5TAKMX1kZmNY7Z0zX6+rEXIQBhAsyA0NnlVmZv8gQq/QEj2+G9wRv46kCFG8 3SCzZma2tpyNB4wQPxhq29kWVk2JYIWHM7zlJxtE9kNsp1QU/VBQCDeKuyTZ00p0gGKzEgxv/c y0Z9rbRsl8PlCfmSJPsUtzc5x24gMebHGCsYpyE573x6C4T8N2ABPgrFdXVyagg1pphM9uJWwJ 8o4=
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>
  • Openpgp: preference=signencrypt

On 29/10/2019 08:25, Norbert Manthey wrote:
> On 10/28/19 18:05, Andrew Cooper wrote:
>> On 25/10/2019 22:56, Norbert Manthey wrote:
>>> On 10/25/19 17:40, Jan Beulich wrote:
>>>> On 25.10.2019 17:27, Andrew Cooper wrote:
>>>>> On 25/10/2019 13:34, Jan Beulich wrote:
>>>>>> On 25.10.2019 14:10, Andrew Cooper wrote:
>>>>>>> The two choices to unblock 4.13 are this patch, or the previous version
>>>>>>> which made CONFIG_HARDEN_BRANCH depend on BROKEN, which was even more
>>>>>>> disliked.
>>>>>> Option 3 is to have just the config option, for people to turn this
>>>>>> off if they feel like doing so.
>>>>> Yes, but no.  A facade of security is worse than no security, and I
>>>>> don't consider doing that an acceptable solution in this case.
>>>> But I thought we all agree that this is something that's presumably
>>>> going to remain incomplete (as in not provably complete) altogether
>>>> anyway. It's just that without the change here it's far more
>>>> incomplete then with it.
>> This is inherently a best-effort approach, but without the
>> always_inline, it is tantamount to useless.
>>
>> Only the grant table operations, and __mfn_valid() are correctly
>> protected with the code in its current form, where as the large changes
>> (in terms of absolute number of protected paths) comes from the predicates.
>>
>>>> In any event I think we should (also) have an opinion from the people
>>>> who had originally contributed this logic. You didn't Cc anyone of
>>>> them; I've added at least Norbert now.
>>> Thanks for adding me. I had a quick look into the discussion. Only
>>> making adding lfence statements around conditionals depending on config
>>> BROKEN does not help, as it would still need to be always_inline to work
>>> as expected, correct?
>> "depends on BROKEN" is a way to unconditionally compile out
>> functionality, which was one option considered to this problem.
>>
>>> Hence, in my opinion, this patch does the right
>>> thing to benefit from the lfences that are placed after evaluation
>>> conditionals.
>> This patch is the alternative to compiling everything out.
>>
>> I'm still holding out hope that we'll find a better compiler based
>> mitigation in the longer term, because I don't see either of these
>> options being viable strategies longterm.
> I fully agree that in the long run, we should have compiler support, to
> e.g. not move lfence statements around.
>
> However, until we have that, we should allow users of Xen to get the
> lfence statements at the correct positions as a best effort practice.
> Hence, the always_inline modification should be there. In case you still
> want to compile out this functionality, you could even add a dependency
> on BROKEN on top, and then users can chose to compile it in, but at
> least get a version where the lfences are placed at the right position.

This is going around in circles.

The following patch in this series is a fully user-selectable Kconfig
option for whether they want to use branch hardening, and is in place
once there is a plausible expectation that the lfences are in suitable
positions.

If this patch series does not agreement, I will unblock livepatching on
4.13 by committing the v2 patch which causes BRANCH_HARDEN to depend on
BROKEN and force it to be compiled out with no user choice at all.

Unbreaking livepatching is strictly more important than keeping a brand
new feature in 4.13 in a broken form.  I've provided two alternative
strategies to fix the 4.13 blockers, but if noone can agree on which
approach to use, I will commit the simpler fix.

~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.