xen-devel.lists.xenproject.org archive mirror
 help / color / mirror / Atom feed
From: He Chen <he.chen@linux.intel.com>
To: Jan Beulich <JBeulich@suse.com>
Cc: Andrew Cooper <andrew.cooper3@citrix.com>,
	Wu Feng <feng.wu@intel.com>, Wang Yong <yong.y.wang@intel.com>,
	Nakajima Jun <jun.nakajima@intel.com>,
	xen-devel@lists.xen.org
Subject: Re: [PATCH V2] xen: support enabling SMEP/SMAP for HVM only
Date: Fri, 12 Aug 2016 18:03:20 +0800	[thread overview]
Message-ID: <20160812100320.GA31791@HE> (raw)
In-Reply-To: <57AC963E0200007800105182@prv-mh.provo.novell.com>

On Thu, Aug 11, 2016 at 07:14:06AM -0600, Jan Beulich wrote:
> >>> On 11.08.16 at 11:17, <he.chen@linux.intel.com> wrote:
> > @@ -1404,12 +1438,20 @@ void __init noreturn __start_xen(unsigned long mbi_p)
> >      if ( !opt_smep )
> >          setup_clear_cpu_cap(X86_FEATURE_SMEP);
> >      if ( cpu_has_smep )
> > +    {
> >          set_in_cr4(X86_CR4_SMEP);
> > +        if ( smep_hvm_only )
> > +            write_cr4(read_cr4() & ~X86_CR4_SMEP);
> > +    }
> 
> So that'll clear CR4.SMEP right here, but won't help with CR4 loads
> from mmu_cr4_features (as e.g. happens indirectly during VM exits,
> since the HOST_CR4 field gets set from this variable).
> 
> Did you in fact test your change, including validation of the features
> correctly remaining off over the lifetime of the system?
> 
> Further, considering that you don't clear the two flags from Xen's
> internal feature bitmap, and taken together with the internal feature
> bitmap driving alternative instruction patching, I'd assume pointless
> (and performance wise perhaps harmful) patching to now take place.
> 
Let me see whether I understand this correctly...

Regarding alternative instruction patching, if enabling SMAP for HVM but
disabling it for Xen, SMAP bit must be set in x86_capability feature
bitmap and cleared in mmu_cr4_features, which means instruction patching
would take place and a #UD may occur (since SMAP is disable in Xen, but
STAC/CLAC are patched and called).

A little dirty solution I can think of now is to temperarily clear SMAP
bit in x86_capability before patching instruction and then set it back
when instruction patching finish, like:

```
if ( opt_smap == SMAP_HVM_ONLY )
    setup_clear_cpu_cap(X86_FEATURE_SMAP);

alternative_instructions();

if ( opt_smap == SMAP_HVM_ONLY )
    __set_bit(X86_FEATURE_SMAP, boot_cpu_data.x86_capability);
```

Appreciate it if you have a better solution.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
https://lists.xen.org/xen-devel

  reply	other threads:[~2016-08-12 10:03 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-08-11  9:17 [PATCH V2] xen: support enabling SMEP/SMAP for HVM only He Chen
2016-08-11 13:14 ` Jan Beulich
2016-08-12 10:03   ` He Chen [this message]
2016-08-12 11:56     ` Jan Beulich
  -- strict thread matches above, loose matches on Subject: below --
2016-08-11  3:02 He Chen

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=20160812100320.GA31791@HE \
    --to=he.chen@linux.intel.com \
    --cc=JBeulich@suse.com \
    --cc=andrew.cooper3@citrix.com \
    --cc=feng.wu@intel.com \
    --cc=jun.nakajima@intel.com \
    --cc=xen-devel@lists.xen.org \
    --cc=yong.y.wang@intel.com \
    /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;
as well as URLs for NNTP newsgroup(s).