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

[Xen-devel] Re: Addback capability check for non-initial features

  • To: "Dong, Eddie" <eddie.dong@xxxxxxxxx>
  • From: Keir Fraser <keir.xen@xxxxxxxxx>
  • Date: Fri, 10 Jun 2011 07:58:46 +0100
  • Cc: "xen-devel@xxxxxxxxxxxxxxxxxxx" <xen-devel@xxxxxxxxxxxxxxxxxxx>
  • Delivery-date: Fri, 10 Jun 2011 00:01:08 -0700
  • Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=user-agent:date:subject:from:to:cc:message-id:thread-topic :thread-index:in-reply-to:mime-version:content-type :content-transfer-encoding; b=qslcE5kgPmL4iW1UxOiKLnSOT5yK9pgnvj03wJwdN95vnVtR1RBT2uW5LlT0D4sJRg BMm7VJOaFNlNNes9yKNg2lKL0LnvXd2eMkwnUvmnImSJVVFfV/DK6PoHa6QV2Lnl5s7r j/fw8AeIOIVvFmOTieldoG1/qczO6mbAB2gnw=
  • List-id: Xen developer discussion <xen-devel.lists.xensource.com>
  • Thread-index: Acwm8ei0Bw+DOsTeG0qq3AI37qSa+QAPjT8AAACHkVAAAItBlwAAxALQAAEXpDQ=
  • Thread-topic: Addback capability check for non-initial features

On 10/06/2011 07:33, "Dong, Eddie" <eddie.dong@xxxxxxxxx> wrote:

>>> add back missing capability check of MSR_IA32_VMX_PROCBASED_CTLS.
>>> Besides initial configuration, adjust_vmx_controls is responsible for
>>> hardware capibility check as well. This patch add back the check.
>> I suppose the CPU_BASED_VIRTUAL_INTR_PENDING addition is correct, for
>> what
>> it's worth (surely every VMX-capable CPU ever has and will support that).
>> The change to CR8 detection looks mad and incorrect. You've inverted it so
>> that CR8 exits get enabled when TPR_SHADOW is available, rather than
>> when it
> CR8 exit is removed later on if TPR_SHADOW exist:)

Not in your patch. You remove it later if TPR_SHADOW *doesn't* exist.

> The only difference is that if there are processors that support TPR_SHADOW
> only, I can check internally if this is the concern.
> Current nested vmx is assuming CR8 exiting is presented to emulate L1 guest
> CR8 exiting. TPR_SHAOW can't trap CR8 read though cr8 write trap is OK w/ TPR
> shadow.

Hmm okay.

> Eventually I want to have a minimal common set of capability that is supported
> by all HW and is presented to L1 guest.
>> isn't, surely? And that can't be correct. I don't see how the CR8-exit
>> detection and enabling is wrong, as it is already.
> The original code for CR8 exit is correct too :)

More correct than yours :)

 -- Keir

> Thx, Eddie

Xen-devel mailing list



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