From: Paolo Bonzini <pbonzini@redhat.com>
To: Gleb Natapov <gleb@cloudius-systems.com>
Cc: linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
jroedel@suse.de, agraf@suse.de, valentine.sinitsyn@gmail.com,
jan.kiszka@siemens.com, avi@cloudius-systems.com
Subject: Re: [PATCH 3/4] KVM: x86: inject nested page faults on emulated instructions
Date: Thu, 04 Sep 2014 16:12:19 +0200 [thread overview]
Message-ID: <54087343.5050409@redhat.com> (raw)
In-Reply-To: <20140904070220.GL9842@cloudius-systems.com>
Il 04/09/2014 09:02, Gleb Natapov ha scritto:
> On Tue, Sep 02, 2014 at 05:13:49PM +0200, Paolo Bonzini wrote:
>> > This is required for the following patch to work correctly. If a nested page
>> > fault happens during emulation, we must inject a vmexit, not a page fault.
>> > Luckily we already have the required machinery: it is enough to return
>> > X86EMUL_INTERCEPTED instead of X86EMUL_PROPAGATE_FAULT.
>> >
> I wonder why this patch is needed. X86EMUL_PROPAGATE_FAULT causes
> ctxt->have_exception to be set to true in x86_emulate_insn().
> x86_emulate_instruction() checks ctxt->have_exception and calls
> inject_emulated_exception() if it is true. inject_emulated_exception()
> calls kvm_propagate_fault() where we check if the fault was nested and
> generate vmexit or a page fault accordingly.
Good question. :)
If you do that, KVM gets down to the "if (writeback)" and writes the
ctxt->eip from L2 into the L1 EIP.
Possibly this patch can be replaced by just this?
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 022513b..475e979 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -5312,7 +5312,7 @@ restart:
if (ctxt->have_exception) {
inject_emulated_exception(vcpu);
- r = EMULATE_DONE;
+ return EMULATE_DONE;
} else if (vcpu->arch.pio.count) {
if (!vcpu->arch.pio.in) {
/* FIXME: return into emulator if single-stepping. */
But I'm not sure how to test it, and I like the idea of treating nested page
faults like other nested vmexits during emulation (which is what this patch
does).
If I included this patch, I could then remove kvm_propagate_fault
like (I think) this:
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 92493e10937c..e096db566ac2 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -4910,9 +4902,10 @@ static void toggle_interruptibility(struct kvm_vcpu *vcpu, u32 mask)
static void inject_emulated_exception(struct kvm_vcpu *vcpu)
{
struct x86_emulate_ctxt *ctxt = &vcpu->arch.emulate_ctxt;
- if (ctxt->exception.vector == PF_VECTOR)
- kvm_propagate_fault(vcpu, &ctxt->exception);
- else if (ctxt->exception.error_code_valid)
+ if (ctxt->exception.vector == PF_VECTOR) {
+ WARN_ON(fault->nested_page_fault);
+ vcpu->arch.walk_mmu->inject_page_fault(vcpu, fault);
+ } else if (ctxt->exception.error_code_valid)
kvm_queue_exception_e(vcpu, ctxt->exception.vector,
ctxt->exception.error_code);
else
What do you think?
Paolo
next prev parent reply other threads:[~2014-09-04 14:12 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-09-02 15:13 [PATCH 0/4] KVM: nested x86: nested page faults fixes Paolo Bonzini
2014-09-02 15:13 ` [PATCH 1/4] KVM: x86: reserve bit 8 of non-leaf PDPEs and PML4Es in 64-bit mode on AMD Paolo Bonzini
2014-09-02 15:13 ` [PATCH 2/4] KVM: nSVM: propagate the NPF EXITINFO to the guest Paolo Bonzini
2014-09-02 16:33 ` Joerg Roedel
2014-09-02 16:46 ` Paolo Bonzini
2014-09-02 17:01 ` Paolo Bonzini
2014-09-02 17:01 ` Joerg Roedel
2014-09-02 17:47 ` Avi Kivity
2014-09-02 15:13 ` [PATCH 3/4] KVM: x86: inject nested page faults on emulated instructions Paolo Bonzini
2014-09-04 7:02 ` Gleb Natapov
2014-09-04 14:12 ` Paolo Bonzini [this message]
2014-09-04 15:05 ` Gleb Natapov
2014-09-04 17:17 ` Paolo Bonzini
2014-09-04 17:44 ` Paolo Bonzini
2014-09-05 9:47 ` Gleb Natapov
2014-09-02 15:13 ` [PATCH 4/4] KVM: x86: propagate exception from permission checks on the nested page fault Paolo Bonzini
2014-09-02 16:02 ` [PATCH 0/4] KVM: nested x86: nested page faults fixes Valentine Sinitsyn
2014-09-02 16:56 ` Paolo Bonzini
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=54087343.5050409@redhat.com \
--to=pbonzini@redhat.com \
--cc=agraf@suse.de \
--cc=avi@cloudius-systems.com \
--cc=gleb@cloudius-systems.com \
--cc=jan.kiszka@siemens.com \
--cc=jroedel@suse.de \
--cc=kvm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=valentine.sinitsyn@gmail.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).