public inbox for kvm@vger.kernel.org
 help / color / mirror / Atom feed
From: Sean Christopherson <seanjc@google.com>
To: Paolo Bonzini <pbonzini@redhat.com>
Cc: kvm@vger.kernel.org, Sean Christopherson <seanjc@google.com>
Subject: [kvm-unit-tests PATCH 01/14] x86/cstart: Don't use MSR_GS_BASE in 32-bit boot code
Date: Wed, 21 Apr 2021 20:04:51 -0700	[thread overview]
Message-ID: <20210422030504.3488253-2-seanjc@google.com> (raw)
In-Reply-To: <20210422030504.3488253-1-seanjc@google.com>

Load the per-cpu GS.base for 32-bit build by building a temporary GDT
and loading a "real" segment.  Using MSR_GS_BASE is wrong and broken,
it's a 64-bit only MSR and does not exist on 32-bit CPUs.  The current
code works only because 32-bit KVM VMX incorrectly disables interception
of MSR_GS_BASE, and no one runs KVM on an actual 32-bit physical CPU,
i.e. the MSR exists in hardware and so everything "works".

32-bit KVM SVM is not buggy and correctly injects #GP on the WRMSR, i.e.
the tests have never worked on 32-bit SVM.

Fixes: dfe6cb6 ("Add 32 bit smp initialization code")
Signed-off-by: Sean Christopherson <seanjc@google.com>
---
 x86/cstart.S | 28 +++++++++++++++++++++++-----
 1 file changed, 23 insertions(+), 5 deletions(-)

diff --git a/x86/cstart.S b/x86/cstart.S
index 489c561..91970a2 100644
--- a/x86/cstart.S
+++ b/x86/cstart.S
@@ -89,13 +89,31 @@ mb_flags = 0x0
 	.long mb_magic, mb_flags, 0 - (mb_magic + mb_flags)
 mb_cmdline = 16
 
-MSR_GS_BASE = 0xc0000101
-
 .macro setup_percpu_area
 	lea -4096(%esp), %eax
-	mov $0, %edx
-	mov $MSR_GS_BASE, %ecx
-	wrmsr
+
+	mov %eax, %edx
+	shl $16, %edx
+	or  $0xffff, %edx
+	mov %edx, 0x10(%eax)
+
+	mov %eax, %edx
+	and $0xff000000, %edx
+	mov %eax, %ecx
+	shr $16, %ecx
+	and $0xff, %ecx
+	or  %ecx, %edx
+	or  $0x00cf9300, %edx
+	mov %edx, 0x14(%eax)
+
+	movw $0x17, 0(%eax)
+	mov %eax, 2(%eax)
+	lgdtl (%eax)
+
+	mov $0x10, %ax
+	mov %ax, %gs
+
+	lgdtl gdt32_descr
 .endm
 
 .macro setup_segments
-- 
2.31.1.498.g6c1eba8ee3d-goog


  reply	other threads:[~2021-04-22  3:05 UTC|newest]

Thread overview: 21+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-04-22  3:04 [kvm-unit-tests PATCH 00/14] x86: MSR_GS_BASE and friends Sean Christopherson
2021-04-22  3:04 ` Sean Christopherson [this message]
2021-04-22  9:44   ` [kvm-unit-tests PATCH 01/14] x86/cstart: Don't use MSR_GS_BASE in 32-bit boot code Paolo Bonzini
2021-04-22 10:02   ` Paolo Bonzini
2021-04-22 17:57     ` Sean Christopherson
2021-04-23  6:57       ` Paolo Bonzini
2021-04-22  3:04 ` [kvm-unit-tests PATCH 02/14] x86: msr: Exclude GS/FS_BASE MSRs from 32-bit builds Sean Christopherson
2021-04-22  3:04 ` [kvm-unit-tests PATCH 03/14] x86: msr: Advertise GenuineIntel as vendor to play nice with SYSENTER Sean Christopherson
2021-04-22 10:11   ` Paolo Bonzini
2021-04-22  3:04 ` [kvm-unit-tests PATCH 04/14] x86: msr: Restore original MSR value after writing arbitrary test value Sean Christopherson
2021-04-22  3:04 ` [kvm-unit-tests PATCH 05/14] x86: Force the compiler to retrieve exception info from per-cpu area Sean Christopherson
2021-04-22  3:04 ` [kvm-unit-tests PATCH 06/14] x86: msr: Replace spaces with tabs in all of msr.c Sean Christopherson
2021-04-22  3:04 ` [kvm-unit-tests PATCH 07/14] x86: msr: Use ARRAY_SIZE() instead of open coded equivalent Sean Christopherson
2021-04-22  3:04 ` [kvm-unit-tests PATCH 08/14] x86: msr: Use the #defined MSR indices in favor of open coding the values Sean Christopherson
2021-04-22  3:04 ` [kvm-unit-tests PATCH 09/14] x86: msr: Drop the explicit expected value Sean Christopherson
2021-04-22  3:05 ` [kvm-unit-tests PATCH 10/14] x86: msr: Add builder macros to define MSR entries Sean Christopherson
2021-04-22  3:05 ` [kvm-unit-tests PATCH 11/14] x86: msr: Pass msr_info instead of doing a lookup at runtime Sean Christopherson
2021-04-22  3:05 ` [kvm-unit-tests PATCH 12/14] x86: msr: Verify 64-bit only MSRs fault on 32-bit hosts Sean Christopherson
2021-04-22 10:32   ` Paolo Bonzini
2021-04-22  3:05 ` [kvm-unit-tests PATCH 13/14] x86: msr: Test that always-canonical MSRs #GP on non-canonical value Sean Christopherson
2021-04-22  3:05 ` [kvm-unit-tests PATCH 14/14] x86: msr: Verify that EFER.SCE can be written on 32-bit vCPUs Sean Christopherson

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=20210422030504.3488253-2-seanjc@google.com \
    --to=seanjc@google.com \
    --cc=kvm@vger.kernel.org \
    --cc=pbonzini@redhat.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