All of lore.kernel.org
 help / color / mirror / Atom feed
From: Peter Zijlstra <peterz@infradead.org>
To: x86@kernel.org
Cc: kys@microsoft.com, haiyangz@microsoft.com, wei.liu@kernel.org,
	decui@microsoft.com, tglx@linutronix.de, mingo@redhat.com,
	bp@alien8.de, dave.hansen@linux.intel.com, hpa@zytor.com,
	seanjc@google.com, pbonzini@redhat.com, ardb@kernel.org,
	kees@kernel.org, Arnd Bergmann <arnd@arndb.de>,
	gregkh@linuxfoundation.org, jpoimboe@kernel.org,
	peterz@infradead.org, linux-hyperv@vger.kernel.org,
	linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
	linux-efi@vger.kernel.org, samitolvanen@google.com,
	ojeda@kernel.org
Subject: [PATCH v2 02/13] x86/kvm/emulate: Introduce COP1
Date: Wed, 30 Apr 2025 13:07:36 +0200	[thread overview]
Message-ID: <20250430112349.208590367@infradead.org> (raw)
In-Reply-To: 20250430110734.392235199@infradead.org

Replace fastops with C-ops. There are a bunch of problems with the
current fastop infrastructure, most all related to their special
calling convention, which bypasses the normal C-ABI.

There are two immediate problems with this at present:

 - it relies on RET preserving EFLAGS; whereas C-ABI does not.

 - it circumvents compiler based control-flow-integrity checking
   because its all asm magic.

The first is a problem for some mitigations where the
x86_indirect_return_thunk needs to include non-trivial work that
clobbers EFLAGS (eg. the Skylake call depth tracking thing).

The second is a problem because it presents a 'naked' indirect call on
kCFI builds, making it a prime target for control flow hijacking.

Additionally, given that a large chunk of virtual machine performance
relies on absolutely avoiding vmexit these days, this emulation stuff
just isn't that critical for performance anymore.

As such, replace the fastop calls with a normal C function using the
'execute' member.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
 arch/x86/kvm/emulate.c |   69 ++++++++++++++++++++++++++++++++++++++++---------
 1 file changed, 57 insertions(+), 12 deletions(-)

--- a/arch/x86/kvm/emulate.c
+++ b/arch/x86/kvm/emulate.c
@@ -267,11 +267,56 @@ static void invalidate_registers(struct
 		     X86_EFLAGS_PF|X86_EFLAGS_CF)
 
 #ifdef CONFIG_X86_64
-#define ON64(x) x
+#define ON64(x...) x
 #else
 #define ON64(x)
 #endif
 
+#define COP_START(op) \
+static int em_##op(struct x86_emulate_ctxt *ctxt) \
+{ \
+	unsigned long flags = (ctxt->eflags & EFLAGS_MASK) | X86_EFLAGS_IF; \
+	int bytes = 1, ok = 1; \
+	if (!(ctxt->d & ByteOp)) \
+		bytes = ctxt->dst.bytes; \
+	switch (bytes) {
+
+#define COP_ASM(str) \
+		asm("push %[flags]; popf \n\t" \
+		    "10: " str \
+		    "pushf; pop %[flags] \n\t" \
+		    "11: \n\t" \
+		    : "+a" (ctxt->dst.val), \
+		      "+d" (ctxt->src.val), \
+		      [flags] "+D" (flags), \
+		      "+S" (ok) \
+		    : "c" (ctxt->src2.val))
+
+#define COP_ASM1(op, dst) \
+		COP_ASM(#op " %%" #dst " \n\t")
+
+#define COP_ASM1_EX(op, dst) \
+		COP_ASM(#op " %%" #dst " \n\t" \
+			_ASM_EXTABLE_TYPE_REG(10b, 11f, EX_TYPE_ZERO_REG, %%esi))
+
+#define COP_ASM2(op, dst, src) \
+		COP_ASM(#op " %" #src ", %" #dst " \n\t")
+
+#define COP_END \
+	} \
+	ctxt->eflags = (ctxt->eflags & ~EFLAGS_MASK) | (flags & EFLAGS_MASK); \
+	return !ok ? emulate_de(ctxt) : X86EMUL_CONTINUE; \
+}
+
+/* 1-operand, using "a" (dst) */
+#define COP1(op) \
+	COP_START(op) \
+	case 1: COP_ASM1(op##b, al); break; \
+	case 2: COP_ASM1(op##w, ax); break; \
+	case 4: COP_ASM1(op##l, eax); break; \
+	ON64(case 8: COP_ASM1(op##q, rax); break;) \
+	COP_END
+
 /*
  * fastop functions have a special calling convention:
  *
@@ -1002,10 +1047,10 @@ FASTOP3WCL(shrd);
 
 FASTOP2W(imul);
 
-FASTOP1(not);
-FASTOP1(neg);
-FASTOP1(inc);
-FASTOP1(dec);
+COP1(not);
+COP1(neg);
+COP1(inc);
+COP1(dec);
 
 FASTOP2CL(rol);
 FASTOP2CL(ror);
@@ -4021,8 +4066,8 @@ static const struct opcode group2[] = {
 static const struct opcode group3[] = {
 	F(DstMem | SrcImm | NoWrite, em_test),
 	F(DstMem | SrcImm | NoWrite, em_test),
-	F(DstMem | SrcNone | Lock, em_not),
-	F(DstMem | SrcNone | Lock, em_neg),
+	I(DstMem | SrcNone | Lock, em_not),
+	I(DstMem | SrcNone | Lock, em_neg),
 	F(DstXacc | Src2Mem, em_mul_ex),
 	F(DstXacc | Src2Mem, em_imul_ex),
 	F(DstXacc | Src2Mem, em_div_ex),
@@ -4030,14 +4075,14 @@ static const struct opcode group3[] = {
 };
 
 static const struct opcode group4[] = {
-	F(ByteOp | DstMem | SrcNone | Lock, em_inc),
-	F(ByteOp | DstMem | SrcNone | Lock, em_dec),
+	I(ByteOp | DstMem | SrcNone | Lock, em_inc),
+	I(ByteOp | DstMem | SrcNone | Lock, em_dec),
 	N, N, N, N, N, N,
 };
 
 static const struct opcode group5[] = {
-	F(DstMem | SrcNone | Lock,		em_inc),
-	F(DstMem | SrcNone | Lock,		em_dec),
+	I(DstMem | SrcNone | Lock,		em_inc),
+	I(DstMem | SrcNone | Lock,		em_dec),
 	I(SrcMem | NearBranch | IsBranch,       em_call_near_abs),
 	I(SrcMemFAddr | ImplicitOps | IsBranch, em_call_far),
 	I(SrcMem | NearBranch | IsBranch,       em_jmp_abs),
@@ -4237,7 +4282,7 @@ static const struct opcode opcode_table[
 	/* 0x38 - 0x3F */
 	F6ALU(NoWrite, em_cmp), N, N,
 	/* 0x40 - 0x4F */
-	X8(F(DstReg, em_inc)), X8(F(DstReg, em_dec)),
+	X8(I(DstReg, em_inc)), X8(I(DstReg, em_dec)),
 	/* 0x50 - 0x57 */
 	X8(I(SrcReg | Stack, em_push)),
 	/* 0x58 - 0x5F */



  parent reply	other threads:[~2025-04-30 11:26 UTC|newest]

Thread overview: 52+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-04-30 11:07 [PATCH v2 00/13] objtool: Detect and warn about indirect calls in __nocfi functions Peter Zijlstra
2025-04-30 11:07 ` [PATCH v2 01/13] x86/kvm/emulate: Implement test_cc() in C Peter Zijlstra
2025-04-30 11:07 ` Peter Zijlstra [this message]
2025-04-30 16:19   ` [PATCH v2 02/13] x86/kvm/emulate: Introduce COP1 Josh Poimboeuf
2025-04-30 19:05     ` Peter Zijlstra
2025-04-30 11:07 ` [PATCH v2 03/13] x86/kvm/emulate: Introduce COP2 Peter Zijlstra
2025-04-30 11:07 ` [PATCH v2 04/13] x86/kvm/emulate: Introduce COP2R Peter Zijlstra
2025-04-30 11:07 ` [PATCH v2 05/13] x86/kvm/emulate: Introduce COP2W Peter Zijlstra
2025-04-30 11:07 ` [PATCH v2 06/13] x86/kvm/emulate: Introduce COP2CL Peter Zijlstra
2025-04-30 11:07 ` [PATCH v2 07/13] x86/kvm/emulate: Introduce COP1SRC2 Peter Zijlstra
2025-04-30 11:07 ` [PATCH v2 08/13] x86/kvm/emulate: Introduce COP3WCL Peter Zijlstra
2025-04-30 11:07 ` [PATCH v2 09/13] x86/kvm/emulate: Convert em_salc() to C Peter Zijlstra
2025-04-30 11:07 ` [PATCH v2 10/13] x86/kvm/emulate: Remove fastops Peter Zijlstra
2025-04-30 11:07 ` [PATCH v2 11/13] x86,hyperv: Clean up hv_do_hypercall() Peter Zijlstra
2025-05-01  2:36   ` Michael Kelley
2025-04-30 11:07 ` [PATCH v2 12/13] x86_64,hyperv: Use direct call to hypercall-page Peter Zijlstra
2025-05-01  2:36   ` Michael Kelley
2025-05-01  8:59     ` Peter Zijlstra
2025-04-30 11:07 ` [PATCH v2 13/13] objtool: Validate kCFI calls Peter Zijlstra
2025-04-30 15:59   ` Josh Poimboeuf
2025-04-30 19:03     ` Peter Zijlstra
2025-05-01 15:56       ` Peter Zijlstra
2025-04-30 14:24 ` [PATCH v2 00/13] objtool: Detect and warn about indirect calls in __nocfi functions H. Peter Anvin
2025-04-30 19:06   ` Peter Zijlstra
2025-05-01 10:30     ` Peter Zijlstra
2025-05-01 15:38       ` Peter Zijlstra
2025-05-01 18:30         ` Sean Christopherson
2025-05-01 18:42           ` H. Peter Anvin
2025-05-01 18:59             ` Sean Christopherson
2025-05-02  6:12               ` Xin Li
2025-05-02  5:46           ` Xin Li
2025-05-02  5:48           ` Xin Li
2025-05-02 19:43             ` H. Peter Anvin
2025-05-02  8:40           ` Peter Zijlstra
2025-05-02 19:53             ` Sean Christopherson
2025-05-03  9:50               ` Peter Zijlstra
2025-05-03 18:28                 ` Josh Poimboeuf
2025-05-06  7:31                   ` Peter Zijlstra
2025-05-06 13:32                     ` Peter Zijlstra
2025-05-06 19:18                       ` Josh Poimboeuf
2025-05-28  7:44                         ` Peter Zijlstra
2025-05-28 16:30                           ` Peter Zijlstra
2025-05-28 16:35                             ` Peter Zijlstra
2025-05-29  9:30                               ` Peter Zijlstra
2025-06-03  5:43                                 ` Josh Poimboeuf
2025-06-03 16:29                                   ` Josh Poimboeuf
2025-06-05 17:19                                     ` Josh Poimboeuf
2025-06-06 10:49                                       ` Peter Zijlstra
2025-06-06 13:15                                         ` Sean Christopherson
2025-06-06 13:20                                           ` Peter Zijlstra
2025-05-01 18:33 ` Paolo Bonzini
2025-05-02 11:08   ` Peter Zijlstra

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=20250430112349.208590367@infradead.org \
    --to=peterz@infradead.org \
    --cc=ardb@kernel.org \
    --cc=arnd@arndb.de \
    --cc=bp@alien8.de \
    --cc=dave.hansen@linux.intel.com \
    --cc=decui@microsoft.com \
    --cc=gregkh@linuxfoundation.org \
    --cc=haiyangz@microsoft.com \
    --cc=hpa@zytor.com \
    --cc=jpoimboe@kernel.org \
    --cc=kees@kernel.org \
    --cc=kvm@vger.kernel.org \
    --cc=kys@microsoft.com \
    --cc=linux-efi@vger.kernel.org \
    --cc=linux-hyperv@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@redhat.com \
    --cc=ojeda@kernel.org \
    --cc=pbonzini@redhat.com \
    --cc=samitolvanen@google.com \
    --cc=seanjc@google.com \
    --cc=tglx@linutronix.de \
    --cc=wei.liu@kernel.org \
    --cc=x86@kernel.org \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.