[RFC] a little disassembly infrastructure

[RFC] a little disassembly infrastructure

[Hollis Blanchard]

[-- Attachment #1: Type: text/plain, Size: 1123 bytes --]

Hi, I wrote this patch for KVM [1], but now that I look closer it seems
like there might be some overlapping functionality.

First there's emulate_instruction(), but since that only handles a few
instructions it's just an ordered list of if ((instruction & MASK_A) ==
INST_A) tests, so it doesn't actually parse out opcodes or anything.

I've also found xmon's ppc-opc.c. That parses the opcode and operands,
so could use some shared macros. Of course, the actual lookup isn't
time-sensitive, so that doesn't make sense to share. On the other hand,
if we do come up with something fast *and* robust for KVM, maybe xmon
could use that.

Of course, these macros alone is pretty small, so maybe it's not a big
deal to make a kvm-specific copy of them, leaving the other uses alone.

Comments?

[1] KVM on PowerPC traps when privileged instructions are executed in
the guest context. We must then (quickly!) disassemble them and emulate
their behavior. Right now we do this with a giant switch statement or
two, but are considering more sophisticated techniques in the future.

-- 
Hollis Blanchard
IBM Linux Technology Center

[-- Attachment #2: ppc-disassemble.diff --]
[-- Type: text/x-patch, Size: 3506 bytes --]

ppc: Create disassemble.h to extract instruction fields

This is used in a couple places in KVM, but isn't KVM-specific.

Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>

diff --git a/arch/powerpc/include/asm/disassemble.h b/arch/powerpc/include/asm/disassemble.h
new file mode 100644
--- /dev/null
+++ b/arch/powerpc/include/asm/disassemble.h
@@ -0,0 +1,80 @@
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+ *
+ * Copyright IBM Corp. 2008
+ *
+ * Authors: Hollis Blanchard <hollisb@us.ibm.com>
+ */
+
+#ifndef __ASM_PPC_DISASSEMBLE_H__
+#define __ASM_PPC_DISASSEMBLE_H__
+
+#include <linux/types.h>
+
+static inline unsigned int get_op(u32 inst)
+{
+	return inst >> 26;
+}
+
+static inline unsigned int get_xop(u32 inst)
+{
+	return (inst >> 1) & 0x3ff;
+}
+
+static inline unsigned int get_sprn(u32 inst)
+{
+	return ((inst >> 16) & 0x1f) | ((inst >> 6) & 0x3e0);
+}
+
+static inline unsigned int get_dcrn(u32 inst)
+{
+	return ((inst >> 16) & 0x1f) | ((inst >> 6) & 0x3e0);
+}
+
+static inline unsigned int get_rt(u32 inst)
+{
+	return (inst >> 21) & 0x1f;
+}
+
+static inline unsigned int get_rs(u32 inst)
+{
+	return (inst >> 21) & 0x1f;
+}
+
+static inline unsigned int get_ra(u32 inst)
+{
+	return (inst >> 16) & 0x1f;
+}
+
+static inline unsigned int get_rb(u32 inst)
+{
+	return (inst >> 11) & 0x1f;
+}
+
+static inline unsigned int get_rc(u32 inst)
+{
+	return inst & 0x1;
+}
+
+static inline unsigned int get_ws(u32 inst)
+{
+	return (inst >> 11) & 0x1f;
+}
+
+static inline unsigned int get_d(u32 inst)
+{
+	return inst & 0xffff;
+}
+
+#endif /* __ASM_PPC_DISASSEMBLE_H__ */
diff --git a/arch/powerpc/kvm/emulate.c b/arch/powerpc/kvm/emulate.c
--- a/arch/powerpc/kvm/emulate.c
+++ b/arch/powerpc/kvm/emulate.c
@@ -28,62 +28,7 @@
 #include <asm/time.h>
 #include <asm/byteorder.h>
 #include <asm/kvm_ppc.h>
-
-/* Instruction decoding */
-static inline unsigned int get_op(u32 inst)
-{
-	return inst >> 26;
-}
-
-static inline unsigned int get_xop(u32 inst)
-{
-	return (inst >> 1) & 0x3ff;
-}
-
-static inline unsigned int get_sprn(u32 inst)
-{
-	return ((inst >> 16) & 0x1f) | ((inst >> 6) & 0x3e0);
-}
-
-static inline unsigned int get_dcrn(u32 inst)
-{
-	return ((inst >> 16) & 0x1f) | ((inst >> 6) & 0x3e0);
-}
-
-static inline unsigned int get_rt(u32 inst)
-{
-	return (inst >> 21) & 0x1f;
-}
-
-static inline unsigned int get_rs(u32 inst)
-{
-	return (inst >> 21) & 0x1f;
-}
-
-static inline unsigned int get_ra(u32 inst)
-{
-	return (inst >> 16) & 0x1f;
-}
-
-static inline unsigned int get_rb(u32 inst)
-{
-	return (inst >> 11) & 0x1f;
-}
-
-static inline unsigned int get_rc(u32 inst)
-{
-	return inst & 0x1;
-}
-
-static inline unsigned int get_ws(u32 inst)
-{
-	return (inst >> 11) & 0x1f;
-}
-
-static inline unsigned int get_d(u32 inst)
-{
-	return inst & 0xffff;
-}
+#include <asm/disassemble.h>
 
 static void kvmppc_emulate_dec(struct kvm_vcpu *vcpu)
 {

Re: [RFC] a little disassembly infrastructure

[sonny]

Hollis Blanchard wrote:
> Hi, I wrote this patch for KVM [1], but now that I look closer it seems
> like there might be some overlapping functionality.
> 
> First there's emulate_instruction(), but since that only handles a few
> instructions it's just an ordered list of if ((instruction & MASK_A) ==
> INST_A) tests, so it doesn't actually parse out opcodes or anything.
> 
> I've also found xmon's ppc-opc.c. That parses the opcode and operands,
> so could use some shared macros. Of course, the actual lookup isn't
> time-sensitive, so that doesn't make sense to share. On the other hand,
> if we do come up with something fast *and* robust for KVM, maybe xmon
> could use that.
> 
> Of course, these macros alone is pretty small, so maybe it's not a big
> deal to make a kvm-specific copy of them, leaving the other uses alone.
> 
> Comments?
> 
> [1] KVM on PowerPC traps when privileged instructions are executed in
> the guest context. We must then (quickly!) disassemble them and emulate
> their behavior. Right now we do this with a giant switch statement or
> two, but are considering more sophisticated techniques in the future.
> 

Yeah, personally I like this change.  I was looking at emulating some 
PPC instructions in a driver and based on inspection I was going to use 
the code you had in for KVM.  Having these macros in a generic header is 
a "Good Thing" IMHO.  I think we should have a generic disassembly 
infrastructure and use that everywhere.

Re: [RFC] a little disassembly infrastructure

[Paul Mackerras]

Hollis Blanchard writes:

> I've also found xmon's ppc-opc.c. That parses the opcode and operands,
> so could use some shared macros.

That's a direct copy from GNU binutils.  I'm reluctant to modify it
because then maintenance becomes more than just copying in the latest
version.

Paul.

Re: [RFC] a little disassembly infrastructure

[Hollis Blanchard]

On Sat, 2008-10-25 at 13:55 +1100, Paul Mackerras wrote:
> Hollis Blanchard writes:
> 
> > I've also found xmon's ppc-opc.c. That parses the opcode and operands,
> > so could use some shared macros.
> 
> That's a direct copy from GNU binutils.  I'm reluctant to modify it
> because then maintenance becomes more than just copying in the latest
> version.

OK, makes sense.

Can I get an Acked-by on this patch, or should I move the code to a
KVM-specific header? Either way I'll submit it via the KVM tree as
usual.

-- 
Hollis Blanchard
IBM Linux Technology Center

Re: [RFC] a little disassembly infrastructure

[Hollis Blanchard]

On Fri, 2008-10-31 at 12:21 -0500, Hollis Blanchard wrote:
> On Sat, 2008-10-25 at 13:55 +1100, Paul Mackerras wrote:
> > Hollis Blanchard writes:
> > 
> > > I've also found xmon's ppc-opc.c. That parses the opcode and operands,
> > > so could use some shared macros.
> > 
> > That's a direct copy from GNU binutils.  I'm reluctant to modify it
> > because then maintenance becomes more than just copying in the latest
> > version.
> 
> OK, makes sense.
> 
> Can I get an Acked-by on this patch, or should I move the code to a
> KVM-specific header? Either way I'll submit it via the KVM tree as
> usual.

Paul, is this OK? I'm  waiting for your ack before sending a number of
dependent KVM patches to Avi.



ppc: Create disassemble.h to extract instruction fields

This is used in a couple places in KVM, but isn't KVM-specific.

However, xmon uses a direct copy of ppc_opc.c from binutils, so xmon won't use
this. emulate_instruction() doesn't need it either, because it can use a series
of mask tests.

Signed-off-by: Hollis Blanchard <hollisb@us.ibm.com>

diff --git a/arch/powerpc/include/asm/disassemble.h b/arch/powerpc/include/asm/disassemble.h
new file mode 100644
--- /dev/null
+++ b/arch/powerpc/include/asm/disassemble.h
@@ -0,0 +1,80 @@
+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+ *
+ * Copyright IBM Corp. 2008
+ *
+ * Authors: Hollis Blanchard <hollisb@us.ibm.com>
+ */
+
+#ifndef __ASM_PPC_DISASSEMBLE_H__
+#define __ASM_PPC_DISASSEMBLE_H__
+
+#include <linux/types.h>
+
+static inline unsigned int get_op(u32 inst)
+{
+	return inst >> 26;
+}
+
+static inline unsigned int get_xop(u32 inst)
+{
+	return (inst >> 1) & 0x3ff;
+}
+
+static inline unsigned int get_sprn(u32 inst)
+{
+	return ((inst >> 16) & 0x1f) | ((inst >> 6) & 0x3e0);
+}
+
+static inline unsigned int get_dcrn(u32 inst)
+{
+	return ((inst >> 16) & 0x1f) | ((inst >> 6) & 0x3e0);
+}
+
+static inline unsigned int get_rt(u32 inst)
+{
+	return (inst >> 21) & 0x1f;
+}
+
+static inline unsigned int get_rs(u32 inst)
+{
+	return (inst >> 21) & 0x1f;
+}
+
+static inline unsigned int get_ra(u32 inst)
+{
+	return (inst >> 16) & 0x1f;
+}
+
+static inline unsigned int get_rb(u32 inst)
+{
+	return (inst >> 11) & 0x1f;
+}
+
+static inline unsigned int get_rc(u32 inst)
+{
+	return inst & 0x1;
+}
+
+static inline unsigned int get_ws(u32 inst)
+{
+	return (inst >> 11) & 0x1f;
+}
+
+static inline unsigned int get_d(u32 inst)
+{
+	return inst & 0xffff;
+}
+
+#endif /* __ASM_PPC_DISASSEMBLE_H__ */
diff --git a/arch/powerpc/kvm/emulate.c b/arch/powerpc/kvm/emulate.c
--- a/arch/powerpc/kvm/emulate.c
+++ b/arch/powerpc/kvm/emulate.c
@@ -28,62 +28,7 @@
 #include <asm/time.h>
 #include <asm/byteorder.h>
 #include <asm/kvm_ppc.h>
-
-/* Instruction decoding */
-static inline unsigned int get_op(u32 inst)
-{
-	return inst >> 26;
-}
-
-static inline unsigned int get_xop(u32 inst)
-{
-	return (inst >> 1) & 0x3ff;
-}
-
-static inline unsigned int get_sprn(u32 inst)
-{
-	return ((inst >> 16) & 0x1f) | ((inst >> 6) & 0x3e0);
-}
-
-static inline unsigned int get_dcrn(u32 inst)
-{
-	return ((inst >> 16) & 0x1f) | ((inst >> 6) & 0x3e0);
-}
-
-static inline unsigned int get_rt(u32 inst)
-{
-	return (inst >> 21) & 0x1f;
-}
-
-static inline unsigned int get_rs(u32 inst)
-{
-	return (inst >> 21) & 0x1f;
-}
-
-static inline unsigned int get_ra(u32 inst)
-{
-	return (inst >> 16) & 0x1f;
-}
-
-static inline unsigned int get_rb(u32 inst)
-{
-	return (inst >> 11) & 0x1f;
-}
-
-static inline unsigned int get_rc(u32 inst)
-{
-	return inst & 0x1;
-}
-
-static inline unsigned int get_ws(u32 inst)
-{
-	return (inst >> 11) & 0x1f;
-}
-
-static inline unsigned int get_d(u32 inst)
-{
-	return inst & 0xffff;
-}
+#include <asm/disassemble.h>
 
 static void kvmppc_emulate_dec(struct kvm_vcpu *vcpu)
 {


-- 
Hollis Blanchard
IBM Linux Technology Center

Re: [RFC] a little disassembly infrastructure

[Paul Mackerras]

Hollis Blanchard writes:

> Hi, I wrote this patch for KVM [1], but now that I look closer it seems
> like there might be some overlapping functionality.

Looks OK -

Acked-by: Paul Mackerras <paulus@samba.org>

subject to you writing a suitable patch description. :)

Paul.