Skip to contentSkip to conversation

Allow emulation of mfpvr on ppc64 kernel

PPC64 David Gibson 3 participants 4 messages Mar 10, 2005 – Mar 11, 2005

lore
Share & tools
lore
#0 lore raw

Andrew, please apply.

Allow userspace programs on ppc64 to use the (privileged) mfpvr instruction to determine the processor type. At the moment it emulates the instruction to provide the real PVR value, though it could be made to lie in future if for some reason we wish to restrict what CPU features userspace uses.

If nothing else this means that some existing ppc32 applications will now run on a 64-bit kernel (the 32-bit kernel has long supported this emulation). It will also be necessary for ppc64 perfctr support, where userspace requires finer-grained cpu type information than the kernel in order to correctly program the performance monitor control registers.

arch/ppc64/kernel/traps.c +19 -7
Index: working-2.6/arch/ppc64/kernel/traps.c
===================================================================
@@ -279,6 +279,9 @@
  * fault.  Return zero on success.
  */
 
+#define INST_MFSPR_PVR		0x7c1f42a6
+#define INST_MFSPR_PVR_MASK	0xfc1fffff
+
 #define INST_DCBA		0x7c0005ec
 #define INST_DCBA_MASK		0x7c0007fe
 
@@ -297,6 +300,15 @@
 	if (get_user(instword, (unsigned int __user *)(regs->nip)))
 		return -EFAULT;
 
+	/* Emulate the mfspr rD, PVR. */
+	if ((instword & INST_MFSPR_PVR_MASK) == INST_MFSPR_PVR) {
+		unsigned int rd;
+
+		rd = (instword >> 21) & 0x1f;
+		regs->gpr[rd] = mfspr(SPRN_PVR);
+		return 0;
+	}
+
 	/* Emulating the dcba insn is just a no-op.  */
 	if ((instword & INST_DCBA_MASK) == INST_DCBA) {
 		static int warned;
@@ -390,11 +402,6 @@
 	if (regs->msr & 0x100000) {
 		/* IEEE FP exception */
 		parse_fpe(regs);
-
-	} else if (regs->msr & 0x40000) {
-		/* Privileged instruction */
-		_exception(SIGILL, regs, ILL_PRVOPC, regs->nip);
-
 	} else if (regs->msr & 0x20000) {
 		/* trap exception */
 
@@ -411,7 +418,7 @@
 		_exception(SIGTRAP, regs, TRAP_BRKPT, regs->nip);
 
 	} else {
-		/* Illegal instruction; try to emulate it.  */
+		/* Privileged or illegal instruction; try to emulate it. */
 		switch (emulate_instruction(regs)) {
 		case 0:
 			regs->nip += 4;
@@ -423,7 +430,12 @@
 			break;
 
 		default:
-			_exception(SIGILL, regs, ILL_ILLOPC, regs->nip);
+			if (regs->msr & 0x40000)
+				/* priveleged */
+				_exception(SIGILL, regs, ILL_PRVOPC, regs->nip);
+			else
+				/* illegal */
+				_exception(SIGILL, regs, ILL_ILLOPC, regs->nip);
 			break;
 		}
 	}
Signature · 4 lines
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist.  NOT _the_ _other_ _way_
				| _around_!
http://www.ozlabs.org/people/dgibson
#1 lore raw
Quoted 7 lines · David Gibson

Andrew, please apply.

Allow userspace programs on ppc64 to use the (privileged) mfpvr instruction to determine the processor type. At the moment it emulates the instruction to provide the real PVR value, though it could be made to lie in future if for some reason we wish to restrict what CPU features userspace uses.

Why not putting the required information into the AUX table when executing your ELF programs? I loved this feature in the ix86 arch.

Regards

Ingo Oeser

#2 lore raw
Quoted 9 lines · Ingo Oeser
Quoted 7 lines · David Gibson

Andrew, please apply.

Allow userspace programs on ppc64 to use the (privileged) mfpvr instruction to determine the processor type. At the moment it emulates the instruction to provide the real PVR value, though it could be made to lie in future if for some reason we wish to restrict what CPU features userspace uses.

Quoted 3 lines

Why not putting the required information into the AUX table when executing your ELF programs? I loved this feature in the ix86 arch.

Because this is easy and is the way we already do it on ppc32..?

Signature · 4 lines
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist.  NOT _the_ _other_ _way_
				| _around_!
http://www.ozlabs.org/people/dgibson
#3 lore raw
In reply to Ingo Oeser
Quoted 3 lines · Ingo Oeser

Why not putting the required information into the AUX table when executing your ELF programs? I loved this feature in the ix86 arch.

We do put an AT_HWCAP entry in the aux table, which is a bitmap of features supported by the cpu. But for some applications, such as programming the performance monitor hardware, you need to know the specific CPU model and version, and this is a way to provide that information.

Paul.