Skip to contentSkip to conversation

2/6 i386 serialize-msr

PATCH In review [email protected] 3 participants 5 messages Jul 30, 2005

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 2/6 i386 serialize-msr

i386 arch cleanup. Introduce the serialize macro to serialize processor state. Why the microcode update needs it I am not quite sure, since wrmsr() is already a serializing instruction, but it is a microcode update, so I will keep the semantic the same, since this could be a timing workaround. As far as I can tell, this has always been there since the original microcode update source.

arch/i386/kernel/microcode.c +5 -2
Index: linux-2.6.13/arch/i386/kernel/microcode.c
===================================================================
@@ -164,7 +164,8 @@
 	}
 
 	wrmsr(MSR_IA32_UCODE_REV, 0, 0);
-	__asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx");
+	/* XXX needed? wrmsr should serialize unless a chip bug */
+	serialize(); 
 	/* get the current revision from MSR 0x8B */
 	rdmsr(MSR_IA32_UCODE_REV, val[0], uci->rev);
 	pr_debug("microcode: collect_cpu_info : sig=0x%x, pf=0x%x, rev=0x%x\n",
@@ -377,7 +378,9 @@
 		(unsigned long) uci->mc->bits >> 16 >> 16);
 	wrmsr(MSR_IA32_UCODE_REV, 0, 0);
 
-	__asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx");
+	/* XXX needed? wrmsr should serialize unless a chip bug */
+	serialize(); 
+
 	/* get the current revision from MSR 0x8B */
 	rdmsr(MSR_IA32_UCODE_REV, val[0], val[1]);
 
include/asm-i386/processor.h +5 -0
Index: linux-2.6.13/include/asm-i386/processor.h
===================================================================
@@ -277,6 +277,11 @@
 	outb((data), 0x23); \
 } while (0)
 
+static inline void serialize(void)
+{
+	 __asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx");
+}
+
 static inline void __monitor(const void *eax, unsigned long ecx,
 		unsigned long edx)
 {
#1 lore raw

Hi!

Quoted 6 lines

i386 arch cleanup. Introduce the serialize macro to serialize processor state. Why the microcode update needs it I am not quite sure, since wrmsr() is already a serializing instruction, but it is a microcode update, so I will keep the semantic the same, since this could be a timing workaround. As far as I can tell, this has always been there since the original microcode update source.

Can we get better name, like "serialize_cpu()"?
Pavel

Quoted 16 lines

Index: linux-2.6.13/arch/i386/kernel/microcode.c
===================================================================

arch/i386/kernel/microcode.c +2 -1
@@ -164,7 +164,8 @@
 	}
 
 	wrmsr(MSR_IA32_UCODE_REV, 0, 0);
-	__asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx");
+	/* XXX needed? wrmsr should serialize unless a chip bug */
+	serialize(); 
 	/* get the current revision from MSR 0x8B */
 	rdmsr(MSR_IA32_UCODE_REV, val[0], uci->rev);
 	pr_debug("microcode: collect_cpu_info : sig=0x%x, pf=0x%x, rev=0x%x\n",
Signature · 1 line
teflon -- maybe it is a trademark, but it should not be.
#2 lore raw
Quoted 13 lines · Pavel Machek

Hi!

Quoted 8 lines

i386 arch cleanup. Introduce the serialize macro to serialize processor state. Why the microcode update needs it I am not quite sure, since wrmsr() is already a serializing instruction, but it is a microcode update, so I will keep the semantic the same, since this could be a timing workaround. As far as I can tell, this has always been there since the original microcode update source.

Quoted 4 lines

Can we get better name, like "serialize_cpu()"?
Pavel

No objections to changing the name by me. In fact, looks like serialize was a rather poor choice in the context of the microcode update source, since there are already references to lock based serialization right around the same code. Updated comments as well.

Zach

i386 arch cleanup. Introduce the serialize macro to serialize processor state. Why the microcode update needs it I am not quite sure, since wrmsr() is already a serializing instruction, but it is a microcode update, so I will keep the semantic the same, since this could be a timing workaround. As far as I can tell, this has always been there since the original microcode update source.

arch/i386/kernel/microcode.c +5 -2
Index: linux-2.6.13/arch/i386/kernel/microcode.c
===================================================================
@@ -164,7 +164,8 @@
 	}
 
 	wrmsr(MSR_IA32_UCODE_REV, 0, 0);
-	__asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx");
+	/* see 1.07.  Apprent chip bug */
+	serialize_cpu(); 
 	/* get the current revision from MSR 0x8B */
 	rdmsr(MSR_IA32_UCODE_REV, val[0], uci->rev);
 	pr_debug("microcode: collect_cpu_info : sig=0x%x, pf=0x%x, rev=0x%x\n",
@@ -377,7 +378,9 @@
 		(unsigned long) uci->mc->bits >> 16 >> 16);
 	wrmsr(MSR_IA32_UCODE_REV, 0, 0);
 
-	__asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx");
+	/* see 1.07.  Apprent chip bug */
+	serialize_cpu(); 
+
 	/* get the current revision from MSR 0x8B */
 	rdmsr(MSR_IA32_UCODE_REV, val[0], val[1]);
 
include/asm-i386/processor.h +5 -0
Index: linux-2.6.13/include/asm-i386/processor.h
===================================================================
@@ -277,6 +277,11 @@
 	outb((data), 0x23); \
 } while (0)
 
+static inline void serialize_cpu(void)
+{
+	 __asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx");
+}
+
 static inline void __monitor(const void *eax, unsigned long ecx,
 		unsigned long edx)
 {
#3 lore raw

wrmsr(MSR_IA32_UCODE_REV, 0, 0);
- __asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx");
+ /* see 1.07. Apprent chip bug */
+ serialize_cpu();

1.07 in which document? Also, please just spell 'apparent' correctly, saving 1 byte really just looks lazy.

#4 lore raw

'Apparently' I need to make some coffee :)

Quoted 9 lines · Zwane Mwaikambo
	wrmsr(MSR_IA32_UCODE_REV, 0, 0);
(quoted diff) +2 -1
-	__asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx");
+	/* see 1.07.  Apprent chip bug */
+	serialize_cpu(); 

1.07 in which document? Also, please just spell 'apparent' correctly, saving 1 byte really just looks lazy.

i386 arch cleanup. Introduce the serialize macro to serialize processor state. Why the microcode update needs it I am not quite sure, since wrmsr() is already a serializing instruction, but it is a microcode update, so I will keep the semantic the same, since this could be a timing workaround. As far as I can tell, this has always been there since the original microcode update source.

arch/i386/kernel/microcode.c +5 -2
Index: linux-2.6.13/arch/i386/kernel/microcode.c
===================================================================
@@ -164,7 +164,8 @@
 	}
 
 	wrmsr(MSR_IA32_UCODE_REV, 0, 0);
-	__asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx");
+	/* see notes above for revision 1.07.  Apparent chip bug */
+	serialize_cpu(); 
 	/* get the current revision from MSR 0x8B */
 	rdmsr(MSR_IA32_UCODE_REV, val[0], uci->rev);
 	pr_debug("microcode: collect_cpu_info : sig=0x%x, pf=0x%x, rev=0x%x\n",
@@ -377,7 +378,9 @@
 		(unsigned long) uci->mc->bits >> 16 >> 16);
 	wrmsr(MSR_IA32_UCODE_REV, 0, 0);
 
-	__asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx");
+	/* see notes above for revision 1.07.  Apparent chip bug */
+	serialize_cpu(); 
+
 	/* get the current revision from MSR 0x8B */
 	rdmsr(MSR_IA32_UCODE_REV, val[0], val[1]);
 
include/asm-i386/processor.h +5 -0
Index: linux-2.6.13/include/asm-i386/processor.h
===================================================================
@@ -277,6 +277,11 @@
 	outb((data), 0x23); \
 } while (0)
 
+static inline void serialize_cpu(void)
+{
+	 __asm__ __volatile__ ("cpuid" : : : "ax", "bx", "cx", "dx");
+}
+
 static inline void __monitor(const void *eax, unsigned long ecx,
 		unsigned long edx)
 {