Skip to contentSkip to conversation

x86: cacheinfo: complete L2/L3 Cache and TLB associativity field definitions

PATCH Superseded Andreas Herrmann 2 participants 7 messages Apr 2, 2009 – Apr 10, 2009 x86

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86: cacheinfo: complete L2/L3 Cache and TLB associativity field definitions

Impact: complete cache information for AMD CPUs

See "CPUID Specification" (AMD Publication #: 25481, Rev. 2.28, April 2008)

---
 arch/x86/kernel/cpu/intel_cacheinfo.c |   16 ++++++++++++----
 1 files changed, 12 insertions(+), 4 deletions(-)

Please apply.

Thanks,

Andreas

arch/x86/kernel/cpu/intel_cacheinfo.c +12 -4
index c471eb1..8f5f084 100644
@@ -207,10 +207,17 @@ union l3_cache {
 };
 
 static const unsigned short __cpuinitconst assocs[] = {
-	[1] = 1, [2] = 2, [4] = 4, [6] = 8,
-	[8] = 16, [0xa] = 32, [0xb] = 48,
+	[1] = 1,
+	[2] = 2,
+	[4] = 4,
+	[6] = 8,
+	[8] = 16,
+	[0xa] = 32,
+	[0xb] = 48,
 	[0xc] = 64,
-	[0xf] = 0xffff // ??
+	[0xd] = 96,
+	[0xe] = 128,
+	[0xf] = 0xffff // fully associative - no way to show this currently
 };
 
 static const unsigned char __cpuinitconst levels[] = { 1, 1, 2, 3 };
@@ -271,7 +278,8 @@ amd_cpuid4(int leaf, union _cpuid4_leaf_eax *eax,
 	eax->split.type = types[leaf];
 	eax->split.level = levels[leaf];
 	if (leaf == 3)
-		eax->split.num_threads_sharing = current_cpu_data.x86_max_cores - 1;
+		eax->split.num_threads_sharing =
+			current_cpu_data.x86_max_cores - 1;
 	else
 		eax->split.num_threads_sharing = 0;
 	eax->split.num_cores_on_die = current_cpu_data.x86_max_cores - 1;
Signature · 1 line
1.6.2
#1 lore raw
Quoted 10 lines · Andreas Herrmann

Impact: complete cache information for AMD CPUs

See "CPUID Specification" (AMD Publication #: 25481, Rev. 2.28, April 2008)

---
 arch/x86/kernel/cpu/intel_cacheinfo.c |   16 ++++++++++++----
 1 files changed, 12 insertions(+), 4 deletions(-)

Please apply.

Hm, the x86/cpu topic is marked broken and blocked by another AMD patch from Mark Langsdorf:

45ca863: x86, cpu: conform L3 Cache Index Disable to Linux standards

We cannot queue up new patches there before this one is fixed. I've attached below my last bugreport about it.

Ingo

----- Forwarded message from Ingo Molnar <[email protected]> -----

Date: Thu, 19 Mar 2009 09:18:37 +0100
From: Ingo Molnar <[email protected]>
To: Mark Langsdorf <[email protected]>
Subject: Re: [PATCH][retry 5] Conform L3 Cache Index Disable to Linux
Standards
Cc: Andrew Morton <[email protected]>, [email protected],
[email protected], [email protected]

Quoted 7 lines · Mark Langsdorf

Add ABI Documentation entry and fix some /sys directory formating issues with the L3 Cache Index Disable feature for future AMD processors. Add a check to disable it for family 0x10 models that do not support it properly. Correct the disable algorithm to reflect erratum 388.

i'm still getting this build failure:

arch/x86/mm/built-in.o:(.bss+0x3640): multiple definition of `k8_northbridges'

i've reported this build failure before. I've excluded this commit from tip:master for now.

Ingo

----- End forwarded message -----

#2 lore raw
Quoted 12 lines · Ingo Molnar
Quoted 9 lines · Andreas Herrmann

Impact: complete cache information for AMD CPUs

See "CPUID Specification" (AMD Publication #: 25481, Rev. 2.28, April 2008)

---
 arch/x86/kernel/cpu/intel_cacheinfo.c |   16 ++++++++++++----
 1 files changed, 12 insertions(+), 4 deletions(-)
Quoted 6 lines
Quoted 1 line

Please apply.

Hm, the x86/cpu topic is marked broken and blocked by another AMD patch from Mark Langsdorf:

45ca863: x86, cpu: conform L3 Cache Index Disable to Linux standards

Oops, wasn't aware that you had applied this patch and I couldn't find it in tip/master. Now I know why.

The code seems broken. Defining k8_northbridges in a header file is the wrong thing to do. I'll look at it.

Andreas

Signature · 5 lines
Operating | Advanced Micro Devices GmbH
  System  | Karl-Hammerschmidt-Str. 34, 85609 Dornach b. München, Germany
 Research | Geschäftsführer: Jochen Polster, Thomas M. McCoy, Giuliano Meroni
  Center  | Sitz: Dornach, Gemeinde Aschheim, Landkreis München
  (OSRC)  | Registergericht München, HRB Nr. 43632
#3 lore raw

PATCH 1/1 x86: cacheinfo: fix build error in cache_disable

In reply to Ingo Molnar

Impact: fix build error

Commit 45ca863 (x86, cpu: conform L3 Cache Index Disable to Linux standard) introduced following build error with CONFIG_K8_NB=n
...
arch/x86/mm/built-in.o:(.bss+0x5c0): multiple definition of `k8_northbridges' arch/x86/kernel/built-in.o:(.bss+0x18c8): first defined here
...

The patch deactivates cache_disable functionality for kernels with CONFIG_K8_NB=n.

---
 arch/x86/include/asm/k8.h             |    4 ----
 arch/x86/kernel/cpu/intel_cacheinfo.c |   19 ++++++++++++++++---
 2 files changed, 16 insertions(+), 7 deletions(-)

The real fix would be to enable cache index disable functionality also for kernels with CONFIG_K8_NB=n but I don't have time to provide such a patch today.

Ingo, please either apply this (compile tested) patch or revert Mark's patch to avoid further blocking of tip/x86/cpu branch.

Thanks,

Andreas

arch/x86/include/asm/k8.h +0 -4
index 0d619c3..54c8cc5 100644
@@ -6,11 +6,7 @@
 extern struct pci_device_id k8_nb_ids[];
 
 extern int early_is_k8_nb(u32 value);
-#ifdef CONFIG_K8_NB
 extern struct pci_dev **k8_northbridges;
-#else
-struct pci_dev **k8_northbridges;
-#endif
 extern int num_k8_northbridges;
 extern int cache_k8_northbridges(void);
 extern void k8_flush_garts(void);
arch/x86/kernel/cpu/intel_cacheinfo.c +16 -3
index 6f4d2a3..63e4468 100644
@@ -640,14 +640,23 @@ static ssize_t show_##file_name						\
 	return sprintf (buf, "%lu\n", (unsigned long)this_leaf->object + val); \
 }
 
+#ifndef CONFIG_K8_NB
+struct pci_dev **k8_northbridges = NULL;
+#endif
+
 static ssize_t show_cache_disable(struct _cpuid4_info *this_leaf, char *buf,
 		 unsigned int index)
 {
 	int cpu = cpumask_first(to_cpumask(this_leaf->shared_cpu_map));
 	int node = cpu_to_node(cpu);
-	struct pci_dev *dev = k8_northbridges[node];
+	struct pci_dev *dev;
 	unsigned int reg = 0;
 
+	if (!k8_northbridges)
+		return -EINVAL;
+
+	dev = k8_northbridges[node];
+
 	if (!this_leaf->can_disable)
 		return -EINVAL;
 
@@ -668,10 +677,15 @@ store_cache_disable(struct _cpuid4_info *this_leaf, const char *buf,
 {
 	int cpu = cpumask_first(to_cpumask(this_leaf->shared_cpu_map));
 	int node = cpu_to_node(cpu);
-	struct pci_dev *dev = k8_northbridges[node];
+	struct pci_dev *dev;
 	unsigned long val = 0;
 	unsigned int scrubber = 0;
 
+	if (!k8_northbridges)
+		return -EINVAL;
+
+	dev = k8_northbridges[node];
+
 	if (!this_leaf->can_disable)
 		return -EINVAL;
 
@@ -786,7 +800,6 @@ static struct _cache_attr cache_disable_0 = __ATTR(cache_disable_0, 0644,
 static struct _cache_attr cache_disable_1 = __ATTR(cache_disable_1, 0644,
 		 show_cache_disable_1, store_cache_disable_1);
 
-
 static struct attribute * default_attrs[] = {
 	&type.attr,
 	&level.attr,
Signature · 1 line
1.6.2
#4 lore raw

Re: [PATCH] [PATCH] x86: cacheinfo: fix build error in cache_disable

Quoted 36 lines · Andreas Herrmann

Impact: fix build error

Commit 45ca863 (x86, cpu: conform L3 Cache Index Disable to Linux standard) introduced following build error with CONFIG_K8_NB=n
...
arch/x86/mm/built-in.o:(.bss+0x5c0): multiple definition of `k8_northbridges' arch/x86/kernel/built-in.o:(.bss+0x18c8): first defined here
...

The patch deactivates cache_disable functionality for kernels with CONFIG_K8_NB=n.

---
 arch/x86/include/asm/k8.h             |    4 ----
 arch/x86/kernel/cpu/intel_cacheinfo.c |   19 ++++++++++++++++---
 2 files changed, 16 insertions(+), 7 deletions(-)

The real fix would be to enable cache index disable functionality also for kernels with CONFIG_K8_NB=n but I don't have time to provide such a patch today.

Ingo, please either apply this (compile tested) patch or revert Mark's patch to avoid further blocking of tip/x86/cpu branch.

Thanks,

Andreas

arch/x86/include/asm/k8.h +0 -0
index 0d619c3..54c8cc5 100644

arch/x86/include/asm/k8.h line 13

Quoted 8 lines
arch/x86/include/asm/k8.h +0 -4
@@ -6,11 +6,7 @@
 extern struct pci_device_id k8_nb_ids[];
 
 extern int early_is_k8_nb(u32 value);
-#ifdef CONFIG_K8_NB
 extern struct pci_dev **k8_northbridges;
-#else
-struct pci_dev **k8_northbridges;
-#endif

hm, what's wrong with the node_to_k8_dev() suggestion i made?

That would be a pretty clean interface, private to k8.c - and it would map a 'return NULL' inline function in the !CONFIG_K8_NB case.

Ingo

#5 lore raw

PATCH v2 1/1 x86: cacheinfo: complete L2/L3 Cache and TLB associativity field definitions

In reply to Ingo Molnar

See "CPUID Specification" (AMD Publication #: 25481, Rev. 2.28, April 2008)

---
 arch/x86/kernel/cpu/intel_cacheinfo.c |   16 ++++++++++++----
 1 files changed, 12 insertions(+), 4 deletions(-)

This patch applies on top of the changes sent with

[PATCH 0/8] x86: cacheinfo: cache_disable fixes/cleanup

Please apply.

Thanks,

Andreas

arch/x86/kernel/cpu/intel_cacheinfo.c +12 -4
index 47e4fb4..3329347 100644
@@ -200,10 +200,17 @@ union l3_cache {
 };
 
 static const unsigned short __cpuinitconst assocs[] = {
-	[1] = 1, [2] = 2, [4] = 4, [6] = 8,
-	[8] = 16, [0xa] = 32, [0xb] = 48,
+	[1] = 1,
+	[2] = 2,
+	[4] = 4,
+	[6] = 8,
+	[8] = 16,
+	[0xa] = 32,
+	[0xb] = 48,
 	[0xc] = 64,
-	[0xf] = 0xffff // ??
+	[0xd] = 96,
+	[0xe] = 128,
+	[0xf] = 0xffff // fully associative - no way to show this currently
 };
 
 static const unsigned char __cpuinitconst levels[] = { 1, 1, 2, 3 };
@@ -264,7 +271,8 @@ amd_cpuid4(int leaf, union _cpuid4_leaf_eax *eax,
 	eax->split.type = types[leaf];
 	eax->split.level = levels[leaf];
 	if (leaf == 3)
-		eax->split.num_threads_sharing = current_cpu_data.x86_max_cores - 1;
+		eax->split.num_threads_sharing =
+			current_cpu_data.x86_max_cores - 1;
 	else
 		eax->split.num_threads_sharing = 0;
 	eax->split.num_cores_on_die = current_cpu_data.x86_max_cores - 1;
Signature · 1 line
1.6.2
#6 lore raw
Commit-ID:  6265ff19ca08df0d96c859ae5e4dc2d9ad07070e
Gitweb:     http://git.kernel.org/tip/6265ff19ca08df0d96c859ae5e4dc2d9ad07070e
Author:     Andreas Herrmann <[email protected]>
AuthorDate: Thu, 9 Apr 2009 15:47:10 +0200
Committer:  Ingo Molnar <[email protected]>
CommitDate: Fri, 10 Apr 2009 15:41:18 +0200

x86: cacheinfo: complete L2/L3 Cache and TLB associativity field definitions

See "CPUID Specification" (AMD Publication #: 25481, Rev. 2.28, April 2008)

LKML-Reference: <[email protected]>

---
 arch/x86/kernel/cpu/intel_cacheinfo.c |   16 ++++++++++++----
 1 files changed, 12 insertions(+), 4 deletions(-)
arch/x86/kernel/cpu/intel_cacheinfo.c +12 -4
index d46a849..789efe2 100644
@@ -200,10 +200,17 @@ union l3_cache {
 };
 
 static const unsigned short __cpuinitconst assocs[] = {
-	[1] = 1, [2] = 2, [4] = 4, [6] = 8,
-	[8] = 16, [0xa] = 32, [0xb] = 48,
+	[1] = 1,
+	[2] = 2,
+	[4] = 4,
+	[6] = 8,
+	[8] = 16,
+	[0xa] = 32,
+	[0xb] = 48,
 	[0xc] = 64,
-	[0xf] = 0xffff // ??
+	[0xd] = 96,
+	[0xe] = 128,
+	[0xf] = 0xffff /* fully associative - no way to show this currently */
 };
 
 static const unsigned char __cpuinitconst levels[] = { 1, 1, 2, 3 };
@@ -264,7 +271,8 @@ amd_cpuid4(int leaf, union _cpuid4_leaf_eax *eax,
 	eax->split.type = types[leaf];
 	eax->split.level = levels[leaf];
 	if (leaf == 3)
-		eax->split.num_threads_sharing = current_cpu_data.x86_max_cores - 1;
+		eax->split.num_threads_sharing =
+			current_cpu_data.x86_max_cores - 1;
 	else
 		eax->split.num_threads_sharing = 0;
 	eax->split.num_cores_on_die = current_cpu_data.x86_max_cores - 1;