Skip to contentSkip to conversation

X86: cpu_debug support for VIA / Centaur CPU's

PATCH In review Harald Welte 6 participants 11 messages Jun 9, 2009 – Jun 10, 2009 x86

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 X86: cpu_debug support for VIA / Centaur CPU's

This patch adds the MSR definitions for the VIA/Centaur CPU models to cpu_debug.c

---
 arch/x86/include/asm/cpu_debug.h |   21 +++++-
 arch/x86/kernel/cpu/cpu_debug.c  |  176 ++++++++++++++++++++++++++++++++++++++
 2 files changed, 196 insertions(+), 1 deletions(-)
arch/x86/include/asm/cpu_debug.h +20 -1
index 2228020..3790657 100644
@@ -134,6 +134,12 @@ enum cpu_processor_bit {
 	CPU_AMD_0F_BIT,
 	CPU_AMD_10_BIT,
 	CPU_AMD_11_BIT,
+/* Centaur */
+	CPU_CENTAUR_C3_BIT,
+	CPU_CENTAUR_C5_BIT,
+	CPU_CENTAUR_C7A_BIT,
+	CPU_CENTAUR_C7D_BIT,
+	CPU_CENTAUR_NANO_BIT,
 };
 
 #define	CPU_INTEL_PENTIUM	(1 << CPU_INTEL_PENTIUM_BIT)
@@ -183,8 +189,21 @@ enum cpu_processor_bit {
 /* Select all supported AMD CPUs */
 #define	CPU_AMD_ALL		(CPU_AMD_K6 | CPU_K7_PLUS)
 
+
+#define CPU_CENT_C3		(1 << CPU_CENTAUR_C3_BIT)
+#define CPU_CENT_C5		(1 << CPU_CENTAUR_C5_BIT)
+#define CPU_CENT_C7A		(1 << CPU_CENTAUR_C7A_BIT)
+#define CPU_CENT_C7D		(1 << CPU_CENTAUR_C7D_BIT)
+#define CPU_CENT_NANO		(1 << CPU_CENTAUR_NANO_BIT)
+
+#define CPU_CENT_C7		(CPU_CENT_C7A | CPU_CENT_C7D)
+#define CPU_CENT_C7_PLUS	(CPU_CENT_C7 | CPU_CENT_NANO)
+
+#define CPU_CENT_ALL		(CPU_CENT_C3 | CPU_CENT_C5 | CPU_CENT_C7 \
+				 | CPU_CENT_NANO)
+
 /* Select all supported CPUs */
-#define	CPU_ALL			(CPU_INTEL_ALL | CPU_AMD_ALL)
+#define	CPU_ALL			(CPU_INTEL_ALL | CPU_AMD_ALL | CPU_CENT_ALL)
 
 #define MAX_CPU_FILES		512
 
arch/x86/kernel/cpu/cpu_debug.c +176 -0

218 lines not on this page · raw message

Signature · 1 line
1.6.3.1
#1 lore raw

Hello Harald,

Quoted 4 lines · Harald Welte

This patch adds the MSR definitions for the VIA/Centaur CPU models to cpu_debug.c

As per Ingo suggestion, we removed model information from cpu_debug and is available in -tip :

http://git.kernel.org/?p=linux/kernel/git/x86/linux-2.6-tip.git;a=commitdiff;h=5095f59bda6793a7b8f0856096d6893fe98e0e51

Please check latest -tip from and see how it works with your chips :
git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-tip.git

Currently I am doing more changes and available in :

http://git.kernel.org/?p=linux/kernel/git/jaswinder/linux-2.6-cpu.git;a=shortlog;h=x86/cpu

Thanks,
--
JSR

#2 lore raw

Thanks for your quick response.

Quoted 8 lines · Jaswinder Singh Rajput
Quoted 4 lines · Harald Welte

This patch adds the MSR definitions for the VIA/Centaur CPU models to cpu_debug.c

As per Ingo suggestion, we removed model information from cpu_debug and is available in -tip :

but what happens if one MSR has a different group from vendor A to vendor B? and what happens if you read MSR's on a CPU that doesn't have them? I suppose you get a segmentation fault, at least that's what I've received in the past when accidentially reading a MSR that doesn't exist. Is that what the _safe_ variant of rdmsr is for?

Quoted 1 line

Please check latest -tip from and see how it works with your chips :

I suppose it works, but it probably lacks all the via/centaur specific MSR's. I'll work on a patch as soon as i find time.

Quoted 5 lines

git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-tip.git

Currently I am doing more changes and available in :

http://git.kernel.org/?p=linux/kernel/git/jaswinder/linux-2.6-cpu.git;a=shortlog;h=x86/cpu

Which means I should base on which of those tree to make sure it can be merged?

Signature · 3 lines
- Harald Welte <[email protected]>	    http://linux.via.com.tw/
============================================================================
VIA Free and Open Source Software Liaison
#3 lore raw
Quoted 8 lines · Harald Welte

Thanks for your quick response.

Quoted 4 lines · Jaswinder Singh Rajput
Quoted 3 lines · Harald Welte

This patch adds the MSR definitions for the VIA/Centaur CPU models to cpu_debug.c

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

As per Ingo suggestion, we removed model information from cpu_debug and is available in -tip :

but what happens if one MSR has a different group from vendor A to vendor B? [...]

MSRs should really be enumerated along CPU features. They will be accessed if a CPU offers that CPU feature.

Quoted 5 lines

[...] and what happens if you read MSR's on a CPU that doesn't have them? I suppose you get a segmentation fault, at least that's what I've received in the past when accidentially reading a MSR that doesn't exist. Is that what the _safe_ variant of rdmsr is for?

Yeah, the safe read should never fault - there should be all zeroes or an error return.

Ingo

#4 lore raw
Quoted 4 lines · Ingo Molnar

MSRs should really be enumerated along CPU features. They will be accessed if a CPU offers that CPU feature.

Nice in theory, but so many MSRs have to be enumerated with obscure test combinations, that it really isn't practical in the general case. That is why we have the safe MSR variants.

Quoted 4 lines

Yeah, the safe read should never fault - there should be all zeroes or an error return.

Error return, MSRs #GP if not present. All zero means a present MSR (which is zero.)

-hpa

#5 lore raw
Quoted 11 lines · H. Peter Anvin
Quoted 4 lines · Ingo Molnar

MSRs should really be enumerated along CPU features. They will be accessed if a CPU offers that CPU feature.

Nice in theory, but so many MSRs have to be enumerated with obscure test combinations, that it really isn't practical in the general case. That is why we have the safe MSR variants.

Quoted 1 line
Quoted 6 lines
Quoted 3 lines

Yeah, the safe read should never fault - there should be all zeroes or an error return.

Error return, MSRs #GP if not present. All zero means a present MSR (which is zero.)

yes, of course - i meant the /debug/x86/cpu/* behavior: it should either result zeroes, or should return -EINVAL. (probably the latter)

Ingo

#6 lore raw
Quoted 18 lines · Ingo Molnar
Quoted 15 lines · H. Peter Anvin
Quoted 4 lines · Ingo Molnar

MSRs should really be enumerated along CPU features. They will be accessed if a CPU offers that CPU feature.

Nice in theory, but so many MSRs have to be enumerated with obscure test combinations, that it really isn't practical in the general case. That is why we have the safe MSR variants.

Quoted 4 lines

Yeah, the safe read should never fault - there should be all zeroes or an error return.

Quoted 7 lines
Quoted 2 lines

Error return, MSRs #GP if not present. All zero means a present MSR (which is zero.)

yes, of course - i meant the /debug/x86/cpu/* behavior: it should either result zeroes, or should return -EINVAL. (probably the latter)

Return zeroes - same as hardware case for bits which can't be set. Returning -EINVAL might match a specific bit pattern caller is looking for.

Mike

Quoted 8 lines

Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#7 lore raw
Quoted 22 lines · Michael S. Zick
Quoted 21 lines · Ingo Molnar
Quoted 17 lines · H. Peter Anvin
Quoted 4 lines · Ingo Molnar

MSRs should really be enumerated along CPU features. They will be accessed if a CPU offers that CPU feature.

Nice in theory, but so many MSRs have to be enumerated with obscure test combinations, that it really isn't practical in the general case. That is why we have the safe MSR variants.

Quoted 4 lines

Yeah, the safe read should never fault - there should be all zeroes or an error return.

Error return, MSRs #GP if not present. All zero means a present MSR (which is zero.)

Quoted 8 lines
Quoted 4 lines

yes, of course - i meant the /debug/x86/cpu/* behavior: it should either result zeroes, or should return -EINVAL. (probably the latter)

Return zeroes - same as hardware case for bits which can't be set. Returning -EINVAL might match a specific bit pattern caller is looking for.

these files are accessed via read(). The -EINVAL is the syscall return value. The value (if any) goes into the buffer that is being read into. So there's no way to 'match a specific bit pattern' - it's two separate spaces.

Ingo

#8 lore raw
Quoted 30 lines · Ingo Molnar
Quoted 27 lines · Michael S. Zick
Quoted 25 lines · Ingo Molnar
Quoted 11 lines · H. Peter Anvin
Quoted 4 lines · Ingo Molnar

MSRs should really be enumerated along CPU features. They will be accessed if a CPU offers that CPU feature.

Nice in theory, but so many MSRs have to be enumerated with obscure test combinations, that it really isn't practical in the general case. That is why we have the safe MSR variants.

Quoted 1 line
Quoted 6 lines
Quoted 3 lines

Yeah, the safe read should never fault - there should be all zeroes or an error return.

Error return, MSRs #GP if not present. All zero means a present MSR (which is zero.)

yes, of course - i meant the /debug/x86/cpu/* behavior: it should either result zeroes, or should return -EINVAL. (probably the latter)

Quoted 9 lines
Quoted 3 lines

Return zeroes - same as hardware case for bits which can't be set. Returning -EINVAL might match a specific bit pattern caller is looking for.

these files are accessed via read(). The -EINVAL is the syscall return value. The value (if any) goes into the buffer that is being read into. So there's no way to 'match a specific bit pattern' - it's two separate spaces.

Right. Prior post was just a bad case of nerves. ;)

Mike

Quoted 8 lines

Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#9 lore raw
In reply to Michael S. Zick
Quoted 8 lines · Michael S. Zick
Quoted 4 lines

yes, of course - i meant the /debug/x86/cpu/* behavior: it should either result zeroes, or should return -EINVAL. (probably the latter)

Return zeroes - same as hardware case for bits which can't be set. Returning -EINVAL might match a specific bit pattern caller is looking for.

It should be -EIO anyway, which is what /dev/cpu/*/msr does.

-hpa

#10 lore raw
In reply to Harald Welte
Quoted 10 lines · Harald Welte

Thanks for your quick response.

Quoted 6 lines · Jaswinder Singh Rajput
Quoted 4 lines · Harald Welte

This patch adds the MSR definitions for the VIA/Centaur CPU models to cpu_debug.c

Quoted 9 lines
Quoted 2 lines

As per Ingo suggestion, we removed model information from cpu_debug and is available in -tip :

but what happens if one MSR has a different group from vendor A to vendor B? and what happens if you read MSR's on a CPU that doesn't have them? I suppose you get a segmentation fault, at least that's what I've received in the past when accidentially reading a MSR that doesn't exist. Is that what the _safe_ variant of rdmsr is for?

AFAIK we are using safe variant. That's why I requested you to test latest -tip and let us know your result and feedback.

Quoted 7 lines
Quoted 4 lines

Currently I am doing more changes and available in :

http://git.kernel.org/?p=linux/kernel/git/jaswinder/linux-2.6-cpu.git;a=shortlog;h=x86/cpu

Which means I should base on which of those tree to make sure it can be merged?

-tip will be better option.

Thanks,
--
JSR