Forcibly enable some MISC_ENABLE features on Intel

PATCH Superseded Andy Lutomirski 5 participants 15 messages Aug 3, 2011 – Aug 6, 2011

Subsystems (2)

x86 ACPI

lore

Share & tools

lore

Andy Lutomirski

author

2011-08-03 15:47 UTC #0 lore raw

cover Forcibly enable some MISC_ENABLE features on Intel

Intel allows BIOS or the OS to enable or disable some CPU fueatures via IA32_MISC_ENABLE. I have machines that don't enable fast strings or monitor/mwait in BIOS, so do it on bootup instead.

The Intel SDM volume 3, appendix B.1 says that the OS should not touch the monitor enable bit if SSE3 is not present, which presumably means that the OS may touch that bit if SSE3 is present. In any case, these patches seem to work.

Andy Lutomirski (2):
x86: Enable fast strings on Intel if BIOS hasn't already
x86: Enable monitor/mwait on Intel if BIOS hasn't already

 arch/x86/kernel/cpu/intel.c |   37 +++++++++++++++++++++++++++++++++++--
 1 files changed, 35 insertions(+), 2 deletions(-)

Signature · 1 line

1.7.6

Andy Lutomirski

author

2011-08-03 15:47 UTC #1 lore raw

PATCH 1/2 x86: Enable fast strings on Intel if BIOS hasn't already

I have two Sandy Bridge systems and neither one enables fast strings in BIOS. I can't find anything in the manual that says that only BIOS should enable or disable the feature, so try to enable it at bootup.

Signed-off-by Andy Lutomirski <[email protected]>

---
 arch/x86/kernel/cpu/intel.c |   17 +++++++++++++++--
 1 files changed, 15 insertions(+), 2 deletions(-)

arch/x86/kernel/cpu/intel.c +15 -2

index ed6086e..6b20fef 100644
@@ -30,6 +30,7 @@
 static void __cpuinit early_init_intel(struct cpuinfo_x86 *c)
 {
 	u64 misc_enable;
+	bool allow_fast_string = true;
 
 	/* Unmask CPUID levels if masked: */
 	if (c->x86 > 6 || (c->x86 == 6 && c->x86_model >= 0xd)) {

@@ -118,10 +119,11 @@ static void __cpuinit early_init_intel(struct cpuinfo_x86 *c)
 	 * (model 2) with the same problem.
 	 */
 	if (c->x86 == 15) {
-		rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
+		allow_fast_string = false;
+		printk(KERN_INFO "kmemcheck: Disabling fast string operations\n");
 
+		rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
 		if (misc_enable & MSR_IA32_MISC_ENABLE_FAST_STRING) {
-			printk(KERN_INFO "kmemcheck: Disabling fast string operations\n");
 
 			misc_enable &= ~MSR_IA32_MISC_ENABLE_FAST_STRING;
 			wrmsrl(MSR_IA32_MISC_ENABLE, misc_enable);

@@ -135,6 +137,17 @@ static void __cpuinit early_init_intel(struct cpuinfo_x86 *c)
 	 */
 	if (c->x86 > 6 || (c->x86 == 6 && c->x86_model >= 0xd)) {
 		rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
+
+		if (allow_fast_string &&
+		    !(misc_enable & MSR_IA32_MISC_ENABLE_FAST_STRING)) {
+			misc_enable |= MSR_IA32_MISC_ENABLE_FAST_STRING;
+			wrmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
+			printk(KERN_INFO "Enabled fast string operations\n");
+
+			/* Re-read to make sure it stuck. */
+			rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
+		}
+
 		if (!(misc_enable & MSR_IA32_MISC_ENABLE_FAST_STRING)) {
 			printk(KERN_INFO "Disabled fast string operations\n");
 			setup_clear_cpu_cap(X86_FEATURE_REP_GOOD);

Signature · 1 line

1.7.6

Yu, Fenghua

2011-08-03 16:35 UTC #2 lore raw

RE: [PATCH 1/2] x86: Enable fast strings on Intel if BIOS hasn't already

Quoted 33 lines

-----Original Message-----
From: Andy Lutomirski [mailto:[email protected]]
Sent: Wednesday, August 03, 2011 8:47 AM
To: [email protected]; [email protected]
Cc: Yu, Fenghua; Andy Lutomirski
Subject: [PATCH 1/2] x86: Enable fast strings on Intel if BIOS hasn't already
I have two Sandy Bridge systems and neither one enables fast strings in BIOS. I can't find anything in the manual that says that only BIOS should enable or disable the feature, so try to enable it at bootup.
Signed-off-by Andy Lutomirski <[email protected]>
---
 arch/x86/kernel/cpu/intel.c |   17 +++++++++++++++--
 1 files changed, 15 insertions(+), 2 deletions(-)
arch/x86/kernel/cpu/intel.c +1 -0
index ed6086e..6b20fef 100644
@@ -30,6 +30,7 @@
 static void __cpuinit early_init_intel(struct cpuinfo_x86 *c)
 {
 	u64 misc_enable;
+	bool allow_fast_string = true;
 
 	/* Unmask CPUID levels if masked: */
 	if (c->x86 > 6 || (c->x86 == 6 && c->x86_model >= 0xd)) {

arch/x86/kernel/cpu/intel.c static void __cpuinit early_init_intel(struct · line 125

Quoted 8 lines

arch/x86/kernel/cpu/intel.c +2 -1

@@ -118,10 +119,11 @@ static void __cpuinit early_init_intel(struct
 cpuinfo_x86 *c)
 	 * (model 2) with the same problem.
 	 */
 	if (c->x86 == 15) {
-		rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
+		allow_fast_string = false;
+		printk(KERN_INFO "kmemcheck: Disabling fast string
 operations\n");

Please don't move this printk() here. Keep it in the original place. If fast string bit is 0 already, we don't need to print the info.

arch/x86/kernel/cpu/intel.c static void __cpuinit early_init_intel(struct · line 137

Quoted 10 lines

arch/x86/kernel/cpu/intel.c +1 -1

+		rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
 		if (misc_enable & MSR_IA32_MISC_ENABLE_FAST_STRING) {
-			printk(KERN_INFO "kmemcheck: Disabling fast string
 operations\n");

arch/x86/kernel/cpu/intel.c +0 -0

 			misc_enable &= ~MSR_IA32_MISC_ENABLE_FAST_STRING;
 			wrmsrl(MSR_IA32_MISC_ENABLE, misc_enable);

@@ -135,6 +137,17 @@ static void __cpuinit early_init_intel(struct
 cpuinfo_x86 *c)

You also need to change the comment right before this code. With your patch, the comment is out-of-date.

Quoted 20 lines

arch/x86/kernel/cpu/intel.c +11 -0

 	 */
 	if (c->x86 > 6 || (c->x86 == 6 && c->x86_model >= 0xd)) {
 		rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
+
+		if (allow_fast_string &&
+		    !(misc_enable & MSR_IA32_MISC_ENABLE_FAST_STRING)) {
+			misc_enable |= MSR_IA32_MISC_ENABLE_FAST_STRING;
+			wrmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
+			printk(KERN_INFO "Enabled fast string operations\n");
+
+			/* Re-read to make sure it stuck. */
+			rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
+		}
+
 		if (!(misc_enable & MSR_IA32_MISC_ENABLE_FAST_STRING)) {
 			printk(KERN_INFO "Disabled fast string

operations\n");
setup_clear_cpu_cap(X86_FEATURE_REP_GOOD);
--
1.7.6

Andrew Lutomirski

author

2011-08-03 16:40 UTC #3 lore raw

Re: [PATCH 1/2] x86: Enable fast strings on Intel if BIOS hasn't already

Quoted 30 lines · Fenghua

Quoted 30 lines
-----Original Message-----
From: Andy Lutomirski [mailto:[email protected]]
Sent: Wednesday, August 03, 2011 8:47 AM
To: [email protected]; [email protected]
Cc: Yu, Fenghua; Andy Lutomirski
Subject: [PATCH 1/2] x86: Enable fast strings on Intel if BIOS hasn't already
I have two Sandy Bridge systems and neither one enables fast strings in BIOS. I can't find anything in the manual that says that only BIOS should enable or disable the feature, so try to enable it at bootup.
Signed-off-by Andy Lutomirski <[email protected]>
---
arch/x86/kernel/cpu/intel.c | 17 +++++++++++++++--
 1 files changed, 15 insertions(+), 2 deletions(-)
arch/x86/kernel/cpu/intel.c +1 -0
index ed6086e..6b20fef 100644
@@ -30,6 +30,7 @@
  static void __cpuinit early_init_intel(struct cpuinfo_x86 *c)
  {
       u64 misc_enable;
+     bool allow_fast_string = true;
 

Quoted 13 lines

arch/x86/kernel/cpu/intel.c static void __cpuinit early_init_intel(struct · line 125

Quoted 11 lines

arch/x86/kernel/cpu/intel.c +2 -1

       /* Unmask CPUID levels if masked: */
       if (c->x86 > 6 || (c->x86 == 6 && c->x86_model >= 0xd)) {

@@ -118,10 +119,11 @@ static void __cpuinit early_init_intel(struct
 cpuinfo_x86 *c)
        * (model 2) with the same problem.
        */
       if (c->x86 == 15) {
-             rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
+             allow_fast_string = false;
+             printk(KERN_INFO "kmemcheck: Disabling fast string
 operations\n");

Please don't move this printk() here. Keep it in the original place. If fast string bit is 0 already, we don't need to print the info.

I'll rework it.

Quoted 7 lines

arch/x86/kernel/cpu/intel.c +1 -1

+             rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
               if (misc_enable & MSR_IA32_MISC_ENABLE_FAST_STRING) {
-                     printk(KERN_INFO "kmemcheck: Disabling fast string
 operations\n");

Quoted 6 lines

arch/x86/kernel/cpu/intel.c static void __cpuinit early_init_intel(struct · line 137
Quoted 4 lines
misc_enable &= ~MSR_IA32_MISC_ENABLE_FAST_STRING; wrmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
arch/x86/kernel/cpu/intel.c +0 -0
@@ -135,6 +137,17 @@ static void __cpuinit early_init_intel(struct
 cpuinfo_x86 *c)
You also need to change the comment right before this code. With your patch, the comment is out-of-date.

Will do.

--Andy

Andy Lutomirski

author

2011-08-03 15:47 UTC #4 lore raw

PATCH 2/2 x86: Enable monitor/mwait on Intel if BIOS hasn't already

In reply to Andy Lutomirski

I have a Sandy Bridge machine with an Intel BIOS that enables monitor/mwait on all but the boot CPU. With this patch, intel_idle works.

Signed-off-by Andy Lutomirski <[email protected]>

---
 arch/x86/kernel/cpu/intel.c |   20 ++++++++++++++++++++
 1 files changed, 20 insertions(+), 0 deletions(-)

arch/x86/kernel/cpu/intel.c +20 -0

index 6b20fef..a3339a1 100644
@@ -487,6 +487,26 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c)
 			wrmsrl(MSR_IA32_ENERGY_PERF_BIAS, epb);
 		}
 	}
+
+	/* Enable monitor/mwait if BIOS didn't do it for us. */
+	if (!cpu_has(c, X86_FEATURE_MWAIT) && cpu_has(c, X86_FEATURE_XMM3)
+	    && c->x86 >= 6 && !(c->x86 == 6 && c->x86_model < 0x1c)
+	    && !(c->x86 == 0xf && c->x86_model < 3)) {
+		u64 misc_enable;
+		rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
+		misc_enable |= MSR_IA32_MISC_ENABLE_MWAIT;
+
+		/*
+		 * Some non-SSE3 cpus will #GP.  We check for that,
+		 * but it can't hurt to be safe.
+		 */
+		wrmsr_safe(MSR_IA32_MISC_ENABLE, (u32)misc_enable,
+			   (u32)(misc_enable >> 32));
+
+		/* Re-read monitor capability. */
+		if (cpuid_ecx(1) & 0x8)
+			set_cpu_cap(c, X86_FEATURE_MWAIT);
+	}
 }
 
 #ifdef CONFIG_X86_32

Signature · 1 line

1.7.6

Len Brown

reviewer

2011-08-03 16:15 UTC #5 lore raw

Re: [PATCH 2/2] x86: Enable monitor/mwait on Intel if BIOS hasn't already

Quoted 3 lines · Andy Lutomirski

I have a Sandy Bridge machine with an Intel BIOS that enables monitor/mwait on all but the boot CPU. With this patch, intel_idle works.

What machine has this BIOS bug, and are you running the latest version of the BIOS?

thanks,
Len Brown, Intel Open Source Technolgy Center

Quoted 45 lines

Signed-off-by Andy Lutomirski <[email protected]>

---
 arch/x86/kernel/cpu/intel.c |   20 ++++++++++++++++++++
 1 files changed, 20 insertions(+), 0 deletions(-)

arch/x86/kernel/cpu/intel.c +20 -0

index 6b20fef..a3339a1 100644
@@ -487,6 +487,26 @@ static void __cpuinit init_intel(struct cpuinfo_x86 *c)
 			wrmsrl(MSR_IA32_ENERGY_PERF_BIAS, epb);
 		}
 	}
+
+	/* Enable monitor/mwait if BIOS didn't do it for us. */
+	if (!cpu_has(c, X86_FEATURE_MWAIT) && cpu_has(c, X86_FEATURE_XMM3)
+	    && c->x86 >= 6 && !(c->x86 == 6 && c->x86_model < 0x1c)
+	    && !(c->x86 == 0xf && c->x86_model < 3)) {
+		u64 misc_enable;
+		rdmsrl(MSR_IA32_MISC_ENABLE, misc_enable);
+		misc_enable |= MSR_IA32_MISC_ENABLE_MWAIT;
+
+		/*
+		 * Some non-SSE3 cpus will #GP.  We check for that,
+		 * but it can't hurt to be safe.
+		 */
+		wrmsr_safe(MSR_IA32_MISC_ENABLE, (u32)misc_enable,
+			   (u32)(misc_enable >> 32));
+
+		/* Re-read monitor capability. */
+		if (cpuid_ecx(1) & 0x8)
+			set_cpu_cap(c, X86_FEATURE_MWAIT);
+	}
 }
 
 #ifdef CONFIG_X86_32

Signature · 7 lines

1.7.6

--
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/

Andrew Lutomirski

author

2011-08-03 16:22 UTC #6 lore raw

Re: [PATCH 2/2] x86: Enable monitor/mwait on Intel if BIOS hasn't already

Quoted 8 lines · Len Brown

Quoted 3 lines · Andy Lutomirski
I have a Sandy Bridge machine with an Intel BIOS that enables monitor/mwait on all but the boot CPU. With this patch, intel_idle works.
What machine has this BIOS bug, and are you running the latest version of the BIOS?

Intel DQ67SW, version SWQ6710H.86A.0051.2011.0413.1154. This is not quite up-to-date -- I'll test a newer BIOS.

(Enabling monitor/mwait on some but not all CPUs is obnoxious, too. I think it's just dumb luck that Linux notices at all and doesn't crash.)

--Andy
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Andrew Lutomirski

author

2011-08-03 16:42 UTC #7 lore raw

Re: [PATCH 2/2] x86: Enable monitor/mwait on Intel if BIOS hasn't already

Quoted 7 lines · Andrew Lutomirski

Quoted 6 lines · Len Brown
Quoted 3 lines · Andy Lutomirski
I have a Sandy Bridge machine with an Intel BIOS that enables monitor/mwait on all but the boot CPU. With this patch, intel_idle works.

Quoted 5 lines

Quoted 2 lines
What machine has this BIOS bug, and are you running the latest version of the BIOS?
Intel DQ67SW, version SWQ6710H.86A.0051.2011.0413.1154. This is not quite up-to-date -- I'll test a newer BIOS.

SWQ6710H.86A.0053.2011.0615.1535 has the same problem. It fixed ASPM, though :)

Should I redo the patch with a printk to warn about the firmware bug?

--Andy

Quoted 7 lines

(Enabling monitor/mwait on some but not all CPUs is obnoxious, too. I think it's just dumb luck that Linux notices at all and doesn't crash.)
--Andy

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

Len Brown

reviewer

2011-08-03 19:18 UTC #8 lore raw

Re: [PATCH 2/2] x86: Enable monitor/mwait on Intel if BIOS hasn't already

Quoted 9 lines

Quoted 5 lines
Quoted 2 lines
What machine has this BIOS bug, and are you running the latest version of the BIOS?
Intel DQ67SW, version SWQ6710H.86A.0051.2011.0413.1154. This is not quite up-to-date -- I'll test a newer BIOS.
SWQ6710H.86A.0053.2011.0615.1535 has the same problem. It fixed ASPM, though :)
Should I redo the patch with a printk to warn about the firmware bug?

I would hope that a patch that _only_ prints [Firmware bug] should be enough. This isn't the kind of workaround that we should really have to carry in the kernel.

This is still a very new platform, and my guess is that they are actively updating the BIOS as we speak. I'll see if I can figure out who to talk to
in order to make sure of that:-)

thanks,
Len Brown, Intel Open Source Technology Center

Andrew Lutomirski

author

2011-08-04 03:18 UTC #9 lore raw

Re: [PATCH 2/2] x86: Enable monitor/mwait on Intel if BIOS hasn't already

Quoted 10 lines · Len Brown

Quoted 9 lines
Quoted 5 lines
Quoted 2 lines
What machine has this BIOS bug, and are you running the latest version of the BIOS?
Intel DQ67SW, version SWQ6710H.86A.0051.2011.0413.1154. This is not quite up-to-date -- I'll test a newer BIOS.
SWQ6710H.86A.0053.2011.0615.1535 has the same problem. It fixed ASPM, though :)
Should I redo the patch with a printk to warn about the firmware bug?

Quoted 8 lines

I would hope that a patch that _only_ prints [Firmware bug] should be enough. This isn't the kind of workaround that we should really have to carry in the kernel.
This is still a very new platform, and my guess is that they are actively updating the BIOS as we speak. I'll see if I can figure out who to talk to
in order to make sure of that:-)

I can add a strange data point: with TXT enabled in BIOS, fast strings are disabled (on all CPUs, I think) and monitor/mwait is disabled on the boot CPU. With TXT disabled, both features are enabled.

I'll see if my laptop does the same thing.

--Andy
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Matthew Garrett

2011-08-04 19:41 UTC #10 lore raw

Re: [PATCH 2/2] x86: Enable monitor/mwait on Intel if BIOS hasn't already

In reply to Len Brown

Quoted 6 lines · Len Brown

Quoted 6 lines
Quoted 5 lines
Quoted 2 lines
What machine has this BIOS bug, and are you running the latest version of the BIOS?
Intel DQ67SW, version SWQ6710H.86A.0051.2011.0413.1154. This is not quite up-to-date -- I'll test a newer BIOS.

Quoted 7 lines

Quoted 3 lines
SWQ6710H.86A.0053.2011.0615.1535 has the same problem. It fixed ASPM, though :)
Should I redo the patch with a printk to warn about the firmware bug?
I would hope that a patch that _only_ prints [Firmware bug] should be enough. This isn't the kind of workaround that we should really have to carry in the kernel.

We can't trust BIOS authors to get things right. Where we can reliably fix up things that are broken, we should do so.

Signature · 5 lines

Matthew Garrett | [email protected]
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Len Brown

reviewer

2011-08-04 21:54 UTC #11 lore raw

Re: [PATCH 2/2] x86: Enable monitor/mwait on Intel if BIOS hasn't already

Quoted 2 lines

We can't trust BIOS authors to get things right. Where we can reliably fix up things that are broken, we should do so.

How about if we can get a BIOS fix for this brand new system, then we shall not clutter the kernel with a workaround, otherwise we shall.

thanks,
-Len

Matthew Garrett

2011-08-04 21:56 UTC #12 lore raw

Re: [PATCH 2/2] x86: Enable monitor/mwait on Intel if BIOS hasn't already

Quoted 6 lines · Len Brown

Quoted 2 lines
We can't trust BIOS authors to get things right. Where we can reliably fix up things that are broken, we should do so.
How about if we can get a BIOS fix for this brand new system, then we shall not clutter the kernel with a workaround, otherwise we shall.

It's a pretty good rule of thumb that if a BIOS bug is present in one BIOS it's present in others, but the users will just never notice.

Signature · 1 line

Matthew Garrett | [email protected]

Ingo Molnar

2011-08-05 09:44 UTC #13 lore raw

Re: [PATCH 2/2] x86: Enable monitor/mwait on Intel if BIOS hasn't already

Quoted 11 lines · Matthew Garrett

Quoted 6 lines · Len Brown
Quoted 2 lines
We can't trust BIOS authors to get things right. Where we can reliably fix up things that are broken, we should do so.
How about if we can get a BIOS fix for this brand new system, then we shall not clutter the kernel with a workaround, otherwise we shall.
It's a pretty good rule of thumb that if a BIOS bug is present in one BIOS it's present in others, but the users will just never notice.

Yes. There's no harm really from working around a BIOS bug and warning about it - when the BIOS gets fixed the warning message will go away.

Btw., few people risk upgrading their BIOSen on already purchased hardware. I almost never do it, i only do it if it's impossible to continue with a current BIOS.

Thanks,

Ingo

Andrew Lutomirski

author

2011-08-06 11:42 UTC #14 lore raw

Re: [PATCH 2/2] x86: Enable monitor/mwait on Intel if BIOS hasn't already

Quoted 15 lines · Ingo Molnar

Quoted 11 lines · Matthew Garrett
Quoted 6 lines · Len Brown
Quoted 2 lines
We can't trust BIOS authors to get things right. Where we can reliably fix up things that are broken, we should do so.
How about if we can get a BIOS fix for this brand new system, then we shall not clutter the kernel with a workaround, otherwise we shall.
It's a pretty good rule of thumb that if a BIOS bug is present in one BIOS it's present in others, but the users will just never notice.

Quoted 7 lines

Yes. There's no harm really from working around a BIOS bug and warning about it - when the BIOS gets fixed the warning message will go away.
Btw., few people risk upgrading their BIOSen on already purchased hardware. I almost never do it, i only do it if it's impossible to continue with a current BIOS.

I do it, but I understand the sentiment. I've destroyed laptops with BIOS upgrades, and the new Intel BIOS seems to insist on spewing garbage into my keyboard input when Linux (but not grub) starts.

Updates patches coming.

Quoted 5 lines

Thanks,
Ingo

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