Skip to contentSkip to conversation

Add support for Penryn mobile CPUs

PATCH In review Rudolf Marek 2 participants 4 messages Jul 6, 2009 – Sep 20, 2009

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Add support for Penryn mobile CPUs

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

Following patch adds support for mobile Penryn CPUs. Intel documents this poorly. I asked the Coretemp author for some help. This is totally untested and may not work. Please test!

Rudolf
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkpSKLgACgkQ3J9wPJqZRNV86gCfbEsfdUfaEEaaS5gsoosTxYDr gWcAoImIpffHIUxRtCJ13GCfMr5jPkly
=2to8
-----END PGP SIGNATURE-----

drivers/hwmon/coretemp.c +18 -4
Index: linux-2.6.30.1/drivers/hwmon/coretemp.c
===================================================================
@@ -157,6 +157,7 @@
 	/* The 100C is default for both mobile and non mobile CPUs */
 
 	int tjmax = 100000;
+	int tjmax_ee = 85000;
 	int usemsr_ee = 1;
 	int err;
 	u32 eax, edx;
@@ -175,20 +176,31 @@
 	}
 
 	if ((c->x86_model > 0xe) && (usemsr_ee)) {
-
+		u8 platform_id;
 		/* Now we can detect the mobile CPU using Intel provided table
 		   http://softwarecommunity.intel.com/Wiki/Mobility/720.htm
 		   For Core2 cores, check MSR 0x17, bit 28 1 = Mobile CPU
 		*/
 
 		err = rdmsr_safe_on_cpu(id, 0x17, &eax, &edx);
+		/* Platform ID bits 52:50 (EDX starts at bit 32) */
+		platform_id = (edx >> 18) & 0x7;
 		if (err) {
 			dev_warn(dev,
 				 "Unable to access MSR 0x17, assuming desktop"
 				 " CPU\n");
 			usemsr_ee = 0;
-		} else if (!(eax & 0x10000000)) {
+		/* trust bit 28 up to Penryn, I could not find any documentation on that
+		   if you happen to know someone at Intel please ask */
+
+		} else if ((!(eax & 0x10000000)) && (c->x86_model < 0x17)) {
 			usemsr_ee = 0;
+		/* Mobile Penryn CPU seems to be platform ID 7 or 5 (guesswork) */
+		} else if ((c->x86_model == 0x17) &&
+				((platform_id == 5) || (platform_id == 7))) {
+			/* If MSR EE bit is set, set it to 90C otherwise 105C */
+			tjmax_ee = 90000;
+			tjmax = 105000;
 		}
 	}
 
@@ -200,7 +212,7 @@
 				 "Unable to access MSR 0xEE, for Tjmax, left"
 				 " at default");
 		} else if (eax & 0x40000000) {
-			tjmax = 85000;
+			tjmax = tjmax_ee;
 		}
 	} else {
 		dev_warn(dev, "Using relative temperature scale!\n");
@@ -420,7 +432,9 @@
 	for_each_online_cpu(i) {
 		struct cpuinfo_x86 *c = &cpu_data(i);
 
-		/* check if family 6, models 0xe, 0xf, 0x16, 0x17, 0x1A, 0x1c */
+		/* check if family 6, models 0xe (Pentium M DC),
+		  0xf (Core 2 DC 65nm), 0x16 (Core 2 SC 65nm), 0x17 (Penryn 45nm)
+		  0x1a (Nehalem), 0x1c (Atom) */
 		if ((c->cpuid_level < 0) || (c->x86 != 0x6) ||
 		    !((c->x86_model == 0xe) || (c->x86_model == 0xf) ||
 			(c->x86_model == 0x16) || (c->x86_model == 0x17) ||
Documentation/hwmon/coretemp +3 -1
Index: linux-2.6.30.1/Documentation/hwmon/coretemp
===================================================================
@@ -4,7 +4,9 @@
 Supported chips:
   * All Intel Core family
     Prefix: 'coretemp'
-    CPUID: family 0x6, models 0xe, 0xf, 0x16, 0x17, 0x1c (Atom)
+    CPUID: family 0x6, models 0xe (Pentium M DC), 0xf (Core 2 DC 65nm),
+                              0x16 (Core 2 SC 65nm), 0x17 (Penryn 45nm),
+                              0x1a (Nehalem), 0x1c (Atom).
     Datasheet: Intel 64 and IA-32 Architectures Software Developer's Manual
                Volume 3A: System Programming Guide
                http://softwarecommunity.intel.com/Wiki/Mobility/720.htm

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#1 lore raw

Hi Ruik,

Quoted 22 lines · Rudolf Marek

Following patch adds support for mobile Penryn CPUs. Intel documents this poorly. I asked the Coretemp author for some help. This is totally untested and may not work. Please test!

Index: linux-2.6.30.1/drivers/hwmon/coretemp.c
=================================> --- linux-2.6.30.1.orig/drivers/hwmon/coretemp.c 2009-07-06 18:25:46.934758067 +0200
+++ linux-2.6.30.1/drivers/hwmon/coretemp.c 2009-07-06 18:35:38.602757417 +0200

drivers/hwmon/coretemp.c +1 -1
@@ -157,6 +157,7 @@
 	/* The 100C is default for both mobile and non mobile CPUs */
 
 	int tjmax = 100000;
+	int tjmax_ee = 85000;
 	int usemsr_ee = 1;
 	int err;
 	u32 eax, edx;
@@ -175,20 +176,31 @@
 	}
 
 	if ((c->x86_model > 0xe) && (usemsr_ee)) {
-

drivers/hwmon/coretemp.c line 188

Quoted 10 lines
drivers/hwmon/coretemp.c +3 -0
+		u8 platform_id;
 		/* Now we can detect the mobile CPU using Intel provided table
 		   http://softwarecommunity.intel.com/Wiki/Mobility/720.htm
 		   For Core2 cores, check MSR 0x17, bit 28 1 = Mobile CPU
 		*/
 
 		err = rdmsr_safe_on_cpu(id, 0x17, &eax, &edx);
+		/* Platform ID bits 52:50 (EDX starts at bit 32) */
+		platform_id = (edx >> 18) & 0x7;
 		if (err) {

Logic is invalid: if there's an error then the value of edx is undefined.

drivers/hwmon/coretemp.c line 195

Quoted 8 lines
drivers/hwmon/coretemp.c +3 -1
 			dev_warn(dev,
 				 "Unable to access MSR 0x17, assuming desktop"
 				 " CPU\n");
 			usemsr_ee = 0;
-		} else if (!(eax & 0x10000000)) {
+		/* trust bit 28 up to Penryn, I could not find any documentation on that
+		   if you happen to know someone at Intel please ask */
+

Indentation is pretty messy, and first comment line is too long.

drivers/hwmon/coretemp.c line 196

Quoted 1 line
drivers/hwmon/coretemp.c +1 -0
+		} else if ((!(eax & 0x10000000)) && (c->x86_model < 0x17)) {

It would make more sense to check model first and eax second IMHO.

Review line 198

Quoted 2 lines
(quoted diff) +1 -0
 			usemsr_ee = 0;
+		/* Mobile Penryn CPU seems to be platform ID 7 or 5 (guesswork) */

Line too long.

Quoted 18 lines
drivers/hwmon/coretemp.c +6 -1
+		} else if ((c->x86_model = 0x17) &&
+				((platform_id = 5) || (platform_id = 7))) {
+			/* If MSR EE bit is set, set it to 90C otherwise 105C */
+			tjmax_ee = 90000;
+			tjmax = 105000;
 		}
 	}
 
@@ -200,7 +212,7 @@
 				 "Unable to access MSR 0xEE, for Tjmax, left"
 				 " at default");
 		} else if (eax & 0x40000000) {
-			tjmax = 85000;
+			tjmax = tjmax_ee;
 		}
 	} else {
 		dev_warn(dev, "Using relative temperature scale!\n");

drivers/hwmon/coretemp.c line 436

Quoted 6 lines
drivers/hwmon/coretemp.c +2 -1
@@ -420,7 +432,9 @@
 	for_each_online_cpu(i) {
 		struct cpuinfo_x86 *c = &cpu_data(i);
 
-		/* check if family 6, models 0xe, 0xf, 0x16, 0x17, 0x1A, 0x1c */
+		/* check if family 6, models 0xe (Pentium M DC),
+		  0xf (Core 2 DC 65nm), 0x16 (Core 2 SC 65nm), 0x17 (Penryn 45nm)

Line too long.

Quoted 18 lines
drivers/hwmon/coretemp.c +1 -0
+		  0x1a (Nehalem), 0x1c (Atom) */
 		if ((c->cpuid_level < 0) || (c->x86 != 0x6) ||
 		    !((c->x86_model = 0xe) || (c->x86_model = 0xf) ||
 			(c->x86_model = 0x16) || (c->x86_model = 0x17) ||

Index: linux-2.6.30.1/Documentation/hwmon/coretemp
=================================> --- linux-2.6.30.1.orig/Documentation/hwmon/coretemp 2009-07-06 18:35:53.765759446 +0200
+++ linux-2.6.30.1/Documentation/hwmon/coretemp 2009-07-06 18:36:39.817757139 +0200

Documentation/hwmon/coretemp +3 -1
@@ -4,7 +4,9 @@
 Supported chips:
   * All Intel Core family
     Prefix: 'coretemp'
-    CPUID: family 0x6, models 0xe, 0xf, 0x16, 0x17, 0x1c (Atom)
+    CPUID: family 0x6, models 0xe (Pentium M DC), 0xf (Core 2 DC 65nm),
+                              0x16 (Core 2 SC 65nm), 0x17 (Penryn 45nm),
+                              0x1a (Nehalem), 0x1c (Atom).
     Datasheet: Intel 64 and IA-32 Architectures Software Developer's Manual
                Volume 3A: System Programming Guide
                http://softwarecommunity.intel.com/Wiki/Mobility/720.htm
Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
#2 lore raw
In reply to Rudolf Marek

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 2 lines

Logic is invalid: if there's an error then the value of edx is undefined.

Yes I know, but there was no place I could create this variable. Perhaps I need to re-think it a bit. Maybe on sunday evening :/

Anyway thanks for the review! I will try to fix it somehow.

Rudolf
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkpVBCUACgkQ3J9wPJqZRNXSIQCfYZdeRqvNDsbbU/navL/P332x BuoAn2aQNMCTsx0Vors75IlnZPZqtK1J
=7GW2
-----END PGP SIGNATURE-----

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors

#3 lore raw
In reply to Rudolf Marek

Hi Rudolf,

Quoted 11 lines · Rudolf Marek

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 2 lines

Logic is invalid: if there's an error then the value of edx is undefined.

Yes I know, but there was no place I could create this variable. Perhaps I need to re-think it a bit. Maybe on sunday evening :/

Anyway thanks for the review! I will try to fix it somehow.

Any news about this? I'd like to push this to Linus, it's been delayed way too long already.

Signature · 6 lines
Jean Delvare

_______________________________________________
lm-sensors mailing list
[email protected]
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors