Skip to contentSkip to conversation

Support generic disabling of all XSAVE features

PATCH Superseded Andi Kleen 1 participant 7 messages Jun 7, 2017 – Jun 13, 2017 x86

v2 of 9 Superseded Latest is v9 →

lore
Share & tools
lore

Andi Kleen

author
#0 lore raw

For performance testing and debugging it can be useful to disable XSAVE features individually. This patchkit hooks up XSAVE with the generic clearcpuid=... option, so that disabling a CPUID feature automatically disables the respective XSAVE feature.

It also cleans up CPUID dependency management. Currently it's possible to generate configurations with cleacpuid that crash.

It replaces an earlier patchkit that did this with special case options.

v1:
Initial post
v2:
Work around broken lguest by exporting set_cpu_cap Repost with cover letter

author
#1 lore raw

PATCH v2 1/5 x86/xsave: Move xsave initialization to after parsing early parameters

From: Andi Kleen <[email protected]>

Move the XSAVE initialization code to be after parsing early parameters. I don't see any reason why the FPU code needs to be initialized that early, nothing else in the initialization phase uses XSAVE. This is useful to be able to handle command line parameters in the XSAVE initialization code.

---
 arch/x86/kernel/cpu/common.c | 1 -
 arch/x86/kernel/setup.c      | 3 +++
 2 files changed, 3 insertions(+), 1 deletion(-)
arch/x86/kernel/cpu/common.c +0 -1
index c8b39870f33e..60b7aa6330ed 100644
@@ -885,7 +885,6 @@ static void __init early_identify_cpu(struct cpuinfo_x86 *c)
 	}
 
 	setup_force_cpu_cap(X86_FEATURE_ALWAYS);
-	fpu__init_system(c);
 }
 
 void __init early_cpu_init(void)
arch/x86/kernel/setup.c +3 -0
index f81823695014..77db31315af3 100644
@@ -91,6 +91,7 @@
 #include <asm/processor.h>
 #include <asm/bugs.h>
 #include <asm/kasan.h>
+#include <asm/fpu/internal.h>
 
 #include <asm/vsyscall.h>
 #include <asm/cpu.h>
@@ -982,6 +983,8 @@ void __init setup_arch(char **cmdline_p)
 
 	parse_early_param();
 
+	fpu__init_system(&boot_cpu_data);
+
 #ifdef CONFIG_MEMORY_HOTPLUG
 	/*
 	 * Memory used by the kernel cannot be hot-removed because Linux
Signature · 1 line
2.9.4
author
#2 lore raw

PATCH v2 2/5 x86/cpuid: Add generic table for cpuid dependencies

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

Some CPUID features depend on other features. Currently it's possible to to clear dependent features, but not clear the base features, which can cause various interesting problems.

This patch implements a generic table to describe dependencies between CPUID features, to be used by all code that clears CPUID.

Some subsystems (like XSAVE) had an own implementation of this, but it's better to do it all in a single place for everyone.

Then clear_cpu_cap and setup_clear_cpu_cap always look up this table and clear all dependencies too.

This is intended to be a practical table: only for features that make sense to clear. If someone for example clears FPU, or other features that are essentially part of the required base feature set, not much is going to work. Handling that is right now out of scope. We're only handling features which can be usefully cleared.

v2: Add EXPORT_SYMBOL for clear_cpu_id for lguest

---
 arch/x86/include/asm/cpufeature.h  |  8 +++-
 arch/x86/include/asm/cpufeatures.h |  5 +++
 arch/x86/kernel/cpu/Makefile       |  1 +
 arch/x86/kernel/cpu/cpuid-deps.c   | 92 ++++++++++++++++++++++++++++++++++++++
 4 files changed, 104 insertions(+), 2 deletions(-)
 create mode 100644 arch/x86/kernel/cpu/cpuid-deps.c
arch/x86/include/asm/cpufeature.h +6 -2
index d59c15c3defd..e6145f383ff8 100644
@@ -125,8 +125,12 @@ extern const char * const x86_bug_flags[NBUGINTS*32];
 #define boot_cpu_has(bit)	cpu_has(&boot_cpu_data, bit)
 
 #define set_cpu_cap(c, bit)	set_bit(bit, (unsigned long *)((c)->x86_capability))
-#define clear_cpu_cap(c, bit)	clear_bit(bit, (unsigned long *)((c)->x86_capability))
-#define setup_clear_cpu_cap(bit) do { \
+#define __clear_cpu_cap(c, bit)	clear_bit(bit, (unsigned long *)((c)->x86_capability))
+
+extern void setup_clear_cpu_cap(int bit);
+extern void clear_cpu_cap(struct cpuinfo_x86 *cpu, int bit);
+
+#define __setup_clear_cpu_cap(bit) do { \
 	clear_cpu_cap(&boot_cpu_data, bit);	\
 	set_bit(bit, (unsigned long *)cpu_caps_cleared); \
 } while (0)
arch/x86/include/asm/cpufeatures.h +5 -0
index 2701e5f8145b..8f371a5966e7 100644
@@ -21,6 +21,11 @@
  * this feature bit is not displayed in /proc/cpuinfo at all.
  */
 
+/*
+ * When adding new features here that depend on other features,
+ * please update the table in kernel/cpu/cpuid-deps.c
+ */
+
 /* Intel-defined CPU features, CPUID level 0x00000001 (edx), word 0 */
 #define X86_FEATURE_FPU		( 0*32+ 0) /* Onboard FPU */
 #define X86_FEATURE_VME		( 0*32+ 1) /* Virtual Mode Extensions */
arch/x86/kernel/cpu/Makefile +1 -0
index 52000010c62e..274fc0fee1e1 100644
@@ -21,6 +21,7 @@ obj-y			+= common.o
 obj-y			+= rdrand.o
 obj-y			+= match.o
 obj-y			+= bugs.o
+obj-y			+= cpuid-deps.o
 
 obj-$(CONFIG_PROC_FS)	+= proc.o
 obj-$(CONFIG_X86_FEATURE_NAMES) += capflags.o powerflags.o
arch/x86/kernel/cpu/cpuid-deps.c new file +92 -0

92 lines not on this page · raw message

Signature · 1 line
2.9.4
author
#3 lore raw

PATCH v2 3/5 x86/cpuid: Make clearcpuid an early param

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

Make clearcpuid= an early param, to make sure it is parsed before the XSAVE initialization. This allows to modify XSAVE state by clearing specific CPUID bits.

---
 arch/x86/kernel/cpu/common.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/kernel/cpu/common.c +1 -1
index 60b7aa6330ed..600e0cf714b9 100644
@@ -1285,7 +1285,7 @@ static __init int setup_disablecpuid(char *arg)
 
 	return 1;
 }
-__setup("clearcpuid=", setup_disablecpuid);
+early_param("clearcpuid", setup_disablecpuid);
 
 #ifdef CONFIG_X86_64
 struct desc_ptr idt_descr __ro_after_init = {
Signature · 1 line
2.9.4
author
#4 lore raw

PATCH v2 4/5 x86/xsave: Make XSAVE check the base CPUID features before enabling

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

Before enabling XSAVE, not only check the XSAVE specific CPUID bits, but also the base CPUID features of the respective XSAVE feature. This allows to disable individual XSAVE states using the existing clearcpuid= option, which can be useful for performance testing and debugging, and also in general avoids inconsistencies.

---
 arch/x86/kernel/fpu/xstate.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
arch/x86/kernel/fpu/xstate.c +22 -0
index c24ac1efb12d..8f0b96320cc9 100644
@@ -15,6 +15,7 @@
 #include <asm/fpu/xstate.h>
 
 #include <asm/tlbflush.h>
+#include <asm/cpufeature.h>
 
 /*
  * Although we spell it out in here, the Processor Trace
@@ -36,6 +37,19 @@ static const char *xfeature_names[] =
 	"unknown xstate feature"	,
 };
 
+static short xsave_cpuid_features[] = {
+	X86_FEATURE_FPU,
+	X86_FEATURE_XMM,
+	X86_FEATURE_AVX,
+	X86_FEATURE_MPX,
+	X86_FEATURE_MPX,
+	X86_FEATURE_AVX512F,
+	X86_FEATURE_AVX512F,
+	X86_FEATURE_AVX512F,
+	X86_FEATURE_INTEL_PT,
+	X86_FEATURE_PKU,
+};
+
 /*
  * Mask of xstate features supported by the CPU and the kernel:
  */
@@ -702,6 +716,7 @@ void __init fpu__init_system_xstate(void)
 	unsigned int eax, ebx, ecx, edx;
 	static int on_boot_cpu __initdata = 1;
 	int err;
+	int i;
 
 	WARN_ON_FPU(!on_boot_cpu);
 	on_boot_cpu = 0;
@@ -735,6 +750,13 @@ void __init fpu__init_system_xstate(void)
 		goto out_disable;
 	}
 
+	/*
+	 * Clear XSAVE features that are disabled in the normal CPUID.
+	 */
+	for (i = 0; i < ARRAY_SIZE(xsave_cpuid_features); i++)
+		if (!boot_cpu_has(xsave_cpuid_features[i]))
+			xfeatures_mask &= ~BIT(i);
+
 	xfeatures_mask &= fpu__get_supported_xfeatures_mask();
 
 	/* Enable xstate instructions to be able to continue with initialization: */
Signature · 1 line
2.9.4
author
#5 lore raw

PATCH v2 5/5 x86/xsave: Using generic CPUID clearing when disabling XSAVE

In reply to Andi Kleen

From: Andi Kleen <[email protected]>

Now that the generic CPUID clearing understands dependencies, it's enough to clear the XSAVE CPUID bit to clear all depending features when XSAVE gets disabled.

So we don't need this hard to maintain explicit list of features depending on XSAVE anymore. Just call the generic clear_cpu_cap() function for XSAVE.

---
 arch/x86/kernel/fpu/xstate.c | 20 --------------------
 1 file changed, 20 deletions(-)
arch/x86/kernel/fpu/xstate.c +0 -20
index 8f0b96320cc9..73c4fda76d32 100644
@@ -73,26 +73,6 @@ unsigned int fpu_user_xstate_size;
 void fpu__xstate_clear_all_cpu_caps(void)
 {
 	setup_clear_cpu_cap(X86_FEATURE_XSAVE);
-	setup_clear_cpu_cap(X86_FEATURE_XSAVEOPT);
-	setup_clear_cpu_cap(X86_FEATURE_XSAVEC);
-	setup_clear_cpu_cap(X86_FEATURE_XSAVES);
-	setup_clear_cpu_cap(X86_FEATURE_AVX);
-	setup_clear_cpu_cap(X86_FEATURE_AVX2);
-	setup_clear_cpu_cap(X86_FEATURE_AVX512F);
-	setup_clear_cpu_cap(X86_FEATURE_AVX512IFMA);
-	setup_clear_cpu_cap(X86_FEATURE_AVX512PF);
-	setup_clear_cpu_cap(X86_FEATURE_AVX512ER);
-	setup_clear_cpu_cap(X86_FEATURE_AVX512CD);
-	setup_clear_cpu_cap(X86_FEATURE_AVX512DQ);
-	setup_clear_cpu_cap(X86_FEATURE_AVX512BW);
-	setup_clear_cpu_cap(X86_FEATURE_AVX512VL);
-	setup_clear_cpu_cap(X86_FEATURE_MPX);
-	setup_clear_cpu_cap(X86_FEATURE_XGETBV1);
-	setup_clear_cpu_cap(X86_FEATURE_AVX512VBMI);
-	setup_clear_cpu_cap(X86_FEATURE_PKU);
-	setup_clear_cpu_cap(X86_FEATURE_AVX512_4VNNIW);
-	setup_clear_cpu_cap(X86_FEATURE_AVX512_4FMAPS);
-	setup_clear_cpu_cap(X86_FEATURE_AVX512_VPOPCNTDQ);
 }
 
 /*
Signature · 1 line
2.9.4
author
#6 lore raw
In reply to Andi Kleen

Andi Kleen <[email protected]> writes:

Any comments on this patchkit? If there are no objections please merge.

-Andi

Quoted 16 lines

For performance testing and debugging it can be useful to disable XSAVE features individually. This patchkit hooks up XSAVE with the generic clearcpuid=... option, so that disabling a CPUID feature automatically disables the respective XSAVE feature.

It also cleans up CPUID dependency management. Currently it's possible to generate configurations with cleacpuid that crash.

It replaces an earlier patchkit that did this with special case options.

v1:
Initial post
v2:
Work around broken lguest by exporting set_cpu_cap Repost with cover letter