[PATCH v13 0/8] x86/arch_prctl Add ARCH_[GET|SET]_CPUID for controlling the CPUID instruction

[PATCH v13 0/8] x86/arch_prctl Add ARCH_[GET|SET]_CPUID for controlling the CPUID instruction

[Kyle Huey]

rr (http://rr-project.org/), a userspace record-and-replay reverse-
execution debugger, would like to trap and emulate the CPUID instruction.
This would allow us to a) mask away certain hardware features that rr does
not support (e.g. RDRAND) and b) enable trace portability across machines
by providing constant results.

Newer Intel CPUs (Ivy Bridge and later) can fault when CPUID is executed at
CPL > 0. Expose this capability to userspace as a new pair of arch_prctls,
ARCH_GET_CPUID and ARCH_SET_CPUID.

Since v12:
Patch 4: x86/syscalls/32: Wire up arch_prctl on x86-32
- compat_sys_arch_prctl prototype has argument names.

Patch 5: x86/cpufeature: Detect CPUID faulting support
- s/init_intel_misc_features_enables/init_intel_misc_features/.
- Added Reviewed-by Borislav.

Patch 6: x86/arch_prctl: Add ARCH_[GET|SET]_CPUID
- Selftest was split out into Patch 7.
- s/msr_misc_features_enables_shadow/msr_misc_features_shadow/.
- ARCH_GET_CPUID now returns its value directly, not in an outparam.
- ARCH_SET_CPUID now takes a simple 0/1 value, ARCH_CPUID_ENABLE and
  ARCH_CPUID_SIGSEGV are gone.
- ARCH_[GET|SET]_CPUID now operate on the enabledness of the CPUID
  instruction, not the CPUID faulting bit. Thus the default return
  value of ARCH_GET_CPUID is 1, and calling ARCH_SET_CPUID with 0
  activates CPUID faulting (and thus disabling the CPUID instruction).

Patch 7: x86/arch_prctl: Selftest for ARCH_[GET|SET]_CPUID
- The self test was updated for the API changes.

[PATCH v13 0/8] x86/arch_prctl Add ARCH_[GET|SET]_CPUID for controlling the CPUID instruction

[Kyle Huey]

rr (http://rr-project.org/), a userspace record-and-replay reverse-
execution debugger, would like to trap and emulate the CPUID instruction.
This would allow us to a) mask away certain hardware features that rr does
not support (e.g. RDRAND) and b) enable trace portability across machines
by providing constant results.

Newer Intel CPUs (Ivy Bridge and later) can fault when CPUID is executed at
CPL > 0. Expose this capability to userspace as a new pair of arch_prctls,
ARCH_GET_CPUID and ARCH_SET_CPUID.

Since v12:
Patch 4: x86/syscalls/32: Wire up arch_prctl on x86-32
- compat_sys_arch_prctl prototype has argument names.

Patch 5: x86/cpufeature: Detect CPUID faulting support
- s/init_intel_misc_features_enables/init_intel_misc_features/.
- Added Reviewed-by Borislav.

Patch 6: x86/arch_prctl: Add ARCH_[GET|SET]_CPUID
- Selftest was split out into Patch 7.
- s/msr_misc_features_enables_shadow/msr_misc_features_shadow/.
- ARCH_GET_CPUID now returns its value directly, not in an outparam.
- ARCH_SET_CPUID now takes a simple 0/1 value, ARCH_CPUID_ENABLE and
  ARCH_CPUID_SIGSEGV are gone.
- ARCH_[GET|SET]_CPUID now operate on the enabledness of the CPUID
  instruction, not the CPUID faulting bit. Thus the default return
  value of ARCH_GET_CPUID is 1, and calling ARCH_SET_CPUID with 0
  activates CPUID faulting (and thus disabling the CPUID instruction).

Patch 7: x86/arch_prctl: Selftest for ARCH_[GET|SET]_CPUID
- The self test was updated for the API changes.

[PATCH v13 1/8] x86/arch_prctl/64: Use SYSCALL_DEFINE2 to define sys_arch_prctl

[Kyle Huey]

Use the SYSCALL_DEFINE2 macro instead of manually defining it.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
---
 arch/x86/kernel/process_64.c | 3 ++-
 arch/x86/um/syscalls_64.c    | 3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index b3760b3..2718cf9 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -30,16 +30,17 @@
 #include <linux/ptrace.h>
 #include <linux/notifier.h>
 #include <linux/kprobes.h>
 #include <linux/kdebug.h>
 #include <linux/prctl.h>
 #include <linux/uaccess.h>
 #include <linux/io.h>
 #include <linux/ftrace.h>
+#include <linux/syscalls.h>
 
 #include <asm/pgtable.h>
 #include <asm/processor.h>
 #include <asm/fpu/internal.h>
 #include <asm/mmu_context.h>
 #include <asm/prctl.h>
 #include <asm/desc.h>
 #include <asm/proto.h>
@@ -607,17 +608,17 @@ long do_arch_prctl(struct task_struct *task, int code, unsigned long addr)
 	default:
 		ret = -EINVAL;
 		break;
 	}
 
 	return ret;
 }
 
-long sys_arch_prctl(int code, unsigned long addr)
+SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, addr)
 {
 	return do_arch_prctl(current, code, addr);
 }
 
 unsigned long KSTK_ESP(struct task_struct *task)
 {
 	return task_pt_regs(task)->sp;
 }
diff --git a/arch/x86/um/syscalls_64.c b/arch/x86/um/syscalls_64.c
index e655227..ab3f7f4 100644
--- a/arch/x86/um/syscalls_64.c
+++ b/arch/x86/um/syscalls_64.c
@@ -1,16 +1,17 @@
 /*
  * Copyright (C) 2003 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
  * Copyright 2003 PathScale, Inc.
  *
  * Licensed under the GPL
  */
 
 #include <linux/sched.h>
+#include <linux/syscalls.h>
 #include <linux/uaccess.h>
 #include <asm/prctl.h> /* XXX This should get the constants from libc */
 #include <os.h>
 
 long arch_prctl(struct task_struct *task, int code, unsigned long __user *addr)
 {
 	unsigned long *ptr = addr, tmp;
 	long ret;
@@ -67,17 +68,17 @@ long arch_prctl(struct task_struct *task, int code, unsigned long __user *addr)
 	case ARCH_GET_GS:
 		ret = put_user(tmp, addr);
 		break;
 	}
 
 	return ret;
 }
 
-long sys_arch_prctl(int code, unsigned long addr)
+SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, addr)
 {
 	return arch_prctl(current, code, (unsigned long __user *) addr);
 }
 
 void arch_switch_to(struct task_struct *to)
 {
 	if ((to->thread.arch.fs == 0) || (to->mm == NULL))
 		return;

base-commit: e5517c2a5a49ed5e99047008629f1cd60246ea0e
-- 
2.10.2

[PATCH v13 1/8] x86/arch_prctl/64: Use SYSCALL_DEFINE2 to define sys_arch_prctl

[Kyle Huey]

Use the SYSCALL_DEFINE2 macro instead of manually defining it.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
---
 arch/x86/kernel/process_64.c | 3 ++-
 arch/x86/um/syscalls_64.c    | 3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index b3760b3..2718cf9 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -30,16 +30,17 @@
 #include <linux/ptrace.h>
 #include <linux/notifier.h>
 #include <linux/kprobes.h>
 #include <linux/kdebug.h>
 #include <linux/prctl.h>
 #include <linux/uaccess.h>
 #include <linux/io.h>
 #include <linux/ftrace.h>
+#include <linux/syscalls.h>
 
 #include <asm/pgtable.h>
 #include <asm/processor.h>
 #include <asm/fpu/internal.h>
 #include <asm/mmu_context.h>
 #include <asm/prctl.h>
 #include <asm/desc.h>
 #include <asm/proto.h>
@@ -607,17 +608,17 @@ long do_arch_prctl(struct task_struct *task, int code, unsigned long addr)
 	default:
 		ret = -EINVAL;
 		break;
 	}
 
 	return ret;
 }
 
-long sys_arch_prctl(int code, unsigned long addr)
+SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, addr)
 {
 	return do_arch_prctl(current, code, addr);
 }
 
 unsigned long KSTK_ESP(struct task_struct *task)
 {
 	return task_pt_regs(task)->sp;
 }
diff --git a/arch/x86/um/syscalls_64.c b/arch/x86/um/syscalls_64.c
index e655227..ab3f7f4 100644
--- a/arch/x86/um/syscalls_64.c
+++ b/arch/x86/um/syscalls_64.c
@@ -1,16 +1,17 @@
 /*
  * Copyright (C) 2003 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
  * Copyright 2003 PathScale, Inc.
  *
  * Licensed under the GPL
  */
 
 #include <linux/sched.h>
+#include <linux/syscalls.h>
 #include <linux/uaccess.h>
 #include <asm/prctl.h> /* XXX This should get the constants from libc */
 #include <os.h>
 
 long arch_prctl(struct task_struct *task, int code, unsigned long __user *addr)
 {
 	unsigned long *ptr = addr, tmp;
 	long ret;
@@ -67,17 +68,17 @@ long arch_prctl(struct task_struct *task, int code, unsigned long __user *addr)
 	case ARCH_GET_GS:
 		ret = put_user(tmp, addr);
 		break;
 	}
 
 	return ret;
 }
 
-long sys_arch_prctl(int code, unsigned long addr)
+SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, addr)
 {
 	return arch_prctl(current, code, (unsigned long __user *) addr);
 }
 
 void arch_switch_to(struct task_struct *to)
 {
 	if ((to->thread.arch.fs == 0) || (to->mm == NULL))
 		return;

base-commit: e5517c2a5a49ed5e99047008629f1cd60246ea0e
-- 
2.10.2

[PATCH v13 2/8] x86/arch_prctl/64: Rename do_arch_prctl to do_arch_prctl_64

[Kyle Huey]

In order to introduce new arch_prctls that are not 64 bit only, rename the
existing 64 bit implementation to do_arch_prctl_64(). Also rename the second
argument to arch_prctl(), which will no longer always be an address.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
Reviewed-by: Andy Lutomirski <luto@kernel.org>
---
 arch/um/include/shared/os.h  |  2 +-
 arch/x86/include/asm/proto.h |  3 +--
 arch/x86/kernel/process_64.c | 32 +++++++++++++++++---------------
 arch/x86/kernel/ptrace.c     |  8 ++++----
 arch/x86/um/os-Linux/prctl.c |  4 ++--
 arch/x86/um/syscalls_64.c    | 12 ++++++------
 6 files changed, 31 insertions(+), 30 deletions(-)

diff --git a/arch/um/include/shared/os.h b/arch/um/include/shared/os.h
index de5d572..2b47e0e 100644
--- a/arch/um/include/shared/os.h
+++ b/arch/um/include/shared/os.h
@@ -298,17 +298,17 @@ extern void os_set_ioignore(void);
 
 /* sigio.c */
 extern int add_sigio_fd(int fd);
 extern int ignore_sigio_fd(int fd);
 extern void maybe_sigio_broken(int fd, int read);
 extern void sigio_broken(int fd, int read);
 
 /* sys-x86_64/prctl.c */
-extern int os_arch_prctl(int pid, int code, unsigned long *addr);
+extern int os_arch_prctl(int pid, int code, unsigned long *arg2);
 
 /* tty.c */
 extern int get_pty(void);
 
 /* sys-$ARCH/task_size.c */
 extern unsigned long os_get_top_address(void);
 
 long syscall(long number, ...);
diff --git a/arch/x86/include/asm/proto.h b/arch/x86/include/asm/proto.h
index 9b9b30b..f8e9194 100644
--- a/arch/x86/include/asm/proto.h
+++ b/arch/x86/include/asm/proto.h
@@ -4,16 +4,17 @@
 #include <asm/ldt.h>
 
 /* misc architecture specific prototypes */
 
 void syscall_init(void);
 
 #ifdef CONFIG_X86_64
 void entry_SYSCALL_64(void);
+long do_arch_prctl_64(struct task_struct *task, int code, unsigned long arg2);
 #endif
 
 #ifdef CONFIG_X86_32
 void entry_INT80_32(void);
 void entry_SYSENTER_32(void);
 void __begin_SYSENTER_singlestep_region(void);
 void __end_SYSENTER_singlestep_region(void);
 #endif
@@ -25,11 +26,9 @@ void entry_SYSCALL_compat(void);
 void entry_INT80_compat(void);
 #endif
 
 void x86_configure_nx(void);
 void x86_report_nx(void);
 
 extern int reboot_force;
 
-long do_arch_prctl(struct task_struct *task, int code, unsigned long addr);
-
 #endif /* _ASM_X86_PROTO_H */
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 2718cf9..611df20 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -193,17 +193,17 @@ int copy_thread_tls(unsigned long clone_flags, unsigned long sp,
 	 */
 	if (clone_flags & CLONE_SETTLS) {
 #ifdef CONFIG_IA32_EMULATION
 		if (in_ia32_syscall())
 			err = do_set_thread_area(p, -1,
 				(struct user_desc __user *)tls, 0);
 		else
 #endif
-			err = do_arch_prctl(p, ARCH_SET_FS, tls);
+			err = do_arch_prctl_64(p, ARCH_SET_FS, tls);
 		if (err)
 			goto out;
 	}
 	err = 0;
 out:
 	if (err && p->thread.io_bitmap_ptr) {
 		kfree(p->thread.io_bitmap_ptr);
 		p->thread.io_bitmap_max = 0;
@@ -534,91 +534,93 @@ static long prctl_map_vdso(const struct vdso_image *image, unsigned long addr)
 	ret = map_vdso_once(image, addr);
 	if (ret)
 		return ret;
 
 	return (long)image->size;
 }
 #endif
 
-long do_arch_prctl(struct task_struct *task, int code, unsigned long addr)
+long do_arch_prctl_64(struct task_struct *task, int code, unsigned long arg2)
 {
 	int ret = 0;
 	int doit = task == current;
 	int cpu;
 
 	switch (code) {
 	case ARCH_SET_GS:
-		if (addr >= TASK_SIZE_MAX)
+		if (arg2 >= TASK_SIZE_MAX)
 			return -EPERM;
 		cpu = get_cpu();
 		task->thread.gsindex = 0;
-		task->thread.gsbase = addr;
+		task->thread.gsbase = arg2;
 		if (doit) {
 			load_gs_index(0);
-			ret = wrmsrl_safe(MSR_KERNEL_GS_BASE, addr);
+			ret = wrmsrl_safe(MSR_KERNEL_GS_BASE, arg2);
 		}
 		put_cpu();
 		break;
 	case ARCH_SET_FS:
 		/* Not strictly needed for fs, but do it for symmetry
 		   with gs */
-		if (addr >= TASK_SIZE_MAX)
+		if (arg2 >= TASK_SIZE_MAX)
 			return -EPERM;
 		cpu = get_cpu();
 		task->thread.fsindex = 0;
-		task->thread.fsbase = addr;
+		task->thread.fsbase = arg2;
 		if (doit) {
 			/* set the selector to 0 to not confuse __switch_to */
 			loadsegment(fs, 0);
-			ret = wrmsrl_safe(MSR_FS_BASE, addr);
+			ret = wrmsrl_safe(MSR_FS_BASE, arg2);
 		}
 		put_cpu();
 		break;
 	case ARCH_GET_FS: {
 		unsigned long base;
+
 		if (doit)
 			rdmsrl(MSR_FS_BASE, base);
 		else
 			base = task->thread.fsbase;
-		ret = put_user(base, (unsigned long __user *)addr);
+		ret = put_user(base, (unsigned long __user *)arg2);
 		break;
 	}
 	case ARCH_GET_GS: {
 		unsigned long base;
+
 		if (doit)
 			rdmsrl(MSR_KERNEL_GS_BASE, base);
 		else
 			base = task->thread.gsbase;
-		ret = put_user(base, (unsigned long __user *)addr);
+		ret = put_user(base, (unsigned long __user *)arg2);
 		break;
 	}
 
 #ifdef CONFIG_CHECKPOINT_RESTORE
 # ifdef CONFIG_X86_X32_ABI
 	case ARCH_MAP_VDSO_X32:
-		return prctl_map_vdso(&vdso_image_x32, addr);
+		return prctl_map_vdso(&vdso_image_x32, arg2);
 # endif
 # if defined CONFIG_X86_32 || defined CONFIG_IA32_EMULATION
 	case ARCH_MAP_VDSO_32:
-		return prctl_map_vdso(&vdso_image_32, addr);
+		return prctl_map_vdso(&vdso_image_32, arg2);
 # endif
 	case ARCH_MAP_VDSO_64:
-		return prctl_map_vdso(&vdso_image_64, addr);
+		return prctl_map_vdso(&vdso_image_64, arg2);
 #endif
 
 	default:
 		ret = -EINVAL;
 		break;
 	}
 
 	return ret;
 }
 
-SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, addr)
+SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
 {
-	return do_arch_prctl(current, code, addr);
+	return do_arch_prctl_64(current, code, arg2);
 }
 
 unsigned long KSTK_ESP(struct task_struct *task)
 {
 	return task_pt_regs(task)->sp;
 }
diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c
index 0e63c02..5004302 100644
--- a/arch/x86/kernel/ptrace.c
+++ b/arch/x86/kernel/ptrace.c
@@ -390,31 +390,31 @@ static int putreg(struct task_struct *child,
 	case offsetof(struct user_regs_struct, flags):
 		return set_flags(child, value);
 
 #ifdef CONFIG_X86_64
 	case offsetof(struct user_regs_struct,fs_base):
 		if (value >= TASK_SIZE_MAX)
 			return -EIO;
 		/*
-		 * When changing the segment base, use do_arch_prctl
+		 * When changing the segment base, use do_arch_prctl_64
 		 * to set either thread.fs or thread.fsindex and the
 		 * corresponding GDT slot.
 		 */
 		if (child->thread.fsbase != value)
-			return do_arch_prctl(child, ARCH_SET_FS, value);
+			return do_arch_prctl_64(child, ARCH_SET_FS, value);
 		return 0;
 	case offsetof(struct user_regs_struct,gs_base):
 		/*
 		 * Exactly the same here as the %fs handling above.
 		 */
 		if (value >= TASK_SIZE_MAX)
 			return -EIO;
 		if (child->thread.gsbase != value)
-			return do_arch_prctl(child, ARCH_SET_GS, value);
+			return do_arch_prctl_64(child, ARCH_SET_GS, value);
 		return 0;
 #endif
 	}
 
 	*pt_regs_access(task_pt_regs(child), offset) = value;
 	return 0;
 }
 
@@ -863,17 +863,17 @@ long arch_ptrace(struct task_struct *child, long request,
 		break;
 #endif
 
 #ifdef CONFIG_X86_64
 		/* normal 64bit interface to access TLS data.
 		   Works just like arch_prctl, except that the arguments
 		   are reversed. */
 	case PTRACE_ARCH_PRCTL:
-		ret = do_arch_prctl(child, data, addr);
+		ret = do_arch_prctl_64(child, data, addr);
 		break;
 #endif
 
 	default:
 		ret = ptrace_request(child, request, addr, data);
 		break;
 	}
 
diff --git a/arch/x86/um/os-Linux/prctl.c b/arch/x86/um/os-Linux/prctl.c
index 96eb2bd..efc9d74 100644
--- a/arch/x86/um/os-Linux/prctl.c
+++ b/arch/x86/um/os-Linux/prctl.c
@@ -1,12 +1,12 @@
 /*
  * Copyright (C) 2007 Jeff Dike (jdike@{addtoit.com,linux.intel.com})
  * Licensed under the GPL
  */
 
 #include <sys/ptrace.h>
 #include <asm/ptrace.h>
 
-int os_arch_prctl(int pid, int code, unsigned long *addr)
+int os_arch_prctl(int pid, int code, unsigned long *arg2)
 {
-        return ptrace(PTRACE_ARCH_PRCTL, pid, (unsigned long) addr, code);
+	return ptrace(PTRACE_ARCH_PRCTL, pid, (unsigned long) arg2, code);
 }
diff --git a/arch/x86/um/syscalls_64.c b/arch/x86/um/syscalls_64.c
index ab3f7f4..d472c6b 100644
--- a/arch/x86/um/syscalls_64.c
+++ b/arch/x86/um/syscalls_64.c
@@ -6,19 +6,19 @@
  */
 
 #include <linux/sched.h>
 #include <linux/syscalls.h>
 #include <linux/uaccess.h>
 #include <asm/prctl.h> /* XXX This should get the constants from libc */
 #include <os.h>
 
-long arch_prctl(struct task_struct *task, int code, unsigned long __user *addr)
+long arch_prctl(struct task_struct *task, int code, unsigned long __user *arg2)
 {
-	unsigned long *ptr = addr, tmp;
+	unsigned long *ptr = arg2, tmp;
 	long ret;
 	int pid = task->mm->context.id.u.pid;
 
 	/*
 	 * With ARCH_SET_FS (and ARCH_SET_GS is treated similarly to
 	 * be safe), we need to call arch_prctl on the host because
 	 * setting %fs may result in something else happening (like a
 	 * GDT or thread.fs being set instead).  So, we let the host
@@ -58,29 +58,29 @@ long arch_prctl(struct task_struct *task, int code, unsigned long __user *addr)
 	case ARCH_SET_FS:
 		current->thread.arch.fs = (unsigned long) ptr;
 		ret = save_registers(pid, &current->thread.regs.regs);
 		break;
 	case ARCH_SET_GS:
 		ret = save_registers(pid, &current->thread.regs.regs);
 		break;
 	case ARCH_GET_FS:
-		ret = put_user(tmp, addr);
+		ret = put_user(tmp, arg2);
 		break;
 	case ARCH_GET_GS:
-		ret = put_user(tmp, addr);
+		ret = put_user(tmp, arg2);
 		break;
 	}
 
 	return ret;
 }
 
-SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, addr)
+SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
 {
-	return arch_prctl(current, code, (unsigned long __user *) addr);
+	return arch_prctl(current, code, (unsigned long __user *) arg2);
 }
 
 void arch_switch_to(struct task_struct *to)
 {
 	if ((to->thread.arch.fs == 0) || (to->mm == NULL))
 		return;
 
 	arch_prctl(to, ARCH_SET_FS, (void __user *) to->thread.arch.fs);
-- 
2.10.2

[PATCH v13 2/8] x86/arch_prctl/64: Rename do_arch_prctl to do_arch_prctl_64

[Kyle Huey]

In order to introduce new arch_prctls that are not 64 bit only, rename the
existing 64 bit implementation to do_arch_prctl_64(). Also rename the second
argument to arch_prctl(), which will no longer always be an address.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
Reviewed-by: Andy Lutomirski <luto@kernel.org>
---
 arch/um/include/shared/os.h  |  2 +-
 arch/x86/include/asm/proto.h |  3 +--
 arch/x86/kernel/process_64.c | 32 +++++++++++++++++---------------
 arch/x86/kernel/ptrace.c     |  8 ++++----
 arch/x86/um/os-Linux/prctl.c |  4 ++--
 arch/x86/um/syscalls_64.c    | 12 ++++++------
 6 files changed, 31 insertions(+), 30 deletions(-)

diff --git a/arch/um/include/shared/os.h b/arch/um/include/shared/os.h
index de5d572..2b47e0e 100644
--- a/arch/um/include/shared/os.h
+++ b/arch/um/include/shared/os.h
@@ -298,17 +298,17 @@ extern void os_set_ioignore(void);
 
 /* sigio.c */
 extern int add_sigio_fd(int fd);
 extern int ignore_sigio_fd(int fd);
 extern void maybe_sigio_broken(int fd, int read);
 extern void sigio_broken(int fd, int read);
 
 /* sys-x86_64/prctl.c */
-extern int os_arch_prctl(int pid, int code, unsigned long *addr);
+extern int os_arch_prctl(int pid, int code, unsigned long *arg2);
 
 /* tty.c */
 extern int get_pty(void);
 
 /* sys-$ARCH/task_size.c */
 extern unsigned long os_get_top_address(void);
 
 long syscall(long number, ...);
diff --git a/arch/x86/include/asm/proto.h b/arch/x86/include/asm/proto.h
index 9b9b30b..f8e9194 100644
--- a/arch/x86/include/asm/proto.h
+++ b/arch/x86/include/asm/proto.h
@@ -4,16 +4,17 @@
 #include <asm/ldt.h>
 
 /* misc architecture specific prototypes */
 
 void syscall_init(void);
 
 #ifdef CONFIG_X86_64
 void entry_SYSCALL_64(void);
+long do_arch_prctl_64(struct task_struct *task, int code, unsigned long arg2);
 #endif
 
 #ifdef CONFIG_X86_32
 void entry_INT80_32(void);
 void entry_SYSENTER_32(void);
 void __begin_SYSENTER_singlestep_region(void);
 void __end_SYSENTER_singlestep_region(void);
 #endif
@@ -25,11 +26,9 @@ void entry_SYSCALL_compat(void);
 void entry_INT80_compat(void);
 #endif
 
 void x86_configure_nx(void);
 void x86_report_nx(void);
 
 extern int reboot_force;
 
-long do_arch_prctl(struct task_struct *task, int code, unsigned long addr);
-
 #endif /* _ASM_X86_PROTO_H */
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 2718cf9..611df20 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -193,17 +193,17 @@ int copy_thread_tls(unsigned long clone_flags, unsigned long sp,
 	 */
 	if (clone_flags & CLONE_SETTLS) {
 #ifdef CONFIG_IA32_EMULATION
 		if (in_ia32_syscall())
 			err = do_set_thread_area(p, -1,
 				(struct user_desc __user *)tls, 0);
 		else
 #endif
-			err = do_arch_prctl(p, ARCH_SET_FS, tls);
+			err = do_arch_prctl_64(p, ARCH_SET_FS, tls);
 		if (err)
 			goto out;
 	}
 	err = 0;
 out:
 	if (err && p->thread.io_bitmap_ptr) {
 		kfree(p->thread.io_bitmap_ptr);
 		p->thread.io_bitmap_max = 0;
@@ -534,91 +534,93 @@ static long prctl_map_vdso(const struct vdso_image *image, unsigned long addr)
 	ret = map_vdso_once(image, addr);
 	if (ret)
 		return ret;
 
 	return (long)image->size;
 }
 #endif
 
-long do_arch_prctl(struct task_struct *task, int code, unsigned long addr)
+long do_arch_prctl_64(struct task_struct *task, int code, unsigned long arg2)
 {
 	int ret = 0;
 	int doit = task == current;
 	int cpu;
 
 	switch (code) {
 	case ARCH_SET_GS:
-		if (addr >= TASK_SIZE_MAX)
+		if (arg2 >= TASK_SIZE_MAX)
 			return -EPERM;
 		cpu = get_cpu();
 		task->thread.gsindex = 0;
-		task->thread.gsbase = addr;
+		task->thread.gsbase = arg2;
 		if (doit) {
 			load_gs_index(0);
-			ret = wrmsrl_safe(MSR_KERNEL_GS_BASE, addr);
+			ret = wrmsrl_safe(MSR_KERNEL_GS_BASE, arg2);
 		}
 		put_cpu();
 		break;
 	case ARCH_SET_FS:
 		/* Not strictly needed for fs, but do it for symmetry
 		   with gs */
-		if (addr >= TASK_SIZE_MAX)
+		if (arg2 >= TASK_SIZE_MAX)
 			return -EPERM;
 		cpu = get_cpu();
 		task->thread.fsindex = 0;
-		task->thread.fsbase = addr;
+		task->thread.fsbase = arg2;
 		if (doit) {
 			/* set the selector to 0 to not confuse __switch_to */
 			loadsegment(fs, 0);
-			ret = wrmsrl_safe(MSR_FS_BASE, addr);
+			ret = wrmsrl_safe(MSR_FS_BASE, arg2);
 		}
 		put_cpu();
 		break;
 	case ARCH_GET_FS: {
 		unsigned long base;
+
 		if (doit)
 			rdmsrl(MSR_FS_BASE, base);
 		else
 			base = task->thread.fsbase;
-		ret = put_user(base, (unsigned long __user *)addr);
+		ret = put_user(base, (unsigned long __user *)arg2);
 		break;
 	}
 	case ARCH_GET_GS: {
 		unsigned long base;
+
 		if (doit)
 			rdmsrl(MSR_KERNEL_GS_BASE, base);
 		else
 			base = task->thread.gsbase;
-		ret = put_user(base, (unsigned long __user *)addr);
+		ret = put_user(base, (unsigned long __user *)arg2);
 		break;
 	}
 
 #ifdef CONFIG_CHECKPOINT_RESTORE
 # ifdef CONFIG_X86_X32_ABI
 	case ARCH_MAP_VDSO_X32:
-		return prctl_map_vdso(&vdso_image_x32, addr);
+		return prctl_map_vdso(&vdso_image_x32, arg2);
 # endif
 # if defined CONFIG_X86_32 || defined CONFIG_IA32_EMULATION
 	case ARCH_MAP_VDSO_32:
-		return prctl_map_vdso(&vdso_image_32, addr);
+		return prctl_map_vdso(&vdso_image_32, arg2);
 # endif
 	case ARCH_MAP_VDSO_64:
-		return prctl_map_vdso(&vdso_image_64, addr);
+		return prctl_map_vdso(&vdso_image_64, arg2);
 #endif
 
 	default:
 		ret = -EINVAL;
 		break;
 	}
 
 	return ret;
 }
 
-SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, addr)
+SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
 {
-	return do_arch_prctl(current, code, addr);
+	return do_arch_prctl_64(current, code, arg2);
 }
 
 unsigned long KSTK_ESP(struct task_struct *task)
 {
 	return task_pt_regs(task)->sp;
 }
diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c
index 0e63c02..5004302 100644
--- a/arch/x86/kernel/ptrace.c
+++ b/arch/x86/kernel/ptrace.c
@@ -390,31 +390,31 @@ static int putreg(struct task_struct *child,
 	case offsetof(struct user_regs_struct, flags):
 		return set_flags(child, value);
 
 #ifdef CONFIG_X86_64
 	case offsetof(struct user_regs_struct,fs_base):
 		if (value >= TASK_SIZE_MAX)
 			return -EIO;
 		/*
-		 * When changing the segment base, use do_arch_prctl
+		 * When changing the segment base, use do_arch_prctl_64
 		 * to set either thread.fs or thread.fsindex and the
 		 * corresponding GDT slot.
 		 */
 		if (child->thread.fsbase != value)
-			return do_arch_prctl(child, ARCH_SET_FS, value);
+			return do_arch_prctl_64(child, ARCH_SET_FS, value);
 		return 0;
 	case offsetof(struct user_regs_struct,gs_base):
 		/*
 		 * Exactly the same here as the %fs handling above.
 		 */
 		if (value >= TASK_SIZE_MAX)
 			return -EIO;
 		if (child->thread.gsbase != value)
-			return do_arch_prctl(child, ARCH_SET_GS, value);
+			return do_arch_prctl_64(child, ARCH_SET_GS, value);
 		return 0;
 #endif
 	}
 
 	*pt_regs_access(task_pt_regs(child), offset) = value;
 	return 0;
 }
 
@@ -863,17 +863,17 @@ long arch_ptrace(struct task_struct *child, long request,
 		break;
 #endif
 
 #ifdef CONFIG_X86_64
 		/* normal 64bit interface to access TLS data.
 		   Works just like arch_prctl, except that the arguments
 		   are reversed. */
 	case PTRACE_ARCH_PRCTL:
-		ret = do_arch_prctl(child, data, addr);
+		ret = do_arch_prctl_64(child, data, addr);
 		break;
 #endif
 
 	default:
 		ret = ptrace_request(child, request, addr, data);
 		break;
 	}
 
diff --git a/arch/x86/um/os-Linux/prctl.c b/arch/x86/um/os-Linux/prctl.c
index 96eb2bd..efc9d74 100644
--- a/arch/x86/um/os-Linux/prctl.c
+++ b/arch/x86/um/os-Linux/prctl.c
@@ -1,12 +1,12 @@
 /*
  * Copyright (C) 2007 Jeff Dike (jdike@{addtoit.com,linux.intel.com})
  * Licensed under the GPL
  */
 
 #include <sys/ptrace.h>
 #include <asm/ptrace.h>
 
-int os_arch_prctl(int pid, int code, unsigned long *addr)
+int os_arch_prctl(int pid, int code, unsigned long *arg2)
 {
-        return ptrace(PTRACE_ARCH_PRCTL, pid, (unsigned long) addr, code);
+	return ptrace(PTRACE_ARCH_PRCTL, pid, (unsigned long) arg2, code);
 }
diff --git a/arch/x86/um/syscalls_64.c b/arch/x86/um/syscalls_64.c
index ab3f7f4..d472c6b 100644
--- a/arch/x86/um/syscalls_64.c
+++ b/arch/x86/um/syscalls_64.c
@@ -6,19 +6,19 @@
  */
 
 #include <linux/sched.h>
 #include <linux/syscalls.h>
 #include <linux/uaccess.h>
 #include <asm/prctl.h> /* XXX This should get the constants from libc */
 #include <os.h>
 
-long arch_prctl(struct task_struct *task, int code, unsigned long __user *addr)
+long arch_prctl(struct task_struct *task, int code, unsigned long __user *arg2)
 {
-	unsigned long *ptr = addr, tmp;
+	unsigned long *ptr = arg2, tmp;
 	long ret;
 	int pid = task->mm->context.id.u.pid;
 
 	/*
 	 * With ARCH_SET_FS (and ARCH_SET_GS is treated similarly to
 	 * be safe), we need to call arch_prctl on the host because
 	 * setting %fs may result in something else happening (like a
 	 * GDT or thread.fs being set instead).  So, we let the host
@@ -58,29 +58,29 @@ long arch_prctl(struct task_struct *task, int code, unsigned long __user *addr)
 	case ARCH_SET_FS:
 		current->thread.arch.fs = (unsigned long) ptr;
 		ret = save_registers(pid, &current->thread.regs.regs);
 		break;
 	case ARCH_SET_GS:
 		ret = save_registers(pid, &current->thread.regs.regs);
 		break;
 	case ARCH_GET_FS:
-		ret = put_user(tmp, addr);
+		ret = put_user(tmp, arg2);
 		break;
 	case ARCH_GET_GS:
-		ret = put_user(tmp, addr);
+		ret = put_user(tmp, arg2);
 		break;
 	}
 
 	return ret;
 }
 
-SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, addr)
+SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
 {
-	return arch_prctl(current, code, (unsigned long __user *) addr);
+	return arch_prctl(current, code, (unsigned long __user *) arg2);
 }
 
 void arch_switch_to(struct task_struct *to)
 {
 	if ((to->thread.arch.fs == 0) || (to->mm == NULL))
 		return;
 
 	arch_prctl(to, ARCH_SET_FS, (void __user *) to->thread.arch.fs);
-- 
2.10.2

[PATCH v13 3/8] x86/arch_prctl: Add do_arch_prctl_common

[Kyle Huey]

Add do_arch_prctl_common() to handle arch_prctls that are not specific to 64
bit mode. Call it from the syscall entry point, but not any of the other
callsites in the kernel, which all want one of the existing 64 bit only
arch_prctls.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
---
 arch/x86/include/asm/proto.h | 3 +++
 arch/x86/kernel/process.c    | 6 ++++++
 arch/x86/kernel/process_64.c | 8 +++++++-
 3 files changed, 16 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/proto.h b/arch/x86/include/asm/proto.h
index f8e9194..99836d9 100644
--- a/arch/x86/include/asm/proto.h
+++ b/arch/x86/include/asm/proto.h
@@ -26,9 +26,12 @@ void entry_SYSCALL_compat(void);
 void entry_INT80_compat(void);
 #endif
 
 void x86_configure_nx(void);
 void x86_report_nx(void);
 
 extern int reboot_force;
 
+long do_arch_prctl_common(struct task_struct *task, int code,
+			  unsigned long cpuid_enabled);
+
 #endif /* _ASM_X86_PROTO_H */
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index 0888a87..5ac20b5 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -579,8 +579,14 @@ unsigned long get_wchan(struct task_struct *p)
 		}
 		fp = READ_ONCE_NOCHECK(*(unsigned long *)fp);
 	} while (count++ < 16 && p->state != TASK_RUNNING);
 
 out:
 	put_task_stack(p);
 	return ret;
 }
+
+long do_arch_prctl_common(struct task_struct *task, int code,
+			  unsigned long cpuid_enabled)
+{
+	return -EINVAL;
+}
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 611df20..bf75d26 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -612,15 +612,21 @@ long do_arch_prctl_64(struct task_struct *task, int code, unsigned long arg2)
 		break;
 	}
 
 	return ret;
 }
 
 SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
 {
-	return do_arch_prctl_64(current, code, arg2);
+	long ret;
+
+	ret = do_arch_prctl_64(current, code, arg2);
+	if (ret == -EINVAL)
+		ret = do_arch_prctl_common(current, code, arg2);
+
+	return ret;
 }
 
 unsigned long KSTK_ESP(struct task_struct *task)
 {
 	return task_pt_regs(task)->sp;
 }
-- 
2.10.2

[PATCH v13 3/8] x86/arch_prctl: Add do_arch_prctl_common

[Kyle Huey]

Add do_arch_prctl_common() to handle arch_prctls that are not specific to 64
bit mode. Call it from the syscall entry point, but not any of the other
callsites in the kernel, which all want one of the existing 64 bit only
arch_prctls.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
---
 arch/x86/include/asm/proto.h | 3 +++
 arch/x86/kernel/process.c    | 6 ++++++
 arch/x86/kernel/process_64.c | 8 +++++++-
 3 files changed, 16 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/proto.h b/arch/x86/include/asm/proto.h
index f8e9194..99836d9 100644
--- a/arch/x86/include/asm/proto.h
+++ b/arch/x86/include/asm/proto.h
@@ -26,9 +26,12 @@ void entry_SYSCALL_compat(void);
 void entry_INT80_compat(void);
 #endif
 
 void x86_configure_nx(void);
 void x86_report_nx(void);
 
 extern int reboot_force;
 
+long do_arch_prctl_common(struct task_struct *task, int code,
+			  unsigned long cpuid_enabled);
+
 #endif /* _ASM_X86_PROTO_H */
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index 0888a87..5ac20b5 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -579,8 +579,14 @@ unsigned long get_wchan(struct task_struct *p)
 		}
 		fp = READ_ONCE_NOCHECK(*(unsigned long *)fp);
 	} while (count++ < 16 && p->state != TASK_RUNNING);
 
 out:
 	put_task_stack(p);
 	return ret;
 }
+
+long do_arch_prctl_common(struct task_struct *task, int code,
+			  unsigned long cpuid_enabled)
+{
+	return -EINVAL;
+}
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 611df20..bf75d26 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -612,15 +612,21 @@ long do_arch_prctl_64(struct task_struct *task, int code, unsigned long arg2)
 		break;
 	}
 
 	return ret;
 }
 
 SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
 {
-	return do_arch_prctl_64(current, code, arg2);
+	long ret;
+
+	ret = do_arch_prctl_64(current, code, arg2);
+	if (ret == -EINVAL)
+		ret = do_arch_prctl_common(current, code, arg2);
+
+	return ret;
 }
 
 unsigned long KSTK_ESP(struct task_struct *task)
 {
 	return task_pt_regs(task)->sp;
 }
-- 
2.10.2

[PATCH v13 4/8] x86/syscalls/32: Wire up arch_prctl on x86-32

[Kyle Huey]

Hook up arch_prctl to call do_arch_prctl() on x86-32, and in 32 bit compat
mode on x86-64. This allows us to have arch_prctls that are not specific to
64 bits.

On UML, simply stub out this syscall.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
---
 arch/x86/entry/syscalls/syscall_32.tbl | 1 +
 arch/x86/kernel/process_32.c           | 7 +++++++
 arch/x86/kernel/process_64.c           | 7 +++++++
 arch/x86/um/Makefile                   | 2 +-
 arch/x86/um/syscalls_32.c              | 7 +++++++
 include/linux/compat.h                 | 2 ++
 6 files changed, 25 insertions(+), 1 deletion(-)
 create mode 100644 arch/x86/um/syscalls_32.c

diff --git a/arch/x86/entry/syscalls/syscall_32.tbl b/arch/x86/entry/syscalls/syscall_32.tbl
index 2b36185..d78c6b5 100644
--- a/arch/x86/entry/syscalls/syscall_32.tbl
+++ b/arch/x86/entry/syscalls/syscall_32.tbl
@@ -384,8 +384,9 @@
 375	i386	membarrier		sys_membarrier
 376	i386	mlock2			sys_mlock2
 377	i386	copy_file_range		sys_copy_file_range
 378	i386	preadv2			sys_preadv2			compat_sys_preadv2
 379	i386	pwritev2		sys_pwritev2			compat_sys_pwritev2
 380	i386	pkey_mprotect		sys_pkey_mprotect
 381	i386	pkey_alloc		sys_pkey_alloc
 382	i386	pkey_free		sys_pkey_free
+385	i386	arch_prctl		sys_arch_prctl			compat_sys_arch_prctl
diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c
index bd7be8e..95d3adc 100644
--- a/arch/x86/kernel/process_32.c
+++ b/arch/x86/kernel/process_32.c
@@ -30,16 +30,17 @@
 #include <linux/ptrace.h>
 #include <linux/personality.h>
 #include <linux/percpu.h>
 #include <linux/prctl.h>
 #include <linux/ftrace.h>
 #include <linux/uaccess.h>
 #include <linux/io.h>
 #include <linux/kdebug.h>
+#include <linux/syscalls.h>
 
 #include <asm/pgtable.h>
 #include <asm/ldt.h>
 #include <asm/processor.h>
 #include <asm/fpu/internal.h>
 #include <asm/desc.h>
 #ifdef CONFIG_MATH_EMULATION
 #include <asm/math_emu.h>
@@ -49,16 +50,17 @@
 
 #include <asm/tlbflush.h>
 #include <asm/cpu.h>
 #include <asm/idle.h>
 #include <asm/syscalls.h>
 #include <asm/debugreg.h>
 #include <asm/switch_to.h>
 #include <asm/vm86.h>
+#include <asm/proto.h>
 
 void __show_regs(struct pt_regs *regs, int all)
 {
 	unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
 	unsigned long d0, d1, d2, d3, d6, d7;
 	unsigned long sp;
 	unsigned short ss, gs;
 
@@ -296,8 +298,13 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 		lazy_load_gs(next->gs);
 
 	switch_fpu_finish(next_fpu, fpu_switch);
 
 	this_cpu_write(current_task, next_p);
 
 	return prev_p;
 }
+
+SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
+{
+	return do_arch_prctl_common(current, code, arg2);
+}
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index bf75d26..3a2a84d 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -621,12 +621,19 @@ SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
 
 	ret = do_arch_prctl_64(current, code, arg2);
 	if (ret == -EINVAL)
 		ret = do_arch_prctl_common(current, code, arg2);
 
 	return ret;
 }
 
+#ifdef CONFIG_IA32_EMULATION
+COMPAT_SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
+{
+	return do_arch_prctl_common(current, code, arg2);
+}
+#endif
+
 unsigned long KSTK_ESP(struct task_struct *task)
 {
 	return task_pt_regs(task)->sp;
 }
diff --git a/arch/x86/um/Makefile b/arch/x86/um/Makefile
index e7e7055..69f0827 100644
--- a/arch/x86/um/Makefile
+++ b/arch/x86/um/Makefile
@@ -11,17 +11,17 @@ endif
 obj-y = bug.o bugs_$(BITS).o delay.o fault.o ldt.o \
 	ptrace_$(BITS).o ptrace_user.o setjmp_$(BITS).o signal.o \
 	stub_$(BITS).o stub_segv.o \
 	sys_call_table_$(BITS).o sysrq_$(BITS).o tls_$(BITS).o \
 	mem_$(BITS).o subarch.o os-$(OS)/
 
 ifeq ($(CONFIG_X86_32),y)
 
-obj-y += checksum_32.o
+obj-y += checksum_32.o syscalls_32.o
 obj-$(CONFIG_ELF_CORE) += elfcore.o
 
 subarch-y = ../lib/string_32.o ../lib/atomic64_32.o ../lib/atomic64_cx8_32.o
 subarch-$(CONFIG_RWSEM_XCHGADD_ALGORITHM) += ../lib/rwsem.o
 
 else
 
 obj-y += syscalls_64.o vdso/
diff --git a/arch/x86/um/syscalls_32.c b/arch/x86/um/syscalls_32.c
new file mode 100644
index 0000000..ccf0598
--- /dev/null
+++ b/arch/x86/um/syscalls_32.c
@@ -0,0 +1,7 @@
+#include <linux/syscalls.h>
+#include <os.h>
+
+SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
+{
+	return -EINVAL;
+}
diff --git a/include/linux/compat.h b/include/linux/compat.h
index 6360939..0ddfc2ba 100644
--- a/include/linux/compat.h
+++ b/include/linux/compat.h
@@ -716,16 +716,18 @@ int __compat_save_altstack(compat_stack_t __user *, unsigned long);
 } while (0);
 
 asmlinkage long compat_sys_sched_rr_get_interval(compat_pid_t pid,
 						 struct compat_timespec __user *interval);
 
 asmlinkage long compat_sys_fanotify_mark(int, unsigned int, __u32, __u32,
 					    int, const char __user *);
 
+asmlinkage long compat_sys_arch_prctl(int code, unsigned long arg2);
+
 /*
  * For most but not all architectures, "am I in a compat syscall?" and
  * "am I a compat task?" are the same question.  For architectures on which
  * they aren't the same question, arch code can override in_compat_syscall.
  */
 
 #ifndef in_compat_syscall
 static inline bool in_compat_syscall(void) { return is_compat_task(); }
-- 
2.10.2

[PATCH v13 4/8] x86/syscalls/32: Wire up arch_prctl on x86-32

[Kyle Huey]

Hook up arch_prctl to call do_arch_prctl() on x86-32, and in 32 bit compat
mode on x86-64. This allows us to have arch_prctls that are not specific to
64 bits.

On UML, simply stub out this syscall.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
---
 arch/x86/entry/syscalls/syscall_32.tbl | 1 +
 arch/x86/kernel/process_32.c           | 7 +++++++
 arch/x86/kernel/process_64.c           | 7 +++++++
 arch/x86/um/Makefile                   | 2 +-
 arch/x86/um/syscalls_32.c              | 7 +++++++
 include/linux/compat.h                 | 2 ++
 6 files changed, 25 insertions(+), 1 deletion(-)
 create mode 100644 arch/x86/um/syscalls_32.c

diff --git a/arch/x86/entry/syscalls/syscall_32.tbl b/arch/x86/entry/syscalls/syscall_32.tbl
index 2b36185..d78c6b5 100644
--- a/arch/x86/entry/syscalls/syscall_32.tbl
+++ b/arch/x86/entry/syscalls/syscall_32.tbl
@@ -384,8 +384,9 @@
 375	i386	membarrier		sys_membarrier
 376	i386	mlock2			sys_mlock2
 377	i386	copy_file_range		sys_copy_file_range
 378	i386	preadv2			sys_preadv2			compat_sys_preadv2
 379	i386	pwritev2		sys_pwritev2			compat_sys_pwritev2
 380	i386	pkey_mprotect		sys_pkey_mprotect
 381	i386	pkey_alloc		sys_pkey_alloc
 382	i386	pkey_free		sys_pkey_free
+385	i386	arch_prctl		sys_arch_prctl			compat_sys_arch_prctl
diff --git a/arch/x86/kernel/process_32.c b/arch/x86/kernel/process_32.c
index bd7be8e..95d3adc 100644
--- a/arch/x86/kernel/process_32.c
+++ b/arch/x86/kernel/process_32.c
@@ -30,16 +30,17 @@
 #include <linux/ptrace.h>
 #include <linux/personality.h>
 #include <linux/percpu.h>
 #include <linux/prctl.h>
 #include <linux/ftrace.h>
 #include <linux/uaccess.h>
 #include <linux/io.h>
 #include <linux/kdebug.h>
+#include <linux/syscalls.h>
 
 #include <asm/pgtable.h>
 #include <asm/ldt.h>
 #include <asm/processor.h>
 #include <asm/fpu/internal.h>
 #include <asm/desc.h>
 #ifdef CONFIG_MATH_EMULATION
 #include <asm/math_emu.h>
@@ -49,16 +50,17 @@
 
 #include <asm/tlbflush.h>
 #include <asm/cpu.h>
 #include <asm/idle.h>
 #include <asm/syscalls.h>
 #include <asm/debugreg.h>
 #include <asm/switch_to.h>
 #include <asm/vm86.h>
+#include <asm/proto.h>
 
 void __show_regs(struct pt_regs *regs, int all)
 {
 	unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;
 	unsigned long d0, d1, d2, d3, d6, d7;
 	unsigned long sp;
 	unsigned short ss, gs;
 
@@ -296,8 +298,13 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 		lazy_load_gs(next->gs);
 
 	switch_fpu_finish(next_fpu, fpu_switch);
 
 	this_cpu_write(current_task, next_p);
 
 	return prev_p;
 }
+
+SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
+{
+	return do_arch_prctl_common(current, code, arg2);
+}
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index bf75d26..3a2a84d 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -621,12 +621,19 @@ SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
 
 	ret = do_arch_prctl_64(current, code, arg2);
 	if (ret == -EINVAL)
 		ret = do_arch_prctl_common(current, code, arg2);
 
 	return ret;
 }
 
+#ifdef CONFIG_IA32_EMULATION
+COMPAT_SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
+{
+	return do_arch_prctl_common(current, code, arg2);
+}
+#endif
+
 unsigned long KSTK_ESP(struct task_struct *task)
 {
 	return task_pt_regs(task)->sp;
 }
diff --git a/arch/x86/um/Makefile b/arch/x86/um/Makefile
index e7e7055..69f0827 100644
--- a/arch/x86/um/Makefile
+++ b/arch/x86/um/Makefile
@@ -11,17 +11,17 @@ endif
 obj-y = bug.o bugs_$(BITS).o delay.o fault.o ldt.o \
 	ptrace_$(BITS).o ptrace_user.o setjmp_$(BITS).o signal.o \
 	stub_$(BITS).o stub_segv.o \
 	sys_call_table_$(BITS).o sysrq_$(BITS).o tls_$(BITS).o \
 	mem_$(BITS).o subarch.o os-$(OS)/
 
 ifeq ($(CONFIG_X86_32),y)
 
-obj-y += checksum_32.o
+obj-y += checksum_32.o syscalls_32.o
 obj-$(CONFIG_ELF_CORE) += elfcore.o
 
 subarch-y = ../lib/string_32.o ../lib/atomic64_32.o ../lib/atomic64_cx8_32.o
 subarch-$(CONFIG_RWSEM_XCHGADD_ALGORITHM) += ../lib/rwsem.o
 
 else
 
 obj-y += syscalls_64.o vdso/
diff --git a/arch/x86/um/syscalls_32.c b/arch/x86/um/syscalls_32.c
new file mode 100644
index 0000000..ccf0598
--- /dev/null
+++ b/arch/x86/um/syscalls_32.c
@@ -0,0 +1,7 @@
+#include <linux/syscalls.h>
+#include <os.h>
+
+SYSCALL_DEFINE2(arch_prctl, int, code, unsigned long, arg2)
+{
+	return -EINVAL;
+}
diff --git a/include/linux/compat.h b/include/linux/compat.h
index 6360939..0ddfc2ba 100644
--- a/include/linux/compat.h
+++ b/include/linux/compat.h
@@ -716,16 +716,18 @@ int __compat_save_altstack(compat_stack_t __user *, unsigned long);
 } while (0);
 
 asmlinkage long compat_sys_sched_rr_get_interval(compat_pid_t pid,
 						 struct compat_timespec __user *interval);
 
 asmlinkage long compat_sys_fanotify_mark(int, unsigned int, __u32, __u32,
 					    int, const char __user *);
 
+asmlinkage long compat_sys_arch_prctl(int code, unsigned long arg2);
+
 /*
  * For most but not all architectures, "am I in a compat syscall?" and
  * "am I a compat task?" are the same question.  For architectures on which
  * they aren't the same question, arch code can override in_compat_syscall.
  */
 
 #ifndef in_compat_syscall
 static inline bool in_compat_syscall(void) { return is_compat_task(); }
-- 
2.10.2

[PATCH v13 5/8] x86/cpufeature: Detect CPUID faulting support

[Kyle Huey]

Intel supports faulting on the CPUID instruction beginning with Ivy Bridge.
When enabled, the processor will fault on attempts to execute the CPUID
instruction with CPL>0. This will allow a ptracer to emulate the CPUID
instruction.

Bit 31 of MSR_PLATFORM_INFO advertises support for this feature. It is
documented in detail in Section 2.3.2 of
https://bugzilla.kernel.org/attachment.cgi?id=243991

Detect support for this feature and expose it as X86_FEATURE_CPUID_FAULT.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
Reviewed-by: Borislav Petkov <bp@suse.de>
---
 arch/x86/include/asm/cpufeatures.h |  1 +
 arch/x86/include/asm/msr-index.h   |  2 ++
 arch/x86/kernel/cpu/intel.c        | 12 ++++++++++++
 3 files changed, 15 insertions(+)

diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h
index a396292..85f853f 100644
--- a/arch/x86/include/asm/cpufeatures.h
+++ b/arch/x86/include/asm/cpufeatures.h
@@ -182,16 +182,17 @@
 
 /*
  * Auxiliary flags: Linux defined - For features scattered in various
  * CPUID levels like 0x6, 0xA etc, word 7.
  *
  * Reuse free bits when adding new feature flags!
  */
 
+#define X86_FEATURE_CPUID_FAULT ( 7*32+ 0) /* Intel CPUID faulting */
 #define X86_FEATURE_CPB		( 7*32+ 2) /* AMD Core Performance Boost */
 #define X86_FEATURE_EPB		( 7*32+ 3) /* IA32_ENERGY_PERF_BIAS support */
 
 #define X86_FEATURE_HW_PSTATE	( 7*32+ 8) /* AMD HW-PState */
 #define X86_FEATURE_PROC_FEEDBACK ( 7*32+ 9) /* AMD ProcFeedbackInterface */
 
 #define X86_FEATURE_INTEL_PT	( 7*32+15) /* Intel Processor Trace */
 #define X86_FEATURE_AVX512_4VNNIW (7*32+16) /* AVX-512 Neural Network Instructions */
diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h
index 78f3760..3ac0acf 100644
--- a/arch/x86/include/asm/msr-index.h
+++ b/arch/x86/include/asm/msr-index.h
@@ -36,16 +36,18 @@
 #define EFER_LMSLE		(1<<_EFER_LMSLE)
 #define EFER_FFXSR		(1<<_EFER_FFXSR)
 
 /* Intel MSRs. Some also available on other CPUs */
 #define MSR_IA32_PERFCTR0		0x000000c1
 #define MSR_IA32_PERFCTR1		0x000000c2
 #define MSR_FSB_FREQ			0x000000cd
 #define MSR_PLATFORM_INFO		0x000000ce
+#define MSR_PLATFORM_INFO_CPUID_FAULT_BIT	31
+#define MSR_PLATFORM_INFO_CPUID_FAULT		BIT_ULL(MSR_PLATFORM_INFO_CPUID_FAULT_BIT)
 
 #define MSR_NHM_SNB_PKG_CST_CFG_CTL	0x000000e2
 #define NHM_C3_AUTO_DEMOTE		(1UL << 25)
 #define NHM_C1_AUTO_DEMOTE		(1UL << 26)
 #define ATM_LNC_C6_AUTO_DEMOTE		(1UL << 25)
 #define SNB_C1_AUTO_UNDEMOTE		(1UL << 27)
 #define SNB_C3_AUTO_UNDEMOTE		(1UL << 28)
 
diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c
index fcd484d..12ddddf 100644
--- a/arch/x86/kernel/cpu/intel.c
+++ b/arch/x86/kernel/cpu/intel.c
@@ -447,16 +447,26 @@ static void intel_bsp_resume(struct cpuinfo_x86 *c)
 {
 	/*
 	 * MSR_IA32_ENERGY_PERF_BIAS is lost across suspend/resume,
 	 * so reinitialize it properly like during bootup:
 	 */
 	init_intel_energy_perf(c);
 }
 
+static void init_intel_misc_features(struct cpuinfo_x86 *c)
+{
+	u64 msr;
+
+	if (!rdmsrl_safe(MSR_PLATFORM_INFO, &msr)) {
+		if (msr & MSR_PLATFORM_INFO_CPUID_FAULT)
+			set_cpu_cap(c, X86_FEATURE_CPUID_FAULT);
+	}
+}
+
 static void init_intel(struct cpuinfo_x86 *c)
 {
 	unsigned int l2 = 0;
 
 	early_init_intel(c);
 
 	intel_workarounds(c);
 
@@ -560,16 +570,18 @@ static void init_intel(struct cpuinfo_x86 *c)
 
 	/* Work around errata */
 	srat_detect_node(c);
 
 	if (cpu_has(c, X86_FEATURE_VMX))
 		detect_vmx_virtcap(c);
 
 	init_intel_energy_perf(c);
+
+	init_intel_misc_features(c);
 }
 
 #ifdef CONFIG_X86_32
 static unsigned int intel_size_cache(struct cpuinfo_x86 *c, unsigned int size)
 {
 	/*
 	 * Intel PIII Tualatin. This comes in two flavours.
 	 * One has 256kb of cache, the other 512. We have no way
-- 
2.10.2

[PATCH v13 5/8] x86/cpufeature: Detect CPUID faulting support

[Kyle Huey]

Intel supports faulting on the CPUID instruction beginning with Ivy Bridge.
When enabled, the processor will fault on attempts to execute the CPUID
instruction with CPL>0. This will allow a ptracer to emulate the CPUID
instruction.

Bit 31 of MSR_PLATFORM_INFO advertises support for this feature. It is
documented in detail in Section 2.3.2 of
https://bugzilla.kernel.org/attachment.cgi?id=243991

Detect support for this feature and expose it as X86_FEATURE_CPUID_FAULT.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
Reviewed-by: Borislav Petkov <bp@suse.de>
---
 arch/x86/include/asm/cpufeatures.h |  1 +
 arch/x86/include/asm/msr-index.h   |  2 ++
 arch/x86/kernel/cpu/intel.c        | 12 ++++++++++++
 3 files changed, 15 insertions(+)

diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h
index a396292..85f853f 100644
--- a/arch/x86/include/asm/cpufeatures.h
+++ b/arch/x86/include/asm/cpufeatures.h
@@ -182,16 +182,17 @@
 
 /*
  * Auxiliary flags: Linux defined - For features scattered in various
  * CPUID levels like 0x6, 0xA etc, word 7.
  *
  * Reuse free bits when adding new feature flags!
  */
 
+#define X86_FEATURE_CPUID_FAULT ( 7*32+ 0) /* Intel CPUID faulting */
 #define X86_FEATURE_CPB		( 7*32+ 2) /* AMD Core Performance Boost */
 #define X86_FEATURE_EPB		( 7*32+ 3) /* IA32_ENERGY_PERF_BIAS support */
 
 #define X86_FEATURE_HW_PSTATE	( 7*32+ 8) /* AMD HW-PState */
 #define X86_FEATURE_PROC_FEEDBACK ( 7*32+ 9) /* AMD ProcFeedbackInterface */
 
 #define X86_FEATURE_INTEL_PT	( 7*32+15) /* Intel Processor Trace */
 #define X86_FEATURE_AVX512_4VNNIW (7*32+16) /* AVX-512 Neural Network Instructions */
diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h
index 78f3760..3ac0acf 100644
--- a/arch/x86/include/asm/msr-index.h
+++ b/arch/x86/include/asm/msr-index.h
@@ -36,16 +36,18 @@
 #define EFER_LMSLE		(1<<_EFER_LMSLE)
 #define EFER_FFXSR		(1<<_EFER_FFXSR)
 
 /* Intel MSRs. Some also available on other CPUs */
 #define MSR_IA32_PERFCTR0		0x000000c1
 #define MSR_IA32_PERFCTR1		0x000000c2
 #define MSR_FSB_FREQ			0x000000cd
 #define MSR_PLATFORM_INFO		0x000000ce
+#define MSR_PLATFORM_INFO_CPUID_FAULT_BIT	31
+#define MSR_PLATFORM_INFO_CPUID_FAULT		BIT_ULL(MSR_PLATFORM_INFO_CPUID_FAULT_BIT)
 
 #define MSR_NHM_SNB_PKG_CST_CFG_CTL	0x000000e2
 #define NHM_C3_AUTO_DEMOTE		(1UL << 25)
 #define NHM_C1_AUTO_DEMOTE		(1UL << 26)
 #define ATM_LNC_C6_AUTO_DEMOTE		(1UL << 25)
 #define SNB_C1_AUTO_UNDEMOTE		(1UL << 27)
 #define SNB_C3_AUTO_UNDEMOTE		(1UL << 28)
 
diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c
index fcd484d..12ddddf 100644
--- a/arch/x86/kernel/cpu/intel.c
+++ b/arch/x86/kernel/cpu/intel.c
@@ -447,16 +447,26 @@ static void intel_bsp_resume(struct cpuinfo_x86 *c)
 {
 	/*
 	 * MSR_IA32_ENERGY_PERF_BIAS is lost across suspend/resume,
 	 * so reinitialize it properly like during bootup:
 	 */
 	init_intel_energy_perf(c);
 }
 
+static void init_intel_misc_features(struct cpuinfo_x86 *c)
+{
+	u64 msr;
+
+	if (!rdmsrl_safe(MSR_PLATFORM_INFO, &msr)) {
+		if (msr & MSR_PLATFORM_INFO_CPUID_FAULT)
+			set_cpu_cap(c, X86_FEATURE_CPUID_FAULT);
+	}
+}
+
 static void init_intel(struct cpuinfo_x86 *c)
 {
 	unsigned int l2 = 0;
 
 	early_init_intel(c);
 
 	intel_workarounds(c);
 
@@ -560,16 +570,18 @@ static void init_intel(struct cpuinfo_x86 *c)
 
 	/* Work around errata */
 	srat_detect_node(c);
 
 	if (cpu_has(c, X86_FEATURE_VMX))
 		detect_vmx_virtcap(c);
 
 	init_intel_energy_perf(c);
+
+	init_intel_misc_features(c);
 }
 
 #ifdef CONFIG_X86_32
 static unsigned int intel_size_cache(struct cpuinfo_x86 *c, unsigned int size)
 {
 	/*
 	 * Intel PIII Tualatin. This comes in two flavours.
 	 * One has 256kb of cache, the other 512. We have no way
-- 
2.10.2

[PATCH v13 6/8] x86/arch_prctl: Add ARCH_[GET|SET]_CPUID

[Kyle Huey]

Intel supports faulting on the CPUID instruction beginning with Ivy Bridge.
When enabled, the processor will fault on attempts to execute the CPUID
instruction with CPL>0. Exposing this feature to userspace will allow a
ptracer to trap and emulate the CPUID instruction.

When supported, this feature is controlled by toggling bit 0 of
MSR_MISC_FEATURES_ENABLES. It is documented in detail in Section 2.3.2 of
https://bugzilla.kernel.org/attachment.cgi?id=243991

Implement a new pair of arch_prctls, available on both x86-32 and x86-64.

ARCH_GET_CPUID: Returns the current CPUID state, either 0 if CPUID faulting
    is enabled (and thus the CPUID instruction is not available) or 1 if
    CPUID faulting is not enabled.

ARCH_SET_CPUID: Set the CPUID state to the second argument. If
    cpuid_enabled is 0 CPUID faulting will be activated, otherwise it will
    be deactivated. Returns ENODEV if CPUID faulting is not supported on
    this system.

The state of the CPUID faulting flag is propagated across forks, but reset
upon exec.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
---
 arch/x86/include/asm/msr-index.h   |  3 ++
 arch/x86/include/asm/processor.h   |  2 +
 arch/x86/include/asm/thread_info.h |  6 ++-
 arch/x86/include/uapi/asm/prctl.h  |  3 ++
 arch/x86/kernel/cpu/intel.c        |  7 ++++
 arch/x86/kernel/process.c          | 82 ++++++++++++++++++++++++++++++++++++++
 fs/exec.c                          |  1 +
 include/linux/thread_info.h        |  4 ++
 8 files changed, 107 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h
index 3ac0acf..b138b8f 100644
--- a/arch/x86/include/asm/msr-index.h
+++ b/arch/x86/include/asm/msr-index.h
@@ -49,16 +49,19 @@
 #define NHM_C1_AUTO_DEMOTE		(1UL << 26)
 #define ATM_LNC_C6_AUTO_DEMOTE		(1UL << 25)
 #define SNB_C1_AUTO_UNDEMOTE		(1UL << 27)
 #define SNB_C3_AUTO_UNDEMOTE		(1UL << 28)
 
 #define MSR_MTRRcap			0x000000fe
 #define MSR_IA32_BBL_CR_CTL		0x00000119
 #define MSR_IA32_BBL_CR_CTL3		0x0000011e
+#define MSR_MISC_FEATURES_ENABLES	0x00000140
+#define MSR_MISC_FEATURES_ENABLES_CPUID_FAULT_BIT	0
+#define MSR_MISC_FEATURES_ENABLES_CPUID_FAULT		BIT_ULL(MSR_MISC_FEATURES_ENABLES_CPUID_FAULT_BIT)
 
 #define MSR_IA32_SYSENTER_CS		0x00000174
 #define MSR_IA32_SYSENTER_ESP		0x00000175
 #define MSR_IA32_SYSENTER_EIP		0x00000176
 
 #define MSR_IA32_MCG_CAP		0x00000179
 #define MSR_IA32_MCG_STATUS		0x0000017a
 #define MSR_IA32_MCG_CTL		0x0000017b
diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index 984a7bf..65b2494 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -803,16 +803,18 @@ extern void start_thread(struct pt_regs *regs, unsigned long new_ip,
 
 /* Get/set a process' ability to use the timestamp counter instruction */
 #define GET_TSC_CTL(adr)	get_tsc_mode((adr))
 #define SET_TSC_CTL(val)	set_tsc_mode((val))
 
 extern int get_tsc_mode(unsigned long adr);
 extern int set_tsc_mode(unsigned int val);
 
+DECLARE_PER_CPU(u64, msr_misc_features_shadow);
+
 /* Register/unregister a process' MPX related resource */
 #define MPX_ENABLE_MANAGEMENT()	mpx_enable_management()
 #define MPX_DISABLE_MANAGEMENT()	mpx_disable_management()
 
 #ifdef CONFIG_X86_INTEL_MPX
 extern int mpx_enable_management(void);
 extern int mpx_disable_management(void);
 #else
diff --git a/arch/x86/include/asm/thread_info.h b/arch/x86/include/asm/thread_info.h
index ad6f5eb0..9fc44b9 100644
--- a/arch/x86/include/asm/thread_info.h
+++ b/arch/x86/include/asm/thread_info.h
@@ -82,16 +82,17 @@ struct thread_info {
 #define TIF_SIGPENDING		2	/* signal pending */
 #define TIF_NEED_RESCHED	3	/* rescheduling necessary */
 #define TIF_SINGLESTEP		4	/* reenable singlestep on user return*/
 #define TIF_SYSCALL_EMU		6	/* syscall emulation active */
 #define TIF_SYSCALL_AUDIT	7	/* syscall auditing active */
 #define TIF_SECCOMP		8	/* secure computing */
 #define TIF_USER_RETURN_NOTIFY	11	/* notify kernel of userspace return */
 #define TIF_UPROBE		12	/* breakpointed or singlestepping */
+#define TIF_NOCPUID		15	/* CPUID is not accessible in userland */
 #define TIF_NOTSC		16	/* TSC is not accessible in userland */
 #define TIF_IA32		17	/* IA32 compatibility process */
 #define TIF_NOHZ		19	/* in adaptive nohz mode */
 #define TIF_MEMDIE		20	/* is terminating due to OOM killer */
 #define TIF_POLLING_NRFLAG	21	/* idle is polling for TIF_NEED_RESCHED */
 #define TIF_IO_BITMAP		22	/* uses I/O bitmap */
 #define TIF_FORCED_TF		24	/* true if TF in eflags artificially */
 #define TIF_BLOCKSTEP		25	/* set when we want DEBUGCTLMSR_BTF */
@@ -105,16 +106,17 @@ struct thread_info {
 #define _TIF_SIGPENDING		(1 << TIF_SIGPENDING)
 #define _TIF_SINGLESTEP		(1 << TIF_SINGLESTEP)
 #define _TIF_NEED_RESCHED	(1 << TIF_NEED_RESCHED)
 #define _TIF_SYSCALL_EMU	(1 << TIF_SYSCALL_EMU)
 #define _TIF_SYSCALL_AUDIT	(1 << TIF_SYSCALL_AUDIT)
 #define _TIF_SECCOMP		(1 << TIF_SECCOMP)
 #define _TIF_USER_RETURN_NOTIFY	(1 << TIF_USER_RETURN_NOTIFY)
 #define _TIF_UPROBE		(1 << TIF_UPROBE)
+#define _TIF_NOCPUID		(1 << TIF_NOCPUID)
 #define _TIF_NOTSC		(1 << TIF_NOTSC)
 #define _TIF_IA32		(1 << TIF_IA32)
 #define _TIF_NOHZ		(1 << TIF_NOHZ)
 #define _TIF_POLLING_NRFLAG	(1 << TIF_POLLING_NRFLAG)
 #define _TIF_IO_BITMAP		(1 << TIF_IO_BITMAP)
 #define _TIF_FORCED_TF		(1 << TIF_FORCED_TF)
 #define _TIF_BLOCKSTEP		(1 << TIF_BLOCKSTEP)
 #define _TIF_LAZY_MMU_UPDATES	(1 << TIF_LAZY_MMU_UPDATES)
@@ -133,17 +135,17 @@ struct thread_info {
 
 /* work to do on any return to user space */
 #define _TIF_ALLWORK_MASK						\
 	((0x0000FFFF & ~_TIF_SECCOMP) | _TIF_SYSCALL_TRACEPOINT |	\
 	_TIF_NOHZ)
 
 /* flags to check in __switch_to() */
 #define _TIF_WORK_CTXSW							\
-	(_TIF_IO_BITMAP|_TIF_NOTSC|_TIF_BLOCKSTEP)
+	(_TIF_IO_BITMAP|_TIF_NOCPUID|_TIF_NOTSC|_TIF_BLOCKSTEP)
 
 #define _TIF_WORK_CTXSW_PREV (_TIF_WORK_CTXSW|_TIF_USER_RETURN_NOTIFY)
 #define _TIF_WORK_CTXSW_NEXT (_TIF_WORK_CTXSW)
 
 #define STACK_WARN		(THREAD_SIZE/8)
 
 /*
  * macros/functions for gaining access to the thread information structure
@@ -234,11 +236,13 @@ static inline int arch_within_stack_frames(const void * const stack,
  * EFLAGS values that other (fast) syscall return instructions
  * are not able to restore properly.
  */
 #define force_iret() set_thread_flag(TIF_NOTIFY_RESUME)
 
 extern void arch_task_cache_init(void);
 extern int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src);
 extern void arch_release_task_struct(struct task_struct *tsk);
+extern void arch_setup_new_exec(void);
+#define arch_setup_new_exec arch_setup_new_exec
 #endif	/* !__ASSEMBLY__ */
 
 #endif /* _ASM_X86_THREAD_INFO_H */
diff --git a/arch/x86/include/uapi/asm/prctl.h b/arch/x86/include/uapi/asm/prctl.h
index ae135de..80240a2 100644
--- a/arch/x86/include/uapi/asm/prctl.h
+++ b/arch/x86/include/uapi/asm/prctl.h
@@ -1,15 +1,18 @@
 #ifndef _ASM_X86_PRCTL_H
 #define _ASM_X86_PRCTL_H
 
 #define ARCH_SET_GS 0x1001
 #define ARCH_SET_FS 0x1002
 #define ARCH_GET_FS 0x1003
 #define ARCH_GET_GS 0x1004
 
+#define ARCH_GET_CPUID 0x1005
+#define ARCH_SET_CPUID 0x1006
+
 #ifdef CONFIG_CHECKPOINT_RESTORE
 # define ARCH_MAP_VDSO_X32	0x2001
 # define ARCH_MAP_VDSO_32	0x2002
 # define ARCH_MAP_VDSO_64	0x2003
 #endif
 
 #endif /* _ASM_X86_PRCTL_H */
diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c
index 12ddddf..11e4a0a 100644
--- a/arch/x86/kernel/cpu/intel.c
+++ b/arch/x86/kernel/cpu/intel.c
@@ -451,16 +451,23 @@ static void intel_bsp_resume(struct cpuinfo_x86 *c)
 	 */
 	init_intel_energy_perf(c);
 }
 
 static void init_intel_misc_features(struct cpuinfo_x86 *c)
 {
 	u64 msr;
 
+	if (rdmsrl_safe(MSR_MISC_FEATURES_ENABLES, &msr))
+		return;
+
+	msr = 0;
+	wrmsrl(MSR_MISC_FEATURES_ENABLES, msr);
+	this_cpu_write(msr_misc_features_shadow, msr);
+
 	if (!rdmsrl_safe(MSR_PLATFORM_INFO, &msr)) {
 		if (msr & MSR_PLATFORM_INFO_CPUID_FAULT)
 			set_cpu_cap(c, X86_FEATURE_CPUID_FAULT);
 	}
 }
 
 static void init_intel(struct cpuinfo_x86 *c)
 {
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index 5ac20b5..ba39090 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -28,16 +28,17 @@
 #include <asm/mwait.h>
 #include <asm/fpu/internal.h>
 #include <asm/debugreg.h>
 #include <asm/nmi.h>
 #include <asm/tlbflush.h>
 #include <asm/mce.h>
 #include <asm/vm86.h>
 #include <asm/switch_to.h>
+#include <asm/prctl.h>
 
 /*
  * per-CPU TSS segments. Threads are completely 'soft' on Linux,
  * no more per-task TSS's. The TSS size is kept cacheline-aligned
  * so they are allowed to end up in the .data..cacheline_aligned
  * section. Since TSS's are completely CPU-local, we want them
  * on exact cacheline boundaries, to eliminate cacheline ping-pong.
  */
@@ -187,16 +188,85 @@ int set_tsc_mode(unsigned int val)
 	else if (val == PR_TSC_ENABLE)
 		enable_TSC();
 	else
 		return -EINVAL;
 
 	return 0;
 }
 
+DEFINE_PER_CPU(u64, msr_misc_features_shadow);
+
+static void set_cpuid_faulting(bool on)
+{
+	u64 msrval;
+
+	DEBUG_LOCKS_WARN_ON(!irqs_disabled());
+
+	msrval = this_cpu_read(msr_misc_features_shadow);
+	msrval &= ~MSR_MISC_FEATURES_ENABLES_CPUID_FAULT;
+	msrval |= (on << MSR_MISC_FEATURES_ENABLES_CPUID_FAULT_BIT);
+	this_cpu_write(msr_misc_features_shadow, msrval);
+	wrmsrl(MSR_MISC_FEATURES_ENABLES, msrval);
+}
+
+static void disable_cpuid(void)
+{
+	preempt_disable();
+	if (!test_and_set_thread_flag(TIF_NOCPUID)) {
+		/*
+		 * Must flip the CPU state synchronously with
+		 * TIF_NOCPUID in the current running context.
+		 */
+		set_cpuid_faulting(true);
+	}
+	preempt_enable();
+}
+
+static void enable_cpuid(void)
+{
+	preempt_disable();
+	if (test_and_clear_thread_flag(TIF_NOCPUID)) {
+		/*
+		 * Must flip the CPU state synchronously with
+		 * TIF_NOCPUID in the current running context.
+		 */
+		set_cpuid_faulting(false);
+	}
+	preempt_enable();
+}
+
+static int get_cpuid_mode(void)
+{
+	return !test_thread_flag(TIF_NOCPUID);
+}
+
+static int set_cpuid_mode(struct task_struct *task, unsigned long cpuid_enabled)
+{
+	if (!static_cpu_has(X86_FEATURE_CPUID_FAULT))
+		return -ENODEV;
+
+	if (cpuid_enabled)
+		enable_cpuid();
+	else
+		disable_cpuid();
+
+	return 0;
+}
+
+/*
+ * Called immediately after a successful exec.
+ */
+void arch_setup_new_exec(void)
+{
+	/* If cpuid was previously disabled for this task, re-enable it. */
+	if (test_thread_flag(TIF_NOCPUID))
+		enable_cpuid();
+}
+
 void __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p,
 		      struct tss_struct *tss)
 {
 	struct thread_struct *prev, *next;
 
 	prev = &prev_p->thread;
 	next = &next_p->thread;
 
@@ -206,16 +276,21 @@ void __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p,
 
 		debugctl &= ~DEBUGCTLMSR_BTF;
 		if (test_tsk_thread_flag(next_p, TIF_BLOCKSTEP))
 			debugctl |= DEBUGCTLMSR_BTF;
 
 		update_debugctlmsr(debugctl);
 	}
 
+	if (test_tsk_thread_flag(prev_p, TIF_NOCPUID) ^
+	    test_tsk_thread_flag(next_p, TIF_NOCPUID)) {
+		set_cpuid_faulting(test_tsk_thread_flag(next_p, TIF_NOCPUID));
+	}
+
 	if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^
 	    test_tsk_thread_flag(next_p, TIF_NOTSC)) {
 		/* prev and next are different */
 		if (test_tsk_thread_flag(next_p, TIF_NOTSC))
 			hard_disable_TSC();
 		else
 			hard_enable_TSC();
 	}
@@ -583,10 +658,17 @@ unsigned long get_wchan(struct task_struct *p)
 out:
 	put_task_stack(p);
 	return ret;
 }
 
 long do_arch_prctl_common(struct task_struct *task, int code,
 			  unsigned long cpuid_enabled)
 {
+	switch (code) {
+	case ARCH_GET_CPUID:
+		return get_cpuid_mode();
+	case ARCH_SET_CPUID:
+		return set_cpuid_mode(task, cpuid_enabled);
+	}
+
 	return -EINVAL;
 }
diff --git a/fs/exec.c b/fs/exec.c
index 4e497b9..3e6872c 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1287,16 +1287,17 @@ void setup_new_exec(struct linux_binprm * bprm)
 	/* This is the point of no return */
 	current->sas_ss_sp = current->sas_ss_size = 0;
 
 	if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
 		set_dumpable(current->mm, SUID_DUMP_USER);
 	else
 		set_dumpable(current->mm, suid_dumpable);
 
+	arch_setup_new_exec();
 	perf_event_exec();
 	__set_task_comm(current, kbasename(bprm->filename), true);
 
 	/* Set the new mm task size. We have to do that late because it may
 	 * depend on TIF_32BIT which is only updated in flush_thread() on
 	 * some architectures like powerpc
 	 */
 	current->mm->task_size = TASK_SIZE;
diff --git a/include/linux/thread_info.h b/include/linux/thread_info.h
index 2873baf..d46f50d 100644
--- a/include/linux/thread_info.h
+++ b/include/linux/thread_info.h
@@ -129,11 +129,15 @@ static __always_inline void check_object_size(const void *ptr, unsigned long n,
 		__check_object_size(ptr, n, to_user);
 }
 #else
 static inline void check_object_size(const void *ptr, unsigned long n,
 				     bool to_user)
 { }
 #endif /* CONFIG_HARDENED_USERCOPY */
 
+#ifndef arch_setup_new_exec
+static inline void arch_setup_new_exec(void) {}
+#endif
+
 #endif	/* __KERNEL__ */
 
 #endif /* _LINUX_THREAD_INFO_H */
-- 
2.10.2

[PATCH v13 6/8] x86/arch_prctl: Add ARCH_[GET|SET]_CPUID

[Kyle Huey]

Intel supports faulting on the CPUID instruction beginning with Ivy Bridge.
When enabled, the processor will fault on attempts to execute the CPUID
instruction with CPL>0. Exposing this feature to userspace will allow a
ptracer to trap and emulate the CPUID instruction.

When supported, this feature is controlled by toggling bit 0 of
MSR_MISC_FEATURES_ENABLES. It is documented in detail in Section 2.3.2 of
https://bugzilla.kernel.org/attachment.cgi?id=243991

Implement a new pair of arch_prctls, available on both x86-32 and x86-64.

ARCH_GET_CPUID: Returns the current CPUID state, either 0 if CPUID faulting
    is enabled (and thus the CPUID instruction is not available) or 1 if
    CPUID faulting is not enabled.

ARCH_SET_CPUID: Set the CPUID state to the second argument. If
    cpuid_enabled is 0 CPUID faulting will be activated, otherwise it will
    be deactivated. Returns ENODEV if CPUID faulting is not supported on
    this system.

The state of the CPUID faulting flag is propagated across forks, but reset
upon exec.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
---
 arch/x86/include/asm/msr-index.h   |  3 ++
 arch/x86/include/asm/processor.h   |  2 +
 arch/x86/include/asm/thread_info.h |  6 ++-
 arch/x86/include/uapi/asm/prctl.h  |  3 ++
 arch/x86/kernel/cpu/intel.c        |  7 ++++
 arch/x86/kernel/process.c          | 82 ++++++++++++++++++++++++++++++++++++++
 fs/exec.c                          |  1 +
 include/linux/thread_info.h        |  4 ++
 8 files changed, 107 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/msr-index.h b/arch/x86/include/asm/msr-index.h
index 3ac0acf..b138b8f 100644
--- a/arch/x86/include/asm/msr-index.h
+++ b/arch/x86/include/asm/msr-index.h
@@ -49,16 +49,19 @@
 #define NHM_C1_AUTO_DEMOTE		(1UL << 26)
 #define ATM_LNC_C6_AUTO_DEMOTE		(1UL << 25)
 #define SNB_C1_AUTO_UNDEMOTE		(1UL << 27)
 #define SNB_C3_AUTO_UNDEMOTE		(1UL << 28)
 
 #define MSR_MTRRcap			0x000000fe
 #define MSR_IA32_BBL_CR_CTL		0x00000119
 #define MSR_IA32_BBL_CR_CTL3		0x0000011e
+#define MSR_MISC_FEATURES_ENABLES	0x00000140
+#define MSR_MISC_FEATURES_ENABLES_CPUID_FAULT_BIT	0
+#define MSR_MISC_FEATURES_ENABLES_CPUID_FAULT		BIT_ULL(MSR_MISC_FEATURES_ENABLES_CPUID_FAULT_BIT)
 
 #define MSR_IA32_SYSENTER_CS		0x00000174
 #define MSR_IA32_SYSENTER_ESP		0x00000175
 #define MSR_IA32_SYSENTER_EIP		0x00000176
 
 #define MSR_IA32_MCG_CAP		0x00000179
 #define MSR_IA32_MCG_STATUS		0x0000017a
 #define MSR_IA32_MCG_CTL		0x0000017b
diff --git a/arch/x86/include/asm/processor.h b/arch/x86/include/asm/processor.h
index 984a7bf..65b2494 100644
--- a/arch/x86/include/asm/processor.h
+++ b/arch/x86/include/asm/processor.h
@@ -803,16 +803,18 @@ extern void start_thread(struct pt_regs *regs, unsigned long new_ip,
 
 /* Get/set a process' ability to use the timestamp counter instruction */
 #define GET_TSC_CTL(adr)	get_tsc_mode((adr))
 #define SET_TSC_CTL(val)	set_tsc_mode((val))
 
 extern int get_tsc_mode(unsigned long adr);
 extern int set_tsc_mode(unsigned int val);
 
+DECLARE_PER_CPU(u64, msr_misc_features_shadow);
+
 /* Register/unregister a process' MPX related resource */
 #define MPX_ENABLE_MANAGEMENT()	mpx_enable_management()
 #define MPX_DISABLE_MANAGEMENT()	mpx_disable_management()
 
 #ifdef CONFIG_X86_INTEL_MPX
 extern int mpx_enable_management(void);
 extern int mpx_disable_management(void);
 #else
diff --git a/arch/x86/include/asm/thread_info.h b/arch/x86/include/asm/thread_info.h
index ad6f5eb0..9fc44b9 100644
--- a/arch/x86/include/asm/thread_info.h
+++ b/arch/x86/include/asm/thread_info.h
@@ -82,16 +82,17 @@ struct thread_info {
 #define TIF_SIGPENDING		2	/* signal pending */
 #define TIF_NEED_RESCHED	3	/* rescheduling necessary */
 #define TIF_SINGLESTEP		4	/* reenable singlestep on user return*/
 #define TIF_SYSCALL_EMU		6	/* syscall emulation active */
 #define TIF_SYSCALL_AUDIT	7	/* syscall auditing active */
 #define TIF_SECCOMP		8	/* secure computing */
 #define TIF_USER_RETURN_NOTIFY	11	/* notify kernel of userspace return */
 #define TIF_UPROBE		12	/* breakpointed or singlestepping */
+#define TIF_NOCPUID		15	/* CPUID is not accessible in userland */
 #define TIF_NOTSC		16	/* TSC is not accessible in userland */
 #define TIF_IA32		17	/* IA32 compatibility process */
 #define TIF_NOHZ		19	/* in adaptive nohz mode */
 #define TIF_MEMDIE		20	/* is terminating due to OOM killer */
 #define TIF_POLLING_NRFLAG	21	/* idle is polling for TIF_NEED_RESCHED */
 #define TIF_IO_BITMAP		22	/* uses I/O bitmap */
 #define TIF_FORCED_TF		24	/* true if TF in eflags artificially */
 #define TIF_BLOCKSTEP		25	/* set when we want DEBUGCTLMSR_BTF */
@@ -105,16 +106,17 @@ struct thread_info {
 #define _TIF_SIGPENDING		(1 << TIF_SIGPENDING)
 #define _TIF_SINGLESTEP		(1 << TIF_SINGLESTEP)
 #define _TIF_NEED_RESCHED	(1 << TIF_NEED_RESCHED)
 #define _TIF_SYSCALL_EMU	(1 << TIF_SYSCALL_EMU)
 #define _TIF_SYSCALL_AUDIT	(1 << TIF_SYSCALL_AUDIT)
 #define _TIF_SECCOMP		(1 << TIF_SECCOMP)
 #define _TIF_USER_RETURN_NOTIFY	(1 << TIF_USER_RETURN_NOTIFY)
 #define _TIF_UPROBE		(1 << TIF_UPROBE)
+#define _TIF_NOCPUID		(1 << TIF_NOCPUID)
 #define _TIF_NOTSC		(1 << TIF_NOTSC)
 #define _TIF_IA32		(1 << TIF_IA32)
 #define _TIF_NOHZ		(1 << TIF_NOHZ)
 #define _TIF_POLLING_NRFLAG	(1 << TIF_POLLING_NRFLAG)
 #define _TIF_IO_BITMAP		(1 << TIF_IO_BITMAP)
 #define _TIF_FORCED_TF		(1 << TIF_FORCED_TF)
 #define _TIF_BLOCKSTEP		(1 << TIF_BLOCKSTEP)
 #define _TIF_LAZY_MMU_UPDATES	(1 << TIF_LAZY_MMU_UPDATES)
@@ -133,17 +135,17 @@ struct thread_info {
 
 /* work to do on any return to user space */
 #define _TIF_ALLWORK_MASK						\
 	((0x0000FFFF & ~_TIF_SECCOMP) | _TIF_SYSCALL_TRACEPOINT |	\
 	_TIF_NOHZ)
 
 /* flags to check in __switch_to() */
 #define _TIF_WORK_CTXSW							\
-	(_TIF_IO_BITMAP|_TIF_NOTSC|_TIF_BLOCKSTEP)
+	(_TIF_IO_BITMAP|_TIF_NOCPUID|_TIF_NOTSC|_TIF_BLOCKSTEP)
 
 #define _TIF_WORK_CTXSW_PREV (_TIF_WORK_CTXSW|_TIF_USER_RETURN_NOTIFY)
 #define _TIF_WORK_CTXSW_NEXT (_TIF_WORK_CTXSW)
 
 #define STACK_WARN		(THREAD_SIZE/8)
 
 /*
  * macros/functions for gaining access to the thread information structure
@@ -234,11 +236,13 @@ static inline int arch_within_stack_frames(const void * const stack,
  * EFLAGS values that other (fast) syscall return instructions
  * are not able to restore properly.
  */
 #define force_iret() set_thread_flag(TIF_NOTIFY_RESUME)
 
 extern void arch_task_cache_init(void);
 extern int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src);
 extern void arch_release_task_struct(struct task_struct *tsk);
+extern void arch_setup_new_exec(void);
+#define arch_setup_new_exec arch_setup_new_exec
 #endif	/* !__ASSEMBLY__ */
 
 #endif /* _ASM_X86_THREAD_INFO_H */
diff --git a/arch/x86/include/uapi/asm/prctl.h b/arch/x86/include/uapi/asm/prctl.h
index ae135de..80240a2 100644
--- a/arch/x86/include/uapi/asm/prctl.h
+++ b/arch/x86/include/uapi/asm/prctl.h
@@ -1,15 +1,18 @@
 #ifndef _ASM_X86_PRCTL_H
 #define _ASM_X86_PRCTL_H
 
 #define ARCH_SET_GS 0x1001
 #define ARCH_SET_FS 0x1002
 #define ARCH_GET_FS 0x1003
 #define ARCH_GET_GS 0x1004
 
+#define ARCH_GET_CPUID 0x1005
+#define ARCH_SET_CPUID 0x1006
+
 #ifdef CONFIG_CHECKPOINT_RESTORE
 # define ARCH_MAP_VDSO_X32	0x2001
 # define ARCH_MAP_VDSO_32	0x2002
 # define ARCH_MAP_VDSO_64	0x2003
 #endif
 
 #endif /* _ASM_X86_PRCTL_H */
diff --git a/arch/x86/kernel/cpu/intel.c b/arch/x86/kernel/cpu/intel.c
index 12ddddf..11e4a0a 100644
--- a/arch/x86/kernel/cpu/intel.c
+++ b/arch/x86/kernel/cpu/intel.c
@@ -451,16 +451,23 @@ static void intel_bsp_resume(struct cpuinfo_x86 *c)
 	 */
 	init_intel_energy_perf(c);
 }
 
 static void init_intel_misc_features(struct cpuinfo_x86 *c)
 {
 	u64 msr;
 
+	if (rdmsrl_safe(MSR_MISC_FEATURES_ENABLES, &msr))
+		return;
+
+	msr = 0;
+	wrmsrl(MSR_MISC_FEATURES_ENABLES, msr);
+	this_cpu_write(msr_misc_features_shadow, msr);
+
 	if (!rdmsrl_safe(MSR_PLATFORM_INFO, &msr)) {
 		if (msr & MSR_PLATFORM_INFO_CPUID_FAULT)
 			set_cpu_cap(c, X86_FEATURE_CPUID_FAULT);
 	}
 }
 
 static void init_intel(struct cpuinfo_x86 *c)
 {
diff --git a/arch/x86/kernel/process.c b/arch/x86/kernel/process.c
index 5ac20b5..ba39090 100644
--- a/arch/x86/kernel/process.c
+++ b/arch/x86/kernel/process.c
@@ -28,16 +28,17 @@
 #include <asm/mwait.h>
 #include <asm/fpu/internal.h>
 #include <asm/debugreg.h>
 #include <asm/nmi.h>
 #include <asm/tlbflush.h>
 #include <asm/mce.h>
 #include <asm/vm86.h>
 #include <asm/switch_to.h>
+#include <asm/prctl.h>
 
 /*
  * per-CPU TSS segments. Threads are completely 'soft' on Linux,
  * no more per-task TSS's. The TSS size is kept cacheline-aligned
  * so they are allowed to end up in the .data..cacheline_aligned
  * section. Since TSS's are completely CPU-local, we want them
  * on exact cacheline boundaries, to eliminate cacheline ping-pong.
  */
@@ -187,16 +188,85 @@ int set_tsc_mode(unsigned int val)
 	else if (val == PR_TSC_ENABLE)
 		enable_TSC();
 	else
 		return -EINVAL;
 
 	return 0;
 }
 
+DEFINE_PER_CPU(u64, msr_misc_features_shadow);
+
+static void set_cpuid_faulting(bool on)
+{
+	u64 msrval;
+
+	DEBUG_LOCKS_WARN_ON(!irqs_disabled());
+
+	msrval = this_cpu_read(msr_misc_features_shadow);
+	msrval &= ~MSR_MISC_FEATURES_ENABLES_CPUID_FAULT;
+	msrval |= (on << MSR_MISC_FEATURES_ENABLES_CPUID_FAULT_BIT);
+	this_cpu_write(msr_misc_features_shadow, msrval);
+	wrmsrl(MSR_MISC_FEATURES_ENABLES, msrval);
+}
+
+static void disable_cpuid(void)
+{
+	preempt_disable();
+	if (!test_and_set_thread_flag(TIF_NOCPUID)) {
+		/*
+		 * Must flip the CPU state synchronously with
+		 * TIF_NOCPUID in the current running context.
+		 */
+		set_cpuid_faulting(true);
+	}
+	preempt_enable();
+}
+
+static void enable_cpuid(void)
+{
+	preempt_disable();
+	if (test_and_clear_thread_flag(TIF_NOCPUID)) {
+		/*
+		 * Must flip the CPU state synchronously with
+		 * TIF_NOCPUID in the current running context.
+		 */
+		set_cpuid_faulting(false);
+	}
+	preempt_enable();
+}
+
+static int get_cpuid_mode(void)
+{
+	return !test_thread_flag(TIF_NOCPUID);
+}
+
+static int set_cpuid_mode(struct task_struct *task, unsigned long cpuid_enabled)
+{
+	if (!static_cpu_has(X86_FEATURE_CPUID_FAULT))
+		return -ENODEV;
+
+	if (cpuid_enabled)
+		enable_cpuid();
+	else
+		disable_cpuid();
+
+	return 0;
+}
+
+/*
+ * Called immediately after a successful exec.
+ */
+void arch_setup_new_exec(void)
+{
+	/* If cpuid was previously disabled for this task, re-enable it. */
+	if (test_thread_flag(TIF_NOCPUID))
+		enable_cpuid();
+}
+
 void __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p,
 		      struct tss_struct *tss)
 {
 	struct thread_struct *prev, *next;
 
 	prev = &prev_p->thread;
 	next = &next_p->thread;
 
@@ -206,16 +276,21 @@ void __switch_to_xtra(struct task_struct *prev_p, struct task_struct *next_p,
 
 		debugctl &= ~DEBUGCTLMSR_BTF;
 		if (test_tsk_thread_flag(next_p, TIF_BLOCKSTEP))
 			debugctl |= DEBUGCTLMSR_BTF;
 
 		update_debugctlmsr(debugctl);
 	}
 
+	if (test_tsk_thread_flag(prev_p, TIF_NOCPUID) ^
+	    test_tsk_thread_flag(next_p, TIF_NOCPUID)) {
+		set_cpuid_faulting(test_tsk_thread_flag(next_p, TIF_NOCPUID));
+	}
+
 	if (test_tsk_thread_flag(prev_p, TIF_NOTSC) ^
 	    test_tsk_thread_flag(next_p, TIF_NOTSC)) {
 		/* prev and next are different */
 		if (test_tsk_thread_flag(next_p, TIF_NOTSC))
 			hard_disable_TSC();
 		else
 			hard_enable_TSC();
 	}
@@ -583,10 +658,17 @@ unsigned long get_wchan(struct task_struct *p)
 out:
 	put_task_stack(p);
 	return ret;
 }
 
 long do_arch_prctl_common(struct task_struct *task, int code,
 			  unsigned long cpuid_enabled)
 {
+	switch (code) {
+	case ARCH_GET_CPUID:
+		return get_cpuid_mode();
+	case ARCH_SET_CPUID:
+		return set_cpuid_mode(task, cpuid_enabled);
+	}
+
 	return -EINVAL;
 }
diff --git a/fs/exec.c b/fs/exec.c
index 4e497b9..3e6872c 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1287,16 +1287,17 @@ void setup_new_exec(struct linux_binprm * bprm)
 	/* This is the point of no return */
 	current->sas_ss_sp = current->sas_ss_size = 0;
 
 	if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
 		set_dumpable(current->mm, SUID_DUMP_USER);
 	else
 		set_dumpable(current->mm, suid_dumpable);
 
+	arch_setup_new_exec();
 	perf_event_exec();
 	__set_task_comm(current, kbasename(bprm->filename), true);
 
 	/* Set the new mm task size. We have to do that late because it may
 	 * depend on TIF_32BIT which is only updated in flush_thread() on
 	 * some architectures like powerpc
 	 */
 	current->mm->task_size = TASK_SIZE;
diff --git a/include/linux/thread_info.h b/include/linux/thread_info.h
index 2873baf..d46f50d 100644
--- a/include/linux/thread_info.h
+++ b/include/linux/thread_info.h
@@ -129,11 +129,15 @@ static __always_inline void check_object_size(const void *ptr, unsigned long n,
 		__check_object_size(ptr, n, to_user);
 }
 #else
 static inline void check_object_size(const void *ptr, unsigned long n,
 				     bool to_user)
 { }
 #endif /* CONFIG_HARDENED_USERCOPY */
 
+#ifndef arch_setup_new_exec
+static inline void arch_setup_new_exec(void) {}
+#endif
+
 #endif	/* __KERNEL__ */
 
 #endif /* _LINUX_THREAD_INFO_H */
-- 
2.10.2

Re: [PATCH v13 6/8] x86/arch_prctl: Add ARCH_[GET|SET]_CPUID

[Thomas Gleixner]

On Sun, 27 Nov 2016, Kyle Huey wrote:
> +DEFINE_PER_CPU(u64, msr_misc_features_shadow);
> +
> +static void set_cpuid_faulting(bool on)
> +{
> +	u64 msrval;
> +
> +	DEBUG_LOCKS_WARN_ON(!irqs_disabled());

The arch_prctl code path does not have interrupts disabled and there is no
reason why it should have.

Thanks,

	tglx

Re: [PATCH v13 6/8] x86/arch_prctl: Add ARCH_[GET|SET]_CPUID

[Thomas Gleixner]

On Sun, 27 Nov 2016, Kyle Huey wrote:
> +DEFINE_PER_CPU(u64, msr_misc_features_shadow);
> +
> +static void set_cpuid_faulting(bool on)
> +{
> +	u64 msrval;
> +
> +	DEBUG_LOCKS_WARN_ON(!irqs_disabled());

The arch_prctl code path does not have interrupts disabled and there is no
reason why it should have.

Thanks,

	tglx

Re: [PATCH v13 6/8] x86/arch_prctl: Add ARCH_[GET|SET]_CPUID

[Kyle Huey]

On Mon, Nov 28, 2016 at 6:06 AM, Thomas Gleixner <tglx@linutronix.de> wrote:
> On Sun, 27 Nov 2016, Kyle Huey wrote:
>> +DEFINE_PER_CPU(u64, msr_misc_features_shadow);
>> +
>> +static void set_cpuid_faulting(bool on)
>> +{
>> +     u64 msrval;
>> +
>> +     DEBUG_LOCKS_WARN_ON(!irqs_disabled());
>
> The arch_prctl code path does not have interrupts disabled and there is no
> reason why it should have.

Yes, indeed. Removed.

- Kyle

Re: [PATCH v13 6/8] x86/arch_prctl: Add ARCH_[GET|SET]_CPUID

[Kyle Huey]

On Mon, Nov 28, 2016 at 6:06 AM, Thomas Gleixner <tglx@linutronix.de> wrote:
> On Sun, 27 Nov 2016, Kyle Huey wrote:
>> +DEFINE_PER_CPU(u64, msr_misc_features_shadow);
>> +
>> +static void set_cpuid_faulting(bool on)
>> +{
>> +     u64 msrval;
>> +
>> +     DEBUG_LOCKS_WARN_ON(!irqs_disabled());
>
> The arch_prctl code path does not have interrupts disabled and there is no
> reason why it should have.

Yes, indeed. Removed.

- Kyle

[PATCH v13 7/8] x86/arch_prctl: Selftest for ARCH_[GET|SET]_CPUID

[Kyle Huey]

Test disabling and reenabling the cpuid instruction via the new arch_prctl
ARCH_SET_CPUID, retrieving the current state via ARCH_GET_CPUID, and the
expected behaviors across fork() and exec().

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
---
 tools/testing/selftests/x86/Makefile      |   2 +-
 tools/testing/selftests/x86/cpuid-fault.c | 251 ++++++++++++++++++++++++++++++
 2 files changed, 252 insertions(+), 1 deletion(-)
 create mode 100644 tools/testing/selftests/x86/cpuid-fault.c

diff --git a/tools/testing/selftests/x86/Makefile b/tools/testing/selftests/x86/Makefile
index a89f80a..3744f28 100644
--- a/tools/testing/selftests/x86/Makefile
+++ b/tools/testing/selftests/x86/Makefile
@@ -1,17 +1,17 @@
 all:
 
 include ../lib.mk
 
 .PHONY: all all_32 all_64 warn_32bit_failure clean
 
 TARGETS_C_BOTHBITS := single_step_syscall sysret_ss_attrs syscall_nt ptrace_syscall test_mremap_vdso \
 			check_initial_reg_state sigreturn ldt_gdt iopl \
-			protection_keys
+			protection_keys cpuid-fault
 TARGETS_C_32BIT_ONLY := entry_from_vm86 syscall_arg_fault test_syscall_vdso unwind_vdso \
 			test_FCMOV test_FCOMI test_FISTTP \
 			vdso_restorer
 TARGETS_C_64BIT_ONLY := fsgsbase
 
 TARGETS_C_32BIT_ALL := $(TARGETS_C_BOTHBITS) $(TARGETS_C_32BIT_ONLY)
 TARGETS_C_64BIT_ALL := $(TARGETS_C_BOTHBITS) $(TARGETS_C_64BIT_ONLY)
 BINARIES_32 := $(TARGETS_C_32BIT_ALL:%=%_32)
diff --git a/tools/testing/selftests/x86/cpuid-fault.c b/tools/testing/selftests/x86/cpuid-fault.c
new file mode 100644
index 0000000..d061abd
--- /dev/null
+++ b/tools/testing/selftests/x86/cpuid-fault.c
@@ -0,0 +1,251 @@
+
+/*
+ * Tests for arch_prctl(ARCH_GET_CPUID, ...) / arch_prctl(ARCH_SET_CPUID, ...)
+ *
+ * Basic test to test behaviour of ARCH_GET_CPUID and ARCH_SET_CPUID
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <signal.h>
+#include <inttypes.h>
+#include <cpuid.h>
+#include <err.h>
+#include <errno.h>
+#include <sys/wait.h>
+
+#include <sys/prctl.h>
+#include <linux/prctl.h>
+
+/*
+#define ARCH_GET_CPUID 0x1005
+#define ARCH_SET_CPUID 0x1006
+#ifdef __x86_64__
+#define SYS_arch_prctl 158
+#else
+#define SYS_arch_prctl 385
+#endif
+*/
+
+const char *cpuid_names[] = {
+	[0] = "[cpuid disabled]",
+	[1] = "[cpuid enabled]",
+};
+
+int arch_prctl(int code, unsigned long arg2)
+{
+	return syscall(SYS_arch_prctl, code, arg2);
+}
+
+int cpuid(unsigned int *eax, unsigned int *ebx, unsigned int *ecx,
+	  unsigned int *edx)
+{
+	return __get_cpuid(0, eax, ebx, ecx, edx);
+}
+
+int do_child_exec_test(int eax, int ebx, int ecx, int edx)
+{
+	int cpuid_val = 0, child = 0, status = 0;
+
+	printf("arch_prctl(ARCH_GET_CPUID); ");
+
+	cpuid_val = arch_prctl(ARCH_GET_CPUID, 0);
+	if (cpuid_val < 0)
+		errx(1, "ARCH_GET_CPUID fails now, but not before?");
+
+	printf("cpuid_val == %s\n", cpuid_names[cpuid_val]);
+	if (cpuid_val != 0)
+		errx(1, "How did cpuid get re-enabled on fork?");
+
+	child = fork();
+	if (child == 0) {
+		cpuid_val = arch_prctl(ARCH_GET_CPUID, 0);
+		if (cpuid_val < 0)
+			errx(1, "ARCH_GET_CPUID fails now, but not before?");
+
+		printf("cpuid_val == %s\n", cpuid_names[cpuid_val]);
+		if (cpuid_val != 0)
+			errx(1, "How did cpuid get re-enabled on fork?");
+
+		printf("exec\n");
+		execl("/proc/self/exe", "cpuid-fault", "-early-return", NULL);
+	}
+
+	if (child != waitpid(child, &status, 0))
+		errx(1, "waitpid failed!?");
+
+	if (WEXITSTATUS(status) != 0)
+		errx(1, "Execed child exited abnormally");
+
+	return 0;
+}
+
+int child_received_signal;
+
+void child_sigsegv_cb(int sig)
+{
+	int cpuid_val = 0;
+
+	child_received_signal = 1;
+	printf("[ SIG_SEGV ]\n");
+	printf("arch_prctl(ARCH_GET_CPUID); ");
+
+	cpuid_val = arch_prctl(ARCH_GET_CPUID, 0);
+	if (cpuid_val < 0)
+		errx(1, "ARCH_GET_CPUID fails now, but not before?");
+
+	printf("cpuid_val == %s\n", cpuid_names[cpuid_val]);
+	printf("arch_prctl(ARCH_SET_CPUID, 1)\n");
+	if (arch_prctl(ARCH_SET_CPUID, 1) != 0)
+		exit(errno);
+
+	printf("cpuid() == ");
+}
+
+int do_child_test(void)
+{
+	unsigned int eax = 0, ebx = 0, ecx = 0, edx = 0;
+
+	signal(SIGSEGV, child_sigsegv_cb);
+
+	/* the child starts out with cpuid disabled, the signal handler
+	 * attempts to enable and retry
+	 */
+	printf("cpuid() == ");
+	cpuid(&eax, &ebx, &ecx, &edx);
+	printf("{%x, %x, %x, %x}\n", eax, ebx, ecx, edx);
+	return child_received_signal ? 0 : 42;
+}
+
+int signal_count;
+
+void sigsegv_cb(int sig)
+{
+	int cpuid_val = 0;
+
+	signal_count++;
+	printf("[ SIG_SEGV ]\n");
+	printf("arch_prctl(ARCH_GET_CPUID); ");
+
+	cpuid_val = arch_prctl(ARCH_GET_CPUID, 0);
+	if (cpuid_val < 0)
+		errx(1, "ARCH_GET_CPUID fails now, but not before?");
+
+	printf("cpuid_val == %s\n", cpuid_names[cpuid_val]);
+	printf("arch_prctl(ARC_SET_CPUID, 1)\n");
+	if (arch_prctl(ARCH_SET_CPUID, 1) != 0)
+		errx(1, "ARCH_SET_CPUID failed!");
+
+	printf("cpuid() == ");
+}
+
+int main(int argc, char **argv)
+{
+	int cpuid_val = 0, child = 0, status = 0;
+	unsigned int eax = 0, ebx = 0, ecx = 0, edx = 0;
+
+	signal(SIGSEGV, sigsegv_cb);
+	setvbuf(stdout, NULL, _IONBF, 0);
+
+	cpuid(&eax, &ebx, &ecx, &edx);
+	printf("cpuid() == {%x, %x, %x, %x}\n", eax, ebx, ecx, edx);
+	printf("arch_prctl(ARCH_GET_CPUID); ");
+
+	cpuid_val = arch_prctl(ARCH_GET_CPUID, 0);
+	if (cpuid_val < 0) {
+		if (errno == EINVAL) {
+			printf("ARCH_GET_CPUID is unsupported on this kernel.\n");
+			fflush(stdout);
+			exit(0); /* no ARCH_GET_CPUID on this system */
+		} else if (errno == ENODEV) {
+			printf("ARCH_GET_CPUID is unsupported on this hardware.\n");
+			fflush(stdout);
+			exit(0); /* no ARCH_GET_CPUID on this system */
+		} else {
+			errx(errno, "ARCH_GET_CPUID failed unexpectedly!");
+		}
+	}
+
+	printf("cpuid_val == %s\n", cpuid_names[cpuid_val]);
+	cpuid(&eax, &ebx, &ecx, &edx);
+	printf("cpuid() == {%x, %x, %x, %x}\n", eax, ebx, ecx, edx);
+	printf("arch_prctl(ARCH_SET_CPUID, 1)\n");
+
+	if (arch_prctl(ARCH_SET_CPUID, 1) != 0) {
+		if (errno == EINVAL) {
+			printf("ARCH_SET_CPUID is unsupported on this kernel.");
+			exit(0); /* no ARCH_SET_CPUID on this system */
+		} else if (errno == ENODEV) {
+			printf("ARCH_SET_CPUID is unsupported on this hardware.");
+			exit(0); /* no ARCH_SET_CPUID on this system */
+		} else {
+			errx(errno, "ARCH_SET_CPUID failed unexpectedly!");
+		}
+	}
+
+
+	cpuid(&eax, &ebx, &ecx, &edx);
+	printf("cpuid() == {%x, %x, %x, %x}\n", eax, ebx, ecx, edx);
+	printf("arch_prctl(ARCH_SET_CPUID, 0)\n");
+	fflush(stdout);
+
+	if (arch_prctl(ARCH_SET_CPUID, 0) == -1)
+		errx(1, "ARCH_SET_CPUID failed!");
+
+	printf("cpuid() == ");
+	eax = ebx = ecx = edx = 0;
+	cpuid(&eax, &ebx, &ecx, &edx);
+	printf("{%x, %x, %x, %x}\n", eax, ebx, ecx, edx);
+	printf("arch_prctl(ARCH_SET_CPUID, 0)\n");
+
+	if (signal_count != 1)
+		errx(1, "cpuid didn't fault!");
+
+	if (arch_prctl(ARCH_SET_CPUID, 0) == -1)
+		errx(1, "ARCH_SET_CPUID failed!");
+
+	if (argc > 1)
+		exit(0); /* Don't run the whole test again if we were execed */
+
+	printf("do_child_test\n");
+	child = fork();
+	if (child == 0)
+		return do_child_test();
+
+	if (child != waitpid(child, &status, 0))
+		errx(1, "waitpid failed!?");
+
+	if (WEXITSTATUS(status) != 0)
+		errx(1, "Child exited abnormally!");
+
+	/* The child enabling cpuid should not have affected us */
+	printf("cpuid() == ");
+	eax = ebx = ecx = edx = 0;
+	cpuid(&eax, &ebx, &ecx, &edx);
+	printf("{%x, %x, %x, %x}\n", eax, ebx, ecx, edx);
+	printf("arch_prctl(ARCH_SET_CPUID, 0)\n");
+
+	if (signal_count != 2)
+		errx(1, "cpuid didn't fault!");
+
+	if (arch_prctl(ARCH_SET_CPUID, 0) == -1)
+		errx(1, "ARCH_SET_CPUID failed!");
+
+	/* Our ARCH_CPUID_SIGSEGV should not propagate through exec */
+	printf("do_child_exec_test\n");
+	fflush(stdout);
+
+	child = fork();
+	if (child == 0)
+		return do_child_exec_test(eax, ebx, ecx, edx);
+
+	if (child != waitpid(child, &status, 0))
+		errx(1, "waitpid failed!?");
+
+	if (WEXITSTATUS(status) != 0)
+		errx(1, "Child exited abnormally!");
+
+	printf("All tests passed!\n");
+	exit(EXIT_SUCCESS);
+}
-- 
2.10.2

[PATCH v13 7/8] x86/arch_prctl: Selftest for ARCH_[GET|SET]_CPUID

[Kyle Huey]

Test disabling and reenabling the cpuid instruction via the new arch_prctl
ARCH_SET_CPUID, retrieving the current state via ARCH_GET_CPUID, and the
expected behaviors across fork() and exec().

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
---
 tools/testing/selftests/x86/Makefile      |   2 +-
 tools/testing/selftests/x86/cpuid-fault.c | 251 ++++++++++++++++++++++++++++++
 2 files changed, 252 insertions(+), 1 deletion(-)
 create mode 100644 tools/testing/selftests/x86/cpuid-fault.c

diff --git a/tools/testing/selftests/x86/Makefile b/tools/testing/selftests/x86/Makefile
index a89f80a..3744f28 100644
--- a/tools/testing/selftests/x86/Makefile
+++ b/tools/testing/selftests/x86/Makefile
@@ -1,17 +1,17 @@
 all:
 
 include ../lib.mk
 
 .PHONY: all all_32 all_64 warn_32bit_failure clean
 
 TARGETS_C_BOTHBITS := single_step_syscall sysret_ss_attrs syscall_nt ptrace_syscall test_mremap_vdso \
 			check_initial_reg_state sigreturn ldt_gdt iopl \
-			protection_keys
+			protection_keys cpuid-fault
 TARGETS_C_32BIT_ONLY := entry_from_vm86 syscall_arg_fault test_syscall_vdso unwind_vdso \
 			test_FCMOV test_FCOMI test_FISTTP \
 			vdso_restorer
 TARGETS_C_64BIT_ONLY := fsgsbase
 
 TARGETS_C_32BIT_ALL := $(TARGETS_C_BOTHBITS) $(TARGETS_C_32BIT_ONLY)
 TARGETS_C_64BIT_ALL := $(TARGETS_C_BOTHBITS) $(TARGETS_C_64BIT_ONLY)
 BINARIES_32 := $(TARGETS_C_32BIT_ALL:%=%_32)
diff --git a/tools/testing/selftests/x86/cpuid-fault.c b/tools/testing/selftests/x86/cpuid-fault.c
new file mode 100644
index 0000000..d061abd
--- /dev/null
+++ b/tools/testing/selftests/x86/cpuid-fault.c
@@ -0,0 +1,251 @@
+
+/*
+ * Tests for arch_prctl(ARCH_GET_CPUID, ...) / arch_prctl(ARCH_SET_CPUID, ...)
+ *
+ * Basic test to test behaviour of ARCH_GET_CPUID and ARCH_SET_CPUID
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <signal.h>
+#include <inttypes.h>
+#include <cpuid.h>
+#include <err.h>
+#include <errno.h>
+#include <sys/wait.h>
+
+#include <sys/prctl.h>
+#include <linux/prctl.h>
+
+/*
+#define ARCH_GET_CPUID 0x1005
+#define ARCH_SET_CPUID 0x1006
+#ifdef __x86_64__
+#define SYS_arch_prctl 158
+#else
+#define SYS_arch_prctl 385
+#endif
+*/
+
+const char *cpuid_names[] = {
+	[0] = "[cpuid disabled]",
+	[1] = "[cpuid enabled]",
+};
+
+int arch_prctl(int code, unsigned long arg2)
+{
+	return syscall(SYS_arch_prctl, code, arg2);
+}
+
+int cpuid(unsigned int *eax, unsigned int *ebx, unsigned int *ecx,
+	  unsigned int *edx)
+{
+	return __get_cpuid(0, eax, ebx, ecx, edx);
+}
+
+int do_child_exec_test(int eax, int ebx, int ecx, int edx)
+{
+	int cpuid_val = 0, child = 0, status = 0;
+
+	printf("arch_prctl(ARCH_GET_CPUID); ");
+
+	cpuid_val = arch_prctl(ARCH_GET_CPUID, 0);
+	if (cpuid_val < 0)
+		errx(1, "ARCH_GET_CPUID fails now, but not before?");
+
+	printf("cpuid_val == %s\n", cpuid_names[cpuid_val]);
+	if (cpuid_val != 0)
+		errx(1, "How did cpuid get re-enabled on fork?");
+
+	child = fork();
+	if (child == 0) {
+		cpuid_val = arch_prctl(ARCH_GET_CPUID, 0);
+		if (cpuid_val < 0)
+			errx(1, "ARCH_GET_CPUID fails now, but not before?");
+
+		printf("cpuid_val == %s\n", cpuid_names[cpuid_val]);
+		if (cpuid_val != 0)
+			errx(1, "How did cpuid get re-enabled on fork?");
+
+		printf("exec\n");
+		execl("/proc/self/exe", "cpuid-fault", "-early-return", NULL);
+	}
+
+	if (child != waitpid(child, &status, 0))
+		errx(1, "waitpid failed!?");
+
+	if (WEXITSTATUS(status) != 0)
+		errx(1, "Execed child exited abnormally");
+
+	return 0;
+}
+
+int child_received_signal;
+
+void child_sigsegv_cb(int sig)
+{
+	int cpuid_val = 0;
+
+	child_received_signal = 1;
+	printf("[ SIG_SEGV ]\n");
+	printf("arch_prctl(ARCH_GET_CPUID); ");
+
+	cpuid_val = arch_prctl(ARCH_GET_CPUID, 0);
+	if (cpuid_val < 0)
+		errx(1, "ARCH_GET_CPUID fails now, but not before?");
+
+	printf("cpuid_val == %s\n", cpuid_names[cpuid_val]);
+	printf("arch_prctl(ARCH_SET_CPUID, 1)\n");
+	if (arch_prctl(ARCH_SET_CPUID, 1) != 0)
+		exit(errno);
+
+	printf("cpuid() == ");
+}
+
+int do_child_test(void)
+{
+	unsigned int eax = 0, ebx = 0, ecx = 0, edx = 0;
+
+	signal(SIGSEGV, child_sigsegv_cb);
+
+	/* the child starts out with cpuid disabled, the signal handler
+	 * attempts to enable and retry
+	 */
+	printf("cpuid() == ");
+	cpuid(&eax, &ebx, &ecx, &edx);
+	printf("{%x, %x, %x, %x}\n", eax, ebx, ecx, edx);
+	return child_received_signal ? 0 : 42;
+}
+
+int signal_count;
+
+void sigsegv_cb(int sig)
+{
+	int cpuid_val = 0;
+
+	signal_count++;
+	printf("[ SIG_SEGV ]\n");
+	printf("arch_prctl(ARCH_GET_CPUID); ");
+
+	cpuid_val = arch_prctl(ARCH_GET_CPUID, 0);
+	if (cpuid_val < 0)
+		errx(1, "ARCH_GET_CPUID fails now, but not before?");
+
+	printf("cpuid_val == %s\n", cpuid_names[cpuid_val]);
+	printf("arch_prctl(ARC_SET_CPUID, 1)\n");
+	if (arch_prctl(ARCH_SET_CPUID, 1) != 0)
+		errx(1, "ARCH_SET_CPUID failed!");
+
+	printf("cpuid() == ");
+}
+
+int main(int argc, char **argv)
+{
+	int cpuid_val = 0, child = 0, status = 0;
+	unsigned int eax = 0, ebx = 0, ecx = 0, edx = 0;
+
+	signal(SIGSEGV, sigsegv_cb);
+	setvbuf(stdout, NULL, _IONBF, 0);
+
+	cpuid(&eax, &ebx, &ecx, &edx);
+	printf("cpuid() == {%x, %x, %x, %x}\n", eax, ebx, ecx, edx);
+	printf("arch_prctl(ARCH_GET_CPUID); ");
+
+	cpuid_val = arch_prctl(ARCH_GET_CPUID, 0);
+	if (cpuid_val < 0) {
+		if (errno == EINVAL) {
+			printf("ARCH_GET_CPUID is unsupported on this kernel.\n");
+			fflush(stdout);
+			exit(0); /* no ARCH_GET_CPUID on this system */
+		} else if (errno == ENODEV) {
+			printf("ARCH_GET_CPUID is unsupported on this hardware.\n");
+			fflush(stdout);
+			exit(0); /* no ARCH_GET_CPUID on this system */
+		} else {
+			errx(errno, "ARCH_GET_CPUID failed unexpectedly!");
+		}
+	}
+
+	printf("cpuid_val == %s\n", cpuid_names[cpuid_val]);
+	cpuid(&eax, &ebx, &ecx, &edx);
+	printf("cpuid() == {%x, %x, %x, %x}\n", eax, ebx, ecx, edx);
+	printf("arch_prctl(ARCH_SET_CPUID, 1)\n");
+
+	if (arch_prctl(ARCH_SET_CPUID, 1) != 0) {
+		if (errno == EINVAL) {
+			printf("ARCH_SET_CPUID is unsupported on this kernel.");
+			exit(0); /* no ARCH_SET_CPUID on this system */
+		} else if (errno == ENODEV) {
+			printf("ARCH_SET_CPUID is unsupported on this hardware.");
+			exit(0); /* no ARCH_SET_CPUID on this system */
+		} else {
+			errx(errno, "ARCH_SET_CPUID failed unexpectedly!");
+		}
+	}
+
+
+	cpuid(&eax, &ebx, &ecx, &edx);
+	printf("cpuid() == {%x, %x, %x, %x}\n", eax, ebx, ecx, edx);
+	printf("arch_prctl(ARCH_SET_CPUID, 0)\n");
+	fflush(stdout);
+
+	if (arch_prctl(ARCH_SET_CPUID, 0) == -1)
+		errx(1, "ARCH_SET_CPUID failed!");
+
+	printf("cpuid() == ");
+	eax = ebx = ecx = edx = 0;
+	cpuid(&eax, &ebx, &ecx, &edx);
+	printf("{%x, %x, %x, %x}\n", eax, ebx, ecx, edx);
+	printf("arch_prctl(ARCH_SET_CPUID, 0)\n");
+
+	if (signal_count != 1)
+		errx(1, "cpuid didn't fault!");
+
+	if (arch_prctl(ARCH_SET_CPUID, 0) == -1)
+		errx(1, "ARCH_SET_CPUID failed!");
+
+	if (argc > 1)
+		exit(0); /* Don't run the whole test again if we were execed */
+
+	printf("do_child_test\n");
+	child = fork();
+	if (child == 0)
+		return do_child_test();
+
+	if (child != waitpid(child, &status, 0))
+		errx(1, "waitpid failed!?");
+
+	if (WEXITSTATUS(status) != 0)
+		errx(1, "Child exited abnormally!");
+
+	/* The child enabling cpuid should not have affected us */
+	printf("cpuid() == ");
+	eax = ebx = ecx = edx = 0;
+	cpuid(&eax, &ebx, &ecx, &edx);
+	printf("{%x, %x, %x, %x}\n", eax, ebx, ecx, edx);
+	printf("arch_prctl(ARCH_SET_CPUID, 0)\n");
+
+	if (signal_count != 2)
+		errx(1, "cpuid didn't fault!");
+
+	if (arch_prctl(ARCH_SET_CPUID, 0) == -1)
+		errx(1, "ARCH_SET_CPUID failed!");
+
+	/* Our ARCH_CPUID_SIGSEGV should not propagate through exec */
+	printf("do_child_exec_test\n");
+	fflush(stdout);
+
+	child = fork();
+	if (child == 0)
+		return do_child_exec_test(eax, ebx, ecx, edx);
+
+	if (child != waitpid(child, &status, 0))
+		errx(1, "waitpid failed!?");
+
+	if (WEXITSTATUS(status) != 0)
+		errx(1, "Child exited abnormally!");
+
+	printf("All tests passed!\n");
+	exit(EXIT_SUCCESS);
+}
-- 
2.10.2

[PATCH v13 8/8] KVM: x86: virtualize cpuid faulting

[Kyle Huey]

Hardware support for faulting on the cpuid instruction is not required to
emulate it, because cpuid triggers a VM exit anyways. KVM handles the relevant
MSRs (MSR_PLATFORM_INFO and MSR_MISC_FEATURES_ENABLE) and upon a
cpuid-induced VM exit checks the cpuid faulting state and the CPL.
kvm_require_cpl is even kind enough to inject the GP fault for us.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
Reviewed-by: David Matlack <dmatlack@google.com>
---
 arch/x86/include/asm/kvm_host.h |  2 ++
 arch/x86/kvm/cpuid.c            |  3 +++
 arch/x86/kvm/cpuid.h            | 11 +++++++++++
 arch/x86/kvm/emulate.c          |  7 +++++++
 arch/x86/kvm/x86.c              | 26 ++++++++++++++++++++++++++
 5 files changed, 49 insertions(+)

diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index bdde807..954f37c 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -592,16 +592,18 @@ struct kvm_vcpu_arch {
 	u64 pat;
 
 	unsigned switch_db_regs;
 	unsigned long db[KVM_NR_DB_REGS];
 	unsigned long dr6;
 	unsigned long dr7;
 	unsigned long eff_db[KVM_NR_DB_REGS];
 	unsigned long guest_debug_dr7;
+	u64 msr_platform_info;
+	u64 msr_misc_features_enables;
 
 	u64 mcg_cap;
 	u64 mcg_status;
 	u64 mcg_ctl;
 	u64 mcg_ext_ctl;
 	u64 *mce_banks;
 
 	/* Cache MMIO info */
diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
index afa7bbb..0109bc0 100644
--- a/arch/x86/kvm/cpuid.c
+++ b/arch/x86/kvm/cpuid.c
@@ -862,16 +862,19 @@ void kvm_cpuid(struct kvm_vcpu *vcpu, u32 *eax, u32 *ebx, u32 *ecx, u32 *edx)
 	trace_kvm_cpuid(function, *eax, *ebx, *ecx, *edx);
 }
 EXPORT_SYMBOL_GPL(kvm_cpuid);
 
 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu)
 {
 	u32 function, eax, ebx, ecx, edx;
 
+	if (cpuid_fault_enabled(vcpu) && !kvm_require_cpl(vcpu, 0))
+		return;
+
 	function = eax = kvm_register_read(vcpu, VCPU_REGS_RAX);
 	ecx = kvm_register_read(vcpu, VCPU_REGS_RCX);
 	kvm_cpuid(vcpu, &eax, &ebx, &ecx, &edx);
 	kvm_register_write(vcpu, VCPU_REGS_RAX, eax);
 	kvm_register_write(vcpu, VCPU_REGS_RBX, ebx);
 	kvm_register_write(vcpu, VCPU_REGS_RCX, ecx);
 	kvm_register_write(vcpu, VCPU_REGS_RDX, edx);
 	kvm_x86_ops->skip_emulated_instruction(vcpu);
diff --git a/arch/x86/kvm/cpuid.h b/arch/x86/kvm/cpuid.h
index 35058c2..a6fd40a 100644
--- a/arch/x86/kvm/cpuid.h
+++ b/arch/x86/kvm/cpuid.h
@@ -200,9 +200,20 @@ static inline int guest_cpuid_stepping(struct kvm_vcpu *vcpu)
 
 	best = kvm_find_cpuid_entry(vcpu, 0x1, 0);
 	if (!best)
 		return -1;
 
 	return x86_stepping(best->eax);
 }
 
+static inline bool supports_cpuid_fault(struct kvm_vcpu *vcpu)
+{
+	return vcpu->arch.msr_platform_info & MSR_PLATFORM_INFO_CPUID_FAULT;
+}
+
+static inline bool cpuid_fault_enabled(struct kvm_vcpu *vcpu)
+{
+	return vcpu->arch.msr_misc_features_enables &
+		  MSR_MISC_FEATURES_ENABLES_CPUID_FAULT;
+}
+
 #endif
diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
index a3ce9d2..7939bd0 100644
--- a/arch/x86/kvm/emulate.c
+++ b/arch/x86/kvm/emulate.c
@@ -3779,16 +3779,23 @@ static int em_sti(struct x86_emulate_ctxt *ctxt)
 	ctxt->interruptibility = KVM_X86_SHADOW_INT_STI;
 	ctxt->eflags |= X86_EFLAGS_IF;
 	return X86EMUL_CONTINUE;
 }
 
 static int em_cpuid(struct x86_emulate_ctxt *ctxt)
 {
 	u32 eax, ebx, ecx, edx;
+	u64 msr = 0;
+
+	ctxt->ops->get_msr(ctxt, MSR_MISC_FEATURES_ENABLES, &msr);
+	if (msr & MSR_MISC_FEATURES_ENABLES_CPUID_FAULT &&
+	    ctxt->ops->cpl(ctxt)) {
+		return emulate_gp(ctxt, 0);
+	}
 
 	eax = reg_read(ctxt, VCPU_REGS_RAX);
 	ecx = reg_read(ctxt, VCPU_REGS_RCX);
 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
 	*reg_write(ctxt, VCPU_REGS_RAX) = eax;
 	*reg_write(ctxt, VCPU_REGS_RBX) = ebx;
 	*reg_write(ctxt, VCPU_REGS_RCX) = ecx;
 	*reg_write(ctxt, VCPU_REGS_RDX) = edx;
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 04c5d96..da50a3f 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -995,16 +995,18 @@ static u32 emulated_msrs[] = {
 
 	MSR_IA32_TSC_ADJUST,
 	MSR_IA32_TSCDEADLINE,
 	MSR_IA32_MISC_ENABLE,
 	MSR_IA32_MCG_STATUS,
 	MSR_IA32_MCG_CTL,
 	MSR_IA32_MCG_EXT_CTL,
 	MSR_IA32_SMBASE,
+	MSR_PLATFORM_INFO,
+	MSR_MISC_FEATURES_ENABLES,
 };
 
 static unsigned num_emulated_msrs;
 
 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer)
 {
 	if (efer & efer_reserved_bits)
 		return false;
@@ -2283,16 +2285,31 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 			return 1;
 		vcpu->arch.osvw.length = data;
 		break;
 	case MSR_AMD64_OSVW_STATUS:
 		if (!guest_cpuid_has_osvw(vcpu))
 			return 1;
 		vcpu->arch.osvw.status = data;
 		break;
+	case MSR_PLATFORM_INFO:
+		if (!msr_info->host_initiated ||
+		    data & ~MSR_PLATFORM_INFO_CPUID_FAULT ||
+		    (!(data & MSR_PLATFORM_INFO_CPUID_FAULT) &&
+		     cpuid_fault_enabled(vcpu)))
+			return 1;
+		vcpu->arch.msr_platform_info = data;
+		break;
+	case MSR_MISC_FEATURES_ENABLES:
+		if (data & ~MSR_MISC_FEATURES_ENABLES_CPUID_FAULT ||
+		    (data & MSR_MISC_FEATURES_ENABLES_CPUID_FAULT &&
+		     !supports_cpuid_fault(vcpu)))
+			return 1;
+		vcpu->arch.msr_misc_features_enables = data;
+		break;
 	default:
 		if (msr && (msr == vcpu->kvm->arch.xen_hvm_config.msr))
 			return xen_hvm_config(vcpu, data);
 		if (kvm_pmu_is_valid_msr(vcpu, msr))
 			return kvm_pmu_set_msr(vcpu, msr_info);
 		if (!ignore_msrs) {
 			vcpu_unimpl(vcpu, "unhandled wrmsr: 0x%x data 0x%llx\n",
 				    msr, data);
@@ -2497,16 +2514,22 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 			return 1;
 		msr_info->data = vcpu->arch.osvw.length;
 		break;
 	case MSR_AMD64_OSVW_STATUS:
 		if (!guest_cpuid_has_osvw(vcpu))
 			return 1;
 		msr_info->data = vcpu->arch.osvw.status;
 		break;
+	case MSR_PLATFORM_INFO:
+		msr_info->data = vcpu->arch.msr_platform_info;
+		break;
+	case MSR_MISC_FEATURES_ENABLES:
+		msr_info->data = vcpu->arch.msr_misc_features_enables;
+		break;
 	default:
 		if (kvm_pmu_is_valid_msr(vcpu, msr_info->index))
 			return kvm_pmu_get_msr(vcpu, msr_info->index, &msr_info->data);
 		if (!ignore_msrs) {
 			vcpu_unimpl(vcpu, "unhandled rdmsr: 0x%x\n", msr_info->index);
 			return 1;
 		} else {
 			vcpu_unimpl(vcpu, "ignored rdmsr: 0x%x\n", msr_info->index);
@@ -7529,16 +7552,19 @@ void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event)
 
 	kvm_clear_async_pf_completion_queue(vcpu);
 	kvm_async_pf_hash_reset(vcpu);
 	vcpu->arch.apf.halted = false;
 
 	if (!init_event) {
 		kvm_pmu_reset(vcpu);
 		vcpu->arch.smbase = 0x30000;
+
+		vcpu->arch.msr_platform_info = MSR_PLATFORM_INFO_CPUID_FAULT;
+		vcpu->arch.msr_misc_features_enables = 0;
 	}
 
 	memset(vcpu->arch.regs, 0, sizeof(vcpu->arch.regs));
 	vcpu->arch.regs_avail = ~0;
 	vcpu->arch.regs_dirty = ~0;
 
 	kvm_x86_ops->vcpu_reset(vcpu, init_event);
 }
-- 
2.10.2

[PATCH v13 8/8] KVM: x86: virtualize cpuid faulting

[Kyle Huey]

Hardware support for faulting on the cpuid instruction is not required to
emulate it, because cpuid triggers a VM exit anyways. KVM handles the relevant
MSRs (MSR_PLATFORM_INFO and MSR_MISC_FEATURES_ENABLE) and upon a
cpuid-induced VM exit checks the cpuid faulting state and the CPL.
kvm_require_cpl is even kind enough to inject the GP fault for us.

Signed-off-by: Kyle Huey <khuey@kylehuey.com>
Reviewed-by: David Matlack <dmatlack@google.com>
---
 arch/x86/include/asm/kvm_host.h |  2 ++
 arch/x86/kvm/cpuid.c            |  3 +++
 arch/x86/kvm/cpuid.h            | 11 +++++++++++
 arch/x86/kvm/emulate.c          |  7 +++++++
 arch/x86/kvm/x86.c              | 26 ++++++++++++++++++++++++++
 5 files changed, 49 insertions(+)

diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index bdde807..954f37c 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -592,16 +592,18 @@ struct kvm_vcpu_arch {
 	u64 pat;
 
 	unsigned switch_db_regs;
 	unsigned long db[KVM_NR_DB_REGS];
 	unsigned long dr6;
 	unsigned long dr7;
 	unsigned long eff_db[KVM_NR_DB_REGS];
 	unsigned long guest_debug_dr7;
+	u64 msr_platform_info;
+	u64 msr_misc_features_enables;
 
 	u64 mcg_cap;
 	u64 mcg_status;
 	u64 mcg_ctl;
 	u64 mcg_ext_ctl;
 	u64 *mce_banks;
 
 	/* Cache MMIO info */
diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
index afa7bbb..0109bc0 100644
--- a/arch/x86/kvm/cpuid.c
+++ b/arch/x86/kvm/cpuid.c
@@ -862,16 +862,19 @@ void kvm_cpuid(struct kvm_vcpu *vcpu, u32 *eax, u32 *ebx, u32 *ecx, u32 *edx)
 	trace_kvm_cpuid(function, *eax, *ebx, *ecx, *edx);
 }
 EXPORT_SYMBOL_GPL(kvm_cpuid);
 
 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu)
 {
 	u32 function, eax, ebx, ecx, edx;
 
+	if (cpuid_fault_enabled(vcpu) && !kvm_require_cpl(vcpu, 0))
+		return;
+
 	function = eax = kvm_register_read(vcpu, VCPU_REGS_RAX);
 	ecx = kvm_register_read(vcpu, VCPU_REGS_RCX);
 	kvm_cpuid(vcpu, &eax, &ebx, &ecx, &edx);
 	kvm_register_write(vcpu, VCPU_REGS_RAX, eax);
 	kvm_register_write(vcpu, VCPU_REGS_RBX, ebx);
 	kvm_register_write(vcpu, VCPU_REGS_RCX, ecx);
 	kvm_register_write(vcpu, VCPU_REGS_RDX, edx);
 	kvm_x86_ops->skip_emulated_instruction(vcpu);
diff --git a/arch/x86/kvm/cpuid.h b/arch/x86/kvm/cpuid.h
index 35058c2..a6fd40a 100644
--- a/arch/x86/kvm/cpuid.h
+++ b/arch/x86/kvm/cpuid.h
@@ -200,9 +200,20 @@ static inline int guest_cpuid_stepping(struct kvm_vcpu *vcpu)
 
 	best = kvm_find_cpuid_entry(vcpu, 0x1, 0);
 	if (!best)
 		return -1;
 
 	return x86_stepping(best->eax);
 }
 
+static inline bool supports_cpuid_fault(struct kvm_vcpu *vcpu)
+{
+	return vcpu->arch.msr_platform_info & MSR_PLATFORM_INFO_CPUID_FAULT;
+}
+
+static inline bool cpuid_fault_enabled(struct kvm_vcpu *vcpu)
+{
+	return vcpu->arch.msr_misc_features_enables &
+		  MSR_MISC_FEATURES_ENABLES_CPUID_FAULT;
+}
+
 #endif
diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
index a3ce9d2..7939bd0 100644
--- a/arch/x86/kvm/emulate.c
+++ b/arch/x86/kvm/emulate.c
@@ -3779,16 +3779,23 @@ static int em_sti(struct x86_emulate_ctxt *ctxt)
 	ctxt->interruptibility = KVM_X86_SHADOW_INT_STI;
 	ctxt->eflags |= X86_EFLAGS_IF;
 	return X86EMUL_CONTINUE;
 }
 
 static int em_cpuid(struct x86_emulate_ctxt *ctxt)
 {
 	u32 eax, ebx, ecx, edx;
+	u64 msr = 0;
+
+	ctxt->ops->get_msr(ctxt, MSR_MISC_FEATURES_ENABLES, &msr);
+	if (msr & MSR_MISC_FEATURES_ENABLES_CPUID_FAULT &&
+	    ctxt->ops->cpl(ctxt)) {
+		return emulate_gp(ctxt, 0);
+	}
 
 	eax = reg_read(ctxt, VCPU_REGS_RAX);
 	ecx = reg_read(ctxt, VCPU_REGS_RCX);
 	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
 	*reg_write(ctxt, VCPU_REGS_RAX) = eax;
 	*reg_write(ctxt, VCPU_REGS_RBX) = ebx;
 	*reg_write(ctxt, VCPU_REGS_RCX) = ecx;
 	*reg_write(ctxt, VCPU_REGS_RDX) = edx;
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 04c5d96..da50a3f 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -995,16 +995,18 @@ static u32 emulated_msrs[] = {
 
 	MSR_IA32_TSC_ADJUST,
 	MSR_IA32_TSCDEADLINE,
 	MSR_IA32_MISC_ENABLE,
 	MSR_IA32_MCG_STATUS,
 	MSR_IA32_MCG_CTL,
 	MSR_IA32_MCG_EXT_CTL,
 	MSR_IA32_SMBASE,
+	MSR_PLATFORM_INFO,
+	MSR_MISC_FEATURES_ENABLES,
 };
 
 static unsigned num_emulated_msrs;
 
 bool kvm_valid_efer(struct kvm_vcpu *vcpu, u64 efer)
 {
 	if (efer & efer_reserved_bits)
 		return false;
@@ -2283,16 +2285,31 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 			return 1;
 		vcpu->arch.osvw.length = data;
 		break;
 	case MSR_AMD64_OSVW_STATUS:
 		if (!guest_cpuid_has_osvw(vcpu))
 			return 1;
 		vcpu->arch.osvw.status = data;
 		break;
+	case MSR_PLATFORM_INFO:
+		if (!msr_info->host_initiated ||
+		    data & ~MSR_PLATFORM_INFO_CPUID_FAULT ||
+		    (!(data & MSR_PLATFORM_INFO_CPUID_FAULT) &&
+		     cpuid_fault_enabled(vcpu)))
+			return 1;
+		vcpu->arch.msr_platform_info = data;
+		break;
+	case MSR_MISC_FEATURES_ENABLES:
+		if (data & ~MSR_MISC_FEATURES_ENABLES_CPUID_FAULT ||
+		    (data & MSR_MISC_FEATURES_ENABLES_CPUID_FAULT &&
+		     !supports_cpuid_fault(vcpu)))
+			return 1;
+		vcpu->arch.msr_misc_features_enables = data;
+		break;
 	default:
 		if (msr && (msr == vcpu->kvm->arch.xen_hvm_config.msr))
 			return xen_hvm_config(vcpu, data);
 		if (kvm_pmu_is_valid_msr(vcpu, msr))
 			return kvm_pmu_set_msr(vcpu, msr_info);
 		if (!ignore_msrs) {
 			vcpu_unimpl(vcpu, "unhandled wrmsr: 0x%x data 0x%llx\n",
 				    msr, data);
@@ -2497,16 +2514,22 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
 			return 1;
 		msr_info->data = vcpu->arch.osvw.length;
 		break;
 	case MSR_AMD64_OSVW_STATUS:
 		if (!guest_cpuid_has_osvw(vcpu))
 			return 1;
 		msr_info->data = vcpu->arch.osvw.status;
 		break;
+	case MSR_PLATFORM_INFO:
+		msr_info->data = vcpu->arch.msr_platform_info;
+		break;
+	case MSR_MISC_FEATURES_ENABLES:
+		msr_info->data = vcpu->arch.msr_misc_features_enables;
+		break;
 	default:
 		if (kvm_pmu_is_valid_msr(vcpu, msr_info->index))
 			return kvm_pmu_get_msr(vcpu, msr_info->index, &msr_info->data);
 		if (!ignore_msrs) {
 			vcpu_unimpl(vcpu, "unhandled rdmsr: 0x%x\n", msr_info->index);
 			return 1;
 		} else {
 			vcpu_unimpl(vcpu, "ignored rdmsr: 0x%x\n", msr_info->index);
@@ -7529,16 +7552,19 @@ void kvm_vcpu_reset(struct kvm_vcpu *vcpu, bool init_event)
 
 	kvm_clear_async_pf_completion_queue(vcpu);
 	kvm_async_pf_hash_reset(vcpu);
 	vcpu->arch.apf.halted = false;
 
 	if (!init_event) {
 		kvm_pmu_reset(vcpu);
 		vcpu->arch.smbase = 0x30000;
+
+		vcpu->arch.msr_platform_info = MSR_PLATFORM_INFO_CPUID_FAULT;
+		vcpu->arch.msr_misc_features_enables = 0;
 	}
 
 	memset(vcpu->arch.regs, 0, sizeof(vcpu->arch.regs));
 	vcpu->arch.regs_avail = ~0;
 	vcpu->arch.regs_dirty = ~0;
 
 	kvm_x86_ops->vcpu_reset(vcpu, init_event);
 }
-- 
2.10.2

Re: [PATCH v13 0/8] x86/arch_prctl Add ARCH_[GET|SET]_CPUID for controlling the CPUID instruction

[Ingo Molnar]

* Kyle Huey <me@kylehuey.com> wrote:

> rr (http://rr-project.org/), a userspace record-and-replay reverse-
> execution debugger, would like to trap and emulate the CPUID instruction.
> This would allow us to a) mask away certain hardware features that rr does
> not support (e.g. RDRAND) and b) enable trace portability across machines
> by providing constant results.
> 
> Newer Intel CPUs (Ivy Bridge and later) can fault when CPUID is executed at
> CPL > 0. Expose this capability to userspace as a new pair of arch_prctls,
> ARCH_GET_CPUID and ARCH_SET_CPUID.
> 
> Since v12:
> Patch 4: x86/syscalls/32: Wire up arch_prctl on x86-32
> - compat_sys_arch_prctl prototype has argument names.

So while I am fine with the feature, I'm still unconvinced about the 
implementation:

1)

As I pointed out before, the arbitrary 'code' argument name x86-ism should be 
changed to 'option' like the canonical core kernel option name is for prctls(). 

This is still unfixed.

2)

As I complained about in my first review, TIF_NOCPUID flag is too far removed from 
the value what will be written into the MSR.

The result is poor code generation on 64-bit defconfig+CONFIG_PREEMPT=y:

        if (test_tsk_thread_flag(prev_p, TIF_NOCPUID) ^
            test_tsk_thread_flag(next_p, TIF_NOCPUID)) {
		set_cpuid_faulting(test_tsk_thread_flag(next_p, TIF_NOCPUID));

is compiled as:

  476:   49 8b 06                mov    (%r14),%rax
  479:   49 8b 55 00             mov    0x0(%r13),%rdx
  47d:   48 c1 e8 0f             shr    $0xf,%rax
  481:   48 c1 ea 0f             shr    $0xf,%rdx
  485:   83 e2 01                and    $0x1,%edx
  488:   83 e0 01                and    $0x1,%eax
  48b:   38 c2                   cmp    %al,%dl
  48d:   74 10                   je     49f <__switch_to_xtra+0x9f>
  48f:   49 8b 7d 00             mov    0x0(%r13),%rdi
  493:   48 c1 ef 0f             shr    $0xf,%rdi
  497:   83 e7 01                and    $0x1,%edi
  49a:   e8 61 fb ff ff          callq  0 <set_cpuid_faulting>

... the first 7 instructions burdens all __switch_to_xtra() users, not just the 
faulting-CPUID users.

The set_cpuid_faulting() call is also unnecessary and the set_cpuid_faulting() 
call generates into an obscene sequence of:

0000000000000000 <set_cpuid_faulting>:
   0:	8b 15 00 00 00 00    	mov    0x0(%rip),%edx        # 6 <set_cpuid_faulting+0x6>
   6:	55                   	push   %rbp
   7:	48 89 e5             	mov    %rsp,%rbp
   a:	53                   	push   %rbx
   b:	40 0f b6 df          	movzbl %dil,%ebx
   f:	85 d2                	test   %edx,%edx
  11:	75 07                	jne    1a <set_cpuid_faulting+0x1a>
  13:	9c                   	pushfq 
  14:	58                   	pop    %rax
  15:	f6 c4 02             	test   $0x2,%ah
  18:	75 48                	jne    62 <set_cpuid_faulting+0x62>
  1a:	65 48 8b 05 00 00 00 	mov    %gs:0x0(%rip),%rax        # 22 <set_cpuid_faulting+0x22>
  21:	00 
  22:	48 83 e0 fe          	and    $0xfffffffffffffffe,%rax
  26:	b9 40 01 00 00       	mov    $0x140,%ecx
  2b:	48 09 d8             	or     %rbx,%rax
  2e:	48 89 c2             	mov    %rax,%rdx
  31:	65 48 89 05 00 00 00 	mov    %rax,%gs:0x0(%rip)        # 39 <set_cpuid_faulting+0x39>
  38:	00 
  39:	48 c1 ea 20          	shr    $0x20,%rdx
  3d:	0f 30                	wrmsr  
  3f:	0f 1f 44 00 00       	nopl   0x0(%rax,%rax,1)
  44:	5b                   	pop    %rbx
  45:	5d                   	pop    %rbp
  46:	c3                   	retq   
  47:	48 c1 e2 20          	shl    $0x20,%rdx
  4b:	89 c0                	mov    %eax,%eax
  4d:	bf 40 01 00 00       	mov    $0x140,%edi
  52:	48 09 d0             	or     %rdx,%rax
  55:	31 d2                	xor    %edx,%edx
  57:	48 89 c6             	mov    %rax,%rsi
  5a:	e8 00 00 00 00       	callq  5f <set_cpuid_faulting+0x5f>
  5f:	5b                   	pop    %rbx
  60:	5d                   	pop    %rbp
  61:	c3                   	retq   
  62:	e8 00 00 00 00       	callq  67 <set_cpuid_faulting+0x67>
  67:	85 c0                	test   %eax,%eax
  69:	74 af                	je     1a <set_cpuid_faulting+0x1a>
  6b:	8b 05 00 00 00 00    	mov    0x0(%rip),%eax        # 71 <set_cpuid_faulting+0x71>
  71:	85 c0                	test   %eax,%eax
  73:	75 a5                	jne    1a <set_cpuid_faulting+0x1a>
  75:	48 c7 c1 00 00 00 00 	mov    $0x0,%rcx
  7c:	48 c7 c2 00 00 00 00 	mov    $0x0,%rdx
  83:	be b9 00 00 00       	mov    $0xb9,%esi
  88:	48 c7 c7 00 00 00 00 	mov    $0x0,%rdi
  8f:	e8 00 00 00 00       	callq  94 <set_cpuid_faulting+0x94>
  94:	eb 84                	jmp    1a <set_cpuid_faulting+0x1a>
  96:	66 2e 0f 1f 84 00 00 	nopw   %cs:0x0(%rax,%rax,1)
  9d:	00 00 00 

The affected object file code size blows up as well, by 17%:

 arch/x86/kernel/process.o:
    text    data     bss     dec     hex filename
    3325    8577      32   11934    2e9e process.o.before
    3889    8609      32   12530    30f2 process.o.after

A good deal of this overhead and complexity comes from the implementation 
inefficiency I pointed out, and all this can be avoided with the method I 
suggested in my previous review, by caching the per task MSR value in the thread 
struct.

So sorry, NAK for this implementation - especially considering how relatively 
straightforward the changes I suggested are to implement.

Thanks,

	Ingo

Re: [PATCH v13 0/8] x86/arch_prctl Add ARCH_[GET|SET]_CPUID for controlling the CPUID instruction

[Ingo Molnar]

* Kyle Huey <me@kylehuey.com> wrote:

> rr (http://rr-project.org/), a userspace record-and-replay reverse-
> execution debugger, would like to trap and emulate the CPUID instruction.
> This would allow us to a) mask away certain hardware features that rr does
> not support (e.g. RDRAND) and b) enable trace portability across machines
> by providing constant results.
> 
> Newer Intel CPUs (Ivy Bridge and later) can fault when CPUID is executed at
> CPL > 0. Expose this capability to userspace as a new pair of arch_prctls,
> ARCH_GET_CPUID and ARCH_SET_CPUID.
> 
> Since v12:
> Patch 4: x86/syscalls/32: Wire up arch_prctl on x86-32
> - compat_sys_arch_prctl prototype has argument names.

So while I am fine with the feature, I'm still unconvinced about the 
implementation:

1)

As I pointed out before, the arbitrary 'code' argument name x86-ism should be 
changed to 'option' like the canonical core kernel option name is for prctls(). 

This is still unfixed.

2)

As I complained about in my first review, TIF_NOCPUID flag is too far removed from 
the value what will be written into the MSR.

The result is poor code generation on 64-bit defconfig+CONFIG_PREEMPT=y:

        if (test_tsk_thread_flag(prev_p, TIF_NOCPUID) ^
            test_tsk_thread_flag(next_p, TIF_NOCPUID)) {
		set_cpuid_faulting(test_tsk_thread_flag(next_p, TIF_NOCPUID));

is compiled as:

  476:   49 8b 06                mov    (%r14),%rax
  479:   49 8b 55 00             mov    0x0(%r13),%rdx
  47d:   48 c1 e8 0f             shr    $0xf,%rax
  481:   48 c1 ea 0f             shr    $0xf,%rdx
  485:   83 e2 01                and    $0x1,%edx
  488:   83 e0 01                and    $0x1,%eax
  48b:   38 c2                   cmp    %al,%dl
  48d:   74 10                   je     49f <__switch_to_xtra+0x9f>
  48f:   49 8b 7d 00             mov    0x0(%r13),%rdi
  493:   48 c1 ef 0f             shr    $0xf,%rdi
  497:   83 e7 01                and    $0x1,%edi
  49a:   e8 61 fb ff ff          callq  0 <set_cpuid_faulting>

... the first 7 instructions burdens all __switch_to_xtra() users, not just the 
faulting-CPUID users.

The set_cpuid_faulting() call is also unnecessary and the set_cpuid_faulting() 
call generates into an obscene sequence of:

0000000000000000 <set_cpuid_faulting>:
   0:	8b 15 00 00 00 00    	mov    0x0(%rip),%edx        # 6 <set_cpuid_faulting+0x6>
   6:	55                   	push   %rbp
   7:	48 89 e5             	mov    %rsp,%rbp
   a:	53                   	push   %rbx
   b:	40 0f b6 df          	movzbl %dil,%ebx
   f:	85 d2                	test   %edx,%edx
  11:	75 07                	jne    1a <set_cpuid_faulting+0x1a>
  13:	9c                   	pushfq 
  14:	58                   	pop    %rax
  15:	f6 c4 02             	test   $0x2,%ah
  18:	75 48                	jne    62 <set_cpuid_faulting+0x62>
  1a:	65 48 8b 05 00 00 00 	mov    %gs:0x0(%rip),%rax        # 22 <set_cpuid_faulting+0x22>
  21:	00 
  22:	48 83 e0 fe          	and    $0xfffffffffffffffe,%rax
  26:	b9 40 01 00 00       	mov    $0x140,%ecx
  2b:	48 09 d8             	or     %rbx,%rax
  2e:	48 89 c2             	mov    %rax,%rdx
  31:	65 48 89 05 00 00 00 	mov    %rax,%gs:0x0(%rip)        # 39 <set_cpuid_faulting+0x39>
  38:	00 
  39:	48 c1 ea 20          	shr    $0x20,%rdx
  3d:	0f 30                	wrmsr  
  3f:	0f 1f 44 00 00       	nopl   0x0(%rax,%rax,1)
  44:	5b                   	pop    %rbx
  45:	5d                   	pop    %rbp
  46:	c3                   	retq   
  47:	48 c1 e2 20          	shl    $0x20,%rdx
  4b:	89 c0                	mov    %eax,%eax
  4d:	bf 40 01 00 00       	mov    $0x140,%edi
  52:	48 09 d0             	or     %rdx,%rax
  55:	31 d2                	xor    %edx,%edx
  57:	48 89 c6             	mov    %rax,%rsi
  5a:	e8 00 00 00 00       	callq  5f <set_cpuid_faulting+0x5f>
  5f:	5b                   	pop    %rbx
  60:	5d                   	pop    %rbp
  61:	c3                   	retq   
  62:	e8 00 00 00 00       	callq  67 <set_cpuid_faulting+0x67>
  67:	85 c0                	test   %eax,%eax
  69:	74 af                	je     1a <set_cpuid_faulting+0x1a>
  6b:	8b 05 00 00 00 00    	mov    0x0(%rip),%eax        # 71 <set_cpuid_faulting+0x71>
  71:	85 c0                	test   %eax,%eax
  73:	75 a5                	jne    1a <set_cpuid_faulting+0x1a>
  75:	48 c7 c1 00 00 00 00 	mov    $0x0,%rcx
  7c:	48 c7 c2 00 00 00 00 	mov    $0x0,%rdx
  83:	be b9 00 00 00       	mov    $0xb9,%esi
  88:	48 c7 c7 00 00 00 00 	mov    $0x0,%rdi
  8f:	e8 00 00 00 00       	callq  94 <set_cpuid_faulting+0x94>
  94:	eb 84                	jmp    1a <set_cpuid_faulting+0x1a>
  96:	66 2e 0f 1f 84 00 00 	nopw   %cs:0x0(%rax,%rax,1)
  9d:	00 00 00 

The affected object file code size blows up as well, by 17%:

 arch/x86/kernel/process.o:
    text    data     bss     dec     hex filename
    3325    8577      32   11934    2e9e process.o.before
    3889    8609      32   12530    30f2 process.o.after

A good deal of this overhead and complexity comes from the implementation 
inefficiency I pointed out, and all this can be avoided with the method I 
suggested in my previous review, by caching the per task MSR value in the thread 
struct.

So sorry, NAK for this implementation - especially considering how relatively 
straightforward the changes I suggested are to implement.

Thanks,

	Ingo

Re: [PATCH v13 0/8] x86/arch_prctl Add ARCH_[GET|SET]_CPUID for controlling the CPUID instruction

[Kyle Huey]

On Fri, Dec 2, 2016 at 2:29 AM, Ingo Molnar <mingo@kernel.org> wrote:
>
> * Kyle Huey <me@kylehuey.com> wrote:
>
>> rr (http://rr-project.org/), a userspace record-and-replay reverse-
>> execution debugger, would like to trap and emulate the CPUID instruction.
>> This would allow us to a) mask away certain hardware features that rr does
>> not support (e.g. RDRAND) and b) enable trace portability across machines
>> by providing constant results.
>>
>> Newer Intel CPUs (Ivy Bridge and later) can fault when CPUID is executed at
>> CPL > 0. Expose this capability to userspace as a new pair of arch_prctls,
>> ARCH_GET_CPUID and ARCH_SET_CPUID.
>>
>> Since v12:
>> Patch 4: x86/syscalls/32: Wire up arch_prctl on x86-32
>> - compat_sys_arch_prctl prototype has argument names.
>
> So while I am fine with the feature, I'm still unconvinced about the
> implementation:
>
> 1)
>
> As I pointed out before, the arbitrary 'code' argument name x86-ism should be
> changed to 'option' like the canonical core kernel option name is for prctls().
>
> This is still unfixed.

Yeah, it'll be fixed next time around.

> 2)
>
> As I complained about in my first review, TIF_NOCPUID flag is too far removed from
> the value what will be written into the MSR.
>
> The result is poor code generation on 64-bit defconfig+CONFIG_PREEMPT=y:
>
>         if (test_tsk_thread_flag(prev_p, TIF_NOCPUID) ^
>             test_tsk_thread_flag(next_p, TIF_NOCPUID)) {
>                 set_cpuid_faulting(test_tsk_thread_flag(next_p, TIF_NOCPUID));
>
> is compiled as:
>
>   476:   49 8b 06                mov    (%r14),%rax
>   479:   49 8b 55 00             mov    0x0(%r13),%rdx
>   47d:   48 c1 e8 0f             shr    $0xf,%rax
>   481:   48 c1 ea 0f             shr    $0xf,%rdx
>   485:   83 e2 01                and    $0x1,%edx
>   488:   83 e0 01                and    $0x1,%eax
>   48b:   38 c2                   cmp    %al,%dl
>   48d:   74 10                   je     49f <__switch_to_xtra+0x9f>
>   48f:   49 8b 7d 00             mov    0x0(%r13),%rdi
>   493:   48 c1 ef 0f             shr    $0xf,%rdi
>   497:   83 e7 01                and    $0x1,%edi
>   49a:   e8 61 fb ff ff          callq  0 <set_cpuid_faulting>
>
> ... the first 7 instructions burdens all __switch_to_xtra() users, not just the
> faulting-CPUID users.

That's fair.  This is certainly suboptimal, and it's suboptimal for
TIF_BLOCKSTEP and TIF_NOTSC too which generate essentially identical
code.  A much better code sequence here would be:

mov (%r14), %rax
xor (%r13), %rax
test $0x80, %ah
jz <elsewhere>
/* do cpuid faulting work */

We could do this by introducing a test_tsk_thread_flag_differs(...),
and supporting infrastructure, that XORs the flags of the two tasks
before doing the bit test.

Once we do that, the non-faulting case is pretty much equivalent to
the mov, mov, cmp, je sequence that would be needed if we stored the
MSR values in the task_struct.  The faulting case becomes a
straightforward time vs space tradeoff, and I'm inclined to think that
calling set_cpuid_faulting (which I don't think is as bad as you
suggest, see below) is better than taking up 8 bytes in every
task_struct for an uncommon feature.

And, as a bonus, we can improve the TIF_BLOCKSTEP and TIF_NOTSC cases too.

> The set_cpuid_faulting() call is also unnecessary and the set_cpuid_faulting()
> call generates into an obscene sequence of:
>
> 0000000000000000 <set_cpuid_faulting>:
>    0:   8b 15 00 00 00 00       mov    0x0(%rip),%edx        # 6 <set_cpuid_faulting+0x6>
>    6:   55                      push   %rbp
>    7:   48 89 e5                mov    %rsp,%rbp
>    a:   53                      push   %rbx
>    b:   40 0f b6 df             movzbl %dil,%ebx
>    f:   85 d2                   test   %edx,%edx
>   11:   75 07                   jne    1a <set_cpuid_faulting+0x1a>
>   13:   9c                      pushfq
>   14:   58                      pop    %rax
>   15:   f6 c4 02                test   $0x2,%ah
>   18:   75 48                   jne    62 <set_cpuid_faulting+0x62>
>   1a:   65 48 8b 05 00 00 00    mov    %gs:0x0(%rip),%rax        # 22 <set_cpuid_faulting+0x22>
>   21:   00
>   22:   48 83 e0 fe             and    $0xfffffffffffffffe,%rax
>   26:   b9 40 01 00 00          mov    $0x140,%ecx
>   2b:   48 09 d8                or     %rbx,%rax
>   2e:   48 89 c2                mov    %rax,%rdx
>   31:   65 48 89 05 00 00 00    mov    %rax,%gs:0x0(%rip)        # 39 <set_cpuid_faulting+0x39>
>   38:   00
>   39:   48 c1 ea 20             shr    $0x20,%rdx
>   3d:   0f 30                   wrmsr
>   3f:   0f 1f 44 00 00          nopl   0x0(%rax,%rax,1)
>   44:   5b                      pop    %rbx
>   45:   5d                      pop    %rbp
>   46:   c3                      retq
>   47:   48 c1 e2 20             shl    $0x20,%rdx
>   4b:   89 c0                   mov    %eax,%eax
>   4d:   bf 40 01 00 00          mov    $0x140,%edi
>   52:   48 09 d0                or     %rdx,%rax
>   55:   31 d2                   xor    %edx,%edx
>   57:   48 89 c6                mov    %rax,%rsi
>   5a:   e8 00 00 00 00          callq  5f <set_cpuid_faulting+0x5f>
>   5f:   5b                      pop    %rbx
>   60:   5d                      pop    %rbp
>   61:   c3                      retq
>   62:   e8 00 00 00 00          callq  67 <set_cpuid_faulting+0x67>
>   67:   85 c0                   test   %eax,%eax
>   69:   74 af                   je     1a <set_cpuid_faulting+0x1a>
>   6b:   8b 05 00 00 00 00       mov    0x0(%rip),%eax        # 71 <set_cpuid_faulting+0x71>
>   71:   85 c0                   test   %eax,%eax
>   73:   75 a5                   jne    1a <set_cpuid_faulting+0x1a>
>   75:   48 c7 c1 00 00 00 00    mov    $0x0,%rcx
>   7c:   48 c7 c2 00 00 00 00    mov    $0x0,%rdx
>   83:   be b9 00 00 00          mov    $0xb9,%esi
>   88:   48 c7 c7 00 00 00 00    mov    $0x0,%rdi
>   8f:   e8 00 00 00 00          callq  94 <set_cpuid_faulting+0x94>
>   94:   eb 84                   jmp    1a <set_cpuid_faulting+0x1a>
>   96:   66 2e 0f 1f 84 00 00    nopw   %cs:0x0(%rax,%rax,1)
>   9d:   00 00 00

I don't know why you're getting that.  Locally (with gcc 5.4) I have,
with CONFIG_PARAVIRT=y:

0000000000000000 <set_cpuid_faulting>:
   0:   e8 00 00 00 00          callq  5 <set_cpuid_faulting+0x5>
   5:   55                      push   %rbp
   6:   65 48 8b 15 00 00 00    mov    %gs:0x0(%rip),%rdx        # e
<set_cpuid_faulting+0xe>
   d:   00
   e:   48 89 d0                mov    %rdx,%rax
  11:   40 0f b6 d7             movzbl %dil,%edx
  15:   48 89 e5                mov    %rsp,%rbp
  18:   48 83 e0 fe             and    $0xfffffffffffffffe,%rax
  1c:   bf 40 01 00 00          mov    $0x140,%edi
  21:   48 09 c2                or     %rax,%rdx
  24:   89 d6                   mov    %edx,%esi
  26:   65 48 89 15 00 00 00    mov    %rdx,%gs:0x0(%rip)        # 2e
<set_cpuid_faulting+0x2e>
  2d:   00
  2e:   48 c1 ea 20             shr    $0x20,%rdx
  32:   ff 14 25 00 00 00 00    callq  *0x0
  39:   5d                      pop    %rbp
  3a:   c3                      retq
  3b:   0f 1f 44 00 00          nopl   0x0(%rax,%rax,1)

and with CONFIG_PARAVIRT=n, set_cpuid_faulting gets inlined into
__switch_to_xtra producing:

 4da:   48 8b 03                mov    (%rbx),%rax
 4dd:   49 33 04 24             xor    (%r12),%rax
 4e2:   f6 c4 80                test   $0x80,%ah
 4e5:   0f 85 ee 00 00 00       jne    5d9 <__switch_to_xtra+0x179>
...
5d9:   48 8b 33                mov    (%rbx),%rsi
 5dc:   b9 40 01 00 00          mov    $0x140,%ecx
 5e1:   65 48 8b 05 00 00 00    mov    %gs:0x0(%rip),%rax        # 5e9
<__switch_to_xtra+0x189>
 5e8:   00
 5e9:   48 83 e0 fe             and    $0xfffffffffffffffe,%rax
 5ed:   48 89 c2                mov    %rax,%rdx
 5f0:   48 89 f0                mov    %rsi,%rax
 5f3:   48 c1 e8 0f             shr    $0xf,%rax
 5f7:   83 e0 01                and    $0x1,%eax
 5fa:   48 09 d0                or     %rdx,%rax
 5fd:   48 89 c2                mov    %rax,%rdx
 600:   65 48 89 05 00 00 00    mov    %rax,%gs:0x0(%rip)        # 608
<__switch_to_xtra+0x1a8>
 607:   00
 608:   48 c1 ea 20             shr    $0x20,%rdx
 60c:   0f 30                   wrmsr
 60e:   0f 1f 44 00 00          nopl   0x0(%rax,%rax,1)
 613:   e9 d3 fe ff ff          jmpq   4eb <__switch_to_xtra+0x8b>

both of which are much saner code.

> The affected object file code size blows up as well, by 17%:
>
>  arch/x86/kernel/process.o:
>     text    data     bss     dec     hex filename
>     3325    8577      32   11934    2e9e process.o.before
>     3889    8609      32   12530    30f2 process.o.after
>
> A good deal of this overhead and complexity comes from the implementation
> inefficiency I pointed out, and all this can be avoided with the method I
> suggested in my previous review, by caching the per task MSR value in the thread
> struct.

12% here, where .before is before any of my changes, .as-is is with
these patches as submitted, and .after is with the changes as
described above.

size process.o*
   text       data        bss        dec        hex    filename
   4669       8521         96      13286       33e6    process.o.after
   4685       8521         96      13302       33f6    process.o.as-is
   4197       8506         96      12799       31ff    process.o.before

> So sorry, NAK for this implementation - especially considering how relatively
> straightforward the changes I suggested are to implement.

Would these proposed changes satisfy you?  Obviously I want to get
this into the kernel or I wouldn't be here.  So if you insist on
caching the MSR in the task_struct I'll do it, but I think this is at
least as good of an approach.

- Kyle

Re: [PATCH v13 0/8] x86/arch_prctl Add ARCH_[GET|SET]_CPUID for controlling the CPUID instruction

[Kyle Huey]

On Fri, Dec 2, 2016 at 2:29 AM, Ingo Molnar <mingo@kernel.org> wrote:
>
> * Kyle Huey <me@kylehuey.com> wrote:
>
>> rr (http://rr-project.org/), a userspace record-and-replay reverse-
>> execution debugger, would like to trap and emulate the CPUID instruction.
>> This would allow us to a) mask away certain hardware features that rr does
>> not support (e.g. RDRAND) and b) enable trace portability across machines
>> by providing constant results.
>>
>> Newer Intel CPUs (Ivy Bridge and later) can fault when CPUID is executed at
>> CPL > 0. Expose this capability to userspace as a new pair of arch_prctls,
>> ARCH_GET_CPUID and ARCH_SET_CPUID.
>>
>> Since v12:
>> Patch 4: x86/syscalls/32: Wire up arch_prctl on x86-32
>> - compat_sys_arch_prctl prototype has argument names.
>
> So while I am fine with the feature, I'm still unconvinced about the
> implementation:
>
> 1)
>
> As I pointed out before, the arbitrary 'code' argument name x86-ism should be
> changed to 'option' like the canonical core kernel option name is for prctls().
>
> This is still unfixed.

Yeah, it'll be fixed next time around.

> 2)
>
> As I complained about in my first review, TIF_NOCPUID flag is too far removed from
> the value what will be written into the MSR.
>
> The result is poor code generation on 64-bit defconfig+CONFIG_PREEMPT=y:
>
>         if (test_tsk_thread_flag(prev_p, TIF_NOCPUID) ^
>             test_tsk_thread_flag(next_p, TIF_NOCPUID)) {
>                 set_cpuid_faulting(test_tsk_thread_flag(next_p, TIF_NOCPUID));
>
> is compiled as:
>
>   476:   49 8b 06                mov    (%r14),%rax
>   479:   49 8b 55 00             mov    0x0(%r13),%rdx
>   47d:   48 c1 e8 0f             shr    $0xf,%rax
>   481:   48 c1 ea 0f             shr    $0xf,%rdx
>   485:   83 e2 01                and    $0x1,%edx
>   488:   83 e0 01                and    $0x1,%eax
>   48b:   38 c2                   cmp    %al,%dl
>   48d:   74 10                   je     49f <__switch_to_xtra+0x9f>
>   48f:   49 8b 7d 00             mov    0x0(%r13),%rdi
>   493:   48 c1 ef 0f             shr    $0xf,%rdi
>   497:   83 e7 01                and    $0x1,%edi
>   49a:   e8 61 fb ff ff          callq  0 <set_cpuid_faulting>
>
> ... the first 7 instructions burdens all __switch_to_xtra() users, not just the
> faulting-CPUID users.

That's fair.  This is certainly suboptimal, and it's suboptimal for
TIF_BLOCKSTEP and TIF_NOTSC too which generate essentially identical
code.  A much better code sequence here would be:

mov (%r14), %rax
xor (%r13), %rax
test $0x80, %ah
jz <elsewhere>
/* do cpuid faulting work */

We could do this by introducing a test_tsk_thread_flag_differs(...),
and supporting infrastructure, that XORs the flags of the two tasks
before doing the bit test.

Once we do that, the non-faulting case is pretty much equivalent to
the mov, mov, cmp, je sequence that would be needed if we stored the
MSR values in the task_struct.  The faulting case becomes a
straightforward time vs space tradeoff, and I'm inclined to think that
calling set_cpuid_faulting (which I don't think is as bad as you
suggest, see below) is better than taking up 8 bytes in every
task_struct for an uncommon feature.

And, as a bonus, we can improve the TIF_BLOCKSTEP and TIF_NOTSC cases too.

> The set_cpuid_faulting() call is also unnecessary and the set_cpuid_faulting()
> call generates into an obscene sequence of:
>
> 0000000000000000 <set_cpuid_faulting>:
>    0:   8b 15 00 00 00 00       mov    0x0(%rip),%edx        # 6 <set_cpuid_faulting+0x6>
>    6:   55                      push   %rbp
>    7:   48 89 e5                mov    %rsp,%rbp
>    a:   53                      push   %rbx
>    b:   40 0f b6 df             movzbl %dil,%ebx
>    f:   85 d2                   test   %edx,%edx
>   11:   75 07                   jne    1a <set_cpuid_faulting+0x1a>
>   13:   9c                      pushfq
>   14:   58                      pop    %rax
>   15:   f6 c4 02                test   $0x2,%ah
>   18:   75 48                   jne    62 <set_cpuid_faulting+0x62>
>   1a:   65 48 8b 05 00 00 00    mov    %gs:0x0(%rip),%rax        # 22 <set_cpuid_faulting+0x22>
>   21:   00
>   22:   48 83 e0 fe             and    $0xfffffffffffffffe,%rax
>   26:   b9 40 01 00 00          mov    $0x140,%ecx
>   2b:   48 09 d8                or     %rbx,%rax
>   2e:   48 89 c2                mov    %rax,%rdx
>   31:   65 48 89 05 00 00 00    mov    %rax,%gs:0x0(%rip)        # 39 <set_cpuid_faulting+0x39>
>   38:   00
>   39:   48 c1 ea 20             shr    $0x20,%rdx
>   3d:   0f 30                   wrmsr
>   3f:   0f 1f 44 00 00          nopl   0x0(%rax,%rax,1)
>   44:   5b                      pop    %rbx
>   45:   5d                      pop    %rbp
>   46:   c3                      retq
>   47:   48 c1 e2 20             shl    $0x20,%rdx
>   4b:   89 c0                   mov    %eax,%eax
>   4d:   bf 40 01 00 00          mov    $0x140,%edi
>   52:   48 09 d0                or     %rdx,%rax
>   55:   31 d2                   xor    %edx,%edx
>   57:   48 89 c6                mov    %rax,%rsi
>   5a:   e8 00 00 00 00          callq  5f <set_cpuid_faulting+0x5f>
>   5f:   5b                      pop    %rbx
>   60:   5d                      pop    %rbp
>   61:   c3                      retq
>   62:   e8 00 00 00 00          callq  67 <set_cpuid_faulting+0x67>
>   67:   85 c0                   test   %eax,%eax
>   69:   74 af                   je     1a <set_cpuid_faulting+0x1a>
>   6b:   8b 05 00 00 00 00       mov    0x0(%rip),%eax        # 71 <set_cpuid_faulting+0x71>
>   71:   85 c0                   test   %eax,%eax
>   73:   75 a5                   jne    1a <set_cpuid_faulting+0x1a>
>   75:   48 c7 c1 00 00 00 00    mov    $0x0,%rcx
>   7c:   48 c7 c2 00 00 00 00    mov    $0x0,%rdx
>   83:   be b9 00 00 00          mov    $0xb9,%esi
>   88:   48 c7 c7 00 00 00 00    mov    $0x0,%rdi
>   8f:   e8 00 00 00 00          callq  94 <set_cpuid_faulting+0x94>
>   94:   eb 84                   jmp    1a <set_cpuid_faulting+0x1a>
>   96:   66 2e 0f 1f 84 00 00    nopw   %cs:0x0(%rax,%rax,1)
>   9d:   00 00 00

I don't know why you're getting that.  Locally (with gcc 5.4) I have,
with CONFIG_PARAVIRT=y:

0000000000000000 <set_cpuid_faulting>:
   0:   e8 00 00 00 00          callq  5 <set_cpuid_faulting+0x5>
   5:   55                      push   %rbp
   6:   65 48 8b 15 00 00 00    mov    %gs:0x0(%rip),%rdx        # e
<set_cpuid_faulting+0xe>
   d:   00
   e:   48 89 d0                mov    %rdx,%rax
  11:   40 0f b6 d7             movzbl %dil,%edx
  15:   48 89 e5                mov    %rsp,%rbp
  18:   48 83 e0 fe             and    $0xfffffffffffffffe,%rax
  1c:   bf 40 01 00 00          mov    $0x140,%edi
  21:   48 09 c2                or     %rax,%rdx
  24:   89 d6                   mov    %edx,%esi
  26:   65 48 89 15 00 00 00    mov    %rdx,%gs:0x0(%rip)        # 2e
<set_cpuid_faulting+0x2e>
  2d:   00
  2e:   48 c1 ea 20             shr    $0x20,%rdx
  32:   ff 14 25 00 00 00 00    callq  *0x0
  39:   5d                      pop    %rbp
  3a:   c3                      retq
  3b:   0f 1f 44 00 00          nopl   0x0(%rax,%rax,1)

and with CONFIG_PARAVIRT=n, set_cpuid_faulting gets inlined into
__switch_to_xtra producing:

 4da:   48 8b 03                mov    (%rbx),%rax
 4dd:   49 33 04 24             xor    (%r12),%rax
 4e2:   f6 c4 80                test   $0x80,%ah
 4e5:   0f 85 ee 00 00 00       jne    5d9 <__switch_to_xtra+0x179>
...
5d9:   48 8b 33                mov    (%rbx),%rsi
 5dc:   b9 40 01 00 00          mov    $0x140,%ecx
 5e1:   65 48 8b 05 00 00 00    mov    %gs:0x0(%rip),%rax        # 5e9
<__switch_to_xtra+0x189>
 5e8:   00
 5e9:   48 83 e0 fe             and    $0xfffffffffffffffe,%rax
 5ed:   48 89 c2                mov    %rax,%rdx
 5f0:   48 89 f0                mov    %rsi,%rax
 5f3:   48 c1 e8 0f             shr    $0xf,%rax
 5f7:   83 e0 01                and    $0x1,%eax
 5fa:   48 09 d0                or     %rdx,%rax
 5fd:   48 89 c2                mov    %rax,%rdx
 600:   65 48 89 05 00 00 00    mov    %rax,%gs:0x0(%rip)        # 608
<__switch_to_xtra+0x1a8>
 607:   00
 608:   48 c1 ea 20             shr    $0x20,%rdx
 60c:   0f 30                   wrmsr
 60e:   0f 1f 44 00 00          nopl   0x0(%rax,%rax,1)
 613:   e9 d3 fe ff ff          jmpq   4eb <__switch_to_xtra+0x8b>

both of which are much saner code.

> The affected object file code size blows up as well, by 17%:
>
>  arch/x86/kernel/process.o:
>     text    data     bss     dec     hex filename
>     3325    8577      32   11934    2e9e process.o.before
>     3889    8609      32   12530    30f2 process.o.after
>
> A good deal of this overhead and complexity comes from the implementation
> inefficiency I pointed out, and all this can be avoided with the method I
> suggested in my previous review, by caching the per task MSR value in the thread
> struct.

12% here, where .before is before any of my changes, .as-is is with
these patches as submitted, and .after is with the changes as
described above.

size process.o*
   text       data        bss        dec        hex    filename
   4669       8521         96      13286       33e6    process.o.after
   4685       8521         96      13302       33f6    process.o.as-is
   4197       8506         96      12799       31ff    process.o.before

> So sorry, NAK for this implementation - especially considering how relatively
> straightforward the changes I suggested are to implement.

Would these proposed changes satisfy you?  Obviously I want to get
this into the kernel or I wouldn't be here.  So if you insist on
caching the MSR in the task_struct I'll do it, but I think this is at
least as good of an approach.

- Kyle