Merge branch 'x86-fpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'x86-fpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, fpu: Merge fpu_save_init()
  x86-32, fpu: Rewrite fpu_save_init()
  x86, fpu: Remove PSHUFB_XMM5_* macros
  x86, fpu: Remove unnecessary ifdefs from i387 code.
  x86-32, fpu: Remove math_emulate stub
  x86-64, fpu: Simplify constraints for fxsave/fxtstor
  x86-64, fpu: Fix %cs value in convert_from_fxsr()
  x86-64, fpu: Disable preemption when using TS_USEDFPU
  x86, fpu: Merge __save_init_fpu()
  x86, fpu: Merge tolerant_fwait()
  x86, fpu: Merge fpu_init()
  x86: Use correct type for %cr4
  x86, xsave: Disable xsave in i387 emulation mode

Fixed up fxsaveq-induced conflict in arch/x86/include/asm/i387.h

4 files changed, 35 insertions, 67 deletions

diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
index 92af7b420ab..4b68bda3093 100644
--- a/arch/x86/kernel/cpu/common.c
+++ b/arch/x86/kernel/cpu/common.c
@@ -1269,13 +1269,6 @@ void __cpuinit cpu_init(void)
 	clear_all_debug_regs();
 	dbg_restore_debug_regs();
 
-	/*
-	 * Force FPU initialization:
-	 */
-	current_thread_info()->status = 0;
-	clear_used_math();
-	mxcsr_feature_mask_init();
-
 	fpu_init();
 	xsave_init();
 }
diff --git a/arch/x86/kernel/i387.c b/arch/x86/kernel/i387.c
index a46cb3522c0..58bb239a2fd 100644
--- a/arch/x86/kernel/i387.c
+++ b/arch/x86/kernel/i387.c
@@ -68,19 +68,22 @@ static void __cpuinit init_thread_xstate(void)
 	 */
 
 	if (!HAVE_HWFP) {
+		/*
+		 * Disable xsave as we do not support it if i387
+		 * emulation is enabled.
+		 */
+		setup_clear_cpu_cap(X86_FEATURE_XSAVE);
+		setup_clear_cpu_cap(X86_FEATURE_XSAVEOPT);
 		xstate_size = sizeof(struct i387_soft_struct);
 		return;
 	}
 
 	if (cpu_has_fxsr)
 		xstate_size = sizeof(struct i387_fxsave_struct);
-#ifdef CONFIG_X86_32
 	else
 		xstate_size = sizeof(struct i387_fsave_struct);
-#endif
 }
 
-#ifdef CONFIG_X86_64
 /*
  * Called at bootup to set up the initial FPU state that is later cloned
  * into all processes.
@@ -88,12 +91,21 @@ static void __cpuinit init_thread_xstate(void)
 
 void __cpuinit fpu_init(void)
 {
-	unsigned long oldcr0 = read_cr0();
-
-	set_in_cr4(X86_CR4_OSFXSR);
-	set_in_cr4(X86_CR4_OSXMMEXCPT);
+	unsigned long cr0;
+	unsigned long cr4_mask = 0;
 
-	write_cr0(oldcr0 & ~(X86_CR0_TS|X86_CR0_EM)); /* clear TS and EM */
+	if (cpu_has_fxsr)
+		cr4_mask |= X86_CR4_OSFXSR;
+	if (cpu_has_xmm)
+		cr4_mask |= X86_CR4_OSXMMEXCPT;
+	if (cr4_mask)
+		set_in_cr4(cr4_mask);
+
+	cr0 = read_cr0();
+	cr0 &= ~(X86_CR0_TS|X86_CR0_EM); /* clear TS and EM */
+	if (!HAVE_HWFP)
+		cr0 |= X86_CR0_EM;
+	write_cr0(cr0);
 
 	if (!smp_processor_id())
 		init_thread_xstate();
@@ -104,24 +116,12 @@ void __cpuinit fpu_init(void)
 	clear_used_math();
 }
 
-#else	/* CONFIG_X86_64 */
-
-void __cpuinit fpu_init(void)
-{
-	if (!smp_processor_id())
-		init_thread_xstate();
-}
-
-#endif	/* CONFIG_X86_32 */
-
 void fpu_finit(struct fpu *fpu)
 {
-#ifdef CONFIG_X86_32
 	if (!HAVE_HWFP) {
 		finit_soft_fpu(&fpu->state->soft);
 		return;
 	}
-#endif
 
 	if (cpu_has_fxsr) {
 		struct i387_fxsave_struct *fx = &fpu->state->fxsave;
@@ -386,19 +386,17 @@ convert_from_fxsr(struct user_i387_ia32_struct *env, struct task_struct *tsk)
 #ifdef CONFIG_X86_64
 	env->fip = fxsave->rip;
 	env->foo = fxsave->rdp;
+	/*
+	 * should be actually ds/cs at fpu exception time, but
+	 * that information is not available in 64bit mode.
+	 */
+	env->fcs = task_pt_regs(tsk)->cs;
 	if (tsk == current) {
-		/*
-		 * should be actually ds/cs at fpu exception time, but
-		 * that information is not available in 64bit mode.
-		 */
-		asm("mov %%ds, %[fos]" : [fos] "=r" (env->fos));
-		asm("mov %%cs, %[fcs]" : [fcs] "=r" (env->fcs));
+		savesegment(ds, env->fos);
 	} else {
-		struct pt_regs *regs = task_pt_regs(tsk);
-
-		env->fos = 0xffff0000 | tsk->thread.ds;
-		env->fcs = regs->cs;
+		env->fos = tsk->thread.ds;
 	}
+	env->fos |= 0xffff0000;
 #else
 	env->fip = fxsave->fip;
 	env->fcs = (u16) fxsave->fcs | ((u32) fxsave->fop << 16);
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 3d9ea531ddd..b3d7a3a04f3 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -424,7 +424,7 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 	load_TLS(next, cpu);
 
 	/* Must be after DS reload */
-	unlazy_fpu(prev_p);
+	__unlazy_fpu(prev_p);
 
 	/* Make sure cpu is ready for new context */
 	if (preload_fpu)
diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c
index 60788dee0f8..d43968503dd 100644
--- a/arch/x86/kernel/traps.c
+++ b/arch/x86/kernel/traps.c
@@ -776,21 +776,10 @@ asmlinkage void math_state_restore(void)
 }
 EXPORT_SYMBOL_GPL(math_state_restore);
 
-#ifndef CONFIG_MATH_EMULATION
-void math_emulate(struct math_emu_info *info)
-{
-	printk(KERN_EMERG
-		"math-emulation not enabled and no coprocessor found.\n");
-	printk(KERN_EMERG "killing %s.\n", current->comm);
-	force_sig(SIGFPE, current);
-	schedule();
-}
-#endif /* CONFIG_MATH_EMULATION */
-
 dotraplinkage void __kprobes
 do_device_not_available(struct pt_regs *regs, long error_code)
 {
-#ifdef CONFIG_X86_32
+#ifdef CONFIG_MATH_EMULATION
 	if (read_cr0() & X86_CR0_EM) {
 		struct math_emu_info info = { };
 
@@ -798,12 +787,12 @@ do_device_not_available(struct pt_regs *regs, long error_code)
 
 		info.regs = regs;
 		math_emulate(&info);
-	} else {
-		math_state_restore(); /* interrupts still off */
-		conditional_sti(regs);
+		return;
 	}
-#else
-	math_state_restore();
+#endif
+	math_state_restore(); /* interrupts still off */
+#ifdef CONFIG_X86_32
+	conditional_sti(regs);
 #endif
 }
 
@@ -881,18 +870,6 @@ void __init trap_init(void)
 #endif
 
 #ifdef CONFIG_X86_32
-	if (cpu_has_fxsr) {
-		printk(KERN_INFO "Enabling fast FPU save and restore... ");
-		set_in_cr4(X86_CR4_OSFXSR);
-		printk("done.\n");
-	}
-	if (cpu_has_xmm) {
-		printk(KERN_INFO
-			"Enabling unmasked SIMD FPU exception support... ");
-		set_in_cr4(X86_CR4_OSXMMEXCPT);
-		printk("done.\n");
-	}
-
 	set_system_trap_gate(SYSCALL_VECTOR, &system_call);
 	set_bit(SYSCALL_VECTOR, used_vectors);
 #endif