x86/fpu: Get rid of PF_USED_MATH usage, convert it to fpu->fpstate_active

Introduce a simple fpu->fpstate_active flag in the fpu context data structure
and use that instead of PF_USED_MATH in task->flags.

Testing for this flag byte should be slightly more efficient than
testing a bit in a bitmask, but the main advantage is that most
FPU functions can now be performed on a 'struct fpu' alone, they
don't need access to 'struct task_struct' anymore.

There's a slight linecount increase, mostly due to the 'fpu' local
variables and due to extra comments. The local variables will go away
once we move most of the FPU methods to pure 'struct fpu' parameters.
Reviewed-by: Borislav Petkov <bp@alien8.de>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

arch/x86/ia32/ia32_signal.c

@@ -307,6 +307,7 @@ static void __user *get_sigframe(struct ksignal *ksig, struct pt_regs *regs,
 				 size_t frame_size,
 				 void __user **fpstate)
 {
+	struct fpu *fpu = &current->thread.fpu;
 	unsigned long sp;
 
 	/* Default to using normal stack */
@@ -321,7 +322,7 @@ static void __user *get_sigframe(struct ksignal *ksig, struct pt_regs *regs,
 		 ksig->ka.sa.sa_restorer)
 		sp = (unsigned long) ksig->ka.sa.sa_restorer;
 
-	if (current->flags & PF_USED_MATH) {
+	if (fpu->fpstate_active) {
 		unsigned long fx_aligned, math_size;
 
 		sp = alloc_mathframe(sp, 1, &fx_aligned, &math_size);

arch/x86/include/asm/fpu-internal.h

@@ -375,7 +375,7 @@ static inline void drop_fpu(struct task_struct *tsk)
 		__thread_fpu_end(fpu);
 	}
 
-	tsk->flags &= ~PF_USED_MATH;
+	fpu->fpstate_active = 0;
 
 	preempt_enable();
 }
@@ -424,7 +424,7 @@ static inline fpu_switch_t switch_fpu_prepare(struct task_struct *old, struct ta
 	 * If the task has used the math, pre-load the FPU on xsave processors
 	 * or if the past 5 consecutive context-switches used math.
 	 */
-	fpu.preload = (new->flags & PF_USED_MATH) &&
+	fpu.preload = new_fpu->fpstate_active &&
 		      (use_eager_fpu() || new->thread.fpu.counter > 5);
 
 	if (old_fpu->has_fpu) {

arch/x86/include/asm/fpu/types.h

@@ -137,6 +137,12 @@ struct fpu {
 	 * deal with bursty apps that only use the FPU for a short time:
 	 */
 	unsigned char			counter;
+	/*
+	 * This flag indicates whether this context is fpstate_active: if the task is
+	 * not running then we can restore from this context, if the task
+	 * is running then we should save into this context.
+	 */
+	unsigned char			fpstate_active;
 };
 
 #endif /* _ASM_X86_FPU_H */

arch/x86/include/asm/processor.h

@@ -385,6 +385,10 @@ struct thread_struct {
 	unsigned long		fs;
 #endif
 	unsigned long		gs;
+
+	/* Floating point and extended processor state */
+	struct fpu		fpu;
+
 	/* Save middle states of ptrace breakpoints */
 	struct perf_event	*ptrace_bps[HBP_NUM];
 	/* Debug status used for traps, single steps, etc... */
@@ -395,8 +399,6 @@ struct thread_struct {
 	unsigned long		cr2;
 	unsigned long		trap_nr;
 	unsigned long		error_code;
-	/* floating point and extended processor state */
-	struct fpu		fpu;
 #ifdef CONFIG_X86_32
 	/* Virtual 86 mode info */
 	struct vm86_struct __user *vm86_info;

arch/x86/kernel/fpu/core.c

@@ -236,14 +236,17 @@ static void fpu_copy(struct task_struct *dst, struct task_struct *src)
 
 int fpu__copy(struct task_struct *dst, struct task_struct *src)
 {
+	struct fpu *dst_fpu = &dst->thread.fpu;
+	struct fpu *src_fpu = &src->thread.fpu;
+
 	dst->thread.fpu.counter = 0;
 	dst->thread.fpu.has_fpu = 0;
 	dst->thread.fpu.state = NULL;
 
 	task_disable_lazy_fpu_restore(dst);
 
-	if (src->flags & PF_USED_MATH) {
-		int err = fpstate_alloc(&dst->thread.fpu);
+	if (src_fpu->fpstate_active) {
+		int err = fpstate_alloc(dst_fpu);
 
 		if (err)
 			return err;
@@ -260,11 +263,12 @@ int fpu__copy(struct task_struct *dst, struct task_struct *src)
  */
 int fpstate_alloc_init(struct task_struct *curr)
 {
+	struct fpu *fpu = &curr->thread.fpu;
 	int ret;
 
 	if (WARN_ON_ONCE(curr != current))
 		return -EINVAL;
-	if (WARN_ON_ONCE(curr->flags & PF_USED_MATH))
+	if (WARN_ON_ONCE(fpu->fpstate_active))
 		return -EINVAL;
 
 	/*
@@ -277,7 +281,7 @@ int fpstate_alloc_init(struct task_struct *curr)
 	fpstate_init(&curr->thread.fpu);
 
 	/* Safe to do for the current task: */
-	curr->flags |= PF_USED_MATH;
+	fpu->fpstate_active = 1;
 
 	return 0;
 }
@@ -308,12 +312,13 @@ EXPORT_SYMBOL_GPL(fpstate_alloc_init);
  */
 static int fpu__unlazy_stopped(struct task_struct *child)
 {
+	struct fpu *child_fpu = &child->thread.fpu;
 	int ret;
 
 	if (WARN_ON_ONCE(child == current))
 		return -EINVAL;
 
-	if (child->flags & PF_USED_MATH) {
+	if (child_fpu->fpstate_active) {
 		task_disable_lazy_fpu_restore(child);
 		return 0;
 	}
@@ -328,7 +333,7 @@ static int fpu__unlazy_stopped(struct task_struct *child)
 	fpstate_init(&child->thread.fpu);
 
 	/* Safe to do for stopped child tasks: */
-	child->flags |= PF_USED_MATH;
+	child_fpu->fpstate_active = 1;
 
 	return 0;
 }
@@ -348,7 +353,7 @@ void fpu__restore(void)
 	struct task_struct *tsk = current;
 	struct fpu *fpu = &tsk->thread.fpu;
 
-	if (!(tsk->flags & PF_USED_MATH)) {
+	if (!fpu->fpstate_active) {
 		local_irq_enable();
 		/*
 		 * does a slab alloc which can sleep
@@ -378,6 +383,8 @@ EXPORT_SYMBOL_GPL(fpu__restore);
 
 void fpu__flush_thread(struct task_struct *tsk)
 {
+	struct fpu *fpu = &tsk->thread.fpu;
+
 	WARN_ON(tsk != current);
 
 	if (!use_eager_fpu()) {
@@ -385,7 +392,7 @@ void fpu__flush_thread(struct task_struct *tsk)
 		drop_fpu(tsk);
 		fpstate_free(&tsk->thread.fpu);
 	} else {
-		if (!(tsk->flags & PF_USED_MATH)) {
+		if (!fpu->fpstate_active) {
 			/* kthread execs. TODO: cleanup this horror. */
 		if (WARN_ON(fpstate_alloc_init(tsk)))
 				force_sig(SIGKILL, tsk);
@@ -402,12 +409,16 @@ void fpu__flush_thread(struct task_struct *tsk)
  */
 int fpregs_active(struct task_struct *target, const struct user_regset *regset)
 {
-	return (target->flags & PF_USED_MATH) ? regset->n : 0;
+	struct fpu *target_fpu = &target->thread.fpu;
+
+	return target_fpu->fpstate_active ? regset->n : 0;
 }
 
 int xfpregs_active(struct task_struct *target, const struct user_regset *regset)
 {
-	return (cpu_has_fxsr && (target->flags & PF_USED_MATH)) ? regset->n : 0;
+	struct fpu *target_fpu = &target->thread.fpu;
+
+	return (cpu_has_fxsr && target_fpu->fpstate_active) ? regset->n : 0;
 }
 
 int xfpregs_get(struct task_struct *target, const struct user_regset *regset,
@@ -733,16 +744,17 @@ int fpregs_set(struct task_struct *target, const struct user_regset *regset,
  * struct user_i387_struct) but is in fact only used for 32-bit
  * dumps, so on 64-bit it is really struct user_i387_ia32_struct.
  */
-int dump_fpu(struct pt_regs *regs, struct user_i387_struct *fpu)
+int dump_fpu(struct pt_regs *regs, struct user_i387_struct *ufpu)
 {
 	struct task_struct *tsk = current;
+	struct fpu *fpu = &tsk->thread.fpu;
 	int fpvalid;
 
-	fpvalid = !!(tsk->flags & PF_USED_MATH);
+	fpvalid = fpu->fpstate_active;
 	if (fpvalid)
 		fpvalid = !fpregs_get(tsk, NULL,
 				      0, sizeof(struct user_i387_ia32_struct),
-				      fpu, NULL);
+				      ufpu, NULL);
 
 	return fpvalid;
 }

arch/x86/kernel/fpu/xsave.c

@@ -334,6 +334,7 @@ int __restore_xstate_sig(void __user *buf, void __user *buf_fx, int size)
 {
 	int ia32_fxstate = (buf != buf_fx);
 	struct task_struct *tsk = current;
+	struct fpu *fpu = &tsk->thread.fpu;
 	int state_size = xstate_size;
 	u64 xstate_bv = 0;
 	int fx_only = 0;
@@ -349,7 +350,7 @@ int __restore_xstate_sig(void __user *buf, void __user *buf_fx, int size)
 	if (!access_ok(VERIFY_READ, buf, size))
 		return -EACCES;
 
-	if (!(tsk->flags & PF_USED_MATH) && fpstate_alloc_init(tsk))
+	if (!fpu->fpstate_active && fpstate_alloc_init(tsk))
 		return -1;
 
 	if (!static_cpu_has(X86_FEATURE_FPU))
@@ -384,12 +385,12 @@ int __restore_xstate_sig(void __user *buf, void __user *buf_fx, int size)
 		int err = 0;
 
 		/*
-		 * Drop the current fpu which clears PF_USED_MATH. This ensures
+		 * Drop the current fpu which clears fpu->fpstate_active. This ensures
 		 * that any context-switch during the copy of the new state,
 		 * avoids the intermediate state from getting restored/saved.
 		 * Thus avoiding the new restored state from getting corrupted.
 		 * We will be ready to restore/save the state only after
-		 * PF_USED_MATH is again set.
+		 * fpu->fpstate_active is again set.
 		 */
 		drop_fpu(tsk);
 
@@ -401,7 +402,7 @@ int __restore_xstate_sig(void __user *buf, void __user *buf_fx, int size)
 			sanitize_restored_xstate(tsk, &env, xstate_bv, fx_only);
 		}
 
-		tsk->flags |= PF_USED_MATH;
+		fpu->fpstate_active = 1;
 		if (use_eager_fpu()) {
 			preempt_disable();
 			fpu__restore();
@@ -685,7 +686,7 @@ void xsave_init(void)
  */
 void __init_refok eager_fpu_init(void)
 {
-	WARN_ON(current->flags & PF_USED_MATH);
+	WARN_ON(current->thread.fpu.fpstate_active);
 	current_thread_info()->status = 0;
 
 	if (eagerfpu == ENABLE)

arch/x86/kernel/signal.c

@@ -198,6 +198,7 @@ get_sigframe(struct k_sigaction *ka, struct pt_regs *regs, size_t frame_size,
 	unsigned long sp = regs->sp;
 	unsigned long buf_fx = 0;
 	int onsigstack = on_sig_stack(sp);
+	struct fpu *fpu = &current->thread.fpu;
 
 	/* redzone */
 	if (config_enabled(CONFIG_X86_64))
@@ -217,7 +218,7 @@ get_sigframe(struct k_sigaction *ka, struct pt_regs *regs, size_t frame_size,
 		}
 	}
 
-	if (current->flags & PF_USED_MATH) {
+	if (fpu->fpstate_active) {
 		sp = alloc_mathframe(sp, config_enabled(CONFIG_X86_32),
 				     &buf_fx, &math_size);
 		*fpstate = (void __user *)sp;
@@ -233,7 +234,7 @@ get_sigframe(struct k_sigaction *ka, struct pt_regs *regs, size_t frame_size,
 		return (void __user *)-1L;
 
 	/* save i387 and extended state */
-	if ((current->flags & PF_USED_MATH) &&
+	if (fpu->fpstate_active &&
 	    save_xstate_sig(*fpstate, (void __user *)buf_fx, math_size) < 0)
 		return (void __user *)-1L;
 
@@ -616,6 +617,7 @@ static void
 handle_signal(struct ksignal *ksig, struct pt_regs *regs)
 {
 	bool stepping, failed;
+	struct fpu *fpu = &current->thread.fpu;
 
 	/* Are we from a system call? */
 	if (syscall_get_nr(current, regs) >= 0) {
@@ -664,7 +666,7 @@ handle_signal(struct ksignal *ksig, struct pt_regs *regs)
 		/*
 		 * Ensure the signal handler starts with the new fpu state.
 		 */
-		if (current->flags & PF_USED_MATH)
+		if (fpu->fpstate_active)
 			fpu_reset_state(current);
 	}
 	signal_setup_done(failed, ksig, stepping);

arch/x86/kvm/x86.c

@@ -6597,10 +6597,11 @@ static int complete_emulated_mmio(struct kvm_vcpu *vcpu)
 
 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
 {
+	struct fpu *fpu = &current->thread.fpu;
 	int r;
 	sigset_t sigsaved;
 
-	if (!(current->flags & PF_USED_MATH) && fpstate_alloc_init(current))
+	if (!fpu->fpstate_active && fpstate_alloc_init(current))
 		return -ENOMEM;
 
 	if (vcpu->sigset_active)

arch/x86/math-emu/fpu_entry.c

@@ -147,8 +147,9 @@ void math_emulate(struct math_emu_info *info)
 	unsigned long code_base = 0;
 	unsigned long code_limit = 0;	/* Initialized to stop compiler warnings */
 	struct desc_struct code_descriptor;
+	struct fpu *fpu = &current->thread.fpu;
 
-	if (!(current->flags & PF_USED_MATH)) {
+	if (!fpu->fpstate_active) {
 		if (fpstate_alloc_init(current)) {
 			do_group_exit(SIGKILL);
 			return;