Commit 61c4628b authored by Suresh Siddha's avatar Suresh Siddha Committed by Ingo Molnar
Browse files

x86, fpu: split FPU state from task struct - v5 



Split the FPU save area from the task struct. This allows easy migration
of FPU context, and it's generally cleaner. It also allows the following
two optimizations:

1) only allocate when the application actually uses FPU, so in the first
lazy FPU trap. This could save memory for non-fpu using apps. Next patch
does this lazy allocation.

2) allocate the right size for the actual cpu rather than 512 bytes always.
Patches enabling xsave/xrstor support (coming shortly) will take advantage
of this.

Signed-off-by: default avatarSuresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: default avatarArjan van de Ven <arjan@linux.intel.com>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
Signed-off-by: default avatarThomas Gleixner <tglx@linutronix.de>
parent fa5c4639

arch/x86/kernel/Makefile

+1 −0
Original line number Diff line number Diff line
@@ -29,6 +29,7 @@ obj-$(CONFIG_X86_64) += pci-nommu_64.o bugs_64.o 
obj-y			+= tsc_$(BITS).o io_delay.o rtc.o



obj-$(CONFIG_X86_TRAMPOLINE)	+= trampoline.o

obj-y				+= process.o

obj-y				+= i387.o

obj-y				+= ptrace.o

obj-y				+= ds.o

arch/x86/kernel/i387.c

+46 −34
Original line number Diff line number Diff line
@@ -8,6 +8,7 @@ 
#include <linux/module.h>

#include <linux/regset.h>

#include <linux/sched.h>

#include <linux/bootmem.h>



#include <asm/sigcontext.h>

#include <asm/processor.h>
@@ -35,17 +36,18 @@ 
#endif



static unsigned int		mxcsr_feature_mask __read_mostly = 0xffffffffu;

unsigned int xstate_size;

static struct i387_fxsave_struct fx_scratch __cpuinitdata;



void mxcsr_feature_mask_init(void)

void __cpuinit mxcsr_feature_mask_init(void)

{

	unsigned long mask = 0;



	clts();

	if (cpu_has_fxsr) {

		memset(&current->thread.i387.fxsave, 0,

		       sizeof(struct i387_fxsave_struct));

		asm volatile("fxsave %0" : : "m" (current->thread.i387.fxsave));

		mask = current->thread.i387.fxsave.mxcsr_mask;

		memset(&fx_scratch, 0, sizeof(struct i387_fxsave_struct));

		asm volatile("fxsave %0" : : "m" (fx_scratch));

		mask = fx_scratch.mxcsr_mask;

		if (mask == 0)

			mask = 0x0000ffbf;

	}
@@ -53,6 +55,17 @@ void mxcsr_feature_mask_init(void) 
	stts();

}



void __init init_thread_xstate(void)

{

	if (cpu_has_fxsr)

		xstate_size = sizeof(struct i387_fxsave_struct);

#ifdef CONFIG_X86_32

	else

		xstate_size = sizeof(struct i387_fsave_struct);

#endif

	init_task.thread.xstate = alloc_bootmem(xstate_size);

}



#ifdef CONFIG_X86_64

/*

 * Called at bootup to set up the initial FPU state that is later cloned
@@ -61,10 +74,6 @@ void mxcsr_feature_mask_init(void) 
void __cpuinit fpu_init(void)

{

	unsigned long oldcr0 = read_cr0();

	extern void __bad_fxsave_alignment(void);



	if (offsetof(struct task_struct, thread.i387.fxsave) & 15)

		__bad_fxsave_alignment();



	set_in_cr4(X86_CR4_OSFXSR);

	set_in_cr4(X86_CR4_OSXMMEXCPT);
@@ -93,18 +102,19 @@ void init_fpu(struct task_struct *tsk) 
	}



	if (cpu_has_fxsr) {

		memset(&tsk->thread.i387.fxsave, 0,

		       sizeof(struct i387_fxsave_struct));

		tsk->thread.i387.fxsave.cwd = 0x37f;

		struct i387_fxsave_struct *fx = &tsk->thread.xstate->fxsave;



		memset(fx, 0, xstate_size);

		fx->cwd = 0x37f;

		if (cpu_has_xmm)

			tsk->thread.i387.fxsave.mxcsr = MXCSR_DEFAULT;

			fx->mxcsr = MXCSR_DEFAULT;

	} else {

		memset(&tsk->thread.i387.fsave, 0,

		       sizeof(struct i387_fsave_struct));

		tsk->thread.i387.fsave.cwd = 0xffff037fu;

		tsk->thread.i387.fsave.swd = 0xffff0000u;

		tsk->thread.i387.fsave.twd = 0xffffffffu;

		tsk->thread.i387.fsave.fos = 0xffff0000u;

		struct i387_fsave_struct *fp = &tsk->thread.xstate->fsave;

		memset(fp, 0, xstate_size);

		fp->cwd = 0xffff037fu;

		fp->swd = 0xffff0000u;

		fp->twd = 0xffffffffu;

		fp->fos = 0xffff0000u;

	}

	/*

	 * Only the device not available exception or ptrace can call init_fpu.
@@ -132,7 +142,7 @@ int xfpregs_get(struct task_struct *target, const struct user_regset *regset, 
	init_fpu(target);



	return user_regset_copyout(&pos, &count, &kbuf, &ubuf,

				   &target->thread.i387.fxsave, 0, -1);

				   &target->thread.xstate->fxsave, 0, -1);

}



int xfpregs_set(struct task_struct *target, const struct user_regset *regset,
@@ -148,12 +158,12 @@ int xfpregs_set(struct task_struct *target, const struct user_regset *regset, 
	set_stopped_child_used_math(target);



	ret = user_regset_copyin(&pos, &count, &kbuf, &ubuf,

				 &target->thread.i387.fxsave, 0, -1);

				 &target->thread.xstate->fxsave, 0, -1);



	/*

	 * mxcsr reserved bits must be masked to zero for security reasons.

	 */

	target->thread.i387.fxsave.mxcsr &= mxcsr_feature_mask;

	target->thread.xstate->fxsave.mxcsr &= mxcsr_feature_mask;



	return ret;

}
@@ -233,7 +243,7 @@ static inline u32 twd_fxsr_to_i387(struct i387_fxsave_struct *fxsave) 
static void

convert_from_fxsr(struct user_i387_ia32_struct *env, struct task_struct *tsk)

{

	struct i387_fxsave_struct *fxsave = &tsk->thread.i387.fxsave;

	struct i387_fxsave_struct *fxsave = &tsk->thread.xstate->fxsave;

	struct _fpreg *to = (struct _fpreg *) &env->st_space[0];

	struct _fpxreg *from = (struct _fpxreg *) &fxsave->st_space[0];

	int i;
@@ -273,7 +283,7 @@ static void convert_to_fxsr(struct task_struct *tsk, 
			    const struct user_i387_ia32_struct *env)



{

	struct i387_fxsave_struct *fxsave = &tsk->thread.i387.fxsave;

	struct i387_fxsave_struct *fxsave = &tsk->thread.xstate->fxsave;

	struct _fpreg *from = (struct _fpreg *) &env->st_space[0];

	struct _fpxreg *to = (struct _fpxreg *) &fxsave->st_space[0];

	int i;
@@ -310,7 +320,8 @@ int fpregs_get(struct task_struct *target, const struct user_regset *regset, 


	if (!cpu_has_fxsr) {

		return user_regset_copyout(&pos, &count, &kbuf, &ubuf,

					   &target->thread.i387.fsave, 0, -1);

					   &target->thread.xstate->fsave, 0,

					   -1);

	}



	if (kbuf && pos == 0 && count == sizeof(env)) {
@@ -338,7 +349,7 @@ int fpregs_set(struct task_struct *target, const struct user_regset *regset, 


	if (!cpu_has_fxsr) {

		return user_regset_copyin(&pos, &count, &kbuf, &ubuf,

					  &target->thread.i387.fsave, 0, -1);

					  &target->thread.xstate->fsave, 0, -1);

	}



	if (pos > 0 || count < sizeof(env))
@@ -358,11 +369,11 @@ int fpregs_set(struct task_struct *target, const struct user_regset *regset, 
static inline int save_i387_fsave(struct _fpstate_ia32 __user *buf)

{

	struct task_struct *tsk = current;

	struct i387_fsave_struct *fp = &tsk->thread.xstate->fsave;



	unlazy_fpu(tsk);

	tsk->thread.i387.fsave.status = tsk->thread.i387.fsave.swd;

	if (__copy_to_user(buf, &tsk->thread.i387.fsave,

			   sizeof(struct i387_fsave_struct)))

	fp->status = fp->swd;

	if (__copy_to_user(buf, fp, sizeof(struct i387_fsave_struct)))

		return -1;

	return 1;

}
@@ -370,6 +381,7 @@ static inline int save_i387_fsave(struct _fpstate_ia32 __user *buf) 
static int save_i387_fxsave(struct _fpstate_ia32 __user *buf)

{

	struct task_struct *tsk = current;

	struct i387_fxsave_struct *fx = &tsk->thread.xstate->fxsave;

	struct user_i387_ia32_struct env;

	int err = 0;
@@ -379,12 +391,12 @@ static int save_i387_fxsave(struct _fpstate_ia32 __user *buf) 
	if (__copy_to_user(buf, &env, sizeof(env)))

		return -1;



	err |= __put_user(tsk->thread.i387.fxsave.swd, &buf->status);

	err |= __put_user(fx->swd, &buf->status);

	err |= __put_user(X86_FXSR_MAGIC, &buf->magic);

	if (err)

		return -1;



	if (__copy_to_user(&buf->_fxsr_env[0], &tsk->thread.i387.fxsave,

	if (__copy_to_user(&buf->_fxsr_env[0], fx,

			   sizeof(struct i387_fxsave_struct)))

		return -1;

	return 1;
@@ -417,7 +429,7 @@ static inline int restore_i387_fsave(struct _fpstate_ia32 __user *buf) 
	struct task_struct *tsk = current;



	clear_fpu(tsk);

	return __copy_from_user(&tsk->thread.i387.fsave, buf,

	return __copy_from_user(&tsk->thread.xstate->fsave, buf,

				sizeof(struct i387_fsave_struct));

}
@@ -428,10 +440,10 @@ static int restore_i387_fxsave(struct _fpstate_ia32 __user *buf) 
	int err;



	clear_fpu(tsk);

	err = __copy_from_user(&tsk->thread.i387.fxsave, &buf->_fxsr_env[0],

	err = __copy_from_user(&tsk->thread.xstate->fxsave, &buf->_fxsr_env[0],

			       sizeof(struct i387_fxsave_struct));

	/* mxcsr reserved bits must be masked to zero for security reasons */

	tsk->thread.i387.fxsave.mxcsr &= mxcsr_feature_mask;

	tsk->thread.xstate->fxsave.mxcsr &= mxcsr_feature_mask;

	if (err || __copy_from_user(&env, buf, sizeof(env)))

		return 1;

	convert_to_fxsr(tsk, &env);

arch/x86/kernel/process.c

0 → 100644
+35 −0
Original line number Diff line number Diff line 
#include <linux/errno.h>

#include <linux/kernel.h>

#include <linux/mm.h>

#include <linux/smp.h>

#include <linux/slab.h>

#include <linux/sched.h>



static struct kmem_cache *task_xstate_cachep;



int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)

{

	*dst = *src;

	dst->thread.xstate = kmem_cache_alloc(task_xstate_cachep, GFP_KERNEL);

	if (!dst->thread.xstate)

		return -ENOMEM;

	WARN_ON((unsigned long)dst->thread.xstate & 15);

	memcpy(dst->thread.xstate, src->thread.xstate, xstate_size);

	return 0;

}



void free_thread_info(struct thread_info *ti)

{

	kmem_cache_free(task_xstate_cachep, ti->task->thread.xstate);

	ti->task->thread.xstate = NULL;



	free_pages((unsigned long)(ti), get_order(THREAD_SIZE));

}



void arch_task_cache_init(void)

{

        task_xstate_cachep =

        	kmem_cache_create("task_xstate", xstate_size,

				  __alignof__(union thread_xstate),

				  SLAB_PANIC, NULL);

}

arch/x86/kernel/process_32.c

+1 −1
Original line number Diff line number Diff line
@@ -703,7 +703,7 @@ struct task_struct * __switch_to(struct task_struct *prev_p, struct task_struct 


	/* we're going to use this soon, after a few expensive things */

	if (next_p->fpu_counter > 5)

		prefetch(&next->i387.fxsave);

		prefetch(next->xstate);



	/*

	 * Reload esp0.

arch/x86/kernel/process_64.c

+1 −1
Original line number Diff line number Diff line
@@ -682,7 +682,7 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p) 


	/* we're going to use this soon, after a few expensive things */

	if (next_p->fpu_counter>5)

		prefetch(&next->i387.fxsave);

		prefetch(next->xstate);



	/*

	 * Reload esp0, LDT and the page table pointer:

CRA Git | Maintained and supported by SUSTech CRA and CCSE