diff --git a/arch/s390/include/asm/fpu-internal.h b/arch/s390/include/asm/fpu-internal.h
index 237f8fcbe46b..55dc2c0fb40a 100644
--- a/arch/s390/include/asm/fpu-internal.h
+++ b/arch/s390/include/asm/fpu-internal.h
@@ -28,7 +28,7 @@ struct fpu {
 	};
 };
 
-void save_fpu_regs(struct fpu *fpu);
+void save_fpu_regs(void);
 
 #define is_vx_fpu(fpu) (!!((fpu)->flags & FPU_USE_VX))
 #define is_vx_task(tsk) (!!((tsk)->thread.fpu.flags & FPU_USE_VX))
diff --git a/arch/s390/include/asm/switch_to.h b/arch/s390/include/asm/switch_to.h
index 0a4a3150b7d7..dcadfde32265 100644
--- a/arch/s390/include/asm/switch_to.h
+++ b/arch/s390/include/asm/switch_to.h
@@ -30,7 +30,7 @@ static inline void restore_access_regs(unsigned int *acrs)
 
 #define switch_to(prev,next,last) do {					\
 	if (prev->mm) {							\
-		save_fpu_regs(&prev->thread.fpu);			\
+		save_fpu_regs();					\
 		save_access_regs(&prev->thread.acrs[0]);		\
 		save_ri_cb(prev->thread.ri_cb);				\
 	}								\
diff --git a/arch/s390/kernel/asm-offsets.c b/arch/s390/kernel/asm-offsets.c
index 6bc42c08be09..48c9af7a7683 100644
--- a/arch/s390/kernel/asm-offsets.c
+++ b/arch/s390/kernel/asm-offsets.c
@@ -28,16 +28,14 @@ int main(void)
 	DEFINE(__TASK_pid, offsetof(struct task_struct, pid));
 	BLANK();
 	DEFINE(__THREAD_ksp, offsetof(struct thread_struct, ksp));
-	DEFINE(__THREAD_fpu, offsetof(struct task_struct, thread.fpu));
+	DEFINE(__THREAD_FPU_fpc, offsetof(struct thread_struct, fpu.fpc));
+	DEFINE(__THREAD_FPU_flags, offsetof(struct thread_struct, fpu.flags));
+	DEFINE(__THREAD_FPU_regs, offsetof(struct thread_struct, fpu.regs));
 	DEFINE(__THREAD_per_cause, offsetof(struct thread_struct, per_event.cause));
 	DEFINE(__THREAD_per_address, offsetof(struct thread_struct, per_event.address));
 	DEFINE(__THREAD_per_paid, offsetof(struct thread_struct, per_event.paid));
 	DEFINE(__THREAD_trap_tdb, offsetof(struct thread_struct, trap_tdb));
 	BLANK();
-	DEFINE(__FPU_fpc, offsetof(struct fpu, fpc));
-	DEFINE(__FPU_flags, offsetof(struct fpu, flags));
-	DEFINE(__FPU_regs, offsetof(struct fpu, regs));
-	BLANK();
 	DEFINE(__TI_task, offsetof(struct thread_info, task));
 	DEFINE(__TI_flags, offsetof(struct thread_info, flags));
 	DEFINE(__TI_sysc_table, offsetof(struct thread_info, sys_call_table));
diff --git a/arch/s390/kernel/compat_signal.c b/arch/s390/kernel/compat_signal.c
index 0b46fd4aa31e..eb4664238613 100644
--- a/arch/s390/kernel/compat_signal.c
+++ b/arch/s390/kernel/compat_signal.c
@@ -154,7 +154,7 @@ int copy_siginfo_from_user32(siginfo_t *to, compat_siginfo_t __user *from)
 static void store_sigregs(void)
 {
 	save_access_regs(current->thread.acrs);
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 }
 
 /* Load registers after signal return */
@@ -286,7 +286,7 @@ COMPAT_SYSCALL_DEFINE0(sigreturn)
 	if (__copy_from_user(&set.sig, &frame->sc.oldmask, _SIGMASK_COPY_SIZE32))
 		goto badframe;
 	set_current_blocked(&set);
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 	if (restore_sigregs32(regs, &frame->sregs))
 		goto badframe;
 	if (restore_sigregs_ext32(regs, &frame->sregs_ext))
@@ -309,7 +309,7 @@ COMPAT_SYSCALL_DEFINE0(rt_sigreturn)
 	set_current_blocked(&set);
 	if (compat_restore_altstack(&frame->uc.uc_stack))
 		goto badframe;
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 	if (restore_sigregs32(regs, &frame->uc.uc_mcontext))
 		goto badframe;
 	if (restore_sigregs_ext32(regs, &frame->uc.uc_mcontext_ext))
diff --git a/arch/s390/kernel/entry.S b/arch/s390/kernel/entry.S
index 21c1219122af..5a966dea937f 100644
--- a/arch/s390/kernel/entry.S
+++ b/arch/s390/kernel/entry.S
@@ -183,7 +183,6 @@ ENTRY(sie64a)
 	xc	__SF_EMPTY+16(16,%r15),__SF_EMPTY+16(%r15) # host id & reason
 	tm	__LC_CPU_FLAGS+7,_CIF_FPU	# load guest fp/vx registers ?
 	jno	.Lsie_load_guest_gprs
-	lg	%r12,__LC_THREAD_INFO		# load fp/vx regs save area
 	brasl	%r14,load_fpu_regs		# load guest fp/vx regs
 .Lsie_load_guest_gprs:
 	lmg	%r0,%r13,0(%r3)			# load guest gprs 0-13
@@ -752,14 +751,16 @@ ENTRY(psw_idle)
  * of the register contents at system call or io return.
  */
 ENTRY(save_fpu_regs)
+	lg	%r2,__LC_CURRENT
+	aghi	%r2,__TASK_thread
 	tm	__LC_CPU_FLAGS+7,_CIF_FPU
 	bor	%r14
-	stfpc	__FPU_fpc(%r2)
+	stfpc	__THREAD_FPU_fpc(%r2)
 .Lsave_fpu_regs_fpc_end:
-	lg	%r3,__FPU_regs(%r2)
+	lg	%r3,__THREAD_FPU_regs(%r2)
 	ltgr	%r3,%r3
 	jz	.Lsave_fpu_regs_done	  # no save area -> set CIF_FPU
-	tm	__FPU_flags+3(%r2),FPU_USE_VX
+	tm	__THREAD_FPU_flags+3(%r2),FPU_USE_VX
 	jz	.Lsave_fpu_regs_fp	  # no -> store FP regs
 .Lsave_fpu_regs_vx_low:
 	VSTM	%v0,%v15,0,%r3		  # vstm 0,15,0(3)
@@ -794,20 +795,19 @@ ENTRY(save_fpu_regs)
  * FP/VX state, the vector-enablement control, CR0.46, is either set or cleared.
  *
  * There are special calling conventions to fit into sysc and io return work:
- *	%r12:	__LC_THREAD_INFO
  *	%r15:	<kernel stack>
  * The function requires:
  *	%r4 and __SF_EMPTY+32(%r15)
  */
 load_fpu_regs:
+	lg	%r4,__LC_CURRENT
+	aghi	%r4,__TASK_thread
 	tm	__LC_CPU_FLAGS+7,_CIF_FPU
 	bnor	%r14
-	lg	%r4,__TI_task(%r12)
-	la	%r4,__THREAD_fpu(%r4)
-	lfpc	__FPU_fpc(%r4)
+	lfpc	__THREAD_FPU_fpc(%r4)
 	stctg	%c0,%c0,__SF_EMPTY+32(%r15)	# store CR0
-	tm	__FPU_flags+3(%r4),FPU_USE_VX	# VX-enabled task ?
-	lg	%r4,__FPU_regs(%r4)		# %r4 <- reg save area
+	tm	__THREAD_FPU_flags+3(%r4),FPU_USE_VX	# VX-enabled task ?
+	lg	%r4,__THREAD_FPU_regs(%r4)	# %r4 <- reg save area
 	jz	.Lload_fpu_regs_fp_ctl		# -> no VX, load FP regs
 .Lload_fpu_regs_vx_ctl:
 	tm	__SF_EMPTY+32+5(%r15),2		# test VX control
@@ -1190,13 +1190,14 @@ cleanup_critical:
 	jhe	2f
 	clg	%r9,BASED(.Lcleanup_save_fpu_fpc_end)
 	jhe	1f
+	lg	%r2,__LC_CURRENT
 0:	# Store floating-point controls
-	stfpc	__FPU_fpc(%r2)
+	stfpc	__THREAD_FPU_fpc(%r2)
 1:	# Load register save area and check if VX is active
-	lg	%r3,__FPU_regs(%r2)
+	lg	%r3,__THREAD_FPU_regs(%r2)
 	ltgr	%r3,%r3
 	jz	5f			  # no save area -> set CIF_FPU
-	tm	__FPU_flags+3(%r2),FPU_USE_VX
+	tm	__THREAD_FPU_flags+3(%r2),FPU_USE_VX
 	jz	4f			  # no VX -> store FP regs
 2:	# Store vector registers (V0-V15)
 	VSTM	%v0,%v15,0,%r3		  # vstm 0,15,0(3)
@@ -1250,11 +1251,10 @@ cleanup_critical:
 	jhe	5f
 	clg	%r9,BASED(.Lcleanup_load_fpu_regs_vx_ctl)
 	jhe	6f
-	lg	%r4,__TI_task(%r12)
-	la	%r4,__THREAD_fpu(%r4)
-	lfpc	__FPU_fpc(%r4)
-	tm	__FPU_flags+3(%r4),FPU_USE_VX	# VX-enabled task ?
-	lg	%r4,__FPU_regs(%r4)		# %r4 <- reg save area
+	lg	%r4,__LC_CURRENT
+	lfpc	__THREAD_FPU_fpc(%r4)
+	tm	__THREAD_FPU_flags+3(%r4),FPU_USE_VX	# VX-enabled task ?
+	lg	%r4,__THREAD_FPU_regs(%r4)	# %r4 <- reg save area
 	jz	3f				# -> no VX, load FP regs
 6:	# Set VX-enablement control
 	stctg	%c0,%c0,__SF_EMPTY+32(%r15)	# store CR0
diff --git a/arch/s390/kernel/process.c b/arch/s390/kernel/process.c
index 9cf0063f920e..f2dac9f0799d 100644
--- a/arch/s390/kernel/process.c
+++ b/arch/s390/kernel/process.c
@@ -104,7 +104,7 @@ int arch_dup_task_struct(struct task_struct *dst, struct task_struct *src)
 	 * The CIF_FPU flag is set in any case to lazy clear or restore a saved
 	 * state when switching to a different task or returning to user space.
 	 */
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 	dst->thread.fpu.fpc = current->thread.fpu.fpc;
 	if (is_vx_task(current))
 		convert_vx_to_fp(dst->thread.fpu.fprs,
@@ -196,7 +196,7 @@ asmlinkage void execve_tail(void)
  */
 int dump_fpu (struct pt_regs * regs, s390_fp_regs *fpregs)
 {
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 	fpregs->fpc = current->thread.fpu.fpc;
 	fpregs->pad = 0;
 	if (is_vx_task(current))
diff --git a/arch/s390/kernel/ptrace.c b/arch/s390/kernel/ptrace.c
index 8c525880a3ff..8b1c8e33f184 100644
--- a/arch/s390/kernel/ptrace.c
+++ b/arch/s390/kernel/ptrace.c
@@ -943,7 +943,7 @@ static int s390_fpregs_get(struct task_struct *target,
 	_s390_fp_regs fp_regs;
 
 	if (target == current)
-		save_fpu_regs(&target->thread.fpu);
+		save_fpu_regs();
 
 	fp_regs.fpc = target->thread.fpu.fpc;
 	fpregs_store(&fp_regs, &target->thread.fpu);
@@ -961,7 +961,7 @@ static int s390_fpregs_set(struct task_struct *target,
 	freg_t fprs[__NUM_FPRS];
 
 	if (target == current)
-		save_fpu_regs(&target->thread.fpu);
+		save_fpu_regs();
 
 	/* If setting FPC, must validate it first. */
 	if (count > 0 && pos < offsetof(s390_fp_regs, fprs)) {
@@ -1049,7 +1049,7 @@ static int s390_vxrs_low_get(struct task_struct *target,
 		return -ENODEV;
 	if (is_vx_task(target)) {
 		if (target == current)
-			save_fpu_regs(&target->thread.fpu);
+			save_fpu_regs();
 		for (i = 0; i < __NUM_VXRS_LOW; i++)
 			vxrs[i] = *((__u64 *)(target->thread.fpu.vxrs + i) + 1);
 	} else
@@ -1072,7 +1072,7 @@ static int s390_vxrs_low_set(struct task_struct *target,
 		if (rc)
 			return rc;
 	} else if (target == current)
-		save_fpu_regs(&target->thread.fpu);
+		save_fpu_regs();
 
 	rc = user_regset_copyin(&pos, &count, &kbuf, &ubuf, vxrs, 0, -1);
 	if (rc == 0)
@@ -1093,7 +1093,7 @@ static int s390_vxrs_high_get(struct task_struct *target,
 		return -ENODEV;
 	if (is_vx_task(target)) {
 		if (target == current)
-			save_fpu_regs(&target->thread.fpu);
+			save_fpu_regs();
 		memcpy(vxrs, target->thread.fpu.vxrs + __NUM_VXRS_LOW,
 		       sizeof(vxrs));
 	} else
@@ -1115,7 +1115,7 @@ static int s390_vxrs_high_set(struct task_struct *target,
 		if (rc)
 			return rc;
 	} else if (target == current)
-		save_fpu_regs(&target->thread.fpu);
+		save_fpu_regs();
 
 	rc = user_regset_copyin(&pos, &count, &kbuf, &ubuf,
 				target->thread.fpu.vxrs + __NUM_VXRS_LOW, 0, -1);
diff --git a/arch/s390/kernel/signal.c b/arch/s390/kernel/signal.c
index 2f4c7e2638c9..9549af102d75 100644
--- a/arch/s390/kernel/signal.c
+++ b/arch/s390/kernel/signal.c
@@ -105,7 +105,7 @@ struct rt_sigframe
 static void store_sigregs(void)
 {
 	save_access_regs(current->thread.acrs);
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 }
 
 /* Load registers after signal return */
@@ -222,7 +222,7 @@ SYSCALL_DEFINE0(sigreturn)
 	if (__copy_from_user(&set.sig, &frame->sc.oldmask, _SIGMASK_COPY_SIZE))
 		goto badframe;
 	set_current_blocked(&set);
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 	if (restore_sigregs(regs, &frame->sregs))
 		goto badframe;
 	if (restore_sigregs_ext(regs, &frame->sregs_ext))
@@ -246,7 +246,7 @@ SYSCALL_DEFINE0(rt_sigreturn)
 	set_current_blocked(&set);
 	if (restore_altstack(&frame->uc.uc_stack))
 		goto badframe;
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 	if (restore_sigregs(regs, &frame->uc.uc_mcontext))
 		goto badframe;
 	if (restore_sigregs_ext(regs, &frame->uc.uc_mcontext_ext))
diff --git a/arch/s390/kernel/traps.c b/arch/s390/kernel/traps.c
index 76f76932ccb9..9861613fb35a 100644
--- a/arch/s390/kernel/traps.c
+++ b/arch/s390/kernel/traps.c
@@ -236,7 +236,7 @@ int alloc_vector_registers(struct task_struct *tsk)
 		return -ENOMEM;
 	preempt_disable();
 	if (tsk == current)
-		save_fpu_regs(&tsk->thread.fpu);
+		save_fpu_regs();
 	/* Copy the 16 floating point registers */
 	convert_fp_to_vx(vxrs, tsk->thread.fpu.fprs);
 	fprs = tsk->thread.fpu.fprs;
@@ -257,7 +257,7 @@ void vector_exception(struct pt_regs *regs)
 	}
 
 	/* get vector interrupt code from fpc */
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 	vic = (current->thread.fpu.fpc & 0xf00) >> 8;
 	switch (vic) {
 	case 1: /* invalid vector operation */
@@ -295,7 +295,7 @@ void data_exception(struct pt_regs *regs)
 
 	location = get_trap_ip(regs);
 
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 	/* Check for vector register enablement */
 	if (MACHINE_HAS_VX && !is_vx_task(current) &&
 	    (current->thread.fpu.fpc & FPC_DXC_MASK) == 0xfe00) {
diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
index c0cceaf4a92e..1903f0212bd0 100644
--- a/arch/s390/kvm/kvm-s390.c
+++ b/arch/s390/kvm/kvm-s390.c
@@ -1224,7 +1224,7 @@ static inline void load_fpu_from(struct fpu *from)
 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 {
 	/* Save host register state */
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 	save_fpu_to(&vcpu->arch.host_fpregs);
 
 	if (test_kvm_facility(vcpu->kvm, 129)) {
@@ -1256,7 +1256,7 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
 	atomic_clear_mask(CPUSTAT_RUNNING, &vcpu->arch.sie_block->cpuflags);
 	gmap_disable(vcpu->arch.gmap);
 
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 
 	if (test_kvm_facility(vcpu->kvm, 129))
 		/*
@@ -1671,7 +1671,7 @@ int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu)
 		return -EINVAL;
 	memcpy(vcpu->arch.guest_fpregs.fprs, &fpu->fprs, sizeof(fpu->fprs));
 	vcpu->arch.guest_fpregs.fpc = fpu->fpc;
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 	load_fpu_from(&vcpu->arch.guest_fpregs);
 	return 0;
 }
@@ -2241,7 +2241,7 @@ int kvm_s390_vcpu_store_status(struct kvm_vcpu *vcpu, unsigned long addr)
 	 * copying in vcpu load/put. Lets update our copies before we save
 	 * it into the save area
 	 */
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 	if (test_kvm_facility(vcpu->kvm, 129)) {
 		/*
 		 * If the vector extension is available, the vector registers
@@ -2288,7 +2288,7 @@ int kvm_s390_vcpu_store_adtl_status(struct kvm_vcpu *vcpu, unsigned long addr)
 	 *
 	 * Let's update our copies before we save it into the save area.
 	 */
-	save_fpu_regs(&current->thread.fpu);
+	save_fpu_regs();
 
 	return kvm_s390_store_adtl_status_unloaded(vcpu, addr);
 }