kernel_optimize_test/arch/mips/kernel/r4k_switch.S
Atsushi Nemoto c138e12f3a [MIPS] Fix fpu_save_double on 64-bit.
> Without this fix, _save_fp() in 64-bit kernel is seriously broken.
>
> ffffffff8010bec0 <_save_fp>:
> ffffffff8010bec0:       400d6000        mfc0    t1,c0_status
> ffffffff8010bec4:       000c7140        sll     t2,t0,0x5
> ffffffff8010bec8:       05c10011        bgez    t2,ffffffff8010bf10 <_save_fp+0x50>
> ffffffff8010becc:       00000000        nop
> ffffffff8010bed0:       f4810328        sdc1    $f1,808(a0)
> ...

Fix register usage in fpu_save_double() and make fpu_restore_double()
more symmetric with fpu_save_double().

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
2006-06-19 17:39:13 +01:00

257 lines
5.1 KiB
ArmAsm

/*
* This file is subject to the terms and conditions of the GNU General Public
* License. See the file "COPYING" in the main directory of this archive
* for more details.
*
* Copyright (C) 1994, 1995, 1996, 1998, 1999, 2002, 2003 Ralf Baechle
* Copyright (C) 1996 David S. Miller (dm@engr.sgi.com)
* Copyright (C) 1994, 1995, 1996, by Andreas Busse
* Copyright (C) 1999 Silicon Graphics, Inc.
* Copyright (C) 2000 MIPS Technologies, Inc.
* written by Carsten Langgaard, carstenl@mips.com
*/
#include <linux/config.h>
#include <asm/asm.h>
#include <asm/cachectl.h>
#include <asm/fpregdef.h>
#include <asm/mipsregs.h>
#include <asm/asm-offsets.h>
#include <asm/page.h>
#include <asm/pgtable-bits.h>
#include <asm/regdef.h>
#include <asm/stackframe.h>
#include <asm/thread_info.h>
#include <asm/asmmacro.h>
/*
* Offset to the current process status flags, the first 32 bytes of the
* stack are not used.
*/
#define ST_OFF (_THREAD_SIZE - 32 - PT_SIZE + PT_STATUS)
/*
* FPU context is saved iff the process has used it's FPU in the current
* time slice as indicated by _TIF_USEDFPU. In any case, the CU1 bit for user
* space STATUS register should be 0, so that a process *always* starts its
* userland with FPU disabled after each context switch.
*
* FPU will be enabled as soon as the process accesses FPU again, through
* do_cpu() trap.
*/
/*
* task_struct *resume(task_struct *prev, task_struct *next,
* struct thread_info *next_ti)
*/
.align 5
LEAF(resume)
#ifndef CONFIG_CPU_HAS_LLSC
sw zero, ll_bit
#endif
mfc0 t1, CP0_STATUS
LONG_S t1, THREAD_STATUS(a0)
cpu_save_nonscratch a0
LONG_S ra, THREAD_REG31(a0)
/*
* check if we need to save FPU registers
*/
PTR_L t3, TASK_THREAD_INFO(a0)
LONG_L t0, TI_FLAGS(t3)
li t1, _TIF_USEDFPU
and t2, t0, t1
beqz t2, 1f
nor t1, zero, t1
and t0, t0, t1
LONG_S t0, TI_FLAGS(t3)
/*
* clear saved user stack CU1 bit
*/
LONG_L t0, ST_OFF(t3)
li t1, ~ST0_CU1
and t0, t0, t1
LONG_S t0, ST_OFF(t3)
fpu_save_double a0 t0 t1 # c0_status passed in t0
# clobbers t1
1:
/*
* The order of restoring the registers takes care of the race
* updating $28, $29 and kernelsp without disabling ints.
*/
move $28, a2
cpu_restore_nonscratch a1
PTR_ADDIU t0, $28, _THREAD_SIZE - 32
set_saved_sp t0, t1, t2
#ifdef CONFIG_MIPS_MT_SMTC
/* Read-modify-writes of Status must be atomic on a VPE */
mfc0 t2, CP0_TCSTATUS
ori t1, t2, TCSTATUS_IXMT
mtc0 t1, CP0_TCSTATUS
andi t2, t2, TCSTATUS_IXMT
ehb
DMT 8 # dmt t0
move t1,ra
jal mips_ihb
move ra,t1
#endif /* CONFIG_MIPS_MT_SMTC */
mfc0 t1, CP0_STATUS /* Do we really need this? */
li a3, 0xff01
and t1, a3
LONG_L a2, THREAD_STATUS(a1)
nor a3, $0, a3
and a2, a3
or a2, t1
mtc0 a2, CP0_STATUS
#ifdef CONFIG_MIPS_MT_SMTC
ehb
andi t0, t0, VPECONTROL_TE
beqz t0, 1f
emt
1:
mfc0 t1, CP0_TCSTATUS
xori t1, t1, TCSTATUS_IXMT
or t1, t1, t2
mtc0 t1, CP0_TCSTATUS
ehb
#endif /* CONFIG_MIPS_MT_SMTC */
move v0, a0
jr ra
END(resume)
/*
* Save a thread's fp context.
*/
LEAF(_save_fp)
#ifdef CONFIG_64BIT
mfc0 t0, CP0_STATUS
#endif
fpu_save_double a0 t0 t1 # clobbers t1
jr ra
END(_save_fp)
/*
* Restore a thread's fp context.
*/
LEAF(_restore_fp)
#ifdef CONFIG_64BIT
mfc0 t0, CP0_STATUS
#endif
fpu_restore_double a0 t0 t1 # clobbers t1
jr ra
END(_restore_fp)
/*
* Load the FPU with signalling NANS. This bit pattern we're using has
* the property that no matter whether considered as single or as double
* precision represents signaling NANS.
*
* We initialize fcr31 to rounding to nearest, no exceptions.
*/
#define FPU_DEFAULT 0x00000000
LEAF(_init_fpu)
#ifdef CONFIG_MIPS_MT_SMTC
/* Rather than manipulate per-VPE Status, set per-TC bit in TCStatus */
mfc0 t0, CP0_TCSTATUS
/* Bit position is the same for Status, TCStatus */
li t1, ST0_CU1
or t0, t1
mtc0 t0, CP0_TCSTATUS
#else /* Normal MIPS CU1 enable */
mfc0 t0, CP0_STATUS
li t1, ST0_CU1
or t0, t1
mtc0 t0, CP0_STATUS
#endif /* CONFIG_MIPS_MT_SMTC */
fpu_enable_hazard
li t1, FPU_DEFAULT
ctc1 t1, fcr31
li t1, -1 # SNaN
#ifdef CONFIG_64BIT
sll t0, t0, 5
bgez t0, 1f # 16 / 32 register mode?
dmtc1 t1, $f1
dmtc1 t1, $f3
dmtc1 t1, $f5
dmtc1 t1, $f7
dmtc1 t1, $f9
dmtc1 t1, $f11
dmtc1 t1, $f13
dmtc1 t1, $f15
dmtc1 t1, $f17
dmtc1 t1, $f19
dmtc1 t1, $f21
dmtc1 t1, $f23
dmtc1 t1, $f25
dmtc1 t1, $f27
dmtc1 t1, $f29
dmtc1 t1, $f31
1:
#endif
#ifdef CONFIG_CPU_MIPS32
mtc1 t1, $f0
mtc1 t1, $f1
mtc1 t1, $f2
mtc1 t1, $f3
mtc1 t1, $f4
mtc1 t1, $f5
mtc1 t1, $f6
mtc1 t1, $f7
mtc1 t1, $f8
mtc1 t1, $f9
mtc1 t1, $f10
mtc1 t1, $f11
mtc1 t1, $f12
mtc1 t1, $f13
mtc1 t1, $f14
mtc1 t1, $f15
mtc1 t1, $f16
mtc1 t1, $f17
mtc1 t1, $f18
mtc1 t1, $f19
mtc1 t1, $f20
mtc1 t1, $f21
mtc1 t1, $f22
mtc1 t1, $f23
mtc1 t1, $f24
mtc1 t1, $f25
mtc1 t1, $f26
mtc1 t1, $f27
mtc1 t1, $f28
mtc1 t1, $f29
mtc1 t1, $f30
mtc1 t1, $f31
#else
.set mips3
dmtc1 t1, $f0
dmtc1 t1, $f2
dmtc1 t1, $f4
dmtc1 t1, $f6
dmtc1 t1, $f8
dmtc1 t1, $f10
dmtc1 t1, $f12
dmtc1 t1, $f14
dmtc1 t1, $f16
dmtc1 t1, $f18
dmtc1 t1, $f20
dmtc1 t1, $f22
dmtc1 t1, $f24
dmtc1 t1, $f26
dmtc1 t1, $f28
dmtc1 t1, $f30
#endif
jr ra
END(_init_fpu)