android_kernel_samsung_hero.../arch/mn10300/mm/misalignment.c
2016-08-17 16:41:52 +08:00

966 lines
29 KiB
C

/* MN10300 Misalignment fixup handler
*
* Copyright (C) 2007 Red Hat, Inc. All Rights Reserved.
* Written by David Howells (dhowells@redhat.com)
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public Licence
* as published by the Free Software Foundation; either version
* 2 of the Licence, or (at your option) any later version.
*/
#include <linux/module.h>
#include <linux/sched.h>
#include <linux/kernel.h>
#include <linux/string.h>
#include <linux/errno.h>
#include <linux/ptrace.h>
#include <linux/timer.h>
#include <linux/mm.h>
#include <linux/smp.h>
#include <linux/init.h>
#include <linux/delay.h>
#include <linux/spinlock.h>
#include <linux/interrupt.h>
#include <linux/pci.h>
#include <asm/processor.h>
#include <asm/uaccess.h>
#include <asm/io.h>
#include <linux/atomic.h>
#include <asm/smp.h>
#include <asm/pgalloc.h>
#include <asm/cpu-regs.h>
#include <asm/busctl-regs.h>
#include <asm/fpu.h>
#include <asm/gdb-stub.h>
#include <asm/asm-offsets.h>
#if 0
#define kdebug(FMT, ...) printk(KERN_DEBUG "MISALIGN: "FMT"\n", ##__VA_ARGS__)
#else
#define kdebug(FMT, ...) do {} while (0)
#endif
static int misalignment_addr(unsigned long *registers, unsigned long sp,
unsigned params, unsigned opcode,
unsigned long disp,
void **_address, unsigned long **_postinc,
unsigned long *_inc);
static int misalignment_reg(unsigned long *registers, unsigned params,
unsigned opcode, unsigned long disp,
unsigned long **_register);
static void misalignment_MOV_Lcc(struct pt_regs *regs, uint32_t opcode);
static const unsigned Dreg_index[] = {
REG_D0 >> 2, REG_D1 >> 2, REG_D2 >> 2, REG_D3 >> 2
};
static const unsigned Areg_index[] = {
REG_A0 >> 2, REG_A1 >> 2, REG_A2 >> 2, REG_A3 >> 2
};
static const unsigned Rreg_index[] = {
REG_E0 >> 2, REG_E1 >> 2, REG_E2 >> 2, REG_E3 >> 2,
REG_E4 >> 2, REG_E5 >> 2, REG_E6 >> 2, REG_E7 >> 2,
REG_A0 >> 2, REG_A1 >> 2, REG_A2 >> 2, REG_A3 >> 2,
REG_D0 >> 2, REG_D1 >> 2, REG_D2 >> 2, REG_D3 >> 2
};
enum format_id {
FMT_S0,
FMT_S1,
FMT_S2,
FMT_S4,
FMT_D0,
FMT_D1,
FMT_D2,
FMT_D4,
FMT_D6,
FMT_D7,
FMT_D8,
FMT_D9,
FMT_D10,
};
static const struct {
u_int8_t opsz, dispsz;
} format_tbl[16] = {
[FMT_S0] = { 8, 0 },
[FMT_S1] = { 8, 8 },
[FMT_S2] = { 8, 16 },
[FMT_S4] = { 8, 32 },
[FMT_D0] = { 16, 0 },
[FMT_D1] = { 16, 8 },
[FMT_D2] = { 16, 16 },
[FMT_D4] = { 16, 32 },
[FMT_D6] = { 24, 0 },
[FMT_D7] = { 24, 8 },
[FMT_D8] = { 24, 24 },
[FMT_D9] = { 24, 32 },
[FMT_D10] = { 32, 0 },
};
enum value_id {
DM0, /* data reg in opcode in bits 0-1 */
DM1, /* data reg in opcode in bits 2-3 */
DM2, /* data reg in opcode in bits 4-5 */
AM0, /* addr reg in opcode in bits 0-1 */
AM1, /* addr reg in opcode in bits 2-3 */
AM2, /* addr reg in opcode in bits 4-5 */
RM0, /* reg in opcode in bits 0-3 */
RM1, /* reg in opcode in bits 2-5 */
RM2, /* reg in opcode in bits 4-7 */
RM4, /* reg in opcode in bits 8-11 */
RM6, /* reg in opcode in bits 12-15 */
RD0, /* reg in displacement in bits 0-3 */
RD2, /* reg in displacement in bits 4-7 */
SP, /* stack pointer */
SD8, /* 8-bit signed displacement */
SD16, /* 16-bit signed displacement */
SD24, /* 24-bit signed displacement */
SIMM4_2, /* 4-bit signed displacement in opcode bits 4-7 */
SIMM8, /* 8-bit signed immediate */
IMM8, /* 8-bit unsigned immediate */
IMM16, /* 16-bit unsigned immediate */
IMM24, /* 24-bit unsigned immediate */
IMM32, /* 32-bit unsigned immediate */
IMM32_HIGH8, /* 32-bit unsigned immediate, LSB in opcode */
IMM32_MEM, /* 32-bit unsigned displacement */
IMM32_HIGH8_MEM, /* 32-bit unsigned displacement, LSB in opcode */
DN0 = DM0,
DN1 = DM1,
DN2 = DM2,
AN0 = AM0,
AN1 = AM1,
AN2 = AM2,
RN0 = RM0,
RN1 = RM1,
RN2 = RM2,
RN4 = RM4,
RN6 = RM6,
DI = DM1,
RI = RM2,
};
struct mn10300_opcode {
const char name[8];
u_int32_t opcode;
u_int32_t opmask;
unsigned exclusion;
enum format_id format;
unsigned cpu_mask;
#define AM33 330
unsigned params[2];
#define MEM(ADDR) (0x80000000 | (ADDR))
#define MEM2(ADDR1, ADDR2) (0x80000000 | (ADDR1) << 8 | (ADDR2))
#define MEMINC(ADDR) (0x81000000 | (ADDR))
#define MEMINC2(ADDR, INC) (0x81000000 | (ADDR) << 8 | (INC))
};
/* LIBOPCODES EXCERPT
Assemble Matsushita MN10300 instructions.
Copyright 1996, 1997, 1998, 1999, 2000 Free Software Foundation, Inc.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public Licence as published by
the Free Software Foundation; either version 2 of the Licence, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public Licence for more details.
You should have received a copy of the GNU General Public Licence
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*/
static const struct mn10300_opcode mn10300_opcodes[] = {
{ "mov", 0x4200, 0xf300, 0, FMT_S1, 0, {DM1, MEM2(IMM8, SP)}},
{ "mov", 0x4300, 0xf300, 0, FMT_S1, 0, {AM1, MEM2(IMM8, SP)}},
{ "mov", 0x5800, 0xfc00, 0, FMT_S1, 0, {MEM2(IMM8, SP), DN0}},
{ "mov", 0x5c00, 0xfc00, 0, FMT_S1, 0, {MEM2(IMM8, SP), AN0}},
{ "mov", 0x60, 0xf0, 0, FMT_S0, 0, {DM1, MEM(AN0)}},
{ "mov", 0x70, 0xf0, 0, FMT_S0, 0, {MEM(AM0), DN1}},
{ "mov", 0xf000, 0xfff0, 0, FMT_D0, 0, {MEM(AM0), AN1}},
{ "mov", 0xf010, 0xfff0, 0, FMT_D0, 0, {AM1, MEM(AN0)}},
{ "mov", 0xf300, 0xffc0, 0, FMT_D0, 0, {MEM2(DI, AM0), DN2}},
{ "mov", 0xf340, 0xffc0, 0, FMT_D0, 0, {DM2, MEM2(DI, AN0)}},
{ "mov", 0xf380, 0xffc0, 0, FMT_D0, 0, {MEM2(DI, AM0), AN2}},
{ "mov", 0xf3c0, 0xffc0, 0, FMT_D0, 0, {AM2, MEM2(DI, AN0)}},
{ "mov", 0xf80000, 0xfff000, 0, FMT_D1, 0, {MEM2(SD8, AM0), DN1}},
{ "mov", 0xf81000, 0xfff000, 0, FMT_D1, 0, {DM1, MEM2(SD8, AN0)}},
{ "mov", 0xf82000, 0xfff000, 0, FMT_D1, 0, {MEM2(SD8,AM0), AN1}},
{ "mov", 0xf83000, 0xfff000, 0, FMT_D1, 0, {AM1, MEM2(SD8, AN0)}},
{ "mov", 0xf90a00, 0xffff00, 0, FMT_D6, AM33, {MEM(RM0), RN2}},
{ "mov", 0xf91a00, 0xffff00, 0, FMT_D6, AM33, {RM2, MEM(RN0)}},
{ "mov", 0xf96a00, 0xffff00, 0x12, FMT_D6, AM33, {MEMINC(RM0), RN2}},
{ "mov", 0xf97a00, 0xffff00, 0, FMT_D6, AM33, {RM2, MEMINC(RN0)}},
{ "mov", 0xfa000000, 0xfff00000, 0, FMT_D2, 0, {MEM2(SD16, AM0), DN1}},
{ "mov", 0xfa100000, 0xfff00000, 0, FMT_D2, 0, {DM1, MEM2(SD16, AN0)}},
{ "mov", 0xfa200000, 0xfff00000, 0, FMT_D2, 0, {MEM2(SD16, AM0), AN1}},
{ "mov", 0xfa300000, 0xfff00000, 0, FMT_D2, 0, {AM1, MEM2(SD16, AN0)}},
{ "mov", 0xfa900000, 0xfff30000, 0, FMT_D2, 0, {AM1, MEM2(IMM16, SP)}},
{ "mov", 0xfa910000, 0xfff30000, 0, FMT_D2, 0, {DM1, MEM2(IMM16, SP)}},
{ "mov", 0xfab00000, 0xfffc0000, 0, FMT_D2, 0, {MEM2(IMM16, SP), AN0}},
{ "mov", 0xfab40000, 0xfffc0000, 0, FMT_D2, 0, {MEM2(IMM16, SP), DN0}},
{ "mov", 0xfb0a0000, 0xffff0000, 0, FMT_D7, AM33, {MEM2(SD8, RM0), RN2}},
{ "mov", 0xfb1a0000, 0xffff0000, 0, FMT_D7, AM33, {RM2, MEM2(SD8, RN0)}},
{ "mov", 0xfb6a0000, 0xffff0000, 0x22, FMT_D7, AM33, {MEMINC2 (RM0, SIMM8), RN2}},
{ "mov", 0xfb7a0000, 0xffff0000, 0, FMT_D7, AM33, {RM2, MEMINC2 (RN0, SIMM8)}},
{ "mov", 0xfb8a0000, 0xffff0f00, 0, FMT_D7, AM33, {MEM2(IMM8, SP), RN2}},
{ "mov", 0xfb8e0000, 0xffff000f, 0, FMT_D7, AM33, {MEM2(RI, RM0), RD2}},
{ "mov", 0xfb9a0000, 0xffff0f00, 0, FMT_D7, AM33, {RM2, MEM2(IMM8, SP)}},
{ "mov", 0xfb9e0000, 0xffff000f, 0, FMT_D7, AM33, {RD2, MEM2(RI, RN0)}},
{ "mov", 0xfc000000, 0xfff00000, 0, FMT_D4, 0, {MEM2(IMM32,AM0), DN1}},
{ "mov", 0xfc100000, 0xfff00000, 0, FMT_D4, 0, {DM1, MEM2(IMM32,AN0)}},
{ "mov", 0xfc200000, 0xfff00000, 0, FMT_D4, 0, {MEM2(IMM32,AM0), AN1}},
{ "mov", 0xfc300000, 0xfff00000, 0, FMT_D4, 0, {AM1, MEM2(IMM32,AN0)}},
{ "mov", 0xfc800000, 0xfff30000, 0, FMT_D4, 0, {AM1, MEM(IMM32_MEM)}},
{ "mov", 0xfc810000, 0xfff30000, 0, FMT_D4, 0, {DM1, MEM(IMM32_MEM)}},
{ "mov", 0xfc900000, 0xfff30000, 0, FMT_D4, 0, {AM1, MEM2(IMM32, SP)}},
{ "mov", 0xfc910000, 0xfff30000, 0, FMT_D4, 0, {DM1, MEM2(IMM32, SP)}},
{ "mov", 0xfca00000, 0xfffc0000, 0, FMT_D4, 0, {MEM(IMM32_MEM), AN0}},
{ "mov", 0xfca40000, 0xfffc0000, 0, FMT_D4, 0, {MEM(IMM32_MEM), DN0}},
{ "mov", 0xfcb00000, 0xfffc0000, 0, FMT_D4, 0, {MEM2(IMM32, SP), AN0}},
{ "mov", 0xfcb40000, 0xfffc0000, 0, FMT_D4, 0, {MEM2(IMM32, SP), DN0}},
{ "mov", 0xfd0a0000, 0xffff0000, 0, FMT_D8, AM33, {MEM2(SD24, RM0), RN2}},
{ "mov", 0xfd1a0000, 0xffff0000, 0, FMT_D8, AM33, {RM2, MEM2(SD24, RN0)}},
{ "mov", 0xfd6a0000, 0xffff0000, 0x22, FMT_D8, AM33, {MEMINC2 (RM0, IMM24), RN2}},
{ "mov", 0xfd7a0000, 0xffff0000, 0, FMT_D8, AM33, {RM2, MEMINC2 (RN0, IMM24)}},
{ "mov", 0xfd8a0000, 0xffff0f00, 0, FMT_D8, AM33, {MEM2(IMM24, SP), RN2}},
{ "mov", 0xfd9a0000, 0xffff0f00, 0, FMT_D8, AM33, {RM2, MEM2(IMM24, SP)}},
{ "mov", 0xfe0a0000, 0xffff0000, 0, FMT_D9, AM33, {MEM2(IMM32_HIGH8,RM0), RN2}},
{ "mov", 0xfe0a0000, 0xffff0000, 0, FMT_D9, AM33, {MEM2(IMM32_HIGH8,RM0), RN2}},
{ "mov", 0xfe0e0000, 0xffff0f00, 0, FMT_D9, AM33, {MEM(IMM32_HIGH8_MEM), RN2}},
{ "mov", 0xfe1a0000, 0xffff0000, 0, FMT_D9, AM33, {RM2, MEM2(IMM32_HIGH8, RN0)}},
{ "mov", 0xfe1a0000, 0xffff0000, 0, FMT_D9, AM33, {RM2, MEM2(IMM32_HIGH8, RN0)}},
{ "mov", 0xfe1e0000, 0xffff0f00, 0, FMT_D9, AM33, {RM2, MEM(IMM32_HIGH8_MEM)}},
{ "mov", 0xfe6a0000, 0xffff0000, 0x22, FMT_D9, AM33, {MEMINC2 (RM0, IMM32_HIGH8), RN2}},
{ "mov", 0xfe7a0000, 0xffff0000, 0, FMT_D9, AM33, {RN2, MEMINC2 (RM0, IMM32_HIGH8)}},
{ "mov", 0xfe8a0000, 0xffff0f00, 0, FMT_D9, AM33, {MEM2(IMM32_HIGH8, SP), RN2}},
{ "mov", 0xfe9a0000, 0xffff0f00, 0, FMT_D9, AM33, {RM2, MEM2(IMM32_HIGH8, SP)}},
{ "movhu", 0xf060, 0xfff0, 0, FMT_D0, 0, {MEM(AM0), DN1}},
{ "movhu", 0xf070, 0xfff0, 0, FMT_D0, 0, {DM1, MEM(AN0)}},
{ "movhu", 0xf480, 0xffc0, 0, FMT_D0, 0, {MEM2(DI, AM0), DN2}},
{ "movhu", 0xf4c0, 0xffc0, 0, FMT_D0, 0, {DM2, MEM2(DI, AN0)}},
{ "movhu", 0xf86000, 0xfff000, 0, FMT_D1, 0, {MEM2(SD8, AM0), DN1}},
{ "movhu", 0xf87000, 0xfff000, 0, FMT_D1, 0, {DM1, MEM2(SD8, AN0)}},
{ "movhu", 0xf89300, 0xfff300, 0, FMT_D1, 0, {DM1, MEM2(IMM8, SP)}},
{ "movhu", 0xf8bc00, 0xfffc00, 0, FMT_D1, 0, {MEM2(IMM8, SP), DN0}},
{ "movhu", 0xf94a00, 0xffff00, 0, FMT_D6, AM33, {MEM(RM0), RN2}},
{ "movhu", 0xf95a00, 0xffff00, 0, FMT_D6, AM33, {RM2, MEM(RN0)}},
{ "movhu", 0xf9ea00, 0xffff00, 0x12, FMT_D6, AM33, {MEMINC(RM0), RN2}},
{ "movhu", 0xf9fa00, 0xffff00, 0, FMT_D6, AM33, {RM2, MEMINC(RN0)}},
{ "movhu", 0xfa600000, 0xfff00000, 0, FMT_D2, 0, {MEM2(SD16, AM0), DN1}},
{ "movhu", 0xfa700000, 0xfff00000, 0, FMT_D2, 0, {DM1, MEM2(SD16, AN0)}},
{ "movhu", 0xfa930000, 0xfff30000, 0, FMT_D2, 0, {DM1, MEM2(IMM16, SP)}},
{ "movhu", 0xfabc0000, 0xfffc0000, 0, FMT_D2, 0, {MEM2(IMM16, SP), DN0}},
{ "movhu", 0xfb4a0000, 0xffff0000, 0, FMT_D7, AM33, {MEM2(SD8, RM0), RN2}},
{ "movhu", 0xfb5a0000, 0xffff0000, 0, FMT_D7, AM33, {RM2, MEM2(SD8, RN0)}},
{ "movhu", 0xfbca0000, 0xffff0f00, 0, FMT_D7, AM33, {MEM2(IMM8, SP), RN2}},
{ "movhu", 0xfbce0000, 0xffff000f, 0, FMT_D7, AM33, {MEM2(RI, RM0), RD2}},
{ "movhu", 0xfbda0000, 0xffff0f00, 0, FMT_D7, AM33, {RM2, MEM2(IMM8, SP)}},
{ "movhu", 0xfbde0000, 0xffff000f, 0, FMT_D7, AM33, {RD2, MEM2(RI, RN0)}},
{ "movhu", 0xfbea0000, 0xffff0000, 0x22, FMT_D7, AM33, {MEMINC2 (RM0, SIMM8), RN2}},
{ "movhu", 0xfbfa0000, 0xffff0000, 0, FMT_D7, AM33, {RM2, MEMINC2 (RN0, SIMM8)}},
{ "movhu", 0xfc600000, 0xfff00000, 0, FMT_D4, 0, {MEM2(IMM32,AM0), DN1}},
{ "movhu", 0xfc700000, 0xfff00000, 0, FMT_D4, 0, {DM1, MEM2(IMM32,AN0)}},
{ "movhu", 0xfc830000, 0xfff30000, 0, FMT_D4, 0, {DM1, MEM(IMM32_MEM)}},
{ "movhu", 0xfc930000, 0xfff30000, 0, FMT_D4, 0, {DM1, MEM2(IMM32, SP)}},
{ "movhu", 0xfcac0000, 0xfffc0000, 0, FMT_D4, 0, {MEM(IMM32_MEM), DN0}},
{ "movhu", 0xfcbc0000, 0xfffc0000, 0, FMT_D4, 0, {MEM2(IMM32, SP), DN0}},
{ "movhu", 0xfd4a0000, 0xffff0000, 0, FMT_D8, AM33, {MEM2(SD24, RM0), RN2}},
{ "movhu", 0xfd5a0000, 0xffff0000, 0, FMT_D8, AM33, {RM2, MEM2(SD24, RN0)}},
{ "movhu", 0xfdca0000, 0xffff0f00, 0, FMT_D8, AM33, {MEM2(IMM24, SP), RN2}},
{ "movhu", 0xfdda0000, 0xffff0f00, 0, FMT_D8, AM33, {RM2, MEM2(IMM24, SP)}},
{ "movhu", 0xfdea0000, 0xffff0000, 0x22, FMT_D8, AM33, {MEMINC2 (RM0, IMM24), RN2}},
{ "movhu", 0xfdfa0000, 0xffff0000, 0, FMT_D8, AM33, {RM2, MEMINC2 (RN0, IMM24)}},
{ "movhu", 0xfe4a0000, 0xffff0000, 0, FMT_D9, AM33, {MEM2(IMM32_HIGH8,RM0), RN2}},
{ "movhu", 0xfe4e0000, 0xffff0f00, 0, FMT_D9, AM33, {MEM(IMM32_HIGH8_MEM), RN2}},
{ "movhu", 0xfe5a0000, 0xffff0000, 0, FMT_D9, AM33, {RM2, MEM2(IMM32_HIGH8, RN0)}},
{ "movhu", 0xfe5e0000, 0xffff0f00, 0, FMT_D9, AM33, {RM2, MEM(IMM32_HIGH8_MEM)}},
{ "movhu", 0xfeca0000, 0xffff0f00, 0, FMT_D9, AM33, {MEM2(IMM32_HIGH8, SP), RN2}},
{ "movhu", 0xfeda0000, 0xffff0f00, 0, FMT_D9, AM33, {RM2, MEM2(IMM32_HIGH8, SP)}},
{ "movhu", 0xfeea0000, 0xffff0000, 0x22, FMT_D9, AM33, {MEMINC2 (RM0, IMM32_HIGH8), RN2}},
{ "movhu", 0xfefa0000, 0xffff0000, 0, FMT_D9, AM33, {RN2, MEMINC2 (RM0, IMM32_HIGH8)}},
{ "mov_llt", 0xf7e00000, 0xffff000f, 0x22, FMT_D10, AM33, {MEMINC2 (RN4,SIMM4_2), RM6}},
{ "mov_lgt", 0xf7e00001, 0xffff000f, 0x22, FMT_D10, AM33, {MEMINC2 (RN4,SIMM4_2), RM6}},
{ "mov_lge", 0xf7e00002, 0xffff000f, 0x22, FMT_D10, AM33, {MEMINC2 (RN4,SIMM4_2), RM6}},
{ "mov_lle", 0xf7e00003, 0xffff000f, 0x22, FMT_D10, AM33, {MEMINC2 (RN4,SIMM4_2), RM6}},
{ "mov_lcs", 0xf7e00004, 0xffff000f, 0x22, FMT_D10, AM33, {MEMINC2 (RN4,SIMM4_2), RM6}},
{ "mov_lhi", 0xf7e00005, 0xffff000f, 0x22, FMT_D10, AM33, {MEMINC2 (RN4,SIMM4_2), RM6}},
{ "mov_lcc", 0xf7e00006, 0xffff000f, 0x22, FMT_D10, AM33, {MEMINC2 (RN4,SIMM4_2), RM6}},
{ "mov_lls", 0xf7e00007, 0xffff000f, 0x22, FMT_D10, AM33, {MEMINC2 (RN4,SIMM4_2), RM6}},
{ "mov_leq", 0xf7e00008, 0xffff000f, 0x22, FMT_D10, AM33, {MEMINC2 (RN4,SIMM4_2), RM6}},
{ "mov_lne", 0xf7e00009, 0xffff000f, 0x22, FMT_D10, AM33, {MEMINC2 (RN4,SIMM4_2), RM6}},
{ "mov_lra", 0xf7e0000a, 0xffff000f, 0x22, FMT_D10, AM33, {MEMINC2 (RN4,SIMM4_2), RM6}},
{ "", 0, 0, 0, 0, 0, {0}},
};
/*
* fix up misalignment problems where possible
*/
asmlinkage void misalignment(struct pt_regs *regs, enum exception_code code)
{
const struct exception_table_entry *fixup;
const struct mn10300_opcode *pop;
unsigned long *registers = (unsigned long *) regs;
unsigned long data, *store, *postinc, disp, inc, sp;
mm_segment_t seg;
siginfo_t info;
uint32_t opcode, noc, xo, xm;
uint8_t *pc, byte, datasz;
void *address;
unsigned tmp, npop, dispsz, loop;
/* we don't fix up userspace misalignment faults */
if (user_mode(regs))
goto bus_error;
sp = (unsigned long) regs + sizeof(*regs);
kdebug("==>misalignment({pc=%lx,sp=%lx})", regs->pc, sp);
if (regs->epsw & EPSW_IE)
asm volatile("or %0,epsw" : : "i"(EPSW_IE));
seg = get_fs();
set_fs(KERNEL_DS);
fixup = search_exception_tables(regs->pc);
/* first thing to do is to match the opcode */
pc = (u_int8_t *) regs->pc;
if (__get_user(byte, pc) != 0)
goto fetch_error;
opcode = byte;
noc = 8;
for (pop = mn10300_opcodes; pop->name[0]; pop++) {
npop = ilog2(pop->opcode | pop->opmask);
if (npop <= 0 || npop > 31)
continue;
npop = (npop + 8) & ~7;
got_more_bits:
if (npop == noc) {
if ((opcode & pop->opmask) == pop->opcode)
goto found_opcode;
} else if (npop > noc) {
xo = pop->opcode >> (npop - noc);
xm = pop->opmask >> (npop - noc);
if ((opcode & xm) != xo)
continue;
/* we've got a partial match (an exact match on the
* first N bytes), so we need to get some more data */
pc++;
if (__get_user(byte, pc) != 0)
goto fetch_error;
opcode = opcode << 8 | byte;
noc += 8;
goto got_more_bits;
} else {
/* there's already been a partial match as long as the
* complete match we're now considering, so this one
* should't match */
continue;
}
}
/* didn't manage to find a fixup */
printk(KERN_CRIT "MISALIGN: %lx: unsupported instruction %x\n",
regs->pc, opcode);
failed:
set_fs(seg);
if (die_if_no_fixup("misalignment error", regs, code))
return;
bus_error:
info.si_signo = SIGBUS;
info.si_errno = 0;
info.si_code = BUS_ADRALN;
info.si_addr = (void *) regs->pc;
force_sig_info(SIGBUS, &info, current);
return;
/* error reading opcodes */
fetch_error:
printk(KERN_CRIT
"MISALIGN: %p: fault whilst reading instruction data\n",
pc);
goto failed;
bad_addr_mode:
printk(KERN_CRIT
"MISALIGN: %lx: unsupported addressing mode %x\n",
regs->pc, opcode);
goto failed;
bad_reg_mode:
printk(KERN_CRIT
"MISALIGN: %lx: unsupported register mode %x\n",
regs->pc, opcode);
goto failed;
unsupported_instruction:
printk(KERN_CRIT
"MISALIGN: %lx: unsupported instruction %x (%s)\n",
regs->pc, opcode, pop->name);
goto failed;
transfer_failed:
set_fs(seg);
if (fixup) {
regs->pc = fixup->fixup;
return;
}
if (die_if_no_fixup("misalignment fixup", regs, code))
return;
info.si_signo = SIGSEGV;
info.si_errno = 0;
info.si_code = 0;
info.si_addr = (void *) regs->pc;
force_sig_info(SIGSEGV, &info, current);
return;
/* we matched the opcode */
found_opcode:
kdebug("%lx: %x==%x { %x, %x }",
regs->pc, opcode, pop->opcode, pop->params[0], pop->params[1]);
tmp = format_tbl[pop->format].opsz;
BUG_ON(tmp > noc); /* match was less complete than it ought to have been */
if (tmp < noc) {
tmp = noc - tmp;
opcode >>= tmp;
pc -= tmp >> 3;
}
/* grab the extra displacement (note it's LSB first) */
disp = 0;
dispsz = format_tbl[pop->format].dispsz;
for (loop = 0; loop < dispsz; loop += 8) {
pc++;
if (__get_user(byte, pc) != 0)
goto fetch_error;
disp |= byte << loop;
kdebug("{%p} disp[%02x]=%02x", pc, loop, byte);
}
kdebug("disp=%lx", disp);
set_fs(KERNEL_XDS);
if (fixup)
set_fs(seg);
tmp = (pop->params[0] ^ pop->params[1]) & 0x80000000;
if (!tmp) {
printk(KERN_CRIT
"MISALIGN: %lx: insn not move to/from memory %x\n",
regs->pc, opcode);
goto failed;
}
/* determine the data transfer size of the move */
if (pop->name[3] == 0 || /* "mov" */
pop->name[4] == 'l') /* mov_lcc */
inc = datasz = 4;
else if (pop->name[3] == 'h') /* movhu */
inc = datasz = 2;
else
goto unsupported_instruction;
if (pop->params[0] & 0x80000000) {
/* move memory to register */
if (!misalignment_addr(registers, sp,
pop->params[0], opcode, disp,
&address, &postinc, &inc))
goto bad_addr_mode;
if (!misalignment_reg(registers, pop->params[1], opcode, disp,
&store))
goto bad_reg_mode;
kdebug("mov%u (%p),DARn", datasz, address);
if (copy_from_user(&data, (void *) address, datasz) != 0)
goto transfer_failed;
if (pop->params[0] & 0x1000000) {
kdebug("inc=%lx", inc);
*postinc += inc;
}
*store = data;
kdebug("loaded %lx", data);
} else {
/* move register to memory */
if (!misalignment_reg(registers, pop->params[0], opcode, disp,
&store))
goto bad_reg_mode;
if (!misalignment_addr(registers, sp,
pop->params[1], opcode, disp,
&address, &postinc, &inc))
goto bad_addr_mode;
data = *store;
kdebug("mov%u %lx,(%p)", datasz, data, address);
if (copy_to_user((void *) address, &data, datasz) != 0)
goto transfer_failed;
if (pop->params[1] & 0x1000000)
*postinc += inc;
}
tmp = format_tbl[pop->format].opsz + format_tbl[pop->format].dispsz;
regs->pc += tmp >> 3;
/* handle MOV_Lcc, which are currently the only FMT_D10 insns that
* access memory */
if (pop->format == FMT_D10)
misalignment_MOV_Lcc(regs, opcode);
set_fs(seg);
}
/*
* determine the address that was being accessed
*/
static int misalignment_addr(unsigned long *registers, unsigned long sp,
unsigned params, unsigned opcode,
unsigned long disp,
void **_address, unsigned long **_postinc,
unsigned long *_inc)
{
unsigned long *postinc = NULL, address = 0, tmp;
if (!(params & 0x1000000)) {
kdebug("noinc");
*_inc = 0;
_inc = NULL;
}
params &= 0x00ffffff;
do {
switch (params & 0xff) {
case DM0:
postinc = &registers[Dreg_index[opcode & 0x03]];
address += *postinc;
break;
case DM1:
postinc = &registers[Dreg_index[opcode >> 2 & 0x03]];
address += *postinc;
break;
case DM2:
postinc = &registers[Dreg_index[opcode >> 4 & 0x03]];
address += *postinc;
break;
case AM0:
postinc = &registers[Areg_index[opcode & 0x03]];
address += *postinc;
break;
case AM1:
postinc = &registers[Areg_index[opcode >> 2 & 0x03]];
address += *postinc;
break;
case AM2:
postinc = &registers[Areg_index[opcode >> 4 & 0x03]];
address += *postinc;
break;
case RM0:
postinc = &registers[Rreg_index[opcode & 0x0f]];
address += *postinc;
break;
case RM1:
postinc = &registers[Rreg_index[opcode >> 2 & 0x0f]];
address += *postinc;
break;
case RM2:
postinc = &registers[Rreg_index[opcode >> 4 & 0x0f]];
address += *postinc;
break;
case RM4:
postinc = &registers[Rreg_index[opcode >> 8 & 0x0f]];
address += *postinc;
break;
case RM6:
postinc = &registers[Rreg_index[opcode >> 12 & 0x0f]];
address += *postinc;
break;
case RD0:
postinc = &registers[Rreg_index[disp & 0x0f]];
address += *postinc;
break;
case RD2:
postinc = &registers[Rreg_index[disp >> 4 & 0x0f]];
address += *postinc;
break;
case SP:
address += sp;
break;
/* displacements are either to be added to the address
* before use, or, in the case of post-inc addressing,
* to be added into the base register after use */
case SD8:
case SIMM8:
disp = (long) (int8_t) (disp & 0xff);
goto displace_or_inc;
case SD16:
disp = (long) (int16_t) (disp & 0xffff);
goto displace_or_inc;
case SD24:
tmp = disp << 8;
asm("asr 8,%0" : "=r"(tmp) : "0"(tmp) : "cc");
disp = (long) tmp;
goto displace_or_inc;
case SIMM4_2:
tmp = opcode >> 4 & 0x0f;
tmp <<= 28;
asm("asr 28,%0" : "=r"(tmp) : "0"(tmp) : "cc");
disp = (long) tmp;
goto displace_or_inc;
case IMM8:
disp &= 0x000000ff;
goto displace_or_inc;
case IMM16:
disp &= 0x0000ffff;
goto displace_or_inc;
case IMM24:
disp &= 0x00ffffff;
goto displace_or_inc;
case IMM32:
case IMM32_MEM:
case IMM32_HIGH8:
case IMM32_HIGH8_MEM:
displace_or_inc:
kdebug("%s %lx", _inc ? "incr" : "disp", disp);
if (!_inc)
address += disp;
else
*_inc = disp;
break;
default:
BUG();
return 0;
}
} while ((params >>= 8));
*_address = (void *) address;
*_postinc = postinc;
return 1;
}
/*
* determine the register that is acting as source/dest
*/
static int misalignment_reg(unsigned long *registers, unsigned params,
unsigned opcode, unsigned long disp,
unsigned long **_register)
{
params &= 0x7fffffff;
if (params & 0xffffff00)
return 0;
switch (params & 0xff) {
case DM0:
*_register = &registers[Dreg_index[opcode & 0x03]];
break;
case DM1:
*_register = &registers[Dreg_index[opcode >> 2 & 0x03]];
break;
case DM2:
*_register = &registers[Dreg_index[opcode >> 4 & 0x03]];
break;
case AM0:
*_register = &registers[Areg_index[opcode & 0x03]];
break;
case AM1:
*_register = &registers[Areg_index[opcode >> 2 & 0x03]];
break;
case AM2:
*_register = &registers[Areg_index[opcode >> 4 & 0x03]];
break;
case RM0:
*_register = &registers[Rreg_index[opcode & 0x0f]];
break;
case RM1:
*_register = &registers[Rreg_index[opcode >> 2 & 0x0f]];
break;
case RM2:
*_register = &registers[Rreg_index[opcode >> 4 & 0x0f]];
break;
case RM4:
*_register = &registers[Rreg_index[opcode >> 8 & 0x0f]];
break;
case RM6:
*_register = &registers[Rreg_index[opcode >> 12 & 0x0f]];
break;
case RD0:
*_register = &registers[Rreg_index[disp & 0x0f]];
break;
case RD2:
*_register = &registers[Rreg_index[disp >> 4 & 0x0f]];
break;
case SP:
*_register = &registers[REG_SP >> 2];
break;
default:
BUG();
return 0;
}
return 1;
}
/*
* handle the conditional loop part of the move-and-loop instructions
*/
static void misalignment_MOV_Lcc(struct pt_regs *regs, uint32_t opcode)
{
unsigned long epsw = regs->epsw;
unsigned long NxorV;
kdebug("MOV_Lcc %x [flags=%lx]", opcode, epsw & 0xf);
/* calculate N^V and shift onto the same bit position as Z */
NxorV = ((epsw >> 3) ^ epsw >> 1) & 1;
switch (opcode & 0xf) {
case 0x0: /* MOV_LLT: N^V */
if (NxorV)
goto take_the_loop;
return;
case 0x1: /* MOV_LGT: ~(Z or (N^V))*/
if (!((epsw & EPSW_FLAG_Z) | NxorV))
goto take_the_loop;
return;
case 0x2: /* MOV_LGE: ~(N^V) */
if (!NxorV)
goto take_the_loop;
return;
case 0x3: /* MOV_LLE: Z or (N^V) */
if ((epsw & EPSW_FLAG_Z) | NxorV)
goto take_the_loop;
return;
case 0x4: /* MOV_LCS: C */
if (epsw & EPSW_FLAG_C)
goto take_the_loop;
return;
case 0x5: /* MOV_LHI: ~(C or Z) */
if (!(epsw & (EPSW_FLAG_C | EPSW_FLAG_Z)))
goto take_the_loop;
return;
case 0x6: /* MOV_LCC: ~C */
if (!(epsw & EPSW_FLAG_C))
goto take_the_loop;
return;
case 0x7: /* MOV_LLS: C or Z */
if (epsw & (EPSW_FLAG_C | EPSW_FLAG_Z))
goto take_the_loop;
return;
case 0x8: /* MOV_LEQ: Z */
if (epsw & EPSW_FLAG_Z)
goto take_the_loop;
return;
case 0x9: /* MOV_LNE: ~Z */
if (!(epsw & EPSW_FLAG_Z))
goto take_the_loop;
return;
case 0xa: /* MOV_LRA: always */
goto take_the_loop;
default:
BUG();
}
take_the_loop:
/* wind the PC back to just after the SETLB insn */
kdebug("loop LAR=%lx", regs->lar);
regs->pc = regs->lar - 4;
}
/*
* misalignment handler tests
*/
#ifdef CONFIG_TEST_MISALIGNMENT_HANDLER
static u8 __initdata testbuf[512] __attribute__((aligned(16))) = {
[257] = 0x11,
[258] = 0x22,
[259] = 0x33,
[260] = 0x44,
};
#define ASSERTCMP(X, OP, Y) \
do { \
if (unlikely(!((X) OP (Y)))) { \
printk(KERN_ERR "\n"); \
printk(KERN_ERR "MISALIGN: Assertion failed at line %u\n", \
__LINE__); \
printk(KERN_ERR "0x%lx " #OP " 0x%lx is false\n", \
(unsigned long)(X), (unsigned long)(Y)); \
BUG(); \
} \
} while(0)
static int __init test_misalignment(void)
{
register void *r asm("e0");
register u32 y asm("e1");
void *p = testbuf, *q;
u32 tmp, tmp2, x;
printk(KERN_NOTICE "==>test_misalignment() [testbuf=%p]\n", p);
p++;
printk(KERN_NOTICE "___ MOV (Am),Dn ___\n");
q = p + 256;
asm volatile("mov (%0),%1" : "+a"(q), "=d"(x));
ASSERTCMP(q, ==, p + 256);
ASSERTCMP(x, ==, 0x44332211);
printk(KERN_NOTICE "___ MOV (256,Am),Dn ___\n");
q = p;
asm volatile("mov (256,%0),%1" : "+a"(q), "=d"(x));
ASSERTCMP(q, ==, p);
ASSERTCMP(x, ==, 0x44332211);
printk(KERN_NOTICE "___ MOV (Di,Am),Dn ___\n");
tmp = 256;
q = p;
asm volatile("mov (%2,%0),%1" : "+a"(q), "=d"(x), "+d"(tmp));
ASSERTCMP(q, ==, p);
ASSERTCMP(x, ==, 0x44332211);
ASSERTCMP(tmp, ==, 256);
printk(KERN_NOTICE "___ MOV (256,Rm),Rn ___\n");
r = p;
asm volatile("mov (256,%0),%1" : "+r"(r), "=r"(y));
ASSERTCMP(r, ==, p);
ASSERTCMP(y, ==, 0x44332211);
printk(KERN_NOTICE "___ MOV (Rm+),Rn ___\n");
r = p + 256;
asm volatile("mov (%0+),%1" : "+r"(r), "=r"(y));
ASSERTCMP(r, ==, p + 256 + 4);
ASSERTCMP(y, ==, 0x44332211);
printk(KERN_NOTICE "___ MOV (Rm+,8),Rn ___\n");
r = p + 256;
asm volatile("mov (%0+,8),%1" : "+r"(r), "=r"(y));
ASSERTCMP(r, ==, p + 256 + 8);
ASSERTCMP(y, ==, 0x44332211);
printk(KERN_NOTICE "___ MOV (7,SP),Rn ___\n");
asm volatile(
"add -16,sp \n"
"mov +0x11,%0 \n"
"movbu %0,(7,sp) \n"
"mov +0x22,%0 \n"
"movbu %0,(8,sp) \n"
"mov +0x33,%0 \n"
"movbu %0,(9,sp) \n"
"mov +0x44,%0 \n"
"movbu %0,(10,sp) \n"
"mov (7,sp),%1 \n"
"add +16,sp \n"
: "+a"(q), "=d"(x));
ASSERTCMP(x, ==, 0x44332211);
printk(KERN_NOTICE "___ MOV (259,SP),Rn ___\n");
asm volatile(
"add -264,sp \n"
"mov +0x11,%0 \n"
"movbu %0,(259,sp) \n"
"mov +0x22,%0 \n"
"movbu %0,(260,sp) \n"
"mov +0x33,%0 \n"
"movbu %0,(261,sp) \n"
"mov +0x55,%0 \n"
"movbu %0,(262,sp) \n"
"mov (259,sp),%1 \n"
"add +264,sp \n"
: "+d"(tmp), "=d"(x));
ASSERTCMP(x, ==, 0x55332211);
printk(KERN_NOTICE "___ MOV (260,SP),Rn ___\n");
asm volatile(
"add -264,sp \n"
"mov +0x11,%0 \n"
"movbu %0,(260,sp) \n"
"mov +0x22,%0 \n"
"movbu %0,(261,sp) \n"
"mov +0x33,%0 \n"
"movbu %0,(262,sp) \n"
"mov +0x55,%0 \n"
"movbu %0,(263,sp) \n"
"mov (260,sp),%1 \n"
"add +264,sp \n"
: "+d"(tmp), "=d"(x));
ASSERTCMP(x, ==, 0x55332211);
printk(KERN_NOTICE "___ MOV_LNE ___\n");
tmp = 1;
tmp2 = 2;
q = p + 256;
asm volatile(
"setlb \n"
"mov %2,%3 \n"
"mov %1,%2 \n"
"cmp +0,%1 \n"
"mov_lne (%0+,4),%1"
: "+r"(q), "+d"(tmp), "+d"(tmp2), "=d"(x)
:
: "cc");
ASSERTCMP(q, ==, p + 256 + 12);
ASSERTCMP(x, ==, 0x44332211);
printk(KERN_NOTICE "___ MOV in SETLB ___\n");
tmp = 1;
tmp2 = 2;
q = p + 256;
asm volatile(
"setlb \n"
"mov %1,%3 \n"
"mov (%0+),%1 \n"
"cmp +0,%1 \n"
"lne "
: "+a"(q), "+d"(tmp), "+d"(tmp2), "=d"(x)
:
: "cc");
ASSERTCMP(q, ==, p + 256 + 8);
ASSERTCMP(x, ==, 0x44332211);
printk(KERN_NOTICE "<==test_misalignment()\n");
return 0;
}
arch_initcall(test_misalignment);
#endif /* CONFIG_TEST_MISALIGNMENT_HANDLER */