wdenk | ef3386f | 2004-10-10 21:27:30 +0000 | [diff] [blame^] | 1 | /* |
| 2 | * (C) Copyright 2004, Psyent Corporation <www.psyent.com> |
| 3 | * Scott McNutt <smcnutt@psyent.com> |
| 4 | * |
| 5 | * See file CREDITS for list of people who contributed to this |
| 6 | * project. |
| 7 | * |
| 8 | * This program is free software; you can redistribute it and/or |
| 9 | * modify it under the terms of the GNU General Public License as |
| 10 | * published by the Free Software Foundation; either version 2 of |
| 11 | * the License, or (at your option) any later version. |
| 12 | * |
| 13 | * This program is distributed in the hope that it will be useful, |
| 14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 | * GNU General Public License for more details. |
| 17 | * |
| 18 | * You should have received a copy of the GNU General Public License |
| 19 | * along with this program; if not, write to the Free Software |
| 20 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, |
| 21 | * MA 02111-1307 USA |
| 22 | */ |
| 23 | |
| 24 | |
| 25 | #include <config.h> |
| 26 | #include <version.h> |
| 27 | |
| 28 | /************************************************************************* |
| 29 | * RESTART |
| 30 | ************************************************************************/ |
| 31 | |
| 32 | .text |
| 33 | .global _start |
| 34 | |
| 35 | _start: |
| 36 | /* ICACHE INIT -- only the icache line at the reset address |
| 37 | * is invalidated at reset. So the init must stay within |
| 38 | * the cache line size (8 words). If GERMS is used, we'll |
| 39 | * just be invalidating the cache a second time. If cache |
| 40 | * is not implemented initi behaves as nop. |
| 41 | */ |
| 42 | movhi r4, %hi(CFG_ICACHELINE_SIZE) |
| 43 | ori r4, r4, %lo(CFG_ICACHELINE_SIZE) |
| 44 | movhi r5, %hi(CFG_ICACHE_SIZE) |
| 45 | ori r5, r5, %lo(CFG_ICACHE_SIZE) |
| 46 | mov r6, r0 |
| 47 | 0: initi r6 |
| 48 | add r6, r6, r4 |
| 49 | bltu r6, r5, 0b |
| 50 | |
| 51 | /* INTERRUPTS -- for now, all interrupts masked and globally |
| 52 | * disabled. |
| 53 | */ |
| 54 | wrctl status, r0 /* Disable interrupts */ |
| 55 | wrctl ienable, r0 /* All disabled */ |
| 56 | |
| 57 | /* DCACHE INIT -- if dcache not implemented, initd behaves as |
| 58 | * nop. |
| 59 | */ |
| 60 | movhi r4, %hi(CFG_DCACHELINE_SIZE) |
| 61 | ori r4, r4, %lo(CFG_DCACHELINE_SIZE) |
| 62 | movhi r5, %hi(CFG_DCACHE_SIZE) |
| 63 | ori r5, r5, %lo(CFG_DCACHE_SIZE) |
| 64 | mov r6, r0 |
| 65 | 1: initd 0(r6) |
| 66 | add r6, r6, r4 |
| 67 | bltu r6, r5, 1b |
| 68 | |
| 69 | /* RELOCATE CODE, DATA & COMMAND TABLE -- the following code |
| 70 | * assumes code, data and the command table are all |
| 71 | * contiguous. This lets us relocate everything as a single |
| 72 | * block. Make sure the linker script matches this ;-) |
| 73 | */ |
| 74 | nextpc r4 |
| 75 | _cur: movhi r5, %hi(_cur - _start) |
| 76 | ori r5, r5, %lo(_cur - _start) |
| 77 | sub r4, r4, r5 /* r4 <- cur _start */ |
| 78 | mov r8, r4 |
| 79 | movhi r5, %hi(_start) |
| 80 | ori r5, r5, %lo(_start) /* r5 <- linked _start */ |
| 81 | beq r4, r5, 3f |
| 82 | |
| 83 | movhi r6, %hi(_edata) |
| 84 | ori r6, r6, %lo(_edata) |
| 85 | 2: ldwio r7, 0(r4) |
| 86 | addi r4, r4, 4 |
| 87 | stwio r7, 0(r5) |
| 88 | addi r5, r5, 4 |
| 89 | bne r5, r6, 2b |
| 90 | 3: |
| 91 | |
| 92 | /* ZERO BSS/SBSS -- bss and sbss are assumed to be adjacent |
| 93 | * and between __bss_start and _end. |
| 94 | */ |
| 95 | movhi r5, %hi(__bss_start) |
| 96 | ori r5, r5, %lo(__bss_start) |
| 97 | movhi r6, %hi(_end) |
| 98 | ori r6, r6, %lo(_end) |
| 99 | beq r5, r6, 5f |
| 100 | |
| 101 | 4: stwio r0, 0(r5) |
| 102 | addi r5, r5, 4 |
| 103 | bne r5, r6, 4b |
| 104 | 5: |
| 105 | |
| 106 | /* GLOBAL POINTER -- the global pointer is used to reference |
| 107 | * "small data" (see -G switch). The linker script must |
| 108 | * provide the gp address. |
| 109 | */ |
| 110 | movhi gp, %hi(_gp) |
| 111 | ori gp, gp, %lo(_gp) |
| 112 | |
| 113 | /* JUMP TO RELOC ADDR */ |
| 114 | movhi r4, %hi(_reloc) |
| 115 | ori r4, r4, %lo(_reloc) |
| 116 | jmp r4 |
| 117 | _reloc: |
| 118 | |
| 119 | /* COPY EXCEPTION TRAMPOLINE -- copy the tramp to the |
| 120 | * exception address. |
| 121 | */ |
| 122 | #if !defined(CONFIG_ROM_STUBS) |
| 123 | movhi r4, %hi(_except_start) |
| 124 | ori r4, r4, %lo(_except_start) |
| 125 | movhi r5, %hi(_except_end) |
| 126 | ori r5, r5, %lo(_except_end) |
| 127 | movhi r6, %hi(CFG_EXCEPTION_ADDR) |
| 128 | ori r6, r6, %lo(CFG_EXCEPTION_ADDR) |
| 129 | |
| 130 | 6: ldwio r7, 0(r4) |
| 131 | stwio r7, 0(r6) |
| 132 | addi r4, r4, 4 |
| 133 | addi r6, r6, 4 |
| 134 | bne r4, r5, 6b |
| 135 | #endif |
| 136 | |
| 137 | /* STACK INIT -- zero top two words for call back chain. |
| 138 | */ |
| 139 | movhi sp, %hi(CFG_INIT_SP) |
| 140 | ori sp, sp, %lo(CFG_INIT_SP) |
| 141 | addi sp, sp, -8 |
| 142 | stw r0, 0(sp) |
| 143 | stw r0, 4(sp) |
| 144 | mov fp, sp |
| 145 | |
| 146 | /* |
| 147 | * Call board_init -- never returns |
| 148 | */ |
| 149 | movhi r4, %hi(board_init@h) |
| 150 | ori r4, r4, %lo(board_init@h) |
| 151 | callr r4 |
| 152 | |
| 153 | /* NEVER RETURNS -- but branch to the _start just |
| 154 | * in case ;-) |
| 155 | */ |
| 156 | br _start |
| 157 | |
| 158 | /* EXCEPTION TRAMPOLINE -- the following gets copied |
| 159 | * to the exception address. |
| 160 | */ |
| 161 | _except_start: |
| 162 | movhi et, %hi(_exception) |
| 163 | ori et, et, %lo(_exception) |
| 164 | jmp et |
| 165 | _except_end: |
| 166 | |
| 167 | |
| 168 | /* |
| 169 | * dly_clks -- Nios2 (like Nios1) doesn't have a timebase in |
| 170 | * the core. For simple delay loops, we do our best by counting |
| 171 | * instruction cycles. |
| 172 | * |
| 173 | * Instruction performance varies based on the core. For cores |
| 174 | * with icache and static/dynamic branch prediction (II/f, II/s): |
| 175 | * |
| 176 | * Normal ALU (e.g. add, cmp, etc): 1 cycle |
| 177 | * Branch (correctly predicted, taken): 2 cycles |
| 178 | * Negative offset is predicted (II/s). |
| 179 | * |
| 180 | * For cores without icache and no branch prediction (II/e): |
| 181 | * |
| 182 | * Normal ALU (e.g. add, cmp, etc): 6 cycles |
| 183 | * Branch (no prediction): 6 cycles |
| 184 | * |
| 185 | * For simplicity, if an instruction cache is implemented we |
| 186 | * assume II/f or II/s. Otherwise, we use the II/e. |
| 187 | * |
| 188 | */ |
| 189 | .globl dly_clks |
| 190 | |
| 191 | dly_clks: |
| 192 | |
| 193 | #if (CFG_ICACHE_SIZE > 0) |
| 194 | subi r4, r4, 3 /* 3 clocks/loop */ |
| 195 | #else |
| 196 | subi r4, r4, 12 /* 12 clocks/loop */ |
| 197 | #endif |
| 198 | bge r4, r0, dly_clks |
| 199 | ret |
| 200 | |
| 201 | |
| 202 | #if !defined(CONFIG_IDENT_STRING) |
| 203 | #define CONFIG_IDENT_STRING "" |
| 204 | #endif |
| 205 | .data |
| 206 | .globl version_string |
| 207 | |
| 208 | version_string: |
| 209 | .ascii U_BOOT_VERSION |
| 210 | .ascii " (", __DATE__, " - ", __TIME__, ")" |
| 211 | .ascii CONFIG_IDENT_STRING, "\0" |