Rev 1975 | Rev 1982 | Go to most recent revision | Details | Compare with Previous | Last modification | View Log | RSS feed
| Rev | Author | Line No. | Line |
|---|---|---|---|
| 418 | jermar | 1 | # |
| 2 | # Copyright (C) 2005 Jakub Jermar |
||
| 3 | # All rights reserved. |
||
| 4 | # |
||
| 5 | # Redistribution and use in source and binary forms, with or without |
||
| 6 | # modification, are permitted provided that the following conditions |
||
| 7 | # are met: |
||
| 8 | # |
||
| 9 | # - Redistributions of source code must retain the above copyright |
||
| 10 | # notice, this list of conditions and the following disclaimer. |
||
| 11 | # - Redistributions in binary form must reproduce the above copyright |
||
| 12 | # notice, this list of conditions and the following disclaimer in the |
||
| 13 | # documentation and/or other materials provided with the distribution. |
||
| 14 | # - The name of the author may not be used to endorse or promote products |
||
| 15 | # derived from this software without specific prior written permission. |
||
| 16 | # |
||
| 17 | # THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR |
||
| 18 | # IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES |
||
| 19 | # OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. |
||
| 20 | # IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, |
||
| 21 | # INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT |
||
| 22 | # NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
||
| 23 | # DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
||
| 24 | # THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
||
| 25 | # (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF |
||
| 26 | # THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
||
| 27 | # |
||
| 28 | |||
| 1903 | jermar | 29 | #include <arch/arch.h> |
| 1789 | jermar | 30 | #include <arch/regdef.h> |
| 1823 | jermar | 31 | #include <arch/boot/boot.h> |
| 1917 | jermar | 32 | #include <arch/stack.h> |
| 846 | jermar | 33 | |
| 1823 | jermar | 34 | #include <arch/mm/mmu.h> |
| 35 | #include <arch/mm/tlb.h> |
||
| 36 | #include <arch/mm/tte.h> |
||
| 37 | |||
| 1903 | jermar | 38 | #ifdef CONFIG_SMP |
| 39 | #include <arch/context_offset.h> |
||
| 40 | #endif |
||
| 41 | |||
| 426 | jermar | 42 | .register %g2, #scratch |
| 43 | .register %g3, #scratch |
||
| 44 | |||
| 418 | jermar | 45 | .section K_TEXT_START, "ax" |
| 46 | |||
| 1978 | jermar | 47 | #define BSP_FLAG 1 |
| 48 | |||
| 847 | jermar | 49 | /* |
| 1978 | jermar | 50 | * Here is where the kernel is passed control from the boot loader. |
| 1790 | jermar | 51 | * |
| 52 | * The registers are expected to be in this state: |
||
| 1978 | jermar | 53 | * - %o0 starting address of physical memory + bootstrap processor flag |
| 54 | * bits 63...1: physical memory starting address / 2 |
||
| 55 | * bit 0: non-zero on BSP processor, zero on AP processors |
||
| 56 | * - %o1 bootinfo structure address (BSP only) |
||
| 57 | * - %o2 bootinfo structure size (BSP only) |
||
| 1792 | jermar | 58 | * |
| 1978 | jermar | 59 | * Moreover, we depend on boot having established the following environment: |
| 1792 | jermar | 60 | * - TLBs are on |
| 61 | * - identity mapping for the kernel image |
||
| 847 | jermar | 62 | */ |
| 63 | |||
| 418 | jermar | 64 | .global kernel_image_start |
| 65 | kernel_image_start: |
||
| 1978 | jermar | 66 | mov BSP_FLAG, %l0 |
| 67 | and %o0, %l0, %l7 ! l7 <= bootstrap processor? |
||
| 68 | andn %o0, %l0, %l6 ! l6 <= start of physical memory |
||
| 846 | jermar | 69 | |
| 1978 | jermar | 70 | sethi %hi(physmem_base), %l5 |
| 71 | stx %l6, [%l5 + %lo(physmem_base)] |
||
| 72 | |||
| 1790 | jermar | 73 | /* |
| 1978 | jermar | 74 | * Get bits 40:13 of physmem_base. |
| 75 | */ |
||
| 76 | sethi %hi(mask_40_13), %l4 |
||
| 77 | sethi %hi(physmem_base_40_13), %l3 |
||
| 78 | ldx [%l4 + %lo(mask_40_13)], %l4 |
||
| 79 | and %l6, %l4, %l5 ! l5 <= physmem_base[40:13] |
||
| 80 | stx %l5, [%l3 + %lo(physmem_base_40_13)] |
||
| 81 | |||
| 82 | /* |
||
| 83 | * Prepare kernel 8K TLB data template. |
||
| 84 | */ |
||
| 85 | sethi %hi(kernel_8k_tlb_data_template), %l4 |
||
| 86 | ldx [%l4 + %lo(kernel_8k_tlb_data_template)], %l3 |
||
| 87 | or %l3, %l5, %l3 |
||
| 88 | stx %l3, [%l4 + %lo(kernel_8k_tlb_data_template)] |
||
| 89 | |||
| 90 | /* |
||
| 1823 | jermar | 91 | * Setup basic runtime environment. |
| 1790 | jermar | 92 | */ |
| 424 | jermar | 93 | |
| 1954 | jermar | 94 | wrpr %g0, NWINDOWS - 2, %cansave ! set maximum saveable windows |
| 1917 | jermar | 95 | wrpr %g0, 0, %canrestore ! get rid of windows we will never need again |
| 96 | wrpr %g0, 0, %otherwin ! make sure the window state is consistent |
||
| 97 | wrpr %g0, NWINDOWS - 1, %cleanwin ! prevent needless clean_window traps for kernel |
||
| 1823 | jermar | 98 | |
| 1881 | jermar | 99 | wrpr %g0, 0, %tl ! TL = 0, primary context register is used |
| 1823 | jermar | 100 | |
| 1881 | jermar | 101 | wrpr %g0, PSTATE_PRIV_BIT, %pstate ! Disable interrupts and disable 32-bit address masking. |
| 1823 | jermar | 102 | |
| 1881 | jermar | 103 | wrpr %g0, 0, %pil ! intialize %pil |
| 104 | |||
| 1790 | jermar | 105 | /* |
| 1823 | jermar | 106 | * Switch to kernel trap table. |
| 107 | */ |
||
| 1880 | jermar | 108 | sethi %hi(trap_table), %g1 |
| 109 | wrpr %g1, %lo(trap_table), %tba |
||
| 1823 | jermar | 110 | |
| 111 | /* |
||
| 112 | * Take over the DMMU by installing global locked |
||
| 113 | * TTE entry identically mapping the first 4M |
||
| 114 | * of memory. |
||
| 1792 | jermar | 115 | * |
| 1823 | jermar | 116 | * In case of DMMU, no FLUSH instructions need to be |
| 117 | * issued. Because of that, the old DTLB contents can |
||
| 118 | * be demapped pretty straightforwardly and without |
||
| 119 | * causing any traps. |
||
| 1792 | jermar | 120 | */ |
| 121 | |||
| 1823 | jermar | 122 | wr %g0, ASI_DMMU, %asi |
| 895 | jermar | 123 | |
| 1823 | jermar | 124 | #define SET_TLB_DEMAP_CMD(r1, context_id) \ |
| 125 | set (TLB_DEMAP_CONTEXT<<TLB_DEMAP_TYPE_SHIFT) | (context_id<<TLB_DEMAP_CONTEXT_SHIFT), %r1 |
||
| 126 | |||
| 127 | ! demap context 0 |
||
| 128 | SET_TLB_DEMAP_CMD(g1, TLB_DEMAP_NUCLEUS) |
||
| 129 | stxa %g0, [%g1] ASI_DMMU_DEMAP |
||
| 130 | membar #Sync |
||
| 131 | |||
| 132 | #define SET_TLB_TAG(r1, context) \ |
||
| 133 | set VMA | (context<<TLB_TAG_ACCESS_CONTEXT_SHIFT), %r1 |
||
| 134 | |||
| 135 | ! write DTLB tag |
||
| 136 | SET_TLB_TAG(g1, MEM_CONTEXT_KERNEL) |
||
| 137 | stxa %g1, [VA_DMMU_TAG_ACCESS] %asi |
||
| 138 | membar #Sync |
||
| 139 | |||
| 140 | #define SET_TLB_DATA(r1, r2, imm) \ |
||
| 1887 | jermar | 141 | set TTE_CV | TTE_CP | TTE_P | LMA | imm, %r1; \ |
| 1978 | jermar | 142 | or %r1, %l5, %r1; \ |
| 143 | mov PAGESIZE_4M, %r2; \ |
||
| 1823 | jermar | 144 | sllx %r2, TTE_SIZE_SHIFT, %r2; \ |
| 145 | or %r1, %r2, %r1; \ |
||
| 1880 | jermar | 146 | mov 1, %r2; \ |
| 1823 | jermar | 147 | sllx %r2, TTE_V_SHIFT, %r2; \ |
| 148 | or %r1, %r2, %r1; |
||
| 149 | |||
| 150 | ! write DTLB data and install the kernel mapping |
||
| 1887 | jermar | 151 | SET_TLB_DATA(g1, g2, TTE_L | TTE_W) ! use non-global mapping |
| 1823 | jermar | 152 | stxa %g1, [%g0] ASI_DTLB_DATA_IN_REG |
| 153 | membar #Sync |
||
| 1868 | jermar | 154 | |
| 155 | /* |
||
| 156 | * Because we cannot use global mappings (because we want to |
||
| 157 | * have separate 64-bit address spaces for both the kernel |
||
| 158 | * and the userspace), we prepare the identity mapping also in |
||
| 159 | * context 1. This step is required by the |
||
| 160 | * code installing the ITLB mapping. |
||
| 161 | */ |
||
| 162 | ! write DTLB tag of context 1 (i.e. MEM_CONTEXT_TEMP) |
||
| 163 | SET_TLB_TAG(g1, MEM_CONTEXT_TEMP) |
||
| 164 | stxa %g1, [VA_DMMU_TAG_ACCESS] %asi |
||
| 165 | membar #Sync |
||
| 166 | |||
| 167 | ! write DTLB data and install the kernel mapping in context 1 |
||
| 1887 | jermar | 168 | SET_TLB_DATA(g1, g2, TTE_W) ! use non-global mapping |
| 1868 | jermar | 169 | stxa %g1, [%g0] ASI_DTLB_DATA_IN_REG |
| 170 | membar #Sync |
||
| 1823 | jermar | 171 | |
| 172 | /* |
||
| 173 | * Now is time to take over the IMMU. |
||
| 174 | * Unfortunatelly, it cannot be done as easily as the DMMU, |
||
| 175 | * because the IMMU is mapping the code it executes. |
||
| 176 | * |
||
| 177 | * [ Note that brave experiments with disabling the IMMU |
||
| 178 | * and using the DMMU approach failed after a dozen |
||
| 179 | * of desparate days with only little success. ] |
||
| 180 | * |
||
| 181 | * The approach used here is inspired from OpenBSD. |
||
| 182 | * First, the kernel creates IMMU mapping for itself |
||
| 183 | * in context 1 (MEM_CONTEXT_TEMP) and switches to |
||
| 184 | * it. Context 0 (MEM_CONTEXT_KERNEL) can be demapped |
||
| 185 | * afterwards and replaced with the kernel permanent |
||
| 186 | * mapping. Finally, the kernel switches back to |
||
| 187 | * context 0 and demaps context 1. |
||
| 188 | * |
||
| 189 | * Moreover, the IMMU requires use of the FLUSH instructions. |
||
| 190 | * But that is OK because we always use operands with |
||
| 191 | * addresses already mapped by the taken over DTLB. |
||
| 192 | */ |
||
| 193 | |||
| 1852 | jermar | 194 | set kernel_image_start, %g5 |
| 1823 | jermar | 195 | |
| 196 | ! write ITLB tag of context 1 |
||
| 197 | SET_TLB_TAG(g1, MEM_CONTEXT_TEMP) |
||
| 1880 | jermar | 198 | mov VA_DMMU_TAG_ACCESS, %g2 |
| 1823 | jermar | 199 | stxa %g1, [%g2] ASI_IMMU |
| 1852 | jermar | 200 | flush %g5 |
| 1823 | jermar | 201 | |
| 202 | ! write ITLB data and install the temporary mapping in context 1 |
||
| 203 | SET_TLB_DATA(g1, g2, 0) ! use non-global mapping |
||
| 204 | stxa %g1, [%g0] ASI_ITLB_DATA_IN_REG |
||
| 1852 | jermar | 205 | flush %g5 |
| 1823 | jermar | 206 | |
| 207 | ! switch to context 1 |
||
| 1880 | jermar | 208 | mov MEM_CONTEXT_TEMP, %g1 |
| 1823 | jermar | 209 | stxa %g1, [VA_PRIMARY_CONTEXT_REG] %asi ! ASI_DMMU is correct here !!! |
| 1852 | jermar | 210 | flush %g5 |
| 1823 | jermar | 211 | |
| 212 | ! demap context 0 |
||
| 213 | SET_TLB_DEMAP_CMD(g1, TLB_DEMAP_NUCLEUS) |
||
| 214 | stxa %g0, [%g1] ASI_IMMU_DEMAP |
||
| 1852 | jermar | 215 | flush %g5 |
| 1823 | jermar | 216 | |
| 217 | ! write ITLB tag of context 0 |
||
| 218 | SET_TLB_TAG(g1, MEM_CONTEXT_KERNEL) |
||
| 1880 | jermar | 219 | mov VA_DMMU_TAG_ACCESS, %g2 |
| 1823 | jermar | 220 | stxa %g1, [%g2] ASI_IMMU |
| 1852 | jermar | 221 | flush %g5 |
| 1823 | jermar | 222 | |
| 223 | ! write ITLB data and install the permanent kernel mapping in context 0 |
||
| 1887 | jermar | 224 | SET_TLB_DATA(g1, g2, TTE_L) ! use non-global mapping |
| 1823 | jermar | 225 | stxa %g1, [%g0] ASI_ITLB_DATA_IN_REG |
| 1852 | jermar | 226 | flush %g5 |
| 1823 | jermar | 227 | |
| 1906 | jermar | 228 | ! enter nucleus - using context 0 |
| 1823 | jermar | 229 | wrpr %g0, 1, %tl |
| 230 | |||
| 231 | ! demap context 1 |
||
| 232 | SET_TLB_DEMAP_CMD(g1, TLB_DEMAP_PRIMARY) |
||
| 233 | stxa %g0, [%g1] ASI_IMMU_DEMAP |
||
| 1852 | jermar | 234 | flush %g5 |
| 1823 | jermar | 235 | |
| 236 | ! set context 0 in the primary context register |
||
| 237 | stxa %g0, [VA_PRIMARY_CONTEXT_REG] %asi ! ASI_DMMU is correct here !!! |
||
| 1852 | jermar | 238 | flush %g5 |
| 1823 | jermar | 239 | |
| 1906 | jermar | 240 | ! leave nucleus - using primary context, i.e. context 0 |
| 1823 | jermar | 241 | wrpr %g0, 0, %tl |
| 1864 | jermar | 242 | |
| 1903 | jermar | 243 | brz %l7, 1f ! skip if you are not the bootstrap CPU |
| 244 | nop |
||
| 1900 | jermar | 245 | |
| 1917 | jermar | 246 | /* |
| 247 | * So far, we have not touched the stack. |
||
| 1975 | jermar | 248 | * It is a good idea to set the kernel stack to a known state now. |
| 1917 | jermar | 249 | */ |
| 250 | sethi %hi(temporary_boot_stack), %sp |
||
| 251 | or %sp, %lo(temporary_boot_stack), %sp |
||
| 252 | sub %sp, STACK_BIAS, %sp |
||
| 253 | |||
| 1906 | jermar | 254 | sethi %hi(bootinfo), %o0 |
| 255 | call memcpy ! copy bootinfo |
||
| 256 | or %o0, %lo(bootinfo), %o0 |
||
| 257 | |||
| 1864 | jermar | 258 | call arch_pre_main |
| 259 | nop |
||
| 1823 | jermar | 260 | |
| 426 | jermar | 261 | call main_bsp |
| 262 | nop |
||
| 263 | |||
| 264 | /* Not reached. */ |
||
| 265 | |||
| 1903 | jermar | 266 | 0: |
| 267 | ba 0b |
||
| 268 | nop |
||
| 269 | |||
| 270 | |||
| 271 | /* |
||
| 272 | * Read MID from the processor. |
||
| 273 | */ |
||
| 274 | 1: |
||
| 275 | ldxa [%g0] ASI_UPA_CONFIG, %g1 |
||
| 276 | srlx %g1, UPA_CONFIG_MID_SHIFT, %g1 |
||
| 277 | and %g1, UPA_CONFIG_MID_MASK, %g1 |
||
| 278 | |||
| 1905 | jermar | 279 | #ifdef CONFIG_SMP |
| 1903 | jermar | 280 | /* |
| 281 | * Active loop for APs until the BSP picks them up. |
||
| 282 | * A processor cannot leave the loop until the |
||
| 283 | * global variable 'waking_up_mid' equals its |
||
| 284 | * MID. |
||
| 285 | */ |
||
| 286 | set waking_up_mid, %g2 |
||
| 424 | jermar | 287 | 2: |
| 1903 | jermar | 288 | ldx [%g2], %g3 |
| 289 | cmp %g3, %g1 |
||
| 290 | bne 2b |
||
| 424 | jermar | 291 | nop |
| 1903 | jermar | 292 | |
| 293 | /* |
||
| 294 | * Configure stack for the AP. |
||
| 295 | * The AP is expected to use the stack saved |
||
| 296 | * in the ctx global variable. |
||
| 297 | */ |
||
| 298 | set ctx, %g1 |
||
| 299 | add %g1, OFFSET_SP, %g1 |
||
| 300 | ldx [%g1], %o6 |
||
| 301 | |||
| 302 | call main_ap |
||
| 303 | nop |
||
| 304 | |||
| 305 | /* Not reached. */ |
||
| 1905 | jermar | 306 | #endif |
| 1903 | jermar | 307 | |
| 308 | 0: |
||
| 309 | ba 0b |
||
| 310 | nop |
||
| 1917 | jermar | 311 | |
| 312 | |||
| 313 | .section K_DATA_START, "aw", @progbits |
||
| 314 | |||
| 315 | /* |
||
| 316 | * Create small stack to be used by the bootstrap processor. |
||
| 317 | * It is going to be used only for a very limited period of |
||
| 318 | * time, but we switch to it anyway, just to be sure we are |
||
| 319 | * properly initialized. |
||
| 320 | * |
||
| 321 | * What is important is that this piece of memory is covered |
||
| 322 | * by the 4M DTLB locked entry and therefore there will be |
||
| 323 | * no surprises like deadly combinations of spill trap and |
||
| 324 | * and TLB miss on the stack address. |
||
| 325 | */ |
||
| 326 | |||
| 327 | #define INITIAL_STACK_SIZE 1024 |
||
| 328 | |||
| 329 | .align STACK_ALIGNMENT |
||
| 1978 | jermar | 330 | .space INITIAL_STACK_SIZE |
| 1917 | jermar | 331 | .align STACK_ALIGNMENT |
| 332 | temporary_boot_stack: |
||
| 1978 | jermar | 333 | .space STACK_WINDOW_SAVE_AREA_SIZE |
| 334 | |||
| 335 | |||
| 336 | .data |
||
| 337 | |||
| 338 | .align 8 |
||
| 339 | .global physmem_base ! copy of the physical memory base address |
||
| 340 | physmem_base: |
||
| 341 | .quad 0 |
||
| 342 | |||
| 343 | .global physmem_base_40_13 |
||
| 344 | physmem_base_40_13: ! physmem_base & mask_40_13 |
||
| 345 | .quad 0 |
||
| 346 | |||
| 347 | .global mask_40_13 |
||
| 348 | mask_40_13: ! constant with bits 40:13 set |
||
| 349 | .quad (((1 << 41) - 1) & ~((1 << 13) - 1)) |
||
| 350 | |||
| 351 | /* |
||
| 352 | * This variable is used by the fast_data_MMU_miss trap handler. |
||
| 353 | * It is initialized to reflect the starting address of physical |
||
| 354 | * memory. |
||
| 355 | */ |
||
| 356 | .global kernel_8k_tlb_data_template |
||
| 357 | kernel_8k_tlb_data_template: |
||
| 358 | .quad ((1 << TTE_V_SHIFT) | TTE_CV | TTE_CP | TTE_P | TTE_W) |
||
| 359 |