| 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 |
|
|---|
| 29 | #include <arch/arch.h>
|
|---|
| 30 | #include <arch/regdef.h>
|
|---|
| 31 | #include <arch/boot/boot.h>
|
|---|
| 32 | #include <arch/stack.h>
|
|---|
| 33 |
|
|---|
| 34 | #include <arch/mm/mmu.h>
|
|---|
| 35 | #include <arch/mm/tlb.h>
|
|---|
| 36 | #include <arch/mm/tte.h>
|
|---|
| 37 |
|
|---|
| 38 | #ifdef CONFIG_SMP
|
|---|
| 39 | #include <arch/context_offset.h>
|
|---|
| 40 | #endif
|
|---|
| 41 |
|
|---|
| 42 | .register %g2, #scratch
|
|---|
| 43 | .register %g3, #scratch
|
|---|
| 44 |
|
|---|
| 45 | .section K_TEXT_START, "ax"
|
|---|
| 46 |
|
|---|
| 47 | #define BSP_FLAG 1
|
|---|
| 48 |
|
|---|
| 49 | /*
|
|---|
| 50 | * Here is where the kernel is passed control from the boot loader.
|
|---|
| 51 | *
|
|---|
| 52 | * The registers are expected to be in this state:
|
|---|
| 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)
|
|---|
| 58 | *
|
|---|
| 59 | * Moreover, we depend on boot having established the following environment:
|
|---|
| 60 | * - TLBs are on
|
|---|
| 61 | * - identity mapping for the kernel image
|
|---|
| 62 | */
|
|---|
| 63 |
|
|---|
| 64 | .global kernel_image_start
|
|---|
| 65 | kernel_image_start:
|
|---|
| 66 | mov BSP_FLAG, %l0
|
|---|
| 67 | and %o0, %l0, %l7 ! l7 <= bootstrap processor?
|
|---|
| 68 | andn %o0, %l0, %l6 ! l6 <= start of physical memory
|
|---|
| 69 |
|
|---|
| 70 | ! Get bits 40:13 of physmem_base.
|
|---|
| 71 | srlx %l6, 13, %l5
|
|---|
| 72 | sllx %l5, 13 + (63 - 40), %l5
|
|---|
| 73 | srlx %l5, 63 - 40, %l5 ! l5 <= physmem_base[40:13]
|
|---|
| 74 |
|
|---|
| 75 | /*
|
|---|
| 76 | * Setup basic runtime environment.
|
|---|
| 77 | */
|
|---|
| 78 |
|
|---|
| 79 | wrpr %g0, NWINDOWS - 2, %cansave ! set maximum saveable windows
|
|---|
| 80 | wrpr %g0, 0, %canrestore ! get rid of windows we will never need again
|
|---|
| 81 | wrpr %g0, 0, %otherwin ! make sure the window state is consistent
|
|---|
| 82 | wrpr %g0, NWINDOWS - 1, %cleanwin ! prevent needless clean_window traps for kernel
|
|---|
| 83 |
|
|---|
| 84 | wrpr %g0, 0, %tl ! TL = 0, primary context register is used
|
|---|
| 85 |
|
|---|
| 86 | wrpr %g0, PSTATE_PRIV_BIT, %pstate ! Disable interrupts and disable 32-bit address masking.
|
|---|
| 87 |
|
|---|
| 88 | wrpr %g0, 0, %pil ! intialize %pil
|
|---|
| 89 |
|
|---|
| 90 | /*
|
|---|
| 91 | * Switch to kernel trap table.
|
|---|
| 92 | */
|
|---|
| 93 | sethi %hi(trap_table), %g1
|
|---|
| 94 | wrpr %g1, %lo(trap_table), %tba
|
|---|
| 95 |
|
|---|
| 96 | /*
|
|---|
| 97 | * Take over the DMMU by installing global locked
|
|---|
| 98 | * TTE entry identically mapping the first 4M
|
|---|
| 99 | * of memory.
|
|---|
| 100 | *
|
|---|
| 101 | * In case of DMMU, no FLUSH instructions need to be
|
|---|
| 102 | * issued. Because of that, the old DTLB contents can
|
|---|
| 103 | * be demapped pretty straightforwardly and without
|
|---|
| 104 | * causing any traps.
|
|---|
| 105 | */
|
|---|
| 106 |
|
|---|
| 107 | wr %g0, ASI_DMMU, %asi
|
|---|
| 108 |
|
|---|
| 109 | #define SET_TLB_DEMAP_CMD(r1, context_id) \
|
|---|
| 110 | set (TLB_DEMAP_CONTEXT<<TLB_DEMAP_TYPE_SHIFT) | (context_id<<TLB_DEMAP_CONTEXT_SHIFT), %r1
|
|---|
| 111 |
|
|---|
| 112 | ! demap context 0
|
|---|
| 113 | SET_TLB_DEMAP_CMD(g1, TLB_DEMAP_NUCLEUS)
|
|---|
| 114 | stxa %g0, [%g1] ASI_DMMU_DEMAP
|
|---|
| 115 | membar #Sync
|
|---|
| 116 |
|
|---|
| 117 | #define SET_TLB_TAG(r1, context) \
|
|---|
| 118 | set VMA | (context<<TLB_TAG_ACCESS_CONTEXT_SHIFT), %r1
|
|---|
| 119 |
|
|---|
| 120 | ! write DTLB tag
|
|---|
| 121 | SET_TLB_TAG(g1, MEM_CONTEXT_KERNEL)
|
|---|
| 122 | stxa %g1, [VA_DMMU_TAG_ACCESS] %asi
|
|---|
| 123 | membar #Sync
|
|---|
| 124 |
|
|---|
| 125 | #ifdef CONFIG_VIRT_IDX_DCACHE
|
|---|
| 126 | #define TTE_LOW_DATA(imm) (TTE_CP | TTE_CV | TTE_P | LMA | (imm))
|
|---|
| 127 | #else /* CONFIG_VIRT_IDX_DCACHE */
|
|---|
| 128 | #define TTE_LOW_DATA(imm) (TTE_CP | TTE_P | LMA | (imm))
|
|---|
| 129 | #endif /* CONFIG_VIRT_IDX_DCACHE */
|
|---|
| 130 |
|
|---|
| 131 | #define SET_TLB_DATA(r1, r2, imm) \
|
|---|
| 132 | set TTE_LOW_DATA(imm), %r1; \
|
|---|
| 133 | or %r1, %l5, %r1; \
|
|---|
| 134 | mov PAGESIZE_4M, %r2; \
|
|---|
| 135 | sllx %r2, TTE_SIZE_SHIFT, %r2; \
|
|---|
| 136 | or %r1, %r2, %r1; \
|
|---|
| 137 | mov 1, %r2; \
|
|---|
| 138 | sllx %r2, TTE_V_SHIFT, %r2; \
|
|---|
| 139 | or %r1, %r2, %r1;
|
|---|
| 140 |
|
|---|
| 141 | ! write DTLB data and install the kernel mapping
|
|---|
| 142 | SET_TLB_DATA(g1, g2, TTE_L | TTE_W) ! use non-global mapping
|
|---|
| 143 | stxa %g1, [%g0] ASI_DTLB_DATA_IN_REG
|
|---|
| 144 | membar #Sync
|
|---|
| 145 |
|
|---|
| 146 | /*
|
|---|
| 147 | * Because we cannot use global mappings (because we want to
|
|---|
| 148 | * have separate 64-bit address spaces for both the kernel
|
|---|
| 149 | * and the userspace), we prepare the identity mapping also in
|
|---|
| 150 | * context 1. This step is required by the
|
|---|
| 151 | * code installing the ITLB mapping.
|
|---|
| 152 | */
|
|---|
| 153 | ! write DTLB tag of context 1 (i.e. MEM_CONTEXT_TEMP)
|
|---|
| 154 | SET_TLB_TAG(g1, MEM_CONTEXT_TEMP)
|
|---|
| 155 | stxa %g1, [VA_DMMU_TAG_ACCESS] %asi
|
|---|
| 156 | membar #Sync
|
|---|
| 157 |
|
|---|
| 158 | ! write DTLB data and install the kernel mapping in context 1
|
|---|
| 159 | SET_TLB_DATA(g1, g2, TTE_W) ! use non-global mapping
|
|---|
| 160 | stxa %g1, [%g0] ASI_DTLB_DATA_IN_REG
|
|---|
| 161 | membar #Sync
|
|---|
| 162 |
|
|---|
| 163 | /*
|
|---|
| 164 | * Now is time to take over the IMMU.
|
|---|
| 165 | * Unfortunatelly, it cannot be done as easily as the DMMU,
|
|---|
| 166 | * because the IMMU is mapping the code it executes.
|
|---|
| 167 | *
|
|---|
| 168 | * [ Note that brave experiments with disabling the IMMU
|
|---|
| 169 | * and using the DMMU approach failed after a dozen
|
|---|
| 170 | * of desparate days with only little success. ]
|
|---|
| 171 | *
|
|---|
| 172 | * The approach used here is inspired from OpenBSD.
|
|---|
| 173 | * First, the kernel creates IMMU mapping for itself
|
|---|
| 174 | * in context 1 (MEM_CONTEXT_TEMP) and switches to
|
|---|
| 175 | * it. Context 0 (MEM_CONTEXT_KERNEL) can be demapped
|
|---|
| 176 | * afterwards and replaced with the kernel permanent
|
|---|
| 177 | * mapping. Finally, the kernel switches back to
|
|---|
| 178 | * context 0 and demaps context 1.
|
|---|
| 179 | *
|
|---|
| 180 | * Moreover, the IMMU requires use of the FLUSH instructions.
|
|---|
| 181 | * But that is OK because we always use operands with
|
|---|
| 182 | * addresses already mapped by the taken over DTLB.
|
|---|
| 183 | */
|
|---|
| 184 |
|
|---|
| 185 | set kernel_image_start, %g5
|
|---|
| 186 |
|
|---|
| 187 | ! write ITLB tag of context 1
|
|---|
| 188 | SET_TLB_TAG(g1, MEM_CONTEXT_TEMP)
|
|---|
| 189 | mov VA_DMMU_TAG_ACCESS, %g2
|
|---|
| 190 | stxa %g1, [%g2] ASI_IMMU
|
|---|
| 191 | flush %g5
|
|---|
| 192 |
|
|---|
| 193 | ! write ITLB data and install the temporary mapping in context 1
|
|---|
| 194 | SET_TLB_DATA(g1, g2, 0) ! use non-global mapping
|
|---|
| 195 | stxa %g1, [%g0] ASI_ITLB_DATA_IN_REG
|
|---|
| 196 | flush %g5
|
|---|
| 197 |
|
|---|
| 198 | ! switch to context 1
|
|---|
| 199 | mov MEM_CONTEXT_TEMP, %g1
|
|---|
| 200 | stxa %g1, [VA_PRIMARY_CONTEXT_REG] %asi ! ASI_DMMU is correct here !!!
|
|---|
| 201 | flush %g5
|
|---|
| 202 |
|
|---|
| 203 | ! demap context 0
|
|---|
| 204 | SET_TLB_DEMAP_CMD(g1, TLB_DEMAP_NUCLEUS)
|
|---|
| 205 | stxa %g0, [%g1] ASI_IMMU_DEMAP
|
|---|
| 206 | flush %g5
|
|---|
| 207 |
|
|---|
| 208 | ! write ITLB tag of context 0
|
|---|
| 209 | SET_TLB_TAG(g1, MEM_CONTEXT_KERNEL)
|
|---|
| 210 | mov VA_DMMU_TAG_ACCESS, %g2
|
|---|
| 211 | stxa %g1, [%g2] ASI_IMMU
|
|---|
| 212 | flush %g5
|
|---|
| 213 |
|
|---|
| 214 | ! write ITLB data and install the permanent kernel mapping in context 0
|
|---|
| 215 | SET_TLB_DATA(g1, g2, TTE_L) ! use non-global mapping
|
|---|
| 216 | stxa %g1, [%g0] ASI_ITLB_DATA_IN_REG
|
|---|
| 217 | flush %g5
|
|---|
| 218 |
|
|---|
| 219 | ! enter nucleus - using context 0
|
|---|
| 220 | wrpr %g0, 1, %tl
|
|---|
| 221 |
|
|---|
| 222 | ! demap context 1
|
|---|
| 223 | SET_TLB_DEMAP_CMD(g1, TLB_DEMAP_PRIMARY)
|
|---|
| 224 | stxa %g0, [%g1] ASI_IMMU_DEMAP
|
|---|
| 225 | flush %g5
|
|---|
| 226 |
|
|---|
| 227 | ! set context 0 in the primary context register
|
|---|
| 228 | stxa %g0, [VA_PRIMARY_CONTEXT_REG] %asi ! ASI_DMMU is correct here !!!
|
|---|
| 229 | flush %g5
|
|---|
| 230 |
|
|---|
| 231 | ! leave nucleus - using primary context, i.e. context 0
|
|---|
| 232 | wrpr %g0, 0, %tl
|
|---|
| 233 |
|
|---|
| 234 | brz %l7, 1f ! skip if you are not the bootstrap CPU
|
|---|
| 235 | nop
|
|---|
| 236 |
|
|---|
| 237 | /*
|
|---|
| 238 | * Save physmem_base for use by the mm subsystem.
|
|---|
| 239 | * %l6 contains starting physical address
|
|---|
| 240 | */
|
|---|
| 241 | sethi %hi(physmem_base), %l4
|
|---|
| 242 | stx %l6, [%l4 + %lo(physmem_base)]
|
|---|
| 243 |
|
|---|
| 244 | /*
|
|---|
| 245 | * Precompute kernel 8K TLB data template.
|
|---|
| 246 | * %l5 contains starting physical address bits [40:13]
|
|---|
| 247 | */
|
|---|
| 248 | sethi %hi(kernel_8k_tlb_data_template), %l4
|
|---|
| 249 | ldx [%l4 + %lo(kernel_8k_tlb_data_template)], %l3
|
|---|
| 250 | or %l3, %l5, %l3
|
|---|
| 251 | stx %l3, [%l4 + %lo(kernel_8k_tlb_data_template)]
|
|---|
| 252 |
|
|---|
| 253 | /*
|
|---|
| 254 | * Flush D-Cache.
|
|---|
| 255 | */
|
|---|
| 256 | call dcache_flush
|
|---|
| 257 | nop
|
|---|
| 258 |
|
|---|
| 259 | /*
|
|---|
| 260 | * So far, we have not touched the stack.
|
|---|
| 261 | * It is a good idea to set the kernel stack to a known state now.
|
|---|
| 262 | */
|
|---|
| 263 | sethi %hi(temporary_boot_stack), %sp
|
|---|
| 264 | or %sp, %lo(temporary_boot_stack), %sp
|
|---|
| 265 | sub %sp, STACK_BIAS, %sp
|
|---|
| 266 |
|
|---|
| 267 | sethi %hi(bootinfo), %o0
|
|---|
| 268 | call memcpy ! copy bootinfo
|
|---|
| 269 | or %o0, %lo(bootinfo), %o0
|
|---|
| 270 |
|
|---|
| 271 | call arch_pre_main
|
|---|
| 272 | nop
|
|---|
| 273 |
|
|---|
| 274 | call main_bsp
|
|---|
| 275 | nop
|
|---|
| 276 |
|
|---|
| 277 | /* Not reached. */
|
|---|
| 278 |
|
|---|
| 279 | 0:
|
|---|
| 280 | ba 0b
|
|---|
| 281 | nop
|
|---|
| 282 |
|
|---|
| 283 |
|
|---|
| 284 | /*
|
|---|
| 285 | * Read MID from the processor.
|
|---|
| 286 | */
|
|---|
| 287 | 1:
|
|---|
| 288 | ldxa [%g0] ASI_UPA_CONFIG, %g1
|
|---|
| 289 | srlx %g1, UPA_CONFIG_MID_SHIFT, %g1
|
|---|
| 290 | and %g1, UPA_CONFIG_MID_MASK, %g1
|
|---|
| 291 |
|
|---|
| 292 | #ifdef CONFIG_SMP
|
|---|
| 293 | /*
|
|---|
| 294 | * Active loop for APs until the BSP picks them up.
|
|---|
| 295 | * A processor cannot leave the loop until the
|
|---|
| 296 | * global variable 'waking_up_mid' equals its
|
|---|
| 297 | * MID.
|
|---|
| 298 | */
|
|---|
| 299 | set waking_up_mid, %g2
|
|---|
| 300 | 2:
|
|---|
| 301 | ldx [%g2], %g3
|
|---|
| 302 | cmp %g3, %g1
|
|---|
| 303 | bne 2b
|
|---|
| 304 | nop
|
|---|
| 305 |
|
|---|
| 306 | /*
|
|---|
| 307 | * Configure stack for the AP.
|
|---|
| 308 | * The AP is expected to use the stack saved
|
|---|
| 309 | * in the ctx global variable.
|
|---|
| 310 | */
|
|---|
| 311 | set ctx, %g1
|
|---|
| 312 | add %g1, OFFSET_SP, %g1
|
|---|
| 313 | ldx [%g1], %o6
|
|---|
| 314 |
|
|---|
| 315 | call main_ap
|
|---|
| 316 | nop
|
|---|
| 317 |
|
|---|
| 318 | /* Not reached. */
|
|---|
| 319 | #endif
|
|---|
| 320 |
|
|---|
| 321 | 0:
|
|---|
| 322 | ba 0b
|
|---|
| 323 | nop
|
|---|
| 324 |
|
|---|
| 325 |
|
|---|
| 326 | .section K_DATA_START, "aw", @progbits
|
|---|
| 327 |
|
|---|
| 328 | /*
|
|---|
| 329 | * Create small stack to be used by the bootstrap processor.
|
|---|
| 330 | * It is going to be used only for a very limited period of
|
|---|
| 331 | * time, but we switch to it anyway, just to be sure we are
|
|---|
| 332 | * properly initialized.
|
|---|
| 333 | *
|
|---|
| 334 | * What is important is that this piece of memory is covered
|
|---|
| 335 | * by the 4M DTLB locked entry and therefore there will be
|
|---|
| 336 | * no surprises like deadly combinations of spill trap and
|
|---|
| 337 | * and TLB miss on the stack address.
|
|---|
| 338 | */
|
|---|
| 339 |
|
|---|
| 340 | #define INITIAL_STACK_SIZE 1024
|
|---|
| 341 |
|
|---|
| 342 | .align STACK_ALIGNMENT
|
|---|
| 343 | .space INITIAL_STACK_SIZE
|
|---|
| 344 | .align STACK_ALIGNMENT
|
|---|
| 345 | temporary_boot_stack:
|
|---|
| 346 | .space STACK_WINDOW_SAVE_AREA_SIZE
|
|---|
| 347 |
|
|---|
| 348 |
|
|---|
| 349 | .data
|
|---|
| 350 |
|
|---|
| 351 | .align 8
|
|---|
| 352 | .global physmem_base ! copy of the physical memory base address
|
|---|
| 353 | physmem_base:
|
|---|
| 354 | .quad 0
|
|---|
| 355 |
|
|---|
| 356 | /*
|
|---|
| 357 | * This variable is used by the fast_data_MMU_miss trap handler.
|
|---|
| 358 | * In runtime, it is further modified to reflect the starting address of
|
|---|
| 359 | * physical memory.
|
|---|
| 360 | */
|
|---|
| 361 | .global kernel_8k_tlb_data_template
|
|---|
| 362 | kernel_8k_tlb_data_template:
|
|---|
| 363 | #ifdef CONFIG_VIRT_IDX_DCACHE
|
|---|
| 364 | .quad ((1 << TTE_V_SHIFT) | (PAGESIZE_8K << TTE_SIZE_SHIFT) | TTE_CP | TTE_CV | TTE_P | TTE_W)
|
|---|
| 365 | #else /* CONFIG_VIRT_IDX_DCACHE */
|
|---|
| 366 | .quad ((1 << TTE_V_SHIFT) | (PAGESIZE_8K << TTE_SIZE_SHIFT) | TTE_CP | TTE_P | TTE_W)
|
|---|
| 367 | #endif /* CONFIG_VIRT_IDX_DCACHE */
|
|---|