riscv_machdep.c revision 1.45
1 /* $NetBSD: riscv_machdep.c,v 1.45 2025/09/06 22:53:48 thorpej Exp $ */ 2 3 /*- 4 * Copyright (c) 2014, 2019, 2022 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Matt Thomas of 3am Software Foundry, and by Nick Hudson. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 #include "opt_ddb.h" 33 #include "opt_modular.h" 34 #include "opt_multiprocessor.h" 35 #include "opt_riscv_debug.h" 36 37 #include <sys/cdefs.h> 38 __RCSID("$NetBSD: riscv_machdep.c,v 1.45 2025/09/06 22:53:48 thorpej Exp $"); 39 40 #include <sys/param.h> 41 42 #include <sys/asan.h> 43 #include <sys/boot_flag.h> 44 #include <sys/cpu.h> 45 #include <sys/exec.h> 46 #include <sys/kmem.h> 47 #include <sys/ktrace.h> 48 #include <sys/lwp.h> 49 #include <sys/module.h> 50 #include <sys/mount.h> 51 #include <sys/msgbuf.h> 52 #include <sys/optstr.h> 53 #include <sys/proc.h> 54 #include <sys/reboot.h> 55 #include <sys/syscall.h> 56 #include <sys/sysctl.h> 57 #include <sys/systm.h> 58 59 #include <dev/cons.h> 60 #ifdef __HAVE_MM_MD_KERNACC 61 #include <dev/mm.h> 62 #endif 63 64 #include <uvm/uvm_extern.h> 65 66 #include <riscv/frame.h> 67 #include <riscv/locore.h> 68 #include <riscv/machdep.h> 69 #include <riscv/pte.h> 70 #include <riscv/sbi.h> 71 #include <riscv/userret.h> 72 73 #include <libfdt.h> 74 #include <dev/fdt/fdtvar.h> 75 #include <dev/fdt/fdt_boot.h> 76 #include <dev/fdt/fdt_console.h> 77 #include <dev/fdt/fdt_memory.h> 78 #include <dev/fdt/fdt_private.h> 79 #include <dev/fdt/fdt_platform.h> 80 81 int cpu_printfataltraps = 1; 82 char machine[] = MACHINE; 83 char machine_arch[] = MACHINE_ARCH; 84 85 #ifdef VERBOSE_INIT_RISCV 86 #define VPRINTF(...) printf(__VA_ARGS__) 87 #else 88 #define VPRINTF(...) __nothing 89 #endif 90 91 /* 64 should be enough, even for a ZFS UUID */ 92 #define MAX_BOOT_DEV_STR 64 93 94 char bootdevstr[MAX_BOOT_DEV_STR] = ""; 95 char *boot_args = NULL; 96 97 paddr_t physical_start; 98 paddr_t physical_end; 99 100 static void 101 earlyconsputc(dev_t dev, int c) 102 { 103 uartputc(c); 104 } 105 106 static int 107 earlyconsgetc(dev_t dev) 108 { 109 return uartgetc(); 110 } 111 112 static struct consdev earlycons = { 113 .cn_putc = earlyconsputc, 114 .cn_getc = earlyconsgetc, 115 .cn_pollc = nullcnpollc, 116 }; 117 118 struct vm_map *phys_map; 119 120 struct trapframe cpu_ddb_regs; 121 const pcu_ops_t * const pcu_ops_md_defs[PCU_UNIT_COUNT] = { 122 #ifdef FPE 123 [PCU_FPU] = &pcu_fpu_ops, 124 #endif 125 }; 126 127 /* 128 * Used by PHYSTOV and VTOPHYS -- Will be set be BSS is zeroed so 129 * keep it in data 130 */ 131 unsigned long kern_vtopdiff __attribute__((__section__(".data"))); 132 133 134 /* 135 * machine dependent system variables. 136 */ 137 SYSCTL_SETUP(sysctl_machdep_setup, "sysctl machdep subtree setup") 138 { 139 sysctl_createv(clog, 0, NULL, NULL, 140 CTLFLAG_PERMANENT, 141 CTLTYPE_NODE, "machdep", NULL, 142 NULL, 0, NULL, 0, 143 CTL_MACHDEP, CTL_EOL); 144 } 145 146 #ifdef MODULAR 147 /* 148 * Push any modules loaded by the boot loader. 149 */ 150 void 151 module_init_md(void) 152 { 153 } 154 #endif /* MODULAR */ 155 156 /* 157 * Set registers on exec. 158 * Clear all registers except sp, pc. 159 * sp is set to the stack pointer passed in. pc is set to the entry 160 * point given by the exec_package passed in. 161 */ 162 void 163 setregs(struct lwp *l, struct exec_package *pack, vaddr_t stack) 164 { 165 struct trapframe * const tf = l->l_md.md_utf; 166 struct proc * const p = l->l_proc; 167 168 memset(tf, 0, sizeof(*tf)); 169 tf->tf_sp = (intptr_t)stack_align(stack); 170 tf->tf_pc = (intptr_t)pack->ep_entry & ~1; 171 #ifdef _LP64 172 tf->tf_sr = (p->p_flag & PK_32) ? SR_USER32 : SR_USER64; 173 #else 174 tf->tf_sr = SR_USER; 175 #endif 176 177 // Set up arguments for ___start(cleanup, ps_strings) 178 tf->tf_a0 = 0; // cleanup 179 tf->tf_a1 = p->p_psstrp; // ps_strings 180 181 /* 182 * Must have interrupts disabled for exception return. 183 * Must be switching to user mode. 184 * Must enable interrupts after sret. 185 */ 186 KASSERT(__SHIFTOUT(tf->tf_sr, SR_SIE) == 0); 187 KASSERT(__SHIFTOUT(tf->tf_sr, SR_SPP) == 0); 188 KASSERT(__SHIFTOUT(tf->tf_sr, SR_SPIE) != 0); 189 } 190 191 void 192 md_child_return(struct lwp *l) 193 { 194 struct trapframe * const tf = lwp_trapframe(l); 195 196 tf->tf_a0 = 0; 197 tf->tf_a1 = 1; 198 #ifdef FPE 199 /* Disable FP as we can't be using it (yet). */ 200 tf->tf_sr &= ~SR_FS; 201 #endif 202 203 /* 204 * Must have interrupts disabled for exception return. 205 * Must be switching to user mode. 206 * Must enable interrupts after sret. 207 */ 208 209 KASSERT(__SHIFTOUT(tf->tf_sr, SR_SIE) == 0); 210 KASSERT(__SHIFTOUT(tf->tf_sr, SR_SPP) == 0); 211 KASSERT(__SHIFTOUT(tf->tf_sr, SR_SPIE) != 0); 212 213 userret(l); 214 } 215 216 /* 217 * Process the tail end of a posix_spawn() for the child. 218 */ 219 void 220 cpu_spawn_return(struct lwp *l) 221 { 222 userret(l); 223 } 224 225 /* 226 * Start a new LWP 227 */ 228 void 229 startlwp(void *arg) 230 { 231 ucontext_t * const uc = arg; 232 lwp_t * const l = curlwp; 233 int error __diagused; 234 235 error = cpu_setmcontext(l, &uc->uc_mcontext, uc->uc_flags); 236 KASSERT(error == 0); 237 238 kmem_free(uc, sizeof(*uc)); 239 userret(l); 240 } 241 242 // We've worked hard to make sure struct reg and __gregset_t are the same. 243 // Ditto for struct fpreg and fregset_t. 244 245 #ifdef _LP64 246 CTASSERT(sizeof(struct reg) == sizeof(__gregset_t)); 247 #endif 248 CTASSERT(sizeof(struct fpreg) == sizeof(__fregset_t)); 249 250 void 251 cpu_getmcontext(struct lwp *l, mcontext_t *mcp, unsigned int *flags) 252 { 253 const struct trapframe * const tf = l->l_md.md_utf; 254 255 /* Save register context. */ 256 *(struct reg *)mcp->__gregs = tf->tf_regs; 257 258 *flags |= _UC_CPU | _UC_TLSBASE; 259 260 /* Save floating point register context, if any. */ 261 KASSERT(l == curlwp); 262 if (fpu_valid_p(l)) { 263 /* 264 * If this process is the current FP owner, dump its 265 * context to the PCB first. 266 */ 267 fpu_save(l); 268 269 struct pcb * const pcb = lwp_getpcb(l); 270 *(struct fpreg *)mcp->__fregs = pcb->pcb_fpregs; 271 *flags |= _UC_FPU; 272 } 273 } 274 275 int 276 cpu_mcontext_validate(struct lwp *l, const mcontext_t *mcp) 277 { 278 /* 279 * Verify that at least the PC and SP are user addresses. 280 */ 281 if ((intptr_t) mcp->__gregs[_REG_PC] < 0 282 || (intptr_t) mcp->__gregs[_REG_SP] < 0 283 || (mcp->__gregs[_REG_PC] & 1)) 284 return EINVAL; 285 286 return 0; 287 } 288 289 int 290 cpu_setmcontext(struct lwp *l, const mcontext_t *mcp, unsigned int flags) 291 { 292 struct trapframe * const tf = l->l_md.md_utf; 293 struct proc * const p = l->l_proc; 294 const __greg_t * const gr = mcp->__gregs; 295 int error; 296 297 /* Restore register context, if any. */ 298 if (flags & _UC_CPU) { 299 error = cpu_mcontext_validate(l, mcp); 300 if (error) 301 return error; 302 303 /* 304 * Avoid updating TLS register here. 305 */ 306 const __greg_t saved_tp = tf->tf_reg[_REG_TP]; 307 tf->tf_regs = *(const struct reg *)gr; 308 tf->tf_reg[_REG_TP] = saved_tp; 309 } 310 311 /* Restore the private thread context */ 312 if (flags & _UC_TLSBASE) { 313 lwp_setprivate(l, (void *)(intptr_t)mcp->__gregs[_X_TP]); 314 } 315 316 /* Restore floating point register context, if any. */ 317 if (flags & _UC_FPU) { 318 KASSERT(l == curlwp); 319 /* Tell PCU we are replacing the FPU contents. */ 320 fpu_replace(l); 321 322 /* 323 * The PCB FP regs struct includes the FP CSR, so use the 324 * proper size of fpreg when copying. 325 */ 326 struct pcb * const pcb = lwp_getpcb(l); 327 pcb->pcb_fpregs = *(const struct fpreg *)mcp->__fregs; 328 } 329 330 mutex_enter(p->p_lock); 331 if (flags & _UC_SETSTACK) 332 l->l_sigstk.ss_flags |= SS_ONSTACK; 333 if (flags & _UC_CLRSTACK) 334 l->l_sigstk.ss_flags &= ~SS_ONSTACK; 335 mutex_exit(p->p_lock); 336 337 return 0; 338 } 339 340 void 341 cpu_need_resched(struct cpu_info *ci, struct lwp *l, int flags) 342 { 343 KASSERT(kpreempt_disabled()); 344 345 if ((flags & RESCHED_KPREEMPT) != 0) { 346 #ifdef __HAVE_PREEMPTION 347 if ((flags & RESCHED_REMOTE) != 0) { 348 cpu_send_ipi(ci, IPI_KPREEMPT); 349 } else { 350 softint_trigger(SOFTINT_KPREEMPT); 351 } 352 #endif 353 return; 354 } 355 if ((flags & RESCHED_REMOTE) != 0) { 356 #ifdef MULTIPROCESSOR 357 cpu_send_ipi(ci, IPI_AST); 358 #endif 359 } else { 360 l->l_md.md_astpending = 1; /* force call to ast() */ 361 } 362 } 363 364 void 365 cpu_signotify(struct lwp *l) 366 { 367 KASSERT(kpreempt_disabled()); 368 #ifdef __HAVE_FAST_SOFTINTS 369 KASSERT(lwp_locked(l, NULL)); 370 #endif 371 372 if (l->l_cpu != curcpu()) { 373 #ifdef MULTIPROCESSOR 374 cpu_send_ipi(l->l_cpu, IPI_AST); 375 #endif 376 } else { 377 l->l_md.md_astpending = 1; /* force call to ast() */ 378 } 379 } 380 381 void 382 cpu_need_proftick(struct lwp *l) 383 { 384 KASSERT(kpreempt_disabled()); 385 KASSERT(l->l_cpu == curcpu()); 386 387 l->l_pflag |= LP_OWEUPC; 388 l->l_md.md_astpending = 1; /* force call to ast() */ 389 } 390 391 392 /* Sync the discs, unmount the filesystems, and adjust the todr */ 393 static void 394 bootsync(void) 395 { 396 static bool bootsyncdone = false; 397 398 if (bootsyncdone) 399 return; 400 401 bootsyncdone = true; 402 403 /* Make sure we can still manage to do things */ 404 if ((csr_sstatus_read() & SR_SIE) == 0) { 405 /* 406 * If we get here then boot has been called without RB_NOSYNC 407 * and interrupts were disabled. This means the boot() call 408 * did not come from a user process e.g. shutdown, but must 409 * have come from somewhere in the kernel. 410 */ 411 ENABLE_INTERRUPTS(); 412 printf("Warning interrupts disabled during boot()\n"); 413 } 414 415 vfs_shutdown(); 416 } 417 418 419 void 420 cpu_reboot(int howto, char *bootstr) 421 { 422 423 /* 424 * If RB_NOSYNC was not specified sync the discs. 425 * Note: Unless cold is set to 1 here, syslogd will die during the 426 * unmount. It looks like syslogd is getting woken up only to find 427 * that it cannot page part of the binary in as the filesystem has 428 * been unmounted. 429 */ 430 if ((howto & RB_NOSYNC) == 0) 431 bootsync(); 432 433 #if 0 434 /* Disable interrupts. */ 435 const int s = splhigh(); 436 437 /* Do a dump if requested. */ 438 if ((howto & (RB_DUMP | RB_HALT)) == RB_DUMP) 439 dumpsys(); 440 441 splx(s); 442 #endif 443 444 pmf_system_shutdown(boothowto); 445 446 /* Say NO to interrupts for good */ 447 splhigh(); 448 449 /* Run any shutdown hooks */ 450 doshutdownhooks(); 451 452 /* Make sure IRQ's are disabled */ 453 DISABLE_INTERRUPTS(); 454 455 if (howto & RB_HALT) { 456 printf("\n"); 457 printf("The operating system has halted.\n"); 458 printf("Please press any key to reboot.\n\n"); 459 cnpollc(true); /* for proper keyboard command handling */ 460 if (cngetc() == 0) { 461 /* no console attached, so just hlt */ 462 printf("No keyboard - cannot reboot after all.\n"); 463 goto spin; 464 } 465 cnpollc(false); 466 } 467 468 printf("rebooting...\n"); 469 470 sbi_system_reset(SBI_RESET_TYPE_COLDREBOOT, SBI_RESET_REASON_NONE); 471 spin: 472 for (;;) { 473 asm volatile("wfi" ::: "memory"); 474 } 475 /* NOTREACHED */ 476 } 477 478 void 479 cpu_dumpconf(void) 480 { 481 // TBD!! 482 } 483 484 485 int 486 cpu_lwp_setprivate(lwp_t *l, void *addr) 487 { 488 struct trapframe * const tf = lwp_trapframe(l); 489 490 tf->tf_reg[_REG_TP] = (register_t)addr; 491 492 return 0; 493 } 494 495 496 void 497 cpu_startup(void) 498 { 499 vaddr_t minaddr, maxaddr; 500 char pbuf[10]; /* "999999 MB" -- But Sv39 is max 512GB */ 501 502 /* 503 * Good {morning,afternoon,evening,night}. 504 */ 505 printf("%s%s", copyright, version); 506 format_bytes(pbuf, sizeof(pbuf), ctob(physmem)); 507 printf("total memory = %s\n", pbuf); 508 509 minaddr = 0; 510 /* 511 * Allocate a submap for physio. 512 */ 513 phys_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr, 514 VM_PHYS_SIZE, 0, FALSE, NULL); 515 516 format_bytes(pbuf, sizeof(pbuf), ptoa(uvm_availmem(false))); 517 printf("avail memory = %s\n", pbuf); 518 519 #ifdef MULTIPROCESSOR 520 kcpuset_create(&cpus_halted, true); 521 KASSERT(cpus_halted != NULL); 522 523 kcpuset_create(&cpus_hatched, true); 524 KASSERT(cpus_hatched != NULL); 525 526 kcpuset_create(&cpus_paused, true); 527 KASSERT(cpus_paused != NULL); 528 529 kcpuset_create(&cpus_resumed, true); 530 KASSERT(cpus_resumed != NULL); 531 532 kcpuset_create(&cpus_running, true); 533 KASSERT(cpus_running != NULL); 534 535 kcpuset_set(cpus_hatched, cpu_index(curcpu())); 536 kcpuset_set(cpus_running, cpu_index(curcpu())); 537 #endif 538 539 fdtbus_intr_init(); 540 541 fdt_setup_rndseed(); 542 fdt_setup_efirng(); 543 } 544 545 static void 546 riscv_add_memory(const struct fdt_memory *m, void *arg) 547 { 548 paddr_t first = atop(m->start); 549 paddr_t last = atop(m->end); 550 int freelist = VM_FREELIST_DEFAULT; 551 552 VPRINTF("adding %#16" PRIxPADDR " - %#16" PRIxPADDR" to freelist %d\n", 553 m->start, m->end, freelist); 554 555 uvm_page_physload(first, last, first, last, freelist); 556 physmem += last - first; 557 } 558 559 560 static void 561 cpu_kernel_vm_init(paddr_t memory_start, paddr_t memory_end) 562 { 563 extern char __kernel_text[]; 564 extern char _end[]; 565 566 vaddr_t kernstart = trunc_page((vaddr_t)__kernel_text); 567 vaddr_t kernend = round_page((vaddr_t)_end); 568 paddr_t kernstart_phys = KERN_VTOPHYS(kernstart); 569 paddr_t kernend_phys = KERN_VTOPHYS(kernend); 570 571 VPRINTF("%s: kernel phys start %#" PRIxPADDR " end %#" PRIxPADDR "\n", 572 __func__, kernstart_phys, kernend_phys); 573 fdt_memory_remove_range(kernstart_phys, 574 kernend_phys - kernstart_phys); 575 576 #if 0 577 /* 578 * Don't give these pages to UVM. 579 * 580 * cpu_kernel_vm_init need to create proper tables then the following 581 * will be true. 582 * 583 * Now we have APs started the pages used for stacks and L1PT can 584 * be given to uvm 585 */ 586 extern char const __start__init_memory[]; 587 extern char const __stop__init_memory[] __weak; 588 if (&__start__init_memory[0] != &__stop__init_memory[0]) { 589 const paddr_t spa = KERN_VTOPHYS((vaddr_t)__start__init_memory); 590 const paddr_t epa = KERN_VTOPHYS((vaddr_t)__stop__init_memory); 591 592 VPRINTF("%s: init phys start %#" PRIxPADDR 593 " end %#" PRIxPADDR "\n", __func__, spa, epa); 594 fdt_memory_remove_range(spa, epa - spa); 595 } 596 #endif 597 598 #ifdef _LP64 599 paddr_t pa = memory_start & ~XSEGOFSET; 600 pmap_direct_base = RISCV_DIRECTMAP_START; 601 extern pd_entry_t l2_pte[PAGE_SIZE / sizeof(pd_entry_t)]; 602 603 604 const vsize_t vshift = XSEGSHIFT; 605 const vaddr_t pdetab_mask = PMAP_PDETABSIZE - 1; 606 const vsize_t inc = 1UL << vshift; 607 608 const vaddr_t sva = RISCV_DIRECTMAP_START + pa; 609 const vaddr_t eva = RISCV_DIRECTMAP_END; 610 const size_t sidx = (sva >> vshift) & pdetab_mask; 611 const size_t eidx = (eva >> vshift) & pdetab_mask; 612 613 /* Allocate gigapages covering all physical memory in the direct map. */ 614 for (size_t i = sidx; i < eidx && pa < memory_end; i++, pa += inc) { 615 l2_pte[i] = PA_TO_PTE(pa) | PTE_KERN | PTE_HARDWIRED | PTE_RW; 616 VPRINTF("dm: %p : %#" PRIxPADDR "\n", &l2_pte[i], l2_pte[i]); 617 } 618 #endif 619 // pt_dump(printf); 620 } 621 622 static void 623 riscv_init_lwp0_uarea(void) 624 { 625 extern char lwp0uspace[]; 626 627 uvm_lwp_setuarea(&lwp0, (vaddr_t)lwp0uspace); 628 memset(&lwp0.l_md, 0, sizeof(lwp0.l_md)); 629 memset(lwp_getpcb(&lwp0), 0, sizeof(struct pcb)); 630 631 struct trapframe *tf = (struct trapframe *)(lwp0uspace + USPACE) - 1; 632 memset(tf, 0, sizeof(*tf)); 633 634 lwp0.l_md.md_utf = lwp0.l_md.md_ktf = tf; 635 } 636 637 638 static void 639 riscv_print_memory(const struct fdt_memory *m, void *arg) 640 { 641 642 VPRINTF("FDT /memory @ 0x%" PRIx64 " size 0x%" PRIx64 "\n", 643 m->start, m->end - m->start); 644 } 645 646 647 static void 648 parse_mi_bootargs(char *args) 649 { 650 int howto; 651 bool found, start, skipping; 652 653 if (args == NULL) 654 return; 655 656 start = true; 657 skipping = false; 658 for (char *cp = args; *cp; cp++) { 659 /* check for "words" starting with a "-" only */ 660 if (start) { 661 if (*cp == '-') { 662 skipping = false; 663 } else { 664 skipping = true; 665 } 666 start = false; 667 continue; 668 } 669 670 if (*cp == ' ') { 671 start = true; 672 skipping = false; 673 continue; 674 } 675 676 if (skipping) { 677 continue; 678 } 679 680 /* Check valid boot flags */ 681 howto = 0; 682 BOOT_FLAG(*cp, howto); 683 if (!howto) 684 printf("bootflag '%c' not recognised\n", *cp); 685 else 686 boothowto |= howto; 687 } 688 689 found = optstr_get(args, "root", bootdevstr, sizeof(bootdevstr)); 690 if (found) { 691 bootspec = bootdevstr; 692 } 693 } 694 695 696 void 697 init_riscv(register_t hartid, paddr_t dtb) 698 { 699 700 /* set temporally to work printf()/panic() even before consinit() */ 701 cn_tab = &earlycons; 702 703 /* Load FDT */ 704 const vaddr_t dtbva = VM_KERNEL_DTB_BASE + (dtb & (NBSEG - 1)); 705 void *fdt_data = (void *)dtbva; 706 int error = fdt_check_header(fdt_data); 707 if (error != 0) 708 panic("fdt_check_header failed: %s", fdt_strerror(error)); 709 710 fdtbus_init(fdt_data); 711 712 /* Lookup platform specific backend */ 713 const struct fdt_platform * const plat = fdt_platform_find(); 714 if (plat == NULL) 715 panic("Kernel does not support this device"); 716 717 /* Early console may be available, announce ourselves. */ 718 VPRINTF("FDT<%p>\n", fdt_data); 719 720 boot_args = fdt_get_bootargs(); 721 722 VPRINTF("devmap %p\n", plat->fp_devmap()); 723 pmap_devmap_bootstrap(0, plat->fp_devmap()); 724 725 VPRINTF("bootstrap\n"); 726 plat->fp_bootstrap(); 727 728 /* 729 * If stdout-path is specified on the command line, override the 730 * value in /chosen/stdout-path before initializing console. 731 */ 732 VPRINTF("stdout\n"); 733 fdt_update_stdout_path(fdt_data, boot_args); 734 735 /* 736 * Done making changes to the FDT. 737 */ 738 fdt_pack(fdt_data); 739 740 const uint32_t dtbsize = round_page(fdt_totalsize(fdt_data)); 741 742 VPRINTF("fdt size %x/%x\n", dtbsize, fdt_totalsize(fdt_data)); 743 744 VPRINTF("consinit "); 745 consinit(); 746 VPRINTF("ok\n"); 747 748 /* Talk to the user */ 749 printf("NetBSD/riscv (fdt) booting ...\n"); 750 751 #ifdef BOOT_ARGS 752 char mi_bootargs[] = BOOT_ARGS; 753 parse_mi_bootargs(mi_bootargs); 754 #endif 755 756 uint64_t memory_start, memory_end; 757 fdt_memory_get(&memory_start, &memory_end); 758 physical_start = memory_start; 759 physical_end = memory_end; 760 761 fdt_memory_foreach(riscv_print_memory, NULL); 762 763 /* Cannot map memory above largest page number */ 764 const uint64_t maxppn = __SHIFTOUT_MASK(PTE_PPN) - 1; 765 const uint64_t memory_limit = ptoa(maxppn); 766 767 if (memory_end > memory_limit) { 768 fdt_memory_remove_range(memory_limit, memory_end); 769 memory_end = memory_limit; 770 } 771 772 uint64_t memory_size __unused = memory_end - memory_start; 773 774 VPRINTF("%s: memory start %" PRIx64 " end %" PRIx64 " (len %" 775 PRIx64 ")\n", __func__, memory_start, memory_end, memory_size); 776 777 /* Parse ramdisk, rndseed, and firmware's RNG from EFI */ 778 fdt_probe_initrd(); 779 fdt_probe_rndseed(); 780 fdt_probe_efirng(); 781 782 fdt_memory_remove_reserved(memory_start, memory_end); 783 784 fdt_memory_remove_range(dtb, dtbsize); 785 fdt_reserve_initrd(); 786 fdt_reserve_rndseed(); 787 fdt_reserve_efirng(); 788 789 /* Perform PT build and VM init */ 790 cpu_kernel_vm_init(memory_start, memory_end); 791 792 VPRINTF("bootargs: %s\n", boot_args); 793 794 parse_mi_bootargs(boot_args); 795 796 #ifdef DDB 797 if (boothowto & RB_KDB) { 798 printf("Entering DDB...\n"); 799 cpu_Debugger(); 800 } 801 #endif 802 803 extern char __kernel_text[]; 804 extern char _end[]; 805 // extern char __data_start[]; 806 // extern char __rodata_start[]; 807 808 vaddr_t kernstart = trunc_page((vaddr_t)__kernel_text); 809 vaddr_t kernend = round_page((vaddr_t)_end); 810 paddr_t kernstart_phys __unused = KERN_VTOPHYS(kernstart); 811 paddr_t kernend_phys __unused = KERN_VTOPHYS(kernend); 812 813 vaddr_t kernelvmstart; 814 815 vaddr_t kernstart_mega __unused = MEGAPAGE_TRUNC(kernstart); 816 vaddr_t kernend_mega = MEGAPAGE_ROUND(kernend); 817 818 kernelvmstart = kernend_mega; 819 820 #if 0 821 #ifdef MODULAR 822 #define MODULE_RESERVED_MAX (1024 * 1024 * 128) 823 #define MODULE_RESERVED_SIZE (1024 * 1024 * 32) /* good enough? */ 824 module_start = kernelvmstart; 825 module_end = kernend_mega + MODULE_RESERVED_SIZE; 826 if (module_end >= kernstart_mega + MODULE_RESERVED_MAX) 827 module_end = kernstart_mega + MODULE_RESERVED_MAX; 828 KASSERT(module_end > kernend_mega); 829 kernelvmstart = module_end; 830 #endif /* MODULAR */ 831 #endif 832 KASSERT(kernelvmstart < VM_KERNEL_VM_BASE); 833 834 kernelvmstart = VM_KERNEL_VM_BASE; 835 836 /* 837 * msgbuf is allocated from the top of the last biggest memory block. 838 */ 839 paddr_t msgbufaddr = 0; 840 841 #ifdef _LP64 842 /* XXX check all ranges for last one with a big enough hole */ 843 msgbufaddr = memory_end - MSGBUFSIZE; 844 KASSERT(msgbufaddr != 0); /* no space for msgbuf */ 845 fdt_memory_remove_range(msgbufaddr, msgbufaddr + MSGBUFSIZE); 846 msgbufaddr = RISCV_PA_TO_KVA(msgbufaddr); 847 VPRINTF("msgbufaddr = %#lx\n", msgbufaddr); 848 initmsgbuf((void *)msgbufaddr, MSGBUFSIZE); 849 #endif 850 851 KASSERT(msgbufaddr != 0); /* no space for msgbuf */ 852 #ifdef _LP64 853 initmsgbuf((void *)RISCV_PA_TO_KVA(msgbufaddr), MSGBUFSIZE); 854 #endif 855 856 #define DPRINTF(v) VPRINTF("%24s = 0x%16lx\n", #v, (unsigned long)v); 857 858 VPRINTF("------------------------------------------\n"); 859 DPRINTF(kern_vtopdiff); 860 DPRINTF(memory_start); 861 DPRINTF(memory_end); 862 DPRINTF(memory_size); 863 DPRINTF(kernstart_phys); 864 DPRINTF(kernend_phys) 865 DPRINTF(msgbufaddr); 866 // DPRINTF(physical_end); 867 DPRINTF(VM_MIN_KERNEL_ADDRESS); 868 DPRINTF(kernstart_mega); 869 DPRINTF(kernstart); 870 DPRINTF(kernend); 871 DPRINTF(kernend_mega); 872 #if 0 873 #ifdef MODULAR 874 DPRINTF(module_start); 875 DPRINTF(module_end); 876 #endif 877 #endif 878 DPRINTF(VM_MAX_KERNEL_ADDRESS); 879 #ifdef _LP64 880 DPRINTF(pmap_direct_base); 881 #endif 882 VPRINTF("------------------------------------------\n"); 883 884 #undef DPRINTF 885 886 uvm_md_init(); 887 888 /* 889 * pass memory pages to uvm 890 */ 891 physmem = 0; 892 fdt_memory_foreach(riscv_add_memory, NULL); 893 894 pmap_bootstrap(kernelvmstart, VM_MAX_KERNEL_ADDRESS); 895 896 kasan_init(); 897 898 /* Finish setting up lwp0 on our end before we call main() */ 899 riscv_init_lwp0_uarea(); 900 901 902 error = 0; 903 if ((boothowto & RB_MD1) == 0) { 904 VPRINTF("mpstart\n"); 905 if (plat->fp_mpstart) 906 error = plat->fp_mpstart(); 907 } 908 if (error) 909 printf("AP startup problems\n"); 910 } 911 912 913 #ifdef __HAVE_MM_MD_KERNACC 914 915 #define IN_RANGE_P(addr, start, end) (start) <= (addr) && (addr) < (end) 916 #ifdef _LP64 917 #define IN_DIRECTMAP_P(va) \ 918 IN_RANGE_P(va, RISCV_DIRECTMAP_START, RISCV_DIRECTMAP_END) 919 #else 920 #define IN_DIRECTMAP_P(va) false 921 #endif 922 923 int 924 mm_md_kernacc(void *ptr, vm_prot_t prot, bool *handled) 925 { 926 extern char __kernel_text[]; 927 extern char _end[]; 928 extern char __data_start[]; 929 930 const vaddr_t kernstart = trunc_page((vaddr_t)__kernel_text); 931 const vaddr_t kernend = round_page((vaddr_t)_end); 932 const vaddr_t data_start = (vaddr_t)__data_start; 933 934 const vaddr_t va = (vaddr_t)ptr; 935 936 *handled = false; 937 if (IN_RANGE_P(va, kernstart, kernend)) { 938 *handled = true; 939 if (va < data_start && (prot & VM_PROT_WRITE) != 0) { 940 return EFAULT; 941 } 942 } else if (IN_DIRECTMAP_P(va)) { 943 *handled = true; 944 } 945 946 return 0; 947 } 948 #endif 949 950 951 #ifdef _LP64 952 static void 953 pte_bits(void (*pr)(const char *, ...), pt_entry_t pte) 954 { 955 (*pr)("%c%c%c%c%c%c%c%c", 956 (pte & PTE_D) ? 'D' : '.', 957 (pte & PTE_A) ? 'A' : '.', 958 (pte & PTE_G) ? 'G' : '.', 959 (pte & PTE_U) ? 'U' : '.', 960 (pte & PTE_X) ? 'X' : '.', 961 (pte & PTE_W) ? 'W' : '.', 962 (pte & PTE_R) ? 'R' : '.', 963 (pte & PTE_V) ? 'V' : '.'); 964 } 965 966 static void 967 dump_ln_table(paddr_t pdp_pa, int topbit, int level, vaddr_t va, 968 void (*pr)(const char *, ...) __printflike(1, 2)) 969 { 970 pd_entry_t *pdp = (void *)PMAP_DIRECT_MAP(pdp_pa); 971 972 (*pr)("l%u @ pa %#16" PRIxREGISTER "\n", level, pdp_pa); 973 for (size_t i = 0; i < PAGE_SIZE / sizeof(pd_entry_t); i++) { 974 pd_entry_t entry = pdp[i]; 975 976 if (topbit) { 977 va = i << (PGSHIFT + level * SEGLENGTH); 978 if (va & __BIT(topbit)) { 979 va |= __BITS(63, topbit); 980 } 981 } 982 if (entry != 0) { 983 paddr_t pa = __SHIFTOUT(entry, PTE_PPN) << PGSHIFT; 984 // check level PPN bits. 985 if (PTE_ISLEAF_P(entry)) { 986 (*pr)("l%u %3zu va 0x%016lx pa 0x%012lx - ", 987 level, i, va, pa); 988 pte_bits(pr, entry); 989 (*pr)("\n"); 990 } else { 991 (*pr)("l%u %3zu va 0x%016lx -> 0x%012lx - ", 992 level, i, va, pa); 993 pte_bits(pr, entry); 994 (*pr)("\n"); 995 if (level == 0) { 996 (*pr)("wtf\n"); 997 continue; 998 } 999 if (pte_pde_valid_p(entry)) 1000 dump_ln_table(pa, 0, level - 1, va, pr); 1001 } 1002 } 1003 va += 1UL << (PGSHIFT + level * SEGLENGTH); 1004 } 1005 } 1006 1007 void 1008 pt_dump(void (*pr)(const char *, ...) __printflike(1, 2)) 1009 { 1010 const register_t satp = csr_satp_read(); 1011 size_t topbit = sizeof(long) * NBBY - 1; 1012 1013 #ifdef _LP64 1014 const paddr_t satp_pa = __SHIFTOUT(satp, SATP_PPN) << PGSHIFT; 1015 const uint8_t mode = __SHIFTOUT(satp, SATP_MODE); 1016 u_int level = 1; 1017 1018 switch (mode) { 1019 case SATP_MODE_SV39: 1020 case SATP_MODE_SV48: 1021 topbit = (39 - 1) + (mode - 8) * SEGLENGTH; 1022 level = mode - 6; 1023 break; 1024 } 1025 #endif 1026 (*pr)("topbit = %zu\n", topbit); 1027 1028 (*pr)("satp = 0x%" PRIxREGISTER "\n", satp); 1029 #ifdef _LP64 1030 dump_ln_table(satp_pa, topbit, level, 0, pr); 1031 #endif 1032 } 1033 #endif 1034 1035 void 1036 consinit(void) 1037 { 1038 static bool initialized = false; 1039 const struct fdt_console *cons = fdtbus_get_console(); 1040 const struct fdt_platform *plat = fdt_platform_find(); 1041 1042 if (initialized || cons == NULL) 1043 return; 1044 1045 u_int uart_freq = 0; 1046 extern struct bus_space riscv_generic_bs_tag; 1047 struct fdt_attach_args faa = { 1048 .faa_bst = &riscv_generic_bs_tag, 1049 }; 1050 1051 faa.faa_phandle = fdtbus_get_stdout_phandle(); 1052 if (plat->fp_uart_freq != NULL) 1053 uart_freq = plat->fp_uart_freq(); 1054 1055 cons->consinit(&faa, uart_freq); 1056 1057 initialized = true; 1058 } 1059
Indexes created Tue Sep 30 07:10:03 GMT 2025