1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * S390 version 4 * Copyright IBM Corp. 1999, 2012 5 * Author(s): Hartmut Penner (hp@de.ibm.com), 6 * Martin Schwidefsky (schwidefsky@de.ibm.com) 7 * 8 * Derived from "arch/i386/kernel/setup.c" 9 * Copyright (C) 1995, Linus Torvalds 10 */ 11 12 /* 13 * This file handles the architecture-dependent parts of initialization 14 */ 15 16 #define KMSG_COMPONENT "setup" 17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt 18 19 #include <linux/errno.h> 20 #include <linux/export.h> 21 #include <linux/sched.h> 22 #include <linux/sched/task.h> 23 #include <linux/cpu.h> 24 #include <linux/kernel.h> 25 #include <linux/memblock.h> 26 #include <linux/mm.h> 27 #include <linux/stddef.h> 28 #include <linux/unistd.h> 29 #include <linux/ptrace.h> 30 #include <linux/random.h> 31 #include <linux/user.h> 32 #include <linux/tty.h> 33 #include <linux/ioport.h> 34 #include <linux/delay.h> 35 #include <linux/init.h> 36 #include <linux/initrd.h> 37 #include <linux/root_dev.h> 38 #include <linux/console.h> 39 #include <linux/kernel_stat.h> 40 #include <linux/dma-map-ops.h> 41 #include <linux/device.h> 42 #include <linux/notifier.h> 43 #include <linux/pfn.h> 44 #include <linux/ctype.h> 45 #include <linux/reboot.h> 46 #include <linux/topology.h> 47 #include <linux/kexec.h> 48 #include <linux/crash_dump.h> 49 #include <linux/memory.h> 50 #include <linux/compat.h> 51 #include <linux/start_kernel.h> 52 #include <linux/hugetlb.h> 53 54 #include <asm/boot_data.h> 55 #include <asm/ipl.h> 56 #include <asm/facility.h> 57 #include <asm/smp.h> 58 #include <asm/mmu_context.h> 59 #include <asm/cpcmd.h> 60 #include <asm/lowcore.h> 61 #include <asm/nmi.h> 62 #include <asm/irq.h> 63 #include <asm/page.h> 64 #include <asm/ptrace.h> 65 #include <asm/sections.h> 66 #include <asm/ebcdic.h> 67 #include <asm/diag.h> 68 #include <asm/os_info.h> 69 #include <asm/sclp.h> 70 #include <asm/stacktrace.h> 71 #include <asm/sysinfo.h> 72 #include <asm/numa.h> 73 #include <asm/alternative.h> 74 #include <asm/nospec-branch.h> 75 #include <asm/mem_detect.h> 76 #include <asm/uv.h> 77 #include <asm/asm-offsets.h> 78 #include "entry.h" 79 80 /* 81 * Machine setup.. 82 */ 83 unsigned int console_mode = 0; 84 EXPORT_SYMBOL(console_mode); 85 86 unsigned int console_devno = -1; 87 EXPORT_SYMBOL(console_devno); 88 89 unsigned int console_irq = -1; 90 EXPORT_SYMBOL(console_irq); 91 92 unsigned long elf_hwcap __read_mostly = 0; 93 char elf_platform[ELF_PLATFORM_SIZE]; 94 95 unsigned long int_hwcap = 0; 96 97 int __bootdata(noexec_disabled); 98 unsigned long __bootdata(ident_map_size); 99 unsigned long __bootdata(vmalloc_size); 100 struct mem_detect_info __bootdata(mem_detect); 101 102 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table); 103 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table); 104 unsigned long __bootdata_preserved(__stext_dma); 105 unsigned long __bootdata_preserved(__etext_dma); 106 unsigned long __bootdata_preserved(__sdma); 107 unsigned long __bootdata_preserved(__edma); 108 unsigned long __bootdata_preserved(__kaslr_offset); 109 unsigned int __bootdata_preserved(zlib_dfltcc_support); 110 EXPORT_SYMBOL(zlib_dfltcc_support); 111 112 unsigned long VMALLOC_START; 113 EXPORT_SYMBOL(VMALLOC_START); 114 115 unsigned long VMALLOC_END; 116 EXPORT_SYMBOL(VMALLOC_END); 117 118 struct page *vmemmap; 119 EXPORT_SYMBOL(vmemmap); 120 unsigned long vmemmap_size; 121 122 unsigned long MODULES_VADDR; 123 unsigned long MODULES_END; 124 125 /* An array with a pointer to the lowcore of every CPU. */ 126 struct lowcore *lowcore_ptr[NR_CPUS]; 127 EXPORT_SYMBOL(lowcore_ptr); 128 129 /* 130 * The Write Back bit position in the physaddr is given by the SLPC PCI. 131 * Leaving the mask zero always uses write through which is safe 132 */ 133 unsigned long mio_wb_bit_mask __ro_after_init; 134 135 /* 136 * This is set up by the setup-routine at boot-time 137 * for S390 need to find out, what we have to setup 138 * using address 0x10400 ... 139 */ 140 141 #include <asm/setup.h> 142 143 /* 144 * condev= and conmode= setup parameter. 145 */ 146 147 static int __init condev_setup(char *str) 148 { 149 int vdev; 150 151 vdev = simple_strtoul(str, &str, 0); 152 if (vdev >= 0 && vdev < 65536) { 153 console_devno = vdev; 154 console_irq = -1; 155 } 156 return 1; 157 } 158 159 __setup("condev=", condev_setup); 160 161 static void __init set_preferred_console(void) 162 { 163 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP) 164 add_preferred_console("ttyS", 0, NULL); 165 else if (CONSOLE_IS_3270) 166 add_preferred_console("tty3270", 0, NULL); 167 else if (CONSOLE_IS_VT220) 168 add_preferred_console("ttyS", 1, NULL); 169 else if (CONSOLE_IS_HVC) 170 add_preferred_console("hvc", 0, NULL); 171 } 172 173 static int __init conmode_setup(char *str) 174 { 175 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 176 if (!strcmp(str, "hwc") || !strcmp(str, "sclp")) 177 SET_CONSOLE_SCLP; 178 #endif 179 #if defined(CONFIG_TN3215_CONSOLE) 180 if (!strcmp(str, "3215")) 181 SET_CONSOLE_3215; 182 #endif 183 #if defined(CONFIG_TN3270_CONSOLE) 184 if (!strcmp(str, "3270")) 185 SET_CONSOLE_3270; 186 #endif 187 set_preferred_console(); 188 return 1; 189 } 190 191 __setup("conmode=", conmode_setup); 192 193 static void __init conmode_default(void) 194 { 195 char query_buffer[1024]; 196 char *ptr; 197 198 if (MACHINE_IS_VM) { 199 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL); 200 console_devno = simple_strtoul(query_buffer + 5, NULL, 16); 201 ptr = strstr(query_buffer, "SUBCHANNEL ="); 202 console_irq = simple_strtoul(ptr + 13, NULL, 16); 203 cpcmd("QUERY TERM", query_buffer, 1024, NULL); 204 ptr = strstr(query_buffer, "CONMODE"); 205 /* 206 * Set the conmode to 3215 so that the device recognition 207 * will set the cu_type of the console to 3215. If the 208 * conmode is 3270 and we don't set it back then both 209 * 3215 and the 3270 driver will try to access the console 210 * device (3215 as console and 3270 as normal tty). 211 */ 212 cpcmd("TERM CONMODE 3215", NULL, 0, NULL); 213 if (ptr == NULL) { 214 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 215 SET_CONSOLE_SCLP; 216 #endif 217 return; 218 } 219 if (str_has_prefix(ptr + 8, "3270")) { 220 #if defined(CONFIG_TN3270_CONSOLE) 221 SET_CONSOLE_3270; 222 #elif defined(CONFIG_TN3215_CONSOLE) 223 SET_CONSOLE_3215; 224 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 225 SET_CONSOLE_SCLP; 226 #endif 227 } else if (str_has_prefix(ptr + 8, "3215")) { 228 #if defined(CONFIG_TN3215_CONSOLE) 229 SET_CONSOLE_3215; 230 #elif defined(CONFIG_TN3270_CONSOLE) 231 SET_CONSOLE_3270; 232 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 233 SET_CONSOLE_SCLP; 234 #endif 235 } 236 } else if (MACHINE_IS_KVM) { 237 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE)) 238 SET_CONSOLE_VT220; 239 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE)) 240 SET_CONSOLE_SCLP; 241 else 242 SET_CONSOLE_HVC; 243 } else { 244 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 245 SET_CONSOLE_SCLP; 246 #endif 247 } 248 } 249 250 #ifdef CONFIG_CRASH_DUMP 251 static void __init setup_zfcpdump(void) 252 { 253 if (!is_ipl_type_dump()) 254 return; 255 if (OLDMEM_BASE) 256 return; 257 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev"); 258 console_loglevel = 2; 259 } 260 #else 261 static inline void setup_zfcpdump(void) {} 262 #endif /* CONFIG_CRASH_DUMP */ 263 264 /* 265 * Reboot, halt and power_off stubs. They just call _machine_restart, 266 * _machine_halt or _machine_power_off. 267 */ 268 269 void machine_restart(char *command) 270 { 271 if ((!in_interrupt() && !in_atomic()) || oops_in_progress) 272 /* 273 * Only unblank the console if we are called in enabled 274 * context or a bust_spinlocks cleared the way for us. 275 */ 276 console_unblank(); 277 _machine_restart(command); 278 } 279 280 void machine_halt(void) 281 { 282 if (!in_interrupt() || oops_in_progress) 283 /* 284 * Only unblank the console if we are called in enabled 285 * context or a bust_spinlocks cleared the way for us. 286 */ 287 console_unblank(); 288 _machine_halt(); 289 } 290 291 void machine_power_off(void) 292 { 293 if (!in_interrupt() || oops_in_progress) 294 /* 295 * Only unblank the console if we are called in enabled 296 * context or a bust_spinlocks cleared the way for us. 297 */ 298 console_unblank(); 299 _machine_power_off(); 300 } 301 302 /* 303 * Dummy power off function. 304 */ 305 void (*pm_power_off)(void) = machine_power_off; 306 EXPORT_SYMBOL_GPL(pm_power_off); 307 308 void *restart_stack; 309 310 unsigned long stack_alloc(void) 311 { 312 #ifdef CONFIG_VMAP_STACK 313 return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE, 314 THREADINFO_GFP, NUMA_NO_NODE, 315 __builtin_return_address(0)); 316 #else 317 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 318 #endif 319 } 320 321 void stack_free(unsigned long stack) 322 { 323 #ifdef CONFIG_VMAP_STACK 324 vfree((void *) stack); 325 #else 326 free_pages(stack, THREAD_SIZE_ORDER); 327 #endif 328 } 329 330 int __init arch_early_irq_init(void) 331 { 332 unsigned long stack; 333 334 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 335 if (!stack) 336 panic("Couldn't allocate async stack"); 337 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET; 338 return 0; 339 } 340 341 static int __init async_stack_realloc(void) 342 { 343 unsigned long old, new; 344 345 old = S390_lowcore.async_stack - STACK_INIT_OFFSET; 346 new = stack_alloc(); 347 if (!new) 348 panic("Couldn't allocate async stack"); 349 S390_lowcore.async_stack = new + STACK_INIT_OFFSET; 350 free_pages(old, THREAD_SIZE_ORDER); 351 return 0; 352 } 353 early_initcall(async_stack_realloc); 354 355 void __init arch_call_rest_init(void) 356 { 357 unsigned long stack; 358 359 stack = stack_alloc(); 360 if (!stack) 361 panic("Couldn't allocate kernel stack"); 362 current->stack = (void *) stack; 363 #ifdef CONFIG_VMAP_STACK 364 current->stack_vm_area = (void *) stack; 365 #endif 366 set_task_stack_end_magic(current); 367 stack += STACK_INIT_OFFSET; 368 S390_lowcore.kernel_stack = stack; 369 CALL_ON_STACK_NORETURN(rest_init, stack); 370 } 371 372 static void __init setup_lowcore_dat_off(void) 373 { 374 unsigned long int_psw_mask = PSW_KERNEL_BITS; 375 struct lowcore *lc; 376 377 if (IS_ENABLED(CONFIG_KASAN)) 378 int_psw_mask |= PSW_MASK_DAT; 379 380 /* 381 * Setup lowcore for boot cpu 382 */ 383 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE); 384 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc)); 385 if (!lc) 386 panic("%s: Failed to allocate %zu bytes align=%zx\n", 387 __func__, sizeof(*lc), sizeof(*lc)); 388 389 lc->restart_psw.mask = PSW_KERNEL_BITS; 390 lc->restart_psw.addr = (unsigned long) restart_int_handler; 391 lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK; 392 lc->external_new_psw.addr = (unsigned long) ext_int_handler; 393 lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK; 394 lc->svc_new_psw.addr = (unsigned long) system_call; 395 lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK; 396 lc->program_new_psw.addr = (unsigned long) pgm_check_handler; 397 lc->mcck_new_psw.mask = PSW_KERNEL_BITS; 398 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler; 399 lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK; 400 lc->io_new_psw.addr = (unsigned long) io_int_handler; 401 lc->clock_comparator = clock_comparator_max; 402 lc->nodat_stack = ((unsigned long) &init_thread_union) 403 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); 404 lc->current_task = (unsigned long)&init_task; 405 lc->lpp = LPP_MAGIC; 406 lc->machine_flags = S390_lowcore.machine_flags; 407 lc->preempt_count = S390_lowcore.preempt_count; 408 lc->stfl_fac_list = S390_lowcore.stfl_fac_list; 409 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list, 410 sizeof(lc->stfle_fac_list)); 411 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list, 412 sizeof(lc->alt_stfle_fac_list)); 413 nmi_alloc_boot_cpu(lc); 414 lc->sync_enter_timer = S390_lowcore.sync_enter_timer; 415 lc->async_enter_timer = S390_lowcore.async_enter_timer; 416 lc->exit_timer = S390_lowcore.exit_timer; 417 lc->user_timer = S390_lowcore.user_timer; 418 lc->system_timer = S390_lowcore.system_timer; 419 lc->steal_timer = S390_lowcore.steal_timer; 420 lc->last_update_timer = S390_lowcore.last_update_timer; 421 lc->last_update_clock = S390_lowcore.last_update_clock; 422 423 /* 424 * Allocate the global restart stack which is the same for 425 * all CPUs in cast *one* of them does a PSW restart. 426 */ 427 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE); 428 if (!restart_stack) 429 panic("%s: Failed to allocate %lu bytes align=0x%lx\n", 430 __func__, THREAD_SIZE, THREAD_SIZE); 431 restart_stack += STACK_INIT_OFFSET; 432 433 /* 434 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 435 * restart data to the absolute zero lowcore. This is necessary if 436 * PSW restart is done on an offline CPU that has lowcore zero. 437 */ 438 lc->restart_stack = (unsigned long) restart_stack; 439 lc->restart_fn = (unsigned long) do_restart; 440 lc->restart_data = 0; 441 lc->restart_source = -1UL; 442 443 /* Setup absolute zero lowcore */ 444 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack); 445 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn); 446 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data); 447 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source); 448 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw); 449 450 lc->spinlock_lockval = arch_spin_lockval(0); 451 lc->spinlock_index = 0; 452 arch_spin_lock_setup(0); 453 lc->br_r1_trampoline = 0x07f1; /* br %r1 */ 454 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW); 455 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW); 456 457 set_prefix((u32)(unsigned long) lc); 458 lowcore_ptr[0] = lc; 459 } 460 461 static void __init setup_lowcore_dat_on(void) 462 { 463 __ctl_clear_bit(0, 28); 464 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT; 465 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT; 466 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT; 467 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT; 468 __ctl_set_bit(0, 28); 469 } 470 471 static struct resource code_resource = { 472 .name = "Kernel code", 473 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 474 }; 475 476 static struct resource data_resource = { 477 .name = "Kernel data", 478 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 479 }; 480 481 static struct resource bss_resource = { 482 .name = "Kernel bss", 483 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 484 }; 485 486 static struct resource __initdata *standard_resources[] = { 487 &code_resource, 488 &data_resource, 489 &bss_resource, 490 }; 491 492 static void __init setup_resources(void) 493 { 494 struct resource *res, *std_res, *sub_res; 495 phys_addr_t start, end; 496 int j; 497 u64 i; 498 499 code_resource.start = (unsigned long) _text; 500 code_resource.end = (unsigned long) _etext - 1; 501 data_resource.start = (unsigned long) _etext; 502 data_resource.end = (unsigned long) _edata - 1; 503 bss_resource.start = (unsigned long) __bss_start; 504 bss_resource.end = (unsigned long) __bss_stop - 1; 505 506 for_each_mem_range(i, &start, &end) { 507 res = memblock_alloc(sizeof(*res), 8); 508 if (!res) 509 panic("%s: Failed to allocate %zu bytes align=0x%x\n", 510 __func__, sizeof(*res), 8); 511 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM; 512 513 res->name = "System RAM"; 514 res->start = start; 515 /* 516 * In memblock, end points to the first byte after the 517 * range while in resourses, end points to the last byte in 518 * the range. 519 */ 520 res->end = end - 1; 521 request_resource(&iomem_resource, res); 522 523 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 524 std_res = standard_resources[j]; 525 if (std_res->start < res->start || 526 std_res->start > res->end) 527 continue; 528 if (std_res->end > res->end) { 529 sub_res = memblock_alloc(sizeof(*sub_res), 8); 530 if (!sub_res) 531 panic("%s: Failed to allocate %zu bytes align=0x%x\n", 532 __func__, sizeof(*sub_res), 8); 533 *sub_res = *std_res; 534 sub_res->end = res->end; 535 std_res->start = res->end + 1; 536 request_resource(res, sub_res); 537 } else { 538 request_resource(res, std_res); 539 } 540 } 541 } 542 #ifdef CONFIG_CRASH_DUMP 543 /* 544 * Re-add removed crash kernel memory as reserved memory. This makes 545 * sure it will be mapped with the identity mapping and struct pages 546 * will be created, so it can be resized later on. 547 * However add it later since the crash kernel resource should not be 548 * part of the System RAM resource. 549 */ 550 if (crashk_res.end) { 551 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0); 552 memblock_reserve(crashk_res.start, resource_size(&crashk_res)); 553 insert_resource(&iomem_resource, &crashk_res); 554 } 555 #endif 556 } 557 558 static void __init setup_ident_map_size(void) 559 { 560 unsigned long vmax, tmp; 561 562 /* Choose kernel address space layout: 3 or 4 levels. */ 563 tmp = ident_map_size / PAGE_SIZE; 564 tmp = tmp * (sizeof(struct page) + PAGE_SIZE); 565 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE) 566 vmax = _REGION2_SIZE; /* 3-level kernel page table */ 567 else 568 vmax = _REGION1_SIZE; /* 4-level kernel page table */ 569 /* module area is at the end of the kernel address space. */ 570 MODULES_END = vmax; 571 if (is_prot_virt_host()) 572 adjust_to_uv_max(&MODULES_END); 573 #ifdef CONFIG_KASAN 574 vmax = _REGION1_SIZE; 575 MODULES_END = kasan_vmax; 576 #endif 577 MODULES_VADDR = MODULES_END - MODULES_LEN; 578 VMALLOC_END = MODULES_VADDR; 579 VMALLOC_START = VMALLOC_END - vmalloc_size; 580 581 /* Split remaining virtual space between 1:1 mapping & vmemmap array */ 582 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page)); 583 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */ 584 tmp = SECTION_ALIGN_UP(tmp); 585 tmp = VMALLOC_START - tmp * sizeof(struct page); 586 tmp &= ~((vmax >> 11) - 1); /* align to page table level */ 587 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS); 588 vmemmap = (struct page *) tmp; 589 590 /* Take care that ident_map_size <= vmemmap */ 591 ident_map_size = min(ident_map_size, (unsigned long)vmemmap); 592 #ifdef CONFIG_KASAN 593 ident_map_size = min(ident_map_size, KASAN_SHADOW_START); 594 #endif 595 vmemmap_size = SECTION_ALIGN_UP(ident_map_size / PAGE_SIZE) * sizeof(struct page); 596 #ifdef CONFIG_KASAN 597 /* move vmemmap above kasan shadow only if stands in a way */ 598 if (KASAN_SHADOW_END > (unsigned long)vmemmap && 599 (unsigned long)vmemmap + vmemmap_size > KASAN_SHADOW_START) 600 vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END); 601 #endif 602 max_pfn = max_low_pfn = PFN_DOWN(ident_map_size); 603 memblock_remove(ident_map_size, ULONG_MAX); 604 605 pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20); 606 } 607 608 #ifdef CONFIG_CRASH_DUMP 609 610 /* 611 * When kdump is enabled, we have to ensure that no memory from the area 612 * [0 - crashkernel memory size] is set offline - it will be exchanged with 613 * the crashkernel memory region when kdump is triggered. The crashkernel 614 * memory region can never get offlined (pages are unmovable). 615 */ 616 static int kdump_mem_notifier(struct notifier_block *nb, 617 unsigned long action, void *data) 618 { 619 struct memory_notify *arg = data; 620 621 if (action != MEM_GOING_OFFLINE) 622 return NOTIFY_OK; 623 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 624 return NOTIFY_BAD; 625 return NOTIFY_OK; 626 } 627 628 static struct notifier_block kdump_mem_nb = { 629 .notifier_call = kdump_mem_notifier, 630 }; 631 632 #endif 633 634 /* 635 * Make sure that the area above identity mapping is protected 636 */ 637 static void __init reserve_above_ident_map(void) 638 { 639 memblock_reserve(ident_map_size, ULONG_MAX); 640 } 641 642 /* 643 * Make sure that oldmem, where the dump is stored, is protected 644 */ 645 static void __init reserve_oldmem(void) 646 { 647 #ifdef CONFIG_CRASH_DUMP 648 if (OLDMEM_BASE) 649 /* Forget all memory above the running kdump system */ 650 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 651 #endif 652 } 653 654 /* 655 * Make sure that oldmem, where the dump is stored, is protected 656 */ 657 static void __init remove_oldmem(void) 658 { 659 #ifdef CONFIG_CRASH_DUMP 660 if (OLDMEM_BASE) 661 /* Forget all memory above the running kdump system */ 662 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX); 663 #endif 664 } 665 666 /* 667 * Reserve memory for kdump kernel to be loaded with kexec 668 */ 669 static void __init reserve_crashkernel(void) 670 { 671 #ifdef CONFIG_CRASH_DUMP 672 unsigned long long crash_base, crash_size; 673 phys_addr_t low, high; 674 int rc; 675 676 rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size, 677 &crash_base); 678 679 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 680 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 681 if (rc || crash_size == 0) 682 return; 683 684 if (memblock.memory.regions[0].size < crash_size) { 685 pr_info("crashkernel reservation failed: %s\n", 686 "first memory chunk must be at least crashkernel size"); 687 return; 688 } 689 690 low = crash_base ?: OLDMEM_BASE; 691 high = low + crash_size; 692 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) { 693 /* The crashkernel fits into OLDMEM, reuse OLDMEM */ 694 crash_base = low; 695 } else { 696 /* Find suitable area in free memory */ 697 low = max_t(unsigned long, crash_size, sclp.hsa_size); 698 high = crash_base ? crash_base + crash_size : ULONG_MAX; 699 700 if (crash_base && crash_base < low) { 701 pr_info("crashkernel reservation failed: %s\n", 702 "crash_base too low"); 703 return; 704 } 705 low = crash_base ?: low; 706 crash_base = memblock_find_in_range(low, high, crash_size, 707 KEXEC_CRASH_MEM_ALIGN); 708 } 709 710 if (!crash_base) { 711 pr_info("crashkernel reservation failed: %s\n", 712 "no suitable area found"); 713 return; 714 } 715 716 if (register_memory_notifier(&kdump_mem_nb)) 717 return; 718 719 if (!OLDMEM_BASE && MACHINE_IS_VM) 720 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 721 crashk_res.start = crash_base; 722 crashk_res.end = crash_base + crash_size - 1; 723 memblock_remove(crash_base, crash_size); 724 pr_info("Reserving %lluMB of memory at %lluMB " 725 "for crashkernel (System RAM: %luMB)\n", 726 crash_size >> 20, crash_base >> 20, 727 (unsigned long)memblock.memory.total_size >> 20); 728 os_info_crashkernel_add(crash_base, crash_size); 729 #endif 730 } 731 732 /* 733 * Reserve the initrd from being used by memblock 734 */ 735 static void __init reserve_initrd(void) 736 { 737 #ifdef CONFIG_BLK_DEV_INITRD 738 if (!INITRD_START || !INITRD_SIZE) 739 return; 740 initrd_start = INITRD_START; 741 initrd_end = initrd_start + INITRD_SIZE; 742 memblock_reserve(INITRD_START, INITRD_SIZE); 743 #endif 744 } 745 746 /* 747 * Reserve the memory area used to pass the certificate lists 748 */ 749 static void __init reserve_certificate_list(void) 750 { 751 if (ipl_cert_list_addr) 752 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size); 753 } 754 755 static void __init reserve_mem_detect_info(void) 756 { 757 unsigned long start, size; 758 759 get_mem_detect_reserved(&start, &size); 760 if (size) 761 memblock_reserve(start, size); 762 } 763 764 static void __init free_mem_detect_info(void) 765 { 766 unsigned long start, size; 767 768 get_mem_detect_reserved(&start, &size); 769 if (size) 770 memblock_free(start, size); 771 } 772 773 static const char * __init get_mem_info_source(void) 774 { 775 switch (mem_detect.info_source) { 776 case MEM_DETECT_SCLP_STOR_INFO: 777 return "sclp storage info"; 778 case MEM_DETECT_DIAG260: 779 return "diag260"; 780 case MEM_DETECT_SCLP_READ_INFO: 781 return "sclp read info"; 782 case MEM_DETECT_BIN_SEARCH: 783 return "binary search"; 784 } 785 return "none"; 786 } 787 788 static void __init memblock_add_mem_detect_info(void) 789 { 790 unsigned long start, end; 791 int i; 792 793 pr_debug("physmem info source: %s (%hhd)\n", 794 get_mem_info_source(), mem_detect.info_source); 795 /* keep memblock lists close to the kernel */ 796 memblock_set_bottom_up(true); 797 for_each_mem_detect_block(i, &start, &end) { 798 memblock_add(start, end - start); 799 memblock_physmem_add(start, end - start); 800 } 801 memblock_set_bottom_up(false); 802 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0); 803 memblock_dump_all(); 804 } 805 806 /* 807 * Check for initrd being in usable memory 808 */ 809 static void __init check_initrd(void) 810 { 811 #ifdef CONFIG_BLK_DEV_INITRD 812 if (INITRD_START && INITRD_SIZE && 813 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) { 814 pr_err("The initial RAM disk does not fit into the memory\n"); 815 memblock_free(INITRD_START, INITRD_SIZE); 816 initrd_start = initrd_end = 0; 817 } 818 #endif 819 } 820 821 /* 822 * Reserve memory used for lowcore/command line/kernel image. 823 */ 824 static void __init reserve_kernel(void) 825 { 826 unsigned long start_pfn = PFN_UP(__pa(_end)); 827 828 memblock_reserve(0, HEAD_END); 829 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn) 830 - (unsigned long)_stext); 831 memblock_reserve(__sdma, __edma - __sdma); 832 } 833 834 static void __init setup_memory(void) 835 { 836 phys_addr_t start, end; 837 u64 i; 838 839 /* 840 * Init storage key for present memory 841 */ 842 for_each_mem_range(i, &start, &end) 843 storage_key_init_range(start, end); 844 845 psw_set_key(PAGE_DEFAULT_KEY); 846 847 /* Only cosmetics */ 848 memblock_enforce_memory_limit(memblock_end_of_DRAM()); 849 } 850 851 /* 852 * Setup hardware capabilities. 853 */ 854 static int __init setup_hwcaps(void) 855 { 856 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 }; 857 struct cpuid cpu_id; 858 int i; 859 860 /* 861 * The store facility list bits numbers as found in the principles 862 * of operation are numbered with bit 1UL<<31 as number 0 to 863 * bit 1UL<<0 as number 31. 864 * Bit 0: instructions named N3, "backported" to esa-mode 865 * Bit 2: z/Architecture mode is active 866 * Bit 7: the store-facility-list-extended facility is installed 867 * Bit 17: the message-security assist is installed 868 * Bit 19: the long-displacement facility is installed 869 * Bit 21: the extended-immediate facility is installed 870 * Bit 22: extended-translation facility 3 is installed 871 * Bit 30: extended-translation facility 3 enhancement facility 872 * These get translated to: 873 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1, 874 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3, 875 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and 876 * HWCAP_S390_ETF3EH bit 8 (22 && 30). 877 */ 878 for (i = 0; i < 6; i++) 879 if (test_facility(stfl_bits[i])) 880 elf_hwcap |= 1UL << i; 881 882 if (test_facility(22) && test_facility(30)) 883 elf_hwcap |= HWCAP_S390_ETF3EH; 884 885 /* 886 * Check for additional facilities with store-facility-list-extended. 887 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0 888 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information 889 * as stored by stfl, bits 32-xxx contain additional facilities. 890 * How many facility words are stored depends on the number of 891 * doublewords passed to the instruction. The additional facilities 892 * are: 893 * Bit 42: decimal floating point facility is installed 894 * Bit 44: perform floating point operation facility is installed 895 * translated to: 896 * HWCAP_S390_DFP bit 6 (42 && 44). 897 */ 898 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44)) 899 elf_hwcap |= HWCAP_S390_DFP; 900 901 /* 902 * Huge page support HWCAP_S390_HPAGE is bit 7. 903 */ 904 if (MACHINE_HAS_EDAT1) 905 elf_hwcap |= HWCAP_S390_HPAGE; 906 907 /* 908 * 64-bit register support for 31-bit processes 909 * HWCAP_S390_HIGH_GPRS is bit 9. 910 */ 911 elf_hwcap |= HWCAP_S390_HIGH_GPRS; 912 913 /* 914 * Transactional execution support HWCAP_S390_TE is bit 10. 915 */ 916 if (MACHINE_HAS_TE) 917 elf_hwcap |= HWCAP_S390_TE; 918 919 /* 920 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension 921 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX 922 * instead of facility bit 129. 923 */ 924 if (MACHINE_HAS_VX) { 925 elf_hwcap |= HWCAP_S390_VXRS; 926 if (test_facility(134)) 927 elf_hwcap |= HWCAP_S390_VXRS_EXT; 928 if (test_facility(135)) 929 elf_hwcap |= HWCAP_S390_VXRS_BCD; 930 if (test_facility(148)) 931 elf_hwcap |= HWCAP_S390_VXRS_EXT2; 932 if (test_facility(152)) 933 elf_hwcap |= HWCAP_S390_VXRS_PDE; 934 } 935 if (test_facility(150)) 936 elf_hwcap |= HWCAP_S390_SORT; 937 if (test_facility(151)) 938 elf_hwcap |= HWCAP_S390_DFLT; 939 940 /* 941 * Guarded storage support HWCAP_S390_GS is bit 12. 942 */ 943 if (MACHINE_HAS_GS) 944 elf_hwcap |= HWCAP_S390_GS; 945 946 get_cpu_id(&cpu_id); 947 add_device_randomness(&cpu_id, sizeof(cpu_id)); 948 switch (cpu_id.machine) { 949 case 0x2064: 950 case 0x2066: 951 default: /* Use "z900" as default for 64 bit kernels. */ 952 strcpy(elf_platform, "z900"); 953 break; 954 case 0x2084: 955 case 0x2086: 956 strcpy(elf_platform, "z990"); 957 break; 958 case 0x2094: 959 case 0x2096: 960 strcpy(elf_platform, "z9-109"); 961 break; 962 case 0x2097: 963 case 0x2098: 964 strcpy(elf_platform, "z10"); 965 break; 966 case 0x2817: 967 case 0x2818: 968 strcpy(elf_platform, "z196"); 969 break; 970 case 0x2827: 971 case 0x2828: 972 strcpy(elf_platform, "zEC12"); 973 break; 974 case 0x2964: 975 case 0x2965: 976 strcpy(elf_platform, "z13"); 977 break; 978 case 0x3906: 979 case 0x3907: 980 strcpy(elf_platform, "z14"); 981 break; 982 case 0x8561: 983 case 0x8562: 984 strcpy(elf_platform, "z15"); 985 break; 986 } 987 988 /* 989 * Virtualization support HWCAP_INT_SIE is bit 0. 990 */ 991 if (sclp.has_sief2) 992 int_hwcap |= HWCAP_INT_SIE; 993 994 return 0; 995 } 996 arch_initcall(setup_hwcaps); 997 998 /* 999 * Add system information as device randomness 1000 */ 1001 static void __init setup_randomness(void) 1002 { 1003 struct sysinfo_3_2_2 *vmms; 1004 1005 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE, 1006 PAGE_SIZE); 1007 if (!vmms) 1008 panic("Failed to allocate memory for sysinfo structure\n"); 1009 1010 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count) 1011 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count); 1012 memblock_free((unsigned long) vmms, PAGE_SIZE); 1013 } 1014 1015 /* 1016 * Find the correct size for the task_struct. This depends on 1017 * the size of the struct fpu at the end of the thread_struct 1018 * which is embedded in the task_struct. 1019 */ 1020 static void __init setup_task_size(void) 1021 { 1022 int task_size = sizeof(struct task_struct); 1023 1024 if (!MACHINE_HAS_VX) { 1025 task_size -= sizeof(__vector128) * __NUM_VXRS; 1026 task_size += sizeof(freg_t) * __NUM_FPRS; 1027 } 1028 arch_task_struct_size = task_size; 1029 } 1030 1031 /* 1032 * Issue diagnose 318 to set the control program name and 1033 * version codes. 1034 */ 1035 static void __init setup_control_program_code(void) 1036 { 1037 union diag318_info diag318_info = { 1038 .cpnc = CPNC_LINUX, 1039 .cpvc = 0, 1040 }; 1041 1042 if (!sclp.has_diag318) 1043 return; 1044 1045 diag_stat_inc(DIAG_STAT_X318); 1046 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val)); 1047 } 1048 1049 /* 1050 * Print the component list from the IPL report 1051 */ 1052 static void __init log_component_list(void) 1053 { 1054 struct ipl_rb_component_entry *ptr, *end; 1055 char *str; 1056 1057 if (!early_ipl_comp_list_addr) 1058 return; 1059 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL) 1060 pr_info("Linux is running with Secure-IPL enabled\n"); 1061 else 1062 pr_info("Linux is running with Secure-IPL disabled\n"); 1063 ptr = (void *) early_ipl_comp_list_addr; 1064 end = (void *) ptr + early_ipl_comp_list_size; 1065 pr_info("The IPL report contains the following components:\n"); 1066 while (ptr < end) { 1067 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) { 1068 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED) 1069 str = "signed, verified"; 1070 else 1071 str = "signed, verification failed"; 1072 } else { 1073 str = "not signed"; 1074 } 1075 pr_info("%016llx - %016llx (%s)\n", 1076 ptr->addr, ptr->addr + ptr->len, str); 1077 ptr++; 1078 } 1079 } 1080 1081 /* 1082 * Setup function called from init/main.c just after the banner 1083 * was printed. 1084 */ 1085 1086 void __init setup_arch(char **cmdline_p) 1087 { 1088 /* 1089 * print what head.S has found out about the machine 1090 */ 1091 if (MACHINE_IS_VM) 1092 pr_info("Linux is running as a z/VM " 1093 "guest operating system in 64-bit mode\n"); 1094 else if (MACHINE_IS_KVM) 1095 pr_info("Linux is running under KVM in 64-bit mode\n"); 1096 else if (MACHINE_IS_LPAR) 1097 pr_info("Linux is running natively in 64-bit mode\n"); 1098 else 1099 pr_info("Linux is running as a guest in 64-bit mode\n"); 1100 1101 log_component_list(); 1102 1103 /* Have one command line that is parsed and saved in /proc/cmdline */ 1104 /* boot_command_line has been already set up in early.c */ 1105 *cmdline_p = boot_command_line; 1106 1107 ROOT_DEV = Root_RAM0; 1108 1109 init_mm.start_code = (unsigned long) _text; 1110 init_mm.end_code = (unsigned long) _etext; 1111 init_mm.end_data = (unsigned long) _edata; 1112 init_mm.brk = (unsigned long) _end; 1113 1114 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO)) 1115 nospec_auto_detect(); 1116 1117 jump_label_init(); 1118 parse_early_param(); 1119 #ifdef CONFIG_CRASH_DUMP 1120 /* Deactivate elfcorehdr= kernel parameter */ 1121 elfcorehdr_addr = ELFCORE_ADDR_MAX; 1122 #endif 1123 1124 os_info_init(); 1125 setup_ipl(); 1126 setup_task_size(); 1127 setup_control_program_code(); 1128 1129 /* Do some memory reservations *before* memory is added to memblock */ 1130 reserve_above_ident_map(); 1131 reserve_oldmem(); 1132 reserve_kernel(); 1133 reserve_initrd(); 1134 reserve_certificate_list(); 1135 reserve_mem_detect_info(); 1136 memblock_allow_resize(); 1137 1138 /* Get information about *all* installed memory */ 1139 memblock_add_mem_detect_info(); 1140 1141 free_mem_detect_info(); 1142 remove_oldmem(); 1143 1144 setup_uv(); 1145 setup_ident_map_size(); 1146 setup_memory(); 1147 dma_contiguous_reserve(ident_map_size); 1148 vmcp_cma_reserve(); 1149 if (MACHINE_HAS_EDAT2) 1150 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT); 1151 1152 check_initrd(); 1153 reserve_crashkernel(); 1154 #ifdef CONFIG_CRASH_DUMP 1155 /* 1156 * Be aware that smp_save_dump_cpus() triggers a system reset. 1157 * Therefore CPU and device initialization should be done afterwards. 1158 */ 1159 smp_save_dump_cpus(); 1160 #endif 1161 1162 setup_resources(); 1163 setup_lowcore_dat_off(); 1164 smp_fill_possible_mask(); 1165 cpu_detect_mhz_feature(); 1166 cpu_init(); 1167 numa_setup(); 1168 smp_detect_cpus(); 1169 topology_init_early(); 1170 1171 /* 1172 * Create kernel page tables and switch to virtual addressing. 1173 */ 1174 paging_init(); 1175 1176 /* 1177 * After paging_init created the kernel page table, the new PSWs 1178 * in lowcore can now run with DAT enabled. 1179 */ 1180 setup_lowcore_dat_on(); 1181 1182 /* Setup default console */ 1183 conmode_default(); 1184 set_preferred_console(); 1185 1186 apply_alternative_instructions(); 1187 if (IS_ENABLED(CONFIG_EXPOLINE)) 1188 nospec_init_branches(); 1189 1190 /* Setup zfcp/nvme dump support */ 1191 setup_zfcpdump(); 1192 1193 /* Add system specific data to the random pool */ 1194 setup_randomness(); 1195 } 1196