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/start_kernel.h> 51 #include <linux/hugetlb.h> 52 #include <linux/kmemleak.h> 53 54 #include <asm/archrandom.h> 55 #include <asm/boot_data.h> 56 #include <asm/machine.h> 57 #include <asm/ipl.h> 58 #include <asm/facility.h> 59 #include <asm/smp.h> 60 #include <asm/mmu_context.h> 61 #include <asm/cpcmd.h> 62 #include <asm/abs_lowcore.h> 63 #include <asm/nmi.h> 64 #include <asm/irq.h> 65 #include <asm/page.h> 66 #include <asm/ptrace.h> 67 #include <asm/sections.h> 68 #include <asm/ebcdic.h> 69 #include <asm/diag.h> 70 #include <asm/os_info.h> 71 #include <asm/sclp.h> 72 #include <asm/stacktrace.h> 73 #include <asm/sysinfo.h> 74 #include <asm/numa.h> 75 #include <asm/alternative.h> 76 #include <asm/nospec-branch.h> 77 #include <asm/physmem_info.h> 78 #include <asm/maccess.h> 79 #include <asm/uv.h> 80 #include <asm/asm-offsets.h> 81 #include "entry.h" 82 83 /* 84 * Machine setup.. 85 */ 86 unsigned int console_mode = 0; 87 EXPORT_SYMBOL(console_mode); 88 89 unsigned int console_devno = -1; 90 EXPORT_SYMBOL(console_devno); 91 92 unsigned int console_irq = -1; 93 EXPORT_SYMBOL(console_irq); 94 95 /* 96 * Some code and data needs to stay below 2 GB, even when the kernel would be 97 * relocated above 2 GB, because it has to use 31 bit addresses. 98 * Such code and data is part of the .amode31 section. 99 */ 100 char __amode31_ref *__samode31 = _samode31; 101 char __amode31_ref *__eamode31 = _eamode31; 102 char __amode31_ref *__stext_amode31 = _stext_amode31; 103 char __amode31_ref *__etext_amode31 = _etext_amode31; 104 struct exception_table_entry __amode31_ref *__start_amode31_ex_table = _start_amode31_ex_table; 105 struct exception_table_entry __amode31_ref *__stop_amode31_ex_table = _stop_amode31_ex_table; 106 107 /* 108 * Control registers CR2, CR5 and CR15 are initialized with addresses 109 * of tables that must be placed below 2G which is handled by the AMODE31 110 * sections. 111 * Because the AMODE31 sections are relocated below 2G at startup, 112 * the content of control registers CR2, CR5 and CR15 must be updated 113 * with new addresses after the relocation. The initial initialization of 114 * control registers occurs in head64.S and then gets updated again after AMODE31 115 * relocation. We must access the relevant AMODE31 tables indirectly via 116 * pointers placed in the .amode31.refs linker section. Those pointers get 117 * updated automatically during AMODE31 relocation and always contain a valid 118 * address within AMODE31 sections. 119 */ 120 121 static __amode31_data u32 __ctl_duct_amode31[16] __aligned(64); 122 123 static __amode31_data u64 __ctl_aste_amode31[8] __aligned(64) = { 124 [1] = 0xffffffffffffffff 125 }; 126 127 static __amode31_data u32 __ctl_duald_amode31[32] __aligned(128) = { 128 0x80000000, 0, 0, 0, 129 0x80000000, 0, 0, 0, 130 0x80000000, 0, 0, 0, 131 0x80000000, 0, 0, 0, 132 0x80000000, 0, 0, 0, 133 0x80000000, 0, 0, 0, 134 0x80000000, 0, 0, 0, 135 0x80000000, 0, 0, 0 136 }; 137 138 static __amode31_data u32 __ctl_linkage_stack_amode31[8] __aligned(64) = { 139 0, 0, 0x89000000, 0, 140 0, 0, 0x8a000000, 0 141 }; 142 143 static u64 __amode31_ref *__ctl_aste = __ctl_aste_amode31; 144 static u32 __amode31_ref *__ctl_duald = __ctl_duald_amode31; 145 static u32 __amode31_ref *__ctl_linkage_stack = __ctl_linkage_stack_amode31; 146 static u32 __amode31_ref *__ctl_duct = __ctl_duct_amode31; 147 148 unsigned long __bootdata_preserved(max_mappable); 149 struct physmem_info __bootdata(physmem_info); 150 151 struct vm_layout __bootdata_preserved(vm_layout); 152 EXPORT_SYMBOL(vm_layout); 153 int __bootdata_preserved(__kaslr_enabled); 154 unsigned int __bootdata_preserved(zlib_dfltcc_support); 155 EXPORT_SYMBOL(zlib_dfltcc_support); 156 u64 __bootdata_preserved(stfle_fac_list[16]); 157 EXPORT_SYMBOL(stfle_fac_list); 158 struct oldmem_data __bootdata_preserved(oldmem_data); 159 160 char __bootdata(boot_rb)[PAGE_SIZE * 2]; 161 bool __bootdata(boot_earlyprintk); 162 size_t __bootdata(boot_rb_off); 163 char __bootdata(bootdebug_filter)[128]; 164 bool __bootdata(bootdebug); 165 166 unsigned long __bootdata_preserved(VMALLOC_START); 167 EXPORT_SYMBOL(VMALLOC_START); 168 169 unsigned long __bootdata_preserved(VMALLOC_END); 170 EXPORT_SYMBOL(VMALLOC_END); 171 172 struct page *__bootdata_preserved(vmemmap); 173 EXPORT_SYMBOL(vmemmap); 174 unsigned long __bootdata_preserved(vmemmap_size); 175 176 unsigned long __bootdata_preserved(MODULES_VADDR); 177 unsigned long __bootdata_preserved(MODULES_END); 178 179 /* An array with a pointer to the lowcore of every CPU. */ 180 struct lowcore *lowcore_ptr[NR_CPUS]; 181 EXPORT_SYMBOL(lowcore_ptr); 182 183 /* 184 * The Write Back bit position in the physaddr is given by the SLPC PCI. 185 * Leaving the mask zero always uses write through which is safe 186 */ 187 unsigned long mio_wb_bit_mask __ro_after_init; 188 189 /* 190 * This is set up by the setup-routine at boot-time 191 * for S390 need to find out, what we have to setup 192 * using address 0x10400 ... 193 */ 194 195 #include <asm/setup.h> 196 197 /* 198 * condev= and conmode= setup parameter. 199 */ 200 201 static int __init condev_setup(char *str) 202 { 203 int vdev; 204 205 vdev = simple_strtoul(str, &str, 0); 206 if (vdev >= 0 && vdev < 65536) { 207 console_devno = vdev; 208 console_irq = -1; 209 } 210 return 1; 211 } 212 213 __setup("condev=", condev_setup); 214 215 static void __init set_preferred_console(void) 216 { 217 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP) 218 add_preferred_console("ttyS", 0, NULL); 219 else if (CONSOLE_IS_3270) 220 add_preferred_console("tty3270", 0, NULL); 221 else if (CONSOLE_IS_VT220) 222 add_preferred_console("ttysclp", 0, NULL); 223 else if (CONSOLE_IS_HVC) 224 add_preferred_console("hvc", 0, NULL); 225 } 226 227 static int __init conmode_setup(char *str) 228 { 229 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 230 if (!strcmp(str, "hwc") || !strcmp(str, "sclp")) 231 SET_CONSOLE_SCLP; 232 #endif 233 #if defined(CONFIG_TN3215_CONSOLE) 234 if (!strcmp(str, "3215")) 235 SET_CONSOLE_3215; 236 #endif 237 #if defined(CONFIG_TN3270_CONSOLE) 238 if (!strcmp(str, "3270")) 239 SET_CONSOLE_3270; 240 #endif 241 set_preferred_console(); 242 return 1; 243 } 244 245 __setup("conmode=", conmode_setup); 246 247 static void __init conmode_default(void) 248 { 249 char query_buffer[1024]; 250 char *ptr; 251 252 if (machine_is_vm()) { 253 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL); 254 console_devno = simple_strtoul(query_buffer + 5, NULL, 16); 255 ptr = strstr(query_buffer, "SUBCHANNEL ="); 256 console_irq = simple_strtoul(ptr + 13, NULL, 16); 257 cpcmd("QUERY TERM", query_buffer, 1024, NULL); 258 ptr = strstr(query_buffer, "CONMODE"); 259 /* 260 * Set the conmode to 3215 so that the device recognition 261 * will set the cu_type of the console to 3215. If the 262 * conmode is 3270 and we don't set it back then both 263 * 3215 and the 3270 driver will try to access the console 264 * device (3215 as console and 3270 as normal tty). 265 */ 266 cpcmd("TERM CONMODE 3215", NULL, 0, NULL); 267 if (ptr == NULL) { 268 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 269 SET_CONSOLE_SCLP; 270 #endif 271 return; 272 } 273 if (str_has_prefix(ptr + 8, "3270")) { 274 #if defined(CONFIG_TN3270_CONSOLE) 275 SET_CONSOLE_3270; 276 #elif defined(CONFIG_TN3215_CONSOLE) 277 SET_CONSOLE_3215; 278 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 279 SET_CONSOLE_SCLP; 280 #endif 281 } else if (str_has_prefix(ptr + 8, "3215")) { 282 #if defined(CONFIG_TN3215_CONSOLE) 283 SET_CONSOLE_3215; 284 #elif defined(CONFIG_TN3270_CONSOLE) 285 SET_CONSOLE_3270; 286 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 287 SET_CONSOLE_SCLP; 288 #endif 289 } 290 } else if (machine_is_kvm()) { 291 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE)) 292 SET_CONSOLE_VT220; 293 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE)) 294 SET_CONSOLE_SCLP; 295 else 296 SET_CONSOLE_HVC; 297 } else { 298 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 299 SET_CONSOLE_SCLP; 300 #endif 301 } 302 } 303 304 #ifdef CONFIG_CRASH_DUMP 305 static void __init setup_zfcpdump(void) 306 { 307 if (!is_ipl_type_dump()) 308 return; 309 if (oldmem_data.start) 310 return; 311 strlcat(boot_command_line, " cio_ignore=all,!ipldev,!condev", COMMAND_LINE_SIZE); 312 console_loglevel = 2; 313 } 314 #else 315 static inline void setup_zfcpdump(void) {} 316 #endif /* CONFIG_CRASH_DUMP */ 317 318 /* 319 * Reboot, halt and power_off stubs. They just call _machine_restart, 320 * _machine_halt or _machine_power_off. 321 */ 322 323 void machine_restart(char *command) 324 { 325 if ((!in_interrupt() && !in_atomic()) || oops_in_progress) 326 /* 327 * Only unblank the console if we are called in enabled 328 * context or a bust_spinlocks cleared the way for us. 329 */ 330 console_unblank(); 331 _machine_restart(command); 332 } 333 334 void machine_halt(void) 335 { 336 if (!in_interrupt() || oops_in_progress) 337 /* 338 * Only unblank the console if we are called in enabled 339 * context or a bust_spinlocks cleared the way for us. 340 */ 341 console_unblank(); 342 _machine_halt(); 343 } 344 345 void machine_power_off(void) 346 { 347 if (!in_interrupt() || oops_in_progress) 348 /* 349 * Only unblank the console if we are called in enabled 350 * context or a bust_spinlocks cleared the way for us. 351 */ 352 console_unblank(); 353 _machine_power_off(); 354 } 355 356 /* 357 * Dummy power off function. 358 */ 359 void (*pm_power_off)(void) = machine_power_off; 360 EXPORT_SYMBOL_GPL(pm_power_off); 361 362 void *restart_stack; 363 364 unsigned long stack_alloc(void) 365 { 366 void *stack; 367 368 stack = __vmalloc_node(THREAD_SIZE, THREAD_SIZE, THREADINFO_GFP, 369 NUMA_NO_NODE, __builtin_return_address(0)); 370 kmemleak_not_leak(stack); 371 return (unsigned long)stack; 372 } 373 374 void stack_free(unsigned long stack) 375 { 376 vfree((void *)stack); 377 } 378 379 static unsigned long __init stack_alloc_early(void) 380 { 381 unsigned long stack; 382 383 stack = (unsigned long)memblock_alloc_or_panic(THREAD_SIZE, THREAD_SIZE); 384 return stack; 385 } 386 387 static void __init setup_lowcore(void) 388 { 389 struct lowcore *lc, *abs_lc; 390 391 /* 392 * Setup lowcore for boot cpu 393 */ 394 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE); 395 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc)); 396 if (!lc) 397 panic("%s: Failed to allocate %zu bytes align=%zx\n", 398 __func__, sizeof(*lc), sizeof(*lc)); 399 400 lc->pcpu = (unsigned long)per_cpu_ptr(&pcpu_devices, 0); 401 lc->restart_psw.mask = PSW_KERNEL_BITS & ~PSW_MASK_DAT; 402 lc->restart_psw.addr = __pa(restart_int_handler); 403 lc->external_new_psw.mask = PSW_KERNEL_BITS; 404 lc->external_new_psw.addr = (unsigned long) ext_int_handler; 405 lc->svc_new_psw.mask = PSW_KERNEL_BITS; 406 lc->svc_new_psw.addr = (unsigned long) system_call; 407 lc->program_new_psw.mask = PSW_KERNEL_BITS; 408 lc->program_new_psw.addr = (unsigned long) pgm_check_handler; 409 lc->mcck_new_psw.mask = PSW_KERNEL_BITS; 410 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler; 411 lc->io_new_psw.mask = PSW_KERNEL_BITS; 412 lc->io_new_psw.addr = (unsigned long) io_int_handler; 413 lc->clock_comparator = clock_comparator_max; 414 lc->current_task = (unsigned long)&init_task; 415 lc->lpp = LPP_MAGIC; 416 lc->preempt_count = get_lowcore()->preempt_count; 417 nmi_alloc_mcesa_early(&lc->mcesad); 418 lc->sys_enter_timer = get_lowcore()->sys_enter_timer; 419 lc->exit_timer = get_lowcore()->exit_timer; 420 lc->user_timer = get_lowcore()->user_timer; 421 lc->system_timer = get_lowcore()->system_timer; 422 lc->steal_timer = get_lowcore()->steal_timer; 423 lc->last_update_timer = get_lowcore()->last_update_timer; 424 lc->last_update_clock = get_lowcore()->last_update_clock; 425 /* 426 * Allocate the global restart stack which is the same for 427 * all CPUs in case *one* of them does a PSW restart. 428 */ 429 restart_stack = (void *)(stack_alloc_early() + STACK_INIT_OFFSET); 430 lc->mcck_stack = stack_alloc_early() + STACK_INIT_OFFSET; 431 lc->async_stack = stack_alloc_early() + STACK_INIT_OFFSET; 432 lc->nodat_stack = stack_alloc_early() + STACK_INIT_OFFSET; 433 lc->kernel_stack = get_lowcore()->kernel_stack; 434 /* 435 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 436 * restart data to the absolute zero lowcore. This is necessary if 437 * PSW restart is done on an offline CPU that has lowcore zero. 438 */ 439 lc->restart_stack = (unsigned long) restart_stack; 440 lc->restart_fn = (unsigned long) do_restart; 441 lc->restart_data = 0; 442 lc->restart_source = -1U; 443 lc->spinlock_lockval = arch_spin_lockval(0); 444 lc->spinlock_index = 0; 445 arch_spin_lock_setup(0); 446 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW); 447 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW); 448 lc->preempt_count = PREEMPT_DISABLED; 449 lc->kernel_asce = get_lowcore()->kernel_asce; 450 lc->user_asce = get_lowcore()->user_asce; 451 452 system_ctlreg_init_save_area(lc); 453 abs_lc = get_abs_lowcore(); 454 abs_lc->restart_stack = lc->restart_stack; 455 abs_lc->restart_fn = lc->restart_fn; 456 abs_lc->restart_data = lc->restart_data; 457 abs_lc->restart_source = lc->restart_source; 458 abs_lc->restart_psw = lc->restart_psw; 459 abs_lc->restart_flags = RESTART_FLAG_CTLREGS; 460 abs_lc->program_new_psw = lc->program_new_psw; 461 abs_lc->mcesad = lc->mcesad; 462 put_abs_lowcore(abs_lc); 463 464 set_prefix(__pa(lc)); 465 lowcore_ptr[0] = lc; 466 if (abs_lowcore_map(0, lowcore_ptr[0], false)) 467 panic("Couldn't setup absolute lowcore"); 468 } 469 470 static struct resource code_resource = { 471 .name = "Kernel code", 472 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 473 }; 474 475 static struct resource data_resource = { 476 .name = "Kernel data", 477 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 478 }; 479 480 static struct resource bss_resource = { 481 .name = "Kernel bss", 482 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 483 }; 484 485 static struct resource __initdata *standard_resources[] = { 486 &code_resource, 487 &data_resource, 488 &bss_resource, 489 }; 490 491 static void __init setup_resources(void) 492 { 493 struct resource *res, *std_res, *sub_res; 494 phys_addr_t start, end; 495 int j; 496 u64 i; 497 498 code_resource.start = __pa_symbol(_text); 499 code_resource.end = __pa_symbol(_etext) - 1; 500 data_resource.start = __pa_symbol(_etext); 501 data_resource.end = __pa_symbol(_edata) - 1; 502 bss_resource.start = __pa_symbol(__bss_start); 503 bss_resource.end = __pa_symbol(__bss_stop) - 1; 504 505 for_each_mem_range(i, &start, &end) { 506 res = memblock_alloc_or_panic(sizeof(*res), 8); 507 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM; 508 509 res->name = "System RAM"; 510 res->start = start; 511 /* 512 * In memblock, end points to the first byte after the 513 * range while in resources, end points to the last byte in 514 * the range. 515 */ 516 res->end = end - 1; 517 request_resource(&iomem_resource, res); 518 519 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 520 std_res = standard_resources[j]; 521 if (std_res->start < res->start || 522 std_res->start > res->end) 523 continue; 524 if (std_res->end > res->end) { 525 sub_res = memblock_alloc_or_panic(sizeof(*sub_res), 8); 526 *sub_res = *std_res; 527 sub_res->end = res->end; 528 std_res->start = res->end + 1; 529 request_resource(res, sub_res); 530 } else { 531 request_resource(res, std_res); 532 } 533 } 534 } 535 #ifdef CONFIG_CRASH_DUMP 536 /* 537 * Re-add removed crash kernel memory as reserved memory. This makes 538 * sure it will be mapped with the identity mapping and struct pages 539 * will be created, so it can be resized later on. 540 * However add it later since the crash kernel resource should not be 541 * part of the System RAM resource. 542 */ 543 if (crashk_res.end) { 544 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 545 0, MEMBLOCK_NONE); 546 memblock_reserve(crashk_res.start, resource_size(&crashk_res)); 547 insert_resource(&iomem_resource, &crashk_res); 548 } 549 #endif 550 } 551 552 static void __init setup_memory_end(void) 553 { 554 max_pfn = max_low_pfn = PFN_DOWN(ident_map_size); 555 pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20); 556 } 557 558 #ifdef CONFIG_CRASH_DUMP 559 560 /* 561 * When kdump is enabled, we have to ensure that no memory from the area 562 * [0 - crashkernel memory size] is set offline - it will be exchanged with 563 * the crashkernel memory region when kdump is triggered. The crashkernel 564 * memory region can never get offlined (pages are unmovable). 565 */ 566 static int kdump_mem_notifier(struct notifier_block *nb, 567 unsigned long action, void *data) 568 { 569 struct memory_notify *arg = data; 570 571 if (action != MEM_GOING_OFFLINE) 572 return NOTIFY_OK; 573 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 574 return NOTIFY_BAD; 575 return NOTIFY_OK; 576 } 577 578 static struct notifier_block kdump_mem_nb = { 579 .notifier_call = kdump_mem_notifier, 580 }; 581 582 #endif 583 584 /* 585 * Reserve page tables created by decompressor 586 */ 587 static void __init reserve_pgtables(void) 588 { 589 unsigned long start, end; 590 struct reserved_range *range; 591 592 for_each_physmem_reserved_type_range(RR_VMEM, range, &start, &end) 593 memblock_reserve(start, end - start); 594 } 595 596 /* 597 * Reserve memory for kdump kernel to be loaded with kexec 598 */ 599 static void __init reserve_crashkernel(void) 600 { 601 #ifdef CONFIG_CRASH_DUMP 602 unsigned long long crash_base, crash_size; 603 phys_addr_t low, high; 604 int rc; 605 606 rc = parse_crashkernel(boot_command_line, ident_map_size, 607 &crash_size, &crash_base, NULL, NULL, NULL); 608 609 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 610 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 611 if (rc || crash_size == 0) 612 return; 613 614 if (memblock.memory.regions[0].size < crash_size) { 615 pr_info("crashkernel reservation failed: %s\n", 616 "first memory chunk must be at least crashkernel size"); 617 return; 618 } 619 620 low = crash_base ?: oldmem_data.start; 621 high = low + crash_size; 622 if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) { 623 /* The crashkernel fits into OLDMEM, reuse OLDMEM */ 624 crash_base = low; 625 } else { 626 /* Find suitable area in free memory */ 627 low = max_t(unsigned long, crash_size, sclp.hsa_size); 628 high = crash_base ? crash_base + crash_size : ULONG_MAX; 629 630 if (crash_base && crash_base < low) { 631 pr_info("crashkernel reservation failed: %s\n", 632 "crash_base too low"); 633 return; 634 } 635 low = crash_base ?: low; 636 crash_base = memblock_phys_alloc_range(crash_size, 637 KEXEC_CRASH_MEM_ALIGN, 638 low, high); 639 } 640 641 if (!crash_base) { 642 pr_info("crashkernel reservation failed: %s\n", 643 "no suitable area found"); 644 return; 645 } 646 647 if (register_memory_notifier(&kdump_mem_nb)) { 648 memblock_phys_free(crash_base, crash_size); 649 return; 650 } 651 652 if (!oldmem_data.start && machine_is_vm()) 653 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 654 crashk_res.start = crash_base; 655 crashk_res.end = crash_base + crash_size - 1; 656 memblock_remove(crash_base, crash_size); 657 pr_info("Reserving %lluMB of memory at %lluMB " 658 "for crashkernel (System RAM: %luMB)\n", 659 crash_size >> 20, crash_base >> 20, 660 (unsigned long)memblock.memory.total_size >> 20); 661 os_info_crashkernel_add(crash_base, crash_size); 662 #endif 663 } 664 665 /* 666 * Reserve the initrd from being used by memblock 667 */ 668 static void __init reserve_initrd(void) 669 { 670 unsigned long addr, size; 671 672 if (!IS_ENABLED(CONFIG_BLK_DEV_INITRD) || !get_physmem_reserved(RR_INITRD, &addr, &size)) 673 return; 674 initrd_start = (unsigned long)__va(addr); 675 initrd_end = initrd_start + size; 676 memblock_reserve(addr, size); 677 } 678 679 /* 680 * Reserve the memory area used to pass the certificate lists 681 */ 682 static void __init reserve_certificate_list(void) 683 { 684 if (ipl_cert_list_addr) 685 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size); 686 } 687 688 static void __init reserve_physmem_info(void) 689 { 690 unsigned long addr, size; 691 692 if (get_physmem_reserved(RR_MEM_DETECT_EXT, &addr, &size)) 693 memblock_reserve(addr, size); 694 } 695 696 static void __init free_physmem_info(void) 697 { 698 unsigned long addr, size; 699 700 if (get_physmem_reserved(RR_MEM_DETECT_EXT, &addr, &size)) 701 memblock_phys_free(addr, size); 702 } 703 704 static void __init memblock_add_physmem_info(void) 705 { 706 unsigned long start, end; 707 int i; 708 709 pr_debug("physmem info source: %s (%hhd)\n", 710 get_physmem_info_source(), physmem_info.info_source); 711 /* keep memblock lists close to the kernel */ 712 memblock_set_bottom_up(true); 713 for_each_physmem_usable_range(i, &start, &end) 714 memblock_add(start, end - start); 715 for_each_physmem_online_range(i, &start, &end) 716 memblock_physmem_add(start, end - start); 717 memblock_set_bottom_up(false); 718 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0); 719 } 720 721 static void __init setup_high_memory(void) 722 { 723 high_memory = __va(ident_map_size); 724 } 725 726 /* 727 * Reserve memory used for lowcore. 728 */ 729 static void __init reserve_lowcore(void) 730 { 731 void *lowcore_start = get_lowcore(); 732 void *lowcore_end = lowcore_start + sizeof(struct lowcore); 733 void *start, *end; 734 735 if (absolute_pointer(__identity_base) < lowcore_end) { 736 start = max(lowcore_start, (void *)__identity_base); 737 end = min(lowcore_end, (void *)(__identity_base + ident_map_size)); 738 memblock_reserve(__pa(start), __pa(end)); 739 } 740 } 741 742 /* 743 * Reserve memory used for absolute lowcore/command line/kernel image. 744 */ 745 static void __init reserve_kernel(void) 746 { 747 memblock_reserve(0, STARTUP_NORMAL_OFFSET); 748 memblock_reserve(OLDMEM_BASE, sizeof(unsigned long)); 749 memblock_reserve(OLDMEM_SIZE, sizeof(unsigned long)); 750 memblock_reserve(physmem_info.reserved[RR_AMODE31].start, __eamode31 - __samode31); 751 memblock_reserve(__pa(sclp_early_sccb), EXT_SCCB_READ_SCP); 752 memblock_reserve(__pa(_stext), _end - _stext); 753 } 754 755 static void __init setup_memory(void) 756 { 757 phys_addr_t start, end; 758 u64 i; 759 760 /* 761 * Init storage key for present memory 762 */ 763 for_each_mem_range(i, &start, &end) 764 storage_key_init_range(start, end); 765 766 psw_set_key(PAGE_DEFAULT_KEY); 767 } 768 769 static void __init relocate_amode31_section(void) 770 { 771 unsigned long amode31_size = __eamode31 - __samode31; 772 long amode31_offset, *ptr; 773 774 amode31_offset = AMODE31_START - (unsigned long)__samode31; 775 pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size); 776 777 /* Move original AMODE31 section to the new one */ 778 memmove((void *)physmem_info.reserved[RR_AMODE31].start, __samode31, amode31_size); 779 /* Zero out the old AMODE31 section to catch invalid accesses within it */ 780 memset(__samode31, 0, amode31_size); 781 782 /* Update all AMODE31 region references */ 783 for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++) 784 *ptr += amode31_offset; 785 } 786 787 /* This must be called after AMODE31 relocation */ 788 static void __init setup_cr(void) 789 { 790 union ctlreg2 cr2; 791 union ctlreg5 cr5; 792 union ctlreg15 cr15; 793 794 __ctl_duct[1] = (unsigned long)__ctl_aste; 795 __ctl_duct[2] = (unsigned long)__ctl_aste; 796 __ctl_duct[4] = (unsigned long)__ctl_duald; 797 798 /* Update control registers CR2, CR5 and CR15 */ 799 local_ctl_store(2, &cr2.reg); 800 local_ctl_store(5, &cr5.reg); 801 local_ctl_store(15, &cr15.reg); 802 cr2.ducto = (unsigned long)__ctl_duct >> 6; 803 cr5.pasteo = (unsigned long)__ctl_duct >> 6; 804 cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3; 805 system_ctl_load(2, &cr2.reg); 806 system_ctl_load(5, &cr5.reg); 807 system_ctl_load(15, &cr15.reg); 808 } 809 810 /* 811 * Add system information as device randomness 812 */ 813 static void __init setup_randomness(void) 814 { 815 struct sysinfo_3_2_2 *vmms; 816 817 vmms = memblock_alloc_or_panic(PAGE_SIZE, PAGE_SIZE); 818 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count) 819 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count); 820 memblock_free(vmms, PAGE_SIZE); 821 822 if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG)) 823 static_branch_enable(&s390_arch_random_available); 824 } 825 826 /* 827 * Issue diagnose 318 to set the control program name and 828 * version codes. 829 */ 830 static void __init setup_control_program_code(void) 831 { 832 union diag318_info diag318_info = { 833 .cpnc = CPNC_LINUX, 834 .cpvc = 0, 835 }; 836 837 if (!sclp.has_diag318) 838 return; 839 840 diag_stat_inc(DIAG_STAT_X318); 841 asm volatile("diag %0,0,0x318" : : "d" (diag318_info.val)); 842 } 843 844 /* 845 * Print the component list from the IPL report 846 */ 847 static void __init log_component_list(void) 848 { 849 struct ipl_rb_component_entry *ptr, *end; 850 char *str; 851 852 if (!early_ipl_comp_list_addr) 853 return; 854 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL) 855 pr_info("Linux is running with Secure-IPL enabled\n"); 856 else 857 pr_info("Linux is running with Secure-IPL disabled\n"); 858 ptr = __va(early_ipl_comp_list_addr); 859 end = (void *) ptr + early_ipl_comp_list_size; 860 pr_info("The IPL report contains the following components:\n"); 861 while (ptr < end) { 862 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) { 863 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED) 864 str = "signed, verified"; 865 else 866 str = "signed, verification failed"; 867 } else { 868 str = "not signed"; 869 } 870 pr_info("%016llx - %016llx (%s)\n", 871 ptr->addr, ptr->addr + ptr->len, str); 872 ptr++; 873 } 874 } 875 876 /* 877 * Print avoiding interpretation of % in buf and taking bootdebug option 878 * into consideration. 879 */ 880 static void __init print_rb_entry(const char *buf) 881 { 882 char fmt[] = KERN_SOH "0boot: %s"; 883 int level = printk_get_level(buf); 884 885 buf = skip_timestamp(printk_skip_level(buf)); 886 if (level == KERN_DEBUG[1] && (!bootdebug || !bootdebug_filter_match(buf))) 887 return; 888 889 fmt[1] = level; 890 printk(fmt, buf); 891 } 892 893 /* 894 * Setup function called from init/main.c just after the banner 895 * was printed. 896 */ 897 898 void __init setup_arch(char **cmdline_p) 899 { 900 /* 901 * print what head.S has found out about the machine 902 */ 903 if (machine_is_vm()) 904 pr_info("Linux is running as a z/VM " 905 "guest operating system in 64-bit mode\n"); 906 else if (machine_is_kvm()) 907 pr_info("Linux is running under KVM in 64-bit mode\n"); 908 else if (machine_is_lpar()) 909 pr_info("Linux is running natively in 64-bit mode\n"); 910 else 911 pr_info("Linux is running as a guest in 64-bit mode\n"); 912 /* Print decompressor messages if not already printed */ 913 if (!boot_earlyprintk) 914 boot_rb_foreach(print_rb_entry); 915 916 if (machine_has_relocated_lowcore()) 917 pr_info("Lowcore relocated to 0x%px\n", get_lowcore()); 918 919 log_component_list(); 920 921 /* Have one command line that is parsed and saved in /proc/cmdline */ 922 /* boot_command_line has been already set up in early.c */ 923 *cmdline_p = boot_command_line; 924 925 ROOT_DEV = Root_RAM0; 926 927 setup_initial_init_mm(_text, _etext, _edata, _end); 928 929 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO)) 930 nospec_auto_detect(); 931 932 jump_label_init(); 933 parse_early_param(); 934 #ifdef CONFIG_CRASH_DUMP 935 /* Deactivate elfcorehdr= kernel parameter */ 936 elfcorehdr_addr = ELFCORE_ADDR_MAX; 937 #endif 938 939 os_info_init(); 940 setup_ipl(); 941 setup_control_program_code(); 942 943 /* Do some memory reservations *before* memory is added to memblock */ 944 reserve_pgtables(); 945 reserve_lowcore(); 946 reserve_kernel(); 947 reserve_initrd(); 948 reserve_certificate_list(); 949 reserve_physmem_info(); 950 memblock_set_current_limit(ident_map_size); 951 memblock_allow_resize(); 952 953 /* Get information about *all* installed memory */ 954 memblock_add_physmem_info(); 955 956 free_physmem_info(); 957 setup_memory_end(); 958 setup_high_memory(); 959 memblock_dump_all(); 960 setup_memory(); 961 962 relocate_amode31_section(); 963 setup_cr(); 964 setup_uv(); 965 dma_contiguous_reserve(ident_map_size); 966 vmcp_cma_reserve(); 967 if (cpu_has_edat2()) 968 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT); 969 970 reserve_crashkernel(); 971 #ifdef CONFIG_CRASH_DUMP 972 /* 973 * Be aware that smp_save_dump_secondary_cpus() triggers a system reset. 974 * Therefore CPU and device initialization should be done afterwards. 975 */ 976 smp_save_dump_secondary_cpus(); 977 #endif 978 979 setup_resources(); 980 setup_lowcore(); 981 smp_fill_possible_mask(); 982 cpu_detect_mhz_feature(); 983 cpu_init(); 984 numa_setup(); 985 smp_detect_cpus(); 986 topology_init_early(); 987 setup_protection_map(); 988 /* 989 * Create kernel page tables. 990 */ 991 paging_init(); 992 993 /* 994 * After paging_init created the kernel page table, the new PSWs 995 * in lowcore can now run with DAT enabled. 996 */ 997 #ifdef CONFIG_CRASH_DUMP 998 smp_save_dump_ipl_cpu(); 999 #endif 1000 1001 /* Setup default console */ 1002 conmode_default(); 1003 set_preferred_console(); 1004 1005 apply_alternative_instructions(); 1006 if (IS_ENABLED(CONFIG_EXPOLINE)) 1007 nospec_init_branches(); 1008 1009 /* Setup zfcp/nvme dump support */ 1010 setup_zfcpdump(); 1011 1012 /* Add system specific data to the random pool */ 1013 setup_randomness(); 1014 } 1015 1016 void __init arch_cpu_finalize_init(void) 1017 { 1018 sclp_init(); 1019 } 1020