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 #include <linux/kmemleak.h> 54 55 #include <asm/boot_data.h> 56 #include <asm/ipl.h> 57 #include <asm/facility.h> 58 #include <asm/smp.h> 59 #include <asm/mmu_context.h> 60 #include <asm/cpcmd.h> 61 #include <asm/lowcore.h> 62 #include <asm/nmi.h> 63 #include <asm/irq.h> 64 #include <asm/page.h> 65 #include <asm/ptrace.h> 66 #include <asm/sections.h> 67 #include <asm/ebcdic.h> 68 #include <asm/diag.h> 69 #include <asm/os_info.h> 70 #include <asm/sclp.h> 71 #include <asm/stacktrace.h> 72 #include <asm/sysinfo.h> 73 #include <asm/numa.h> 74 #include <asm/alternative.h> 75 #include <asm/nospec-branch.h> 76 #include <asm/mem_detect.h> 77 #include <asm/uv.h> 78 #include <asm/asm-offsets.h> 79 #include "entry.h" 80 81 /* 82 * Machine setup.. 83 */ 84 unsigned int console_mode = 0; 85 EXPORT_SYMBOL(console_mode); 86 87 unsigned int console_devno = -1; 88 EXPORT_SYMBOL(console_devno); 89 90 unsigned int console_irq = -1; 91 EXPORT_SYMBOL(console_irq); 92 93 /* 94 * Some code and data needs to stay below 2 GB, even when the kernel would be 95 * relocated above 2 GB, because it has to use 31 bit addresses. 96 * Such code and data is part of the .amode31 section. 97 */ 98 unsigned long __amode31_ref __samode31 = (unsigned long)&_samode31; 99 unsigned long __amode31_ref __eamode31 = (unsigned long)&_eamode31; 100 unsigned long __amode31_ref __stext_amode31 = (unsigned long)&_stext_amode31; 101 unsigned long __amode31_ref __etext_amode31 = (unsigned long)&_etext_amode31; 102 struct exception_table_entry __amode31_ref *__start_amode31_ex_table = _start_amode31_ex_table; 103 struct exception_table_entry __amode31_ref *__stop_amode31_ex_table = _stop_amode31_ex_table; 104 105 /* 106 * Control registers CR2, CR5 and CR15 are initialized with addresses 107 * of tables that must be placed below 2G which is handled by the AMODE31 108 * sections. 109 * Because the AMODE31 sections are relocated below 2G at startup, 110 * the content of control registers CR2, CR5 and CR15 must be updated 111 * with new addresses after the relocation. The initial initialization of 112 * control registers occurs in head64.S and then gets updated again after AMODE31 113 * relocation. We must access the relevant AMODE31 tables indirectly via 114 * pointers placed in the .amode31.refs linker section. Those pointers get 115 * updated automatically during AMODE31 relocation and always contain a valid 116 * address within AMODE31 sections. 117 */ 118 119 static __amode31_data u32 __ctl_duct_amode31[16] __aligned(64); 120 121 static __amode31_data u64 __ctl_aste_amode31[8] __aligned(64) = { 122 [1] = 0xffffffffffffffff 123 }; 124 125 static __amode31_data u32 __ctl_duald_amode31[32] __aligned(128) = { 126 0x80000000, 0, 0, 0, 127 0x80000000, 0, 0, 0, 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 }; 135 136 static __amode31_data u32 __ctl_linkage_stack_amode31[8] __aligned(64) = { 137 0, 0, 0x89000000, 0, 138 0, 0, 0x8a000000, 0 139 }; 140 141 static u64 __amode31_ref *__ctl_aste = __ctl_aste_amode31; 142 static u32 __amode31_ref *__ctl_duald = __ctl_duald_amode31; 143 static u32 __amode31_ref *__ctl_linkage_stack = __ctl_linkage_stack_amode31; 144 static u32 __amode31_ref *__ctl_duct = __ctl_duct_amode31; 145 146 int __bootdata(noexec_disabled); 147 unsigned long __bootdata(ident_map_size); 148 struct mem_detect_info __bootdata(mem_detect); 149 struct initrd_data __bootdata(initrd_data); 150 151 unsigned long __bootdata_preserved(__kaslr_offset); 152 unsigned long __bootdata(__amode31_base); 153 unsigned int __bootdata_preserved(zlib_dfltcc_support); 154 EXPORT_SYMBOL(zlib_dfltcc_support); 155 u64 __bootdata_preserved(stfle_fac_list[16]); 156 EXPORT_SYMBOL(stfle_fac_list); 157 u64 __bootdata_preserved(alt_stfle_fac_list[16]); 158 struct oldmem_data __bootdata_preserved(oldmem_data); 159 160 unsigned long VMALLOC_START; 161 EXPORT_SYMBOL(VMALLOC_START); 162 163 unsigned long VMALLOC_END; 164 EXPORT_SYMBOL(VMALLOC_END); 165 166 struct page *vmemmap; 167 EXPORT_SYMBOL(vmemmap); 168 unsigned long vmemmap_size; 169 170 unsigned long MODULES_VADDR; 171 unsigned long MODULES_END; 172 173 /* An array with a pointer to the lowcore of every CPU. */ 174 struct lowcore *lowcore_ptr[NR_CPUS]; 175 EXPORT_SYMBOL(lowcore_ptr); 176 177 DEFINE_STATIC_KEY_FALSE(cpu_has_bear); 178 179 /* 180 * The Write Back bit position in the physaddr is given by the SLPC PCI. 181 * Leaving the mask zero always uses write through which is safe 182 */ 183 unsigned long mio_wb_bit_mask __ro_after_init; 184 185 /* 186 * This is set up by the setup-routine at boot-time 187 * for S390 need to find out, what we have to setup 188 * using address 0x10400 ... 189 */ 190 191 #include <asm/setup.h> 192 193 /* 194 * condev= and conmode= setup parameter. 195 */ 196 197 static int __init condev_setup(char *str) 198 { 199 int vdev; 200 201 vdev = simple_strtoul(str, &str, 0); 202 if (vdev >= 0 && vdev < 65536) { 203 console_devno = vdev; 204 console_irq = -1; 205 } 206 return 1; 207 } 208 209 __setup("condev=", condev_setup); 210 211 static void __init set_preferred_console(void) 212 { 213 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP) 214 add_preferred_console("ttyS", 0, NULL); 215 else if (CONSOLE_IS_3270) 216 add_preferred_console("tty3270", 0, NULL); 217 else if (CONSOLE_IS_VT220) 218 add_preferred_console("ttysclp", 0, NULL); 219 else if (CONSOLE_IS_HVC) 220 add_preferred_console("hvc", 0, NULL); 221 } 222 223 static int __init conmode_setup(char *str) 224 { 225 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 226 if (!strcmp(str, "hwc") || !strcmp(str, "sclp")) 227 SET_CONSOLE_SCLP; 228 #endif 229 #if defined(CONFIG_TN3215_CONSOLE) 230 if (!strcmp(str, "3215")) 231 SET_CONSOLE_3215; 232 #endif 233 #if defined(CONFIG_TN3270_CONSOLE) 234 if (!strcmp(str, "3270")) 235 SET_CONSOLE_3270; 236 #endif 237 set_preferred_console(); 238 return 1; 239 } 240 241 __setup("conmode=", conmode_setup); 242 243 static void __init conmode_default(void) 244 { 245 char query_buffer[1024]; 246 char *ptr; 247 248 if (MACHINE_IS_VM) { 249 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL); 250 console_devno = simple_strtoul(query_buffer + 5, NULL, 16); 251 ptr = strstr(query_buffer, "SUBCHANNEL ="); 252 console_irq = simple_strtoul(ptr + 13, NULL, 16); 253 cpcmd("QUERY TERM", query_buffer, 1024, NULL); 254 ptr = strstr(query_buffer, "CONMODE"); 255 /* 256 * Set the conmode to 3215 so that the device recognition 257 * will set the cu_type of the console to 3215. If the 258 * conmode is 3270 and we don't set it back then both 259 * 3215 and the 3270 driver will try to access the console 260 * device (3215 as console and 3270 as normal tty). 261 */ 262 cpcmd("TERM CONMODE 3215", NULL, 0, NULL); 263 if (ptr == NULL) { 264 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 265 SET_CONSOLE_SCLP; 266 #endif 267 return; 268 } 269 if (str_has_prefix(ptr + 8, "3270")) { 270 #if defined(CONFIG_TN3270_CONSOLE) 271 SET_CONSOLE_3270; 272 #elif defined(CONFIG_TN3215_CONSOLE) 273 SET_CONSOLE_3215; 274 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 275 SET_CONSOLE_SCLP; 276 #endif 277 } else if (str_has_prefix(ptr + 8, "3215")) { 278 #if defined(CONFIG_TN3215_CONSOLE) 279 SET_CONSOLE_3215; 280 #elif defined(CONFIG_TN3270_CONSOLE) 281 SET_CONSOLE_3270; 282 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 283 SET_CONSOLE_SCLP; 284 #endif 285 } 286 } else if (MACHINE_IS_KVM) { 287 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE)) 288 SET_CONSOLE_VT220; 289 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE)) 290 SET_CONSOLE_SCLP; 291 else 292 SET_CONSOLE_HVC; 293 } else { 294 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE) 295 SET_CONSOLE_SCLP; 296 #endif 297 } 298 } 299 300 #ifdef CONFIG_CRASH_DUMP 301 static void __init setup_zfcpdump(void) 302 { 303 if (!is_ipl_type_dump()) 304 return; 305 if (oldmem_data.start) 306 return; 307 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev"); 308 console_loglevel = 2; 309 } 310 #else 311 static inline void setup_zfcpdump(void) {} 312 #endif /* CONFIG_CRASH_DUMP */ 313 314 /* 315 * Reboot, halt and power_off stubs. They just call _machine_restart, 316 * _machine_halt or _machine_power_off. 317 */ 318 319 void machine_restart(char *command) 320 { 321 if ((!in_interrupt() && !in_atomic()) || oops_in_progress) 322 /* 323 * Only unblank the console if we are called in enabled 324 * context or a bust_spinlocks cleared the way for us. 325 */ 326 console_unblank(); 327 _machine_restart(command); 328 } 329 330 void machine_halt(void) 331 { 332 if (!in_interrupt() || oops_in_progress) 333 /* 334 * Only unblank the console if we are called in enabled 335 * context or a bust_spinlocks cleared the way for us. 336 */ 337 console_unblank(); 338 _machine_halt(); 339 } 340 341 void machine_power_off(void) 342 { 343 if (!in_interrupt() || oops_in_progress) 344 /* 345 * Only unblank the console if we are called in enabled 346 * context or a bust_spinlocks cleared the way for us. 347 */ 348 console_unblank(); 349 _machine_power_off(); 350 } 351 352 /* 353 * Dummy power off function. 354 */ 355 void (*pm_power_off)(void) = machine_power_off; 356 EXPORT_SYMBOL_GPL(pm_power_off); 357 358 void *restart_stack; 359 360 unsigned long stack_alloc(void) 361 { 362 #ifdef CONFIG_VMAP_STACK 363 void *ret; 364 365 ret = __vmalloc_node(THREAD_SIZE, THREAD_SIZE, THREADINFO_GFP, 366 NUMA_NO_NODE, __builtin_return_address(0)); 367 kmemleak_not_leak(ret); 368 return (unsigned long)ret; 369 #else 370 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 371 #endif 372 } 373 374 void stack_free(unsigned long stack) 375 { 376 #ifdef CONFIG_VMAP_STACK 377 vfree((void *) stack); 378 #else 379 free_pages(stack, THREAD_SIZE_ORDER); 380 #endif 381 } 382 383 int __init arch_early_irq_init(void) 384 { 385 unsigned long stack; 386 387 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER); 388 if (!stack) 389 panic("Couldn't allocate async stack"); 390 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET; 391 return 0; 392 } 393 394 void __init arch_call_rest_init(void) 395 { 396 unsigned long stack; 397 398 stack = stack_alloc(); 399 if (!stack) 400 panic("Couldn't allocate kernel stack"); 401 current->stack = (void *) stack; 402 #ifdef CONFIG_VMAP_STACK 403 current->stack_vm_area = (void *) stack; 404 #endif 405 set_task_stack_end_magic(current); 406 stack += STACK_INIT_OFFSET; 407 S390_lowcore.kernel_stack = stack; 408 call_on_stack_noreturn(rest_init, stack); 409 } 410 411 static void __init setup_lowcore_dat_off(void) 412 { 413 unsigned long int_psw_mask = PSW_KERNEL_BITS; 414 unsigned long mcck_stack; 415 struct lowcore *lc; 416 417 if (IS_ENABLED(CONFIG_KASAN)) 418 int_psw_mask |= PSW_MASK_DAT; 419 420 /* 421 * Setup lowcore for boot cpu 422 */ 423 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE); 424 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc)); 425 if (!lc) 426 panic("%s: Failed to allocate %zu bytes align=%zx\n", 427 __func__, sizeof(*lc), sizeof(*lc)); 428 429 lc->restart_psw.mask = PSW_KERNEL_BITS; 430 lc->restart_psw.addr = (unsigned long) restart_int_handler; 431 lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK; 432 lc->external_new_psw.addr = (unsigned long) ext_int_handler; 433 lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK; 434 lc->svc_new_psw.addr = (unsigned long) system_call; 435 lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK; 436 lc->program_new_psw.addr = (unsigned long) pgm_check_handler; 437 lc->mcck_new_psw.mask = PSW_KERNEL_BITS; 438 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler; 439 lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK; 440 lc->io_new_psw.addr = (unsigned long) io_int_handler; 441 lc->clock_comparator = clock_comparator_max; 442 lc->nodat_stack = ((unsigned long) &init_thread_union) 443 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); 444 lc->current_task = (unsigned long)&init_task; 445 lc->lpp = LPP_MAGIC; 446 lc->machine_flags = S390_lowcore.machine_flags; 447 lc->preempt_count = S390_lowcore.preempt_count; 448 nmi_alloc_mcesa_early(&lc->mcesad); 449 lc->sys_enter_timer = S390_lowcore.sys_enter_timer; 450 lc->exit_timer = S390_lowcore.exit_timer; 451 lc->user_timer = S390_lowcore.user_timer; 452 lc->system_timer = S390_lowcore.system_timer; 453 lc->steal_timer = S390_lowcore.steal_timer; 454 lc->last_update_timer = S390_lowcore.last_update_timer; 455 lc->last_update_clock = S390_lowcore.last_update_clock; 456 457 /* 458 * Allocate the global restart stack which is the same for 459 * all CPUs in cast *one* of them does a PSW restart. 460 */ 461 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE); 462 if (!restart_stack) 463 panic("%s: Failed to allocate %lu bytes align=0x%lx\n", 464 __func__, THREAD_SIZE, THREAD_SIZE); 465 restart_stack += STACK_INIT_OFFSET; 466 467 /* 468 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant 469 * restart data to the absolute zero lowcore. This is necessary if 470 * PSW restart is done on an offline CPU that has lowcore zero. 471 */ 472 lc->restart_stack = (unsigned long) restart_stack; 473 lc->restart_fn = (unsigned long) do_restart; 474 lc->restart_data = 0; 475 lc->restart_source = -1U; 476 477 put_abs_lowcore(restart_stack, lc->restart_stack); 478 put_abs_lowcore(restart_fn, lc->restart_fn); 479 put_abs_lowcore(restart_data, lc->restart_data); 480 put_abs_lowcore(restart_source, lc->restart_source); 481 put_abs_lowcore(restart_psw, lc->restart_psw); 482 483 mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE); 484 if (!mcck_stack) 485 panic("%s: Failed to allocate %lu bytes align=0x%lx\n", 486 __func__, THREAD_SIZE, THREAD_SIZE); 487 lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET; 488 489 lc->spinlock_lockval = arch_spin_lockval(0); 490 lc->spinlock_index = 0; 491 arch_spin_lock_setup(0); 492 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW); 493 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW); 494 lc->preempt_count = PREEMPT_DISABLED; 495 496 set_prefix(__pa(lc)); 497 lowcore_ptr[0] = lc; 498 } 499 500 static void __init setup_lowcore_dat_on(void) 501 { 502 struct lowcore *lc = lowcore_ptr[0]; 503 int cr; 504 505 __ctl_clear_bit(0, 28); 506 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT; 507 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT; 508 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT; 509 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT; 510 __ctl_store(S390_lowcore.cregs_save_area, 0, 15); 511 __ctl_set_bit(0, 28); 512 put_abs_lowcore(restart_flags, RESTART_FLAG_CTLREGS); 513 put_abs_lowcore(program_new_psw, lc->program_new_psw); 514 for (cr = 0; cr < ARRAY_SIZE(lc->cregs_save_area); cr++) 515 put_abs_lowcore(cregs_save_area[cr], lc->cregs_save_area[cr]); 516 } 517 518 static struct resource code_resource = { 519 .name = "Kernel code", 520 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 521 }; 522 523 static struct resource data_resource = { 524 .name = "Kernel data", 525 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 526 }; 527 528 static struct resource bss_resource = { 529 .name = "Kernel bss", 530 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM, 531 }; 532 533 static struct resource __initdata *standard_resources[] = { 534 &code_resource, 535 &data_resource, 536 &bss_resource, 537 }; 538 539 static void __init setup_resources(void) 540 { 541 struct resource *res, *std_res, *sub_res; 542 phys_addr_t start, end; 543 int j; 544 u64 i; 545 546 code_resource.start = (unsigned long) _text; 547 code_resource.end = (unsigned long) _etext - 1; 548 data_resource.start = (unsigned long) _etext; 549 data_resource.end = (unsigned long) _edata - 1; 550 bss_resource.start = (unsigned long) __bss_start; 551 bss_resource.end = (unsigned long) __bss_stop - 1; 552 553 for_each_mem_range(i, &start, &end) { 554 res = memblock_alloc(sizeof(*res), 8); 555 if (!res) 556 panic("%s: Failed to allocate %zu bytes align=0x%x\n", 557 __func__, sizeof(*res), 8); 558 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM; 559 560 res->name = "System RAM"; 561 res->start = start; 562 /* 563 * In memblock, end points to the first byte after the 564 * range while in resourses, end points to the last byte in 565 * the range. 566 */ 567 res->end = end - 1; 568 request_resource(&iomem_resource, res); 569 570 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) { 571 std_res = standard_resources[j]; 572 if (std_res->start < res->start || 573 std_res->start > res->end) 574 continue; 575 if (std_res->end > res->end) { 576 sub_res = memblock_alloc(sizeof(*sub_res), 8); 577 if (!sub_res) 578 panic("%s: Failed to allocate %zu bytes align=0x%x\n", 579 __func__, sizeof(*sub_res), 8); 580 *sub_res = *std_res; 581 sub_res->end = res->end; 582 std_res->start = res->end + 1; 583 request_resource(res, sub_res); 584 } else { 585 request_resource(res, std_res); 586 } 587 } 588 } 589 #ifdef CONFIG_CRASH_DUMP 590 /* 591 * Re-add removed crash kernel memory as reserved memory. This makes 592 * sure it will be mapped with the identity mapping and struct pages 593 * will be created, so it can be resized later on. 594 * However add it later since the crash kernel resource should not be 595 * part of the System RAM resource. 596 */ 597 if (crashk_res.end) { 598 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 599 0, MEMBLOCK_NONE); 600 memblock_reserve(crashk_res.start, resource_size(&crashk_res)); 601 insert_resource(&iomem_resource, &crashk_res); 602 } 603 #endif 604 } 605 606 static void __init setup_memory_end(void) 607 { 608 memblock_remove(ident_map_size, PHYS_ADDR_MAX - ident_map_size); 609 max_pfn = max_low_pfn = PFN_DOWN(ident_map_size); 610 pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20); 611 } 612 613 #ifdef CONFIG_CRASH_DUMP 614 615 /* 616 * When kdump is enabled, we have to ensure that no memory from the area 617 * [0 - crashkernel memory size] is set offline - it will be exchanged with 618 * the crashkernel memory region when kdump is triggered. The crashkernel 619 * memory region can never get offlined (pages are unmovable). 620 */ 621 static int kdump_mem_notifier(struct notifier_block *nb, 622 unsigned long action, void *data) 623 { 624 struct memory_notify *arg = data; 625 626 if (action != MEM_GOING_OFFLINE) 627 return NOTIFY_OK; 628 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res))) 629 return NOTIFY_BAD; 630 return NOTIFY_OK; 631 } 632 633 static struct notifier_block kdump_mem_nb = { 634 .notifier_call = kdump_mem_notifier, 635 }; 636 637 #endif 638 639 /* 640 * Reserve memory for kdump kernel to be loaded with kexec 641 */ 642 static void __init reserve_crashkernel(void) 643 { 644 #ifdef CONFIG_CRASH_DUMP 645 unsigned long long crash_base, crash_size; 646 phys_addr_t low, high; 647 int rc; 648 649 rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size, 650 &crash_base); 651 652 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN); 653 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN); 654 if (rc || crash_size == 0) 655 return; 656 657 if (memblock.memory.regions[0].size < crash_size) { 658 pr_info("crashkernel reservation failed: %s\n", 659 "first memory chunk must be at least crashkernel size"); 660 return; 661 } 662 663 low = crash_base ?: oldmem_data.start; 664 high = low + crash_size; 665 if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) { 666 /* The crashkernel fits into OLDMEM, reuse OLDMEM */ 667 crash_base = low; 668 } else { 669 /* Find suitable area in free memory */ 670 low = max_t(unsigned long, crash_size, sclp.hsa_size); 671 high = crash_base ? crash_base + crash_size : ULONG_MAX; 672 673 if (crash_base && crash_base < low) { 674 pr_info("crashkernel reservation failed: %s\n", 675 "crash_base too low"); 676 return; 677 } 678 low = crash_base ?: low; 679 crash_base = memblock_phys_alloc_range(crash_size, 680 KEXEC_CRASH_MEM_ALIGN, 681 low, high); 682 } 683 684 if (!crash_base) { 685 pr_info("crashkernel reservation failed: %s\n", 686 "no suitable area found"); 687 return; 688 } 689 690 if (register_memory_notifier(&kdump_mem_nb)) { 691 memblock_phys_free(crash_base, crash_size); 692 return; 693 } 694 695 if (!oldmem_data.start && MACHINE_IS_VM) 696 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size)); 697 crashk_res.start = crash_base; 698 crashk_res.end = crash_base + crash_size - 1; 699 memblock_remove(crash_base, crash_size); 700 pr_info("Reserving %lluMB of memory at %lluMB " 701 "for crashkernel (System RAM: %luMB)\n", 702 crash_size >> 20, crash_base >> 20, 703 (unsigned long)memblock.memory.total_size >> 20); 704 os_info_crashkernel_add(crash_base, crash_size); 705 #endif 706 } 707 708 /* 709 * Reserve the initrd from being used by memblock 710 */ 711 static void __init reserve_initrd(void) 712 { 713 #ifdef CONFIG_BLK_DEV_INITRD 714 if (!initrd_data.start || !initrd_data.size) 715 return; 716 initrd_start = (unsigned long)__va(initrd_data.start); 717 initrd_end = initrd_start + initrd_data.size; 718 memblock_reserve(initrd_data.start, initrd_data.size); 719 #endif 720 } 721 722 /* 723 * Reserve the memory area used to pass the certificate lists 724 */ 725 static void __init reserve_certificate_list(void) 726 { 727 if (ipl_cert_list_addr) 728 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size); 729 } 730 731 static void __init reserve_mem_detect_info(void) 732 { 733 unsigned long start, size; 734 735 get_mem_detect_reserved(&start, &size); 736 if (size) 737 memblock_reserve(start, size); 738 } 739 740 static void __init free_mem_detect_info(void) 741 { 742 unsigned long start, size; 743 744 get_mem_detect_reserved(&start, &size); 745 if (size) 746 memblock_phys_free(start, size); 747 } 748 749 static const char * __init get_mem_info_source(void) 750 { 751 switch (mem_detect.info_source) { 752 case MEM_DETECT_SCLP_STOR_INFO: 753 return "sclp storage info"; 754 case MEM_DETECT_DIAG260: 755 return "diag260"; 756 case MEM_DETECT_SCLP_READ_INFO: 757 return "sclp read info"; 758 case MEM_DETECT_BIN_SEARCH: 759 return "binary search"; 760 } 761 return "none"; 762 } 763 764 static void __init memblock_add_mem_detect_info(void) 765 { 766 unsigned long start, end; 767 int i; 768 769 pr_debug("physmem info source: %s (%hhd)\n", 770 get_mem_info_source(), mem_detect.info_source); 771 /* keep memblock lists close to the kernel */ 772 memblock_set_bottom_up(true); 773 for_each_mem_detect_block(i, &start, &end) { 774 memblock_add(start, end - start); 775 memblock_physmem_add(start, end - start); 776 } 777 memblock_set_bottom_up(false); 778 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0); 779 } 780 781 /* 782 * Check for initrd being in usable memory 783 */ 784 static void __init check_initrd(void) 785 { 786 #ifdef CONFIG_BLK_DEV_INITRD 787 if (initrd_data.start && initrd_data.size && 788 !memblock_is_region_memory(initrd_data.start, initrd_data.size)) { 789 pr_err("The initial RAM disk does not fit into the memory\n"); 790 memblock_phys_free(initrd_data.start, initrd_data.size); 791 initrd_start = initrd_end = 0; 792 } 793 #endif 794 } 795 796 /* 797 * Reserve memory used for lowcore/command line/kernel image. 798 */ 799 static void __init reserve_kernel(void) 800 { 801 memblock_reserve(0, STARTUP_NORMAL_OFFSET); 802 memblock_reserve(OLDMEM_BASE, sizeof(unsigned long)); 803 memblock_reserve(OLDMEM_SIZE, sizeof(unsigned long)); 804 memblock_reserve(__amode31_base, __eamode31 - __samode31); 805 memblock_reserve(__pa(sclp_early_sccb), EXT_SCCB_READ_SCP); 806 memblock_reserve(__pa(_stext), _end - _stext); 807 } 808 809 static void __init setup_memory(void) 810 { 811 phys_addr_t start, end; 812 u64 i; 813 814 /* 815 * Init storage key for present memory 816 */ 817 for_each_mem_range(i, &start, &end) 818 storage_key_init_range(start, end); 819 820 psw_set_key(PAGE_DEFAULT_KEY); 821 } 822 823 static void __init relocate_amode31_section(void) 824 { 825 unsigned long amode31_size = __eamode31 - __samode31; 826 long amode31_offset = __amode31_base - __samode31; 827 long *ptr; 828 829 pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size); 830 831 /* Move original AMODE31 section to the new one */ 832 memmove((void *)__amode31_base, (void *)__samode31, amode31_size); 833 /* Zero out the old AMODE31 section to catch invalid accesses within it */ 834 memset((void *)__samode31, 0, amode31_size); 835 836 /* Update all AMODE31 region references */ 837 for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++) 838 *ptr += amode31_offset; 839 } 840 841 /* This must be called after AMODE31 relocation */ 842 static void __init setup_cr(void) 843 { 844 union ctlreg2 cr2; 845 union ctlreg5 cr5; 846 union ctlreg15 cr15; 847 848 __ctl_duct[1] = (unsigned long)__ctl_aste; 849 __ctl_duct[2] = (unsigned long)__ctl_aste; 850 __ctl_duct[4] = (unsigned long)__ctl_duald; 851 852 /* Update control registers CR2, CR5 and CR15 */ 853 __ctl_store(cr2.val, 2, 2); 854 __ctl_store(cr5.val, 5, 5); 855 __ctl_store(cr15.val, 15, 15); 856 cr2.ducto = (unsigned long)__ctl_duct >> 6; 857 cr5.pasteo = (unsigned long)__ctl_duct >> 6; 858 cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3; 859 __ctl_load(cr2.val, 2, 2); 860 __ctl_load(cr5.val, 5, 5); 861 __ctl_load(cr15.val, 15, 15); 862 } 863 864 /* 865 * Add system information as device randomness 866 */ 867 static void __init setup_randomness(void) 868 { 869 struct sysinfo_3_2_2 *vmms; 870 871 vmms = memblock_alloc(PAGE_SIZE, PAGE_SIZE); 872 if (!vmms) 873 panic("Failed to allocate memory for sysinfo structure\n"); 874 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count) 875 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count); 876 memblock_free(vmms, PAGE_SIZE); 877 878 if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG)) 879 static_branch_enable(&s390_arch_random_available); 880 } 881 882 /* 883 * Find the correct size for the task_struct. This depends on 884 * the size of the struct fpu at the end of the thread_struct 885 * which is embedded in the task_struct. 886 */ 887 static void __init setup_task_size(void) 888 { 889 int task_size = sizeof(struct task_struct); 890 891 if (!MACHINE_HAS_VX) { 892 task_size -= sizeof(__vector128) * __NUM_VXRS; 893 task_size += sizeof(freg_t) * __NUM_FPRS; 894 } 895 arch_task_struct_size = task_size; 896 } 897 898 /* 899 * Issue diagnose 318 to set the control program name and 900 * version codes. 901 */ 902 static void __init setup_control_program_code(void) 903 { 904 union diag318_info diag318_info = { 905 .cpnc = CPNC_LINUX, 906 .cpvc = 0, 907 }; 908 909 if (!sclp.has_diag318) 910 return; 911 912 diag_stat_inc(DIAG_STAT_X318); 913 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val)); 914 } 915 916 /* 917 * Print the component list from the IPL report 918 */ 919 static void __init log_component_list(void) 920 { 921 struct ipl_rb_component_entry *ptr, *end; 922 char *str; 923 924 if (!early_ipl_comp_list_addr) 925 return; 926 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL) 927 pr_info("Linux is running with Secure-IPL enabled\n"); 928 else 929 pr_info("Linux is running with Secure-IPL disabled\n"); 930 ptr = (void *) early_ipl_comp_list_addr; 931 end = (void *) ptr + early_ipl_comp_list_size; 932 pr_info("The IPL report contains the following components:\n"); 933 while (ptr < end) { 934 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) { 935 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED) 936 str = "signed, verified"; 937 else 938 str = "signed, verification failed"; 939 } else { 940 str = "not signed"; 941 } 942 pr_info("%016llx - %016llx (%s)\n", 943 ptr->addr, ptr->addr + ptr->len, str); 944 ptr++; 945 } 946 } 947 948 /* 949 * Setup function called from init/main.c just after the banner 950 * was printed. 951 */ 952 953 void __init setup_arch(char **cmdline_p) 954 { 955 /* 956 * print what head.S has found out about the machine 957 */ 958 if (MACHINE_IS_VM) 959 pr_info("Linux is running as a z/VM " 960 "guest operating system in 64-bit mode\n"); 961 else if (MACHINE_IS_KVM) 962 pr_info("Linux is running under KVM in 64-bit mode\n"); 963 else if (MACHINE_IS_LPAR) 964 pr_info("Linux is running natively in 64-bit mode\n"); 965 else 966 pr_info("Linux is running as a guest in 64-bit mode\n"); 967 968 log_component_list(); 969 970 /* Have one command line that is parsed and saved in /proc/cmdline */ 971 /* boot_command_line has been already set up in early.c */ 972 *cmdline_p = boot_command_line; 973 974 ROOT_DEV = Root_RAM0; 975 976 setup_initial_init_mm(_text, _etext, _edata, _end); 977 978 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO)) 979 nospec_auto_detect(); 980 981 jump_label_init(); 982 parse_early_param(); 983 #ifdef CONFIG_CRASH_DUMP 984 /* Deactivate elfcorehdr= kernel parameter */ 985 elfcorehdr_addr = ELFCORE_ADDR_MAX; 986 #endif 987 988 os_info_init(); 989 setup_ipl(); 990 setup_task_size(); 991 setup_control_program_code(); 992 993 /* Do some memory reservations *before* memory is added to memblock */ 994 reserve_kernel(); 995 reserve_initrd(); 996 reserve_certificate_list(); 997 reserve_mem_detect_info(); 998 memblock_set_current_limit(ident_map_size); 999 memblock_allow_resize(); 1000 1001 /* Get information about *all* installed memory */ 1002 memblock_add_mem_detect_info(); 1003 1004 free_mem_detect_info(); 1005 setup_memory_end(); 1006 memblock_dump_all(); 1007 setup_memory(); 1008 1009 relocate_amode31_section(); 1010 setup_cr(); 1011 setup_uv(); 1012 dma_contiguous_reserve(ident_map_size); 1013 vmcp_cma_reserve(); 1014 if (MACHINE_HAS_EDAT2) 1015 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT); 1016 1017 check_initrd(); 1018 reserve_crashkernel(); 1019 #ifdef CONFIG_CRASH_DUMP 1020 /* 1021 * Be aware that smp_save_dump_cpus() triggers a system reset. 1022 * Therefore CPU and device initialization should be done afterwards. 1023 */ 1024 smp_save_dump_cpus(); 1025 #endif 1026 1027 setup_resources(); 1028 setup_lowcore_dat_off(); 1029 smp_fill_possible_mask(); 1030 cpu_detect_mhz_feature(); 1031 cpu_init(); 1032 numa_setup(); 1033 smp_detect_cpus(); 1034 topology_init_early(); 1035 1036 if (test_facility(193)) 1037 static_branch_enable(&cpu_has_bear); 1038 1039 /* 1040 * Create kernel page tables and switch to virtual addressing. 1041 */ 1042 paging_init(); 1043 1044 /* 1045 * After paging_init created the kernel page table, the new PSWs 1046 * in lowcore can now run with DAT enabled. 1047 */ 1048 setup_lowcore_dat_on(); 1049 1050 /* Setup default console */ 1051 conmode_default(); 1052 set_preferred_console(); 1053 1054 apply_alternative_instructions(); 1055 if (IS_ENABLED(CONFIG_EXPOLINE)) 1056 nospec_init_branches(); 1057 1058 /* Setup zfcp/nvme dump support */ 1059 setup_zfcpdump(); 1060 1061 /* Add system specific data to the random pool */ 1062 setup_randomness(); 1063 } 1064