www.infradead.org Git - users/jedix/linux-maple.git/blob

   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-contiguous.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
  53 #include <asm/boot_data.h>
  54 #include <asm/ipl.h>
  55 #include <asm/facility.h>
  56 #include <asm/smp.h>
  57 #include <asm/mmu_context.h>
  58 #include <asm/cpcmd.h>
  59 #include <asm/lowcore.h>
  60 #include <asm/nmi.h>
  61 #include <asm/irq.h>
  62 #include <asm/page.h>
  63 #include <asm/ptrace.h>
  64 #include <asm/sections.h>
  65 #include <asm/ebcdic.h>
  66 #include <asm/diag.h>
  67 #include <asm/os_info.h>
  68 #include <asm/sclp.h>
  69 #include <asm/stacktrace.h>
  70 #include <asm/sysinfo.h>
  71 #include <asm/numa.h>
  72 #include <asm/alternative.h>
  73 #include <asm/nospec-branch.h>
  74 #include <asm/mem_detect.h>
  75 #include <asm/uv.h>
  76 #include "entry.h"
  77
  78 /*
  79  * Machine setup..
  80  */
  81 unsigned int console_mode = 0;
  82 EXPORT_SYMBOL(console_mode);
  83
  84 unsigned int console_devno = -1;
  85 EXPORT_SYMBOL(console_devno);
  86
  87 unsigned int console_irq = -1;
  88 EXPORT_SYMBOL(console_irq);
  89
  90 unsigned long elf_hwcap __read_mostly = 0;
  91 char elf_platform[ELF_PLATFORM_SIZE];
  92
  93 unsigned long int_hwcap = 0;
  94
  95 int __bootdata(noexec_disabled);
  96 int __bootdata(memory_end_set);
  97 unsigned long __bootdata(memory_end);
  98 unsigned long __bootdata(vmalloc_size);
  99 unsigned long __bootdata(max_physmem_end);
 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(__swsusp_reset_dma);
 105 unsigned long __bootdata_preserved(__stext_dma);
 106 unsigned long __bootdata_preserved(__etext_dma);
 107 unsigned long __bootdata_preserved(__sdma);
 108 unsigned long __bootdata_preserved(__edma);
 109 unsigned long __bootdata_preserved(__kaslr_offset);
 110 unsigned int __bootdata_preserved(zlib_dfltcc_support);
 111 EXPORT_SYMBOL(zlib_dfltcc_support);
 112
 113 unsigned long VMALLOC_START;
 114 EXPORT_SYMBOL(VMALLOC_START);
 115
 116 unsigned long VMALLOC_END;
 117 EXPORT_SYMBOL(VMALLOC_END);
 118
 119 struct page *vmemmap;
 120 EXPORT_SYMBOL(vmemmap);
 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  * This is set up by the setup-routine at boot-time
 131  * for S390 need to find out, what we have to setup
 132  * using address 0x10400 ...
 133  */
 134
 135 #include <asm/setup.h>
 136
 137 /*
 138  * condev= and conmode= setup parameter.
 139  */
 140
 141 static int __init condev_setup(char *str)
 142 {
 143         int vdev;
 144
 145         vdev = simple_strtoul(str, &str, 0);
 146         if (vdev >= 0 && vdev < 65536) {
 147                 console_devno = vdev;
 148                 console_irq = -1;
 149         }
 150         return 1;
 151 }
 152
 153 __setup("condev=", condev_setup);
 154
 155 static void __init set_preferred_console(void)
 156 {
 157         if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 158                 add_preferred_console("ttyS", 0, NULL);
 159         else if (CONSOLE_IS_3270)
 160                 add_preferred_console("tty3270", 0, NULL);
 161         else if (CONSOLE_IS_VT220)
 162                 add_preferred_console("ttyS", 1, NULL);
 163         else if (CONSOLE_IS_HVC)
 164                 add_preferred_console("hvc", 0, NULL);
 165 }
 166
 167 static int __init conmode_setup(char *str)
 168 {
 169 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 170         if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
 171                 SET_CONSOLE_SCLP;
 172 #endif
 173 #if defined(CONFIG_TN3215_CONSOLE)
 174         if (!strcmp(str, "3215"))
 175                 SET_CONSOLE_3215;
 176 #endif
 177 #if defined(CONFIG_TN3270_CONSOLE)
 178         if (!strcmp(str, "3270"))
 179                 SET_CONSOLE_3270;
 180 #endif
 181         set_preferred_console();
 182         return 1;
 183 }
 184
 185 __setup("conmode=", conmode_setup);
 186
 187 static void __init conmode_default(void)
 188 {
 189         char query_buffer[1024];
 190         char *ptr;
 191
 192         if (MACHINE_IS_VM) {
 193                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 194                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 195                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 196                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 197                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 198                 ptr = strstr(query_buffer, "CONMODE");
 199                 /*
 200                  * Set the conmode to 3215 so that the device recognition
 201                  * will set the cu_type of the console to 3215. If the
 202                  * conmode is 3270 and we don't set it back then both
 203                  * 3215 and the 3270 driver will try to access the console
 204                  * device (3215 as console and 3270 as normal tty).
 205                  */
 206                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 207                 if (ptr == NULL) {
 208 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 209                         SET_CONSOLE_SCLP;
 210 #endif
 211                         return;
 212                 }
 213                 if (str_has_prefix(ptr + 8, "3270")) {
 214 #if defined(CONFIG_TN3270_CONSOLE)
 215                         SET_CONSOLE_3270;
 216 #elif defined(CONFIG_TN3215_CONSOLE)
 217                         SET_CONSOLE_3215;
 218 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 219                         SET_CONSOLE_SCLP;
 220 #endif
 221                 } else if (str_has_prefix(ptr + 8, "3215")) {
 222 #if defined(CONFIG_TN3215_CONSOLE)
 223                         SET_CONSOLE_3215;
 224 #elif defined(CONFIG_TN3270_CONSOLE)
 225                         SET_CONSOLE_3270;
 226 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 227                         SET_CONSOLE_SCLP;
 228 #endif
 229                 }
 230         } else if (MACHINE_IS_KVM) {
 231                 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
 232                         SET_CONSOLE_VT220;
 233                 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
 234                         SET_CONSOLE_SCLP;
 235                 else
 236                         SET_CONSOLE_HVC;
 237         } else {
 238 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 239                 SET_CONSOLE_SCLP;
 240 #endif
 241         }
 242 }
 243
 244 #ifdef CONFIG_CRASH_DUMP
 245 static void __init setup_zfcpdump(void)
 246 {
 247         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 248                 return;
 249         if (OLDMEM_BASE)
 250                 return;
 251         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 252         console_loglevel = 2;
 253 }
 254 #else
 255 static inline void setup_zfcpdump(void) {}
 256 #endif /* CONFIG_CRASH_DUMP */
 257
 258  /*
 259  * Reboot, halt and power_off stubs. They just call _machine_restart,
 260  * _machine_halt or _machine_power_off.
 261  */
 262
 263 void machine_restart(char *command)
 264 {
 265         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 266                 /*
 267                  * Only unblank the console if we are called in enabled
 268                  * context or a bust_spinlocks cleared the way for us.
 269                  */
 270                 console_unblank();
 271         _machine_restart(command);
 272 }
 273
 274 void machine_halt(void)
 275 {
 276         if (!in_interrupt() || oops_in_progress)
 277                 /*
 278                  * Only unblank the console if we are called in enabled
 279                  * context or a bust_spinlocks cleared the way for us.
 280                  */
 281                 console_unblank();
 282         _machine_halt();
 283 }
 284
 285 void machine_power_off(void)
 286 {
 287         if (!in_interrupt() || oops_in_progress)
 288                 /*
 289                  * Only unblank the console if we are called in enabled
 290                  * context or a bust_spinlocks cleared the way for us.
 291                  */
 292                 console_unblank();
 293         _machine_power_off();
 294 }
 295
 296 /*
 297  * Dummy power off function.
 298  */
 299 void (*pm_power_off)(void) = machine_power_off;
 300 EXPORT_SYMBOL_GPL(pm_power_off);
 301
 302 void *restart_stack __section(.data);
 303
 304 unsigned long stack_alloc(void)
 305 {
 306 #ifdef CONFIG_VMAP_STACK
 307         return (unsigned long)
 308                 __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
 309                                      VMALLOC_START, VMALLOC_END,
 310                                      THREADINFO_GFP,
 311                                      PAGE_KERNEL, 0, NUMA_NO_NODE,
 312                                      __builtin_return_address(0));
 313 #else
 314         return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 315 #endif
 316 }
 317
 318 void stack_free(unsigned long stack)
 319 {
 320 #ifdef CONFIG_VMAP_STACK
 321         vfree((void *) stack);
 322 #else
 323         free_pages(stack, THREAD_SIZE_ORDER);
 324 #endif
 325 }
 326
 327 int __init arch_early_irq_init(void)
 328 {
 329         unsigned long stack;
 330
 331         stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
 332         if (!stack)
 333                 panic("Couldn't allocate async stack");
 334         S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
 335         return 0;
 336 }
 337
 338 static int __init async_stack_realloc(void)
 339 {
 340         unsigned long old, new;
 341
 342         old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
 343         new = stack_alloc();
 344         if (!new)
 345                 panic("Couldn't allocate async stack");
 346         S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
 347         free_pages(old, THREAD_SIZE_ORDER);
 348         return 0;
 349 }
 350 early_initcall(async_stack_realloc);
 351
 352 void __init arch_call_rest_init(void)
 353 {
 354         unsigned long stack;
 355
 356         stack = stack_alloc();
 357         if (!stack)
 358                 panic("Couldn't allocate kernel stack");
 359         current->stack = (void *) stack;
 360 #ifdef CONFIG_VMAP_STACK
 361         current->stack_vm_area = (void *) stack;
 362 #endif
 363         set_task_stack_end_magic(current);
 364         stack += STACK_INIT_OFFSET;
 365         S390_lowcore.kernel_stack = stack;
 366         CALL_ON_STACK_NORETURN(rest_init, stack);
 367 }
 368
 369 static void __init setup_lowcore_dat_off(void)
 370 {
 371         struct lowcore *lc;
 372
 373         /*
 374          * Setup lowcore for boot cpu
 375          */
 376         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 377         lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
 378         if (!lc)
 379                 panic("%s: Failed to allocate %zu bytes align=%zx\n",
 380                       __func__, sizeof(*lc), sizeof(*lc));
 381
 382         lc->restart_psw.mask = PSW_KERNEL_BITS;
 383         lc->restart_psw.addr = (unsigned long) restart_int_handler;
 384         lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 385         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 386         lc->svc_new_psw.mask = PSW_KERNEL_BITS |
 387                 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 388         lc->svc_new_psw.addr = (unsigned long) system_call;
 389         lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 390         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 391         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 392         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 393         lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
 394         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 395         lc->clock_comparator = clock_comparator_max;
 396         lc->nodat_stack = ((unsigned long) &init_thread_union)
 397                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 398         lc->current_task = (unsigned long)&init_task;
 399         lc->lpp = LPP_MAGIC;
 400         lc->machine_flags = S390_lowcore.machine_flags;
 401         lc->preempt_count = S390_lowcore.preempt_count;
 402         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 403         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 404                sizeof(lc->stfle_fac_list));
 405         memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
 406                sizeof(lc->alt_stfle_fac_list));
 407         nmi_alloc_boot_cpu(lc);
 408         vdso_alloc_boot_cpu(lc);
 409         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 410         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 411         lc->exit_timer = S390_lowcore.exit_timer;
 412         lc->user_timer = S390_lowcore.user_timer;
 413         lc->system_timer = S390_lowcore.system_timer;
 414         lc->steal_timer = S390_lowcore.steal_timer;
 415         lc->last_update_timer = S390_lowcore.last_update_timer;
 416         lc->last_update_clock = S390_lowcore.last_update_clock;
 417
 418         /*
 419          * Allocate the global restart stack which is the same for
 420          * all CPUs in cast *one* of them does a PSW restart.
 421          */
 422         restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 423         if (!restart_stack)
 424                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 425                       __func__, THREAD_SIZE, THREAD_SIZE);
 426         restart_stack += STACK_INIT_OFFSET;
 427
 428         /*
 429          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 430          * restart data to the absolute zero lowcore. This is necessary if
 431          * PSW restart is done on an offline CPU that has lowcore zero.
 432          */
 433         lc->restart_stack = (unsigned long) restart_stack;
 434         lc->restart_fn = (unsigned long) do_restart;
 435         lc->restart_data = 0;
 436         lc->restart_source = -1UL;
 437
 438         /* Setup absolute zero lowcore */
 439         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 440         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 441         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 442         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 443         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 444
 445         lc->spinlock_lockval = arch_spin_lockval(0);
 446         lc->spinlock_index = 0;
 447         arch_spin_lock_setup(0);
 448         lc->br_r1_trampoline = 0x07f1;  /* br %r1 */
 449
 450         set_prefix((u32)(unsigned long) lc);
 451         lowcore_ptr[0] = lc;
 452 }
 453
 454 static void __init setup_lowcore_dat_on(void)
 455 {
 456         __ctl_clear_bit(0, 28);
 457         S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
 458         S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
 459         S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
 460         S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
 461         __ctl_set_bit(0, 28);
 462 }
 463
 464 static struct resource code_resource = {
 465         .name  = "Kernel code",
 466         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 467 };
 468
 469 static struct resource data_resource = {
 470         .name = "Kernel data",
 471         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 472 };
 473
 474 static struct resource bss_resource = {
 475         .name = "Kernel bss",
 476         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 477 };
 478
 479 static struct resource __initdata *standard_resources[] = {
 480         &code_resource,
 481         &data_resource,
 482         &bss_resource,
 483 };
 484
 485 static void __init setup_resources(void)
 486 {
 487         struct resource *res, *std_res, *sub_res;
 488         struct memblock_region *reg;
 489         int j;
 490
 491         code_resource.start = (unsigned long) _text;
 492         code_resource.end = (unsigned long) _etext - 1;
 493         data_resource.start = (unsigned long) _etext;
 494         data_resource.end = (unsigned long) _edata - 1;
 495         bss_resource.start = (unsigned long) __bss_start;
 496         bss_resource.end = (unsigned long) __bss_stop - 1;
 497
 498         for_each_memblock(memory, reg) {
 499                 res = memblock_alloc(sizeof(*res), 8);
 500                 if (!res)
 501                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 502                               __func__, sizeof(*res), 8);
 503                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 504
 505                 res->name = "System RAM";
 506                 res->start = reg->base;
 507                 res->end = reg->base + reg->size - 1;
 508                 request_resource(&iomem_resource, res);
 509
 510                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 511                         std_res = standard_resources[j];
 512                         if (std_res->start < res->start ||
 513                             std_res->start > res->end)
 514                                 continue;
 515                         if (std_res->end > res->end) {
 516                                 sub_res = memblock_alloc(sizeof(*sub_res), 8);
 517                                 if (!sub_res)
 518                                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 519                                               __func__, sizeof(*sub_res), 8);
 520                                 *sub_res = *std_res;
 521                                 sub_res->end = res->end;
 522                                 std_res->start = res->end + 1;
 523                                 request_resource(res, sub_res);
 524                         } else {
 525                                 request_resource(res, std_res);
 526                         }
 527                 }
 528         }
 529 #ifdef CONFIG_CRASH_DUMP
 530         /*
 531          * Re-add removed crash kernel memory as reserved memory. This makes
 532          * sure it will be mapped with the identity mapping and struct pages
 533          * will be created, so it can be resized later on.
 534          * However add it later since the crash kernel resource should not be
 535          * part of the System RAM resource.
 536          */
 537         if (crashk_res.end) {
 538                 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
 539                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 540                 insert_resource(&iomem_resource, &crashk_res);
 541         }
 542 #endif
 543 }
 544
 545 static void __init setup_memory_end(void)
 546 {
 547         unsigned long vmax, tmp;
 548
 549         /* Choose kernel address space layout: 3 or 4 levels. */
 550         if (IS_ENABLED(CONFIG_KASAN)) {
 551                 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
 552                            ? _REGION1_SIZE
 553                            : _REGION2_SIZE;
 554         } else {
 555                 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
 556                 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
 557                 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
 558                         vmax = _REGION2_SIZE; /* 3-level kernel page table */
 559                 else
 560                         vmax = _REGION1_SIZE; /* 4-level kernel page table */
 561         }
 562
 563         if (is_prot_virt_host())
 564                 adjust_to_uv_max(&vmax);
 565
 566         /* module area is at the end of the kernel address space. */
 567         MODULES_END = vmax;
 568         MODULES_VADDR = MODULES_END - MODULES_LEN;
 569         VMALLOC_END = MODULES_VADDR;
 570         VMALLOC_START = VMALLOC_END - vmalloc_size;
 571
 572         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 573         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 574         /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
 575         tmp = SECTION_ALIGN_UP(tmp);
 576         tmp = VMALLOC_START - tmp * sizeof(struct page);
 577         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 578         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 579         vmemmap = (struct page *) tmp;
 580
 581         /* Take care that memory_end is set and <= vmemmap */
 582         memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
 583 #ifdef CONFIG_KASAN
 584         /* fit in kasan shadow memory region between 1:1 and vmemmap */
 585         memory_end = min(memory_end, KASAN_SHADOW_START);
 586         vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
 587 #endif
 588         max_pfn = max_low_pfn = PFN_DOWN(memory_end);
 589         memblock_remove(memory_end, ULONG_MAX);
 590
 591         pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
 592 }
 593
 594 #ifdef CONFIG_CRASH_DUMP
 595
 596 /*
 597  * When kdump is enabled, we have to ensure that no memory from
 598  * the area [0 - crashkernel memory size] and
 599  * [crashk_res.start - crashk_res.end] is set offline.
 600  */
 601 static int kdump_mem_notifier(struct notifier_block *nb,
 602                               unsigned long action, void *data)
 603 {
 604         struct memory_notify *arg = data;
 605
 606         if (action != MEM_GOING_OFFLINE)
 607                 return NOTIFY_OK;
 608         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 609                 return NOTIFY_BAD;
 610         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 611                 return NOTIFY_OK;
 612         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 613                 return NOTIFY_OK;
 614         return NOTIFY_BAD;
 615 }
 616
 617 static struct notifier_block kdump_mem_nb = {
 618         .notifier_call = kdump_mem_notifier,
 619 };
 620
 621 #endif
 622
 623 /*
 624  * Make sure that the area behind memory_end is protected
 625  */
 626 static void reserve_memory_end(void)
 627 {
 628         if (memory_end_set)
 629                 memblock_reserve(memory_end, ULONG_MAX);
 630 }
 631
 632 /*
 633  * Make sure that oldmem, where the dump is stored, is protected
 634  */
 635 static void reserve_oldmem(void)
 636 {
 637 #ifdef CONFIG_CRASH_DUMP
 638         if (OLDMEM_BASE)
 639                 /* Forget all memory above the running kdump system */
 640                 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 641 #endif
 642 }
 643
 644 /*
 645  * Make sure that oldmem, where the dump is stored, is protected
 646  */
 647 static void remove_oldmem(void)
 648 {
 649 #ifdef CONFIG_CRASH_DUMP
 650         if (OLDMEM_BASE)
 651                 /* Forget all memory above the running kdump system */
 652                 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 653 #endif
 654 }
 655
 656 /*
 657  * Reserve memory for kdump kernel to be loaded with kexec
 658  */
 659 static void __init reserve_crashkernel(void)
 660 {
 661 #ifdef CONFIG_CRASH_DUMP
 662         unsigned long long crash_base, crash_size;
 663         phys_addr_t low, high;
 664         int rc;
 665
 666         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 667                                &crash_base);
 668
 669         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 670         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 671         if (rc || crash_size == 0)
 672                 return;
 673
 674         if (memblock.memory.regions[0].size < crash_size) {
 675                 pr_info("crashkernel reservation failed: %s\n",
 676                         "first memory chunk must be at least crashkernel size");
 677                 return;
 678         }
 679
 680         low = crash_base ?: OLDMEM_BASE;
 681         high = low + crash_size;
 682         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
 683                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 684                 crash_base = low;
 685         } else {
 686                 /* Find suitable area in free memory */
 687                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 688                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 689
 690                 if (crash_base && crash_base < low) {
 691                         pr_info("crashkernel reservation failed: %s\n",
 692                                 "crash_base too low");
 693                         return;
 694                 }
 695                 low = crash_base ?: low;
 696                 crash_base = memblock_find_in_range(low, high, crash_size,
 697                                                     KEXEC_CRASH_MEM_ALIGN);
 698         }
 699
 700         if (!crash_base) {
 701                 pr_info("crashkernel reservation failed: %s\n",
 702                         "no suitable area found");
 703                 return;
 704         }
 705
 706         if (register_memory_notifier(&kdump_mem_nb))
 707                 return;
 708
 709         if (!OLDMEM_BASE && MACHINE_IS_VM)
 710                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 711         crashk_res.start = crash_base;
 712         crashk_res.end = crash_base + crash_size - 1;
 713         memblock_remove(crash_base, crash_size);
 714         pr_info("Reserving %lluMB of memory at %lluMB "
 715                 "for crashkernel (System RAM: %luMB)\n",
 716                 crash_size >> 20, crash_base >> 20,
 717                 (unsigned long)memblock.memory.total_size >> 20);
 718         os_info_crashkernel_add(crash_base, crash_size);
 719 #endif
 720 }
 721
 722 /*
 723  * Reserve the initrd from being used by memblock
 724  */
 725 static void __init reserve_initrd(void)
 726 {
 727 #ifdef CONFIG_BLK_DEV_INITRD
 728         if (!INITRD_START || !INITRD_SIZE)
 729                 return;
 730         initrd_start = INITRD_START;
 731         initrd_end = initrd_start + INITRD_SIZE;
 732         memblock_reserve(INITRD_START, INITRD_SIZE);
 733 #endif
 734 }
 735
 736 /*
 737  * Reserve the memory area used to pass the certificate lists
 738  */
 739 static void __init reserve_certificate_list(void)
 740 {
 741         if (ipl_cert_list_addr)
 742                 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
 743 }
 744
 745 static void __init reserve_mem_detect_info(void)
 746 {
 747         unsigned long start, size;
 748
 749         get_mem_detect_reserved(&start, &size);
 750         if (size)
 751                 memblock_reserve(start, size);
 752 }
 753
 754 static void __init free_mem_detect_info(void)
 755 {
 756         unsigned long start, size;
 757
 758         get_mem_detect_reserved(&start, &size);
 759         if (size)
 760                 memblock_free(start, size);
 761 }
 762
 763 static const char * __init get_mem_info_source(void)
 764 {
 765         switch (mem_detect.info_source) {
 766         case MEM_DETECT_SCLP_STOR_INFO:
 767                 return "sclp storage info";
 768         case MEM_DETECT_DIAG260:
 769                 return "diag260";
 770         case MEM_DETECT_SCLP_READ_INFO:
 771                 return "sclp read info";
 772         case MEM_DETECT_BIN_SEARCH:
 773                 return "binary search";
 774         }
 775         return "none";
 776 }
 777
 778 static void __init memblock_add_mem_detect_info(void)
 779 {
 780         unsigned long start, end;
 781         int i;
 782
 783         memblock_dbg("physmem info source: %s (%hhd)\n",
 784                      get_mem_info_source(), mem_detect.info_source);
 785         /* keep memblock lists close to the kernel */
 786         memblock_set_bottom_up(true);
 787         for_each_mem_detect_block(i, &start, &end) {
 788                 memblock_add(start, end - start);
 789                 memblock_physmem_add(start, end - start);
 790         }
 791         memblock_set_bottom_up(false);
 792         memblock_dump_all();
 793 }
 794
 795 /*
 796  * Check for initrd being in usable memory
 797  */
 798 static void __init check_initrd(void)
 799 {
 800 #ifdef CONFIG_BLK_DEV_INITRD
 801         if (INITRD_START && INITRD_SIZE &&
 802             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
 803                 pr_err("The initial RAM disk does not fit into the memory\n");
 804                 memblock_free(INITRD_START, INITRD_SIZE);
 805                 initrd_start = initrd_end = 0;
 806         }
 807 #endif
 808 }
 809
 810 /*
 811  * Reserve memory used for lowcore/command line/kernel image.
 812  */
 813 static void __init reserve_kernel(void)
 814 {
 815         unsigned long start_pfn = PFN_UP(__pa(_end));
 816
 817         memblock_reserve(0, HEAD_END);
 818         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 819                          - (unsigned long)_stext);
 820         memblock_reserve(__sdma, __edma - __sdma);
 821 }
 822
 823 static void __init setup_memory(void)
 824 {
 825         struct memblock_region *reg;
 826
 827         /*
 828          * Init storage key for present memory
 829          */
 830         for_each_memblock(memory, reg) {
 831                 storage_key_init_range(reg->base, reg->base + reg->size);
 832         }
 833         psw_set_key(PAGE_DEFAULT_KEY);
 834
 835         /* Only cosmetics */
 836         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 837 }
 838
 839 /*
 840  * Setup hardware capabilities.
 841  */
 842 static int __init setup_hwcaps(void)
 843 {
 844         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 845         struct cpuid cpu_id;
 846         int i;
 847
 848         /*
 849          * The store facility list bits numbers as found in the principles
 850          * of operation are numbered with bit 1UL<<31 as number 0 to
 851          * bit 1UL<<0 as number 31.
 852          *   Bit 0: instructions named N3, "backported" to esa-mode
 853          *   Bit 2: z/Architecture mode is active
 854          *   Bit 7: the store-facility-list-extended facility is installed
 855          *   Bit 17: the message-security assist is installed
 856          *   Bit 19: the long-displacement facility is installed
 857          *   Bit 21: the extended-immediate facility is installed
 858          *   Bit 22: extended-translation facility 3 is installed
 859          *   Bit 30: extended-translation facility 3 enhancement facility
 860          * These get translated to:
 861          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 862          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 863          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 864          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 865          */
 866         for (i = 0; i < 6; i++)
 867                 if (test_facility(stfl_bits[i]))
 868                         elf_hwcap |= 1UL << i;
 869
 870         if (test_facility(22) && test_facility(30))
 871                 elf_hwcap |= HWCAP_S390_ETF3EH;
 872
 873         /*
 874          * Check for additional facilities with store-facility-list-extended.
 875          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 876          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 877          * as stored by stfl, bits 32-xxx contain additional facilities.
 878          * How many facility words are stored depends on the number of
 879          * doublewords passed to the instruction. The additional facilities
 880          * are:
 881          *   Bit 42: decimal floating point facility is installed
 882          *   Bit 44: perform floating point operation facility is installed
 883          * translated to:
 884          *   HWCAP_S390_DFP bit 6 (42 && 44).
 885          */
 886         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 887                 elf_hwcap |= HWCAP_S390_DFP;
 888
 889         /*
 890          * Huge page support HWCAP_S390_HPAGE is bit 7.
 891          */
 892         if (MACHINE_HAS_EDAT1)
 893                 elf_hwcap |= HWCAP_S390_HPAGE;
 894
 895         /*
 896          * 64-bit register support for 31-bit processes
 897          * HWCAP_S390_HIGH_GPRS is bit 9.
 898          */
 899         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 900
 901         /*
 902          * Transactional execution support HWCAP_S390_TE is bit 10.
 903          */
 904         if (MACHINE_HAS_TE)
 905                 elf_hwcap |= HWCAP_S390_TE;
 906
 907         /*
 908          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
 909          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
 910          * instead of facility bit 129.
 911          */
 912         if (MACHINE_HAS_VX) {
 913                 elf_hwcap |= HWCAP_S390_VXRS;
 914                 if (test_facility(134))
 915                         elf_hwcap |= HWCAP_S390_VXRS_EXT;
 916                 if (test_facility(135))
 917                         elf_hwcap |= HWCAP_S390_VXRS_BCD;
 918                 if (test_facility(148))
 919                         elf_hwcap |= HWCAP_S390_VXRS_EXT2;
 920                 if (test_facility(152))
 921                         elf_hwcap |= HWCAP_S390_VXRS_PDE;
 922         }
 923         if (test_facility(150))
 924                 elf_hwcap |= HWCAP_S390_SORT;
 925         if (test_facility(151))
 926                 elf_hwcap |= HWCAP_S390_DFLT;
 927
 928         /*
 929          * Guarded storage support HWCAP_S390_GS is bit 12.
 930          */
 931         if (MACHINE_HAS_GS)
 932                 elf_hwcap |= HWCAP_S390_GS;
 933
 934         get_cpu_id(&cpu_id);
 935         add_device_randomness(&cpu_id, sizeof(cpu_id));
 936         switch (cpu_id.machine) {
 937         case 0x2064:
 938         case 0x2066:
 939         default:        /* Use "z900" as default for 64 bit kernels. */
 940                 strcpy(elf_platform, "z900");
 941                 break;
 942         case 0x2084:
 943         case 0x2086:
 944                 strcpy(elf_platform, "z990");
 945                 break;
 946         case 0x2094:
 947         case 0x2096:
 948                 strcpy(elf_platform, "z9-109");
 949                 break;
 950         case 0x2097:
 951         case 0x2098:
 952                 strcpy(elf_platform, "z10");
 953                 break;
 954         case 0x2817:
 955         case 0x2818:
 956                 strcpy(elf_platform, "z196");
 957                 break;
 958         case 0x2827:
 959         case 0x2828:
 960                 strcpy(elf_platform, "zEC12");
 961                 break;
 962         case 0x2964:
 963         case 0x2965:
 964                 strcpy(elf_platform, "z13");
 965                 break;
 966         case 0x3906:
 967         case 0x3907:
 968                 strcpy(elf_platform, "z14");
 969                 break;
 970         case 0x8561:
 971         case 0x8562:
 972                 strcpy(elf_platform, "z15");
 973                 break;
 974         }
 975
 976         /*
 977          * Virtualization support HWCAP_INT_SIE is bit 0.
 978          */
 979         if (sclp.has_sief2)
 980                 int_hwcap |= HWCAP_INT_SIE;
 981
 982         return 0;
 983 }
 984 arch_initcall(setup_hwcaps);
 985
 986 /*
 987  * Add system information as device randomness
 988  */
 989 static void __init setup_randomness(void)
 990 {
 991         struct sysinfo_3_2_2 *vmms;
 992
 993         vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
 994                                                             PAGE_SIZE);
 995         if (!vmms)
 996                 panic("Failed to allocate memory for sysinfo structure\n");
 997
 998         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
 999                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1000         memblock_free((unsigned long) vmms, PAGE_SIZE);
1001 }
1002
1003 /*
1004  * Find the correct size for the task_struct. This depends on
1005  * the size of the struct fpu at the end of the thread_struct
1006  * which is embedded in the task_struct.
1007  */
1008 static void __init setup_task_size(void)
1009 {
1010         int task_size = sizeof(struct task_struct);
1011
1012         if (!MACHINE_HAS_VX) {
1013                 task_size -= sizeof(__vector128) * __NUM_VXRS;
1014                 task_size += sizeof(freg_t) * __NUM_FPRS;
1015         }
1016         arch_task_struct_size = task_size;
1017 }
1018
1019 /*
1020  * Issue diagnose 318 to set the control program name and
1021  * version codes.
1022  */
1023 static void __init setup_control_program_code(void)
1024 {
1025         union diag318_info diag318_info = {
1026                 .cpnc = CPNC_LINUX,
1027                 .cpvc_linux = 0,
1028                 .cpvc_distro = {0},
1029         };
1030
1031         if (!sclp.has_diag318)
1032                 return;
1033
1034         diag_stat_inc(DIAG_STAT_X318);
1035         asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1036 }
1037
1038 /*
1039  * Print the component list from the IPL report
1040  */
1041 static void __init log_component_list(void)
1042 {
1043         struct ipl_rb_component_entry *ptr, *end;
1044         char *str;
1045
1046         if (!early_ipl_comp_list_addr)
1047                 return;
1048         if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1049                 pr_info("Linux is running with Secure-IPL enabled\n");
1050         else
1051                 pr_info("Linux is running with Secure-IPL disabled\n");
1052         ptr = (void *) early_ipl_comp_list_addr;
1053         end = (void *) ptr + early_ipl_comp_list_size;
1054         pr_info("The IPL report contains the following components:\n");
1055         while (ptr < end) {
1056                 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1057                         if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1058                                 str = "signed, verified";
1059                         else
1060                                 str = "signed, verification failed";
1061                 } else {
1062                         str = "not signed";
1063                 }
1064                 pr_info("%016llx - %016llx (%s)\n",
1065                         ptr->addr, ptr->addr + ptr->len, str);
1066                 ptr++;
1067         }
1068 }
1069
1070 /*
1071  * Setup function called from init/main.c just after the banner
1072  * was printed.
1073  */
1074
1075 void __init setup_arch(char **cmdline_p)
1076 {
1077         /*
1078          * print what head.S has found out about the machine
1079          */
1080         if (MACHINE_IS_VM)
1081                 pr_info("Linux is running as a z/VM "
1082                         "guest operating system in 64-bit mode\n");
1083         else if (MACHINE_IS_KVM)
1084                 pr_info("Linux is running under KVM in 64-bit mode\n");
1085         else if (MACHINE_IS_LPAR)
1086                 pr_info("Linux is running natively in 64-bit mode\n");
1087         else
1088                 pr_info("Linux is running as a guest in 64-bit mode\n");
1089
1090         log_component_list();
1091
1092         /* Have one command line that is parsed and saved in /proc/cmdline */
1093         /* boot_command_line has been already set up in early.c */
1094         *cmdline_p = boot_command_line;
1095
1096         ROOT_DEV = Root_RAM0;
1097
1098         init_mm.start_code = (unsigned long) _text;
1099         init_mm.end_code = (unsigned long) _etext;
1100         init_mm.end_data = (unsigned long) _edata;
1101         init_mm.brk = (unsigned long) _end;
1102
1103         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1104                 nospec_auto_detect();
1105
1106         parse_early_param();
1107 #ifdef CONFIG_CRASH_DUMP
1108         /* Deactivate elfcorehdr= kernel parameter */
1109         elfcorehdr_addr = ELFCORE_ADDR_MAX;
1110 #endif
1111
1112         os_info_init();
1113         setup_ipl();
1114         setup_task_size();
1115         setup_control_program_code();
1116
1117         /* Do some memory reservations *before* memory is added to memblock */
1118         reserve_memory_end();
1119         reserve_oldmem();
1120         reserve_kernel();
1121         reserve_initrd();
1122         reserve_certificate_list();
1123         reserve_mem_detect_info();
1124         memblock_allow_resize();
1125
1126         /* Get information about *all* installed memory */
1127         memblock_add_mem_detect_info();
1128
1129         free_mem_detect_info();
1130         remove_oldmem();
1131
1132         /*
1133          * Make sure all chunks are MAX_ORDER aligned so we don't need the
1134          * extra checks that HOLES_IN_ZONE would require.
1135          *
1136          * Is this still required?
1137          */
1138         memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
1139
1140         if (is_prot_virt_host())
1141                 setup_uv();
1142         setup_memory_end();
1143         setup_memory();
1144         dma_contiguous_reserve(memory_end);
1145         vmcp_cma_reserve();
1146
1147         check_initrd();
1148         reserve_crashkernel();
1149 #ifdef CONFIG_CRASH_DUMP
1150         /*
1151          * Be aware that smp_save_dump_cpus() triggers a system reset.
1152          * Therefore CPU and device initialization should be done afterwards.
1153          */
1154         smp_save_dump_cpus();
1155 #endif
1156
1157         setup_resources();
1158         setup_lowcore_dat_off();
1159         smp_fill_possible_mask();
1160         cpu_detect_mhz_feature();
1161         cpu_init();
1162         numa_setup();
1163         smp_detect_cpus();
1164         topology_init_early();
1165
1166         /*
1167          * Create kernel page tables and switch to virtual addressing.
1168          */
1169         paging_init();
1170
1171         /*
1172          * After paging_init created the kernel page table, the new PSWs
1173          * in lowcore can now run with DAT enabled.
1174          */
1175         setup_lowcore_dat_on();
1176
1177         /* Setup default console */
1178         conmode_default();
1179         set_preferred_console();
1180
1181         apply_alternative_instructions();
1182         if (IS_ENABLED(CONFIG_EXPOLINE))
1183                 nospec_init_branches();
1184
1185         /* Setup zfcpdump support */
1186         setup_zfcpdump();
1187
1188         /* Add system specific data to the random pool */
1189         setup_randomness();
1190 }