arch/s390/kernel/setup.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /*
   3  *  S390 version
   4  *    Copyright IBM Corp. 1999, 2012
   5  *    Author(s): Hartmut Penner ([email protected]),
   6  *               Martin Schwidefsky ([email protected])
   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/archrandom.h>
  56 #include <asm/boot_data.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 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         strlcat(boot_command_line, " cio_ignore=all,!ipldev,!condev", COMMAND_LINE_SIZE);
 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 static unsigned long __init stack_alloc_early(void)
 384 {
 385         unsigned long stack;
 386
 387         stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
 388         if (!stack) {
 389                 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
 390                       __func__, THREAD_SIZE, THREAD_SIZE);
 391         }
 392         return stack;
 393 }
 394
 395 static void __init setup_lowcore(void)
 396 {
 397         struct lowcore *lc, *abs_lc;
 398
 399         /*
 400          * Setup lowcore for boot cpu
 401          */
 402         BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
 403         lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
 404         if (!lc)
 405                 panic("%s: Failed to allocate %zu bytes align=%zx\n",
 406                       __func__, sizeof(*lc), sizeof(*lc));
 407
 408         lc->pcpu = (unsigned long)per_cpu_ptr(&pcpu_devices, 0);
 409         lc->restart_psw.mask = PSW_KERNEL_BITS & ~PSW_MASK_DAT;
 410         lc->restart_psw.addr = __pa(restart_int_handler);
 411         lc->external_new_psw.mask = PSW_KERNEL_BITS;
 412         lc->external_new_psw.addr = (unsigned long) ext_int_handler;
 413         lc->svc_new_psw.mask = PSW_KERNEL_BITS;
 414         lc->svc_new_psw.addr = (unsigned long) system_call;
 415         lc->program_new_psw.mask = PSW_KERNEL_BITS;
 416         lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
 417         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 418         lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
 419         lc->io_new_psw.mask = PSW_KERNEL_BITS;
 420         lc->io_new_psw.addr = (unsigned long) io_int_handler;
 421         lc->clock_comparator = clock_comparator_max;
 422         lc->current_task = (unsigned long)&init_task;
 423         lc->lpp = LPP_MAGIC;
 424         lc->machine_flags = get_lowcore()->machine_flags;
 425         lc->preempt_count = get_lowcore()->preempt_count;
 426         nmi_alloc_mcesa_early(&lc->mcesad);
 427         lc->sys_enter_timer = get_lowcore()->sys_enter_timer;
 428         lc->exit_timer = get_lowcore()->exit_timer;
 429         lc->user_timer = get_lowcore()->user_timer;
 430         lc->system_timer = get_lowcore()->system_timer;
 431         lc->steal_timer = get_lowcore()->steal_timer;
 432         lc->last_update_timer = get_lowcore()->last_update_timer;
 433         lc->last_update_clock = get_lowcore()->last_update_clock;
 434         /*
 435          * Allocate the global restart stack which is the same for
 436          * all CPUs in case *one* of them does a PSW restart.
 437          */
 438         restart_stack = (void *)(stack_alloc_early() + STACK_INIT_OFFSET);
 439         lc->mcck_stack = stack_alloc_early() + STACK_INIT_OFFSET;
 440         lc->async_stack = stack_alloc_early() + STACK_INIT_OFFSET;
 441         lc->nodat_stack = stack_alloc_early() + STACK_INIT_OFFSET;
 442         lc->kernel_stack = get_lowcore()->kernel_stack;
 443         /*
 444          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 445          * restart data to the absolute zero lowcore. This is necessary if
 446          * PSW restart is done on an offline CPU that has lowcore zero.
 447          */
 448         lc->restart_stack = (unsigned long) restart_stack;
 449         lc->restart_fn = (unsigned long) do_restart;
 450         lc->restart_data = 0;
 451         lc->restart_source = -1U;
 452         lc->spinlock_lockval = arch_spin_lockval(0);
 453         lc->spinlock_index = 0;
 454         arch_spin_lock_setup(0);
 455         lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
 456         lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
 457         lc->preempt_count = PREEMPT_DISABLED;
 458         lc->kernel_asce = get_lowcore()->kernel_asce;
 459         lc->user_asce = get_lowcore()->user_asce;
 460
 461         system_ctlreg_init_save_area(lc);
 462         abs_lc = get_abs_lowcore();
 463         abs_lc->restart_stack = lc->restart_stack;
 464         abs_lc->restart_fn = lc->restart_fn;
 465         abs_lc->restart_data = lc->restart_data;
 466         abs_lc->restart_source = lc->restart_source;
 467         abs_lc->restart_psw = lc->restart_psw;
 468         abs_lc->restart_flags = RESTART_FLAG_CTLREGS;
 469         abs_lc->program_new_psw = lc->program_new_psw;
 470         abs_lc->mcesad = lc->mcesad;
 471         put_abs_lowcore(abs_lc);
 472
 473         set_prefix(__pa(lc));
 474         lowcore_ptr[0] = lc;
 475         if (abs_lowcore_map(0, lowcore_ptr[0], false))
 476                 panic("Couldn't setup absolute lowcore");
 477 }
 478
 479 static struct resource code_resource = {
 480         .name  = "Kernel code",
 481         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 482 };
 483
 484 static struct resource data_resource = {
 485         .name = "Kernel data",
 486         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 487 };
 488
 489 static struct resource bss_resource = {
 490         .name = "Kernel bss",
 491         .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
 492 };
 493
 494 static struct resource __initdata *standard_resources[] = {
 495         &code_resource,
 496         &data_resource,
 497         &bss_resource,
 498 };
 499
 500 static void __init setup_resources(void)
 501 {
 502         struct resource *res, *std_res, *sub_res;
 503         phys_addr_t start, end;
 504         int j;
 505         u64 i;
 506
 507         code_resource.start = __pa_symbol(_text);
 508         code_resource.end = __pa_symbol(_etext) - 1;
 509         data_resource.start = __pa_symbol(_etext);
 510         data_resource.end = __pa_symbol(_edata) - 1;
 511         bss_resource.start = __pa_symbol(__bss_start);
 512         bss_resource.end = __pa_symbol(__bss_stop) - 1;
 513
 514         for_each_mem_range(i, &start, &end) {
 515                 res = memblock_alloc(sizeof(*res), 8);
 516                 if (!res)
 517                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 518                               __func__, sizeof(*res), 8);
 519                 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
 520
 521                 res->name = "System RAM";
 522                 res->start = start;
 523                 /*
 524                  * In memblock, end points to the first byte after the
 525                  * range while in resources, end points to the last byte in
 526                  * the range.
 527                  */
 528                 res->end = end - 1;
 529                 request_resource(&iomem_resource, res);
 530
 531                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 532                         std_res = standard_resources[j];
 533                         if (std_res->start < res->start ||
 534                             std_res->start > res->end)
 535                                 continue;
 536                         if (std_res->end > res->end) {
 537                                 sub_res = memblock_alloc(sizeof(*sub_res), 8);
 538                                 if (!sub_res)
 539                                         panic("%s: Failed to allocate %zu bytes align=0x%x\n",
 540                                               __func__, sizeof(*sub_res), 8);
 541                                 *sub_res = *std_res;
 542                                 sub_res->end = res->end;
 543                                 std_res->start = res->end + 1;
 544                                 request_resource(res, sub_res);
 545                         } else {
 546                                 request_resource(res, std_res);
 547                         }
 548                 }
 549         }
 550 #ifdef CONFIG_CRASH_DUMP
 551         /*
 552          * Re-add removed crash kernel memory as reserved memory. This makes
 553          * sure it will be mapped with the identity mapping and struct pages
 554          * will be created, so it can be resized later on.
 555          * However add it later since the crash kernel resource should not be
 556          * part of the System RAM resource.
 557          */
 558         if (crashk_res.end) {
 559                 memblock_add_node(crashk_res.start, resource_size(&crashk_res),
 560                                   0, MEMBLOCK_NONE);
 561                 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
 562                 insert_resource(&iomem_resource, &crashk_res);
 563         }
 564 #endif
 565 }
 566
 567 static void __init setup_memory_end(void)
 568 {
 569         max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
 570         pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
 571 }
 572
 573 #ifdef CONFIG_CRASH_DUMP
 574
 575 /*
 576  * When kdump is enabled, we have to ensure that no memory from the area
 577  * [0 - crashkernel memory size] is set offline - it will be exchanged with
 578  * the crashkernel memory region when kdump is triggered. The crashkernel
 579  * memory region can never get offlined (pages are unmovable).
 580  */
 581 static int kdump_mem_notifier(struct notifier_block *nb,
 582                               unsigned long action, void *data)
 583 {
 584         struct memory_notify *arg = data;
 585
 586         if (action != MEM_GOING_OFFLINE)
 587                 return NOTIFY_OK;
 588         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 589                 return NOTIFY_BAD;
 590         return NOTIFY_OK;
 591 }
 592
 593 static struct notifier_block kdump_mem_nb = {
 594         .notifier_call = kdump_mem_notifier,
 595 };
 596
 597 #endif
 598
 599 /*
 600  * Reserve page tables created by decompressor
 601  */
 602 static void __init reserve_pgtables(void)
 603 {
 604         unsigned long start, end;
 605         struct reserved_range *range;
 606
 607         for_each_physmem_reserved_type_range(RR_VMEM, range, &start, &end)
 608                 memblock_reserve(start, end - start);
 609 }
 610
 611 /*
 612  * Reserve memory for kdump kernel to be loaded with kexec
 613  */
 614 static void __init reserve_crashkernel(void)
 615 {
 616 #ifdef CONFIG_CRASH_DUMP
 617         unsigned long long crash_base, crash_size;
 618         phys_addr_t low, high;
 619         int rc;
 620
 621         rc = parse_crashkernel(boot_command_line, ident_map_size,
 622                                &crash_size, &crash_base, NULL, NULL);
 623
 624         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 625         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 626         if (rc || crash_size == 0)
 627                 return;
 628
 629         if (memblock.memory.regions[0].size < crash_size) {
 630                 pr_info("crashkernel reservation failed: %s\n",
 631                         "first memory chunk must be at least crashkernel size");
 632                 return;
 633         }
 634
 635         low = crash_base ?: oldmem_data.start;
 636         high = low + crash_size;
 637         if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
 638                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 639                 crash_base = low;
 640         } else {
 641                 /* Find suitable area in free memory */
 642                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 643                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 644
 645                 if (crash_base && crash_base < low) {
 646                         pr_info("crashkernel reservation failed: %s\n",
 647                                 "crash_base too low");
 648                         return;
 649                 }
 650                 low = crash_base ?: low;
 651                 crash_base = memblock_phys_alloc_range(crash_size,
 652                                                        KEXEC_CRASH_MEM_ALIGN,
 653                                                        low, high);
 654         }
 655
 656         if (!crash_base) {
 657                 pr_info("crashkernel reservation failed: %s\n",
 658                         "no suitable area found");
 659                 return;
 660         }
 661
 662         if (register_memory_notifier(&kdump_mem_nb)) {
 663                 memblock_phys_free(crash_base, crash_size);
 664                 return;
 665         }
 666
 667         if (!oldmem_data.start && MACHINE_IS_VM)
 668                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 669         crashk_res.start = crash_base;
 670         crashk_res.end = crash_base + crash_size - 1;
 671         memblock_remove(crash_base, crash_size);
 672         pr_info("Reserving %lluMB of memory at %lluMB "
 673                 "for crashkernel (System RAM: %luMB)\n",
 674                 crash_size >> 20, crash_base >> 20,
 675                 (unsigned long)memblock.memory.total_size >> 20);
 676         os_info_crashkernel_add(crash_base, crash_size);
 677 #endif
 678 }
 679
 680 /*
 681  * Reserve the initrd from being used by memblock
 682  */
 683 static void __init reserve_initrd(void)
 684 {
 685         unsigned long addr, size;
 686
 687         if (!IS_ENABLED(CONFIG_BLK_DEV_INITRD) || !get_physmem_reserved(RR_INITRD, &addr, &size))
 688                 return;
 689         initrd_start = (unsigned long)__va(addr);
 690         initrd_end = initrd_start + size;
 691         memblock_reserve(addr, size);
 692 }
 693
 694 /*
 695  * Reserve the memory area used to pass the certificate lists
 696  */
 697 static void __init reserve_certificate_list(void)
 698 {
 699         if (ipl_cert_list_addr)
 700                 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
 701 }
 702
 703 static void __init reserve_physmem_info(void)
 704 {
 705         unsigned long addr, size;
 706
 707         if (get_physmem_reserved(RR_MEM_DETECT_EXTENDED, &addr, &size))
 708                 memblock_reserve(addr, size);
 709 }
 710
 711 static void __init free_physmem_info(void)
 712 {
 713         unsigned long addr, size;
 714
 715         if (get_physmem_reserved(RR_MEM_DETECT_EXTENDED, &addr, &size))
 716                 memblock_phys_free(addr, size);
 717 }
 718
 719 static void __init memblock_add_physmem_info(void)
 720 {
 721         unsigned long start, end;
 722         int i;
 723
 724         pr_debug("physmem info source: %s (%hhd)\n",
 725                  get_physmem_info_source(), physmem_info.info_source);
 726         /* keep memblock lists close to the kernel */
 727         memblock_set_bottom_up(true);
 728         for_each_physmem_usable_range(i, &start, &end)
 729                 memblock_add(start, end - start);
 730         for_each_physmem_online_range(i, &start, &end)
 731                 memblock_physmem_add(start, end - start);
 732         memblock_set_bottom_up(false);
 733         memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
 734 }
 735
 736 /*
 737  * Reserve memory used for lowcore.
 738  */
 739 static void __init reserve_lowcore(void)
 740 {
 741         void *lowcore_start = get_lowcore();
 742         void *lowcore_end = lowcore_start + sizeof(struct lowcore);
 743         void *start, *end;
 744
 745         if ((void *)__identity_base < lowcore_end) {
 746                 start = max(lowcore_start, (void *)__identity_base);
 747                 end = min(lowcore_end, (void *)(__identity_base + ident_map_size));
 748                 memblock_reserve(__pa(start), __pa(end));
 749         }
 750 }
 751
 752 /*
 753  * Reserve memory used for absolute lowcore/command line/kernel image.
 754  */
 755 static void __init reserve_kernel(void)
 756 {
 757         memblock_reserve(0, STARTUP_NORMAL_OFFSET);
 758         memblock_reserve(OLDMEM_BASE, sizeof(unsigned long));
 759         memblock_reserve(OLDMEM_SIZE, sizeof(unsigned long));
 760         memblock_reserve(physmem_info.reserved[RR_AMODE31].start, __eamode31 - __samode31);
 761         memblock_reserve(__pa(sclp_early_sccb), EXT_SCCB_READ_SCP);
 762         memblock_reserve(__pa(_stext), _end - _stext);
 763 }
 764
 765 static void __init setup_memory(void)
 766 {
 767         phys_addr_t start, end;
 768         u64 i;
 769
 770         /*
 771          * Init storage key for present memory
 772          */
 773         for_each_mem_range(i, &start, &end)
 774                 storage_key_init_range(start, end);
 775
 776         psw_set_key(PAGE_DEFAULT_KEY);
 777 }
 778
 779 static void __init relocate_amode31_section(void)
 780 {
 781         unsigned long amode31_size = __eamode31 - __samode31;
 782         long amode31_offset, *ptr;
 783
 784         amode31_offset = AMODE31_START - (unsigned long)__samode31;
 785         pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size);
 786
 787         /* Move original AMODE31 section to the new one */
 788         memmove((void *)physmem_info.reserved[RR_AMODE31].start, __samode31, amode31_size);
 789         /* Zero out the old AMODE31 section to catch invalid accesses within it */
 790         memset(__samode31, 0, amode31_size);
 791
 792         /* Update all AMODE31 region references */
 793         for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++)
 794                 *ptr += amode31_offset;
 795 }
 796
 797 /* This must be called after AMODE31 relocation */
 798 static void __init setup_cr(void)
 799 {
 800         union ctlreg2 cr2;
 801         union ctlreg5 cr5;
 802         union ctlreg15 cr15;
 803
 804         __ctl_duct[1] = (unsigned long)__ctl_aste;
 805         __ctl_duct[2] = (unsigned long)__ctl_aste;
 806         __ctl_duct[4] = (unsigned long)__ctl_duald;
 807
 808         /* Update control registers CR2, CR5 and CR15 */
 809         local_ctl_store(2, &cr2.reg);
 810         local_ctl_store(5, &cr5.reg);
 811         local_ctl_store(15, &cr15.reg);
 812         cr2.ducto = (unsigned long)__ctl_duct >> 6;
 813         cr5.pasteo = (unsigned long)__ctl_duct >> 6;
 814         cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
 815         system_ctl_load(2, &cr2.reg);
 816         system_ctl_load(5, &cr5.reg);
 817         system_ctl_load(15, &cr15.reg);
 818 }
 819
 820 /*
 821  * Add system information as device randomness
 822  */
 823 static void __init setup_randomness(void)
 824 {
 825         struct sysinfo_3_2_2 *vmms;
 826
 827         vmms = memblock_alloc(PAGE_SIZE, PAGE_SIZE);
 828         if (!vmms)
 829                 panic("Failed to allocate memory for sysinfo structure\n");
 830         if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
 831                 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
 832         memblock_free(vmms, PAGE_SIZE);
 833
 834         if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG))
 835                 static_branch_enable(&s390_arch_random_available);
 836 }
 837
 838 /*
 839  * Issue diagnose 318 to set the control program name and
 840  * version codes.
 841  */
 842 static void __init setup_control_program_code(void)
 843 {
 844         union diag318_info diag318_info = {
 845                 .cpnc = CPNC_LINUX,
 846                 .cpvc = 0,
 847         };
 848
 849         if (!sclp.has_diag318)
 850                 return;
 851
 852         diag_stat_inc(DIAG_STAT_X318);
 853         asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
 854 }
 855
 856 /*
 857  * Print the component list from the IPL report
 858  */
 859 static void __init log_component_list(void)
 860 {
 861         struct ipl_rb_component_entry *ptr, *end;
 862         char *str;
 863
 864         if (!early_ipl_comp_list_addr)
 865                 return;
 866         if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
 867                 pr_info("Linux is running with Secure-IPL enabled\n");
 868         else
 869                 pr_info("Linux is running with Secure-IPL disabled\n");
 870         ptr = __va(early_ipl_comp_list_addr);
 871         end = (void *) ptr + early_ipl_comp_list_size;
 872         pr_info("The IPL report contains the following components:\n");
 873         while (ptr < end) {
 874                 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
 875                         if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
 876                                 str = "signed, verified";
 877                         else
 878                                 str = "signed, verification failed";
 879                 } else {
 880                         str = "not signed";
 881                 }
 882                 pr_info("%016llx - %016llx (%s)\n",
 883                         ptr->addr, ptr->addr + ptr->len, str);
 884                 ptr++;
 885         }
 886 }
 887
 888 /*
 889  * Setup function called from init/main.c just after the banner
 890  * was printed.
 891  */
 892
 893 void __init setup_arch(char **cmdline_p)
 894 {
 895         /*
 896          * print what head.S has found out about the machine
 897          */
 898         if (MACHINE_IS_VM)
 899                 pr_info("Linux is running as a z/VM "
 900                         "guest operating system in 64-bit mode\n");
 901         else if (MACHINE_IS_KVM)
 902                 pr_info("Linux is running under KVM in 64-bit mode\n");
 903         else if (MACHINE_IS_LPAR)
 904                 pr_info("Linux is running natively in 64-bit mode\n");
 905         else
 906                 pr_info("Linux is running as a guest in 64-bit mode\n");
 907
 908         if (have_relocated_lowcore())
 909                 pr_info("Lowcore relocated to 0x%px\n", get_lowcore());
 910
 911         log_component_list();
 912
 913         /* Have one command line that is parsed and saved in /proc/cmdline */
 914         /* boot_command_line has been already set up in early.c */
 915         *cmdline_p = boot_command_line;
 916
 917         ROOT_DEV = Root_RAM0;
 918
 919         setup_initial_init_mm(_text, _etext, _edata, _end);
 920
 921         if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
 922                 nospec_auto_detect();
 923
 924         jump_label_init();
 925         parse_early_param();
 926 #ifdef CONFIG_CRASH_DUMP
 927         /* Deactivate elfcorehdr= kernel parameter */
 928         elfcorehdr_addr = ELFCORE_ADDR_MAX;
 929 #endif
 930
 931         os_info_init();
 932         setup_ipl();
 933         setup_control_program_code();
 934
 935         /* Do some memory reservations *before* memory is added to memblock */
 936         reserve_pgtables();
 937         reserve_lowcore();
 938         reserve_kernel();
 939         reserve_initrd();
 940         reserve_certificate_list();
 941         reserve_physmem_info();
 942         memblock_set_current_limit(ident_map_size);
 943         memblock_allow_resize();
 944
 945         /* Get information about *all* installed memory */
 946         memblock_add_physmem_info();
 947
 948         free_physmem_info();
 949         setup_memory_end();
 950         memblock_dump_all();
 951         setup_memory();
 952
 953         relocate_amode31_section();
 954         setup_cr();
 955         setup_uv();
 956         dma_contiguous_reserve(ident_map_size);
 957         vmcp_cma_reserve();
 958         if (MACHINE_HAS_EDAT2)
 959                 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
 960
 961         reserve_crashkernel();
 962 #ifdef CONFIG_CRASH_DUMP
 963         /*
 964          * Be aware that smp_save_dump_secondary_cpus() triggers a system reset.
 965          * Therefore CPU and device initialization should be done afterwards.
 966          */
 967         smp_save_dump_secondary_cpus();
 968 #endif
 969
 970         setup_resources();
 971         setup_lowcore();
 972         smp_fill_possible_mask();
 973         cpu_detect_mhz_feature();
 974         cpu_init();
 975         numa_setup();
 976         smp_detect_cpus();
 977         topology_init_early();
 978
 979         if (test_facility(193))
 980                 static_branch_enable(&cpu_has_bear);
 981
 982         /*
 983          * Create kernel page tables.
 984          */
 985         paging_init();
 986
 987         /*
 988          * After paging_init created the kernel page table, the new PSWs
 989          * in lowcore can now run with DAT enabled.
 990          */
 991 #ifdef CONFIG_CRASH_DUMP
 992         smp_save_dump_ipl_cpu();
 993 #endif
 994
 995         /* Setup default console */
 996         conmode_default();
 997         set_preferred_console();
 998
 999         apply_alternative_instructions();
1000         if (IS_ENABLED(CONFIG_EXPOLINE))
1001                 nospec_init_branches();
1002
1003         /* Setup zfcp/nvme dump support */
1004         setup_zfcpdump();
1005
1006         /* Add system specific data to the random pool */
1007         setup_randomness();
1008 }