1 // SPDX-License-Identifier: GPL-2.0
4 * Copyright IBM Corp. 1999, 2012
8 * Derived from "arch/i386/kernel/setup.c"
9 * Copyright (C) 1995, Linus Torvalds
13 * This file handles the architecture-dependent parts of initialization
16 #define KMSG_COMPONENT "setup"
17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
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>
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>
54 #include <asm/boot_data.h>
56 #include <asm/facility.h>
58 #include <asm/mmu_context.h>
59 #include <asm/cpcmd.h>
60 #include <asm/lowcore.h>
64 #include <asm/ptrace.h>
65 #include <asm/sections.h>
66 #include <asm/ebcdic.h>
68 #include <asm/os_info.h>
70 #include <asm/stacktrace.h>
71 #include <asm/sysinfo.h>
73 #include <asm/alternative.h>
74 #include <asm/nospec-branch.h>
75 #include <asm/mem_detect.h>
77 #include <asm/asm-offsets.h>
83 unsigned int console_mode = 0;
84 EXPORT_SYMBOL(console_mode);
86 unsigned int console_devno = -1;
87 EXPORT_SYMBOL(console_devno);
89 unsigned int console_irq = -1;
90 EXPORT_SYMBOL(console_irq);
92 unsigned long elf_hwcap __read_mostly = 0;
93 char elf_platform[ELF_PLATFORM_SIZE];
95 unsigned long int_hwcap = 0;
97 int __bootdata(noexec_disabled);
98 unsigned long __bootdata(ident_map_size);
99 unsigned long __bootdata(vmalloc_size);
100 struct mem_detect_info __bootdata(mem_detect);
102 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table);
103 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table);
104 unsigned long __bootdata_preserved(__stext_dma);
105 unsigned long __bootdata_preserved(__etext_dma);
106 unsigned long __bootdata_preserved(__sdma);
107 unsigned long __bootdata_preserved(__edma);
108 unsigned long __bootdata_preserved(__kaslr_offset);
109 unsigned int __bootdata_preserved(zlib_dfltcc_support);
110 EXPORT_SYMBOL(zlib_dfltcc_support);
112 unsigned long VMALLOC_START;
113 EXPORT_SYMBOL(VMALLOC_START);
115 unsigned long VMALLOC_END;
116 EXPORT_SYMBOL(VMALLOC_END);
118 struct page *vmemmap;
119 EXPORT_SYMBOL(vmemmap);
120 unsigned long vmemmap_size;
122 unsigned long MODULES_VADDR;
123 unsigned long MODULES_END;
125 /* An array with a pointer to the lowcore of every CPU. */
126 struct lowcore *lowcore_ptr[NR_CPUS];
127 EXPORT_SYMBOL(lowcore_ptr);
130 * The Write Back bit position in the physaddr is given by the SLPC PCI.
131 * Leaving the mask zero always uses write through which is safe
133 unsigned long mio_wb_bit_mask __ro_after_init;
136 * This is set up by the setup-routine at boot-time
137 * for S390 need to find out, what we have to setup
138 * using address 0x10400 ...
141 #include <asm/setup.h>
144 * condev= and conmode= setup parameter.
147 static int __init condev_setup(char *str)
151 vdev = simple_strtoul(str, &str, 0);
152 if (vdev >= 0 && vdev < 65536) {
153 console_devno = vdev;
159 __setup("condev=", condev_setup);
161 static void __init set_preferred_console(void)
163 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
164 add_preferred_console("ttyS", 0, NULL);
165 else if (CONSOLE_IS_3270)
166 add_preferred_console("tty3270", 0, NULL);
167 else if (CONSOLE_IS_VT220)
168 add_preferred_console("ttyS", 1, NULL);
169 else if (CONSOLE_IS_HVC)
170 add_preferred_console("hvc", 0, NULL);
173 static int __init conmode_setup(char *str)
175 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
176 if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
179 #if defined(CONFIG_TN3215_CONSOLE)
180 if (!strcmp(str, "3215"))
183 #if defined(CONFIG_TN3270_CONSOLE)
184 if (!strcmp(str, "3270"))
187 set_preferred_console();
191 __setup("conmode=", conmode_setup);
193 static void __init conmode_default(void)
195 char query_buffer[1024];
199 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
200 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
201 ptr = strstr(query_buffer, "SUBCHANNEL =");
202 console_irq = simple_strtoul(ptr + 13, NULL, 16);
203 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
204 ptr = strstr(query_buffer, "CONMODE");
206 * Set the conmode to 3215 so that the device recognition
207 * will set the cu_type of the console to 3215. If the
208 * conmode is 3270 and we don't set it back then both
209 * 3215 and the 3270 driver will try to access the console
210 * device (3215 as console and 3270 as normal tty).
212 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
214 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
219 if (str_has_prefix(ptr + 8, "3270")) {
220 #if defined(CONFIG_TN3270_CONSOLE)
222 #elif defined(CONFIG_TN3215_CONSOLE)
224 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
227 } else if (str_has_prefix(ptr + 8, "3215")) {
228 #if defined(CONFIG_TN3215_CONSOLE)
230 #elif defined(CONFIG_TN3270_CONSOLE)
232 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
236 } else if (MACHINE_IS_KVM) {
237 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
239 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
244 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
250 #ifdef CONFIG_CRASH_DUMP
251 static void __init setup_zfcpdump(void)
253 if (!is_ipl_type_dump())
257 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
258 console_loglevel = 2;
261 static inline void setup_zfcpdump(void) {}
262 #endif /* CONFIG_CRASH_DUMP */
265 * Reboot, halt and power_off stubs. They just call _machine_restart,
266 * _machine_halt or _machine_power_off.
269 void machine_restart(char *command)
271 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
273 * Only unblank the console if we are called in enabled
274 * context or a bust_spinlocks cleared the way for us.
277 _machine_restart(command);
280 void machine_halt(void)
282 if (!in_interrupt() || oops_in_progress)
284 * Only unblank the console if we are called in enabled
285 * context or a bust_spinlocks cleared the way for us.
291 void machine_power_off(void)
293 if (!in_interrupt() || oops_in_progress)
295 * Only unblank the console if we are called in enabled
296 * context or a bust_spinlocks cleared the way for us.
299 _machine_power_off();
303 * Dummy power off function.
305 void (*pm_power_off)(void) = machine_power_off;
306 EXPORT_SYMBOL_GPL(pm_power_off);
310 unsigned long stack_alloc(void)
312 #ifdef CONFIG_VMAP_STACK
313 return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE,
314 THREADINFO_GFP, NUMA_NO_NODE,
315 __builtin_return_address(0));
317 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
321 void stack_free(unsigned long stack)
323 #ifdef CONFIG_VMAP_STACK
324 vfree((void *) stack);
326 free_pages(stack, THREAD_SIZE_ORDER);
330 int __init arch_early_irq_init(void)
334 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
336 panic("Couldn't allocate async stack");
337 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
341 static int __init stack_realloc(void)
343 unsigned long old, new;
345 old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
348 panic("Couldn't allocate async stack");
349 WRITE_ONCE(S390_lowcore.async_stack, new + STACK_INIT_OFFSET);
350 free_pages(old, THREAD_SIZE_ORDER);
352 old = S390_lowcore.mcck_stack - STACK_INIT_OFFSET;
355 panic("Couldn't allocate machine check stack");
356 WRITE_ONCE(S390_lowcore.mcck_stack, new + STACK_INIT_OFFSET);
357 memblock_free_late(old, THREAD_SIZE);
360 early_initcall(stack_realloc);
362 void __init arch_call_rest_init(void)
366 stack = stack_alloc();
368 panic("Couldn't allocate kernel stack");
369 current->stack = (void *) stack;
370 #ifdef CONFIG_VMAP_STACK
371 current->stack_vm_area = (void *) stack;
373 set_task_stack_end_magic(current);
374 stack += STACK_INIT_OFFSET;
375 S390_lowcore.kernel_stack = stack;
376 CALL_ON_STACK_NORETURN(rest_init, stack);
379 static void __init setup_lowcore_dat_off(void)
381 unsigned long int_psw_mask = PSW_KERNEL_BITS;
382 unsigned long mcck_stack;
385 if (IS_ENABLED(CONFIG_KASAN))
386 int_psw_mask |= PSW_MASK_DAT;
389 * Setup lowcore for boot cpu
391 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
392 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
394 panic("%s: Failed to allocate %zu bytes align=%zx\n",
395 __func__, sizeof(*lc), sizeof(*lc));
397 lc->restart_psw.mask = PSW_KERNEL_BITS;
398 lc->restart_psw.addr = (unsigned long) restart_int_handler;
399 lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
400 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
401 lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
402 lc->svc_new_psw.addr = (unsigned long) system_call;
403 lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
404 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
405 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
406 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
407 lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
408 lc->io_new_psw.addr = (unsigned long) io_int_handler;
409 lc->clock_comparator = clock_comparator_max;
410 lc->nodat_stack = ((unsigned long) &init_thread_union)
411 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
412 lc->current_task = (unsigned long)&init_task;
414 lc->machine_flags = S390_lowcore.machine_flags;
415 lc->preempt_count = S390_lowcore.preempt_count;
416 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
417 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
418 sizeof(lc->stfle_fac_list));
419 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
420 sizeof(lc->alt_stfle_fac_list));
421 nmi_alloc_boot_cpu(lc);
422 lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
423 lc->exit_timer = S390_lowcore.exit_timer;
424 lc->user_timer = S390_lowcore.user_timer;
425 lc->system_timer = S390_lowcore.system_timer;
426 lc->steal_timer = S390_lowcore.steal_timer;
427 lc->last_update_timer = S390_lowcore.last_update_timer;
428 lc->last_update_clock = S390_lowcore.last_update_clock;
431 * Allocate the global restart stack which is the same for
432 * all CPUs in cast *one* of them does a PSW restart.
434 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
436 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
437 __func__, THREAD_SIZE, THREAD_SIZE);
438 restart_stack += STACK_INIT_OFFSET;
441 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
442 * restart data to the absolute zero lowcore. This is necessary if
443 * PSW restart is done on an offline CPU that has lowcore zero.
445 lc->restart_stack = (unsigned long) restart_stack;
446 lc->restart_fn = (unsigned long) do_restart;
447 lc->restart_data = 0;
448 lc->restart_source = -1UL;
450 mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
452 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
453 __func__, THREAD_SIZE, THREAD_SIZE);
454 lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
456 /* Setup absolute zero lowcore */
457 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
458 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
459 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
460 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
461 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
463 lc->spinlock_lockval = arch_spin_lockval(0);
464 lc->spinlock_index = 0;
465 arch_spin_lock_setup(0);
466 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
467 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
468 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
470 set_prefix((u32)(unsigned long) lc);
474 static void __init setup_lowcore_dat_on(void)
476 __ctl_clear_bit(0, 28);
477 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
478 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
479 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
480 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
481 __ctl_set_bit(0, 28);
484 static struct resource code_resource = {
485 .name = "Kernel code",
486 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
489 static struct resource data_resource = {
490 .name = "Kernel data",
491 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
494 static struct resource bss_resource = {
495 .name = "Kernel bss",
496 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
499 static struct resource __initdata *standard_resources[] = {
505 static void __init setup_resources(void)
507 struct resource *res, *std_res, *sub_res;
508 phys_addr_t start, end;
512 code_resource.start = (unsigned long) _text;
513 code_resource.end = (unsigned long) _etext - 1;
514 data_resource.start = (unsigned long) _etext;
515 data_resource.end = (unsigned long) _edata - 1;
516 bss_resource.start = (unsigned long) __bss_start;
517 bss_resource.end = (unsigned long) __bss_stop - 1;
519 for_each_mem_range(i, &start, &end) {
520 res = memblock_alloc(sizeof(*res), 8);
522 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
523 __func__, sizeof(*res), 8);
524 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
526 res->name = "System RAM";
529 * In memblock, end points to the first byte after the
530 * range while in resourses, end points to the last byte in
534 request_resource(&iomem_resource, res);
536 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
537 std_res = standard_resources[j];
538 if (std_res->start < res->start ||
539 std_res->start > res->end)
541 if (std_res->end > res->end) {
542 sub_res = memblock_alloc(sizeof(*sub_res), 8);
544 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
545 __func__, sizeof(*sub_res), 8);
547 sub_res->end = res->end;
548 std_res->start = res->end + 1;
549 request_resource(res, sub_res);
551 request_resource(res, std_res);
555 #ifdef CONFIG_CRASH_DUMP
557 * Re-add removed crash kernel memory as reserved memory. This makes
558 * sure it will be mapped with the identity mapping and struct pages
559 * will be created, so it can be resized later on.
560 * However add it later since the crash kernel resource should not be
561 * part of the System RAM resource.
563 if (crashk_res.end) {
564 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
565 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
566 insert_resource(&iomem_resource, &crashk_res);
571 static void __init setup_ident_map_size(void)
573 unsigned long vmax, tmp;
575 /* Choose kernel address space layout: 3 or 4 levels. */
576 tmp = ident_map_size / PAGE_SIZE;
577 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
578 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
579 vmax = _REGION2_SIZE; /* 3-level kernel page table */
581 vmax = _REGION1_SIZE; /* 4-level kernel page table */
582 /* module area is at the end of the kernel address space. */
584 if (is_prot_virt_host())
585 adjust_to_uv_max(&MODULES_END);
587 vmax = _REGION1_SIZE;
588 MODULES_END = kasan_vmax;
590 MODULES_VADDR = MODULES_END - MODULES_LEN;
591 VMALLOC_END = MODULES_VADDR;
592 VMALLOC_START = VMALLOC_END - vmalloc_size;
594 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
595 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
596 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
597 tmp = SECTION_ALIGN_UP(tmp);
598 tmp = VMALLOC_START - tmp * sizeof(struct page);
599 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
600 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
601 vmemmap = (struct page *) tmp;
603 /* Take care that ident_map_size <= vmemmap */
604 ident_map_size = min(ident_map_size, (unsigned long)vmemmap);
606 ident_map_size = min(ident_map_size, KASAN_SHADOW_START);
608 vmemmap_size = SECTION_ALIGN_UP(ident_map_size / PAGE_SIZE) * sizeof(struct page);
610 /* move vmemmap above kasan shadow only if stands in a way */
611 if (KASAN_SHADOW_END > (unsigned long)vmemmap &&
612 (unsigned long)vmemmap + vmemmap_size > KASAN_SHADOW_START)
613 vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
615 max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
616 memblock_remove(ident_map_size, ULONG_MAX);
618 pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
621 #ifdef CONFIG_CRASH_DUMP
624 * When kdump is enabled, we have to ensure that no memory from the area
625 * [0 - crashkernel memory size] is set offline - it will be exchanged with
626 * the crashkernel memory region when kdump is triggered. The crashkernel
627 * memory region can never get offlined (pages are unmovable).
629 static int kdump_mem_notifier(struct notifier_block *nb,
630 unsigned long action, void *data)
632 struct memory_notify *arg = data;
634 if (action != MEM_GOING_OFFLINE)
636 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
641 static struct notifier_block kdump_mem_nb = {
642 .notifier_call = kdump_mem_notifier,
648 * Make sure that the area above identity mapping is protected
650 static void __init reserve_above_ident_map(void)
652 memblock_reserve(ident_map_size, ULONG_MAX);
656 * Make sure that oldmem, where the dump is stored, is protected
658 static void __init reserve_oldmem(void)
660 #ifdef CONFIG_CRASH_DUMP
662 /* Forget all memory above the running kdump system */
663 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
668 * Make sure that oldmem, where the dump is stored, is protected
670 static void __init remove_oldmem(void)
672 #ifdef CONFIG_CRASH_DUMP
674 /* Forget all memory above the running kdump system */
675 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
680 * Reserve memory for kdump kernel to be loaded with kexec
682 static void __init reserve_crashkernel(void)
684 #ifdef CONFIG_CRASH_DUMP
685 unsigned long long crash_base, crash_size;
686 phys_addr_t low, high;
689 rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
692 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
693 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
694 if (rc || crash_size == 0)
697 if (memblock.memory.regions[0].size < crash_size) {
698 pr_info("crashkernel reservation failed: %s\n",
699 "first memory chunk must be at least crashkernel size");
703 low = crash_base ?: OLDMEM_BASE;
704 high = low + crash_size;
705 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
706 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
709 /* Find suitable area in free memory */
710 low = max_t(unsigned long, crash_size, sclp.hsa_size);
711 high = crash_base ? crash_base + crash_size : ULONG_MAX;
713 if (crash_base && crash_base < low) {
714 pr_info("crashkernel reservation failed: %s\n",
715 "crash_base too low");
718 low = crash_base ?: low;
719 crash_base = memblock_find_in_range(low, high, crash_size,
720 KEXEC_CRASH_MEM_ALIGN);
724 pr_info("crashkernel reservation failed: %s\n",
725 "no suitable area found");
729 if (register_memory_notifier(&kdump_mem_nb))
732 if (!OLDMEM_BASE && MACHINE_IS_VM)
733 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
734 crashk_res.start = crash_base;
735 crashk_res.end = crash_base + crash_size - 1;
736 memblock_remove(crash_base, crash_size);
737 pr_info("Reserving %lluMB of memory at %lluMB "
738 "for crashkernel (System RAM: %luMB)\n",
739 crash_size >> 20, crash_base >> 20,
740 (unsigned long)memblock.memory.total_size >> 20);
741 os_info_crashkernel_add(crash_base, crash_size);
746 * Reserve the initrd from being used by memblock
748 static void __init reserve_initrd(void)
750 #ifdef CONFIG_BLK_DEV_INITRD
751 if (!INITRD_START || !INITRD_SIZE)
753 initrd_start = INITRD_START;
754 initrd_end = initrd_start + INITRD_SIZE;
755 memblock_reserve(INITRD_START, INITRD_SIZE);
760 * Reserve the memory area used to pass the certificate lists
762 static void __init reserve_certificate_list(void)
764 if (ipl_cert_list_addr)
765 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
768 static void __init reserve_mem_detect_info(void)
770 unsigned long start, size;
772 get_mem_detect_reserved(&start, &size);
774 memblock_reserve(start, size);
777 static void __init free_mem_detect_info(void)
779 unsigned long start, size;
781 get_mem_detect_reserved(&start, &size);
783 memblock_free(start, size);
786 static const char * __init get_mem_info_source(void)
788 switch (mem_detect.info_source) {
789 case MEM_DETECT_SCLP_STOR_INFO:
790 return "sclp storage info";
791 case MEM_DETECT_DIAG260:
793 case MEM_DETECT_SCLP_READ_INFO:
794 return "sclp read info";
795 case MEM_DETECT_BIN_SEARCH:
796 return "binary search";
801 static void __init memblock_add_mem_detect_info(void)
803 unsigned long start, end;
806 pr_debug("physmem info source: %s (%hhd)\n",
807 get_mem_info_source(), mem_detect.info_source);
808 /* keep memblock lists close to the kernel */
809 memblock_set_bottom_up(true);
810 for_each_mem_detect_block(i, &start, &end) {
811 memblock_add(start, end - start);
812 memblock_physmem_add(start, end - start);
814 memblock_set_bottom_up(false);
815 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
820 * Check for initrd being in usable memory
822 static void __init check_initrd(void)
824 #ifdef CONFIG_BLK_DEV_INITRD
825 if (INITRD_START && INITRD_SIZE &&
826 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
827 pr_err("The initial RAM disk does not fit into the memory\n");
828 memblock_free(INITRD_START, INITRD_SIZE);
829 initrd_start = initrd_end = 0;
835 * Reserve memory used for lowcore/command line/kernel image.
837 static void __init reserve_kernel(void)
839 unsigned long start_pfn = PFN_UP(__pa(_end));
841 memblock_reserve(0, HEAD_END);
842 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
843 - (unsigned long)_stext);
844 memblock_reserve(__sdma, __edma - __sdma);
847 static void __init setup_memory(void)
849 phys_addr_t start, end;
853 * Init storage key for present memory
855 for_each_mem_range(i, &start, &end)
856 storage_key_init_range(start, end);
858 psw_set_key(PAGE_DEFAULT_KEY);
861 memblock_enforce_memory_limit(memblock_end_of_DRAM());
865 * Setup hardware capabilities.
867 static int __init setup_hwcaps(void)
869 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
874 * The store facility list bits numbers as found in the principles
875 * of operation are numbered with bit 1UL<<31 as number 0 to
876 * bit 1UL<<0 as number 31.
877 * Bit 0: instructions named N3, "backported" to esa-mode
878 * Bit 2: z/Architecture mode is active
879 * Bit 7: the store-facility-list-extended facility is installed
880 * Bit 17: the message-security assist is installed
881 * Bit 19: the long-displacement facility is installed
882 * Bit 21: the extended-immediate facility is installed
883 * Bit 22: extended-translation facility 3 is installed
884 * Bit 30: extended-translation facility 3 enhancement facility
885 * These get translated to:
886 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
887 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
888 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
889 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
891 for (i = 0; i < 6; i++)
892 if (test_facility(stfl_bits[i]))
893 elf_hwcap |= 1UL << i;
895 if (test_facility(22) && test_facility(30))
896 elf_hwcap |= HWCAP_S390_ETF3EH;
899 * Check for additional facilities with store-facility-list-extended.
900 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
901 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
902 * as stored by stfl, bits 32-xxx contain additional facilities.
903 * How many facility words are stored depends on the number of
904 * doublewords passed to the instruction. The additional facilities
906 * Bit 42: decimal floating point facility is installed
907 * Bit 44: perform floating point operation facility is installed
909 * HWCAP_S390_DFP bit 6 (42 && 44).
911 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
912 elf_hwcap |= HWCAP_S390_DFP;
915 * Huge page support HWCAP_S390_HPAGE is bit 7.
917 if (MACHINE_HAS_EDAT1)
918 elf_hwcap |= HWCAP_S390_HPAGE;
921 * 64-bit register support for 31-bit processes
922 * HWCAP_S390_HIGH_GPRS is bit 9.
924 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
927 * Transactional execution support HWCAP_S390_TE is bit 10.
930 elf_hwcap |= HWCAP_S390_TE;
933 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
934 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
935 * instead of facility bit 129.
937 if (MACHINE_HAS_VX) {
938 elf_hwcap |= HWCAP_S390_VXRS;
939 if (test_facility(134))
940 elf_hwcap |= HWCAP_S390_VXRS_EXT;
941 if (test_facility(135))
942 elf_hwcap |= HWCAP_S390_VXRS_BCD;
943 if (test_facility(148))
944 elf_hwcap |= HWCAP_S390_VXRS_EXT2;
945 if (test_facility(152))
946 elf_hwcap |= HWCAP_S390_VXRS_PDE;
948 if (test_facility(150))
949 elf_hwcap |= HWCAP_S390_SORT;
950 if (test_facility(151))
951 elf_hwcap |= HWCAP_S390_DFLT;
954 * Guarded storage support HWCAP_S390_GS is bit 12.
957 elf_hwcap |= HWCAP_S390_GS;
960 add_device_randomness(&cpu_id, sizeof(cpu_id));
961 switch (cpu_id.machine) {
964 default: /* Use "z900" as default for 64 bit kernels. */
965 strcpy(elf_platform, "z900");
969 strcpy(elf_platform, "z990");
973 strcpy(elf_platform, "z9-109");
977 strcpy(elf_platform, "z10");
981 strcpy(elf_platform, "z196");
985 strcpy(elf_platform, "zEC12");
989 strcpy(elf_platform, "z13");
993 strcpy(elf_platform, "z14");
997 strcpy(elf_platform, "z15");
1002 * Virtualization support HWCAP_INT_SIE is bit 0.
1005 int_hwcap |= HWCAP_INT_SIE;
1009 arch_initcall(setup_hwcaps);
1012 * Add system information as device randomness
1014 static void __init setup_randomness(void)
1016 struct sysinfo_3_2_2 *vmms;
1018 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
1021 panic("Failed to allocate memory for sysinfo structure\n");
1023 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1024 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1025 memblock_free((unsigned long) vmms, PAGE_SIZE);
1029 * Find the correct size for the task_struct. This depends on
1030 * the size of the struct fpu at the end of the thread_struct
1031 * which is embedded in the task_struct.
1033 static void __init setup_task_size(void)
1035 int task_size = sizeof(struct task_struct);
1037 if (!MACHINE_HAS_VX) {
1038 task_size -= sizeof(__vector128) * __NUM_VXRS;
1039 task_size += sizeof(freg_t) * __NUM_FPRS;
1041 arch_task_struct_size = task_size;
1045 * Issue diagnose 318 to set the control program name and
1048 static void __init setup_control_program_code(void)
1050 union diag318_info diag318_info = {
1055 if (!sclp.has_diag318)
1058 diag_stat_inc(DIAG_STAT_X318);
1059 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1063 * Print the component list from the IPL report
1065 static void __init log_component_list(void)
1067 struct ipl_rb_component_entry *ptr, *end;
1070 if (!early_ipl_comp_list_addr)
1072 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1073 pr_info("Linux is running with Secure-IPL enabled\n");
1075 pr_info("Linux is running with Secure-IPL disabled\n");
1076 ptr = (void *) early_ipl_comp_list_addr;
1077 end = (void *) ptr + early_ipl_comp_list_size;
1078 pr_info("The IPL report contains the following components:\n");
1080 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1081 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1082 str = "signed, verified";
1084 str = "signed, verification failed";
1088 pr_info("%016llx - %016llx (%s)\n",
1089 ptr->addr, ptr->addr + ptr->len, str);
1095 * Setup function called from init/main.c just after the banner
1099 void __init setup_arch(char **cmdline_p)
1102 * print what head.S has found out about the machine
1105 pr_info("Linux is running as a z/VM "
1106 "guest operating system in 64-bit mode\n");
1107 else if (MACHINE_IS_KVM)
1108 pr_info("Linux is running under KVM in 64-bit mode\n");
1109 else if (MACHINE_IS_LPAR)
1110 pr_info("Linux is running natively in 64-bit mode\n");
1112 pr_info("Linux is running as a guest in 64-bit mode\n");
1114 log_component_list();
1116 /* Have one command line that is parsed and saved in /proc/cmdline */
1117 /* boot_command_line has been already set up in early.c */
1118 *cmdline_p = boot_command_line;
1120 ROOT_DEV = Root_RAM0;
1122 init_mm.start_code = (unsigned long) _text;
1123 init_mm.end_code = (unsigned long) _etext;
1124 init_mm.end_data = (unsigned long) _edata;
1125 init_mm.brk = (unsigned long) _end;
1127 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1128 nospec_auto_detect();
1131 parse_early_param();
1132 #ifdef CONFIG_CRASH_DUMP
1133 /* Deactivate elfcorehdr= kernel parameter */
1134 elfcorehdr_addr = ELFCORE_ADDR_MAX;
1140 setup_control_program_code();
1142 /* Do some memory reservations *before* memory is added to memblock */
1143 reserve_above_ident_map();
1147 reserve_certificate_list();
1148 reserve_mem_detect_info();
1149 memblock_allow_resize();
1151 /* Get information about *all* installed memory */
1152 memblock_add_mem_detect_info();
1154 free_mem_detect_info();
1158 setup_ident_map_size();
1160 dma_contiguous_reserve(ident_map_size);
1162 if (MACHINE_HAS_EDAT2)
1163 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1166 reserve_crashkernel();
1167 #ifdef CONFIG_CRASH_DUMP
1169 * Be aware that smp_save_dump_cpus() triggers a system reset.
1170 * Therefore CPU and device initialization should be done afterwards.
1172 smp_save_dump_cpus();
1176 setup_lowcore_dat_off();
1177 smp_fill_possible_mask();
1178 cpu_detect_mhz_feature();
1182 topology_init_early();
1185 * Create kernel page tables and switch to virtual addressing.
1190 * After paging_init created the kernel page table, the new PSWs
1191 * in lowcore can now run with DAT enabled.
1193 setup_lowcore_dat_on();
1195 /* Setup default console */
1197 set_preferred_console();
1199 apply_alternative_instructions();
1200 if (IS_ENABLED(CONFIG_EXPOLINE))
1201 nospec_init_branches();
1203 /* Setup zfcp/nvme dump support */
1206 /* Add system specific data to the random pool */