1 // SPDX-License-Identifier: GPL-2.0
4 * Copyright IBM Corp. 1999, 2012
5 * Author(s): Hartmut Penner (hp@de.ibm.com),
6 * Martin Schwidefsky (schwidefsky@de.ibm.com)
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-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>
53 #include <asm/boot_data.h>
55 #include <asm/facility.h>
57 #include <asm/mmu_context.h>
58 #include <asm/cpcmd.h>
59 #include <asm/lowcore.h>
63 #include <asm/ptrace.h>
64 #include <asm/sections.h>
65 #include <asm/ebcdic.h>
67 #include <asm/os_info.h>
69 #include <asm/sysinfo.h>
71 #include <asm/alternative.h>
72 #include <asm/nospec-branch.h>
73 #include <asm/mem_detect.h>
80 unsigned int console_mode = 0;
81 EXPORT_SYMBOL(console_mode);
83 unsigned int console_devno = -1;
84 EXPORT_SYMBOL(console_devno);
86 unsigned int console_irq = -1;
87 EXPORT_SYMBOL(console_irq);
89 unsigned long elf_hwcap __read_mostly = 0;
90 char elf_platform[ELF_PLATFORM_SIZE];
92 unsigned long int_hwcap = 0;
94 #ifdef CONFIG_PROTECTED_VIRTUALIZATION_GUEST
95 int __bootdata_preserved(prot_virt_guest);
98 int __bootdata(noexec_disabled);
99 int __bootdata(memory_end_set);
100 unsigned long __bootdata(memory_end);
101 unsigned long __bootdata(max_physmem_end);
102 struct mem_detect_info __bootdata(mem_detect);
104 unsigned long VMALLOC_START;
105 EXPORT_SYMBOL(VMALLOC_START);
107 unsigned long VMALLOC_END;
108 EXPORT_SYMBOL(VMALLOC_END);
110 struct page *vmemmap;
111 EXPORT_SYMBOL(vmemmap);
113 unsigned long MODULES_VADDR;
114 unsigned long MODULES_END;
116 /* An array with a pointer to the lowcore of every CPU. */
117 struct lowcore *lowcore_ptr[NR_CPUS];
118 EXPORT_SYMBOL(lowcore_ptr);
121 * This is set up by the setup-routine at boot-time
122 * for S390 need to find out, what we have to setup
123 * using address 0x10400 ...
126 #include <asm/setup.h>
129 * condev= and conmode= setup parameter.
132 static int __init condev_setup(char *str)
136 vdev = simple_strtoul(str, &str, 0);
137 if (vdev >= 0 && vdev < 65536) {
138 console_devno = vdev;
144 __setup("condev=", condev_setup);
146 static void __init set_preferred_console(void)
148 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
149 add_preferred_console("ttyS", 0, NULL);
150 else if (CONSOLE_IS_3270)
151 add_preferred_console("tty3270", 0, NULL);
152 else if (CONSOLE_IS_VT220)
153 add_preferred_console("ttyS", 1, NULL);
154 else if (CONSOLE_IS_HVC)
155 add_preferred_console("hvc", 0, NULL);
158 static int __init conmode_setup(char *str)
160 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
161 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
164 #if defined(CONFIG_TN3215_CONSOLE)
165 if (strncmp(str, "3215", 5) == 0)
168 #if defined(CONFIG_TN3270_CONSOLE)
169 if (strncmp(str, "3270", 5) == 0)
172 set_preferred_console();
176 __setup("conmode=", conmode_setup);
178 static void __init conmode_default(void)
180 char query_buffer[1024];
184 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
185 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
186 ptr = strstr(query_buffer, "SUBCHANNEL =");
187 console_irq = simple_strtoul(ptr + 13, NULL, 16);
188 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
189 ptr = strstr(query_buffer, "CONMODE");
191 * Set the conmode to 3215 so that the device recognition
192 * will set the cu_type of the console to 3215. If the
193 * conmode is 3270 and we don't set it back then both
194 * 3215 and the 3270 driver will try to access the console
195 * device (3215 as console and 3270 as normal tty).
197 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
199 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
204 if (strncmp(ptr + 8, "3270", 4) == 0) {
205 #if defined(CONFIG_TN3270_CONSOLE)
207 #elif defined(CONFIG_TN3215_CONSOLE)
209 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
212 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
213 #if defined(CONFIG_TN3215_CONSOLE)
215 #elif defined(CONFIG_TN3270_CONSOLE)
217 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
221 } else if (MACHINE_IS_KVM) {
222 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
224 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
229 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
233 if (IS_ENABLED(CONFIG_VT) && IS_ENABLED(CONFIG_DUMMY_CONSOLE))
234 conswitchp = &dummy_con;
237 #ifdef CONFIG_CRASH_DUMP
238 static void __init setup_zfcpdump(void)
240 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
244 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
245 console_loglevel = 2;
248 static inline void setup_zfcpdump(void) {}
249 #endif /* CONFIG_CRASH_DUMP */
252 * Reboot, halt and power_off stubs. They just call _machine_restart,
253 * _machine_halt or _machine_power_off.
256 void machine_restart(char *command)
258 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
260 * Only unblank the console if we are called in enabled
261 * context or a bust_spinlocks cleared the way for us.
264 _machine_restart(command);
267 void machine_halt(void)
269 if (!in_interrupt() || oops_in_progress)
271 * Only unblank the console if we are called in enabled
272 * context or a bust_spinlocks cleared the way for us.
278 void machine_power_off(void)
280 if (!in_interrupt() || oops_in_progress)
282 * Only unblank the console if we are called in enabled
283 * context or a bust_spinlocks cleared the way for us.
286 _machine_power_off();
290 * Dummy power off function.
292 void (*pm_power_off)(void) = machine_power_off;
293 EXPORT_SYMBOL_GPL(pm_power_off);
295 static int __init parse_vmalloc(char *arg)
299 VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
302 early_param("vmalloc", parse_vmalloc);
304 void *restart_stack __section(.data);
306 unsigned long stack_alloc(void)
308 #ifdef CONFIG_VMAP_STACK
309 return (unsigned long)
310 __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
311 VMALLOC_START, VMALLOC_END,
313 PAGE_KERNEL, 0, NUMA_NO_NODE,
314 __builtin_return_address(0));
316 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
320 void stack_free(unsigned long stack)
322 #ifdef CONFIG_VMAP_STACK
323 vfree((void *) stack);
325 free_pages(stack, THREAD_SIZE_ORDER);
329 int __init arch_early_irq_init(void)
333 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
335 panic("Couldn't allocate async stack");
336 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
340 static int __init async_stack_realloc(void)
342 unsigned long old, new;
344 old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
347 panic("Couldn't allocate async stack");
348 S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
349 free_pages(old, THREAD_SIZE_ORDER);
352 early_initcall(async_stack_realloc);
354 void __init arch_call_rest_init(void)
356 struct stack_frame *frame;
359 stack = stack_alloc();
361 panic("Couldn't allocate kernel stack");
362 current->stack = (void *) stack;
363 #ifdef CONFIG_VMAP_STACK
364 current->stack_vm_area = (void *) stack;
366 set_task_stack_end_magic(current);
367 stack += STACK_INIT_OFFSET;
368 S390_lowcore.kernel_stack = stack;
369 frame = (struct stack_frame *) stack;
370 memset(frame, 0, sizeof(*frame));
371 /* Branch to rest_init on the new stack, never returns */
373 " la 15,0(%[_frame])\n"
375 : : [_frame] "a" (frame));
378 static void __init setup_lowcore_dat_off(void)
383 * Setup lowcore for boot cpu
385 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
386 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
388 panic("%s: Failed to allocate %zu bytes align=%zx\n",
389 __func__, sizeof(*lc), sizeof(*lc));
391 lc->restart_psw.mask = PSW_KERNEL_BITS;
392 lc->restart_psw.addr = (unsigned long) restart_int_handler;
393 lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
394 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
395 lc->svc_new_psw.mask = PSW_KERNEL_BITS |
396 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
397 lc->svc_new_psw.addr = (unsigned long) system_call;
398 lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
399 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
400 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
401 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
402 lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
403 lc->io_new_psw.addr = (unsigned long) io_int_handler;
404 lc->clock_comparator = clock_comparator_max;
405 lc->nodat_stack = ((unsigned long) &init_thread_union)
406 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
407 lc->current_task = (unsigned long)&init_task;
409 lc->machine_flags = S390_lowcore.machine_flags;
410 lc->preempt_count = S390_lowcore.preempt_count;
411 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
412 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
413 sizeof(lc->stfle_fac_list));
414 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
415 sizeof(lc->alt_stfle_fac_list));
416 nmi_alloc_boot_cpu(lc);
417 vdso_alloc_boot_cpu(lc);
418 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
419 lc->async_enter_timer = S390_lowcore.async_enter_timer;
420 lc->exit_timer = S390_lowcore.exit_timer;
421 lc->user_timer = S390_lowcore.user_timer;
422 lc->system_timer = S390_lowcore.system_timer;
423 lc->steal_timer = S390_lowcore.steal_timer;
424 lc->last_update_timer = S390_lowcore.last_update_timer;
425 lc->last_update_clock = S390_lowcore.last_update_clock;
428 * Allocate the global restart stack which is the same for
429 * all CPUs in cast *one* of them does a PSW restart.
431 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
433 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
434 __func__, THREAD_SIZE, THREAD_SIZE);
435 restart_stack += STACK_INIT_OFFSET;
438 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
439 * restart data to the absolute zero lowcore. This is necessary if
440 * PSW restart is done on an offline CPU that has lowcore zero.
442 lc->restart_stack = (unsigned long) restart_stack;
443 lc->restart_fn = (unsigned long) do_restart;
444 lc->restart_data = 0;
445 lc->restart_source = -1UL;
447 /* Setup absolute zero lowcore */
448 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
449 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
450 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
451 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
452 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
455 lc->spinlock_lockval = arch_spin_lockval(0);
456 lc->spinlock_index = 0;
457 arch_spin_lock_setup(0);
459 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
461 set_prefix((u32)(unsigned long) lc);
465 static void __init setup_lowcore_dat_on(void)
467 __ctl_clear_bit(0, 28);
468 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
469 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
470 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
471 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
472 __ctl_set_bit(0, 28);
475 static struct resource code_resource = {
476 .name = "Kernel code",
477 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
480 static struct resource data_resource = {
481 .name = "Kernel data",
482 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
485 static struct resource bss_resource = {
486 .name = "Kernel bss",
487 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
490 static struct resource __initdata *standard_resources[] = {
496 static void __init setup_resources(void)
498 struct resource *res, *std_res, *sub_res;
499 struct memblock_region *reg;
502 code_resource.start = (unsigned long) _text;
503 code_resource.end = (unsigned long) _etext - 1;
504 data_resource.start = (unsigned long) _etext;
505 data_resource.end = (unsigned long) _edata - 1;
506 bss_resource.start = (unsigned long) __bss_start;
507 bss_resource.end = (unsigned long) __bss_stop - 1;
509 for_each_memblock(memory, reg) {
510 res = memblock_alloc(sizeof(*res), 8);
512 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
513 __func__, sizeof(*res), 8);
514 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
516 res->name = "System RAM";
517 res->start = reg->base;
518 res->end = reg->base + reg->size - 1;
519 request_resource(&iomem_resource, res);
521 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
522 std_res = standard_resources[j];
523 if (std_res->start < res->start ||
524 std_res->start > res->end)
526 if (std_res->end > res->end) {
527 sub_res = memblock_alloc(sizeof(*sub_res), 8);
529 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
530 __func__, sizeof(*sub_res), 8);
532 sub_res->end = res->end;
533 std_res->start = res->end + 1;
534 request_resource(res, sub_res);
536 request_resource(res, std_res);
540 #ifdef CONFIG_CRASH_DUMP
542 * Re-add removed crash kernel memory as reserved memory. This makes
543 * sure it will be mapped with the identity mapping and struct pages
544 * will be created, so it can be resized later on.
545 * However add it later since the crash kernel resource should not be
546 * part of the System RAM resource.
548 if (crashk_res.end) {
549 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
550 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
551 insert_resource(&iomem_resource, &crashk_res);
556 static void __init setup_memory_end(void)
558 unsigned long vmax, vmalloc_size, tmp;
560 /* Choose kernel address space layout: 3 or 4 levels. */
561 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
562 if (IS_ENABLED(CONFIG_KASAN)) {
563 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
567 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
568 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
569 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
570 vmax = _REGION2_SIZE; /* 3-level kernel page table */
572 vmax = _REGION1_SIZE; /* 4-level kernel page table */
575 /* module area is at the end of the kernel address space. */
577 MODULES_VADDR = MODULES_END - MODULES_LEN;
578 VMALLOC_END = MODULES_VADDR;
579 VMALLOC_START = VMALLOC_END - vmalloc_size;
581 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
582 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
583 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
584 tmp = SECTION_ALIGN_UP(tmp);
585 tmp = VMALLOC_START - tmp * sizeof(struct page);
586 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
587 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
588 vmemmap = (struct page *) tmp;
590 /* Take care that memory_end is set and <= vmemmap */
591 memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
593 /* fit in kasan shadow memory region between 1:1 and vmemmap */
594 memory_end = min(memory_end, KASAN_SHADOW_START);
595 vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
597 max_pfn = max_low_pfn = PFN_DOWN(memory_end);
598 memblock_remove(memory_end, ULONG_MAX);
600 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
603 #ifdef CONFIG_CRASH_DUMP
606 * When kdump is enabled, we have to ensure that no memory from
607 * the area [0 - crashkernel memory size] and
608 * [crashk_res.start - crashk_res.end] is set offline.
610 static int kdump_mem_notifier(struct notifier_block *nb,
611 unsigned long action, void *data)
613 struct memory_notify *arg = data;
615 if (action != MEM_GOING_OFFLINE)
617 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
619 if (arg->start_pfn > PFN_DOWN(crashk_res.end))
621 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
626 static struct notifier_block kdump_mem_nb = {
627 .notifier_call = kdump_mem_notifier,
633 * Make sure that the area behind memory_end is protected
635 static void reserve_memory_end(void)
638 memblock_reserve(memory_end, ULONG_MAX);
642 * Make sure that oldmem, where the dump is stored, is protected
644 static void reserve_oldmem(void)
646 #ifdef CONFIG_CRASH_DUMP
648 /* Forget all memory above the running kdump system */
649 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
654 * Make sure that oldmem, where the dump is stored, is protected
656 static void remove_oldmem(void)
658 #ifdef CONFIG_CRASH_DUMP
660 /* Forget all memory above the running kdump system */
661 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
666 * Reserve memory for kdump kernel to be loaded with kexec
668 static void __init reserve_crashkernel(void)
670 #ifdef CONFIG_CRASH_DUMP
671 unsigned long long crash_base, crash_size;
672 phys_addr_t low, high;
675 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
678 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
679 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
680 if (rc || crash_size == 0)
683 if (memblock.memory.regions[0].size < crash_size) {
684 pr_info("crashkernel reservation failed: %s\n",
685 "first memory chunk must be at least crashkernel size");
689 low = crash_base ?: OLDMEM_BASE;
690 high = low + crash_size;
691 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
692 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
695 /* Find suitable area in free memory */
696 low = max_t(unsigned long, crash_size, sclp.hsa_size);
697 high = crash_base ? crash_base + crash_size : ULONG_MAX;
699 if (crash_base && crash_base < low) {
700 pr_info("crashkernel reservation failed: %s\n",
701 "crash_base too low");
704 low = crash_base ?: low;
705 crash_base = memblock_find_in_range(low, high, crash_size,
706 KEXEC_CRASH_MEM_ALIGN);
710 pr_info("crashkernel reservation failed: %s\n",
711 "no suitable area found");
715 if (register_memory_notifier(&kdump_mem_nb))
718 if (!OLDMEM_BASE && MACHINE_IS_VM)
719 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
720 crashk_res.start = crash_base;
721 crashk_res.end = crash_base + crash_size - 1;
722 memblock_remove(crash_base, crash_size);
723 pr_info("Reserving %lluMB of memory at %lluMB "
724 "for crashkernel (System RAM: %luMB)\n",
725 crash_size >> 20, crash_base >> 20,
726 (unsigned long)memblock.memory.total_size >> 20);
727 os_info_crashkernel_add(crash_base, crash_size);
732 * Reserve the initrd from being used by memblock
734 static void __init reserve_initrd(void)
736 #ifdef CONFIG_BLK_DEV_INITRD
737 if (!INITRD_START || !INITRD_SIZE)
739 initrd_start = INITRD_START;
740 initrd_end = initrd_start + INITRD_SIZE;
741 memblock_reserve(INITRD_START, INITRD_SIZE);
746 * Reserve the memory area used to pass the certificate lists
748 static void __init reserve_certificate_list(void)
750 if (ipl_cert_list_addr)
751 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
754 static void __init reserve_mem_detect_info(void)
756 unsigned long start, size;
758 get_mem_detect_reserved(&start, &size);
760 memblock_reserve(start, size);
763 static void __init free_mem_detect_info(void)
765 unsigned long start, size;
767 get_mem_detect_reserved(&start, &size);
769 memblock_free(start, size);
772 static void __init memblock_physmem_add(phys_addr_t start, phys_addr_t size)
774 memblock_dbg("memblock_physmem_add: [%#016llx-%#016llx]\n",
775 start, start + size - 1);
776 memblock_add_range(&memblock.memory, start, size, 0, 0);
777 memblock_add_range(&memblock.physmem, start, size, 0, 0);
780 static const char * __init get_mem_info_source(void)
782 switch (mem_detect.info_source) {
783 case MEM_DETECT_SCLP_STOR_INFO:
784 return "sclp storage info";
785 case MEM_DETECT_DIAG260:
787 case MEM_DETECT_SCLP_READ_INFO:
788 return "sclp read info";
789 case MEM_DETECT_BIN_SEARCH:
790 return "binary search";
795 static void __init memblock_add_mem_detect_info(void)
797 unsigned long start, end;
800 memblock_dbg("physmem info source: %s (%hhd)\n",
801 get_mem_info_source(), mem_detect.info_source);
802 /* keep memblock lists close to the kernel */
803 memblock_set_bottom_up(true);
804 for_each_mem_detect_block(i, &start, &end)
805 memblock_physmem_add(start, end - start);
806 memblock_set_bottom_up(false);
811 * Check for initrd being in usable memory
813 static void __init check_initrd(void)
815 #ifdef CONFIG_BLK_DEV_INITRD
816 if (INITRD_START && INITRD_SIZE &&
817 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
818 pr_err("The initial RAM disk does not fit into the memory\n");
819 memblock_free(INITRD_START, INITRD_SIZE);
820 initrd_start = initrd_end = 0;
826 * Reserve memory used for lowcore/command line/kernel image.
828 static void __init reserve_kernel(void)
830 unsigned long start_pfn = PFN_UP(__pa(_end));
832 memblock_reserve(0, HEAD_END);
833 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
834 - (unsigned long)_stext);
837 static void __init setup_memory(void)
839 struct memblock_region *reg;
842 * Init storage key for present memory
844 for_each_memblock(memory, reg) {
845 storage_key_init_range(reg->base, reg->base + reg->size);
847 psw_set_key(PAGE_DEFAULT_KEY);
850 memblock_enforce_memory_limit(memblock_end_of_DRAM());
854 * Setup hardware capabilities.
856 static int __init setup_hwcaps(void)
858 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
863 * The store facility list bits numbers as found in the principles
864 * of operation are numbered with bit 1UL<<31 as number 0 to
865 * bit 1UL<<0 as number 31.
866 * Bit 0: instructions named N3, "backported" to esa-mode
867 * Bit 2: z/Architecture mode is active
868 * Bit 7: the store-facility-list-extended facility is installed
869 * Bit 17: the message-security assist is installed
870 * Bit 19: the long-displacement facility is installed
871 * Bit 21: the extended-immediate facility is installed
872 * Bit 22: extended-translation facility 3 is installed
873 * Bit 30: extended-translation facility 3 enhancement facility
874 * These get translated to:
875 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
876 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
877 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
878 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
880 for (i = 0; i < 6; i++)
881 if (test_facility(stfl_bits[i]))
882 elf_hwcap |= 1UL << i;
884 if (test_facility(22) && test_facility(30))
885 elf_hwcap |= HWCAP_S390_ETF3EH;
888 * Check for additional facilities with store-facility-list-extended.
889 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
890 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
891 * as stored by stfl, bits 32-xxx contain additional facilities.
892 * How many facility words are stored depends on the number of
893 * doublewords passed to the instruction. The additional facilities
895 * Bit 42: decimal floating point facility is installed
896 * Bit 44: perform floating point operation facility is installed
898 * HWCAP_S390_DFP bit 6 (42 && 44).
900 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
901 elf_hwcap |= HWCAP_S390_DFP;
904 * Huge page support HWCAP_S390_HPAGE is bit 7.
906 if (MACHINE_HAS_EDAT1)
907 elf_hwcap |= HWCAP_S390_HPAGE;
910 * 64-bit register support for 31-bit processes
911 * HWCAP_S390_HIGH_GPRS is bit 9.
913 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
916 * Transactional execution support HWCAP_S390_TE is bit 10.
919 elf_hwcap |= HWCAP_S390_TE;
922 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
923 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
924 * instead of facility bit 129.
926 if (MACHINE_HAS_VX) {
927 elf_hwcap |= HWCAP_S390_VXRS;
928 if (test_facility(134))
929 elf_hwcap |= HWCAP_S390_VXRS_EXT;
930 if (test_facility(135))
931 elf_hwcap |= HWCAP_S390_VXRS_BCD;
932 if (test_facility(148))
933 elf_hwcap |= HWCAP_S390_VXRS_EXT2;
934 if (test_facility(152))
935 elf_hwcap |= HWCAP_S390_VXRS_PDE;
937 if (test_facility(150))
938 elf_hwcap |= HWCAP_S390_SORT;
939 if (test_facility(151))
940 elf_hwcap |= HWCAP_S390_DFLT;
943 * Guarded storage support HWCAP_S390_GS is bit 12.
946 elf_hwcap |= HWCAP_S390_GS;
949 add_device_randomness(&cpu_id, sizeof(cpu_id));
950 switch (cpu_id.machine) {
953 default: /* Use "z900" as default for 64 bit kernels. */
954 strcpy(elf_platform, "z900");
958 strcpy(elf_platform, "z990");
962 strcpy(elf_platform, "z9-109");
966 strcpy(elf_platform, "z10");
970 strcpy(elf_platform, "z196");
974 strcpy(elf_platform, "zEC12");
978 strcpy(elf_platform, "z13");
982 strcpy(elf_platform, "z14");
987 * Virtualization support HWCAP_INT_SIE is bit 0.
990 int_hwcap |= HWCAP_INT_SIE;
994 arch_initcall(setup_hwcaps);
997 * Add system information as device randomness
999 static void __init setup_randomness(void)
1001 struct sysinfo_3_2_2 *vmms;
1003 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
1006 panic("Failed to allocate memory for sysinfo structure\n");
1008 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1009 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1010 memblock_free((unsigned long) vmms, PAGE_SIZE);
1014 * Find the correct size for the task_struct. This depends on
1015 * the size of the struct fpu at the end of the thread_struct
1016 * which is embedded in the task_struct.
1018 static void __init setup_task_size(void)
1020 int task_size = sizeof(struct task_struct);
1022 if (!MACHINE_HAS_VX) {
1023 task_size -= sizeof(__vector128) * __NUM_VXRS;
1024 task_size += sizeof(freg_t) * __NUM_FPRS;
1026 arch_task_struct_size = task_size;
1030 * Issue diagnose 318 to set the control program name and
1033 static void __init setup_control_program_code(void)
1035 union diag318_info diag318_info = {
1041 if (!sclp.has_diag318)
1044 diag_stat_inc(DIAG_STAT_X318);
1045 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1049 * Print the component list from the IPL report
1051 static void __init log_component_list(void)
1053 struct ipl_rb_component_entry *ptr, *end;
1056 if (!early_ipl_comp_list_addr)
1058 if (ipl_block.hdr.flags & IPL_PL_FLAG_IPLSR)
1059 pr_info("Linux is running with Secure-IPL enabled\n");
1061 pr_info("Linux is running with Secure-IPL disabled\n");
1062 ptr = (void *) early_ipl_comp_list_addr;
1063 end = (void *) ptr + early_ipl_comp_list_size;
1064 pr_info("The IPL report contains the following components:\n");
1066 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1067 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1068 str = "signed, verified";
1070 str = "signed, verification failed";
1074 pr_info("%016llx - %016llx (%s)\n",
1075 ptr->addr, ptr->addr + ptr->len, str);
1081 * Setup function called from init/main.c just after the banner
1085 void __init setup_arch(char **cmdline_p)
1088 * print what head.S has found out about the machine
1091 pr_info("Linux is running as a z/VM "
1092 "guest operating system in 64-bit mode\n");
1093 else if (MACHINE_IS_KVM)
1094 pr_info("Linux is running under KVM in 64-bit mode\n");
1095 else if (MACHINE_IS_LPAR)
1096 pr_info("Linux is running natively in 64-bit mode\n");
1098 pr_info("Linux is running as a guest in 64-bit mode\n");
1100 log_component_list();
1102 /* Have one command line that is parsed and saved in /proc/cmdline */
1103 /* boot_command_line has been already set up in early.c */
1104 *cmdline_p = boot_command_line;
1106 ROOT_DEV = Root_RAM0;
1108 /* Is init_mm really needed? */
1109 init_mm.start_code = PAGE_OFFSET;
1110 init_mm.end_code = (unsigned long) _etext;
1111 init_mm.end_data = (unsigned long) _edata;
1112 init_mm.brk = (unsigned long) _end;
1114 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1115 nospec_auto_detect();
1117 parse_early_param();
1118 #ifdef CONFIG_CRASH_DUMP
1119 /* Deactivate elfcorehdr= kernel parameter */
1120 elfcorehdr_addr = ELFCORE_ADDR_MAX;
1126 setup_control_program_code();
1128 /* Do some memory reservations *before* memory is added to memblock */
1129 reserve_memory_end();
1133 reserve_certificate_list();
1134 reserve_mem_detect_info();
1135 memblock_allow_resize();
1137 /* Get information about *all* installed memory */
1138 memblock_add_mem_detect_info();
1140 free_mem_detect_info();
1144 * Make sure all chunks are MAX_ORDER aligned so we don't need the
1145 * extra checks that HOLES_IN_ZONE would require.
1147 * Is this still required?
1149 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
1153 dma_contiguous_reserve(memory_end);
1157 reserve_crashkernel();
1158 #ifdef CONFIG_CRASH_DUMP
1160 * Be aware that smp_save_dump_cpus() triggers a system reset.
1161 * Therefore CPU and device initialization should be done afterwards.
1163 smp_save_dump_cpus();
1167 setup_lowcore_dat_off();
1168 smp_fill_possible_mask();
1169 cpu_detect_mhz_feature();
1173 topology_init_early();
1176 * Create kernel page tables and switch to virtual addressing.
1181 * After paging_init created the kernel page table, the new PSWs
1182 * in lowcore can now run with DAT enabled.
1184 setup_lowcore_dat_on();
1186 /* Setup default console */
1188 set_preferred_console();
1190 apply_alternative_instructions();
1191 if (IS_ENABLED(CONFIG_EXPOLINE))
1192 nospec_init_branches();
1194 /* Setup zfcpdump support */
1197 /* Add system specific data to the random pool */