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-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 struct mem_detect_info __bootdata(mem_detect);
101 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table);
102 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table);
103 unsigned long __bootdata_preserved(__stext_dma);
104 unsigned long __bootdata_preserved(__etext_dma);
105 unsigned long __bootdata_preserved(__sdma);
106 unsigned long __bootdata_preserved(__edma);
107 unsigned long __bootdata_preserved(__kaslr_offset);
108 unsigned int __bootdata_preserved(zlib_dfltcc_support);
109 EXPORT_SYMBOL(zlib_dfltcc_support);
110 u64 __bootdata_preserved(stfle_fac_list[16]);
111 EXPORT_SYMBOL(stfle_fac_list);
112 u64 __bootdata_preserved(alt_stfle_fac_list[16]);
114 unsigned long VMALLOC_START;
115 EXPORT_SYMBOL(VMALLOC_START);
117 unsigned long VMALLOC_END;
118 EXPORT_SYMBOL(VMALLOC_END);
120 struct page *vmemmap;
121 EXPORT_SYMBOL(vmemmap);
122 unsigned long vmemmap_size;
124 unsigned long MODULES_VADDR;
125 unsigned long MODULES_END;
127 /* An array with a pointer to the lowcore of every CPU. */
128 struct lowcore *lowcore_ptr[NR_CPUS];
129 EXPORT_SYMBOL(lowcore_ptr);
132 * The Write Back bit position in the physaddr is given by the SLPC PCI.
133 * Leaving the mask zero always uses write through which is safe
135 unsigned long mio_wb_bit_mask __ro_after_init;
138 * This is set up by the setup-routine at boot-time
139 * for S390 need to find out, what we have to setup
140 * using address 0x10400 ...
143 #include <asm/setup.h>
146 * condev= and conmode= setup parameter.
149 static int __init condev_setup(char *str)
153 vdev = simple_strtoul(str, &str, 0);
154 if (vdev >= 0 && vdev < 65536) {
155 console_devno = vdev;
161 __setup("condev=", condev_setup);
163 static void __init set_preferred_console(void)
165 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
166 add_preferred_console("ttyS", 0, NULL);
167 else if (CONSOLE_IS_3270)
168 add_preferred_console("tty3270", 0, NULL);
169 else if (CONSOLE_IS_VT220)
170 add_preferred_console("ttysclp", 0, NULL);
171 else if (CONSOLE_IS_HVC)
172 add_preferred_console("hvc", 0, NULL);
175 static int __init conmode_setup(char *str)
177 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
178 if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
181 #if defined(CONFIG_TN3215_CONSOLE)
182 if (!strcmp(str, "3215"))
185 #if defined(CONFIG_TN3270_CONSOLE)
186 if (!strcmp(str, "3270"))
189 set_preferred_console();
193 __setup("conmode=", conmode_setup);
195 static void __init conmode_default(void)
197 char query_buffer[1024];
201 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
202 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
203 ptr = strstr(query_buffer, "SUBCHANNEL =");
204 console_irq = simple_strtoul(ptr + 13, NULL, 16);
205 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
206 ptr = strstr(query_buffer, "CONMODE");
208 * Set the conmode to 3215 so that the device recognition
209 * will set the cu_type of the console to 3215. If the
210 * conmode is 3270 and we don't set it back then both
211 * 3215 and the 3270 driver will try to access the console
212 * device (3215 as console and 3270 as normal tty).
214 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
216 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
221 if (str_has_prefix(ptr + 8, "3270")) {
222 #if defined(CONFIG_TN3270_CONSOLE)
224 #elif defined(CONFIG_TN3215_CONSOLE)
226 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
229 } else if (str_has_prefix(ptr + 8, "3215")) {
230 #if defined(CONFIG_TN3215_CONSOLE)
232 #elif defined(CONFIG_TN3270_CONSOLE)
234 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
238 } else if (MACHINE_IS_KVM) {
239 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
241 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
246 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
252 #ifdef CONFIG_CRASH_DUMP
253 static void __init setup_zfcpdump(void)
255 if (!is_ipl_type_dump())
259 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
260 console_loglevel = 2;
263 static inline void setup_zfcpdump(void) {}
264 #endif /* CONFIG_CRASH_DUMP */
267 * Reboot, halt and power_off stubs. They just call _machine_restart,
268 * _machine_halt or _machine_power_off.
271 void machine_restart(char *command)
273 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
275 * Only unblank the console if we are called in enabled
276 * context or a bust_spinlocks cleared the way for us.
279 _machine_restart(command);
282 void machine_halt(void)
284 if (!in_interrupt() || oops_in_progress)
286 * Only unblank the console if we are called in enabled
287 * context or a bust_spinlocks cleared the way for us.
293 void machine_power_off(void)
295 if (!in_interrupt() || oops_in_progress)
297 * Only unblank the console if we are called in enabled
298 * context or a bust_spinlocks cleared the way for us.
301 _machine_power_off();
305 * Dummy power off function.
307 void (*pm_power_off)(void) = machine_power_off;
308 EXPORT_SYMBOL_GPL(pm_power_off);
312 unsigned long stack_alloc(void)
314 #ifdef CONFIG_VMAP_STACK
315 return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE,
316 THREADINFO_GFP, NUMA_NO_NODE,
317 __builtin_return_address(0));
319 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
323 void stack_free(unsigned long stack)
325 #ifdef CONFIG_VMAP_STACK
326 vfree((void *) stack);
328 free_pages(stack, THREAD_SIZE_ORDER);
332 int __init arch_early_irq_init(void)
336 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
338 panic("Couldn't allocate async stack");
339 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
343 void __init arch_call_rest_init(void)
347 stack = stack_alloc();
349 panic("Couldn't allocate kernel stack");
350 current->stack = (void *) stack;
351 #ifdef CONFIG_VMAP_STACK
352 current->stack_vm_area = (void *) stack;
354 set_task_stack_end_magic(current);
355 stack += STACK_INIT_OFFSET;
356 S390_lowcore.kernel_stack = stack;
357 CALL_ON_STACK_NORETURN(rest_init, stack);
360 static void __init setup_lowcore_dat_off(void)
362 unsigned long int_psw_mask = PSW_KERNEL_BITS;
363 unsigned long mcck_stack;
366 if (IS_ENABLED(CONFIG_KASAN))
367 int_psw_mask |= PSW_MASK_DAT;
370 * Setup lowcore for boot cpu
372 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
373 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
375 panic("%s: Failed to allocate %zu bytes align=%zx\n",
376 __func__, sizeof(*lc), sizeof(*lc));
378 lc->restart_psw.mask = PSW_KERNEL_BITS;
379 lc->restart_psw.addr = (unsigned long) restart_int_handler;
380 lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
381 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
382 lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
383 lc->svc_new_psw.addr = (unsigned long) system_call;
384 lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
385 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
386 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
387 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
388 lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
389 lc->io_new_psw.addr = (unsigned long) io_int_handler;
390 lc->clock_comparator = clock_comparator_max;
391 lc->nodat_stack = ((unsigned long) &init_thread_union)
392 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
393 lc->current_task = (unsigned long)&init_task;
395 lc->machine_flags = S390_lowcore.machine_flags;
396 lc->preempt_count = S390_lowcore.preempt_count;
397 nmi_alloc_boot_cpu(lc);
398 lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
399 lc->exit_timer = S390_lowcore.exit_timer;
400 lc->user_timer = S390_lowcore.user_timer;
401 lc->system_timer = S390_lowcore.system_timer;
402 lc->steal_timer = S390_lowcore.steal_timer;
403 lc->last_update_timer = S390_lowcore.last_update_timer;
404 lc->last_update_clock = S390_lowcore.last_update_clock;
407 * Allocate the global restart stack which is the same for
408 * all CPUs in cast *one* of them does a PSW restart.
410 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
412 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
413 __func__, THREAD_SIZE, THREAD_SIZE);
414 restart_stack += STACK_INIT_OFFSET;
417 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
418 * restart data to the absolute zero lowcore. This is necessary if
419 * PSW restart is done on an offline CPU that has lowcore zero.
421 lc->restart_stack = (unsigned long) restart_stack;
422 lc->restart_fn = (unsigned long) do_restart;
423 lc->restart_data = 0;
424 lc->restart_source = -1UL;
426 mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
428 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
429 __func__, THREAD_SIZE, THREAD_SIZE);
430 lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
432 /* Setup absolute zero lowcore */
433 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
434 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
435 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
436 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
437 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
439 lc->spinlock_lockval = arch_spin_lockval(0);
440 lc->spinlock_index = 0;
441 arch_spin_lock_setup(0);
442 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
443 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
444 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
446 set_prefix((u32)(unsigned long) lc);
450 static void __init setup_lowcore_dat_on(void)
452 __ctl_clear_bit(0, 28);
453 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
454 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
455 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
456 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
457 __ctl_set_bit(0, 28);
460 static struct resource code_resource = {
461 .name = "Kernel code",
462 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
465 static struct resource data_resource = {
466 .name = "Kernel data",
467 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
470 static struct resource bss_resource = {
471 .name = "Kernel bss",
472 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
475 static struct resource __initdata *standard_resources[] = {
481 static void __init setup_resources(void)
483 struct resource *res, *std_res, *sub_res;
484 phys_addr_t start, end;
488 code_resource.start = (unsigned long) _text;
489 code_resource.end = (unsigned long) _etext - 1;
490 data_resource.start = (unsigned long) _etext;
491 data_resource.end = (unsigned long) _edata - 1;
492 bss_resource.start = (unsigned long) __bss_start;
493 bss_resource.end = (unsigned long) __bss_stop - 1;
495 for_each_mem_range(i, &start, &end) {
496 res = memblock_alloc(sizeof(*res), 8);
498 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
499 __func__, sizeof(*res), 8);
500 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
502 res->name = "System RAM";
505 * In memblock, end points to the first byte after the
506 * range while in resourses, end points to the last byte in
510 request_resource(&iomem_resource, res);
512 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
513 std_res = standard_resources[j];
514 if (std_res->start < res->start ||
515 std_res->start > res->end)
517 if (std_res->end > res->end) {
518 sub_res = memblock_alloc(sizeof(*sub_res), 8);
520 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
521 __func__, sizeof(*sub_res), 8);
523 sub_res->end = res->end;
524 std_res->start = res->end + 1;
525 request_resource(res, sub_res);
527 request_resource(res, std_res);
531 #ifdef CONFIG_CRASH_DUMP
533 * Re-add removed crash kernel memory as reserved memory. This makes
534 * sure it will be mapped with the identity mapping and struct pages
535 * will be created, so it can be resized later on.
536 * However add it later since the crash kernel resource should not be
537 * part of the System RAM resource.
539 if (crashk_res.end) {
540 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
541 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
542 insert_resource(&iomem_resource, &crashk_res);
547 static void __init setup_memory_end(void)
549 memblock_remove(ident_map_size, ULONG_MAX);
550 max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
551 pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
554 #ifdef CONFIG_CRASH_DUMP
557 * When kdump is enabled, we have to ensure that no memory from the area
558 * [0 - crashkernel memory size] is set offline - it will be exchanged with
559 * the crashkernel memory region when kdump is triggered. The crashkernel
560 * memory region can never get offlined (pages are unmovable).
562 static int kdump_mem_notifier(struct notifier_block *nb,
563 unsigned long action, void *data)
565 struct memory_notify *arg = data;
567 if (action != MEM_GOING_OFFLINE)
569 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
574 static struct notifier_block kdump_mem_nb = {
575 .notifier_call = kdump_mem_notifier,
581 * Make sure that the area above identity mapping is protected
583 static void __init reserve_above_ident_map(void)
585 memblock_reserve(ident_map_size, ULONG_MAX);
589 * Make sure that oldmem, where the dump is stored, is protected
591 static void __init reserve_oldmem(void)
593 #ifdef CONFIG_CRASH_DUMP
595 /* Forget all memory above the running kdump system */
596 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
601 * Make sure that oldmem, where the dump is stored, is protected
603 static void __init remove_oldmem(void)
605 #ifdef CONFIG_CRASH_DUMP
607 /* Forget all memory above the running kdump system */
608 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
613 * Reserve memory for kdump kernel to be loaded with kexec
615 static void __init reserve_crashkernel(void)
617 #ifdef CONFIG_CRASH_DUMP
618 unsigned long long crash_base, crash_size;
619 phys_addr_t low, high;
622 rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
625 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
626 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
627 if (rc || crash_size == 0)
630 if (memblock.memory.regions[0].size < crash_size) {
631 pr_info("crashkernel reservation failed: %s\n",
632 "first memory chunk must be at least crashkernel size");
636 low = crash_base ?: OLDMEM_BASE;
637 high = low + crash_size;
638 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
639 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
642 /* Find suitable area in free memory */
643 low = max_t(unsigned long, crash_size, sclp.hsa_size);
644 high = crash_base ? crash_base + crash_size : ULONG_MAX;
646 if (crash_base && crash_base < low) {
647 pr_info("crashkernel reservation failed: %s\n",
648 "crash_base too low");
651 low = crash_base ?: low;
652 crash_base = memblock_find_in_range(low, high, crash_size,
653 KEXEC_CRASH_MEM_ALIGN);
657 pr_info("crashkernel reservation failed: %s\n",
658 "no suitable area found");
662 if (register_memory_notifier(&kdump_mem_nb))
665 if (!OLDMEM_BASE && MACHINE_IS_VM)
666 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
667 crashk_res.start = crash_base;
668 crashk_res.end = crash_base + crash_size - 1;
669 memblock_remove(crash_base, crash_size);
670 pr_info("Reserving %lluMB of memory at %lluMB "
671 "for crashkernel (System RAM: %luMB)\n",
672 crash_size >> 20, crash_base >> 20,
673 (unsigned long)memblock.memory.total_size >> 20);
674 os_info_crashkernel_add(crash_base, crash_size);
679 * Reserve the initrd from being used by memblock
681 static void __init reserve_initrd(void)
683 #ifdef CONFIG_BLK_DEV_INITRD
684 if (!INITRD_START || !INITRD_SIZE)
686 initrd_start = INITRD_START;
687 initrd_end = initrd_start + INITRD_SIZE;
688 memblock_reserve(INITRD_START, INITRD_SIZE);
693 * Reserve the memory area used to pass the certificate lists
695 static void __init reserve_certificate_list(void)
697 if (ipl_cert_list_addr)
698 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
701 static void __init reserve_mem_detect_info(void)
703 unsigned long start, size;
705 get_mem_detect_reserved(&start, &size);
707 memblock_reserve(start, size);
710 static void __init free_mem_detect_info(void)
712 unsigned long start, size;
714 get_mem_detect_reserved(&start, &size);
716 memblock_free(start, size);
719 static const char * __init get_mem_info_source(void)
721 switch (mem_detect.info_source) {
722 case MEM_DETECT_SCLP_STOR_INFO:
723 return "sclp storage info";
724 case MEM_DETECT_DIAG260:
726 case MEM_DETECT_SCLP_READ_INFO:
727 return "sclp read info";
728 case MEM_DETECT_BIN_SEARCH:
729 return "binary search";
734 static void __init memblock_add_mem_detect_info(void)
736 unsigned long start, end;
739 pr_debug("physmem info source: %s (%hhd)\n",
740 get_mem_info_source(), mem_detect.info_source);
741 /* keep memblock lists close to the kernel */
742 memblock_set_bottom_up(true);
743 for_each_mem_detect_block(i, &start, &end) {
744 memblock_add(start, end - start);
745 memblock_physmem_add(start, end - start);
747 memblock_set_bottom_up(false);
748 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
753 * Check for initrd being in usable memory
755 static void __init check_initrd(void)
757 #ifdef CONFIG_BLK_DEV_INITRD
758 if (INITRD_START && INITRD_SIZE &&
759 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
760 pr_err("The initial RAM disk does not fit into the memory\n");
761 memblock_free(INITRD_START, INITRD_SIZE);
762 initrd_start = initrd_end = 0;
768 * Reserve memory used for lowcore/command line/kernel image.
770 static void __init reserve_kernel(void)
772 unsigned long start_pfn = PFN_UP(__pa(_end));
774 memblock_reserve(0, HEAD_END);
775 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
776 - (unsigned long)_stext);
777 memblock_reserve(__sdma, __edma - __sdma);
780 static void __init setup_memory(void)
782 phys_addr_t start, end;
786 * Init storage key for present memory
788 for_each_mem_range(i, &start, &end)
789 storage_key_init_range(start, end);
791 psw_set_key(PAGE_DEFAULT_KEY);
794 memblock_enforce_memory_limit(memblock_end_of_DRAM());
798 * Setup hardware capabilities.
800 static int __init setup_hwcaps(void)
802 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
807 * The store facility list bits numbers as found in the principles
808 * of operation are numbered with bit 1UL<<31 as number 0 to
809 * bit 1UL<<0 as number 31.
810 * Bit 0: instructions named N3, "backported" to esa-mode
811 * Bit 2: z/Architecture mode is active
812 * Bit 7: the store-facility-list-extended facility is installed
813 * Bit 17: the message-security assist is installed
814 * Bit 19: the long-displacement facility is installed
815 * Bit 21: the extended-immediate facility is installed
816 * Bit 22: extended-translation facility 3 is installed
817 * Bit 30: extended-translation facility 3 enhancement facility
818 * These get translated to:
819 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
820 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
821 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
822 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
824 for (i = 0; i < 6; i++)
825 if (test_facility(stfl_bits[i]))
826 elf_hwcap |= 1UL << i;
828 if (test_facility(22) && test_facility(30))
829 elf_hwcap |= HWCAP_S390_ETF3EH;
832 * Check for additional facilities with store-facility-list-extended.
833 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
834 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
835 * as stored by stfl, bits 32-xxx contain additional facilities.
836 * How many facility words are stored depends on the number of
837 * doublewords passed to the instruction. The additional facilities
839 * Bit 42: decimal floating point facility is installed
840 * Bit 44: perform floating point operation facility is installed
842 * HWCAP_S390_DFP bit 6 (42 && 44).
844 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
845 elf_hwcap |= HWCAP_S390_DFP;
848 * Huge page support HWCAP_S390_HPAGE is bit 7.
850 if (MACHINE_HAS_EDAT1)
851 elf_hwcap |= HWCAP_S390_HPAGE;
854 * 64-bit register support for 31-bit processes
855 * HWCAP_S390_HIGH_GPRS is bit 9.
857 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
860 * Transactional execution support HWCAP_S390_TE is bit 10.
863 elf_hwcap |= HWCAP_S390_TE;
866 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
867 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
868 * instead of facility bit 129.
870 if (MACHINE_HAS_VX) {
871 elf_hwcap |= HWCAP_S390_VXRS;
872 if (test_facility(134))
873 elf_hwcap |= HWCAP_S390_VXRS_BCD;
874 if (test_facility(135))
875 elf_hwcap |= HWCAP_S390_VXRS_EXT;
876 if (test_facility(148))
877 elf_hwcap |= HWCAP_S390_VXRS_EXT2;
878 if (test_facility(152))
879 elf_hwcap |= HWCAP_S390_VXRS_PDE;
881 if (test_facility(150))
882 elf_hwcap |= HWCAP_S390_SORT;
883 if (test_facility(151))
884 elf_hwcap |= HWCAP_S390_DFLT;
887 * Guarded storage support HWCAP_S390_GS is bit 12.
890 elf_hwcap |= HWCAP_S390_GS;
893 add_device_randomness(&cpu_id, sizeof(cpu_id));
894 switch (cpu_id.machine) {
897 default: /* Use "z900" as default for 64 bit kernels. */
898 strcpy(elf_platform, "z900");
902 strcpy(elf_platform, "z990");
906 strcpy(elf_platform, "z9-109");
910 strcpy(elf_platform, "z10");
914 strcpy(elf_platform, "z196");
918 strcpy(elf_platform, "zEC12");
922 strcpy(elf_platform, "z13");
926 strcpy(elf_platform, "z14");
930 strcpy(elf_platform, "z15");
935 * Virtualization support HWCAP_INT_SIE is bit 0.
938 int_hwcap |= HWCAP_INT_SIE;
942 arch_initcall(setup_hwcaps);
945 * Add system information as device randomness
947 static void __init setup_randomness(void)
949 struct sysinfo_3_2_2 *vmms;
951 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
954 panic("Failed to allocate memory for sysinfo structure\n");
956 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
957 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
958 memblock_free((unsigned long) vmms, PAGE_SIZE);
962 * Find the correct size for the task_struct. This depends on
963 * the size of the struct fpu at the end of the thread_struct
964 * which is embedded in the task_struct.
966 static void __init setup_task_size(void)
968 int task_size = sizeof(struct task_struct);
970 if (!MACHINE_HAS_VX) {
971 task_size -= sizeof(__vector128) * __NUM_VXRS;
972 task_size += sizeof(freg_t) * __NUM_FPRS;
974 arch_task_struct_size = task_size;
978 * Issue diagnose 318 to set the control program name and
981 static void __init setup_control_program_code(void)
983 union diag318_info diag318_info = {
988 if (!sclp.has_diag318)
991 diag_stat_inc(DIAG_STAT_X318);
992 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
996 * Print the component list from the IPL report
998 static void __init log_component_list(void)
1000 struct ipl_rb_component_entry *ptr, *end;
1003 if (!early_ipl_comp_list_addr)
1005 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1006 pr_info("Linux is running with Secure-IPL enabled\n");
1008 pr_info("Linux is running with Secure-IPL disabled\n");
1009 ptr = (void *) early_ipl_comp_list_addr;
1010 end = (void *) ptr + early_ipl_comp_list_size;
1011 pr_info("The IPL report contains the following components:\n");
1013 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1014 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1015 str = "signed, verified";
1017 str = "signed, verification failed";
1021 pr_info("%016llx - %016llx (%s)\n",
1022 ptr->addr, ptr->addr + ptr->len, str);
1028 * Setup function called from init/main.c just after the banner
1032 void __init setup_arch(char **cmdline_p)
1035 * print what head.S has found out about the machine
1038 pr_info("Linux is running as a z/VM "
1039 "guest operating system in 64-bit mode\n");
1040 else if (MACHINE_IS_KVM)
1041 pr_info("Linux is running under KVM in 64-bit mode\n");
1042 else if (MACHINE_IS_LPAR)
1043 pr_info("Linux is running natively in 64-bit mode\n");
1045 pr_info("Linux is running as a guest in 64-bit mode\n");
1047 log_component_list();
1049 /* Have one command line that is parsed and saved in /proc/cmdline */
1050 /* boot_command_line has been already set up in early.c */
1051 *cmdline_p = boot_command_line;
1053 ROOT_DEV = Root_RAM0;
1055 init_mm.start_code = (unsigned long) _text;
1056 init_mm.end_code = (unsigned long) _etext;
1057 init_mm.end_data = (unsigned long) _edata;
1058 init_mm.brk = (unsigned long) _end;
1060 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1061 nospec_auto_detect();
1064 parse_early_param();
1065 #ifdef CONFIG_CRASH_DUMP
1066 /* Deactivate elfcorehdr= kernel parameter */
1067 elfcorehdr_addr = ELFCORE_ADDR_MAX;
1073 setup_control_program_code();
1075 /* Do some memory reservations *before* memory is added to memblock */
1076 reserve_above_ident_map();
1080 reserve_certificate_list();
1081 reserve_mem_detect_info();
1082 memblock_allow_resize();
1084 /* Get information about *all* installed memory */
1085 memblock_add_mem_detect_info();
1087 free_mem_detect_info();
1093 dma_contiguous_reserve(ident_map_size);
1095 if (MACHINE_HAS_EDAT2)
1096 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1099 reserve_crashkernel();
1100 #ifdef CONFIG_CRASH_DUMP
1102 * Be aware that smp_save_dump_cpus() triggers a system reset.
1103 * Therefore CPU and device initialization should be done afterwards.
1105 smp_save_dump_cpus();
1109 setup_lowcore_dat_off();
1110 smp_fill_possible_mask();
1111 cpu_detect_mhz_feature();
1115 topology_init_early();
1118 * Create kernel page tables and switch to virtual addressing.
1123 * After paging_init created the kernel page table, the new PSWs
1124 * in lowcore can now run with DAT enabled.
1126 setup_lowcore_dat_on();
1128 /* Setup default console */
1130 set_preferred_console();
1132 apply_alternative_instructions();
1133 if (IS_ENABLED(CONFIG_EXPOLINE))
1134 nospec_init_branches();
1136 /* Setup zfcp/nvme dump support */
1139 /* Add system specific data to the random pool */