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;
98 * Some code and data needs to stay below 2 GB, even when the kernel would be
99 * relocated above 2 GB, because it has to use 31 bit addresses.
100 * Such code and data is part of the .dma section.
102 unsigned long __dma_ref __sdma = __pa(&_sdma);
103 unsigned long __dma_ref __edma = __pa(&_edma);
104 unsigned long __dma_ref __stext_dma = __pa(&_stext_dma);
105 unsigned long __dma_ref __etext_dma = __pa(&_etext_dma);
106 struct exception_table_entry __dma_ref *__start_dma_ex_table = _start_dma_ex_table;
107 struct exception_table_entry __dma_ref *__stop_dma_ex_table = _stop_dma_ex_table;
110 * Control registers CR2, CR5 and CR15 are initialized with addresses
111 * of tables that must be placed below 2G which is handled by the DMA
113 * Because the DMA sections are relocated below 2G at startup,
114 * the content of control registers CR2, CR5 and CR15 must be updated
115 * with new addresses after the relocation. The initial initialization of
116 * control registers occurs in head64.S and then gets updated again after DMA
117 * relocation. We must access the relevant DMA tables indirectly via
118 * pointers placed in the .dma.refs linker section. Those pointers get
119 * updated automatically during DMA relocation and always contain a valid
120 * address within DMA sections.
123 static __dma_data u32 __ctl_duct_dma[16] __aligned(64);
125 static __dma_data u64 __ctl_aste_dma[8] __aligned(64) = {
126 [1] = 0xffffffffffffffff
129 static __dma_data u32 __ctl_duald_dma[32] __aligned(128) = {
140 static __dma_data u32 __ctl_linkage_stack_dma[8] __aligned(64) = {
145 static u64 __dma_ref *__ctl_aste = __ctl_aste_dma;
146 static u32 __dma_ref *__ctl_duald = __ctl_duald_dma;
147 static u32 __dma_ref *__ctl_linkage_stack = __ctl_linkage_stack_dma;
148 static u32 __dma_ref *__ctl_duct = __ctl_duct_dma;
150 int __bootdata(noexec_disabled);
151 unsigned long __bootdata(ident_map_size);
152 struct mem_detect_info __bootdata(mem_detect);
153 struct initrd_data __bootdata(initrd_data);
155 unsigned long __bootdata_preserved(__kaslr_offset);
156 unsigned int __bootdata_preserved(zlib_dfltcc_support);
157 EXPORT_SYMBOL(zlib_dfltcc_support);
158 u64 __bootdata_preserved(stfle_fac_list[16]);
159 EXPORT_SYMBOL(stfle_fac_list);
160 u64 __bootdata_preserved(alt_stfle_fac_list[16]);
161 struct oldmem_data __bootdata_preserved(oldmem_data);
163 unsigned long VMALLOC_START;
164 EXPORT_SYMBOL(VMALLOC_START);
166 unsigned long VMALLOC_END;
167 EXPORT_SYMBOL(VMALLOC_END);
169 struct page *vmemmap;
170 EXPORT_SYMBOL(vmemmap);
171 unsigned long vmemmap_size;
173 unsigned long MODULES_VADDR;
174 unsigned long MODULES_END;
176 /* An array with a pointer to the lowcore of every CPU. */
177 struct lowcore *lowcore_ptr[NR_CPUS];
178 EXPORT_SYMBOL(lowcore_ptr);
181 * The Write Back bit position in the physaddr is given by the SLPC PCI.
182 * Leaving the mask zero always uses write through which is safe
184 unsigned long mio_wb_bit_mask __ro_after_init;
187 * This is set up by the setup-routine at boot-time
188 * for S390 need to find out, what we have to setup
189 * using address 0x10400 ...
192 #include <asm/setup.h>
195 * condev= and conmode= setup parameter.
198 static int __init condev_setup(char *str)
202 vdev = simple_strtoul(str, &str, 0);
203 if (vdev >= 0 && vdev < 65536) {
204 console_devno = vdev;
210 __setup("condev=", condev_setup);
212 static void __init set_preferred_console(void)
214 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
215 add_preferred_console("ttyS", 0, NULL);
216 else if (CONSOLE_IS_3270)
217 add_preferred_console("tty3270", 0, NULL);
218 else if (CONSOLE_IS_VT220)
219 add_preferred_console("ttysclp", 0, NULL);
220 else if (CONSOLE_IS_HVC)
221 add_preferred_console("hvc", 0, NULL);
224 static int __init conmode_setup(char *str)
226 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
227 if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
230 #if defined(CONFIG_TN3215_CONSOLE)
231 if (!strcmp(str, "3215"))
234 #if defined(CONFIG_TN3270_CONSOLE)
235 if (!strcmp(str, "3270"))
238 set_preferred_console();
242 __setup("conmode=", conmode_setup);
244 static void __init conmode_default(void)
246 char query_buffer[1024];
250 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
251 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
252 ptr = strstr(query_buffer, "SUBCHANNEL =");
253 console_irq = simple_strtoul(ptr + 13, NULL, 16);
254 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
255 ptr = strstr(query_buffer, "CONMODE");
257 * Set the conmode to 3215 so that the device recognition
258 * will set the cu_type of the console to 3215. If the
259 * conmode is 3270 and we don't set it back then both
260 * 3215 and the 3270 driver will try to access the console
261 * device (3215 as console and 3270 as normal tty).
263 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
265 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
270 if (str_has_prefix(ptr + 8, "3270")) {
271 #if defined(CONFIG_TN3270_CONSOLE)
273 #elif defined(CONFIG_TN3215_CONSOLE)
275 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
278 } else if (str_has_prefix(ptr + 8, "3215")) {
279 #if defined(CONFIG_TN3215_CONSOLE)
281 #elif defined(CONFIG_TN3270_CONSOLE)
283 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
287 } else if (MACHINE_IS_KVM) {
288 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
290 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
295 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
301 #ifdef CONFIG_CRASH_DUMP
302 static void __init setup_zfcpdump(void)
304 if (!is_ipl_type_dump())
306 if (oldmem_data.start)
308 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
309 console_loglevel = 2;
312 static inline void setup_zfcpdump(void) {}
313 #endif /* CONFIG_CRASH_DUMP */
316 * Reboot, halt and power_off stubs. They just call _machine_restart,
317 * _machine_halt or _machine_power_off.
320 void machine_restart(char *command)
322 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
324 * Only unblank the console if we are called in enabled
325 * context or a bust_spinlocks cleared the way for us.
328 _machine_restart(command);
331 void machine_halt(void)
333 if (!in_interrupt() || oops_in_progress)
335 * Only unblank the console if we are called in enabled
336 * context or a bust_spinlocks cleared the way for us.
342 void machine_power_off(void)
344 if (!in_interrupt() || oops_in_progress)
346 * Only unblank the console if we are called in enabled
347 * context or a bust_spinlocks cleared the way for us.
350 _machine_power_off();
354 * Dummy power off function.
356 void (*pm_power_off)(void) = machine_power_off;
357 EXPORT_SYMBOL_GPL(pm_power_off);
361 unsigned long stack_alloc(void)
363 #ifdef CONFIG_VMAP_STACK
364 return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE,
365 THREADINFO_GFP, NUMA_NO_NODE,
366 __builtin_return_address(0));
368 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
372 void stack_free(unsigned long stack)
374 #ifdef CONFIG_VMAP_STACK
375 vfree((void *) stack);
377 free_pages(stack, THREAD_SIZE_ORDER);
381 int __init arch_early_irq_init(void)
385 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
387 panic("Couldn't allocate async stack");
388 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
392 void __init arch_call_rest_init(void)
396 stack = stack_alloc();
398 panic("Couldn't allocate kernel stack");
399 current->stack = (void *) stack;
400 #ifdef CONFIG_VMAP_STACK
401 current->stack_vm_area = (void *) stack;
403 set_task_stack_end_magic(current);
404 stack += STACK_INIT_OFFSET;
405 S390_lowcore.kernel_stack = stack;
406 call_on_stack_noreturn(rest_init, stack);
409 static void __init setup_lowcore_dat_off(void)
411 unsigned long int_psw_mask = PSW_KERNEL_BITS;
412 unsigned long mcck_stack;
415 if (IS_ENABLED(CONFIG_KASAN))
416 int_psw_mask |= PSW_MASK_DAT;
419 * Setup lowcore for boot cpu
421 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
422 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
424 panic("%s: Failed to allocate %zu bytes align=%zx\n",
425 __func__, sizeof(*lc), sizeof(*lc));
427 lc->restart_psw.mask = PSW_KERNEL_BITS;
428 lc->restart_psw.addr = (unsigned long) restart_int_handler;
429 lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
430 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
431 lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
432 lc->svc_new_psw.addr = (unsigned long) system_call;
433 lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
434 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
435 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
436 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
437 lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
438 lc->io_new_psw.addr = (unsigned long) io_int_handler;
439 lc->clock_comparator = clock_comparator_max;
440 lc->nodat_stack = ((unsigned long) &init_thread_union)
441 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
442 lc->current_task = (unsigned long)&init_task;
444 lc->machine_flags = S390_lowcore.machine_flags;
445 lc->preempt_count = S390_lowcore.preempt_count;
446 nmi_alloc_boot_cpu(lc);
447 lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
448 lc->exit_timer = S390_lowcore.exit_timer;
449 lc->user_timer = S390_lowcore.user_timer;
450 lc->system_timer = S390_lowcore.system_timer;
451 lc->steal_timer = S390_lowcore.steal_timer;
452 lc->last_update_timer = S390_lowcore.last_update_timer;
453 lc->last_update_clock = S390_lowcore.last_update_clock;
456 * Allocate the global restart stack which is the same for
457 * all CPUs in cast *one* of them does a PSW restart.
459 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
461 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
462 __func__, THREAD_SIZE, THREAD_SIZE);
463 restart_stack += STACK_INIT_OFFSET;
466 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
467 * restart data to the absolute zero lowcore. This is necessary if
468 * PSW restart is done on an offline CPU that has lowcore zero.
470 lc->restart_stack = (unsigned long) restart_stack;
471 lc->restart_fn = (unsigned long) do_restart;
472 lc->restart_data = 0;
473 lc->restart_source = -1UL;
475 mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
477 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
478 __func__, THREAD_SIZE, THREAD_SIZE);
479 lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
481 /* Setup absolute zero lowcore */
482 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
483 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
484 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
485 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
486 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
488 lc->spinlock_lockval = arch_spin_lockval(0);
489 lc->spinlock_index = 0;
490 arch_spin_lock_setup(0);
491 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
492 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
493 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
494 lc->preempt_count = PREEMPT_DISABLED;
496 set_prefix((u32)(unsigned long) lc);
500 static void __init setup_lowcore_dat_on(void)
502 __ctl_clear_bit(0, 28);
503 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
504 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
505 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
506 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
507 __ctl_set_bit(0, 28);
510 static struct resource code_resource = {
511 .name = "Kernel code",
512 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
515 static struct resource data_resource = {
516 .name = "Kernel data",
517 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
520 static struct resource bss_resource = {
521 .name = "Kernel bss",
522 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
525 static struct resource __initdata *standard_resources[] = {
531 static void __init setup_resources(void)
533 struct resource *res, *std_res, *sub_res;
534 phys_addr_t start, end;
538 code_resource.start = (unsigned long) _text;
539 code_resource.end = (unsigned long) _etext - 1;
540 data_resource.start = (unsigned long) _etext;
541 data_resource.end = (unsigned long) _edata - 1;
542 bss_resource.start = (unsigned long) __bss_start;
543 bss_resource.end = (unsigned long) __bss_stop - 1;
545 for_each_mem_range(i, &start, &end) {
546 res = memblock_alloc(sizeof(*res), 8);
548 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
549 __func__, sizeof(*res), 8);
550 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
552 res->name = "System RAM";
555 * In memblock, end points to the first byte after the
556 * range while in resourses, end points to the last byte in
560 request_resource(&iomem_resource, res);
562 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
563 std_res = standard_resources[j];
564 if (std_res->start < res->start ||
565 std_res->start > res->end)
567 if (std_res->end > res->end) {
568 sub_res = memblock_alloc(sizeof(*sub_res), 8);
570 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
571 __func__, sizeof(*sub_res), 8);
573 sub_res->end = res->end;
574 std_res->start = res->end + 1;
575 request_resource(res, sub_res);
577 request_resource(res, std_res);
581 #ifdef CONFIG_CRASH_DUMP
583 * Re-add removed crash kernel memory as reserved memory. This makes
584 * sure it will be mapped with the identity mapping and struct pages
585 * will be created, so it can be resized later on.
586 * However add it later since the crash kernel resource should not be
587 * part of the System RAM resource.
589 if (crashk_res.end) {
590 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
591 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
592 insert_resource(&iomem_resource, &crashk_res);
597 static void __init setup_memory_end(void)
599 memblock_remove(ident_map_size, ULONG_MAX);
600 max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
601 pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
604 #ifdef CONFIG_CRASH_DUMP
607 * When kdump is enabled, we have to ensure that no memory from the area
608 * [0 - crashkernel memory size] is set offline - it will be exchanged with
609 * the crashkernel memory region when kdump is triggered. The crashkernel
610 * memory region can never get offlined (pages are unmovable).
612 static int kdump_mem_notifier(struct notifier_block *nb,
613 unsigned long action, void *data)
615 struct memory_notify *arg = data;
617 if (action != MEM_GOING_OFFLINE)
619 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
624 static struct notifier_block kdump_mem_nb = {
625 .notifier_call = kdump_mem_notifier,
631 * Make sure that the area above identity mapping is protected
633 static void __init reserve_above_ident_map(void)
635 memblock_reserve(ident_map_size, ULONG_MAX);
639 * Reserve memory for kdump kernel to be loaded with kexec
641 static void __init reserve_crashkernel(void)
643 #ifdef CONFIG_CRASH_DUMP
644 unsigned long long crash_base, crash_size;
645 phys_addr_t low, high;
648 rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
651 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
652 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
653 if (rc || crash_size == 0)
656 if (memblock.memory.regions[0].size < crash_size) {
657 pr_info("crashkernel reservation failed: %s\n",
658 "first memory chunk must be at least crashkernel size");
662 low = crash_base ?: oldmem_data.start;
663 high = low + crash_size;
664 if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
665 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
668 /* Find suitable area in free memory */
669 low = max_t(unsigned long, crash_size, sclp.hsa_size);
670 high = crash_base ? crash_base + crash_size : ULONG_MAX;
672 if (crash_base && crash_base < low) {
673 pr_info("crashkernel reservation failed: %s\n",
674 "crash_base too low");
677 low = crash_base ?: low;
678 crash_base = memblock_find_in_range(low, high, crash_size,
679 KEXEC_CRASH_MEM_ALIGN);
683 pr_info("crashkernel reservation failed: %s\n",
684 "no suitable area found");
688 if (register_memory_notifier(&kdump_mem_nb))
691 if (!oldmem_data.start && MACHINE_IS_VM)
692 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
693 crashk_res.start = crash_base;
694 crashk_res.end = crash_base + crash_size - 1;
695 memblock_remove(crash_base, crash_size);
696 pr_info("Reserving %lluMB of memory at %lluMB "
697 "for crashkernel (System RAM: %luMB)\n",
698 crash_size >> 20, crash_base >> 20,
699 (unsigned long)memblock.memory.total_size >> 20);
700 os_info_crashkernel_add(crash_base, crash_size);
705 * Reserve the initrd from being used by memblock
707 static void __init reserve_initrd(void)
709 #ifdef CONFIG_BLK_DEV_INITRD
710 if (!initrd_data.start || !initrd_data.size)
712 initrd_start = initrd_data.start;
713 initrd_end = initrd_start + initrd_data.size;
714 memblock_reserve(initrd_data.start, initrd_data.size);
719 * Reserve the memory area used to pass the certificate lists
721 static void __init reserve_certificate_list(void)
723 if (ipl_cert_list_addr)
724 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
727 static void __init reserve_mem_detect_info(void)
729 unsigned long start, size;
731 get_mem_detect_reserved(&start, &size);
733 memblock_reserve(start, size);
736 static void __init free_mem_detect_info(void)
738 unsigned long start, size;
740 get_mem_detect_reserved(&start, &size);
742 memblock_free(start, size);
745 static const char * __init get_mem_info_source(void)
747 switch (mem_detect.info_source) {
748 case MEM_DETECT_SCLP_STOR_INFO:
749 return "sclp storage info";
750 case MEM_DETECT_DIAG260:
752 case MEM_DETECT_SCLP_READ_INFO:
753 return "sclp read info";
754 case MEM_DETECT_BIN_SEARCH:
755 return "binary search";
760 static void __init memblock_add_mem_detect_info(void)
762 unsigned long start, end;
765 pr_debug("physmem info source: %s (%hhd)\n",
766 get_mem_info_source(), mem_detect.info_source);
767 /* keep memblock lists close to the kernel */
768 memblock_set_bottom_up(true);
769 for_each_mem_detect_block(i, &start, &end) {
770 memblock_add(start, end - start);
771 memblock_physmem_add(start, end - start);
773 memblock_set_bottom_up(false);
774 memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
779 * Check for initrd being in usable memory
781 static void __init check_initrd(void)
783 #ifdef CONFIG_BLK_DEV_INITRD
784 if (initrd_data.start && initrd_data.size &&
785 !memblock_is_region_memory(initrd_data.start, initrd_data.size)) {
786 pr_err("The initial RAM disk does not fit into the memory\n");
787 memblock_free(initrd_data.start, initrd_data.size);
788 initrd_start = initrd_end = 0;
794 * Reserve memory used for lowcore/command line/kernel image.
796 static void __init reserve_kernel(void)
798 unsigned long start_pfn = PFN_UP(__pa(_end));
800 memblock_reserve(0, HEAD_END);
801 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
802 - (unsigned long)_stext);
805 static void __init setup_memory(void)
807 phys_addr_t start, end;
811 * Init storage key for present memory
813 for_each_mem_range(i, &start, &end)
814 storage_key_init_range(start, end);
816 psw_set_key(PAGE_DEFAULT_KEY);
819 memblock_enforce_memory_limit(memblock_end_of_DRAM());
822 static void __init relocate_dma_section(void)
824 unsigned long dma_addr, dma_size;
828 /* Allocate a new DMA capable memory region */
829 dma_size = __edma - __sdma;
830 pr_info("Relocating DMA section of size 0x%08lx\n", dma_size);
831 dma_addr = (unsigned long)memblock_alloc_low(dma_size, PAGE_SIZE);
833 panic("Failed to allocate memory for DMA section\n");
834 dma_offset = dma_addr - __sdma;
836 /* Move original DMA section to the new one */
837 memmove((void *)dma_addr, (void *)__sdma, dma_size);
838 /* Zero out the old DMA section to catch invalid accesses within it */
839 memset((void *)__sdma, 0, dma_size);
841 /* Update all DMA region references */
842 for (ptr = _start_dma_refs; ptr != _end_dma_refs; ptr++)
846 /* This must be called after DMA relocation */
847 static void __init setup_cr(void)
853 __ctl_duct[1] = (unsigned long)__ctl_aste;
854 __ctl_duct[2] = (unsigned long)__ctl_aste;
855 __ctl_duct[4] = (unsigned long)__ctl_duald;
857 /* Update control registers CR2, CR5 and CR15 */
858 __ctl_store(cr2.val, 2, 2);
859 __ctl_store(cr5.val, 5, 5);
860 __ctl_store(cr15.val, 15, 15);
861 cr2.ducto = (unsigned long)__ctl_duct >> 6;
862 cr5.pasteo = (unsigned long)__ctl_duct >> 6;
863 cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
864 __ctl_load(cr2.val, 2, 2);
865 __ctl_load(cr5.val, 5, 5);
866 __ctl_load(cr15.val, 15, 15);
870 * Setup hardware capabilities.
872 static int __init setup_hwcaps(void)
874 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
879 * The store facility list bits numbers as found in the principles
880 * of operation are numbered with bit 1UL<<31 as number 0 to
881 * bit 1UL<<0 as number 31.
882 * Bit 0: instructions named N3, "backported" to esa-mode
883 * Bit 2: z/Architecture mode is active
884 * Bit 7: the store-facility-list-extended facility is installed
885 * Bit 17: the message-security assist is installed
886 * Bit 19: the long-displacement facility is installed
887 * Bit 21: the extended-immediate facility is installed
888 * Bit 22: extended-translation facility 3 is installed
889 * Bit 30: extended-translation facility 3 enhancement facility
890 * These get translated to:
891 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
892 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
893 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
894 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
896 for (i = 0; i < 6; i++)
897 if (test_facility(stfl_bits[i]))
898 elf_hwcap |= 1UL << i;
900 if (test_facility(22) && test_facility(30))
901 elf_hwcap |= HWCAP_S390_ETF3EH;
904 * Check for additional facilities with store-facility-list-extended.
905 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
906 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
907 * as stored by stfl, bits 32-xxx contain additional facilities.
908 * How many facility words are stored depends on the number of
909 * doublewords passed to the instruction. The additional facilities
911 * Bit 42: decimal floating point facility is installed
912 * Bit 44: perform floating point operation facility is installed
914 * HWCAP_S390_DFP bit 6 (42 && 44).
916 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
917 elf_hwcap |= HWCAP_S390_DFP;
920 * Huge page support HWCAP_S390_HPAGE is bit 7.
922 if (MACHINE_HAS_EDAT1)
923 elf_hwcap |= HWCAP_S390_HPAGE;
926 * 64-bit register support for 31-bit processes
927 * HWCAP_S390_HIGH_GPRS is bit 9.
929 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
932 * Transactional execution support HWCAP_S390_TE is bit 10.
935 elf_hwcap |= HWCAP_S390_TE;
938 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
939 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
940 * instead of facility bit 129.
942 if (MACHINE_HAS_VX) {
943 elf_hwcap |= HWCAP_S390_VXRS;
944 if (test_facility(134))
945 elf_hwcap |= HWCAP_S390_VXRS_BCD;
946 if (test_facility(135))
947 elf_hwcap |= HWCAP_S390_VXRS_EXT;
948 if (test_facility(148))
949 elf_hwcap |= HWCAP_S390_VXRS_EXT2;
950 if (test_facility(152))
951 elf_hwcap |= HWCAP_S390_VXRS_PDE;
953 if (test_facility(150))
954 elf_hwcap |= HWCAP_S390_SORT;
955 if (test_facility(151))
956 elf_hwcap |= HWCAP_S390_DFLT;
959 * Guarded storage support HWCAP_S390_GS is bit 12.
962 elf_hwcap |= HWCAP_S390_GS;
965 add_device_randomness(&cpu_id, sizeof(cpu_id));
966 switch (cpu_id.machine) {
969 default: /* Use "z900" as default for 64 bit kernels. */
970 strcpy(elf_platform, "z900");
974 strcpy(elf_platform, "z990");
978 strcpy(elf_platform, "z9-109");
982 strcpy(elf_platform, "z10");
986 strcpy(elf_platform, "z196");
990 strcpy(elf_platform, "zEC12");
994 strcpy(elf_platform, "z13");
998 strcpy(elf_platform, "z14");
1002 strcpy(elf_platform, "z15");
1007 * Virtualization support HWCAP_INT_SIE is bit 0.
1010 int_hwcap |= HWCAP_INT_SIE;
1014 arch_initcall(setup_hwcaps);
1017 * Add system information as device randomness
1019 static void __init setup_randomness(void)
1021 struct sysinfo_3_2_2 *vmms;
1023 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
1026 panic("Failed to allocate memory for sysinfo structure\n");
1028 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1029 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1030 memblock_free((unsigned long) vmms, PAGE_SIZE);
1034 * Find the correct size for the task_struct. This depends on
1035 * the size of the struct fpu at the end of the thread_struct
1036 * which is embedded in the task_struct.
1038 static void __init setup_task_size(void)
1040 int task_size = sizeof(struct task_struct);
1042 if (!MACHINE_HAS_VX) {
1043 task_size -= sizeof(__vector128) * __NUM_VXRS;
1044 task_size += sizeof(freg_t) * __NUM_FPRS;
1046 arch_task_struct_size = task_size;
1050 * Issue diagnose 318 to set the control program name and
1053 static void __init setup_control_program_code(void)
1055 union diag318_info diag318_info = {
1060 if (!sclp.has_diag318)
1063 diag_stat_inc(DIAG_STAT_X318);
1064 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1068 * Print the component list from the IPL report
1070 static void __init log_component_list(void)
1072 struct ipl_rb_component_entry *ptr, *end;
1075 if (!early_ipl_comp_list_addr)
1077 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1078 pr_info("Linux is running with Secure-IPL enabled\n");
1080 pr_info("Linux is running with Secure-IPL disabled\n");
1081 ptr = (void *) early_ipl_comp_list_addr;
1082 end = (void *) ptr + early_ipl_comp_list_size;
1083 pr_info("The IPL report contains the following components:\n");
1085 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1086 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1087 str = "signed, verified";
1089 str = "signed, verification failed";
1093 pr_info("%016llx - %016llx (%s)\n",
1094 ptr->addr, ptr->addr + ptr->len, str);
1100 * Setup function called from init/main.c just after the banner
1104 void __init setup_arch(char **cmdline_p)
1107 * print what head.S has found out about the machine
1110 pr_info("Linux is running as a z/VM "
1111 "guest operating system in 64-bit mode\n");
1112 else if (MACHINE_IS_KVM)
1113 pr_info("Linux is running under KVM in 64-bit mode\n");
1114 else if (MACHINE_IS_LPAR)
1115 pr_info("Linux is running natively in 64-bit mode\n");
1117 pr_info("Linux is running as a guest in 64-bit mode\n");
1119 log_component_list();
1121 /* Have one command line that is parsed and saved in /proc/cmdline */
1122 /* boot_command_line has been already set up in early.c */
1123 *cmdline_p = boot_command_line;
1125 ROOT_DEV = Root_RAM0;
1127 setup_initial_init_mm(_text, _etext, _edata, _end);
1129 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1130 nospec_auto_detect();
1133 parse_early_param();
1134 #ifdef CONFIG_CRASH_DUMP
1135 /* Deactivate elfcorehdr= kernel parameter */
1136 elfcorehdr_addr = ELFCORE_ADDR_MAX;
1142 setup_control_program_code();
1144 /* Do some memory reservations *before* memory is added to memblock */
1145 reserve_above_ident_map();
1148 reserve_certificate_list();
1149 reserve_mem_detect_info();
1150 memblock_allow_resize();
1152 /* Get information about *all* installed memory */
1153 memblock_add_mem_detect_info();
1155 free_mem_detect_info();
1157 relocate_dma_section();
1163 dma_contiguous_reserve(ident_map_size);
1165 if (MACHINE_HAS_EDAT2)
1166 hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1169 reserve_crashkernel();
1170 #ifdef CONFIG_CRASH_DUMP
1172 * Be aware that smp_save_dump_cpus() triggers a system reset.
1173 * Therefore CPU and device initialization should be done afterwards.
1175 smp_save_dump_cpus();
1179 setup_lowcore_dat_off();
1180 smp_fill_possible_mask();
1181 cpu_detect_mhz_feature();
1185 topology_init_early();
1188 * Create kernel page tables and switch to virtual addressing.
1193 * After paging_init created the kernel page table, the new PSWs
1194 * in lowcore can now run with DAT enabled.
1196 setup_lowcore_dat_on();
1198 /* Setup default console */
1200 set_preferred_console();
1202 apply_alternative_instructions();
1203 if (IS_ENABLED(CONFIG_EXPOLINE))
1204 nospec_init_branches();
1206 /* Setup zfcp/nvme dump support */
1209 /* Add system specific data to the random pool */