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>
52 #include <linux/security.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 int __bootdata(memory_end_set
);
99 unsigned long __bootdata(memory_end
);
100 unsigned long __bootdata(vmalloc_size
);
101 unsigned long __bootdata(max_physmem_end
);
102 struct mem_detect_info
__bootdata(mem_detect
);
104 struct exception_table_entry
*__bootdata_preserved(__start_dma_ex_table
);
105 struct exception_table_entry
*__bootdata_preserved(__stop_dma_ex_table
);
106 unsigned long __bootdata_preserved(__swsusp_reset_dma
);
107 unsigned long __bootdata_preserved(__stext_dma
);
108 unsigned long __bootdata_preserved(__etext_dma
);
109 unsigned long __bootdata_preserved(__sdma
);
110 unsigned long __bootdata_preserved(__edma
);
111 unsigned long __bootdata_preserved(__kaslr_offset
);
112 unsigned int __bootdata_preserved(zlib_dfltcc_support
);
113 EXPORT_SYMBOL(zlib_dfltcc_support
);
115 unsigned long VMALLOC_START
;
116 EXPORT_SYMBOL(VMALLOC_START
);
118 unsigned long VMALLOC_END
;
119 EXPORT_SYMBOL(VMALLOC_END
);
121 struct page
*vmemmap
;
122 EXPORT_SYMBOL(vmemmap
);
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 * This is set up by the setup-routine at boot-time
133 * for S390 need to find out, what we have to setup
134 * using address 0x10400 ...
137 #include <asm/setup.h>
140 * condev= and conmode= setup parameter.
143 static int __init
condev_setup(char *str
)
147 vdev
= simple_strtoul(str
, &str
, 0);
148 if (vdev
>= 0 && vdev
< 65536) {
149 console_devno
= vdev
;
155 __setup("condev=", condev_setup
);
157 static void __init
set_preferred_console(void)
159 if (CONSOLE_IS_3215
|| CONSOLE_IS_SCLP
)
160 add_preferred_console("ttyS", 0, NULL
);
161 else if (CONSOLE_IS_3270
)
162 add_preferred_console("tty3270", 0, NULL
);
163 else if (CONSOLE_IS_VT220
)
164 add_preferred_console("ttysclp", 0, NULL
);
165 else if (CONSOLE_IS_HVC
)
166 add_preferred_console("hvc", 0, NULL
);
169 static int __init
conmode_setup(char *str
)
171 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
172 if (!strcmp(str
, "hwc") || !strcmp(str
, "sclp"))
175 #if defined(CONFIG_TN3215_CONSOLE)
176 if (!strcmp(str
, "3215"))
179 #if defined(CONFIG_TN3270_CONSOLE)
180 if (!strcmp(str
, "3270"))
183 set_preferred_console();
187 __setup("conmode=", conmode_setup
);
189 static void __init
conmode_default(void)
191 char query_buffer
[1024];
195 cpcmd("QUERY CONSOLE", query_buffer
, 1024, NULL
);
196 console_devno
= simple_strtoul(query_buffer
+ 5, NULL
, 16);
197 ptr
= strstr(query_buffer
, "SUBCHANNEL =");
198 console_irq
= simple_strtoul(ptr
+ 13, NULL
, 16);
199 cpcmd("QUERY TERM", query_buffer
, 1024, NULL
);
200 ptr
= strstr(query_buffer
, "CONMODE");
202 * Set the conmode to 3215 so that the device recognition
203 * will set the cu_type of the console to 3215. If the
204 * conmode is 3270 and we don't set it back then both
205 * 3215 and the 3270 driver will try to access the console
206 * device (3215 as console and 3270 as normal tty).
208 cpcmd("TERM CONMODE 3215", NULL
, 0, NULL
);
210 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
215 if (str_has_prefix(ptr
+ 8, "3270")) {
216 #if defined(CONFIG_TN3270_CONSOLE)
218 #elif defined(CONFIG_TN3215_CONSOLE)
220 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
223 } else if (str_has_prefix(ptr
+ 8, "3215")) {
224 #if defined(CONFIG_TN3215_CONSOLE)
226 #elif defined(CONFIG_TN3270_CONSOLE)
228 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
232 } else if (MACHINE_IS_KVM
) {
233 if (sclp
.has_vt220
&& IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE
))
235 else if (sclp
.has_linemode
&& IS_ENABLED(CONFIG_SCLP_CONSOLE
))
240 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
244 if (IS_ENABLED(CONFIG_VT
) && IS_ENABLED(CONFIG_DUMMY_CONSOLE
))
245 conswitchp
= &dummy_con
;
248 #ifdef CONFIG_CRASH_DUMP
249 static void __init
setup_zfcpdump(void)
251 if (ipl_info
.type
!= IPL_TYPE_FCP_DUMP
)
255 strcat(boot_command_line
, " cio_ignore=all,!ipldev,!condev");
256 console_loglevel
= 2;
259 static inline void setup_zfcpdump(void) {}
260 #endif /* CONFIG_CRASH_DUMP */
263 * Reboot, halt and power_off stubs. They just call _machine_restart,
264 * _machine_halt or _machine_power_off.
267 void machine_restart(char *command
)
269 if ((!in_interrupt() && !in_atomic()) || 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.
275 _machine_restart(command
);
278 void machine_halt(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.
289 void machine_power_off(void)
291 if (!in_interrupt() || oops_in_progress
)
293 * Only unblank the console if we are called in enabled
294 * context or a bust_spinlocks cleared the way for us.
297 _machine_power_off();
301 * Dummy power off function.
303 void (*pm_power_off
)(void) = machine_power_off
;
304 EXPORT_SYMBOL_GPL(pm_power_off
);
306 void *restart_stack
__section(.data
);
308 unsigned long stack_alloc(void)
310 #ifdef CONFIG_VMAP_STACK
311 return (unsigned long)
312 __vmalloc_node_range(THREAD_SIZE
, THREAD_SIZE
,
313 VMALLOC_START
, VMALLOC_END
,
315 PAGE_KERNEL
, 0, NUMA_NO_NODE
,
316 __builtin_return_address(0));
318 return __get_free_pages(GFP_KERNEL
, THREAD_SIZE_ORDER
);
322 void stack_free(unsigned long stack
)
324 #ifdef CONFIG_VMAP_STACK
325 vfree((void *) stack
);
327 free_pages(stack
, THREAD_SIZE_ORDER
);
331 int __init
arch_early_irq_init(void)
335 stack
= __get_free_pages(GFP_KERNEL
, THREAD_SIZE_ORDER
);
337 panic("Couldn't allocate async stack");
338 S390_lowcore
.async_stack
= stack
+ STACK_INIT_OFFSET
;
342 static int __init
async_stack_realloc(void)
344 unsigned long old
, new;
346 old
= S390_lowcore
.async_stack
- STACK_INIT_OFFSET
;
349 panic("Couldn't allocate async stack");
350 S390_lowcore
.async_stack
= new + STACK_INIT_OFFSET
;
351 free_pages(old
, THREAD_SIZE_ORDER
);
354 early_initcall(async_stack_realloc
);
356 void __init
arch_call_rest_init(void)
360 stack
= stack_alloc();
362 panic("Couldn't allocate kernel stack");
363 current
->stack
= (void *) stack
;
364 #ifdef CONFIG_VMAP_STACK
365 current
->stack_vm_area
= (void *) stack
;
367 set_task_stack_end_magic(current
);
368 stack
+= STACK_INIT_OFFSET
;
369 S390_lowcore
.kernel_stack
= stack
;
370 CALL_ON_STACK_NORETURN(rest_init
, stack
);
373 static void __init
setup_lowcore_dat_off(void)
378 * Setup lowcore for boot cpu
380 BUILD_BUG_ON(sizeof(struct lowcore
) != LC_PAGES
* PAGE_SIZE
);
381 lc
= memblock_alloc_low(sizeof(*lc
), sizeof(*lc
));
383 panic("%s: Failed to allocate %zu bytes align=%zx\n",
384 __func__
, sizeof(*lc
), sizeof(*lc
));
386 lc
->restart_psw
.mask
= PSW_KERNEL_BITS
;
387 lc
->restart_psw
.addr
= (unsigned long) restart_int_handler
;
388 lc
->external_new_psw
.mask
= PSW_KERNEL_BITS
| PSW_MASK_MCHECK
;
389 lc
->external_new_psw
.addr
= (unsigned long) ext_int_handler
;
390 lc
->svc_new_psw
.mask
= PSW_KERNEL_BITS
|
391 PSW_MASK_IO
| PSW_MASK_EXT
| PSW_MASK_MCHECK
;
392 lc
->svc_new_psw
.addr
= (unsigned long) system_call
;
393 lc
->program_new_psw
.mask
= PSW_KERNEL_BITS
| PSW_MASK_MCHECK
;
394 lc
->program_new_psw
.addr
= (unsigned long) pgm_check_handler
;
395 lc
->mcck_new_psw
.mask
= PSW_KERNEL_BITS
;
396 lc
->mcck_new_psw
.addr
= (unsigned long) mcck_int_handler
;
397 lc
->io_new_psw
.mask
= PSW_KERNEL_BITS
| PSW_MASK_MCHECK
;
398 lc
->io_new_psw
.addr
= (unsigned long) io_int_handler
;
399 lc
->clock_comparator
= clock_comparator_max
;
400 lc
->nodat_stack
= ((unsigned long) &init_thread_union
)
401 + THREAD_SIZE
- STACK_FRAME_OVERHEAD
- sizeof(struct pt_regs
);
402 lc
->current_task
= (unsigned long)&init_task
;
404 lc
->machine_flags
= S390_lowcore
.machine_flags
;
405 lc
->preempt_count
= S390_lowcore
.preempt_count
;
406 lc
->stfl_fac_list
= S390_lowcore
.stfl_fac_list
;
407 memcpy(lc
->stfle_fac_list
, S390_lowcore
.stfle_fac_list
,
408 sizeof(lc
->stfle_fac_list
));
409 memcpy(lc
->alt_stfle_fac_list
, S390_lowcore
.alt_stfle_fac_list
,
410 sizeof(lc
->alt_stfle_fac_list
));
411 nmi_alloc_boot_cpu(lc
);
412 vdso_alloc_boot_cpu(lc
);
413 lc
->sync_enter_timer
= S390_lowcore
.sync_enter_timer
;
414 lc
->async_enter_timer
= S390_lowcore
.async_enter_timer
;
415 lc
->exit_timer
= S390_lowcore
.exit_timer
;
416 lc
->user_timer
= S390_lowcore
.user_timer
;
417 lc
->system_timer
= S390_lowcore
.system_timer
;
418 lc
->steal_timer
= S390_lowcore
.steal_timer
;
419 lc
->last_update_timer
= S390_lowcore
.last_update_timer
;
420 lc
->last_update_clock
= S390_lowcore
.last_update_clock
;
423 * Allocate the global restart stack which is the same for
424 * all CPUs in cast *one* of them does a PSW restart.
426 restart_stack
= 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 restart_stack
+= STACK_INIT_OFFSET
;
433 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
434 * restart data to the absolute zero lowcore. This is necessary if
435 * PSW restart is done on an offline CPU that has lowcore zero.
437 lc
->restart_stack
= (unsigned long) restart_stack
;
438 lc
->restart_fn
= (unsigned long) do_restart
;
439 lc
->restart_data
= 0;
440 lc
->restart_source
= -1UL;
442 /* Setup absolute zero lowcore */
443 mem_assign_absolute(S390_lowcore
.restart_stack
, lc
->restart_stack
);
444 mem_assign_absolute(S390_lowcore
.restart_fn
, lc
->restart_fn
);
445 mem_assign_absolute(S390_lowcore
.restart_data
, lc
->restart_data
);
446 mem_assign_absolute(S390_lowcore
.restart_source
, lc
->restart_source
);
447 mem_assign_absolute(S390_lowcore
.restart_psw
, lc
->restart_psw
);
449 lc
->spinlock_lockval
= arch_spin_lockval(0);
450 lc
->spinlock_index
= 0;
451 arch_spin_lock_setup(0);
452 lc
->br_r1_trampoline
= 0x07f1; /* br %r1 */
453 lc
->return_lpswe
= gen_lpswe(__LC_RETURN_PSW
);
454 lc
->return_mcck_lpswe
= gen_lpswe(__LC_RETURN_MCCK_PSW
);
456 set_prefix((u32
)(unsigned long) lc
);
460 static void __init
setup_lowcore_dat_on(void)
462 __ctl_clear_bit(0, 28);
463 S390_lowcore
.external_new_psw
.mask
|= PSW_MASK_DAT
;
464 S390_lowcore
.svc_new_psw
.mask
|= PSW_MASK_DAT
;
465 S390_lowcore
.program_new_psw
.mask
|= PSW_MASK_DAT
;
466 S390_lowcore
.io_new_psw
.mask
|= PSW_MASK_DAT
;
467 __ctl_set_bit(0, 28);
470 static struct resource code_resource
= {
471 .name
= "Kernel code",
472 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
,
475 static struct resource data_resource
= {
476 .name
= "Kernel data",
477 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
,
480 static struct resource bss_resource
= {
481 .name
= "Kernel bss",
482 .flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
,
485 static struct resource __initdata
*standard_resources
[] = {
491 static void __init
setup_resources(void)
493 struct resource
*res
, *std_res
, *sub_res
;
494 struct memblock_region
*reg
;
497 code_resource
.start
= (unsigned long) _text
;
498 code_resource
.end
= (unsigned long) _etext
- 1;
499 data_resource
.start
= (unsigned long) _etext
;
500 data_resource
.end
= (unsigned long) _edata
- 1;
501 bss_resource
.start
= (unsigned long) __bss_start
;
502 bss_resource
.end
= (unsigned long) __bss_stop
- 1;
504 for_each_memblock(memory
, reg
) {
505 res
= memblock_alloc(sizeof(*res
), 8);
507 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
508 __func__
, sizeof(*res
), 8);
509 res
->flags
= IORESOURCE_BUSY
| IORESOURCE_SYSTEM_RAM
;
511 res
->name
= "System RAM";
512 res
->start
= reg
->base
;
513 res
->end
= reg
->base
+ reg
->size
- 1;
514 request_resource(&iomem_resource
, res
);
516 for (j
= 0; j
< ARRAY_SIZE(standard_resources
); j
++) {
517 std_res
= standard_resources
[j
];
518 if (std_res
->start
< res
->start
||
519 std_res
->start
> res
->end
)
521 if (std_res
->end
> res
->end
) {
522 sub_res
= memblock_alloc(sizeof(*sub_res
), 8);
524 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
525 __func__
, sizeof(*sub_res
), 8);
527 sub_res
->end
= res
->end
;
528 std_res
->start
= res
->end
+ 1;
529 request_resource(res
, sub_res
);
531 request_resource(res
, std_res
);
535 #ifdef CONFIG_CRASH_DUMP
537 * Re-add removed crash kernel memory as reserved memory. This makes
538 * sure it will be mapped with the identity mapping and struct pages
539 * will be created, so it can be resized later on.
540 * However add it later since the crash kernel resource should not be
541 * part of the System RAM resource.
543 if (crashk_res
.end
) {
544 memblock_add_node(crashk_res
.start
, resource_size(&crashk_res
), 0);
545 memblock_reserve(crashk_res
.start
, resource_size(&crashk_res
));
546 insert_resource(&iomem_resource
, &crashk_res
);
551 static void __init
setup_memory_end(void)
553 unsigned long vmax
, tmp
;
555 /* Choose kernel address space layout: 3 or 4 levels. */
556 if (IS_ENABLED(CONFIG_KASAN
)) {
557 vmax
= IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING
)
561 tmp
= (memory_end
?: max_physmem_end
) / PAGE_SIZE
;
562 tmp
= tmp
* (sizeof(struct page
) + PAGE_SIZE
);
563 if (tmp
+ vmalloc_size
+ MODULES_LEN
<= _REGION2_SIZE
)
564 vmax
= _REGION2_SIZE
; /* 3-level kernel page table */
566 vmax
= _REGION1_SIZE
; /* 4-level kernel page table */
569 if (is_prot_virt_host())
570 adjust_to_uv_max(&vmax
);
572 /* module area is at the end of the kernel address space. */
574 MODULES_VADDR
= MODULES_END
- MODULES_LEN
;
575 VMALLOC_END
= MODULES_VADDR
;
576 VMALLOC_START
= VMALLOC_END
- vmalloc_size
;
578 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
579 tmp
= VMALLOC_START
/ (PAGE_SIZE
+ sizeof(struct page
));
580 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
581 tmp
= SECTION_ALIGN_UP(tmp
);
582 tmp
= VMALLOC_START
- tmp
* sizeof(struct page
);
583 tmp
&= ~((vmax
>> 11) - 1); /* align to page table level */
584 tmp
= min(tmp
, 1UL << MAX_PHYSMEM_BITS
);
585 vmemmap
= (struct page
*) tmp
;
587 /* Take care that memory_end is set and <= vmemmap */
588 memory_end
= min(memory_end
?: max_physmem_end
, (unsigned long)vmemmap
);
590 /* fit in kasan shadow memory region between 1:1 and vmemmap */
591 memory_end
= min(memory_end
, KASAN_SHADOW_START
);
592 vmemmap
= max(vmemmap
, (struct page
*)KASAN_SHADOW_END
);
594 max_pfn
= max_low_pfn
= PFN_DOWN(memory_end
);
595 memblock_remove(memory_end
, ULONG_MAX
);
597 pr_notice("The maximum memory size is %luMB\n", memory_end
>> 20);
600 #ifdef CONFIG_CRASH_DUMP
603 * When kdump is enabled, we have to ensure that no memory from
604 * the area [0 - crashkernel memory size] and
605 * [crashk_res.start - crashk_res.end] is set offline.
607 static int kdump_mem_notifier(struct notifier_block
*nb
,
608 unsigned long action
, void *data
)
610 struct memory_notify
*arg
= data
;
612 if (action
!= MEM_GOING_OFFLINE
)
614 if (arg
->start_pfn
< PFN_DOWN(resource_size(&crashk_res
)))
616 if (arg
->start_pfn
> PFN_DOWN(crashk_res
.end
))
618 if (arg
->start_pfn
+ arg
->nr_pages
- 1 < PFN_DOWN(crashk_res
.start
))
623 static struct notifier_block kdump_mem_nb
= {
624 .notifier_call
= kdump_mem_notifier
,
630 * Make sure that the area behind memory_end is protected
632 static void __init
reserve_memory_end(void)
635 memblock_reserve(memory_end
, ULONG_MAX
);
639 * Make sure that oldmem, where the dump is stored, is protected
641 static void __init
reserve_oldmem(void)
643 #ifdef CONFIG_CRASH_DUMP
645 /* Forget all memory above the running kdump system */
646 memblock_reserve(OLDMEM_SIZE
, (phys_addr_t
)ULONG_MAX
);
651 * Make sure that oldmem, where the dump is stored, is protected
653 static void __init
remove_oldmem(void)
655 #ifdef CONFIG_CRASH_DUMP
657 /* Forget all memory above the running kdump system */
658 memblock_remove(OLDMEM_SIZE
, (phys_addr_t
)ULONG_MAX
);
663 * Reserve memory for kdump kernel to be loaded with kexec
665 static void __init
reserve_crashkernel(void)
667 #ifdef CONFIG_CRASH_DUMP
668 unsigned long long crash_base
, crash_size
;
669 phys_addr_t low
, high
;
672 rc
= parse_crashkernel(boot_command_line
, memory_end
, &crash_size
,
675 crash_base
= ALIGN(crash_base
, KEXEC_CRASH_MEM_ALIGN
);
676 crash_size
= ALIGN(crash_size
, KEXEC_CRASH_MEM_ALIGN
);
677 if (rc
|| crash_size
== 0)
680 if (memblock
.memory
.regions
[0].size
< crash_size
) {
681 pr_info("crashkernel reservation failed: %s\n",
682 "first memory chunk must be at least crashkernel size");
686 low
= crash_base
?: OLDMEM_BASE
;
687 high
= low
+ crash_size
;
688 if (low
>= OLDMEM_BASE
&& high
<= OLDMEM_BASE
+ OLDMEM_SIZE
) {
689 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
692 /* Find suitable area in free memory */
693 low
= max_t(unsigned long, crash_size
, sclp
.hsa_size
);
694 high
= crash_base
? crash_base
+ crash_size
: ULONG_MAX
;
696 if (crash_base
&& crash_base
< low
) {
697 pr_info("crashkernel reservation failed: %s\n",
698 "crash_base too low");
701 low
= crash_base
?: low
;
702 crash_base
= memblock_find_in_range(low
, high
, crash_size
,
703 KEXEC_CRASH_MEM_ALIGN
);
707 pr_info("crashkernel reservation failed: %s\n",
708 "no suitable area found");
712 if (register_memory_notifier(&kdump_mem_nb
))
715 if (!OLDMEM_BASE
&& MACHINE_IS_VM
)
716 diag10_range(PFN_DOWN(crash_base
), PFN_DOWN(crash_size
));
717 crashk_res
.start
= crash_base
;
718 crashk_res
.end
= crash_base
+ crash_size
- 1;
719 memblock_remove(crash_base
, crash_size
);
720 pr_info("Reserving %lluMB of memory at %lluMB "
721 "for crashkernel (System RAM: %luMB)\n",
722 crash_size
>> 20, crash_base
>> 20,
723 (unsigned long)memblock
.memory
.total_size
>> 20);
724 os_info_crashkernel_add(crash_base
, crash_size
);
729 * Reserve the initrd from being used by memblock
731 static void __init
reserve_initrd(void)
733 #ifdef CONFIG_BLK_DEV_INITRD
734 if (!INITRD_START
|| !INITRD_SIZE
)
736 initrd_start
= INITRD_START
;
737 initrd_end
= initrd_start
+ INITRD_SIZE
;
738 memblock_reserve(INITRD_START
, INITRD_SIZE
);
743 * Reserve the memory area used to pass the certificate lists
745 static void __init
reserve_certificate_list(void)
747 if (ipl_cert_list_addr
)
748 memblock_reserve(ipl_cert_list_addr
, ipl_cert_list_size
);
751 static void __init
reserve_mem_detect_info(void)
753 unsigned long start
, size
;
755 get_mem_detect_reserved(&start
, &size
);
757 memblock_reserve(start
, size
);
760 static void __init
free_mem_detect_info(void)
762 unsigned long start
, size
;
764 get_mem_detect_reserved(&start
, &size
);
766 memblock_free(start
, size
);
769 static void __init
memblock_physmem_add(phys_addr_t start
, phys_addr_t size
)
771 memblock_dbg("memblock_physmem_add: [%#016llx-%#016llx]\n",
772 start
, start
+ size
- 1);
773 memblock_add_range(&memblock
.memory
, start
, size
, 0, 0);
774 memblock_add_range(&memblock
.physmem
, start
, size
, 0, 0);
777 static const char * __init
get_mem_info_source(void)
779 switch (mem_detect
.info_source
) {
780 case MEM_DETECT_SCLP_STOR_INFO
:
781 return "sclp storage info";
782 case MEM_DETECT_DIAG260
:
784 case MEM_DETECT_SCLP_READ_INFO
:
785 return "sclp read info";
786 case MEM_DETECT_BIN_SEARCH
:
787 return "binary search";
792 static void __init
memblock_add_mem_detect_info(void)
794 unsigned long start
, end
;
797 memblock_dbg("physmem info source: %s (%hhd)\n",
798 get_mem_info_source(), mem_detect
.info_source
);
799 /* keep memblock lists close to the kernel */
800 memblock_set_bottom_up(true);
801 for_each_mem_detect_block(i
, &start
, &end
)
802 memblock_physmem_add(start
, end
- start
);
803 memblock_set_bottom_up(false);
808 * Check for initrd being in usable memory
810 static void __init
check_initrd(void)
812 #ifdef CONFIG_BLK_DEV_INITRD
813 if (INITRD_START
&& INITRD_SIZE
&&
814 !memblock_is_region_memory(INITRD_START
, INITRD_SIZE
)) {
815 pr_err("The initial RAM disk does not fit into the memory\n");
816 memblock_free(INITRD_START
, INITRD_SIZE
);
817 initrd_start
= initrd_end
= 0;
823 * Reserve memory used for lowcore/command line/kernel image.
825 static void __init
reserve_kernel(void)
827 unsigned long start_pfn
= PFN_UP(__pa(_end
));
829 memblock_reserve(0, HEAD_END
);
830 memblock_reserve((unsigned long)_stext
, PFN_PHYS(start_pfn
)
831 - (unsigned long)_stext
);
832 memblock_reserve(__sdma
, __edma
- __sdma
);
835 static void __init
setup_memory(void)
837 struct memblock_region
*reg
;
840 * Init storage key for present memory
842 for_each_memblock(memory
, reg
) {
843 storage_key_init_range(reg
->base
, reg
->base
+ reg
->size
);
845 psw_set_key(PAGE_DEFAULT_KEY
);
849 * Setup hardware capabilities.
851 static int __init
setup_hwcaps(void)
853 static const int stfl_bits
[6] = { 0, 2, 7, 17, 19, 21 };
858 * The store facility list bits numbers as found in the principles
859 * of operation are numbered with bit 1UL<<31 as number 0 to
860 * bit 1UL<<0 as number 31.
861 * Bit 0: instructions named N3, "backported" to esa-mode
862 * Bit 2: z/Architecture mode is active
863 * Bit 7: the store-facility-list-extended facility is installed
864 * Bit 17: the message-security assist is installed
865 * Bit 19: the long-displacement facility is installed
866 * Bit 21: the extended-immediate facility is installed
867 * Bit 22: extended-translation facility 3 is installed
868 * Bit 30: extended-translation facility 3 enhancement facility
869 * These get translated to:
870 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
871 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
872 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
873 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
875 for (i
= 0; i
< 6; i
++)
876 if (test_facility(stfl_bits
[i
]))
877 elf_hwcap
|= 1UL << i
;
879 if (test_facility(22) && test_facility(30))
880 elf_hwcap
|= HWCAP_S390_ETF3EH
;
883 * Check for additional facilities with store-facility-list-extended.
884 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
885 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
886 * as stored by stfl, bits 32-xxx contain additional facilities.
887 * How many facility words are stored depends on the number of
888 * doublewords passed to the instruction. The additional facilities
890 * Bit 42: decimal floating point facility is installed
891 * Bit 44: perform floating point operation facility is installed
893 * HWCAP_S390_DFP bit 6 (42 && 44).
895 if ((elf_hwcap
& (1UL << 2)) && test_facility(42) && test_facility(44))
896 elf_hwcap
|= HWCAP_S390_DFP
;
899 * Huge page support HWCAP_S390_HPAGE is bit 7.
901 if (MACHINE_HAS_EDAT1
)
902 elf_hwcap
|= HWCAP_S390_HPAGE
;
905 * 64-bit register support for 31-bit processes
906 * HWCAP_S390_HIGH_GPRS is bit 9.
908 elf_hwcap
|= HWCAP_S390_HIGH_GPRS
;
911 * Transactional execution support HWCAP_S390_TE is bit 10.
914 elf_hwcap
|= HWCAP_S390_TE
;
917 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
918 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
919 * instead of facility bit 129.
921 if (MACHINE_HAS_VX
) {
922 elf_hwcap
|= HWCAP_S390_VXRS
;
923 if (test_facility(134))
924 elf_hwcap
|= HWCAP_S390_VXRS_BCD
;
925 if (test_facility(135))
926 elf_hwcap
|= HWCAP_S390_VXRS_EXT
;
927 if (test_facility(148))
928 elf_hwcap
|= HWCAP_S390_VXRS_EXT2
;
929 if (test_facility(152))
930 elf_hwcap
|= HWCAP_S390_VXRS_PDE
;
932 if (test_facility(150))
933 elf_hwcap
|= HWCAP_S390_SORT
;
934 if (test_facility(151))
935 elf_hwcap
|= HWCAP_S390_DFLT
;
938 * Guarded storage support HWCAP_S390_GS is bit 12.
941 elf_hwcap
|= HWCAP_S390_GS
;
944 add_device_randomness(&cpu_id
, sizeof(cpu_id
));
945 switch (cpu_id
.machine
) {
948 default: /* Use "z900" as default for 64 bit kernels. */
949 strcpy(elf_platform
, "z900");
953 strcpy(elf_platform
, "z990");
957 strcpy(elf_platform
, "z9-109");
961 strcpy(elf_platform
, "z10");
965 strcpy(elf_platform
, "z196");
969 strcpy(elf_platform
, "zEC12");
973 strcpy(elf_platform
, "z13");
977 strcpy(elf_platform
, "z14");
981 strcpy(elf_platform
, "z15");
986 * Virtualization support HWCAP_INT_SIE is bit 0.
989 int_hwcap
|= HWCAP_INT_SIE
;
993 arch_initcall(setup_hwcaps
);
996 * Add system information as device randomness
998 static void __init
setup_randomness(void)
1000 struct sysinfo_3_2_2
*vmms
;
1002 vmms
= (struct sysinfo_3_2_2
*) memblock_phys_alloc(PAGE_SIZE
,
1005 panic("Failed to allocate memory for sysinfo structure\n");
1007 if (stsi(vmms
, 3, 2, 2) == 0 && vmms
->count
)
1008 add_device_randomness(&vmms
->vm
, sizeof(vmms
->vm
[0]) * vmms
->count
);
1009 memblock_free((unsigned long) vmms
, PAGE_SIZE
);
1013 * Find the correct size for the task_struct. This depends on
1014 * the size of the struct fpu at the end of the thread_struct
1015 * which is embedded in the task_struct.
1017 static void __init
setup_task_size(void)
1019 int task_size
= sizeof(struct task_struct
);
1021 if (!MACHINE_HAS_VX
) {
1022 task_size
-= sizeof(__vector128
) * __NUM_VXRS
;
1023 task_size
+= sizeof(freg_t
) * __NUM_FPRS
;
1025 arch_task_struct_size
= task_size
;
1029 * Issue diagnose 318 to set the control program name and
1032 static void __init
setup_control_program_code(void)
1034 union diag318_info diag318_info
= {
1039 if (!sclp
.has_diag318
)
1042 diag_stat_inc(DIAG_STAT_X318
);
1043 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info
.val
));
1047 * Print the component list from the IPL report
1049 static void __init
log_component_list(void)
1051 struct ipl_rb_component_entry
*ptr
, *end
;
1054 if (!early_ipl_comp_list_addr
)
1056 if (ipl_block
.hdr
.flags
& IPL_PL_FLAG_SIPL
)
1057 pr_info("Linux is running with Secure-IPL enabled\n");
1059 pr_info("Linux is running with Secure-IPL disabled\n");
1060 ptr
= (void *) early_ipl_comp_list_addr
;
1061 end
= (void *) ptr
+ early_ipl_comp_list_size
;
1062 pr_info("The IPL report contains the following components:\n");
1064 if (ptr
->flags
& IPL_RB_COMPONENT_FLAG_SIGNED
) {
1065 if (ptr
->flags
& IPL_RB_COMPONENT_FLAG_VERIFIED
)
1066 str
= "signed, verified";
1068 str
= "signed, verification failed";
1072 pr_info("%016llx - %016llx (%s)\n",
1073 ptr
->addr
, ptr
->addr
+ ptr
->len
, str
);
1079 * Setup function called from init/main.c just after the banner
1083 void __init
setup_arch(char **cmdline_p
)
1086 * print what head.S has found out about the machine
1089 pr_info("Linux is running as a z/VM "
1090 "guest operating system in 64-bit mode\n");
1091 else if (MACHINE_IS_KVM
)
1092 pr_info("Linux is running under KVM in 64-bit mode\n");
1093 else if (MACHINE_IS_LPAR
)
1094 pr_info("Linux is running natively in 64-bit mode\n");
1096 pr_info("Linux is running as a guest in 64-bit mode\n");
1098 log_component_list();
1100 #ifdef CONFIG_LOCK_DOWN_IN_SECURE_BOOT
1101 if (ipl_get_secureboot())
1102 security_lock_kernel_down("Secure IPL",
1103 LOCKDOWN_INTEGRITY_MAX
);
1106 /* Have one command line that is parsed and saved in /proc/cmdline */
1107 /* boot_command_line has been already set up in early.c */
1108 *cmdline_p
= boot_command_line
;
1110 ROOT_DEV
= Root_RAM0
;
1112 init_mm
.start_code
= (unsigned long) _text
;
1113 init_mm
.end_code
= (unsigned long) _etext
;
1114 init_mm
.end_data
= (unsigned long) _edata
;
1115 init_mm
.brk
= (unsigned long) _end
;
1117 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO
))
1118 nospec_auto_detect();
1121 parse_early_param();
1122 #ifdef CONFIG_CRASH_DUMP
1123 /* Deactivate elfcorehdr= kernel parameter */
1124 elfcorehdr_addr
= ELFCORE_ADDR_MAX
;
1130 setup_control_program_code();
1132 /* Do some memory reservations *before* memory is added to memblock */
1133 reserve_memory_end();
1137 reserve_certificate_list();
1138 reserve_mem_detect_info();
1139 memblock_allow_resize();
1141 /* Get information about *all* installed memory */
1142 memblock_add_mem_detect_info();
1144 free_mem_detect_info();
1148 * Make sure all chunks are MAX_ORDER aligned so we don't need the
1149 * extra checks that HOLES_IN_ZONE would require.
1151 * Is this still required?
1153 memblock_trim_memory(1UL << (MAX_ORDER
- 1 + PAGE_SHIFT
));
1155 if (is_prot_virt_host())
1159 dma_contiguous_reserve(memory_end
);
1163 reserve_crashkernel();
1164 #ifdef CONFIG_CRASH_DUMP
1166 * Be aware that smp_save_dump_cpus() triggers a system reset.
1167 * Therefore CPU and device initialization should be done afterwards.
1169 smp_save_dump_cpus();
1173 setup_lowcore_dat_off();
1174 smp_fill_possible_mask();
1175 cpu_detect_mhz_feature();
1179 topology_init_early();
1182 * Create kernel page tables and switch to virtual addressing.
1187 * After paging_init created the kernel page table, the new PSWs
1188 * in lowcore can now run with DAT enabled.
1190 setup_lowcore_dat_on();
1192 /* Setup default console */
1194 set_preferred_console();
1196 apply_alternative_instructions();
1197 if (IS_ENABLED(CONFIG_EXPOLINE
))
1198 nospec_init_branches();
1200 /* Setup zfcpdump support */
1203 /* Add system specific data to the random pool */