xref: /openbmc/linux/arch/s390/kernel/setup.c (revision 37002bc6)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  S390 version
4  *    Copyright IBM Corp. 1999, 2012
5  *    Author(s): Hartmut Penner (hp@de.ibm.com),
6  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
7  *
8  *  Derived from "arch/i386/kernel/setup.c"
9  *    Copyright (C) 1995, Linus Torvalds
10  */
11 
12 /*
13  * This file handles the architecture-dependent parts of initialization
14  */
15 
16 #define KMSG_COMPONENT "setup"
17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
18 
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>
26 #include <linux/mm.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>
53 #include <linux/kmemleak.h>
54 
55 #include <asm/archrandom.h>
56 #include <asm/boot_data.h>
57 #include <asm/ipl.h>
58 #include <asm/facility.h>
59 #include <asm/smp.h>
60 #include <asm/mmu_context.h>
61 #include <asm/cpcmd.h>
62 #include <asm/abs_lowcore.h>
63 #include <asm/nmi.h>
64 #include <asm/irq.h>
65 #include <asm/page.h>
66 #include <asm/ptrace.h>
67 #include <asm/sections.h>
68 #include <asm/ebcdic.h>
69 #include <asm/diag.h>
70 #include <asm/os_info.h>
71 #include <asm/sclp.h>
72 #include <asm/stacktrace.h>
73 #include <asm/sysinfo.h>
74 #include <asm/numa.h>
75 #include <asm/alternative.h>
76 #include <asm/nospec-branch.h>
77 #include <asm/physmem_info.h>
78 #include <asm/maccess.h>
79 #include <asm/uv.h>
80 #include <asm/asm-offsets.h>
81 #include "entry.h"
82 
83 /*
84  * Machine setup..
85  */
86 unsigned int console_mode = 0;
87 EXPORT_SYMBOL(console_mode);
88 
89 unsigned int console_devno = -1;
90 EXPORT_SYMBOL(console_devno);
91 
92 unsigned int console_irq = -1;
93 EXPORT_SYMBOL(console_irq);
94 
95 /*
96  * Some code and data needs to stay below 2 GB, even when the kernel would be
97  * relocated above 2 GB, because it has to use 31 bit addresses.
98  * Such code and data is part of the .amode31 section.
99  */
100 unsigned long __amode31_ref __samode31 = (unsigned long)&_samode31;
101 unsigned long __amode31_ref __eamode31 = (unsigned long)&_eamode31;
102 unsigned long __amode31_ref __stext_amode31 = (unsigned long)&_stext_amode31;
103 unsigned long __amode31_ref __etext_amode31 = (unsigned long)&_etext_amode31;
104 struct exception_table_entry __amode31_ref *__start_amode31_ex_table = _start_amode31_ex_table;
105 struct exception_table_entry __amode31_ref *__stop_amode31_ex_table = _stop_amode31_ex_table;
106 
107 /*
108  * Control registers CR2, CR5 and CR15 are initialized with addresses
109  * of tables that must be placed below 2G which is handled by the AMODE31
110  * sections.
111  * Because the AMODE31 sections are relocated below 2G at startup,
112  * the content of control registers CR2, CR5 and CR15 must be updated
113  * with new addresses after the relocation. The initial initialization of
114  * control registers occurs in head64.S and then gets updated again after AMODE31
115  * relocation. We must access the relevant AMODE31 tables indirectly via
116  * pointers placed in the .amode31.refs linker section. Those pointers get
117  * updated automatically during AMODE31 relocation and always contain a valid
118  * address within AMODE31 sections.
119  */
120 
121 static __amode31_data u32 __ctl_duct_amode31[16] __aligned(64);
122 
123 static __amode31_data u64 __ctl_aste_amode31[8] __aligned(64) = {
124 	[1] = 0xffffffffffffffff
125 };
126 
127 static __amode31_data u32 __ctl_duald_amode31[32] __aligned(128) = {
128 	0x80000000, 0, 0, 0,
129 	0x80000000, 0, 0, 0,
130 	0x80000000, 0, 0, 0,
131 	0x80000000, 0, 0, 0,
132 	0x80000000, 0, 0, 0,
133 	0x80000000, 0, 0, 0,
134 	0x80000000, 0, 0, 0,
135 	0x80000000, 0, 0, 0
136 };
137 
138 static __amode31_data u32 __ctl_linkage_stack_amode31[8] __aligned(64) = {
139 	0, 0, 0x89000000, 0,
140 	0, 0, 0x8a000000, 0
141 };
142 
143 static u64 __amode31_ref *__ctl_aste = __ctl_aste_amode31;
144 static u32 __amode31_ref *__ctl_duald = __ctl_duald_amode31;
145 static u32 __amode31_ref *__ctl_linkage_stack = __ctl_linkage_stack_amode31;
146 static u32 __amode31_ref *__ctl_duct = __ctl_duct_amode31;
147 
148 int __bootdata(noexec_disabled);
149 unsigned long __bootdata_preserved(max_mappable);
150 unsigned long __bootdata(ident_map_size);
151 struct physmem_info __bootdata(physmem_info);
152 
153 unsigned long __bootdata_preserved(__kaslr_offset);
154 int __bootdata_preserved(__kaslr_enabled);
155 unsigned int __bootdata_preserved(zlib_dfltcc_support);
156 EXPORT_SYMBOL(zlib_dfltcc_support);
157 u64 __bootdata_preserved(stfle_fac_list[16]);
158 EXPORT_SYMBOL(stfle_fac_list);
159 u64 __bootdata_preserved(alt_stfle_fac_list[16]);
160 struct oldmem_data __bootdata_preserved(oldmem_data);
161 
162 unsigned long VMALLOC_START;
163 EXPORT_SYMBOL(VMALLOC_START);
164 
165 unsigned long VMALLOC_END;
166 EXPORT_SYMBOL(VMALLOC_END);
167 
168 struct page *vmemmap;
169 EXPORT_SYMBOL(vmemmap);
170 unsigned long vmemmap_size;
171 
172 unsigned long MODULES_VADDR;
173 unsigned long MODULES_END;
174 
175 /* An array with a pointer to the lowcore of every CPU. */
176 struct lowcore *lowcore_ptr[NR_CPUS];
177 EXPORT_SYMBOL(lowcore_ptr);
178 
179 DEFINE_STATIC_KEY_FALSE(cpu_has_bear);
180 
181 /*
182  * The Write Back bit position in the physaddr is given by the SLPC PCI.
183  * Leaving the mask zero always uses write through which is safe
184  */
185 unsigned long mio_wb_bit_mask __ro_after_init;
186 
187 /*
188  * This is set up by the setup-routine at boot-time
189  * for S390 need to find out, what we have to setup
190  * using address 0x10400 ...
191  */
192 
193 #include <asm/setup.h>
194 
195 /*
196  * condev= and conmode= setup parameter.
197  */
198 
199 static int __init condev_setup(char *str)
200 {
201 	int vdev;
202 
203 	vdev = simple_strtoul(str, &str, 0);
204 	if (vdev >= 0 && vdev < 65536) {
205 		console_devno = vdev;
206 		console_irq = -1;
207 	}
208 	return 1;
209 }
210 
211 __setup("condev=", condev_setup);
212 
213 static void __init set_preferred_console(void)
214 {
215 	if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
216 		add_preferred_console("ttyS", 0, NULL);
217 	else if (CONSOLE_IS_3270)
218 		add_preferred_console("tty3270", 0, NULL);
219 	else if (CONSOLE_IS_VT220)
220 		add_preferred_console("ttysclp", 0, NULL);
221 	else if (CONSOLE_IS_HVC)
222 		add_preferred_console("hvc", 0, NULL);
223 }
224 
225 static int __init conmode_setup(char *str)
226 {
227 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
228 	if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
229                 SET_CONSOLE_SCLP;
230 #endif
231 #if defined(CONFIG_TN3215_CONSOLE)
232 	if (!strcmp(str, "3215"))
233 		SET_CONSOLE_3215;
234 #endif
235 #if defined(CONFIG_TN3270_CONSOLE)
236 	if (!strcmp(str, "3270"))
237 		SET_CONSOLE_3270;
238 #endif
239 	set_preferred_console();
240         return 1;
241 }
242 
243 __setup("conmode=", conmode_setup);
244 
245 static void __init conmode_default(void)
246 {
247 	char query_buffer[1024];
248 	char *ptr;
249 
250         if (MACHINE_IS_VM) {
251 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
252 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
253 		ptr = strstr(query_buffer, "SUBCHANNEL =");
254 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
255 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
256 		ptr = strstr(query_buffer, "CONMODE");
257 		/*
258 		 * Set the conmode to 3215 so that the device recognition
259 		 * will set the cu_type of the console to 3215. If the
260 		 * conmode is 3270 and we don't set it back then both
261 		 * 3215 and the 3270 driver will try to access the console
262 		 * device (3215 as console and 3270 as normal tty).
263 		 */
264 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
265 		if (ptr == NULL) {
266 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
267 			SET_CONSOLE_SCLP;
268 #endif
269 			return;
270 		}
271 		if (str_has_prefix(ptr + 8, "3270")) {
272 #if defined(CONFIG_TN3270_CONSOLE)
273 			SET_CONSOLE_3270;
274 #elif defined(CONFIG_TN3215_CONSOLE)
275 			SET_CONSOLE_3215;
276 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
277 			SET_CONSOLE_SCLP;
278 #endif
279 		} else if (str_has_prefix(ptr + 8, "3215")) {
280 #if defined(CONFIG_TN3215_CONSOLE)
281 			SET_CONSOLE_3215;
282 #elif defined(CONFIG_TN3270_CONSOLE)
283 			SET_CONSOLE_3270;
284 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
285 			SET_CONSOLE_SCLP;
286 #endif
287 		}
288 	} else if (MACHINE_IS_KVM) {
289 		if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
290 			SET_CONSOLE_VT220;
291 		else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
292 			SET_CONSOLE_SCLP;
293 		else
294 			SET_CONSOLE_HVC;
295 	} else {
296 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
297 		SET_CONSOLE_SCLP;
298 #endif
299 	}
300 }
301 
302 #ifdef CONFIG_CRASH_DUMP
303 static void __init setup_zfcpdump(void)
304 {
305 	if (!is_ipl_type_dump())
306 		return;
307 	if (oldmem_data.start)
308 		return;
309 	strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
310 	console_loglevel = 2;
311 }
312 #else
313 static inline void setup_zfcpdump(void) {}
314 #endif /* CONFIG_CRASH_DUMP */
315 
316  /*
317  * Reboot, halt and power_off stubs. They just call _machine_restart,
318  * _machine_halt or _machine_power_off.
319  */
320 
321 void machine_restart(char *command)
322 {
323 	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
324 		/*
325 		 * Only unblank the console if we are called in enabled
326 		 * context or a bust_spinlocks cleared the way for us.
327 		 */
328 		console_unblank();
329 	_machine_restart(command);
330 }
331 
332 void machine_halt(void)
333 {
334 	if (!in_interrupt() || oops_in_progress)
335 		/*
336 		 * Only unblank the console if we are called in enabled
337 		 * context or a bust_spinlocks cleared the way for us.
338 		 */
339 		console_unblank();
340 	_machine_halt();
341 }
342 
343 void machine_power_off(void)
344 {
345 	if (!in_interrupt() || oops_in_progress)
346 		/*
347 		 * Only unblank the console if we are called in enabled
348 		 * context or a bust_spinlocks cleared the way for us.
349 		 */
350 		console_unblank();
351 	_machine_power_off();
352 }
353 
354 /*
355  * Dummy power off function.
356  */
357 void (*pm_power_off)(void) = machine_power_off;
358 EXPORT_SYMBOL_GPL(pm_power_off);
359 
360 void *restart_stack;
361 
362 unsigned long stack_alloc(void)
363 {
364 #ifdef CONFIG_VMAP_STACK
365 	void *ret;
366 
367 	ret = __vmalloc_node(THREAD_SIZE, THREAD_SIZE, THREADINFO_GFP,
368 			     NUMA_NO_NODE, __builtin_return_address(0));
369 	kmemleak_not_leak(ret);
370 	return (unsigned long)ret;
371 #else
372 	return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
373 #endif
374 }
375 
376 void stack_free(unsigned long stack)
377 {
378 #ifdef CONFIG_VMAP_STACK
379 	vfree((void *) stack);
380 #else
381 	free_pages(stack, THREAD_SIZE_ORDER);
382 #endif
383 }
384 
385 void __init __noreturn arch_call_rest_init(void)
386 {
387 	smp_reinit_ipl_cpu();
388 	rest_init();
389 }
390 
391 static unsigned long __init stack_alloc_early(void)
392 {
393 	unsigned long stack;
394 
395 	stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
396 	if (!stack) {
397 		panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
398 		      __func__, THREAD_SIZE, THREAD_SIZE);
399 	}
400 	return stack;
401 }
402 
403 static void __init setup_lowcore(void)
404 {
405 	struct lowcore *lc, *abs_lc;
406 
407 	/*
408 	 * Setup lowcore for boot cpu
409 	 */
410 	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
411 	lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
412 	if (!lc)
413 		panic("%s: Failed to allocate %zu bytes align=%zx\n",
414 		      __func__, sizeof(*lc), sizeof(*lc));
415 
416 	lc->restart_psw.mask = PSW_KERNEL_BITS & ~PSW_MASK_DAT;
417 	lc->restart_psw.addr = __pa(restart_int_handler);
418 	lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
419 	lc->external_new_psw.addr = (unsigned long) ext_int_handler;
420 	lc->svc_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
421 	lc->svc_new_psw.addr = (unsigned long) system_call;
422 	lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
423 	lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
424 	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
425 	lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
426 	lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
427 	lc->io_new_psw.addr = (unsigned long) io_int_handler;
428 	lc->clock_comparator = clock_comparator_max;
429 	lc->current_task = (unsigned long)&init_task;
430 	lc->lpp = LPP_MAGIC;
431 	lc->machine_flags = S390_lowcore.machine_flags;
432 	lc->preempt_count = S390_lowcore.preempt_count;
433 	nmi_alloc_mcesa_early(&lc->mcesad);
434 	lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
435 	lc->exit_timer = S390_lowcore.exit_timer;
436 	lc->user_timer = S390_lowcore.user_timer;
437 	lc->system_timer = S390_lowcore.system_timer;
438 	lc->steal_timer = S390_lowcore.steal_timer;
439 	lc->last_update_timer = S390_lowcore.last_update_timer;
440 	lc->last_update_clock = S390_lowcore.last_update_clock;
441 	/*
442 	 * Allocate the global restart stack which is the same for
443 	 * all CPUs in case *one* of them does a PSW restart.
444 	 */
445 	restart_stack = (void *)(stack_alloc_early() + STACK_INIT_OFFSET);
446 	lc->mcck_stack = stack_alloc_early() + STACK_INIT_OFFSET;
447 	lc->async_stack = stack_alloc_early() + STACK_INIT_OFFSET;
448 	lc->nodat_stack = stack_alloc_early() + STACK_INIT_OFFSET;
449 	lc->kernel_stack = S390_lowcore.kernel_stack;
450 	/*
451 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
452 	 * restart data to the absolute zero lowcore. This is necessary if
453 	 * PSW restart is done on an offline CPU that has lowcore zero.
454 	 */
455 	lc->restart_stack = (unsigned long) restart_stack;
456 	lc->restart_fn = (unsigned long) do_restart;
457 	lc->restart_data = 0;
458 	lc->restart_source = -1U;
459 	__ctl_store(lc->cregs_save_area, 0, 15);
460 	lc->spinlock_lockval = arch_spin_lockval(0);
461 	lc->spinlock_index = 0;
462 	arch_spin_lock_setup(0);
463 	lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
464 	lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
465 	lc->preempt_count = PREEMPT_DISABLED;
466 	lc->kernel_asce = S390_lowcore.kernel_asce;
467 	lc->user_asce = S390_lowcore.user_asce;
468 
469 	abs_lc = get_abs_lowcore();
470 	abs_lc->restart_stack = lc->restart_stack;
471 	abs_lc->restart_fn = lc->restart_fn;
472 	abs_lc->restart_data = lc->restart_data;
473 	abs_lc->restart_source = lc->restart_source;
474 	abs_lc->restart_psw = lc->restart_psw;
475 	abs_lc->restart_flags = RESTART_FLAG_CTLREGS;
476 	memcpy(abs_lc->cregs_save_area, lc->cregs_save_area, sizeof(abs_lc->cregs_save_area));
477 	abs_lc->program_new_psw = lc->program_new_psw;
478 	abs_lc->mcesad = lc->mcesad;
479 	put_abs_lowcore(abs_lc);
480 
481 	set_prefix(__pa(lc));
482 	lowcore_ptr[0] = lc;
483 	if (abs_lowcore_map(0, lowcore_ptr[0], false))
484 		panic("Couldn't setup absolute lowcore");
485 }
486 
487 static struct resource code_resource = {
488 	.name  = "Kernel code",
489 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
490 };
491 
492 static struct resource data_resource = {
493 	.name = "Kernel data",
494 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
495 };
496 
497 static struct resource bss_resource = {
498 	.name = "Kernel bss",
499 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
500 };
501 
502 static struct resource __initdata *standard_resources[] = {
503 	&code_resource,
504 	&data_resource,
505 	&bss_resource,
506 };
507 
508 static void __init setup_resources(void)
509 {
510 	struct resource *res, *std_res, *sub_res;
511 	phys_addr_t start, end;
512 	int j;
513 	u64 i;
514 
515 	code_resource.start = (unsigned long) _text;
516 	code_resource.end = (unsigned long) _etext - 1;
517 	data_resource.start = (unsigned long) _etext;
518 	data_resource.end = (unsigned long) _edata - 1;
519 	bss_resource.start = (unsigned long) __bss_start;
520 	bss_resource.end = (unsigned long) __bss_stop - 1;
521 
522 	for_each_mem_range(i, &start, &end) {
523 		res = memblock_alloc(sizeof(*res), 8);
524 		if (!res)
525 			panic("%s: Failed to allocate %zu bytes align=0x%x\n",
526 			      __func__, sizeof(*res), 8);
527 		res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
528 
529 		res->name = "System RAM";
530 		res->start = start;
531 		/*
532 		 * In memblock, end points to the first byte after the
533 		 * range while in resources, end points to the last byte in
534 		 * the range.
535 		 */
536 		res->end = end - 1;
537 		request_resource(&iomem_resource, res);
538 
539 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
540 			std_res = standard_resources[j];
541 			if (std_res->start < res->start ||
542 			    std_res->start > res->end)
543 				continue;
544 			if (std_res->end > res->end) {
545 				sub_res = memblock_alloc(sizeof(*sub_res), 8);
546 				if (!sub_res)
547 					panic("%s: Failed to allocate %zu bytes align=0x%x\n",
548 					      __func__, sizeof(*sub_res), 8);
549 				*sub_res = *std_res;
550 				sub_res->end = res->end;
551 				std_res->start = res->end + 1;
552 				request_resource(res, sub_res);
553 			} else {
554 				request_resource(res, std_res);
555 			}
556 		}
557 	}
558 #ifdef CONFIG_CRASH_DUMP
559 	/*
560 	 * Re-add removed crash kernel memory as reserved memory. This makes
561 	 * sure it will be mapped with the identity mapping and struct pages
562 	 * will be created, so it can be resized later on.
563 	 * However add it later since the crash kernel resource should not be
564 	 * part of the System RAM resource.
565 	 */
566 	if (crashk_res.end) {
567 		memblock_add_node(crashk_res.start, resource_size(&crashk_res),
568 				  0, MEMBLOCK_NONE);
569 		memblock_reserve(crashk_res.start, resource_size(&crashk_res));
570 		insert_resource(&iomem_resource, &crashk_res);
571 	}
572 #endif
573 }
574 
575 static void __init setup_memory_end(void)
576 {
577 	max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
578 	pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
579 }
580 
581 #ifdef CONFIG_CRASH_DUMP
582 
583 /*
584  * When kdump is enabled, we have to ensure that no memory from the area
585  * [0 - crashkernel memory size] is set offline - it will be exchanged with
586  * the crashkernel memory region when kdump is triggered. The crashkernel
587  * memory region can never get offlined (pages are unmovable).
588  */
589 static int kdump_mem_notifier(struct notifier_block *nb,
590 			      unsigned long action, void *data)
591 {
592 	struct memory_notify *arg = data;
593 
594 	if (action != MEM_GOING_OFFLINE)
595 		return NOTIFY_OK;
596 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
597 		return NOTIFY_BAD;
598 	return NOTIFY_OK;
599 }
600 
601 static struct notifier_block kdump_mem_nb = {
602 	.notifier_call = kdump_mem_notifier,
603 };
604 
605 #endif
606 
607 /*
608  * Reserve page tables created by decompressor
609  */
610 static void __init reserve_pgtables(void)
611 {
612 	unsigned long start, end;
613 	struct reserved_range *range;
614 
615 	for_each_physmem_reserved_type_range(RR_VMEM, range, &start, &end)
616 		memblock_reserve(start, end - start);
617 }
618 
619 /*
620  * Reserve memory for kdump kernel to be loaded with kexec
621  */
622 static void __init reserve_crashkernel(void)
623 {
624 #ifdef CONFIG_CRASH_DUMP
625 	unsigned long long crash_base, crash_size;
626 	phys_addr_t low, high;
627 	int rc;
628 
629 	rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
630 			       &crash_base);
631 
632 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
633 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
634 	if (rc || crash_size == 0)
635 		return;
636 
637 	if (memblock.memory.regions[0].size < crash_size) {
638 		pr_info("crashkernel reservation failed: %s\n",
639 			"first memory chunk must be at least crashkernel size");
640 		return;
641 	}
642 
643 	low = crash_base ?: oldmem_data.start;
644 	high = low + crash_size;
645 	if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
646 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
647 		crash_base = low;
648 	} else {
649 		/* Find suitable area in free memory */
650 		low = max_t(unsigned long, crash_size, sclp.hsa_size);
651 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
652 
653 		if (crash_base && crash_base < low) {
654 			pr_info("crashkernel reservation failed: %s\n",
655 				"crash_base too low");
656 			return;
657 		}
658 		low = crash_base ?: low;
659 		crash_base = memblock_phys_alloc_range(crash_size,
660 						       KEXEC_CRASH_MEM_ALIGN,
661 						       low, high);
662 	}
663 
664 	if (!crash_base) {
665 		pr_info("crashkernel reservation failed: %s\n",
666 			"no suitable area found");
667 		return;
668 	}
669 
670 	if (register_memory_notifier(&kdump_mem_nb)) {
671 		memblock_phys_free(crash_base, crash_size);
672 		return;
673 	}
674 
675 	if (!oldmem_data.start && MACHINE_IS_VM)
676 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
677 	crashk_res.start = crash_base;
678 	crashk_res.end = crash_base + crash_size - 1;
679 	memblock_remove(crash_base, crash_size);
680 	pr_info("Reserving %lluMB of memory at %lluMB "
681 		"for crashkernel (System RAM: %luMB)\n",
682 		crash_size >> 20, crash_base >> 20,
683 		(unsigned long)memblock.memory.total_size >> 20);
684 	os_info_crashkernel_add(crash_base, crash_size);
685 #endif
686 }
687 
688 /*
689  * Reserve the initrd from being used by memblock
690  */
691 static void __init reserve_initrd(void)
692 {
693 	unsigned long addr, size;
694 
695 	if (!IS_ENABLED(CONFIG_BLK_DEV_INITRD) || !get_physmem_reserved(RR_INITRD, &addr, &size))
696 		return;
697 	initrd_start = (unsigned long)__va(addr);
698 	initrd_end = initrd_start + size;
699 	memblock_reserve(addr, size);
700 }
701 
702 /*
703  * Reserve the memory area used to pass the certificate lists
704  */
705 static void __init reserve_certificate_list(void)
706 {
707 	if (ipl_cert_list_addr)
708 		memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
709 }
710 
711 static void __init reserve_physmem_info(void)
712 {
713 	unsigned long addr, size;
714 
715 	if (get_physmem_reserved(RR_MEM_DETECT_EXTENDED, &addr, &size))
716 		memblock_reserve(addr, size);
717 }
718 
719 static void __init free_physmem_info(void)
720 {
721 	unsigned long addr, size;
722 
723 	if (get_physmem_reserved(RR_MEM_DETECT_EXTENDED, &addr, &size))
724 		memblock_phys_free(addr, size);
725 }
726 
727 static void __init memblock_add_physmem_info(void)
728 {
729 	unsigned long start, end;
730 	int i;
731 
732 	pr_debug("physmem info source: %s (%hhd)\n",
733 		 get_physmem_info_source(), physmem_info.info_source);
734 	/* keep memblock lists close to the kernel */
735 	memblock_set_bottom_up(true);
736 	for_each_physmem_usable_range(i, &start, &end)
737 		memblock_add(start, end - start);
738 	for_each_physmem_online_range(i, &start, &end)
739 		memblock_physmem_add(start, end - start);
740 	memblock_set_bottom_up(false);
741 	memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
742 }
743 
744 /*
745  * Reserve memory used for lowcore/command line/kernel image.
746  */
747 static void __init reserve_kernel(void)
748 {
749 	memblock_reserve(0, STARTUP_NORMAL_OFFSET);
750 	memblock_reserve(OLDMEM_BASE, sizeof(unsigned long));
751 	memblock_reserve(OLDMEM_SIZE, sizeof(unsigned long));
752 	memblock_reserve(physmem_info.reserved[RR_AMODE31].start, __eamode31 - __samode31);
753 	memblock_reserve(__pa(sclp_early_sccb), EXT_SCCB_READ_SCP);
754 	memblock_reserve(__pa(_stext), _end - _stext);
755 }
756 
757 static void __init setup_memory(void)
758 {
759 	phys_addr_t start, end;
760 	u64 i;
761 
762 	/*
763 	 * Init storage key for present memory
764 	 */
765 	for_each_mem_range(i, &start, &end)
766 		storage_key_init_range(start, end);
767 
768 	psw_set_key(PAGE_DEFAULT_KEY);
769 }
770 
771 static void __init relocate_amode31_section(void)
772 {
773 	unsigned long amode31_size = __eamode31 - __samode31;
774 	long amode31_offset = physmem_info.reserved[RR_AMODE31].start - __samode31;
775 	long *ptr;
776 
777 	pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size);
778 
779 	/* Move original AMODE31 section to the new one */
780 	memmove((void *)physmem_info.reserved[RR_AMODE31].start, (void *)__samode31, amode31_size);
781 	/* Zero out the old AMODE31 section to catch invalid accesses within it */
782 	memset((void *)__samode31, 0, amode31_size);
783 
784 	/* Update all AMODE31 region references */
785 	for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++)
786 		*ptr += amode31_offset;
787 }
788 
789 /* This must be called after AMODE31 relocation */
790 static void __init setup_cr(void)
791 {
792 	union ctlreg2 cr2;
793 	union ctlreg5 cr5;
794 	union ctlreg15 cr15;
795 
796 	__ctl_duct[1] = (unsigned long)__ctl_aste;
797 	__ctl_duct[2] = (unsigned long)__ctl_aste;
798 	__ctl_duct[4] = (unsigned long)__ctl_duald;
799 
800 	/* Update control registers CR2, CR5 and CR15 */
801 	__ctl_store(cr2.val, 2, 2);
802 	__ctl_store(cr5.val, 5, 5);
803 	__ctl_store(cr15.val, 15, 15);
804 	cr2.ducto = (unsigned long)__ctl_duct >> 6;
805 	cr5.pasteo = (unsigned long)__ctl_duct >> 6;
806 	cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
807 	__ctl_load(cr2.val, 2, 2);
808 	__ctl_load(cr5.val, 5, 5);
809 	__ctl_load(cr15.val, 15, 15);
810 }
811 
812 /*
813  * Add system information as device randomness
814  */
815 static void __init setup_randomness(void)
816 {
817 	struct sysinfo_3_2_2 *vmms;
818 
819 	vmms = memblock_alloc(PAGE_SIZE, PAGE_SIZE);
820 	if (!vmms)
821 		panic("Failed to allocate memory for sysinfo structure\n");
822 	if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
823 		add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
824 	memblock_free(vmms, PAGE_SIZE);
825 
826 	if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG))
827 		static_branch_enable(&s390_arch_random_available);
828 }
829 
830 /*
831  * Find the correct size for the task_struct. This depends on
832  * the size of the struct fpu at the end of the thread_struct
833  * which is embedded in the task_struct.
834  */
835 static void __init setup_task_size(void)
836 {
837 	int task_size = sizeof(struct task_struct);
838 
839 	if (!MACHINE_HAS_VX) {
840 		task_size -= sizeof(__vector128) * __NUM_VXRS;
841 		task_size += sizeof(freg_t) * __NUM_FPRS;
842 	}
843 	arch_task_struct_size = task_size;
844 }
845 
846 /*
847  * Issue diagnose 318 to set the control program name and
848  * version codes.
849  */
850 static void __init setup_control_program_code(void)
851 {
852 	union diag318_info diag318_info = {
853 		.cpnc = CPNC_LINUX,
854 		.cpvc = 0,
855 	};
856 
857 	if (!sclp.has_diag318)
858 		return;
859 
860 	diag_stat_inc(DIAG_STAT_X318);
861 	asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
862 }
863 
864 /*
865  * Print the component list from the IPL report
866  */
867 static void __init log_component_list(void)
868 {
869 	struct ipl_rb_component_entry *ptr, *end;
870 	char *str;
871 
872 	if (!early_ipl_comp_list_addr)
873 		return;
874 	if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
875 		pr_info("Linux is running with Secure-IPL enabled\n");
876 	else
877 		pr_info("Linux is running with Secure-IPL disabled\n");
878 	ptr = (void *) early_ipl_comp_list_addr;
879 	end = (void *) ptr + early_ipl_comp_list_size;
880 	pr_info("The IPL report contains the following components:\n");
881 	while (ptr < end) {
882 		if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
883 			if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
884 				str = "signed, verified";
885 			else
886 				str = "signed, verification failed";
887 		} else {
888 			str = "not signed";
889 		}
890 		pr_info("%016llx - %016llx (%s)\n",
891 			ptr->addr, ptr->addr + ptr->len, str);
892 		ptr++;
893 	}
894 }
895 
896 /*
897  * Setup function called from init/main.c just after the banner
898  * was printed.
899  */
900 
901 void __init setup_arch(char **cmdline_p)
902 {
903         /*
904          * print what head.S has found out about the machine
905          */
906 	if (MACHINE_IS_VM)
907 		pr_info("Linux is running as a z/VM "
908 			"guest operating system in 64-bit mode\n");
909 	else if (MACHINE_IS_KVM)
910 		pr_info("Linux is running under KVM in 64-bit mode\n");
911 	else if (MACHINE_IS_LPAR)
912 		pr_info("Linux is running natively in 64-bit mode\n");
913 	else
914 		pr_info("Linux is running as a guest in 64-bit mode\n");
915 
916 	log_component_list();
917 
918 	/* Have one command line that is parsed and saved in /proc/cmdline */
919 	/* boot_command_line has been already set up in early.c */
920 	*cmdline_p = boot_command_line;
921 
922         ROOT_DEV = Root_RAM0;
923 
924 	setup_initial_init_mm(_text, _etext, _edata, _end);
925 
926 	if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
927 		nospec_auto_detect();
928 
929 	jump_label_init();
930 	parse_early_param();
931 #ifdef CONFIG_CRASH_DUMP
932 	/* Deactivate elfcorehdr= kernel parameter */
933 	elfcorehdr_addr = ELFCORE_ADDR_MAX;
934 #endif
935 
936 	os_info_init();
937 	setup_ipl();
938 	setup_task_size();
939 	setup_control_program_code();
940 
941 	/* Do some memory reservations *before* memory is added to memblock */
942 	reserve_pgtables();
943 	reserve_kernel();
944 	reserve_initrd();
945 	reserve_certificate_list();
946 	reserve_physmem_info();
947 	memblock_set_current_limit(ident_map_size);
948 	memblock_allow_resize();
949 
950 	/* Get information about *all* installed memory */
951 	memblock_add_physmem_info();
952 
953 	free_physmem_info();
954 	setup_memory_end();
955 	memblock_dump_all();
956 	setup_memory();
957 
958 	relocate_amode31_section();
959 	setup_cr();
960 	setup_uv();
961 	dma_contiguous_reserve(ident_map_size);
962 	vmcp_cma_reserve();
963 	if (MACHINE_HAS_EDAT2)
964 		hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
965 
966 	reserve_crashkernel();
967 #ifdef CONFIG_CRASH_DUMP
968 	/*
969 	 * Be aware that smp_save_dump_secondary_cpus() triggers a system reset.
970 	 * Therefore CPU and device initialization should be done afterwards.
971 	 */
972 	smp_save_dump_secondary_cpus();
973 #endif
974 
975 	setup_resources();
976 	setup_lowcore();
977 	smp_fill_possible_mask();
978 	cpu_detect_mhz_feature();
979         cpu_init();
980 	numa_setup();
981 	smp_detect_cpus();
982 	topology_init_early();
983 
984 	if (test_facility(193))
985 		static_branch_enable(&cpu_has_bear);
986 
987 	/*
988 	 * Create kernel page tables.
989 	 */
990         paging_init();
991 
992 	/*
993 	 * After paging_init created the kernel page table, the new PSWs
994 	 * in lowcore can now run with DAT enabled.
995 	 */
996 #ifdef CONFIG_CRASH_DUMP
997 	smp_save_dump_ipl_cpu();
998 #endif
999 
1000         /* Setup default console */
1001 	conmode_default();
1002 	set_preferred_console();
1003 
1004 	apply_alternative_instructions();
1005 	if (IS_ENABLED(CONFIG_EXPOLINE))
1006 		nospec_init_branches();
1007 
1008 	/* Setup zfcp/nvme dump support */
1009 	setup_zfcpdump();
1010 
1011 	/* Add system specific data to the random pool */
1012 	setup_randomness();
1013 }
1014