xref: /linux/arch/s390/kernel/setup.c (revision e55f67391fa986f7357edba0ca59e668d99c3a5f)
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/mem_detect.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(ident_map_size);
150 struct mem_detect_info __bootdata(mem_detect);
151 struct initrd_data __bootdata(initrd_data);
152 
153 unsigned long __bootdata_preserved(__kaslr_offset);
154 unsigned long __bootdata(__amode31_base);
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 int __init arch_early_irq_init(void)
386 {
387 	unsigned long stack;
388 
389 	stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
390 	if (!stack)
391 		panic("Couldn't allocate async stack");
392 	S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
393 	return 0;
394 }
395 
396 void __init arch_call_rest_init(void)
397 {
398 	unsigned long stack;
399 
400 	smp_reinit_ipl_cpu();
401 	stack = stack_alloc();
402 	if (!stack)
403 		panic("Couldn't allocate kernel stack");
404 	current->stack = (void *) stack;
405 #ifdef CONFIG_VMAP_STACK
406 	current->stack_vm_area = (void *) stack;
407 #endif
408 	set_task_stack_end_magic(current);
409 	stack += STACK_INIT_OFFSET;
410 	S390_lowcore.kernel_stack = stack;
411 	call_on_stack_noreturn(rest_init, stack);
412 }
413 
414 static void __init setup_lowcore_dat_off(void)
415 {
416 	unsigned long int_psw_mask = PSW_KERNEL_BITS;
417 	struct lowcore *abs_lc, *lc;
418 	unsigned long mcck_stack;
419 	unsigned long flags;
420 
421 	if (IS_ENABLED(CONFIG_KASAN))
422 		int_psw_mask |= PSW_MASK_DAT;
423 
424 	/*
425 	 * Setup lowcore for boot cpu
426 	 */
427 	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
428 	lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
429 	if (!lc)
430 		panic("%s: Failed to allocate %zu bytes align=%zx\n",
431 		      __func__, sizeof(*lc), sizeof(*lc));
432 
433 	lc->restart_psw.mask = PSW_KERNEL_BITS;
434 	lc->restart_psw.addr = (unsigned long) restart_int_handler;
435 	lc->external_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
436 	lc->external_new_psw.addr = (unsigned long) ext_int_handler;
437 	lc->svc_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
438 	lc->svc_new_psw.addr = (unsigned long) system_call;
439 	lc->program_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
440 	lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
441 	lc->mcck_new_psw.mask = int_psw_mask;
442 	lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
443 	lc->io_new_psw.mask = int_psw_mask | PSW_MASK_MCHECK;
444 	lc->io_new_psw.addr = (unsigned long) io_int_handler;
445 	lc->clock_comparator = clock_comparator_max;
446 	lc->nodat_stack = ((unsigned long) &init_thread_union)
447 		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
448 	lc->current_task = (unsigned long)&init_task;
449 	lc->lpp = LPP_MAGIC;
450 	lc->machine_flags = S390_lowcore.machine_flags;
451 	lc->preempt_count = S390_lowcore.preempt_count;
452 	nmi_alloc_mcesa_early(&lc->mcesad);
453 	lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
454 	lc->exit_timer = S390_lowcore.exit_timer;
455 	lc->user_timer = S390_lowcore.user_timer;
456 	lc->system_timer = S390_lowcore.system_timer;
457 	lc->steal_timer = S390_lowcore.steal_timer;
458 	lc->last_update_timer = S390_lowcore.last_update_timer;
459 	lc->last_update_clock = S390_lowcore.last_update_clock;
460 
461 	/*
462 	 * Allocate the global restart stack which is the same for
463 	 * all CPUs in cast *one* of them does a PSW restart.
464 	 */
465 	restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
466 	if (!restart_stack)
467 		panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
468 		      __func__, THREAD_SIZE, THREAD_SIZE);
469 	restart_stack += STACK_INIT_OFFSET;
470 
471 	/*
472 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
473 	 * restart data to the absolute zero lowcore. This is necessary if
474 	 * PSW restart is done on an offline CPU that has lowcore zero.
475 	 */
476 	lc->restart_stack = (unsigned long) restart_stack;
477 	lc->restart_fn = (unsigned long) do_restart;
478 	lc->restart_data = 0;
479 	lc->restart_source = -1U;
480 
481 	abs_lc = get_abs_lowcore(&flags);
482 	abs_lc->restart_stack = lc->restart_stack;
483 	abs_lc->restart_fn = lc->restart_fn;
484 	abs_lc->restart_data = lc->restart_data;
485 	abs_lc->restart_source = lc->restart_source;
486 	abs_lc->restart_psw = lc->restart_psw;
487 	abs_lc->mcesad = lc->mcesad;
488 	put_abs_lowcore(abs_lc, flags);
489 
490 	mcck_stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
491 	if (!mcck_stack)
492 		panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
493 		      __func__, THREAD_SIZE, THREAD_SIZE);
494 	lc->mcck_stack = mcck_stack + STACK_INIT_OFFSET;
495 
496 	lc->spinlock_lockval = arch_spin_lockval(0);
497 	lc->spinlock_index = 0;
498 	arch_spin_lock_setup(0);
499 	lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
500 	lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
501 	lc->preempt_count = PREEMPT_DISABLED;
502 
503 	set_prefix(__pa(lc));
504 	lowcore_ptr[0] = lc;
505 }
506 
507 static void __init setup_lowcore_dat_on(void)
508 {
509 	struct lowcore *abs_lc;
510 	unsigned long flags;
511 	int i;
512 
513 	__ctl_clear_bit(0, 28);
514 	S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
515 	S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
516 	S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
517 	S390_lowcore.mcck_new_psw.mask |= PSW_MASK_DAT;
518 	S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
519 	__ctl_set_bit(0, 28);
520 	__ctl_store(S390_lowcore.cregs_save_area, 0, 15);
521 	if (abs_lowcore_map(0, lowcore_ptr[0], true))
522 		panic("Couldn't setup absolute lowcore");
523 	abs_lowcore_mapped = true;
524 	abs_lc = get_abs_lowcore(&flags);
525 	abs_lc->restart_flags = RESTART_FLAG_CTLREGS;
526 	abs_lc->program_new_psw = S390_lowcore.program_new_psw;
527 	for (i = 0; i < 16; i++)
528 		abs_lc->cregs_save_area[i] = S390_lowcore.cregs_save_area[i];
529 	put_abs_lowcore(abs_lc, flags);
530 }
531 
532 static struct resource code_resource = {
533 	.name  = "Kernel code",
534 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
535 };
536 
537 static struct resource data_resource = {
538 	.name = "Kernel data",
539 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
540 };
541 
542 static struct resource bss_resource = {
543 	.name = "Kernel bss",
544 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
545 };
546 
547 static struct resource __initdata *standard_resources[] = {
548 	&code_resource,
549 	&data_resource,
550 	&bss_resource,
551 };
552 
553 static void __init setup_resources(void)
554 {
555 	struct resource *res, *std_res, *sub_res;
556 	phys_addr_t start, end;
557 	int j;
558 	u64 i;
559 
560 	code_resource.start = (unsigned long) _text;
561 	code_resource.end = (unsigned long) _etext - 1;
562 	data_resource.start = (unsigned long) _etext;
563 	data_resource.end = (unsigned long) _edata - 1;
564 	bss_resource.start = (unsigned long) __bss_start;
565 	bss_resource.end = (unsigned long) __bss_stop - 1;
566 
567 	for_each_mem_range(i, &start, &end) {
568 		res = memblock_alloc(sizeof(*res), 8);
569 		if (!res)
570 			panic("%s: Failed to allocate %zu bytes align=0x%x\n",
571 			      __func__, sizeof(*res), 8);
572 		res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
573 
574 		res->name = "System RAM";
575 		res->start = start;
576 		/*
577 		 * In memblock, end points to the first byte after the
578 		 * range while in resourses, end points to the last byte in
579 		 * the range.
580 		 */
581 		res->end = end - 1;
582 		request_resource(&iomem_resource, res);
583 
584 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
585 			std_res = standard_resources[j];
586 			if (std_res->start < res->start ||
587 			    std_res->start > res->end)
588 				continue;
589 			if (std_res->end > res->end) {
590 				sub_res = memblock_alloc(sizeof(*sub_res), 8);
591 				if (!sub_res)
592 					panic("%s: Failed to allocate %zu bytes align=0x%x\n",
593 					      __func__, sizeof(*sub_res), 8);
594 				*sub_res = *std_res;
595 				sub_res->end = res->end;
596 				std_res->start = res->end + 1;
597 				request_resource(res, sub_res);
598 			} else {
599 				request_resource(res, std_res);
600 			}
601 		}
602 	}
603 #ifdef CONFIG_CRASH_DUMP
604 	/*
605 	 * Re-add removed crash kernel memory as reserved memory. This makes
606 	 * sure it will be mapped with the identity mapping and struct pages
607 	 * will be created, so it can be resized later on.
608 	 * However add it later since the crash kernel resource should not be
609 	 * part of the System RAM resource.
610 	 */
611 	if (crashk_res.end) {
612 		memblock_add_node(crashk_res.start, resource_size(&crashk_res),
613 				  0, MEMBLOCK_NONE);
614 		memblock_reserve(crashk_res.start, resource_size(&crashk_res));
615 		insert_resource(&iomem_resource, &crashk_res);
616 	}
617 #endif
618 }
619 
620 static void __init setup_memory_end(void)
621 {
622 	memblock_remove(ident_map_size, PHYS_ADDR_MAX - ident_map_size);
623 	max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
624 	pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
625 }
626 
627 #ifdef CONFIG_CRASH_DUMP
628 
629 /*
630  * When kdump is enabled, we have to ensure that no memory from the area
631  * [0 - crashkernel memory size] is set offline - it will be exchanged with
632  * the crashkernel memory region when kdump is triggered. The crashkernel
633  * memory region can never get offlined (pages are unmovable).
634  */
635 static int kdump_mem_notifier(struct notifier_block *nb,
636 			      unsigned long action, void *data)
637 {
638 	struct memory_notify *arg = data;
639 
640 	if (action != MEM_GOING_OFFLINE)
641 		return NOTIFY_OK;
642 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
643 		return NOTIFY_BAD;
644 	return NOTIFY_OK;
645 }
646 
647 static struct notifier_block kdump_mem_nb = {
648 	.notifier_call = kdump_mem_notifier,
649 };
650 
651 #endif
652 
653 /*
654  * Reserve memory for kdump kernel to be loaded with kexec
655  */
656 static void __init reserve_crashkernel(void)
657 {
658 #ifdef CONFIG_CRASH_DUMP
659 	unsigned long long crash_base, crash_size;
660 	phys_addr_t low, high;
661 	int rc;
662 
663 	rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
664 			       &crash_base);
665 
666 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
667 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
668 	if (rc || crash_size == 0)
669 		return;
670 
671 	if (memblock.memory.regions[0].size < crash_size) {
672 		pr_info("crashkernel reservation failed: %s\n",
673 			"first memory chunk must be at least crashkernel size");
674 		return;
675 	}
676 
677 	low = crash_base ?: oldmem_data.start;
678 	high = low + crash_size;
679 	if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
680 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
681 		crash_base = low;
682 	} else {
683 		/* Find suitable area in free memory */
684 		low = max_t(unsigned long, crash_size, sclp.hsa_size);
685 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
686 
687 		if (crash_base && crash_base < low) {
688 			pr_info("crashkernel reservation failed: %s\n",
689 				"crash_base too low");
690 			return;
691 		}
692 		low = crash_base ?: low;
693 		crash_base = memblock_phys_alloc_range(crash_size,
694 						       KEXEC_CRASH_MEM_ALIGN,
695 						       low, high);
696 	}
697 
698 	if (!crash_base) {
699 		pr_info("crashkernel reservation failed: %s\n",
700 			"no suitable area found");
701 		return;
702 	}
703 
704 	if (register_memory_notifier(&kdump_mem_nb)) {
705 		memblock_phys_free(crash_base, crash_size);
706 		return;
707 	}
708 
709 	if (!oldmem_data.start && MACHINE_IS_VM)
710 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
711 	crashk_res.start = crash_base;
712 	crashk_res.end = crash_base + crash_size - 1;
713 	memblock_remove(crash_base, crash_size);
714 	pr_info("Reserving %lluMB of memory at %lluMB "
715 		"for crashkernel (System RAM: %luMB)\n",
716 		crash_size >> 20, crash_base >> 20,
717 		(unsigned long)memblock.memory.total_size >> 20);
718 	os_info_crashkernel_add(crash_base, crash_size);
719 #endif
720 }
721 
722 /*
723  * Reserve the initrd from being used by memblock
724  */
725 static void __init reserve_initrd(void)
726 {
727 #ifdef CONFIG_BLK_DEV_INITRD
728 	if (!initrd_data.start || !initrd_data.size)
729 		return;
730 	initrd_start = (unsigned long)__va(initrd_data.start);
731 	initrd_end = initrd_start + initrd_data.size;
732 	memblock_reserve(initrd_data.start, initrd_data.size);
733 #endif
734 }
735 
736 /*
737  * Reserve the memory area used to pass the certificate lists
738  */
739 static void __init reserve_certificate_list(void)
740 {
741 	if (ipl_cert_list_addr)
742 		memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
743 }
744 
745 static void __init reserve_mem_detect_info(void)
746 {
747 	unsigned long start, size;
748 
749 	get_mem_detect_reserved(&start, &size);
750 	if (size)
751 		memblock_reserve(start, size);
752 }
753 
754 static void __init free_mem_detect_info(void)
755 {
756 	unsigned long start, size;
757 
758 	get_mem_detect_reserved(&start, &size);
759 	if (size)
760 		memblock_phys_free(start, size);
761 }
762 
763 static const char * __init get_mem_info_source(void)
764 {
765 	switch (mem_detect.info_source) {
766 	case MEM_DETECT_SCLP_STOR_INFO:
767 		return "sclp storage info";
768 	case MEM_DETECT_DIAG260:
769 		return "diag260";
770 	case MEM_DETECT_SCLP_READ_INFO:
771 		return "sclp read info";
772 	case MEM_DETECT_BIN_SEARCH:
773 		return "binary search";
774 	}
775 	return "none";
776 }
777 
778 static void __init memblock_add_mem_detect_info(void)
779 {
780 	unsigned long start, end;
781 	int i;
782 
783 	pr_debug("physmem info source: %s (%hhd)\n",
784 		 get_mem_info_source(), mem_detect.info_source);
785 	/* keep memblock lists close to the kernel */
786 	memblock_set_bottom_up(true);
787 	for_each_mem_detect_block(i, &start, &end) {
788 		memblock_add(start, end - start);
789 		memblock_physmem_add(start, end - start);
790 	}
791 	memblock_set_bottom_up(false);
792 	memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
793 }
794 
795 /*
796  * Check for initrd being in usable memory
797  */
798 static void __init check_initrd(void)
799 {
800 #ifdef CONFIG_BLK_DEV_INITRD
801 	if (initrd_data.start && initrd_data.size &&
802 	    !memblock_is_region_memory(initrd_data.start, initrd_data.size)) {
803 		pr_err("The initial RAM disk does not fit into the memory\n");
804 		memblock_phys_free(initrd_data.start, initrd_data.size);
805 		initrd_start = initrd_end = 0;
806 	}
807 #endif
808 }
809 
810 /*
811  * Reserve memory used for lowcore/command line/kernel image.
812  */
813 static void __init reserve_kernel(void)
814 {
815 	memblock_reserve(0, STARTUP_NORMAL_OFFSET);
816 	memblock_reserve(OLDMEM_BASE, sizeof(unsigned long));
817 	memblock_reserve(OLDMEM_SIZE, sizeof(unsigned long));
818 	memblock_reserve(__amode31_base, __eamode31 - __samode31);
819 	memblock_reserve(__pa(sclp_early_sccb), EXT_SCCB_READ_SCP);
820 	memblock_reserve(__pa(_stext), _end - _stext);
821 }
822 
823 static void __init setup_memory(void)
824 {
825 	phys_addr_t start, end;
826 	u64 i;
827 
828 	/*
829 	 * Init storage key for present memory
830 	 */
831 	for_each_mem_range(i, &start, &end)
832 		storage_key_init_range(start, end);
833 
834 	psw_set_key(PAGE_DEFAULT_KEY);
835 }
836 
837 static void __init relocate_amode31_section(void)
838 {
839 	unsigned long amode31_size = __eamode31 - __samode31;
840 	long amode31_offset = __amode31_base - __samode31;
841 	long *ptr;
842 
843 	pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size);
844 
845 	/* Move original AMODE31 section to the new one */
846 	memmove((void *)__amode31_base, (void *)__samode31, amode31_size);
847 	/* Zero out the old AMODE31 section to catch invalid accesses within it */
848 	memset((void *)__samode31, 0, amode31_size);
849 
850 	/* Update all AMODE31 region references */
851 	for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++)
852 		*ptr += amode31_offset;
853 }
854 
855 /* This must be called after AMODE31 relocation */
856 static void __init setup_cr(void)
857 {
858 	union ctlreg2 cr2;
859 	union ctlreg5 cr5;
860 	union ctlreg15 cr15;
861 
862 	__ctl_duct[1] = (unsigned long)__ctl_aste;
863 	__ctl_duct[2] = (unsigned long)__ctl_aste;
864 	__ctl_duct[4] = (unsigned long)__ctl_duald;
865 
866 	/* Update control registers CR2, CR5 and CR15 */
867 	__ctl_store(cr2.val, 2, 2);
868 	__ctl_store(cr5.val, 5, 5);
869 	__ctl_store(cr15.val, 15, 15);
870 	cr2.ducto = (unsigned long)__ctl_duct >> 6;
871 	cr5.pasteo = (unsigned long)__ctl_duct >> 6;
872 	cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
873 	__ctl_load(cr2.val, 2, 2);
874 	__ctl_load(cr5.val, 5, 5);
875 	__ctl_load(cr15.val, 15, 15);
876 }
877 
878 /*
879  * Add system information as device randomness
880  */
881 static void __init setup_randomness(void)
882 {
883 	struct sysinfo_3_2_2 *vmms;
884 
885 	vmms = memblock_alloc(PAGE_SIZE, PAGE_SIZE);
886 	if (!vmms)
887 		panic("Failed to allocate memory for sysinfo structure\n");
888 	if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
889 		add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
890 	memblock_free(vmms, PAGE_SIZE);
891 
892 	if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG))
893 		static_branch_enable(&s390_arch_random_available);
894 }
895 
896 /*
897  * Find the correct size for the task_struct. This depends on
898  * the size of the struct fpu at the end of the thread_struct
899  * which is embedded in the task_struct.
900  */
901 static void __init setup_task_size(void)
902 {
903 	int task_size = sizeof(struct task_struct);
904 
905 	if (!MACHINE_HAS_VX) {
906 		task_size -= sizeof(__vector128) * __NUM_VXRS;
907 		task_size += sizeof(freg_t) * __NUM_FPRS;
908 	}
909 	arch_task_struct_size = task_size;
910 }
911 
912 /*
913  * Issue diagnose 318 to set the control program name and
914  * version codes.
915  */
916 static void __init setup_control_program_code(void)
917 {
918 	union diag318_info diag318_info = {
919 		.cpnc = CPNC_LINUX,
920 		.cpvc = 0,
921 	};
922 
923 	if (!sclp.has_diag318)
924 		return;
925 
926 	diag_stat_inc(DIAG_STAT_X318);
927 	asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
928 }
929 
930 /*
931  * Print the component list from the IPL report
932  */
933 static void __init log_component_list(void)
934 {
935 	struct ipl_rb_component_entry *ptr, *end;
936 	char *str;
937 
938 	if (!early_ipl_comp_list_addr)
939 		return;
940 	if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
941 		pr_info("Linux is running with Secure-IPL enabled\n");
942 	else
943 		pr_info("Linux is running with Secure-IPL disabled\n");
944 	ptr = (void *) early_ipl_comp_list_addr;
945 	end = (void *) ptr + early_ipl_comp_list_size;
946 	pr_info("The IPL report contains the following components:\n");
947 	while (ptr < end) {
948 		if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
949 			if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
950 				str = "signed, verified";
951 			else
952 				str = "signed, verification failed";
953 		} else {
954 			str = "not signed";
955 		}
956 		pr_info("%016llx - %016llx (%s)\n",
957 			ptr->addr, ptr->addr + ptr->len, str);
958 		ptr++;
959 	}
960 }
961 
962 /*
963  * Setup function called from init/main.c just after the banner
964  * was printed.
965  */
966 
967 void __init setup_arch(char **cmdline_p)
968 {
969         /*
970          * print what head.S has found out about the machine
971          */
972 	if (MACHINE_IS_VM)
973 		pr_info("Linux is running as a z/VM "
974 			"guest operating system in 64-bit mode\n");
975 	else if (MACHINE_IS_KVM)
976 		pr_info("Linux is running under KVM in 64-bit mode\n");
977 	else if (MACHINE_IS_LPAR)
978 		pr_info("Linux is running natively in 64-bit mode\n");
979 	else
980 		pr_info("Linux is running as a guest in 64-bit mode\n");
981 
982 	log_component_list();
983 
984 	/* Have one command line that is parsed and saved in /proc/cmdline */
985 	/* boot_command_line has been already set up in early.c */
986 	*cmdline_p = boot_command_line;
987 
988         ROOT_DEV = Root_RAM0;
989 
990 	setup_initial_init_mm(_text, _etext, _edata, _end);
991 
992 	if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
993 		nospec_auto_detect();
994 
995 	jump_label_init();
996 	parse_early_param();
997 #ifdef CONFIG_CRASH_DUMP
998 	/* Deactivate elfcorehdr= kernel parameter */
999 	elfcorehdr_addr = ELFCORE_ADDR_MAX;
1000 #endif
1001 
1002 	os_info_init();
1003 	setup_ipl();
1004 	setup_task_size();
1005 	setup_control_program_code();
1006 
1007 	/* Do some memory reservations *before* memory is added to memblock */
1008 	reserve_kernel();
1009 	reserve_initrd();
1010 	reserve_certificate_list();
1011 	reserve_mem_detect_info();
1012 	memblock_set_current_limit(ident_map_size);
1013 	memblock_allow_resize();
1014 
1015 	/* Get information about *all* installed memory */
1016 	memblock_add_mem_detect_info();
1017 
1018 	free_mem_detect_info();
1019 	setup_memory_end();
1020 	memblock_dump_all();
1021 	setup_memory();
1022 
1023 	relocate_amode31_section();
1024 	setup_cr();
1025 	setup_uv();
1026 	dma_contiguous_reserve(ident_map_size);
1027 	vmcp_cma_reserve();
1028 	if (MACHINE_HAS_EDAT2)
1029 		hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1030 
1031 	check_initrd();
1032 	reserve_crashkernel();
1033 #ifdef CONFIG_CRASH_DUMP
1034 	/*
1035 	 * Be aware that smp_save_dump_secondary_cpus() triggers a system reset.
1036 	 * Therefore CPU and device initialization should be done afterwards.
1037 	 */
1038 	smp_save_dump_secondary_cpus();
1039 #endif
1040 
1041 	setup_resources();
1042 	setup_lowcore_dat_off();
1043 	smp_fill_possible_mask();
1044 	cpu_detect_mhz_feature();
1045         cpu_init();
1046 	numa_setup();
1047 	smp_detect_cpus();
1048 	topology_init_early();
1049 
1050 	if (test_facility(193))
1051 		static_branch_enable(&cpu_has_bear);
1052 
1053 	/*
1054 	 * Create kernel page tables and switch to virtual addressing.
1055 	 */
1056         paging_init();
1057 	memcpy_real_init();
1058 	/*
1059 	 * After paging_init created the kernel page table, the new PSWs
1060 	 * in lowcore can now run with DAT enabled.
1061 	 */
1062 	setup_lowcore_dat_on();
1063 #ifdef CONFIG_CRASH_DUMP
1064 	smp_save_dump_ipl_cpu();
1065 #endif
1066 
1067         /* Setup default console */
1068 	conmode_default();
1069 	set_preferred_console();
1070 
1071 	apply_alternative_instructions();
1072 	if (IS_ENABLED(CONFIG_EXPOLINE))
1073 		nospec_init_branches();
1074 
1075 	/* Setup zfcp/nvme dump support */
1076 	setup_zfcpdump();
1077 
1078 	/* Add system specific data to the random pool */
1079 	setup_randomness();
1080 }
1081