xref: /openbmc/linux/arch/s390/kernel/setup.c (revision f3531d1a)
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 
512 	__ctl_clear_bit(0, 28);
513 	S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
514 	S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
515 	S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
516 	S390_lowcore.mcck_new_psw.mask |= PSW_MASK_DAT;
517 	S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
518 	__ctl_set_bit(0, 28);
519 	__ctl_store(S390_lowcore.cregs_save_area, 0, 15);
520 	if (abs_lowcore_map(0, lowcore_ptr[0], true))
521 		panic("Couldn't setup absolute lowcore");
522 	abs_lowcore_mapped = true;
523 	abs_lc = get_abs_lowcore(&flags);
524 	abs_lc->restart_flags = RESTART_FLAG_CTLREGS;
525 	abs_lc->program_new_psw = S390_lowcore.program_new_psw;
526 	memcpy(abs_lc->cregs_save_area, S390_lowcore.cregs_save_area,
527 	       sizeof(abs_lc->cregs_save_area));
528 	put_abs_lowcore(abs_lc, flags);
529 }
530 
531 static struct resource code_resource = {
532 	.name  = "Kernel code",
533 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
534 };
535 
536 static struct resource data_resource = {
537 	.name = "Kernel data",
538 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
539 };
540 
541 static struct resource bss_resource = {
542 	.name = "Kernel bss",
543 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
544 };
545 
546 static struct resource __initdata *standard_resources[] = {
547 	&code_resource,
548 	&data_resource,
549 	&bss_resource,
550 };
551 
552 static void __init setup_resources(void)
553 {
554 	struct resource *res, *std_res, *sub_res;
555 	phys_addr_t start, end;
556 	int j;
557 	u64 i;
558 
559 	code_resource.start = (unsigned long) _text;
560 	code_resource.end = (unsigned long) _etext - 1;
561 	data_resource.start = (unsigned long) _etext;
562 	data_resource.end = (unsigned long) _edata - 1;
563 	bss_resource.start = (unsigned long) __bss_start;
564 	bss_resource.end = (unsigned long) __bss_stop - 1;
565 
566 	for_each_mem_range(i, &start, &end) {
567 		res = memblock_alloc(sizeof(*res), 8);
568 		if (!res)
569 			panic("%s: Failed to allocate %zu bytes align=0x%x\n",
570 			      __func__, sizeof(*res), 8);
571 		res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
572 
573 		res->name = "System RAM";
574 		res->start = start;
575 		/*
576 		 * In memblock, end points to the first byte after the
577 		 * range while in resourses, end points to the last byte in
578 		 * the range.
579 		 */
580 		res->end = end - 1;
581 		request_resource(&iomem_resource, res);
582 
583 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
584 			std_res = standard_resources[j];
585 			if (std_res->start < res->start ||
586 			    std_res->start > res->end)
587 				continue;
588 			if (std_res->end > res->end) {
589 				sub_res = memblock_alloc(sizeof(*sub_res), 8);
590 				if (!sub_res)
591 					panic("%s: Failed to allocate %zu bytes align=0x%x\n",
592 					      __func__, sizeof(*sub_res), 8);
593 				*sub_res = *std_res;
594 				sub_res->end = res->end;
595 				std_res->start = res->end + 1;
596 				request_resource(res, sub_res);
597 			} else {
598 				request_resource(res, std_res);
599 			}
600 		}
601 	}
602 #ifdef CONFIG_CRASH_DUMP
603 	/*
604 	 * Re-add removed crash kernel memory as reserved memory. This makes
605 	 * sure it will be mapped with the identity mapping and struct pages
606 	 * will be created, so it can be resized later on.
607 	 * However add it later since the crash kernel resource should not be
608 	 * part of the System RAM resource.
609 	 */
610 	if (crashk_res.end) {
611 		memblock_add_node(crashk_res.start, resource_size(&crashk_res),
612 				  0, MEMBLOCK_NONE);
613 		memblock_reserve(crashk_res.start, resource_size(&crashk_res));
614 		insert_resource(&iomem_resource, &crashk_res);
615 	}
616 #endif
617 }
618 
619 static void __init setup_memory_end(void)
620 {
621 	memblock_remove(ident_map_size, PHYS_ADDR_MAX - ident_map_size);
622 	max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
623 	pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
624 }
625 
626 #ifdef CONFIG_CRASH_DUMP
627 
628 /*
629  * When kdump is enabled, we have to ensure that no memory from the area
630  * [0 - crashkernel memory size] is set offline - it will be exchanged with
631  * the crashkernel memory region when kdump is triggered. The crashkernel
632  * memory region can never get offlined (pages are unmovable).
633  */
634 static int kdump_mem_notifier(struct notifier_block *nb,
635 			      unsigned long action, void *data)
636 {
637 	struct memory_notify *arg = data;
638 
639 	if (action != MEM_GOING_OFFLINE)
640 		return NOTIFY_OK;
641 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
642 		return NOTIFY_BAD;
643 	return NOTIFY_OK;
644 }
645 
646 static struct notifier_block kdump_mem_nb = {
647 	.notifier_call = kdump_mem_notifier,
648 };
649 
650 #endif
651 
652 /*
653  * Reserve memory for kdump kernel to be loaded with kexec
654  */
655 static void __init reserve_crashkernel(void)
656 {
657 #ifdef CONFIG_CRASH_DUMP
658 	unsigned long long crash_base, crash_size;
659 	phys_addr_t low, high;
660 	int rc;
661 
662 	rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
663 			       &crash_base);
664 
665 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
666 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
667 	if (rc || crash_size == 0)
668 		return;
669 
670 	if (memblock.memory.regions[0].size < crash_size) {
671 		pr_info("crashkernel reservation failed: %s\n",
672 			"first memory chunk must be at least crashkernel size");
673 		return;
674 	}
675 
676 	low = crash_base ?: oldmem_data.start;
677 	high = low + crash_size;
678 	if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
679 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
680 		crash_base = low;
681 	} else {
682 		/* Find suitable area in free memory */
683 		low = max_t(unsigned long, crash_size, sclp.hsa_size);
684 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
685 
686 		if (crash_base && crash_base < low) {
687 			pr_info("crashkernel reservation failed: %s\n",
688 				"crash_base too low");
689 			return;
690 		}
691 		low = crash_base ?: low;
692 		crash_base = memblock_phys_alloc_range(crash_size,
693 						       KEXEC_CRASH_MEM_ALIGN,
694 						       low, high);
695 	}
696 
697 	if (!crash_base) {
698 		pr_info("crashkernel reservation failed: %s\n",
699 			"no suitable area found");
700 		return;
701 	}
702 
703 	if (register_memory_notifier(&kdump_mem_nb)) {
704 		memblock_phys_free(crash_base, crash_size);
705 		return;
706 	}
707 
708 	if (!oldmem_data.start && MACHINE_IS_VM)
709 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
710 	crashk_res.start = crash_base;
711 	crashk_res.end = crash_base + crash_size - 1;
712 	memblock_remove(crash_base, crash_size);
713 	pr_info("Reserving %lluMB of memory at %lluMB "
714 		"for crashkernel (System RAM: %luMB)\n",
715 		crash_size >> 20, crash_base >> 20,
716 		(unsigned long)memblock.memory.total_size >> 20);
717 	os_info_crashkernel_add(crash_base, crash_size);
718 #endif
719 }
720 
721 /*
722  * Reserve the initrd from being used by memblock
723  */
724 static void __init reserve_initrd(void)
725 {
726 #ifdef CONFIG_BLK_DEV_INITRD
727 	if (!initrd_data.start || !initrd_data.size)
728 		return;
729 	initrd_start = (unsigned long)__va(initrd_data.start);
730 	initrd_end = initrd_start + initrd_data.size;
731 	memblock_reserve(initrd_data.start, initrd_data.size);
732 #endif
733 }
734 
735 /*
736  * Reserve the memory area used to pass the certificate lists
737  */
738 static void __init reserve_certificate_list(void)
739 {
740 	if (ipl_cert_list_addr)
741 		memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
742 }
743 
744 static void __init reserve_mem_detect_info(void)
745 {
746 	unsigned long start, size;
747 
748 	get_mem_detect_reserved(&start, &size);
749 	if (size)
750 		memblock_reserve(start, size);
751 }
752 
753 static void __init free_mem_detect_info(void)
754 {
755 	unsigned long start, size;
756 
757 	get_mem_detect_reserved(&start, &size);
758 	if (size)
759 		memblock_phys_free(start, size);
760 }
761 
762 static const char * __init get_mem_info_source(void)
763 {
764 	switch (mem_detect.info_source) {
765 	case MEM_DETECT_SCLP_STOR_INFO:
766 		return "sclp storage info";
767 	case MEM_DETECT_DIAG260:
768 		return "diag260";
769 	case MEM_DETECT_SCLP_READ_INFO:
770 		return "sclp read info";
771 	case MEM_DETECT_BIN_SEARCH:
772 		return "binary search";
773 	}
774 	return "none";
775 }
776 
777 static void __init memblock_add_mem_detect_info(void)
778 {
779 	unsigned long start, end;
780 	int i;
781 
782 	pr_debug("physmem info source: %s (%hhd)\n",
783 		 get_mem_info_source(), mem_detect.info_source);
784 	/* keep memblock lists close to the kernel */
785 	memblock_set_bottom_up(true);
786 	for_each_mem_detect_block(i, &start, &end) {
787 		memblock_add(start, end - start);
788 		memblock_physmem_add(start, end - start);
789 	}
790 	memblock_set_bottom_up(false);
791 	memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
792 }
793 
794 /*
795  * Check for initrd being in usable memory
796  */
797 static void __init check_initrd(void)
798 {
799 #ifdef CONFIG_BLK_DEV_INITRD
800 	if (initrd_data.start && initrd_data.size &&
801 	    !memblock_is_region_memory(initrd_data.start, initrd_data.size)) {
802 		pr_err("The initial RAM disk does not fit into the memory\n");
803 		memblock_phys_free(initrd_data.start, initrd_data.size);
804 		initrd_start = initrd_end = 0;
805 	}
806 #endif
807 }
808 
809 /*
810  * Reserve memory used for lowcore/command line/kernel image.
811  */
812 static void __init reserve_kernel(void)
813 {
814 	memblock_reserve(0, STARTUP_NORMAL_OFFSET);
815 	memblock_reserve(OLDMEM_BASE, sizeof(unsigned long));
816 	memblock_reserve(OLDMEM_SIZE, sizeof(unsigned long));
817 	memblock_reserve(__amode31_base, __eamode31 - __samode31);
818 	memblock_reserve(__pa(sclp_early_sccb), EXT_SCCB_READ_SCP);
819 	memblock_reserve(__pa(_stext), _end - _stext);
820 }
821 
822 static void __init setup_memory(void)
823 {
824 	phys_addr_t start, end;
825 	u64 i;
826 
827 	/*
828 	 * Init storage key for present memory
829 	 */
830 	for_each_mem_range(i, &start, &end)
831 		storage_key_init_range(start, end);
832 
833 	psw_set_key(PAGE_DEFAULT_KEY);
834 }
835 
836 static void __init relocate_amode31_section(void)
837 {
838 	unsigned long amode31_size = __eamode31 - __samode31;
839 	long amode31_offset = __amode31_base - __samode31;
840 	long *ptr;
841 
842 	pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size);
843 
844 	/* Move original AMODE31 section to the new one */
845 	memmove((void *)__amode31_base, (void *)__samode31, amode31_size);
846 	/* Zero out the old AMODE31 section to catch invalid accesses within it */
847 	memset((void *)__samode31, 0, amode31_size);
848 
849 	/* Update all AMODE31 region references */
850 	for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++)
851 		*ptr += amode31_offset;
852 }
853 
854 /* This must be called after AMODE31 relocation */
855 static void __init setup_cr(void)
856 {
857 	union ctlreg2 cr2;
858 	union ctlreg5 cr5;
859 	union ctlreg15 cr15;
860 
861 	__ctl_duct[1] = (unsigned long)__ctl_aste;
862 	__ctl_duct[2] = (unsigned long)__ctl_aste;
863 	__ctl_duct[4] = (unsigned long)__ctl_duald;
864 
865 	/* Update control registers CR2, CR5 and CR15 */
866 	__ctl_store(cr2.val, 2, 2);
867 	__ctl_store(cr5.val, 5, 5);
868 	__ctl_store(cr15.val, 15, 15);
869 	cr2.ducto = (unsigned long)__ctl_duct >> 6;
870 	cr5.pasteo = (unsigned long)__ctl_duct >> 6;
871 	cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
872 	__ctl_load(cr2.val, 2, 2);
873 	__ctl_load(cr5.val, 5, 5);
874 	__ctl_load(cr15.val, 15, 15);
875 }
876 
877 /*
878  * Add system information as device randomness
879  */
880 static void __init setup_randomness(void)
881 {
882 	struct sysinfo_3_2_2 *vmms;
883 
884 	vmms = memblock_alloc(PAGE_SIZE, PAGE_SIZE);
885 	if (!vmms)
886 		panic("Failed to allocate memory for sysinfo structure\n");
887 	if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
888 		add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
889 	memblock_free(vmms, PAGE_SIZE);
890 
891 	if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG))
892 		static_branch_enable(&s390_arch_random_available);
893 }
894 
895 /*
896  * Find the correct size for the task_struct. This depends on
897  * the size of the struct fpu at the end of the thread_struct
898  * which is embedded in the task_struct.
899  */
900 static void __init setup_task_size(void)
901 {
902 	int task_size = sizeof(struct task_struct);
903 
904 	if (!MACHINE_HAS_VX) {
905 		task_size -= sizeof(__vector128) * __NUM_VXRS;
906 		task_size += sizeof(freg_t) * __NUM_FPRS;
907 	}
908 	arch_task_struct_size = task_size;
909 }
910 
911 /*
912  * Issue diagnose 318 to set the control program name and
913  * version codes.
914  */
915 static void __init setup_control_program_code(void)
916 {
917 	union diag318_info diag318_info = {
918 		.cpnc = CPNC_LINUX,
919 		.cpvc = 0,
920 	};
921 
922 	if (!sclp.has_diag318)
923 		return;
924 
925 	diag_stat_inc(DIAG_STAT_X318);
926 	asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
927 }
928 
929 /*
930  * Print the component list from the IPL report
931  */
932 static void __init log_component_list(void)
933 {
934 	struct ipl_rb_component_entry *ptr, *end;
935 	char *str;
936 
937 	if (!early_ipl_comp_list_addr)
938 		return;
939 	if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
940 		pr_info("Linux is running with Secure-IPL enabled\n");
941 	else
942 		pr_info("Linux is running with Secure-IPL disabled\n");
943 	ptr = (void *) early_ipl_comp_list_addr;
944 	end = (void *) ptr + early_ipl_comp_list_size;
945 	pr_info("The IPL report contains the following components:\n");
946 	while (ptr < end) {
947 		if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
948 			if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
949 				str = "signed, verified";
950 			else
951 				str = "signed, verification failed";
952 		} else {
953 			str = "not signed";
954 		}
955 		pr_info("%016llx - %016llx (%s)\n",
956 			ptr->addr, ptr->addr + ptr->len, str);
957 		ptr++;
958 	}
959 }
960 
961 /*
962  * Setup function called from init/main.c just after the banner
963  * was printed.
964  */
965 
966 void __init setup_arch(char **cmdline_p)
967 {
968         /*
969          * print what head.S has found out about the machine
970          */
971 	if (MACHINE_IS_VM)
972 		pr_info("Linux is running as a z/VM "
973 			"guest operating system in 64-bit mode\n");
974 	else if (MACHINE_IS_KVM)
975 		pr_info("Linux is running under KVM in 64-bit mode\n");
976 	else if (MACHINE_IS_LPAR)
977 		pr_info("Linux is running natively in 64-bit mode\n");
978 	else
979 		pr_info("Linux is running as a guest in 64-bit mode\n");
980 
981 	log_component_list();
982 
983 	/* Have one command line that is parsed and saved in /proc/cmdline */
984 	/* boot_command_line has been already set up in early.c */
985 	*cmdline_p = boot_command_line;
986 
987         ROOT_DEV = Root_RAM0;
988 
989 	setup_initial_init_mm(_text, _etext, _edata, _end);
990 
991 	if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
992 		nospec_auto_detect();
993 
994 	jump_label_init();
995 	parse_early_param();
996 #ifdef CONFIG_CRASH_DUMP
997 	/* Deactivate elfcorehdr= kernel parameter */
998 	elfcorehdr_addr = ELFCORE_ADDR_MAX;
999 #endif
1000 
1001 	os_info_init();
1002 	setup_ipl();
1003 	setup_task_size();
1004 	setup_control_program_code();
1005 
1006 	/* Do some memory reservations *before* memory is added to memblock */
1007 	reserve_kernel();
1008 	reserve_initrd();
1009 	reserve_certificate_list();
1010 	reserve_mem_detect_info();
1011 	memblock_set_current_limit(ident_map_size);
1012 	memblock_allow_resize();
1013 
1014 	/* Get information about *all* installed memory */
1015 	memblock_add_mem_detect_info();
1016 
1017 	free_mem_detect_info();
1018 	setup_memory_end();
1019 	memblock_dump_all();
1020 	setup_memory();
1021 
1022 	relocate_amode31_section();
1023 	setup_cr();
1024 	setup_uv();
1025 	dma_contiguous_reserve(ident_map_size);
1026 	vmcp_cma_reserve();
1027 	if (MACHINE_HAS_EDAT2)
1028 		hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
1029 
1030 	check_initrd();
1031 	reserve_crashkernel();
1032 #ifdef CONFIG_CRASH_DUMP
1033 	/*
1034 	 * Be aware that smp_save_dump_secondary_cpus() triggers a system reset.
1035 	 * Therefore CPU and device initialization should be done afterwards.
1036 	 */
1037 	smp_save_dump_secondary_cpus();
1038 #endif
1039 
1040 	setup_resources();
1041 	setup_lowcore_dat_off();
1042 	smp_fill_possible_mask();
1043 	cpu_detect_mhz_feature();
1044         cpu_init();
1045 	numa_setup();
1046 	smp_detect_cpus();
1047 	topology_init_early();
1048 
1049 	if (test_facility(193))
1050 		static_branch_enable(&cpu_has_bear);
1051 
1052 	/*
1053 	 * Create kernel page tables and switch to virtual addressing.
1054 	 */
1055         paging_init();
1056 	memcpy_real_init();
1057 	/*
1058 	 * After paging_init created the kernel page table, the new PSWs
1059 	 * in lowcore can now run with DAT enabled.
1060 	 */
1061 	setup_lowcore_dat_on();
1062 #ifdef CONFIG_CRASH_DUMP
1063 	smp_save_dump_ipl_cpu();
1064 #endif
1065 
1066         /* Setup default console */
1067 	conmode_default();
1068 	set_preferred_console();
1069 
1070 	apply_alternative_instructions();
1071 	if (IS_ENABLED(CONFIG_EXPOLINE))
1072 		nospec_init_branches();
1073 
1074 	/* Setup zfcp/nvme dump support */
1075 	setup_zfcpdump();
1076 
1077 	/* Add system specific data to the random pool */
1078 	setup_randomness();
1079 }
1080