xref: /openbmc/linux/arch/s390/kernel/setup.c (revision 7af6fbdd)
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 
53 #include <asm/boot_data.h>
54 #include <asm/ipl.h>
55 #include <asm/facility.h>
56 #include <asm/smp.h>
57 #include <asm/mmu_context.h>
58 #include <asm/cpcmd.h>
59 #include <asm/lowcore.h>
60 #include <asm/nmi.h>
61 #include <asm/irq.h>
62 #include <asm/page.h>
63 #include <asm/ptrace.h>
64 #include <asm/sections.h>
65 #include <asm/ebcdic.h>
66 #include <asm/diag.h>
67 #include <asm/os_info.h>
68 #include <asm/sclp.h>
69 #include <asm/stacktrace.h>
70 #include <asm/sysinfo.h>
71 #include <asm/numa.h>
72 #include <asm/alternative.h>
73 #include <asm/nospec-branch.h>
74 #include <asm/mem_detect.h>
75 #include <asm/uv.h>
76 #include <asm/asm-offsets.h>
77 #include "entry.h"
78 
79 /*
80  * Machine setup..
81  */
82 unsigned int console_mode = 0;
83 EXPORT_SYMBOL(console_mode);
84 
85 unsigned int console_devno = -1;
86 EXPORT_SYMBOL(console_devno);
87 
88 unsigned int console_irq = -1;
89 EXPORT_SYMBOL(console_irq);
90 
91 unsigned long elf_hwcap __read_mostly = 0;
92 char elf_platform[ELF_PLATFORM_SIZE];
93 
94 unsigned long int_hwcap = 0;
95 
96 int __bootdata(noexec_disabled);
97 int __bootdata(memory_end_set);
98 unsigned long __bootdata(memory_end);
99 unsigned long __bootdata(vmalloc_size);
100 unsigned long __bootdata(max_physmem_end);
101 struct mem_detect_info __bootdata(mem_detect);
102 
103 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table);
104 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table);
105 unsigned long __bootdata_preserved(__swsusp_reset_dma);
106 unsigned long __bootdata_preserved(__stext_dma);
107 unsigned long __bootdata_preserved(__etext_dma);
108 unsigned long __bootdata_preserved(__sdma);
109 unsigned long __bootdata_preserved(__edma);
110 unsigned long __bootdata_preserved(__kaslr_offset);
111 unsigned int __bootdata_preserved(zlib_dfltcc_support);
112 EXPORT_SYMBOL(zlib_dfltcc_support);
113 
114 unsigned long VMALLOC_START;
115 EXPORT_SYMBOL(VMALLOC_START);
116 
117 unsigned long VMALLOC_END;
118 EXPORT_SYMBOL(VMALLOC_END);
119 
120 struct page *vmemmap;
121 EXPORT_SYMBOL(vmemmap);
122 
123 unsigned long MODULES_VADDR;
124 unsigned long MODULES_END;
125 
126 /* An array with a pointer to the lowcore of every CPU. */
127 struct lowcore *lowcore_ptr[NR_CPUS];
128 EXPORT_SYMBOL(lowcore_ptr);
129 
130 /*
131  * This is set up by the setup-routine at boot-time
132  * for S390 need to find out, what we have to setup
133  * using address 0x10400 ...
134  */
135 
136 #include <asm/setup.h>
137 
138 /*
139  * condev= and conmode= setup parameter.
140  */
141 
142 static int __init condev_setup(char *str)
143 {
144 	int vdev;
145 
146 	vdev = simple_strtoul(str, &str, 0);
147 	if (vdev >= 0 && vdev < 65536) {
148 		console_devno = vdev;
149 		console_irq = -1;
150 	}
151 	return 1;
152 }
153 
154 __setup("condev=", condev_setup);
155 
156 static void __init set_preferred_console(void)
157 {
158 	if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
159 		add_preferred_console("ttyS", 0, NULL);
160 	else if (CONSOLE_IS_3270)
161 		add_preferred_console("tty3270", 0, NULL);
162 	else if (CONSOLE_IS_VT220)
163 		add_preferred_console("ttyS", 1, NULL);
164 	else if (CONSOLE_IS_HVC)
165 		add_preferred_console("hvc", 0, NULL);
166 }
167 
168 static int __init conmode_setup(char *str)
169 {
170 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
171 	if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
172                 SET_CONSOLE_SCLP;
173 #endif
174 #if defined(CONFIG_TN3215_CONSOLE)
175 	if (!strcmp(str, "3215"))
176 		SET_CONSOLE_3215;
177 #endif
178 #if defined(CONFIG_TN3270_CONSOLE)
179 	if (!strcmp(str, "3270"))
180 		SET_CONSOLE_3270;
181 #endif
182 	set_preferred_console();
183         return 1;
184 }
185 
186 __setup("conmode=", conmode_setup);
187 
188 static void __init conmode_default(void)
189 {
190 	char query_buffer[1024];
191 	char *ptr;
192 
193         if (MACHINE_IS_VM) {
194 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
195 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
196 		ptr = strstr(query_buffer, "SUBCHANNEL =");
197 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
198 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
199 		ptr = strstr(query_buffer, "CONMODE");
200 		/*
201 		 * Set the conmode to 3215 so that the device recognition
202 		 * will set the cu_type of the console to 3215. If the
203 		 * conmode is 3270 and we don't set it back then both
204 		 * 3215 and the 3270 driver will try to access the console
205 		 * device (3215 as console and 3270 as normal tty).
206 		 */
207 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
208 		if (ptr == NULL) {
209 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
210 			SET_CONSOLE_SCLP;
211 #endif
212 			return;
213 		}
214 		if (str_has_prefix(ptr + 8, "3270")) {
215 #if defined(CONFIG_TN3270_CONSOLE)
216 			SET_CONSOLE_3270;
217 #elif defined(CONFIG_TN3215_CONSOLE)
218 			SET_CONSOLE_3215;
219 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
220 			SET_CONSOLE_SCLP;
221 #endif
222 		} else if (str_has_prefix(ptr + 8, "3215")) {
223 #if defined(CONFIG_TN3215_CONSOLE)
224 			SET_CONSOLE_3215;
225 #elif defined(CONFIG_TN3270_CONSOLE)
226 			SET_CONSOLE_3270;
227 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
228 			SET_CONSOLE_SCLP;
229 #endif
230 		}
231 	} else if (MACHINE_IS_KVM) {
232 		if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
233 			SET_CONSOLE_VT220;
234 		else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
235 			SET_CONSOLE_SCLP;
236 		else
237 			SET_CONSOLE_HVC;
238 	} else {
239 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
240 		SET_CONSOLE_SCLP;
241 #endif
242 	}
243 }
244 
245 #ifdef CONFIG_CRASH_DUMP
246 static void __init setup_zfcpdump(void)
247 {
248 	if (ipl_info.type != IPL_TYPE_FCP_DUMP)
249 		return;
250 	if (OLDMEM_BASE)
251 		return;
252 	strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
253 	console_loglevel = 2;
254 }
255 #else
256 static inline void setup_zfcpdump(void) {}
257 #endif /* CONFIG_CRASH_DUMP */
258 
259  /*
260  * Reboot, halt and power_off stubs. They just call _machine_restart,
261  * _machine_halt or _machine_power_off.
262  */
263 
264 void machine_restart(char *command)
265 {
266 	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
267 		/*
268 		 * Only unblank the console if we are called in enabled
269 		 * context or a bust_spinlocks cleared the way for us.
270 		 */
271 		console_unblank();
272 	_machine_restart(command);
273 }
274 
275 void machine_halt(void)
276 {
277 	if (!in_interrupt() || oops_in_progress)
278 		/*
279 		 * Only unblank the console if we are called in enabled
280 		 * context or a bust_spinlocks cleared the way for us.
281 		 */
282 		console_unblank();
283 	_machine_halt();
284 }
285 
286 void machine_power_off(void)
287 {
288 	if (!in_interrupt() || oops_in_progress)
289 		/*
290 		 * Only unblank the console if we are called in enabled
291 		 * context or a bust_spinlocks cleared the way for us.
292 		 */
293 		console_unblank();
294 	_machine_power_off();
295 }
296 
297 /*
298  * Dummy power off function.
299  */
300 void (*pm_power_off)(void) = machine_power_off;
301 EXPORT_SYMBOL_GPL(pm_power_off);
302 
303 void *restart_stack __section(.data);
304 
305 unsigned long stack_alloc(void)
306 {
307 #ifdef CONFIG_VMAP_STACK
308 	return (unsigned long)__vmalloc_node(THREAD_SIZE, THREAD_SIZE,
309 			THREADINFO_GFP, NUMA_NO_NODE,
310 			__builtin_return_address(0));
311 #else
312 	return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
313 #endif
314 }
315 
316 void stack_free(unsigned long stack)
317 {
318 #ifdef CONFIG_VMAP_STACK
319 	vfree((void *) stack);
320 #else
321 	free_pages(stack, THREAD_SIZE_ORDER);
322 #endif
323 }
324 
325 int __init arch_early_irq_init(void)
326 {
327 	unsigned long stack;
328 
329 	stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
330 	if (!stack)
331 		panic("Couldn't allocate async stack");
332 	S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
333 	return 0;
334 }
335 
336 static int __init async_stack_realloc(void)
337 {
338 	unsigned long old, new;
339 
340 	old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
341 	new = stack_alloc();
342 	if (!new)
343 		panic("Couldn't allocate async stack");
344 	S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
345 	free_pages(old, THREAD_SIZE_ORDER);
346 	return 0;
347 }
348 early_initcall(async_stack_realloc);
349 
350 void __init arch_call_rest_init(void)
351 {
352 	unsigned long stack;
353 
354 	stack = stack_alloc();
355 	if (!stack)
356 		panic("Couldn't allocate kernel stack");
357 	current->stack = (void *) stack;
358 #ifdef CONFIG_VMAP_STACK
359 	current->stack_vm_area = (void *) stack;
360 #endif
361 	set_task_stack_end_magic(current);
362 	stack += STACK_INIT_OFFSET;
363 	S390_lowcore.kernel_stack = stack;
364 	CALL_ON_STACK_NORETURN(rest_init, stack);
365 }
366 
367 static void __init setup_lowcore_dat_off(void)
368 {
369 	struct lowcore *lc;
370 
371 	/*
372 	 * Setup lowcore for boot cpu
373 	 */
374 	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
375 	lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
376 	if (!lc)
377 		panic("%s: Failed to allocate %zu bytes align=%zx\n",
378 		      __func__, sizeof(*lc), sizeof(*lc));
379 
380 	lc->restart_psw.mask = PSW_KERNEL_BITS;
381 	lc->restart_psw.addr = (unsigned long) restart_int_handler;
382 	lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
383 	lc->external_new_psw.addr = (unsigned long) ext_int_handler;
384 	lc->svc_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
385 	lc->svc_new_psw.addr = (unsigned long) system_call;
386 	lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
387 	lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
388 	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
389 	lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
390 	lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
391 	lc->io_new_psw.addr = (unsigned long) io_int_handler;
392 	lc->clock_comparator = clock_comparator_max;
393 	lc->nodat_stack = ((unsigned long) &init_thread_union)
394 		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
395 	lc->current_task = (unsigned long)&init_task;
396 	lc->lpp = LPP_MAGIC;
397 	lc->machine_flags = S390_lowcore.machine_flags;
398 	lc->preempt_count = S390_lowcore.preempt_count;
399 	lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
400 	memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
401 	       sizeof(lc->stfle_fac_list));
402 	memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
403 	       sizeof(lc->alt_stfle_fac_list));
404 	nmi_alloc_boot_cpu(lc);
405 	vdso_alloc_boot_cpu(lc);
406 	lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
407 	lc->async_enter_timer = S390_lowcore.async_enter_timer;
408 	lc->exit_timer = S390_lowcore.exit_timer;
409 	lc->user_timer = S390_lowcore.user_timer;
410 	lc->system_timer = S390_lowcore.system_timer;
411 	lc->steal_timer = S390_lowcore.steal_timer;
412 	lc->last_update_timer = S390_lowcore.last_update_timer;
413 	lc->last_update_clock = S390_lowcore.last_update_clock;
414 
415 	/*
416 	 * Allocate the global restart stack which is the same for
417 	 * all CPUs in cast *one* of them does a PSW restart.
418 	 */
419 	restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
420 	if (!restart_stack)
421 		panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
422 		      __func__, THREAD_SIZE, THREAD_SIZE);
423 	restart_stack += STACK_INIT_OFFSET;
424 
425 	/*
426 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
427 	 * restart data to the absolute zero lowcore. This is necessary if
428 	 * PSW restart is done on an offline CPU that has lowcore zero.
429 	 */
430 	lc->restart_stack = (unsigned long) restart_stack;
431 	lc->restart_fn = (unsigned long) do_restart;
432 	lc->restart_data = 0;
433 	lc->restart_source = -1UL;
434 
435 	/* Setup absolute zero lowcore */
436 	mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
437 	mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
438 	mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
439 	mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
440 	mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
441 
442 	lc->spinlock_lockval = arch_spin_lockval(0);
443 	lc->spinlock_index = 0;
444 	arch_spin_lock_setup(0);
445 	lc->br_r1_trampoline = 0x07f1;	/* br %r1 */
446 	lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
447 	lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
448 
449 	set_prefix((u32)(unsigned long) lc);
450 	lowcore_ptr[0] = lc;
451 }
452 
453 static void __init setup_lowcore_dat_on(void)
454 {
455 	__ctl_clear_bit(0, 28);
456 	S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
457 	S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
458 	S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
459 	S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
460 	__ctl_set_bit(0, 28);
461 }
462 
463 static struct resource code_resource = {
464 	.name  = "Kernel code",
465 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
466 };
467 
468 static struct resource data_resource = {
469 	.name = "Kernel data",
470 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
471 };
472 
473 static struct resource bss_resource = {
474 	.name = "Kernel bss",
475 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
476 };
477 
478 static struct resource __initdata *standard_resources[] = {
479 	&code_resource,
480 	&data_resource,
481 	&bss_resource,
482 };
483 
484 static void __init setup_resources(void)
485 {
486 	struct resource *res, *std_res, *sub_res;
487 	phys_addr_t start, end;
488 	int j;
489 	u64 i;
490 
491 	code_resource.start = (unsigned long) _text;
492 	code_resource.end = (unsigned long) _etext - 1;
493 	data_resource.start = (unsigned long) _etext;
494 	data_resource.end = (unsigned long) _edata - 1;
495 	bss_resource.start = (unsigned long) __bss_start;
496 	bss_resource.end = (unsigned long) __bss_stop - 1;
497 
498 	for_each_mem_range(i, &start, &end) {
499 		res = memblock_alloc(sizeof(*res), 8);
500 		if (!res)
501 			panic("%s: Failed to allocate %zu bytes align=0x%x\n",
502 			      __func__, sizeof(*res), 8);
503 		res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
504 
505 		res->name = "System RAM";
506 		res->start = start;
507 		/*
508 		 * In memblock, end points to the first byte after the
509 		 * range while in resourses, end points to the last byte in
510 		 * the range.
511 		 */
512 		res->end = end - 1;
513 		request_resource(&iomem_resource, res);
514 
515 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
516 			std_res = standard_resources[j];
517 			if (std_res->start < res->start ||
518 			    std_res->start > res->end)
519 				continue;
520 			if (std_res->end > res->end) {
521 				sub_res = memblock_alloc(sizeof(*sub_res), 8);
522 				if (!sub_res)
523 					panic("%s: Failed to allocate %zu bytes align=0x%x\n",
524 					      __func__, sizeof(*sub_res), 8);
525 				*sub_res = *std_res;
526 				sub_res->end = res->end;
527 				std_res->start = res->end + 1;
528 				request_resource(res, sub_res);
529 			} else {
530 				request_resource(res, std_res);
531 			}
532 		}
533 	}
534 #ifdef CONFIG_CRASH_DUMP
535 	/*
536 	 * Re-add removed crash kernel memory as reserved memory. This makes
537 	 * sure it will be mapped with the identity mapping and struct pages
538 	 * will be created, so it can be resized later on.
539 	 * However add it later since the crash kernel resource should not be
540 	 * part of the System RAM resource.
541 	 */
542 	if (crashk_res.end) {
543 		memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
544 		memblock_reserve(crashk_res.start, resource_size(&crashk_res));
545 		insert_resource(&iomem_resource, &crashk_res);
546 	}
547 #endif
548 }
549 
550 static void __init setup_memory_end(void)
551 {
552 	unsigned long vmax, tmp;
553 
554 	/* Choose kernel address space layout: 3 or 4 levels. */
555 	if (IS_ENABLED(CONFIG_KASAN)) {
556 		vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
557 			   ? _REGION1_SIZE
558 			   : _REGION2_SIZE;
559 	} else {
560 		tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
561 		tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
562 		if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
563 			vmax = _REGION2_SIZE; /* 3-level kernel page table */
564 		else
565 			vmax = _REGION1_SIZE; /* 4-level kernel page table */
566 	}
567 
568 	if (is_prot_virt_host())
569 		adjust_to_uv_max(&vmax);
570 
571 	/* module area is at the end of the kernel address space. */
572 	MODULES_END = vmax;
573 	MODULES_VADDR = MODULES_END - MODULES_LEN;
574 	VMALLOC_END = MODULES_VADDR;
575 	VMALLOC_START = VMALLOC_END - vmalloc_size;
576 
577 	/* Split remaining virtual space between 1:1 mapping & vmemmap array */
578 	tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
579 	/* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
580 	tmp = SECTION_ALIGN_UP(tmp);
581 	tmp = VMALLOC_START - tmp * sizeof(struct page);
582 	tmp &= ~((vmax >> 11) - 1);	/* align to page table level */
583 	tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
584 	vmemmap = (struct page *) tmp;
585 
586 	/* Take care that memory_end is set and <= vmemmap */
587 	memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
588 #ifdef CONFIG_KASAN
589 	/* fit in kasan shadow memory region between 1:1 and vmemmap */
590 	memory_end = min(memory_end, KASAN_SHADOW_START);
591 	vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
592 #endif
593 	max_pfn = max_low_pfn = PFN_DOWN(memory_end);
594 	memblock_remove(memory_end, ULONG_MAX);
595 
596 	pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
597 }
598 
599 #ifdef CONFIG_CRASH_DUMP
600 
601 /*
602  * When kdump is enabled, we have to ensure that no memory from the area
603  * [0 - crashkernel memory size] is set offline - it will be exchanged with
604  * the crashkernel memory region when kdump is triggered. The crashkernel
605  * memory region can never get offlined (pages are unmovable).
606  */
607 static int kdump_mem_notifier(struct notifier_block *nb,
608 			      unsigned long action, void *data)
609 {
610 	struct memory_notify *arg = data;
611 
612 	if (action != MEM_GOING_OFFLINE)
613 		return NOTIFY_OK;
614 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
615 		return NOTIFY_BAD;
616 	return NOTIFY_OK;
617 }
618 
619 static struct notifier_block kdump_mem_nb = {
620 	.notifier_call = kdump_mem_notifier,
621 };
622 
623 #endif
624 
625 /*
626  * Make sure that the area behind memory_end is protected
627  */
628 static void __init reserve_memory_end(void)
629 {
630 	if (memory_end_set)
631 		memblock_reserve(memory_end, ULONG_MAX);
632 }
633 
634 /*
635  * Make sure that oldmem, where the dump is stored, is protected
636  */
637 static void __init reserve_oldmem(void)
638 {
639 #ifdef CONFIG_CRASH_DUMP
640 	if (OLDMEM_BASE)
641 		/* Forget all memory above the running kdump system */
642 		memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
643 #endif
644 }
645 
646 /*
647  * Make sure that oldmem, where the dump is stored, is protected
648  */
649 static void __init remove_oldmem(void)
650 {
651 #ifdef CONFIG_CRASH_DUMP
652 	if (OLDMEM_BASE)
653 		/* Forget all memory above the running kdump system */
654 		memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
655 #endif
656 }
657 
658 /*
659  * Reserve memory for kdump kernel to be loaded with kexec
660  */
661 static void __init reserve_crashkernel(void)
662 {
663 #ifdef CONFIG_CRASH_DUMP
664 	unsigned long long crash_base, crash_size;
665 	phys_addr_t low, high;
666 	int rc;
667 
668 	rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
669 			       &crash_base);
670 
671 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
672 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
673 	if (rc || crash_size == 0)
674 		return;
675 
676 	if (memblock.memory.regions[0].size < crash_size) {
677 		pr_info("crashkernel reservation failed: %s\n",
678 			"first memory chunk must be at least crashkernel size");
679 		return;
680 	}
681 
682 	low = crash_base ?: OLDMEM_BASE;
683 	high = low + crash_size;
684 	if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
685 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
686 		crash_base = low;
687 	} else {
688 		/* Find suitable area in free memory */
689 		low = max_t(unsigned long, crash_size, sclp.hsa_size);
690 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
691 
692 		if (crash_base && crash_base < low) {
693 			pr_info("crashkernel reservation failed: %s\n",
694 				"crash_base too low");
695 			return;
696 		}
697 		low = crash_base ?: low;
698 		crash_base = memblock_find_in_range(low, high, crash_size,
699 						    KEXEC_CRASH_MEM_ALIGN);
700 	}
701 
702 	if (!crash_base) {
703 		pr_info("crashkernel reservation failed: %s\n",
704 			"no suitable area found");
705 		return;
706 	}
707 
708 	if (register_memory_notifier(&kdump_mem_nb))
709 		return;
710 
711 	if (!OLDMEM_BASE && MACHINE_IS_VM)
712 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
713 	crashk_res.start = crash_base;
714 	crashk_res.end = crash_base + crash_size - 1;
715 	memblock_remove(crash_base, crash_size);
716 	pr_info("Reserving %lluMB of memory at %lluMB "
717 		"for crashkernel (System RAM: %luMB)\n",
718 		crash_size >> 20, crash_base >> 20,
719 		(unsigned long)memblock.memory.total_size >> 20);
720 	os_info_crashkernel_add(crash_base, crash_size);
721 #endif
722 }
723 
724 /*
725  * Reserve the initrd from being used by memblock
726  */
727 static void __init reserve_initrd(void)
728 {
729 #ifdef CONFIG_BLK_DEV_INITRD
730 	if (!INITRD_START || !INITRD_SIZE)
731 		return;
732 	initrd_start = INITRD_START;
733 	initrd_end = initrd_start + INITRD_SIZE;
734 	memblock_reserve(INITRD_START, INITRD_SIZE);
735 #endif
736 }
737 
738 /*
739  * Reserve the memory area used to pass the certificate lists
740  */
741 static void __init reserve_certificate_list(void)
742 {
743 	if (ipl_cert_list_addr)
744 		memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
745 }
746 
747 static void __init reserve_mem_detect_info(void)
748 {
749 	unsigned long start, size;
750 
751 	get_mem_detect_reserved(&start, &size);
752 	if (size)
753 		memblock_reserve(start, size);
754 }
755 
756 static void __init free_mem_detect_info(void)
757 {
758 	unsigned long start, size;
759 
760 	get_mem_detect_reserved(&start, &size);
761 	if (size)
762 		memblock_free(start, size);
763 }
764 
765 static const char * __init get_mem_info_source(void)
766 {
767 	switch (mem_detect.info_source) {
768 	case MEM_DETECT_SCLP_STOR_INFO:
769 		return "sclp storage info";
770 	case MEM_DETECT_DIAG260:
771 		return "diag260";
772 	case MEM_DETECT_SCLP_READ_INFO:
773 		return "sclp read info";
774 	case MEM_DETECT_BIN_SEARCH:
775 		return "binary search";
776 	}
777 	return "none";
778 }
779 
780 static void __init memblock_add_mem_detect_info(void)
781 {
782 	unsigned long start, end;
783 	int i;
784 
785 	pr_debug("physmem info source: %s (%hhd)\n",
786 		 get_mem_info_source(), mem_detect.info_source);
787 	/* keep memblock lists close to the kernel */
788 	memblock_set_bottom_up(true);
789 	for_each_mem_detect_block(i, &start, &end) {
790 		memblock_add(start, end - start);
791 		memblock_physmem_add(start, end - start);
792 	}
793 	memblock_set_bottom_up(false);
794 	memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
795 	memblock_dump_all();
796 }
797 
798 /*
799  * Check for initrd being in usable memory
800  */
801 static void __init check_initrd(void)
802 {
803 #ifdef CONFIG_BLK_DEV_INITRD
804 	if (INITRD_START && INITRD_SIZE &&
805 	    !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
806 		pr_err("The initial RAM disk does not fit into the memory\n");
807 		memblock_free(INITRD_START, INITRD_SIZE);
808 		initrd_start = initrd_end = 0;
809 	}
810 #endif
811 }
812 
813 /*
814  * Reserve memory used for lowcore/command line/kernel image.
815  */
816 static void __init reserve_kernel(void)
817 {
818 	unsigned long start_pfn = PFN_UP(__pa(_end));
819 
820 	memblock_reserve(0, HEAD_END);
821 	memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
822 			 - (unsigned long)_stext);
823 	memblock_reserve(__sdma, __edma - __sdma);
824 }
825 
826 static void __init setup_memory(void)
827 {
828 	phys_addr_t start, end;
829 	u64 i;
830 
831 	/*
832 	 * Init storage key for present memory
833 	 */
834 	for_each_mem_range(i, &start, &end)
835 		storage_key_init_range(start, end);
836 
837 	psw_set_key(PAGE_DEFAULT_KEY);
838 
839 	/* Only cosmetics */
840 	memblock_enforce_memory_limit(memblock_end_of_DRAM());
841 }
842 
843 /*
844  * Setup hardware capabilities.
845  */
846 static int __init setup_hwcaps(void)
847 {
848 	static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
849 	struct cpuid cpu_id;
850 	int i;
851 
852 	/*
853 	 * The store facility list bits numbers as found in the principles
854 	 * of operation are numbered with bit 1UL<<31 as number 0 to
855 	 * bit 1UL<<0 as number 31.
856 	 *   Bit 0: instructions named N3, "backported" to esa-mode
857 	 *   Bit 2: z/Architecture mode is active
858 	 *   Bit 7: the store-facility-list-extended facility is installed
859 	 *   Bit 17: the message-security assist is installed
860 	 *   Bit 19: the long-displacement facility is installed
861 	 *   Bit 21: the extended-immediate facility is installed
862 	 *   Bit 22: extended-translation facility 3 is installed
863 	 *   Bit 30: extended-translation facility 3 enhancement facility
864 	 * These get translated to:
865 	 *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
866 	 *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
867 	 *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
868 	 *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
869 	 */
870 	for (i = 0; i < 6; i++)
871 		if (test_facility(stfl_bits[i]))
872 			elf_hwcap |= 1UL << i;
873 
874 	if (test_facility(22) && test_facility(30))
875 		elf_hwcap |= HWCAP_S390_ETF3EH;
876 
877 	/*
878 	 * Check for additional facilities with store-facility-list-extended.
879 	 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
880 	 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
881 	 * as stored by stfl, bits 32-xxx contain additional facilities.
882 	 * How many facility words are stored depends on the number of
883 	 * doublewords passed to the instruction. The additional facilities
884 	 * are:
885 	 *   Bit 42: decimal floating point facility is installed
886 	 *   Bit 44: perform floating point operation facility is installed
887 	 * translated to:
888 	 *   HWCAP_S390_DFP bit 6 (42 && 44).
889 	 */
890 	if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
891 		elf_hwcap |= HWCAP_S390_DFP;
892 
893 	/*
894 	 * Huge page support HWCAP_S390_HPAGE is bit 7.
895 	 */
896 	if (MACHINE_HAS_EDAT1)
897 		elf_hwcap |= HWCAP_S390_HPAGE;
898 
899 	/*
900 	 * 64-bit register support for 31-bit processes
901 	 * HWCAP_S390_HIGH_GPRS is bit 9.
902 	 */
903 	elf_hwcap |= HWCAP_S390_HIGH_GPRS;
904 
905 	/*
906 	 * Transactional execution support HWCAP_S390_TE is bit 10.
907 	 */
908 	if (MACHINE_HAS_TE)
909 		elf_hwcap |= HWCAP_S390_TE;
910 
911 	/*
912 	 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
913 	 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
914 	 * instead of facility bit 129.
915 	 */
916 	if (MACHINE_HAS_VX) {
917 		elf_hwcap |= HWCAP_S390_VXRS;
918 		if (test_facility(134))
919 			elf_hwcap |= HWCAP_S390_VXRS_EXT;
920 		if (test_facility(135))
921 			elf_hwcap |= HWCAP_S390_VXRS_BCD;
922 		if (test_facility(148))
923 			elf_hwcap |= HWCAP_S390_VXRS_EXT2;
924 		if (test_facility(152))
925 			elf_hwcap |= HWCAP_S390_VXRS_PDE;
926 	}
927 	if (test_facility(150))
928 		elf_hwcap |= HWCAP_S390_SORT;
929 	if (test_facility(151))
930 		elf_hwcap |= HWCAP_S390_DFLT;
931 
932 	/*
933 	 * Guarded storage support HWCAP_S390_GS is bit 12.
934 	 */
935 	if (MACHINE_HAS_GS)
936 		elf_hwcap |= HWCAP_S390_GS;
937 
938 	get_cpu_id(&cpu_id);
939 	add_device_randomness(&cpu_id, sizeof(cpu_id));
940 	switch (cpu_id.machine) {
941 	case 0x2064:
942 	case 0x2066:
943 	default:	/* Use "z900" as default for 64 bit kernels. */
944 		strcpy(elf_platform, "z900");
945 		break;
946 	case 0x2084:
947 	case 0x2086:
948 		strcpy(elf_platform, "z990");
949 		break;
950 	case 0x2094:
951 	case 0x2096:
952 		strcpy(elf_platform, "z9-109");
953 		break;
954 	case 0x2097:
955 	case 0x2098:
956 		strcpy(elf_platform, "z10");
957 		break;
958 	case 0x2817:
959 	case 0x2818:
960 		strcpy(elf_platform, "z196");
961 		break;
962 	case 0x2827:
963 	case 0x2828:
964 		strcpy(elf_platform, "zEC12");
965 		break;
966 	case 0x2964:
967 	case 0x2965:
968 		strcpy(elf_platform, "z13");
969 		break;
970 	case 0x3906:
971 	case 0x3907:
972 		strcpy(elf_platform, "z14");
973 		break;
974 	case 0x8561:
975 	case 0x8562:
976 		strcpy(elf_platform, "z15");
977 		break;
978 	}
979 
980 	/*
981 	 * Virtualization support HWCAP_INT_SIE is bit 0.
982 	 */
983 	if (sclp.has_sief2)
984 		int_hwcap |= HWCAP_INT_SIE;
985 
986 	return 0;
987 }
988 arch_initcall(setup_hwcaps);
989 
990 /*
991  * Add system information as device randomness
992  */
993 static void __init setup_randomness(void)
994 {
995 	struct sysinfo_3_2_2 *vmms;
996 
997 	vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
998 							    PAGE_SIZE);
999 	if (!vmms)
1000 		panic("Failed to allocate memory for sysinfo structure\n");
1001 
1002 	if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1003 		add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1004 	memblock_free((unsigned long) vmms, PAGE_SIZE);
1005 }
1006 
1007 /*
1008  * Find the correct size for the task_struct. This depends on
1009  * the size of the struct fpu at the end of the thread_struct
1010  * which is embedded in the task_struct.
1011  */
1012 static void __init setup_task_size(void)
1013 {
1014 	int task_size = sizeof(struct task_struct);
1015 
1016 	if (!MACHINE_HAS_VX) {
1017 		task_size -= sizeof(__vector128) * __NUM_VXRS;
1018 		task_size += sizeof(freg_t) * __NUM_FPRS;
1019 	}
1020 	arch_task_struct_size = task_size;
1021 }
1022 
1023 /*
1024  * Issue diagnose 318 to set the control program name and
1025  * version codes.
1026  */
1027 static void __init setup_control_program_code(void)
1028 {
1029 	union diag318_info diag318_info = {
1030 		.cpnc = CPNC_LINUX,
1031 		.cpvc = 0,
1032 	};
1033 
1034 	if (!sclp.has_diag318)
1035 		return;
1036 
1037 	diag_stat_inc(DIAG_STAT_X318);
1038 	asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1039 }
1040 
1041 /*
1042  * Print the component list from the IPL report
1043  */
1044 static void __init log_component_list(void)
1045 {
1046 	struct ipl_rb_component_entry *ptr, *end;
1047 	char *str;
1048 
1049 	if (!early_ipl_comp_list_addr)
1050 		return;
1051 	if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1052 		pr_info("Linux is running with Secure-IPL enabled\n");
1053 	else
1054 		pr_info("Linux is running with Secure-IPL disabled\n");
1055 	ptr = (void *) early_ipl_comp_list_addr;
1056 	end = (void *) ptr + early_ipl_comp_list_size;
1057 	pr_info("The IPL report contains the following components:\n");
1058 	while (ptr < end) {
1059 		if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1060 			if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1061 				str = "signed, verified";
1062 			else
1063 				str = "signed, verification failed";
1064 		} else {
1065 			str = "not signed";
1066 		}
1067 		pr_info("%016llx - %016llx (%s)\n",
1068 			ptr->addr, ptr->addr + ptr->len, str);
1069 		ptr++;
1070 	}
1071 }
1072 
1073 /*
1074  * Setup function called from init/main.c just after the banner
1075  * was printed.
1076  */
1077 
1078 void __init setup_arch(char **cmdline_p)
1079 {
1080         /*
1081          * print what head.S has found out about the machine
1082          */
1083 	if (MACHINE_IS_VM)
1084 		pr_info("Linux is running as a z/VM "
1085 			"guest operating system in 64-bit mode\n");
1086 	else if (MACHINE_IS_KVM)
1087 		pr_info("Linux is running under KVM in 64-bit mode\n");
1088 	else if (MACHINE_IS_LPAR)
1089 		pr_info("Linux is running natively in 64-bit mode\n");
1090 	else
1091 		pr_info("Linux is running as a guest in 64-bit mode\n");
1092 
1093 	log_component_list();
1094 
1095 	/* Have one command line that is parsed and saved in /proc/cmdline */
1096 	/* boot_command_line has been already set up in early.c */
1097 	*cmdline_p = boot_command_line;
1098 
1099         ROOT_DEV = Root_RAM0;
1100 
1101 	init_mm.start_code = (unsigned long) _text;
1102 	init_mm.end_code = (unsigned long) _etext;
1103 	init_mm.end_data = (unsigned long) _edata;
1104 	init_mm.brk = (unsigned long) _end;
1105 
1106 	if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1107 		nospec_auto_detect();
1108 
1109 	jump_label_init();
1110 	parse_early_param();
1111 #ifdef CONFIG_CRASH_DUMP
1112 	/* Deactivate elfcorehdr= kernel parameter */
1113 	elfcorehdr_addr = ELFCORE_ADDR_MAX;
1114 #endif
1115 
1116 	os_info_init();
1117 	setup_ipl();
1118 	setup_task_size();
1119 	setup_control_program_code();
1120 
1121 	/* Do some memory reservations *before* memory is added to memblock */
1122 	reserve_memory_end();
1123 	reserve_oldmem();
1124 	reserve_kernel();
1125 	reserve_initrd();
1126 	reserve_certificate_list();
1127 	reserve_mem_detect_info();
1128 	memblock_allow_resize();
1129 
1130 	/* Get information about *all* installed memory */
1131 	memblock_add_mem_detect_info();
1132 
1133 	free_mem_detect_info();
1134 	remove_oldmem();
1135 
1136 	if (is_prot_virt_host())
1137 		setup_uv();
1138 	setup_memory_end();
1139 	setup_memory();
1140 	dma_contiguous_reserve(memory_end);
1141 	vmcp_cma_reserve();
1142 
1143 	check_initrd();
1144 	reserve_crashkernel();
1145 #ifdef CONFIG_CRASH_DUMP
1146 	/*
1147 	 * Be aware that smp_save_dump_cpus() triggers a system reset.
1148 	 * Therefore CPU and device initialization should be done afterwards.
1149 	 */
1150 	smp_save_dump_cpus();
1151 #endif
1152 
1153 	setup_resources();
1154 	setup_lowcore_dat_off();
1155 	smp_fill_possible_mask();
1156 	cpu_detect_mhz_feature();
1157         cpu_init();
1158 	numa_setup();
1159 	smp_detect_cpus();
1160 	topology_init_early();
1161 
1162 	/*
1163 	 * Create kernel page tables and switch to virtual addressing.
1164 	 */
1165         paging_init();
1166 
1167 	/*
1168 	 * After paging_init created the kernel page table, the new PSWs
1169 	 * in lowcore can now run with DAT enabled.
1170 	 */
1171 	setup_lowcore_dat_on();
1172 
1173         /* Setup default console */
1174 	conmode_default();
1175 	set_preferred_console();
1176 
1177 	apply_alternative_instructions();
1178 	if (IS_ENABLED(CONFIG_EXPOLINE))
1179 		nospec_init_branches();
1180 
1181 	/* Setup zfcpdump support */
1182 	setup_zfcpdump();
1183 
1184 	/* Add system specific data to the random pool */
1185 	setup_randomness();
1186 }
1187