xref: /openbmc/linux/arch/s390/kernel/setup.c (revision 916cf5d6)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  S390 version
4  *    Copyright IBM Corp. 1999, 2012
5  *    Author(s): Hartmut Penner (hp@de.ibm.com),
6  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
7  *
8  *  Derived from "arch/i386/kernel/setup.c"
9  *    Copyright (C) 1995, Linus Torvalds
10  */
11 
12 /*
13  * This file handles the architecture-dependent parts of initialization
14  */
15 
16 #define KMSG_COMPONENT "setup"
17 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
18 
19 #include <linux/errno.h>
20 #include <linux/export.h>
21 #include <linux/sched.h>
22 #include <linux/sched/task.h>
23 #include <linux/cpu.h>
24 #include <linux/kernel.h>
25 #include <linux/memblock.h>
26 #include <linux/mm.h>
27 #include <linux/stddef.h>
28 #include <linux/unistd.h>
29 #include <linux/ptrace.h>
30 #include <linux/random.h>
31 #include <linux/user.h>
32 #include <linux/tty.h>
33 #include <linux/ioport.h>
34 #include <linux/delay.h>
35 #include <linux/init.h>
36 #include <linux/initrd.h>
37 #include <linux/root_dev.h>
38 #include <linux/console.h>
39 #include <linux/kernel_stat.h>
40 #include <linux/dma-map-ops.h>
41 #include <linux/device.h>
42 #include <linux/notifier.h>
43 #include <linux/pfn.h>
44 #include <linux/ctype.h>
45 #include <linux/reboot.h>
46 #include <linux/topology.h>
47 #include <linux/kexec.h>
48 #include <linux/crash_dump.h>
49 #include <linux/memory.h>
50 #include <linux/compat.h>
51 #include <linux/start_kernel.h>
52 #include <linux/hugetlb.h>
53 #include <linux/kmemleak.h>
54 
55 #include <asm/archrandom.h>
56 #include <asm/boot_data.h>
57 #include <asm/ipl.h>
58 #include <asm/facility.h>
59 #include <asm/smp.h>
60 #include <asm/mmu_context.h>
61 #include <asm/cpcmd.h>
62 #include <asm/abs_lowcore.h>
63 #include <asm/nmi.h>
64 #include <asm/irq.h>
65 #include <asm/page.h>
66 #include <asm/ptrace.h>
67 #include <asm/sections.h>
68 #include <asm/ebcdic.h>
69 #include <asm/diag.h>
70 #include <asm/os_info.h>
71 #include <asm/sclp.h>
72 #include <asm/stacktrace.h>
73 #include <asm/sysinfo.h>
74 #include <asm/numa.h>
75 #include <asm/alternative.h>
76 #include <asm/nospec-branch.h>
77 #include <asm/physmem_info.h>
78 #include <asm/maccess.h>
79 #include <asm/uv.h>
80 #include <asm/asm-offsets.h>
81 #include "entry.h"
82 
83 /*
84  * Machine setup..
85  */
86 unsigned int console_mode = 0;
87 EXPORT_SYMBOL(console_mode);
88 
89 unsigned int console_devno = -1;
90 EXPORT_SYMBOL(console_devno);
91 
92 unsigned int console_irq = -1;
93 EXPORT_SYMBOL(console_irq);
94 
95 /*
96  * Some code and data needs to stay below 2 GB, even when the kernel would be
97  * relocated above 2 GB, because it has to use 31 bit addresses.
98  * Such code and data is part of the .amode31 section.
99  */
100 char __amode31_ref *__samode31 = _samode31;
101 char __amode31_ref *__eamode31 = _eamode31;
102 char __amode31_ref *__stext_amode31 = _stext_amode31;
103 char __amode31_ref *__etext_amode31 = _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 unsigned long __bootdata_preserved(max_mappable);
149 unsigned long __bootdata(ident_map_size);
150 struct physmem_info __bootdata(physmem_info);
151 
152 unsigned long __bootdata_preserved(__kaslr_offset);
153 int __bootdata_preserved(__kaslr_enabled);
154 unsigned int __bootdata_preserved(zlib_dfltcc_support);
155 EXPORT_SYMBOL(zlib_dfltcc_support);
156 u64 __bootdata_preserved(stfle_fac_list[16]);
157 EXPORT_SYMBOL(stfle_fac_list);
158 u64 alt_stfle_fac_list[16];
159 struct oldmem_data __bootdata_preserved(oldmem_data);
160 
161 unsigned long VMALLOC_START;
162 EXPORT_SYMBOL(VMALLOC_START);
163 
164 unsigned long VMALLOC_END;
165 EXPORT_SYMBOL(VMALLOC_END);
166 
167 struct page *vmemmap;
168 EXPORT_SYMBOL(vmemmap);
169 unsigned long vmemmap_size;
170 
171 unsigned long MODULES_VADDR;
172 unsigned long MODULES_END;
173 
174 /* An array with a pointer to the lowcore of every CPU. */
175 struct lowcore *lowcore_ptr[NR_CPUS];
176 EXPORT_SYMBOL(lowcore_ptr);
177 
178 DEFINE_STATIC_KEY_FALSE(cpu_has_bear);
179 
180 /*
181  * The Write Back bit position in the physaddr is given by the SLPC PCI.
182  * Leaving the mask zero always uses write through which is safe
183  */
184 unsigned long mio_wb_bit_mask __ro_after_init;
185 
186 /*
187  * This is set up by the setup-routine at boot-time
188  * for S390 need to find out, what we have to setup
189  * using address 0x10400 ...
190  */
191 
192 #include <asm/setup.h>
193 
194 /*
195  * condev= and conmode= setup parameter.
196  */
197 
condev_setup(char * str)198 static int __init condev_setup(char *str)
199 {
200 	int vdev;
201 
202 	vdev = simple_strtoul(str, &str, 0);
203 	if (vdev >= 0 && vdev < 65536) {
204 		console_devno = vdev;
205 		console_irq = -1;
206 	}
207 	return 1;
208 }
209 
210 __setup("condev=", condev_setup);
211 
set_preferred_console(void)212 static void __init set_preferred_console(void)
213 {
214 	if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
215 		add_preferred_console("ttyS", 0, NULL);
216 	else if (CONSOLE_IS_3270)
217 		add_preferred_console("tty3270", 0, NULL);
218 	else if (CONSOLE_IS_VT220)
219 		add_preferred_console("ttysclp", 0, NULL);
220 	else if (CONSOLE_IS_HVC)
221 		add_preferred_console("hvc", 0, NULL);
222 }
223 
conmode_setup(char * str)224 static int __init conmode_setup(char *str)
225 {
226 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
227 	if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
228                 SET_CONSOLE_SCLP;
229 #endif
230 #if defined(CONFIG_TN3215_CONSOLE)
231 	if (!strcmp(str, "3215"))
232 		SET_CONSOLE_3215;
233 #endif
234 #if defined(CONFIG_TN3270_CONSOLE)
235 	if (!strcmp(str, "3270"))
236 		SET_CONSOLE_3270;
237 #endif
238 	set_preferred_console();
239         return 1;
240 }
241 
242 __setup("conmode=", conmode_setup);
243 
conmode_default(void)244 static void __init conmode_default(void)
245 {
246 	char query_buffer[1024];
247 	char *ptr;
248 
249         if (MACHINE_IS_VM) {
250 		cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
251 		console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
252 		ptr = strstr(query_buffer, "SUBCHANNEL =");
253 		console_irq = simple_strtoul(ptr + 13, NULL, 16);
254 		cpcmd("QUERY TERM", query_buffer, 1024, NULL);
255 		ptr = strstr(query_buffer, "CONMODE");
256 		/*
257 		 * Set the conmode to 3215 so that the device recognition
258 		 * will set the cu_type of the console to 3215. If the
259 		 * conmode is 3270 and we don't set it back then both
260 		 * 3215 and the 3270 driver will try to access the console
261 		 * device (3215 as console and 3270 as normal tty).
262 		 */
263 		cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
264 		if (ptr == NULL) {
265 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
266 			SET_CONSOLE_SCLP;
267 #endif
268 			return;
269 		}
270 		if (str_has_prefix(ptr + 8, "3270")) {
271 #if defined(CONFIG_TN3270_CONSOLE)
272 			SET_CONSOLE_3270;
273 #elif defined(CONFIG_TN3215_CONSOLE)
274 			SET_CONSOLE_3215;
275 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
276 			SET_CONSOLE_SCLP;
277 #endif
278 		} else if (str_has_prefix(ptr + 8, "3215")) {
279 #if defined(CONFIG_TN3215_CONSOLE)
280 			SET_CONSOLE_3215;
281 #elif defined(CONFIG_TN3270_CONSOLE)
282 			SET_CONSOLE_3270;
283 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
284 			SET_CONSOLE_SCLP;
285 #endif
286 		}
287 	} else if (MACHINE_IS_KVM) {
288 		if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
289 			SET_CONSOLE_VT220;
290 		else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
291 			SET_CONSOLE_SCLP;
292 		else
293 			SET_CONSOLE_HVC;
294 	} else {
295 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
296 		SET_CONSOLE_SCLP;
297 #endif
298 	}
299 }
300 
301 #ifdef CONFIG_CRASH_DUMP
setup_zfcpdump(void)302 static void __init setup_zfcpdump(void)
303 {
304 	if (!is_ipl_type_dump())
305 		return;
306 	if (oldmem_data.start)
307 		return;
308 	strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
309 	console_loglevel = 2;
310 }
311 #else
setup_zfcpdump(void)312 static inline void setup_zfcpdump(void) {}
313 #endif /* CONFIG_CRASH_DUMP */
314 
315  /*
316  * Reboot, halt and power_off stubs. They just call _machine_restart,
317  * _machine_halt or _machine_power_off.
318  */
319 
machine_restart(char * command)320 void machine_restart(char *command)
321 {
322 	if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
323 		/*
324 		 * Only unblank the console if we are called in enabled
325 		 * context or a bust_spinlocks cleared the way for us.
326 		 */
327 		console_unblank();
328 	_machine_restart(command);
329 }
330 
machine_halt(void)331 void machine_halt(void)
332 {
333 	if (!in_interrupt() || oops_in_progress)
334 		/*
335 		 * Only unblank the console if we are called in enabled
336 		 * context or a bust_spinlocks cleared the way for us.
337 		 */
338 		console_unblank();
339 	_machine_halt();
340 }
341 
machine_power_off(void)342 void machine_power_off(void)
343 {
344 	if (!in_interrupt() || oops_in_progress)
345 		/*
346 		 * Only unblank the console if we are called in enabled
347 		 * context or a bust_spinlocks cleared the way for us.
348 		 */
349 		console_unblank();
350 	_machine_power_off();
351 }
352 
353 /*
354  * Dummy power off function.
355  */
356 void (*pm_power_off)(void) = machine_power_off;
357 EXPORT_SYMBOL_GPL(pm_power_off);
358 
359 void *restart_stack;
360 
stack_alloc(void)361 unsigned long stack_alloc(void)
362 {
363 #ifdef CONFIG_VMAP_STACK
364 	void *ret;
365 
366 	ret = __vmalloc_node(THREAD_SIZE, THREAD_SIZE, THREADINFO_GFP,
367 			     NUMA_NO_NODE, __builtin_return_address(0));
368 	kmemleak_not_leak(ret);
369 	return (unsigned long)ret;
370 #else
371 	return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
372 #endif
373 }
374 
stack_free(unsigned long stack)375 void stack_free(unsigned long stack)
376 {
377 #ifdef CONFIG_VMAP_STACK
378 	vfree((void *) stack);
379 #else
380 	free_pages(stack, THREAD_SIZE_ORDER);
381 #endif
382 }
383 
arch_call_rest_init(void)384 void __init __noreturn arch_call_rest_init(void)
385 {
386 	smp_reinit_ipl_cpu();
387 	rest_init();
388 }
389 
stack_alloc_early(void)390 static unsigned long __init stack_alloc_early(void)
391 {
392 	unsigned long stack;
393 
394 	stack = (unsigned long)memblock_alloc(THREAD_SIZE, THREAD_SIZE);
395 	if (!stack) {
396 		panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
397 		      __func__, THREAD_SIZE, THREAD_SIZE);
398 	}
399 	return stack;
400 }
401 
setup_lowcore(void)402 static void __init setup_lowcore(void)
403 {
404 	struct lowcore *lc, *abs_lc;
405 
406 	/*
407 	 * Setup lowcore for boot cpu
408 	 */
409 	BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
410 	lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
411 	if (!lc)
412 		panic("%s: Failed to allocate %zu bytes align=%zx\n",
413 		      __func__, sizeof(*lc), sizeof(*lc));
414 
415 	lc->restart_psw.mask = PSW_KERNEL_BITS & ~PSW_MASK_DAT;
416 	lc->restart_psw.addr = __pa(restart_int_handler);
417 	lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
418 	lc->external_new_psw.addr = (unsigned long) ext_int_handler;
419 	lc->svc_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
420 	lc->svc_new_psw.addr = (unsigned long) system_call;
421 	lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
422 	lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
423 	lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
424 	lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
425 	lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
426 	lc->io_new_psw.addr = (unsigned long) io_int_handler;
427 	lc->clock_comparator = clock_comparator_max;
428 	lc->current_task = (unsigned long)&init_task;
429 	lc->lpp = LPP_MAGIC;
430 	lc->machine_flags = S390_lowcore.machine_flags;
431 	lc->preempt_count = S390_lowcore.preempt_count;
432 	nmi_alloc_mcesa_early(&lc->mcesad);
433 	lc->sys_enter_timer = S390_lowcore.sys_enter_timer;
434 	lc->exit_timer = S390_lowcore.exit_timer;
435 	lc->user_timer = S390_lowcore.user_timer;
436 	lc->system_timer = S390_lowcore.system_timer;
437 	lc->steal_timer = S390_lowcore.steal_timer;
438 	lc->last_update_timer = S390_lowcore.last_update_timer;
439 	lc->last_update_clock = S390_lowcore.last_update_clock;
440 	/*
441 	 * Allocate the global restart stack which is the same for
442 	 * all CPUs in case *one* of them does a PSW restart.
443 	 */
444 	restart_stack = (void *)(stack_alloc_early() + STACK_INIT_OFFSET);
445 	lc->mcck_stack = stack_alloc_early() + STACK_INIT_OFFSET;
446 	lc->async_stack = stack_alloc_early() + STACK_INIT_OFFSET;
447 	lc->nodat_stack = stack_alloc_early() + STACK_INIT_OFFSET;
448 	lc->kernel_stack = S390_lowcore.kernel_stack;
449 	/*
450 	 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
451 	 * restart data to the absolute zero lowcore. This is necessary if
452 	 * PSW restart is done on an offline CPU that has lowcore zero.
453 	 */
454 	lc->restart_stack = (unsigned long) restart_stack;
455 	lc->restart_fn = (unsigned long) do_restart;
456 	lc->restart_data = 0;
457 	lc->restart_source = -1U;
458 	__ctl_store(lc->cregs_save_area, 0, 15);
459 	lc->spinlock_lockval = arch_spin_lockval(0);
460 	lc->spinlock_index = 0;
461 	arch_spin_lock_setup(0);
462 	lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
463 	lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
464 	lc->preempt_count = PREEMPT_DISABLED;
465 	lc->kernel_asce = S390_lowcore.kernel_asce;
466 	lc->user_asce = S390_lowcore.user_asce;
467 
468 	abs_lc = get_abs_lowcore();
469 	abs_lc->restart_stack = lc->restart_stack;
470 	abs_lc->restart_fn = lc->restart_fn;
471 	abs_lc->restart_data = lc->restart_data;
472 	abs_lc->restart_source = lc->restart_source;
473 	abs_lc->restart_psw = lc->restart_psw;
474 	abs_lc->restart_flags = RESTART_FLAG_CTLREGS;
475 	memcpy(abs_lc->cregs_save_area, lc->cregs_save_area, sizeof(abs_lc->cregs_save_area));
476 	abs_lc->program_new_psw = lc->program_new_psw;
477 	abs_lc->mcesad = lc->mcesad;
478 	put_abs_lowcore(abs_lc);
479 
480 	set_prefix(__pa(lc));
481 	lowcore_ptr[0] = lc;
482 	if (abs_lowcore_map(0, lowcore_ptr[0], false))
483 		panic("Couldn't setup absolute lowcore");
484 }
485 
486 static struct resource code_resource = {
487 	.name  = "Kernel code",
488 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
489 };
490 
491 static struct resource data_resource = {
492 	.name = "Kernel data",
493 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
494 };
495 
496 static struct resource bss_resource = {
497 	.name = "Kernel bss",
498 	.flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
499 };
500 
501 static struct resource __initdata *standard_resources[] = {
502 	&code_resource,
503 	&data_resource,
504 	&bss_resource,
505 };
506 
setup_resources(void)507 static void __init setup_resources(void)
508 {
509 	struct resource *res, *std_res, *sub_res;
510 	phys_addr_t start, end;
511 	int j;
512 	u64 i;
513 
514 	code_resource.start = (unsigned long) _text;
515 	code_resource.end = (unsigned long) _etext - 1;
516 	data_resource.start = (unsigned long) _etext;
517 	data_resource.end = (unsigned long) _edata - 1;
518 	bss_resource.start = (unsigned long) __bss_start;
519 	bss_resource.end = (unsigned long) __bss_stop - 1;
520 
521 	for_each_mem_range(i, &start, &end) {
522 		res = memblock_alloc(sizeof(*res), 8);
523 		if (!res)
524 			panic("%s: Failed to allocate %zu bytes align=0x%x\n",
525 			      __func__, sizeof(*res), 8);
526 		res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
527 
528 		res->name = "System RAM";
529 		res->start = start;
530 		/*
531 		 * In memblock, end points to the first byte after the
532 		 * range while in resources, end points to the last byte in
533 		 * the range.
534 		 */
535 		res->end = end - 1;
536 		request_resource(&iomem_resource, res);
537 
538 		for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
539 			std_res = standard_resources[j];
540 			if (std_res->start < res->start ||
541 			    std_res->start > res->end)
542 				continue;
543 			if (std_res->end > res->end) {
544 				sub_res = memblock_alloc(sizeof(*sub_res), 8);
545 				if (!sub_res)
546 					panic("%s: Failed to allocate %zu bytes align=0x%x\n",
547 					      __func__, sizeof(*sub_res), 8);
548 				*sub_res = *std_res;
549 				sub_res->end = res->end;
550 				std_res->start = res->end + 1;
551 				request_resource(res, sub_res);
552 			} else {
553 				request_resource(res, std_res);
554 			}
555 		}
556 	}
557 #ifdef CONFIG_CRASH_DUMP
558 	/*
559 	 * Re-add removed crash kernel memory as reserved memory. This makes
560 	 * sure it will be mapped with the identity mapping and struct pages
561 	 * will be created, so it can be resized later on.
562 	 * However add it later since the crash kernel resource should not be
563 	 * part of the System RAM resource.
564 	 */
565 	if (crashk_res.end) {
566 		memblock_add_node(crashk_res.start, resource_size(&crashk_res),
567 				  0, MEMBLOCK_NONE);
568 		memblock_reserve(crashk_res.start, resource_size(&crashk_res));
569 		insert_resource(&iomem_resource, &crashk_res);
570 	}
571 #endif
572 }
573 
setup_memory_end(void)574 static void __init setup_memory_end(void)
575 {
576 	max_pfn = max_low_pfn = PFN_DOWN(ident_map_size);
577 	pr_notice("The maximum memory size is %luMB\n", ident_map_size >> 20);
578 }
579 
580 #ifdef CONFIG_CRASH_DUMP
581 
582 /*
583  * When kdump is enabled, we have to ensure that no memory from the area
584  * [0 - crashkernel memory size] is set offline - it will be exchanged with
585  * the crashkernel memory region when kdump is triggered. The crashkernel
586  * memory region can never get offlined (pages are unmovable).
587  */
kdump_mem_notifier(struct notifier_block * nb,unsigned long action,void * data)588 static int kdump_mem_notifier(struct notifier_block *nb,
589 			      unsigned long action, void *data)
590 {
591 	struct memory_notify *arg = data;
592 
593 	if (action != MEM_GOING_OFFLINE)
594 		return NOTIFY_OK;
595 	if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
596 		return NOTIFY_BAD;
597 	return NOTIFY_OK;
598 }
599 
600 static struct notifier_block kdump_mem_nb = {
601 	.notifier_call = kdump_mem_notifier,
602 };
603 
604 #endif
605 
606 /*
607  * Reserve page tables created by decompressor
608  */
reserve_pgtables(void)609 static void __init reserve_pgtables(void)
610 {
611 	unsigned long start, end;
612 	struct reserved_range *range;
613 
614 	for_each_physmem_reserved_type_range(RR_VMEM, range, &start, &end)
615 		memblock_reserve(start, end - start);
616 }
617 
618 /*
619  * Reserve memory for kdump kernel to be loaded with kexec
620  */
reserve_crashkernel(void)621 static void __init reserve_crashkernel(void)
622 {
623 #ifdef CONFIG_CRASH_DUMP
624 	unsigned long long crash_base, crash_size;
625 	phys_addr_t low, high;
626 	int rc;
627 
628 	rc = parse_crashkernel(boot_command_line, ident_map_size, &crash_size,
629 			       &crash_base);
630 
631 	crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
632 	crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
633 	if (rc || crash_size == 0)
634 		return;
635 
636 	if (memblock.memory.regions[0].size < crash_size) {
637 		pr_info("crashkernel reservation failed: %s\n",
638 			"first memory chunk must be at least crashkernel size");
639 		return;
640 	}
641 
642 	low = crash_base ?: oldmem_data.start;
643 	high = low + crash_size;
644 	if (low >= oldmem_data.start && high <= oldmem_data.start + oldmem_data.size) {
645 		/* The crashkernel fits into OLDMEM, reuse OLDMEM */
646 		crash_base = low;
647 	} else {
648 		/* Find suitable area in free memory */
649 		low = max_t(unsigned long, crash_size, sclp.hsa_size);
650 		high = crash_base ? crash_base + crash_size : ULONG_MAX;
651 
652 		if (crash_base && crash_base < low) {
653 			pr_info("crashkernel reservation failed: %s\n",
654 				"crash_base too low");
655 			return;
656 		}
657 		low = crash_base ?: low;
658 		crash_base = memblock_phys_alloc_range(crash_size,
659 						       KEXEC_CRASH_MEM_ALIGN,
660 						       low, high);
661 	}
662 
663 	if (!crash_base) {
664 		pr_info("crashkernel reservation failed: %s\n",
665 			"no suitable area found");
666 		return;
667 	}
668 
669 	if (register_memory_notifier(&kdump_mem_nb)) {
670 		memblock_phys_free(crash_base, crash_size);
671 		return;
672 	}
673 
674 	if (!oldmem_data.start && MACHINE_IS_VM)
675 		diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
676 	crashk_res.start = crash_base;
677 	crashk_res.end = crash_base + crash_size - 1;
678 	memblock_remove(crash_base, crash_size);
679 	pr_info("Reserving %lluMB of memory at %lluMB "
680 		"for crashkernel (System RAM: %luMB)\n",
681 		crash_size >> 20, crash_base >> 20,
682 		(unsigned long)memblock.memory.total_size >> 20);
683 	os_info_crashkernel_add(crash_base, crash_size);
684 #endif
685 }
686 
687 /*
688  * Reserve the initrd from being used by memblock
689  */
reserve_initrd(void)690 static void __init reserve_initrd(void)
691 {
692 	unsigned long addr, size;
693 
694 	if (!IS_ENABLED(CONFIG_BLK_DEV_INITRD) || !get_physmem_reserved(RR_INITRD, &addr, &size))
695 		return;
696 	initrd_start = (unsigned long)__va(addr);
697 	initrd_end = initrd_start + size;
698 	memblock_reserve(addr, size);
699 }
700 
701 /*
702  * Reserve the memory area used to pass the certificate lists
703  */
reserve_certificate_list(void)704 static void __init reserve_certificate_list(void)
705 {
706 	if (ipl_cert_list_addr)
707 		memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
708 }
709 
reserve_physmem_info(void)710 static void __init reserve_physmem_info(void)
711 {
712 	unsigned long addr, size;
713 
714 	if (get_physmem_reserved(RR_MEM_DETECT_EXTENDED, &addr, &size))
715 		memblock_reserve(addr, size);
716 }
717 
free_physmem_info(void)718 static void __init free_physmem_info(void)
719 {
720 	unsigned long addr, size;
721 
722 	if (get_physmem_reserved(RR_MEM_DETECT_EXTENDED, &addr, &size))
723 		memblock_phys_free(addr, size);
724 }
725 
memblock_add_physmem_info(void)726 static void __init memblock_add_physmem_info(void)
727 {
728 	unsigned long start, end;
729 	int i;
730 
731 	pr_debug("physmem info source: %s (%hhd)\n",
732 		 get_physmem_info_source(), physmem_info.info_source);
733 	/* keep memblock lists close to the kernel */
734 	memblock_set_bottom_up(true);
735 	for_each_physmem_usable_range(i, &start, &end)
736 		memblock_add(start, end - start);
737 	for_each_physmem_online_range(i, &start, &end)
738 		memblock_physmem_add(start, end - start);
739 	memblock_set_bottom_up(false);
740 	memblock_set_node(0, ULONG_MAX, &memblock.memory, 0);
741 }
742 
743 /*
744  * Reserve memory used for lowcore/command line/kernel image.
745  */
reserve_kernel(void)746 static void __init reserve_kernel(void)
747 {
748 	memblock_reserve(0, STARTUP_NORMAL_OFFSET);
749 	memblock_reserve(OLDMEM_BASE, sizeof(unsigned long));
750 	memblock_reserve(OLDMEM_SIZE, sizeof(unsigned long));
751 	memblock_reserve(physmem_info.reserved[RR_AMODE31].start, __eamode31 - __samode31);
752 	memblock_reserve(__pa(sclp_early_sccb), EXT_SCCB_READ_SCP);
753 	memblock_reserve(__pa(_stext), _end - _stext);
754 }
755 
setup_memory(void)756 static void __init setup_memory(void)
757 {
758 	phys_addr_t start, end;
759 	u64 i;
760 
761 	/*
762 	 * Init storage key for present memory
763 	 */
764 	for_each_mem_range(i, &start, &end)
765 		storage_key_init_range(start, end);
766 
767 	psw_set_key(PAGE_DEFAULT_KEY);
768 }
769 
relocate_amode31_section(void)770 static void __init relocate_amode31_section(void)
771 {
772 	unsigned long amode31_size = __eamode31 - __samode31;
773 	long amode31_offset, *ptr;
774 
775 	amode31_offset = physmem_info.reserved[RR_AMODE31].start - (unsigned long)__samode31;
776 	pr_info("Relocating AMODE31 section of size 0x%08lx\n", amode31_size);
777 
778 	/* Move original AMODE31 section to the new one */
779 	memmove((void *)physmem_info.reserved[RR_AMODE31].start, __samode31, amode31_size);
780 	/* Zero out the old AMODE31 section to catch invalid accesses within it */
781 	memset(__samode31, 0, amode31_size);
782 
783 	/* Update all AMODE31 region references */
784 	for (ptr = _start_amode31_refs; ptr != _end_amode31_refs; ptr++)
785 		*ptr += amode31_offset;
786 }
787 
788 /* This must be called after AMODE31 relocation */
setup_cr(void)789 static void __init setup_cr(void)
790 {
791 	union ctlreg2 cr2;
792 	union ctlreg5 cr5;
793 	union ctlreg15 cr15;
794 
795 	__ctl_duct[1] = (unsigned long)__ctl_aste;
796 	__ctl_duct[2] = (unsigned long)__ctl_aste;
797 	__ctl_duct[4] = (unsigned long)__ctl_duald;
798 
799 	/* Update control registers CR2, CR5 and CR15 */
800 	__ctl_store(cr2.val, 2, 2);
801 	__ctl_store(cr5.val, 5, 5);
802 	__ctl_store(cr15.val, 15, 15);
803 	cr2.ducto = (unsigned long)__ctl_duct >> 6;
804 	cr5.pasteo = (unsigned long)__ctl_duct >> 6;
805 	cr15.lsea = (unsigned long)__ctl_linkage_stack >> 3;
806 	__ctl_load(cr2.val, 2, 2);
807 	__ctl_load(cr5.val, 5, 5);
808 	__ctl_load(cr15.val, 15, 15);
809 }
810 
811 /*
812  * Add system information as device randomness
813  */
setup_randomness(void)814 static void __init setup_randomness(void)
815 {
816 	struct sysinfo_3_2_2 *vmms;
817 
818 	vmms = memblock_alloc(PAGE_SIZE, PAGE_SIZE);
819 	if (!vmms)
820 		panic("Failed to allocate memory for sysinfo structure\n");
821 	if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
822 		add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
823 	memblock_free(vmms, PAGE_SIZE);
824 
825 	if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG))
826 		static_branch_enable(&s390_arch_random_available);
827 }
828 
829 /*
830  * Find the correct size for the task_struct. This depends on
831  * the size of the struct fpu at the end of the thread_struct
832  * which is embedded in the task_struct.
833  */
setup_task_size(void)834 static void __init setup_task_size(void)
835 {
836 	int task_size = sizeof(struct task_struct);
837 
838 	if (!MACHINE_HAS_VX) {
839 		task_size -= sizeof(__vector128) * __NUM_VXRS;
840 		task_size += sizeof(freg_t) * __NUM_FPRS;
841 	}
842 	arch_task_struct_size = task_size;
843 }
844 
845 /*
846  * Issue diagnose 318 to set the control program name and
847  * version codes.
848  */
setup_control_program_code(void)849 static void __init setup_control_program_code(void)
850 {
851 	union diag318_info diag318_info = {
852 		.cpnc = CPNC_LINUX,
853 		.cpvc = 0,
854 	};
855 
856 	if (!sclp.has_diag318)
857 		return;
858 
859 	diag_stat_inc(DIAG_STAT_X318);
860 	asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
861 }
862 
863 /*
864  * Print the component list from the IPL report
865  */
log_component_list(void)866 static void __init log_component_list(void)
867 {
868 	struct ipl_rb_component_entry *ptr, *end;
869 	char *str;
870 
871 	if (!early_ipl_comp_list_addr)
872 		return;
873 	if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
874 		pr_info("Linux is running with Secure-IPL enabled\n");
875 	else
876 		pr_info("Linux is running with Secure-IPL disabled\n");
877 	ptr = __va(early_ipl_comp_list_addr);
878 	end = (void *) ptr + early_ipl_comp_list_size;
879 	pr_info("The IPL report contains the following components:\n");
880 	while (ptr < end) {
881 		if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
882 			if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
883 				str = "signed, verified";
884 			else
885 				str = "signed, verification failed";
886 		} else {
887 			str = "not signed";
888 		}
889 		pr_info("%016llx - %016llx (%s)\n",
890 			ptr->addr, ptr->addr + ptr->len, str);
891 		ptr++;
892 	}
893 }
894 
895 /*
896  * Setup function called from init/main.c just after the banner
897  * was printed.
898  */
899 
setup_arch(char ** cmdline_p)900 void __init setup_arch(char **cmdline_p)
901 {
902         /*
903          * print what head.S has found out about the machine
904          */
905 	if (MACHINE_IS_VM)
906 		pr_info("Linux is running as a z/VM "
907 			"guest operating system in 64-bit mode\n");
908 	else if (MACHINE_IS_KVM)
909 		pr_info("Linux is running under KVM in 64-bit mode\n");
910 	else if (MACHINE_IS_LPAR)
911 		pr_info("Linux is running natively in 64-bit mode\n");
912 	else
913 		pr_info("Linux is running as a guest in 64-bit mode\n");
914 
915 	log_component_list();
916 
917 	/* Have one command line that is parsed and saved in /proc/cmdline */
918 	/* boot_command_line has been already set up in early.c */
919 	*cmdline_p = boot_command_line;
920 
921         ROOT_DEV = Root_RAM0;
922 
923 	setup_initial_init_mm(_text, _etext, _edata, _end);
924 
925 	if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
926 		nospec_auto_detect();
927 
928 	jump_label_init();
929 	parse_early_param();
930 #ifdef CONFIG_CRASH_DUMP
931 	/* Deactivate elfcorehdr= kernel parameter */
932 	elfcorehdr_addr = ELFCORE_ADDR_MAX;
933 #endif
934 
935 	os_info_init();
936 	setup_ipl();
937 	setup_task_size();
938 	setup_control_program_code();
939 
940 	/* Do some memory reservations *before* memory is added to memblock */
941 	reserve_pgtables();
942 	reserve_kernel();
943 	reserve_initrd();
944 	reserve_certificate_list();
945 	reserve_physmem_info();
946 	memblock_set_current_limit(ident_map_size);
947 	memblock_allow_resize();
948 
949 	/* Get information about *all* installed memory */
950 	memblock_add_physmem_info();
951 
952 	free_physmem_info();
953 	setup_memory_end();
954 	memblock_dump_all();
955 	setup_memory();
956 
957 	relocate_amode31_section();
958 	setup_cr();
959 	setup_uv();
960 	dma_contiguous_reserve(ident_map_size);
961 	vmcp_cma_reserve();
962 	if (MACHINE_HAS_EDAT2)
963 		hugetlb_cma_reserve(PUD_SHIFT - PAGE_SHIFT);
964 
965 	reserve_crashkernel();
966 #ifdef CONFIG_CRASH_DUMP
967 	/*
968 	 * Be aware that smp_save_dump_secondary_cpus() triggers a system reset.
969 	 * Therefore CPU and device initialization should be done afterwards.
970 	 */
971 	smp_save_dump_secondary_cpus();
972 #endif
973 
974 	setup_resources();
975 	setup_lowcore();
976 	smp_fill_possible_mask();
977 	cpu_detect_mhz_feature();
978         cpu_init();
979 	numa_setup();
980 	smp_detect_cpus();
981 	topology_init_early();
982 
983 	if (test_facility(193))
984 		static_branch_enable(&cpu_has_bear);
985 
986 	/*
987 	 * Create kernel page tables.
988 	 */
989         paging_init();
990 
991 	/*
992 	 * After paging_init created the kernel page table, the new PSWs
993 	 * in lowcore can now run with DAT enabled.
994 	 */
995 #ifdef CONFIG_CRASH_DUMP
996 	smp_save_dump_ipl_cpu();
997 #endif
998 
999         /* Setup default console */
1000 	conmode_default();
1001 	set_preferred_console();
1002 
1003 	apply_alternative_instructions();
1004 	if (IS_ENABLED(CONFIG_EXPOLINE))
1005 		nospec_init_branches();
1006 
1007 	/* Setup zfcp/nvme dump support */
1008 	setup_zfcpdump();
1009 
1010 	/* Add system specific data to the random pool */
1011 	setup_randomness();
1012 }
1013