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-contiguous.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 #ifdef CONFIG_PROTECTED_VIRTUALIZATION_GUEST
97 int __bootdata_preserved(prot_virt_guest);
98 #endif
99
100 int __bootdata(noexec_disabled);
101 int __bootdata(memory_end_set);
102 unsigned long __bootdata(memory_end);
103 unsigned long __bootdata(vmalloc_size);
104 unsigned long __bootdata(max_physmem_end);
105 struct mem_detect_info __bootdata(mem_detect);
106
107 struct exception_table_entry *__bootdata_preserved(__start_dma_ex_table);
108 struct exception_table_entry *__bootdata_preserved(__stop_dma_ex_table);
109 unsigned long __bootdata_preserved(__swsusp_reset_dma);
110 unsigned long __bootdata_preserved(__stext_dma);
111 unsigned long __bootdata_preserved(__etext_dma);
112 unsigned long __bootdata_preserved(__sdma);
113 unsigned long __bootdata_preserved(__edma);
114 unsigned long __bootdata_preserved(__kaslr_offset);
115
116 unsigned long VMALLOC_START;
117 EXPORT_SYMBOL(VMALLOC_START);
118
119 unsigned long VMALLOC_END;
120 EXPORT_SYMBOL(VMALLOC_END);
121
122 struct page *vmemmap;
123 EXPORT_SYMBOL(vmemmap);
124
125 unsigned long MODULES_VADDR;
126 unsigned long MODULES_END;
127
128 /* An array with a pointer to the lowcore of every CPU. */
129 struct lowcore *lowcore_ptr[NR_CPUS];
130 EXPORT_SYMBOL(lowcore_ptr);
131
132 /*
133 * This is set up by the setup-routine at boot-time
134 * for S390 need to find out, what we have to setup
135 * using address 0x10400 ...
136 */
137
138 #include <asm/setup.h>
139
140 /*
141 * condev= and conmode= setup parameter.
142 */
143
condev_setup(char * str)144 static int __init condev_setup(char *str)
145 {
146 int vdev;
147
148 vdev = simple_strtoul(str, &str, 0);
149 if (vdev >= 0 && vdev < 65536) {
150 console_devno = vdev;
151 console_irq = -1;
152 }
153 return 1;
154 }
155
156 __setup("condev=", condev_setup);
157
set_preferred_console(void)158 static void __init set_preferred_console(void)
159 {
160 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
161 add_preferred_console("ttyS", 0, NULL);
162 else if (CONSOLE_IS_3270)
163 add_preferred_console("tty3270", 0, NULL);
164 else if (CONSOLE_IS_VT220)
165 add_preferred_console("ttysclp", 0, NULL);
166 else if (CONSOLE_IS_HVC)
167 add_preferred_console("hvc", 0, NULL);
168 }
169
conmode_setup(char * str)170 static int __init conmode_setup(char *str)
171 {
172 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
173 if (!strcmp(str, "hwc") || !strcmp(str, "sclp"))
174 SET_CONSOLE_SCLP;
175 #endif
176 #if defined(CONFIG_TN3215_CONSOLE)
177 if (!strcmp(str, "3215"))
178 SET_CONSOLE_3215;
179 #endif
180 #if defined(CONFIG_TN3270_CONSOLE)
181 if (!strcmp(str, "3270"))
182 SET_CONSOLE_3270;
183 #endif
184 set_preferred_console();
185 return 1;
186 }
187
188 __setup("conmode=", conmode_setup);
189
conmode_default(void)190 static void __init conmode_default(void)
191 {
192 char query_buffer[1024];
193 char *ptr;
194
195 if (MACHINE_IS_VM) {
196 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
197 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
198 ptr = strstr(query_buffer, "SUBCHANNEL =");
199 console_irq = simple_strtoul(ptr + 13, NULL, 16);
200 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
201 ptr = strstr(query_buffer, "CONMODE");
202 /*
203 * Set the conmode to 3215 so that the device recognition
204 * will set the cu_type of the console to 3215. If the
205 * conmode is 3270 and we don't set it back then both
206 * 3215 and the 3270 driver will try to access the console
207 * device (3215 as console and 3270 as normal tty).
208 */
209 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
210 if (ptr == NULL) {
211 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
212 SET_CONSOLE_SCLP;
213 #endif
214 return;
215 }
216 if (str_has_prefix(ptr + 8, "3270")) {
217 #if defined(CONFIG_TN3270_CONSOLE)
218 SET_CONSOLE_3270;
219 #elif defined(CONFIG_TN3215_CONSOLE)
220 SET_CONSOLE_3215;
221 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
222 SET_CONSOLE_SCLP;
223 #endif
224 } else if (str_has_prefix(ptr + 8, "3215")) {
225 #if defined(CONFIG_TN3215_CONSOLE)
226 SET_CONSOLE_3215;
227 #elif defined(CONFIG_TN3270_CONSOLE)
228 SET_CONSOLE_3270;
229 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
230 SET_CONSOLE_SCLP;
231 #endif
232 }
233 } else if (MACHINE_IS_KVM) {
234 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
235 SET_CONSOLE_VT220;
236 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
237 SET_CONSOLE_SCLP;
238 else
239 SET_CONSOLE_HVC;
240 } else {
241 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
242 SET_CONSOLE_SCLP;
243 #endif
244 }
245 if (IS_ENABLED(CONFIG_VT) && IS_ENABLED(CONFIG_DUMMY_CONSOLE))
246 conswitchp = &dummy_con;
247 }
248
249 #ifdef CONFIG_CRASH_DUMP
setup_zfcpdump(void)250 static void __init setup_zfcpdump(void)
251 {
252 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
253 return;
254 if (OLDMEM_BASE)
255 return;
256 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
257 console_loglevel = 2;
258 }
259 #else
setup_zfcpdump(void)260 static inline void setup_zfcpdump(void) {}
261 #endif /* CONFIG_CRASH_DUMP */
262
263 /*
264 * Reboot, halt and power_off stubs. They just call _machine_restart,
265 * _machine_halt or _machine_power_off.
266 */
267
machine_restart(char * command)268 void machine_restart(char *command)
269 {
270 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
271 /*
272 * Only unblank the console if we are called in enabled
273 * context or a bust_spinlocks cleared the way for us.
274 */
275 console_unblank();
276 _machine_restart(command);
277 }
278
machine_halt(void)279 void machine_halt(void)
280 {
281 if (!in_interrupt() || oops_in_progress)
282 /*
283 * Only unblank the console if we are called in enabled
284 * context or a bust_spinlocks cleared the way for us.
285 */
286 console_unblank();
287 _machine_halt();
288 }
289
machine_power_off(void)290 void machine_power_off(void)
291 {
292 if (!in_interrupt() || oops_in_progress)
293 /*
294 * Only unblank the console if we are called in enabled
295 * context or a bust_spinlocks cleared the way for us.
296 */
297 console_unblank();
298 _machine_power_off();
299 }
300
301 /*
302 * Dummy power off function.
303 */
304 void (*pm_power_off)(void) = machine_power_off;
305 EXPORT_SYMBOL_GPL(pm_power_off);
306
307 void *restart_stack __section(.data);
308
stack_alloc(void)309 unsigned long stack_alloc(void)
310 {
311 #ifdef CONFIG_VMAP_STACK
312 return (unsigned long)
313 __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
314 VMALLOC_START, VMALLOC_END,
315 THREADINFO_GFP,
316 PAGE_KERNEL, 0, NUMA_NO_NODE,
317 __builtin_return_address(0));
318 #else
319 return __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
320 #endif
321 }
322
stack_free(unsigned long stack)323 void stack_free(unsigned long stack)
324 {
325 #ifdef CONFIG_VMAP_STACK
326 vfree((void *) stack);
327 #else
328 free_pages(stack, THREAD_SIZE_ORDER);
329 #endif
330 }
331
arch_early_irq_init(void)332 int __init arch_early_irq_init(void)
333 {
334 unsigned long stack;
335
336 stack = __get_free_pages(GFP_KERNEL, THREAD_SIZE_ORDER);
337 if (!stack)
338 panic("Couldn't allocate async stack");
339 S390_lowcore.async_stack = stack + STACK_INIT_OFFSET;
340 return 0;
341 }
342
async_stack_realloc(void)343 static int __init async_stack_realloc(void)
344 {
345 unsigned long old, new;
346
347 old = S390_lowcore.async_stack - STACK_INIT_OFFSET;
348 new = stack_alloc();
349 if (!new)
350 panic("Couldn't allocate async stack");
351 S390_lowcore.async_stack = new + STACK_INIT_OFFSET;
352 free_pages(old, THREAD_SIZE_ORDER);
353 return 0;
354 }
355 early_initcall(async_stack_realloc);
356
arch_call_rest_init(void)357 void __init arch_call_rest_init(void)
358 {
359 unsigned long stack;
360
361 stack = stack_alloc();
362 if (!stack)
363 panic("Couldn't allocate kernel stack");
364 current->stack = (void *) stack;
365 #ifdef CONFIG_VMAP_STACK
366 current->stack_vm_area = (void *) stack;
367 #endif
368 set_task_stack_end_magic(current);
369 stack += STACK_INIT_OFFSET;
370 S390_lowcore.kernel_stack = stack;
371 CALL_ON_STACK_NORETURN(rest_init, stack);
372 }
373
setup_lowcore_dat_off(void)374 static void __init setup_lowcore_dat_off(void)
375 {
376 struct lowcore *lc;
377
378 /*
379 * Setup lowcore for boot cpu
380 */
381 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
382 lc = memblock_alloc_low(sizeof(*lc), sizeof(*lc));
383 if (!lc)
384 panic("%s: Failed to allocate %zu bytes align=%zx\n",
385 __func__, sizeof(*lc), sizeof(*lc));
386
387 lc->restart_psw.mask = PSW_KERNEL_BITS;
388 lc->restart_psw.addr = (unsigned long) restart_int_handler;
389 lc->external_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
390 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
391 lc->svc_new_psw.mask = PSW_KERNEL_BITS |
392 PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
393 lc->svc_new_psw.addr = (unsigned long) system_call;
394 lc->program_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
395 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
396 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
397 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
398 lc->io_new_psw.mask = PSW_KERNEL_BITS | PSW_MASK_MCHECK;
399 lc->io_new_psw.addr = (unsigned long) io_int_handler;
400 lc->clock_comparator = clock_comparator_max;
401 lc->nodat_stack = ((unsigned long) &init_thread_union)
402 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
403 lc->current_task = (unsigned long)&init_task;
404 lc->lpp = LPP_MAGIC;
405 lc->machine_flags = S390_lowcore.machine_flags;
406 lc->preempt_count = S390_lowcore.preempt_count;
407 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
408 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
409 sizeof(lc->stfle_fac_list));
410 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
411 sizeof(lc->alt_stfle_fac_list));
412 nmi_alloc_boot_cpu(lc);
413 vdso_alloc_boot_cpu(lc);
414 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
415 lc->async_enter_timer = S390_lowcore.async_enter_timer;
416 lc->exit_timer = S390_lowcore.exit_timer;
417 lc->user_timer = S390_lowcore.user_timer;
418 lc->system_timer = S390_lowcore.system_timer;
419 lc->steal_timer = S390_lowcore.steal_timer;
420 lc->last_update_timer = S390_lowcore.last_update_timer;
421 lc->last_update_clock = S390_lowcore.last_update_clock;
422
423 /*
424 * Allocate the global restart stack which is the same for
425 * all CPUs in cast *one* of them does a PSW restart.
426 */
427 restart_stack = memblock_alloc(THREAD_SIZE, THREAD_SIZE);
428 if (!restart_stack)
429 panic("%s: Failed to allocate %lu bytes align=0x%lx\n",
430 __func__, THREAD_SIZE, THREAD_SIZE);
431 restart_stack += STACK_INIT_OFFSET;
432
433 /*
434 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
435 * restart data to the absolute zero lowcore. This is necessary if
436 * PSW restart is done on an offline CPU that has lowcore zero.
437 */
438 lc->restart_stack = (unsigned long) restart_stack;
439 lc->restart_fn = (unsigned long) do_restart;
440 lc->restart_data = 0;
441 lc->restart_source = -1UL;
442
443 /* Setup absolute zero lowcore */
444 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
445 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
446 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
447 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
448 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
449
450 lc->spinlock_lockval = arch_spin_lockval(0);
451 lc->spinlock_index = 0;
452 arch_spin_lock_setup(0);
453 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
454 lc->return_lpswe = gen_lpswe(__LC_RETURN_PSW);
455 lc->return_mcck_lpswe = gen_lpswe(__LC_RETURN_MCCK_PSW);
456
457 set_prefix((u32)(unsigned long) lc);
458 lowcore_ptr[0] = lc;
459 }
460
setup_lowcore_dat_on(void)461 static void __init setup_lowcore_dat_on(void)
462 {
463 __ctl_clear_bit(0, 28);
464 S390_lowcore.external_new_psw.mask |= PSW_MASK_DAT;
465 S390_lowcore.svc_new_psw.mask |= PSW_MASK_DAT;
466 S390_lowcore.program_new_psw.mask |= PSW_MASK_DAT;
467 S390_lowcore.io_new_psw.mask |= PSW_MASK_DAT;
468 __ctl_set_bit(0, 28);
469 }
470
471 static struct resource code_resource = {
472 .name = "Kernel code",
473 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
474 };
475
476 static struct resource data_resource = {
477 .name = "Kernel data",
478 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
479 };
480
481 static struct resource bss_resource = {
482 .name = "Kernel bss",
483 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
484 };
485
486 static struct resource __initdata *standard_resources[] = {
487 &code_resource,
488 &data_resource,
489 &bss_resource,
490 };
491
setup_resources(void)492 static void __init setup_resources(void)
493 {
494 struct resource *res, *std_res, *sub_res;
495 struct memblock_region *reg;
496 int j;
497
498 code_resource.start = (unsigned long) _text;
499 code_resource.end = (unsigned long) _etext - 1;
500 data_resource.start = (unsigned long) _etext;
501 data_resource.end = (unsigned long) _edata - 1;
502 bss_resource.start = (unsigned long) __bss_start;
503 bss_resource.end = (unsigned long) __bss_stop - 1;
504
505 for_each_memblock(memory, reg) {
506 res = memblock_alloc(sizeof(*res), 8);
507 if (!res)
508 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
509 __func__, sizeof(*res), 8);
510 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
511
512 res->name = "System RAM";
513 res->start = reg->base;
514 res->end = reg->base + reg->size - 1;
515 request_resource(&iomem_resource, res);
516
517 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
518 std_res = standard_resources[j];
519 if (std_res->start < res->start ||
520 std_res->start > res->end)
521 continue;
522 if (std_res->end > res->end) {
523 sub_res = memblock_alloc(sizeof(*sub_res), 8);
524 if (!sub_res)
525 panic("%s: Failed to allocate %zu bytes align=0x%x\n",
526 __func__, sizeof(*sub_res), 8);
527 *sub_res = *std_res;
528 sub_res->end = res->end;
529 std_res->start = res->end + 1;
530 request_resource(res, sub_res);
531 } else {
532 request_resource(res, std_res);
533 }
534 }
535 }
536 #ifdef CONFIG_CRASH_DUMP
537 /*
538 * Re-add removed crash kernel memory as reserved memory. This makes
539 * sure it will be mapped with the identity mapping and struct pages
540 * will be created, so it can be resized later on.
541 * However add it later since the crash kernel resource should not be
542 * part of the System RAM resource.
543 */
544 if (crashk_res.end) {
545 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
546 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
547 insert_resource(&iomem_resource, &crashk_res);
548 }
549 #endif
550 }
551
setup_memory_end(void)552 static void __init setup_memory_end(void)
553 {
554 unsigned long vmax, tmp;
555
556 /* Choose kernel address space layout: 3 or 4 levels. */
557 if (IS_ENABLED(CONFIG_KASAN)) {
558 vmax = IS_ENABLED(CONFIG_KASAN_S390_4_LEVEL_PAGING)
559 ? _REGION1_SIZE
560 : _REGION2_SIZE;
561 } else {
562 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
563 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
564 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
565 vmax = _REGION2_SIZE; /* 3-level kernel page table */
566 else
567 vmax = _REGION1_SIZE; /* 4-level kernel page table */
568 }
569
570 /* module area is at the end of the kernel address space. */
571 MODULES_END = vmax;
572 MODULES_VADDR = MODULES_END - MODULES_LEN;
573 VMALLOC_END = MODULES_VADDR;
574 VMALLOC_START = VMALLOC_END - vmalloc_size;
575
576 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
577 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
578 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
579 tmp = SECTION_ALIGN_UP(tmp);
580 tmp = VMALLOC_START - tmp * sizeof(struct page);
581 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
582 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
583 vmemmap = (struct page *) tmp;
584
585 /* Take care that memory_end is set and <= vmemmap */
586 memory_end = min(memory_end ?: max_physmem_end, (unsigned long)vmemmap);
587 #ifdef CONFIG_KASAN
588 /* fit in kasan shadow memory region between 1:1 and vmemmap */
589 memory_end = min(memory_end, KASAN_SHADOW_START);
590 vmemmap = max(vmemmap, (struct page *)KASAN_SHADOW_END);
591 #endif
592 max_pfn = max_low_pfn = PFN_DOWN(memory_end);
593 memblock_remove(memory_end, ULONG_MAX);
594
595 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
596 }
597
598 #ifdef CONFIG_CRASH_DUMP
599
600 /*
601 * When kdump is enabled, we have to ensure that no memory from
602 * the area [0 - crashkernel memory size] and
603 * [crashk_res.start - crashk_res.end] is set offline.
604 */
kdump_mem_notifier(struct notifier_block * nb,unsigned long action,void * data)605 static int kdump_mem_notifier(struct notifier_block *nb,
606 unsigned long action, void *data)
607 {
608 struct memory_notify *arg = data;
609
610 if (action != MEM_GOING_OFFLINE)
611 return NOTIFY_OK;
612 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
613 return NOTIFY_BAD;
614 if (arg->start_pfn > PFN_DOWN(crashk_res.end))
615 return NOTIFY_OK;
616 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
617 return NOTIFY_OK;
618 return NOTIFY_BAD;
619 }
620
621 static struct notifier_block kdump_mem_nb = {
622 .notifier_call = kdump_mem_notifier,
623 };
624
625 #endif
626
627 /*
628 * Make sure that the area behind memory_end is protected
629 */
reserve_memory_end(void)630 static void __init reserve_memory_end(void)
631 {
632 if (memory_end_set)
633 memblock_reserve(memory_end, ULONG_MAX);
634 }
635
636 /*
637 * Make sure that oldmem, where the dump is stored, is protected
638 */
reserve_oldmem(void)639 static void __init reserve_oldmem(void)
640 {
641 #ifdef CONFIG_CRASH_DUMP
642 if (OLDMEM_BASE)
643 /* Forget all memory above the running kdump system */
644 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
645 #endif
646 }
647
648 /*
649 * Make sure that oldmem, where the dump is stored, is protected
650 */
remove_oldmem(void)651 static void __init remove_oldmem(void)
652 {
653 #ifdef CONFIG_CRASH_DUMP
654 if (OLDMEM_BASE)
655 /* Forget all memory above the running kdump system */
656 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
657 #endif
658 }
659
660 /*
661 * Reserve memory for kdump kernel to be loaded with kexec
662 */
reserve_crashkernel(void)663 static void __init reserve_crashkernel(void)
664 {
665 #ifdef CONFIG_CRASH_DUMP
666 unsigned long long crash_base, crash_size;
667 phys_addr_t low, high;
668 int rc;
669
670 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
671 &crash_base);
672
673 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
674 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
675 if (rc || crash_size == 0)
676 return;
677
678 if (memblock.memory.regions[0].size < crash_size) {
679 pr_info("crashkernel reservation failed: %s\n",
680 "first memory chunk must be at least crashkernel size");
681 return;
682 }
683
684 low = crash_base ?: OLDMEM_BASE;
685 high = low + crash_size;
686 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
687 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
688 crash_base = low;
689 } else {
690 /* Find suitable area in free memory */
691 low = max_t(unsigned long, crash_size, sclp.hsa_size);
692 high = crash_base ? crash_base + crash_size : ULONG_MAX;
693
694 if (crash_base && crash_base < low) {
695 pr_info("crashkernel reservation failed: %s\n",
696 "crash_base too low");
697 return;
698 }
699 low = crash_base ?: low;
700 crash_base = memblock_find_in_range(low, high, crash_size,
701 KEXEC_CRASH_MEM_ALIGN);
702 }
703
704 if (!crash_base) {
705 pr_info("crashkernel reservation failed: %s\n",
706 "no suitable area found");
707 return;
708 }
709
710 if (register_memory_notifier(&kdump_mem_nb))
711 return;
712
713 if (!OLDMEM_BASE && MACHINE_IS_VM)
714 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
715 crashk_res.start = crash_base;
716 crashk_res.end = crash_base + crash_size - 1;
717 memblock_remove(crash_base, crash_size);
718 pr_info("Reserving %lluMB of memory at %lluMB "
719 "for crashkernel (System RAM: %luMB)\n",
720 crash_size >> 20, crash_base >> 20,
721 (unsigned long)memblock.memory.total_size >> 20);
722 os_info_crashkernel_add(crash_base, crash_size);
723 #endif
724 }
725
726 /*
727 * Reserve the initrd from being used by memblock
728 */
reserve_initrd(void)729 static void __init reserve_initrd(void)
730 {
731 #ifdef CONFIG_BLK_DEV_INITRD
732 if (!INITRD_START || !INITRD_SIZE)
733 return;
734 initrd_start = INITRD_START;
735 initrd_end = initrd_start + INITRD_SIZE;
736 memblock_reserve(INITRD_START, INITRD_SIZE);
737 #endif
738 }
739
740 /*
741 * Reserve the memory area used to pass the certificate lists
742 */
reserve_certificate_list(void)743 static void __init reserve_certificate_list(void)
744 {
745 if (ipl_cert_list_addr)
746 memblock_reserve(ipl_cert_list_addr, ipl_cert_list_size);
747 }
748
reserve_mem_detect_info(void)749 static void __init reserve_mem_detect_info(void)
750 {
751 unsigned long start, size;
752
753 get_mem_detect_reserved(&start, &size);
754 if (size)
755 memblock_reserve(start, size);
756 }
757
free_mem_detect_info(void)758 static void __init free_mem_detect_info(void)
759 {
760 unsigned long start, size;
761
762 get_mem_detect_reserved(&start, &size);
763 if (size)
764 memblock_free(start, size);
765 }
766
memblock_physmem_add(phys_addr_t start,phys_addr_t size)767 static void __init memblock_physmem_add(phys_addr_t start, phys_addr_t size)
768 {
769 memblock_dbg("memblock_physmem_add: [%#016llx-%#016llx]\n",
770 start, start + size - 1);
771 memblock_add_range(&memblock.memory, start, size, 0, 0);
772 memblock_add_range(&memblock.physmem, start, size, 0, 0);
773 }
774
get_mem_info_source(void)775 static const char * __init get_mem_info_source(void)
776 {
777 switch (mem_detect.info_source) {
778 case MEM_DETECT_SCLP_STOR_INFO:
779 return "sclp storage info";
780 case MEM_DETECT_DIAG260:
781 return "diag260";
782 case MEM_DETECT_SCLP_READ_INFO:
783 return "sclp read info";
784 case MEM_DETECT_BIN_SEARCH:
785 return "binary search";
786 }
787 return "none";
788 }
789
memblock_add_mem_detect_info(void)790 static void __init memblock_add_mem_detect_info(void)
791 {
792 unsigned long start, end;
793 int i;
794
795 memblock_dbg("physmem info source: %s (%hhd)\n",
796 get_mem_info_source(), mem_detect.info_source);
797 /* keep memblock lists close to the kernel */
798 memblock_set_bottom_up(true);
799 for_each_mem_detect_block(i, &start, &end)
800 memblock_physmem_add(start, end - start);
801 memblock_set_bottom_up(false);
802 memblock_dump_all();
803 }
804
805 /*
806 * Check for initrd being in usable memory
807 */
check_initrd(void)808 static void __init check_initrd(void)
809 {
810 #ifdef CONFIG_BLK_DEV_INITRD
811 if (INITRD_START && INITRD_SIZE &&
812 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
813 pr_err("The initial RAM disk does not fit into the memory\n");
814 memblock_free(INITRD_START, INITRD_SIZE);
815 initrd_start = initrd_end = 0;
816 }
817 #endif
818 }
819
820 /*
821 * Reserve memory used for lowcore/command line/kernel image.
822 */
reserve_kernel(void)823 static void __init reserve_kernel(void)
824 {
825 unsigned long start_pfn = PFN_UP(__pa(_end));
826
827 memblock_reserve(0, HEAD_END);
828 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
829 - (unsigned long)_stext);
830 memblock_reserve(__sdma, __edma - __sdma);
831 }
832
setup_memory(void)833 static void __init setup_memory(void)
834 {
835 struct memblock_region *reg;
836
837 /*
838 * Init storage key for present memory
839 */
840 for_each_memblock(memory, reg) {
841 storage_key_init_range(reg->base, reg->base + reg->size);
842 }
843 psw_set_key(PAGE_DEFAULT_KEY);
844 }
845
846 /*
847 * Setup hardware capabilities.
848 */
setup_hwcaps(void)849 static int __init setup_hwcaps(void)
850 {
851 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
852 struct cpuid cpu_id;
853 int i;
854
855 /*
856 * The store facility list bits numbers as found in the principles
857 * of operation are numbered with bit 1UL<<31 as number 0 to
858 * bit 1UL<<0 as number 31.
859 * Bit 0: instructions named N3, "backported" to esa-mode
860 * Bit 2: z/Architecture mode is active
861 * Bit 7: the store-facility-list-extended facility is installed
862 * Bit 17: the message-security assist is installed
863 * Bit 19: the long-displacement facility is installed
864 * Bit 21: the extended-immediate facility is installed
865 * Bit 22: extended-translation facility 3 is installed
866 * Bit 30: extended-translation facility 3 enhancement facility
867 * These get translated to:
868 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
869 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
870 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
871 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
872 */
873 for (i = 0; i < 6; i++)
874 if (test_facility(stfl_bits[i]))
875 elf_hwcap |= 1UL << i;
876
877 if (test_facility(22) && test_facility(30))
878 elf_hwcap |= HWCAP_S390_ETF3EH;
879
880 /*
881 * Check for additional facilities with store-facility-list-extended.
882 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
883 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
884 * as stored by stfl, bits 32-xxx contain additional facilities.
885 * How many facility words are stored depends on the number of
886 * doublewords passed to the instruction. The additional facilities
887 * are:
888 * Bit 42: decimal floating point facility is installed
889 * Bit 44: perform floating point operation facility is installed
890 * translated to:
891 * HWCAP_S390_DFP bit 6 (42 && 44).
892 */
893 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
894 elf_hwcap |= HWCAP_S390_DFP;
895
896 /*
897 * Huge page support HWCAP_S390_HPAGE is bit 7.
898 */
899 if (MACHINE_HAS_EDAT1)
900 elf_hwcap |= HWCAP_S390_HPAGE;
901
902 /*
903 * 64-bit register support for 31-bit processes
904 * HWCAP_S390_HIGH_GPRS is bit 9.
905 */
906 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
907
908 /*
909 * Transactional execution support HWCAP_S390_TE is bit 10.
910 */
911 if (MACHINE_HAS_TE)
912 elf_hwcap |= HWCAP_S390_TE;
913
914 /*
915 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
916 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
917 * instead of facility bit 129.
918 */
919 if (MACHINE_HAS_VX) {
920 elf_hwcap |= HWCAP_S390_VXRS;
921 if (test_facility(134))
922 elf_hwcap |= HWCAP_S390_VXRS_BCD;
923 if (test_facility(135))
924 elf_hwcap |= HWCAP_S390_VXRS_EXT;
925 if (test_facility(148))
926 elf_hwcap |= HWCAP_S390_VXRS_EXT2;
927 if (test_facility(152))
928 elf_hwcap |= HWCAP_S390_VXRS_PDE;
929 }
930 if (test_facility(150))
931 elf_hwcap |= HWCAP_S390_SORT;
932 if (test_facility(151))
933 elf_hwcap |= HWCAP_S390_DFLT;
934
935 /*
936 * Guarded storage support HWCAP_S390_GS is bit 12.
937 */
938 if (MACHINE_HAS_GS)
939 elf_hwcap |= HWCAP_S390_GS;
940
941 get_cpu_id(&cpu_id);
942 add_device_randomness(&cpu_id, sizeof(cpu_id));
943 switch (cpu_id.machine) {
944 case 0x2064:
945 case 0x2066:
946 default: /* Use "z900" as default for 64 bit kernels. */
947 strcpy(elf_platform, "z900");
948 break;
949 case 0x2084:
950 case 0x2086:
951 strcpy(elf_platform, "z990");
952 break;
953 case 0x2094:
954 case 0x2096:
955 strcpy(elf_platform, "z9-109");
956 break;
957 case 0x2097:
958 case 0x2098:
959 strcpy(elf_platform, "z10");
960 break;
961 case 0x2817:
962 case 0x2818:
963 strcpy(elf_platform, "z196");
964 break;
965 case 0x2827:
966 case 0x2828:
967 strcpy(elf_platform, "zEC12");
968 break;
969 case 0x2964:
970 case 0x2965:
971 strcpy(elf_platform, "z13");
972 break;
973 case 0x3906:
974 case 0x3907:
975 strcpy(elf_platform, "z14");
976 break;
977 case 0x8561:
978 case 0x8562:
979 strcpy(elf_platform, "z15");
980 break;
981 }
982
983 /*
984 * Virtualization support HWCAP_INT_SIE is bit 0.
985 */
986 if (sclp.has_sief2)
987 int_hwcap |= HWCAP_INT_SIE;
988
989 return 0;
990 }
991 arch_initcall(setup_hwcaps);
992
993 /*
994 * Add system information as device randomness
995 */
setup_randomness(void)996 static void __init setup_randomness(void)
997 {
998 struct sysinfo_3_2_2 *vmms;
999
1000 vmms = (struct sysinfo_3_2_2 *) memblock_phys_alloc(PAGE_SIZE,
1001 PAGE_SIZE);
1002 if (!vmms)
1003 panic("Failed to allocate memory for sysinfo structure\n");
1004
1005 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
1006 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
1007 memblock_free((unsigned long) vmms, PAGE_SIZE);
1008
1009 #ifdef CONFIG_ARCH_RANDOM
1010 if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG))
1011 static_branch_enable(&s390_arch_random_available);
1012 #endif
1013 }
1014
1015 /*
1016 * Find the correct size for the task_struct. This depends on
1017 * the size of the struct fpu at the end of the thread_struct
1018 * which is embedded in the task_struct.
1019 */
setup_task_size(void)1020 static void __init setup_task_size(void)
1021 {
1022 int task_size = sizeof(struct task_struct);
1023
1024 if (!MACHINE_HAS_VX) {
1025 task_size -= sizeof(__vector128) * __NUM_VXRS;
1026 task_size += sizeof(freg_t) * __NUM_FPRS;
1027 }
1028 arch_task_struct_size = task_size;
1029 }
1030
1031 /*
1032 * Issue diagnose 318 to set the control program name and
1033 * version codes.
1034 */
setup_control_program_code(void)1035 static void __init setup_control_program_code(void)
1036 {
1037 union diag318_info diag318_info = {
1038 .cpnc = CPNC_LINUX,
1039 .cpvc_linux = 0,
1040 .cpvc_distro = {0},
1041 };
1042
1043 if (!sclp.has_diag318)
1044 return;
1045
1046 diag_stat_inc(DIAG_STAT_X318);
1047 asm volatile("diag %0,0,0x318\n" : : "d" (diag318_info.val));
1048 }
1049
1050 /*
1051 * Print the component list from the IPL report
1052 */
log_component_list(void)1053 static void __init log_component_list(void)
1054 {
1055 struct ipl_rb_component_entry *ptr, *end;
1056 char *str;
1057
1058 if (!early_ipl_comp_list_addr)
1059 return;
1060 if (ipl_block.hdr.flags & IPL_PL_FLAG_SIPL)
1061 pr_info("Linux is running with Secure-IPL enabled\n");
1062 else
1063 pr_info("Linux is running with Secure-IPL disabled\n");
1064 ptr = (void *) early_ipl_comp_list_addr;
1065 end = (void *) ptr + early_ipl_comp_list_size;
1066 pr_info("The IPL report contains the following components:\n");
1067 while (ptr < end) {
1068 if (ptr->flags & IPL_RB_COMPONENT_FLAG_SIGNED) {
1069 if (ptr->flags & IPL_RB_COMPONENT_FLAG_VERIFIED)
1070 str = "signed, verified";
1071 else
1072 str = "signed, verification failed";
1073 } else {
1074 str = "not signed";
1075 }
1076 pr_info("%016llx - %016llx (%s)\n",
1077 ptr->addr, ptr->addr + ptr->len, str);
1078 ptr++;
1079 }
1080 }
1081
1082 /*
1083 * Setup function called from init/main.c just after the banner
1084 * was printed.
1085 */
1086
setup_arch(char ** cmdline_p)1087 void __init setup_arch(char **cmdline_p)
1088 {
1089 /*
1090 * print what head.S has found out about the machine
1091 */
1092 if (MACHINE_IS_VM)
1093 pr_info("Linux is running as a z/VM "
1094 "guest operating system in 64-bit mode\n");
1095 else if (MACHINE_IS_KVM)
1096 pr_info("Linux is running under KVM in 64-bit mode\n");
1097 else if (MACHINE_IS_LPAR)
1098 pr_info("Linux is running natively in 64-bit mode\n");
1099 else
1100 pr_info("Linux is running as a guest in 64-bit mode\n");
1101
1102 log_component_list();
1103
1104 /* Have one command line that is parsed and saved in /proc/cmdline */
1105 /* boot_command_line has been already set up in early.c */
1106 *cmdline_p = boot_command_line;
1107
1108 ROOT_DEV = Root_RAM0;
1109
1110 init_mm.start_code = (unsigned long) _text;
1111 init_mm.end_code = (unsigned long) _etext;
1112 init_mm.end_data = (unsigned long) _edata;
1113 init_mm.brk = (unsigned long) _end;
1114
1115 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
1116 nospec_auto_detect();
1117
1118 jump_label_init();
1119 parse_early_param();
1120 #ifdef CONFIG_CRASH_DUMP
1121 /* Deactivate elfcorehdr= kernel parameter */
1122 elfcorehdr_addr = ELFCORE_ADDR_MAX;
1123 #endif
1124
1125 os_info_init();
1126 setup_ipl();
1127 setup_task_size();
1128 setup_control_program_code();
1129
1130 /* Do some memory reservations *before* memory is added to memblock */
1131 reserve_memory_end();
1132 reserve_oldmem();
1133 reserve_kernel();
1134 reserve_initrd();
1135 reserve_certificate_list();
1136 reserve_mem_detect_info();
1137 memblock_allow_resize();
1138
1139 /* Get information about *all* installed memory */
1140 memblock_add_mem_detect_info();
1141
1142 free_mem_detect_info();
1143 remove_oldmem();
1144
1145 /*
1146 * Make sure all chunks are MAX_ORDER aligned so we don't need the
1147 * extra checks that HOLES_IN_ZONE would require.
1148 *
1149 * Is this still required?
1150 */
1151 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
1152
1153 setup_memory_end();
1154 setup_memory();
1155 dma_contiguous_reserve(memory_end);
1156 vmcp_cma_reserve();
1157
1158 check_initrd();
1159 reserve_crashkernel();
1160 #ifdef CONFIG_CRASH_DUMP
1161 /*
1162 * Be aware that smp_save_dump_cpus() triggers a system reset.
1163 * Therefore CPU and device initialization should be done afterwards.
1164 */
1165 smp_save_dump_cpus();
1166 #endif
1167
1168 setup_resources();
1169 setup_lowcore_dat_off();
1170 smp_fill_possible_mask();
1171 cpu_detect_mhz_feature();
1172 cpu_init();
1173 numa_setup();
1174 smp_detect_cpus();
1175 topology_init_early();
1176
1177 /*
1178 * Create kernel page tables and switch to virtual addressing.
1179 */
1180 paging_init();
1181
1182 /*
1183 * After paging_init created the kernel page table, the new PSWs
1184 * in lowcore can now run with DAT enabled.
1185 */
1186 setup_lowcore_dat_on();
1187
1188 /* Setup default console */
1189 conmode_default();
1190 set_preferred_console();
1191
1192 apply_alternative_instructions();
1193 if (IS_ENABLED(CONFIG_EXPOLINE))
1194 nospec_init_branches();
1195
1196 /* Setup zfcpdump support */
1197 setup_zfcpdump();
1198
1199 /* Add system specific data to the random pool */
1200 setup_randomness();
1201 }
1202