1 /*
2 * S390 version
3 * Copyright IBM Corp. 1999, 2012
4 * Author(s): Hartmut Penner (hp@de.ibm.com),
5 * Martin Schwidefsky (schwidefsky@de.ibm.com)
6 *
7 * Derived from "arch/i386/kernel/setup.c"
8 * Copyright (C) 1995, Linus Torvalds
9 */
10
11 /*
12 * This file handles the architecture-dependent parts of initialization
13 */
14
15 #define KMSG_COMPONENT "setup"
16 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
17
18 #include <linux/errno.h>
19 #include <linux/export.h>
20 #include <linux/sched.h>
21 #include <linux/kernel.h>
22 #include <linux/memblock.h>
23 #include <linux/mm.h>
24 #include <linux/stddef.h>
25 #include <linux/unistd.h>
26 #include <linux/ptrace.h>
27 #include <linux/random.h>
28 #include <linux/user.h>
29 #include <linux/tty.h>
30 #include <linux/ioport.h>
31 #include <linux/delay.h>
32 #include <linux/init.h>
33 #include <linux/initrd.h>
34 #include <linux/bootmem.h>
35 #include <linux/root_dev.h>
36 #include <linux/console.h>
37 #include <linux/kernel_stat.h>
38 #include <linux/device.h>
39 #include <linux/notifier.h>
40 #include <linux/pfn.h>
41 #include <linux/ctype.h>
42 #include <linux/reboot.h>
43 #include <linux/topology.h>
44 #include <linux/kexec.h>
45 #include <linux/crash_dump.h>
46 #include <linux/memory.h>
47 #include <linux/compat.h>
48
49 #include <asm/ipl.h>
50 #include <asm/facility.h>
51 #include <asm/smp.h>
52 #include <asm/mmu_context.h>
53 #include <asm/cpcmd.h>
54 #include <asm/lowcore.h>
55 #include <asm/irq.h>
56 #include <asm/page.h>
57 #include <asm/ptrace.h>
58 #include <asm/sections.h>
59 #include <asm/ebcdic.h>
60 #include <asm/kvm_virtio.h>
61 #include <asm/diag.h>
62 #include <asm/os_info.h>
63 #include <asm/sclp.h>
64 #include <asm/sysinfo.h>
65 #include <asm/numa.h>
66 #include <asm/alternative.h>
67 #include <asm/nospec-branch.h>
68 #include "entry.h"
69
70 /*
71 * Machine setup..
72 */
73 unsigned int console_mode = 0;
74 EXPORT_SYMBOL(console_mode);
75
76 unsigned int console_devno = -1;
77 EXPORT_SYMBOL(console_devno);
78
79 unsigned int console_irq = -1;
80 EXPORT_SYMBOL(console_irq);
81
82 unsigned long elf_hwcap __read_mostly = 0;
83 char elf_platform[ELF_PLATFORM_SIZE];
84
85 unsigned long int_hwcap = 0;
86
87 int __initdata memory_end_set;
88 unsigned long __initdata memory_end;
89 unsigned long __initdata max_physmem_end;
90
91 unsigned long VMALLOC_START;
92 EXPORT_SYMBOL(VMALLOC_START);
93
94 unsigned long VMALLOC_END;
95 EXPORT_SYMBOL(VMALLOC_END);
96
97 struct page *vmemmap;
98 EXPORT_SYMBOL(vmemmap);
99
100 unsigned long MODULES_VADDR;
101 unsigned long MODULES_END;
102
103 /* An array with a pointer to the lowcore of every CPU. */
104 struct lowcore *lowcore_ptr[NR_CPUS];
105 EXPORT_SYMBOL(lowcore_ptr);
106
107 /*
108 * This is set up by the setup-routine at boot-time
109 * for S390 need to find out, what we have to setup
110 * using address 0x10400 ...
111 */
112
113 #include <asm/setup.h>
114
115 /*
116 * condev= and conmode= setup parameter.
117 */
118
condev_setup(char * str)119 static int __init condev_setup(char *str)
120 {
121 int vdev;
122
123 vdev = simple_strtoul(str, &str, 0);
124 if (vdev >= 0 && vdev < 65536) {
125 console_devno = vdev;
126 console_irq = -1;
127 }
128 return 1;
129 }
130
131 __setup("condev=", condev_setup);
132
set_preferred_console(void)133 static void __init set_preferred_console(void)
134 {
135 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
136 add_preferred_console("ttyS", 0, NULL);
137 else if (CONSOLE_IS_3270)
138 add_preferred_console("tty3270", 0, NULL);
139 else if (CONSOLE_IS_VT220)
140 add_preferred_console("ttyS", 1, NULL);
141 else if (CONSOLE_IS_HVC)
142 add_preferred_console("hvc", 0, NULL);
143 }
144
conmode_setup(char * str)145 static int __init conmode_setup(char *str)
146 {
147 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
148 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
149 SET_CONSOLE_SCLP;
150 #endif
151 #if defined(CONFIG_TN3215_CONSOLE)
152 if (strncmp(str, "3215", 5) == 0)
153 SET_CONSOLE_3215;
154 #endif
155 #if defined(CONFIG_TN3270_CONSOLE)
156 if (strncmp(str, "3270", 5) == 0)
157 SET_CONSOLE_3270;
158 #endif
159 set_preferred_console();
160 return 1;
161 }
162
163 __setup("conmode=", conmode_setup);
164
conmode_default(void)165 static void __init conmode_default(void)
166 {
167 char query_buffer[1024];
168 char *ptr;
169
170 if (MACHINE_IS_VM) {
171 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
172 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
173 ptr = strstr(query_buffer, "SUBCHANNEL =");
174 console_irq = simple_strtoul(ptr + 13, NULL, 16);
175 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
176 ptr = strstr(query_buffer, "CONMODE");
177 /*
178 * Set the conmode to 3215 so that the device recognition
179 * will set the cu_type of the console to 3215. If the
180 * conmode is 3270 and we don't set it back then both
181 * 3215 and the 3270 driver will try to access the console
182 * device (3215 as console and 3270 as normal tty).
183 */
184 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
185 if (ptr == NULL) {
186 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
187 SET_CONSOLE_SCLP;
188 #endif
189 return;
190 }
191 if (strncmp(ptr + 8, "3270", 4) == 0) {
192 #if defined(CONFIG_TN3270_CONSOLE)
193 SET_CONSOLE_3270;
194 #elif defined(CONFIG_TN3215_CONSOLE)
195 SET_CONSOLE_3215;
196 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
197 SET_CONSOLE_SCLP;
198 #endif
199 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
200 #if defined(CONFIG_TN3215_CONSOLE)
201 SET_CONSOLE_3215;
202 #elif defined(CONFIG_TN3270_CONSOLE)
203 SET_CONSOLE_3270;
204 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
205 SET_CONSOLE_SCLP;
206 #endif
207 }
208 } else if (MACHINE_IS_KVM) {
209 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
210 SET_CONSOLE_VT220;
211 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
212 SET_CONSOLE_SCLP;
213 else
214 SET_CONSOLE_HVC;
215 } else {
216 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
217 SET_CONSOLE_SCLP;
218 #endif
219 }
220 }
221
222 #ifdef CONFIG_CRASH_DUMP
setup_zfcpdump(void)223 static void __init setup_zfcpdump(void)
224 {
225 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
226 return;
227 if (OLDMEM_BASE)
228 return;
229 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
230 console_loglevel = 2;
231 }
232 #else
setup_zfcpdump(void)233 static inline void setup_zfcpdump(void) {}
234 #endif /* CONFIG_CRASH_DUMP */
235
236 /*
237 * Reboot, halt and power_off stubs. They just call _machine_restart,
238 * _machine_halt or _machine_power_off.
239 */
240
machine_restart(char * command)241 void machine_restart(char *command)
242 {
243 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
244 /*
245 * Only unblank the console if we are called in enabled
246 * context or a bust_spinlocks cleared the way for us.
247 */
248 console_unblank();
249 _machine_restart(command);
250 }
251
machine_halt(void)252 void machine_halt(void)
253 {
254 if (!in_interrupt() || oops_in_progress)
255 /*
256 * Only unblank the console if we are called in enabled
257 * context or a bust_spinlocks cleared the way for us.
258 */
259 console_unblank();
260 _machine_halt();
261 }
262
machine_power_off(void)263 void machine_power_off(void)
264 {
265 if (!in_interrupt() || oops_in_progress)
266 /*
267 * Only unblank the console if we are called in enabled
268 * context or a bust_spinlocks cleared the way for us.
269 */
270 console_unblank();
271 _machine_power_off();
272 }
273
274 /*
275 * Dummy power off function.
276 */
277 void (*pm_power_off)(void) = machine_power_off;
278 EXPORT_SYMBOL_GPL(pm_power_off);
279
early_parse_mem(char * p)280 static int __init early_parse_mem(char *p)
281 {
282 memory_end = memparse(p, &p);
283 memory_end &= PAGE_MASK;
284 memory_end_set = 1;
285 return 0;
286 }
287 early_param("mem", early_parse_mem);
288
parse_vmalloc(char * arg)289 static int __init parse_vmalloc(char *arg)
290 {
291 if (!arg)
292 return -EINVAL;
293 VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
294 return 0;
295 }
296 early_param("vmalloc", parse_vmalloc);
297
298 void *restart_stack __section(.data);
299
setup_lowcore(void)300 static void __init setup_lowcore(void)
301 {
302 struct lowcore *lc;
303
304 /*
305 * Setup lowcore for boot cpu
306 */
307 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * 4096);
308 lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
309 lc->restart_psw.mask = PSW_KERNEL_BITS;
310 lc->restart_psw.addr = (unsigned long) restart_int_handler;
311 lc->external_new_psw.mask = PSW_KERNEL_BITS |
312 PSW_MASK_DAT | PSW_MASK_MCHECK;
313 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
314 lc->svc_new_psw.mask = PSW_KERNEL_BITS |
315 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
316 lc->svc_new_psw.addr = (unsigned long) system_call;
317 lc->program_new_psw.mask = PSW_KERNEL_BITS |
318 PSW_MASK_DAT | PSW_MASK_MCHECK;
319 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
320 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
321 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
322 lc->io_new_psw.mask = PSW_KERNEL_BITS |
323 PSW_MASK_DAT | PSW_MASK_MCHECK;
324 lc->io_new_psw.addr = (unsigned long) io_int_handler;
325 lc->clock_comparator = -1ULL;
326 lc->kernel_stack = ((unsigned long) &init_thread_union)
327 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
328 lc->async_stack = (unsigned long)
329 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0)
330 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
331 lc->panic_stack = (unsigned long)
332 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0)
333 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
334 lc->current_task = (unsigned long) init_thread_union.thread_info.task;
335 lc->thread_info = (unsigned long) &init_thread_union;
336 lc->lpp = LPP_MAGIC;
337 lc->machine_flags = S390_lowcore.machine_flags;
338 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
339 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
340 sizeof(lc->stfle_fac_list));
341 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
342 sizeof(lc->alt_stfle_fac_list));
343 if (MACHINE_HAS_VX)
344 lc->vector_save_area_addr =
345 (unsigned long) &lc->vector_save_area;
346 lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
347 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
348 lc->async_enter_timer = S390_lowcore.async_enter_timer;
349 lc->exit_timer = S390_lowcore.exit_timer;
350 lc->user_timer = S390_lowcore.user_timer;
351 lc->system_timer = S390_lowcore.system_timer;
352 lc->steal_timer = S390_lowcore.steal_timer;
353 lc->last_update_timer = S390_lowcore.last_update_timer;
354 lc->last_update_clock = S390_lowcore.last_update_clock;
355
356 restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
357 restart_stack += ASYNC_SIZE;
358
359 /*
360 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
361 * restart data to the absolute zero lowcore. This is necessary if
362 * PSW restart is done on an offline CPU that has lowcore zero.
363 */
364 lc->restart_stack = (unsigned long) restart_stack;
365 lc->restart_fn = (unsigned long) do_restart;
366 lc->restart_data = 0;
367 lc->restart_source = -1UL;
368
369 /* Setup absolute zero lowcore */
370 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
371 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
372 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
373 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
374 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
375
376 #ifdef CONFIG_SMP
377 lc->spinlock_lockval = arch_spin_lockval(0);
378 #endif
379 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
380
381 set_prefix((u32)(unsigned long) lc);
382 lowcore_ptr[0] = lc;
383 }
384
385 static struct resource code_resource = {
386 .name = "Kernel code",
387 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
388 };
389
390 static struct resource data_resource = {
391 .name = "Kernel data",
392 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
393 };
394
395 static struct resource bss_resource = {
396 .name = "Kernel bss",
397 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
398 };
399
400 static struct resource __initdata *standard_resources[] = {
401 &code_resource,
402 &data_resource,
403 &bss_resource,
404 };
405
setup_resources(void)406 static void __init setup_resources(void)
407 {
408 struct resource *res, *std_res, *sub_res;
409 struct memblock_region *reg;
410 int j;
411
412 code_resource.start = (unsigned long) &_text;
413 code_resource.end = (unsigned long) &_etext - 1;
414 data_resource.start = (unsigned long) &_etext;
415 data_resource.end = (unsigned long) &_edata - 1;
416 bss_resource.start = (unsigned long) &__bss_start;
417 bss_resource.end = (unsigned long) &__bss_stop - 1;
418
419 for_each_memblock(memory, reg) {
420 res = alloc_bootmem_low(sizeof(*res));
421 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
422
423 res->name = "System RAM";
424 res->start = reg->base;
425 res->end = reg->base + reg->size - 1;
426 request_resource(&iomem_resource, res);
427
428 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
429 std_res = standard_resources[j];
430 if (std_res->start < res->start ||
431 std_res->start > res->end)
432 continue;
433 if (std_res->end > res->end) {
434 sub_res = alloc_bootmem_low(sizeof(*sub_res));
435 *sub_res = *std_res;
436 sub_res->end = res->end;
437 std_res->start = res->end + 1;
438 request_resource(res, sub_res);
439 } else {
440 request_resource(res, std_res);
441 }
442 }
443 }
444 #ifdef CONFIG_CRASH_DUMP
445 /*
446 * Re-add removed crash kernel memory as reserved memory. This makes
447 * sure it will be mapped with the identity mapping and struct pages
448 * will be created, so it can be resized later on.
449 * However add it later since the crash kernel resource should not be
450 * part of the System RAM resource.
451 */
452 if (crashk_res.end) {
453 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
454 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
455 insert_resource(&iomem_resource, &crashk_res);
456 }
457 #endif
458 }
459
setup_memory_end(void)460 static void __init setup_memory_end(void)
461 {
462 unsigned long vmax, vmalloc_size, tmp;
463
464 /* Choose kernel address space layout: 2, 3, or 4 levels. */
465 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
466 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
467 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
468 if (tmp + vmalloc_size + MODULES_LEN <= (1UL << 42))
469 vmax = 1UL << 42; /* 3-level kernel page table */
470 else
471 vmax = 1UL << 53; /* 4-level kernel page table */
472 /* module area is at the end of the kernel address space. */
473 MODULES_END = vmax;
474 MODULES_VADDR = MODULES_END - MODULES_LEN;
475 VMALLOC_END = MODULES_VADDR;
476 VMALLOC_START = vmax - vmalloc_size;
477
478 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
479 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
480 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
481 tmp = SECTION_ALIGN_UP(tmp);
482 tmp = VMALLOC_START - tmp * sizeof(struct page);
483 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
484 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
485 vmemmap = (struct page *) tmp;
486
487 /* Take care that memory_end is set and <= vmemmap */
488 memory_end = min(memory_end ?: max_physmem_end, tmp);
489 max_pfn = max_low_pfn = PFN_DOWN(memory_end);
490 memblock_remove(memory_end, ULONG_MAX);
491
492 pr_notice("Max memory size: %luMB\n", memory_end >> 20);
493 }
494
setup_vmcoreinfo(void)495 static void __init setup_vmcoreinfo(void)
496 {
497 mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
498 }
499
500 #ifdef CONFIG_CRASH_DUMP
501
502 /*
503 * When kdump is enabled, we have to ensure that no memory from
504 * the area [0 - crashkernel memory size] and
505 * [crashk_res.start - crashk_res.end] is set offline.
506 */
kdump_mem_notifier(struct notifier_block * nb,unsigned long action,void * data)507 static int kdump_mem_notifier(struct notifier_block *nb,
508 unsigned long action, void *data)
509 {
510 struct memory_notify *arg = data;
511
512 if (action != MEM_GOING_OFFLINE)
513 return NOTIFY_OK;
514 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
515 return NOTIFY_BAD;
516 if (arg->start_pfn > PFN_DOWN(crashk_res.end))
517 return NOTIFY_OK;
518 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
519 return NOTIFY_OK;
520 return NOTIFY_BAD;
521 }
522
523 static struct notifier_block kdump_mem_nb = {
524 .notifier_call = kdump_mem_notifier,
525 };
526
527 #endif
528
529 /*
530 * Make sure that the area behind memory_end is protected
531 */
reserve_memory_end(void)532 static void reserve_memory_end(void)
533 {
534 #ifdef CONFIG_CRASH_DUMP
535 if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
536 !OLDMEM_BASE && sclp.hsa_size) {
537 memory_end = sclp.hsa_size;
538 memory_end &= PAGE_MASK;
539 memory_end_set = 1;
540 }
541 #endif
542 if (!memory_end_set)
543 return;
544 memblock_reserve(memory_end, ULONG_MAX);
545 }
546
547 /*
548 * Make sure that oldmem, where the dump is stored, is protected
549 */
reserve_oldmem(void)550 static void reserve_oldmem(void)
551 {
552 #ifdef CONFIG_CRASH_DUMP
553 if (OLDMEM_BASE)
554 /* Forget all memory above the running kdump system */
555 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
556 #endif
557 }
558
559 /*
560 * Make sure that oldmem, where the dump is stored, is protected
561 */
remove_oldmem(void)562 static void remove_oldmem(void)
563 {
564 #ifdef CONFIG_CRASH_DUMP
565 if (OLDMEM_BASE)
566 /* Forget all memory above the running kdump system */
567 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
568 #endif
569 }
570
571 /*
572 * Reserve memory for kdump kernel to be loaded with kexec
573 */
reserve_crashkernel(void)574 static void __init reserve_crashkernel(void)
575 {
576 #ifdef CONFIG_CRASH_DUMP
577 unsigned long long crash_base, crash_size;
578 phys_addr_t low, high;
579 int rc;
580
581 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
582 &crash_base);
583
584 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
585 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
586 if (rc || crash_size == 0)
587 return;
588
589 if (memblock.memory.regions[0].size < crash_size) {
590 pr_info("crashkernel reservation failed: %s\n",
591 "first memory chunk must be at least crashkernel size");
592 return;
593 }
594
595 low = crash_base ?: OLDMEM_BASE;
596 high = low + crash_size;
597 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
598 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
599 crash_base = low;
600 } else {
601 /* Find suitable area in free memory */
602 low = max_t(unsigned long, crash_size, sclp.hsa_size);
603 high = crash_base ? crash_base + crash_size : ULONG_MAX;
604
605 if (crash_base && crash_base < low) {
606 pr_info("crashkernel reservation failed: %s\n",
607 "crash_base too low");
608 return;
609 }
610 low = crash_base ?: low;
611 crash_base = memblock_find_in_range(low, high, crash_size,
612 KEXEC_CRASH_MEM_ALIGN);
613 }
614
615 if (!crash_base) {
616 pr_info("crashkernel reservation failed: %s\n",
617 "no suitable area found");
618 return;
619 }
620
621 if (register_memory_notifier(&kdump_mem_nb))
622 return;
623
624 if (!OLDMEM_BASE && MACHINE_IS_VM)
625 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
626 crashk_res.start = crash_base;
627 crashk_res.end = crash_base + crash_size - 1;
628 memblock_remove(crash_base, crash_size);
629 pr_info("Reserving %lluMB of memory at %lluMB "
630 "for crashkernel (System RAM: %luMB)\n",
631 crash_size >> 20, crash_base >> 20,
632 (unsigned long)memblock.memory.total_size >> 20);
633 os_info_crashkernel_add(crash_base, crash_size);
634 #endif
635 }
636
637 /*
638 * Reserve the initrd from being used by memblock
639 */
reserve_initrd(void)640 static void __init reserve_initrd(void)
641 {
642 #ifdef CONFIG_BLK_DEV_INITRD
643 initrd_start = INITRD_START;
644 initrd_end = initrd_start + INITRD_SIZE;
645 memblock_reserve(INITRD_START, INITRD_SIZE);
646 #endif
647 }
648
649 /*
650 * Check for initrd being in usable memory
651 */
check_initrd(void)652 static void __init check_initrd(void)
653 {
654 #ifdef CONFIG_BLK_DEV_INITRD
655 if (INITRD_START && INITRD_SIZE &&
656 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
657 pr_err("initrd does not fit memory.\n");
658 memblock_free(INITRD_START, INITRD_SIZE);
659 initrd_start = initrd_end = 0;
660 }
661 #endif
662 }
663
664 /*
665 * Reserve memory used for lowcore/command line/kernel image.
666 */
reserve_kernel(void)667 static void __init reserve_kernel(void)
668 {
669 unsigned long start_pfn = PFN_UP(__pa(&_end));
670
671 #ifdef CONFIG_DMA_API_DEBUG
672 /*
673 * DMA_API_DEBUG code stumbles over addresses from the
674 * range [_ehead, _stext]. Mark the memory as reserved
675 * so it is not used for CONFIG_DMA_API_DEBUG=y.
676 */
677 memblock_reserve(0, PFN_PHYS(start_pfn));
678 #else
679 memblock_reserve(0, (unsigned long)_ehead);
680 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
681 - (unsigned long)_stext);
682 #endif
683 }
684
setup_memory(void)685 static void __init setup_memory(void)
686 {
687 struct memblock_region *reg;
688
689 /*
690 * Init storage key for present memory
691 */
692 for_each_memblock(memory, reg) {
693 storage_key_init_range(reg->base, reg->base + reg->size);
694 }
695 psw_set_key(PAGE_DEFAULT_KEY);
696
697 /* Only cosmetics */
698 memblock_enforce_memory_limit(memblock_end_of_DRAM());
699 }
700
701 /*
702 * Setup hardware capabilities.
703 */
setup_hwcaps(void)704 static int __init setup_hwcaps(void)
705 {
706 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
707 struct cpuid cpu_id;
708 int i;
709
710 /*
711 * The store facility list bits numbers as found in the principles
712 * of operation are numbered with bit 1UL<<31 as number 0 to
713 * bit 1UL<<0 as number 31.
714 * Bit 0: instructions named N3, "backported" to esa-mode
715 * Bit 2: z/Architecture mode is active
716 * Bit 7: the store-facility-list-extended facility is installed
717 * Bit 17: the message-security assist is installed
718 * Bit 19: the long-displacement facility is installed
719 * Bit 21: the extended-immediate facility is installed
720 * Bit 22: extended-translation facility 3 is installed
721 * Bit 30: extended-translation facility 3 enhancement facility
722 * These get translated to:
723 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
724 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
725 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
726 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
727 */
728 for (i = 0; i < 6; i++)
729 if (test_facility(stfl_bits[i]))
730 elf_hwcap |= 1UL << i;
731
732 if (test_facility(22) && test_facility(30))
733 elf_hwcap |= HWCAP_S390_ETF3EH;
734
735 /*
736 * Check for additional facilities with store-facility-list-extended.
737 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
738 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
739 * as stored by stfl, bits 32-xxx contain additional facilities.
740 * How many facility words are stored depends on the number of
741 * doublewords passed to the instruction. The additional facilities
742 * are:
743 * Bit 42: decimal floating point facility is installed
744 * Bit 44: perform floating point operation facility is installed
745 * translated to:
746 * HWCAP_S390_DFP bit 6 (42 && 44).
747 */
748 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
749 elf_hwcap |= HWCAP_S390_DFP;
750
751 /*
752 * Huge page support HWCAP_S390_HPAGE is bit 7.
753 */
754 if (MACHINE_HAS_HPAGE)
755 elf_hwcap |= HWCAP_S390_HPAGE;
756
757 /*
758 * 64-bit register support for 31-bit processes
759 * HWCAP_S390_HIGH_GPRS is bit 9.
760 */
761 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
762
763 /*
764 * Transactional execution support HWCAP_S390_TE is bit 10.
765 */
766 if (test_facility(50) && test_facility(73))
767 elf_hwcap |= HWCAP_S390_TE;
768
769 /*
770 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
771 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
772 * instead of facility bit 129.
773 */
774 if (MACHINE_HAS_VX)
775 elf_hwcap |= HWCAP_S390_VXRS;
776 get_cpu_id(&cpu_id);
777 add_device_randomness(&cpu_id, sizeof(cpu_id));
778 switch (cpu_id.machine) {
779 case 0x2064:
780 case 0x2066:
781 default: /* Use "z900" as default for 64 bit kernels. */
782 strcpy(elf_platform, "z900");
783 break;
784 case 0x2084:
785 case 0x2086:
786 strcpy(elf_platform, "z990");
787 break;
788 case 0x2094:
789 case 0x2096:
790 strcpy(elf_platform, "z9-109");
791 break;
792 case 0x2097:
793 case 0x2098:
794 strcpy(elf_platform, "z10");
795 break;
796 case 0x2817:
797 case 0x2818:
798 strcpy(elf_platform, "z196");
799 break;
800 case 0x2827:
801 case 0x2828:
802 strcpy(elf_platform, "zEC12");
803 break;
804 case 0x2964:
805 case 0x2965:
806 strcpy(elf_platform, "z13");
807 break;
808 }
809
810 /*
811 * Virtualization support HWCAP_INT_SIE is bit 0.
812 */
813 if (sclp.has_sief2)
814 int_hwcap |= HWCAP_INT_SIE;
815
816 return 0;
817 }
818 arch_initcall(setup_hwcaps);
819
820 /*
821 * Add system information as device randomness
822 */
setup_randomness(void)823 static void __init setup_randomness(void)
824 {
825 struct sysinfo_3_2_2 *vmms;
826
827 vmms = (struct sysinfo_3_2_2 *) memblock_alloc(PAGE_SIZE, PAGE_SIZE);
828 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
829 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
830 memblock_free((unsigned long) vmms, PAGE_SIZE);
831 }
832
833 /*
834 * Find the correct size for the task_struct. This depends on
835 * the size of the struct fpu at the end of the thread_struct
836 * which is embedded in the task_struct.
837 */
setup_task_size(void)838 static void __init setup_task_size(void)
839 {
840 int task_size = sizeof(struct task_struct);
841
842 if (!MACHINE_HAS_VX) {
843 task_size -= sizeof(__vector128) * __NUM_VXRS;
844 task_size += sizeof(freg_t) * __NUM_FPRS;
845 }
846 arch_task_struct_size = task_size;
847 }
848
849 /*
850 * Setup function called from init/main.c just after the banner
851 * was printed.
852 */
853
setup_arch(char ** cmdline_p)854 void __init setup_arch(char **cmdline_p)
855 {
856 /*
857 * print what head.S has found out about the machine
858 */
859 if (MACHINE_IS_VM)
860 pr_info("Linux is running as a z/VM "
861 "guest operating system in 64-bit mode\n");
862 else if (MACHINE_IS_KVM)
863 pr_info("Linux is running under KVM in 64-bit mode\n");
864 else if (MACHINE_IS_LPAR)
865 pr_info("Linux is running natively in 64-bit mode\n");
866
867 /* Have one command line that is parsed and saved in /proc/cmdline */
868 /* boot_command_line has been already set up in early.c */
869 *cmdline_p = boot_command_line;
870
871 ROOT_DEV = Root_RAM0;
872
873 /* Is init_mm really needed? */
874 init_mm.start_code = PAGE_OFFSET;
875 init_mm.end_code = (unsigned long) &_etext;
876 init_mm.end_data = (unsigned long) &_edata;
877 init_mm.brk = (unsigned long) &_end;
878
879 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
880 nospec_auto_detect();
881
882 parse_early_param();
883 #ifdef CONFIG_CRASH_DUMP
884 /* Deactivate elfcorehdr= kernel parameter */
885 elfcorehdr_addr = ELFCORE_ADDR_MAX;
886 #endif
887
888 os_info_init();
889 setup_ipl();
890 setup_task_size();
891
892 /* Do some memory reservations *before* memory is added to memblock */
893 reserve_memory_end();
894 reserve_oldmem();
895 reserve_kernel();
896 reserve_initrd();
897 memblock_allow_resize();
898
899 /* Get information about *all* installed memory */
900 detect_memory_memblock();
901
902 remove_oldmem();
903
904 /*
905 * Make sure all chunks are MAX_ORDER aligned so we don't need the
906 * extra checks that HOLES_IN_ZONE would require.
907 *
908 * Is this still required?
909 */
910 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
911
912 setup_memory_end();
913 setup_memory();
914
915 check_initrd();
916 reserve_crashkernel();
917 #ifdef CONFIG_CRASH_DUMP
918 /*
919 * Be aware that smp_save_dump_cpus() triggers a system reset.
920 * Therefore CPU and device initialization should be done afterwards.
921 */
922 smp_save_dump_cpus();
923 #endif
924
925 setup_resources();
926 setup_vmcoreinfo();
927 setup_lowcore();
928 smp_fill_possible_mask();
929 cpu_detect_mhz_feature();
930 cpu_init();
931 numa_setup();
932
933 /*
934 * Create kernel page tables and switch to virtual addressing.
935 */
936 paging_init();
937
938 /* Setup default console */
939 conmode_default();
940 set_preferred_console();
941
942 apply_alternative_instructions();
943 if (IS_ENABLED(CONFIG_EXPOLINE))
944 nospec_init_branches();
945
946 /* Setup zfcpdump support */
947 setup_zfcpdump();
948
949 /* Add system specific data to the random pool */
950 setup_randomness();
951 }
952