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