]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - arch/s390/kernel/setup.c
Merge remote-tracking branches 'asoc/topic/sgtl5000', 'asoc/topic/simple', 'asoc...
[mirror_ubuntu-zesty-kernel.git] / arch / s390 / kernel / setup.c
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 "entry.h"
67
68 /*
69 * Machine setup..
70 */
71 unsigned int console_mode = 0;
72 EXPORT_SYMBOL(console_mode);
73
74 unsigned int console_devno = -1;
75 EXPORT_SYMBOL(console_devno);
76
77 unsigned int console_irq = -1;
78 EXPORT_SYMBOL(console_irq);
79
80 unsigned long elf_hwcap __read_mostly = 0;
81 char elf_platform[ELF_PLATFORM_SIZE];
82
83 unsigned long int_hwcap = 0;
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
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
131 static void __init set_preferred_console(void)
132 {
133 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
134 add_preferred_console("ttyS", 0, NULL);
135 else if (CONSOLE_IS_3270)
136 add_preferred_console("tty3270", 0, NULL);
137 else if (CONSOLE_IS_VT220)
138 add_preferred_console("ttyS", 1, NULL);
139 else if (CONSOLE_IS_HVC)
140 add_preferred_console("hvc", 0, NULL);
141 }
142
143 static int __init conmode_setup(char *str)
144 {
145 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
146 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
147 SET_CONSOLE_SCLP;
148 #endif
149 #if defined(CONFIG_TN3215_CONSOLE)
150 if (strncmp(str, "3215", 5) == 0)
151 SET_CONSOLE_3215;
152 #endif
153 #if defined(CONFIG_TN3270_CONSOLE)
154 if (strncmp(str, "3270", 5) == 0)
155 SET_CONSOLE_3270;
156 #endif
157 set_preferred_console();
158 return 1;
159 }
160
161 __setup("conmode=", conmode_setup);
162
163 static void __init conmode_default(void)
164 {
165 char query_buffer[1024];
166 char *ptr;
167
168 if (MACHINE_IS_VM) {
169 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
170 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
171 ptr = strstr(query_buffer, "SUBCHANNEL =");
172 console_irq = simple_strtoul(ptr + 13, NULL, 16);
173 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
174 ptr = strstr(query_buffer, "CONMODE");
175 /*
176 * Set the conmode to 3215 so that the device recognition
177 * will set the cu_type of the console to 3215. If the
178 * conmode is 3270 and we don't set it back then both
179 * 3215 and the 3270 driver will try to access the console
180 * device (3215 as console and 3270 as normal tty).
181 */
182 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
183 if (ptr == NULL) {
184 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
185 SET_CONSOLE_SCLP;
186 #endif
187 return;
188 }
189 if (strncmp(ptr + 8, "3270", 4) == 0) {
190 #if defined(CONFIG_TN3270_CONSOLE)
191 SET_CONSOLE_3270;
192 #elif defined(CONFIG_TN3215_CONSOLE)
193 SET_CONSOLE_3215;
194 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
195 SET_CONSOLE_SCLP;
196 #endif
197 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
198 #if defined(CONFIG_TN3215_CONSOLE)
199 SET_CONSOLE_3215;
200 #elif defined(CONFIG_TN3270_CONSOLE)
201 SET_CONSOLE_3270;
202 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
203 SET_CONSOLE_SCLP;
204 #endif
205 }
206 } else if (MACHINE_IS_KVM) {
207 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
208 SET_CONSOLE_VT220;
209 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
210 SET_CONSOLE_SCLP;
211 else
212 SET_CONSOLE_HVC;
213 } else {
214 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
215 SET_CONSOLE_SCLP;
216 #endif
217 }
218 }
219
220 #ifdef CONFIG_CRASH_DUMP
221 static void __init setup_zfcpdump(void)
222 {
223 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
224 return;
225 if (OLDMEM_BASE)
226 return;
227 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
228 console_loglevel = 2;
229 }
230 #else
231 static inline void setup_zfcpdump(void) {}
232 #endif /* CONFIG_CRASH_DUMP */
233
234 /*
235 * Reboot, halt and power_off stubs. They just call _machine_restart,
236 * _machine_halt or _machine_power_off.
237 */
238
239 void machine_restart(char *command)
240 {
241 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
242 /*
243 * Only unblank the console if we are called in enabled
244 * context or a bust_spinlocks cleared the way for us.
245 */
246 console_unblank();
247 _machine_restart(command);
248 }
249
250 void machine_halt(void)
251 {
252 if (!in_interrupt() || oops_in_progress)
253 /*
254 * Only unblank the console if we are called in enabled
255 * context or a bust_spinlocks cleared the way for us.
256 */
257 console_unblank();
258 _machine_halt();
259 }
260
261 void machine_power_off(void)
262 {
263 if (!in_interrupt() || oops_in_progress)
264 /*
265 * Only unblank the console if we are called in enabled
266 * context or a bust_spinlocks cleared the way for us.
267 */
268 console_unblank();
269 _machine_power_off();
270 }
271
272 /*
273 * Dummy power off function.
274 */
275 void (*pm_power_off)(void) = machine_power_off;
276 EXPORT_SYMBOL_GPL(pm_power_off);
277
278 static int __init early_parse_mem(char *p)
279 {
280 memory_end = memparse(p, &p);
281 memory_end &= PAGE_MASK;
282 memory_end_set = 1;
283 return 0;
284 }
285 early_param("mem", early_parse_mem);
286
287 static int __init parse_vmalloc(char *arg)
288 {
289 if (!arg)
290 return -EINVAL;
291 VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
292 return 0;
293 }
294 early_param("vmalloc", parse_vmalloc);
295
296 void *restart_stack __section(.data);
297
298 static void __init setup_lowcore(void)
299 {
300 struct lowcore *lc;
301
302 /*
303 * Setup lowcore for boot cpu
304 */
305 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * 4096);
306 lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
307 lc->restart_psw.mask = PSW_KERNEL_BITS;
308 lc->restart_psw.addr = (unsigned long) restart_int_handler;
309 lc->external_new_psw.mask = PSW_KERNEL_BITS |
310 PSW_MASK_DAT | PSW_MASK_MCHECK;
311 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
312 lc->svc_new_psw.mask = PSW_KERNEL_BITS |
313 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
314 lc->svc_new_psw.addr = (unsigned long) system_call;
315 lc->program_new_psw.mask = PSW_KERNEL_BITS |
316 PSW_MASK_DAT | PSW_MASK_MCHECK;
317 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
318 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
319 lc->mcck_new_psw.addr = (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 = (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 MAX_FACILITY_BIT/8);
339 if (MACHINE_HAS_VX)
340 lc->vector_save_area_addr =
341 (unsigned long) &lc->vector_save_area;
342 lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
343 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
344 lc->async_enter_timer = S390_lowcore.async_enter_timer;
345 lc->exit_timer = S390_lowcore.exit_timer;
346 lc->user_timer = S390_lowcore.user_timer;
347 lc->system_timer = S390_lowcore.system_timer;
348 lc->steal_timer = S390_lowcore.steal_timer;
349 lc->last_update_timer = S390_lowcore.last_update_timer;
350 lc->last_update_clock = S390_lowcore.last_update_clock;
351
352 restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
353 restart_stack += ASYNC_SIZE;
354
355 /*
356 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
357 * restart data to the absolute zero lowcore. This is necessary if
358 * PSW restart is done on an offline CPU that has lowcore zero.
359 */
360 lc->restart_stack = (unsigned long) restart_stack;
361 lc->restart_fn = (unsigned long) do_restart;
362 lc->restart_data = 0;
363 lc->restart_source = -1UL;
364
365 /* Setup absolute zero lowcore */
366 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
367 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
368 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
369 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
370 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
371
372 #ifdef CONFIG_SMP
373 lc->spinlock_lockval = arch_spin_lockval(0);
374 #endif
375
376 set_prefix((u32)(unsigned long) lc);
377 lowcore_ptr[0] = lc;
378 }
379
380 static struct resource code_resource = {
381 .name = "Kernel code",
382 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
383 };
384
385 static struct resource data_resource = {
386 .name = "Kernel data",
387 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
388 };
389
390 static struct resource bss_resource = {
391 .name = "Kernel bss",
392 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
393 };
394
395 static struct resource __initdata *standard_resources[] = {
396 &code_resource,
397 &data_resource,
398 &bss_resource,
399 };
400
401 static void __init setup_resources(void)
402 {
403 struct resource *res, *std_res, *sub_res;
404 struct memblock_region *reg;
405 int j;
406
407 code_resource.start = (unsigned long) &_text;
408 code_resource.end = (unsigned long) &_etext - 1;
409 data_resource.start = (unsigned long) &_etext;
410 data_resource.end = (unsigned long) &_edata - 1;
411 bss_resource.start = (unsigned long) &__bss_start;
412 bss_resource.end = (unsigned long) &__bss_stop - 1;
413
414 for_each_memblock(memory, reg) {
415 res = alloc_bootmem_low(sizeof(*res));
416 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
417
418 res->name = "System RAM";
419 res->start = reg->base;
420 res->end = reg->base + reg->size - 1;
421 request_resource(&iomem_resource, res);
422
423 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
424 std_res = standard_resources[j];
425 if (std_res->start < res->start ||
426 std_res->start > res->end)
427 continue;
428 if (std_res->end > res->end) {
429 sub_res = alloc_bootmem_low(sizeof(*sub_res));
430 *sub_res = *std_res;
431 sub_res->end = res->end;
432 std_res->start = res->end + 1;
433 request_resource(res, sub_res);
434 } else {
435 request_resource(res, std_res);
436 }
437 }
438 }
439 #ifdef CONFIG_CRASH_DUMP
440 /*
441 * Re-add removed crash kernel memory as reserved memory. This makes
442 * sure it will be mapped with the identity mapping and struct pages
443 * will be created, so it can be resized later on.
444 * However add it later since the crash kernel resource should not be
445 * part of the System RAM resource.
446 */
447 if (crashk_res.end) {
448 memblock_add(crashk_res.start, resource_size(&crashk_res));
449 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
450 insert_resource(&iomem_resource, &crashk_res);
451 }
452 #endif
453 }
454
455 static void __init setup_memory_end(void)
456 {
457 unsigned long vmax, vmalloc_size, tmp;
458
459 /* Choose kernel address space layout: 2, 3, or 4 levels. */
460 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
461 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
462 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
463 if (tmp + vmalloc_size + MODULES_LEN <= (1UL << 42))
464 vmax = 1UL << 42; /* 3-level kernel page table */
465 else
466 vmax = 1UL << 53; /* 4-level kernel page table */
467 /* module area is at the end of the kernel address space. */
468 MODULES_END = vmax;
469 MODULES_VADDR = MODULES_END - MODULES_LEN;
470 VMALLOC_END = MODULES_VADDR;
471 VMALLOC_START = vmax - vmalloc_size;
472
473 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
474 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
475 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
476 tmp = SECTION_ALIGN_UP(tmp);
477 tmp = VMALLOC_START - tmp * sizeof(struct page);
478 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
479 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
480 vmemmap = (struct page *) tmp;
481
482 /* Take care that memory_end is set and <= vmemmap */
483 memory_end = min(memory_end ?: max_physmem_end, tmp);
484 max_pfn = max_low_pfn = PFN_DOWN(memory_end);
485 memblock_remove(memory_end, ULONG_MAX);
486
487 pr_notice("Max memory size: %luMB\n", memory_end >> 20);
488 }
489
490 static void __init setup_vmcoreinfo(void)
491 {
492 mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
493 }
494
495 #ifdef CONFIG_CRASH_DUMP
496
497 /*
498 * When kdump is enabled, we have to ensure that no memory from
499 * the area [0 - crashkernel memory size] and
500 * [crashk_res.start - crashk_res.end] is set offline.
501 */
502 static int kdump_mem_notifier(struct notifier_block *nb,
503 unsigned long action, void *data)
504 {
505 struct memory_notify *arg = data;
506
507 if (action != MEM_GOING_OFFLINE)
508 return NOTIFY_OK;
509 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
510 return NOTIFY_BAD;
511 if (arg->start_pfn > PFN_DOWN(crashk_res.end))
512 return NOTIFY_OK;
513 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
514 return NOTIFY_OK;
515 return NOTIFY_BAD;
516 }
517
518 static struct notifier_block kdump_mem_nb = {
519 .notifier_call = kdump_mem_notifier,
520 };
521
522 #endif
523
524 /*
525 * Make sure that the area behind memory_end is protected
526 */
527 static void reserve_memory_end(void)
528 {
529 #ifdef CONFIG_CRASH_DUMP
530 if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
531 !OLDMEM_BASE && sclp.hsa_size) {
532 memory_end = sclp.hsa_size;
533 memory_end &= PAGE_MASK;
534 memory_end_set = 1;
535 }
536 #endif
537 if (!memory_end_set)
538 return;
539 memblock_reserve(memory_end, ULONG_MAX);
540 }
541
542 /*
543 * Make sure that oldmem, where the dump is stored, is protected
544 */
545 static void reserve_oldmem(void)
546 {
547 #ifdef CONFIG_CRASH_DUMP
548 if (OLDMEM_BASE)
549 /* Forget all memory above the running kdump system */
550 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
551 #endif
552 }
553
554 /*
555 * Make sure that oldmem, where the dump is stored, is protected
556 */
557 static void remove_oldmem(void)
558 {
559 #ifdef CONFIG_CRASH_DUMP
560 if (OLDMEM_BASE)
561 /* Forget all memory above the running kdump system */
562 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
563 #endif
564 }
565
566 /*
567 * Reserve memory for kdump kernel to be loaded with kexec
568 */
569 static void __init reserve_crashkernel(void)
570 {
571 #ifdef CONFIG_CRASH_DUMP
572 unsigned long long crash_base, crash_size;
573 phys_addr_t low, high;
574 int rc;
575
576 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
577 &crash_base);
578
579 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
580 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
581 if (rc || crash_size == 0)
582 return;
583
584 if (memblock.memory.regions[0].size < crash_size) {
585 pr_info("crashkernel reservation failed: %s\n",
586 "first memory chunk must be at least crashkernel size");
587 return;
588 }
589
590 low = crash_base ?: OLDMEM_BASE;
591 high = low + crash_size;
592 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
593 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
594 crash_base = low;
595 } else {
596 /* Find suitable area in free memory */
597 low = max_t(unsigned long, crash_size, sclp.hsa_size);
598 high = crash_base ? crash_base + crash_size : ULONG_MAX;
599
600 if (crash_base && crash_base < low) {
601 pr_info("crashkernel reservation failed: %s\n",
602 "crash_base too low");
603 return;
604 }
605 low = crash_base ?: low;
606 crash_base = memblock_find_in_range(low, high, crash_size,
607 KEXEC_CRASH_MEM_ALIGN);
608 }
609
610 if (!crash_base) {
611 pr_info("crashkernel reservation failed: %s\n",
612 "no suitable area found");
613 return;
614 }
615
616 if (register_memory_notifier(&kdump_mem_nb))
617 return;
618
619 if (!OLDMEM_BASE && MACHINE_IS_VM)
620 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
621 crashk_res.start = crash_base;
622 crashk_res.end = crash_base + crash_size - 1;
623 memblock_remove(crash_base, crash_size);
624 pr_info("Reserving %lluMB of memory at %lluMB "
625 "for crashkernel (System RAM: %luMB)\n",
626 crash_size >> 20, crash_base >> 20,
627 (unsigned long)memblock.memory.total_size >> 20);
628 os_info_crashkernel_add(crash_base, crash_size);
629 #endif
630 }
631
632 /*
633 * Reserve the initrd from being used by memblock
634 */
635 static void __init reserve_initrd(void)
636 {
637 #ifdef CONFIG_BLK_DEV_INITRD
638 initrd_start = INITRD_START;
639 initrd_end = initrd_start + INITRD_SIZE;
640 memblock_reserve(INITRD_START, INITRD_SIZE);
641 #endif
642 }
643
644 /*
645 * Check for initrd being in usable memory
646 */
647 static void __init check_initrd(void)
648 {
649 #ifdef CONFIG_BLK_DEV_INITRD
650 if (INITRD_START && INITRD_SIZE &&
651 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
652 pr_err("initrd does not fit memory.\n");
653 memblock_free(INITRD_START, INITRD_SIZE);
654 initrd_start = initrd_end = 0;
655 }
656 #endif
657 }
658
659 /*
660 * Reserve memory used for lowcore/command line/kernel image.
661 */
662 static void __init reserve_kernel(void)
663 {
664 unsigned long start_pfn = PFN_UP(__pa(&_end));
665
666 #ifdef CONFIG_DMA_API_DEBUG
667 /*
668 * DMA_API_DEBUG code stumbles over addresses from the
669 * range [_ehead, _stext]. Mark the memory as reserved
670 * so it is not used for CONFIG_DMA_API_DEBUG=y.
671 */
672 memblock_reserve(0, PFN_PHYS(start_pfn));
673 #else
674 memblock_reserve(0, (unsigned long)_ehead);
675 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
676 - (unsigned long)_stext);
677 #endif
678 }
679
680 static void __init setup_memory(void)
681 {
682 struct memblock_region *reg;
683
684 /*
685 * Init storage key for present memory
686 */
687 for_each_memblock(memory, reg) {
688 storage_key_init_range(reg->base, reg->base + reg->size);
689 }
690 psw_set_key(PAGE_DEFAULT_KEY);
691
692 /* Only cosmetics */
693 memblock_enforce_memory_limit(memblock_end_of_DRAM());
694 }
695
696 /*
697 * Setup hardware capabilities.
698 */
699 static int __init setup_hwcaps(void)
700 {
701 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
702 struct cpuid cpu_id;
703 int i;
704
705 /*
706 * The store facility list bits numbers as found in the principles
707 * of operation are numbered with bit 1UL<<31 as number 0 to
708 * bit 1UL<<0 as number 31.
709 * Bit 0: instructions named N3, "backported" to esa-mode
710 * Bit 2: z/Architecture mode is active
711 * Bit 7: the store-facility-list-extended facility is installed
712 * Bit 17: the message-security assist is installed
713 * Bit 19: the long-displacement facility is installed
714 * Bit 21: the extended-immediate facility is installed
715 * Bit 22: extended-translation facility 3 is installed
716 * Bit 30: extended-translation facility 3 enhancement facility
717 * These get translated to:
718 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
719 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
720 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
721 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
722 */
723 for (i = 0; i < 6; i++)
724 if (test_facility(stfl_bits[i]))
725 elf_hwcap |= 1UL << i;
726
727 if (test_facility(22) && test_facility(30))
728 elf_hwcap |= HWCAP_S390_ETF3EH;
729
730 /*
731 * Check for additional facilities with store-facility-list-extended.
732 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
733 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
734 * as stored by stfl, bits 32-xxx contain additional facilities.
735 * How many facility words are stored depends on the number of
736 * doublewords passed to the instruction. The additional facilities
737 * are:
738 * Bit 42: decimal floating point facility is installed
739 * Bit 44: perform floating point operation facility is installed
740 * translated to:
741 * HWCAP_S390_DFP bit 6 (42 && 44).
742 */
743 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
744 elf_hwcap |= HWCAP_S390_DFP;
745
746 /*
747 * Huge page support HWCAP_S390_HPAGE is bit 7.
748 */
749 if (MACHINE_HAS_HPAGE)
750 elf_hwcap |= HWCAP_S390_HPAGE;
751
752 /*
753 * 64-bit register support for 31-bit processes
754 * HWCAP_S390_HIGH_GPRS is bit 9.
755 */
756 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
757
758 /*
759 * Transactional execution support HWCAP_S390_TE is bit 10.
760 */
761 if (test_facility(50) && test_facility(73))
762 elf_hwcap |= HWCAP_S390_TE;
763
764 /*
765 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
766 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
767 * instead of facility bit 129.
768 */
769 if (MACHINE_HAS_VX)
770 elf_hwcap |= HWCAP_S390_VXRS;
771 get_cpu_id(&cpu_id);
772 add_device_randomness(&cpu_id, sizeof(cpu_id));
773 switch (cpu_id.machine) {
774 case 0x2064:
775 case 0x2066:
776 default: /* Use "z900" as default for 64 bit kernels. */
777 strcpy(elf_platform, "z900");
778 break;
779 case 0x2084:
780 case 0x2086:
781 strcpy(elf_platform, "z990");
782 break;
783 case 0x2094:
784 case 0x2096:
785 strcpy(elf_platform, "z9-109");
786 break;
787 case 0x2097:
788 case 0x2098:
789 strcpy(elf_platform, "z10");
790 break;
791 case 0x2817:
792 case 0x2818:
793 strcpy(elf_platform, "z196");
794 break;
795 case 0x2827:
796 case 0x2828:
797 strcpy(elf_platform, "zEC12");
798 break;
799 case 0x2964:
800 case 0x2965:
801 strcpy(elf_platform, "z13");
802 break;
803 }
804
805 /*
806 * Virtualization support HWCAP_INT_SIE is bit 0.
807 */
808 if (sclp.has_sief2)
809 int_hwcap |= HWCAP_INT_SIE;
810
811 return 0;
812 }
813 arch_initcall(setup_hwcaps);
814
815 /*
816 * Add system information as device randomness
817 */
818 static void __init setup_randomness(void)
819 {
820 struct sysinfo_3_2_2 *vmms;
821
822 vmms = (struct sysinfo_3_2_2 *) alloc_page(GFP_KERNEL);
823 if (vmms && stsi(vmms, 3, 2, 2) == 0 && vmms->count)
824 add_device_randomness(&vmms, vmms->count);
825 free_page((unsigned long) vmms);
826 }
827
828 /*
829 * Find the correct size for the task_struct. This depends on
830 * the size of the struct fpu at the end of the thread_struct
831 * which is embedded in the task_struct.
832 */
833 static void __init setup_task_size(void)
834 {
835 int task_size = sizeof(struct task_struct);
836
837 if (!MACHINE_HAS_VX) {
838 task_size -= sizeof(__vector128) * __NUM_VXRS;
839 task_size += sizeof(freg_t) * __NUM_FPRS;
840 }
841 arch_task_struct_size = task_size;
842 }
843
844 /*
845 * Setup function called from init/main.c just after the banner
846 * was printed.
847 */
848
849 void __init setup_arch(char **cmdline_p)
850 {
851 /*
852 * print what head.S has found out about the machine
853 */
854 if (MACHINE_IS_VM)
855 pr_info("Linux is running as a z/VM "
856 "guest operating system in 64-bit mode\n");
857 else if (MACHINE_IS_KVM)
858 pr_info("Linux is running under KVM in 64-bit mode\n");
859 else if (MACHINE_IS_LPAR)
860 pr_info("Linux is running natively in 64-bit mode\n");
861
862 /* Have one command line that is parsed and saved in /proc/cmdline */
863 /* boot_command_line has been already set up in early.c */
864 *cmdline_p = boot_command_line;
865
866 ROOT_DEV = Root_RAM0;
867
868 /* Is init_mm really needed? */
869 init_mm.start_code = PAGE_OFFSET;
870 init_mm.end_code = (unsigned long) &_etext;
871 init_mm.end_data = (unsigned long) &_edata;
872 init_mm.brk = (unsigned long) &_end;
873
874 parse_early_param();
875 #ifdef CONFIG_CRASH_DUMP
876 /* Deactivate elfcorehdr= kernel parameter */
877 elfcorehdr_addr = ELFCORE_ADDR_MAX;
878 #endif
879
880 os_info_init();
881 setup_ipl();
882 setup_task_size();
883
884 /* Do some memory reservations *before* memory is added to memblock */
885 reserve_memory_end();
886 reserve_oldmem();
887 reserve_kernel();
888 reserve_initrd();
889 memblock_allow_resize();
890
891 /* Get information about *all* installed memory */
892 detect_memory_memblock();
893
894 remove_oldmem();
895
896 /*
897 * Make sure all chunks are MAX_ORDER aligned so we don't need the
898 * extra checks that HOLES_IN_ZONE would require.
899 *
900 * Is this still required?
901 */
902 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
903
904 setup_memory_end();
905 setup_memory();
906
907 check_initrd();
908 reserve_crashkernel();
909 #ifdef CONFIG_CRASH_DUMP
910 /*
911 * Be aware that smp_save_dump_cpus() triggers a system reset.
912 * Therefore CPU and device initialization should be done afterwards.
913 */
914 smp_save_dump_cpus();
915 #endif
916
917 setup_resources();
918 setup_vmcoreinfo();
919 setup_lowcore();
920 smp_fill_possible_mask();
921 cpu_detect_mhz_feature();
922 cpu_init();
923 numa_setup();
924
925 /*
926 * Create kernel page tables and switch to virtual addressing.
927 */
928 paging_init();
929
930 /* Setup default console */
931 conmode_default();
932 set_preferred_console();
933
934 /* Setup zfcpdump support */
935 setup_zfcpdump();
936
937 /* Add system specific data to the random pool */
938 setup_randomness();
939 }