Loading...
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/bootmem.h>
38#include <linux/root_dev.h>
39#include <linux/console.h>
40#include <linux/kernel_stat.h>
41#include <linux/dma-contiguous.h>
42#include <linux/device.h>
43#include <linux/notifier.h>
44#include <linux/pfn.h>
45#include <linux/ctype.h>
46#include <linux/reboot.h>
47#include <linux/topology.h>
48#include <linux/kexec.h>
49#include <linux/crash_dump.h>
50#include <linux/memory.h>
51#include <linux/compat.h>
52
53#include <asm/ipl.h>
54#include <asm/facility.h>
55#include <asm/smp.h>
56#include <asm/mmu_context.h>
57#include <asm/cpcmd.h>
58#include <asm/lowcore.h>
59#include <asm/nmi.h>
60#include <asm/irq.h>
61#include <asm/page.h>
62#include <asm/ptrace.h>
63#include <asm/sections.h>
64#include <asm/ebcdic.h>
65#include <asm/diag.h>
66#include <asm/os_info.h>
67#include <asm/sclp.h>
68#include <asm/sysinfo.h>
69#include <asm/numa.h>
70#include <asm/alternative.h>
71#include <asm/nospec-branch.h>
72#include "entry.h"
73
74/*
75 * Machine setup..
76 */
77unsigned int console_mode = 0;
78EXPORT_SYMBOL(console_mode);
79
80unsigned int console_devno = -1;
81EXPORT_SYMBOL(console_devno);
82
83unsigned int console_irq = -1;
84EXPORT_SYMBOL(console_irq);
85
86unsigned long elf_hwcap __read_mostly = 0;
87char elf_platform[ELF_PLATFORM_SIZE];
88
89unsigned long int_hwcap = 0;
90
91int __initdata memory_end_set;
92unsigned long __initdata memory_end;
93unsigned long __initdata max_physmem_end;
94
95unsigned long VMALLOC_START;
96EXPORT_SYMBOL(VMALLOC_START);
97
98unsigned long VMALLOC_END;
99EXPORT_SYMBOL(VMALLOC_END);
100
101struct page *vmemmap;
102EXPORT_SYMBOL(vmemmap);
103
104unsigned long MODULES_VADDR;
105unsigned long MODULES_END;
106
107/* An array with a pointer to the lowcore of every CPU. */
108struct lowcore *lowcore_ptr[NR_CPUS];
109EXPORT_SYMBOL(lowcore_ptr);
110
111/*
112 * This is set up by the setup-routine at boot-time
113 * for S390 need to find out, what we have to setup
114 * using address 0x10400 ...
115 */
116
117#include <asm/setup.h>
118
119/*
120 * condev= and conmode= setup parameter.
121 */
122
123static int __init condev_setup(char *str)
124{
125 int vdev;
126
127 vdev = simple_strtoul(str, &str, 0);
128 if (vdev >= 0 && vdev < 65536) {
129 console_devno = vdev;
130 console_irq = -1;
131 }
132 return 1;
133}
134
135__setup("condev=", condev_setup);
136
137static void __init set_preferred_console(void)
138{
139 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
140 add_preferred_console("ttyS", 0, NULL);
141 else if (CONSOLE_IS_3270)
142 add_preferred_console("tty3270", 0, NULL);
143 else if (CONSOLE_IS_VT220)
144 add_preferred_console("ttyS", 1, NULL);
145 else if (CONSOLE_IS_HVC)
146 add_preferred_console("hvc", 0, NULL);
147}
148
149static int __init conmode_setup(char *str)
150{
151#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
152 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
153 SET_CONSOLE_SCLP;
154#endif
155#if defined(CONFIG_TN3215_CONSOLE)
156 if (strncmp(str, "3215", 5) == 0)
157 SET_CONSOLE_3215;
158#endif
159#if defined(CONFIG_TN3270_CONSOLE)
160 if (strncmp(str, "3270", 5) == 0)
161 SET_CONSOLE_3270;
162#endif
163 set_preferred_console();
164 return 1;
165}
166
167__setup("conmode=", conmode_setup);
168
169static void __init conmode_default(void)
170{
171 char query_buffer[1024];
172 char *ptr;
173
174 if (MACHINE_IS_VM) {
175 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
176 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
177 ptr = strstr(query_buffer, "SUBCHANNEL =");
178 console_irq = simple_strtoul(ptr + 13, NULL, 16);
179 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
180 ptr = strstr(query_buffer, "CONMODE");
181 /*
182 * Set the conmode to 3215 so that the device recognition
183 * will set the cu_type of the console to 3215. If the
184 * conmode is 3270 and we don't set it back then both
185 * 3215 and the 3270 driver will try to access the console
186 * device (3215 as console and 3270 as normal tty).
187 */
188 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
189 if (ptr == NULL) {
190#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
191 SET_CONSOLE_SCLP;
192#endif
193 return;
194 }
195 if (strncmp(ptr + 8, "3270", 4) == 0) {
196#if defined(CONFIG_TN3270_CONSOLE)
197 SET_CONSOLE_3270;
198#elif defined(CONFIG_TN3215_CONSOLE)
199 SET_CONSOLE_3215;
200#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
201 SET_CONSOLE_SCLP;
202#endif
203 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
204#if defined(CONFIG_TN3215_CONSOLE)
205 SET_CONSOLE_3215;
206#elif defined(CONFIG_TN3270_CONSOLE)
207 SET_CONSOLE_3270;
208#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
209 SET_CONSOLE_SCLP;
210#endif
211 }
212 } else if (MACHINE_IS_KVM) {
213 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
214 SET_CONSOLE_VT220;
215 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
216 SET_CONSOLE_SCLP;
217 else
218 SET_CONSOLE_HVC;
219 } else {
220#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
221 SET_CONSOLE_SCLP;
222#endif
223 }
224 if (IS_ENABLED(CONFIG_VT) && IS_ENABLED(CONFIG_DUMMY_CONSOLE))
225 conswitchp = &dummy_con;
226}
227
228#ifdef CONFIG_CRASH_DUMP
229static void __init setup_zfcpdump(void)
230{
231 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
232 return;
233 if (OLDMEM_BASE)
234 return;
235 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
236 console_loglevel = 2;
237}
238#else
239static inline void setup_zfcpdump(void) {}
240#endif /* CONFIG_CRASH_DUMP */
241
242 /*
243 * Reboot, halt and power_off stubs. They just call _machine_restart,
244 * _machine_halt or _machine_power_off.
245 */
246
247void machine_restart(char *command)
248{
249 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
250 /*
251 * Only unblank the console if we are called in enabled
252 * context or a bust_spinlocks cleared the way for us.
253 */
254 console_unblank();
255 _machine_restart(command);
256}
257
258void machine_halt(void)
259{
260 if (!in_interrupt() || oops_in_progress)
261 /*
262 * Only unblank the console if we are called in enabled
263 * context or a bust_spinlocks cleared the way for us.
264 */
265 console_unblank();
266 _machine_halt();
267}
268
269void machine_power_off(void)
270{
271 if (!in_interrupt() || oops_in_progress)
272 /*
273 * Only unblank the console if we are called in enabled
274 * context or a bust_spinlocks cleared the way for us.
275 */
276 console_unblank();
277 _machine_power_off();
278}
279
280/*
281 * Dummy power off function.
282 */
283void (*pm_power_off)(void) = machine_power_off;
284EXPORT_SYMBOL_GPL(pm_power_off);
285
286static int __init early_parse_mem(char *p)
287{
288 memory_end = memparse(p, &p);
289 memory_end &= PAGE_MASK;
290 memory_end_set = 1;
291 return 0;
292}
293early_param("mem", early_parse_mem);
294
295static int __init parse_vmalloc(char *arg)
296{
297 if (!arg)
298 return -EINVAL;
299 VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
300 return 0;
301}
302early_param("vmalloc", parse_vmalloc);
303
304void *restart_stack __section(.data);
305
306static void __init setup_lowcore(void)
307{
308 struct lowcore *lc;
309
310 /*
311 * Setup lowcore for boot cpu
312 */
313 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * PAGE_SIZE);
314 lc = memblock_virt_alloc_low(sizeof(*lc), sizeof(*lc));
315 lc->restart_psw.mask = PSW_KERNEL_BITS;
316 lc->restart_psw.addr = (unsigned long) restart_int_handler;
317 lc->external_new_psw.mask = PSW_KERNEL_BITS |
318 PSW_MASK_DAT | PSW_MASK_MCHECK;
319 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
320 lc->svc_new_psw.mask = PSW_KERNEL_BITS |
321 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
322 lc->svc_new_psw.addr = (unsigned long) system_call;
323 lc->program_new_psw.mask = PSW_KERNEL_BITS |
324 PSW_MASK_DAT | PSW_MASK_MCHECK;
325 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
326 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
327 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
328 lc->io_new_psw.mask = PSW_KERNEL_BITS |
329 PSW_MASK_DAT | PSW_MASK_MCHECK;
330 lc->io_new_psw.addr = (unsigned long) io_int_handler;
331 lc->clock_comparator = clock_comparator_max;
332 lc->kernel_stack = ((unsigned long) &init_thread_union)
333 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
334 lc->async_stack = (unsigned long)
335 memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE)
336 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
337 lc->panic_stack = (unsigned long)
338 memblock_virt_alloc(PAGE_SIZE, PAGE_SIZE)
339 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
340 lc->current_task = (unsigned long)&init_task;
341 lc->lpp = LPP_MAGIC;
342 lc->machine_flags = S390_lowcore.machine_flags;
343 lc->preempt_count = S390_lowcore.preempt_count;
344 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
345 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
346 sizeof(lc->stfle_fac_list));
347 memcpy(lc->alt_stfle_fac_list, S390_lowcore.alt_stfle_fac_list,
348 sizeof(lc->alt_stfle_fac_list));
349 nmi_alloc_boot_cpu(lc);
350 vdso_alloc_boot_cpu(lc);
351 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
352 lc->async_enter_timer = S390_lowcore.async_enter_timer;
353 lc->exit_timer = S390_lowcore.exit_timer;
354 lc->user_timer = S390_lowcore.user_timer;
355 lc->system_timer = S390_lowcore.system_timer;
356 lc->steal_timer = S390_lowcore.steal_timer;
357 lc->last_update_timer = S390_lowcore.last_update_timer;
358 lc->last_update_clock = S390_lowcore.last_update_clock;
359
360 restart_stack = memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE);
361 restart_stack += ASYNC_SIZE;
362
363 /*
364 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
365 * restart data to the absolute zero lowcore. This is necessary if
366 * PSW restart is done on an offline CPU that has lowcore zero.
367 */
368 lc->restart_stack = (unsigned long) restart_stack;
369 lc->restart_fn = (unsigned long) do_restart;
370 lc->restart_data = 0;
371 lc->restart_source = -1UL;
372
373 /* Setup absolute zero lowcore */
374 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
375 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
376 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
377 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
378 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
379
380#ifdef CONFIG_SMP
381 lc->spinlock_lockval = arch_spin_lockval(0);
382 lc->spinlock_index = 0;
383 arch_spin_lock_setup(0);
384#endif
385 lc->br_r1_trampoline = 0x07f1; /* br %r1 */
386
387 set_prefix((u32)(unsigned long) lc);
388 lowcore_ptr[0] = lc;
389}
390
391static struct resource code_resource = {
392 .name = "Kernel code",
393 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
394};
395
396static struct resource data_resource = {
397 .name = "Kernel data",
398 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
399};
400
401static struct resource bss_resource = {
402 .name = "Kernel bss",
403 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
404};
405
406static struct resource __initdata *standard_resources[] = {
407 &code_resource,
408 &data_resource,
409 &bss_resource,
410};
411
412static void __init setup_resources(void)
413{
414 struct resource *res, *std_res, *sub_res;
415 struct memblock_region *reg;
416 int j;
417
418 code_resource.start = (unsigned long) _text;
419 code_resource.end = (unsigned long) _etext - 1;
420 data_resource.start = (unsigned long) _etext;
421 data_resource.end = (unsigned long) _edata - 1;
422 bss_resource.start = (unsigned long) __bss_start;
423 bss_resource.end = (unsigned long) __bss_stop - 1;
424
425 for_each_memblock(memory, reg) {
426 res = memblock_virt_alloc(sizeof(*res), 8);
427 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
428
429 res->name = "System RAM";
430 res->start = reg->base;
431 res->end = reg->base + reg->size - 1;
432 request_resource(&iomem_resource, res);
433
434 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
435 std_res = standard_resources[j];
436 if (std_res->start < res->start ||
437 std_res->start > res->end)
438 continue;
439 if (std_res->end > res->end) {
440 sub_res = memblock_virt_alloc(sizeof(*sub_res), 8);
441 *sub_res = *std_res;
442 sub_res->end = res->end;
443 std_res->start = res->end + 1;
444 request_resource(res, sub_res);
445 } else {
446 request_resource(res, std_res);
447 }
448 }
449 }
450#ifdef CONFIG_CRASH_DUMP
451 /*
452 * Re-add removed crash kernel memory as reserved memory. This makes
453 * sure it will be mapped with the identity mapping and struct pages
454 * will be created, so it can be resized later on.
455 * However add it later since the crash kernel resource should not be
456 * part of the System RAM resource.
457 */
458 if (crashk_res.end) {
459 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
460 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
461 insert_resource(&iomem_resource, &crashk_res);
462 }
463#endif
464}
465
466static void __init setup_memory_end(void)
467{
468 unsigned long vmax, vmalloc_size, tmp;
469
470 /* Choose kernel address space layout: 2, 3, or 4 levels. */
471 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
472 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
473 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
474 if (tmp + vmalloc_size + MODULES_LEN <= _REGION2_SIZE)
475 vmax = _REGION2_SIZE; /* 3-level kernel page table */
476 else
477 vmax = _REGION1_SIZE; /* 4-level kernel page table */
478 /* module area is at the end of the kernel address space. */
479 MODULES_END = vmax;
480 MODULES_VADDR = MODULES_END - MODULES_LEN;
481 VMALLOC_END = MODULES_VADDR;
482 VMALLOC_START = vmax - vmalloc_size;
483
484 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
485 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
486 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
487 tmp = SECTION_ALIGN_UP(tmp);
488 tmp = VMALLOC_START - tmp * sizeof(struct page);
489 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
490 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
491 vmemmap = (struct page *) tmp;
492
493 /* Take care that memory_end is set and <= vmemmap */
494 memory_end = min(memory_end ?: max_physmem_end, tmp);
495 max_pfn = max_low_pfn = PFN_DOWN(memory_end);
496 memblock_remove(memory_end, ULONG_MAX);
497
498 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
499}
500
501#ifdef CONFIG_CRASH_DUMP
502
503/*
504 * When kdump is enabled, we have to ensure that no memory from
505 * the area [0 - crashkernel memory size] and
506 * [crashk_res.start - crashk_res.end] is set offline.
507 */
508static int kdump_mem_notifier(struct notifier_block *nb,
509 unsigned long action, void *data)
510{
511 struct memory_notify *arg = data;
512
513 if (action != MEM_GOING_OFFLINE)
514 return NOTIFY_OK;
515 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
516 return NOTIFY_BAD;
517 if (arg->start_pfn > PFN_DOWN(crashk_res.end))
518 return NOTIFY_OK;
519 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
520 return NOTIFY_OK;
521 return NOTIFY_BAD;
522}
523
524static struct notifier_block kdump_mem_nb = {
525 .notifier_call = kdump_mem_notifier,
526};
527
528#endif
529
530/*
531 * Make sure that the area behind memory_end is protected
532 */
533static void reserve_memory_end(void)
534{
535#ifdef CONFIG_CRASH_DUMP
536 if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
537 !OLDMEM_BASE && sclp.hsa_size) {
538 memory_end = sclp.hsa_size;
539 memory_end &= PAGE_MASK;
540 memory_end_set = 1;
541 }
542#endif
543 if (!memory_end_set)
544 return;
545 memblock_reserve(memory_end, ULONG_MAX);
546}
547
548/*
549 * Make sure that oldmem, where the dump is stored, is protected
550 */
551static void reserve_oldmem(void)
552{
553#ifdef CONFIG_CRASH_DUMP
554 if (OLDMEM_BASE)
555 /* Forget all memory above the running kdump system */
556 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
557#endif
558}
559
560/*
561 * Make sure that oldmem, where the dump is stored, is protected
562 */
563static void remove_oldmem(void)
564{
565#ifdef CONFIG_CRASH_DUMP
566 if (OLDMEM_BASE)
567 /* Forget all memory above the running kdump system */
568 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
569#endif
570}
571
572/*
573 * Reserve memory for kdump kernel to be loaded with kexec
574 */
575static void __init reserve_crashkernel(void)
576{
577#ifdef CONFIG_CRASH_DUMP
578 unsigned long long crash_base, crash_size;
579 phys_addr_t low, high;
580 int rc;
581
582 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
583 &crash_base);
584
585 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
586 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
587 if (rc || crash_size == 0)
588 return;
589
590 if (memblock.memory.regions[0].size < crash_size) {
591 pr_info("crashkernel reservation failed: %s\n",
592 "first memory chunk must be at least crashkernel size");
593 return;
594 }
595
596 low = crash_base ?: OLDMEM_BASE;
597 high = low + crash_size;
598 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
599 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
600 crash_base = low;
601 } else {
602 /* Find suitable area in free memory */
603 low = max_t(unsigned long, crash_size, sclp.hsa_size);
604 high = crash_base ? crash_base + crash_size : ULONG_MAX;
605
606 if (crash_base && crash_base < low) {
607 pr_info("crashkernel reservation failed: %s\n",
608 "crash_base too low");
609 return;
610 }
611 low = crash_base ?: low;
612 crash_base = memblock_find_in_range(low, high, crash_size,
613 KEXEC_CRASH_MEM_ALIGN);
614 }
615
616 if (!crash_base) {
617 pr_info("crashkernel reservation failed: %s\n",
618 "no suitable area found");
619 return;
620 }
621
622 if (register_memory_notifier(&kdump_mem_nb))
623 return;
624
625 if (!OLDMEM_BASE && MACHINE_IS_VM)
626 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
627 crashk_res.start = crash_base;
628 crashk_res.end = crash_base + crash_size - 1;
629 memblock_remove(crash_base, crash_size);
630 pr_info("Reserving %lluMB of memory at %lluMB "
631 "for crashkernel (System RAM: %luMB)\n",
632 crash_size >> 20, crash_base >> 20,
633 (unsigned long)memblock.memory.total_size >> 20);
634 os_info_crashkernel_add(crash_base, crash_size);
635#endif
636}
637
638/*
639 * Reserve the initrd from being used by memblock
640 */
641static void __init reserve_initrd(void)
642{
643#ifdef CONFIG_BLK_DEV_INITRD
644 if (!INITRD_START || !INITRD_SIZE)
645 return;
646 initrd_start = INITRD_START;
647 initrd_end = initrd_start + INITRD_SIZE;
648 memblock_reserve(INITRD_START, INITRD_SIZE);
649#endif
650}
651
652/*
653 * Check for initrd being in usable memory
654 */
655static void __init check_initrd(void)
656{
657#ifdef CONFIG_BLK_DEV_INITRD
658 if (INITRD_START && INITRD_SIZE &&
659 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
660 pr_err("The initial RAM disk does not fit into the memory\n");
661 memblock_free(INITRD_START, INITRD_SIZE);
662 initrd_start = initrd_end = 0;
663 }
664#endif
665}
666
667/*
668 * Reserve memory used for lowcore/command line/kernel image.
669 */
670static void __init reserve_kernel(void)
671{
672 unsigned long start_pfn = PFN_UP(__pa(_end));
673
674#ifdef CONFIG_DMA_API_DEBUG
675 /*
676 * DMA_API_DEBUG code stumbles over addresses from the
677 * range [_ehead, _stext]. Mark the memory as reserved
678 * so it is not used for CONFIG_DMA_API_DEBUG=y.
679 */
680 memblock_reserve(0, PFN_PHYS(start_pfn));
681#else
682 memblock_reserve(0, (unsigned long)_ehead);
683 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
684 - (unsigned long)_stext);
685#endif
686}
687
688static void __init setup_memory(void)
689{
690 struct memblock_region *reg;
691
692 /*
693 * Init storage key for present memory
694 */
695 for_each_memblock(memory, reg) {
696 storage_key_init_range(reg->base, reg->base + reg->size);
697 }
698 psw_set_key(PAGE_DEFAULT_KEY);
699
700 /* Only cosmetics */
701 memblock_enforce_memory_limit(memblock_end_of_DRAM());
702}
703
704/*
705 * Setup hardware capabilities.
706 */
707static int __init setup_hwcaps(void)
708{
709 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
710 struct cpuid cpu_id;
711 int i;
712
713 /*
714 * The store facility list bits numbers as found in the principles
715 * of operation are numbered with bit 1UL<<31 as number 0 to
716 * bit 1UL<<0 as number 31.
717 * Bit 0: instructions named N3, "backported" to esa-mode
718 * Bit 2: z/Architecture mode is active
719 * Bit 7: the store-facility-list-extended facility is installed
720 * Bit 17: the message-security assist is installed
721 * Bit 19: the long-displacement facility is installed
722 * Bit 21: the extended-immediate facility is installed
723 * Bit 22: extended-translation facility 3 is installed
724 * Bit 30: extended-translation facility 3 enhancement facility
725 * These get translated to:
726 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
727 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
728 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
729 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
730 */
731 for (i = 0; i < 6; i++)
732 if (test_facility(stfl_bits[i]))
733 elf_hwcap |= 1UL << i;
734
735 if (test_facility(22) && test_facility(30))
736 elf_hwcap |= HWCAP_S390_ETF3EH;
737
738 /*
739 * Check for additional facilities with store-facility-list-extended.
740 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
741 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
742 * as stored by stfl, bits 32-xxx contain additional facilities.
743 * How many facility words are stored depends on the number of
744 * doublewords passed to the instruction. The additional facilities
745 * are:
746 * Bit 42: decimal floating point facility is installed
747 * Bit 44: perform floating point operation facility is installed
748 * translated to:
749 * HWCAP_S390_DFP bit 6 (42 && 44).
750 */
751 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
752 elf_hwcap |= HWCAP_S390_DFP;
753
754 /*
755 * Huge page support HWCAP_S390_HPAGE is bit 7.
756 */
757 if (MACHINE_HAS_EDAT1)
758 elf_hwcap |= HWCAP_S390_HPAGE;
759
760 /*
761 * 64-bit register support for 31-bit processes
762 * HWCAP_S390_HIGH_GPRS is bit 9.
763 */
764 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
765
766 /*
767 * Transactional execution support HWCAP_S390_TE is bit 10.
768 */
769 if (MACHINE_HAS_TE)
770 elf_hwcap |= HWCAP_S390_TE;
771
772 /*
773 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
774 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
775 * instead of facility bit 129.
776 */
777 if (MACHINE_HAS_VX) {
778 elf_hwcap |= HWCAP_S390_VXRS;
779 if (test_facility(134))
780 elf_hwcap |= HWCAP_S390_VXRS_EXT;
781 if (test_facility(135))
782 elf_hwcap |= HWCAP_S390_VXRS_BCD;
783 }
784
785 /*
786 * Guarded storage support HWCAP_S390_GS is bit 12.
787 */
788 if (MACHINE_HAS_GS)
789 elf_hwcap |= HWCAP_S390_GS;
790
791 get_cpu_id(&cpu_id);
792 add_device_randomness(&cpu_id, sizeof(cpu_id));
793 switch (cpu_id.machine) {
794 case 0x2064:
795 case 0x2066:
796 default: /* Use "z900" as default for 64 bit kernels. */
797 strcpy(elf_platform, "z900");
798 break;
799 case 0x2084:
800 case 0x2086:
801 strcpy(elf_platform, "z990");
802 break;
803 case 0x2094:
804 case 0x2096:
805 strcpy(elf_platform, "z9-109");
806 break;
807 case 0x2097:
808 case 0x2098:
809 strcpy(elf_platform, "z10");
810 break;
811 case 0x2817:
812 case 0x2818:
813 strcpy(elf_platform, "z196");
814 break;
815 case 0x2827:
816 case 0x2828:
817 strcpy(elf_platform, "zEC12");
818 break;
819 case 0x2964:
820 case 0x2965:
821 strcpy(elf_platform, "z13");
822 break;
823 case 0x3906:
824 case 0x3907:
825 strcpy(elf_platform, "z14");
826 break;
827 }
828
829 /*
830 * Virtualization support HWCAP_INT_SIE is bit 0.
831 */
832 if (sclp.has_sief2)
833 int_hwcap |= HWCAP_INT_SIE;
834
835 return 0;
836}
837arch_initcall(setup_hwcaps);
838
839/*
840 * Add system information as device randomness
841 */
842static void __init setup_randomness(void)
843{
844 struct sysinfo_3_2_2 *vmms;
845
846 vmms = (struct sysinfo_3_2_2 *) memblock_alloc(PAGE_SIZE, PAGE_SIZE);
847 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
848 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
849 memblock_free((unsigned long) vmms, PAGE_SIZE);
850}
851
852/*
853 * Find the correct size for the task_struct. This depends on
854 * the size of the struct fpu at the end of the thread_struct
855 * which is embedded in the task_struct.
856 */
857static void __init setup_task_size(void)
858{
859 int task_size = sizeof(struct task_struct);
860
861 if (!MACHINE_HAS_VX) {
862 task_size -= sizeof(__vector128) * __NUM_VXRS;
863 task_size += sizeof(freg_t) * __NUM_FPRS;
864 }
865 arch_task_struct_size = task_size;
866}
867
868/*
869 * Setup function called from init/main.c just after the banner
870 * was printed.
871 */
872
873void __init setup_arch(char **cmdline_p)
874{
875 /*
876 * print what head.S has found out about the machine
877 */
878 if (MACHINE_IS_VM)
879 pr_info("Linux is running as a z/VM "
880 "guest operating system in 64-bit mode\n");
881 else if (MACHINE_IS_KVM)
882 pr_info("Linux is running under KVM in 64-bit mode\n");
883 else if (MACHINE_IS_LPAR)
884 pr_info("Linux is running natively in 64-bit mode\n");
885
886 /* Have one command line that is parsed and saved in /proc/cmdline */
887 /* boot_command_line has been already set up in early.c */
888 *cmdline_p = boot_command_line;
889
890 ROOT_DEV = Root_RAM0;
891
892 /* Is init_mm really needed? */
893 init_mm.start_code = PAGE_OFFSET;
894 init_mm.end_code = (unsigned long) _etext;
895 init_mm.end_data = (unsigned long) _edata;
896 init_mm.brk = (unsigned long) _end;
897
898 if (IS_ENABLED(CONFIG_EXPOLINE_AUTO))
899 nospec_auto_detect();
900
901 parse_early_param();
902#ifdef CONFIG_CRASH_DUMP
903 /* Deactivate elfcorehdr= kernel parameter */
904 elfcorehdr_addr = ELFCORE_ADDR_MAX;
905#endif
906
907 os_info_init();
908 setup_ipl();
909 setup_task_size();
910
911 /* Do some memory reservations *before* memory is added to memblock */
912 reserve_memory_end();
913 reserve_oldmem();
914 reserve_kernel();
915 reserve_initrd();
916 memblock_allow_resize();
917
918 /* Get information about *all* installed memory */
919 detect_memory_memblock();
920
921 remove_oldmem();
922
923 /*
924 * Make sure all chunks are MAX_ORDER aligned so we don't need the
925 * extra checks that HOLES_IN_ZONE would require.
926 *
927 * Is this still required?
928 */
929 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
930
931 setup_memory_end();
932 setup_memory();
933 dma_contiguous_reserve(memory_end);
934 vmcp_cma_reserve();
935
936 check_initrd();
937 reserve_crashkernel();
938#ifdef CONFIG_CRASH_DUMP
939 /*
940 * Be aware that smp_save_dump_cpus() triggers a system reset.
941 * Therefore CPU and device initialization should be done afterwards.
942 */
943 smp_save_dump_cpus();
944#endif
945
946 setup_resources();
947 setup_lowcore();
948 smp_fill_possible_mask();
949 cpu_detect_mhz_feature();
950 cpu_init();
951 numa_setup();
952 smp_detect_cpus();
953 topology_init_early();
954
955 /*
956 * Create kernel page tables and switch to virtual addressing.
957 */
958 paging_init();
959
960 /* Setup default console */
961 conmode_default();
962 set_preferred_console();
963
964 apply_alternative_instructions();
965 if (IS_ENABLED(CONFIG_EXPOLINE))
966 nospec_init_branches();
967
968 /* Setup zfcpdump support */
969 setup_zfcpdump();
970
971 /* Add system specific data to the random pool */
972 setup_randomness();
973}
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/dma-contiguous.h>
39#include <linux/device.h>
40#include <linux/notifier.h>
41#include <linux/pfn.h>
42#include <linux/ctype.h>
43#include <linux/reboot.h>
44#include <linux/topology.h>
45#include <linux/kexec.h>
46#include <linux/crash_dump.h>
47#include <linux/memory.h>
48#include <linux/compat.h>
49
50#include <asm/ipl.h>
51#include <asm/facility.h>
52#include <asm/smp.h>
53#include <asm/mmu_context.h>
54#include <asm/cpcmd.h>
55#include <asm/lowcore.h>
56#include <asm/irq.h>
57#include <asm/page.h>
58#include <asm/ptrace.h>
59#include <asm/sections.h>
60#include <asm/ebcdic.h>
61#include <asm/kvm_virtio.h>
62#include <asm/diag.h>
63#include <asm/os_info.h>
64#include <asm/sclp.h>
65#include <asm/sysinfo.h>
66#include <asm/numa.h>
67#include "entry.h"
68
69/*
70 * Machine setup..
71 */
72unsigned int console_mode = 0;
73EXPORT_SYMBOL(console_mode);
74
75unsigned int console_devno = -1;
76EXPORT_SYMBOL(console_devno);
77
78unsigned int console_irq = -1;
79EXPORT_SYMBOL(console_irq);
80
81unsigned long elf_hwcap __read_mostly = 0;
82char elf_platform[ELF_PLATFORM_SIZE];
83
84unsigned long int_hwcap = 0;
85
86int __initdata memory_end_set;
87unsigned long __initdata memory_end;
88unsigned long __initdata max_physmem_end;
89
90unsigned long VMALLOC_START;
91EXPORT_SYMBOL(VMALLOC_START);
92
93unsigned long VMALLOC_END;
94EXPORT_SYMBOL(VMALLOC_END);
95
96struct page *vmemmap;
97EXPORT_SYMBOL(vmemmap);
98
99unsigned long MODULES_VADDR;
100unsigned long MODULES_END;
101
102/* An array with a pointer to the lowcore of every CPU. */
103struct lowcore *lowcore_ptr[NR_CPUS];
104EXPORT_SYMBOL(lowcore_ptr);
105
106/*
107 * This is set up by the setup-routine at boot-time
108 * for S390 need to find out, what we have to setup
109 * using address 0x10400 ...
110 */
111
112#include <asm/setup.h>
113
114/*
115 * condev= and conmode= setup parameter.
116 */
117
118static int __init condev_setup(char *str)
119{
120 int vdev;
121
122 vdev = simple_strtoul(str, &str, 0);
123 if (vdev >= 0 && vdev < 65536) {
124 console_devno = vdev;
125 console_irq = -1;
126 }
127 return 1;
128}
129
130__setup("condev=", condev_setup);
131
132static void __init set_preferred_console(void)
133{
134 if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
135 add_preferred_console("ttyS", 0, NULL);
136 else if (CONSOLE_IS_3270)
137 add_preferred_console("tty3270", 0, NULL);
138 else if (CONSOLE_IS_VT220)
139 add_preferred_console("ttyS", 1, NULL);
140 else if (CONSOLE_IS_HVC)
141 add_preferred_console("hvc", 0, NULL);
142}
143
144static int __init conmode_setup(char *str)
145{
146#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
147 if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
148 SET_CONSOLE_SCLP;
149#endif
150#if defined(CONFIG_TN3215_CONSOLE)
151 if (strncmp(str, "3215", 5) == 0)
152 SET_CONSOLE_3215;
153#endif
154#if defined(CONFIG_TN3270_CONSOLE)
155 if (strncmp(str, "3270", 5) == 0)
156 SET_CONSOLE_3270;
157#endif
158 set_preferred_console();
159 return 1;
160}
161
162__setup("conmode=", conmode_setup);
163
164static void __init conmode_default(void)
165{
166 char query_buffer[1024];
167 char *ptr;
168
169 if (MACHINE_IS_VM) {
170 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
171 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
172 ptr = strstr(query_buffer, "SUBCHANNEL =");
173 console_irq = simple_strtoul(ptr + 13, NULL, 16);
174 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
175 ptr = strstr(query_buffer, "CONMODE");
176 /*
177 * Set the conmode to 3215 so that the device recognition
178 * will set the cu_type of the console to 3215. If the
179 * conmode is 3270 and we don't set it back then both
180 * 3215 and the 3270 driver will try to access the console
181 * device (3215 as console and 3270 as normal tty).
182 */
183 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
184 if (ptr == NULL) {
185#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
186 SET_CONSOLE_SCLP;
187#endif
188 return;
189 }
190 if (strncmp(ptr + 8, "3270", 4) == 0) {
191#if defined(CONFIG_TN3270_CONSOLE)
192 SET_CONSOLE_3270;
193#elif defined(CONFIG_TN3215_CONSOLE)
194 SET_CONSOLE_3215;
195#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
196 SET_CONSOLE_SCLP;
197#endif
198 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
199#if defined(CONFIG_TN3215_CONSOLE)
200 SET_CONSOLE_3215;
201#elif defined(CONFIG_TN3270_CONSOLE)
202 SET_CONSOLE_3270;
203#elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
204 SET_CONSOLE_SCLP;
205#endif
206 }
207 } else if (MACHINE_IS_KVM) {
208 if (sclp.has_vt220 && IS_ENABLED(CONFIG_SCLP_VT220_CONSOLE))
209 SET_CONSOLE_VT220;
210 else if (sclp.has_linemode && IS_ENABLED(CONFIG_SCLP_CONSOLE))
211 SET_CONSOLE_SCLP;
212 else
213 SET_CONSOLE_HVC;
214 } else {
215#if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
216 SET_CONSOLE_SCLP;
217#endif
218 }
219}
220
221#ifdef CONFIG_CRASH_DUMP
222static void __init setup_zfcpdump(void)
223{
224 if (ipl_info.type != IPL_TYPE_FCP_DUMP)
225 return;
226 if (OLDMEM_BASE)
227 return;
228 strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
229 console_loglevel = 2;
230}
231#else
232static inline void setup_zfcpdump(void) {}
233#endif /* CONFIG_CRASH_DUMP */
234
235 /*
236 * Reboot, halt and power_off stubs. They just call _machine_restart,
237 * _machine_halt or _machine_power_off.
238 */
239
240void machine_restart(char *command)
241{
242 if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
243 /*
244 * Only unblank the console if we are called in enabled
245 * context or a bust_spinlocks cleared the way for us.
246 */
247 console_unblank();
248 _machine_restart(command);
249}
250
251void machine_halt(void)
252{
253 if (!in_interrupt() || oops_in_progress)
254 /*
255 * Only unblank the console if we are called in enabled
256 * context or a bust_spinlocks cleared the way for us.
257 */
258 console_unblank();
259 _machine_halt();
260}
261
262void machine_power_off(void)
263{
264 if (!in_interrupt() || oops_in_progress)
265 /*
266 * Only unblank the console if we are called in enabled
267 * context or a bust_spinlocks cleared the way for us.
268 */
269 console_unblank();
270 _machine_power_off();
271}
272
273/*
274 * Dummy power off function.
275 */
276void (*pm_power_off)(void) = machine_power_off;
277EXPORT_SYMBOL_GPL(pm_power_off);
278
279static int __init early_parse_mem(char *p)
280{
281 memory_end = memparse(p, &p);
282 memory_end &= PAGE_MASK;
283 memory_end_set = 1;
284 return 0;
285}
286early_param("mem", early_parse_mem);
287
288static int __init parse_vmalloc(char *arg)
289{
290 if (!arg)
291 return -EINVAL;
292 VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
293 return 0;
294}
295early_param("vmalloc", parse_vmalloc);
296
297void *restart_stack __section(.data);
298
299static void __init setup_lowcore(void)
300{
301 struct lowcore *lc;
302
303 /*
304 * Setup lowcore for boot cpu
305 */
306 BUILD_BUG_ON(sizeof(struct lowcore) != LC_PAGES * 4096);
307 lc = memblock_virt_alloc_low(sizeof(*lc), sizeof(*lc));
308 lc->restart_psw.mask = PSW_KERNEL_BITS;
309 lc->restart_psw.addr = (unsigned long) restart_int_handler;
310 lc->external_new_psw.mask = PSW_KERNEL_BITS |
311 PSW_MASK_DAT | PSW_MASK_MCHECK;
312 lc->external_new_psw.addr = (unsigned long) ext_int_handler;
313 lc->svc_new_psw.mask = PSW_KERNEL_BITS |
314 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
315 lc->svc_new_psw.addr = (unsigned long) system_call;
316 lc->program_new_psw.mask = PSW_KERNEL_BITS |
317 PSW_MASK_DAT | PSW_MASK_MCHECK;
318 lc->program_new_psw.addr = (unsigned long) pgm_check_handler;
319 lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
320 lc->mcck_new_psw.addr = (unsigned long) mcck_int_handler;
321 lc->io_new_psw.mask = PSW_KERNEL_BITS |
322 PSW_MASK_DAT | PSW_MASK_MCHECK;
323 lc->io_new_psw.addr = (unsigned long) io_int_handler;
324 lc->clock_comparator = -1ULL;
325 lc->kernel_stack = ((unsigned long) &init_thread_union)
326 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
327 lc->async_stack = (unsigned long)
328 memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE)
329 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
330 lc->panic_stack = (unsigned long)
331 memblock_virt_alloc(PAGE_SIZE, PAGE_SIZE)
332 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
333 lc->current_task = (unsigned long)&init_task;
334 lc->lpp = LPP_MAGIC;
335 lc->machine_flags = S390_lowcore.machine_flags;
336 lc->preempt_count = S390_lowcore.preempt_count;
337 lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
338 memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
339 MAX_FACILITY_BIT/8);
340 if (MACHINE_HAS_VX)
341 lc->vector_save_area_addr =
342 (unsigned long) &lc->vector_save_area;
343 lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
344 lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
345 lc->async_enter_timer = S390_lowcore.async_enter_timer;
346 lc->exit_timer = S390_lowcore.exit_timer;
347 lc->user_timer = S390_lowcore.user_timer;
348 lc->system_timer = S390_lowcore.system_timer;
349 lc->steal_timer = S390_lowcore.steal_timer;
350 lc->last_update_timer = S390_lowcore.last_update_timer;
351 lc->last_update_clock = S390_lowcore.last_update_clock;
352
353 restart_stack = memblock_virt_alloc(ASYNC_SIZE, ASYNC_SIZE);
354 restart_stack += ASYNC_SIZE;
355
356 /*
357 * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
358 * restart data to the absolute zero lowcore. This is necessary if
359 * PSW restart is done on an offline CPU that has lowcore zero.
360 */
361 lc->restart_stack = (unsigned long) restart_stack;
362 lc->restart_fn = (unsigned long) do_restart;
363 lc->restart_data = 0;
364 lc->restart_source = -1UL;
365
366 /* Setup absolute zero lowcore */
367 mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
368 mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
369 mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
370 mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
371 mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
372
373#ifdef CONFIG_SMP
374 lc->spinlock_lockval = arch_spin_lockval(0);
375#endif
376
377 set_prefix((u32)(unsigned long) lc);
378 lowcore_ptr[0] = lc;
379}
380
381static struct resource code_resource = {
382 .name = "Kernel code",
383 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
384};
385
386static struct resource data_resource = {
387 .name = "Kernel data",
388 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
389};
390
391static struct resource bss_resource = {
392 .name = "Kernel bss",
393 .flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM,
394};
395
396static struct resource __initdata *standard_resources[] = {
397 &code_resource,
398 &data_resource,
399 &bss_resource,
400};
401
402static void __init setup_resources(void)
403{
404 struct resource *res, *std_res, *sub_res;
405 struct memblock_region *reg;
406 int j;
407
408 code_resource.start = (unsigned long) &_text;
409 code_resource.end = (unsigned long) &_etext - 1;
410 data_resource.start = (unsigned long) &_etext;
411 data_resource.end = (unsigned long) &_edata - 1;
412 bss_resource.start = (unsigned long) &__bss_start;
413 bss_resource.end = (unsigned long) &__bss_stop - 1;
414
415 for_each_memblock(memory, reg) {
416 res = memblock_virt_alloc(sizeof(*res), 8);
417 res->flags = IORESOURCE_BUSY | IORESOURCE_SYSTEM_RAM;
418
419 res->name = "System RAM";
420 res->start = reg->base;
421 res->end = reg->base + reg->size - 1;
422 request_resource(&iomem_resource, res);
423
424 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
425 std_res = standard_resources[j];
426 if (std_res->start < res->start ||
427 std_res->start > res->end)
428 continue;
429 if (std_res->end > res->end) {
430 sub_res = memblock_virt_alloc(sizeof(*sub_res), 8);
431 *sub_res = *std_res;
432 sub_res->end = res->end;
433 std_res->start = res->end + 1;
434 request_resource(res, sub_res);
435 } else {
436 request_resource(res, std_res);
437 }
438 }
439 }
440#ifdef CONFIG_CRASH_DUMP
441 /*
442 * Re-add removed crash kernel memory as reserved memory. This makes
443 * sure it will be mapped with the identity mapping and struct pages
444 * will be created, so it can be resized later on.
445 * However add it later since the crash kernel resource should not be
446 * part of the System RAM resource.
447 */
448 if (crashk_res.end) {
449 memblock_add_node(crashk_res.start, resource_size(&crashk_res), 0);
450 memblock_reserve(crashk_res.start, resource_size(&crashk_res));
451 insert_resource(&iomem_resource, &crashk_res);
452 }
453#endif
454}
455
456static void __init setup_memory_end(void)
457{
458 unsigned long vmax, vmalloc_size, tmp;
459
460 /* Choose kernel address space layout: 2, 3, or 4 levels. */
461 vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
462 tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
463 tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
464 if (tmp + vmalloc_size + MODULES_LEN <= (1UL << 42))
465 vmax = 1UL << 42; /* 3-level kernel page table */
466 else
467 vmax = 1UL << 53; /* 4-level kernel page table */
468 /* module area is at the end of the kernel address space. */
469 MODULES_END = vmax;
470 MODULES_VADDR = MODULES_END - MODULES_LEN;
471 VMALLOC_END = MODULES_VADDR;
472 VMALLOC_START = vmax - vmalloc_size;
473
474 /* Split remaining virtual space between 1:1 mapping & vmemmap array */
475 tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
476 /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
477 tmp = SECTION_ALIGN_UP(tmp);
478 tmp = VMALLOC_START - tmp * sizeof(struct page);
479 tmp &= ~((vmax >> 11) - 1); /* align to page table level */
480 tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
481 vmemmap = (struct page *) tmp;
482
483 /* Take care that memory_end is set and <= vmemmap */
484 memory_end = min(memory_end ?: max_physmem_end, tmp);
485 max_pfn = max_low_pfn = PFN_DOWN(memory_end);
486 memblock_remove(memory_end, ULONG_MAX);
487
488 pr_notice("The maximum memory size is %luMB\n", memory_end >> 20);
489}
490
491static void __init setup_vmcoreinfo(void)
492{
493 mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
494}
495
496#ifdef CONFIG_CRASH_DUMP
497
498/*
499 * When kdump is enabled, we have to ensure that no memory from
500 * the area [0 - crashkernel memory size] and
501 * [crashk_res.start - crashk_res.end] is set offline.
502 */
503static int kdump_mem_notifier(struct notifier_block *nb,
504 unsigned long action, void *data)
505{
506 struct memory_notify *arg = data;
507
508 if (action != MEM_GOING_OFFLINE)
509 return NOTIFY_OK;
510 if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
511 return NOTIFY_BAD;
512 if (arg->start_pfn > PFN_DOWN(crashk_res.end))
513 return NOTIFY_OK;
514 if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
515 return NOTIFY_OK;
516 return NOTIFY_BAD;
517}
518
519static struct notifier_block kdump_mem_nb = {
520 .notifier_call = kdump_mem_notifier,
521};
522
523#endif
524
525/*
526 * Make sure that the area behind memory_end is protected
527 */
528static void reserve_memory_end(void)
529{
530#ifdef CONFIG_CRASH_DUMP
531 if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
532 !OLDMEM_BASE && sclp.hsa_size) {
533 memory_end = sclp.hsa_size;
534 memory_end &= PAGE_MASK;
535 memory_end_set = 1;
536 }
537#endif
538 if (!memory_end_set)
539 return;
540 memblock_reserve(memory_end, ULONG_MAX);
541}
542
543/*
544 * Make sure that oldmem, where the dump is stored, is protected
545 */
546static void reserve_oldmem(void)
547{
548#ifdef CONFIG_CRASH_DUMP
549 if (OLDMEM_BASE)
550 /* Forget all memory above the running kdump system */
551 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
552#endif
553}
554
555/*
556 * Make sure that oldmem, where the dump is stored, is protected
557 */
558static void remove_oldmem(void)
559{
560#ifdef CONFIG_CRASH_DUMP
561 if (OLDMEM_BASE)
562 /* Forget all memory above the running kdump system */
563 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
564#endif
565}
566
567/*
568 * Reserve memory for kdump kernel to be loaded with kexec
569 */
570static void __init reserve_crashkernel(void)
571{
572#ifdef CONFIG_CRASH_DUMP
573 unsigned long long crash_base, crash_size;
574 phys_addr_t low, high;
575 int rc;
576
577 rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
578 &crash_base);
579
580 crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
581 crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
582 if (rc || crash_size == 0)
583 return;
584
585 if (memblock.memory.regions[0].size < crash_size) {
586 pr_info("crashkernel reservation failed: %s\n",
587 "first memory chunk must be at least crashkernel size");
588 return;
589 }
590
591 low = crash_base ?: OLDMEM_BASE;
592 high = low + crash_size;
593 if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
594 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
595 crash_base = low;
596 } else {
597 /* Find suitable area in free memory */
598 low = max_t(unsigned long, crash_size, sclp.hsa_size);
599 high = crash_base ? crash_base + crash_size : ULONG_MAX;
600
601 if (crash_base && crash_base < low) {
602 pr_info("crashkernel reservation failed: %s\n",
603 "crash_base too low");
604 return;
605 }
606 low = crash_base ?: low;
607 crash_base = memblock_find_in_range(low, high, crash_size,
608 KEXEC_CRASH_MEM_ALIGN);
609 }
610
611 if (!crash_base) {
612 pr_info("crashkernel reservation failed: %s\n",
613 "no suitable area found");
614 return;
615 }
616
617 if (register_memory_notifier(&kdump_mem_nb))
618 return;
619
620 if (!OLDMEM_BASE && MACHINE_IS_VM)
621 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
622 crashk_res.start = crash_base;
623 crashk_res.end = crash_base + crash_size - 1;
624 memblock_remove(crash_base, crash_size);
625 pr_info("Reserving %lluMB of memory at %lluMB "
626 "for crashkernel (System RAM: %luMB)\n",
627 crash_size >> 20, crash_base >> 20,
628 (unsigned long)memblock.memory.total_size >> 20);
629 os_info_crashkernel_add(crash_base, crash_size);
630#endif
631}
632
633/*
634 * Reserve the initrd from being used by memblock
635 */
636static void __init reserve_initrd(void)
637{
638#ifdef CONFIG_BLK_DEV_INITRD
639 initrd_start = INITRD_START;
640 initrd_end = initrd_start + INITRD_SIZE;
641 memblock_reserve(INITRD_START, INITRD_SIZE);
642#endif
643}
644
645/*
646 * Check for initrd being in usable memory
647 */
648static void __init check_initrd(void)
649{
650#ifdef CONFIG_BLK_DEV_INITRD
651 if (INITRD_START && INITRD_SIZE &&
652 !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
653 pr_err("The initial RAM disk does not fit into the memory\n");
654 memblock_free(INITRD_START, INITRD_SIZE);
655 initrd_start = initrd_end = 0;
656 }
657#endif
658}
659
660/*
661 * Reserve memory used for lowcore/command line/kernel image.
662 */
663static void __init reserve_kernel(void)
664{
665 unsigned long start_pfn = PFN_UP(__pa(&_end));
666
667#ifdef CONFIG_DMA_API_DEBUG
668 /*
669 * DMA_API_DEBUG code stumbles over addresses from the
670 * range [_ehead, _stext]. Mark the memory as reserved
671 * so it is not used for CONFIG_DMA_API_DEBUG=y.
672 */
673 memblock_reserve(0, PFN_PHYS(start_pfn));
674#else
675 memblock_reserve(0, (unsigned long)_ehead);
676 memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
677 - (unsigned long)_stext);
678#endif
679}
680
681static void __init setup_memory(void)
682{
683 struct memblock_region *reg;
684
685 /*
686 * Init storage key for present memory
687 */
688 for_each_memblock(memory, reg) {
689 storage_key_init_range(reg->base, reg->base + reg->size);
690 }
691 psw_set_key(PAGE_DEFAULT_KEY);
692
693 /* Only cosmetics */
694 memblock_enforce_memory_limit(memblock_end_of_DRAM());
695}
696
697/*
698 * Setup hardware capabilities.
699 */
700static int __init setup_hwcaps(void)
701{
702 static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
703 struct cpuid cpu_id;
704 int i;
705
706 /*
707 * The store facility list bits numbers as found in the principles
708 * of operation are numbered with bit 1UL<<31 as number 0 to
709 * bit 1UL<<0 as number 31.
710 * Bit 0: instructions named N3, "backported" to esa-mode
711 * Bit 2: z/Architecture mode is active
712 * Bit 7: the store-facility-list-extended facility is installed
713 * Bit 17: the message-security assist is installed
714 * Bit 19: the long-displacement facility is installed
715 * Bit 21: the extended-immediate facility is installed
716 * Bit 22: extended-translation facility 3 is installed
717 * Bit 30: extended-translation facility 3 enhancement facility
718 * These get translated to:
719 * HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
720 * HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
721 * HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
722 * HWCAP_S390_ETF3EH bit 8 (22 && 30).
723 */
724 for (i = 0; i < 6; i++)
725 if (test_facility(stfl_bits[i]))
726 elf_hwcap |= 1UL << i;
727
728 if (test_facility(22) && test_facility(30))
729 elf_hwcap |= HWCAP_S390_ETF3EH;
730
731 /*
732 * Check for additional facilities with store-facility-list-extended.
733 * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
734 * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
735 * as stored by stfl, bits 32-xxx contain additional facilities.
736 * How many facility words are stored depends on the number of
737 * doublewords passed to the instruction. The additional facilities
738 * are:
739 * Bit 42: decimal floating point facility is installed
740 * Bit 44: perform floating point operation facility is installed
741 * translated to:
742 * HWCAP_S390_DFP bit 6 (42 && 44).
743 */
744 if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
745 elf_hwcap |= HWCAP_S390_DFP;
746
747 /*
748 * Huge page support HWCAP_S390_HPAGE is bit 7.
749 */
750 if (MACHINE_HAS_HPAGE)
751 elf_hwcap |= HWCAP_S390_HPAGE;
752
753 /*
754 * 64-bit register support for 31-bit processes
755 * HWCAP_S390_HIGH_GPRS is bit 9.
756 */
757 elf_hwcap |= HWCAP_S390_HIGH_GPRS;
758
759 /*
760 * Transactional execution support HWCAP_S390_TE is bit 10.
761 */
762 if (test_facility(50) && test_facility(73))
763 elf_hwcap |= HWCAP_S390_TE;
764
765 /*
766 * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
767 * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
768 * instead of facility bit 129.
769 */
770 if (MACHINE_HAS_VX)
771 elf_hwcap |= HWCAP_S390_VXRS;
772 get_cpu_id(&cpu_id);
773 add_device_randomness(&cpu_id, sizeof(cpu_id));
774 switch (cpu_id.machine) {
775 case 0x2064:
776 case 0x2066:
777 default: /* Use "z900" as default for 64 bit kernels. */
778 strcpy(elf_platform, "z900");
779 break;
780 case 0x2084:
781 case 0x2086:
782 strcpy(elf_platform, "z990");
783 break;
784 case 0x2094:
785 case 0x2096:
786 strcpy(elf_platform, "z9-109");
787 break;
788 case 0x2097:
789 case 0x2098:
790 strcpy(elf_platform, "z10");
791 break;
792 case 0x2817:
793 case 0x2818:
794 strcpy(elf_platform, "z196");
795 break;
796 case 0x2827:
797 case 0x2828:
798 strcpy(elf_platform, "zEC12");
799 break;
800 case 0x2964:
801 case 0x2965:
802 strcpy(elf_platform, "z13");
803 break;
804 }
805
806 /*
807 * Virtualization support HWCAP_INT_SIE is bit 0.
808 */
809 if (sclp.has_sief2)
810 int_hwcap |= HWCAP_INT_SIE;
811
812 return 0;
813}
814arch_initcall(setup_hwcaps);
815
816/*
817 * Add system information as device randomness
818 */
819static void __init setup_randomness(void)
820{
821 struct sysinfo_3_2_2 *vmms;
822
823 vmms = (struct sysinfo_3_2_2 *) memblock_alloc(PAGE_SIZE, PAGE_SIZE);
824 if (stsi(vmms, 3, 2, 2) == 0 && vmms->count)
825 add_device_randomness(&vmms->vm, sizeof(vmms->vm[0]) * vmms->count);
826 memblock_free((unsigned long) vmms, PAGE_SIZE);
827}
828
829/*
830 * Find the correct size for the task_struct. This depends on
831 * the size of the struct fpu at the end of the thread_struct
832 * which is embedded in the task_struct.
833 */
834static void __init setup_task_size(void)
835{
836 int task_size = sizeof(struct task_struct);
837
838 if (!MACHINE_HAS_VX) {
839 task_size -= sizeof(__vector128) * __NUM_VXRS;
840 task_size += sizeof(freg_t) * __NUM_FPRS;
841 }
842 arch_task_struct_size = task_size;
843}
844
845/*
846 * Setup function called from init/main.c just after the banner
847 * was printed.
848 */
849
850void __init setup_arch(char **cmdline_p)
851{
852 /*
853 * print what head.S has found out about the machine
854 */
855 if (MACHINE_IS_VM)
856 pr_info("Linux is running as a z/VM "
857 "guest operating system in 64-bit mode\n");
858 else if (MACHINE_IS_KVM)
859 pr_info("Linux is running under KVM in 64-bit mode\n");
860 else if (MACHINE_IS_LPAR)
861 pr_info("Linux is running natively in 64-bit mode\n");
862
863 /* Have one command line that is parsed and saved in /proc/cmdline */
864 /* boot_command_line has been already set up in early.c */
865 *cmdline_p = boot_command_line;
866
867 ROOT_DEV = Root_RAM0;
868
869 /* Is init_mm really needed? */
870 init_mm.start_code = PAGE_OFFSET;
871 init_mm.end_code = (unsigned long) &_etext;
872 init_mm.end_data = (unsigned long) &_edata;
873 init_mm.brk = (unsigned long) &_end;
874
875 parse_early_param();
876#ifdef CONFIG_CRASH_DUMP
877 /* Deactivate elfcorehdr= kernel parameter */
878 elfcorehdr_addr = ELFCORE_ADDR_MAX;
879#endif
880
881 os_info_init();
882 setup_ipl();
883 setup_task_size();
884
885 /* Do some memory reservations *before* memory is added to memblock */
886 reserve_memory_end();
887 reserve_oldmem();
888 reserve_kernel();
889 reserve_initrd();
890 memblock_allow_resize();
891
892 /* Get information about *all* installed memory */
893 detect_memory_memblock();
894
895 remove_oldmem();
896
897 /*
898 * Make sure all chunks are MAX_ORDER aligned so we don't need the
899 * extra checks that HOLES_IN_ZONE would require.
900 *
901 * Is this still required?
902 */
903 memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
904
905 setup_memory_end();
906 setup_memory();
907 dma_contiguous_reserve(memory_end);
908
909 check_initrd();
910 reserve_crashkernel();
911#ifdef CONFIG_CRASH_DUMP
912 /*
913 * Be aware that smp_save_dump_cpus() triggers a system reset.
914 * Therefore CPU and device initialization should be done afterwards.
915 */
916 smp_save_dump_cpus();
917#endif
918
919 setup_resources();
920 setup_vmcoreinfo();
921 setup_lowcore();
922 smp_fill_possible_mask();
923 cpu_detect_mhz_feature();
924 cpu_init();
925 numa_setup();
926 smp_detect_cpus();
927 topology_init_early();
928
929 /*
930 * Create kernel page tables and switch to virtual addressing.
931 */
932 paging_init();
933
934 /* Setup default console */
935 conmode_default();
936 set_preferred_console();
937
938 /* Setup zfcpdump support */
939 setup_zfcpdump();
940
941 /* Add system specific data to the random pool */
942 setup_randomness();
943}