Loading...
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Architecture-specific setup.
4 *
5 * Copyright (C) 1998-2003 Hewlett-Packard Co
6 * David Mosberger-Tang <davidm@hpl.hp.com>
7 * 04/11/17 Ashok Raj <ashok.raj@intel.com> Added CPU Hotplug Support
8 *
9 * 2005-10-07 Keith Owens <kaos@sgi.com>
10 * Add notify_die() hooks.
11 */
12#include <linux/cpu.h>
13#include <linux/pm.h>
14#include <linux/elf.h>
15#include <linux/errno.h>
16#include <linux/kernel.h>
17#include <linux/mm.h>
18#include <linux/slab.h>
19#include <linux/module.h>
20#include <linux/notifier.h>
21#include <linux/personality.h>
22#include <linux/sched.h>
23#include <linux/sched/debug.h>
24#include <linux/sched/hotplug.h>
25#include <linux/sched/task.h>
26#include <linux/sched/task_stack.h>
27#include <linux/stddef.h>
28#include <linux/thread_info.h>
29#include <linux/unistd.h>
30#include <linux/efi.h>
31#include <linux/interrupt.h>
32#include <linux/delay.h>
33#include <linux/kdebug.h>
34#include <linux/utsname.h>
35#include <linux/tracehook.h>
36#include <linux/rcupdate.h>
37
38#include <asm/cpu.h>
39#include <asm/delay.h>
40#include <asm/elf.h>
41#include <asm/irq.h>
42#include <asm/kexec.h>
43#include <asm/processor.h>
44#include <asm/sal.h>
45#include <asm/switch_to.h>
46#include <asm/tlbflush.h>
47#include <linux/uaccess.h>
48#include <asm/unwind.h>
49#include <asm/user.h>
50#include <asm/xtp.h>
51
52#include "entry.h"
53
54#ifdef CONFIG_PERFMON
55# include <asm/perfmon.h>
56#endif
57
58#include "sigframe.h"
59
60void (*ia64_mark_idle)(int);
61
62unsigned long boot_option_idle_override = IDLE_NO_OVERRIDE;
63EXPORT_SYMBOL(boot_option_idle_override);
64void (*pm_power_off) (void);
65EXPORT_SYMBOL(pm_power_off);
66
67static void
68ia64_do_show_stack (struct unw_frame_info *info, void *arg)
69{
70 unsigned long ip, sp, bsp;
71 const char *loglvl = arg;
72
73 printk("%s\nCall Trace:\n", loglvl);
74 do {
75 unw_get_ip(info, &ip);
76 if (ip == 0)
77 break;
78
79 unw_get_sp(info, &sp);
80 unw_get_bsp(info, &bsp);
81 printk("%s [<%016lx>] %pS\n"
82 " sp=%016lx bsp=%016lx\n",
83 loglvl, ip, (void *)ip, sp, bsp);
84 } while (unw_unwind(info) >= 0);
85}
86
87void
88show_stack (struct task_struct *task, unsigned long *sp, const char *loglvl)
89{
90 if (!task)
91 unw_init_running(ia64_do_show_stack, (void *)loglvl);
92 else {
93 struct unw_frame_info info;
94
95 unw_init_from_blocked_task(&info, task);
96 ia64_do_show_stack(&info, (void *)loglvl);
97 }
98}
99
100void
101show_regs (struct pt_regs *regs)
102{
103 unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
104
105 print_modules();
106 printk("\n");
107 show_regs_print_info(KERN_DEFAULT);
108 printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s (%s)\n",
109 regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(),
110 init_utsname()->release);
111 printk("ip is at %pS\n", (void *)ip);
112 printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
113 regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
114 printk("rnat: %016lx bsps: %016lx pr : %016lx\n",
115 regs->ar_rnat, regs->ar_bspstore, regs->pr);
116 printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
117 regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
118 printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
119 printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs->b0, regs->b6, regs->b7);
120 printk("f6 : %05lx%016lx f7 : %05lx%016lx\n",
121 regs->f6.u.bits[1], regs->f6.u.bits[0],
122 regs->f7.u.bits[1], regs->f7.u.bits[0]);
123 printk("f8 : %05lx%016lx f9 : %05lx%016lx\n",
124 regs->f8.u.bits[1], regs->f8.u.bits[0],
125 regs->f9.u.bits[1], regs->f9.u.bits[0]);
126 printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
127 regs->f10.u.bits[1], regs->f10.u.bits[0],
128 regs->f11.u.bits[1], regs->f11.u.bits[0]);
129
130 printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs->r1, regs->r2, regs->r3);
131 printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
132 printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
133 printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
134 printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
135 printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
136 printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
137 printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
138 printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
139
140 if (user_mode(regs)) {
141 /* print the stacked registers */
142 unsigned long val, *bsp, ndirty;
143 int i, sof, is_nat = 0;
144
145 sof = regs->cr_ifs & 0x7f; /* size of frame */
146 ndirty = (regs->loadrs >> 19);
147 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
148 for (i = 0; i < sof; ++i) {
149 get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
150 printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
151 ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
152 }
153 } else
154 show_stack(NULL, NULL, KERN_DEFAULT);
155}
156
157/* local support for deprecated console_print */
158void
159console_print(const char *s)
160{
161 printk(KERN_EMERG "%s", s);
162}
163
164void
165do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
166{
167 if (fsys_mode(current, &scr->pt)) {
168 /*
169 * defer signal-handling etc. until we return to
170 * privilege-level 0.
171 */
172 if (!ia64_psr(&scr->pt)->lp)
173 ia64_psr(&scr->pt)->lp = 1;
174 return;
175 }
176
177#ifdef CONFIG_PERFMON
178 if (current->thread.pfm_needs_checking)
179 /*
180 * Note: pfm_handle_work() allow us to call it with interrupts
181 * disabled, and may enable interrupts within the function.
182 */
183 pfm_handle_work();
184#endif
185
186 /* deal with pending signal delivery */
187 if (test_thread_flag(TIF_SIGPENDING)) {
188 local_irq_enable(); /* force interrupt enable */
189 ia64_do_signal(scr, in_syscall);
190 }
191
192 if (test_and_clear_thread_flag(TIF_NOTIFY_RESUME)) {
193 local_irq_enable(); /* force interrupt enable */
194 tracehook_notify_resume(&scr->pt);
195 }
196
197 /* copy user rbs to kernel rbs */
198 if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
199 local_irq_enable(); /* force interrupt enable */
200 ia64_sync_krbs();
201 }
202
203 local_irq_disable(); /* force interrupt disable */
204}
205
206static int __init nohalt_setup(char * str)
207{
208 cpu_idle_poll_ctrl(true);
209 return 1;
210}
211__setup("nohalt", nohalt_setup);
212
213#ifdef CONFIG_HOTPLUG_CPU
214/* We don't actually take CPU down, just spin without interrupts. */
215static inline void play_dead(void)
216{
217 unsigned int this_cpu = smp_processor_id();
218
219 /* Ack it */
220 __this_cpu_write(cpu_state, CPU_DEAD);
221
222 max_xtp();
223 local_irq_disable();
224 idle_task_exit();
225 ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
226 /*
227 * The above is a point of no-return, the processor is
228 * expected to be in SAL loop now.
229 */
230 BUG();
231}
232#else
233static inline void play_dead(void)
234{
235 BUG();
236}
237#endif /* CONFIG_HOTPLUG_CPU */
238
239void arch_cpu_idle_dead(void)
240{
241 play_dead();
242}
243
244void arch_cpu_idle(void)
245{
246 void (*mark_idle)(int) = ia64_mark_idle;
247
248#ifdef CONFIG_SMP
249 min_xtp();
250#endif
251 rmb();
252 if (mark_idle)
253 (*mark_idle)(1);
254
255 safe_halt();
256
257 if (mark_idle)
258 (*mark_idle)(0);
259#ifdef CONFIG_SMP
260 normal_xtp();
261#endif
262}
263
264void
265ia64_save_extra (struct task_struct *task)
266{
267#ifdef CONFIG_PERFMON
268 unsigned long info;
269#endif
270
271 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
272 ia64_save_debug_regs(&task->thread.dbr[0]);
273
274#ifdef CONFIG_PERFMON
275 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
276 pfm_save_regs(task);
277
278 info = __this_cpu_read(pfm_syst_info);
279 if (info & PFM_CPUINFO_SYST_WIDE)
280 pfm_syst_wide_update_task(task, info, 0);
281#endif
282}
283
284void
285ia64_load_extra (struct task_struct *task)
286{
287#ifdef CONFIG_PERFMON
288 unsigned long info;
289#endif
290
291 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
292 ia64_load_debug_regs(&task->thread.dbr[0]);
293
294#ifdef CONFIG_PERFMON
295 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
296 pfm_load_regs(task);
297
298 info = __this_cpu_read(pfm_syst_info);
299 if (info & PFM_CPUINFO_SYST_WIDE)
300 pfm_syst_wide_update_task(task, info, 1);
301#endif
302}
303
304/*
305 * Copy the state of an ia-64 thread.
306 *
307 * We get here through the following call chain:
308 *
309 * from user-level: from kernel:
310 *
311 * <clone syscall> <some kernel call frames>
312 * sys_clone :
313 * _do_fork _do_fork
314 * copy_thread copy_thread
315 *
316 * This means that the stack layout is as follows:
317 *
318 * +---------------------+ (highest addr)
319 * | struct pt_regs |
320 * +---------------------+
321 * | struct switch_stack |
322 * +---------------------+
323 * | |
324 * | memory stack |
325 * | | <-- sp (lowest addr)
326 * +---------------------+
327 *
328 * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an
329 * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
330 * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the
331 * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since
332 * the stack is page aligned and the page size is at least 4KB, this is always the case,
333 * so there is nothing to worry about.
334 */
335int
336copy_thread(unsigned long clone_flags, unsigned long user_stack_base,
337 unsigned long user_stack_size, struct task_struct *p, unsigned long tls)
338{
339 extern char ia64_ret_from_clone;
340 struct switch_stack *child_stack, *stack;
341 unsigned long rbs, child_rbs, rbs_size;
342 struct pt_regs *child_ptregs;
343 struct pt_regs *regs = current_pt_regs();
344 int retval = 0;
345
346 child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
347 child_stack = (struct switch_stack *) child_ptregs - 1;
348
349 rbs = (unsigned long) current + IA64_RBS_OFFSET;
350 child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
351
352 /* copy parts of thread_struct: */
353 p->thread.ksp = (unsigned long) child_stack - 16;
354
355 /*
356 * NOTE: The calling convention considers all floating point
357 * registers in the high partition (fph) to be scratch. Since
358 * the only way to get to this point is through a system call,
359 * we know that the values in fph are all dead. Hence, there
360 * is no need to inherit the fph state from the parent to the
361 * child and all we have to do is to make sure that
362 * IA64_THREAD_FPH_VALID is cleared in the child.
363 *
364 * XXX We could push this optimization a bit further by
365 * clearing IA64_THREAD_FPH_VALID on ANY system call.
366 * However, it's not clear this is worth doing. Also, it
367 * would be a slight deviation from the normal Linux system
368 * call behavior where scratch registers are preserved across
369 * system calls (unless used by the system call itself).
370 */
371# define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
372 | IA64_THREAD_PM_VALID)
373# define THREAD_FLAGS_TO_SET 0
374 p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
375 | THREAD_FLAGS_TO_SET);
376
377 ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */
378
379 if (unlikely(p->flags & PF_KTHREAD)) {
380 if (unlikely(!user_stack_base)) {
381 /* fork_idle() called us */
382 return 0;
383 }
384 memset(child_stack, 0, sizeof(*child_ptregs) + sizeof(*child_stack));
385 child_stack->r4 = user_stack_base; /* payload */
386 child_stack->r5 = user_stack_size; /* argument */
387 /*
388 * Preserve PSR bits, except for bits 32-34 and 37-45,
389 * which we can't read.
390 */
391 child_ptregs->cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
392 /* mark as valid, empty frame */
393 child_ptregs->cr_ifs = 1UL << 63;
394 child_stack->ar_fpsr = child_ptregs->ar_fpsr
395 = ia64_getreg(_IA64_REG_AR_FPSR);
396 child_stack->pr = (1 << PRED_KERNEL_STACK);
397 child_stack->ar_bspstore = child_rbs;
398 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
399
400 /* stop some PSR bits from being inherited.
401 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
402 * therefore we must specify them explicitly here and not include them in
403 * IA64_PSR_BITS_TO_CLEAR.
404 */
405 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
406 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
407
408 return 0;
409 }
410 stack = ((struct switch_stack *) regs) - 1;
411 /* copy parent's switch_stack & pt_regs to child: */
412 memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
413
414 /* copy the parent's register backing store to the child: */
415 rbs_size = stack->ar_bspstore - rbs;
416 memcpy((void *) child_rbs, (void *) rbs, rbs_size);
417 if (clone_flags & CLONE_SETTLS)
418 child_ptregs->r13 = tls;
419 if (user_stack_base) {
420 child_ptregs->r12 = user_stack_base + user_stack_size - 16;
421 child_ptregs->ar_bspstore = user_stack_base;
422 child_ptregs->ar_rnat = 0;
423 child_ptregs->loadrs = 0;
424 }
425 child_stack->ar_bspstore = child_rbs + rbs_size;
426 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
427
428 /* stop some PSR bits from being inherited.
429 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
430 * therefore we must specify them explicitly here and not include them in
431 * IA64_PSR_BITS_TO_CLEAR.
432 */
433 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
434 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
435
436#ifdef CONFIG_PERFMON
437 if (current->thread.pfm_context)
438 pfm_inherit(p, child_ptregs);
439#endif
440 return retval;
441}
442
443asmlinkage long ia64_clone(unsigned long clone_flags, unsigned long stack_start,
444 unsigned long stack_size, unsigned long parent_tidptr,
445 unsigned long child_tidptr, unsigned long tls)
446{
447 struct kernel_clone_args args = {
448 .flags = (lower_32_bits(clone_flags) & ~CSIGNAL),
449 .pidfd = (int __user *)parent_tidptr,
450 .child_tid = (int __user *)child_tidptr,
451 .parent_tid = (int __user *)parent_tidptr,
452 .exit_signal = (lower_32_bits(clone_flags) & CSIGNAL),
453 .stack = stack_start,
454 .stack_size = stack_size,
455 .tls = tls,
456 };
457
458 return _do_fork(&args);
459}
460
461static void
462do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
463{
464 unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
465 unsigned long ip;
466 elf_greg_t *dst = arg;
467 struct pt_regs *pt;
468 char nat;
469 int i;
470
471 memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */
472
473 if (unw_unwind_to_user(info) < 0)
474 return;
475
476 unw_get_sp(info, &sp);
477 pt = (struct pt_regs *) (sp + 16);
478
479 urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
480
481 if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
482 return;
483
484 ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
485 &ar_rnat);
486
487 /*
488 * coredump format:
489 * r0-r31
490 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
491 * predicate registers (p0-p63)
492 * b0-b7
493 * ip cfm user-mask
494 * ar.rsc ar.bsp ar.bspstore ar.rnat
495 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
496 */
497
498 /* r0 is zero */
499 for (i = 1, mask = (1UL << i); i < 32; ++i) {
500 unw_get_gr(info, i, &dst[i], &nat);
501 if (nat)
502 nat_bits |= mask;
503 mask <<= 1;
504 }
505 dst[32] = nat_bits;
506 unw_get_pr(info, &dst[33]);
507
508 for (i = 0; i < 8; ++i)
509 unw_get_br(info, i, &dst[34 + i]);
510
511 unw_get_rp(info, &ip);
512 dst[42] = ip + ia64_psr(pt)->ri;
513 dst[43] = cfm;
514 dst[44] = pt->cr_ipsr & IA64_PSR_UM;
515
516 unw_get_ar(info, UNW_AR_RSC, &dst[45]);
517 /*
518 * For bsp and bspstore, unw_get_ar() would return the kernel
519 * addresses, but we need the user-level addresses instead:
520 */
521 dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */
522 dst[47] = pt->ar_bspstore;
523 dst[48] = ar_rnat;
524 unw_get_ar(info, UNW_AR_CCV, &dst[49]);
525 unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
526 unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
527 dst[52] = pt->ar_pfs; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
528 unw_get_ar(info, UNW_AR_LC, &dst[53]);
529 unw_get_ar(info, UNW_AR_EC, &dst[54]);
530 unw_get_ar(info, UNW_AR_CSD, &dst[55]);
531 unw_get_ar(info, UNW_AR_SSD, &dst[56]);
532}
533
534void
535do_copy_regs (struct unw_frame_info *info, void *arg)
536{
537 do_copy_task_regs(current, info, arg);
538}
539
540void
541ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
542{
543 unw_init_running(do_copy_regs, dst);
544}
545
546/*
547 * Flush thread state. This is called when a thread does an execve().
548 */
549void
550flush_thread (void)
551{
552 /* drop floating-point and debug-register state if it exists: */
553 current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
554 ia64_drop_fpu(current);
555}
556
557/*
558 * Clean up state associated with a thread. This is called when
559 * the thread calls exit().
560 */
561void
562exit_thread (struct task_struct *tsk)
563{
564
565 ia64_drop_fpu(tsk);
566#ifdef CONFIG_PERFMON
567 /* if needed, stop monitoring and flush state to perfmon context */
568 if (tsk->thread.pfm_context)
569 pfm_exit_thread(tsk);
570
571 /* free debug register resources */
572 if (tsk->thread.flags & IA64_THREAD_DBG_VALID)
573 pfm_release_debug_registers(tsk);
574#endif
575}
576
577unsigned long
578get_wchan (struct task_struct *p)
579{
580 struct unw_frame_info info;
581 unsigned long ip;
582 int count = 0;
583
584 if (!p || p == current || p->state == TASK_RUNNING)
585 return 0;
586
587 /*
588 * Note: p may not be a blocked task (it could be current or
589 * another process running on some other CPU. Rather than
590 * trying to determine if p is really blocked, we just assume
591 * it's blocked and rely on the unwind routines to fail
592 * gracefully if the process wasn't really blocked after all.
593 * --davidm 99/12/15
594 */
595 unw_init_from_blocked_task(&info, p);
596 do {
597 if (p->state == TASK_RUNNING)
598 return 0;
599 if (unw_unwind(&info) < 0)
600 return 0;
601 unw_get_ip(&info, &ip);
602 if (!in_sched_functions(ip))
603 return ip;
604 } while (count++ < 16);
605 return 0;
606}
607
608void
609cpu_halt (void)
610{
611 pal_power_mgmt_info_u_t power_info[8];
612 unsigned long min_power;
613 int i, min_power_state;
614
615 if (ia64_pal_halt_info(power_info) != 0)
616 return;
617
618 min_power_state = 0;
619 min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
620 for (i = 1; i < 8; ++i)
621 if (power_info[i].pal_power_mgmt_info_s.im
622 && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
623 min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
624 min_power_state = i;
625 }
626
627 while (1)
628 ia64_pal_halt(min_power_state);
629}
630
631void machine_shutdown(void)
632{
633 smp_shutdown_nonboot_cpus(reboot_cpu);
634
635#ifdef CONFIG_KEXEC
636 kexec_disable_iosapic();
637#endif
638}
639
640void
641machine_restart (char *restart_cmd)
642{
643 (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
644 efi_reboot(REBOOT_WARM, NULL);
645}
646
647void
648machine_halt (void)
649{
650 (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
651 cpu_halt();
652}
653
654void
655machine_power_off (void)
656{
657 if (pm_power_off)
658 pm_power_off();
659 machine_halt();
660}
661
662EXPORT_SYMBOL(ia64_delay_loop);
1/*
2 * Architecture-specific setup.
3 *
4 * Copyright (C) 1998-2003 Hewlett-Packard Co
5 * David Mosberger-Tang <davidm@hpl.hp.com>
6 * 04/11/17 Ashok Raj <ashok.raj@intel.com> Added CPU Hotplug Support
7 *
8 * 2005-10-07 Keith Owens <kaos@sgi.com>
9 * Add notify_die() hooks.
10 */
11#include <linux/cpu.h>
12#include <linux/pm.h>
13#include <linux/elf.h>
14#include <linux/errno.h>
15#include <linux/kallsyms.h>
16#include <linux/kernel.h>
17#include <linux/mm.h>
18#include <linux/slab.h>
19#include <linux/module.h>
20#include <linux/notifier.h>
21#include <linux/personality.h>
22#include <linux/sched.h>
23#include <linux/stddef.h>
24#include <linux/thread_info.h>
25#include <linux/unistd.h>
26#include <linux/efi.h>
27#include <linux/interrupt.h>
28#include <linux/delay.h>
29#include <linux/kdebug.h>
30#include <linux/utsname.h>
31#include <linux/tracehook.h>
32
33#include <asm/cpu.h>
34#include <asm/delay.h>
35#include <asm/elf.h>
36#include <asm/irq.h>
37#include <asm/kexec.h>
38#include <asm/pgalloc.h>
39#include <asm/processor.h>
40#include <asm/sal.h>
41#include <asm/switch_to.h>
42#include <asm/tlbflush.h>
43#include <asm/uaccess.h>
44#include <asm/unwind.h>
45#include <asm/user.h>
46
47#include "entry.h"
48
49#ifdef CONFIG_PERFMON
50# include <asm/perfmon.h>
51#endif
52
53#include "sigframe.h"
54
55void (*ia64_mark_idle)(int);
56
57unsigned long boot_option_idle_override = IDLE_NO_OVERRIDE;
58EXPORT_SYMBOL(boot_option_idle_override);
59void (*pm_idle) (void);
60EXPORT_SYMBOL(pm_idle);
61void (*pm_power_off) (void);
62EXPORT_SYMBOL(pm_power_off);
63
64void
65ia64_do_show_stack (struct unw_frame_info *info, void *arg)
66{
67 unsigned long ip, sp, bsp;
68 char buf[128]; /* don't make it so big that it overflows the stack! */
69
70 printk("\nCall Trace:\n");
71 do {
72 unw_get_ip(info, &ip);
73 if (ip == 0)
74 break;
75
76 unw_get_sp(info, &sp);
77 unw_get_bsp(info, &bsp);
78 snprintf(buf, sizeof(buf),
79 " [<%016lx>] %%s\n"
80 " sp=%016lx bsp=%016lx\n",
81 ip, sp, bsp);
82 print_symbol(buf, ip);
83 } while (unw_unwind(info) >= 0);
84}
85
86void
87show_stack (struct task_struct *task, unsigned long *sp)
88{
89 if (!task)
90 unw_init_running(ia64_do_show_stack, NULL);
91 else {
92 struct unw_frame_info info;
93
94 unw_init_from_blocked_task(&info, task);
95 ia64_do_show_stack(&info, NULL);
96 }
97}
98
99void
100dump_stack (void)
101{
102 show_stack(NULL, NULL);
103}
104
105EXPORT_SYMBOL(dump_stack);
106
107void
108show_regs (struct pt_regs *regs)
109{
110 unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
111
112 print_modules();
113 printk("\nPid: %d, CPU %d, comm: %20s\n", task_pid_nr(current),
114 smp_processor_id(), current->comm);
115 printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s (%s)\n",
116 regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(),
117 init_utsname()->release);
118 print_symbol("ip is at %s\n", ip);
119 printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
120 regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
121 printk("rnat: %016lx bsps: %016lx pr : %016lx\n",
122 regs->ar_rnat, regs->ar_bspstore, regs->pr);
123 printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
124 regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
125 printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
126 printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs->b0, regs->b6, regs->b7);
127 printk("f6 : %05lx%016lx f7 : %05lx%016lx\n",
128 regs->f6.u.bits[1], regs->f6.u.bits[0],
129 regs->f7.u.bits[1], regs->f7.u.bits[0]);
130 printk("f8 : %05lx%016lx f9 : %05lx%016lx\n",
131 regs->f8.u.bits[1], regs->f8.u.bits[0],
132 regs->f9.u.bits[1], regs->f9.u.bits[0]);
133 printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
134 regs->f10.u.bits[1], regs->f10.u.bits[0],
135 regs->f11.u.bits[1], regs->f11.u.bits[0]);
136
137 printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs->r1, regs->r2, regs->r3);
138 printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
139 printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
140 printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
141 printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
142 printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
143 printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
144 printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
145 printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
146
147 if (user_mode(regs)) {
148 /* print the stacked registers */
149 unsigned long val, *bsp, ndirty;
150 int i, sof, is_nat = 0;
151
152 sof = regs->cr_ifs & 0x7f; /* size of frame */
153 ndirty = (regs->loadrs >> 19);
154 bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
155 for (i = 0; i < sof; ++i) {
156 get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
157 printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
158 ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
159 }
160 } else
161 show_stack(NULL, NULL);
162}
163
164/* local support for deprecated console_print */
165void
166console_print(const char *s)
167{
168 printk(KERN_EMERG "%s", s);
169}
170
171void
172do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
173{
174 if (fsys_mode(current, &scr->pt)) {
175 /*
176 * defer signal-handling etc. until we return to
177 * privilege-level 0.
178 */
179 if (!ia64_psr(&scr->pt)->lp)
180 ia64_psr(&scr->pt)->lp = 1;
181 return;
182 }
183
184#ifdef CONFIG_PERFMON
185 if (current->thread.pfm_needs_checking)
186 /*
187 * Note: pfm_handle_work() allow us to call it with interrupts
188 * disabled, and may enable interrupts within the function.
189 */
190 pfm_handle_work();
191#endif
192
193 /* deal with pending signal delivery */
194 if (test_thread_flag(TIF_SIGPENDING)) {
195 local_irq_enable(); /* force interrupt enable */
196 ia64_do_signal(scr, in_syscall);
197 }
198
199 if (test_thread_flag(TIF_NOTIFY_RESUME)) {
200 clear_thread_flag(TIF_NOTIFY_RESUME);
201 tracehook_notify_resume(&scr->pt);
202 }
203
204 /* copy user rbs to kernel rbs */
205 if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
206 local_irq_enable(); /* force interrupt enable */
207 ia64_sync_krbs();
208 }
209
210 local_irq_disable(); /* force interrupt disable */
211}
212
213static int pal_halt = 1;
214static int can_do_pal_halt = 1;
215
216static int __init nohalt_setup(char * str)
217{
218 pal_halt = can_do_pal_halt = 0;
219 return 1;
220}
221__setup("nohalt", nohalt_setup);
222
223void
224update_pal_halt_status(int status)
225{
226 can_do_pal_halt = pal_halt && status;
227}
228
229/*
230 * We use this if we don't have any better idle routine..
231 */
232void
233default_idle (void)
234{
235 local_irq_enable();
236 while (!need_resched()) {
237 if (can_do_pal_halt) {
238 local_irq_disable();
239 if (!need_resched()) {
240 safe_halt();
241 }
242 local_irq_enable();
243 } else
244 cpu_relax();
245 }
246}
247
248#ifdef CONFIG_HOTPLUG_CPU
249/* We don't actually take CPU down, just spin without interrupts. */
250static inline void play_dead(void)
251{
252 unsigned int this_cpu = smp_processor_id();
253
254 /* Ack it */
255 __get_cpu_var(cpu_state) = CPU_DEAD;
256
257 max_xtp();
258 local_irq_disable();
259 idle_task_exit();
260 ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
261 /*
262 * The above is a point of no-return, the processor is
263 * expected to be in SAL loop now.
264 */
265 BUG();
266}
267#else
268static inline void play_dead(void)
269{
270 BUG();
271}
272#endif /* CONFIG_HOTPLUG_CPU */
273
274void __attribute__((noreturn))
275cpu_idle (void)
276{
277 void (*mark_idle)(int) = ia64_mark_idle;
278 int cpu = smp_processor_id();
279
280 /* endless idle loop with no priority at all */
281 while (1) {
282 if (can_do_pal_halt) {
283 current_thread_info()->status &= ~TS_POLLING;
284 /*
285 * TS_POLLING-cleared state must be visible before we
286 * test NEED_RESCHED:
287 */
288 smp_mb();
289 } else {
290 current_thread_info()->status |= TS_POLLING;
291 }
292
293 if (!need_resched()) {
294 void (*idle)(void);
295#ifdef CONFIG_SMP
296 min_xtp();
297#endif
298 rmb();
299 if (mark_idle)
300 (*mark_idle)(1);
301
302 idle = pm_idle;
303 if (!idle)
304 idle = default_idle;
305 (*idle)();
306 if (mark_idle)
307 (*mark_idle)(0);
308#ifdef CONFIG_SMP
309 normal_xtp();
310#endif
311 }
312 schedule_preempt_disabled();
313 check_pgt_cache();
314 if (cpu_is_offline(cpu))
315 play_dead();
316 }
317}
318
319void
320ia64_save_extra (struct task_struct *task)
321{
322#ifdef CONFIG_PERFMON
323 unsigned long info;
324#endif
325
326 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
327 ia64_save_debug_regs(&task->thread.dbr[0]);
328
329#ifdef CONFIG_PERFMON
330 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
331 pfm_save_regs(task);
332
333 info = __get_cpu_var(pfm_syst_info);
334 if (info & PFM_CPUINFO_SYST_WIDE)
335 pfm_syst_wide_update_task(task, info, 0);
336#endif
337}
338
339void
340ia64_load_extra (struct task_struct *task)
341{
342#ifdef CONFIG_PERFMON
343 unsigned long info;
344#endif
345
346 if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
347 ia64_load_debug_regs(&task->thread.dbr[0]);
348
349#ifdef CONFIG_PERFMON
350 if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
351 pfm_load_regs(task);
352
353 info = __get_cpu_var(pfm_syst_info);
354 if (info & PFM_CPUINFO_SYST_WIDE)
355 pfm_syst_wide_update_task(task, info, 1);
356#endif
357}
358
359/*
360 * Copy the state of an ia-64 thread.
361 *
362 * We get here through the following call chain:
363 *
364 * from user-level: from kernel:
365 *
366 * <clone syscall> <some kernel call frames>
367 * sys_clone :
368 * do_fork do_fork
369 * copy_thread copy_thread
370 *
371 * This means that the stack layout is as follows:
372 *
373 * +---------------------+ (highest addr)
374 * | struct pt_regs |
375 * +---------------------+
376 * | struct switch_stack |
377 * +---------------------+
378 * | |
379 * | memory stack |
380 * | | <-- sp (lowest addr)
381 * +---------------------+
382 *
383 * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an
384 * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
385 * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the
386 * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since
387 * the stack is page aligned and the page size is at least 4KB, this is always the case,
388 * so there is nothing to worry about.
389 */
390int
391copy_thread(unsigned long clone_flags,
392 unsigned long user_stack_base, unsigned long user_stack_size,
393 struct task_struct *p, struct pt_regs *regs)
394{
395 extern char ia64_ret_from_clone;
396 struct switch_stack *child_stack, *stack;
397 unsigned long rbs, child_rbs, rbs_size;
398 struct pt_regs *child_ptregs;
399 int retval = 0;
400
401#ifdef CONFIG_SMP
402 /*
403 * For SMP idle threads, fork_by_hand() calls do_fork with
404 * NULL regs.
405 */
406 if (!regs)
407 return 0;
408#endif
409
410 stack = ((struct switch_stack *) regs) - 1;
411
412 child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
413 child_stack = (struct switch_stack *) child_ptregs - 1;
414
415 /* copy parent's switch_stack & pt_regs to child: */
416 memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
417
418 rbs = (unsigned long) current + IA64_RBS_OFFSET;
419 child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
420 rbs_size = stack->ar_bspstore - rbs;
421
422 /* copy the parent's register backing store to the child: */
423 memcpy((void *) child_rbs, (void *) rbs, rbs_size);
424
425 if (likely(user_mode(child_ptregs))) {
426 if (clone_flags & CLONE_SETTLS)
427 child_ptregs->r13 = regs->r16; /* see sys_clone2() in entry.S */
428 if (user_stack_base) {
429 child_ptregs->r12 = user_stack_base + user_stack_size - 16;
430 child_ptregs->ar_bspstore = user_stack_base;
431 child_ptregs->ar_rnat = 0;
432 child_ptregs->loadrs = 0;
433 }
434 } else {
435 /*
436 * Note: we simply preserve the relative position of
437 * the stack pointer here. There is no need to
438 * allocate a scratch area here, since that will have
439 * been taken care of by the caller of sys_clone()
440 * already.
441 */
442 child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
443 child_ptregs->r13 = (unsigned long) p; /* set `current' pointer */
444 }
445 child_stack->ar_bspstore = child_rbs + rbs_size;
446 child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
447
448 /* copy parts of thread_struct: */
449 p->thread.ksp = (unsigned long) child_stack - 16;
450
451 /* stop some PSR bits from being inherited.
452 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
453 * therefore we must specify them explicitly here and not include them in
454 * IA64_PSR_BITS_TO_CLEAR.
455 */
456 child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
457 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
458
459 /*
460 * NOTE: The calling convention considers all floating point
461 * registers in the high partition (fph) to be scratch. Since
462 * the only way to get to this point is through a system call,
463 * we know that the values in fph are all dead. Hence, there
464 * is no need to inherit the fph state from the parent to the
465 * child and all we have to do is to make sure that
466 * IA64_THREAD_FPH_VALID is cleared in the child.
467 *
468 * XXX We could push this optimization a bit further by
469 * clearing IA64_THREAD_FPH_VALID on ANY system call.
470 * However, it's not clear this is worth doing. Also, it
471 * would be a slight deviation from the normal Linux system
472 * call behavior where scratch registers are preserved across
473 * system calls (unless used by the system call itself).
474 */
475# define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
476 | IA64_THREAD_PM_VALID)
477# define THREAD_FLAGS_TO_SET 0
478 p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
479 | THREAD_FLAGS_TO_SET);
480 ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */
481
482#ifdef CONFIG_PERFMON
483 if (current->thread.pfm_context)
484 pfm_inherit(p, child_ptregs);
485#endif
486 return retval;
487}
488
489static void
490do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
491{
492 unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
493 unsigned long uninitialized_var(ip); /* GCC be quiet */
494 elf_greg_t *dst = arg;
495 struct pt_regs *pt;
496 char nat;
497 int i;
498
499 memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */
500
501 if (unw_unwind_to_user(info) < 0)
502 return;
503
504 unw_get_sp(info, &sp);
505 pt = (struct pt_regs *) (sp + 16);
506
507 urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
508
509 if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
510 return;
511
512 ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
513 &ar_rnat);
514
515 /*
516 * coredump format:
517 * r0-r31
518 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
519 * predicate registers (p0-p63)
520 * b0-b7
521 * ip cfm user-mask
522 * ar.rsc ar.bsp ar.bspstore ar.rnat
523 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
524 */
525
526 /* r0 is zero */
527 for (i = 1, mask = (1UL << i); i < 32; ++i) {
528 unw_get_gr(info, i, &dst[i], &nat);
529 if (nat)
530 nat_bits |= mask;
531 mask <<= 1;
532 }
533 dst[32] = nat_bits;
534 unw_get_pr(info, &dst[33]);
535
536 for (i = 0; i < 8; ++i)
537 unw_get_br(info, i, &dst[34 + i]);
538
539 unw_get_rp(info, &ip);
540 dst[42] = ip + ia64_psr(pt)->ri;
541 dst[43] = cfm;
542 dst[44] = pt->cr_ipsr & IA64_PSR_UM;
543
544 unw_get_ar(info, UNW_AR_RSC, &dst[45]);
545 /*
546 * For bsp and bspstore, unw_get_ar() would return the kernel
547 * addresses, but we need the user-level addresses instead:
548 */
549 dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */
550 dst[47] = pt->ar_bspstore;
551 dst[48] = ar_rnat;
552 unw_get_ar(info, UNW_AR_CCV, &dst[49]);
553 unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
554 unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
555 dst[52] = pt->ar_pfs; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
556 unw_get_ar(info, UNW_AR_LC, &dst[53]);
557 unw_get_ar(info, UNW_AR_EC, &dst[54]);
558 unw_get_ar(info, UNW_AR_CSD, &dst[55]);
559 unw_get_ar(info, UNW_AR_SSD, &dst[56]);
560}
561
562void
563do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
564{
565 elf_fpreg_t *dst = arg;
566 int i;
567
568 memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */
569
570 if (unw_unwind_to_user(info) < 0)
571 return;
572
573 /* f0 is 0.0, f1 is 1.0 */
574
575 for (i = 2; i < 32; ++i)
576 unw_get_fr(info, i, dst + i);
577
578 ia64_flush_fph(task);
579 if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
580 memcpy(dst + 32, task->thread.fph, 96*16);
581}
582
583void
584do_copy_regs (struct unw_frame_info *info, void *arg)
585{
586 do_copy_task_regs(current, info, arg);
587}
588
589void
590do_dump_fpu (struct unw_frame_info *info, void *arg)
591{
592 do_dump_task_fpu(current, info, arg);
593}
594
595void
596ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
597{
598 unw_init_running(do_copy_regs, dst);
599}
600
601int
602dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
603{
604 unw_init_running(do_dump_fpu, dst);
605 return 1; /* f0-f31 are always valid so we always return 1 */
606}
607
608long
609sys_execve (const char __user *filename,
610 const char __user *const __user *argv,
611 const char __user *const __user *envp,
612 struct pt_regs *regs)
613{
614 char *fname;
615 int error;
616
617 fname = getname(filename);
618 error = PTR_ERR(fname);
619 if (IS_ERR(fname))
620 goto out;
621 error = do_execve(fname, argv, envp, regs);
622 putname(fname);
623out:
624 return error;
625}
626
627pid_t
628kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
629{
630 extern void start_kernel_thread (void);
631 unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
632 struct {
633 struct switch_stack sw;
634 struct pt_regs pt;
635 } regs;
636
637 memset(®s, 0, sizeof(regs));
638 regs.pt.cr_iip = helper_fptr[0]; /* set entry point (IP) */
639 regs.pt.r1 = helper_fptr[1]; /* set GP */
640 regs.pt.r9 = (unsigned long) fn; /* 1st argument */
641 regs.pt.r11 = (unsigned long) arg; /* 2nd argument */
642 /* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read. */
643 regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
644 regs.pt.cr_ifs = 1UL << 63; /* mark as valid, empty frame */
645 regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
646 regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
647 regs.sw.pr = (1 << PRED_KERNEL_STACK);
648 return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, ®s.pt, 0, NULL, NULL);
649}
650EXPORT_SYMBOL(kernel_thread);
651
652/* This gets called from kernel_thread() via ia64_invoke_thread_helper(). */
653int
654kernel_thread_helper (int (*fn)(void *), void *arg)
655{
656 return (*fn)(arg);
657}
658
659/*
660 * Flush thread state. This is called when a thread does an execve().
661 */
662void
663flush_thread (void)
664{
665 /* drop floating-point and debug-register state if it exists: */
666 current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
667 ia64_drop_fpu(current);
668}
669
670/*
671 * Clean up state associated with current thread. This is called when
672 * the thread calls exit().
673 */
674void
675exit_thread (void)
676{
677
678 ia64_drop_fpu(current);
679#ifdef CONFIG_PERFMON
680 /* if needed, stop monitoring and flush state to perfmon context */
681 if (current->thread.pfm_context)
682 pfm_exit_thread(current);
683
684 /* free debug register resources */
685 if (current->thread.flags & IA64_THREAD_DBG_VALID)
686 pfm_release_debug_registers(current);
687#endif
688}
689
690unsigned long
691get_wchan (struct task_struct *p)
692{
693 struct unw_frame_info info;
694 unsigned long ip;
695 int count = 0;
696
697 if (!p || p == current || p->state == TASK_RUNNING)
698 return 0;
699
700 /*
701 * Note: p may not be a blocked task (it could be current or
702 * another process running on some other CPU. Rather than
703 * trying to determine if p is really blocked, we just assume
704 * it's blocked and rely on the unwind routines to fail
705 * gracefully if the process wasn't really blocked after all.
706 * --davidm 99/12/15
707 */
708 unw_init_from_blocked_task(&info, p);
709 do {
710 if (p->state == TASK_RUNNING)
711 return 0;
712 if (unw_unwind(&info) < 0)
713 return 0;
714 unw_get_ip(&info, &ip);
715 if (!in_sched_functions(ip))
716 return ip;
717 } while (count++ < 16);
718 return 0;
719}
720
721void
722cpu_halt (void)
723{
724 pal_power_mgmt_info_u_t power_info[8];
725 unsigned long min_power;
726 int i, min_power_state;
727
728 if (ia64_pal_halt_info(power_info) != 0)
729 return;
730
731 min_power_state = 0;
732 min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
733 for (i = 1; i < 8; ++i)
734 if (power_info[i].pal_power_mgmt_info_s.im
735 && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
736 min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
737 min_power_state = i;
738 }
739
740 while (1)
741 ia64_pal_halt(min_power_state);
742}
743
744void machine_shutdown(void)
745{
746#ifdef CONFIG_HOTPLUG_CPU
747 int cpu;
748
749 for_each_online_cpu(cpu) {
750 if (cpu != smp_processor_id())
751 cpu_down(cpu);
752 }
753#endif
754#ifdef CONFIG_KEXEC
755 kexec_disable_iosapic();
756#endif
757}
758
759void
760machine_restart (char *restart_cmd)
761{
762 (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
763 (*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
764}
765
766void
767machine_halt (void)
768{
769 (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
770 cpu_halt();
771}
772
773void
774machine_power_off (void)
775{
776 if (pm_power_off)
777 pm_power_off();
778 machine_halt();
779}
780