xref: /openbmc/linux/arch/ia64/kernel/process.c (revision 63dc02bd)
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 
55 void (*ia64_mark_idle)(int);
56 
57 unsigned long boot_option_idle_override = IDLE_NO_OVERRIDE;
58 EXPORT_SYMBOL(boot_option_idle_override);
59 void (*pm_idle) (void);
60 EXPORT_SYMBOL(pm_idle);
61 void (*pm_power_off) (void);
62 EXPORT_SYMBOL(pm_power_off);
63 
64 void
65 ia64_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 
86 void
87 show_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 
99 void
100 dump_stack (void)
101 {
102 	show_stack(NULL, NULL);
103 }
104 
105 EXPORT_SYMBOL(dump_stack);
106 
107 void
108 show_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 */
165 void
166 console_print(const char *s)
167 {
168 	printk(KERN_EMERG "%s", s);
169 }
170 
171 void
172 do_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 		if (current->replacement_session_keyring)
203 			key_replace_session_keyring();
204 	}
205 
206 	/* copy user rbs to kernel rbs */
207 	if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
208 		local_irq_enable();	/* force interrupt enable */
209 		ia64_sync_krbs();
210 	}
211 
212 	local_irq_disable();	/* force interrupt disable */
213 }
214 
215 static int pal_halt        = 1;
216 static int can_do_pal_halt = 1;
217 
218 static int __init nohalt_setup(char * str)
219 {
220 	pal_halt = can_do_pal_halt = 0;
221 	return 1;
222 }
223 __setup("nohalt", nohalt_setup);
224 
225 void
226 update_pal_halt_status(int status)
227 {
228 	can_do_pal_halt = pal_halt && status;
229 }
230 
231 /*
232  * We use this if we don't have any better idle routine..
233  */
234 void
235 default_idle (void)
236 {
237 	local_irq_enable();
238 	while (!need_resched()) {
239 		if (can_do_pal_halt) {
240 			local_irq_disable();
241 			if (!need_resched()) {
242 				safe_halt();
243 			}
244 			local_irq_enable();
245 		} else
246 			cpu_relax();
247 	}
248 }
249 
250 #ifdef CONFIG_HOTPLUG_CPU
251 /* We don't actually take CPU down, just spin without interrupts. */
252 static inline void play_dead(void)
253 {
254 	unsigned int this_cpu = smp_processor_id();
255 
256 	/* Ack it */
257 	__get_cpu_var(cpu_state) = CPU_DEAD;
258 
259 	max_xtp();
260 	local_irq_disable();
261 	idle_task_exit();
262 	ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
263 	/*
264 	 * The above is a point of no-return, the processor is
265 	 * expected to be in SAL loop now.
266 	 */
267 	BUG();
268 }
269 #else
270 static inline void play_dead(void)
271 {
272 	BUG();
273 }
274 #endif /* CONFIG_HOTPLUG_CPU */
275 
276 static void do_nothing(void *unused)
277 {
278 }
279 
280 /*
281  * cpu_idle_wait - Used to ensure that all the CPUs discard old value of
282  * pm_idle and update to new pm_idle value. Required while changing pm_idle
283  * handler on SMP systems.
284  *
285  * Caller must have changed pm_idle to the new value before the call. Old
286  * pm_idle value will not be used by any CPU after the return of this function.
287  */
288 void cpu_idle_wait(void)
289 {
290 	smp_mb();
291 	/* kick all the CPUs so that they exit out of pm_idle */
292 	smp_call_function(do_nothing, NULL, 1);
293 }
294 EXPORT_SYMBOL_GPL(cpu_idle_wait);
295 
296 void __attribute__((noreturn))
297 cpu_idle (void)
298 {
299 	void (*mark_idle)(int) = ia64_mark_idle;
300   	int cpu = smp_processor_id();
301 
302 	/* endless idle loop with no priority at all */
303 	while (1) {
304 		if (can_do_pal_halt) {
305 			current_thread_info()->status &= ~TS_POLLING;
306 			/*
307 			 * TS_POLLING-cleared state must be visible before we
308 			 * test NEED_RESCHED:
309 			 */
310 			smp_mb();
311 		} else {
312 			current_thread_info()->status |= TS_POLLING;
313 		}
314 
315 		if (!need_resched()) {
316 			void (*idle)(void);
317 #ifdef CONFIG_SMP
318 			min_xtp();
319 #endif
320 			rmb();
321 			if (mark_idle)
322 				(*mark_idle)(1);
323 
324 			idle = pm_idle;
325 			if (!idle)
326 				idle = default_idle;
327 			(*idle)();
328 			if (mark_idle)
329 				(*mark_idle)(0);
330 #ifdef CONFIG_SMP
331 			normal_xtp();
332 #endif
333 		}
334 		schedule_preempt_disabled();
335 		check_pgt_cache();
336 		if (cpu_is_offline(cpu))
337 			play_dead();
338 	}
339 }
340 
341 void
342 ia64_save_extra (struct task_struct *task)
343 {
344 #ifdef CONFIG_PERFMON
345 	unsigned long info;
346 #endif
347 
348 	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
349 		ia64_save_debug_regs(&task->thread.dbr[0]);
350 
351 #ifdef CONFIG_PERFMON
352 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
353 		pfm_save_regs(task);
354 
355 	info = __get_cpu_var(pfm_syst_info);
356 	if (info & PFM_CPUINFO_SYST_WIDE)
357 		pfm_syst_wide_update_task(task, info, 0);
358 #endif
359 }
360 
361 void
362 ia64_load_extra (struct task_struct *task)
363 {
364 #ifdef CONFIG_PERFMON
365 	unsigned long info;
366 #endif
367 
368 	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
369 		ia64_load_debug_regs(&task->thread.dbr[0]);
370 
371 #ifdef CONFIG_PERFMON
372 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
373 		pfm_load_regs(task);
374 
375 	info = __get_cpu_var(pfm_syst_info);
376 	if (info & PFM_CPUINFO_SYST_WIDE)
377 		pfm_syst_wide_update_task(task, info, 1);
378 #endif
379 }
380 
381 /*
382  * Copy the state of an ia-64 thread.
383  *
384  * We get here through the following  call chain:
385  *
386  *	from user-level:	from kernel:
387  *
388  *	<clone syscall>	        <some kernel call frames>
389  *	sys_clone		   :
390  *	do_fork			do_fork
391  *	copy_thread		copy_thread
392  *
393  * This means that the stack layout is as follows:
394  *
395  *	+---------------------+ (highest addr)
396  *	|   struct pt_regs    |
397  *	+---------------------+
398  *	| struct switch_stack |
399  *	+---------------------+
400  *	|                     |
401  *	|    memory stack     |
402  *	|                     | <-- sp (lowest addr)
403  *	+---------------------+
404  *
405  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
406  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
407  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
408  * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
409  * the stack is page aligned and the page size is at least 4KB, this is always the case,
410  * so there is nothing to worry about.
411  */
412 int
413 copy_thread(unsigned long clone_flags,
414 	     unsigned long user_stack_base, unsigned long user_stack_size,
415 	     struct task_struct *p, struct pt_regs *regs)
416 {
417 	extern char ia64_ret_from_clone;
418 	struct switch_stack *child_stack, *stack;
419 	unsigned long rbs, child_rbs, rbs_size;
420 	struct pt_regs *child_ptregs;
421 	int retval = 0;
422 
423 #ifdef CONFIG_SMP
424 	/*
425 	 * For SMP idle threads, fork_by_hand() calls do_fork with
426 	 * NULL regs.
427 	 */
428 	if (!regs)
429 		return 0;
430 #endif
431 
432 	stack = ((struct switch_stack *) regs) - 1;
433 
434 	child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
435 	child_stack = (struct switch_stack *) child_ptregs - 1;
436 
437 	/* copy parent's switch_stack & pt_regs to child: */
438 	memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
439 
440 	rbs = (unsigned long) current + IA64_RBS_OFFSET;
441 	child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
442 	rbs_size = stack->ar_bspstore - rbs;
443 
444 	/* copy the parent's register backing store to the child: */
445 	memcpy((void *) child_rbs, (void *) rbs, rbs_size);
446 
447 	if (likely(user_mode(child_ptregs))) {
448 		if (clone_flags & CLONE_SETTLS)
449 			child_ptregs->r13 = regs->r16;	/* see sys_clone2() in entry.S */
450 		if (user_stack_base) {
451 			child_ptregs->r12 = user_stack_base + user_stack_size - 16;
452 			child_ptregs->ar_bspstore = user_stack_base;
453 			child_ptregs->ar_rnat = 0;
454 			child_ptregs->loadrs = 0;
455 		}
456 	} else {
457 		/*
458 		 * Note: we simply preserve the relative position of
459 		 * the stack pointer here.  There is no need to
460 		 * allocate a scratch area here, since that will have
461 		 * been taken care of by the caller of sys_clone()
462 		 * already.
463 		 */
464 		child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
465 		child_ptregs->r13 = (unsigned long) p;		/* set `current' pointer */
466 	}
467 	child_stack->ar_bspstore = child_rbs + rbs_size;
468 	child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
469 
470 	/* copy parts of thread_struct: */
471 	p->thread.ksp = (unsigned long) child_stack - 16;
472 
473 	/* stop some PSR bits from being inherited.
474 	 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
475 	 * therefore we must specify them explicitly here and not include them in
476 	 * IA64_PSR_BITS_TO_CLEAR.
477 	 */
478 	child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
479 				 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
480 
481 	/*
482 	 * NOTE: The calling convention considers all floating point
483 	 * registers in the high partition (fph) to be scratch.  Since
484 	 * the only way to get to this point is through a system call,
485 	 * we know that the values in fph are all dead.  Hence, there
486 	 * is no need to inherit the fph state from the parent to the
487 	 * child and all we have to do is to make sure that
488 	 * IA64_THREAD_FPH_VALID is cleared in the child.
489 	 *
490 	 * XXX We could push this optimization a bit further by
491 	 * clearing IA64_THREAD_FPH_VALID on ANY system call.
492 	 * However, it's not clear this is worth doing.  Also, it
493 	 * would be a slight deviation from the normal Linux system
494 	 * call behavior where scratch registers are preserved across
495 	 * system calls (unless used by the system call itself).
496 	 */
497 #	define THREAD_FLAGS_TO_CLEAR	(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
498 					 | IA64_THREAD_PM_VALID)
499 #	define THREAD_FLAGS_TO_SET	0
500 	p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
501 			   | THREAD_FLAGS_TO_SET);
502 	ia64_drop_fpu(p);	/* don't pick up stale state from a CPU's fph */
503 
504 #ifdef CONFIG_PERFMON
505 	if (current->thread.pfm_context)
506 		pfm_inherit(p, child_ptregs);
507 #endif
508 	return retval;
509 }
510 
511 static void
512 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
513 {
514 	unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
515 	unsigned long uninitialized_var(ip);	/* GCC be quiet */
516 	elf_greg_t *dst = arg;
517 	struct pt_regs *pt;
518 	char nat;
519 	int i;
520 
521 	memset(dst, 0, sizeof(elf_gregset_t));	/* don't leak any kernel bits to user-level */
522 
523 	if (unw_unwind_to_user(info) < 0)
524 		return;
525 
526 	unw_get_sp(info, &sp);
527 	pt = (struct pt_regs *) (sp + 16);
528 
529 	urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
530 
531 	if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
532 		return;
533 
534 	ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
535 		  &ar_rnat);
536 
537 	/*
538 	 * coredump format:
539 	 *	r0-r31
540 	 *	NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
541 	 *	predicate registers (p0-p63)
542 	 *	b0-b7
543 	 *	ip cfm user-mask
544 	 *	ar.rsc ar.bsp ar.bspstore ar.rnat
545 	 *	ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
546 	 */
547 
548 	/* r0 is zero */
549 	for (i = 1, mask = (1UL << i); i < 32; ++i) {
550 		unw_get_gr(info, i, &dst[i], &nat);
551 		if (nat)
552 			nat_bits |= mask;
553 		mask <<= 1;
554 	}
555 	dst[32] = nat_bits;
556 	unw_get_pr(info, &dst[33]);
557 
558 	for (i = 0; i < 8; ++i)
559 		unw_get_br(info, i, &dst[34 + i]);
560 
561 	unw_get_rp(info, &ip);
562 	dst[42] = ip + ia64_psr(pt)->ri;
563 	dst[43] = cfm;
564 	dst[44] = pt->cr_ipsr & IA64_PSR_UM;
565 
566 	unw_get_ar(info, UNW_AR_RSC, &dst[45]);
567 	/*
568 	 * For bsp and bspstore, unw_get_ar() would return the kernel
569 	 * addresses, but we need the user-level addresses instead:
570 	 */
571 	dst[46] = urbs_end;	/* note: by convention PT_AR_BSP points to the end of the urbs! */
572 	dst[47] = pt->ar_bspstore;
573 	dst[48] = ar_rnat;
574 	unw_get_ar(info, UNW_AR_CCV, &dst[49]);
575 	unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
576 	unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
577 	dst[52] = pt->ar_pfs;	/* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
578 	unw_get_ar(info, UNW_AR_LC, &dst[53]);
579 	unw_get_ar(info, UNW_AR_EC, &dst[54]);
580 	unw_get_ar(info, UNW_AR_CSD, &dst[55]);
581 	unw_get_ar(info, UNW_AR_SSD, &dst[56]);
582 }
583 
584 void
585 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
586 {
587 	elf_fpreg_t *dst = arg;
588 	int i;
589 
590 	memset(dst, 0, sizeof(elf_fpregset_t));	/* don't leak any "random" bits */
591 
592 	if (unw_unwind_to_user(info) < 0)
593 		return;
594 
595 	/* f0 is 0.0, f1 is 1.0 */
596 
597 	for (i = 2; i < 32; ++i)
598 		unw_get_fr(info, i, dst + i);
599 
600 	ia64_flush_fph(task);
601 	if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
602 		memcpy(dst + 32, task->thread.fph, 96*16);
603 }
604 
605 void
606 do_copy_regs (struct unw_frame_info *info, void *arg)
607 {
608 	do_copy_task_regs(current, info, arg);
609 }
610 
611 void
612 do_dump_fpu (struct unw_frame_info *info, void *arg)
613 {
614 	do_dump_task_fpu(current, info, arg);
615 }
616 
617 void
618 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
619 {
620 	unw_init_running(do_copy_regs, dst);
621 }
622 
623 int
624 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
625 {
626 	unw_init_running(do_dump_fpu, dst);
627 	return 1;	/* f0-f31 are always valid so we always return 1 */
628 }
629 
630 long
631 sys_execve (const char __user *filename,
632 	    const char __user *const __user *argv,
633 	    const char __user *const __user *envp,
634 	    struct pt_regs *regs)
635 {
636 	char *fname;
637 	int error;
638 
639 	fname = getname(filename);
640 	error = PTR_ERR(fname);
641 	if (IS_ERR(fname))
642 		goto out;
643 	error = do_execve(fname, argv, envp, regs);
644 	putname(fname);
645 out:
646 	return error;
647 }
648 
649 pid_t
650 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
651 {
652 	extern void start_kernel_thread (void);
653 	unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
654 	struct {
655 		struct switch_stack sw;
656 		struct pt_regs pt;
657 	} regs;
658 
659 	memset(&regs, 0, sizeof(regs));
660 	regs.pt.cr_iip = helper_fptr[0];	/* set entry point (IP) */
661 	regs.pt.r1 = helper_fptr[1];		/* set GP */
662 	regs.pt.r9 = (unsigned long) fn;	/* 1st argument */
663 	regs.pt.r11 = (unsigned long) arg;	/* 2nd argument */
664 	/* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
665 	regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
666 	regs.pt.cr_ifs = 1UL << 63;		/* mark as valid, empty frame */
667 	regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
668 	regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
669 	regs.sw.pr = (1 << PRED_KERNEL_STACK);
670 	return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
671 }
672 EXPORT_SYMBOL(kernel_thread);
673 
674 /* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
675 int
676 kernel_thread_helper (int (*fn)(void *), void *arg)
677 {
678 	return (*fn)(arg);
679 }
680 
681 /*
682  * Flush thread state.  This is called when a thread does an execve().
683  */
684 void
685 flush_thread (void)
686 {
687 	/* drop floating-point and debug-register state if it exists: */
688 	current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
689 	ia64_drop_fpu(current);
690 }
691 
692 /*
693  * Clean up state associated with current thread.  This is called when
694  * the thread calls exit().
695  */
696 void
697 exit_thread (void)
698 {
699 
700 	ia64_drop_fpu(current);
701 #ifdef CONFIG_PERFMON
702        /* if needed, stop monitoring and flush state to perfmon context */
703 	if (current->thread.pfm_context)
704 		pfm_exit_thread(current);
705 
706 	/* free debug register resources */
707 	if (current->thread.flags & IA64_THREAD_DBG_VALID)
708 		pfm_release_debug_registers(current);
709 #endif
710 }
711 
712 unsigned long
713 get_wchan (struct task_struct *p)
714 {
715 	struct unw_frame_info info;
716 	unsigned long ip;
717 	int count = 0;
718 
719 	if (!p || p == current || p->state == TASK_RUNNING)
720 		return 0;
721 
722 	/*
723 	 * Note: p may not be a blocked task (it could be current or
724 	 * another process running on some other CPU.  Rather than
725 	 * trying to determine if p is really blocked, we just assume
726 	 * it's blocked and rely on the unwind routines to fail
727 	 * gracefully if the process wasn't really blocked after all.
728 	 * --davidm 99/12/15
729 	 */
730 	unw_init_from_blocked_task(&info, p);
731 	do {
732 		if (p->state == TASK_RUNNING)
733 			return 0;
734 		if (unw_unwind(&info) < 0)
735 			return 0;
736 		unw_get_ip(&info, &ip);
737 		if (!in_sched_functions(ip))
738 			return ip;
739 	} while (count++ < 16);
740 	return 0;
741 }
742 
743 void
744 cpu_halt (void)
745 {
746 	pal_power_mgmt_info_u_t power_info[8];
747 	unsigned long min_power;
748 	int i, min_power_state;
749 
750 	if (ia64_pal_halt_info(power_info) != 0)
751 		return;
752 
753 	min_power_state = 0;
754 	min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
755 	for (i = 1; i < 8; ++i)
756 		if (power_info[i].pal_power_mgmt_info_s.im
757 		    && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
758 			min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
759 			min_power_state = i;
760 		}
761 
762 	while (1)
763 		ia64_pal_halt(min_power_state);
764 }
765 
766 void machine_shutdown(void)
767 {
768 #ifdef CONFIG_HOTPLUG_CPU
769 	int cpu;
770 
771 	for_each_online_cpu(cpu) {
772 		if (cpu != smp_processor_id())
773 			cpu_down(cpu);
774 	}
775 #endif
776 #ifdef CONFIG_KEXEC
777 	kexec_disable_iosapic();
778 #endif
779 }
780 
781 void
782 machine_restart (char *restart_cmd)
783 {
784 	(void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
785 	(*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
786 }
787 
788 void
789 machine_halt (void)
790 {
791 	(void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
792 	cpu_halt();
793 }
794 
795 void
796 machine_power_off (void)
797 {
798 	if (pm_power_off)
799 		pm_power_off();
800 	machine_halt();
801 }
802 
803