xref: /openbmc/linux/arch/ia64/kernel/process.c (revision bb8416bf8b93d88e23cbbfde962ef85acda5bd5f)
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/module.h>
19 #include <linux/notifier.h>
20 #include <linux/personality.h>
21 #include <linux/sched.h>
22 #include <linux/slab.h>
23 #include <linux/smp_lock.h>
24 #include <linux/stddef.h>
25 #include <linux/thread_info.h>
26 #include <linux/unistd.h>
27 #include <linux/efi.h>
28 #include <linux/interrupt.h>
29 #include <linux/delay.h>
30 
31 #include <asm/cpu.h>
32 #include <asm/delay.h>
33 #include <asm/elf.h>
34 #include <asm/ia32.h>
35 #include <asm/irq.h>
36 #include <asm/kdebug.h>
37 #include <asm/kexec.h>
38 #include <asm/pgalloc.h>
39 #include <asm/processor.h>
40 #include <asm/sal.h>
41 #include <asm/tlbflush.h>
42 #include <asm/uaccess.h>
43 #include <asm/unwind.h>
44 #include <asm/user.h>
45 
46 #include "entry.h"
47 
48 #ifdef CONFIG_PERFMON
49 # include <asm/perfmon.h>
50 #endif
51 
52 #include "sigframe.h"
53 
54 void (*ia64_mark_idle)(int);
55 static DEFINE_PER_CPU(unsigned int, cpu_idle_state);
56 
57 unsigned long boot_option_idle_override = 0;
58 EXPORT_SYMBOL(boot_option_idle_override);
59 
60 void
61 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
62 {
63 	unsigned long ip, sp, bsp;
64 	char buf[128];			/* don't make it so big that it overflows the stack! */
65 
66 	printk("\nCall Trace:\n");
67 	do {
68 		unw_get_ip(info, &ip);
69 		if (ip == 0)
70 			break;
71 
72 		unw_get_sp(info, &sp);
73 		unw_get_bsp(info, &bsp);
74 		snprintf(buf, sizeof(buf),
75 			 " [<%016lx>] %%s\n"
76 			 "                                sp=%016lx bsp=%016lx\n",
77 			 ip, sp, bsp);
78 		print_symbol(buf, ip);
79 	} while (unw_unwind(info) >= 0);
80 }
81 
82 void
83 show_stack (struct task_struct *task, unsigned long *sp)
84 {
85 	if (!task)
86 		unw_init_running(ia64_do_show_stack, NULL);
87 	else {
88 		struct unw_frame_info info;
89 
90 		unw_init_from_blocked_task(&info, task);
91 		ia64_do_show_stack(&info, NULL);
92 	}
93 }
94 
95 void
96 dump_stack (void)
97 {
98 	show_stack(NULL, NULL);
99 }
100 
101 EXPORT_SYMBOL(dump_stack);
102 
103 void
104 show_regs (struct pt_regs *regs)
105 {
106 	unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
107 
108 	print_modules();
109 	printk("\nPid: %d, CPU %d, comm: %20s\n", current->pid, smp_processor_id(), current->comm);
110 	printk("psr : %016lx ifs : %016lx ip  : [<%016lx>]    %s\n",
111 	       regs->cr_ipsr, regs->cr_ifs, ip, print_tainted());
112 	print_symbol("ip is at %s\n", ip);
113 	printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
114 	       regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
115 	printk("rnat: %016lx bsps: %016lx pr  : %016lx\n",
116 	       regs->ar_rnat, regs->ar_bspstore, regs->pr);
117 	printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
118 	       regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
119 	printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
120 	printk("b0  : %016lx b6  : %016lx b7  : %016lx\n", regs->b0, regs->b6, regs->b7);
121 	printk("f6  : %05lx%016lx f7  : %05lx%016lx\n",
122 	       regs->f6.u.bits[1], regs->f6.u.bits[0],
123 	       regs->f7.u.bits[1], regs->f7.u.bits[0]);
124 	printk("f8  : %05lx%016lx f9  : %05lx%016lx\n",
125 	       regs->f8.u.bits[1], regs->f8.u.bits[0],
126 	       regs->f9.u.bits[1], regs->f9.u.bits[0]);
127 	printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
128 	       regs->f10.u.bits[1], regs->f10.u.bits[0],
129 	       regs->f11.u.bits[1], regs->f11.u.bits[0]);
130 
131 	printk("r1  : %016lx r2  : %016lx r3  : %016lx\n", regs->r1, regs->r2, regs->r3);
132 	printk("r8  : %016lx r9  : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
133 	printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
134 	printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
135 	printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
136 	printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
137 	printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
138 	printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
139 	printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
140 
141 	if (user_mode(regs)) {
142 		/* print the stacked registers */
143 		unsigned long val, *bsp, ndirty;
144 		int i, sof, is_nat = 0;
145 
146 		sof = regs->cr_ifs & 0x7f;	/* size of frame */
147 		ndirty = (regs->loadrs >> 19);
148 		bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
149 		for (i = 0; i < sof; ++i) {
150 			get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
151 			printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
152 			       ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
153 		}
154 	} else
155 		show_stack(NULL, NULL);
156 }
157 
158 void
159 do_notify_resume_user (sigset_t *oldset, struct sigscratch *scr, long in_syscall)
160 {
161 	if (fsys_mode(current, &scr->pt)) {
162 		/* defer signal-handling etc. until we return to privilege-level 0.  */
163 		if (!ia64_psr(&scr->pt)->lp)
164 			ia64_psr(&scr->pt)->lp = 1;
165 		return;
166 	}
167 
168 #ifdef CONFIG_PERFMON
169 	if (current->thread.pfm_needs_checking)
170 		pfm_handle_work();
171 #endif
172 
173 	/* deal with pending signal delivery */
174 	if (test_thread_flag(TIF_SIGPENDING))
175 		ia64_do_signal(oldset, scr, in_syscall);
176 }
177 
178 static int pal_halt        = 1;
179 static int can_do_pal_halt = 1;
180 
181 static int __init nohalt_setup(char * str)
182 {
183 	pal_halt = can_do_pal_halt = 0;
184 	return 1;
185 }
186 __setup("nohalt", nohalt_setup);
187 
188 void
189 update_pal_halt_status(int status)
190 {
191 	can_do_pal_halt = pal_halt && status;
192 }
193 
194 /*
195  * We use this if we don't have any better idle routine..
196  */
197 void
198 default_idle (void)
199 {
200 	local_irq_enable();
201 	while (!need_resched()) {
202 		if (can_do_pal_halt)
203 			safe_halt();
204 		else
205 			cpu_relax();
206 	}
207 }
208 
209 #ifdef CONFIG_HOTPLUG_CPU
210 /* We don't actually take CPU down, just spin without interrupts. */
211 static inline void play_dead(void)
212 {
213 	extern void ia64_cpu_local_tick (void);
214 	unsigned int this_cpu = smp_processor_id();
215 
216 	/* Ack it */
217 	__get_cpu_var(cpu_state) = CPU_DEAD;
218 
219 	max_xtp();
220 	local_irq_disable();
221 	idle_task_exit();
222 	ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
223 	/*
224 	 * The above is a point of no-return, the processor is
225 	 * expected to be in SAL loop now.
226 	 */
227 	BUG();
228 }
229 #else
230 static inline void play_dead(void)
231 {
232 	BUG();
233 }
234 #endif /* CONFIG_HOTPLUG_CPU */
235 
236 void cpu_idle_wait(void)
237 {
238 	unsigned int cpu, this_cpu = get_cpu();
239 	cpumask_t map;
240 	cpumask_t tmp = current->cpus_allowed;
241 
242 	set_cpus_allowed(current, cpumask_of_cpu(this_cpu));
243 	put_cpu();
244 
245 	cpus_clear(map);
246 	for_each_online_cpu(cpu) {
247 		per_cpu(cpu_idle_state, cpu) = 1;
248 		cpu_set(cpu, map);
249 	}
250 
251 	__get_cpu_var(cpu_idle_state) = 0;
252 
253 	wmb();
254 	do {
255 		ssleep(1);
256 		for_each_online_cpu(cpu) {
257 			if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu))
258 				cpu_clear(cpu, map);
259 		}
260 		cpus_and(map, map, cpu_online_map);
261 	} while (!cpus_empty(map));
262 	set_cpus_allowed(current, tmp);
263 }
264 EXPORT_SYMBOL_GPL(cpu_idle_wait);
265 
266 void __attribute__((noreturn))
267 cpu_idle (void)
268 {
269 	void (*mark_idle)(int) = ia64_mark_idle;
270   	int cpu = smp_processor_id();
271 
272 	/* endless idle loop with no priority at all */
273 	while (1) {
274 		if (can_do_pal_halt) {
275 			current_thread_info()->status &= ~TS_POLLING;
276 			/*
277 			 * TS_POLLING-cleared state must be visible before we
278 			 * test NEED_RESCHED:
279 			 */
280 			smp_mb();
281 		} else {
282 			current_thread_info()->status |= TS_POLLING;
283 		}
284 
285 		if (!need_resched()) {
286 			void (*idle)(void);
287 #ifdef CONFIG_SMP
288 			min_xtp();
289 #endif
290 			if (__get_cpu_var(cpu_idle_state))
291 				__get_cpu_var(cpu_idle_state) = 0;
292 
293 			rmb();
294 			if (mark_idle)
295 				(*mark_idle)(1);
296 
297 			idle = pm_idle;
298 			if (!idle)
299 				idle = default_idle;
300 			(*idle)();
301 			if (mark_idle)
302 				(*mark_idle)(0);
303 #ifdef CONFIG_SMP
304 			normal_xtp();
305 #endif
306 		}
307 		preempt_enable_no_resched();
308 		schedule();
309 		preempt_disable();
310 		check_pgt_cache();
311 		if (cpu_is_offline(cpu))
312 			play_dead();
313 	}
314 }
315 
316 void
317 ia64_save_extra (struct task_struct *task)
318 {
319 #ifdef CONFIG_PERFMON
320 	unsigned long info;
321 #endif
322 
323 	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
324 		ia64_save_debug_regs(&task->thread.dbr[0]);
325 
326 #ifdef CONFIG_PERFMON
327 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
328 		pfm_save_regs(task);
329 
330 	info = __get_cpu_var(pfm_syst_info);
331 	if (info & PFM_CPUINFO_SYST_WIDE)
332 		pfm_syst_wide_update_task(task, info, 0);
333 #endif
334 
335 #ifdef CONFIG_IA32_SUPPORT
336 	if (IS_IA32_PROCESS(task_pt_regs(task)))
337 		ia32_save_state(task);
338 #endif
339 }
340 
341 void
342 ia64_load_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_load_debug_regs(&task->thread.dbr[0]);
350 
351 #ifdef CONFIG_PERFMON
352 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
353 		pfm_load_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, 1);
358 #endif
359 
360 #ifdef CONFIG_IA32_SUPPORT
361 	if (IS_IA32_PROCESS(task_pt_regs(task)))
362 		ia32_load_state(task);
363 #endif
364 }
365 
366 /*
367  * Copy the state of an ia-64 thread.
368  *
369  * We get here through the following  call chain:
370  *
371  *	from user-level:	from kernel:
372  *
373  *	<clone syscall>	        <some kernel call frames>
374  *	sys_clone		   :
375  *	do_fork			do_fork
376  *	copy_thread		copy_thread
377  *
378  * This means that the stack layout is as follows:
379  *
380  *	+---------------------+ (highest addr)
381  *	|   struct pt_regs    |
382  *	+---------------------+
383  *	| struct switch_stack |
384  *	+---------------------+
385  *	|                     |
386  *	|    memory stack     |
387  *	|                     | <-- sp (lowest addr)
388  *	+---------------------+
389  *
390  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
391  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
392  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
393  * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
394  * the stack is page aligned and the page size is at least 4KB, this is always the case,
395  * so there is nothing to worry about.
396  */
397 int
398 copy_thread (int nr, unsigned long clone_flags,
399 	     unsigned long user_stack_base, unsigned long user_stack_size,
400 	     struct task_struct *p, struct pt_regs *regs)
401 {
402 	extern char ia64_ret_from_clone, ia32_ret_from_clone;
403 	struct switch_stack *child_stack, *stack;
404 	unsigned long rbs, child_rbs, rbs_size;
405 	struct pt_regs *child_ptregs;
406 	int retval = 0;
407 
408 #ifdef CONFIG_SMP
409 	/*
410 	 * For SMP idle threads, fork_by_hand() calls do_fork with
411 	 * NULL regs.
412 	 */
413 	if (!regs)
414 		return 0;
415 #endif
416 
417 	stack = ((struct switch_stack *) regs) - 1;
418 
419 	child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
420 	child_stack = (struct switch_stack *) child_ptregs - 1;
421 
422 	/* copy parent's switch_stack & pt_regs to child: */
423 	memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
424 
425 	rbs = (unsigned long) current + IA64_RBS_OFFSET;
426 	child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
427 	rbs_size = stack->ar_bspstore - rbs;
428 
429 	/* copy the parent's register backing store to the child: */
430 	memcpy((void *) child_rbs, (void *) rbs, rbs_size);
431 
432 	if (likely(user_mode(child_ptregs))) {
433 		if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs))
434 			child_ptregs->r13 = regs->r16;	/* see sys_clone2() in entry.S */
435 		if (user_stack_base) {
436 			child_ptregs->r12 = user_stack_base + user_stack_size - 16;
437 			child_ptregs->ar_bspstore = user_stack_base;
438 			child_ptregs->ar_rnat = 0;
439 			child_ptregs->loadrs = 0;
440 		}
441 	} else {
442 		/*
443 		 * Note: we simply preserve the relative position of
444 		 * the stack pointer here.  There is no need to
445 		 * allocate a scratch area here, since that will have
446 		 * been taken care of by the caller of sys_clone()
447 		 * already.
448 		 */
449 		child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
450 		child_ptregs->r13 = (unsigned long) p;		/* set `current' pointer */
451 	}
452 	child_stack->ar_bspstore = child_rbs + rbs_size;
453 	if (IS_IA32_PROCESS(regs))
454 		child_stack->b0 = (unsigned long) &ia32_ret_from_clone;
455 	else
456 		child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
457 
458 	/* copy parts of thread_struct: */
459 	p->thread.ksp = (unsigned long) child_stack - 16;
460 
461 	/* stop some PSR bits from being inherited.
462 	 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
463 	 * therefore we must specify them explicitly here and not include them in
464 	 * IA64_PSR_BITS_TO_CLEAR.
465 	 */
466 	child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
467 				 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
468 
469 	/*
470 	 * NOTE: The calling convention considers all floating point
471 	 * registers in the high partition (fph) to be scratch.  Since
472 	 * the only way to get to this point is through a system call,
473 	 * we know that the values in fph are all dead.  Hence, there
474 	 * is no need to inherit the fph state from the parent to the
475 	 * child and all we have to do is to make sure that
476 	 * IA64_THREAD_FPH_VALID is cleared in the child.
477 	 *
478 	 * XXX We could push this optimization a bit further by
479 	 * clearing IA64_THREAD_FPH_VALID on ANY system call.
480 	 * However, it's not clear this is worth doing.  Also, it
481 	 * would be a slight deviation from the normal Linux system
482 	 * call behavior where scratch registers are preserved across
483 	 * system calls (unless used by the system call itself).
484 	 */
485 #	define THREAD_FLAGS_TO_CLEAR	(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
486 					 | IA64_THREAD_PM_VALID)
487 #	define THREAD_FLAGS_TO_SET	0
488 	p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
489 			   | THREAD_FLAGS_TO_SET);
490 	ia64_drop_fpu(p);	/* don't pick up stale state from a CPU's fph */
491 #ifdef CONFIG_IA32_SUPPORT
492 	/*
493 	 * If we're cloning an IA32 task then save the IA32 extra
494 	 * state from the current task to the new task
495 	 */
496 	if (IS_IA32_PROCESS(task_pt_regs(current))) {
497 		ia32_save_state(p);
498 		if (clone_flags & CLONE_SETTLS)
499 			retval = ia32_clone_tls(p, child_ptregs);
500 
501 		/* Copy partially mapped page list */
502 		if (!retval)
503 			retval = ia32_copy_partial_page_list(p, clone_flags);
504 	}
505 #endif
506 
507 #ifdef CONFIG_PERFMON
508 	if (current->thread.pfm_context)
509 		pfm_inherit(p, child_ptregs);
510 #endif
511 	return retval;
512 }
513 
514 static void
515 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
516 {
517 	unsigned long mask, sp, nat_bits = 0, ip, ar_rnat, urbs_end, cfm;
518 	elf_greg_t *dst = arg;
519 	struct pt_regs *pt;
520 	char nat;
521 	int i;
522 
523 	memset(dst, 0, sizeof(elf_gregset_t));	/* don't leak any kernel bits to user-level */
524 
525 	if (unw_unwind_to_user(info) < 0)
526 		return;
527 
528 	unw_get_sp(info, &sp);
529 	pt = (struct pt_regs *) (sp + 16);
530 
531 	urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
532 
533 	if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
534 		return;
535 
536 	ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
537 		  &ar_rnat);
538 
539 	/*
540 	 * coredump format:
541 	 *	r0-r31
542 	 *	NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
543 	 *	predicate registers (p0-p63)
544 	 *	b0-b7
545 	 *	ip cfm user-mask
546 	 *	ar.rsc ar.bsp ar.bspstore ar.rnat
547 	 *	ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
548 	 */
549 
550 	/* r0 is zero */
551 	for (i = 1, mask = (1UL << i); i < 32; ++i) {
552 		unw_get_gr(info, i, &dst[i], &nat);
553 		if (nat)
554 			nat_bits |= mask;
555 		mask <<= 1;
556 	}
557 	dst[32] = nat_bits;
558 	unw_get_pr(info, &dst[33]);
559 
560 	for (i = 0; i < 8; ++i)
561 		unw_get_br(info, i, &dst[34 + i]);
562 
563 	unw_get_rp(info, &ip);
564 	dst[42] = ip + ia64_psr(pt)->ri;
565 	dst[43] = cfm;
566 	dst[44] = pt->cr_ipsr & IA64_PSR_UM;
567 
568 	unw_get_ar(info, UNW_AR_RSC, &dst[45]);
569 	/*
570 	 * For bsp and bspstore, unw_get_ar() would return the kernel
571 	 * addresses, but we need the user-level addresses instead:
572 	 */
573 	dst[46] = urbs_end;	/* note: by convention PT_AR_BSP points to the end of the urbs! */
574 	dst[47] = pt->ar_bspstore;
575 	dst[48] = ar_rnat;
576 	unw_get_ar(info, UNW_AR_CCV, &dst[49]);
577 	unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
578 	unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
579 	dst[52] = pt->ar_pfs;	/* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
580 	unw_get_ar(info, UNW_AR_LC, &dst[53]);
581 	unw_get_ar(info, UNW_AR_EC, &dst[54]);
582 	unw_get_ar(info, UNW_AR_CSD, &dst[55]);
583 	unw_get_ar(info, UNW_AR_SSD, &dst[56]);
584 }
585 
586 void
587 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
588 {
589 	elf_fpreg_t *dst = arg;
590 	int i;
591 
592 	memset(dst, 0, sizeof(elf_fpregset_t));	/* don't leak any "random" bits */
593 
594 	if (unw_unwind_to_user(info) < 0)
595 		return;
596 
597 	/* f0 is 0.0, f1 is 1.0 */
598 
599 	for (i = 2; i < 32; ++i)
600 		unw_get_fr(info, i, dst + i);
601 
602 	ia64_flush_fph(task);
603 	if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
604 		memcpy(dst + 32, task->thread.fph, 96*16);
605 }
606 
607 void
608 do_copy_regs (struct unw_frame_info *info, void *arg)
609 {
610 	do_copy_task_regs(current, info, arg);
611 }
612 
613 void
614 do_dump_fpu (struct unw_frame_info *info, void *arg)
615 {
616 	do_dump_task_fpu(current, info, arg);
617 }
618 
619 int
620 dump_task_regs(struct task_struct *task, elf_gregset_t *regs)
621 {
622 	struct unw_frame_info tcore_info;
623 
624 	if (current == task) {
625 		unw_init_running(do_copy_regs, regs);
626 	} else {
627 		memset(&tcore_info, 0, sizeof(tcore_info));
628 		unw_init_from_blocked_task(&tcore_info, task);
629 		do_copy_task_regs(task, &tcore_info, regs);
630 	}
631 	return 1;
632 }
633 
634 void
635 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
636 {
637 	unw_init_running(do_copy_regs, dst);
638 }
639 
640 int
641 dump_task_fpu (struct task_struct *task, elf_fpregset_t *dst)
642 {
643 	struct unw_frame_info tcore_info;
644 
645 	if (current == task) {
646 		unw_init_running(do_dump_fpu, dst);
647 	} else {
648 		memset(&tcore_info, 0, sizeof(tcore_info));
649 		unw_init_from_blocked_task(&tcore_info, task);
650 		do_dump_task_fpu(task, &tcore_info, dst);
651 	}
652 	return 1;
653 }
654 
655 int
656 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
657 {
658 	unw_init_running(do_dump_fpu, dst);
659 	return 1;	/* f0-f31 are always valid so we always return 1 */
660 }
661 
662 long
663 sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp,
664 	    struct pt_regs *regs)
665 {
666 	char *fname;
667 	int error;
668 
669 	fname = getname(filename);
670 	error = PTR_ERR(fname);
671 	if (IS_ERR(fname))
672 		goto out;
673 	error = do_execve(fname, argv, envp, regs);
674 	putname(fname);
675 out:
676 	return error;
677 }
678 
679 pid_t
680 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
681 {
682 	extern void start_kernel_thread (void);
683 	unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
684 	struct {
685 		struct switch_stack sw;
686 		struct pt_regs pt;
687 	} regs;
688 
689 	memset(&regs, 0, sizeof(regs));
690 	regs.pt.cr_iip = helper_fptr[0];	/* set entry point (IP) */
691 	regs.pt.r1 = helper_fptr[1];		/* set GP */
692 	regs.pt.r9 = (unsigned long) fn;	/* 1st argument */
693 	regs.pt.r11 = (unsigned long) arg;	/* 2nd argument */
694 	/* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
695 	regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
696 	regs.pt.cr_ifs = 1UL << 63;		/* mark as valid, empty frame */
697 	regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
698 	regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
699 	regs.sw.pr = (1 << PRED_KERNEL_STACK);
700 	return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
701 }
702 EXPORT_SYMBOL(kernel_thread);
703 
704 /* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
705 int
706 kernel_thread_helper (int (*fn)(void *), void *arg)
707 {
708 #ifdef CONFIG_IA32_SUPPORT
709 	if (IS_IA32_PROCESS(task_pt_regs(current))) {
710 		/* A kernel thread is always a 64-bit process. */
711 		current->thread.map_base  = DEFAULT_MAP_BASE;
712 		current->thread.task_size = DEFAULT_TASK_SIZE;
713 		ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
714 		ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
715 	}
716 #endif
717 	return (*fn)(arg);
718 }
719 
720 /*
721  * Flush thread state.  This is called when a thread does an execve().
722  */
723 void
724 flush_thread (void)
725 {
726 	/* drop floating-point and debug-register state if it exists: */
727 	current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
728 	ia64_drop_fpu(current);
729 #ifdef CONFIG_IA32_SUPPORT
730 	if (IS_IA32_PROCESS(task_pt_regs(current))) {
731 		ia32_drop_partial_page_list(current);
732 		current->thread.task_size = IA32_PAGE_OFFSET;
733 		set_fs(USER_DS);
734 	}
735 #endif
736 }
737 
738 /*
739  * Clean up state associated with current thread.  This is called when
740  * the thread calls exit().
741  */
742 void
743 exit_thread (void)
744 {
745 
746 	ia64_drop_fpu(current);
747 #ifdef CONFIG_PERFMON
748        /* if needed, stop monitoring and flush state to perfmon context */
749 	if (current->thread.pfm_context)
750 		pfm_exit_thread(current);
751 
752 	/* free debug register resources */
753 	if (current->thread.flags & IA64_THREAD_DBG_VALID)
754 		pfm_release_debug_registers(current);
755 #endif
756 	if (IS_IA32_PROCESS(task_pt_regs(current)))
757 		ia32_drop_partial_page_list(current);
758 }
759 
760 unsigned long
761 get_wchan (struct task_struct *p)
762 {
763 	struct unw_frame_info info;
764 	unsigned long ip;
765 	int count = 0;
766 
767 	/*
768 	 * Note: p may not be a blocked task (it could be current or
769 	 * another process running on some other CPU.  Rather than
770 	 * trying to determine if p is really blocked, we just assume
771 	 * it's blocked and rely on the unwind routines to fail
772 	 * gracefully if the process wasn't really blocked after all.
773 	 * --davidm 99/12/15
774 	 */
775 	unw_init_from_blocked_task(&info, p);
776 	do {
777 		if (unw_unwind(&info) < 0)
778 			return 0;
779 		unw_get_ip(&info, &ip);
780 		if (!in_sched_functions(ip))
781 			return ip;
782 	} while (count++ < 16);
783 	return 0;
784 }
785 
786 void
787 cpu_halt (void)
788 {
789 	pal_power_mgmt_info_u_t power_info[8];
790 	unsigned long min_power;
791 	int i, min_power_state;
792 
793 	if (ia64_pal_halt_info(power_info) != 0)
794 		return;
795 
796 	min_power_state = 0;
797 	min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
798 	for (i = 1; i < 8; ++i)
799 		if (power_info[i].pal_power_mgmt_info_s.im
800 		    && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
801 			min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
802 			min_power_state = i;
803 		}
804 
805 	while (1)
806 		ia64_pal_halt(min_power_state);
807 }
808 
809 void machine_shutdown(void)
810 {
811 #ifdef CONFIG_HOTPLUG_CPU
812 	int cpu;
813 
814 	for_each_online_cpu(cpu) {
815 		if (cpu != smp_processor_id())
816 			cpu_down(cpu);
817 	}
818 #endif
819 #ifdef CONFIG_KEXEC
820 	kexec_disable_iosapic();
821 #endif
822 }
823 
824 void
825 machine_restart (char *restart_cmd)
826 {
827 	(void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
828 	(*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
829 }
830 
831 void
832 machine_halt (void)
833 {
834 	(void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
835 	cpu_halt();
836 }
837 
838 void
839 machine_power_off (void)
840 {
841 	if (pm_power_off)
842 		pm_power_off();
843 	machine_halt();
844 }
845 
846