xref: /openbmc/linux/arch/ia64/kernel/process.c (revision e63340ae6b6205fef26b40a75673d1c9c0c8bb90)
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/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 
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/kexec.h>
37 #include <asm/pgalloc.h>
38 #include <asm/processor.h>
39 #include <asm/sal.h>
40 #include <asm/tlbflush.h>
41 #include <asm/uaccess.h>
42 #include <asm/unwind.h>
43 #include <asm/user.h>
44 
45 #include "entry.h"
46 
47 #ifdef CONFIG_PERFMON
48 # include <asm/perfmon.h>
49 #endif
50 
51 #include "sigframe.h"
52 
53 void (*ia64_mark_idle)(int);
54 static DEFINE_PER_CPU(unsigned int, cpu_idle_state);
55 
56 unsigned long boot_option_idle_override = 0;
57 EXPORT_SYMBOL(boot_option_idle_override);
58 
59 void
60 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
61 {
62 	unsigned long ip, sp, bsp;
63 	char buf[128];			/* don't make it so big that it overflows the stack! */
64 
65 	printk("\nCall Trace:\n");
66 	do {
67 		unw_get_ip(info, &ip);
68 		if (ip == 0)
69 			break;
70 
71 		unw_get_sp(info, &sp);
72 		unw_get_bsp(info, &bsp);
73 		snprintf(buf, sizeof(buf),
74 			 " [<%016lx>] %%s\n"
75 			 "                                sp=%016lx bsp=%016lx\n",
76 			 ip, sp, bsp);
77 		print_symbol(buf, ip);
78 	} while (unw_unwind(info) >= 0);
79 }
80 
81 void
82 show_stack (struct task_struct *task, unsigned long *sp)
83 {
84 	if (!task)
85 		unw_init_running(ia64_do_show_stack, NULL);
86 	else {
87 		struct unw_frame_info info;
88 
89 		unw_init_from_blocked_task(&info, task);
90 		ia64_do_show_stack(&info, NULL);
91 	}
92 }
93 
94 void
95 dump_stack (void)
96 {
97 	show_stack(NULL, NULL);
98 }
99 
100 EXPORT_SYMBOL(dump_stack);
101 
102 void
103 show_regs (struct pt_regs *regs)
104 {
105 	unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
106 
107 	print_modules();
108 	printk("\nPid: %d, CPU %d, comm: %20s\n", current->pid, smp_processor_id(), current->comm);
109 	printk("psr : %016lx ifs : %016lx ip  : [<%016lx>]    %s\n",
110 	       regs->cr_ipsr, regs->cr_ifs, ip, print_tainted());
111 	print_symbol("ip is at %s\n", 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);
155 }
156 
157 void
158 do_notify_resume_user (sigset_t *oldset, struct sigscratch *scr, long in_syscall)
159 {
160 	if (fsys_mode(current, &scr->pt)) {
161 		/* defer signal-handling etc. until we return to privilege-level 0.  */
162 		if (!ia64_psr(&scr->pt)->lp)
163 			ia64_psr(&scr->pt)->lp = 1;
164 		return;
165 	}
166 
167 #ifdef CONFIG_PERFMON
168 	if (current->thread.pfm_needs_checking)
169 		pfm_handle_work();
170 #endif
171 
172 	/* deal with pending signal delivery */
173 	if (test_thread_flag(TIF_SIGPENDING))
174 		ia64_do_signal(oldset, scr, in_syscall);
175 }
176 
177 static int pal_halt        = 1;
178 static int can_do_pal_halt = 1;
179 
180 static int __init nohalt_setup(char * str)
181 {
182 	pal_halt = can_do_pal_halt = 0;
183 	return 1;
184 }
185 __setup("nohalt", nohalt_setup);
186 
187 void
188 update_pal_halt_status(int status)
189 {
190 	can_do_pal_halt = pal_halt && status;
191 }
192 
193 /*
194  * We use this if we don't have any better idle routine..
195  */
196 void
197 default_idle (void)
198 {
199 	local_irq_enable();
200 	while (!need_resched()) {
201 		if (can_do_pal_halt)
202 			safe_halt();
203 		else
204 			cpu_relax();
205 	}
206 }
207 
208 #ifdef CONFIG_HOTPLUG_CPU
209 /* We don't actually take CPU down, just spin without interrupts. */
210 static inline void play_dead(void)
211 {
212 	extern void ia64_cpu_local_tick (void);
213 	unsigned int this_cpu = smp_processor_id();
214 
215 	/* Ack it */
216 	__get_cpu_var(cpu_state) = CPU_DEAD;
217 
218 	max_xtp();
219 	local_irq_disable();
220 	idle_task_exit();
221 	ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
222 	/*
223 	 * The above is a point of no-return, the processor is
224 	 * expected to be in SAL loop now.
225 	 */
226 	BUG();
227 }
228 #else
229 static inline void play_dead(void)
230 {
231 	BUG();
232 }
233 #endif /* CONFIG_HOTPLUG_CPU */
234 
235 void cpu_idle_wait(void)
236 {
237 	unsigned int cpu, this_cpu = get_cpu();
238 	cpumask_t map;
239 
240 	set_cpus_allowed(current, cpumask_of_cpu(this_cpu));
241 	put_cpu();
242 
243 	cpus_clear(map);
244 	for_each_online_cpu(cpu) {
245 		per_cpu(cpu_idle_state, cpu) = 1;
246 		cpu_set(cpu, map);
247 	}
248 
249 	__get_cpu_var(cpu_idle_state) = 0;
250 
251 	wmb();
252 	do {
253 		ssleep(1);
254 		for_each_online_cpu(cpu) {
255 			if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu))
256 				cpu_clear(cpu, map);
257 		}
258 		cpus_and(map, map, cpu_online_map);
259 	} while (!cpus_empty(map));
260 }
261 EXPORT_SYMBOL_GPL(cpu_idle_wait);
262 
263 void __attribute__((noreturn))
264 cpu_idle (void)
265 {
266 	void (*mark_idle)(int) = ia64_mark_idle;
267   	int cpu = smp_processor_id();
268 
269 	/* endless idle loop with no priority at all */
270 	while (1) {
271 		if (can_do_pal_halt) {
272 			current_thread_info()->status &= ~TS_POLLING;
273 			/*
274 			 * TS_POLLING-cleared state must be visible before we
275 			 * test NEED_RESCHED:
276 			 */
277 			smp_mb();
278 		} else {
279 			current_thread_info()->status |= TS_POLLING;
280 		}
281 
282 		if (!need_resched()) {
283 			void (*idle)(void);
284 #ifdef CONFIG_SMP
285 			min_xtp();
286 #endif
287 			if (__get_cpu_var(cpu_idle_state))
288 				__get_cpu_var(cpu_idle_state) = 0;
289 
290 			rmb();
291 			if (mark_idle)
292 				(*mark_idle)(1);
293 
294 			idle = pm_idle;
295 			if (!idle)
296 				idle = default_idle;
297 			(*idle)();
298 			if (mark_idle)
299 				(*mark_idle)(0);
300 #ifdef CONFIG_SMP
301 			normal_xtp();
302 #endif
303 		}
304 		preempt_enable_no_resched();
305 		schedule();
306 		preempt_disable();
307 		check_pgt_cache();
308 		if (cpu_is_offline(cpu))
309 			play_dead();
310 	}
311 }
312 
313 void
314 ia64_save_extra (struct task_struct *task)
315 {
316 #ifdef CONFIG_PERFMON
317 	unsigned long info;
318 #endif
319 
320 	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
321 		ia64_save_debug_regs(&task->thread.dbr[0]);
322 
323 #ifdef CONFIG_PERFMON
324 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
325 		pfm_save_regs(task);
326 
327 	info = __get_cpu_var(pfm_syst_info);
328 	if (info & PFM_CPUINFO_SYST_WIDE)
329 		pfm_syst_wide_update_task(task, info, 0);
330 #endif
331 
332 #ifdef CONFIG_IA32_SUPPORT
333 	if (IS_IA32_PROCESS(task_pt_regs(task)))
334 		ia32_save_state(task);
335 #endif
336 }
337 
338 void
339 ia64_load_extra (struct task_struct *task)
340 {
341 #ifdef CONFIG_PERFMON
342 	unsigned long info;
343 #endif
344 
345 	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
346 		ia64_load_debug_regs(&task->thread.dbr[0]);
347 
348 #ifdef CONFIG_PERFMON
349 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
350 		pfm_load_regs(task);
351 
352 	info = __get_cpu_var(pfm_syst_info);
353 	if (info & PFM_CPUINFO_SYST_WIDE)
354 		pfm_syst_wide_update_task(task, info, 1);
355 #endif
356 
357 #ifdef CONFIG_IA32_SUPPORT
358 	if (IS_IA32_PROCESS(task_pt_regs(task)))
359 		ia32_load_state(task);
360 #endif
361 }
362 
363 /*
364  * Copy the state of an ia-64 thread.
365  *
366  * We get here through the following  call chain:
367  *
368  *	from user-level:	from kernel:
369  *
370  *	<clone syscall>	        <some kernel call frames>
371  *	sys_clone		   :
372  *	do_fork			do_fork
373  *	copy_thread		copy_thread
374  *
375  * This means that the stack layout is as follows:
376  *
377  *	+---------------------+ (highest addr)
378  *	|   struct pt_regs    |
379  *	+---------------------+
380  *	| struct switch_stack |
381  *	+---------------------+
382  *	|                     |
383  *	|    memory stack     |
384  *	|                     | <-- sp (lowest addr)
385  *	+---------------------+
386  *
387  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
388  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
389  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
390  * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
391  * the stack is page aligned and the page size is at least 4KB, this is always the case,
392  * so there is nothing to worry about.
393  */
394 int
395 copy_thread (int nr, unsigned long clone_flags,
396 	     unsigned long user_stack_base, unsigned long user_stack_size,
397 	     struct task_struct *p, struct pt_regs *regs)
398 {
399 	extern char ia64_ret_from_clone, ia32_ret_from_clone;
400 	struct switch_stack *child_stack, *stack;
401 	unsigned long rbs, child_rbs, rbs_size;
402 	struct pt_regs *child_ptregs;
403 	int retval = 0;
404 
405 #ifdef CONFIG_SMP
406 	/*
407 	 * For SMP idle threads, fork_by_hand() calls do_fork with
408 	 * NULL regs.
409 	 */
410 	if (!regs)
411 		return 0;
412 #endif
413 
414 	stack = ((struct switch_stack *) regs) - 1;
415 
416 	child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
417 	child_stack = (struct switch_stack *) child_ptregs - 1;
418 
419 	/* copy parent's switch_stack & pt_regs to child: */
420 	memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
421 
422 	rbs = (unsigned long) current + IA64_RBS_OFFSET;
423 	child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
424 	rbs_size = stack->ar_bspstore - rbs;
425 
426 	/* copy the parent's register backing store to the child: */
427 	memcpy((void *) child_rbs, (void *) rbs, rbs_size);
428 
429 	if (likely(user_mode(child_ptregs))) {
430 		if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs))
431 			child_ptregs->r13 = regs->r16;	/* see sys_clone2() in entry.S */
432 		if (user_stack_base) {
433 			child_ptregs->r12 = user_stack_base + user_stack_size - 16;
434 			child_ptregs->ar_bspstore = user_stack_base;
435 			child_ptregs->ar_rnat = 0;
436 			child_ptregs->loadrs = 0;
437 		}
438 	} else {
439 		/*
440 		 * Note: we simply preserve the relative position of
441 		 * the stack pointer here.  There is no need to
442 		 * allocate a scratch area here, since that will have
443 		 * been taken care of by the caller of sys_clone()
444 		 * already.
445 		 */
446 		child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
447 		child_ptregs->r13 = (unsigned long) p;		/* set `current' pointer */
448 	}
449 	child_stack->ar_bspstore = child_rbs + rbs_size;
450 	if (IS_IA32_PROCESS(regs))
451 		child_stack->b0 = (unsigned long) &ia32_ret_from_clone;
452 	else
453 		child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
454 
455 	/* copy parts of thread_struct: */
456 	p->thread.ksp = (unsigned long) child_stack - 16;
457 
458 	/* stop some PSR bits from being inherited.
459 	 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
460 	 * therefore we must specify them explicitly here and not include them in
461 	 * IA64_PSR_BITS_TO_CLEAR.
462 	 */
463 	child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
464 				 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
465 
466 	/*
467 	 * NOTE: The calling convention considers all floating point
468 	 * registers in the high partition (fph) to be scratch.  Since
469 	 * the only way to get to this point is through a system call,
470 	 * we know that the values in fph are all dead.  Hence, there
471 	 * is no need to inherit the fph state from the parent to the
472 	 * child and all we have to do is to make sure that
473 	 * IA64_THREAD_FPH_VALID is cleared in the child.
474 	 *
475 	 * XXX We could push this optimization a bit further by
476 	 * clearing IA64_THREAD_FPH_VALID on ANY system call.
477 	 * However, it's not clear this is worth doing.  Also, it
478 	 * would be a slight deviation from the normal Linux system
479 	 * call behavior where scratch registers are preserved across
480 	 * system calls (unless used by the system call itself).
481 	 */
482 #	define THREAD_FLAGS_TO_CLEAR	(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
483 					 | IA64_THREAD_PM_VALID)
484 #	define THREAD_FLAGS_TO_SET	0
485 	p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
486 			   | THREAD_FLAGS_TO_SET);
487 	ia64_drop_fpu(p);	/* don't pick up stale state from a CPU's fph */
488 #ifdef CONFIG_IA32_SUPPORT
489 	/*
490 	 * If we're cloning an IA32 task then save the IA32 extra
491 	 * state from the current task to the new task
492 	 */
493 	if (IS_IA32_PROCESS(task_pt_regs(current))) {
494 		ia32_save_state(p);
495 		if (clone_flags & CLONE_SETTLS)
496 			retval = ia32_clone_tls(p, child_ptregs);
497 
498 		/* Copy partially mapped page list */
499 		if (!retval)
500 			retval = ia32_copy_partial_page_list(p, clone_flags);
501 	}
502 #endif
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, ip, ar_rnat, urbs_end, cfm;
515 	elf_greg_t *dst = arg;
516 	struct pt_regs *pt;
517 	char nat;
518 	int i;
519 
520 	memset(dst, 0, sizeof(elf_gregset_t));	/* don't leak any kernel bits to user-level */
521 
522 	if (unw_unwind_to_user(info) < 0)
523 		return;
524 
525 	unw_get_sp(info, &sp);
526 	pt = (struct pt_regs *) (sp + 16);
527 
528 	urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
529 
530 	if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
531 		return;
532 
533 	ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
534 		  &ar_rnat);
535 
536 	/*
537 	 * coredump format:
538 	 *	r0-r31
539 	 *	NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
540 	 *	predicate registers (p0-p63)
541 	 *	b0-b7
542 	 *	ip cfm user-mask
543 	 *	ar.rsc ar.bsp ar.bspstore ar.rnat
544 	 *	ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
545 	 */
546 
547 	/* r0 is zero */
548 	for (i = 1, mask = (1UL << i); i < 32; ++i) {
549 		unw_get_gr(info, i, &dst[i], &nat);
550 		if (nat)
551 			nat_bits |= mask;
552 		mask <<= 1;
553 	}
554 	dst[32] = nat_bits;
555 	unw_get_pr(info, &dst[33]);
556 
557 	for (i = 0; i < 8; ++i)
558 		unw_get_br(info, i, &dst[34 + i]);
559 
560 	unw_get_rp(info, &ip);
561 	dst[42] = ip + ia64_psr(pt)->ri;
562 	dst[43] = cfm;
563 	dst[44] = pt->cr_ipsr & IA64_PSR_UM;
564 
565 	unw_get_ar(info, UNW_AR_RSC, &dst[45]);
566 	/*
567 	 * For bsp and bspstore, unw_get_ar() would return the kernel
568 	 * addresses, but we need the user-level addresses instead:
569 	 */
570 	dst[46] = urbs_end;	/* note: by convention PT_AR_BSP points to the end of the urbs! */
571 	dst[47] = pt->ar_bspstore;
572 	dst[48] = ar_rnat;
573 	unw_get_ar(info, UNW_AR_CCV, &dst[49]);
574 	unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
575 	unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
576 	dst[52] = pt->ar_pfs;	/* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
577 	unw_get_ar(info, UNW_AR_LC, &dst[53]);
578 	unw_get_ar(info, UNW_AR_EC, &dst[54]);
579 	unw_get_ar(info, UNW_AR_CSD, &dst[55]);
580 	unw_get_ar(info, UNW_AR_SSD, &dst[56]);
581 }
582 
583 void
584 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
585 {
586 	elf_fpreg_t *dst = arg;
587 	int i;
588 
589 	memset(dst, 0, sizeof(elf_fpregset_t));	/* don't leak any "random" bits */
590 
591 	if (unw_unwind_to_user(info) < 0)
592 		return;
593 
594 	/* f0 is 0.0, f1 is 1.0 */
595 
596 	for (i = 2; i < 32; ++i)
597 		unw_get_fr(info, i, dst + i);
598 
599 	ia64_flush_fph(task);
600 	if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
601 		memcpy(dst + 32, task->thread.fph, 96*16);
602 }
603 
604 void
605 do_copy_regs (struct unw_frame_info *info, void *arg)
606 {
607 	do_copy_task_regs(current, info, arg);
608 }
609 
610 void
611 do_dump_fpu (struct unw_frame_info *info, void *arg)
612 {
613 	do_dump_task_fpu(current, info, arg);
614 }
615 
616 int
617 dump_task_regs(struct task_struct *task, elf_gregset_t *regs)
618 {
619 	struct unw_frame_info tcore_info;
620 
621 	if (current == task) {
622 		unw_init_running(do_copy_regs, regs);
623 	} else {
624 		memset(&tcore_info, 0, sizeof(tcore_info));
625 		unw_init_from_blocked_task(&tcore_info, task);
626 		do_copy_task_regs(task, &tcore_info, regs);
627 	}
628 	return 1;
629 }
630 
631 void
632 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
633 {
634 	unw_init_running(do_copy_regs, dst);
635 }
636 
637 int
638 dump_task_fpu (struct task_struct *task, elf_fpregset_t *dst)
639 {
640 	struct unw_frame_info tcore_info;
641 
642 	if (current == task) {
643 		unw_init_running(do_dump_fpu, dst);
644 	} else {
645 		memset(&tcore_info, 0, sizeof(tcore_info));
646 		unw_init_from_blocked_task(&tcore_info, task);
647 		do_dump_task_fpu(task, &tcore_info, dst);
648 	}
649 	return 1;
650 }
651 
652 int
653 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
654 {
655 	unw_init_running(do_dump_fpu, dst);
656 	return 1;	/* f0-f31 are always valid so we always return 1 */
657 }
658 
659 long
660 sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp,
661 	    struct pt_regs *regs)
662 {
663 	char *fname;
664 	int error;
665 
666 	fname = getname(filename);
667 	error = PTR_ERR(fname);
668 	if (IS_ERR(fname))
669 		goto out;
670 	error = do_execve(fname, argv, envp, regs);
671 	putname(fname);
672 out:
673 	return error;
674 }
675 
676 pid_t
677 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
678 {
679 	extern void start_kernel_thread (void);
680 	unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
681 	struct {
682 		struct switch_stack sw;
683 		struct pt_regs pt;
684 	} regs;
685 
686 	memset(&regs, 0, sizeof(regs));
687 	regs.pt.cr_iip = helper_fptr[0];	/* set entry point (IP) */
688 	regs.pt.r1 = helper_fptr[1];		/* set GP */
689 	regs.pt.r9 = (unsigned long) fn;	/* 1st argument */
690 	regs.pt.r11 = (unsigned long) arg;	/* 2nd argument */
691 	/* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
692 	regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
693 	regs.pt.cr_ifs = 1UL << 63;		/* mark as valid, empty frame */
694 	regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
695 	regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
696 	regs.sw.pr = (1 << PRED_KERNEL_STACK);
697 	return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
698 }
699 EXPORT_SYMBOL(kernel_thread);
700 
701 /* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
702 int
703 kernel_thread_helper (int (*fn)(void *), void *arg)
704 {
705 #ifdef CONFIG_IA32_SUPPORT
706 	if (IS_IA32_PROCESS(task_pt_regs(current))) {
707 		/* A kernel thread is always a 64-bit process. */
708 		current->thread.map_base  = DEFAULT_MAP_BASE;
709 		current->thread.task_size = DEFAULT_TASK_SIZE;
710 		ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
711 		ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
712 	}
713 #endif
714 	return (*fn)(arg);
715 }
716 
717 /*
718  * Flush thread state.  This is called when a thread does an execve().
719  */
720 void
721 flush_thread (void)
722 {
723 	/* drop floating-point and debug-register state if it exists: */
724 	current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
725 	ia64_drop_fpu(current);
726 #ifdef CONFIG_IA32_SUPPORT
727 	if (IS_IA32_PROCESS(task_pt_regs(current))) {
728 		ia32_drop_partial_page_list(current);
729 		current->thread.task_size = IA32_PAGE_OFFSET;
730 		set_fs(USER_DS);
731 	}
732 #endif
733 }
734 
735 /*
736  * Clean up state associated with current thread.  This is called when
737  * the thread calls exit().
738  */
739 void
740 exit_thread (void)
741 {
742 
743 	ia64_drop_fpu(current);
744 #ifdef CONFIG_PERFMON
745        /* if needed, stop monitoring and flush state to perfmon context */
746 	if (current->thread.pfm_context)
747 		pfm_exit_thread(current);
748 
749 	/* free debug register resources */
750 	if (current->thread.flags & IA64_THREAD_DBG_VALID)
751 		pfm_release_debug_registers(current);
752 #endif
753 	if (IS_IA32_PROCESS(task_pt_regs(current)))
754 		ia32_drop_partial_page_list(current);
755 }
756 
757 unsigned long
758 get_wchan (struct task_struct *p)
759 {
760 	struct unw_frame_info info;
761 	unsigned long ip;
762 	int count = 0;
763 
764 	/*
765 	 * Note: p may not be a blocked task (it could be current or
766 	 * another process running on some other CPU.  Rather than
767 	 * trying to determine if p is really blocked, we just assume
768 	 * it's blocked and rely on the unwind routines to fail
769 	 * gracefully if the process wasn't really blocked after all.
770 	 * --davidm 99/12/15
771 	 */
772 	unw_init_from_blocked_task(&info, p);
773 	do {
774 		if (unw_unwind(&info) < 0)
775 			return 0;
776 		unw_get_ip(&info, &ip);
777 		if (!in_sched_functions(ip))
778 			return ip;
779 	} while (count++ < 16);
780 	return 0;
781 }
782 
783 void
784 cpu_halt (void)
785 {
786 	pal_power_mgmt_info_u_t power_info[8];
787 	unsigned long min_power;
788 	int i, min_power_state;
789 
790 	if (ia64_pal_halt_info(power_info) != 0)
791 		return;
792 
793 	min_power_state = 0;
794 	min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
795 	for (i = 1; i < 8; ++i)
796 		if (power_info[i].pal_power_mgmt_info_s.im
797 		    && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
798 			min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
799 			min_power_state = i;
800 		}
801 
802 	while (1)
803 		ia64_pal_halt(min_power_state);
804 }
805 
806 void machine_shutdown(void)
807 {
808 #ifdef CONFIG_HOTPLUG_CPU
809 	int cpu;
810 
811 	for_each_online_cpu(cpu) {
812 		if (cpu != smp_processor_id())
813 			cpu_down(cpu);
814 	}
815 #endif
816 #ifdef CONFIG_KEXEC
817 	kexec_disable_iosapic();
818 #endif
819 }
820 
821 void
822 machine_restart (char *restart_cmd)
823 {
824 	(void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
825 	(*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
826 }
827 
828 void
829 machine_halt (void)
830 {
831 	(void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
832 	cpu_halt();
833 }
834 
835 void
836 machine_power_off (void)
837 {
838 	if (pm_power_off)
839 		pm_power_off();
840 	machine_halt();
841 }
842 
843