]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Architecture-specific setup. | |
3 | * | |
4 | * Copyright (C) 1998-2003 Hewlett-Packard Co | |
5 | * David Mosberger-Tang <davidm@hpl.hp.com> | |
b8d8b883 | 6 | * 04/11/17 Ashok Raj <ashok.raj@intel.com> Added CPU Hotplug Support |
9138d581 KO |
7 | * |
8 | * 2005-10-07 Keith Owens <kaos@sgi.com> | |
9 | * Add notify_die() hooks. | |
1da177e4 | 10 | */ |
1da177e4 LT |
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> | |
1da177e4 LT |
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> | |
1eeb66a1 | 29 | #include <linux/kdebug.h> |
ee211b37 | 30 | #include <linux/utsname.h> |
1da177e4 LT |
31 | |
32 | #include <asm/cpu.h> | |
33 | #include <asm/delay.h> | |
34 | #include <asm/elf.h> | |
35 | #include <asm/ia32.h> | |
36 | #include <asm/irq.h> | |
c237508a | 37 | #include <asm/kexec.h> |
1da177e4 LT |
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); | |
7d5f9c0f | 55 | static DEFINE_PER_CPU(unsigned int, cpu_idle_state); |
1da177e4 LT |
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(); | |
19c5870c AD |
109 | printk("\nPid: %d, CPU %d, comm: %20s\n", task_pid_nr(current), |
110 | smp_processor_id(), current->comm); | |
ee211b37 TL |
111 | printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s (%s)\n", |
112 | regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(), | |
113 | init_utsname()->release); | |
1da177e4 LT |
114 | print_symbol("ip is at %s\n", ip); |
115 | printk("unat: %016lx pfs : %016lx rsc : %016lx\n", | |
116 | regs->ar_unat, regs->ar_pfs, regs->ar_rsc); | |
117 | printk("rnat: %016lx bsps: %016lx pr : %016lx\n", | |
118 | regs->ar_rnat, regs->ar_bspstore, regs->pr); | |
119 | printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n", | |
120 | regs->loadrs, regs->ar_ccv, regs->ar_fpsr); | |
121 | printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd); | |
122 | printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs->b0, regs->b6, regs->b7); | |
123 | printk("f6 : %05lx%016lx f7 : %05lx%016lx\n", | |
124 | regs->f6.u.bits[1], regs->f6.u.bits[0], | |
125 | regs->f7.u.bits[1], regs->f7.u.bits[0]); | |
126 | printk("f8 : %05lx%016lx f9 : %05lx%016lx\n", | |
127 | regs->f8.u.bits[1], regs->f8.u.bits[0], | |
128 | regs->f9.u.bits[1], regs->f9.u.bits[0]); | |
129 | printk("f10 : %05lx%016lx f11 : %05lx%016lx\n", | |
130 | regs->f10.u.bits[1], regs->f10.u.bits[0], | |
131 | regs->f11.u.bits[1], regs->f11.u.bits[0]); | |
132 | ||
133 | printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs->r1, regs->r2, regs->r3); | |
134 | printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10); | |
135 | printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13); | |
136 | printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16); | |
137 | printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19); | |
138 | printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22); | |
139 | printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25); | |
140 | printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28); | |
141 | printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31); | |
142 | ||
143 | if (user_mode(regs)) { | |
144 | /* print the stacked registers */ | |
145 | unsigned long val, *bsp, ndirty; | |
146 | int i, sof, is_nat = 0; | |
147 | ||
148 | sof = regs->cr_ifs & 0x7f; /* size of frame */ | |
149 | ndirty = (regs->loadrs >> 19); | |
150 | bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty); | |
151 | for (i = 0; i < sof; ++i) { | |
152 | get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i)); | |
153 | printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val, | |
154 | ((i == sof - 1) || (i % 3) == 2) ? "\n" : " "); | |
155 | } | |
156 | } else | |
157 | show_stack(NULL, NULL); | |
158 | } | |
159 | ||
5aa92ffd PT |
160 | void tsk_clear_notify_resume(struct task_struct *tsk) |
161 | { | |
162 | #ifdef CONFIG_PERFMON | |
163 | if (tsk->thread.pfm_needs_checking) | |
164 | return; | |
165 | #endif | |
3b2ce0b1 PT |
166 | if (test_ti_thread_flag(task_thread_info(tsk), TIF_RESTORE_RSE)) |
167 | return; | |
5aa92ffd PT |
168 | clear_ti_thread_flag(task_thread_info(tsk), TIF_NOTIFY_RESUME); |
169 | } | |
170 | ||
1da177e4 | 171 | void |
4a177cbf | 172 | do_notify_resume_user (sigset_t *unused, struct sigscratch *scr, long in_syscall) |
1da177e4 LT |
173 | { |
174 | if (fsys_mode(current, &scr->pt)) { | |
175 | /* defer signal-handling etc. until we return to privilege-level 0. */ | |
176 | if (!ia64_psr(&scr->pt)->lp) | |
177 | ia64_psr(&scr->pt)->lp = 1; | |
178 | return; | |
179 | } | |
180 | ||
181 | #ifdef CONFIG_PERFMON | |
182 | if (current->thread.pfm_needs_checking) | |
183 | pfm_handle_work(); | |
184 | #endif | |
185 | ||
186 | /* deal with pending signal delivery */ | |
4a177cbf AD |
187 | if (test_thread_flag(TIF_SIGPENDING)||test_thread_flag(TIF_RESTORE_SIGMASK)) |
188 | ia64_do_signal(scr, in_syscall); | |
3b2ce0b1 PT |
189 | |
190 | /* copy user rbs to kernel rbs */ | |
191 | if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) | |
192 | ia64_sync_krbs(); | |
1da177e4 LT |
193 | } |
194 | ||
8df5a500 SE |
195 | static int pal_halt = 1; |
196 | static int can_do_pal_halt = 1; | |
197 | ||
1da177e4 LT |
198 | static int __init nohalt_setup(char * str) |
199 | { | |
fb573856 | 200 | pal_halt = can_do_pal_halt = 0; |
1da177e4 LT |
201 | return 1; |
202 | } | |
203 | __setup("nohalt", nohalt_setup); | |
204 | ||
a71f62ed | 205 | void |
8df5a500 SE |
206 | update_pal_halt_status(int status) |
207 | { | |
208 | can_do_pal_halt = pal_halt && status; | |
209 | } | |
210 | ||
1da177e4 LT |
211 | /* |
212 | * We use this if we don't have any better idle routine.. | |
213 | */ | |
214 | void | |
215 | default_idle (void) | |
216 | { | |
6c4fa560 | 217 | local_irq_enable(); |
64c7c8f8 | 218 | while (!need_resched()) { |
71416bea DS |
219 | if (can_do_pal_halt) { |
220 | local_irq_disable(); | |
221 | if (!need_resched()) { | |
222 | safe_halt(); | |
223 | } | |
224 | local_irq_enable(); | |
225 | } else | |
1da177e4 | 226 | cpu_relax(); |
64c7c8f8 | 227 | } |
1da177e4 LT |
228 | } |
229 | ||
230 | #ifdef CONFIG_HOTPLUG_CPU | |
231 | /* We don't actually take CPU down, just spin without interrupts. */ | |
232 | static inline void play_dead(void) | |
233 | { | |
234 | extern void ia64_cpu_local_tick (void); | |
b8d8b883 AR |
235 | unsigned int this_cpu = smp_processor_id(); |
236 | ||
1da177e4 LT |
237 | /* Ack it */ |
238 | __get_cpu_var(cpu_state) = CPU_DEAD; | |
239 | ||
1da177e4 LT |
240 | max_xtp(); |
241 | local_irq_disable(); | |
b8d8b883 AR |
242 | idle_task_exit(); |
243 | ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]); | |
1da177e4 | 244 | /* |
b8d8b883 AR |
245 | * The above is a point of no-return, the processor is |
246 | * expected to be in SAL loop now. | |
1da177e4 | 247 | */ |
b8d8b883 | 248 | BUG(); |
1da177e4 LT |
249 | } |
250 | #else | |
251 | static inline void play_dead(void) | |
252 | { | |
253 | BUG(); | |
254 | } | |
255 | #endif /* CONFIG_HOTPLUG_CPU */ | |
256 | ||
1da177e4 LT |
257 | void cpu_idle_wait(void) |
258 | { | |
7d5f9c0f ZM |
259 | unsigned int cpu, this_cpu = get_cpu(); |
260 | cpumask_t map; | |
bb8416bf | 261 | cpumask_t tmp = current->cpus_allowed; |
7d5f9c0f ZM |
262 | |
263 | set_cpus_allowed(current, cpumask_of_cpu(this_cpu)); | |
264 | put_cpu(); | |
1da177e4 | 265 | |
7d5f9c0f ZM |
266 | cpus_clear(map); |
267 | for_each_online_cpu(cpu) { | |
268 | per_cpu(cpu_idle_state, cpu) = 1; | |
269 | cpu_set(cpu, map); | |
270 | } | |
1da177e4 | 271 | |
7d5f9c0f ZM |
272 | __get_cpu_var(cpu_idle_state) = 0; |
273 | ||
274 | wmb(); | |
275 | do { | |
276 | ssleep(1); | |
277 | for_each_online_cpu(cpu) { | |
278 | if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu)) | |
279 | cpu_clear(cpu, map); | |
280 | } | |
281 | cpus_and(map, map, cpu_online_map); | |
282 | } while (!cpus_empty(map)); | |
bb8416bf | 283 | set_cpus_allowed(current, tmp); |
1da177e4 LT |
284 | } |
285 | EXPORT_SYMBOL_GPL(cpu_idle_wait); | |
286 | ||
287 | void __attribute__((noreturn)) | |
288 | cpu_idle (void) | |
289 | { | |
290 | void (*mark_idle)(int) = ia64_mark_idle; | |
64c7c8f8 | 291 | int cpu = smp_processor_id(); |
1da177e4 LT |
292 | |
293 | /* endless idle loop with no priority at all */ | |
294 | while (1) { | |
0888f06a | 295 | if (can_do_pal_halt) { |
495ab9c0 | 296 | current_thread_info()->status &= ~TS_POLLING; |
0888f06a IM |
297 | /* |
298 | * TS_POLLING-cleared state must be visible before we | |
299 | * test NEED_RESCHED: | |
300 | */ | |
301 | smp_mb(); | |
302 | } else { | |
495ab9c0 | 303 | current_thread_info()->status |= TS_POLLING; |
0888f06a | 304 | } |
1e185b97 | 305 | |
64c7c8f8 NP |
306 | if (!need_resched()) { |
307 | void (*idle)(void); | |
1da177e4 | 308 | #ifdef CONFIG_SMP |
1da177e4 LT |
309 | min_xtp(); |
310 | #endif | |
7d5f9c0f ZM |
311 | if (__get_cpu_var(cpu_idle_state)) |
312 | __get_cpu_var(cpu_idle_state) = 0; | |
313 | ||
314 | rmb(); | |
1da177e4 LT |
315 | if (mark_idle) |
316 | (*mark_idle)(1); | |
317 | ||
1da177e4 LT |
318 | idle = pm_idle; |
319 | if (!idle) | |
320 | idle = default_idle; | |
321 | (*idle)(); | |
64c7c8f8 NP |
322 | if (mark_idle) |
323 | (*mark_idle)(0); | |
1da177e4 | 324 | #ifdef CONFIG_SMP |
64c7c8f8 | 325 | normal_xtp(); |
1da177e4 | 326 | #endif |
64c7c8f8 | 327 | } |
5bfb5d69 | 328 | preempt_enable_no_resched(); |
1da177e4 | 329 | schedule(); |
5bfb5d69 | 330 | preempt_disable(); |
1da177e4 | 331 | check_pgt_cache(); |
64c7c8f8 | 332 | if (cpu_is_offline(cpu)) |
1da177e4 LT |
333 | play_dead(); |
334 | } | |
335 | } | |
336 | ||
337 | void | |
338 | ia64_save_extra (struct task_struct *task) | |
339 | { | |
340 | #ifdef CONFIG_PERFMON | |
341 | unsigned long info; | |
342 | #endif | |
343 | ||
344 | if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0) | |
345 | ia64_save_debug_regs(&task->thread.dbr[0]); | |
346 | ||
347 | #ifdef CONFIG_PERFMON | |
348 | if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0) | |
349 | pfm_save_regs(task); | |
350 | ||
351 | info = __get_cpu_var(pfm_syst_info); | |
352 | if (info & PFM_CPUINFO_SYST_WIDE) | |
353 | pfm_syst_wide_update_task(task, info, 0); | |
354 | #endif | |
355 | ||
356 | #ifdef CONFIG_IA32_SUPPORT | |
6450578f | 357 | if (IS_IA32_PROCESS(task_pt_regs(task))) |
1da177e4 LT |
358 | ia32_save_state(task); |
359 | #endif | |
360 | } | |
361 | ||
362 | void | |
363 | ia64_load_extra (struct task_struct *task) | |
364 | { | |
365 | #ifdef CONFIG_PERFMON | |
366 | unsigned long info; | |
367 | #endif | |
368 | ||
369 | if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0) | |
370 | ia64_load_debug_regs(&task->thread.dbr[0]); | |
371 | ||
372 | #ifdef CONFIG_PERFMON | |
373 | if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0) | |
374 | pfm_load_regs(task); | |
375 | ||
376 | info = __get_cpu_var(pfm_syst_info); | |
377 | if (info & PFM_CPUINFO_SYST_WIDE) | |
378 | pfm_syst_wide_update_task(task, info, 1); | |
379 | #endif | |
380 | ||
381 | #ifdef CONFIG_IA32_SUPPORT | |
6450578f | 382 | if (IS_IA32_PROCESS(task_pt_regs(task))) |
1da177e4 LT |
383 | ia32_load_state(task); |
384 | #endif | |
385 | } | |
386 | ||
387 | /* | |
388 | * Copy the state of an ia-64 thread. | |
389 | * | |
390 | * We get here through the following call chain: | |
391 | * | |
392 | * from user-level: from kernel: | |
393 | * | |
394 | * <clone syscall> <some kernel call frames> | |
395 | * sys_clone : | |
396 | * do_fork do_fork | |
397 | * copy_thread copy_thread | |
398 | * | |
399 | * This means that the stack layout is as follows: | |
400 | * | |
401 | * +---------------------+ (highest addr) | |
402 | * | struct pt_regs | | |
403 | * +---------------------+ | |
404 | * | struct switch_stack | | |
405 | * +---------------------+ | |
406 | * | | | |
407 | * | memory stack | | |
408 | * | | <-- sp (lowest addr) | |
409 | * +---------------------+ | |
410 | * | |
411 | * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an | |
412 | * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register, | |
413 | * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the | |
414 | * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since | |
415 | * the stack is page aligned and the page size is at least 4KB, this is always the case, | |
416 | * so there is nothing to worry about. | |
417 | */ | |
418 | int | |
419 | copy_thread (int nr, unsigned long clone_flags, | |
420 | unsigned long user_stack_base, unsigned long user_stack_size, | |
421 | struct task_struct *p, struct pt_regs *regs) | |
422 | { | |
423 | extern char ia64_ret_from_clone, ia32_ret_from_clone; | |
424 | struct switch_stack *child_stack, *stack; | |
425 | unsigned long rbs, child_rbs, rbs_size; | |
426 | struct pt_regs *child_ptregs; | |
427 | int retval = 0; | |
428 | ||
429 | #ifdef CONFIG_SMP | |
430 | /* | |
431 | * For SMP idle threads, fork_by_hand() calls do_fork with | |
432 | * NULL regs. | |
433 | */ | |
434 | if (!regs) | |
435 | return 0; | |
436 | #endif | |
437 | ||
438 | stack = ((struct switch_stack *) regs) - 1; | |
439 | ||
440 | child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1; | |
441 | child_stack = (struct switch_stack *) child_ptregs - 1; | |
442 | ||
443 | /* copy parent's switch_stack & pt_regs to child: */ | |
444 | memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack)); | |
445 | ||
446 | rbs = (unsigned long) current + IA64_RBS_OFFSET; | |
447 | child_rbs = (unsigned long) p + IA64_RBS_OFFSET; | |
448 | rbs_size = stack->ar_bspstore - rbs; | |
449 | ||
450 | /* copy the parent's register backing store to the child: */ | |
451 | memcpy((void *) child_rbs, (void *) rbs, rbs_size); | |
452 | ||
453 | if (likely(user_mode(child_ptregs))) { | |
454 | if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs)) | |
455 | child_ptregs->r13 = regs->r16; /* see sys_clone2() in entry.S */ | |
456 | if (user_stack_base) { | |
457 | child_ptregs->r12 = user_stack_base + user_stack_size - 16; | |
458 | child_ptregs->ar_bspstore = user_stack_base; | |
459 | child_ptregs->ar_rnat = 0; | |
460 | child_ptregs->loadrs = 0; | |
461 | } | |
462 | } else { | |
463 | /* | |
464 | * Note: we simply preserve the relative position of | |
465 | * the stack pointer here. There is no need to | |
466 | * allocate a scratch area here, since that will have | |
467 | * been taken care of by the caller of sys_clone() | |
468 | * already. | |
469 | */ | |
470 | child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */ | |
471 | child_ptregs->r13 = (unsigned long) p; /* set `current' pointer */ | |
472 | } | |
473 | child_stack->ar_bspstore = child_rbs + rbs_size; | |
474 | if (IS_IA32_PROCESS(regs)) | |
475 | child_stack->b0 = (unsigned long) &ia32_ret_from_clone; | |
476 | else | |
477 | child_stack->b0 = (unsigned long) &ia64_ret_from_clone; | |
478 | ||
479 | /* copy parts of thread_struct: */ | |
480 | p->thread.ksp = (unsigned long) child_stack - 16; | |
481 | ||
482 | /* stop some PSR bits from being inherited. | |
483 | * the psr.up/psr.pp bits must be cleared on fork but inherited on execve() | |
484 | * therefore we must specify them explicitly here and not include them in | |
485 | * IA64_PSR_BITS_TO_CLEAR. | |
486 | */ | |
487 | child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET) | |
488 | & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP)); | |
489 | ||
490 | /* | |
491 | * NOTE: The calling convention considers all floating point | |
492 | * registers in the high partition (fph) to be scratch. Since | |
493 | * the only way to get to this point is through a system call, | |
494 | * we know that the values in fph are all dead. Hence, there | |
495 | * is no need to inherit the fph state from the parent to the | |
496 | * child and all we have to do is to make sure that | |
497 | * IA64_THREAD_FPH_VALID is cleared in the child. | |
498 | * | |
499 | * XXX We could push this optimization a bit further by | |
500 | * clearing IA64_THREAD_FPH_VALID on ANY system call. | |
501 | * However, it's not clear this is worth doing. Also, it | |
502 | * would be a slight deviation from the normal Linux system | |
503 | * call behavior where scratch registers are preserved across | |
504 | * system calls (unless used by the system call itself). | |
505 | */ | |
506 | # define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \ | |
507 | | IA64_THREAD_PM_VALID) | |
508 | # define THREAD_FLAGS_TO_SET 0 | |
509 | p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR) | |
510 | | THREAD_FLAGS_TO_SET); | |
511 | ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */ | |
512 | #ifdef CONFIG_IA32_SUPPORT | |
513 | /* | |
514 | * If we're cloning an IA32 task then save the IA32 extra | |
515 | * state from the current task to the new task | |
516 | */ | |
6450578f | 517 | if (IS_IA32_PROCESS(task_pt_regs(current))) { |
1da177e4 LT |
518 | ia32_save_state(p); |
519 | if (clone_flags & CLONE_SETTLS) | |
520 | retval = ia32_clone_tls(p, child_ptregs); | |
521 | ||
522 | /* Copy partially mapped page list */ | |
523 | if (!retval) | |
3b74d18e | 524 | retval = ia32_copy_ia64_partial_page_list(p, |
525 | clone_flags); | |
1da177e4 LT |
526 | } |
527 | #endif | |
528 | ||
529 | #ifdef CONFIG_PERFMON | |
530 | if (current->thread.pfm_context) | |
531 | pfm_inherit(p, child_ptregs); | |
532 | #endif | |
533 | return retval; | |
534 | } | |
535 | ||
536 | static void | |
537 | do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg) | |
538 | { | |
256a7e09 JS |
539 | unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm; |
540 | unsigned long uninitialized_var(ip); /* GCC be quiet */ | |
1da177e4 LT |
541 | elf_greg_t *dst = arg; |
542 | struct pt_regs *pt; | |
543 | char nat; | |
544 | int i; | |
545 | ||
546 | memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */ | |
547 | ||
548 | if (unw_unwind_to_user(info) < 0) | |
549 | return; | |
550 | ||
551 | unw_get_sp(info, &sp); | |
552 | pt = (struct pt_regs *) (sp + 16); | |
553 | ||
554 | urbs_end = ia64_get_user_rbs_end(task, pt, &cfm); | |
555 | ||
556 | if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0) | |
557 | return; | |
558 | ||
559 | ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end), | |
560 | &ar_rnat); | |
561 | ||
562 | /* | |
563 | * coredump format: | |
564 | * r0-r31 | |
565 | * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT) | |
566 | * predicate registers (p0-p63) | |
567 | * b0-b7 | |
568 | * ip cfm user-mask | |
569 | * ar.rsc ar.bsp ar.bspstore ar.rnat | |
570 | * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec | |
571 | */ | |
572 | ||
573 | /* r0 is zero */ | |
574 | for (i = 1, mask = (1UL << i); i < 32; ++i) { | |
575 | unw_get_gr(info, i, &dst[i], &nat); | |
576 | if (nat) | |
577 | nat_bits |= mask; | |
578 | mask <<= 1; | |
579 | } | |
580 | dst[32] = nat_bits; | |
581 | unw_get_pr(info, &dst[33]); | |
582 | ||
583 | for (i = 0; i < 8; ++i) | |
584 | unw_get_br(info, i, &dst[34 + i]); | |
585 | ||
586 | unw_get_rp(info, &ip); | |
587 | dst[42] = ip + ia64_psr(pt)->ri; | |
588 | dst[43] = cfm; | |
589 | dst[44] = pt->cr_ipsr & IA64_PSR_UM; | |
590 | ||
591 | unw_get_ar(info, UNW_AR_RSC, &dst[45]); | |
592 | /* | |
593 | * For bsp and bspstore, unw_get_ar() would return the kernel | |
594 | * addresses, but we need the user-level addresses instead: | |
595 | */ | |
596 | dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */ | |
597 | dst[47] = pt->ar_bspstore; | |
598 | dst[48] = ar_rnat; | |
599 | unw_get_ar(info, UNW_AR_CCV, &dst[49]); | |
600 | unw_get_ar(info, UNW_AR_UNAT, &dst[50]); | |
601 | unw_get_ar(info, UNW_AR_FPSR, &dst[51]); | |
602 | dst[52] = pt->ar_pfs; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */ | |
603 | unw_get_ar(info, UNW_AR_LC, &dst[53]); | |
604 | unw_get_ar(info, UNW_AR_EC, &dst[54]); | |
605 | unw_get_ar(info, UNW_AR_CSD, &dst[55]); | |
606 | unw_get_ar(info, UNW_AR_SSD, &dst[56]); | |
607 | } | |
608 | ||
609 | void | |
610 | do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg) | |
611 | { | |
612 | elf_fpreg_t *dst = arg; | |
613 | int i; | |
614 | ||
615 | memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */ | |
616 | ||
617 | if (unw_unwind_to_user(info) < 0) | |
618 | return; | |
619 | ||
620 | /* f0 is 0.0, f1 is 1.0 */ | |
621 | ||
622 | for (i = 2; i < 32; ++i) | |
623 | unw_get_fr(info, i, dst + i); | |
624 | ||
625 | ia64_flush_fph(task); | |
626 | if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0) | |
627 | memcpy(dst + 32, task->thread.fph, 96*16); | |
628 | } | |
629 | ||
630 | void | |
631 | do_copy_regs (struct unw_frame_info *info, void *arg) | |
632 | { | |
633 | do_copy_task_regs(current, info, arg); | |
634 | } | |
635 | ||
636 | void | |
637 | do_dump_fpu (struct unw_frame_info *info, void *arg) | |
638 | { | |
639 | do_dump_task_fpu(current, info, arg); | |
640 | } | |
641 | ||
642 | int | |
643 | dump_task_regs(struct task_struct *task, elf_gregset_t *regs) | |
644 | { | |
645 | struct unw_frame_info tcore_info; | |
646 | ||
647 | if (current == task) { | |
648 | unw_init_running(do_copy_regs, regs); | |
649 | } else { | |
650 | memset(&tcore_info, 0, sizeof(tcore_info)); | |
651 | unw_init_from_blocked_task(&tcore_info, task); | |
652 | do_copy_task_regs(task, &tcore_info, regs); | |
653 | } | |
654 | return 1; | |
655 | } | |
656 | ||
657 | void | |
658 | ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst) | |
659 | { | |
660 | unw_init_running(do_copy_regs, dst); | |
661 | } | |
662 | ||
663 | int | |
664 | dump_task_fpu (struct task_struct *task, elf_fpregset_t *dst) | |
665 | { | |
666 | struct unw_frame_info tcore_info; | |
667 | ||
668 | if (current == task) { | |
669 | unw_init_running(do_dump_fpu, dst); | |
670 | } else { | |
671 | memset(&tcore_info, 0, sizeof(tcore_info)); | |
672 | unw_init_from_blocked_task(&tcore_info, task); | |
673 | do_dump_task_fpu(task, &tcore_info, dst); | |
674 | } | |
675 | return 1; | |
676 | } | |
677 | ||
678 | int | |
679 | dump_fpu (struct pt_regs *pt, elf_fpregset_t dst) | |
680 | { | |
681 | unw_init_running(do_dump_fpu, dst); | |
682 | return 1; /* f0-f31 are always valid so we always return 1 */ | |
683 | } | |
684 | ||
685 | long | |
686 | sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp, | |
687 | struct pt_regs *regs) | |
688 | { | |
689 | char *fname; | |
690 | int error; | |
691 | ||
692 | fname = getname(filename); | |
693 | error = PTR_ERR(fname); | |
694 | if (IS_ERR(fname)) | |
695 | goto out; | |
696 | error = do_execve(fname, argv, envp, regs); | |
697 | putname(fname); | |
698 | out: | |
699 | return error; | |
700 | } | |
701 | ||
702 | pid_t | |
703 | kernel_thread (int (*fn)(void *), void *arg, unsigned long flags) | |
704 | { | |
705 | extern void start_kernel_thread (void); | |
706 | unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread; | |
707 | struct { | |
708 | struct switch_stack sw; | |
709 | struct pt_regs pt; | |
710 | } regs; | |
711 | ||
712 | memset(®s, 0, sizeof(regs)); | |
713 | regs.pt.cr_iip = helper_fptr[0]; /* set entry point (IP) */ | |
714 | regs.pt.r1 = helper_fptr[1]; /* set GP */ | |
715 | regs.pt.r9 = (unsigned long) fn; /* 1st argument */ | |
716 | regs.pt.r11 = (unsigned long) arg; /* 2nd argument */ | |
717 | /* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read. */ | |
718 | regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN; | |
719 | regs.pt.cr_ifs = 1UL << 63; /* mark as valid, empty frame */ | |
720 | regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR); | |
721 | regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET; | |
722 | regs.sw.pr = (1 << PRED_KERNEL_STACK); | |
723 | return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, ®s.pt, 0, NULL, NULL); | |
724 | } | |
725 | EXPORT_SYMBOL(kernel_thread); | |
726 | ||
727 | /* This gets called from kernel_thread() via ia64_invoke_thread_helper(). */ | |
728 | int | |
729 | kernel_thread_helper (int (*fn)(void *), void *arg) | |
730 | { | |
731 | #ifdef CONFIG_IA32_SUPPORT | |
6450578f | 732 | if (IS_IA32_PROCESS(task_pt_regs(current))) { |
1da177e4 LT |
733 | /* A kernel thread is always a 64-bit process. */ |
734 | current->thread.map_base = DEFAULT_MAP_BASE; | |
735 | current->thread.task_size = DEFAULT_TASK_SIZE; | |
736 | ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob); | |
737 | ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1); | |
738 | } | |
739 | #endif | |
740 | return (*fn)(arg); | |
741 | } | |
742 | ||
743 | /* | |
744 | * Flush thread state. This is called when a thread does an execve(). | |
745 | */ | |
746 | void | |
747 | flush_thread (void) | |
748 | { | |
749 | /* drop floating-point and debug-register state if it exists: */ | |
750 | current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID); | |
751 | ia64_drop_fpu(current); | |
27af4cfd | 752 | #ifdef CONFIG_IA32_SUPPORT |
6450578f | 753 | if (IS_IA32_PROCESS(task_pt_regs(current))) { |
3b74d18e | 754 | ia32_drop_ia64_partial_page_list(current); |
bd1d6e24 RH |
755 | current->thread.task_size = IA32_PAGE_OFFSET; |
756 | set_fs(USER_DS); | |
e384f414 | 757 | memset(current->thread.tls_array, 0, sizeof(current->thread.tls_array)); |
bd1d6e24 | 758 | } |
27af4cfd | 759 | #endif |
1da177e4 LT |
760 | } |
761 | ||
762 | /* | |
763 | * Clean up state associated with current thread. This is called when | |
764 | * the thread calls exit(). | |
765 | */ | |
766 | void | |
767 | exit_thread (void) | |
768 | { | |
9508dbfe | 769 | |
1da177e4 LT |
770 | ia64_drop_fpu(current); |
771 | #ifdef CONFIG_PERFMON | |
772 | /* if needed, stop monitoring and flush state to perfmon context */ | |
773 | if (current->thread.pfm_context) | |
774 | pfm_exit_thread(current); | |
775 | ||
776 | /* free debug register resources */ | |
777 | if (current->thread.flags & IA64_THREAD_DBG_VALID) | |
778 | pfm_release_debug_registers(current); | |
779 | #endif | |
6450578f | 780 | if (IS_IA32_PROCESS(task_pt_regs(current))) |
3b74d18e | 781 | ia32_drop_ia64_partial_page_list(current); |
1da177e4 LT |
782 | } |
783 | ||
784 | unsigned long | |
785 | get_wchan (struct task_struct *p) | |
786 | { | |
787 | struct unw_frame_info info; | |
788 | unsigned long ip; | |
789 | int count = 0; | |
790 | ||
6ae38488 RH |
791 | if (!p || p == current || p->state == TASK_RUNNING) |
792 | return 0; | |
793 | ||
1da177e4 LT |
794 | /* |
795 | * Note: p may not be a blocked task (it could be current or | |
796 | * another process running on some other CPU. Rather than | |
797 | * trying to determine if p is really blocked, we just assume | |
798 | * it's blocked and rely on the unwind routines to fail | |
799 | * gracefully if the process wasn't really blocked after all. | |
800 | * --davidm 99/12/15 | |
801 | */ | |
802 | unw_init_from_blocked_task(&info, p); | |
803 | do { | |
6ae38488 RH |
804 | if (p->state == TASK_RUNNING) |
805 | return 0; | |
1da177e4 LT |
806 | if (unw_unwind(&info) < 0) |
807 | return 0; | |
808 | unw_get_ip(&info, &ip); | |
809 | if (!in_sched_functions(ip)) | |
810 | return ip; | |
811 | } while (count++ < 16); | |
812 | return 0; | |
813 | } | |
814 | ||
815 | void | |
816 | cpu_halt (void) | |
817 | { | |
818 | pal_power_mgmt_info_u_t power_info[8]; | |
819 | unsigned long min_power; | |
820 | int i, min_power_state; | |
821 | ||
822 | if (ia64_pal_halt_info(power_info) != 0) | |
823 | return; | |
824 | ||
825 | min_power_state = 0; | |
826 | min_power = power_info[0].pal_power_mgmt_info_s.power_consumption; | |
827 | for (i = 1; i < 8; ++i) | |
828 | if (power_info[i].pal_power_mgmt_info_s.im | |
829 | && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) { | |
830 | min_power = power_info[i].pal_power_mgmt_info_s.power_consumption; | |
831 | min_power_state = i; | |
832 | } | |
833 | ||
834 | while (1) | |
835 | ia64_pal_halt(min_power_state); | |
836 | } | |
837 | ||
c237508a H |
838 | void machine_shutdown(void) |
839 | { | |
840 | #ifdef CONFIG_HOTPLUG_CPU | |
841 | int cpu; | |
842 | ||
843 | for_each_online_cpu(cpu) { | |
844 | if (cpu != smp_processor_id()) | |
845 | cpu_down(cpu); | |
846 | } | |
847 | #endif | |
848 | #ifdef CONFIG_KEXEC | |
849 | kexec_disable_iosapic(); | |
850 | #endif | |
851 | } | |
852 | ||
1da177e4 LT |
853 | void |
854 | machine_restart (char *restart_cmd) | |
855 | { | |
9138d581 | 856 | (void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0); |
1da177e4 LT |
857 | (*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL); |
858 | } | |
859 | ||
1da177e4 LT |
860 | void |
861 | machine_halt (void) | |
862 | { | |
9138d581 | 863 | (void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0); |
1da177e4 LT |
864 | cpu_halt(); |
865 | } | |
866 | ||
1da177e4 LT |
867 | void |
868 | machine_power_off (void) | |
869 | { | |
870 | if (pm_power_off) | |
871 | pm_power_off(); | |
872 | machine_halt(); | |
873 | } | |
874 |