Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * Architecture-specific setup. |
| 3 | * |
| 4 | * Copyright (C) 1998-2003 Hewlett-Packard Co |
| 5 | * David Mosberger-Tang <davidm@hpl.hp.com> |
Ashok Raj | b8d8b88 | 2005-04-22 14:44:40 -0700 | [diff] [blame] | 6 | * 04/11/17 Ashok Raj <ashok.raj@intel.com> Added CPU Hotplug Support |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 7 | */ |
| 8 | #define __KERNEL_SYSCALLS__ /* see <asm/unistd.h> */ |
| 9 | #include <linux/config.h> |
| 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/pgalloc.h> |
| 37 | #include <asm/processor.h> |
| 38 | #include <asm/sal.h> |
| 39 | #include <asm/tlbflush.h> |
| 40 | #include <asm/uaccess.h> |
| 41 | #include <asm/unwind.h> |
| 42 | #include <asm/user.h> |
| 43 | |
| 44 | #include "entry.h" |
| 45 | |
| 46 | #ifdef CONFIG_PERFMON |
| 47 | # include <asm/perfmon.h> |
| 48 | #endif |
| 49 | |
| 50 | #include "sigframe.h" |
| 51 | |
| 52 | void (*ia64_mark_idle)(int); |
Zwane Mwaikambo | 7d5f9c0 | 2005-03-30 21:40:00 -0700 | [diff] [blame] | 53 | static DEFINE_PER_CPU(unsigned int, cpu_idle_state); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 54 | |
| 55 | unsigned long boot_option_idle_override = 0; |
| 56 | EXPORT_SYMBOL(boot_option_idle_override); |
| 57 | |
| 58 | void |
| 59 | ia64_do_show_stack (struct unw_frame_info *info, void *arg) |
| 60 | { |
| 61 | unsigned long ip, sp, bsp; |
| 62 | char buf[128]; /* don't make it so big that it overflows the stack! */ |
| 63 | |
| 64 | printk("\nCall Trace:\n"); |
| 65 | do { |
| 66 | unw_get_ip(info, &ip); |
| 67 | if (ip == 0) |
| 68 | break; |
| 69 | |
| 70 | unw_get_sp(info, &sp); |
| 71 | unw_get_bsp(info, &bsp); |
| 72 | snprintf(buf, sizeof(buf), |
| 73 | " [<%016lx>] %%s\n" |
| 74 | " sp=%016lx bsp=%016lx\n", |
| 75 | ip, sp, bsp); |
| 76 | print_symbol(buf, ip); |
| 77 | } while (unw_unwind(info) >= 0); |
| 78 | } |
| 79 | |
| 80 | void |
| 81 | show_stack (struct task_struct *task, unsigned long *sp) |
| 82 | { |
| 83 | if (!task) |
| 84 | unw_init_running(ia64_do_show_stack, NULL); |
| 85 | else { |
| 86 | struct unw_frame_info info; |
| 87 | |
| 88 | unw_init_from_blocked_task(&info, task); |
| 89 | ia64_do_show_stack(&info, NULL); |
| 90 | } |
| 91 | } |
| 92 | |
| 93 | void |
| 94 | dump_stack (void) |
| 95 | { |
| 96 | show_stack(NULL, NULL); |
| 97 | } |
| 98 | |
| 99 | EXPORT_SYMBOL(dump_stack); |
| 100 | |
| 101 | void |
| 102 | show_regs (struct pt_regs *regs) |
| 103 | { |
| 104 | unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri; |
| 105 | |
| 106 | print_modules(); |
| 107 | printk("\nPid: %d, CPU %d, comm: %20s\n", current->pid, smp_processor_id(), current->comm); |
| 108 | printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s\n", |
| 109 | regs->cr_ipsr, regs->cr_ifs, ip, print_tainted()); |
| 110 | print_symbol("ip is at %s\n", ip); |
| 111 | printk("unat: %016lx pfs : %016lx rsc : %016lx\n", |
| 112 | regs->ar_unat, regs->ar_pfs, regs->ar_rsc); |
| 113 | printk("rnat: %016lx bsps: %016lx pr : %016lx\n", |
| 114 | regs->ar_rnat, regs->ar_bspstore, regs->pr); |
| 115 | printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n", |
| 116 | regs->loadrs, regs->ar_ccv, regs->ar_fpsr); |
| 117 | printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd); |
| 118 | printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs->b0, regs->b6, regs->b7); |
| 119 | printk("f6 : %05lx%016lx f7 : %05lx%016lx\n", |
| 120 | regs->f6.u.bits[1], regs->f6.u.bits[0], |
| 121 | regs->f7.u.bits[1], regs->f7.u.bits[0]); |
| 122 | printk("f8 : %05lx%016lx f9 : %05lx%016lx\n", |
| 123 | regs->f8.u.bits[1], regs->f8.u.bits[0], |
| 124 | regs->f9.u.bits[1], regs->f9.u.bits[0]); |
| 125 | printk("f10 : %05lx%016lx f11 : %05lx%016lx\n", |
| 126 | regs->f10.u.bits[1], regs->f10.u.bits[0], |
| 127 | regs->f11.u.bits[1], regs->f11.u.bits[0]); |
| 128 | |
| 129 | printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs->r1, regs->r2, regs->r3); |
| 130 | printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10); |
| 131 | printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13); |
| 132 | printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16); |
| 133 | printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19); |
| 134 | printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22); |
| 135 | printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25); |
| 136 | printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28); |
| 137 | printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31); |
| 138 | |
| 139 | if (user_mode(regs)) { |
| 140 | /* print the stacked registers */ |
| 141 | unsigned long val, *bsp, ndirty; |
| 142 | int i, sof, is_nat = 0; |
| 143 | |
| 144 | sof = regs->cr_ifs & 0x7f; /* size of frame */ |
| 145 | ndirty = (regs->loadrs >> 19); |
| 146 | bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty); |
| 147 | for (i = 0; i < sof; ++i) { |
| 148 | get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i)); |
| 149 | printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val, |
| 150 | ((i == sof - 1) || (i % 3) == 2) ? "\n" : " "); |
| 151 | } |
| 152 | } else |
| 153 | show_stack(NULL, NULL); |
| 154 | } |
| 155 | |
| 156 | void |
| 157 | do_notify_resume_user (sigset_t *oldset, struct sigscratch *scr, long in_syscall) |
| 158 | { |
| 159 | if (fsys_mode(current, &scr->pt)) { |
| 160 | /* defer signal-handling etc. until we return to privilege-level 0. */ |
| 161 | if (!ia64_psr(&scr->pt)->lp) |
| 162 | ia64_psr(&scr->pt)->lp = 1; |
| 163 | return; |
| 164 | } |
| 165 | |
| 166 | #ifdef CONFIG_PERFMON |
| 167 | if (current->thread.pfm_needs_checking) |
| 168 | pfm_handle_work(); |
| 169 | #endif |
| 170 | |
| 171 | /* deal with pending signal delivery */ |
| 172 | if (test_thread_flag(TIF_SIGPENDING)) |
| 173 | ia64_do_signal(oldset, scr, in_syscall); |
| 174 | } |
| 175 | |
Stephane Eranian | 8df5a50 | 2005-04-11 13:45:00 -0700 | [diff] [blame] | 176 | static int pal_halt = 1; |
| 177 | static int can_do_pal_halt = 1; |
| 178 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 179 | static int __init nohalt_setup(char * str) |
| 180 | { |
| 181 | pal_halt = 0; |
| 182 | return 1; |
| 183 | } |
| 184 | __setup("nohalt", nohalt_setup); |
| 185 | |
Tony Luck | a71f62e | 2005-05-03 16:21:45 -0700 | [diff] [blame] | 186 | void |
Stephane Eranian | 8df5a50 | 2005-04-11 13:45:00 -0700 | [diff] [blame] | 187 | update_pal_halt_status(int status) |
| 188 | { |
| 189 | can_do_pal_halt = pal_halt && status; |
| 190 | } |
| 191 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 192 | /* |
| 193 | * We use this if we don't have any better idle routine.. |
| 194 | */ |
| 195 | void |
| 196 | default_idle (void) |
| 197 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 198 | while (!need_resched()) |
Stephane Eranian | 8df5a50 | 2005-04-11 13:45:00 -0700 | [diff] [blame] | 199 | if (can_do_pal_halt) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 200 | safe_halt(); |
| 201 | else |
| 202 | cpu_relax(); |
| 203 | } |
| 204 | |
| 205 | #ifdef CONFIG_HOTPLUG_CPU |
| 206 | /* We don't actually take CPU down, just spin without interrupts. */ |
| 207 | static inline void play_dead(void) |
| 208 | { |
| 209 | extern void ia64_cpu_local_tick (void); |
Ashok Raj | b8d8b88 | 2005-04-22 14:44:40 -0700 | [diff] [blame] | 210 | unsigned int this_cpu = smp_processor_id(); |
| 211 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 212 | /* Ack it */ |
| 213 | __get_cpu_var(cpu_state) = CPU_DEAD; |
| 214 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 215 | max_xtp(); |
| 216 | local_irq_disable(); |
Ashok Raj | b8d8b88 | 2005-04-22 14:44:40 -0700 | [diff] [blame] | 217 | idle_task_exit(); |
| 218 | ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 219 | /* |
Ashok Raj | b8d8b88 | 2005-04-22 14:44:40 -0700 | [diff] [blame] | 220 | * The above is a point of no-return, the processor is |
| 221 | * expected to be in SAL loop now. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 222 | */ |
Ashok Raj | b8d8b88 | 2005-04-22 14:44:40 -0700 | [diff] [blame] | 223 | BUG(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 224 | } |
| 225 | #else |
| 226 | static inline void play_dead(void) |
| 227 | { |
| 228 | BUG(); |
| 229 | } |
| 230 | #endif /* CONFIG_HOTPLUG_CPU */ |
| 231 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 232 | void cpu_idle_wait(void) |
| 233 | { |
Zwane Mwaikambo | 7d5f9c0 | 2005-03-30 21:40:00 -0700 | [diff] [blame] | 234 | unsigned int cpu, this_cpu = get_cpu(); |
| 235 | cpumask_t map; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 236 | |
Zwane Mwaikambo | 7d5f9c0 | 2005-03-30 21:40:00 -0700 | [diff] [blame] | 237 | set_cpus_allowed(current, cpumask_of_cpu(this_cpu)); |
| 238 | put_cpu(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 239 | |
Zwane Mwaikambo | 7d5f9c0 | 2005-03-30 21:40:00 -0700 | [diff] [blame] | 240 | cpus_clear(map); |
| 241 | for_each_online_cpu(cpu) { |
| 242 | per_cpu(cpu_idle_state, cpu) = 1; |
| 243 | cpu_set(cpu, map); |
| 244 | } |
| 245 | |
| 246 | __get_cpu_var(cpu_idle_state) = 0; |
| 247 | |
| 248 | wmb(); |
| 249 | do { |
| 250 | ssleep(1); |
| 251 | for_each_online_cpu(cpu) { |
| 252 | if (cpu_isset(cpu, map) && !per_cpu(cpu_idle_state, cpu)) |
| 253 | cpu_clear(cpu, map); |
| 254 | } |
| 255 | cpus_and(map, map, cpu_online_map); |
| 256 | } while (!cpus_empty(map)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 257 | } |
| 258 | EXPORT_SYMBOL_GPL(cpu_idle_wait); |
| 259 | |
| 260 | void __attribute__((noreturn)) |
| 261 | cpu_idle (void) |
| 262 | { |
| 263 | void (*mark_idle)(int) = ia64_mark_idle; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 264 | |
| 265 | /* endless idle loop with no priority at all */ |
| 266 | while (1) { |
| 267 | #ifdef CONFIG_SMP |
| 268 | if (!need_resched()) |
| 269 | min_xtp(); |
| 270 | #endif |
| 271 | while (!need_resched()) { |
| 272 | void (*idle)(void); |
| 273 | |
Zwane Mwaikambo | 7d5f9c0 | 2005-03-30 21:40:00 -0700 | [diff] [blame] | 274 | if (__get_cpu_var(cpu_idle_state)) |
| 275 | __get_cpu_var(cpu_idle_state) = 0; |
| 276 | |
| 277 | rmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 278 | if (mark_idle) |
| 279 | (*mark_idle)(1); |
| 280 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 281 | idle = pm_idle; |
| 282 | if (!idle) |
| 283 | idle = default_idle; |
| 284 | (*idle)(); |
| 285 | } |
| 286 | |
| 287 | if (mark_idle) |
| 288 | (*mark_idle)(0); |
| 289 | |
| 290 | #ifdef CONFIG_SMP |
| 291 | normal_xtp(); |
| 292 | #endif |
| 293 | schedule(); |
| 294 | check_pgt_cache(); |
| 295 | if (cpu_is_offline(smp_processor_id())) |
| 296 | play_dead(); |
| 297 | } |
| 298 | } |
| 299 | |
| 300 | void |
| 301 | ia64_save_extra (struct task_struct *task) |
| 302 | { |
| 303 | #ifdef CONFIG_PERFMON |
| 304 | unsigned long info; |
| 305 | #endif |
| 306 | |
| 307 | if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0) |
| 308 | ia64_save_debug_regs(&task->thread.dbr[0]); |
| 309 | |
| 310 | #ifdef CONFIG_PERFMON |
| 311 | if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0) |
| 312 | pfm_save_regs(task); |
| 313 | |
| 314 | info = __get_cpu_var(pfm_syst_info); |
| 315 | if (info & PFM_CPUINFO_SYST_WIDE) |
| 316 | pfm_syst_wide_update_task(task, info, 0); |
| 317 | #endif |
| 318 | |
| 319 | #ifdef CONFIG_IA32_SUPPORT |
| 320 | if (IS_IA32_PROCESS(ia64_task_regs(task))) |
| 321 | ia32_save_state(task); |
| 322 | #endif |
| 323 | } |
| 324 | |
| 325 | void |
| 326 | ia64_load_extra (struct task_struct *task) |
| 327 | { |
| 328 | #ifdef CONFIG_PERFMON |
| 329 | unsigned long info; |
| 330 | #endif |
| 331 | |
| 332 | if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0) |
| 333 | ia64_load_debug_regs(&task->thread.dbr[0]); |
| 334 | |
| 335 | #ifdef CONFIG_PERFMON |
| 336 | if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0) |
| 337 | pfm_load_regs(task); |
| 338 | |
| 339 | info = __get_cpu_var(pfm_syst_info); |
| 340 | if (info & PFM_CPUINFO_SYST_WIDE) |
| 341 | pfm_syst_wide_update_task(task, info, 1); |
| 342 | #endif |
| 343 | |
| 344 | #ifdef CONFIG_IA32_SUPPORT |
| 345 | if (IS_IA32_PROCESS(ia64_task_regs(task))) |
| 346 | ia32_load_state(task); |
| 347 | #endif |
| 348 | } |
| 349 | |
| 350 | /* |
| 351 | * Copy the state of an ia-64 thread. |
| 352 | * |
| 353 | * We get here through the following call chain: |
| 354 | * |
| 355 | * from user-level: from kernel: |
| 356 | * |
| 357 | * <clone syscall> <some kernel call frames> |
| 358 | * sys_clone : |
| 359 | * do_fork do_fork |
| 360 | * copy_thread copy_thread |
| 361 | * |
| 362 | * This means that the stack layout is as follows: |
| 363 | * |
| 364 | * +---------------------+ (highest addr) |
| 365 | * | struct pt_regs | |
| 366 | * +---------------------+ |
| 367 | * | struct switch_stack | |
| 368 | * +---------------------+ |
| 369 | * | | |
| 370 | * | memory stack | |
| 371 | * | | <-- sp (lowest addr) |
| 372 | * +---------------------+ |
| 373 | * |
| 374 | * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an |
| 375 | * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register, |
| 376 | * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the |
| 377 | * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since |
| 378 | * the stack is page aligned and the page size is at least 4KB, this is always the case, |
| 379 | * so there is nothing to worry about. |
| 380 | */ |
| 381 | int |
| 382 | copy_thread (int nr, unsigned long clone_flags, |
| 383 | unsigned long user_stack_base, unsigned long user_stack_size, |
| 384 | struct task_struct *p, struct pt_regs *regs) |
| 385 | { |
| 386 | extern char ia64_ret_from_clone, ia32_ret_from_clone; |
| 387 | struct switch_stack *child_stack, *stack; |
| 388 | unsigned long rbs, child_rbs, rbs_size; |
| 389 | struct pt_regs *child_ptregs; |
| 390 | int retval = 0; |
| 391 | |
| 392 | #ifdef CONFIG_SMP |
| 393 | /* |
| 394 | * For SMP idle threads, fork_by_hand() calls do_fork with |
| 395 | * NULL regs. |
| 396 | */ |
| 397 | if (!regs) |
| 398 | return 0; |
| 399 | #endif |
| 400 | |
| 401 | stack = ((struct switch_stack *) regs) - 1; |
| 402 | |
| 403 | child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1; |
| 404 | child_stack = (struct switch_stack *) child_ptregs - 1; |
| 405 | |
| 406 | /* copy parent's switch_stack & pt_regs to child: */ |
| 407 | memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack)); |
| 408 | |
| 409 | rbs = (unsigned long) current + IA64_RBS_OFFSET; |
| 410 | child_rbs = (unsigned long) p + IA64_RBS_OFFSET; |
| 411 | rbs_size = stack->ar_bspstore - rbs; |
| 412 | |
| 413 | /* copy the parent's register backing store to the child: */ |
| 414 | memcpy((void *) child_rbs, (void *) rbs, rbs_size); |
| 415 | |
| 416 | if (likely(user_mode(child_ptregs))) { |
| 417 | if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs)) |
| 418 | child_ptregs->r13 = regs->r16; /* see sys_clone2() in entry.S */ |
| 419 | if (user_stack_base) { |
| 420 | child_ptregs->r12 = user_stack_base + user_stack_size - 16; |
| 421 | child_ptregs->ar_bspstore = user_stack_base; |
| 422 | child_ptregs->ar_rnat = 0; |
| 423 | child_ptregs->loadrs = 0; |
| 424 | } |
| 425 | } else { |
| 426 | /* |
| 427 | * Note: we simply preserve the relative position of |
| 428 | * the stack pointer here. There is no need to |
| 429 | * allocate a scratch area here, since that will have |
| 430 | * been taken care of by the caller of sys_clone() |
| 431 | * already. |
| 432 | */ |
| 433 | child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */ |
| 434 | child_ptregs->r13 = (unsigned long) p; /* set `current' pointer */ |
| 435 | } |
| 436 | child_stack->ar_bspstore = child_rbs + rbs_size; |
| 437 | if (IS_IA32_PROCESS(regs)) |
| 438 | child_stack->b0 = (unsigned long) &ia32_ret_from_clone; |
| 439 | else |
| 440 | child_stack->b0 = (unsigned long) &ia64_ret_from_clone; |
| 441 | |
| 442 | /* copy parts of thread_struct: */ |
| 443 | p->thread.ksp = (unsigned long) child_stack - 16; |
| 444 | |
| 445 | /* stop some PSR bits from being inherited. |
| 446 | * the psr.up/psr.pp bits must be cleared on fork but inherited on execve() |
| 447 | * therefore we must specify them explicitly here and not include them in |
| 448 | * IA64_PSR_BITS_TO_CLEAR. |
| 449 | */ |
| 450 | child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET) |
| 451 | & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP)); |
| 452 | |
| 453 | /* |
| 454 | * NOTE: The calling convention considers all floating point |
| 455 | * registers in the high partition (fph) to be scratch. Since |
| 456 | * the only way to get to this point is through a system call, |
| 457 | * we know that the values in fph are all dead. Hence, there |
| 458 | * is no need to inherit the fph state from the parent to the |
| 459 | * child and all we have to do is to make sure that |
| 460 | * IA64_THREAD_FPH_VALID is cleared in the child. |
| 461 | * |
| 462 | * XXX We could push this optimization a bit further by |
| 463 | * clearing IA64_THREAD_FPH_VALID on ANY system call. |
| 464 | * However, it's not clear this is worth doing. Also, it |
| 465 | * would be a slight deviation from the normal Linux system |
| 466 | * call behavior where scratch registers are preserved across |
| 467 | * system calls (unless used by the system call itself). |
| 468 | */ |
| 469 | # define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \ |
| 470 | | IA64_THREAD_PM_VALID) |
| 471 | # define THREAD_FLAGS_TO_SET 0 |
| 472 | p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR) |
| 473 | | THREAD_FLAGS_TO_SET); |
| 474 | ia64_drop_fpu(p); /* don't pick up stale state from a CPU's fph */ |
| 475 | #ifdef CONFIG_IA32_SUPPORT |
| 476 | /* |
| 477 | * If we're cloning an IA32 task then save the IA32 extra |
| 478 | * state from the current task to the new task |
| 479 | */ |
| 480 | if (IS_IA32_PROCESS(ia64_task_regs(current))) { |
| 481 | ia32_save_state(p); |
| 482 | if (clone_flags & CLONE_SETTLS) |
| 483 | retval = ia32_clone_tls(p, child_ptregs); |
| 484 | |
| 485 | /* Copy partially mapped page list */ |
| 486 | if (!retval) |
| 487 | retval = ia32_copy_partial_page_list(p, clone_flags); |
| 488 | } |
| 489 | #endif |
| 490 | |
| 491 | #ifdef CONFIG_PERFMON |
| 492 | if (current->thread.pfm_context) |
| 493 | pfm_inherit(p, child_ptregs); |
| 494 | #endif |
| 495 | return retval; |
| 496 | } |
| 497 | |
| 498 | static void |
| 499 | do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg) |
| 500 | { |
| 501 | unsigned long mask, sp, nat_bits = 0, ip, ar_rnat, urbs_end, cfm; |
| 502 | elf_greg_t *dst = arg; |
| 503 | struct pt_regs *pt; |
| 504 | char nat; |
| 505 | int i; |
| 506 | |
| 507 | memset(dst, 0, sizeof(elf_gregset_t)); /* don't leak any kernel bits to user-level */ |
| 508 | |
| 509 | if (unw_unwind_to_user(info) < 0) |
| 510 | return; |
| 511 | |
| 512 | unw_get_sp(info, &sp); |
| 513 | pt = (struct pt_regs *) (sp + 16); |
| 514 | |
| 515 | urbs_end = ia64_get_user_rbs_end(task, pt, &cfm); |
| 516 | |
| 517 | if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0) |
| 518 | return; |
| 519 | |
| 520 | ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end), |
| 521 | &ar_rnat); |
| 522 | |
| 523 | /* |
| 524 | * coredump format: |
| 525 | * r0-r31 |
| 526 | * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT) |
| 527 | * predicate registers (p0-p63) |
| 528 | * b0-b7 |
| 529 | * ip cfm user-mask |
| 530 | * ar.rsc ar.bsp ar.bspstore ar.rnat |
| 531 | * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec |
| 532 | */ |
| 533 | |
| 534 | /* r0 is zero */ |
| 535 | for (i = 1, mask = (1UL << i); i < 32; ++i) { |
| 536 | unw_get_gr(info, i, &dst[i], &nat); |
| 537 | if (nat) |
| 538 | nat_bits |= mask; |
| 539 | mask <<= 1; |
| 540 | } |
| 541 | dst[32] = nat_bits; |
| 542 | unw_get_pr(info, &dst[33]); |
| 543 | |
| 544 | for (i = 0; i < 8; ++i) |
| 545 | unw_get_br(info, i, &dst[34 + i]); |
| 546 | |
| 547 | unw_get_rp(info, &ip); |
| 548 | dst[42] = ip + ia64_psr(pt)->ri; |
| 549 | dst[43] = cfm; |
| 550 | dst[44] = pt->cr_ipsr & IA64_PSR_UM; |
| 551 | |
| 552 | unw_get_ar(info, UNW_AR_RSC, &dst[45]); |
| 553 | /* |
| 554 | * For bsp and bspstore, unw_get_ar() would return the kernel |
| 555 | * addresses, but we need the user-level addresses instead: |
| 556 | */ |
| 557 | dst[46] = urbs_end; /* note: by convention PT_AR_BSP points to the end of the urbs! */ |
| 558 | dst[47] = pt->ar_bspstore; |
| 559 | dst[48] = ar_rnat; |
| 560 | unw_get_ar(info, UNW_AR_CCV, &dst[49]); |
| 561 | unw_get_ar(info, UNW_AR_UNAT, &dst[50]); |
| 562 | unw_get_ar(info, UNW_AR_FPSR, &dst[51]); |
| 563 | dst[52] = pt->ar_pfs; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */ |
| 564 | unw_get_ar(info, UNW_AR_LC, &dst[53]); |
| 565 | unw_get_ar(info, UNW_AR_EC, &dst[54]); |
| 566 | unw_get_ar(info, UNW_AR_CSD, &dst[55]); |
| 567 | unw_get_ar(info, UNW_AR_SSD, &dst[56]); |
| 568 | } |
| 569 | |
| 570 | void |
| 571 | do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg) |
| 572 | { |
| 573 | elf_fpreg_t *dst = arg; |
| 574 | int i; |
| 575 | |
| 576 | memset(dst, 0, sizeof(elf_fpregset_t)); /* don't leak any "random" bits */ |
| 577 | |
| 578 | if (unw_unwind_to_user(info) < 0) |
| 579 | return; |
| 580 | |
| 581 | /* f0 is 0.0, f1 is 1.0 */ |
| 582 | |
| 583 | for (i = 2; i < 32; ++i) |
| 584 | unw_get_fr(info, i, dst + i); |
| 585 | |
| 586 | ia64_flush_fph(task); |
| 587 | if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0) |
| 588 | memcpy(dst + 32, task->thread.fph, 96*16); |
| 589 | } |
| 590 | |
| 591 | void |
| 592 | do_copy_regs (struct unw_frame_info *info, void *arg) |
| 593 | { |
| 594 | do_copy_task_regs(current, info, arg); |
| 595 | } |
| 596 | |
| 597 | void |
| 598 | do_dump_fpu (struct unw_frame_info *info, void *arg) |
| 599 | { |
| 600 | do_dump_task_fpu(current, info, arg); |
| 601 | } |
| 602 | |
| 603 | int |
| 604 | dump_task_regs(struct task_struct *task, elf_gregset_t *regs) |
| 605 | { |
| 606 | struct unw_frame_info tcore_info; |
| 607 | |
| 608 | if (current == task) { |
| 609 | unw_init_running(do_copy_regs, regs); |
| 610 | } else { |
| 611 | memset(&tcore_info, 0, sizeof(tcore_info)); |
| 612 | unw_init_from_blocked_task(&tcore_info, task); |
| 613 | do_copy_task_regs(task, &tcore_info, regs); |
| 614 | } |
| 615 | return 1; |
| 616 | } |
| 617 | |
| 618 | void |
| 619 | ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst) |
| 620 | { |
| 621 | unw_init_running(do_copy_regs, dst); |
| 622 | } |
| 623 | |
| 624 | int |
| 625 | dump_task_fpu (struct task_struct *task, elf_fpregset_t *dst) |
| 626 | { |
| 627 | struct unw_frame_info tcore_info; |
| 628 | |
| 629 | if (current == task) { |
| 630 | unw_init_running(do_dump_fpu, dst); |
| 631 | } else { |
| 632 | memset(&tcore_info, 0, sizeof(tcore_info)); |
| 633 | unw_init_from_blocked_task(&tcore_info, task); |
| 634 | do_dump_task_fpu(task, &tcore_info, dst); |
| 635 | } |
| 636 | return 1; |
| 637 | } |
| 638 | |
| 639 | int |
| 640 | dump_fpu (struct pt_regs *pt, elf_fpregset_t dst) |
| 641 | { |
| 642 | unw_init_running(do_dump_fpu, dst); |
| 643 | return 1; /* f0-f31 are always valid so we always return 1 */ |
| 644 | } |
| 645 | |
| 646 | long |
| 647 | sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp, |
| 648 | struct pt_regs *regs) |
| 649 | { |
| 650 | char *fname; |
| 651 | int error; |
| 652 | |
| 653 | fname = getname(filename); |
| 654 | error = PTR_ERR(fname); |
| 655 | if (IS_ERR(fname)) |
| 656 | goto out; |
| 657 | error = do_execve(fname, argv, envp, regs); |
| 658 | putname(fname); |
| 659 | out: |
| 660 | return error; |
| 661 | } |
| 662 | |
| 663 | pid_t |
| 664 | kernel_thread (int (*fn)(void *), void *arg, unsigned long flags) |
| 665 | { |
| 666 | extern void start_kernel_thread (void); |
| 667 | unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread; |
| 668 | struct { |
| 669 | struct switch_stack sw; |
| 670 | struct pt_regs pt; |
| 671 | } regs; |
| 672 | |
| 673 | memset(®s, 0, sizeof(regs)); |
| 674 | regs.pt.cr_iip = helper_fptr[0]; /* set entry point (IP) */ |
| 675 | regs.pt.r1 = helper_fptr[1]; /* set GP */ |
| 676 | regs.pt.r9 = (unsigned long) fn; /* 1st argument */ |
| 677 | regs.pt.r11 = (unsigned long) arg; /* 2nd argument */ |
| 678 | /* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read. */ |
| 679 | regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN; |
| 680 | regs.pt.cr_ifs = 1UL << 63; /* mark as valid, empty frame */ |
| 681 | regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR); |
| 682 | regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET; |
| 683 | regs.sw.pr = (1 << PRED_KERNEL_STACK); |
| 684 | return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, ®s.pt, 0, NULL, NULL); |
| 685 | } |
| 686 | EXPORT_SYMBOL(kernel_thread); |
| 687 | |
| 688 | /* This gets called from kernel_thread() via ia64_invoke_thread_helper(). */ |
| 689 | int |
| 690 | kernel_thread_helper (int (*fn)(void *), void *arg) |
| 691 | { |
| 692 | #ifdef CONFIG_IA32_SUPPORT |
| 693 | if (IS_IA32_PROCESS(ia64_task_regs(current))) { |
| 694 | /* A kernel thread is always a 64-bit process. */ |
| 695 | current->thread.map_base = DEFAULT_MAP_BASE; |
| 696 | current->thread.task_size = DEFAULT_TASK_SIZE; |
| 697 | ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob); |
| 698 | ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1); |
| 699 | } |
| 700 | #endif |
| 701 | return (*fn)(arg); |
| 702 | } |
| 703 | |
| 704 | /* |
| 705 | * Flush thread state. This is called when a thread does an execve(). |
| 706 | */ |
| 707 | void |
| 708 | flush_thread (void) |
| 709 | { |
| 710 | /* drop floating-point and debug-register state if it exists: */ |
| 711 | current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID); |
| 712 | ia64_drop_fpu(current); |
| 713 | if (IS_IA32_PROCESS(ia64_task_regs(current))) |
| 714 | ia32_drop_partial_page_list(current); |
| 715 | } |
| 716 | |
| 717 | /* |
| 718 | * Clean up state associated with current thread. This is called when |
| 719 | * the thread calls exit(). |
| 720 | */ |
| 721 | void |
| 722 | exit_thread (void) |
| 723 | { |
| 724 | ia64_drop_fpu(current); |
| 725 | #ifdef CONFIG_PERFMON |
| 726 | /* if needed, stop monitoring and flush state to perfmon context */ |
| 727 | if (current->thread.pfm_context) |
| 728 | pfm_exit_thread(current); |
| 729 | |
| 730 | /* free debug register resources */ |
| 731 | if (current->thread.flags & IA64_THREAD_DBG_VALID) |
| 732 | pfm_release_debug_registers(current); |
| 733 | #endif |
| 734 | if (IS_IA32_PROCESS(ia64_task_regs(current))) |
| 735 | ia32_drop_partial_page_list(current); |
| 736 | } |
| 737 | |
| 738 | unsigned long |
| 739 | get_wchan (struct task_struct *p) |
| 740 | { |
| 741 | struct unw_frame_info info; |
| 742 | unsigned long ip; |
| 743 | int count = 0; |
| 744 | |
| 745 | /* |
| 746 | * Note: p may not be a blocked task (it could be current or |
| 747 | * another process running on some other CPU. Rather than |
| 748 | * trying to determine if p is really blocked, we just assume |
| 749 | * it's blocked and rely on the unwind routines to fail |
| 750 | * gracefully if the process wasn't really blocked after all. |
| 751 | * --davidm 99/12/15 |
| 752 | */ |
| 753 | unw_init_from_blocked_task(&info, p); |
| 754 | do { |
| 755 | if (unw_unwind(&info) < 0) |
| 756 | return 0; |
| 757 | unw_get_ip(&info, &ip); |
| 758 | if (!in_sched_functions(ip)) |
| 759 | return ip; |
| 760 | } while (count++ < 16); |
| 761 | return 0; |
| 762 | } |
| 763 | |
| 764 | void |
| 765 | cpu_halt (void) |
| 766 | { |
| 767 | pal_power_mgmt_info_u_t power_info[8]; |
| 768 | unsigned long min_power; |
| 769 | int i, min_power_state; |
| 770 | |
| 771 | if (ia64_pal_halt_info(power_info) != 0) |
| 772 | return; |
| 773 | |
| 774 | min_power_state = 0; |
| 775 | min_power = power_info[0].pal_power_mgmt_info_s.power_consumption; |
| 776 | for (i = 1; i < 8; ++i) |
| 777 | if (power_info[i].pal_power_mgmt_info_s.im |
| 778 | && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) { |
| 779 | min_power = power_info[i].pal_power_mgmt_info_s.power_consumption; |
| 780 | min_power_state = i; |
| 781 | } |
| 782 | |
| 783 | while (1) |
| 784 | ia64_pal_halt(min_power_state); |
| 785 | } |
| 786 | |
| 787 | void |
| 788 | machine_restart (char *restart_cmd) |
| 789 | { |
| 790 | (*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL); |
| 791 | } |
| 792 | |
| 793 | EXPORT_SYMBOL(machine_restart); |
| 794 | |
| 795 | void |
| 796 | machine_halt (void) |
| 797 | { |
| 798 | cpu_halt(); |
| 799 | } |
| 800 | |
| 801 | EXPORT_SYMBOL(machine_halt); |
| 802 | |
| 803 | void |
| 804 | machine_power_off (void) |
| 805 | { |
| 806 | if (pm_power_off) |
| 807 | pm_power_off(); |
| 808 | machine_halt(); |
| 809 | } |
| 810 | |
| 811 | EXPORT_SYMBOL(machine_power_off); |