"[PATCH] Fix leaks on /proc/{*/sched,sched_debug,timer_list,timer_stats}" and
[mmotm.git] / arch / mips / kernel / process.c
blobf3d73e1831c1567ccb674429eb83492feb2942e1
1 /*
2 * This file is subject to the terms and conditions of the GNU General Public
3 * License. See the file "COPYING" in the main directory of this archive
4 * for more details.
6 * Copyright (C) 1994 - 1999, 2000 by Ralf Baechle and others.
7 * Copyright (C) 2005, 2006 by Ralf Baechle (ralf@linux-mips.org)
8 * Copyright (C) 1999, 2000 Silicon Graphics, Inc.
9 * Copyright (C) 2004 Thiemo Seufer
11 #include <linux/errno.h>
12 #include <linux/module.h>
13 #include <linux/sched.h>
14 #include <linux/tick.h>
15 #include <linux/kernel.h>
16 #include <linux/mm.h>
17 #include <linux/stddef.h>
18 #include <linux/unistd.h>
19 #include <linux/ptrace.h>
20 #include <linux/slab.h>
21 #include <linux/mman.h>
22 #include <linux/personality.h>
23 #include <linux/sys.h>
24 #include <linux/user.h>
25 #include <linux/init.h>
26 #include <linux/completion.h>
27 #include <linux/kallsyms.h>
28 #include <linux/random.h>
30 #include <asm/asm.h>
31 #include <asm/bootinfo.h>
32 #include <asm/cpu.h>
33 #include <asm/dsp.h>
34 #include <asm/fpu.h>
35 #include <asm/pgtable.h>
36 #include <asm/system.h>
37 #include <asm/mipsregs.h>
38 #include <asm/processor.h>
39 #include <asm/uaccess.h>
40 #include <asm/io.h>
41 #include <asm/elf.h>
42 #include <asm/isadep.h>
43 #include <asm/inst.h>
44 #include <asm/stacktrace.h>
47 * The idle thread. There's no useful work to be done, so just try to conserve
48 * power and have a low exit latency (ie sit in a loop waiting for somebody to
49 * say that they'd like to reschedule)
51 void __noreturn cpu_idle(void)
53 int cpu;
55 /* CPU is going idle. */
56 cpu = smp_processor_id();
58 /* endless idle loop with no priority at all */
59 while (1) {
60 tick_nohz_stop_sched_tick(1);
61 while (!need_resched() && cpu_online(cpu)) {
62 #ifdef CONFIG_MIPS_MT_SMTC
63 extern void smtc_idle_loop_hook(void);
65 smtc_idle_loop_hook();
66 #endif
67 if (cpu_wait)
68 (*cpu_wait)();
70 #ifdef CONFIG_HOTPLUG_CPU
71 if (!cpu_online(cpu) && !cpu_isset(cpu, cpu_callin_map) &&
72 (system_state == SYSTEM_RUNNING ||
73 system_state == SYSTEM_BOOTING))
74 play_dead();
75 #endif
76 tick_nohz_restart_sched_tick();
77 preempt_enable_no_resched();
78 schedule();
79 preempt_disable();
83 asmlinkage void ret_from_fork(void);
85 void start_thread(struct pt_regs * regs, unsigned long pc, unsigned long sp)
87 unsigned long status;
89 /* New thread loses kernel privileges. */
90 status = regs->cp0_status & ~(ST0_CU0|ST0_CU1|ST0_FR|KU_MASK);
91 #ifdef CONFIG_64BIT
92 status |= test_thread_flag(TIF_32BIT_REGS) ? 0 : ST0_FR;
93 #endif
94 status |= KU_USER;
95 regs->cp0_status = status;
96 clear_used_math();
97 clear_fpu_owner();
98 if (cpu_has_dsp)
99 __init_dsp();
100 regs->cp0_epc = pc;
101 regs->regs[29] = sp;
102 current_thread_info()->addr_limit = USER_DS;
105 void exit_thread(void)
109 void flush_thread(void)
113 int copy_thread(unsigned long clone_flags, unsigned long usp,
114 unsigned long unused, struct task_struct *p, struct pt_regs *regs)
116 struct thread_info *ti = task_thread_info(p);
117 struct pt_regs *childregs;
118 unsigned long childksp;
119 p->set_child_tid = p->clear_child_tid = NULL;
121 childksp = (unsigned long)task_stack_page(p) + THREAD_SIZE - 32;
123 preempt_disable();
125 if (is_fpu_owner())
126 save_fp(p);
128 if (cpu_has_dsp)
129 save_dsp(p);
131 preempt_enable();
133 /* set up new TSS. */
134 childregs = (struct pt_regs *) childksp - 1;
135 /* Put the stack after the struct pt_regs. */
136 childksp = (unsigned long) childregs;
137 *childregs = *regs;
138 childregs->regs[7] = 0; /* Clear error flag */
140 childregs->regs[2] = 0; /* Child gets zero as return value */
141 regs->regs[2] = p->pid;
143 if (childregs->cp0_status & ST0_CU0) {
144 childregs->regs[28] = (unsigned long) ti;
145 childregs->regs[29] = childksp;
146 ti->addr_limit = KERNEL_DS;
147 } else {
148 childregs->regs[29] = usp;
149 ti->addr_limit = USER_DS;
151 p->thread.reg29 = (unsigned long) childregs;
152 p->thread.reg31 = (unsigned long) ret_from_fork;
155 * New tasks lose permission to use the fpu. This accelerates context
156 * switching for most programs since they don't use the fpu.
158 p->thread.cp0_status = read_c0_status() & ~(ST0_CU2|ST0_CU1);
159 childregs->cp0_status &= ~(ST0_CU2|ST0_CU1);
161 #ifdef CONFIG_MIPS_MT_SMTC
163 * SMTC restores TCStatus after Status, and the CU bits
164 * are aliased there.
166 childregs->cp0_tcstatus &= ~(ST0_CU2|ST0_CU1);
167 #endif
168 clear_tsk_thread_flag(p, TIF_USEDFPU);
170 #ifdef CONFIG_MIPS_MT_FPAFF
171 clear_tsk_thread_flag(p, TIF_FPUBOUND);
172 #endif /* CONFIG_MIPS_MT_FPAFF */
174 if (clone_flags & CLONE_SETTLS)
175 ti->tp_value = regs->regs[7];
177 return 0;
180 /* Fill in the fpu structure for a core dump.. */
181 int dump_fpu(struct pt_regs *regs, elf_fpregset_t *r)
183 memcpy(r, &current->thread.fpu, sizeof(current->thread.fpu));
185 return 1;
188 void elf_dump_regs(elf_greg_t *gp, struct pt_regs *regs)
190 int i;
192 for (i = 0; i < EF_R0; i++)
193 gp[i] = 0;
194 gp[EF_R0] = 0;
195 for (i = 1; i <= 31; i++)
196 gp[EF_R0 + i] = regs->regs[i];
197 gp[EF_R26] = 0;
198 gp[EF_R27] = 0;
199 gp[EF_LO] = regs->lo;
200 gp[EF_HI] = regs->hi;
201 gp[EF_CP0_EPC] = regs->cp0_epc;
202 gp[EF_CP0_BADVADDR] = regs->cp0_badvaddr;
203 gp[EF_CP0_STATUS] = regs->cp0_status;
204 gp[EF_CP0_CAUSE] = regs->cp0_cause;
205 #ifdef EF_UNUSED0
206 gp[EF_UNUSED0] = 0;
207 #endif
210 int dump_task_regs(struct task_struct *tsk, elf_gregset_t *regs)
212 elf_dump_regs(*regs, task_pt_regs(tsk));
213 return 1;
216 int dump_task_fpu(struct task_struct *t, elf_fpregset_t *fpr)
218 memcpy(fpr, &t->thread.fpu, sizeof(current->thread.fpu));
220 return 1;
224 * Create a kernel thread
226 static void __noreturn kernel_thread_helper(void *arg, int (*fn)(void *))
228 do_exit(fn(arg));
231 long kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
233 struct pt_regs regs;
235 memset(&regs, 0, sizeof(regs));
237 regs.regs[4] = (unsigned long) arg;
238 regs.regs[5] = (unsigned long) fn;
239 regs.cp0_epc = (unsigned long) kernel_thread_helper;
240 regs.cp0_status = read_c0_status();
241 #if defined(CONFIG_CPU_R3000) || defined(CONFIG_CPU_TX39XX)
242 regs.cp0_status = (regs.cp0_status & ~(ST0_KUP | ST0_IEP | ST0_IEC)) |
243 ((regs.cp0_status & (ST0_KUC | ST0_IEC)) << 2);
244 #else
245 regs.cp0_status |= ST0_EXL;
246 #endif
248 /* Ok, create the new process.. */
249 return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs, 0, NULL, NULL);
255 struct mips_frame_info {
256 void *func;
257 unsigned long func_size;
258 int frame_size;
259 int pc_offset;
262 static inline int is_ra_save_ins(union mips_instruction *ip)
264 /* sw / sd $ra, offset($sp) */
265 return (ip->i_format.opcode == sw_op || ip->i_format.opcode == sd_op) &&
266 ip->i_format.rs == 29 &&
267 ip->i_format.rt == 31;
270 static inline int is_jal_jalr_jr_ins(union mips_instruction *ip)
272 if (ip->j_format.opcode == jal_op)
273 return 1;
274 if (ip->r_format.opcode != spec_op)
275 return 0;
276 return ip->r_format.func == jalr_op || ip->r_format.func == jr_op;
279 static inline int is_sp_move_ins(union mips_instruction *ip)
281 /* addiu/daddiu sp,sp,-imm */
282 if (ip->i_format.rs != 29 || ip->i_format.rt != 29)
283 return 0;
284 if (ip->i_format.opcode == addiu_op || ip->i_format.opcode == daddiu_op)
285 return 1;
286 return 0;
289 static int get_frame_info(struct mips_frame_info *info)
291 union mips_instruction *ip = info->func;
292 unsigned max_insns = info->func_size / sizeof(union mips_instruction);
293 unsigned i;
295 info->pc_offset = -1;
296 info->frame_size = 0;
298 if (!ip)
299 goto err;
301 if (max_insns == 0)
302 max_insns = 128U; /* unknown function size */
303 max_insns = min(128U, max_insns);
305 for (i = 0; i < max_insns; i++, ip++) {
307 if (is_jal_jalr_jr_ins(ip))
308 break;
309 if (!info->frame_size) {
310 if (is_sp_move_ins(ip))
311 info->frame_size = - ip->i_format.simmediate;
312 continue;
314 if (info->pc_offset == -1 && is_ra_save_ins(ip)) {
315 info->pc_offset =
316 ip->i_format.simmediate / sizeof(long);
317 break;
320 if (info->frame_size && info->pc_offset >= 0) /* nested */
321 return 0;
322 if (info->pc_offset < 0) /* leaf */
323 return 1;
324 /* prologue seems boggus... */
325 err:
326 return -1;
329 static struct mips_frame_info schedule_mfi __read_mostly;
331 static int __init frame_info_init(void)
333 unsigned long size = 0;
334 #ifdef CONFIG_KALLSYMS
335 unsigned long ofs;
337 kallsyms_lookup_size_offset((unsigned long)schedule, &size, &ofs);
338 #endif
339 schedule_mfi.func = schedule;
340 schedule_mfi.func_size = size;
342 get_frame_info(&schedule_mfi);
345 * Without schedule() frame info, result given by
346 * thread_saved_pc() and get_wchan() are not reliable.
348 if (schedule_mfi.pc_offset < 0)
349 printk("Can't analyze schedule() prologue at %p\n", schedule);
351 return 0;
354 arch_initcall(frame_info_init);
357 * Return saved PC of a blocked thread.
359 unsigned long thread_saved_pc(struct task_struct *tsk)
361 struct thread_struct *t = &tsk->thread;
363 /* New born processes are a special case */
364 if (t->reg31 == (unsigned long) ret_from_fork)
365 return t->reg31;
366 if (schedule_mfi.pc_offset < 0)
367 return 0;
368 return ((unsigned long *)t->reg29)[schedule_mfi.pc_offset];
372 #ifdef CONFIG_KALLSYMS
373 /* used by show_backtrace() */
374 unsigned long unwind_stack(struct task_struct *task, unsigned long *sp,
375 unsigned long pc, unsigned long *ra)
377 unsigned long stack_page;
378 struct mips_frame_info info;
379 unsigned long size, ofs;
380 int leaf;
381 extern void ret_from_irq(void);
382 extern void ret_from_exception(void);
384 stack_page = (unsigned long)task_stack_page(task);
385 if (!stack_page)
386 return 0;
389 * If we reached the bottom of interrupt context,
390 * return saved pc in pt_regs.
392 if (pc == (unsigned long)ret_from_irq ||
393 pc == (unsigned long)ret_from_exception) {
394 struct pt_regs *regs;
395 if (*sp >= stack_page &&
396 *sp + sizeof(*regs) <= stack_page + THREAD_SIZE - 32) {
397 regs = (struct pt_regs *)*sp;
398 pc = regs->cp0_epc;
399 if (__kernel_text_address(pc)) {
400 *sp = regs->regs[29];
401 *ra = regs->regs[31];
402 return pc;
405 return 0;
407 if (!kallsyms_lookup_size_offset(pc, &size, &ofs))
408 return 0;
410 * Return ra if an exception occured at the first instruction
412 if (unlikely(ofs == 0)) {
413 pc = *ra;
414 *ra = 0;
415 return pc;
418 info.func = (void *)(pc - ofs);
419 info.func_size = ofs; /* analyze from start to ofs */
420 leaf = get_frame_info(&info);
421 if (leaf < 0)
422 return 0;
424 if (*sp < stack_page ||
425 *sp + info.frame_size > stack_page + THREAD_SIZE - 32)
426 return 0;
428 if (leaf)
430 * For some extreme cases, get_frame_info() can
431 * consider wrongly a nested function as a leaf
432 * one. In that cases avoid to return always the
433 * same value.
435 pc = pc != *ra ? *ra : 0;
436 else
437 pc = ((unsigned long *)(*sp))[info.pc_offset];
439 *sp += info.frame_size;
440 *ra = 0;
441 return __kernel_text_address(pc) ? pc : 0;
443 #endif
446 * get_wchan - a maintenance nightmare^W^Wpain in the ass ...
448 unsigned long get_wchan(struct task_struct *task)
450 unsigned long pc = 0;
451 #ifdef CONFIG_KALLSYMS
452 unsigned long sp;
453 unsigned long ra = 0;
454 #endif
456 if (!task || task == current || task->state == TASK_RUNNING)
457 goto out;
458 if (!task_stack_page(task))
459 goto out;
461 pc = thread_saved_pc(task);
463 #ifdef CONFIG_KALLSYMS
464 sp = task->thread.reg29 + schedule_mfi.frame_size;
466 while (in_sched_functions(pc))
467 pc = unwind_stack(task, &sp, pc, &ra);
468 #endif
470 out:
471 return pc;
475 * Don't forget that the stack pointer must be aligned on a 8 bytes
476 * boundary for 32-bits ABI and 16 bytes for 64-bits ABI.
478 unsigned long arch_align_stack(unsigned long sp)
480 if (!(current->personality & ADDR_NO_RANDOMIZE) && randomize_va_space)
481 sp -= get_random_int() & ~PAGE_MASK;
483 return sp & ALMASK;