]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/fork.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * 'fork.c' contains the help-routines for the 'fork' system call | |
9 | * (see also entry.S and others). | |
10 | * Fork is rather simple, once you get the hang of it, but the memory | |
11 | * management can be a bitch. See 'mm/memory.c': 'copy_page_range()' | |
12 | */ | |
13 | ||
1da177e4 LT |
14 | #include <linux/slab.h> |
15 | #include <linux/init.h> | |
16 | #include <linux/unistd.h> | |
1da177e4 LT |
17 | #include <linux/module.h> |
18 | #include <linux/vmalloc.h> | |
19 | #include <linux/completion.h> | |
1da177e4 LT |
20 | #include <linux/personality.h> |
21 | #include <linux/mempolicy.h> | |
22 | #include <linux/sem.h> | |
23 | #include <linux/file.h> | |
9f3acc31 | 24 | #include <linux/fdtable.h> |
da9cbc87 | 25 | #include <linux/iocontext.h> |
1da177e4 LT |
26 | #include <linux/key.h> |
27 | #include <linux/binfmts.h> | |
28 | #include <linux/mman.h> | |
cddb8a5c | 29 | #include <linux/mmu_notifier.h> |
1da177e4 | 30 | #include <linux/fs.h> |
ab516013 | 31 | #include <linux/nsproxy.h> |
c59ede7b | 32 | #include <linux/capability.h> |
1da177e4 | 33 | #include <linux/cpu.h> |
b4f48b63 | 34 | #include <linux/cgroup.h> |
1da177e4 | 35 | #include <linux/security.h> |
a1e78772 | 36 | #include <linux/hugetlb.h> |
1da177e4 LT |
37 | #include <linux/swap.h> |
38 | #include <linux/syscalls.h> | |
39 | #include <linux/jiffies.h> | |
40 | #include <linux/futex.h> | |
8141c7f3 | 41 | #include <linux/compat.h> |
207205a2 | 42 | #include <linux/kthread.h> |
7c3ab738 | 43 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 44 | #include <linux/rcupdate.h> |
1da177e4 LT |
45 | #include <linux/ptrace.h> |
46 | #include <linux/mount.h> | |
47 | #include <linux/audit.h> | |
78fb7466 | 48 | #include <linux/memcontrol.h> |
f201ae23 | 49 | #include <linux/ftrace.h> |
1da177e4 LT |
50 | #include <linux/profile.h> |
51 | #include <linux/rmap.h> | |
f8af4da3 | 52 | #include <linux/ksm.h> |
1da177e4 | 53 | #include <linux/acct.h> |
8f0ab514 | 54 | #include <linux/tsacct_kern.h> |
9f46080c | 55 | #include <linux/cn_proc.h> |
ba96a0c8 | 56 | #include <linux/freezer.h> |
ca74e92b | 57 | #include <linux/delayacct.h> |
ad4ecbcb | 58 | #include <linux/taskstats_kern.h> |
0a425405 | 59 | #include <linux/random.h> |
522ed776 | 60 | #include <linux/tty.h> |
fd0928df | 61 | #include <linux/blkdev.h> |
5ad4e53b | 62 | #include <linux/fs_struct.h> |
7c9f8861 | 63 | #include <linux/magic.h> |
cdd6c482 | 64 | #include <linux/perf_event.h> |
42c4ab41 | 65 | #include <linux/posix-timers.h> |
8e7cac79 | 66 | #include <linux/user-return-notifier.h> |
3d5992d2 | 67 | #include <linux/oom.h> |
ba76149f | 68 | #include <linux/khugepaged.h> |
1da177e4 LT |
69 | |
70 | #include <asm/pgtable.h> | |
71 | #include <asm/pgalloc.h> | |
72 | #include <asm/uaccess.h> | |
73 | #include <asm/mmu_context.h> | |
74 | #include <asm/cacheflush.h> | |
75 | #include <asm/tlbflush.h> | |
76 | ||
ad8d75ff SR |
77 | #include <trace/events/sched.h> |
78 | ||
43d2b113 KH |
79 | #define CREATE_TRACE_POINTS |
80 | #include <trace/events/task.h> | |
81 | ||
1da177e4 LT |
82 | /* |
83 | * Protected counters by write_lock_irq(&tasklist_lock) | |
84 | */ | |
85 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 86 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
87 | |
88 | int max_threads; /* tunable limit on nr_threads */ | |
89 | ||
90 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
91 | ||
c59923a1 | 92 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
93 | |
94 | #ifdef CONFIG_PROVE_RCU | |
95 | int lockdep_tasklist_lock_is_held(void) | |
96 | { | |
97 | return lockdep_is_held(&tasklist_lock); | |
98 | } | |
99 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
100 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
101 | |
102 | int nr_processes(void) | |
103 | { | |
104 | int cpu; | |
105 | int total = 0; | |
106 | ||
1d510750 | 107 | for_each_possible_cpu(cpu) |
1da177e4 LT |
108 | total += per_cpu(process_counts, cpu); |
109 | ||
110 | return total; | |
111 | } | |
112 | ||
113 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
504f52b5 ED |
114 | # define alloc_task_struct_node(node) \ |
115 | kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node) | |
116 | # define free_task_struct(tsk) \ | |
117 | kmem_cache_free(task_struct_cachep, (tsk)) | |
e18b890b | 118 | static struct kmem_cache *task_struct_cachep; |
1da177e4 LT |
119 | #endif |
120 | ||
b69c49b7 | 121 | #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR |
b6a84016 ED |
122 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, |
123 | int node) | |
b69c49b7 FT |
124 | { |
125 | #ifdef CONFIG_DEBUG_STACK_USAGE | |
126 | gfp_t mask = GFP_KERNEL | __GFP_ZERO; | |
127 | #else | |
128 | gfp_t mask = GFP_KERNEL; | |
129 | #endif | |
b6a84016 ED |
130 | struct page *page = alloc_pages_node(node, mask, THREAD_SIZE_ORDER); |
131 | ||
132 | return page ? page_address(page) : NULL; | |
b69c49b7 FT |
133 | } |
134 | ||
135 | static inline void free_thread_info(struct thread_info *ti) | |
136 | { | |
137 | free_pages((unsigned long)ti, THREAD_SIZE_ORDER); | |
138 | } | |
139 | #endif | |
140 | ||
1da177e4 | 141 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 142 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
143 | |
144 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 145 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
146 | |
147 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 148 | struct kmem_cache *files_cachep; |
1da177e4 LT |
149 | |
150 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 151 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
152 | |
153 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 154 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
155 | |
156 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 157 | static struct kmem_cache *mm_cachep; |
1da177e4 | 158 | |
c6a7f572 KM |
159 | static void account_kernel_stack(struct thread_info *ti, int account) |
160 | { | |
161 | struct zone *zone = page_zone(virt_to_page(ti)); | |
162 | ||
163 | mod_zone_page_state(zone, NR_KERNEL_STACK, account); | |
164 | } | |
165 | ||
1da177e4 LT |
166 | void free_task(struct task_struct *tsk) |
167 | { | |
c6a7f572 | 168 | account_kernel_stack(tsk->stack, -1); |
f7e4217b | 169 | free_thread_info(tsk->stack); |
23f78d4a | 170 | rt_mutex_debug_task_free(tsk); |
fb52607a | 171 | ftrace_graph_exit_task(tsk); |
1da177e4 LT |
172 | free_task_struct(tsk); |
173 | } | |
174 | EXPORT_SYMBOL(free_task); | |
175 | ||
ea6d290c ON |
176 | static inline void free_signal_struct(struct signal_struct *sig) |
177 | { | |
97101eb4 | 178 | taskstats_tgid_free(sig); |
1c5354de | 179 | sched_autogroup_exit(sig); |
ea6d290c ON |
180 | kmem_cache_free(signal_cachep, sig); |
181 | } | |
182 | ||
183 | static inline void put_signal_struct(struct signal_struct *sig) | |
184 | { | |
1c5354de | 185 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
186 | free_signal_struct(sig); |
187 | } | |
188 | ||
158d9ebd | 189 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 190 | { |
270f722d | 191 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
192 | WARN_ON(atomic_read(&tsk->usage)); |
193 | WARN_ON(tsk == current); | |
194 | ||
e0e81739 | 195 | exit_creds(tsk); |
35df17c5 | 196 | delayacct_tsk_free(tsk); |
ea6d290c | 197 | put_signal_struct(tsk->signal); |
1da177e4 LT |
198 | |
199 | if (!profile_handoff_task(tsk)) | |
200 | free_task(tsk); | |
201 | } | |
77c100c8 | 202 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 203 | |
2adee9b3 SS |
204 | /* |
205 | * macro override instead of weak attribute alias, to workaround | |
206 | * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions. | |
207 | */ | |
208 | #ifndef arch_task_cache_init | |
209 | #define arch_task_cache_init() | |
210 | #endif | |
61c4628b | 211 | |
1da177e4 LT |
212 | void __init fork_init(unsigned long mempages) |
213 | { | |
214 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
215 | #ifndef ARCH_MIN_TASKALIGN | |
216 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
217 | #endif | |
218 | /* create a slab on which task_structs can be allocated */ | |
219 | task_struct_cachep = | |
220 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
2dff4405 | 221 | ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL); |
1da177e4 LT |
222 | #endif |
223 | ||
61c4628b SS |
224 | /* do the arch specific task caches init */ |
225 | arch_task_cache_init(); | |
226 | ||
1da177e4 LT |
227 | /* |
228 | * The default maximum number of threads is set to a safe | |
229 | * value: the thread structures can take up at most half | |
230 | * of memory. | |
231 | */ | |
232 | max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE); | |
233 | ||
234 | /* | |
235 | * we need to allow at least 20 threads to boot a system | |
236 | */ | |
fb0a685c | 237 | if (max_threads < 20) |
1da177e4 LT |
238 | max_threads = 20; |
239 | ||
240 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
241 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
242 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
243 | init_task.signal->rlim[RLIMIT_NPROC]; | |
244 | } | |
245 | ||
61c4628b SS |
246 | int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst, |
247 | struct task_struct *src) | |
248 | { | |
249 | *dst = *src; | |
250 | return 0; | |
251 | } | |
252 | ||
1da177e4 LT |
253 | static struct task_struct *dup_task_struct(struct task_struct *orig) |
254 | { | |
255 | struct task_struct *tsk; | |
256 | struct thread_info *ti; | |
7c9f8861 | 257 | unsigned long *stackend; |
207205a2 | 258 | int node = tsk_fork_get_node(orig); |
3e26c149 | 259 | int err; |
1da177e4 LT |
260 | |
261 | prepare_to_copy(orig); | |
262 | ||
504f52b5 | 263 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
264 | if (!tsk) |
265 | return NULL; | |
266 | ||
b6a84016 | 267 | ti = alloc_thread_info_node(tsk, node); |
1da177e4 LT |
268 | if (!ti) { |
269 | free_task_struct(tsk); | |
270 | return NULL; | |
271 | } | |
272 | ||
fb0a685c | 273 | err = arch_dup_task_struct(tsk, orig); |
61c4628b SS |
274 | if (err) |
275 | goto out; | |
276 | ||
f7e4217b | 277 | tsk->stack = ti; |
3e26c149 | 278 | |
10ebffde | 279 | setup_thread_stack(tsk, orig); |
8e7cac79 | 280 | clear_user_return_notifier(tsk); |
f26f9aff | 281 | clear_tsk_need_resched(tsk); |
7c9f8861 ES |
282 | stackend = end_of_stack(tsk); |
283 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
1da177e4 | 284 | |
0a425405 AV |
285 | #ifdef CONFIG_CC_STACKPROTECTOR |
286 | tsk->stack_canary = get_random_int(); | |
287 | #endif | |
288 | ||
fb0a685c DRO |
289 | /* |
290 | * One for us, one for whoever does the "release_task()" (usually | |
291 | * parent) | |
292 | */ | |
293 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 294 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 295 | tsk->btrace_seq = 0; |
6c5c9341 | 296 | #endif |
a0aa7f68 | 297 | tsk->splice_pipe = NULL; |
c6a7f572 KM |
298 | |
299 | account_kernel_stack(ti, 1); | |
300 | ||
1da177e4 | 301 | return tsk; |
61c4628b SS |
302 | |
303 | out: | |
304 | free_thread_info(ti); | |
305 | free_task_struct(tsk); | |
306 | return NULL; | |
1da177e4 LT |
307 | } |
308 | ||
309 | #ifdef CONFIG_MMU | |
a39bc516 | 310 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 311 | { |
297c5eee | 312 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
313 | struct rb_node **rb_link, *rb_parent; |
314 | int retval; | |
315 | unsigned long charge; | |
316 | struct mempolicy *pol; | |
317 | ||
318 | down_write(&oldmm->mmap_sem); | |
ec8c0446 | 319 | flush_cache_dup_mm(oldmm); |
ad339451 IM |
320 | /* |
321 | * Not linked in yet - no deadlock potential: | |
322 | */ | |
323 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 324 | |
1da177e4 LT |
325 | mm->locked_vm = 0; |
326 | mm->mmap = NULL; | |
327 | mm->mmap_cache = NULL; | |
328 | mm->free_area_cache = oldmm->mmap_base; | |
1363c3cd | 329 | mm->cached_hole_size = ~0UL; |
1da177e4 | 330 | mm->map_count = 0; |
94894244 | 331 | cpumask_clear(mm_cpumask(mm)); |
1da177e4 LT |
332 | mm->mm_rb = RB_ROOT; |
333 | rb_link = &mm->mm_rb.rb_node; | |
334 | rb_parent = NULL; | |
335 | pprev = &mm->mmap; | |
f8af4da3 | 336 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
337 | if (retval) |
338 | goto out; | |
339 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
340 | if (retval) |
341 | goto out; | |
1da177e4 | 342 | |
297c5eee | 343 | prev = NULL; |
fd3e42fc | 344 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
345 | struct file *file; |
346 | ||
347 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
3b6bfcdb HD |
348 | long pages = vma_pages(mpnt); |
349 | mm->total_vm -= pages; | |
ab50b8ed | 350 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
3b6bfcdb | 351 | -pages); |
1da177e4 LT |
352 | continue; |
353 | } | |
354 | charge = 0; | |
355 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
356 | unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT; | |
357 | if (security_vm_enough_memory(len)) | |
358 | goto fail_nomem; | |
359 | charge = len; | |
360 | } | |
e94b1766 | 361 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
362 | if (!tmp) |
363 | goto fail_nomem; | |
364 | *tmp = *mpnt; | |
5beb4930 | 365 | INIT_LIST_HEAD(&tmp->anon_vma_chain); |
846a16bf | 366 | pol = mpol_dup(vma_policy(mpnt)); |
1da177e4 LT |
367 | retval = PTR_ERR(pol); |
368 | if (IS_ERR(pol)) | |
369 | goto fail_nomem_policy; | |
370 | vma_set_policy(tmp, pol); | |
a247c3a9 | 371 | tmp->vm_mm = mm; |
5beb4930 RR |
372 | if (anon_vma_fork(tmp, mpnt)) |
373 | goto fail_nomem_anon_vma_fork; | |
1da177e4 | 374 | tmp->vm_flags &= ~VM_LOCKED; |
297c5eee | 375 | tmp->vm_next = tmp->vm_prev = NULL; |
1da177e4 LT |
376 | file = tmp->vm_file; |
377 | if (file) { | |
f3a43f3f | 378 | struct inode *inode = file->f_path.dentry->d_inode; |
b88ed205 HD |
379 | struct address_space *mapping = file->f_mapping; |
380 | ||
1da177e4 LT |
381 | get_file(file); |
382 | if (tmp->vm_flags & VM_DENYWRITE) | |
383 | atomic_dec(&inode->i_writecount); | |
3d48ae45 | 384 | mutex_lock(&mapping->i_mmap_mutex); |
b88ed205 HD |
385 | if (tmp->vm_flags & VM_SHARED) |
386 | mapping->i_mmap_writable++; | |
b88ed205 HD |
387 | flush_dcache_mmap_lock(mapping); |
388 | /* insert tmp into the share list, just after mpnt */ | |
1da177e4 | 389 | vma_prio_tree_add(tmp, mpnt); |
b88ed205 | 390 | flush_dcache_mmap_unlock(mapping); |
3d48ae45 | 391 | mutex_unlock(&mapping->i_mmap_mutex); |
1da177e4 LT |
392 | } |
393 | ||
a1e78772 MG |
394 | /* |
395 | * Clear hugetlb-related page reserves for children. This only | |
396 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
397 | * are not guaranteed to succeed, even if read-only | |
398 | */ | |
399 | if (is_vm_hugetlb_page(tmp)) | |
400 | reset_vma_resv_huge_pages(tmp); | |
401 | ||
1da177e4 | 402 | /* |
7ee78232 | 403 | * Link in the new vma and copy the page table entries. |
1da177e4 | 404 | */ |
1da177e4 LT |
405 | *pprev = tmp; |
406 | pprev = &tmp->vm_next; | |
297c5eee LT |
407 | tmp->vm_prev = prev; |
408 | prev = tmp; | |
1da177e4 LT |
409 | |
410 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
411 | rb_link = &tmp->vm_rb.rb_right; | |
412 | rb_parent = &tmp->vm_rb; | |
413 | ||
414 | mm->map_count++; | |
0b0db14c | 415 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
416 | |
417 | if (tmp->vm_ops && tmp->vm_ops->open) | |
418 | tmp->vm_ops->open(tmp); | |
419 | ||
420 | if (retval) | |
421 | goto out; | |
422 | } | |
d6dd61c8 JF |
423 | /* a new mm has just been created */ |
424 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 425 | retval = 0; |
1da177e4 | 426 | out: |
7ee78232 | 427 | up_write(&mm->mmap_sem); |
fd3e42fc | 428 | flush_tlb_mm(oldmm); |
1da177e4 LT |
429 | up_write(&oldmm->mmap_sem); |
430 | return retval; | |
5beb4930 RR |
431 | fail_nomem_anon_vma_fork: |
432 | mpol_put(pol); | |
1da177e4 LT |
433 | fail_nomem_policy: |
434 | kmem_cache_free(vm_area_cachep, tmp); | |
435 | fail_nomem: | |
436 | retval = -ENOMEM; | |
437 | vm_unacct_memory(charge); | |
438 | goto out; | |
439 | } | |
440 | ||
fb0a685c | 441 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
442 | { |
443 | mm->pgd = pgd_alloc(mm); | |
444 | if (unlikely(!mm->pgd)) | |
445 | return -ENOMEM; | |
446 | return 0; | |
447 | } | |
448 | ||
fb0a685c | 449 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 450 | { |
5e541973 | 451 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
452 | } |
453 | #else | |
454 | #define dup_mmap(mm, oldmm) (0) | |
455 | #define mm_alloc_pgd(mm) (0) | |
456 | #define mm_free_pgd(mm) | |
457 | #endif /* CONFIG_MMU */ | |
458 | ||
23ff4440 | 459 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 460 | |
e94b1766 | 461 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
462 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
463 | ||
4cb0e11b HK |
464 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
465 | ||
466 | static int __init coredump_filter_setup(char *s) | |
467 | { | |
468 | default_dump_filter = | |
469 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
470 | MMF_DUMP_FILTER_MASK; | |
471 | return 1; | |
472 | } | |
473 | ||
474 | __setup("coredump_filter=", coredump_filter_setup); | |
475 | ||
1da177e4 LT |
476 | #include <linux/init_task.h> |
477 | ||
858f0993 AD |
478 | static void mm_init_aio(struct mm_struct *mm) |
479 | { | |
480 | #ifdef CONFIG_AIO | |
481 | spin_lock_init(&mm->ioctx_lock); | |
482 | INIT_HLIST_HEAD(&mm->ioctx_list); | |
483 | #endif | |
484 | } | |
485 | ||
fb0a685c | 486 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) |
1da177e4 LT |
487 | { |
488 | atomic_set(&mm->mm_users, 1); | |
489 | atomic_set(&mm->mm_count, 1); | |
490 | init_rwsem(&mm->mmap_sem); | |
491 | INIT_LIST_HEAD(&mm->mmlist); | |
f8af4da3 HD |
492 | mm->flags = (current->mm) ? |
493 | (current->mm->flags & MMF_INIT_MASK) : default_dump_filter; | |
999d9fc1 | 494 | mm->core_state = NULL; |
1da177e4 | 495 | mm->nr_ptes = 0; |
d559db08 | 496 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 497 | spin_lock_init(&mm->page_table_lock); |
1da177e4 | 498 | mm->free_area_cache = TASK_UNMAPPED_BASE; |
1363c3cd | 499 | mm->cached_hole_size = ~0UL; |
858f0993 | 500 | mm_init_aio(mm); |
cf475ad2 | 501 | mm_init_owner(mm, p); |
1da177e4 LT |
502 | |
503 | if (likely(!mm_alloc_pgd(mm))) { | |
504 | mm->def_flags = 0; | |
cddb8a5c | 505 | mmu_notifier_mm_init(mm); |
1da177e4 LT |
506 | return mm; |
507 | } | |
78fb7466 | 508 | |
1da177e4 LT |
509 | free_mm(mm); |
510 | return NULL; | |
511 | } | |
512 | ||
513 | /* | |
514 | * Allocate and initialize an mm_struct. | |
515 | */ | |
fb0a685c | 516 | struct mm_struct *mm_alloc(void) |
1da177e4 | 517 | { |
fb0a685c | 518 | struct mm_struct *mm; |
1da177e4 LT |
519 | |
520 | mm = allocate_mm(); | |
de03c72c KM |
521 | if (!mm) |
522 | return NULL; | |
523 | ||
524 | memset(mm, 0, sizeof(*mm)); | |
6345d24d LT |
525 | mm_init_cpumask(mm); |
526 | return mm_init(mm, current); | |
1da177e4 LT |
527 | } |
528 | ||
529 | /* | |
530 | * Called when the last reference to the mm | |
531 | * is dropped: either by a lazy thread or by | |
532 | * mmput. Free the page directory and the mm. | |
533 | */ | |
7ad5b3a5 | 534 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
535 | { |
536 | BUG_ON(mm == &init_mm); | |
537 | mm_free_pgd(mm); | |
538 | destroy_context(mm); | |
cddb8a5c | 539 | mmu_notifier_mm_destroy(mm); |
e7a00c45 AA |
540 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
541 | VM_BUG_ON(mm->pmd_huge_pte); | |
542 | #endif | |
1da177e4 LT |
543 | free_mm(mm); |
544 | } | |
6d4e4c4f | 545 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 LT |
546 | |
547 | /* | |
548 | * Decrement the use count and release all resources for an mm. | |
549 | */ | |
550 | void mmput(struct mm_struct *mm) | |
551 | { | |
0ae26f1b AM |
552 | might_sleep(); |
553 | ||
1da177e4 LT |
554 | if (atomic_dec_and_test(&mm->mm_users)) { |
555 | exit_aio(mm); | |
1c2fb7a4 | 556 | ksm_exit(mm); |
ba76149f | 557 | khugepaged_exit(mm); /* must run before exit_mmap */ |
1da177e4 | 558 | exit_mmap(mm); |
925d1c40 | 559 | set_mm_exe_file(mm, NULL); |
1da177e4 LT |
560 | if (!list_empty(&mm->mmlist)) { |
561 | spin_lock(&mmlist_lock); | |
562 | list_del(&mm->mmlist); | |
563 | spin_unlock(&mmlist_lock); | |
564 | } | |
565 | put_swap_token(mm); | |
801460d0 HS |
566 | if (mm->binfmt) |
567 | module_put(mm->binfmt->module); | |
1da177e4 LT |
568 | mmdrop(mm); |
569 | } | |
570 | } | |
571 | EXPORT_SYMBOL_GPL(mmput); | |
572 | ||
38646013 JS |
573 | /* |
574 | * We added or removed a vma mapping the executable. The vmas are only mapped | |
575 | * during exec and are not mapped with the mmap system call. | |
576 | * Callers must hold down_write() on the mm's mmap_sem for these | |
577 | */ | |
578 | void added_exe_file_vma(struct mm_struct *mm) | |
579 | { | |
580 | mm->num_exe_file_vmas++; | |
581 | } | |
582 | ||
583 | void removed_exe_file_vma(struct mm_struct *mm) | |
584 | { | |
585 | mm->num_exe_file_vmas--; | |
fb0a685c | 586 | if ((mm->num_exe_file_vmas == 0) && mm->exe_file) { |
38646013 JS |
587 | fput(mm->exe_file); |
588 | mm->exe_file = NULL; | |
589 | } | |
590 | ||
591 | } | |
592 | ||
593 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) | |
594 | { | |
595 | if (new_exe_file) | |
596 | get_file(new_exe_file); | |
597 | if (mm->exe_file) | |
598 | fput(mm->exe_file); | |
599 | mm->exe_file = new_exe_file; | |
600 | mm->num_exe_file_vmas = 0; | |
601 | } | |
602 | ||
603 | struct file *get_mm_exe_file(struct mm_struct *mm) | |
604 | { | |
605 | struct file *exe_file; | |
606 | ||
607 | /* We need mmap_sem to protect against races with removal of | |
608 | * VM_EXECUTABLE vmas */ | |
609 | down_read(&mm->mmap_sem); | |
610 | exe_file = mm->exe_file; | |
611 | if (exe_file) | |
612 | get_file(exe_file); | |
613 | up_read(&mm->mmap_sem); | |
614 | return exe_file; | |
615 | } | |
616 | ||
617 | static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm) | |
618 | { | |
619 | /* It's safe to write the exe_file pointer without exe_file_lock because | |
620 | * this is called during fork when the task is not yet in /proc */ | |
621 | newmm->exe_file = get_mm_exe_file(oldmm); | |
622 | } | |
623 | ||
1da177e4 LT |
624 | /** |
625 | * get_task_mm - acquire a reference to the task's mm | |
626 | * | |
246bb0b1 | 627 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
628 | * this kernel workthread has transiently adopted a user mm with use_mm, |
629 | * to do its AIO) is not set and if so returns a reference to it, after | |
630 | * bumping up the use count. User must release the mm via mmput() | |
631 | * after use. Typically used by /proc and ptrace. | |
632 | */ | |
633 | struct mm_struct *get_task_mm(struct task_struct *task) | |
634 | { | |
635 | struct mm_struct *mm; | |
636 | ||
637 | task_lock(task); | |
638 | mm = task->mm; | |
639 | if (mm) { | |
246bb0b1 | 640 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
641 | mm = NULL; |
642 | else | |
643 | atomic_inc(&mm->mm_users); | |
644 | } | |
645 | task_unlock(task); | |
646 | return mm; | |
647 | } | |
648 | EXPORT_SYMBOL_GPL(get_task_mm); | |
649 | ||
8cdb878d CY |
650 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
651 | { | |
652 | struct mm_struct *mm; | |
653 | int err; | |
654 | ||
655 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
656 | if (err) | |
657 | return ERR_PTR(err); | |
658 | ||
659 | mm = get_task_mm(task); | |
660 | if (mm && mm != current->mm && | |
661 | !ptrace_may_access(task, mode)) { | |
662 | mmput(mm); | |
663 | mm = ERR_PTR(-EACCES); | |
664 | } | |
665 | mutex_unlock(&task->signal->cred_guard_mutex); | |
666 | ||
667 | return mm; | |
668 | } | |
669 | ||
1da177e4 LT |
670 | /* Please note the differences between mmput and mm_release. |
671 | * mmput is called whenever we stop holding onto a mm_struct, | |
672 | * error success whatever. | |
673 | * | |
674 | * mm_release is called after a mm_struct has been removed | |
675 | * from the current process. | |
676 | * | |
677 | * This difference is important for error handling, when we | |
678 | * only half set up a mm_struct for a new process and need to restore | |
679 | * the old one. Because we mmput the new mm_struct before | |
680 | * restoring the old one. . . | |
681 | * Eric Biederman 10 January 1998 | |
682 | */ | |
683 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
684 | { | |
685 | struct completion *vfork_done = tsk->vfork_done; | |
686 | ||
8141c7f3 LT |
687 | /* Get rid of any futexes when releasing the mm */ |
688 | #ifdef CONFIG_FUTEX | |
fc6b177d | 689 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 690 | exit_robust_list(tsk); |
fc6b177d PZ |
691 | tsk->robust_list = NULL; |
692 | } | |
8141c7f3 | 693 | #ifdef CONFIG_COMPAT |
fc6b177d | 694 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 695 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
696 | tsk->compat_robust_list = NULL; |
697 | } | |
8141c7f3 | 698 | #endif |
322a2c10 TG |
699 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
700 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
701 | #endif |
702 | ||
1da177e4 LT |
703 | /* Get rid of any cached register state */ |
704 | deactivate_mm(tsk, mm); | |
705 | ||
706 | /* notify parent sleeping on vfork() */ | |
707 | if (vfork_done) { | |
708 | tsk->vfork_done = NULL; | |
709 | complete(vfork_done); | |
710 | } | |
fec1d011 RM |
711 | |
712 | /* | |
713 | * If we're exiting normally, clear a user-space tid field if | |
714 | * requested. We leave this alone when dying by signal, to leave | |
715 | * the value intact in a core dump, and to save the unnecessary | |
716 | * trouble otherwise. Userland only wants this done for a sys_exit. | |
717 | */ | |
9c8a8228 ED |
718 | if (tsk->clear_child_tid) { |
719 | if (!(tsk->flags & PF_SIGNALED) && | |
720 | atomic_read(&mm->mm_users) > 1) { | |
721 | /* | |
722 | * We don't check the error code - if userspace has | |
723 | * not set up a proper pointer then tough luck. | |
724 | */ | |
725 | put_user(0, tsk->clear_child_tid); | |
726 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
727 | 1, NULL, NULL, 0); | |
728 | } | |
1da177e4 | 729 | tsk->clear_child_tid = NULL; |
1da177e4 LT |
730 | } |
731 | } | |
732 | ||
a0a7ec30 JD |
733 | /* |
734 | * Allocate a new mm structure and copy contents from the | |
735 | * mm structure of the passed in task structure. | |
736 | */ | |
402b0862 | 737 | struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
738 | { |
739 | struct mm_struct *mm, *oldmm = current->mm; | |
740 | int err; | |
741 | ||
742 | if (!oldmm) | |
743 | return NULL; | |
744 | ||
745 | mm = allocate_mm(); | |
746 | if (!mm) | |
747 | goto fail_nomem; | |
748 | ||
749 | memcpy(mm, oldmm, sizeof(*mm)); | |
6345d24d | 750 | mm_init_cpumask(mm); |
a0a7ec30 | 751 | |
7602bdf2 AC |
752 | /* Initializing for Swap token stuff */ |
753 | mm->token_priority = 0; | |
754 | mm->last_interval = 0; | |
755 | ||
e7a00c45 AA |
756 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
757 | mm->pmd_huge_pte = NULL; | |
758 | #endif | |
759 | ||
78fb7466 | 760 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
761 | goto fail_nomem; |
762 | ||
763 | if (init_new_context(tsk, mm)) | |
764 | goto fail_nocontext; | |
765 | ||
925d1c40 MH |
766 | dup_mm_exe_file(oldmm, mm); |
767 | ||
a0a7ec30 JD |
768 | err = dup_mmap(mm, oldmm); |
769 | if (err) | |
770 | goto free_pt; | |
771 | ||
772 | mm->hiwater_rss = get_mm_rss(mm); | |
773 | mm->hiwater_vm = mm->total_vm; | |
774 | ||
801460d0 HS |
775 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
776 | goto free_pt; | |
777 | ||
a0a7ec30 JD |
778 | return mm; |
779 | ||
780 | free_pt: | |
801460d0 HS |
781 | /* don't put binfmt in mmput, we haven't got module yet */ |
782 | mm->binfmt = NULL; | |
a0a7ec30 JD |
783 | mmput(mm); |
784 | ||
785 | fail_nomem: | |
786 | return NULL; | |
787 | ||
788 | fail_nocontext: | |
789 | /* | |
790 | * If init_new_context() failed, we cannot use mmput() to free the mm | |
791 | * because it calls destroy_context() | |
792 | */ | |
793 | mm_free_pgd(mm); | |
794 | free_mm(mm); | |
795 | return NULL; | |
796 | } | |
797 | ||
fb0a685c | 798 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 799 | { |
fb0a685c | 800 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
801 | int retval; |
802 | ||
803 | tsk->min_flt = tsk->maj_flt = 0; | |
804 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
805 | #ifdef CONFIG_DETECT_HUNG_TASK |
806 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
807 | #endif | |
1da177e4 LT |
808 | |
809 | tsk->mm = NULL; | |
810 | tsk->active_mm = NULL; | |
811 | ||
812 | /* | |
813 | * Are we cloning a kernel thread? | |
814 | * | |
815 | * We need to steal a active VM for that.. | |
816 | */ | |
817 | oldmm = current->mm; | |
818 | if (!oldmm) | |
819 | return 0; | |
820 | ||
821 | if (clone_flags & CLONE_VM) { | |
822 | atomic_inc(&oldmm->mm_users); | |
823 | mm = oldmm; | |
1da177e4 LT |
824 | goto good_mm; |
825 | } | |
826 | ||
827 | retval = -ENOMEM; | |
a0a7ec30 | 828 | mm = dup_mm(tsk); |
1da177e4 LT |
829 | if (!mm) |
830 | goto fail_nomem; | |
831 | ||
1da177e4 | 832 | good_mm: |
7602bdf2 AC |
833 | /* Initializing for Swap token stuff */ |
834 | mm->token_priority = 0; | |
835 | mm->last_interval = 0; | |
836 | ||
1da177e4 LT |
837 | tsk->mm = mm; |
838 | tsk->active_mm = mm; | |
839 | return 0; | |
840 | ||
1da177e4 LT |
841 | fail_nomem: |
842 | return retval; | |
1da177e4 LT |
843 | } |
844 | ||
a39bc516 | 845 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 846 | { |
498052bb | 847 | struct fs_struct *fs = current->fs; |
1da177e4 | 848 | if (clone_flags & CLONE_FS) { |
498052bb | 849 | /* tsk->fs is already what we want */ |
2a4419b5 | 850 | spin_lock(&fs->lock); |
498052bb | 851 | if (fs->in_exec) { |
2a4419b5 | 852 | spin_unlock(&fs->lock); |
498052bb AV |
853 | return -EAGAIN; |
854 | } | |
855 | fs->users++; | |
2a4419b5 | 856 | spin_unlock(&fs->lock); |
1da177e4 LT |
857 | return 0; |
858 | } | |
498052bb | 859 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
860 | if (!tsk->fs) |
861 | return -ENOMEM; | |
862 | return 0; | |
863 | } | |
864 | ||
fb0a685c | 865 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
866 | { |
867 | struct files_struct *oldf, *newf; | |
868 | int error = 0; | |
869 | ||
870 | /* | |
871 | * A background process may not have any files ... | |
872 | */ | |
873 | oldf = current->files; | |
874 | if (!oldf) | |
875 | goto out; | |
876 | ||
877 | if (clone_flags & CLONE_FILES) { | |
878 | atomic_inc(&oldf->count); | |
879 | goto out; | |
880 | } | |
881 | ||
a016f338 JD |
882 | newf = dup_fd(oldf, &error); |
883 | if (!newf) | |
884 | goto out; | |
885 | ||
886 | tsk->files = newf; | |
887 | error = 0; | |
888 | out: | |
889 | return error; | |
890 | } | |
891 | ||
fadad878 | 892 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
893 | { |
894 | #ifdef CONFIG_BLOCK | |
895 | struct io_context *ioc = current->io_context; | |
6e736be7 | 896 | struct io_context *new_ioc; |
fd0928df JA |
897 | |
898 | if (!ioc) | |
899 | return 0; | |
fadad878 JA |
900 | /* |
901 | * Share io context with parent, if CLONE_IO is set | |
902 | */ | |
903 | if (clone_flags & CLONE_IO) { | |
904 | tsk->io_context = ioc_task_link(ioc); | |
905 | if (unlikely(!tsk->io_context)) | |
906 | return -ENOMEM; | |
907 | } else if (ioprio_valid(ioc->ioprio)) { | |
6e736be7 TH |
908 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
909 | if (unlikely(!new_ioc)) | |
fd0928df JA |
910 | return -ENOMEM; |
911 | ||
6e736be7 | 912 | new_ioc->ioprio = ioc->ioprio; |
b2efa052 | 913 | put_io_context(new_ioc, NULL); |
fd0928df JA |
914 | } |
915 | #endif | |
916 | return 0; | |
917 | } | |
918 | ||
a39bc516 | 919 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
920 | { |
921 | struct sighand_struct *sig; | |
922 | ||
60348802 | 923 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
924 | atomic_inc(¤t->sighand->count); |
925 | return 0; | |
926 | } | |
927 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 928 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
929 | if (!sig) |
930 | return -ENOMEM; | |
1da177e4 LT |
931 | atomic_set(&sig->count, 1); |
932 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
933 | return 0; | |
934 | } | |
935 | ||
a7e5328a | 936 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 937 | { |
c81addc9 ON |
938 | if (atomic_dec_and_test(&sighand->count)) |
939 | kmem_cache_free(sighand_cachep, sighand); | |
940 | } | |
941 | ||
f06febc9 FM |
942 | |
943 | /* | |
944 | * Initialize POSIX timer handling for a thread group. | |
945 | */ | |
946 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
947 | { | |
78d7d407 JS |
948 | unsigned long cpu_limit; |
949 | ||
f06febc9 FM |
950 | /* Thread group counters. */ |
951 | thread_group_cputime_init(sig); | |
952 | ||
78d7d407 JS |
953 | cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
954 | if (cpu_limit != RLIM_INFINITY) { | |
955 | sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit); | |
6279a751 ON |
956 | sig->cputimer.running = 1; |
957 | } | |
958 | ||
f06febc9 FM |
959 | /* The timer lists. */ |
960 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
961 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
962 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
963 | } | |
964 | ||
a39bc516 | 965 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
966 | { |
967 | struct signal_struct *sig; | |
1da177e4 | 968 | |
4ab6c083 | 969 | if (clone_flags & CLONE_THREAD) |
490dea45 | 970 | return 0; |
490dea45 | 971 | |
a56704ef | 972 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
973 | tsk->signal = sig; |
974 | if (!sig) | |
975 | return -ENOMEM; | |
976 | ||
b3ac022c | 977 | sig->nr_threads = 1; |
1da177e4 | 978 | atomic_set(&sig->live, 1); |
b3ac022c | 979 | atomic_set(&sig->sigcnt, 1); |
1da177e4 | 980 | init_waitqueue_head(&sig->wait_chldexit); |
b3bfa0cb SB |
981 | if (clone_flags & CLONE_NEWPID) |
982 | sig->flags |= SIGNAL_UNKILLABLE; | |
db51aecc | 983 | sig->curr_target = tsk; |
1da177e4 LT |
984 | init_sigpending(&sig->shared_pending); |
985 | INIT_LIST_HEAD(&sig->posix_timers); | |
986 | ||
c9cb2e3d | 987 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 988 | sig->real_timer.function = it_real_fn; |
1da177e4 | 989 | |
1da177e4 LT |
990 | task_lock(current->group_leader); |
991 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
992 | task_unlock(current->group_leader); | |
993 | ||
6279a751 ON |
994 | posix_cpu_timers_init_group(sig); |
995 | ||
522ed776 | 996 | tty_audit_fork(sig); |
5091faa4 | 997 | sched_autogroup_fork(sig); |
522ed776 | 998 | |
4714d1d3 | 999 | #ifdef CONFIG_CGROUPS |
257058ae | 1000 | init_rwsem(&sig->group_rwsem); |
4714d1d3 BB |
1001 | #endif |
1002 | ||
28b83c51 | 1003 | sig->oom_adj = current->signal->oom_adj; |
a63d83f4 | 1004 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1005 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1006 | |
9b1bf12d KM |
1007 | mutex_init(&sig->cred_guard_mutex); |
1008 | ||
1da177e4 LT |
1009 | return 0; |
1010 | } | |
1011 | ||
a39bc516 | 1012 | static void copy_flags(unsigned long clone_flags, struct task_struct *p) |
1da177e4 LT |
1013 | { |
1014 | unsigned long new_flags = p->flags; | |
1015 | ||
21aa9af0 | 1016 | new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER); |
1da177e4 | 1017 | new_flags |= PF_FORKNOEXEC; |
09a05394 | 1018 | new_flags |= PF_STARTING; |
1da177e4 LT |
1019 | p->flags = new_flags; |
1020 | } | |
1021 | ||
17da2bd9 | 1022 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1023 | { |
1024 | current->clear_child_tid = tidptr; | |
1025 | ||
b488893a | 1026 | return task_pid_vnr(current); |
1da177e4 LT |
1027 | } |
1028 | ||
a39bc516 | 1029 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1030 | { |
1d615482 | 1031 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1032 | #ifdef CONFIG_RT_MUTEXES |
732375c6 | 1033 | plist_head_init(&p->pi_waiters); |
23f78d4a | 1034 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1035 | #endif |
1036 | } | |
1037 | ||
cf475ad2 BS |
1038 | #ifdef CONFIG_MM_OWNER |
1039 | void mm_init_owner(struct mm_struct *mm, struct task_struct *p) | |
1040 | { | |
1041 | mm->owner = p; | |
1042 | } | |
1043 | #endif /* CONFIG_MM_OWNER */ | |
1044 | ||
f06febc9 FM |
1045 | /* |
1046 | * Initialize POSIX timer handling for a single task. | |
1047 | */ | |
1048 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1049 | { | |
64861634 MS |
1050 | tsk->cputime_expires.prof_exp = 0; |
1051 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1052 | tsk->cputime_expires.sched_exp = 0; |
1053 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1054 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1055 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1056 | } | |
1057 | ||
1da177e4 LT |
1058 | /* |
1059 | * This creates a new process as a copy of the old one, | |
1060 | * but does not actually start it yet. | |
1061 | * | |
1062 | * It copies the registers, and all the appropriate | |
1063 | * parts of the process environment (as per the clone | |
1064 | * flags). The actual kick-off is left to the caller. | |
1065 | */ | |
36c8b586 IM |
1066 | static struct task_struct *copy_process(unsigned long clone_flags, |
1067 | unsigned long stack_start, | |
1068 | struct pt_regs *regs, | |
1069 | unsigned long stack_size, | |
36c8b586 | 1070 | int __user *child_tidptr, |
09a05394 RM |
1071 | struct pid *pid, |
1072 | int trace) | |
1da177e4 LT |
1073 | { |
1074 | int retval; | |
a24efe62 | 1075 | struct task_struct *p; |
b4f48b63 | 1076 | int cgroup_callbacks_done = 0; |
1da177e4 LT |
1077 | |
1078 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1079 | return ERR_PTR(-EINVAL); | |
1080 | ||
1081 | /* | |
1082 | * Thread groups must share signals as well, and detached threads | |
1083 | * can only be started up within the thread group. | |
1084 | */ | |
1085 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1086 | return ERR_PTR(-EINVAL); | |
1087 | ||
1088 | /* | |
1089 | * Shared signal handlers imply shared VM. By way of the above, | |
1090 | * thread groups also imply shared VM. Blocking this case allows | |
1091 | * for various simplifications in other code. | |
1092 | */ | |
1093 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1094 | return ERR_PTR(-EINVAL); | |
1095 | ||
123be07b SB |
1096 | /* |
1097 | * Siblings of global init remain as zombies on exit since they are | |
1098 | * not reaped by their parent (swapper). To solve this and to avoid | |
1099 | * multi-rooted process trees, prevent global and container-inits | |
1100 | * from creating siblings. | |
1101 | */ | |
1102 | if ((clone_flags & CLONE_PARENT) && | |
1103 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1104 | return ERR_PTR(-EINVAL); | |
1105 | ||
1da177e4 LT |
1106 | retval = security_task_create(clone_flags); |
1107 | if (retval) | |
1108 | goto fork_out; | |
1109 | ||
1110 | retval = -ENOMEM; | |
1111 | p = dup_task_struct(current); | |
1112 | if (!p) | |
1113 | goto fork_out; | |
1114 | ||
f7e8b616 SR |
1115 | ftrace_graph_init_task(p); |
1116 | ||
bea493a0 PZ |
1117 | rt_mutex_init_task(p); |
1118 | ||
d12c1a37 | 1119 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1120 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1121 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1122 | #endif | |
1da177e4 | 1123 | retval = -EAGAIN; |
3b11a1de | 1124 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1125 | task_rlimit(p, RLIMIT_NPROC)) { |
1da177e4 | 1126 | if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) && |
18b6e041 | 1127 | p->real_cred->user != INIT_USER) |
1da177e4 LT |
1128 | goto bad_fork_free; |
1129 | } | |
72fa5997 | 1130 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1131 | |
f1752eec DH |
1132 | retval = copy_creds(p, clone_flags); |
1133 | if (retval < 0) | |
1134 | goto bad_fork_free; | |
1da177e4 LT |
1135 | |
1136 | /* | |
1137 | * If multiple threads are within copy_process(), then this check | |
1138 | * triggers too late. This doesn't hurt, the check is only there | |
1139 | * to stop root fork bombs. | |
1140 | */ | |
04ec93fe | 1141 | retval = -EAGAIN; |
1da177e4 LT |
1142 | if (nr_threads >= max_threads) |
1143 | goto bad_fork_cleanup_count; | |
1144 | ||
a1261f54 | 1145 | if (!try_module_get(task_thread_info(p)->exec_domain->module)) |
1da177e4 LT |
1146 | goto bad_fork_cleanup_count; |
1147 | ||
1da177e4 | 1148 | p->did_exec = 0; |
ca74e92b | 1149 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
1da177e4 | 1150 | copy_flags(clone_flags, p); |
1da177e4 LT |
1151 | INIT_LIST_HEAD(&p->children); |
1152 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1153 | rcu_copy_process(p); |
1da177e4 LT |
1154 | p->vfork_done = NULL; |
1155 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1156 | |
1da177e4 LT |
1157 | init_sigpending(&p->pending); |
1158 | ||
64861634 MS |
1159 | p->utime = p->stime = p->gtime = 0; |
1160 | p->utimescaled = p->stimescaled = 0; | |
d99ca3b9 | 1161 | #ifndef CONFIG_VIRT_CPU_ACCOUNTING |
64861634 | 1162 | p->prev_utime = p->prev_stime = 0; |
d99ca3b9 | 1163 | #endif |
a3a2e76c KH |
1164 | #if defined(SPLIT_RSS_COUNTING) |
1165 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1166 | #endif | |
172ba844 | 1167 | |
6976675d AV |
1168 | p->default_timer_slack_ns = current->timer_slack_ns; |
1169 | ||
5995477a | 1170 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1171 | acct_clear_integrals(p); |
1172 | ||
f06febc9 | 1173 | posix_cpu_timers_init(p); |
1da177e4 | 1174 | |
1da177e4 | 1175 | do_posix_clock_monotonic_gettime(&p->start_time); |
924b42d5 TJ |
1176 | p->real_start_time = p->start_time; |
1177 | monotonic_to_bootbased(&p->real_start_time); | |
1da177e4 | 1178 | p->io_context = NULL; |
1da177e4 | 1179 | p->audit_context = NULL; |
4714d1d3 | 1180 | if (clone_flags & CLONE_THREAD) |
257058ae | 1181 | threadgroup_change_begin(current); |
b4f48b63 | 1182 | cgroup_fork(p); |
1da177e4 | 1183 | #ifdef CONFIG_NUMA |
846a16bf | 1184 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1185 | if (IS_ERR(p->mempolicy)) { |
1186 | retval = PTR_ERR(p->mempolicy); | |
1187 | p->mempolicy = NULL; | |
1188 | goto bad_fork_cleanup_cgroup; | |
1189 | } | |
c61afb18 | 1190 | mpol_fix_fork_child_flag(p); |
1da177e4 | 1191 | #endif |
778d3b0f MH |
1192 | #ifdef CONFIG_CPUSETS |
1193 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1194 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
1195 | #endif | |
de30a2b3 IM |
1196 | #ifdef CONFIG_TRACE_IRQFLAGS |
1197 | p->irq_events = 0; | |
b36e4758 RK |
1198 | #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW |
1199 | p->hardirqs_enabled = 1; | |
1200 | #else | |
de30a2b3 | 1201 | p->hardirqs_enabled = 0; |
b36e4758 | 1202 | #endif |
de30a2b3 IM |
1203 | p->hardirq_enable_ip = 0; |
1204 | p->hardirq_enable_event = 0; | |
1205 | p->hardirq_disable_ip = _THIS_IP_; | |
1206 | p->hardirq_disable_event = 0; | |
1207 | p->softirqs_enabled = 1; | |
1208 | p->softirq_enable_ip = _THIS_IP_; | |
1209 | p->softirq_enable_event = 0; | |
1210 | p->softirq_disable_ip = 0; | |
1211 | p->softirq_disable_event = 0; | |
1212 | p->hardirq_context = 0; | |
1213 | p->softirq_context = 0; | |
1214 | #endif | |
fbb9ce95 IM |
1215 | #ifdef CONFIG_LOCKDEP |
1216 | p->lockdep_depth = 0; /* no locks held yet */ | |
1217 | p->curr_chain_key = 0; | |
1218 | p->lockdep_recursion = 0; | |
1219 | #endif | |
1da177e4 | 1220 | |
408894ee IM |
1221 | #ifdef CONFIG_DEBUG_MUTEXES |
1222 | p->blocked_on = NULL; /* not blocked yet */ | |
1223 | #endif | |
569b846d KH |
1224 | #ifdef CONFIG_CGROUP_MEM_RES_CTLR |
1225 | p->memcg_batch.do_batch = 0; | |
1226 | p->memcg_batch.memcg = NULL; | |
1227 | #endif | |
0f481406 | 1228 | |
3c90e6e9 | 1229 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
3e51e3ed | 1230 | sched_fork(p); |
6ab423e0 | 1231 | |
cdd6c482 | 1232 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1233 | if (retval) |
1234 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1235 | retval = audit_alloc(p); |
1236 | if (retval) | |
f1752eec | 1237 | goto bad_fork_cleanup_policy; |
1da177e4 | 1238 | /* copy all the process information */ |
fb0a685c DRO |
1239 | retval = copy_semundo(clone_flags, p); |
1240 | if (retval) | |
1da177e4 | 1241 | goto bad_fork_cleanup_audit; |
fb0a685c DRO |
1242 | retval = copy_files(clone_flags, p); |
1243 | if (retval) | |
1da177e4 | 1244 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1245 | retval = copy_fs(clone_flags, p); |
1246 | if (retval) | |
1da177e4 | 1247 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1248 | retval = copy_sighand(clone_flags, p); |
1249 | if (retval) | |
1da177e4 | 1250 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1251 | retval = copy_signal(clone_flags, p); |
1252 | if (retval) | |
1da177e4 | 1253 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1254 | retval = copy_mm(clone_flags, p); |
1255 | if (retval) | |
1da177e4 | 1256 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1257 | retval = copy_namespaces(clone_flags, p); |
1258 | if (retval) | |
d84f4f99 | 1259 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1260 | retval = copy_io(clone_flags, p); |
1261 | if (retval) | |
fd0928df | 1262 | goto bad_fork_cleanup_namespaces; |
6f2c55b8 | 1263 | retval = copy_thread(clone_flags, stack_start, stack_size, p, regs); |
1da177e4 | 1264 | if (retval) |
fd0928df | 1265 | goto bad_fork_cleanup_io; |
1da177e4 | 1266 | |
425fb2b4 PE |
1267 | if (pid != &init_struct_pid) { |
1268 | retval = -ENOMEM; | |
61bce0f1 | 1269 | pid = alloc_pid(p->nsproxy->pid_ns); |
425fb2b4 | 1270 | if (!pid) |
fd0928df | 1271 | goto bad_fork_cleanup_io; |
425fb2b4 PE |
1272 | } |
1273 | ||
1274 | p->pid = pid_nr(pid); | |
1275 | p->tgid = p->pid; | |
1276 | if (clone_flags & CLONE_THREAD) | |
1277 | p->tgid = current->tgid; | |
1278 | ||
1da177e4 LT |
1279 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1280 | /* | |
1281 | * Clear TID on mm_release()? | |
1282 | */ | |
fb0a685c | 1283 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; |
73c10101 JA |
1284 | #ifdef CONFIG_BLOCK |
1285 | p->plug = NULL; | |
1286 | #endif | |
42b2dd0a | 1287 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1288 | p->robust_list = NULL; |
1289 | #ifdef CONFIG_COMPAT | |
1290 | p->compat_robust_list = NULL; | |
1291 | #endif | |
c87e2837 IM |
1292 | INIT_LIST_HEAD(&p->pi_state_list); |
1293 | p->pi_state_cache = NULL; | |
42b2dd0a | 1294 | #endif |
f9a3879a GM |
1295 | /* |
1296 | * sigaltstack should be cleared when sharing the same VM | |
1297 | */ | |
1298 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1299 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1300 | ||
1da177e4 | 1301 | /* |
6580807d ON |
1302 | * Syscall tracing and stepping should be turned off in the |
1303 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1304 | */ |
6580807d | 1305 | user_disable_single_step(p); |
1da177e4 | 1306 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1307 | #ifdef TIF_SYSCALL_EMU |
1308 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1309 | #endif | |
9745512c | 1310 | clear_all_latency_tracing(p); |
1da177e4 | 1311 | |
1da177e4 LT |
1312 | /* ok, now we should be set up.. */ |
1313 | p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL); | |
1314 | p->pdeath_signal = 0; | |
1315 | p->exit_state = 0; | |
1316 | ||
9d823e8f WF |
1317 | p->nr_dirtied = 0; |
1318 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1319 | p->dirty_paused_when = 0; |
9d823e8f | 1320 | |
1da177e4 LT |
1321 | /* |
1322 | * Ok, make it visible to the rest of the system. | |
1323 | * We dont wake it up yet. | |
1324 | */ | |
1325 | p->group_leader = p; | |
47e65328 | 1326 | INIT_LIST_HEAD(&p->thread_group); |
1da177e4 | 1327 | |
b4f48b63 PM |
1328 | /* Now that the task is set up, run cgroup callbacks if |
1329 | * necessary. We need to run them before the task is visible | |
1330 | * on the tasklist. */ | |
1331 | cgroup_fork_callbacks(p); | |
1332 | cgroup_callbacks_done = 1; | |
1333 | ||
1da177e4 LT |
1334 | /* Need tasklist lock for parent etc handling! */ |
1335 | write_lock_irq(&tasklist_lock); | |
1336 | ||
1da177e4 | 1337 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1338 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1339 | p->real_parent = current->real_parent; |
2d5516cb ON |
1340 | p->parent_exec_id = current->parent_exec_id; |
1341 | } else { | |
1da177e4 | 1342 | p->real_parent = current; |
2d5516cb ON |
1343 | p->parent_exec_id = current->self_exec_id; |
1344 | } | |
1da177e4 | 1345 | |
3f17da69 | 1346 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 ON |
1347 | |
1348 | /* | |
1349 | * Process group and session signals need to be delivered to just the | |
1350 | * parent before the fork or both the parent and the child after the | |
1351 | * fork. Restart if a signal comes in before we add the new process to | |
1352 | * it's process group. | |
1353 | * A fatal signal pending means that current will exit, so the new | |
1354 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1355 | */ |
23ff4440 | 1356 | recalc_sigpending(); |
4a2c7a78 ON |
1357 | if (signal_pending(current)) { |
1358 | spin_unlock(¤t->sighand->siglock); | |
1359 | write_unlock_irq(&tasklist_lock); | |
1360 | retval = -ERESTARTNOINTR; | |
f7e8b616 | 1361 | goto bad_fork_free_pid; |
4a2c7a78 ON |
1362 | } |
1363 | ||
1da177e4 | 1364 | if (clone_flags & CLONE_THREAD) { |
b3ac022c | 1365 | current->signal->nr_threads++; |
4ab6c083 | 1366 | atomic_inc(¤t->signal->live); |
b3ac022c | 1367 | atomic_inc(¤t->signal->sigcnt); |
1da177e4 | 1368 | p->group_leader = current->group_leader; |
47e65328 | 1369 | list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group); |
1da177e4 LT |
1370 | } |
1371 | ||
73b9ebfe | 1372 | if (likely(p->pid)) { |
4b9d33e6 | 1373 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe ON |
1374 | |
1375 | if (thread_group_leader(p)) { | |
45a68628 | 1376 | if (is_child_reaper(pid)) |
30e49c26 | 1377 | p->nsproxy->pid_ns->child_reaper = p; |
73b9ebfe | 1378 | |
fea9d175 | 1379 | p->signal->leader_pid = pid; |
9c9f4ded | 1380 | p->signal->tty = tty_kref_get(current->signal->tty); |
5cd17569 EB |
1381 | attach_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1382 | attach_pid(p, PIDTYPE_SID, task_session(current)); | |
9cd80bbb | 1383 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1384 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
909ea964 | 1385 | __this_cpu_inc(process_counts); |
73b9ebfe | 1386 | } |
85868995 | 1387 | attach_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1388 | nr_threads++; |
1da177e4 LT |
1389 | } |
1390 | ||
1da177e4 | 1391 | total_forks++; |
3f17da69 | 1392 | spin_unlock(¤t->sighand->siglock); |
1da177e4 | 1393 | write_unlock_irq(&tasklist_lock); |
c13cf856 | 1394 | proc_fork_connector(p); |
817929ec | 1395 | cgroup_post_fork(p); |
4714d1d3 | 1396 | if (clone_flags & CLONE_THREAD) |
257058ae | 1397 | threadgroup_change_end(current); |
cdd6c482 | 1398 | perf_event_fork(p); |
43d2b113 KH |
1399 | |
1400 | trace_task_newtask(p, clone_flags); | |
1401 | ||
1da177e4 LT |
1402 | return p; |
1403 | ||
425fb2b4 PE |
1404 | bad_fork_free_pid: |
1405 | if (pid != &init_struct_pid) | |
1406 | free_pid(pid); | |
fd0928df | 1407 | bad_fork_cleanup_io: |
b69f2292 LR |
1408 | if (p->io_context) |
1409 | exit_io_context(p); | |
ab516013 | 1410 | bad_fork_cleanup_namespaces: |
444f378b | 1411 | exit_task_namespaces(p); |
1da177e4 | 1412 | bad_fork_cleanup_mm: |
c9f01245 | 1413 | if (p->mm) |
1da177e4 LT |
1414 | mmput(p->mm); |
1415 | bad_fork_cleanup_signal: | |
4ab6c083 | 1416 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 1417 | free_signal_struct(p->signal); |
1da177e4 | 1418 | bad_fork_cleanup_sighand: |
a7e5328a | 1419 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1420 | bad_fork_cleanup_fs: |
1421 | exit_fs(p); /* blocking */ | |
1422 | bad_fork_cleanup_files: | |
1423 | exit_files(p); /* blocking */ | |
1424 | bad_fork_cleanup_semundo: | |
1425 | exit_sem(p); | |
1426 | bad_fork_cleanup_audit: | |
1427 | audit_free(p); | |
1da177e4 | 1428 | bad_fork_cleanup_policy: |
cdd6c482 | 1429 | perf_event_free_task(p); |
1da177e4 | 1430 | #ifdef CONFIG_NUMA |
f0be3d32 | 1431 | mpol_put(p->mempolicy); |
b4f48b63 | 1432 | bad_fork_cleanup_cgroup: |
1da177e4 | 1433 | #endif |
4714d1d3 | 1434 | if (clone_flags & CLONE_THREAD) |
257058ae | 1435 | threadgroup_change_end(current); |
b4f48b63 | 1436 | cgroup_exit(p, cgroup_callbacks_done); |
35df17c5 | 1437 | delayacct_tsk_free(p); |
a1261f54 | 1438 | module_put(task_thread_info(p)->exec_domain->module); |
1da177e4 | 1439 | bad_fork_cleanup_count: |
d84f4f99 | 1440 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1441 | exit_creds(p); |
1da177e4 LT |
1442 | bad_fork_free: |
1443 | free_task(p); | |
fe7d37d1 ON |
1444 | fork_out: |
1445 | return ERR_PTR(retval); | |
1da177e4 LT |
1446 | } |
1447 | ||
6b2fb3c6 | 1448 | noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs) |
1da177e4 LT |
1449 | { |
1450 | memset(regs, 0, sizeof(struct pt_regs)); | |
1451 | return regs; | |
1452 | } | |
1453 | ||
f106eee1 ON |
1454 | static inline void init_idle_pids(struct pid_link *links) |
1455 | { | |
1456 | enum pid_type type; | |
1457 | ||
1458 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1459 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
1460 | links[type].pid = &init_struct_pid; | |
1461 | } | |
1462 | } | |
1463 | ||
9abcf40b | 1464 | struct task_struct * __cpuinit fork_idle(int cpu) |
1da177e4 | 1465 | { |
36c8b586 | 1466 | struct task_struct *task; |
1da177e4 LT |
1467 | struct pt_regs regs; |
1468 | ||
30e49c26 | 1469 | task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL, |
09a05394 | 1470 | &init_struct_pid, 0); |
f106eee1 ON |
1471 | if (!IS_ERR(task)) { |
1472 | init_idle_pids(task->pids); | |
753ca4f3 | 1473 | init_idle(task, cpu); |
f106eee1 | 1474 | } |
73b9ebfe | 1475 | |
1da177e4 LT |
1476 | return task; |
1477 | } | |
1478 | ||
1da177e4 LT |
1479 | /* |
1480 | * Ok, this is the main fork-routine. | |
1481 | * | |
1482 | * It copies the process, and if successful kick-starts | |
1483 | * it and waits for it to finish using the VM if required. | |
1484 | */ | |
1485 | long do_fork(unsigned long clone_flags, | |
1486 | unsigned long stack_start, | |
1487 | struct pt_regs *regs, | |
1488 | unsigned long stack_size, | |
1489 | int __user *parent_tidptr, | |
1490 | int __user *child_tidptr) | |
1491 | { | |
1492 | struct task_struct *p; | |
1493 | int trace = 0; | |
92476d7f | 1494 | long nr; |
1da177e4 | 1495 | |
18b6e041 SH |
1496 | /* |
1497 | * Do some preliminary argument and permissions checking before we | |
1498 | * actually start allocating stuff | |
1499 | */ | |
1500 | if (clone_flags & CLONE_NEWUSER) { | |
1501 | if (clone_flags & CLONE_THREAD) | |
1502 | return -EINVAL; | |
1503 | /* hopefully this check will go away when userns support is | |
1504 | * complete | |
1505 | */ | |
7657d904 SH |
1506 | if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) || |
1507 | !capable(CAP_SETGID)) | |
18b6e041 SH |
1508 | return -EPERM; |
1509 | } | |
1510 | ||
09a05394 | 1511 | /* |
4b9d33e6 TH |
1512 | * Determine whether and which event to report to ptracer. When |
1513 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
1514 | * requested, no event is reported; otherwise, report if the event | |
1515 | * for the type of forking is enabled. | |
09a05394 | 1516 | */ |
4b9d33e6 TH |
1517 | if (likely(user_mode(regs)) && !(clone_flags & CLONE_UNTRACED)) { |
1518 | if (clone_flags & CLONE_VFORK) | |
1519 | trace = PTRACE_EVENT_VFORK; | |
1520 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
1521 | trace = PTRACE_EVENT_CLONE; | |
1522 | else | |
1523 | trace = PTRACE_EVENT_FORK; | |
1524 | ||
1525 | if (likely(!ptrace_event_enabled(current, trace))) | |
1526 | trace = 0; | |
1527 | } | |
1da177e4 | 1528 | |
a6f5e063 | 1529 | p = copy_process(clone_flags, stack_start, regs, stack_size, |
09a05394 | 1530 | child_tidptr, NULL, trace); |
1da177e4 LT |
1531 | /* |
1532 | * Do this prior waking up the new thread - the thread pointer | |
1533 | * might get invalid after that point, if the thread exits quickly. | |
1534 | */ | |
1535 | if (!IS_ERR(p)) { | |
1536 | struct completion vfork; | |
1537 | ||
0a16b607 MD |
1538 | trace_sched_process_fork(current, p); |
1539 | ||
6c5f3e7b | 1540 | nr = task_pid_vnr(p); |
30e49c26 PE |
1541 | |
1542 | if (clone_flags & CLONE_PARENT_SETTID) | |
1543 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1544 | |
1da177e4 LT |
1545 | if (clone_flags & CLONE_VFORK) { |
1546 | p->vfork_done = &vfork; | |
1547 | init_completion(&vfork); | |
1548 | } | |
1549 | ||
09a05394 RM |
1550 | /* |
1551 | * We set PF_STARTING at creation in case tracing wants to | |
1552 | * use this to distinguish a fully live task from one that | |
4b9d33e6 TH |
1553 | * hasn't finished SIGSTOP raising yet. Now we clear it |
1554 | * and set the child going. | |
09a05394 RM |
1555 | */ |
1556 | p->flags &= ~PF_STARTING; | |
1557 | ||
3e51e3ed | 1558 | wake_up_new_task(p); |
1da177e4 | 1559 | |
4b9d33e6 TH |
1560 | /* forking complete and child started to run, tell ptracer */ |
1561 | if (unlikely(trace)) | |
1562 | ptrace_event(trace, nr); | |
09a05394 | 1563 | |
1da177e4 | 1564 | if (clone_flags & CLONE_VFORK) { |
ba96a0c8 | 1565 | freezer_do_not_count(); |
1da177e4 | 1566 | wait_for_completion(&vfork); |
ba96a0c8 | 1567 | freezer_count(); |
a288eecc | 1568 | ptrace_event(PTRACE_EVENT_VFORK_DONE, nr); |
1da177e4 LT |
1569 | } |
1570 | } else { | |
92476d7f | 1571 | nr = PTR_ERR(p); |
1da177e4 | 1572 | } |
92476d7f | 1573 | return nr; |
1da177e4 LT |
1574 | } |
1575 | ||
5fd63b30 RT |
1576 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1577 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1578 | #endif | |
1579 | ||
51cc5068 | 1580 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1581 | { |
1582 | struct sighand_struct *sighand = data; | |
1583 | ||
a35afb83 | 1584 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1585 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1586 | } |
1587 | ||
1da177e4 LT |
1588 | void __init proc_caches_init(void) |
1589 | { | |
1590 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1591 | sizeof(struct sighand_struct), 0, | |
2dff4405 VN |
1592 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
1593 | SLAB_NOTRACK, sighand_ctor); | |
1da177e4 LT |
1594 | signal_cachep = kmem_cache_create("signal_cache", |
1595 | sizeof(struct signal_struct), 0, | |
2dff4405 | 1596 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1597 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 1598 | sizeof(struct files_struct), 0, |
2dff4405 | 1599 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1600 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 1601 | sizeof(struct fs_struct), 0, |
2dff4405 | 1602 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
6345d24d LT |
1603 | /* |
1604 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
1605 | * whole struct cpumask for the OFFSTACK case. We could change | |
1606 | * this to *only* allocate as much of it as required by the | |
1607 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
1608 | * is at the end of the structure, exactly for that reason. | |
1609 | */ | |
1da177e4 | 1610 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1611 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
2dff4405 | 1612 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
33e5d769 | 1613 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC); |
8feae131 | 1614 | mmap_init(); |
66577193 | 1615 | nsproxy_cache_init(); |
1da177e4 | 1616 | } |
cf2e340f | 1617 | |
cf2e340f | 1618 | /* |
9bfb23fc | 1619 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 1620 | */ |
9bfb23fc | 1621 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 1622 | { |
9bfb23fc ON |
1623 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
1624 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
1625 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET)) | |
1626 | return -EINVAL; | |
cf2e340f | 1627 | /* |
9bfb23fc ON |
1628 | * Not implemented, but pretend it works if there is nothing to |
1629 | * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND | |
1630 | * needs to unshare vm. | |
cf2e340f | 1631 | */ |
9bfb23fc ON |
1632 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
1633 | /* FIXME: get_task_mm() increments ->mm_users */ | |
1634 | if (atomic_read(¤t->mm->mm_users) > 1) | |
1635 | return -EINVAL; | |
1636 | } | |
cf2e340f JD |
1637 | |
1638 | return 0; | |
1639 | } | |
1640 | ||
1641 | /* | |
99d1419d | 1642 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1643 | */ |
1644 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1645 | { | |
1646 | struct fs_struct *fs = current->fs; | |
1647 | ||
498052bb AV |
1648 | if (!(unshare_flags & CLONE_FS) || !fs) |
1649 | return 0; | |
1650 | ||
1651 | /* don't need lock here; in the worst case we'll do useless copy */ | |
1652 | if (fs->users == 1) | |
1653 | return 0; | |
1654 | ||
1655 | *new_fsp = copy_fs_struct(fs); | |
1656 | if (!*new_fsp) | |
1657 | return -ENOMEM; | |
cf2e340f JD |
1658 | |
1659 | return 0; | |
1660 | } | |
1661 | ||
cf2e340f | 1662 | /* |
a016f338 | 1663 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1664 | */ |
1665 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1666 | { | |
1667 | struct files_struct *fd = current->files; | |
a016f338 | 1668 | int error = 0; |
cf2e340f JD |
1669 | |
1670 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1671 | (fd && atomic_read(&fd->count) > 1)) { |
1672 | *new_fdp = dup_fd(fd, &error); | |
1673 | if (!*new_fdp) | |
1674 | return error; | |
1675 | } | |
cf2e340f JD |
1676 | |
1677 | return 0; | |
1678 | } | |
1679 | ||
cf2e340f JD |
1680 | /* |
1681 | * unshare allows a process to 'unshare' part of the process | |
1682 | * context which was originally shared using clone. copy_* | |
1683 | * functions used by do_fork() cannot be used here directly | |
1684 | * because they modify an inactive task_struct that is being | |
1685 | * constructed. Here we are modifying the current, active, | |
1686 | * task_struct. | |
1687 | */ | |
6559eed8 | 1688 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 1689 | { |
cf2e340f | 1690 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 1691 | struct files_struct *fd, *new_fd = NULL; |
cf7b708c | 1692 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 1693 | int do_sysvsem = 0; |
9bfb23fc | 1694 | int err; |
cf2e340f | 1695 | |
9bfb23fc ON |
1696 | err = check_unshare_flags(unshare_flags); |
1697 | if (err) | |
06f9d4f9 EB |
1698 | goto bad_unshare_out; |
1699 | ||
9bfb23fc ON |
1700 | /* |
1701 | * If unsharing namespace, must also unshare filesystem information. | |
1702 | */ | |
1703 | if (unshare_flags & CLONE_NEWNS) | |
1704 | unshare_flags |= CLONE_FS; | |
6013f67f MS |
1705 | /* |
1706 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
1707 | * to a new ipc namespace, the semaphore arrays from the old | |
1708 | * namespace are unreachable. | |
1709 | */ | |
1710 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 1711 | do_sysvsem = 1; |
fb0a685c DRO |
1712 | err = unshare_fs(unshare_flags, &new_fs); |
1713 | if (err) | |
9bfb23fc | 1714 | goto bad_unshare_out; |
fb0a685c DRO |
1715 | err = unshare_fd(unshare_flags, &new_fd); |
1716 | if (err) | |
9bfb23fc | 1717 | goto bad_unshare_cleanup_fs; |
fb0a685c DRO |
1718 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, new_fs); |
1719 | if (err) | |
9edff4ab | 1720 | goto bad_unshare_cleanup_fd; |
c0b2fc31 | 1721 | |
9bfb23fc | 1722 | if (new_fs || new_fd || do_sysvsem || new_nsproxy) { |
9edff4ab MS |
1723 | if (do_sysvsem) { |
1724 | /* | |
1725 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
1726 | */ | |
1727 | exit_sem(current); | |
1728 | } | |
ab516013 | 1729 | |
c0b2fc31 | 1730 | if (new_nsproxy) { |
cf7b708c PE |
1731 | switch_task_namespaces(current, new_nsproxy); |
1732 | new_nsproxy = NULL; | |
c0b2fc31 | 1733 | } |
cf2e340f | 1734 | |
cf7b708c PE |
1735 | task_lock(current); |
1736 | ||
cf2e340f JD |
1737 | if (new_fs) { |
1738 | fs = current->fs; | |
2a4419b5 | 1739 | spin_lock(&fs->lock); |
cf2e340f | 1740 | current->fs = new_fs; |
498052bb AV |
1741 | if (--fs->users) |
1742 | new_fs = NULL; | |
1743 | else | |
1744 | new_fs = fs; | |
2a4419b5 | 1745 | spin_unlock(&fs->lock); |
cf2e340f JD |
1746 | } |
1747 | ||
cf2e340f JD |
1748 | if (new_fd) { |
1749 | fd = current->files; | |
1750 | current->files = new_fd; | |
1751 | new_fd = fd; | |
1752 | } | |
1753 | ||
1754 | task_unlock(current); | |
1755 | } | |
1756 | ||
c0b2fc31 | 1757 | if (new_nsproxy) |
444f378b | 1758 | put_nsproxy(new_nsproxy); |
c0b2fc31 | 1759 | |
cf2e340f JD |
1760 | bad_unshare_cleanup_fd: |
1761 | if (new_fd) | |
1762 | put_files_struct(new_fd); | |
1763 | ||
cf2e340f JD |
1764 | bad_unshare_cleanup_fs: |
1765 | if (new_fs) | |
498052bb | 1766 | free_fs_struct(new_fs); |
cf2e340f | 1767 | |
cf2e340f JD |
1768 | bad_unshare_out: |
1769 | return err; | |
1770 | } | |
3b125388 AV |
1771 | |
1772 | /* | |
1773 | * Helper to unshare the files of the current task. | |
1774 | * We don't want to expose copy_files internals to | |
1775 | * the exec layer of the kernel. | |
1776 | */ | |
1777 | ||
1778 | int unshare_files(struct files_struct **displaced) | |
1779 | { | |
1780 | struct task_struct *task = current; | |
50704516 | 1781 | struct files_struct *copy = NULL; |
3b125388 AV |
1782 | int error; |
1783 | ||
1784 | error = unshare_fd(CLONE_FILES, ©); | |
1785 | if (error || !copy) { | |
1786 | *displaced = NULL; | |
1787 | return error; | |
1788 | } | |
1789 | *displaced = task->files; | |
1790 | task_lock(task); | |
1791 | task->files = copy; | |
1792 | task_unlock(task); | |
1793 | return 0; | |
1794 | } |