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> |
e2cfabdf | 37 | #include <linux/seccomp.h> |
1da177e4 LT |
38 | #include <linux/swap.h> |
39 | #include <linux/syscalls.h> | |
40 | #include <linux/jiffies.h> | |
41 | #include <linux/futex.h> | |
8141c7f3 | 42 | #include <linux/compat.h> |
207205a2 | 43 | #include <linux/kthread.h> |
7c3ab738 | 44 | #include <linux/task_io_accounting_ops.h> |
ab2af1f5 | 45 | #include <linux/rcupdate.h> |
1da177e4 LT |
46 | #include <linux/ptrace.h> |
47 | #include <linux/mount.h> | |
48 | #include <linux/audit.h> | |
78fb7466 | 49 | #include <linux/memcontrol.h> |
f201ae23 | 50 | #include <linux/ftrace.h> |
5e2bf014 | 51 | #include <linux/proc_fs.h> |
1da177e4 LT |
52 | #include <linux/profile.h> |
53 | #include <linux/rmap.h> | |
f8af4da3 | 54 | #include <linux/ksm.h> |
1da177e4 | 55 | #include <linux/acct.h> |
8f0ab514 | 56 | #include <linux/tsacct_kern.h> |
9f46080c | 57 | #include <linux/cn_proc.h> |
ba96a0c8 | 58 | #include <linux/freezer.h> |
ca74e92b | 59 | #include <linux/delayacct.h> |
ad4ecbcb | 60 | #include <linux/taskstats_kern.h> |
0a425405 | 61 | #include <linux/random.h> |
522ed776 | 62 | #include <linux/tty.h> |
fd0928df | 63 | #include <linux/blkdev.h> |
5ad4e53b | 64 | #include <linux/fs_struct.h> |
7c9f8861 | 65 | #include <linux/magic.h> |
cdd6c482 | 66 | #include <linux/perf_event.h> |
42c4ab41 | 67 | #include <linux/posix-timers.h> |
8e7cac79 | 68 | #include <linux/user-return-notifier.h> |
3d5992d2 | 69 | #include <linux/oom.h> |
ba76149f | 70 | #include <linux/khugepaged.h> |
d80e731e | 71 | #include <linux/signalfd.h> |
0326f5a9 | 72 | #include <linux/uprobes.h> |
1da177e4 LT |
73 | |
74 | #include <asm/pgtable.h> | |
75 | #include <asm/pgalloc.h> | |
76 | #include <asm/uaccess.h> | |
77 | #include <asm/mmu_context.h> | |
78 | #include <asm/cacheflush.h> | |
79 | #include <asm/tlbflush.h> | |
80 | ||
ad8d75ff SR |
81 | #include <trace/events/sched.h> |
82 | ||
43d2b113 KH |
83 | #define CREATE_TRACE_POINTS |
84 | #include <trace/events/task.h> | |
85 | ||
1da177e4 LT |
86 | /* |
87 | * Protected counters by write_lock_irq(&tasklist_lock) | |
88 | */ | |
89 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
fb0a685c | 90 | int nr_threads; /* The idle threads do not count.. */ |
1da177e4 LT |
91 | |
92 | int max_threads; /* tunable limit on nr_threads */ | |
93 | ||
94 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
95 | ||
c59923a1 | 96 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
db1466b3 PM |
97 | |
98 | #ifdef CONFIG_PROVE_RCU | |
99 | int lockdep_tasklist_lock_is_held(void) | |
100 | { | |
101 | return lockdep_is_held(&tasklist_lock); | |
102 | } | |
103 | EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held); | |
104 | #endif /* #ifdef CONFIG_PROVE_RCU */ | |
1da177e4 LT |
105 | |
106 | int nr_processes(void) | |
107 | { | |
108 | int cpu; | |
109 | int total = 0; | |
110 | ||
1d510750 | 111 | for_each_possible_cpu(cpu) |
1da177e4 LT |
112 | total += per_cpu(process_counts, cpu); |
113 | ||
114 | return total; | |
115 | } | |
116 | ||
f19b9f74 AM |
117 | void __weak arch_release_task_struct(struct task_struct *tsk) |
118 | { | |
119 | } | |
120 | ||
f5e10287 | 121 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
e18b890b | 122 | static struct kmem_cache *task_struct_cachep; |
41101809 TG |
123 | |
124 | static inline struct task_struct *alloc_task_struct_node(int node) | |
125 | { | |
126 | return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node); | |
127 | } | |
128 | ||
41101809 TG |
129 | static inline void free_task_struct(struct task_struct *tsk) |
130 | { | |
41101809 TG |
131 | kmem_cache_free(task_struct_cachep, tsk); |
132 | } | |
1da177e4 LT |
133 | #endif |
134 | ||
f19b9f74 AM |
135 | void __weak arch_release_thread_info(struct thread_info *ti) |
136 | { | |
137 | } | |
138 | ||
f5e10287 | 139 | #ifndef CONFIG_ARCH_THREAD_INFO_ALLOCATOR |
41101809 | 140 | |
0d15d74a TG |
141 | /* |
142 | * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a | |
143 | * kmemcache based allocator. | |
144 | */ | |
145 | # if THREAD_SIZE >= PAGE_SIZE | |
b6a84016 ED |
146 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, |
147 | int node) | |
b69c49b7 | 148 | { |
2889f608 TG |
149 | struct page *page = alloc_pages_node(node, THREADINFO_GFP, |
150 | THREAD_SIZE_ORDER); | |
b6a84016 ED |
151 | |
152 | return page ? page_address(page) : NULL; | |
b69c49b7 FT |
153 | } |
154 | ||
155 | static inline void free_thread_info(struct thread_info *ti) | |
156 | { | |
157 | free_pages((unsigned long)ti, THREAD_SIZE_ORDER); | |
158 | } | |
0d15d74a TG |
159 | # else |
160 | static struct kmem_cache *thread_info_cache; | |
161 | ||
162 | static struct thread_info *alloc_thread_info_node(struct task_struct *tsk, | |
163 | int node) | |
164 | { | |
165 | return kmem_cache_alloc_node(thread_info_cache, THREADINFO_GFP, node); | |
166 | } | |
167 | ||
168 | static void free_thread_info(struct thread_info *ti) | |
169 | { | |
0d15d74a TG |
170 | kmem_cache_free(thread_info_cache, ti); |
171 | } | |
172 | ||
173 | void thread_info_cache_init(void) | |
174 | { | |
175 | thread_info_cache = kmem_cache_create("thread_info", THREAD_SIZE, | |
176 | THREAD_SIZE, 0, NULL); | |
177 | BUG_ON(thread_info_cache == NULL); | |
178 | } | |
179 | # endif | |
b69c49b7 FT |
180 | #endif |
181 | ||
1da177e4 | 182 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 183 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
184 | |
185 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 186 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
187 | |
188 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 189 | struct kmem_cache *files_cachep; |
1da177e4 LT |
190 | |
191 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 192 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
193 | |
194 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 195 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
196 | |
197 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 198 | static struct kmem_cache *mm_cachep; |
1da177e4 | 199 | |
c6a7f572 KM |
200 | static void account_kernel_stack(struct thread_info *ti, int account) |
201 | { | |
202 | struct zone *zone = page_zone(virt_to_page(ti)); | |
203 | ||
204 | mod_zone_page_state(zone, NR_KERNEL_STACK, account); | |
205 | } | |
206 | ||
1da177e4 LT |
207 | void free_task(struct task_struct *tsk) |
208 | { | |
c6a7f572 | 209 | account_kernel_stack(tsk->stack, -1); |
f19b9f74 | 210 | arch_release_thread_info(tsk->stack); |
f7e4217b | 211 | free_thread_info(tsk->stack); |
23f78d4a | 212 | rt_mutex_debug_task_free(tsk); |
fb52607a | 213 | ftrace_graph_exit_task(tsk); |
e2cfabdf | 214 | put_seccomp_filter(tsk); |
f19b9f74 | 215 | arch_release_task_struct(tsk); |
1da177e4 LT |
216 | free_task_struct(tsk); |
217 | } | |
218 | EXPORT_SYMBOL(free_task); | |
219 | ||
ea6d290c ON |
220 | static inline void free_signal_struct(struct signal_struct *sig) |
221 | { | |
97101eb4 | 222 | taskstats_tgid_free(sig); |
1c5354de | 223 | sched_autogroup_exit(sig); |
ea6d290c ON |
224 | kmem_cache_free(signal_cachep, sig); |
225 | } | |
226 | ||
227 | static inline void put_signal_struct(struct signal_struct *sig) | |
228 | { | |
1c5354de | 229 | if (atomic_dec_and_test(&sig->sigcnt)) |
ea6d290c ON |
230 | free_signal_struct(sig); |
231 | } | |
232 | ||
158d9ebd | 233 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 234 | { |
270f722d | 235 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
236 | WARN_ON(atomic_read(&tsk->usage)); |
237 | WARN_ON(tsk == current); | |
238 | ||
1a2a4d06 | 239 | security_task_free(tsk); |
e0e81739 | 240 | exit_creds(tsk); |
35df17c5 | 241 | delayacct_tsk_free(tsk); |
ea6d290c | 242 | put_signal_struct(tsk->signal); |
1da177e4 LT |
243 | |
244 | if (!profile_handoff_task(tsk)) | |
245 | free_task(tsk); | |
246 | } | |
77c100c8 | 247 | EXPORT_SYMBOL_GPL(__put_task_struct); |
1da177e4 | 248 | |
6c0a9fa6 | 249 | void __init __weak arch_task_cache_init(void) { } |
61c4628b | 250 | |
1da177e4 LT |
251 | void __init fork_init(unsigned long mempages) |
252 | { | |
f5e10287 | 253 | #ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR |
1da177e4 LT |
254 | #ifndef ARCH_MIN_TASKALIGN |
255 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
256 | #endif | |
257 | /* create a slab on which task_structs can be allocated */ | |
258 | task_struct_cachep = | |
259 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
2dff4405 | 260 | ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL); |
1da177e4 LT |
261 | #endif |
262 | ||
61c4628b SS |
263 | /* do the arch specific task caches init */ |
264 | arch_task_cache_init(); | |
265 | ||
1da177e4 LT |
266 | /* |
267 | * The default maximum number of threads is set to a safe | |
268 | * value: the thread structures can take up at most half | |
269 | * of memory. | |
270 | */ | |
271 | max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE); | |
272 | ||
273 | /* | |
274 | * we need to allow at least 20 threads to boot a system | |
275 | */ | |
fb0a685c | 276 | if (max_threads < 20) |
1da177e4 LT |
277 | max_threads = 20; |
278 | ||
279 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
280 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
281 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
282 | init_task.signal->rlim[RLIMIT_NPROC]; | |
283 | } | |
284 | ||
61c4628b SS |
285 | int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst, |
286 | struct task_struct *src) | |
287 | { | |
288 | *dst = *src; | |
289 | return 0; | |
290 | } | |
291 | ||
1da177e4 LT |
292 | static struct task_struct *dup_task_struct(struct task_struct *orig) |
293 | { | |
294 | struct task_struct *tsk; | |
295 | struct thread_info *ti; | |
7c9f8861 | 296 | unsigned long *stackend; |
207205a2 | 297 | int node = tsk_fork_get_node(orig); |
3e26c149 | 298 | int err; |
1da177e4 | 299 | |
504f52b5 | 300 | tsk = alloc_task_struct_node(node); |
1da177e4 LT |
301 | if (!tsk) |
302 | return NULL; | |
303 | ||
b6a84016 | 304 | ti = alloc_thread_info_node(tsk, node); |
f19b9f74 AM |
305 | if (!ti) |
306 | goto free_tsk; | |
1da177e4 | 307 | |
fb0a685c | 308 | err = arch_dup_task_struct(tsk, orig); |
164c33c6 | 309 | if (err) |
f19b9f74 | 310 | goto free_ti; |
164c33c6 | 311 | |
87bec58a AM |
312 | tsk->stack = ti; |
313 | ||
314 | setup_thread_stack(tsk, orig); | |
8e7cac79 | 315 | clear_user_return_notifier(tsk); |
f26f9aff | 316 | clear_tsk_need_resched(tsk); |
7c9f8861 ES |
317 | stackend = end_of_stack(tsk); |
318 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
1da177e4 | 319 | |
0a425405 AV |
320 | #ifdef CONFIG_CC_STACKPROTECTOR |
321 | tsk->stack_canary = get_random_int(); | |
322 | #endif | |
323 | ||
fb0a685c DRO |
324 | /* |
325 | * One for us, one for whoever does the "release_task()" (usually | |
326 | * parent) | |
327 | */ | |
328 | atomic_set(&tsk->usage, 2); | |
6c5c9341 | 329 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 330 | tsk->btrace_seq = 0; |
6c5c9341 | 331 | #endif |
a0aa7f68 | 332 | tsk->splice_pipe = NULL; |
5640f768 | 333 | tsk->task_frag.page = NULL; |
c6a7f572 KM |
334 | |
335 | account_kernel_stack(ti, 1); | |
336 | ||
1da177e4 | 337 | return tsk; |
61c4628b | 338 | |
f19b9f74 | 339 | free_ti: |
61c4628b | 340 | free_thread_info(ti); |
f19b9f74 | 341 | free_tsk: |
61c4628b SS |
342 | free_task_struct(tsk); |
343 | return NULL; | |
1da177e4 LT |
344 | } |
345 | ||
346 | #ifdef CONFIG_MMU | |
a39bc516 | 347 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 348 | { |
297c5eee | 349 | struct vm_area_struct *mpnt, *tmp, *prev, **pprev; |
1da177e4 LT |
350 | struct rb_node **rb_link, *rb_parent; |
351 | int retval; | |
352 | unsigned long charge; | |
353 | struct mempolicy *pol; | |
354 | ||
355 | down_write(&oldmm->mmap_sem); | |
ec8c0446 | 356 | flush_cache_dup_mm(oldmm); |
ad339451 IM |
357 | /* |
358 | * Not linked in yet - no deadlock potential: | |
359 | */ | |
360 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 361 | |
1da177e4 LT |
362 | mm->locked_vm = 0; |
363 | mm->mmap = NULL; | |
364 | mm->mmap_cache = NULL; | |
365 | mm->free_area_cache = oldmm->mmap_base; | |
1363c3cd | 366 | mm->cached_hole_size = ~0UL; |
1da177e4 | 367 | mm->map_count = 0; |
94894244 | 368 | cpumask_clear(mm_cpumask(mm)); |
1da177e4 LT |
369 | mm->mm_rb = RB_ROOT; |
370 | rb_link = &mm->mm_rb.rb_node; | |
371 | rb_parent = NULL; | |
372 | pprev = &mm->mmap; | |
f8af4da3 | 373 | retval = ksm_fork(mm, oldmm); |
ba76149f AA |
374 | if (retval) |
375 | goto out; | |
376 | retval = khugepaged_fork(mm, oldmm); | |
f8af4da3 HD |
377 | if (retval) |
378 | goto out; | |
1da177e4 | 379 | |
297c5eee | 380 | prev = NULL; |
fd3e42fc | 381 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
382 | struct file *file; |
383 | ||
384 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
ab50b8ed | 385 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
44de9d0c | 386 | -vma_pages(mpnt)); |
1da177e4 LT |
387 | continue; |
388 | } | |
389 | charge = 0; | |
390 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
b2412b7f HS |
391 | unsigned long len = vma_pages(mpnt); |
392 | ||
191c5424 | 393 | if (security_vm_enough_memory_mm(oldmm, len)) /* sic */ |
1da177e4 LT |
394 | goto fail_nomem; |
395 | charge = len; | |
396 | } | |
e94b1766 | 397 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
398 | if (!tmp) |
399 | goto fail_nomem; | |
400 | *tmp = *mpnt; | |
5beb4930 | 401 | INIT_LIST_HEAD(&tmp->anon_vma_chain); |
846a16bf | 402 | pol = mpol_dup(vma_policy(mpnt)); |
1da177e4 LT |
403 | retval = PTR_ERR(pol); |
404 | if (IS_ERR(pol)) | |
405 | goto fail_nomem_policy; | |
406 | vma_set_policy(tmp, pol); | |
a247c3a9 | 407 | tmp->vm_mm = mm; |
5beb4930 RR |
408 | if (anon_vma_fork(tmp, mpnt)) |
409 | goto fail_nomem_anon_vma_fork; | |
1da177e4 | 410 | tmp->vm_flags &= ~VM_LOCKED; |
297c5eee | 411 | tmp->vm_next = tmp->vm_prev = NULL; |
1da177e4 LT |
412 | file = tmp->vm_file; |
413 | if (file) { | |
f3a43f3f | 414 | struct inode *inode = file->f_path.dentry->d_inode; |
b88ed205 HD |
415 | struct address_space *mapping = file->f_mapping; |
416 | ||
1da177e4 LT |
417 | get_file(file); |
418 | if (tmp->vm_flags & VM_DENYWRITE) | |
419 | atomic_dec(&inode->i_writecount); | |
3d48ae45 | 420 | mutex_lock(&mapping->i_mmap_mutex); |
b88ed205 HD |
421 | if (tmp->vm_flags & VM_SHARED) |
422 | mapping->i_mmap_writable++; | |
b88ed205 HD |
423 | flush_dcache_mmap_lock(mapping); |
424 | /* insert tmp into the share list, just after mpnt */ | |
1da177e4 | 425 | vma_prio_tree_add(tmp, mpnt); |
b88ed205 | 426 | flush_dcache_mmap_unlock(mapping); |
3d48ae45 | 427 | mutex_unlock(&mapping->i_mmap_mutex); |
1da177e4 LT |
428 | } |
429 | ||
a1e78772 MG |
430 | /* |
431 | * Clear hugetlb-related page reserves for children. This only | |
432 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
433 | * are not guaranteed to succeed, even if read-only | |
434 | */ | |
435 | if (is_vm_hugetlb_page(tmp)) | |
436 | reset_vma_resv_huge_pages(tmp); | |
437 | ||
1da177e4 | 438 | /* |
7ee78232 | 439 | * Link in the new vma and copy the page table entries. |
1da177e4 | 440 | */ |
1da177e4 LT |
441 | *pprev = tmp; |
442 | pprev = &tmp->vm_next; | |
297c5eee LT |
443 | tmp->vm_prev = prev; |
444 | prev = tmp; | |
1da177e4 LT |
445 | |
446 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
447 | rb_link = &tmp->vm_rb.rb_right; | |
448 | rb_parent = &tmp->vm_rb; | |
449 | ||
450 | mm->map_count++; | |
0b0db14c | 451 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
452 | |
453 | if (tmp->vm_ops && tmp->vm_ops->open) | |
454 | tmp->vm_ops->open(tmp); | |
455 | ||
456 | if (retval) | |
457 | goto out; | |
682968e0 | 458 | |
c7a3a88c ON |
459 | if (file) |
460 | uprobe_mmap(tmp); | |
1da177e4 | 461 | } |
d6dd61c8 JF |
462 | /* a new mm has just been created */ |
463 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 464 | retval = 0; |
1da177e4 | 465 | out: |
7ee78232 | 466 | up_write(&mm->mmap_sem); |
fd3e42fc | 467 | flush_tlb_mm(oldmm); |
1da177e4 LT |
468 | up_write(&oldmm->mmap_sem); |
469 | return retval; | |
5beb4930 RR |
470 | fail_nomem_anon_vma_fork: |
471 | mpol_put(pol); | |
1da177e4 LT |
472 | fail_nomem_policy: |
473 | kmem_cache_free(vm_area_cachep, tmp); | |
474 | fail_nomem: | |
475 | retval = -ENOMEM; | |
476 | vm_unacct_memory(charge); | |
477 | goto out; | |
478 | } | |
479 | ||
fb0a685c | 480 | static inline int mm_alloc_pgd(struct mm_struct *mm) |
1da177e4 LT |
481 | { |
482 | mm->pgd = pgd_alloc(mm); | |
483 | if (unlikely(!mm->pgd)) | |
484 | return -ENOMEM; | |
485 | return 0; | |
486 | } | |
487 | ||
fb0a685c | 488 | static inline void mm_free_pgd(struct mm_struct *mm) |
1da177e4 | 489 | { |
5e541973 | 490 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
491 | } |
492 | #else | |
493 | #define dup_mmap(mm, oldmm) (0) | |
494 | #define mm_alloc_pgd(mm) (0) | |
495 | #define mm_free_pgd(mm) | |
496 | #endif /* CONFIG_MMU */ | |
497 | ||
23ff4440 | 498 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 499 | |
e94b1766 | 500 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
501 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
502 | ||
4cb0e11b HK |
503 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
504 | ||
505 | static int __init coredump_filter_setup(char *s) | |
506 | { | |
507 | default_dump_filter = | |
508 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
509 | MMF_DUMP_FILTER_MASK; | |
510 | return 1; | |
511 | } | |
512 | ||
513 | __setup("coredump_filter=", coredump_filter_setup); | |
514 | ||
1da177e4 LT |
515 | #include <linux/init_task.h> |
516 | ||
858f0993 AD |
517 | static void mm_init_aio(struct mm_struct *mm) |
518 | { | |
519 | #ifdef CONFIG_AIO | |
520 | spin_lock_init(&mm->ioctx_lock); | |
521 | INIT_HLIST_HEAD(&mm->ioctx_list); | |
522 | #endif | |
523 | } | |
524 | ||
fb0a685c | 525 | static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p) |
1da177e4 LT |
526 | { |
527 | atomic_set(&mm->mm_users, 1); | |
528 | atomic_set(&mm->mm_count, 1); | |
529 | init_rwsem(&mm->mmap_sem); | |
530 | INIT_LIST_HEAD(&mm->mmlist); | |
f8af4da3 HD |
531 | mm->flags = (current->mm) ? |
532 | (current->mm->flags & MMF_INIT_MASK) : default_dump_filter; | |
999d9fc1 | 533 | mm->core_state = NULL; |
1da177e4 | 534 | mm->nr_ptes = 0; |
d559db08 | 535 | memset(&mm->rss_stat, 0, sizeof(mm->rss_stat)); |
1da177e4 | 536 | spin_lock_init(&mm->page_table_lock); |
1da177e4 | 537 | mm->free_area_cache = TASK_UNMAPPED_BASE; |
1363c3cd | 538 | mm->cached_hole_size = ~0UL; |
858f0993 | 539 | mm_init_aio(mm); |
cf475ad2 | 540 | mm_init_owner(mm, p); |
1da177e4 LT |
541 | |
542 | if (likely(!mm_alloc_pgd(mm))) { | |
543 | mm->def_flags = 0; | |
cddb8a5c | 544 | mmu_notifier_mm_init(mm); |
1da177e4 LT |
545 | return mm; |
546 | } | |
78fb7466 | 547 | |
1da177e4 LT |
548 | free_mm(mm); |
549 | return NULL; | |
550 | } | |
551 | ||
c3f0327f KK |
552 | static void check_mm(struct mm_struct *mm) |
553 | { | |
554 | int i; | |
555 | ||
556 | for (i = 0; i < NR_MM_COUNTERS; i++) { | |
557 | long x = atomic_long_read(&mm->rss_stat.count[i]); | |
558 | ||
559 | if (unlikely(x)) | |
560 | printk(KERN_ALERT "BUG: Bad rss-counter state " | |
561 | "mm:%p idx:%d val:%ld\n", mm, i, x); | |
562 | } | |
563 | ||
564 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
565 | VM_BUG_ON(mm->pmd_huge_pte); | |
566 | #endif | |
567 | } | |
568 | ||
1da177e4 LT |
569 | /* |
570 | * Allocate and initialize an mm_struct. | |
571 | */ | |
fb0a685c | 572 | struct mm_struct *mm_alloc(void) |
1da177e4 | 573 | { |
fb0a685c | 574 | struct mm_struct *mm; |
1da177e4 LT |
575 | |
576 | mm = allocate_mm(); | |
de03c72c KM |
577 | if (!mm) |
578 | return NULL; | |
579 | ||
580 | memset(mm, 0, sizeof(*mm)); | |
6345d24d LT |
581 | mm_init_cpumask(mm); |
582 | return mm_init(mm, current); | |
1da177e4 LT |
583 | } |
584 | ||
585 | /* | |
586 | * Called when the last reference to the mm | |
587 | * is dropped: either by a lazy thread or by | |
588 | * mmput. Free the page directory and the mm. | |
589 | */ | |
7ad5b3a5 | 590 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
591 | { |
592 | BUG_ON(mm == &init_mm); | |
593 | mm_free_pgd(mm); | |
594 | destroy_context(mm); | |
cddb8a5c | 595 | mmu_notifier_mm_destroy(mm); |
c3f0327f | 596 | check_mm(mm); |
1da177e4 LT |
597 | free_mm(mm); |
598 | } | |
6d4e4c4f | 599 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 LT |
600 | |
601 | /* | |
602 | * Decrement the use count and release all resources for an mm. | |
603 | */ | |
604 | void mmput(struct mm_struct *mm) | |
605 | { | |
0ae26f1b AM |
606 | might_sleep(); |
607 | ||
1da177e4 | 608 | if (atomic_dec_and_test(&mm->mm_users)) { |
d4b3b638 | 609 | uprobe_clear_state(mm); |
1da177e4 | 610 | exit_aio(mm); |
1c2fb7a4 | 611 | ksm_exit(mm); |
ba76149f | 612 | khugepaged_exit(mm); /* must run before exit_mmap */ |
1da177e4 | 613 | exit_mmap(mm); |
925d1c40 | 614 | set_mm_exe_file(mm, NULL); |
1da177e4 LT |
615 | if (!list_empty(&mm->mmlist)) { |
616 | spin_lock(&mmlist_lock); | |
617 | list_del(&mm->mmlist); | |
618 | spin_unlock(&mmlist_lock); | |
619 | } | |
801460d0 HS |
620 | if (mm->binfmt) |
621 | module_put(mm->binfmt->module); | |
1da177e4 LT |
622 | mmdrop(mm); |
623 | } | |
624 | } | |
625 | EXPORT_SYMBOL_GPL(mmput); | |
626 | ||
38646013 JS |
627 | /* |
628 | * We added or removed a vma mapping the executable. The vmas are only mapped | |
629 | * during exec and are not mapped with the mmap system call. | |
630 | * Callers must hold down_write() on the mm's mmap_sem for these | |
631 | */ | |
632 | void added_exe_file_vma(struct mm_struct *mm) | |
633 | { | |
634 | mm->num_exe_file_vmas++; | |
635 | } | |
636 | ||
637 | void removed_exe_file_vma(struct mm_struct *mm) | |
638 | { | |
639 | mm->num_exe_file_vmas--; | |
fb0a685c | 640 | if ((mm->num_exe_file_vmas == 0) && mm->exe_file) { |
38646013 JS |
641 | fput(mm->exe_file); |
642 | mm->exe_file = NULL; | |
643 | } | |
644 | ||
645 | } | |
646 | ||
647 | void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file) | |
648 | { | |
649 | if (new_exe_file) | |
650 | get_file(new_exe_file); | |
651 | if (mm->exe_file) | |
652 | fput(mm->exe_file); | |
653 | mm->exe_file = new_exe_file; | |
654 | mm->num_exe_file_vmas = 0; | |
655 | } | |
656 | ||
657 | struct file *get_mm_exe_file(struct mm_struct *mm) | |
658 | { | |
659 | struct file *exe_file; | |
660 | ||
661 | /* We need mmap_sem to protect against races with removal of | |
662 | * VM_EXECUTABLE vmas */ | |
663 | down_read(&mm->mmap_sem); | |
664 | exe_file = mm->exe_file; | |
665 | if (exe_file) | |
666 | get_file(exe_file); | |
667 | up_read(&mm->mmap_sem); | |
668 | return exe_file; | |
669 | } | |
670 | ||
671 | static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm) | |
672 | { | |
673 | /* It's safe to write the exe_file pointer without exe_file_lock because | |
674 | * this is called during fork when the task is not yet in /proc */ | |
675 | newmm->exe_file = get_mm_exe_file(oldmm); | |
676 | } | |
677 | ||
1da177e4 LT |
678 | /** |
679 | * get_task_mm - acquire a reference to the task's mm | |
680 | * | |
246bb0b1 | 681 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
682 | * this kernel workthread has transiently adopted a user mm with use_mm, |
683 | * to do its AIO) is not set and if so returns a reference to it, after | |
684 | * bumping up the use count. User must release the mm via mmput() | |
685 | * after use. Typically used by /proc and ptrace. | |
686 | */ | |
687 | struct mm_struct *get_task_mm(struct task_struct *task) | |
688 | { | |
689 | struct mm_struct *mm; | |
690 | ||
691 | task_lock(task); | |
692 | mm = task->mm; | |
693 | if (mm) { | |
246bb0b1 | 694 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
695 | mm = NULL; |
696 | else | |
697 | atomic_inc(&mm->mm_users); | |
698 | } | |
699 | task_unlock(task); | |
700 | return mm; | |
701 | } | |
702 | EXPORT_SYMBOL_GPL(get_task_mm); | |
703 | ||
8cdb878d CY |
704 | struct mm_struct *mm_access(struct task_struct *task, unsigned int mode) |
705 | { | |
706 | struct mm_struct *mm; | |
707 | int err; | |
708 | ||
709 | err = mutex_lock_killable(&task->signal->cred_guard_mutex); | |
710 | if (err) | |
711 | return ERR_PTR(err); | |
712 | ||
713 | mm = get_task_mm(task); | |
714 | if (mm && mm != current->mm && | |
715 | !ptrace_may_access(task, mode)) { | |
716 | mmput(mm); | |
717 | mm = ERR_PTR(-EACCES); | |
718 | } | |
719 | mutex_unlock(&task->signal->cred_guard_mutex); | |
720 | ||
721 | return mm; | |
722 | } | |
723 | ||
57b59c4a | 724 | static void complete_vfork_done(struct task_struct *tsk) |
c415c3b4 | 725 | { |
d68b46fe | 726 | struct completion *vfork; |
c415c3b4 | 727 | |
d68b46fe ON |
728 | task_lock(tsk); |
729 | vfork = tsk->vfork_done; | |
730 | if (likely(vfork)) { | |
731 | tsk->vfork_done = NULL; | |
732 | complete(vfork); | |
733 | } | |
734 | task_unlock(tsk); | |
735 | } | |
736 | ||
737 | static int wait_for_vfork_done(struct task_struct *child, | |
738 | struct completion *vfork) | |
739 | { | |
740 | int killed; | |
741 | ||
742 | freezer_do_not_count(); | |
743 | killed = wait_for_completion_killable(vfork); | |
744 | freezer_count(); | |
745 | ||
746 | if (killed) { | |
747 | task_lock(child); | |
748 | child->vfork_done = NULL; | |
749 | task_unlock(child); | |
750 | } | |
751 | ||
752 | put_task_struct(child); | |
753 | return killed; | |
c415c3b4 ON |
754 | } |
755 | ||
1da177e4 LT |
756 | /* Please note the differences between mmput and mm_release. |
757 | * mmput is called whenever we stop holding onto a mm_struct, | |
758 | * error success whatever. | |
759 | * | |
760 | * mm_release is called after a mm_struct has been removed | |
761 | * from the current process. | |
762 | * | |
763 | * This difference is important for error handling, when we | |
764 | * only half set up a mm_struct for a new process and need to restore | |
765 | * the old one. Because we mmput the new mm_struct before | |
766 | * restoring the old one. . . | |
767 | * Eric Biederman 10 January 1998 | |
768 | */ | |
769 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
770 | { | |
8141c7f3 LT |
771 | /* Get rid of any futexes when releasing the mm */ |
772 | #ifdef CONFIG_FUTEX | |
fc6b177d | 773 | if (unlikely(tsk->robust_list)) { |
8141c7f3 | 774 | exit_robust_list(tsk); |
fc6b177d PZ |
775 | tsk->robust_list = NULL; |
776 | } | |
8141c7f3 | 777 | #ifdef CONFIG_COMPAT |
fc6b177d | 778 | if (unlikely(tsk->compat_robust_list)) { |
8141c7f3 | 779 | compat_exit_robust_list(tsk); |
fc6b177d PZ |
780 | tsk->compat_robust_list = NULL; |
781 | } | |
8141c7f3 | 782 | #endif |
322a2c10 TG |
783 | if (unlikely(!list_empty(&tsk->pi_state_list))) |
784 | exit_pi_state_list(tsk); | |
8141c7f3 LT |
785 | #endif |
786 | ||
0326f5a9 SD |
787 | uprobe_free_utask(tsk); |
788 | ||
1da177e4 LT |
789 | /* Get rid of any cached register state */ |
790 | deactivate_mm(tsk, mm); | |
791 | ||
fec1d011 RM |
792 | /* |
793 | * If we're exiting normally, clear a user-space tid field if | |
794 | * requested. We leave this alone when dying by signal, to leave | |
795 | * the value intact in a core dump, and to save the unnecessary | |
d68b46fe ON |
796 | * trouble, say, a killed vfork parent shouldn't touch this mm. |
797 | * Userland only wants this done for a sys_exit. | |
fec1d011 | 798 | */ |
9c8a8228 ED |
799 | if (tsk->clear_child_tid) { |
800 | if (!(tsk->flags & PF_SIGNALED) && | |
801 | atomic_read(&mm->mm_users) > 1) { | |
802 | /* | |
803 | * We don't check the error code - if userspace has | |
804 | * not set up a proper pointer then tough luck. | |
805 | */ | |
806 | put_user(0, tsk->clear_child_tid); | |
807 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
808 | 1, NULL, NULL, 0); | |
809 | } | |
1da177e4 | 810 | tsk->clear_child_tid = NULL; |
1da177e4 | 811 | } |
f7505d64 KK |
812 | |
813 | /* | |
814 | * All done, finally we can wake up parent and return this mm to him. | |
815 | * Also kthread_stop() uses this completion for synchronization. | |
816 | */ | |
817 | if (tsk->vfork_done) | |
818 | complete_vfork_done(tsk); | |
1da177e4 LT |
819 | } |
820 | ||
a0a7ec30 JD |
821 | /* |
822 | * Allocate a new mm structure and copy contents from the | |
823 | * mm structure of the passed in task structure. | |
824 | */ | |
402b0862 | 825 | struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
826 | { |
827 | struct mm_struct *mm, *oldmm = current->mm; | |
828 | int err; | |
829 | ||
830 | if (!oldmm) | |
831 | return NULL; | |
832 | ||
833 | mm = allocate_mm(); | |
834 | if (!mm) | |
835 | goto fail_nomem; | |
836 | ||
837 | memcpy(mm, oldmm, sizeof(*mm)); | |
6345d24d | 838 | mm_init_cpumask(mm); |
a0a7ec30 | 839 | |
e7a00c45 AA |
840 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
841 | mm->pmd_huge_pte = NULL; | |
842 | #endif | |
d4b3b638 | 843 | uprobe_reset_state(mm); |
e7a00c45 | 844 | |
78fb7466 | 845 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
846 | goto fail_nomem; |
847 | ||
848 | if (init_new_context(tsk, mm)) | |
849 | goto fail_nocontext; | |
850 | ||
925d1c40 MH |
851 | dup_mm_exe_file(oldmm, mm); |
852 | ||
a0a7ec30 JD |
853 | err = dup_mmap(mm, oldmm); |
854 | if (err) | |
855 | goto free_pt; | |
856 | ||
857 | mm->hiwater_rss = get_mm_rss(mm); | |
858 | mm->hiwater_vm = mm->total_vm; | |
859 | ||
801460d0 HS |
860 | if (mm->binfmt && !try_module_get(mm->binfmt->module)) |
861 | goto free_pt; | |
862 | ||
a0a7ec30 JD |
863 | return mm; |
864 | ||
865 | free_pt: | |
801460d0 HS |
866 | /* don't put binfmt in mmput, we haven't got module yet */ |
867 | mm->binfmt = NULL; | |
a0a7ec30 JD |
868 | mmput(mm); |
869 | ||
870 | fail_nomem: | |
871 | return NULL; | |
872 | ||
873 | fail_nocontext: | |
874 | /* | |
875 | * If init_new_context() failed, we cannot use mmput() to free the mm | |
876 | * because it calls destroy_context() | |
877 | */ | |
878 | mm_free_pgd(mm); | |
879 | free_mm(mm); | |
880 | return NULL; | |
881 | } | |
882 | ||
fb0a685c | 883 | static int copy_mm(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 884 | { |
fb0a685c | 885 | struct mm_struct *mm, *oldmm; |
1da177e4 LT |
886 | int retval; |
887 | ||
888 | tsk->min_flt = tsk->maj_flt = 0; | |
889 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
890 | #ifdef CONFIG_DETECT_HUNG_TASK |
891 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
892 | #endif | |
1da177e4 LT |
893 | |
894 | tsk->mm = NULL; | |
895 | tsk->active_mm = NULL; | |
896 | ||
897 | /* | |
898 | * Are we cloning a kernel thread? | |
899 | * | |
900 | * We need to steal a active VM for that.. | |
901 | */ | |
902 | oldmm = current->mm; | |
903 | if (!oldmm) | |
904 | return 0; | |
905 | ||
906 | if (clone_flags & CLONE_VM) { | |
907 | atomic_inc(&oldmm->mm_users); | |
908 | mm = oldmm; | |
1da177e4 LT |
909 | goto good_mm; |
910 | } | |
911 | ||
912 | retval = -ENOMEM; | |
a0a7ec30 | 913 | mm = dup_mm(tsk); |
1da177e4 LT |
914 | if (!mm) |
915 | goto fail_nomem; | |
916 | ||
1da177e4 LT |
917 | good_mm: |
918 | tsk->mm = mm; | |
919 | tsk->active_mm = mm; | |
920 | return 0; | |
921 | ||
1da177e4 LT |
922 | fail_nomem: |
923 | return retval; | |
1da177e4 LT |
924 | } |
925 | ||
a39bc516 | 926 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 927 | { |
498052bb | 928 | struct fs_struct *fs = current->fs; |
1da177e4 | 929 | if (clone_flags & CLONE_FS) { |
498052bb | 930 | /* tsk->fs is already what we want */ |
2a4419b5 | 931 | spin_lock(&fs->lock); |
498052bb | 932 | if (fs->in_exec) { |
2a4419b5 | 933 | spin_unlock(&fs->lock); |
498052bb AV |
934 | return -EAGAIN; |
935 | } | |
936 | fs->users++; | |
2a4419b5 | 937 | spin_unlock(&fs->lock); |
1da177e4 LT |
938 | return 0; |
939 | } | |
498052bb | 940 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
941 | if (!tsk->fs) |
942 | return -ENOMEM; | |
943 | return 0; | |
944 | } | |
945 | ||
fb0a685c | 946 | static int copy_files(unsigned long clone_flags, struct task_struct *tsk) |
a016f338 JD |
947 | { |
948 | struct files_struct *oldf, *newf; | |
949 | int error = 0; | |
950 | ||
951 | /* | |
952 | * A background process may not have any files ... | |
953 | */ | |
954 | oldf = current->files; | |
955 | if (!oldf) | |
956 | goto out; | |
957 | ||
958 | if (clone_flags & CLONE_FILES) { | |
959 | atomic_inc(&oldf->count); | |
960 | goto out; | |
961 | } | |
962 | ||
a016f338 JD |
963 | newf = dup_fd(oldf, &error); |
964 | if (!newf) | |
965 | goto out; | |
966 | ||
967 | tsk->files = newf; | |
968 | error = 0; | |
969 | out: | |
970 | return error; | |
971 | } | |
972 | ||
fadad878 | 973 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
974 | { |
975 | #ifdef CONFIG_BLOCK | |
976 | struct io_context *ioc = current->io_context; | |
6e736be7 | 977 | struct io_context *new_ioc; |
fd0928df JA |
978 | |
979 | if (!ioc) | |
980 | return 0; | |
fadad878 JA |
981 | /* |
982 | * Share io context with parent, if CLONE_IO is set | |
983 | */ | |
984 | if (clone_flags & CLONE_IO) { | |
3d48749d TH |
985 | ioc_task_link(ioc); |
986 | tsk->io_context = ioc; | |
fadad878 | 987 | } else if (ioprio_valid(ioc->ioprio)) { |
6e736be7 TH |
988 | new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE); |
989 | if (unlikely(!new_ioc)) | |
fd0928df JA |
990 | return -ENOMEM; |
991 | ||
6e736be7 | 992 | new_ioc->ioprio = ioc->ioprio; |
11a3122f | 993 | put_io_context(new_ioc); |
fd0928df JA |
994 | } |
995 | #endif | |
996 | return 0; | |
997 | } | |
998 | ||
a39bc516 | 999 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1000 | { |
1001 | struct sighand_struct *sig; | |
1002 | ||
60348802 | 1003 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
1004 | atomic_inc(¤t->sighand->count); |
1005 | return 0; | |
1006 | } | |
1007 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 1008 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
1009 | if (!sig) |
1010 | return -ENOMEM; | |
1da177e4 LT |
1011 | atomic_set(&sig->count, 1); |
1012 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
1013 | return 0; | |
1014 | } | |
1015 | ||
a7e5328a | 1016 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 1017 | { |
d80e731e ON |
1018 | if (atomic_dec_and_test(&sighand->count)) { |
1019 | signalfd_cleanup(sighand); | |
c81addc9 | 1020 | kmem_cache_free(sighand_cachep, sighand); |
d80e731e | 1021 | } |
c81addc9 ON |
1022 | } |
1023 | ||
f06febc9 FM |
1024 | |
1025 | /* | |
1026 | * Initialize POSIX timer handling for a thread group. | |
1027 | */ | |
1028 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
1029 | { | |
78d7d407 JS |
1030 | unsigned long cpu_limit; |
1031 | ||
f06febc9 FM |
1032 | /* Thread group counters. */ |
1033 | thread_group_cputime_init(sig); | |
1034 | ||
78d7d407 JS |
1035 | cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur); |
1036 | if (cpu_limit != RLIM_INFINITY) { | |
1037 | sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit); | |
6279a751 ON |
1038 | sig->cputimer.running = 1; |
1039 | } | |
1040 | ||
f06febc9 FM |
1041 | /* The timer lists. */ |
1042 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
1043 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
1044 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
1045 | } | |
1046 | ||
a39bc516 | 1047 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
1048 | { |
1049 | struct signal_struct *sig; | |
1da177e4 | 1050 | |
4ab6c083 | 1051 | if (clone_flags & CLONE_THREAD) |
490dea45 | 1052 | return 0; |
490dea45 | 1053 | |
a56704ef | 1054 | sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
1055 | tsk->signal = sig; |
1056 | if (!sig) | |
1057 | return -ENOMEM; | |
1058 | ||
b3ac022c | 1059 | sig->nr_threads = 1; |
1da177e4 | 1060 | atomic_set(&sig->live, 1); |
b3ac022c | 1061 | atomic_set(&sig->sigcnt, 1); |
1da177e4 | 1062 | init_waitqueue_head(&sig->wait_chldexit); |
b3bfa0cb SB |
1063 | if (clone_flags & CLONE_NEWPID) |
1064 | sig->flags |= SIGNAL_UNKILLABLE; | |
db51aecc | 1065 | sig->curr_target = tsk; |
1da177e4 LT |
1066 | init_sigpending(&sig->shared_pending); |
1067 | INIT_LIST_HEAD(&sig->posix_timers); | |
1068 | ||
c9cb2e3d | 1069 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
1da177e4 | 1070 | sig->real_timer.function = it_real_fn; |
1da177e4 | 1071 | |
1da177e4 LT |
1072 | task_lock(current->group_leader); |
1073 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
1074 | task_unlock(current->group_leader); | |
1075 | ||
6279a751 ON |
1076 | posix_cpu_timers_init_group(sig); |
1077 | ||
522ed776 | 1078 | tty_audit_fork(sig); |
5091faa4 | 1079 | sched_autogroup_fork(sig); |
522ed776 | 1080 | |
4714d1d3 | 1081 | #ifdef CONFIG_CGROUPS |
257058ae | 1082 | init_rwsem(&sig->group_rwsem); |
4714d1d3 BB |
1083 | #endif |
1084 | ||
28b83c51 | 1085 | sig->oom_adj = current->signal->oom_adj; |
a63d83f4 | 1086 | sig->oom_score_adj = current->signal->oom_score_adj; |
dabb16f6 | 1087 | sig->oom_score_adj_min = current->signal->oom_score_adj_min; |
28b83c51 | 1088 | |
ebec18a6 LP |
1089 | sig->has_child_subreaper = current->signal->has_child_subreaper || |
1090 | current->signal->is_child_subreaper; | |
1091 | ||
9b1bf12d KM |
1092 | mutex_init(&sig->cred_guard_mutex); |
1093 | ||
1da177e4 LT |
1094 | return 0; |
1095 | } | |
1096 | ||
a39bc516 | 1097 | static void copy_flags(unsigned long clone_flags, struct task_struct *p) |
1da177e4 LT |
1098 | { |
1099 | unsigned long new_flags = p->flags; | |
1100 | ||
21aa9af0 | 1101 | new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER); |
1da177e4 | 1102 | new_flags |= PF_FORKNOEXEC; |
1da177e4 LT |
1103 | p->flags = new_flags; |
1104 | } | |
1105 | ||
17da2bd9 | 1106 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
1107 | { |
1108 | current->clear_child_tid = tidptr; | |
1109 | ||
b488893a | 1110 | return task_pid_vnr(current); |
1da177e4 LT |
1111 | } |
1112 | ||
a39bc516 | 1113 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 1114 | { |
1d615482 | 1115 | raw_spin_lock_init(&p->pi_lock); |
e29e175b | 1116 | #ifdef CONFIG_RT_MUTEXES |
732375c6 | 1117 | plist_head_init(&p->pi_waiters); |
23f78d4a | 1118 | p->pi_blocked_on = NULL; |
23f78d4a IM |
1119 | #endif |
1120 | } | |
1121 | ||
cf475ad2 BS |
1122 | #ifdef CONFIG_MM_OWNER |
1123 | void mm_init_owner(struct mm_struct *mm, struct task_struct *p) | |
1124 | { | |
1125 | mm->owner = p; | |
1126 | } | |
1127 | #endif /* CONFIG_MM_OWNER */ | |
1128 | ||
f06febc9 FM |
1129 | /* |
1130 | * Initialize POSIX timer handling for a single task. | |
1131 | */ | |
1132 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
1133 | { | |
64861634 MS |
1134 | tsk->cputime_expires.prof_exp = 0; |
1135 | tsk->cputime_expires.virt_exp = 0; | |
f06febc9 FM |
1136 | tsk->cputime_expires.sched_exp = 0; |
1137 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
1138 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
1139 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
1140 | } | |
1141 | ||
1da177e4 LT |
1142 | /* |
1143 | * This creates a new process as a copy of the old one, | |
1144 | * but does not actually start it yet. | |
1145 | * | |
1146 | * It copies the registers, and all the appropriate | |
1147 | * parts of the process environment (as per the clone | |
1148 | * flags). The actual kick-off is left to the caller. | |
1149 | */ | |
36c8b586 IM |
1150 | static struct task_struct *copy_process(unsigned long clone_flags, |
1151 | unsigned long stack_start, | |
1152 | struct pt_regs *regs, | |
1153 | unsigned long stack_size, | |
36c8b586 | 1154 | int __user *child_tidptr, |
09a05394 RM |
1155 | struct pid *pid, |
1156 | int trace) | |
1da177e4 LT |
1157 | { |
1158 | int retval; | |
a24efe62 | 1159 | struct task_struct *p; |
b4f48b63 | 1160 | int cgroup_callbacks_done = 0; |
1da177e4 LT |
1161 | |
1162 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
1163 | return ERR_PTR(-EINVAL); | |
1164 | ||
1165 | /* | |
1166 | * Thread groups must share signals as well, and detached threads | |
1167 | * can only be started up within the thread group. | |
1168 | */ | |
1169 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
1170 | return ERR_PTR(-EINVAL); | |
1171 | ||
1172 | /* | |
1173 | * Shared signal handlers imply shared VM. By way of the above, | |
1174 | * thread groups also imply shared VM. Blocking this case allows | |
1175 | * for various simplifications in other code. | |
1176 | */ | |
1177 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
1178 | return ERR_PTR(-EINVAL); | |
1179 | ||
123be07b SB |
1180 | /* |
1181 | * Siblings of global init remain as zombies on exit since they are | |
1182 | * not reaped by their parent (swapper). To solve this and to avoid | |
1183 | * multi-rooted process trees, prevent global and container-inits | |
1184 | * from creating siblings. | |
1185 | */ | |
1186 | if ((clone_flags & CLONE_PARENT) && | |
1187 | current->signal->flags & SIGNAL_UNKILLABLE) | |
1188 | return ERR_PTR(-EINVAL); | |
1189 | ||
1da177e4 LT |
1190 | retval = security_task_create(clone_flags); |
1191 | if (retval) | |
1192 | goto fork_out; | |
1193 | ||
1194 | retval = -ENOMEM; | |
1195 | p = dup_task_struct(current); | |
1196 | if (!p) | |
1197 | goto fork_out; | |
1198 | ||
f7e8b616 | 1199 | ftrace_graph_init_task(p); |
e2cfabdf | 1200 | get_seccomp_filter(p); |
f7e8b616 | 1201 | |
bea493a0 PZ |
1202 | rt_mutex_init_task(p); |
1203 | ||
d12c1a37 | 1204 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
1205 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
1206 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
1207 | #endif | |
1da177e4 | 1208 | retval = -EAGAIN; |
3b11a1de | 1209 | if (atomic_read(&p->real_cred->user->processes) >= |
78d7d407 | 1210 | task_rlimit(p, RLIMIT_NPROC)) { |
1da177e4 | 1211 | if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) && |
18b6e041 | 1212 | p->real_cred->user != INIT_USER) |
1da177e4 LT |
1213 | goto bad_fork_free; |
1214 | } | |
72fa5997 | 1215 | current->flags &= ~PF_NPROC_EXCEEDED; |
1da177e4 | 1216 | |
f1752eec DH |
1217 | retval = copy_creds(p, clone_flags); |
1218 | if (retval < 0) | |
1219 | goto bad_fork_free; | |
1da177e4 LT |
1220 | |
1221 | /* | |
1222 | * If multiple threads are within copy_process(), then this check | |
1223 | * triggers too late. This doesn't hurt, the check is only there | |
1224 | * to stop root fork bombs. | |
1225 | */ | |
04ec93fe | 1226 | retval = -EAGAIN; |
1da177e4 LT |
1227 | if (nr_threads >= max_threads) |
1228 | goto bad_fork_cleanup_count; | |
1229 | ||
a1261f54 | 1230 | if (!try_module_get(task_thread_info(p)->exec_domain->module)) |
1da177e4 LT |
1231 | goto bad_fork_cleanup_count; |
1232 | ||
1da177e4 | 1233 | p->did_exec = 0; |
ca74e92b | 1234 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
1da177e4 | 1235 | copy_flags(clone_flags, p); |
1da177e4 LT |
1236 | INIT_LIST_HEAD(&p->children); |
1237 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1238 | rcu_copy_process(p); |
1da177e4 LT |
1239 | p->vfork_done = NULL; |
1240 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1241 | |
1da177e4 LT |
1242 | init_sigpending(&p->pending); |
1243 | ||
64861634 MS |
1244 | p->utime = p->stime = p->gtime = 0; |
1245 | p->utimescaled = p->stimescaled = 0; | |
d99ca3b9 | 1246 | #ifndef CONFIG_VIRT_CPU_ACCOUNTING |
64861634 | 1247 | p->prev_utime = p->prev_stime = 0; |
d99ca3b9 | 1248 | #endif |
a3a2e76c KH |
1249 | #if defined(SPLIT_RSS_COUNTING) |
1250 | memset(&p->rss_stat, 0, sizeof(p->rss_stat)); | |
1251 | #endif | |
172ba844 | 1252 | |
6976675d AV |
1253 | p->default_timer_slack_ns = current->timer_slack_ns; |
1254 | ||
5995477a | 1255 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1256 | acct_clear_integrals(p); |
1257 | ||
f06febc9 | 1258 | posix_cpu_timers_init(p); |
1da177e4 | 1259 | |
1da177e4 | 1260 | do_posix_clock_monotonic_gettime(&p->start_time); |
924b42d5 TJ |
1261 | p->real_start_time = p->start_time; |
1262 | monotonic_to_bootbased(&p->real_start_time); | |
1da177e4 | 1263 | p->io_context = NULL; |
1da177e4 | 1264 | p->audit_context = NULL; |
4714d1d3 | 1265 | if (clone_flags & CLONE_THREAD) |
257058ae | 1266 | threadgroup_change_begin(current); |
b4f48b63 | 1267 | cgroup_fork(p); |
1da177e4 | 1268 | #ifdef CONFIG_NUMA |
846a16bf | 1269 | p->mempolicy = mpol_dup(p->mempolicy); |
fb0a685c DRO |
1270 | if (IS_ERR(p->mempolicy)) { |
1271 | retval = PTR_ERR(p->mempolicy); | |
1272 | p->mempolicy = NULL; | |
1273 | goto bad_fork_cleanup_cgroup; | |
1274 | } | |
c61afb18 | 1275 | mpol_fix_fork_child_flag(p); |
1da177e4 | 1276 | #endif |
778d3b0f MH |
1277 | #ifdef CONFIG_CPUSETS |
1278 | p->cpuset_mem_spread_rotor = NUMA_NO_NODE; | |
1279 | p->cpuset_slab_spread_rotor = NUMA_NO_NODE; | |
cc9a6c87 | 1280 | seqcount_init(&p->mems_allowed_seq); |
778d3b0f | 1281 | #endif |
de30a2b3 IM |
1282 | #ifdef CONFIG_TRACE_IRQFLAGS |
1283 | p->irq_events = 0; | |
b36e4758 RK |
1284 | #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW |
1285 | p->hardirqs_enabled = 1; | |
1286 | #else | |
de30a2b3 | 1287 | p->hardirqs_enabled = 0; |
b36e4758 | 1288 | #endif |
de30a2b3 IM |
1289 | p->hardirq_enable_ip = 0; |
1290 | p->hardirq_enable_event = 0; | |
1291 | p->hardirq_disable_ip = _THIS_IP_; | |
1292 | p->hardirq_disable_event = 0; | |
1293 | p->softirqs_enabled = 1; | |
1294 | p->softirq_enable_ip = _THIS_IP_; | |
1295 | p->softirq_enable_event = 0; | |
1296 | p->softirq_disable_ip = 0; | |
1297 | p->softirq_disable_event = 0; | |
1298 | p->hardirq_context = 0; | |
1299 | p->softirq_context = 0; | |
1300 | #endif | |
fbb9ce95 IM |
1301 | #ifdef CONFIG_LOCKDEP |
1302 | p->lockdep_depth = 0; /* no locks held yet */ | |
1303 | p->curr_chain_key = 0; | |
1304 | p->lockdep_recursion = 0; | |
1305 | #endif | |
1da177e4 | 1306 | |
408894ee IM |
1307 | #ifdef CONFIG_DEBUG_MUTEXES |
1308 | p->blocked_on = NULL; /* not blocked yet */ | |
1309 | #endif | |
c255a458 | 1310 | #ifdef CONFIG_MEMCG |
569b846d KH |
1311 | p->memcg_batch.do_batch = 0; |
1312 | p->memcg_batch.memcg = NULL; | |
1313 | #endif | |
0f481406 | 1314 | |
3c90e6e9 | 1315 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
3e51e3ed | 1316 | sched_fork(p); |
6ab423e0 | 1317 | |
cdd6c482 | 1318 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1319 | if (retval) |
1320 | goto bad_fork_cleanup_policy; | |
fb0a685c DRO |
1321 | retval = audit_alloc(p); |
1322 | if (retval) | |
f1752eec | 1323 | goto bad_fork_cleanup_policy; |
1da177e4 | 1324 | /* copy all the process information */ |
fb0a685c DRO |
1325 | retval = copy_semundo(clone_flags, p); |
1326 | if (retval) | |
1da177e4 | 1327 | goto bad_fork_cleanup_audit; |
fb0a685c DRO |
1328 | retval = copy_files(clone_flags, p); |
1329 | if (retval) | |
1da177e4 | 1330 | goto bad_fork_cleanup_semundo; |
fb0a685c DRO |
1331 | retval = copy_fs(clone_flags, p); |
1332 | if (retval) | |
1da177e4 | 1333 | goto bad_fork_cleanup_files; |
fb0a685c DRO |
1334 | retval = copy_sighand(clone_flags, p); |
1335 | if (retval) | |
1da177e4 | 1336 | goto bad_fork_cleanup_fs; |
fb0a685c DRO |
1337 | retval = copy_signal(clone_flags, p); |
1338 | if (retval) | |
1da177e4 | 1339 | goto bad_fork_cleanup_sighand; |
fb0a685c DRO |
1340 | retval = copy_mm(clone_flags, p); |
1341 | if (retval) | |
1da177e4 | 1342 | goto bad_fork_cleanup_signal; |
fb0a685c DRO |
1343 | retval = copy_namespaces(clone_flags, p); |
1344 | if (retval) | |
d84f4f99 | 1345 | goto bad_fork_cleanup_mm; |
fb0a685c DRO |
1346 | retval = copy_io(clone_flags, p); |
1347 | if (retval) | |
fd0928df | 1348 | goto bad_fork_cleanup_namespaces; |
6f2c55b8 | 1349 | retval = copy_thread(clone_flags, stack_start, stack_size, p, regs); |
1da177e4 | 1350 | if (retval) |
fd0928df | 1351 | goto bad_fork_cleanup_io; |
1da177e4 | 1352 | |
425fb2b4 PE |
1353 | if (pid != &init_struct_pid) { |
1354 | retval = -ENOMEM; | |
61bce0f1 | 1355 | pid = alloc_pid(p->nsproxy->pid_ns); |
425fb2b4 | 1356 | if (!pid) |
fd0928df | 1357 | goto bad_fork_cleanup_io; |
425fb2b4 PE |
1358 | } |
1359 | ||
1360 | p->pid = pid_nr(pid); | |
1361 | p->tgid = p->pid; | |
1362 | if (clone_flags & CLONE_THREAD) | |
1363 | p->tgid = current->tgid; | |
1364 | ||
1da177e4 LT |
1365 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1366 | /* | |
1367 | * Clear TID on mm_release()? | |
1368 | */ | |
fb0a685c | 1369 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL; |
73c10101 JA |
1370 | #ifdef CONFIG_BLOCK |
1371 | p->plug = NULL; | |
1372 | #endif | |
42b2dd0a | 1373 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1374 | p->robust_list = NULL; |
1375 | #ifdef CONFIG_COMPAT | |
1376 | p->compat_robust_list = NULL; | |
1377 | #endif | |
c87e2837 IM |
1378 | INIT_LIST_HEAD(&p->pi_state_list); |
1379 | p->pi_state_cache = NULL; | |
42b2dd0a | 1380 | #endif |
0326f5a9 | 1381 | uprobe_copy_process(p); |
f9a3879a GM |
1382 | /* |
1383 | * sigaltstack should be cleared when sharing the same VM | |
1384 | */ | |
1385 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1386 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1387 | ||
1da177e4 | 1388 | /* |
6580807d ON |
1389 | * Syscall tracing and stepping should be turned off in the |
1390 | * child regardless of CLONE_PTRACE. | |
1da177e4 | 1391 | */ |
6580807d | 1392 | user_disable_single_step(p); |
1da177e4 | 1393 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); |
ed75e8d5 LV |
1394 | #ifdef TIF_SYSCALL_EMU |
1395 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1396 | #endif | |
9745512c | 1397 | clear_all_latency_tracing(p); |
1da177e4 | 1398 | |
1da177e4 | 1399 | /* ok, now we should be set up.. */ |
5f8aadd8 ON |
1400 | if (clone_flags & CLONE_THREAD) |
1401 | p->exit_signal = -1; | |
1402 | else if (clone_flags & CLONE_PARENT) | |
1403 | p->exit_signal = current->group_leader->exit_signal; | |
1404 | else | |
1405 | p->exit_signal = (clone_flags & CSIGNAL); | |
1406 | ||
1da177e4 LT |
1407 | p->pdeath_signal = 0; |
1408 | p->exit_state = 0; | |
1409 | ||
9d823e8f WF |
1410 | p->nr_dirtied = 0; |
1411 | p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10); | |
83712358 | 1412 | p->dirty_paused_when = 0; |
9d823e8f | 1413 | |
1da177e4 LT |
1414 | /* |
1415 | * Ok, make it visible to the rest of the system. | |
1416 | * We dont wake it up yet. | |
1417 | */ | |
1418 | p->group_leader = p; | |
47e65328 | 1419 | INIT_LIST_HEAD(&p->thread_group); |
158e1645 | 1420 | p->task_works = NULL; |
1da177e4 | 1421 | |
b4f48b63 PM |
1422 | /* Now that the task is set up, run cgroup callbacks if |
1423 | * necessary. We need to run them before the task is visible | |
1424 | * on the tasklist. */ | |
1425 | cgroup_fork_callbacks(p); | |
1426 | cgroup_callbacks_done = 1; | |
1427 | ||
1da177e4 LT |
1428 | /* Need tasklist lock for parent etc handling! */ |
1429 | write_lock_irq(&tasklist_lock); | |
1430 | ||
1da177e4 | 1431 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1432 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1433 | p->real_parent = current->real_parent; |
2d5516cb ON |
1434 | p->parent_exec_id = current->parent_exec_id; |
1435 | } else { | |
1da177e4 | 1436 | p->real_parent = current; |
2d5516cb ON |
1437 | p->parent_exec_id = current->self_exec_id; |
1438 | } | |
1da177e4 | 1439 | |
3f17da69 | 1440 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 ON |
1441 | |
1442 | /* | |
1443 | * Process group and session signals need to be delivered to just the | |
1444 | * parent before the fork or both the parent and the child after the | |
1445 | * fork. Restart if a signal comes in before we add the new process to | |
1446 | * it's process group. | |
1447 | * A fatal signal pending means that current will exit, so the new | |
1448 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
fb0a685c | 1449 | */ |
23ff4440 | 1450 | recalc_sigpending(); |
4a2c7a78 ON |
1451 | if (signal_pending(current)) { |
1452 | spin_unlock(¤t->sighand->siglock); | |
1453 | write_unlock_irq(&tasklist_lock); | |
1454 | retval = -ERESTARTNOINTR; | |
f7e8b616 | 1455 | goto bad_fork_free_pid; |
4a2c7a78 ON |
1456 | } |
1457 | ||
1da177e4 | 1458 | if (clone_flags & CLONE_THREAD) { |
b3ac022c | 1459 | current->signal->nr_threads++; |
4ab6c083 | 1460 | atomic_inc(¤t->signal->live); |
b3ac022c | 1461 | atomic_inc(¤t->signal->sigcnt); |
1da177e4 | 1462 | p->group_leader = current->group_leader; |
47e65328 | 1463 | list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group); |
1da177e4 LT |
1464 | } |
1465 | ||
73b9ebfe | 1466 | if (likely(p->pid)) { |
4b9d33e6 | 1467 | ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace); |
73b9ebfe ON |
1468 | |
1469 | if (thread_group_leader(p)) { | |
45a68628 | 1470 | if (is_child_reaper(pid)) |
30e49c26 | 1471 | p->nsproxy->pid_ns->child_reaper = p; |
73b9ebfe | 1472 | |
fea9d175 | 1473 | p->signal->leader_pid = pid; |
9c9f4ded | 1474 | p->signal->tty = tty_kref_get(current->signal->tty); |
5cd17569 EB |
1475 | attach_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1476 | attach_pid(p, PIDTYPE_SID, task_session(current)); | |
9cd80bbb | 1477 | list_add_tail(&p->sibling, &p->real_parent->children); |
5e85d4ab | 1478 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
909ea964 | 1479 | __this_cpu_inc(process_counts); |
73b9ebfe | 1480 | } |
85868995 | 1481 | attach_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1482 | nr_threads++; |
1da177e4 LT |
1483 | } |
1484 | ||
1da177e4 | 1485 | total_forks++; |
3f17da69 | 1486 | spin_unlock(¤t->sighand->siglock); |
1da177e4 | 1487 | write_unlock_irq(&tasklist_lock); |
c13cf856 | 1488 | proc_fork_connector(p); |
817929ec | 1489 | cgroup_post_fork(p); |
4714d1d3 | 1490 | if (clone_flags & CLONE_THREAD) |
257058ae | 1491 | threadgroup_change_end(current); |
cdd6c482 | 1492 | perf_event_fork(p); |
43d2b113 KH |
1493 | |
1494 | trace_task_newtask(p, clone_flags); | |
1495 | ||
1da177e4 LT |
1496 | return p; |
1497 | ||
425fb2b4 PE |
1498 | bad_fork_free_pid: |
1499 | if (pid != &init_struct_pid) | |
1500 | free_pid(pid); | |
fd0928df | 1501 | bad_fork_cleanup_io: |
b69f2292 LR |
1502 | if (p->io_context) |
1503 | exit_io_context(p); | |
ab516013 | 1504 | bad_fork_cleanup_namespaces: |
5e2bf014 MG |
1505 | if (unlikely(clone_flags & CLONE_NEWPID)) |
1506 | pid_ns_release_proc(p->nsproxy->pid_ns); | |
444f378b | 1507 | exit_task_namespaces(p); |
1da177e4 | 1508 | bad_fork_cleanup_mm: |
c9f01245 | 1509 | if (p->mm) |
1da177e4 LT |
1510 | mmput(p->mm); |
1511 | bad_fork_cleanup_signal: | |
4ab6c083 | 1512 | if (!(clone_flags & CLONE_THREAD)) |
1c5354de | 1513 | free_signal_struct(p->signal); |
1da177e4 | 1514 | bad_fork_cleanup_sighand: |
a7e5328a | 1515 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1516 | bad_fork_cleanup_fs: |
1517 | exit_fs(p); /* blocking */ | |
1518 | bad_fork_cleanup_files: | |
1519 | exit_files(p); /* blocking */ | |
1520 | bad_fork_cleanup_semundo: | |
1521 | exit_sem(p); | |
1522 | bad_fork_cleanup_audit: | |
1523 | audit_free(p); | |
1da177e4 | 1524 | bad_fork_cleanup_policy: |
cdd6c482 | 1525 | perf_event_free_task(p); |
1da177e4 | 1526 | #ifdef CONFIG_NUMA |
f0be3d32 | 1527 | mpol_put(p->mempolicy); |
b4f48b63 | 1528 | bad_fork_cleanup_cgroup: |
1da177e4 | 1529 | #endif |
4714d1d3 | 1530 | if (clone_flags & CLONE_THREAD) |
257058ae | 1531 | threadgroup_change_end(current); |
b4f48b63 | 1532 | cgroup_exit(p, cgroup_callbacks_done); |
35df17c5 | 1533 | delayacct_tsk_free(p); |
a1261f54 | 1534 | module_put(task_thread_info(p)->exec_domain->module); |
1da177e4 | 1535 | bad_fork_cleanup_count: |
d84f4f99 | 1536 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1537 | exit_creds(p); |
1da177e4 LT |
1538 | bad_fork_free: |
1539 | free_task(p); | |
fe7d37d1 ON |
1540 | fork_out: |
1541 | return ERR_PTR(retval); | |
1da177e4 LT |
1542 | } |
1543 | ||
6b2fb3c6 | 1544 | noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs) |
1da177e4 LT |
1545 | { |
1546 | memset(regs, 0, sizeof(struct pt_regs)); | |
1547 | return regs; | |
1548 | } | |
1549 | ||
f106eee1 ON |
1550 | static inline void init_idle_pids(struct pid_link *links) |
1551 | { | |
1552 | enum pid_type type; | |
1553 | ||
1554 | for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) { | |
1555 | INIT_HLIST_NODE(&links[type].node); /* not really needed */ | |
1556 | links[type].pid = &init_struct_pid; | |
1557 | } | |
1558 | } | |
1559 | ||
9abcf40b | 1560 | struct task_struct * __cpuinit fork_idle(int cpu) |
1da177e4 | 1561 | { |
36c8b586 | 1562 | struct task_struct *task; |
1da177e4 LT |
1563 | struct pt_regs regs; |
1564 | ||
30e49c26 | 1565 | task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL, |
09a05394 | 1566 | &init_struct_pid, 0); |
f106eee1 ON |
1567 | if (!IS_ERR(task)) { |
1568 | init_idle_pids(task->pids); | |
753ca4f3 | 1569 | init_idle(task, cpu); |
f106eee1 | 1570 | } |
73b9ebfe | 1571 | |
1da177e4 LT |
1572 | return task; |
1573 | } | |
1574 | ||
1da177e4 LT |
1575 | /* |
1576 | * Ok, this is the main fork-routine. | |
1577 | * | |
1578 | * It copies the process, and if successful kick-starts | |
1579 | * it and waits for it to finish using the VM if required. | |
1580 | */ | |
1581 | long do_fork(unsigned long clone_flags, | |
1582 | unsigned long stack_start, | |
1583 | struct pt_regs *regs, | |
1584 | unsigned long stack_size, | |
1585 | int __user *parent_tidptr, | |
1586 | int __user *child_tidptr) | |
1587 | { | |
1588 | struct task_struct *p; | |
1589 | int trace = 0; | |
92476d7f | 1590 | long nr; |
1da177e4 | 1591 | |
18b6e041 SH |
1592 | /* |
1593 | * Do some preliminary argument and permissions checking before we | |
1594 | * actually start allocating stuff | |
1595 | */ | |
1596 | if (clone_flags & CLONE_NEWUSER) { | |
1597 | if (clone_flags & CLONE_THREAD) | |
1598 | return -EINVAL; | |
1599 | /* hopefully this check will go away when userns support is | |
1600 | * complete | |
1601 | */ | |
7657d904 SH |
1602 | if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) || |
1603 | !capable(CAP_SETGID)) | |
18b6e041 SH |
1604 | return -EPERM; |
1605 | } | |
1606 | ||
09a05394 | 1607 | /* |
4b9d33e6 TH |
1608 | * Determine whether and which event to report to ptracer. When |
1609 | * called from kernel_thread or CLONE_UNTRACED is explicitly | |
1610 | * requested, no event is reported; otherwise, report if the event | |
1611 | * for the type of forking is enabled. | |
09a05394 | 1612 | */ |
4b9d33e6 TH |
1613 | if (likely(user_mode(regs)) && !(clone_flags & CLONE_UNTRACED)) { |
1614 | if (clone_flags & CLONE_VFORK) | |
1615 | trace = PTRACE_EVENT_VFORK; | |
1616 | else if ((clone_flags & CSIGNAL) != SIGCHLD) | |
1617 | trace = PTRACE_EVENT_CLONE; | |
1618 | else | |
1619 | trace = PTRACE_EVENT_FORK; | |
1620 | ||
1621 | if (likely(!ptrace_event_enabled(current, trace))) | |
1622 | trace = 0; | |
1623 | } | |
1da177e4 | 1624 | |
a6f5e063 | 1625 | p = copy_process(clone_flags, stack_start, regs, stack_size, |
09a05394 | 1626 | child_tidptr, NULL, trace); |
1da177e4 LT |
1627 | /* |
1628 | * Do this prior waking up the new thread - the thread pointer | |
1629 | * might get invalid after that point, if the thread exits quickly. | |
1630 | */ | |
1631 | if (!IS_ERR(p)) { | |
1632 | struct completion vfork; | |
1633 | ||
0a16b607 MD |
1634 | trace_sched_process_fork(current, p); |
1635 | ||
6c5f3e7b | 1636 | nr = task_pid_vnr(p); |
30e49c26 PE |
1637 | |
1638 | if (clone_flags & CLONE_PARENT_SETTID) | |
1639 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1640 | |
1da177e4 LT |
1641 | if (clone_flags & CLONE_VFORK) { |
1642 | p->vfork_done = &vfork; | |
1643 | init_completion(&vfork); | |
d68b46fe | 1644 | get_task_struct(p); |
1da177e4 LT |
1645 | } |
1646 | ||
3e51e3ed | 1647 | wake_up_new_task(p); |
1da177e4 | 1648 | |
4b9d33e6 TH |
1649 | /* forking complete and child started to run, tell ptracer */ |
1650 | if (unlikely(trace)) | |
1651 | ptrace_event(trace, nr); | |
09a05394 | 1652 | |
1da177e4 | 1653 | if (clone_flags & CLONE_VFORK) { |
d68b46fe ON |
1654 | if (!wait_for_vfork_done(p, &vfork)) |
1655 | ptrace_event(PTRACE_EVENT_VFORK_DONE, nr); | |
1da177e4 LT |
1656 | } |
1657 | } else { | |
92476d7f | 1658 | nr = PTR_ERR(p); |
1da177e4 | 1659 | } |
92476d7f | 1660 | return nr; |
1da177e4 LT |
1661 | } |
1662 | ||
5fd63b30 RT |
1663 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1664 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1665 | #endif | |
1666 | ||
51cc5068 | 1667 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1668 | { |
1669 | struct sighand_struct *sighand = data; | |
1670 | ||
a35afb83 | 1671 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1672 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1673 | } |
1674 | ||
1da177e4 LT |
1675 | void __init proc_caches_init(void) |
1676 | { | |
1677 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1678 | sizeof(struct sighand_struct), 0, | |
2dff4405 VN |
1679 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
1680 | SLAB_NOTRACK, sighand_ctor); | |
1da177e4 LT |
1681 | signal_cachep = kmem_cache_create("signal_cache", |
1682 | sizeof(struct signal_struct), 0, | |
2dff4405 | 1683 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1684 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 1685 | sizeof(struct files_struct), 0, |
2dff4405 | 1686 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1687 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 1688 | sizeof(struct fs_struct), 0, |
2dff4405 | 1689 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
6345d24d LT |
1690 | /* |
1691 | * FIXME! The "sizeof(struct mm_struct)" currently includes the | |
1692 | * whole struct cpumask for the OFFSTACK case. We could change | |
1693 | * this to *only* allocate as much of it as required by the | |
1694 | * maximum number of CPU's we can ever have. The cpumask_allocation | |
1695 | * is at the end of the structure, exactly for that reason. | |
1696 | */ | |
1da177e4 | 1697 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1698 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
2dff4405 | 1699 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
33e5d769 | 1700 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC); |
8feae131 | 1701 | mmap_init(); |
66577193 | 1702 | nsproxy_cache_init(); |
1da177e4 | 1703 | } |
cf2e340f | 1704 | |
cf2e340f | 1705 | /* |
9bfb23fc | 1706 | * Check constraints on flags passed to the unshare system call. |
cf2e340f | 1707 | */ |
9bfb23fc | 1708 | static int check_unshare_flags(unsigned long unshare_flags) |
cf2e340f | 1709 | { |
9bfb23fc ON |
1710 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| |
1711 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| | |
1712 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET)) | |
1713 | return -EINVAL; | |
cf2e340f | 1714 | /* |
9bfb23fc ON |
1715 | * Not implemented, but pretend it works if there is nothing to |
1716 | * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND | |
1717 | * needs to unshare vm. | |
cf2e340f | 1718 | */ |
9bfb23fc ON |
1719 | if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) { |
1720 | /* FIXME: get_task_mm() increments ->mm_users */ | |
1721 | if (atomic_read(¤t->mm->mm_users) > 1) | |
1722 | return -EINVAL; | |
1723 | } | |
cf2e340f JD |
1724 | |
1725 | return 0; | |
1726 | } | |
1727 | ||
1728 | /* | |
99d1419d | 1729 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1730 | */ |
1731 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1732 | { | |
1733 | struct fs_struct *fs = current->fs; | |
1734 | ||
498052bb AV |
1735 | if (!(unshare_flags & CLONE_FS) || !fs) |
1736 | return 0; | |
1737 | ||
1738 | /* don't need lock here; in the worst case we'll do useless copy */ | |
1739 | if (fs->users == 1) | |
1740 | return 0; | |
1741 | ||
1742 | *new_fsp = copy_fs_struct(fs); | |
1743 | if (!*new_fsp) | |
1744 | return -ENOMEM; | |
cf2e340f JD |
1745 | |
1746 | return 0; | |
1747 | } | |
1748 | ||
cf2e340f | 1749 | /* |
a016f338 | 1750 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1751 | */ |
1752 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1753 | { | |
1754 | struct files_struct *fd = current->files; | |
a016f338 | 1755 | int error = 0; |
cf2e340f JD |
1756 | |
1757 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1758 | (fd && atomic_read(&fd->count) > 1)) { |
1759 | *new_fdp = dup_fd(fd, &error); | |
1760 | if (!*new_fdp) | |
1761 | return error; | |
1762 | } | |
cf2e340f JD |
1763 | |
1764 | return 0; | |
1765 | } | |
1766 | ||
cf2e340f JD |
1767 | /* |
1768 | * unshare allows a process to 'unshare' part of the process | |
1769 | * context which was originally shared using clone. copy_* | |
1770 | * functions used by do_fork() cannot be used here directly | |
1771 | * because they modify an inactive task_struct that is being | |
1772 | * constructed. Here we are modifying the current, active, | |
1773 | * task_struct. | |
1774 | */ | |
6559eed8 | 1775 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f | 1776 | { |
cf2e340f | 1777 | struct fs_struct *fs, *new_fs = NULL; |
cf2e340f | 1778 | struct files_struct *fd, *new_fd = NULL; |
cf7b708c | 1779 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 1780 | int do_sysvsem = 0; |
9bfb23fc | 1781 | int err; |
cf2e340f | 1782 | |
9bfb23fc ON |
1783 | err = check_unshare_flags(unshare_flags); |
1784 | if (err) | |
06f9d4f9 EB |
1785 | goto bad_unshare_out; |
1786 | ||
9bfb23fc ON |
1787 | /* |
1788 | * If unsharing namespace, must also unshare filesystem information. | |
1789 | */ | |
1790 | if (unshare_flags & CLONE_NEWNS) | |
1791 | unshare_flags |= CLONE_FS; | |
6013f67f MS |
1792 | /* |
1793 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
1794 | * to a new ipc namespace, the semaphore arrays from the old | |
1795 | * namespace are unreachable. | |
1796 | */ | |
1797 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 1798 | do_sysvsem = 1; |
fb0a685c DRO |
1799 | err = unshare_fs(unshare_flags, &new_fs); |
1800 | if (err) | |
9bfb23fc | 1801 | goto bad_unshare_out; |
fb0a685c DRO |
1802 | err = unshare_fd(unshare_flags, &new_fd); |
1803 | if (err) | |
9bfb23fc | 1804 | goto bad_unshare_cleanup_fs; |
fb0a685c DRO |
1805 | err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, new_fs); |
1806 | if (err) | |
9edff4ab | 1807 | goto bad_unshare_cleanup_fd; |
c0b2fc31 | 1808 | |
9bfb23fc | 1809 | if (new_fs || new_fd || do_sysvsem || new_nsproxy) { |
9edff4ab MS |
1810 | if (do_sysvsem) { |
1811 | /* | |
1812 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
1813 | */ | |
1814 | exit_sem(current); | |
1815 | } | |
ab516013 | 1816 | |
c0b2fc31 | 1817 | if (new_nsproxy) { |
cf7b708c PE |
1818 | switch_task_namespaces(current, new_nsproxy); |
1819 | new_nsproxy = NULL; | |
c0b2fc31 | 1820 | } |
cf2e340f | 1821 | |
cf7b708c PE |
1822 | task_lock(current); |
1823 | ||
cf2e340f JD |
1824 | if (new_fs) { |
1825 | fs = current->fs; | |
2a4419b5 | 1826 | spin_lock(&fs->lock); |
cf2e340f | 1827 | current->fs = new_fs; |
498052bb AV |
1828 | if (--fs->users) |
1829 | new_fs = NULL; | |
1830 | else | |
1831 | new_fs = fs; | |
2a4419b5 | 1832 | spin_unlock(&fs->lock); |
cf2e340f JD |
1833 | } |
1834 | ||
cf2e340f JD |
1835 | if (new_fd) { |
1836 | fd = current->files; | |
1837 | current->files = new_fd; | |
1838 | new_fd = fd; | |
1839 | } | |
1840 | ||
1841 | task_unlock(current); | |
1842 | } | |
1843 | ||
c0b2fc31 | 1844 | if (new_nsproxy) |
444f378b | 1845 | put_nsproxy(new_nsproxy); |
c0b2fc31 | 1846 | |
cf2e340f JD |
1847 | bad_unshare_cleanup_fd: |
1848 | if (new_fd) | |
1849 | put_files_struct(new_fd); | |
1850 | ||
cf2e340f JD |
1851 | bad_unshare_cleanup_fs: |
1852 | if (new_fs) | |
498052bb | 1853 | free_fs_struct(new_fs); |
cf2e340f | 1854 | |
cf2e340f JD |
1855 | bad_unshare_out: |
1856 | return err; | |
1857 | } | |
3b125388 AV |
1858 | |
1859 | /* | |
1860 | * Helper to unshare the files of the current task. | |
1861 | * We don't want to expose copy_files internals to | |
1862 | * the exec layer of the kernel. | |
1863 | */ | |
1864 | ||
1865 | int unshare_files(struct files_struct **displaced) | |
1866 | { | |
1867 | struct task_struct *task = current; | |
50704516 | 1868 | struct files_struct *copy = NULL; |
3b125388 AV |
1869 | int error; |
1870 | ||
1871 | error = unshare_fd(CLONE_FILES, ©); | |
1872 | if (error || !copy) { | |
1873 | *displaced = NULL; | |
1874 | return error; | |
1875 | } | |
1876 | *displaced = task->files; | |
1877 | task_lock(task); | |
1878 | task->files = copy; | |
1879 | task_unlock(task); | |
1880 | return 0; | |
1881 | } |