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> | |
09a05394 | 40 | #include <linux/tracehook.h> |
1da177e4 | 41 | #include <linux/futex.h> |
8141c7f3 | 42 | #include <linux/compat.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> |
6f4e6433 | 61 | #include <linux/proc_fs.h> |
fd0928df | 62 | #include <linux/blkdev.h> |
5ad4e53b | 63 | #include <linux/fs_struct.h> |
7c9f8861 | 64 | #include <linux/magic.h> |
cdd6c482 | 65 | #include <linux/perf_event.h> |
1da177e4 LT |
66 | |
67 | #include <asm/pgtable.h> | |
68 | #include <asm/pgalloc.h> | |
69 | #include <asm/uaccess.h> | |
70 | #include <asm/mmu_context.h> | |
71 | #include <asm/cacheflush.h> | |
72 | #include <asm/tlbflush.h> | |
73 | ||
ad8d75ff SR |
74 | #include <trace/events/sched.h> |
75 | ||
1da177e4 LT |
76 | /* |
77 | * Protected counters by write_lock_irq(&tasklist_lock) | |
78 | */ | |
79 | unsigned long total_forks; /* Handle normal Linux uptimes. */ | |
80 | int nr_threads; /* The idle threads do not count.. */ | |
81 | ||
82 | int max_threads; /* tunable limit on nr_threads */ | |
83 | ||
84 | DEFINE_PER_CPU(unsigned long, process_counts) = 0; | |
85 | ||
c59923a1 | 86 | __cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */ |
1da177e4 LT |
87 | |
88 | int nr_processes(void) | |
89 | { | |
90 | int cpu; | |
91 | int total = 0; | |
92 | ||
93 | for_each_online_cpu(cpu) | |
94 | total += per_cpu(process_counts, cpu); | |
95 | ||
96 | return total; | |
97 | } | |
98 | ||
99 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
100 | # define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL) | |
101 | # define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk)) | |
e18b890b | 102 | static struct kmem_cache *task_struct_cachep; |
1da177e4 LT |
103 | #endif |
104 | ||
b69c49b7 FT |
105 | #ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR |
106 | static inline struct thread_info *alloc_thread_info(struct task_struct *tsk) | |
107 | { | |
108 | #ifdef CONFIG_DEBUG_STACK_USAGE | |
109 | gfp_t mask = GFP_KERNEL | __GFP_ZERO; | |
110 | #else | |
111 | gfp_t mask = GFP_KERNEL; | |
112 | #endif | |
113 | return (struct thread_info *)__get_free_pages(mask, THREAD_SIZE_ORDER); | |
114 | } | |
115 | ||
116 | static inline void free_thread_info(struct thread_info *ti) | |
117 | { | |
118 | free_pages((unsigned long)ti, THREAD_SIZE_ORDER); | |
119 | } | |
120 | #endif | |
121 | ||
1da177e4 | 122 | /* SLAB cache for signal_struct structures (tsk->signal) */ |
e18b890b | 123 | static struct kmem_cache *signal_cachep; |
1da177e4 LT |
124 | |
125 | /* SLAB cache for sighand_struct structures (tsk->sighand) */ | |
e18b890b | 126 | struct kmem_cache *sighand_cachep; |
1da177e4 LT |
127 | |
128 | /* SLAB cache for files_struct structures (tsk->files) */ | |
e18b890b | 129 | struct kmem_cache *files_cachep; |
1da177e4 LT |
130 | |
131 | /* SLAB cache for fs_struct structures (tsk->fs) */ | |
e18b890b | 132 | struct kmem_cache *fs_cachep; |
1da177e4 LT |
133 | |
134 | /* SLAB cache for vm_area_struct structures */ | |
e18b890b | 135 | struct kmem_cache *vm_area_cachep; |
1da177e4 LT |
136 | |
137 | /* SLAB cache for mm_struct structures (tsk->mm) */ | |
e18b890b | 138 | static struct kmem_cache *mm_cachep; |
1da177e4 | 139 | |
c6a7f572 KM |
140 | static void account_kernel_stack(struct thread_info *ti, int account) |
141 | { | |
142 | struct zone *zone = page_zone(virt_to_page(ti)); | |
143 | ||
144 | mod_zone_page_state(zone, NR_KERNEL_STACK, account); | |
145 | } | |
146 | ||
1da177e4 LT |
147 | void free_task(struct task_struct *tsk) |
148 | { | |
3e26c149 | 149 | prop_local_destroy_single(&tsk->dirties); |
c6a7f572 | 150 | account_kernel_stack(tsk->stack, -1); |
f7e4217b | 151 | free_thread_info(tsk->stack); |
23f78d4a | 152 | rt_mutex_debug_task_free(tsk); |
fb52607a | 153 | ftrace_graph_exit_task(tsk); |
1da177e4 LT |
154 | free_task_struct(tsk); |
155 | } | |
156 | EXPORT_SYMBOL(free_task); | |
157 | ||
158d9ebd | 158 | void __put_task_struct(struct task_struct *tsk) |
1da177e4 | 159 | { |
270f722d | 160 | WARN_ON(!tsk->exit_state); |
1da177e4 LT |
161 | WARN_ON(atomic_read(&tsk->usage)); |
162 | WARN_ON(tsk == current); | |
163 | ||
e0e81739 | 164 | exit_creds(tsk); |
35df17c5 | 165 | delayacct_tsk_free(tsk); |
1da177e4 LT |
166 | |
167 | if (!profile_handoff_task(tsk)) | |
168 | free_task(tsk); | |
169 | } | |
170 | ||
2adee9b3 SS |
171 | /* |
172 | * macro override instead of weak attribute alias, to workaround | |
173 | * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions. | |
174 | */ | |
175 | #ifndef arch_task_cache_init | |
176 | #define arch_task_cache_init() | |
177 | #endif | |
61c4628b | 178 | |
1da177e4 LT |
179 | void __init fork_init(unsigned long mempages) |
180 | { | |
181 | #ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR | |
182 | #ifndef ARCH_MIN_TASKALIGN | |
183 | #define ARCH_MIN_TASKALIGN L1_CACHE_BYTES | |
184 | #endif | |
185 | /* create a slab on which task_structs can be allocated */ | |
186 | task_struct_cachep = | |
187 | kmem_cache_create("task_struct", sizeof(struct task_struct), | |
2dff4405 | 188 | ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL); |
1da177e4 LT |
189 | #endif |
190 | ||
61c4628b SS |
191 | /* do the arch specific task caches init */ |
192 | arch_task_cache_init(); | |
193 | ||
1da177e4 LT |
194 | /* |
195 | * The default maximum number of threads is set to a safe | |
196 | * value: the thread structures can take up at most half | |
197 | * of memory. | |
198 | */ | |
199 | max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE); | |
200 | ||
201 | /* | |
202 | * we need to allow at least 20 threads to boot a system | |
203 | */ | |
204 | if(max_threads < 20) | |
205 | max_threads = 20; | |
206 | ||
207 | init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2; | |
208 | init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2; | |
209 | init_task.signal->rlim[RLIMIT_SIGPENDING] = | |
210 | init_task.signal->rlim[RLIMIT_NPROC]; | |
211 | } | |
212 | ||
61c4628b SS |
213 | int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst, |
214 | struct task_struct *src) | |
215 | { | |
216 | *dst = *src; | |
217 | return 0; | |
218 | } | |
219 | ||
1da177e4 LT |
220 | static struct task_struct *dup_task_struct(struct task_struct *orig) |
221 | { | |
222 | struct task_struct *tsk; | |
223 | struct thread_info *ti; | |
7c9f8861 ES |
224 | unsigned long *stackend; |
225 | ||
3e26c149 | 226 | int err; |
1da177e4 LT |
227 | |
228 | prepare_to_copy(orig); | |
229 | ||
230 | tsk = alloc_task_struct(); | |
231 | if (!tsk) | |
232 | return NULL; | |
233 | ||
234 | ti = alloc_thread_info(tsk); | |
235 | if (!ti) { | |
236 | free_task_struct(tsk); | |
237 | return NULL; | |
238 | } | |
239 | ||
61c4628b SS |
240 | err = arch_dup_task_struct(tsk, orig); |
241 | if (err) | |
242 | goto out; | |
243 | ||
f7e4217b | 244 | tsk->stack = ti; |
3e26c149 PZ |
245 | |
246 | err = prop_local_init_single(&tsk->dirties); | |
61c4628b SS |
247 | if (err) |
248 | goto out; | |
3e26c149 | 249 | |
10ebffde | 250 | setup_thread_stack(tsk, orig); |
7c9f8861 ES |
251 | stackend = end_of_stack(tsk); |
252 | *stackend = STACK_END_MAGIC; /* for overflow detection */ | |
1da177e4 | 253 | |
0a425405 AV |
254 | #ifdef CONFIG_CC_STACKPROTECTOR |
255 | tsk->stack_canary = get_random_int(); | |
256 | #endif | |
257 | ||
1da177e4 LT |
258 | /* One for us, one for whoever does the "release_task()" (usually parent) */ |
259 | atomic_set(&tsk->usage,2); | |
4b5d37ac | 260 | atomic_set(&tsk->fs_excl, 0); |
6c5c9341 | 261 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
2056a782 | 262 | tsk->btrace_seq = 0; |
6c5c9341 | 263 | #endif |
a0aa7f68 | 264 | tsk->splice_pipe = NULL; |
c6a7f572 KM |
265 | |
266 | account_kernel_stack(ti, 1); | |
267 | ||
1da177e4 | 268 | return tsk; |
61c4628b SS |
269 | |
270 | out: | |
271 | free_thread_info(ti); | |
272 | free_task_struct(tsk); | |
273 | return NULL; | |
1da177e4 LT |
274 | } |
275 | ||
276 | #ifdef CONFIG_MMU | |
a39bc516 | 277 | static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm) |
1da177e4 | 278 | { |
fd3e42fc | 279 | struct vm_area_struct *mpnt, *tmp, **pprev; |
1da177e4 LT |
280 | struct rb_node **rb_link, *rb_parent; |
281 | int retval; | |
282 | unsigned long charge; | |
283 | struct mempolicy *pol; | |
284 | ||
285 | down_write(&oldmm->mmap_sem); | |
ec8c0446 | 286 | flush_cache_dup_mm(oldmm); |
ad339451 IM |
287 | /* |
288 | * Not linked in yet - no deadlock potential: | |
289 | */ | |
290 | down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING); | |
7ee78232 | 291 | |
1da177e4 LT |
292 | mm->locked_vm = 0; |
293 | mm->mmap = NULL; | |
294 | mm->mmap_cache = NULL; | |
295 | mm->free_area_cache = oldmm->mmap_base; | |
1363c3cd | 296 | mm->cached_hole_size = ~0UL; |
1da177e4 | 297 | mm->map_count = 0; |
94894244 | 298 | cpumask_clear(mm_cpumask(mm)); |
1da177e4 LT |
299 | mm->mm_rb = RB_ROOT; |
300 | rb_link = &mm->mm_rb.rb_node; | |
301 | rb_parent = NULL; | |
302 | pprev = &mm->mmap; | |
f8af4da3 HD |
303 | retval = ksm_fork(mm, oldmm); |
304 | if (retval) | |
305 | goto out; | |
1da177e4 | 306 | |
fd3e42fc | 307 | for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) { |
1da177e4 LT |
308 | struct file *file; |
309 | ||
310 | if (mpnt->vm_flags & VM_DONTCOPY) { | |
3b6bfcdb HD |
311 | long pages = vma_pages(mpnt); |
312 | mm->total_vm -= pages; | |
ab50b8ed | 313 | vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file, |
3b6bfcdb | 314 | -pages); |
1da177e4 LT |
315 | continue; |
316 | } | |
317 | charge = 0; | |
318 | if (mpnt->vm_flags & VM_ACCOUNT) { | |
319 | unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT; | |
320 | if (security_vm_enough_memory(len)) | |
321 | goto fail_nomem; | |
322 | charge = len; | |
323 | } | |
e94b1766 | 324 | tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL); |
1da177e4 LT |
325 | if (!tmp) |
326 | goto fail_nomem; | |
327 | *tmp = *mpnt; | |
846a16bf | 328 | pol = mpol_dup(vma_policy(mpnt)); |
1da177e4 LT |
329 | retval = PTR_ERR(pol); |
330 | if (IS_ERR(pol)) | |
331 | goto fail_nomem_policy; | |
332 | vma_set_policy(tmp, pol); | |
333 | tmp->vm_flags &= ~VM_LOCKED; | |
334 | tmp->vm_mm = mm; | |
335 | tmp->vm_next = NULL; | |
336 | anon_vma_link(tmp); | |
337 | file = tmp->vm_file; | |
338 | if (file) { | |
f3a43f3f | 339 | struct inode *inode = file->f_path.dentry->d_inode; |
b88ed205 HD |
340 | struct address_space *mapping = file->f_mapping; |
341 | ||
1da177e4 LT |
342 | get_file(file); |
343 | if (tmp->vm_flags & VM_DENYWRITE) | |
344 | atomic_dec(&inode->i_writecount); | |
b88ed205 HD |
345 | spin_lock(&mapping->i_mmap_lock); |
346 | if (tmp->vm_flags & VM_SHARED) | |
347 | mapping->i_mmap_writable++; | |
1da177e4 | 348 | tmp->vm_truncate_count = mpnt->vm_truncate_count; |
b88ed205 HD |
349 | flush_dcache_mmap_lock(mapping); |
350 | /* insert tmp into the share list, just after mpnt */ | |
1da177e4 | 351 | vma_prio_tree_add(tmp, mpnt); |
b88ed205 HD |
352 | flush_dcache_mmap_unlock(mapping); |
353 | spin_unlock(&mapping->i_mmap_lock); | |
1da177e4 LT |
354 | } |
355 | ||
a1e78772 MG |
356 | /* |
357 | * Clear hugetlb-related page reserves for children. This only | |
358 | * affects MAP_PRIVATE mappings. Faults generated by the child | |
359 | * are not guaranteed to succeed, even if read-only | |
360 | */ | |
361 | if (is_vm_hugetlb_page(tmp)) | |
362 | reset_vma_resv_huge_pages(tmp); | |
363 | ||
1da177e4 | 364 | /* |
7ee78232 | 365 | * Link in the new vma and copy the page table entries. |
1da177e4 | 366 | */ |
1da177e4 LT |
367 | *pprev = tmp; |
368 | pprev = &tmp->vm_next; | |
369 | ||
370 | __vma_link_rb(mm, tmp, rb_link, rb_parent); | |
371 | rb_link = &tmp->vm_rb.rb_right; | |
372 | rb_parent = &tmp->vm_rb; | |
373 | ||
374 | mm->map_count++; | |
0b0db14c | 375 | retval = copy_page_range(mm, oldmm, mpnt); |
1da177e4 LT |
376 | |
377 | if (tmp->vm_ops && tmp->vm_ops->open) | |
378 | tmp->vm_ops->open(tmp); | |
379 | ||
380 | if (retval) | |
381 | goto out; | |
382 | } | |
d6dd61c8 JF |
383 | /* a new mm has just been created */ |
384 | arch_dup_mmap(oldmm, mm); | |
1da177e4 | 385 | retval = 0; |
1da177e4 | 386 | out: |
7ee78232 | 387 | up_write(&mm->mmap_sem); |
fd3e42fc | 388 | flush_tlb_mm(oldmm); |
1da177e4 LT |
389 | up_write(&oldmm->mmap_sem); |
390 | return retval; | |
391 | fail_nomem_policy: | |
392 | kmem_cache_free(vm_area_cachep, tmp); | |
393 | fail_nomem: | |
394 | retval = -ENOMEM; | |
395 | vm_unacct_memory(charge); | |
396 | goto out; | |
397 | } | |
398 | ||
399 | static inline int mm_alloc_pgd(struct mm_struct * mm) | |
400 | { | |
401 | mm->pgd = pgd_alloc(mm); | |
402 | if (unlikely(!mm->pgd)) | |
403 | return -ENOMEM; | |
404 | return 0; | |
405 | } | |
406 | ||
407 | static inline void mm_free_pgd(struct mm_struct * mm) | |
408 | { | |
5e541973 | 409 | pgd_free(mm, mm->pgd); |
1da177e4 LT |
410 | } |
411 | #else | |
412 | #define dup_mmap(mm, oldmm) (0) | |
413 | #define mm_alloc_pgd(mm) (0) | |
414 | #define mm_free_pgd(mm) | |
415 | #endif /* CONFIG_MMU */ | |
416 | ||
23ff4440 | 417 | __cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock); |
1da177e4 | 418 | |
e94b1766 | 419 | #define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL)) |
1da177e4 LT |
420 | #define free_mm(mm) (kmem_cache_free(mm_cachep, (mm))) |
421 | ||
4cb0e11b HK |
422 | static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT; |
423 | ||
424 | static int __init coredump_filter_setup(char *s) | |
425 | { | |
426 | default_dump_filter = | |
427 | (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) & | |
428 | MMF_DUMP_FILTER_MASK; | |
429 | return 1; | |
430 | } | |
431 | ||
432 | __setup("coredump_filter=", coredump_filter_setup); | |
433 | ||
1da177e4 LT |
434 | #include <linux/init_task.h> |
435 | ||
78fb7466 | 436 | static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p) |
1da177e4 LT |
437 | { |
438 | atomic_set(&mm->mm_users, 1); | |
439 | atomic_set(&mm->mm_count, 1); | |
440 | init_rwsem(&mm->mmap_sem); | |
441 | INIT_LIST_HEAD(&mm->mmlist); | |
f8af4da3 HD |
442 | mm->flags = (current->mm) ? |
443 | (current->mm->flags & MMF_INIT_MASK) : default_dump_filter; | |
999d9fc1 | 444 | mm->core_state = NULL; |
1da177e4 | 445 | mm->nr_ptes = 0; |
4294621f | 446 | set_mm_counter(mm, file_rss, 0); |
404351e6 | 447 | set_mm_counter(mm, anon_rss, 0); |
1da177e4 | 448 | spin_lock_init(&mm->page_table_lock); |
abf137dd JA |
449 | spin_lock_init(&mm->ioctx_lock); |
450 | INIT_HLIST_HEAD(&mm->ioctx_list); | |
1da177e4 | 451 | mm->free_area_cache = TASK_UNMAPPED_BASE; |
1363c3cd | 452 | mm->cached_hole_size = ~0UL; |
cf475ad2 | 453 | mm_init_owner(mm, p); |
1da177e4 LT |
454 | |
455 | if (likely(!mm_alloc_pgd(mm))) { | |
456 | mm->def_flags = 0; | |
cddb8a5c | 457 | mmu_notifier_mm_init(mm); |
1da177e4 LT |
458 | return mm; |
459 | } | |
78fb7466 | 460 | |
1da177e4 LT |
461 | free_mm(mm); |
462 | return NULL; | |
463 | } | |
464 | ||
465 | /* | |
466 | * Allocate and initialize an mm_struct. | |
467 | */ | |
468 | struct mm_struct * mm_alloc(void) | |
469 | { | |
470 | struct mm_struct * mm; | |
471 | ||
472 | mm = allocate_mm(); | |
473 | if (mm) { | |
474 | memset(mm, 0, sizeof(*mm)); | |
78fb7466 | 475 | mm = mm_init(mm, current); |
1da177e4 LT |
476 | } |
477 | return mm; | |
478 | } | |
479 | ||
480 | /* | |
481 | * Called when the last reference to the mm | |
482 | * is dropped: either by a lazy thread or by | |
483 | * mmput. Free the page directory and the mm. | |
484 | */ | |
7ad5b3a5 | 485 | void __mmdrop(struct mm_struct *mm) |
1da177e4 LT |
486 | { |
487 | BUG_ON(mm == &init_mm); | |
488 | mm_free_pgd(mm); | |
489 | destroy_context(mm); | |
cddb8a5c | 490 | mmu_notifier_mm_destroy(mm); |
1da177e4 LT |
491 | free_mm(mm); |
492 | } | |
6d4e4c4f | 493 | EXPORT_SYMBOL_GPL(__mmdrop); |
1da177e4 LT |
494 | |
495 | /* | |
496 | * Decrement the use count and release all resources for an mm. | |
497 | */ | |
498 | void mmput(struct mm_struct *mm) | |
499 | { | |
0ae26f1b AM |
500 | might_sleep(); |
501 | ||
1da177e4 LT |
502 | if (atomic_dec_and_test(&mm->mm_users)) { |
503 | exit_aio(mm); | |
1c2fb7a4 | 504 | ksm_exit(mm); |
1da177e4 | 505 | exit_mmap(mm); |
925d1c40 | 506 | set_mm_exe_file(mm, NULL); |
1da177e4 LT |
507 | if (!list_empty(&mm->mmlist)) { |
508 | spin_lock(&mmlist_lock); | |
509 | list_del(&mm->mmlist); | |
510 | spin_unlock(&mmlist_lock); | |
511 | } | |
512 | put_swap_token(mm); | |
513 | mmdrop(mm); | |
514 | } | |
515 | } | |
516 | EXPORT_SYMBOL_GPL(mmput); | |
517 | ||
518 | /** | |
519 | * get_task_mm - acquire a reference to the task's mm | |
520 | * | |
246bb0b1 | 521 | * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning |
1da177e4 LT |
522 | * this kernel workthread has transiently adopted a user mm with use_mm, |
523 | * to do its AIO) is not set and if so returns a reference to it, after | |
524 | * bumping up the use count. User must release the mm via mmput() | |
525 | * after use. Typically used by /proc and ptrace. | |
526 | */ | |
527 | struct mm_struct *get_task_mm(struct task_struct *task) | |
528 | { | |
529 | struct mm_struct *mm; | |
530 | ||
531 | task_lock(task); | |
532 | mm = task->mm; | |
533 | if (mm) { | |
246bb0b1 | 534 | if (task->flags & PF_KTHREAD) |
1da177e4 LT |
535 | mm = NULL; |
536 | else | |
537 | atomic_inc(&mm->mm_users); | |
538 | } | |
539 | task_unlock(task); | |
540 | return mm; | |
541 | } | |
542 | EXPORT_SYMBOL_GPL(get_task_mm); | |
543 | ||
544 | /* Please note the differences between mmput and mm_release. | |
545 | * mmput is called whenever we stop holding onto a mm_struct, | |
546 | * error success whatever. | |
547 | * | |
548 | * mm_release is called after a mm_struct has been removed | |
549 | * from the current process. | |
550 | * | |
551 | * This difference is important for error handling, when we | |
552 | * only half set up a mm_struct for a new process and need to restore | |
553 | * the old one. Because we mmput the new mm_struct before | |
554 | * restoring the old one. . . | |
555 | * Eric Biederman 10 January 1998 | |
556 | */ | |
557 | void mm_release(struct task_struct *tsk, struct mm_struct *mm) | |
558 | { | |
559 | struct completion *vfork_done = tsk->vfork_done; | |
560 | ||
8141c7f3 LT |
561 | /* Get rid of any futexes when releasing the mm */ |
562 | #ifdef CONFIG_FUTEX | |
563 | if (unlikely(tsk->robust_list)) | |
564 | exit_robust_list(tsk); | |
565 | #ifdef CONFIG_COMPAT | |
566 | if (unlikely(tsk->compat_robust_list)) | |
567 | compat_exit_robust_list(tsk); | |
568 | #endif | |
569 | #endif | |
570 | ||
1da177e4 LT |
571 | /* Get rid of any cached register state */ |
572 | deactivate_mm(tsk, mm); | |
573 | ||
574 | /* notify parent sleeping on vfork() */ | |
575 | if (vfork_done) { | |
576 | tsk->vfork_done = NULL; | |
577 | complete(vfork_done); | |
578 | } | |
fec1d011 RM |
579 | |
580 | /* | |
581 | * If we're exiting normally, clear a user-space tid field if | |
582 | * requested. We leave this alone when dying by signal, to leave | |
583 | * the value intact in a core dump, and to save the unnecessary | |
584 | * trouble otherwise. Userland only wants this done for a sys_exit. | |
585 | */ | |
9c8a8228 ED |
586 | if (tsk->clear_child_tid) { |
587 | if (!(tsk->flags & PF_SIGNALED) && | |
588 | atomic_read(&mm->mm_users) > 1) { | |
589 | /* | |
590 | * We don't check the error code - if userspace has | |
591 | * not set up a proper pointer then tough luck. | |
592 | */ | |
593 | put_user(0, tsk->clear_child_tid); | |
594 | sys_futex(tsk->clear_child_tid, FUTEX_WAKE, | |
595 | 1, NULL, NULL, 0); | |
596 | } | |
1da177e4 | 597 | tsk->clear_child_tid = NULL; |
1da177e4 LT |
598 | } |
599 | } | |
600 | ||
a0a7ec30 JD |
601 | /* |
602 | * Allocate a new mm structure and copy contents from the | |
603 | * mm structure of the passed in task structure. | |
604 | */ | |
402b0862 | 605 | struct mm_struct *dup_mm(struct task_struct *tsk) |
a0a7ec30 JD |
606 | { |
607 | struct mm_struct *mm, *oldmm = current->mm; | |
608 | int err; | |
609 | ||
610 | if (!oldmm) | |
611 | return NULL; | |
612 | ||
613 | mm = allocate_mm(); | |
614 | if (!mm) | |
615 | goto fail_nomem; | |
616 | ||
617 | memcpy(mm, oldmm, sizeof(*mm)); | |
618 | ||
7602bdf2 AC |
619 | /* Initializing for Swap token stuff */ |
620 | mm->token_priority = 0; | |
621 | mm->last_interval = 0; | |
622 | ||
78fb7466 | 623 | if (!mm_init(mm, tsk)) |
a0a7ec30 JD |
624 | goto fail_nomem; |
625 | ||
626 | if (init_new_context(tsk, mm)) | |
627 | goto fail_nocontext; | |
628 | ||
925d1c40 MH |
629 | dup_mm_exe_file(oldmm, mm); |
630 | ||
a0a7ec30 JD |
631 | err = dup_mmap(mm, oldmm); |
632 | if (err) | |
633 | goto free_pt; | |
634 | ||
635 | mm->hiwater_rss = get_mm_rss(mm); | |
636 | mm->hiwater_vm = mm->total_vm; | |
637 | ||
638 | return mm; | |
639 | ||
640 | free_pt: | |
641 | mmput(mm); | |
642 | ||
643 | fail_nomem: | |
644 | return NULL; | |
645 | ||
646 | fail_nocontext: | |
647 | /* | |
648 | * If init_new_context() failed, we cannot use mmput() to free the mm | |
649 | * because it calls destroy_context() | |
650 | */ | |
651 | mm_free_pgd(mm); | |
652 | free_mm(mm); | |
653 | return NULL; | |
654 | } | |
655 | ||
1da177e4 LT |
656 | static int copy_mm(unsigned long clone_flags, struct task_struct * tsk) |
657 | { | |
658 | struct mm_struct * mm, *oldmm; | |
659 | int retval; | |
660 | ||
661 | tsk->min_flt = tsk->maj_flt = 0; | |
662 | tsk->nvcsw = tsk->nivcsw = 0; | |
17406b82 MSB |
663 | #ifdef CONFIG_DETECT_HUNG_TASK |
664 | tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw; | |
665 | #endif | |
1da177e4 LT |
666 | |
667 | tsk->mm = NULL; | |
668 | tsk->active_mm = NULL; | |
669 | ||
670 | /* | |
671 | * Are we cloning a kernel thread? | |
672 | * | |
673 | * We need to steal a active VM for that.. | |
674 | */ | |
675 | oldmm = current->mm; | |
676 | if (!oldmm) | |
677 | return 0; | |
678 | ||
679 | if (clone_flags & CLONE_VM) { | |
680 | atomic_inc(&oldmm->mm_users); | |
681 | mm = oldmm; | |
1da177e4 LT |
682 | goto good_mm; |
683 | } | |
684 | ||
685 | retval = -ENOMEM; | |
a0a7ec30 | 686 | mm = dup_mm(tsk); |
1da177e4 LT |
687 | if (!mm) |
688 | goto fail_nomem; | |
689 | ||
1da177e4 | 690 | good_mm: |
7602bdf2 AC |
691 | /* Initializing for Swap token stuff */ |
692 | mm->token_priority = 0; | |
693 | mm->last_interval = 0; | |
694 | ||
1da177e4 LT |
695 | tsk->mm = mm; |
696 | tsk->active_mm = mm; | |
697 | return 0; | |
698 | ||
1da177e4 LT |
699 | fail_nomem: |
700 | return retval; | |
1da177e4 LT |
701 | } |
702 | ||
a39bc516 | 703 | static int copy_fs(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 | 704 | { |
498052bb | 705 | struct fs_struct *fs = current->fs; |
1da177e4 | 706 | if (clone_flags & CLONE_FS) { |
498052bb AV |
707 | /* tsk->fs is already what we want */ |
708 | write_lock(&fs->lock); | |
709 | if (fs->in_exec) { | |
710 | write_unlock(&fs->lock); | |
711 | return -EAGAIN; | |
712 | } | |
713 | fs->users++; | |
714 | write_unlock(&fs->lock); | |
1da177e4 LT |
715 | return 0; |
716 | } | |
498052bb | 717 | tsk->fs = copy_fs_struct(fs); |
1da177e4 LT |
718 | if (!tsk->fs) |
719 | return -ENOMEM; | |
720 | return 0; | |
721 | } | |
722 | ||
a016f338 JD |
723 | static int copy_files(unsigned long clone_flags, struct task_struct * tsk) |
724 | { | |
725 | struct files_struct *oldf, *newf; | |
726 | int error = 0; | |
727 | ||
728 | /* | |
729 | * A background process may not have any files ... | |
730 | */ | |
731 | oldf = current->files; | |
732 | if (!oldf) | |
733 | goto out; | |
734 | ||
735 | if (clone_flags & CLONE_FILES) { | |
736 | atomic_inc(&oldf->count); | |
737 | goto out; | |
738 | } | |
739 | ||
a016f338 JD |
740 | newf = dup_fd(oldf, &error); |
741 | if (!newf) | |
742 | goto out; | |
743 | ||
744 | tsk->files = newf; | |
745 | error = 0; | |
746 | out: | |
747 | return error; | |
748 | } | |
749 | ||
fadad878 | 750 | static int copy_io(unsigned long clone_flags, struct task_struct *tsk) |
fd0928df JA |
751 | { |
752 | #ifdef CONFIG_BLOCK | |
753 | struct io_context *ioc = current->io_context; | |
754 | ||
755 | if (!ioc) | |
756 | return 0; | |
fadad878 JA |
757 | /* |
758 | * Share io context with parent, if CLONE_IO is set | |
759 | */ | |
760 | if (clone_flags & CLONE_IO) { | |
761 | tsk->io_context = ioc_task_link(ioc); | |
762 | if (unlikely(!tsk->io_context)) | |
763 | return -ENOMEM; | |
764 | } else if (ioprio_valid(ioc->ioprio)) { | |
fd0928df JA |
765 | tsk->io_context = alloc_io_context(GFP_KERNEL, -1); |
766 | if (unlikely(!tsk->io_context)) | |
767 | return -ENOMEM; | |
768 | ||
fd0928df JA |
769 | tsk->io_context->ioprio = ioc->ioprio; |
770 | } | |
771 | #endif | |
772 | return 0; | |
773 | } | |
774 | ||
a39bc516 | 775 | static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
776 | { |
777 | struct sighand_struct *sig; | |
778 | ||
60348802 | 779 | if (clone_flags & CLONE_SIGHAND) { |
1da177e4 LT |
780 | atomic_inc(¤t->sighand->count); |
781 | return 0; | |
782 | } | |
783 | sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
e56d0903 | 784 | rcu_assign_pointer(tsk->sighand, sig); |
1da177e4 LT |
785 | if (!sig) |
786 | return -ENOMEM; | |
1da177e4 LT |
787 | atomic_set(&sig->count, 1); |
788 | memcpy(sig->action, current->sighand->action, sizeof(sig->action)); | |
789 | return 0; | |
790 | } | |
791 | ||
a7e5328a | 792 | void __cleanup_sighand(struct sighand_struct *sighand) |
c81addc9 | 793 | { |
c81addc9 ON |
794 | if (atomic_dec_and_test(&sighand->count)) |
795 | kmem_cache_free(sighand_cachep, sighand); | |
796 | } | |
797 | ||
f06febc9 FM |
798 | |
799 | /* | |
800 | * Initialize POSIX timer handling for a thread group. | |
801 | */ | |
802 | static void posix_cpu_timers_init_group(struct signal_struct *sig) | |
803 | { | |
804 | /* Thread group counters. */ | |
805 | thread_group_cputime_init(sig); | |
806 | ||
807 | /* Expiration times and increments. */ | |
808 | sig->it_virt_expires = cputime_zero; | |
809 | sig->it_virt_incr = cputime_zero; | |
810 | sig->it_prof_expires = cputime_zero; | |
811 | sig->it_prof_incr = cputime_zero; | |
812 | ||
813 | /* Cached expiration times. */ | |
814 | sig->cputime_expires.prof_exp = cputime_zero; | |
815 | sig->cputime_expires.virt_exp = cputime_zero; | |
816 | sig->cputime_expires.sched_exp = 0; | |
817 | ||
6279a751 ON |
818 | if (sig->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) { |
819 | sig->cputime_expires.prof_exp = | |
820 | secs_to_cputime(sig->rlim[RLIMIT_CPU].rlim_cur); | |
821 | sig->cputimer.running = 1; | |
822 | } | |
823 | ||
f06febc9 FM |
824 | /* The timer lists. */ |
825 | INIT_LIST_HEAD(&sig->cpu_timers[0]); | |
826 | INIT_LIST_HEAD(&sig->cpu_timers[1]); | |
827 | INIT_LIST_HEAD(&sig->cpu_timers[2]); | |
828 | } | |
829 | ||
a39bc516 | 830 | static int copy_signal(unsigned long clone_flags, struct task_struct *tsk) |
1da177e4 LT |
831 | { |
832 | struct signal_struct *sig; | |
1da177e4 | 833 | |
4ab6c083 | 834 | if (clone_flags & CLONE_THREAD) |
490dea45 | 835 | return 0; |
490dea45 | 836 | |
6279a751 | 837 | sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL); |
1da177e4 LT |
838 | tsk->signal = sig; |
839 | if (!sig) | |
840 | return -ENOMEM; | |
841 | ||
1da177e4 LT |
842 | atomic_set(&sig->count, 1); |
843 | atomic_set(&sig->live, 1); | |
844 | init_waitqueue_head(&sig->wait_chldexit); | |
845 | sig->flags = 0; | |
b3bfa0cb SB |
846 | if (clone_flags & CLONE_NEWPID) |
847 | sig->flags |= SIGNAL_UNKILLABLE; | |
1da177e4 LT |
848 | sig->group_exit_code = 0; |
849 | sig->group_exit_task = NULL; | |
850 | sig->group_stop_count = 0; | |
db51aecc | 851 | sig->curr_target = tsk; |
1da177e4 LT |
852 | init_sigpending(&sig->shared_pending); |
853 | INIT_LIST_HEAD(&sig->posix_timers); | |
854 | ||
c9cb2e3d | 855 | hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); |
2ff678b8 | 856 | sig->it_real_incr.tv64 = 0; |
1da177e4 | 857 | sig->real_timer.function = it_real_fn; |
1da177e4 | 858 | |
1da177e4 | 859 | sig->leader = 0; /* session leadership doesn't inherit */ |
ab521dc0 | 860 | sig->tty_old_pgrp = NULL; |
9c9f4ded | 861 | sig->tty = NULL; |
1da177e4 | 862 | |
32bd671d | 863 | sig->utime = sig->stime = sig->cutime = sig->cstime = cputime_zero; |
9ac52315 LV |
864 | sig->gtime = cputime_zero; |
865 | sig->cgtime = cputime_zero; | |
1da177e4 LT |
866 | sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0; |
867 | sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0; | |
6eaeeaba | 868 | sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0; |
1f10206c | 869 | sig->maxrss = sig->cmaxrss = 0; |
5995477a | 870 | task_io_accounting_init(&sig->ioac); |
32bd671d | 871 | sig->sum_sched_runtime = 0; |
ad4ecbcb | 872 | taskstats_tgid_init(sig); |
1da177e4 LT |
873 | |
874 | task_lock(current->group_leader); | |
875 | memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim); | |
876 | task_unlock(current->group_leader); | |
877 | ||
6279a751 ON |
878 | posix_cpu_timers_init_group(sig); |
879 | ||
0e464814 | 880 | acct_init_pacct(&sig->pacct); |
1da177e4 | 881 | |
522ed776 MT |
882 | tty_audit_fork(sig); |
883 | ||
28b83c51 KM |
884 | sig->oom_adj = current->signal->oom_adj; |
885 | ||
1da177e4 LT |
886 | return 0; |
887 | } | |
888 | ||
6b3934ef ON |
889 | void __cleanup_signal(struct signal_struct *sig) |
890 | { | |
f06febc9 | 891 | thread_group_cputime_free(sig); |
9c9f4ded | 892 | tty_kref_put(sig->tty); |
6b3934ef ON |
893 | kmem_cache_free(signal_cachep, sig); |
894 | } | |
895 | ||
a39bc516 | 896 | static void copy_flags(unsigned long clone_flags, struct task_struct *p) |
1da177e4 LT |
897 | { |
898 | unsigned long new_flags = p->flags; | |
899 | ||
83144186 | 900 | new_flags &= ~PF_SUPERPRIV; |
1da177e4 | 901 | new_flags |= PF_FORKNOEXEC; |
09a05394 | 902 | new_flags |= PF_STARTING; |
1da177e4 | 903 | p->flags = new_flags; |
2e131895 | 904 | clear_freeze_flag(p); |
1da177e4 LT |
905 | } |
906 | ||
17da2bd9 | 907 | SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr) |
1da177e4 LT |
908 | { |
909 | current->clear_child_tid = tidptr; | |
910 | ||
b488893a | 911 | return task_pid_vnr(current); |
1da177e4 LT |
912 | } |
913 | ||
a39bc516 | 914 | static void rt_mutex_init_task(struct task_struct *p) |
23f78d4a | 915 | { |
23f78d4a | 916 | spin_lock_init(&p->pi_lock); |
e29e175b | 917 | #ifdef CONFIG_RT_MUTEXES |
23f78d4a IM |
918 | plist_head_init(&p->pi_waiters, &p->pi_lock); |
919 | p->pi_blocked_on = NULL; | |
23f78d4a IM |
920 | #endif |
921 | } | |
922 | ||
cf475ad2 BS |
923 | #ifdef CONFIG_MM_OWNER |
924 | void mm_init_owner(struct mm_struct *mm, struct task_struct *p) | |
925 | { | |
926 | mm->owner = p; | |
927 | } | |
928 | #endif /* CONFIG_MM_OWNER */ | |
929 | ||
f06febc9 FM |
930 | /* |
931 | * Initialize POSIX timer handling for a single task. | |
932 | */ | |
933 | static void posix_cpu_timers_init(struct task_struct *tsk) | |
934 | { | |
935 | tsk->cputime_expires.prof_exp = cputime_zero; | |
936 | tsk->cputime_expires.virt_exp = cputime_zero; | |
937 | tsk->cputime_expires.sched_exp = 0; | |
938 | INIT_LIST_HEAD(&tsk->cpu_timers[0]); | |
939 | INIT_LIST_HEAD(&tsk->cpu_timers[1]); | |
940 | INIT_LIST_HEAD(&tsk->cpu_timers[2]); | |
941 | } | |
942 | ||
1da177e4 LT |
943 | /* |
944 | * This creates a new process as a copy of the old one, | |
945 | * but does not actually start it yet. | |
946 | * | |
947 | * It copies the registers, and all the appropriate | |
948 | * parts of the process environment (as per the clone | |
949 | * flags). The actual kick-off is left to the caller. | |
950 | */ | |
36c8b586 IM |
951 | static struct task_struct *copy_process(unsigned long clone_flags, |
952 | unsigned long stack_start, | |
953 | struct pt_regs *regs, | |
954 | unsigned long stack_size, | |
36c8b586 | 955 | int __user *child_tidptr, |
09a05394 RM |
956 | struct pid *pid, |
957 | int trace) | |
1da177e4 LT |
958 | { |
959 | int retval; | |
a24efe62 | 960 | struct task_struct *p; |
b4f48b63 | 961 | int cgroup_callbacks_done = 0; |
1da177e4 LT |
962 | |
963 | if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS)) | |
964 | return ERR_PTR(-EINVAL); | |
965 | ||
966 | /* | |
967 | * Thread groups must share signals as well, and detached threads | |
968 | * can only be started up within the thread group. | |
969 | */ | |
970 | if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND)) | |
971 | return ERR_PTR(-EINVAL); | |
972 | ||
973 | /* | |
974 | * Shared signal handlers imply shared VM. By way of the above, | |
975 | * thread groups also imply shared VM. Blocking this case allows | |
976 | * for various simplifications in other code. | |
977 | */ | |
978 | if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM)) | |
979 | return ERR_PTR(-EINVAL); | |
980 | ||
981 | retval = security_task_create(clone_flags); | |
982 | if (retval) | |
983 | goto fork_out; | |
984 | ||
985 | retval = -ENOMEM; | |
986 | p = dup_task_struct(current); | |
987 | if (!p) | |
988 | goto fork_out; | |
989 | ||
f7e8b616 SR |
990 | ftrace_graph_init_task(p); |
991 | ||
bea493a0 PZ |
992 | rt_mutex_init_task(p); |
993 | ||
d12c1a37 | 994 | #ifdef CONFIG_PROVE_LOCKING |
de30a2b3 IM |
995 | DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled); |
996 | DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled); | |
997 | #endif | |
1da177e4 | 998 | retval = -EAGAIN; |
3b11a1de | 999 | if (atomic_read(&p->real_cred->user->processes) >= |
1da177e4 LT |
1000 | p->signal->rlim[RLIMIT_NPROC].rlim_cur) { |
1001 | if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) && | |
18b6e041 | 1002 | p->real_cred->user != INIT_USER) |
1da177e4 LT |
1003 | goto bad_fork_free; |
1004 | } | |
1005 | ||
f1752eec DH |
1006 | retval = copy_creds(p, clone_flags); |
1007 | if (retval < 0) | |
1008 | goto bad_fork_free; | |
1da177e4 LT |
1009 | |
1010 | /* | |
1011 | * If multiple threads are within copy_process(), then this check | |
1012 | * triggers too late. This doesn't hurt, the check is only there | |
1013 | * to stop root fork bombs. | |
1014 | */ | |
04ec93fe | 1015 | retval = -EAGAIN; |
1da177e4 LT |
1016 | if (nr_threads >= max_threads) |
1017 | goto bad_fork_cleanup_count; | |
1018 | ||
a1261f54 | 1019 | if (!try_module_get(task_thread_info(p)->exec_domain->module)) |
1da177e4 LT |
1020 | goto bad_fork_cleanup_count; |
1021 | ||
1022 | if (p->binfmt && !try_module_get(p->binfmt->module)) | |
1023 | goto bad_fork_cleanup_put_domain; | |
1024 | ||
1025 | p->did_exec = 0; | |
ca74e92b | 1026 | delayacct_tsk_init(p); /* Must remain after dup_task_struct() */ |
1da177e4 | 1027 | copy_flags(clone_flags, p); |
1da177e4 LT |
1028 | INIT_LIST_HEAD(&p->children); |
1029 | INIT_LIST_HEAD(&p->sibling); | |
f41d911f | 1030 | rcu_copy_process(p); |
1da177e4 LT |
1031 | p->vfork_done = NULL; |
1032 | spin_lock_init(&p->alloc_lock); | |
1da177e4 | 1033 | |
1da177e4 LT |
1034 | init_sigpending(&p->pending); |
1035 | ||
1036 | p->utime = cputime_zero; | |
1037 | p->stime = cputime_zero; | |
9ac52315 | 1038 | p->gtime = cputime_zero; |
c66f08be MN |
1039 | p->utimescaled = cputime_zero; |
1040 | p->stimescaled = cputime_zero; | |
73a2bcb0 | 1041 | p->prev_utime = cputime_zero; |
9301899b | 1042 | p->prev_stime = cputime_zero; |
172ba844 | 1043 | |
6976675d AV |
1044 | p->default_timer_slack_ns = current->timer_slack_ns; |
1045 | ||
5995477a | 1046 | task_io_accounting_init(&p->ioac); |
1da177e4 LT |
1047 | acct_clear_integrals(p); |
1048 | ||
f06febc9 | 1049 | posix_cpu_timers_init(p); |
1da177e4 LT |
1050 | |
1051 | p->lock_depth = -1; /* -1 = no lock */ | |
1052 | do_posix_clock_monotonic_gettime(&p->start_time); | |
924b42d5 TJ |
1053 | p->real_start_time = p->start_time; |
1054 | monotonic_to_bootbased(&p->real_start_time); | |
1da177e4 | 1055 | p->io_context = NULL; |
1da177e4 | 1056 | p->audit_context = NULL; |
b4f48b63 | 1057 | cgroup_fork(p); |
1da177e4 | 1058 | #ifdef CONFIG_NUMA |
846a16bf | 1059 | p->mempolicy = mpol_dup(p->mempolicy); |
1da177e4 LT |
1060 | if (IS_ERR(p->mempolicy)) { |
1061 | retval = PTR_ERR(p->mempolicy); | |
1062 | p->mempolicy = NULL; | |
b4f48b63 | 1063 | goto bad_fork_cleanup_cgroup; |
1da177e4 | 1064 | } |
c61afb18 | 1065 | mpol_fix_fork_child_flag(p); |
1da177e4 | 1066 | #endif |
de30a2b3 IM |
1067 | #ifdef CONFIG_TRACE_IRQFLAGS |
1068 | p->irq_events = 0; | |
b36e4758 RK |
1069 | #ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW |
1070 | p->hardirqs_enabled = 1; | |
1071 | #else | |
de30a2b3 | 1072 | p->hardirqs_enabled = 0; |
b36e4758 | 1073 | #endif |
de30a2b3 IM |
1074 | p->hardirq_enable_ip = 0; |
1075 | p->hardirq_enable_event = 0; | |
1076 | p->hardirq_disable_ip = _THIS_IP_; | |
1077 | p->hardirq_disable_event = 0; | |
1078 | p->softirqs_enabled = 1; | |
1079 | p->softirq_enable_ip = _THIS_IP_; | |
1080 | p->softirq_enable_event = 0; | |
1081 | p->softirq_disable_ip = 0; | |
1082 | p->softirq_disable_event = 0; | |
1083 | p->hardirq_context = 0; | |
1084 | p->softirq_context = 0; | |
1085 | #endif | |
fbb9ce95 IM |
1086 | #ifdef CONFIG_LOCKDEP |
1087 | p->lockdep_depth = 0; /* no locks held yet */ | |
1088 | p->curr_chain_key = 0; | |
1089 | p->lockdep_recursion = 0; | |
1090 | #endif | |
1da177e4 | 1091 | |
408894ee IM |
1092 | #ifdef CONFIG_DEBUG_MUTEXES |
1093 | p->blocked_on = NULL; /* not blocked yet */ | |
1094 | #endif | |
0f481406 MM |
1095 | |
1096 | p->bts = NULL; | |
408894ee | 1097 | |
3c90e6e9 SV |
1098 | /* Perform scheduler related setup. Assign this task to a CPU. */ |
1099 | sched_fork(p, clone_flags); | |
6ab423e0 | 1100 | |
cdd6c482 | 1101 | retval = perf_event_init_task(p); |
6ab423e0 PZ |
1102 | if (retval) |
1103 | goto bad_fork_cleanup_policy; | |
3c90e6e9 | 1104 | |
1da177e4 | 1105 | if ((retval = audit_alloc(p))) |
f1752eec | 1106 | goto bad_fork_cleanup_policy; |
1da177e4 LT |
1107 | /* copy all the process information */ |
1108 | if ((retval = copy_semundo(clone_flags, p))) | |
1109 | goto bad_fork_cleanup_audit; | |
1110 | if ((retval = copy_files(clone_flags, p))) | |
1111 | goto bad_fork_cleanup_semundo; | |
1112 | if ((retval = copy_fs(clone_flags, p))) | |
1113 | goto bad_fork_cleanup_files; | |
1114 | if ((retval = copy_sighand(clone_flags, p))) | |
1115 | goto bad_fork_cleanup_fs; | |
1116 | if ((retval = copy_signal(clone_flags, p))) | |
1117 | goto bad_fork_cleanup_sighand; | |
1118 | if ((retval = copy_mm(clone_flags, p))) | |
1119 | goto bad_fork_cleanup_signal; | |
ab516013 | 1120 | if ((retval = copy_namespaces(clone_flags, p))) |
d84f4f99 | 1121 | goto bad_fork_cleanup_mm; |
fadad878 | 1122 | if ((retval = copy_io(clone_flags, p))) |
fd0928df | 1123 | goto bad_fork_cleanup_namespaces; |
6f2c55b8 | 1124 | retval = copy_thread(clone_flags, stack_start, stack_size, p, regs); |
1da177e4 | 1125 | if (retval) |
fd0928df | 1126 | goto bad_fork_cleanup_io; |
1da177e4 | 1127 | |
425fb2b4 PE |
1128 | if (pid != &init_struct_pid) { |
1129 | retval = -ENOMEM; | |
61bce0f1 | 1130 | pid = alloc_pid(p->nsproxy->pid_ns); |
425fb2b4 | 1131 | if (!pid) |
fd0928df | 1132 | goto bad_fork_cleanup_io; |
6f4e6433 PE |
1133 | |
1134 | if (clone_flags & CLONE_NEWPID) { | |
61bce0f1 | 1135 | retval = pid_ns_prepare_proc(p->nsproxy->pid_ns); |
6f4e6433 PE |
1136 | if (retval < 0) |
1137 | goto bad_fork_free_pid; | |
1138 | } | |
425fb2b4 PE |
1139 | } |
1140 | ||
1141 | p->pid = pid_nr(pid); | |
1142 | p->tgid = p->pid; | |
1143 | if (clone_flags & CLONE_THREAD) | |
1144 | p->tgid = current->tgid; | |
1145 | ||
e885dcde SH |
1146 | if (current->nsproxy != p->nsproxy) { |
1147 | retval = ns_cgroup_clone(p, pid); | |
1148 | if (retval) | |
f7e8b616 | 1149 | goto bad_fork_free_pid; |
e885dcde SH |
1150 | } |
1151 | ||
1da177e4 LT |
1152 | p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL; |
1153 | /* | |
1154 | * Clear TID on mm_release()? | |
1155 | */ | |
1156 | p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL; | |
42b2dd0a | 1157 | #ifdef CONFIG_FUTEX |
8f17d3a5 IM |
1158 | p->robust_list = NULL; |
1159 | #ifdef CONFIG_COMPAT | |
1160 | p->compat_robust_list = NULL; | |
1161 | #endif | |
c87e2837 IM |
1162 | INIT_LIST_HEAD(&p->pi_state_list); |
1163 | p->pi_state_cache = NULL; | |
42b2dd0a | 1164 | #endif |
f9a3879a GM |
1165 | /* |
1166 | * sigaltstack should be cleared when sharing the same VM | |
1167 | */ | |
1168 | if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM) | |
1169 | p->sas_ss_sp = p->sas_ss_size = 0; | |
1170 | ||
1da177e4 LT |
1171 | /* |
1172 | * Syscall tracing should be turned off in the child regardless | |
1173 | * of CLONE_PTRACE. | |
1174 | */ | |
1175 | clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE); | |
ed75e8d5 LV |
1176 | #ifdef TIF_SYSCALL_EMU |
1177 | clear_tsk_thread_flag(p, TIF_SYSCALL_EMU); | |
1178 | #endif | |
9745512c | 1179 | clear_all_latency_tracing(p); |
1da177e4 | 1180 | |
1da177e4 LT |
1181 | /* ok, now we should be set up.. */ |
1182 | p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL); | |
1183 | p->pdeath_signal = 0; | |
1184 | p->exit_state = 0; | |
1185 | ||
1da177e4 LT |
1186 | /* |
1187 | * Ok, make it visible to the rest of the system. | |
1188 | * We dont wake it up yet. | |
1189 | */ | |
1190 | p->group_leader = p; | |
47e65328 | 1191 | INIT_LIST_HEAD(&p->thread_group); |
1da177e4 | 1192 | |
b4f48b63 PM |
1193 | /* Now that the task is set up, run cgroup callbacks if |
1194 | * necessary. We need to run them before the task is visible | |
1195 | * on the tasklist. */ | |
1196 | cgroup_fork_callbacks(p); | |
1197 | cgroup_callbacks_done = 1; | |
1198 | ||
1da177e4 LT |
1199 | /* Need tasklist lock for parent etc handling! */ |
1200 | write_lock_irq(&tasklist_lock); | |
1201 | ||
1202 | /* | |
476d139c NP |
1203 | * The task hasn't been attached yet, so its cpus_allowed mask will |
1204 | * not be changed, nor will its assigned CPU. | |
1205 | * | |
1206 | * The cpus_allowed mask of the parent may have changed after it was | |
1207 | * copied first time - so re-copy it here, then check the child's CPU | |
1208 | * to ensure it is on a valid CPU (and if not, just force it back to | |
1209 | * parent's CPU). This avoids alot of nasty races. | |
1da177e4 LT |
1210 | */ |
1211 | p->cpus_allowed = current->cpus_allowed; | |
6f505b16 | 1212 | p->rt.nr_cpus_allowed = current->rt.nr_cpus_allowed; |
26ff6ad9 SV |
1213 | if (unlikely(!cpu_isset(task_cpu(p), p->cpus_allowed) || |
1214 | !cpu_online(task_cpu(p)))) | |
476d139c | 1215 | set_task_cpu(p, smp_processor_id()); |
1da177e4 | 1216 | |
1da177e4 | 1217 | /* CLONE_PARENT re-uses the old parent */ |
2d5516cb | 1218 | if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) { |
1da177e4 | 1219 | p->real_parent = current->real_parent; |
2d5516cb ON |
1220 | p->parent_exec_id = current->parent_exec_id; |
1221 | } else { | |
1da177e4 | 1222 | p->real_parent = current; |
2d5516cb ON |
1223 | p->parent_exec_id = current->self_exec_id; |
1224 | } | |
1da177e4 | 1225 | |
3f17da69 | 1226 | spin_lock(¤t->sighand->siglock); |
4a2c7a78 ON |
1227 | |
1228 | /* | |
1229 | * Process group and session signals need to be delivered to just the | |
1230 | * parent before the fork or both the parent and the child after the | |
1231 | * fork. Restart if a signal comes in before we add the new process to | |
1232 | * it's process group. | |
1233 | * A fatal signal pending means that current will exit, so the new | |
1234 | * thread can't slip out of an OOM kill (or normal SIGKILL). | |
1235 | */ | |
23ff4440 | 1236 | recalc_sigpending(); |
4a2c7a78 ON |
1237 | if (signal_pending(current)) { |
1238 | spin_unlock(¤t->sighand->siglock); | |
1239 | write_unlock_irq(&tasklist_lock); | |
1240 | retval = -ERESTARTNOINTR; | |
f7e8b616 | 1241 | goto bad_fork_free_pid; |
4a2c7a78 ON |
1242 | } |
1243 | ||
1da177e4 | 1244 | if (clone_flags & CLONE_THREAD) { |
4ab6c083 ON |
1245 | atomic_inc(¤t->signal->count); |
1246 | atomic_inc(¤t->signal->live); | |
1da177e4 | 1247 | p->group_leader = current->group_leader; |
47e65328 | 1248 | list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group); |
1da177e4 LT |
1249 | } |
1250 | ||
73b9ebfe | 1251 | if (likely(p->pid)) { |
f470021a | 1252 | list_add_tail(&p->sibling, &p->real_parent->children); |
09a05394 | 1253 | tracehook_finish_clone(p, clone_flags, trace); |
73b9ebfe ON |
1254 | |
1255 | if (thread_group_leader(p)) { | |
5cd17569 | 1256 | if (clone_flags & CLONE_NEWPID) |
30e49c26 | 1257 | p->nsproxy->pid_ns->child_reaper = p; |
73b9ebfe | 1258 | |
fea9d175 | 1259 | p->signal->leader_pid = pid; |
9c9f4ded AC |
1260 | tty_kref_put(p->signal->tty); |
1261 | p->signal->tty = tty_kref_get(current->signal->tty); | |
5cd17569 EB |
1262 | attach_pid(p, PIDTYPE_PGID, task_pgrp(current)); |
1263 | attach_pid(p, PIDTYPE_SID, task_session(current)); | |
5e85d4ab | 1264 | list_add_tail_rcu(&p->tasks, &init_task.tasks); |
1da177e4 | 1265 | __get_cpu_var(process_counts)++; |
73b9ebfe | 1266 | } |
85868995 | 1267 | attach_pid(p, PIDTYPE_PID, pid); |
73b9ebfe | 1268 | nr_threads++; |
1da177e4 LT |
1269 | } |
1270 | ||
1da177e4 | 1271 | total_forks++; |
3f17da69 | 1272 | spin_unlock(¤t->sighand->siglock); |
1da177e4 | 1273 | write_unlock_irq(&tasklist_lock); |
c13cf856 | 1274 | proc_fork_connector(p); |
817929ec | 1275 | cgroup_post_fork(p); |
cdd6c482 | 1276 | perf_event_fork(p); |
1da177e4 LT |
1277 | return p; |
1278 | ||
425fb2b4 PE |
1279 | bad_fork_free_pid: |
1280 | if (pid != &init_struct_pid) | |
1281 | free_pid(pid); | |
fd0928df JA |
1282 | bad_fork_cleanup_io: |
1283 | put_io_context(p->io_context); | |
ab516013 | 1284 | bad_fork_cleanup_namespaces: |
444f378b | 1285 | exit_task_namespaces(p); |
1da177e4 LT |
1286 | bad_fork_cleanup_mm: |
1287 | if (p->mm) | |
1288 | mmput(p->mm); | |
1289 | bad_fork_cleanup_signal: | |
4ab6c083 ON |
1290 | if (!(clone_flags & CLONE_THREAD)) |
1291 | __cleanup_signal(p->signal); | |
1da177e4 | 1292 | bad_fork_cleanup_sighand: |
a7e5328a | 1293 | __cleanup_sighand(p->sighand); |
1da177e4 LT |
1294 | bad_fork_cleanup_fs: |
1295 | exit_fs(p); /* blocking */ | |
1296 | bad_fork_cleanup_files: | |
1297 | exit_files(p); /* blocking */ | |
1298 | bad_fork_cleanup_semundo: | |
1299 | exit_sem(p); | |
1300 | bad_fork_cleanup_audit: | |
1301 | audit_free(p); | |
1da177e4 | 1302 | bad_fork_cleanup_policy: |
cdd6c482 | 1303 | perf_event_free_task(p); |
1da177e4 | 1304 | #ifdef CONFIG_NUMA |
f0be3d32 | 1305 | mpol_put(p->mempolicy); |
b4f48b63 | 1306 | bad_fork_cleanup_cgroup: |
1da177e4 | 1307 | #endif |
b4f48b63 | 1308 | cgroup_exit(p, cgroup_callbacks_done); |
35df17c5 | 1309 | delayacct_tsk_free(p); |
1da177e4 LT |
1310 | if (p->binfmt) |
1311 | module_put(p->binfmt->module); | |
1312 | bad_fork_cleanup_put_domain: | |
a1261f54 | 1313 | module_put(task_thread_info(p)->exec_domain->module); |
1da177e4 | 1314 | bad_fork_cleanup_count: |
d84f4f99 | 1315 | atomic_dec(&p->cred->user->processes); |
e0e81739 | 1316 | exit_creds(p); |
1da177e4 LT |
1317 | bad_fork_free: |
1318 | free_task(p); | |
fe7d37d1 ON |
1319 | fork_out: |
1320 | return ERR_PTR(retval); | |
1da177e4 LT |
1321 | } |
1322 | ||
6b2fb3c6 | 1323 | noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs) |
1da177e4 LT |
1324 | { |
1325 | memset(regs, 0, sizeof(struct pt_regs)); | |
1326 | return regs; | |
1327 | } | |
1328 | ||
9abcf40b | 1329 | struct task_struct * __cpuinit fork_idle(int cpu) |
1da177e4 | 1330 | { |
36c8b586 | 1331 | struct task_struct *task; |
1da177e4 LT |
1332 | struct pt_regs regs; |
1333 | ||
30e49c26 | 1334 | task = copy_process(CLONE_VM, 0, idle_regs(®s), 0, NULL, |
09a05394 | 1335 | &init_struct_pid, 0); |
753ca4f3 AM |
1336 | if (!IS_ERR(task)) |
1337 | init_idle(task, cpu); | |
73b9ebfe | 1338 | |
1da177e4 LT |
1339 | return task; |
1340 | } | |
1341 | ||
1da177e4 LT |
1342 | /* |
1343 | * Ok, this is the main fork-routine. | |
1344 | * | |
1345 | * It copies the process, and if successful kick-starts | |
1346 | * it and waits for it to finish using the VM if required. | |
1347 | */ | |
1348 | long do_fork(unsigned long clone_flags, | |
1349 | unsigned long stack_start, | |
1350 | struct pt_regs *regs, | |
1351 | unsigned long stack_size, | |
1352 | int __user *parent_tidptr, | |
1353 | int __user *child_tidptr) | |
1354 | { | |
1355 | struct task_struct *p; | |
1356 | int trace = 0; | |
92476d7f | 1357 | long nr; |
1da177e4 | 1358 | |
18b6e041 SH |
1359 | /* |
1360 | * Do some preliminary argument and permissions checking before we | |
1361 | * actually start allocating stuff | |
1362 | */ | |
1363 | if (clone_flags & CLONE_NEWUSER) { | |
1364 | if (clone_flags & CLONE_THREAD) | |
1365 | return -EINVAL; | |
1366 | /* hopefully this check will go away when userns support is | |
1367 | * complete | |
1368 | */ | |
7657d904 SH |
1369 | if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) || |
1370 | !capable(CAP_SETGID)) | |
18b6e041 SH |
1371 | return -EPERM; |
1372 | } | |
1373 | ||
bdff746a AM |
1374 | /* |
1375 | * We hope to recycle these flags after 2.6.26 | |
1376 | */ | |
1377 | if (unlikely(clone_flags & CLONE_STOPPED)) { | |
1378 | static int __read_mostly count = 100; | |
1379 | ||
1380 | if (count > 0 && printk_ratelimit()) { | |
1381 | char comm[TASK_COMM_LEN]; | |
1382 | ||
1383 | count--; | |
1384 | printk(KERN_INFO "fork(): process `%s' used deprecated " | |
1385 | "clone flags 0x%lx\n", | |
1386 | get_task_comm(comm, current), | |
1387 | clone_flags & CLONE_STOPPED); | |
1388 | } | |
1389 | } | |
1390 | ||
09a05394 RM |
1391 | /* |
1392 | * When called from kernel_thread, don't do user tracing stuff. | |
1393 | */ | |
1394 | if (likely(user_mode(regs))) | |
1395 | trace = tracehook_prepare_clone(clone_flags); | |
1da177e4 | 1396 | |
a6f5e063 | 1397 | p = copy_process(clone_flags, stack_start, regs, stack_size, |
09a05394 | 1398 | child_tidptr, NULL, trace); |
1da177e4 LT |
1399 | /* |
1400 | * Do this prior waking up the new thread - the thread pointer | |
1401 | * might get invalid after that point, if the thread exits quickly. | |
1402 | */ | |
1403 | if (!IS_ERR(p)) { | |
1404 | struct completion vfork; | |
1405 | ||
0a16b607 MD |
1406 | trace_sched_process_fork(current, p); |
1407 | ||
6c5f3e7b | 1408 | nr = task_pid_vnr(p); |
30e49c26 PE |
1409 | |
1410 | if (clone_flags & CLONE_PARENT_SETTID) | |
1411 | put_user(nr, parent_tidptr); | |
a6f5e063 | 1412 | |
1da177e4 LT |
1413 | if (clone_flags & CLONE_VFORK) { |
1414 | p->vfork_done = &vfork; | |
1415 | init_completion(&vfork); | |
1416 | } | |
1417 | ||
a64e6494 | 1418 | audit_finish_fork(p); |
087eb437 | 1419 | tracehook_report_clone(regs, clone_flags, nr, p); |
09a05394 RM |
1420 | |
1421 | /* | |
1422 | * We set PF_STARTING at creation in case tracing wants to | |
1423 | * use this to distinguish a fully live task from one that | |
1424 | * hasn't gotten to tracehook_report_clone() yet. Now we | |
1425 | * clear it and set the child going. | |
1426 | */ | |
1427 | p->flags &= ~PF_STARTING; | |
1428 | ||
1429 | if (unlikely(clone_flags & CLONE_STOPPED)) { | |
1da177e4 LT |
1430 | /* |
1431 | * We'll start up with an immediate SIGSTOP. | |
1432 | */ | |
1433 | sigaddset(&p->pending.signal, SIGSTOP); | |
1434 | set_tsk_thread_flag(p, TIF_SIGPENDING); | |
d9ae90ac | 1435 | __set_task_state(p, TASK_STOPPED); |
09a05394 RM |
1436 | } else { |
1437 | wake_up_new_task(p, clone_flags); | |
1da177e4 LT |
1438 | } |
1439 | ||
09a05394 RM |
1440 | tracehook_report_clone_complete(trace, regs, |
1441 | clone_flags, nr, p); | |
1442 | ||
1da177e4 | 1443 | if (clone_flags & CLONE_VFORK) { |
ba96a0c8 | 1444 | freezer_do_not_count(); |
1da177e4 | 1445 | wait_for_completion(&vfork); |
ba96a0c8 | 1446 | freezer_count(); |
daded34b | 1447 | tracehook_report_vfork_done(p, nr); |
1da177e4 LT |
1448 | } |
1449 | } else { | |
92476d7f | 1450 | nr = PTR_ERR(p); |
1da177e4 | 1451 | } |
92476d7f | 1452 | return nr; |
1da177e4 LT |
1453 | } |
1454 | ||
5fd63b30 RT |
1455 | #ifndef ARCH_MIN_MMSTRUCT_ALIGN |
1456 | #define ARCH_MIN_MMSTRUCT_ALIGN 0 | |
1457 | #endif | |
1458 | ||
51cc5068 | 1459 | static void sighand_ctor(void *data) |
aa1757f9 ON |
1460 | { |
1461 | struct sighand_struct *sighand = data; | |
1462 | ||
a35afb83 | 1463 | spin_lock_init(&sighand->siglock); |
b8fceee1 | 1464 | init_waitqueue_head(&sighand->signalfd_wqh); |
aa1757f9 ON |
1465 | } |
1466 | ||
1da177e4 LT |
1467 | void __init proc_caches_init(void) |
1468 | { | |
1469 | sighand_cachep = kmem_cache_create("sighand_cache", | |
1470 | sizeof(struct sighand_struct), 0, | |
2dff4405 VN |
1471 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU| |
1472 | SLAB_NOTRACK, sighand_ctor); | |
1da177e4 LT |
1473 | signal_cachep = kmem_cache_create("signal_cache", |
1474 | sizeof(struct signal_struct), 0, | |
2dff4405 | 1475 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1476 | files_cachep = kmem_cache_create("files_cache", |
1da177e4 | 1477 | sizeof(struct files_struct), 0, |
2dff4405 | 1478 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
20c2df83 | 1479 | fs_cachep = kmem_cache_create("fs_cache", |
1da177e4 | 1480 | sizeof(struct fs_struct), 0, |
2dff4405 | 1481 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
1da177e4 | 1482 | mm_cachep = kmem_cache_create("mm_struct", |
5fd63b30 | 1483 | sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN, |
2dff4405 | 1484 | SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL); |
33e5d769 | 1485 | vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC); |
8feae131 | 1486 | mmap_init(); |
1da177e4 | 1487 | } |
cf2e340f | 1488 | |
cf2e340f JD |
1489 | /* |
1490 | * Check constraints on flags passed to the unshare system call and | |
1491 | * force unsharing of additional process context as appropriate. | |
1492 | */ | |
a39bc516 | 1493 | static void check_unshare_flags(unsigned long *flags_ptr) |
cf2e340f JD |
1494 | { |
1495 | /* | |
1496 | * If unsharing a thread from a thread group, must also | |
1497 | * unshare vm. | |
1498 | */ | |
1499 | if (*flags_ptr & CLONE_THREAD) | |
1500 | *flags_ptr |= CLONE_VM; | |
1501 | ||
1502 | /* | |
1503 | * If unsharing vm, must also unshare signal handlers. | |
1504 | */ | |
1505 | if (*flags_ptr & CLONE_VM) | |
1506 | *flags_ptr |= CLONE_SIGHAND; | |
1507 | ||
1508 | /* | |
1509 | * If unsharing signal handlers and the task was created | |
1510 | * using CLONE_THREAD, then must unshare the thread | |
1511 | */ | |
1512 | if ((*flags_ptr & CLONE_SIGHAND) && | |
1513 | (atomic_read(¤t->signal->count) > 1)) | |
1514 | *flags_ptr |= CLONE_THREAD; | |
1515 | ||
1516 | /* | |
1517 | * If unsharing namespace, must also unshare filesystem information. | |
1518 | */ | |
1519 | if (*flags_ptr & CLONE_NEWNS) | |
1520 | *flags_ptr |= CLONE_FS; | |
1521 | } | |
1522 | ||
1523 | /* | |
1524 | * Unsharing of tasks created with CLONE_THREAD is not supported yet | |
1525 | */ | |
1526 | static int unshare_thread(unsigned long unshare_flags) | |
1527 | { | |
1528 | if (unshare_flags & CLONE_THREAD) | |
1529 | return -EINVAL; | |
1530 | ||
1531 | return 0; | |
1532 | } | |
1533 | ||
1534 | /* | |
99d1419d | 1535 | * Unshare the filesystem structure if it is being shared |
cf2e340f JD |
1536 | */ |
1537 | static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp) | |
1538 | { | |
1539 | struct fs_struct *fs = current->fs; | |
1540 | ||
498052bb AV |
1541 | if (!(unshare_flags & CLONE_FS) || !fs) |
1542 | return 0; | |
1543 | ||
1544 | /* don't need lock here; in the worst case we'll do useless copy */ | |
1545 | if (fs->users == 1) | |
1546 | return 0; | |
1547 | ||
1548 | *new_fsp = copy_fs_struct(fs); | |
1549 | if (!*new_fsp) | |
1550 | return -ENOMEM; | |
cf2e340f JD |
1551 | |
1552 | return 0; | |
1553 | } | |
1554 | ||
cf2e340f | 1555 | /* |
dae3c5a0 | 1556 | * Unsharing of sighand is not supported yet |
cf2e340f JD |
1557 | */ |
1558 | static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp) | |
1559 | { | |
1560 | struct sighand_struct *sigh = current->sighand; | |
1561 | ||
dae3c5a0 | 1562 | if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1) |
cf2e340f JD |
1563 | return -EINVAL; |
1564 | else | |
1565 | return 0; | |
1566 | } | |
1567 | ||
1568 | /* | |
a0a7ec30 | 1569 | * Unshare vm if it is being shared |
cf2e340f JD |
1570 | */ |
1571 | static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp) | |
1572 | { | |
1573 | struct mm_struct *mm = current->mm; | |
1574 | ||
1575 | if ((unshare_flags & CLONE_VM) && | |
a0a7ec30 | 1576 | (mm && atomic_read(&mm->mm_users) > 1)) { |
2d61b867 | 1577 | return -EINVAL; |
a0a7ec30 | 1578 | } |
cf2e340f JD |
1579 | |
1580 | return 0; | |
cf2e340f JD |
1581 | } |
1582 | ||
1583 | /* | |
a016f338 | 1584 | * Unshare file descriptor table if it is being shared |
cf2e340f JD |
1585 | */ |
1586 | static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp) | |
1587 | { | |
1588 | struct files_struct *fd = current->files; | |
a016f338 | 1589 | int error = 0; |
cf2e340f JD |
1590 | |
1591 | if ((unshare_flags & CLONE_FILES) && | |
a016f338 JD |
1592 | (fd && atomic_read(&fd->count) > 1)) { |
1593 | *new_fdp = dup_fd(fd, &error); | |
1594 | if (!*new_fdp) | |
1595 | return error; | |
1596 | } | |
cf2e340f JD |
1597 | |
1598 | return 0; | |
1599 | } | |
1600 | ||
cf2e340f JD |
1601 | /* |
1602 | * unshare allows a process to 'unshare' part of the process | |
1603 | * context which was originally shared using clone. copy_* | |
1604 | * functions used by do_fork() cannot be used here directly | |
1605 | * because they modify an inactive task_struct that is being | |
1606 | * constructed. Here we are modifying the current, active, | |
1607 | * task_struct. | |
1608 | */ | |
6559eed8 | 1609 | SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags) |
cf2e340f JD |
1610 | { |
1611 | int err = 0; | |
1612 | struct fs_struct *fs, *new_fs = NULL; | |
dae3c5a0 | 1613 | struct sighand_struct *new_sigh = NULL; |
cf2e340f JD |
1614 | struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL; |
1615 | struct files_struct *fd, *new_fd = NULL; | |
cf7b708c | 1616 | struct nsproxy *new_nsproxy = NULL; |
9edff4ab | 1617 | int do_sysvsem = 0; |
cf2e340f JD |
1618 | |
1619 | check_unshare_flags(&unshare_flags); | |
1620 | ||
06f9d4f9 EB |
1621 | /* Return -EINVAL for all unsupported flags */ |
1622 | err = -EINVAL; | |
1623 | if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND| | |
25b21cb2 | 1624 | CLONE_VM|CLONE_FILES|CLONE_SYSVSEM| |
18b6e041 | 1625 | CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET)) |
06f9d4f9 EB |
1626 | goto bad_unshare_out; |
1627 | ||
6013f67f MS |
1628 | /* |
1629 | * CLONE_NEWIPC must also detach from the undolist: after switching | |
1630 | * to a new ipc namespace, the semaphore arrays from the old | |
1631 | * namespace are unreachable. | |
1632 | */ | |
1633 | if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM)) | |
9edff4ab | 1634 | do_sysvsem = 1; |
cf2e340f JD |
1635 | if ((err = unshare_thread(unshare_flags))) |
1636 | goto bad_unshare_out; | |
1637 | if ((err = unshare_fs(unshare_flags, &new_fs))) | |
1638 | goto bad_unshare_cleanup_thread; | |
cf2e340f | 1639 | if ((err = unshare_sighand(unshare_flags, &new_sigh))) |
e3222c4e | 1640 | goto bad_unshare_cleanup_fs; |
cf2e340f JD |
1641 | if ((err = unshare_vm(unshare_flags, &new_mm))) |
1642 | goto bad_unshare_cleanup_sigh; | |
1643 | if ((err = unshare_fd(unshare_flags, &new_fd))) | |
1644 | goto bad_unshare_cleanup_vm; | |
e3222c4e BP |
1645 | if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, |
1646 | new_fs))) | |
9edff4ab | 1647 | goto bad_unshare_cleanup_fd; |
c0b2fc31 | 1648 | |
9edff4ab MS |
1649 | if (new_fs || new_mm || new_fd || do_sysvsem || new_nsproxy) { |
1650 | if (do_sysvsem) { | |
1651 | /* | |
1652 | * CLONE_SYSVSEM is equivalent to sys_exit(). | |
1653 | */ | |
1654 | exit_sem(current); | |
1655 | } | |
ab516013 | 1656 | |
c0b2fc31 | 1657 | if (new_nsproxy) { |
cf7b708c PE |
1658 | switch_task_namespaces(current, new_nsproxy); |
1659 | new_nsproxy = NULL; | |
c0b2fc31 | 1660 | } |
cf2e340f | 1661 | |
cf7b708c PE |
1662 | task_lock(current); |
1663 | ||
cf2e340f JD |
1664 | if (new_fs) { |
1665 | fs = current->fs; | |
498052bb | 1666 | write_lock(&fs->lock); |
cf2e340f | 1667 | current->fs = new_fs; |
498052bb AV |
1668 | if (--fs->users) |
1669 | new_fs = NULL; | |
1670 | else | |
1671 | new_fs = fs; | |
1672 | write_unlock(&fs->lock); | |
cf2e340f JD |
1673 | } |
1674 | ||
cf2e340f JD |
1675 | if (new_mm) { |
1676 | mm = current->mm; | |
1677 | active_mm = current->active_mm; | |
1678 | current->mm = new_mm; | |
1679 | current->active_mm = new_mm; | |
1680 | activate_mm(active_mm, new_mm); | |
1681 | new_mm = mm; | |
1682 | } | |
1683 | ||
1684 | if (new_fd) { | |
1685 | fd = current->files; | |
1686 | current->files = new_fd; | |
1687 | new_fd = fd; | |
1688 | } | |
1689 | ||
1690 | task_unlock(current); | |
1691 | } | |
1692 | ||
c0b2fc31 | 1693 | if (new_nsproxy) |
444f378b | 1694 | put_nsproxy(new_nsproxy); |
c0b2fc31 | 1695 | |
cf2e340f JD |
1696 | bad_unshare_cleanup_fd: |
1697 | if (new_fd) | |
1698 | put_files_struct(new_fd); | |
1699 | ||
1700 | bad_unshare_cleanup_vm: | |
1701 | if (new_mm) | |
1702 | mmput(new_mm); | |
1703 | ||
1704 | bad_unshare_cleanup_sigh: | |
1705 | if (new_sigh) | |
1706 | if (atomic_dec_and_test(&new_sigh->count)) | |
1707 | kmem_cache_free(sighand_cachep, new_sigh); | |
1708 | ||
cf2e340f JD |
1709 | bad_unshare_cleanup_fs: |
1710 | if (new_fs) | |
498052bb | 1711 | free_fs_struct(new_fs); |
cf2e340f JD |
1712 | |
1713 | bad_unshare_cleanup_thread: | |
1714 | bad_unshare_out: | |
1715 | return err; | |
1716 | } | |
3b125388 AV |
1717 | |
1718 | /* | |
1719 | * Helper to unshare the files of the current task. | |
1720 | * We don't want to expose copy_files internals to | |
1721 | * the exec layer of the kernel. | |
1722 | */ | |
1723 | ||
1724 | int unshare_files(struct files_struct **displaced) | |
1725 | { | |
1726 | struct task_struct *task = current; | |
50704516 | 1727 | struct files_struct *copy = NULL; |
3b125388 AV |
1728 | int error; |
1729 | ||
1730 | error = unshare_fd(CLONE_FILES, ©); | |
1731 | if (error || !copy) { | |
1732 | *displaced = NULL; | |
1733 | return error; | |
1734 | } | |
1735 | *displaced = task->files; | |
1736 | task_lock(task); | |
1737 | task->files = copy; | |
1738 | task_unlock(task); | |
1739 | return 0; | |
1740 | } |