blob: 027c80e5162fa750db320b876d5d288342c3fb96 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/slab.h>
15#include <linux/init.h>
16#include <linux/unistd.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/module.h>
18#include <linux/vmalloc.h>
19#include <linux/completion.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/personality.h>
21#include <linux/mempolicy.h>
22#include <linux/sem.h>
23#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040024#include <linux/fdtable.h>
Jens Axboeda9cbc82008-06-30 20:42:08 +020025#include <linux/iocontext.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/key.h>
27#include <linux/binfmts.h>
28#include <linux/mman.h>
Andrea Arcangelicddb8a52008-07-28 15:46:29 -070029#include <linux/mmu_notifier.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include <linux/fs.h>
Serge E. Hallynab516012006-10-02 02:18:06 -070031#include <linux/nsproxy.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080032#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <linux/cpu.h>
Paul Menageb4f48b62007-10-18 23:39:33 -070034#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035#include <linux/security.h>
Mel Gormana1e78772008-07-23 21:27:23 -070036#include <linux/hugetlb.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#include <linux/swap.h>
38#include <linux/syscalls.h>
39#include <linux/jiffies.h>
Roland McGrath09a05392008-07-25 19:45:47 -070040#include <linux/tracehook.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#include <linux/futex.h>
Linus Torvalds8141c7f2008-11-15 10:20:36 -080042#include <linux/compat.h>
Andrew Morton7c3ab732006-12-10 02:19:19 -080043#include <linux/task_io_accounting_ops.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070044#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070045#include <linux/ptrace.h>
46#include <linux/mount.h>
47#include <linux/audit.h>
Pavel Emelianov78fb7462008-02-07 00:13:51 -080048#include <linux/memcontrol.h>
Frederic Weisbeckerf201ae22008-11-23 06:22:56 +010049#include <linux/ftrace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070050#include <linux/profile.h>
51#include <linux/rmap.h>
Hugh Dickinsf8af4da2009-09-21 17:01:57 -070052#include <linux/ksm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070053#include <linux/acct.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070054#include <linux/tsacct_kern.h>
Matt Helsley9f460802005-11-07 00:59:16 -080055#include <linux/cn_proc.h>
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -070056#include <linux/freezer.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070057#include <linux/delayacct.h>
Shailabh Nagarad4ecbc2006-07-14 00:24:44 -070058#include <linux/taskstats_kern.h>
Arjan van de Ven0a425402006-09-26 10:52:38 +020059#include <linux/random.h>
Miloslav Trmac522ed772007-07-15 23:40:56 -070060#include <linux/tty.h>
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -070061#include <linux/proc_fs.h>
Jens Axboefd0928d2008-01-24 08:52:45 +010062#include <linux/blkdev.h>
Al Viro5ad4e532009-03-29 19:50:06 -040063#include <linux/fs_struct.h>
Eric Sandeen7c9f8862008-04-22 16:38:23 -050064#include <linux/magic.h>
Ingo Molnarcdd6c482009-09-21 12:02:48 +020065#include <linux/perf_event.h>
Stanislaw Gruszka42c4ab42009-07-29 12:15:26 +020066#include <linux/posix-timers.h>
Avi Kivity8e7cac72009-11-29 16:34:48 +020067#include <linux/user-return-notifier.h>
Ying Han3d5992d2010-10-26 14:21:23 -070068#include <linux/oom.h>
Andrea Arcangeliba761492011-01-13 15:46:58 -080069#include <linux/khugepaged.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070070
71#include <asm/pgtable.h>
72#include <asm/pgalloc.h>
73#include <asm/uaccess.h>
74#include <asm/mmu_context.h>
75#include <asm/cacheflush.h>
76#include <asm/tlbflush.h>
77
Steven Rostedtad8d75f2009-04-14 19:39:12 -040078#include <trace/events/sched.h>
79
Linus Torvalds1da177e2005-04-16 15:20:36 -070080/*
81 * Protected counters by write_lock_irq(&tasklist_lock)
82 */
83unsigned long total_forks; /* Handle normal Linux uptimes. */
84int nr_threads; /* The idle threads do not count.. */
85
86int max_threads; /* tunable limit on nr_threads */
87
88DEFINE_PER_CPU(unsigned long, process_counts) = 0;
89
Christoph Hellwigc59923a2006-07-10 04:45:40 -070090__cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
Paul E. McKenneydb1466b2010-03-03 07:46:56 -080091
92#ifdef CONFIG_PROVE_RCU
93int lockdep_tasklist_lock_is_held(void)
94{
95 return lockdep_is_held(&tasklist_lock);
96}
97EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
98#endif /* #ifdef CONFIG_PROVE_RCU */
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
100int nr_processes(void)
101{
102 int cpu;
103 int total = 0;
104
Ian Campbell1d510752009-11-03 10:11:14 +0000105 for_each_possible_cpu(cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 total += per_cpu(process_counts, cpu);
107
108 return total;
109}
110
111#ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
112# define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
113# define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
Christoph Lametere18b8902006-12-06 20:33:20 -0800114static struct kmem_cache *task_struct_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115#endif
116
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700117#ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
118static inline struct thread_info *alloc_thread_info(struct task_struct *tsk)
119{
120#ifdef CONFIG_DEBUG_STACK_USAGE
121 gfp_t mask = GFP_KERNEL | __GFP_ZERO;
122#else
123 gfp_t mask = GFP_KERNEL;
124#endif
125 return (struct thread_info *)__get_free_pages(mask, THREAD_SIZE_ORDER);
126}
127
128static inline void free_thread_info(struct thread_info *ti)
129{
130 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
131}
132#endif
133
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134/* SLAB cache for signal_struct structures (tsk->signal) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800135static struct kmem_cache *signal_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136
137/* SLAB cache for sighand_struct structures (tsk->sighand) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800138struct kmem_cache *sighand_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139
140/* SLAB cache for files_struct structures (tsk->files) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800141struct kmem_cache *files_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142
143/* SLAB cache for fs_struct structures (tsk->fs) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800144struct kmem_cache *fs_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145
146/* SLAB cache for vm_area_struct structures */
Christoph Lametere18b8902006-12-06 20:33:20 -0800147struct kmem_cache *vm_area_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148
149/* SLAB cache for mm_struct structures (tsk->mm) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800150static struct kmem_cache *mm_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700152static void account_kernel_stack(struct thread_info *ti, int account)
153{
154 struct zone *zone = page_zone(virt_to_page(ti));
155
156 mod_zone_page_state(zone, NR_KERNEL_STACK, account);
157}
158
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159void free_task(struct task_struct *tsk)
160{
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700161 prop_local_destroy_single(&tsk->dirties);
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700162 account_kernel_stack(tsk->stack, -1);
Roman Zippelf7e42172007-05-09 02:35:17 -0700163 free_thread_info(tsk->stack);
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700164 rt_mutex_debug_task_free(tsk);
Frederic Weisbeckerfb526072008-11-25 21:07:04 +0100165 ftrace_graph_exit_task(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 free_task_struct(tsk);
167}
168EXPORT_SYMBOL(free_task);
169
Oleg Nesterovea6d2902010-05-26 14:43:16 -0700170static inline void free_signal_struct(struct signal_struct *sig)
171{
Oleg Nesterov97101eb2010-05-26 14:43:20 -0700172 taskstats_tgid_free(sig);
Mike Galbraith1c5354d2011-01-05 11:16:04 +0100173 sched_autogroup_exit(sig);
Oleg Nesterovea6d2902010-05-26 14:43:16 -0700174 kmem_cache_free(signal_cachep, sig);
175}
176
177static inline void put_signal_struct(struct signal_struct *sig)
178{
Mike Galbraith1c5354d2011-01-05 11:16:04 +0100179 if (atomic_dec_and_test(&sig->sigcnt))
Oleg Nesterovea6d2902010-05-26 14:43:16 -0700180 free_signal_struct(sig);
181}
182
Andrew Morton158d9eb2006-03-31 02:31:34 -0800183void __put_task_struct(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184{
Eugene Teo270f7222007-10-18 23:40:38 -0700185 WARN_ON(!tsk->exit_state);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 WARN_ON(atomic_read(&tsk->usage));
187 WARN_ON(tsk == current);
188
David Howellse0e81732009-09-02 09:13:40 +0100189 exit_creds(tsk);
Shailabh Nagar35df17c2006-08-31 21:27:38 -0700190 delayacct_tsk_free(tsk);
Oleg Nesterovea6d2902010-05-26 14:43:16 -0700191 put_signal_struct(tsk->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192
193 if (!profile_handoff_task(tsk))
194 free_task(tsk);
195}
196
Suresh Siddha2adee9b2008-04-16 10:25:35 +0200197/*
198 * macro override instead of weak attribute alias, to workaround
199 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
200 */
201#ifndef arch_task_cache_init
202#define arch_task_cache_init()
203#endif
Suresh Siddha61c46282008-03-10 15:28:04 -0700204
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205void __init fork_init(unsigned long mempages)
206{
207#ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
208#ifndef ARCH_MIN_TASKALIGN
209#define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
210#endif
211 /* create a slab on which task_structs can be allocated */
212 task_struct_cachep =
213 kmem_cache_create("task_struct", sizeof(struct task_struct),
Vegard Nossum2dff4402008-05-31 15:56:17 +0200214 ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215#endif
216
Suresh Siddha61c46282008-03-10 15:28:04 -0700217 /* do the arch specific task caches init */
218 arch_task_cache_init();
219
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 /*
221 * The default maximum number of threads is set to a safe
222 * value: the thread structures can take up at most half
223 * of memory.
224 */
225 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
226
227 /*
228 * we need to allow at least 20 threads to boot a system
229 */
230 if(max_threads < 20)
231 max_threads = 20;
232
233 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
234 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
235 init_task.signal->rlim[RLIMIT_SIGPENDING] =
236 init_task.signal->rlim[RLIMIT_NPROC];
237}
238
Suresh Siddha61c46282008-03-10 15:28:04 -0700239int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
240 struct task_struct *src)
241{
242 *dst = *src;
243 return 0;
244}
245
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246static struct task_struct *dup_task_struct(struct task_struct *orig)
247{
248 struct task_struct *tsk;
249 struct thread_info *ti;
Eric Sandeen7c9f8862008-04-22 16:38:23 -0500250 unsigned long *stackend;
251
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700252 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253
254 prepare_to_copy(orig);
255
256 tsk = alloc_task_struct();
257 if (!tsk)
258 return NULL;
259
260 ti = alloc_thread_info(tsk);
261 if (!ti) {
262 free_task_struct(tsk);
263 return NULL;
264 }
265
Suresh Siddha61c46282008-03-10 15:28:04 -0700266 err = arch_dup_task_struct(tsk, orig);
267 if (err)
268 goto out;
269
Roman Zippelf7e42172007-05-09 02:35:17 -0700270 tsk->stack = ti;
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700271
272 err = prop_local_init_single(&tsk->dirties);
Suresh Siddha61c46282008-03-10 15:28:04 -0700273 if (err)
274 goto out;
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700275
Al Viro10ebffd2005-11-13 16:06:56 -0800276 setup_thread_stack(tsk, orig);
Avi Kivity8e7cac72009-11-29 16:34:48 +0200277 clear_user_return_notifier(tsk);
Mike Galbraithf26f9af2010-12-08 11:05:42 +0100278 clear_tsk_need_resched(tsk);
Eric Sandeen7c9f8862008-04-22 16:38:23 -0500279 stackend = end_of_stack(tsk);
280 *stackend = STACK_END_MAGIC; /* for overflow detection */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281
Arjan van de Ven0a425402006-09-26 10:52:38 +0200282#ifdef CONFIG_CC_STACKPROTECTOR
283 tsk->stack_canary = get_random_int();
284#endif
285
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 /* One for us, one for whoever does the "release_task()" (usually parent) */
287 atomic_set(&tsk->usage,2);
Giancarlo Formicuccia4b5d37a2005-09-09 13:01:22 -0700288 atomic_set(&tsk->fs_excl, 0);
Alexey Dobriyan6c5c9342006-09-29 01:59:40 -0700289#ifdef CONFIG_BLK_DEV_IO_TRACE
Jens Axboe2056a782006-03-23 20:00:26 +0100290 tsk->btrace_seq = 0;
Alexey Dobriyan6c5c9342006-09-29 01:59:40 -0700291#endif
Jens Axboea0aa7f62006-04-20 13:05:33 +0200292 tsk->splice_pipe = NULL;
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700293
294 account_kernel_stack(ti, 1);
295
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296 return tsk;
Suresh Siddha61c46282008-03-10 15:28:04 -0700297
298out:
299 free_thread_info(ti);
300 free_task_struct(tsk);
301 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302}
303
304#ifdef CONFIG_MMU
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700305static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306{
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700307 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 struct rb_node **rb_link, *rb_parent;
309 int retval;
310 unsigned long charge;
311 struct mempolicy *pol;
312
313 down_write(&oldmm->mmap_sem);
Ralf Baechleec8c0442006-12-12 17:14:57 +0000314 flush_cache_dup_mm(oldmm);
Ingo Molnarad339452006-07-03 00:25:15 -0700315 /*
316 * Not linked in yet - no deadlock potential:
317 */
318 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
Hugh Dickins7ee78232005-10-29 18:16:08 -0700319
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 mm->locked_vm = 0;
321 mm->mmap = NULL;
322 mm->mmap_cache = NULL;
323 mm->free_area_cache = oldmm->mmap_base;
Wolfgang Wander1363c3c2005-06-21 17:14:49 -0700324 mm->cached_hole_size = ~0UL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325 mm->map_count = 0;
Rusty Russell94894242009-03-30 22:05:12 -0600326 cpumask_clear(mm_cpumask(mm));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 mm->mm_rb = RB_ROOT;
328 rb_link = &mm->mm_rb.rb_node;
329 rb_parent = NULL;
330 pprev = &mm->mmap;
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700331 retval = ksm_fork(mm, oldmm);
332 if (retval)
333 goto out;
Andrea Arcangeliba761492011-01-13 15:46:58 -0800334 retval = khugepaged_fork(mm, oldmm);
335 if (retval)
336 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700338 prev = NULL;
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700339 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 struct file *file;
341
342 if (mpnt->vm_flags & VM_DONTCOPY) {
Hugh Dickins3b6bfcd2005-07-12 13:58:09 -0700343 long pages = vma_pages(mpnt);
344 mm->total_vm -= pages;
Hugh Dickinsab50b8e2005-10-29 18:15:56 -0700345 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
Hugh Dickins3b6bfcd2005-07-12 13:58:09 -0700346 -pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 continue;
348 }
349 charge = 0;
350 if (mpnt->vm_flags & VM_ACCOUNT) {
351 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
352 if (security_vm_enough_memory(len))
353 goto fail_nomem;
354 charge = len;
355 }
Christoph Lametere94b1762006-12-06 20:33:17 -0800356 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 if (!tmp)
358 goto fail_nomem;
359 *tmp = *mpnt;
Rik van Riel5beb4932010-03-05 13:42:07 -0800360 INIT_LIST_HEAD(&tmp->anon_vma_chain);
Lee Schermerhorn846a16b2008-04-28 02:13:09 -0700361 pol = mpol_dup(vma_policy(mpnt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 retval = PTR_ERR(pol);
363 if (IS_ERR(pol))
364 goto fail_nomem_policy;
365 vma_set_policy(tmp, pol);
Andrea Arcangelia247c3a2010-09-22 13:05:12 -0700366 tmp->vm_mm = mm;
Rik van Riel5beb4932010-03-05 13:42:07 -0800367 if (anon_vma_fork(tmp, mpnt))
368 goto fail_nomem_anon_vma_fork;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 tmp->vm_flags &= ~VM_LOCKED;
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700370 tmp->vm_next = tmp->vm_prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 file = tmp->vm_file;
372 if (file) {
Josef "Jeff" Sipekf3a43f32006-12-08 02:36:43 -0800373 struct inode *inode = file->f_path.dentry->d_inode;
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000374 struct address_space *mapping = file->f_mapping;
375
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 get_file(file);
377 if (tmp->vm_flags & VM_DENYWRITE)
378 atomic_dec(&inode->i_writecount);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000379 spin_lock(&mapping->i_mmap_lock);
380 if (tmp->vm_flags & VM_SHARED)
381 mapping->i_mmap_writable++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 tmp->vm_truncate_count = mpnt->vm_truncate_count;
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000383 flush_dcache_mmap_lock(mapping);
384 /* insert tmp into the share list, just after mpnt */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 vma_prio_tree_add(tmp, mpnt);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000386 flush_dcache_mmap_unlock(mapping);
387 spin_unlock(&mapping->i_mmap_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 }
389
390 /*
Mel Gormana1e78772008-07-23 21:27:23 -0700391 * Clear hugetlb-related page reserves for children. This only
392 * affects MAP_PRIVATE mappings. Faults generated by the child
393 * are not guaranteed to succeed, even if read-only
394 */
395 if (is_vm_hugetlb_page(tmp))
396 reset_vma_resv_huge_pages(tmp);
397
398 /*
Hugh Dickins7ee78232005-10-29 18:16:08 -0700399 * Link in the new vma and copy the page table entries.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 *pprev = tmp;
402 pprev = &tmp->vm_next;
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700403 tmp->vm_prev = prev;
404 prev = tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405
406 __vma_link_rb(mm, tmp, rb_link, rb_parent);
407 rb_link = &tmp->vm_rb.rb_right;
408 rb_parent = &tmp->vm_rb;
409
410 mm->map_count++;
Hugh Dickins0b0db142005-11-21 21:32:20 -0800411 retval = copy_page_range(mm, oldmm, mpnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412
413 if (tmp->vm_ops && tmp->vm_ops->open)
414 tmp->vm_ops->open(tmp);
415
416 if (retval)
417 goto out;
418 }
Jeremy Fitzhardinged6dd61c2007-05-02 19:27:14 +0200419 /* a new mm has just been created */
420 arch_dup_mmap(oldmm, mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422out:
Hugh Dickins7ee78232005-10-29 18:16:08 -0700423 up_write(&mm->mmap_sem);
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700424 flush_tlb_mm(oldmm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 up_write(&oldmm->mmap_sem);
426 return retval;
Rik van Riel5beb4932010-03-05 13:42:07 -0800427fail_nomem_anon_vma_fork:
428 mpol_put(pol);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429fail_nomem_policy:
430 kmem_cache_free(vm_area_cachep, tmp);
431fail_nomem:
432 retval = -ENOMEM;
433 vm_unacct_memory(charge);
434 goto out;
435}
436
437static inline int mm_alloc_pgd(struct mm_struct * mm)
438{
439 mm->pgd = pgd_alloc(mm);
440 if (unlikely(!mm->pgd))
441 return -ENOMEM;
442 return 0;
443}
444
445static inline void mm_free_pgd(struct mm_struct * mm)
446{
Benjamin Herrenschmidt5e541972008-02-04 22:29:14 -0800447 pgd_free(mm, mm->pgd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448}
449#else
450#define dup_mmap(mm, oldmm) (0)
451#define mm_alloc_pgd(mm) (0)
452#define mm_free_pgd(mm)
453#endif /* CONFIG_MMU */
454
Daniel Walker23ff4442007-10-18 03:06:07 -0700455__cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456
Christoph Lametere94b1762006-12-06 20:33:17 -0800457#define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458#define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
459
Hidehiro Kawai4cb0e112009-01-06 14:42:47 -0800460static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
461
462static int __init coredump_filter_setup(char *s)
463{
464 default_dump_filter =
465 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
466 MMF_DUMP_FILTER_MASK;
467 return 1;
468}
469
470__setup("coredump_filter=", coredump_filter_setup);
471
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472#include <linux/init_task.h>
473
Alexey Dobriyan858f0992009-09-23 15:57:32 -0700474static void mm_init_aio(struct mm_struct *mm)
475{
476#ifdef CONFIG_AIO
477 spin_lock_init(&mm->ioctx_lock);
478 INIT_HLIST_HEAD(&mm->ioctx_list);
479#endif
480}
481
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800482static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483{
484 atomic_set(&mm->mm_users, 1);
485 atomic_set(&mm->mm_count, 1);
486 init_rwsem(&mm->mmap_sem);
487 INIT_LIST_HEAD(&mm->mmlist);
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700488 mm->flags = (current->mm) ?
489 (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700490 mm->core_state = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 mm->nr_ptes = 0;
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -0800492 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 spin_lock_init(&mm->page_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 mm->free_area_cache = TASK_UNMAPPED_BASE;
Wolfgang Wander1363c3c2005-06-21 17:14:49 -0700495 mm->cached_hole_size = ~0UL;
Alexey Dobriyan858f0992009-09-23 15:57:32 -0700496 mm_init_aio(mm);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700497 mm_init_owner(mm, p);
Ying Han3d5992d2010-10-26 14:21:23 -0700498 atomic_set(&mm->oom_disable_count, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499
500 if (likely(!mm_alloc_pgd(mm))) {
501 mm->def_flags = 0;
Andrea Arcangelicddb8a52008-07-28 15:46:29 -0700502 mmu_notifier_mm_init(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 return mm;
504 }
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800505
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 free_mm(mm);
507 return NULL;
508}
509
510/*
511 * Allocate and initialize an mm_struct.
512 */
513struct mm_struct * mm_alloc(void)
514{
515 struct mm_struct * mm;
516
517 mm = allocate_mm();
518 if (mm) {
519 memset(mm, 0, sizeof(*mm));
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800520 mm = mm_init(mm, current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 }
522 return mm;
523}
524
525/*
526 * Called when the last reference to the mm
527 * is dropped: either by a lazy thread or by
528 * mmput. Free the page directory and the mm.
529 */
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800530void __mmdrop(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531{
532 BUG_ON(mm == &init_mm);
533 mm_free_pgd(mm);
534 destroy_context(mm);
Andrea Arcangelicddb8a52008-07-28 15:46:29 -0700535 mmu_notifier_mm_destroy(mm);
Andrea Arcangelie7a00c42011-01-13 15:46:45 -0800536#ifdef CONFIG_TRANSPARENT_HUGEPAGE
537 VM_BUG_ON(mm->pmd_huge_pte);
538#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 free_mm(mm);
540}
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200541EXPORT_SYMBOL_GPL(__mmdrop);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542
543/*
544 * Decrement the use count and release all resources for an mm.
545 */
546void mmput(struct mm_struct *mm)
547{
Andrew Morton0ae26f12006-06-23 02:05:15 -0700548 might_sleep();
549
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 if (atomic_dec_and_test(&mm->mm_users)) {
551 exit_aio(mm);
Andrea Arcangeli1c2fb7a2009-09-21 17:02:22 -0700552 ksm_exit(mm);
Andrea Arcangeliba761492011-01-13 15:46:58 -0800553 khugepaged_exit(mm); /* must run before exit_mmap */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 exit_mmap(mm);
Matt Helsley925d1c42008-04-29 01:01:36 -0700555 set_mm_exe_file(mm, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 if (!list_empty(&mm->mmlist)) {
557 spin_lock(&mmlist_lock);
558 list_del(&mm->mmlist);
559 spin_unlock(&mmlist_lock);
560 }
561 put_swap_token(mm);
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -0700562 if (mm->binfmt)
563 module_put(mm->binfmt->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 mmdrop(mm);
565 }
566}
567EXPORT_SYMBOL_GPL(mmput);
568
569/**
570 * get_task_mm - acquire a reference to the task's mm
571 *
Oleg Nesterov246bb0b2008-07-25 01:47:38 -0700572 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 * this kernel workthread has transiently adopted a user mm with use_mm,
574 * to do its AIO) is not set and if so returns a reference to it, after
575 * bumping up the use count. User must release the mm via mmput()
576 * after use. Typically used by /proc and ptrace.
577 */
578struct mm_struct *get_task_mm(struct task_struct *task)
579{
580 struct mm_struct *mm;
581
582 task_lock(task);
583 mm = task->mm;
584 if (mm) {
Oleg Nesterov246bb0b2008-07-25 01:47:38 -0700585 if (task->flags & PF_KTHREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 mm = NULL;
587 else
588 atomic_inc(&mm->mm_users);
589 }
590 task_unlock(task);
591 return mm;
592}
593EXPORT_SYMBOL_GPL(get_task_mm);
594
595/* Please note the differences between mmput and mm_release.
596 * mmput is called whenever we stop holding onto a mm_struct,
597 * error success whatever.
598 *
599 * mm_release is called after a mm_struct has been removed
600 * from the current process.
601 *
602 * This difference is important for error handling, when we
603 * only half set up a mm_struct for a new process and need to restore
604 * the old one. Because we mmput the new mm_struct before
605 * restoring the old one. . .
606 * Eric Biederman 10 January 1998
607 */
608void mm_release(struct task_struct *tsk, struct mm_struct *mm)
609{
610 struct completion *vfork_done = tsk->vfork_done;
611
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800612 /* Get rid of any futexes when releasing the mm */
613#ifdef CONFIG_FUTEX
Peter Zijlstrafc6b1772009-10-05 18:17:32 +0200614 if (unlikely(tsk->robust_list)) {
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800615 exit_robust_list(tsk);
Peter Zijlstrafc6b1772009-10-05 18:17:32 +0200616 tsk->robust_list = NULL;
617 }
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800618#ifdef CONFIG_COMPAT
Peter Zijlstrafc6b1772009-10-05 18:17:32 +0200619 if (unlikely(tsk->compat_robust_list)) {
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800620 compat_exit_robust_list(tsk);
Peter Zijlstrafc6b1772009-10-05 18:17:32 +0200621 tsk->compat_robust_list = NULL;
622 }
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800623#endif
Thomas Gleixner322a2c12009-10-05 18:18:03 +0200624 if (unlikely(!list_empty(&tsk->pi_state_list)))
625 exit_pi_state_list(tsk);
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800626#endif
627
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 /* Get rid of any cached register state */
629 deactivate_mm(tsk, mm);
630
631 /* notify parent sleeping on vfork() */
632 if (vfork_done) {
633 tsk->vfork_done = NULL;
634 complete(vfork_done);
635 }
Roland McGrathfec1d012006-12-06 20:36:34 -0800636
637 /*
638 * If we're exiting normally, clear a user-space tid field if
639 * requested. We leave this alone when dying by signal, to leave
640 * the value intact in a core dump, and to save the unnecessary
641 * trouble otherwise. Userland only wants this done for a sys_exit.
642 */
Eric Dumazet9c8a8222009-08-06 15:09:28 -0700643 if (tsk->clear_child_tid) {
644 if (!(tsk->flags & PF_SIGNALED) &&
645 atomic_read(&mm->mm_users) > 1) {
646 /*
647 * We don't check the error code - if userspace has
648 * not set up a proper pointer then tough luck.
649 */
650 put_user(0, tsk->clear_child_tid);
651 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
652 1, NULL, NULL, 0);
653 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 tsk->clear_child_tid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 }
656}
657
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800658/*
659 * Allocate a new mm structure and copy contents from the
660 * mm structure of the passed in task structure.
661 */
Carsten Otte402b0862008-03-25 18:47:10 +0100662struct mm_struct *dup_mm(struct task_struct *tsk)
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800663{
664 struct mm_struct *mm, *oldmm = current->mm;
665 int err;
666
667 if (!oldmm)
668 return NULL;
669
670 mm = allocate_mm();
671 if (!mm)
672 goto fail_nomem;
673
674 memcpy(mm, oldmm, sizeof(*mm));
675
Ashwin Chaugule7602bdf2006-12-06 20:31:57 -0800676 /* Initializing for Swap token stuff */
677 mm->token_priority = 0;
678 mm->last_interval = 0;
679
Andrea Arcangelie7a00c42011-01-13 15:46:45 -0800680#ifdef CONFIG_TRANSPARENT_HUGEPAGE
681 mm->pmd_huge_pte = NULL;
682#endif
683
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800684 if (!mm_init(mm, tsk))
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800685 goto fail_nomem;
686
687 if (init_new_context(tsk, mm))
688 goto fail_nocontext;
689
Matt Helsley925d1c42008-04-29 01:01:36 -0700690 dup_mm_exe_file(oldmm, mm);
691
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800692 err = dup_mmap(mm, oldmm);
693 if (err)
694 goto free_pt;
695
696 mm->hiwater_rss = get_mm_rss(mm);
697 mm->hiwater_vm = mm->total_vm;
698
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -0700699 if (mm->binfmt && !try_module_get(mm->binfmt->module))
700 goto free_pt;
701
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800702 return mm;
703
704free_pt:
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -0700705 /* don't put binfmt in mmput, we haven't got module yet */
706 mm->binfmt = NULL;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800707 mmput(mm);
708
709fail_nomem:
710 return NULL;
711
712fail_nocontext:
713 /*
714 * If init_new_context() failed, we cannot use mmput() to free the mm
715 * because it calls destroy_context()
716 */
717 mm_free_pgd(mm);
718 free_mm(mm);
719 return NULL;
720}
721
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722static int copy_mm(unsigned long clone_flags, struct task_struct * tsk)
723{
724 struct mm_struct * mm, *oldmm;
725 int retval;
726
727 tsk->min_flt = tsk->maj_flt = 0;
728 tsk->nvcsw = tsk->nivcsw = 0;
Mandeep Singh Baines17406b82009-02-06 15:37:47 -0800729#ifdef CONFIG_DETECT_HUNG_TASK
730 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
731#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732
733 tsk->mm = NULL;
734 tsk->active_mm = NULL;
735
736 /*
737 * Are we cloning a kernel thread?
738 *
739 * We need to steal a active VM for that..
740 */
741 oldmm = current->mm;
742 if (!oldmm)
743 return 0;
744
745 if (clone_flags & CLONE_VM) {
746 atomic_inc(&oldmm->mm_users);
747 mm = oldmm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 goto good_mm;
749 }
750
751 retval = -ENOMEM;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800752 mm = dup_mm(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 if (!mm)
754 goto fail_nomem;
755
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756good_mm:
Ashwin Chaugule7602bdf2006-12-06 20:31:57 -0800757 /* Initializing for Swap token stuff */
758 mm->token_priority = 0;
759 mm->last_interval = 0;
Ying Han3d5992d2010-10-26 14:21:23 -0700760 if (tsk->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
761 atomic_inc(&mm->oom_disable_count);
Ashwin Chaugule7602bdf2006-12-06 20:31:57 -0800762
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763 tsk->mm = mm;
764 tsk->active_mm = mm;
765 return 0;
766
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767fail_nomem:
768 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769}
770
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700771static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772{
Al Viro498052b2009-03-30 07:20:30 -0400773 struct fs_struct *fs = current->fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 if (clone_flags & CLONE_FS) {
Al Viro498052b2009-03-30 07:20:30 -0400775 /* tsk->fs is already what we want */
Nick Piggin2a4419b2010-08-18 04:37:33 +1000776 spin_lock(&fs->lock);
Al Viro498052b2009-03-30 07:20:30 -0400777 if (fs->in_exec) {
Nick Piggin2a4419b2010-08-18 04:37:33 +1000778 spin_unlock(&fs->lock);
Al Viro498052b2009-03-30 07:20:30 -0400779 return -EAGAIN;
780 }
781 fs->users++;
Nick Piggin2a4419b2010-08-18 04:37:33 +1000782 spin_unlock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 return 0;
784 }
Al Viro498052b2009-03-30 07:20:30 -0400785 tsk->fs = copy_fs_struct(fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 if (!tsk->fs)
787 return -ENOMEM;
788 return 0;
789}
790
JANAK DESAIa016f332006-02-07 12:59:02 -0800791static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
792{
793 struct files_struct *oldf, *newf;
794 int error = 0;
795
796 /*
797 * A background process may not have any files ...
798 */
799 oldf = current->files;
800 if (!oldf)
801 goto out;
802
803 if (clone_flags & CLONE_FILES) {
804 atomic_inc(&oldf->count);
805 goto out;
806 }
807
JANAK DESAIa016f332006-02-07 12:59:02 -0800808 newf = dup_fd(oldf, &error);
809 if (!newf)
810 goto out;
811
812 tsk->files = newf;
813 error = 0;
814out:
815 return error;
816}
817
Jens Axboefadad8782008-01-24 08:54:47 +0100818static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
Jens Axboefd0928d2008-01-24 08:52:45 +0100819{
820#ifdef CONFIG_BLOCK
821 struct io_context *ioc = current->io_context;
822
823 if (!ioc)
824 return 0;
Jens Axboefadad8782008-01-24 08:54:47 +0100825 /*
826 * Share io context with parent, if CLONE_IO is set
827 */
828 if (clone_flags & CLONE_IO) {
829 tsk->io_context = ioc_task_link(ioc);
830 if (unlikely(!tsk->io_context))
831 return -ENOMEM;
832 } else if (ioprio_valid(ioc->ioprio)) {
Jens Axboefd0928d2008-01-24 08:52:45 +0100833 tsk->io_context = alloc_io_context(GFP_KERNEL, -1);
834 if (unlikely(!tsk->io_context))
835 return -ENOMEM;
836
Jens Axboefd0928d2008-01-24 08:52:45 +0100837 tsk->io_context->ioprio = ioc->ioprio;
838 }
839#endif
840 return 0;
841}
842
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700843static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844{
845 struct sighand_struct *sig;
846
Zhaolei60348802009-01-06 14:40:46 -0800847 if (clone_flags & CLONE_SIGHAND) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 atomic_inc(&current->sighand->count);
849 return 0;
850 }
851 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
Ingo Molnare56d0902006-01-08 01:01:37 -0800852 rcu_assign_pointer(tsk->sighand, sig);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 if (!sig)
854 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 atomic_set(&sig->count, 1);
856 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
857 return 0;
858}
859
Oleg Nesterova7e53282006-03-28 16:11:27 -0800860void __cleanup_sighand(struct sighand_struct *sighand)
Oleg Nesterovc81addc2006-03-28 16:11:17 -0800861{
Oleg Nesterovc81addc2006-03-28 16:11:17 -0800862 if (atomic_dec_and_test(&sighand->count))
863 kmem_cache_free(sighand_cachep, sighand);
864}
865
Frank Mayharf06febc2008-09-12 09:54:39 -0700866
867/*
868 * Initialize POSIX timer handling for a thread group.
869 */
870static void posix_cpu_timers_init_group(struct signal_struct *sig)
871{
Jiri Slaby78d7d402010-03-05 13:42:54 -0800872 unsigned long cpu_limit;
873
Frank Mayharf06febc2008-09-12 09:54:39 -0700874 /* Thread group counters. */
875 thread_group_cputime_init(sig);
876
Jiri Slaby78d7d402010-03-05 13:42:54 -0800877 cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
878 if (cpu_limit != RLIM_INFINITY) {
879 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
Oleg Nesterov6279a7512009-03-27 01:06:07 +0100880 sig->cputimer.running = 1;
881 }
882
Frank Mayharf06febc2008-09-12 09:54:39 -0700883 /* The timer lists. */
884 INIT_LIST_HEAD(&sig->cpu_timers[0]);
885 INIT_LIST_HEAD(&sig->cpu_timers[1]);
886 INIT_LIST_HEAD(&sig->cpu_timers[2]);
887}
888
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700889static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890{
891 struct signal_struct *sig;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892
Oleg Nesterov4ab6c082009-08-26 14:29:24 -0700893 if (clone_flags & CLONE_THREAD)
Peter Zijlstra490dea42008-11-24 17:06:57 +0100894 return 0;
Oleg Nesterov6279a7512009-03-27 01:06:07 +0100895
Veaceslav Falicoa56704e2010-03-10 15:23:01 -0800896 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 tsk->signal = sig;
898 if (!sig)
899 return -ENOMEM;
900
Oleg Nesterovb3ac0222010-05-26 14:43:24 -0700901 sig->nr_threads = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 atomic_set(&sig->live, 1);
Oleg Nesterovb3ac0222010-05-26 14:43:24 -0700903 atomic_set(&sig->sigcnt, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 init_waitqueue_head(&sig->wait_chldexit);
Sukadev Bhattiprolub3bfa0c2009-04-02 16:58:08 -0700905 if (clone_flags & CLONE_NEWPID)
906 sig->flags |= SIGNAL_UNKILLABLE;
Oleg Nesterovdb51aec2008-04-30 00:52:52 -0700907 sig->curr_target = tsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 init_sigpending(&sig->shared_pending);
909 INIT_LIST_HEAD(&sig->posix_timers);
910
Thomas Gleixnerc9cb2e32007-02-16 01:27:49 -0800911 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 sig->real_timer.function = it_real_fn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 task_lock(current->group_leader);
915 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
916 task_unlock(current->group_leader);
917
Oleg Nesterov6279a7512009-03-27 01:06:07 +0100918 posix_cpu_timers_init_group(sig);
919
Miloslav Trmac522ed772007-07-15 23:40:56 -0700920 tty_audit_fork(sig);
Mike Galbraith5091faa2010-11-30 14:18:03 +0100921 sched_autogroup_fork(sig);
Miloslav Trmac522ed772007-07-15 23:40:56 -0700922
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -0700923 sig->oom_adj = current->signal->oom_adj;
David Rientjesa63d83f2010-08-09 17:19:46 -0700924 sig->oom_score_adj = current->signal->oom_score_adj;
Mandeep Singh Bainesdabb16f2011-01-13 15:46:05 -0800925 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -0700926
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -0700927 mutex_init(&sig->cred_guard_mutex);
928
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 return 0;
930}
931
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700932static void copy_flags(unsigned long clone_flags, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933{
934 unsigned long new_flags = p->flags;
935
Tejun Heo21aa9af2010-06-08 21:40:37 +0200936 new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 new_flags |= PF_FORKNOEXEC;
Roland McGrath09a05392008-07-25 19:45:47 -0700938 new_flags |= PF_STARTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 p->flags = new_flags;
Rafael J. Wysocki2e131892007-10-18 03:04:45 -0700940 clear_freeze_flag(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941}
942
Heiko Carstens17da2bd2009-01-14 14:14:10 +0100943SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944{
945 current->clear_child_tid = tidptr;
946
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700947 return task_pid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948}
949
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700950static void rt_mutex_init_task(struct task_struct *p)
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700951{
Thomas Gleixner1d615482009-11-17 14:54:03 +0100952 raw_spin_lock_init(&p->pi_lock);
Zilvinas Valinskase29e1752007-03-16 13:38:34 -0800953#ifdef CONFIG_RT_MUTEXES
Thomas Gleixner1d615482009-11-17 14:54:03 +0100954 plist_head_init_raw(&p->pi_waiters, &p->pi_lock);
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700955 p->pi_blocked_on = NULL;
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700956#endif
957}
958
Balbir Singhcf475ad2008-04-29 01:00:16 -0700959#ifdef CONFIG_MM_OWNER
960void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
961{
962 mm->owner = p;
963}
964#endif /* CONFIG_MM_OWNER */
965
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966/*
Frank Mayharf06febc2008-09-12 09:54:39 -0700967 * Initialize POSIX timer handling for a single task.
968 */
969static void posix_cpu_timers_init(struct task_struct *tsk)
970{
971 tsk->cputime_expires.prof_exp = cputime_zero;
972 tsk->cputime_expires.virt_exp = cputime_zero;
973 tsk->cputime_expires.sched_exp = 0;
974 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
975 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
976 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
977}
978
979/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 * This creates a new process as a copy of the old one,
981 * but does not actually start it yet.
982 *
983 * It copies the registers, and all the appropriate
984 * parts of the process environment (as per the clone
985 * flags). The actual kick-off is left to the caller.
986 */
Ingo Molnar36c8b582006-07-03 00:25:41 -0700987static struct task_struct *copy_process(unsigned long clone_flags,
988 unsigned long stack_start,
989 struct pt_regs *regs,
990 unsigned long stack_size,
Ingo Molnar36c8b582006-07-03 00:25:41 -0700991 int __user *child_tidptr,
Roland McGrath09a05392008-07-25 19:45:47 -0700992 struct pid *pid,
993 int trace)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994{
995 int retval;
Mariusz Kozlowskia24efe62007-10-18 23:41:09 -0700996 struct task_struct *p;
Paul Menageb4f48b62007-10-18 23:39:33 -0700997 int cgroup_callbacks_done = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998
999 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1000 return ERR_PTR(-EINVAL);
1001
1002 /*
1003 * Thread groups must share signals as well, and detached threads
1004 * can only be started up within the thread group.
1005 */
1006 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1007 return ERR_PTR(-EINVAL);
1008
1009 /*
1010 * Shared signal handlers imply shared VM. By way of the above,
1011 * thread groups also imply shared VM. Blocking this case allows
1012 * for various simplifications in other code.
1013 */
1014 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1015 return ERR_PTR(-EINVAL);
1016
Sukadev Bhattiprolu123be072009-09-23 15:57:20 -07001017 /*
1018 * Siblings of global init remain as zombies on exit since they are
1019 * not reaped by their parent (swapper). To solve this and to avoid
1020 * multi-rooted process trees, prevent global and container-inits
1021 * from creating siblings.
1022 */
1023 if ((clone_flags & CLONE_PARENT) &&
1024 current->signal->flags & SIGNAL_UNKILLABLE)
1025 return ERR_PTR(-EINVAL);
1026
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 retval = security_task_create(clone_flags);
1028 if (retval)
1029 goto fork_out;
1030
1031 retval = -ENOMEM;
1032 p = dup_task_struct(current);
1033 if (!p)
1034 goto fork_out;
1035
Steven Rostedtf7e8b612009-06-02 16:39:48 -04001036 ftrace_graph_init_task(p);
1037
Peter Zijlstrabea493a2006-10-17 00:10:33 -07001038 rt_mutex_init_task(p);
1039
Ingo Molnard12c1a32008-07-14 12:09:28 +02001040#ifdef CONFIG_PROVE_LOCKING
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001041 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1042 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1043#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 retval = -EAGAIN;
David Howells3b11a1d2008-11-14 10:39:26 +11001045 if (atomic_read(&p->real_cred->user->processes) >=
Jiri Slaby78d7d402010-03-05 13:42:54 -08001046 task_rlimit(p, RLIMIT_NPROC)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
Serge Hallyn18b6e042008-10-15 16:38:45 -05001048 p->real_cred->user != INIT_USER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 goto bad_fork_free;
1050 }
1051
David Howellsf1752ee2008-11-14 10:39:17 +11001052 retval = copy_creds(p, clone_flags);
1053 if (retval < 0)
1054 goto bad_fork_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055
1056 /*
1057 * If multiple threads are within copy_process(), then this check
1058 * triggers too late. This doesn't hurt, the check is only there
1059 * to stop root fork bombs.
1060 */
Li Zefan04ec93f2009-02-06 08:17:19 +00001061 retval = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062 if (nr_threads >= max_threads)
1063 goto bad_fork_cleanup_count;
1064
Al Viroa1261f52005-11-13 16:06:55 -08001065 if (!try_module_get(task_thread_info(p)->exec_domain->module))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066 goto bad_fork_cleanup_count;
1067
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 p->did_exec = 0;
Shailabh Nagarca74e922006-07-14 00:24:36 -07001069 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 copy_flags(clone_flags, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 INIT_LIST_HEAD(&p->children);
1072 INIT_LIST_HEAD(&p->sibling);
Paul E. McKenneyf41d9112009-08-22 13:56:52 -07001073 rcu_copy_process(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 p->vfork_done = NULL;
1075 spin_lock_init(&p->alloc_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 init_sigpending(&p->pending);
1078
1079 p->utime = cputime_zero;
1080 p->stime = cputime_zero;
Laurent Vivier9ac52312007-10-15 17:00:19 +02001081 p->gtime = cputime_zero;
Michael Neulingc66f08b2007-10-18 03:06:34 -07001082 p->utimescaled = cputime_zero;
1083 p->stimescaled = cputime_zero;
Hidetoshi Setod99ca3b2009-12-02 17:26:47 +09001084#ifndef CONFIG_VIRT_CPU_ACCOUNTING
Peter Zijlstra73a2bcb2007-10-29 21:18:11 +01001085 p->prev_utime = cputime_zero;
Balbir Singh93018992007-10-30 00:26:32 +01001086 p->prev_stime = cputime_zero;
Hidetoshi Setod99ca3b2009-12-02 17:26:47 +09001087#endif
KAMEZAWA Hiroyukia3a2e762010-04-06 14:34:42 -07001088#if defined(SPLIT_RSS_COUNTING)
1089 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1090#endif
Balbir Singh172ba842007-07-09 18:52:00 +02001091
Arjan van de Ven69766752008-09-01 15:52:40 -07001092 p->default_timer_slack_ns = current->timer_slack_ns;
1093
Andrea Righi59954772008-07-27 17:29:15 +02001094 task_io_accounting_init(&p->ioac);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 acct_clear_integrals(p);
1096
Frank Mayharf06febc2008-09-12 09:54:39 -07001097 posix_cpu_timers_init(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098
1099 p->lock_depth = -1; /* -1 = no lock */
1100 do_posix_clock_monotonic_gettime(&p->start_time);
Tomas Janousek924b42d2007-07-15 23:39:42 -07001101 p->real_start_time = p->start_time;
1102 monotonic_to_bootbased(&p->real_start_time);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 p->io_context = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 p->audit_context = NULL;
Paul Menageb4f48b62007-10-18 23:39:33 -07001105 cgroup_fork(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106#ifdef CONFIG_NUMA
Lee Schermerhorn846a16b2008-04-28 02:13:09 -07001107 p->mempolicy = mpol_dup(p->mempolicy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 if (IS_ERR(p->mempolicy)) {
1109 retval = PTR_ERR(p->mempolicy);
1110 p->mempolicy = NULL;
Paul Menageb4f48b62007-10-18 23:39:33 -07001111 goto bad_fork_cleanup_cgroup;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 }
Paul Jacksonc61afb12006-03-24 03:16:08 -08001113 mpol_fix_fork_child_flag(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114#endif
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001115#ifdef CONFIG_TRACE_IRQFLAGS
1116 p->irq_events = 0;
Russell Kingb36e4752006-08-27 12:26:34 +01001117#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1118 p->hardirqs_enabled = 1;
1119#else
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001120 p->hardirqs_enabled = 0;
Russell Kingb36e4752006-08-27 12:26:34 +01001121#endif
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001122 p->hardirq_enable_ip = 0;
1123 p->hardirq_enable_event = 0;
1124 p->hardirq_disable_ip = _THIS_IP_;
1125 p->hardirq_disable_event = 0;
1126 p->softirqs_enabled = 1;
1127 p->softirq_enable_ip = _THIS_IP_;
1128 p->softirq_enable_event = 0;
1129 p->softirq_disable_ip = 0;
1130 p->softirq_disable_event = 0;
1131 p->hardirq_context = 0;
1132 p->softirq_context = 0;
1133#endif
Ingo Molnarfbb9ce952006-07-03 00:24:50 -07001134#ifdef CONFIG_LOCKDEP
1135 p->lockdep_depth = 0; /* no locks held yet */
1136 p->curr_chain_key = 0;
1137 p->lockdep_recursion = 0;
1138#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139
Ingo Molnar408894e2006-01-09 15:59:20 -08001140#ifdef CONFIG_DEBUG_MUTEXES
1141 p->blocked_on = NULL; /* not blocked yet */
1142#endif
KAMEZAWA Hiroyuki569b8462009-12-15 16:47:03 -08001143#ifdef CONFIG_CGROUP_MEM_RES_CTLR
1144 p->memcg_batch.do_batch = 0;
1145 p->memcg_batch.memcg = NULL;
1146#endif
Markus Metzger0f481402009-04-03 16:43:48 +02001147
Srivatsa Vaddagiri3c90e6e2007-11-09 22:39:39 +01001148 /* Perform scheduler related setup. Assign this task to a CPU. */
1149 sched_fork(p, clone_flags);
Peter Zijlstra6ab423e2009-05-25 14:45:27 +02001150
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001151 retval = perf_event_init_task(p);
Peter Zijlstra6ab423e2009-05-25 14:45:27 +02001152 if (retval)
1153 goto bad_fork_cleanup_policy;
Srivatsa Vaddagiri3c90e6e2007-11-09 22:39:39 +01001154
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155 if ((retval = audit_alloc(p)))
David Howellsf1752ee2008-11-14 10:39:17 +11001156 goto bad_fork_cleanup_policy;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 /* copy all the process information */
1158 if ((retval = copy_semundo(clone_flags, p)))
1159 goto bad_fork_cleanup_audit;
1160 if ((retval = copy_files(clone_flags, p)))
1161 goto bad_fork_cleanup_semundo;
1162 if ((retval = copy_fs(clone_flags, p)))
1163 goto bad_fork_cleanup_files;
1164 if ((retval = copy_sighand(clone_flags, p)))
1165 goto bad_fork_cleanup_fs;
1166 if ((retval = copy_signal(clone_flags, p)))
1167 goto bad_fork_cleanup_sighand;
1168 if ((retval = copy_mm(clone_flags, p)))
1169 goto bad_fork_cleanup_signal;
Serge E. Hallynab516012006-10-02 02:18:06 -07001170 if ((retval = copy_namespaces(clone_flags, p)))
David Howellsd84f4f92008-11-14 10:39:23 +11001171 goto bad_fork_cleanup_mm;
Jens Axboefadad8782008-01-24 08:54:47 +01001172 if ((retval = copy_io(clone_flags, p)))
Jens Axboefd0928d2008-01-24 08:52:45 +01001173 goto bad_fork_cleanup_namespaces;
Alexey Dobriyan6f2c55b2009-04-02 16:56:59 -07001174 retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 if (retval)
Jens Axboefd0928d2008-01-24 08:52:45 +01001176 goto bad_fork_cleanup_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001178 if (pid != &init_struct_pid) {
1179 retval = -ENOMEM;
Eric W. Biederman61bce0f2009-01-07 18:08:49 -08001180 pid = alloc_pid(p->nsproxy->pid_ns);
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001181 if (!pid)
Jens Axboefd0928d2008-01-24 08:52:45 +01001182 goto bad_fork_cleanup_io;
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07001183
1184 if (clone_flags & CLONE_NEWPID) {
Eric W. Biederman61bce0f2009-01-07 18:08:49 -08001185 retval = pid_ns_prepare_proc(p->nsproxy->pid_ns);
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07001186 if (retval < 0)
1187 goto bad_fork_free_pid;
1188 }
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001189 }
1190
1191 p->pid = pid_nr(pid);
1192 p->tgid = p->pid;
1193 if (clone_flags & CLONE_THREAD)
1194 p->tgid = current->tgid;
1195
Serge E. Hallyne885dcd2008-07-25 01:47:06 -07001196 if (current->nsproxy != p->nsproxy) {
1197 retval = ns_cgroup_clone(p, pid);
1198 if (retval)
Steven Rostedtf7e8b612009-06-02 16:39:48 -04001199 goto bad_fork_free_pid;
Serge E. Hallyne885dcd2008-07-25 01:47:06 -07001200 }
1201
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1203 /*
1204 * Clear TID on mm_release()?
1205 */
1206 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
Jens Axboe73c10102011-03-08 13:19:51 +01001207#ifdef CONFIG_BLOCK
1208 p->plug = NULL;
1209#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001210#ifdef CONFIG_FUTEX
Ingo Molnar8f17d3a2006-03-27 01:16:27 -08001211 p->robust_list = NULL;
1212#ifdef CONFIG_COMPAT
1213 p->compat_robust_list = NULL;
1214#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07001215 INIT_LIST_HEAD(&p->pi_state_list);
1216 p->pi_state_cache = NULL;
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001217#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 /*
GOTO Masanorif9a38792006-03-13 21:20:44 -08001219 * sigaltstack should be cleared when sharing the same VM
1220 */
1221 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1222 p->sas_ss_sp = p->sas_ss_size = 0;
1223
1224 /*
Oleg Nesterov65808072009-12-15 16:47:16 -08001225 * Syscall tracing and stepping should be turned off in the
1226 * child regardless of CLONE_PTRACE.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 */
Oleg Nesterov65808072009-12-15 16:47:16 -08001228 user_disable_single_step(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
Laurent Viviered75e8d2005-09-03 15:57:18 -07001230#ifdef TIF_SYSCALL_EMU
1231 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1232#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01001233 clear_all_latency_tracing(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 /* ok, now we should be set up.. */
1236 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1237 p->pdeath_signal = 0;
1238 p->exit_state = 0;
1239
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 /*
1241 * Ok, make it visible to the rest of the system.
1242 * We dont wake it up yet.
1243 */
1244 p->group_leader = p;
Oleg Nesterov47e65322006-03-28 16:11:25 -08001245 INIT_LIST_HEAD(&p->thread_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246
Paul Menageb4f48b62007-10-18 23:39:33 -07001247 /* Now that the task is set up, run cgroup callbacks if
1248 * necessary. We need to run them before the task is visible
1249 * on the tasklist. */
1250 cgroup_fork_callbacks(p);
1251 cgroup_callbacks_done = 1;
1252
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 /* Need tasklist lock for parent etc handling! */
1254 write_lock_irq(&tasklist_lock);
1255
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 /* CLONE_PARENT re-uses the old parent */
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01001257 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 p->real_parent = current->real_parent;
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01001259 p->parent_exec_id = current->parent_exec_id;
1260 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 p->real_parent = current;
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01001262 p->parent_exec_id = current->self_exec_id;
1263 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264
Oleg Nesterov3f17da62006-02-15 22:13:24 +03001265 spin_lock(&current->sighand->siglock);
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001266
1267 /*
1268 * Process group and session signals need to be delivered to just the
1269 * parent before the fork or both the parent and the child after the
1270 * fork. Restart if a signal comes in before we add the new process to
1271 * it's process group.
1272 * A fatal signal pending means that current will exit, so the new
1273 * thread can't slip out of an OOM kill (or normal SIGKILL).
1274 */
Daniel Walker23ff4442007-10-18 03:06:07 -07001275 recalc_sigpending();
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001276 if (signal_pending(current)) {
1277 spin_unlock(&current->sighand->siglock);
1278 write_unlock_irq(&tasklist_lock);
1279 retval = -ERESTARTNOINTR;
Steven Rostedtf7e8b612009-06-02 16:39:48 -04001280 goto bad_fork_free_pid;
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001281 }
1282
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 if (clone_flags & CLONE_THREAD) {
Oleg Nesterovb3ac0222010-05-26 14:43:24 -07001284 current->signal->nr_threads++;
Oleg Nesterov4ab6c082009-08-26 14:29:24 -07001285 atomic_inc(&current->signal->live);
Oleg Nesterovb3ac0222010-05-26 14:43:24 -07001286 atomic_inc(&current->signal->sigcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 p->group_leader = current->group_leader;
Oleg Nesterov47e65322006-03-28 16:11:25 -08001288 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 }
1290
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001291 if (likely(p->pid)) {
Roland McGrath09a05392008-07-25 19:45:47 -07001292 tracehook_finish_clone(p, clone_flags, trace);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001294 if (thread_group_leader(p)) {
Eric W. Biederman5cd17562007-12-04 23:45:04 -08001295 if (clone_flags & CLONE_NEWPID)
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001296 p->nsproxy->pid_ns->child_reaper = p;
Oleg Nesterovc97d9892006-03-28 16:11:06 -08001297
Oleg Nesterovfea9d172008-02-08 04:19:19 -08001298 p->signal->leader_pid = pid;
Alan Cox9c9f4de2008-10-13 10:37:26 +01001299 p->signal->tty = tty_kref_get(current->signal->tty);
Eric W. Biederman5cd17562007-12-04 23:45:04 -08001300 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1301 attach_pid(p, PIDTYPE_SID, task_session(current));
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001302 list_add_tail(&p->sibling, &p->real_parent->children);
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -07001303 list_add_tail_rcu(&p->tasks, &init_task.tasks);
Christoph Lameter909ea962010-12-08 16:22:55 +01001304 __this_cpu_inc(process_counts);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001305 }
Sukadev Bhattiprolu85868992007-05-10 22:23:03 -07001306 attach_pid(p, PIDTYPE_PID, pid);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001307 nr_threads++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308 }
1309
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310 total_forks++;
Oleg Nesterov3f17da62006-02-15 22:13:24 +03001311 spin_unlock(&current->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 write_unlock_irq(&tasklist_lock);
Andrew Mortonc13cf852005-11-28 13:43:48 -08001313 proc_fork_connector(p);
Paul Menage817929e2007-10-18 23:39:36 -07001314 cgroup_post_fork(p);
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001315 perf_event_fork(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 return p;
1317
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001318bad_fork_free_pid:
1319 if (pid != &init_struct_pid)
1320 free_pid(pid);
Jens Axboefd0928d2008-01-24 08:52:45 +01001321bad_fork_cleanup_io:
Louis Rillingb69f2292009-12-04 14:52:42 +01001322 if (p->io_context)
1323 exit_io_context(p);
Serge E. Hallynab516012006-10-02 02:18:06 -07001324bad_fork_cleanup_namespaces:
Linus Torvalds444f3782007-01-30 13:35:18 -08001325 exit_task_namespaces(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326bad_fork_cleanup_mm:
Ying Han3d5992d2010-10-26 14:21:23 -07001327 if (p->mm) {
1328 task_lock(p);
1329 if (p->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1330 atomic_dec(&p->mm->oom_disable_count);
1331 task_unlock(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 mmput(p->mm);
Ying Han3d5992d2010-10-26 14:21:23 -07001333 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334bad_fork_cleanup_signal:
Oleg Nesterov4ab6c082009-08-26 14:29:24 -07001335 if (!(clone_flags & CLONE_THREAD))
Mike Galbraith1c5354d2011-01-05 11:16:04 +01001336 free_signal_struct(p->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337bad_fork_cleanup_sighand:
Oleg Nesterova7e53282006-03-28 16:11:27 -08001338 __cleanup_sighand(p->sighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339bad_fork_cleanup_fs:
1340 exit_fs(p); /* blocking */
1341bad_fork_cleanup_files:
1342 exit_files(p); /* blocking */
1343bad_fork_cleanup_semundo:
1344 exit_sem(p);
1345bad_fork_cleanup_audit:
1346 audit_free(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347bad_fork_cleanup_policy:
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001348 perf_event_free_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349#ifdef CONFIG_NUMA
Lee Schermerhornf0be3d32008-04-28 02:13:08 -07001350 mpol_put(p->mempolicy);
Paul Menageb4f48b62007-10-18 23:39:33 -07001351bad_fork_cleanup_cgroup:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352#endif
Paul Menageb4f48b62007-10-18 23:39:33 -07001353 cgroup_exit(p, cgroup_callbacks_done);
Shailabh Nagar35df17c2006-08-31 21:27:38 -07001354 delayacct_tsk_free(p);
Al Viroa1261f52005-11-13 16:06:55 -08001355 module_put(task_thread_info(p)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356bad_fork_cleanup_count:
David Howellsd84f4f92008-11-14 10:39:23 +11001357 atomic_dec(&p->cred->user->processes);
David Howellse0e81732009-09-02 09:13:40 +01001358 exit_creds(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359bad_fork_free:
1360 free_task(p);
Oleg Nesterovfe7d37d2006-01-08 01:04:02 -08001361fork_out:
1362 return ERR_PTR(retval);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363}
1364
Adrian Bunk6b2fb3c2008-02-06 01:37:55 -08001365noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366{
1367 memset(regs, 0, sizeof(struct pt_regs));
1368 return regs;
1369}
1370
Oleg Nesterovf106eee2010-05-26 14:44:11 -07001371static inline void init_idle_pids(struct pid_link *links)
1372{
1373 enum pid_type type;
1374
1375 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1376 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1377 links[type].pid = &init_struct_pid;
1378 }
1379}
1380
Al Viro9abcf402007-02-01 13:52:48 +00001381struct task_struct * __cpuinit fork_idle(int cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382{
Ingo Molnar36c8b582006-07-03 00:25:41 -07001383 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 struct pt_regs regs;
1385
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001386 task = copy_process(CLONE_VM, 0, idle_regs(&regs), 0, NULL,
Roland McGrath09a05392008-07-25 19:45:47 -07001387 &init_struct_pid, 0);
Oleg Nesterovf106eee2010-05-26 14:44:11 -07001388 if (!IS_ERR(task)) {
1389 init_idle_pids(task->pids);
Akinobu Mita753ca4f2006-11-25 11:09:34 -08001390 init_idle(task, cpu);
Oleg Nesterovf106eee2010-05-26 14:44:11 -07001391 }
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001392
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393 return task;
1394}
1395
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396/*
1397 * Ok, this is the main fork-routine.
1398 *
1399 * It copies the process, and if successful kick-starts
1400 * it and waits for it to finish using the VM if required.
1401 */
1402long do_fork(unsigned long clone_flags,
1403 unsigned long stack_start,
1404 struct pt_regs *regs,
1405 unsigned long stack_size,
1406 int __user *parent_tidptr,
1407 int __user *child_tidptr)
1408{
1409 struct task_struct *p;
1410 int trace = 0;
Eric W. Biederman92476d72006-03-31 02:31:42 -08001411 long nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412
Andrew Mortonbdff7462008-02-04 22:27:22 -08001413 /*
Serge Hallyn18b6e042008-10-15 16:38:45 -05001414 * Do some preliminary argument and permissions checking before we
1415 * actually start allocating stuff
1416 */
1417 if (clone_flags & CLONE_NEWUSER) {
1418 if (clone_flags & CLONE_THREAD)
1419 return -EINVAL;
1420 /* hopefully this check will go away when userns support is
1421 * complete
1422 */
Serge E. Hallyn7657d902008-12-03 13:17:33 -06001423 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1424 !capable(CAP_SETGID))
Serge Hallyn18b6e042008-10-15 16:38:45 -05001425 return -EPERM;
1426 }
1427
1428 /*
Roland McGrath09a05392008-07-25 19:45:47 -07001429 * When called from kernel_thread, don't do user tracing stuff.
1430 */
1431 if (likely(user_mode(regs)))
1432 trace = tracehook_prepare_clone(clone_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433
Sukadev Bhattiprolua6f5e062007-10-18 23:39:53 -07001434 p = copy_process(clone_flags, stack_start, regs, stack_size,
Roland McGrath09a05392008-07-25 19:45:47 -07001435 child_tidptr, NULL, trace);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436 /*
1437 * Do this prior waking up the new thread - the thread pointer
1438 * might get invalid after that point, if the thread exits quickly.
1439 */
1440 if (!IS_ERR(p)) {
1441 struct completion vfork;
1442
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04001443 trace_sched_process_fork(current, p);
1444
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001445 nr = task_pid_vnr(p);
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001446
1447 if (clone_flags & CLONE_PARENT_SETTID)
1448 put_user(nr, parent_tidptr);
Sukadev Bhattiprolua6f5e062007-10-18 23:39:53 -07001449
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450 if (clone_flags & CLONE_VFORK) {
1451 p->vfork_done = &vfork;
1452 init_completion(&vfork);
1453 }
1454
Al Viroa64e6492008-11-12 18:37:41 -05001455 audit_finish_fork(p);
Oleg Nesterov087eb432009-06-04 16:29:07 -07001456 tracehook_report_clone(regs, clone_flags, nr, p);
Roland McGrath09a05392008-07-25 19:45:47 -07001457
1458 /*
1459 * We set PF_STARTING at creation in case tracing wants to
1460 * use this to distinguish a fully live task from one that
1461 * hasn't gotten to tracehook_report_clone() yet. Now we
1462 * clear it and set the child going.
1463 */
1464 p->flags &= ~PF_STARTING;
1465
Dave Jones43bb40c2011-01-13 15:45:40 -08001466 wake_up_new_task(p, clone_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467
Roland McGrath09a05392008-07-25 19:45:47 -07001468 tracehook_report_clone_complete(trace, regs,
1469 clone_flags, nr, p);
1470
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 if (clone_flags & CLONE_VFORK) {
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -07001472 freezer_do_not_count();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 wait_for_completion(&vfork);
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -07001474 freezer_count();
Roland McGrathdaded342008-07-25 19:45:47 -07001475 tracehook_report_vfork_done(p, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 }
1477 } else {
Eric W. Biederman92476d72006-03-31 02:31:42 -08001478 nr = PTR_ERR(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 }
Eric W. Biederman92476d72006-03-31 02:31:42 -08001480 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481}
1482
Ravikiran G Thirumalai5fd63b32006-01-11 22:46:15 +01001483#ifndef ARCH_MIN_MMSTRUCT_ALIGN
1484#define ARCH_MIN_MMSTRUCT_ALIGN 0
1485#endif
1486
Alexey Dobriyan51cc5062008-07-25 19:45:34 -07001487static void sighand_ctor(void *data)
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08001488{
1489 struct sighand_struct *sighand = data;
1490
Christoph Lametera35afb82007-05-16 22:10:57 -07001491 spin_lock_init(&sighand->siglock);
Davide Libenzib8fceee2007-09-20 12:40:16 -07001492 init_waitqueue_head(&sighand->signalfd_wqh);
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08001493}
1494
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495void __init proc_caches_init(void)
1496{
1497 sighand_cachep = kmem_cache_create("sighand_cache",
1498 sizeof(struct sighand_struct), 0,
Vegard Nossum2dff4402008-05-31 15:56:17 +02001499 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1500 SLAB_NOTRACK, sighand_ctor);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 signal_cachep = kmem_cache_create("signal_cache",
1502 sizeof(struct signal_struct), 0,
Vegard Nossum2dff4402008-05-31 15:56:17 +02001503 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
Paul Mundt20c2df82007-07-20 10:11:58 +09001504 files_cachep = kmem_cache_create("files_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 sizeof(struct files_struct), 0,
Vegard Nossum2dff4402008-05-31 15:56:17 +02001506 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
Paul Mundt20c2df82007-07-20 10:11:58 +09001507 fs_cachep = kmem_cache_create("fs_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508 sizeof(struct fs_struct), 0,
Vegard Nossum2dff4402008-05-31 15:56:17 +02001509 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 mm_cachep = kmem_cache_create("mm_struct",
Ravikiran G Thirumalai5fd63b32006-01-11 22:46:15 +01001511 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
Vegard Nossum2dff4402008-05-31 15:56:17 +02001512 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
David Howells33e5d7692009-04-02 16:56:32 -07001513 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
David Howells8feae132009-01-08 12:04:47 +00001514 mmap_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515}
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001516
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001517/*
1518 * Check constraints on flags passed to the unshare system call and
1519 * force unsharing of additional process context as appropriate.
1520 */
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001521static void check_unshare_flags(unsigned long *flags_ptr)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001522{
1523 /*
1524 * If unsharing a thread from a thread group, must also
1525 * unshare vm.
1526 */
1527 if (*flags_ptr & CLONE_THREAD)
1528 *flags_ptr |= CLONE_VM;
1529
1530 /*
1531 * If unsharing vm, must also unshare signal handlers.
1532 */
1533 if (*flags_ptr & CLONE_VM)
1534 *flags_ptr |= CLONE_SIGHAND;
1535
1536 /*
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001537 * If unsharing namespace, must also unshare filesystem information.
1538 */
1539 if (*flags_ptr & CLONE_NEWNS)
1540 *flags_ptr |= CLONE_FS;
1541}
1542
1543/*
1544 * Unsharing of tasks created with CLONE_THREAD is not supported yet
1545 */
1546static int unshare_thread(unsigned long unshare_flags)
1547{
1548 if (unshare_flags & CLONE_THREAD)
1549 return -EINVAL;
1550
1551 return 0;
1552}
1553
1554/*
JANAK DESAI99d14192006-02-07 12:58:59 -08001555 * Unshare the filesystem structure if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001556 */
1557static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1558{
1559 struct fs_struct *fs = current->fs;
1560
Al Viro498052b2009-03-30 07:20:30 -04001561 if (!(unshare_flags & CLONE_FS) || !fs)
1562 return 0;
1563
1564 /* don't need lock here; in the worst case we'll do useless copy */
1565 if (fs->users == 1)
1566 return 0;
1567
1568 *new_fsp = copy_fs_struct(fs);
1569 if (!*new_fsp)
1570 return -ENOMEM;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001571
1572 return 0;
1573}
1574
1575/*
Oleg Nesterovdae3c5a2006-12-08 02:36:09 -08001576 * Unsharing of sighand is not supported yet
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001577 */
1578static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp)
1579{
1580 struct sighand_struct *sigh = current->sighand;
1581
Oleg Nesterovdae3c5a2006-12-08 02:36:09 -08001582 if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001583 return -EINVAL;
1584 else
1585 return 0;
1586}
1587
1588/*
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001589 * Unshare vm if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001590 */
1591static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp)
1592{
1593 struct mm_struct *mm = current->mm;
1594
1595 if ((unshare_flags & CLONE_VM) &&
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001596 (mm && atomic_read(&mm->mm_users) > 1)) {
Oleg Nesterov2d61b862006-03-18 20:41:10 +03001597 return -EINVAL;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001598 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001599
1600 return 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001601}
1602
1603/*
JANAK DESAIa016f332006-02-07 12:59:02 -08001604 * Unshare file descriptor table if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001605 */
1606static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1607{
1608 struct files_struct *fd = current->files;
JANAK DESAIa016f332006-02-07 12:59:02 -08001609 int error = 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001610
1611 if ((unshare_flags & CLONE_FILES) &&
JANAK DESAIa016f332006-02-07 12:59:02 -08001612 (fd && atomic_read(&fd->count) > 1)) {
1613 *new_fdp = dup_fd(fd, &error);
1614 if (!*new_fdp)
1615 return error;
1616 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001617
1618 return 0;
1619}
1620
1621/*
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001622 * unshare allows a process to 'unshare' part of the process
1623 * context which was originally shared using clone. copy_*
1624 * functions used by do_fork() cannot be used here directly
1625 * because they modify an inactive task_struct that is being
1626 * constructed. Here we are modifying the current, active,
1627 * task_struct.
1628 */
Heiko Carstens6559eed82009-01-14 14:14:32 +01001629SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001630{
1631 int err = 0;
1632 struct fs_struct *fs, *new_fs = NULL;
Oleg Nesterovdae3c5a2006-12-08 02:36:09 -08001633 struct sighand_struct *new_sigh = NULL;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001634 struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL;
1635 struct files_struct *fd, *new_fd = NULL;
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07001636 struct nsproxy *new_nsproxy = NULL;
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001637 int do_sysvsem = 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001638
1639 check_unshare_flags(&unshare_flags);
1640
Eric W. Biederman06f9d4f2006-03-22 00:07:40 -08001641 /* Return -EINVAL for all unsupported flags */
1642 err = -EINVAL;
1643 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
Kirill Korotaev25b21cb2006-10-02 02:18:19 -07001644 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
Serge Hallyn18b6e042008-10-15 16:38:45 -05001645 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
Eric W. Biederman06f9d4f2006-03-22 00:07:40 -08001646 goto bad_unshare_out;
1647
Manfred Spraul6013f672008-04-29 01:00:59 -07001648 /*
1649 * CLONE_NEWIPC must also detach from the undolist: after switching
1650 * to a new ipc namespace, the semaphore arrays from the old
1651 * namespace are unreachable.
1652 */
1653 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001654 do_sysvsem = 1;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001655 if ((err = unshare_thread(unshare_flags)))
1656 goto bad_unshare_out;
1657 if ((err = unshare_fs(unshare_flags, &new_fs)))
1658 goto bad_unshare_cleanup_thread;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001659 if ((err = unshare_sighand(unshare_flags, &new_sigh)))
Badari Pulavartye3222c42007-05-08 00:25:21 -07001660 goto bad_unshare_cleanup_fs;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001661 if ((err = unshare_vm(unshare_flags, &new_mm)))
1662 goto bad_unshare_cleanup_sigh;
1663 if ((err = unshare_fd(unshare_flags, &new_fd)))
1664 goto bad_unshare_cleanup_vm;
Badari Pulavartye3222c42007-05-08 00:25:21 -07001665 if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
1666 new_fs)))
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001667 goto bad_unshare_cleanup_fd;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001668
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001669 if (new_fs || new_mm || new_fd || do_sysvsem || new_nsproxy) {
1670 if (do_sysvsem) {
1671 /*
1672 * CLONE_SYSVSEM is equivalent to sys_exit().
1673 */
1674 exit_sem(current);
1675 }
Serge E. Hallynab516012006-10-02 02:18:06 -07001676
Serge Hallync0b2fc32006-10-02 02:18:18 -07001677 if (new_nsproxy) {
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07001678 switch_task_namespaces(current, new_nsproxy);
1679 new_nsproxy = NULL;
Serge Hallync0b2fc32006-10-02 02:18:18 -07001680 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001681
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07001682 task_lock(current);
1683
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001684 if (new_fs) {
1685 fs = current->fs;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001686 spin_lock(&fs->lock);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001687 current->fs = new_fs;
Al Viro498052b2009-03-30 07:20:30 -04001688 if (--fs->users)
1689 new_fs = NULL;
1690 else
1691 new_fs = fs;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001692 spin_unlock(&fs->lock);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001693 }
1694
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001695 if (new_mm) {
1696 mm = current->mm;
1697 active_mm = current->active_mm;
1698 current->mm = new_mm;
1699 current->active_mm = new_mm;
Ying Han3d5992d2010-10-26 14:21:23 -07001700 if (current->signal->oom_score_adj == OOM_SCORE_ADJ_MIN) {
1701 atomic_dec(&mm->oom_disable_count);
1702 atomic_inc(&new_mm->oom_disable_count);
1703 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001704 activate_mm(active_mm, new_mm);
1705 new_mm = mm;
1706 }
1707
1708 if (new_fd) {
1709 fd = current->files;
1710 current->files = new_fd;
1711 new_fd = fd;
1712 }
1713
1714 task_unlock(current);
1715 }
1716
Serge Hallync0b2fc32006-10-02 02:18:18 -07001717 if (new_nsproxy)
Linus Torvalds444f3782007-01-30 13:35:18 -08001718 put_nsproxy(new_nsproxy);
Serge Hallync0b2fc32006-10-02 02:18:18 -07001719
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001720bad_unshare_cleanup_fd:
1721 if (new_fd)
1722 put_files_struct(new_fd);
1723
1724bad_unshare_cleanup_vm:
1725 if (new_mm)
1726 mmput(new_mm);
1727
1728bad_unshare_cleanup_sigh:
1729 if (new_sigh)
1730 if (atomic_dec_and_test(&new_sigh->count))
1731 kmem_cache_free(sighand_cachep, new_sigh);
1732
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001733bad_unshare_cleanup_fs:
1734 if (new_fs)
Al Viro498052b2009-03-30 07:20:30 -04001735 free_fs_struct(new_fs);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001736
1737bad_unshare_cleanup_thread:
1738bad_unshare_out:
1739 return err;
1740}
Al Viro3b125382008-04-22 05:31:30 -04001741
1742/*
1743 * Helper to unshare the files of the current task.
1744 * We don't want to expose copy_files internals to
1745 * the exec layer of the kernel.
1746 */
1747
1748int unshare_files(struct files_struct **displaced)
1749{
1750 struct task_struct *task = current;
Al Viro50704512008-04-26 05:25:00 +01001751 struct files_struct *copy = NULL;
Al Viro3b125382008-04-22 05:31:30 -04001752 int error;
1753
1754 error = unshare_fd(CLONE_FILES, &copy);
1755 if (error || !copy) {
1756 *displaced = NULL;
1757 return error;
1758 }
1759 *displaced = task->files;
1760 task_lock(task);
1761 task->files = copy;
1762 task_unlock(task);
1763 return 0;
1764}