blob: 4854c2c4a82eae9ed1032ee1b66abc46124499ad [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>
Kirill Korotaev6b3286e2006-12-08 02:37:56 -080020#include <linux/mnt_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/personality.h>
22#include <linux/mempolicy.h>
23#include <linux/sem.h>
24#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040025#include <linux/fdtable.h>
Jens Axboeda9cbc82008-06-30 20:42:08 +020026#include <linux/iocontext.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/key.h>
28#include <linux/binfmts.h>
29#include <linux/mman.h>
Andrea Arcangelicddb8a52008-07-28 15:46:29 -070030#include <linux/mmu_notifier.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <linux/fs.h>
Serge E. Hallynab516012006-10-02 02:18:06 -070032#include <linux/nsproxy.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080033#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#include <linux/cpu.h>
Paul Menageb4f48b62007-10-18 23:39:33 -070035#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#include <linux/security.h>
Mel Gormana1e78772008-07-23 21:27:23 -070037#include <linux/hugetlb.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038#include <linux/swap.h>
39#include <linux/syscalls.h>
40#include <linux/jiffies.h>
Roland McGrath09a05392008-07-25 19:45:47 -070041#include <linux/tracehook.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042#include <linux/futex.h>
Linus Torvalds8141c7f2008-11-15 10:20:36 -080043#include <linux/compat.h>
Andrew Morton7c3ab7382006-12-10 02:19:19 -080044#include <linux/task_io_accounting_ops.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070045#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070046#include <linux/ptrace.h>
47#include <linux/mount.h>
48#include <linux/audit.h>
Pavel Emelianov78fb7462008-02-07 00:13:51 -080049#include <linux/memcontrol.h>
Frederic Weisbeckerf201ae22008-11-23 06:22:56 +010050#include <linux/ftrace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070051#include <linux/profile.h>
52#include <linux/rmap.h>
53#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 Ven0a4254052006-09-26 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>
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -040063#include <trace/sched.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070064
65#include <asm/pgtable.h>
66#include <asm/pgalloc.h>
67#include <asm/uaccess.h>
68#include <asm/mmu_context.h>
69#include <asm/cacheflush.h>
70#include <asm/tlbflush.h>
71
72/*
73 * Protected counters by write_lock_irq(&tasklist_lock)
74 */
75unsigned long total_forks; /* Handle normal Linux uptimes. */
76int nr_threads; /* The idle threads do not count.. */
77
78int max_threads; /* tunable limit on nr_threads */
79
80DEFINE_PER_CPU(unsigned long, process_counts) = 0;
81
Christoph Hellwigc59923a2006-07-10 04:45:40 -070082__cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
Linus Torvalds1da177e2005-04-16 15:20:36 -070083
Mathieu Desnoyers7e066fb2008-11-14 17:47:47 -050084DEFINE_TRACE(sched_process_fork);
85
Linus Torvalds1da177e2005-04-16 15:20:36 -070086int nr_processes(void)
87{
88 int cpu;
89 int total = 0;
90
91 for_each_online_cpu(cpu)
92 total += per_cpu(process_counts, cpu);
93
94 return total;
95}
96
97#ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
98# define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
99# define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
Christoph Lametere18b8902006-12-06 20:33:20 -0800100static struct kmem_cache *task_struct_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101#endif
102
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700103#ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
104static inline struct thread_info *alloc_thread_info(struct task_struct *tsk)
105{
106#ifdef CONFIG_DEBUG_STACK_USAGE
107 gfp_t mask = GFP_KERNEL | __GFP_ZERO;
108#else
109 gfp_t mask = GFP_KERNEL;
110#endif
111 return (struct thread_info *)__get_free_pages(mask, THREAD_SIZE_ORDER);
112}
113
114static inline void free_thread_info(struct thread_info *ti)
115{
116 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
117}
118#endif
119
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120/* SLAB cache for signal_struct structures (tsk->signal) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800121static struct kmem_cache *signal_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122
123/* SLAB cache for sighand_struct structures (tsk->sighand) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800124struct kmem_cache *sighand_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125
126/* SLAB cache for files_struct structures (tsk->files) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800127struct kmem_cache *files_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128
129/* SLAB cache for fs_struct structures (tsk->fs) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800130struct kmem_cache *fs_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131
132/* SLAB cache for vm_area_struct structures */
Christoph Lametere18b8902006-12-06 20:33:20 -0800133struct kmem_cache *vm_area_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134
135/* SLAB cache for mm_struct structures (tsk->mm) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800136static struct kmem_cache *mm_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137
138void free_task(struct task_struct *tsk)
139{
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700140 prop_local_destroy_single(&tsk->dirties);
Roman Zippelf7e42172007-05-09 02:35:17 -0700141 free_thread_info(tsk->stack);
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700142 rt_mutex_debug_task_free(tsk);
Frederic Weisbeckerfb526072008-11-25 21:07:04 +0100143 ftrace_graph_exit_task(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 free_task_struct(tsk);
145}
146EXPORT_SYMBOL(free_task);
147
Andrew Morton158d9eb2006-03-31 02:31:34 -0800148void __put_task_struct(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149{
Eugene Teo270f7222007-10-18 23:40:38 -0700150 WARN_ON(!tsk->exit_state);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 WARN_ON(atomic_read(&tsk->usage));
152 WARN_ON(tsk == current);
153
David Howells3b11a1d2008-11-14 10:39:26 +1100154 put_cred(tsk->real_cred);
David Howellsf1752ee2008-11-14 10:39:17 +1100155 put_cred(tsk->cred);
Shailabh Nagar35df17c2006-08-31 21:27:38 -0700156 delayacct_tsk_free(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157
158 if (!profile_handoff_task(tsk))
159 free_task(tsk);
160}
161
Suresh Siddha2adee9b2008-04-16 10:25:35 +0200162/*
163 * macro override instead of weak attribute alias, to workaround
164 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
165 */
166#ifndef arch_task_cache_init
167#define arch_task_cache_init()
168#endif
Suresh Siddha61c46282008-03-10 15:28:04 -0700169
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170void __init fork_init(unsigned long mempages)
171{
172#ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
173#ifndef ARCH_MIN_TASKALIGN
174#define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
175#endif
176 /* create a slab on which task_structs can be allocated */
177 task_struct_cachep =
178 kmem_cache_create("task_struct", sizeof(struct task_struct),
Paul Mundt20c2df82007-07-20 10:11:58 +0900179 ARCH_MIN_TASKALIGN, SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180#endif
181
Suresh Siddha61c46282008-03-10 15:28:04 -0700182 /* do the arch specific task caches init */
183 arch_task_cache_init();
184
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 /*
186 * The default maximum number of threads is set to a safe
187 * value: the thread structures can take up at most half
188 * of memory.
189 */
190 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
191
192 /*
193 * we need to allow at least 20 threads to boot a system
194 */
195 if(max_threads < 20)
196 max_threads = 20;
197
198 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
199 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
200 init_task.signal->rlim[RLIMIT_SIGPENDING] =
201 init_task.signal->rlim[RLIMIT_NPROC];
202}
203
Suresh Siddha61c46282008-03-10 15:28:04 -0700204int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
205 struct task_struct *src)
206{
207 *dst = *src;
208 return 0;
209}
210
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211static struct task_struct *dup_task_struct(struct task_struct *orig)
212{
213 struct task_struct *tsk;
214 struct thread_info *ti;
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700215 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216
217 prepare_to_copy(orig);
218
219 tsk = alloc_task_struct();
220 if (!tsk)
221 return NULL;
222
223 ti = alloc_thread_info(tsk);
224 if (!ti) {
225 free_task_struct(tsk);
226 return NULL;
227 }
228
Suresh Siddha61c46282008-03-10 15:28:04 -0700229 err = arch_dup_task_struct(tsk, orig);
230 if (err)
231 goto out;
232
Roman Zippelf7e42172007-05-09 02:35:17 -0700233 tsk->stack = ti;
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700234
235 err = prop_local_init_single(&tsk->dirties);
Suresh Siddha61c46282008-03-10 15:28:04 -0700236 if (err)
237 goto out;
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700238
Al Viro10ebffd2005-11-13 16:06:56 -0800239 setup_thread_stack(tsk, orig);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240
Arjan van de Ven0a4254052006-09-26 10:52:38 +0200241#ifdef CONFIG_CC_STACKPROTECTOR
242 tsk->stack_canary = get_random_int();
243#endif
244
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 /* One for us, one for whoever does the "release_task()" (usually parent) */
246 atomic_set(&tsk->usage,2);
Giancarlo Formicuccia4b5d37a2005-09-09 13:01:22 -0700247 atomic_set(&tsk->fs_excl, 0);
Alexey Dobriyan6c5c9342006-09-29 01:59:40 -0700248#ifdef CONFIG_BLK_DEV_IO_TRACE
Jens Axboe2056a782006-03-23 20:00:26 +0100249 tsk->btrace_seq = 0;
Alexey Dobriyan6c5c9342006-09-29 01:59:40 -0700250#endif
Jens Axboea0aa7f62006-04-20 13:05:33 +0200251 tsk->splice_pipe = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 return tsk;
Suresh Siddha61c46282008-03-10 15:28:04 -0700253
254out:
255 free_thread_info(ti);
256 free_task_struct(tsk);
257 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258}
259
260#ifdef CONFIG_MMU
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700261static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262{
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700263 struct vm_area_struct *mpnt, *tmp, **pprev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 struct rb_node **rb_link, *rb_parent;
265 int retval;
266 unsigned long charge;
267 struct mempolicy *pol;
268
269 down_write(&oldmm->mmap_sem);
Ralf Baechleec8c0442006-12-12 17:14:57 +0000270 flush_cache_dup_mm(oldmm);
Ingo Molnarad339452006-07-03 00:25:15 -0700271 /*
272 * Not linked in yet - no deadlock potential:
273 */
274 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
Hugh Dickins7ee78232005-10-29 18:16:08 -0700275
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276 mm->locked_vm = 0;
277 mm->mmap = NULL;
278 mm->mmap_cache = NULL;
279 mm->free_area_cache = oldmm->mmap_base;
Wolfgang Wander1363c3c2005-06-21 17:14:49 -0700280 mm->cached_hole_size = ~0UL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 mm->map_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 cpus_clear(mm->cpu_vm_mask);
283 mm->mm_rb = RB_ROOT;
284 rb_link = &mm->mm_rb.rb_node;
285 rb_parent = NULL;
286 pprev = &mm->mmap;
287
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700288 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289 struct file *file;
290
291 if (mpnt->vm_flags & VM_DONTCOPY) {
Hugh Dickins3b6bfcd2005-07-12 13:58:09 -0700292 long pages = vma_pages(mpnt);
293 mm->total_vm -= pages;
Hugh Dickinsab50b8e2005-10-29 18:15:56 -0700294 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
Hugh Dickins3b6bfcd2005-07-12 13:58:09 -0700295 -pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296 continue;
297 }
298 charge = 0;
299 if (mpnt->vm_flags & VM_ACCOUNT) {
300 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
301 if (security_vm_enough_memory(len))
302 goto fail_nomem;
303 charge = len;
304 }
Christoph Lametere94b1762006-12-06 20:33:17 -0800305 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306 if (!tmp)
307 goto fail_nomem;
308 *tmp = *mpnt;
Lee Schermerhorn846a16b2008-04-28 02:13:09 -0700309 pol = mpol_dup(vma_policy(mpnt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 retval = PTR_ERR(pol);
311 if (IS_ERR(pol))
312 goto fail_nomem_policy;
313 vma_set_policy(tmp, pol);
314 tmp->vm_flags &= ~VM_LOCKED;
315 tmp->vm_mm = mm;
316 tmp->vm_next = NULL;
317 anon_vma_link(tmp);
318 file = tmp->vm_file;
319 if (file) {
Josef "Jeff" Sipekf3a43f32006-12-08 02:36:43 -0800320 struct inode *inode = file->f_path.dentry->d_inode;
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000321 struct address_space *mapping = file->f_mapping;
322
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 get_file(file);
324 if (tmp->vm_flags & VM_DENYWRITE)
325 atomic_dec(&inode->i_writecount);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000326 spin_lock(&mapping->i_mmap_lock);
327 if (tmp->vm_flags & VM_SHARED)
328 mapping->i_mmap_writable++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329 tmp->vm_truncate_count = mpnt->vm_truncate_count;
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000330 flush_dcache_mmap_lock(mapping);
331 /* insert tmp into the share list, just after mpnt */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 vma_prio_tree_add(tmp, mpnt);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000333 flush_dcache_mmap_unlock(mapping);
334 spin_unlock(&mapping->i_mmap_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 }
336
337 /*
Mel Gormana1e78772008-07-23 21:27:23 -0700338 * Clear hugetlb-related page reserves for children. This only
339 * affects MAP_PRIVATE mappings. Faults generated by the child
340 * are not guaranteed to succeed, even if read-only
341 */
342 if (is_vm_hugetlb_page(tmp))
343 reset_vma_resv_huge_pages(tmp);
344
345 /*
Hugh Dickins7ee78232005-10-29 18:16:08 -0700346 * Link in the new vma and copy the page table entries.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 *pprev = tmp;
349 pprev = &tmp->vm_next;
350
351 __vma_link_rb(mm, tmp, rb_link, rb_parent);
352 rb_link = &tmp->vm_rb.rb_right;
353 rb_parent = &tmp->vm_rb;
354
355 mm->map_count++;
Hugh Dickins0b0db142005-11-21 21:32:20 -0800356 retval = copy_page_range(mm, oldmm, mpnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357
358 if (tmp->vm_ops && tmp->vm_ops->open)
359 tmp->vm_ops->open(tmp);
360
361 if (retval)
362 goto out;
363 }
Jeremy Fitzhardinged6dd61c2007-05-02 19:27:14 +0200364 /* a new mm has just been created */
365 arch_dup_mmap(oldmm, mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367out:
Hugh Dickins7ee78232005-10-29 18:16:08 -0700368 up_write(&mm->mmap_sem);
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700369 flush_tlb_mm(oldmm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 up_write(&oldmm->mmap_sem);
371 return retval;
372fail_nomem_policy:
373 kmem_cache_free(vm_area_cachep, tmp);
374fail_nomem:
375 retval = -ENOMEM;
376 vm_unacct_memory(charge);
377 goto out;
378}
379
380static inline int mm_alloc_pgd(struct mm_struct * mm)
381{
382 mm->pgd = pgd_alloc(mm);
383 if (unlikely(!mm->pgd))
384 return -ENOMEM;
385 return 0;
386}
387
388static inline void mm_free_pgd(struct mm_struct * mm)
389{
Benjamin Herrenschmidt5e541972008-02-04 22:29:14 -0800390 pgd_free(mm, mm->pgd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391}
392#else
393#define dup_mmap(mm, oldmm) (0)
394#define mm_alloc_pgd(mm) (0)
395#define mm_free_pgd(mm)
396#endif /* CONFIG_MMU */
397
Daniel Walker23ff4442007-10-18 03:06:07 -0700398__cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399
Christoph Lametere94b1762006-12-06 20:33:17 -0800400#define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401#define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
402
Hidehiro Kawai4cb0e112009-01-06 14:42:47 -0800403static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
404
405static int __init coredump_filter_setup(char *s)
406{
407 default_dump_filter =
408 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
409 MMF_DUMP_FILTER_MASK;
410 return 1;
411}
412
413__setup("coredump_filter=", coredump_filter_setup);
414
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415#include <linux/init_task.h>
416
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800417static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418{
419 atomic_set(&mm->mm_users, 1);
420 atomic_set(&mm->mm_count, 1);
421 init_rwsem(&mm->mmap_sem);
422 INIT_LIST_HEAD(&mm->mmlist);
Hidehiro Kawai4cb0e112009-01-06 14:42:47 -0800423 mm->flags = (current->mm) ? current->mm->flags : default_dump_filter;
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700424 mm->core_state = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 mm->nr_ptes = 0;
Hugh Dickins42946212005-10-29 18:16:05 -0700426 set_mm_counter(mm, file_rss, 0);
Hugh Dickins404351e2005-10-29 18:16:04 -0700427 set_mm_counter(mm, anon_rss, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 spin_lock_init(&mm->page_table_lock);
Jens Axboeabf137d2008-12-09 08:11:22 +0100429 spin_lock_init(&mm->ioctx_lock);
430 INIT_HLIST_HEAD(&mm->ioctx_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 mm->free_area_cache = TASK_UNMAPPED_BASE;
Wolfgang Wander1363c3c2005-06-21 17:14:49 -0700432 mm->cached_hole_size = ~0UL;
Balbir Singhcf475ad2008-04-29 01:00:16 -0700433 mm_init_owner(mm, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434
435 if (likely(!mm_alloc_pgd(mm))) {
436 mm->def_flags = 0;
Andrea Arcangelicddb8a52008-07-28 15:46:29 -0700437 mmu_notifier_mm_init(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 return mm;
439 }
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800440
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 free_mm(mm);
442 return NULL;
443}
444
445/*
446 * Allocate and initialize an mm_struct.
447 */
448struct mm_struct * mm_alloc(void)
449{
450 struct mm_struct * mm;
451
452 mm = allocate_mm();
453 if (mm) {
454 memset(mm, 0, sizeof(*mm));
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800455 mm = mm_init(mm, current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 }
457 return mm;
458}
459
460/*
461 * Called when the last reference to the mm
462 * is dropped: either by a lazy thread or by
463 * mmput. Free the page directory and the mm.
464 */
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800465void __mmdrop(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466{
467 BUG_ON(mm == &init_mm);
468 mm_free_pgd(mm);
469 destroy_context(mm);
Andrea Arcangelicddb8a52008-07-28 15:46:29 -0700470 mmu_notifier_mm_destroy(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 free_mm(mm);
472}
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200473EXPORT_SYMBOL_GPL(__mmdrop);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474
475/*
476 * Decrement the use count and release all resources for an mm.
477 */
478void mmput(struct mm_struct *mm)
479{
Andrew Morton0ae26f12006-06-23 02:05:15 -0700480 might_sleep();
481
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 if (atomic_dec_and_test(&mm->mm_users)) {
483 exit_aio(mm);
484 exit_mmap(mm);
Matt Helsley925d1c42008-04-29 01:01:36 -0700485 set_mm_exe_file(mm, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 if (!list_empty(&mm->mmlist)) {
487 spin_lock(&mmlist_lock);
488 list_del(&mm->mmlist);
489 spin_unlock(&mmlist_lock);
490 }
491 put_swap_token(mm);
492 mmdrop(mm);
493 }
494}
495EXPORT_SYMBOL_GPL(mmput);
496
497/**
498 * get_task_mm - acquire a reference to the task's mm
499 *
Oleg Nesterov246bb0b2008-07-25 01:47:38 -0700500 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 * this kernel workthread has transiently adopted a user mm with use_mm,
502 * to do its AIO) is not set and if so returns a reference to it, after
503 * bumping up the use count. User must release the mm via mmput()
504 * after use. Typically used by /proc and ptrace.
505 */
506struct mm_struct *get_task_mm(struct task_struct *task)
507{
508 struct mm_struct *mm;
509
510 task_lock(task);
511 mm = task->mm;
512 if (mm) {
Oleg Nesterov246bb0b2008-07-25 01:47:38 -0700513 if (task->flags & PF_KTHREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 mm = NULL;
515 else
516 atomic_inc(&mm->mm_users);
517 }
518 task_unlock(task);
519 return mm;
520}
521EXPORT_SYMBOL_GPL(get_task_mm);
522
523/* Please note the differences between mmput and mm_release.
524 * mmput is called whenever we stop holding onto a mm_struct,
525 * error success whatever.
526 *
527 * mm_release is called after a mm_struct has been removed
528 * from the current process.
529 *
530 * This difference is important for error handling, when we
531 * only half set up a mm_struct for a new process and need to restore
532 * the old one. Because we mmput the new mm_struct before
533 * restoring the old one. . .
534 * Eric Biederman 10 January 1998
535 */
536void mm_release(struct task_struct *tsk, struct mm_struct *mm)
537{
538 struct completion *vfork_done = tsk->vfork_done;
539
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800540 /* Get rid of any futexes when releasing the mm */
541#ifdef CONFIG_FUTEX
542 if (unlikely(tsk->robust_list))
543 exit_robust_list(tsk);
544#ifdef CONFIG_COMPAT
545 if (unlikely(tsk->compat_robust_list))
546 compat_exit_robust_list(tsk);
547#endif
548#endif
549
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 /* Get rid of any cached register state */
551 deactivate_mm(tsk, mm);
552
553 /* notify parent sleeping on vfork() */
554 if (vfork_done) {
555 tsk->vfork_done = NULL;
556 complete(vfork_done);
557 }
Roland McGrathfec1d012006-12-06 20:36:34 -0800558
559 /*
560 * If we're exiting normally, clear a user-space tid field if
561 * requested. We leave this alone when dying by signal, to leave
562 * the value intact in a core dump, and to save the unnecessary
563 * trouble otherwise. Userland only wants this done for a sys_exit.
564 */
565 if (tsk->clear_child_tid
566 && !(tsk->flags & PF_SIGNALED)
567 && atomic_read(&mm->mm_users) > 1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 u32 __user * tidptr = tsk->clear_child_tid;
569 tsk->clear_child_tid = NULL;
570
571 /*
572 * We don't check the error code - if userspace has
573 * not set up a proper pointer then tough luck.
574 */
575 put_user(0, tidptr);
576 sys_futex(tidptr, FUTEX_WAKE, 1, NULL, NULL, 0);
577 }
578}
579
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800580/*
581 * Allocate a new mm structure and copy contents from the
582 * mm structure of the passed in task structure.
583 */
Carsten Otte402b0862008-03-25 18:47:10 +0100584struct mm_struct *dup_mm(struct task_struct *tsk)
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800585{
586 struct mm_struct *mm, *oldmm = current->mm;
587 int err;
588
589 if (!oldmm)
590 return NULL;
591
592 mm = allocate_mm();
593 if (!mm)
594 goto fail_nomem;
595
596 memcpy(mm, oldmm, sizeof(*mm));
597
Ashwin Chaugule7602bdf2006-12-06 20:31:57 -0800598 /* Initializing for Swap token stuff */
599 mm->token_priority = 0;
600 mm->last_interval = 0;
601
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800602 if (!mm_init(mm, tsk))
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800603 goto fail_nomem;
604
605 if (init_new_context(tsk, mm))
606 goto fail_nocontext;
607
Matt Helsley925d1c42008-04-29 01:01:36 -0700608 dup_mm_exe_file(oldmm, mm);
609
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800610 err = dup_mmap(mm, oldmm);
611 if (err)
612 goto free_pt;
613
614 mm->hiwater_rss = get_mm_rss(mm);
615 mm->hiwater_vm = mm->total_vm;
616
617 return mm;
618
619free_pt:
620 mmput(mm);
621
622fail_nomem:
623 return NULL;
624
625fail_nocontext:
626 /*
627 * If init_new_context() failed, we cannot use mmput() to free the mm
628 * because it calls destroy_context()
629 */
630 mm_free_pgd(mm);
631 free_mm(mm);
632 return NULL;
633}
634
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635static int copy_mm(unsigned long clone_flags, struct task_struct * tsk)
636{
637 struct mm_struct * mm, *oldmm;
638 int retval;
639
640 tsk->min_flt = tsk->maj_flt = 0;
641 tsk->nvcsw = tsk->nivcsw = 0;
642
643 tsk->mm = NULL;
644 tsk->active_mm = NULL;
645
646 /*
647 * Are we cloning a kernel thread?
648 *
649 * We need to steal a active VM for that..
650 */
651 oldmm = current->mm;
652 if (!oldmm)
653 return 0;
654
655 if (clone_flags & CLONE_VM) {
656 atomic_inc(&oldmm->mm_users);
657 mm = oldmm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 goto good_mm;
659 }
660
661 retval = -ENOMEM;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800662 mm = dup_mm(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 if (!mm)
664 goto fail_nomem;
665
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666good_mm:
Ashwin Chaugule7602bdf2006-12-06 20:31:57 -0800667 /* Initializing for Swap token stuff */
668 mm->token_priority = 0;
669 mm->last_interval = 0;
670
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 tsk->mm = mm;
672 tsk->active_mm = mm;
673 return 0;
674
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675fail_nomem:
676 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677}
678
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700679static struct fs_struct *__copy_fs_struct(struct fs_struct *old)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680{
681 struct fs_struct *fs = kmem_cache_alloc(fs_cachep, GFP_KERNEL);
682 /* We don't need to lock fs - think why ;-) */
683 if (fs) {
684 atomic_set(&fs->count, 1);
685 rwlock_init(&fs->lock);
686 fs->umask = old->umask;
687 read_lock(&old->lock);
Jan Blunck6ac08c32008-02-14 19:34:38 -0800688 fs->root = old->root;
689 path_get(&old->root);
690 fs->pwd = old->pwd;
691 path_get(&old->pwd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 read_unlock(&old->lock);
693 }
694 return fs;
695}
696
697struct fs_struct *copy_fs_struct(struct fs_struct *old)
698{
699 return __copy_fs_struct(old);
700}
701
702EXPORT_SYMBOL_GPL(copy_fs_struct);
703
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700704static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705{
706 if (clone_flags & CLONE_FS) {
707 atomic_inc(&current->fs->count);
708 return 0;
709 }
710 tsk->fs = __copy_fs_struct(current->fs);
711 if (!tsk->fs)
712 return -ENOMEM;
713 return 0;
714}
715
JANAK DESAIa016f332006-02-07 12:59:02 -0800716static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
717{
718 struct files_struct *oldf, *newf;
719 int error = 0;
720
721 /*
722 * A background process may not have any files ...
723 */
724 oldf = current->files;
725 if (!oldf)
726 goto out;
727
728 if (clone_flags & CLONE_FILES) {
729 atomic_inc(&oldf->count);
730 goto out;
731 }
732
JANAK DESAIa016f332006-02-07 12:59:02 -0800733 newf = dup_fd(oldf, &error);
734 if (!newf)
735 goto out;
736
737 tsk->files = newf;
738 error = 0;
739out:
740 return error;
741}
742
Jens Axboefadad8782008-01-24 08:54:47 +0100743static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
Jens Axboefd0928d2008-01-24 08:52:45 +0100744{
745#ifdef CONFIG_BLOCK
746 struct io_context *ioc = current->io_context;
747
748 if (!ioc)
749 return 0;
Jens Axboefadad8782008-01-24 08:54:47 +0100750 /*
751 * Share io context with parent, if CLONE_IO is set
752 */
753 if (clone_flags & CLONE_IO) {
754 tsk->io_context = ioc_task_link(ioc);
755 if (unlikely(!tsk->io_context))
756 return -ENOMEM;
757 } else if (ioprio_valid(ioc->ioprio)) {
Jens Axboefd0928d2008-01-24 08:52:45 +0100758 tsk->io_context = alloc_io_context(GFP_KERNEL, -1);
759 if (unlikely(!tsk->io_context))
760 return -ENOMEM;
761
Jens Axboefd0928d2008-01-24 08:52:45 +0100762 tsk->io_context->ioprio = ioc->ioprio;
763 }
764#endif
765 return 0;
766}
767
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700768static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769{
770 struct sighand_struct *sig;
771
Zhaolei60348802009-01-06 14:40:46 -0800772 if (clone_flags & CLONE_SIGHAND) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 atomic_inc(&current->sighand->count);
774 return 0;
775 }
776 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
Ingo Molnare56d0902006-01-08 01:01:37 -0800777 rcu_assign_pointer(tsk->sighand, sig);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 if (!sig)
779 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 atomic_set(&sig->count, 1);
781 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
782 return 0;
783}
784
Oleg Nesterova7e53282006-03-28 16:11:27 -0800785void __cleanup_sighand(struct sighand_struct *sighand)
Oleg Nesterovc81addc2006-03-28 16:11:17 -0800786{
Oleg Nesterovc81addc2006-03-28 16:11:17 -0800787 if (atomic_dec_and_test(&sighand->count))
788 kmem_cache_free(sighand_cachep, sighand);
789}
790
Frank Mayharf06febc2008-09-12 09:54:39 -0700791
792/*
793 * Initialize POSIX timer handling for a thread group.
794 */
795static void posix_cpu_timers_init_group(struct signal_struct *sig)
796{
797 /* Thread group counters. */
798 thread_group_cputime_init(sig);
799
800 /* Expiration times and increments. */
801 sig->it_virt_expires = cputime_zero;
802 sig->it_virt_incr = cputime_zero;
803 sig->it_prof_expires = cputime_zero;
804 sig->it_prof_incr = cputime_zero;
805
806 /* Cached expiration times. */
807 sig->cputime_expires.prof_exp = cputime_zero;
808 sig->cputime_expires.virt_exp = cputime_zero;
809 sig->cputime_expires.sched_exp = 0;
810
811 /* The timer lists. */
812 INIT_LIST_HEAD(&sig->cpu_timers[0]);
813 INIT_LIST_HEAD(&sig->cpu_timers[1]);
814 INIT_LIST_HEAD(&sig->cpu_timers[2]);
815}
816
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700817static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818{
819 struct signal_struct *sig;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820
821 if (clone_flags & CLONE_THREAD) {
Peter Zijlstra490dea42008-11-24 17:06:57 +0100822 atomic_inc(&current->signal->count);
823 atomic_inc(&current->signal->live);
824 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 }
826 sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
Peter Zijlstra490dea42008-11-24 17:06:57 +0100827
828 if (sig)
829 posix_cpu_timers_init_group(sig);
830
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831 tsk->signal = sig;
832 if (!sig)
833 return -ENOMEM;
834
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 atomic_set(&sig->count, 1);
836 atomic_set(&sig->live, 1);
837 init_waitqueue_head(&sig->wait_chldexit);
838 sig->flags = 0;
839 sig->group_exit_code = 0;
840 sig->group_exit_task = NULL;
841 sig->group_stop_count = 0;
Oleg Nesterovdb51aec2008-04-30 00:52:52 -0700842 sig->curr_target = tsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 init_sigpending(&sig->shared_pending);
844 INIT_LIST_HEAD(&sig->posix_timers);
845
Thomas Gleixnerc9cb2e32007-02-16 01:27:49 -0800846 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
Thomas Gleixner2ff678b2006-01-09 20:52:34 -0800847 sig->it_real_incr.tv64 = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 sig->real_timer.function = it_real_fn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 sig->leader = 0; /* session leadership doesn't inherit */
Eric W. Biedermanab521dc2007-02-12 00:53:00 -0800851 sig->tty_old_pgrp = NULL;
Alan Cox9c9f4de2008-10-13 10:37:26 +0100852 sig->tty = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853
Peter Zijlstra32bd6712009-02-05 12:24:15 +0100854 sig->utime = sig->stime = sig->cutime = sig->cstime = cputime_zero;
Laurent Vivier9ac52312007-10-15 17:00:19 +0200855 sig->gtime = cputime_zero;
856 sig->cgtime = cputime_zero;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0;
858 sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -0700859 sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0;
Andrea Righi59954772008-07-27 17:29:15 +0200860 task_io_accounting_init(&sig->ioac);
Peter Zijlstra32bd6712009-02-05 12:24:15 +0100861 sig->sum_sched_runtime = 0;
Shailabh Nagarad4ecbc2006-07-14 00:24:44 -0700862 taskstats_tgid_init(sig);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863
864 task_lock(current->group_leader);
865 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
866 task_unlock(current->group_leader);
867
KaiGai Kohei0e464812006-06-25 05:49:24 -0700868 acct_init_pacct(&sig->pacct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869
Miloslav Trmac522ed772007-07-15 23:40:56 -0700870 tty_audit_fork(sig);
871
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 return 0;
873}
874
Oleg Nesterov6b3934ef2006-03-28 16:11:16 -0800875void __cleanup_signal(struct signal_struct *sig)
876{
Frank Mayharf06febc2008-09-12 09:54:39 -0700877 thread_group_cputime_free(sig);
Alan Cox9c9f4de2008-10-13 10:37:26 +0100878 tty_kref_put(sig->tty);
Oleg Nesterov6b3934ef2006-03-28 16:11:16 -0800879 kmem_cache_free(signal_cachep, sig);
880}
881
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700882static void cleanup_signal(struct task_struct *tsk)
Oleg Nesterov6b3934ef2006-03-28 16:11:16 -0800883{
884 struct signal_struct *sig = tsk->signal;
885
886 atomic_dec(&sig->live);
887
888 if (atomic_dec_and_test(&sig->count))
889 __cleanup_signal(sig);
890}
891
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700892static void copy_flags(unsigned long clone_flags, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893{
894 unsigned long new_flags = p->flags;
895
Rafael J. Wysocki83144182007-07-17 04:03:35 -0700896 new_flags &= ~PF_SUPERPRIV;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 new_flags |= PF_FORKNOEXEC;
Roland McGrath09a05392008-07-25 19:45:47 -0700898 new_flags |= PF_STARTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 p->flags = new_flags;
Rafael J. Wysocki2e131892007-10-18 03:04:45 -0700900 clear_freeze_flag(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901}
902
Heiko Carstens17da2bd2009-01-14 14:14:10 +0100903SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904{
905 current->clear_child_tid = tidptr;
906
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700907 return task_pid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908}
909
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700910static void rt_mutex_init_task(struct task_struct *p)
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700911{
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700912 spin_lock_init(&p->pi_lock);
Zilvinas Valinskase29e1752007-03-16 13:38:34 -0800913#ifdef CONFIG_RT_MUTEXES
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700914 plist_head_init(&p->pi_waiters, &p->pi_lock);
915 p->pi_blocked_on = NULL;
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700916#endif
917}
918
Balbir Singhcf475ad2008-04-29 01:00:16 -0700919#ifdef CONFIG_MM_OWNER
920void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
921{
922 mm->owner = p;
923}
924#endif /* CONFIG_MM_OWNER */
925
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926/*
Frank Mayharf06febc2008-09-12 09:54:39 -0700927 * Initialize POSIX timer handling for a single task.
928 */
929static void posix_cpu_timers_init(struct task_struct *tsk)
930{
931 tsk->cputime_expires.prof_exp = cputime_zero;
932 tsk->cputime_expires.virt_exp = cputime_zero;
933 tsk->cputime_expires.sched_exp = 0;
934 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
935 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
936 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
937}
938
939/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 * This creates a new process as a copy of the old one,
941 * but does not actually start it yet.
942 *
943 * It copies the registers, and all the appropriate
944 * parts of the process environment (as per the clone
945 * flags). The actual kick-off is left to the caller.
946 */
Ingo Molnar36c8b582006-07-03 00:25:41 -0700947static struct task_struct *copy_process(unsigned long clone_flags,
948 unsigned long stack_start,
949 struct pt_regs *regs,
950 unsigned long stack_size,
Ingo Molnar36c8b582006-07-03 00:25:41 -0700951 int __user *child_tidptr,
Roland McGrath09a05392008-07-25 19:45:47 -0700952 struct pid *pid,
953 int trace)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954{
955 int retval;
Mariusz Kozlowskia24efe62007-10-18 23:41:09 -0700956 struct task_struct *p;
Paul Menageb4f48b62007-10-18 23:39:33 -0700957 int cgroup_callbacks_done = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958
959 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
960 return ERR_PTR(-EINVAL);
961
962 /*
963 * Thread groups must share signals as well, and detached threads
964 * can only be started up within the thread group.
965 */
966 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
967 return ERR_PTR(-EINVAL);
968
969 /*
970 * Shared signal handlers imply shared VM. By way of the above,
971 * thread groups also imply shared VM. Blocking this case allows
972 * for various simplifications in other code.
973 */
974 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
975 return ERR_PTR(-EINVAL);
976
977 retval = security_task_create(clone_flags);
978 if (retval)
979 goto fork_out;
980
981 retval = -ENOMEM;
982 p = dup_task_struct(current);
983 if (!p)
984 goto fork_out;
985
Peter Zijlstrabea493a2006-10-17 00:10:33 -0700986 rt_mutex_init_task(p);
987
Ingo Molnard12c1a32008-07-14 12:09:28 +0200988#ifdef CONFIG_PROVE_LOCKING
Ingo Molnarde30a2b2006-07-03 00:24:42 -0700989 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
990 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
991#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 retval = -EAGAIN;
David Howells3b11a1d2008-11-14 10:39:26 +1100993 if (atomic_read(&p->real_cred->user->processes) >=
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 p->signal->rlim[RLIMIT_NPROC].rlim_cur) {
995 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
Serge Hallyn18b6e042008-10-15 16:38:45 -0500996 p->real_cred->user != INIT_USER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 goto bad_fork_free;
998 }
999
David Howellsf1752ee2008-11-14 10:39:17 +11001000 retval = copy_creds(p, clone_flags);
1001 if (retval < 0)
1002 goto bad_fork_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
1004 /*
1005 * If multiple threads are within copy_process(), then this check
1006 * triggers too late. This doesn't hurt, the check is only there
1007 * to stop root fork bombs.
1008 */
Li Zefan04ec93f2009-02-06 08:17:19 +00001009 retval = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 if (nr_threads >= max_threads)
1011 goto bad_fork_cleanup_count;
1012
Al Viroa1261f542005-11-13 16:06:55 -08001013 if (!try_module_get(task_thread_info(p)->exec_domain->module))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 goto bad_fork_cleanup_count;
1015
1016 if (p->binfmt && !try_module_get(p->binfmt->module))
1017 goto bad_fork_cleanup_put_domain;
1018
1019 p->did_exec = 0;
Shailabh Nagarca74e922006-07-14 00:24:36 -07001020 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 copy_flags(clone_flags, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 INIT_LIST_HEAD(&p->children);
1023 INIT_LIST_HEAD(&p->sibling);
Paul E. McKenneye260be62008-01-25 21:08:24 +01001024#ifdef CONFIG_PREEMPT_RCU
1025 p->rcu_read_lock_nesting = 0;
1026 p->rcu_flipctr_idx = 0;
1027#endif /* #ifdef CONFIG_PREEMPT_RCU */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 p->vfork_done = NULL;
1029 spin_lock_init(&p->alloc_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030
1031 clear_tsk_thread_flag(p, TIF_SIGPENDING);
1032 init_sigpending(&p->pending);
1033
1034 p->utime = cputime_zero;
1035 p->stime = cputime_zero;
Laurent Vivier9ac52312007-10-15 17:00:19 +02001036 p->gtime = cputime_zero;
Michael Neulingc66f08b2007-10-18 03:06:34 -07001037 p->utimescaled = cputime_zero;
1038 p->stimescaled = cputime_zero;
Peter Zijlstra73a2bcb2007-10-29 21:18:11 +01001039 p->prev_utime = cputime_zero;
Balbir Singh93018992007-10-30 00:26:32 +01001040 p->prev_stime = cputime_zero;
Balbir Singh172ba842007-07-09 18:52:00 +02001041
Arjan van de Ven69766752008-09-01 15:52:40 -07001042 p->default_timer_slack_ns = current->timer_slack_ns;
1043
Ingo Molnar82a1fcb2008-01-25 21:08:02 +01001044#ifdef CONFIG_DETECT_SOFTLOCKUP
1045 p->last_switch_count = 0;
1046 p->last_switch_timestamp = 0;
1047#endif
1048
Andrea Righi59954772008-07-27 17:29:15 +02001049 task_io_accounting_init(&p->ioac);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 acct_clear_integrals(p);
1051
Frank Mayharf06febc2008-09-12 09:54:39 -07001052 posix_cpu_timers_init(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053
1054 p->lock_depth = -1; /* -1 = no lock */
1055 do_posix_clock_monotonic_gettime(&p->start_time);
Tomas Janousek924b42d2007-07-15 23:39:42 -07001056 p->real_start_time = p->start_time;
1057 monotonic_to_bootbased(&p->real_start_time);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 p->io_context = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059 p->audit_context = NULL;
Paul Menageb4f48b62007-10-18 23:39:33 -07001060 cgroup_fork(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061#ifdef CONFIG_NUMA
Lee Schermerhorn846a16b2008-04-28 02:13:09 -07001062 p->mempolicy = mpol_dup(p->mempolicy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 if (IS_ERR(p->mempolicy)) {
1064 retval = PTR_ERR(p->mempolicy);
1065 p->mempolicy = NULL;
Paul Menageb4f48b62007-10-18 23:39:33 -07001066 goto bad_fork_cleanup_cgroup;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 }
Paul Jacksonc61afb12006-03-24 03:16:08 -08001068 mpol_fix_fork_child_flag(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069#endif
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001070#ifdef CONFIG_TRACE_IRQFLAGS
1071 p->irq_events = 0;
Russell Kingb36e4752006-08-27 12:26:34 +01001072#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1073 p->hardirqs_enabled = 1;
1074#else
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001075 p->hardirqs_enabled = 0;
Russell Kingb36e4752006-08-27 12:26:34 +01001076#endif
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001077 p->hardirq_enable_ip = 0;
1078 p->hardirq_enable_event = 0;
1079 p->hardirq_disable_ip = _THIS_IP_;
1080 p->hardirq_disable_event = 0;
1081 p->softirqs_enabled = 1;
1082 p->softirq_enable_ip = _THIS_IP_;
1083 p->softirq_enable_event = 0;
1084 p->softirq_disable_ip = 0;
1085 p->softirq_disable_event = 0;
1086 p->hardirq_context = 0;
1087 p->softirq_context = 0;
1088#endif
Ingo Molnarfbb9ce952006-07-03 00:24:50 -07001089#ifdef CONFIG_LOCKDEP
1090 p->lockdep_depth = 0; /* no locks held yet */
1091 p->curr_chain_key = 0;
1092 p->lockdep_recursion = 0;
1093#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094
Ingo Molnar408894e2006-01-09 15:59:20 -08001095#ifdef CONFIG_DEBUG_MUTEXES
1096 p->blocked_on = NULL; /* not blocked yet */
1097#endif
Oleg Nesterov06eb23b2009-02-09 02:02:33 +01001098 if (unlikely(current->ptrace))
Markus Metzgerbf53de92008-12-19 15:10:24 +01001099 ptrace_fork(p, clone_flags);
Ingo Molnar408894e2006-01-09 15:59:20 -08001100
Srivatsa Vaddagiri3c90e6e2007-11-09 22:39:39 +01001101 /* Perform scheduler related setup. Assign this task to a CPU. */
1102 sched_fork(p, clone_flags);
1103
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 if ((retval = audit_alloc(p)))
David Howellsf1752ee2008-11-14 10:39:17 +11001105 goto bad_fork_cleanup_policy;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 /* copy all the process information */
1107 if ((retval = copy_semundo(clone_flags, p)))
1108 goto bad_fork_cleanup_audit;
1109 if ((retval = copy_files(clone_flags, p)))
1110 goto bad_fork_cleanup_semundo;
1111 if ((retval = copy_fs(clone_flags, p)))
1112 goto bad_fork_cleanup_files;
1113 if ((retval = copy_sighand(clone_flags, p)))
1114 goto bad_fork_cleanup_fs;
1115 if ((retval = copy_signal(clone_flags, p)))
1116 goto bad_fork_cleanup_sighand;
1117 if ((retval = copy_mm(clone_flags, p)))
1118 goto bad_fork_cleanup_signal;
Serge E. Hallynab516012006-10-02 02:18:06 -07001119 if ((retval = copy_namespaces(clone_flags, p)))
David Howellsd84f4f92008-11-14 10:39:23 +11001120 goto bad_fork_cleanup_mm;
Jens Axboefadad8782008-01-24 08:54:47 +01001121 if ((retval = copy_io(clone_flags, p)))
Jens Axboefd0928d2008-01-24 08:52:45 +01001122 goto bad_fork_cleanup_namespaces;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 retval = copy_thread(0, clone_flags, stack_start, stack_size, p, regs);
1124 if (retval)
Jens Axboefd0928d2008-01-24 08:52:45 +01001125 goto bad_fork_cleanup_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001127 if (pid != &init_struct_pid) {
1128 retval = -ENOMEM;
Eric W. Biederman61bce0f2009-01-07 18:08:49 -08001129 pid = alloc_pid(p->nsproxy->pid_ns);
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001130 if (!pid)
Jens Axboefd0928d2008-01-24 08:52:45 +01001131 goto bad_fork_cleanup_io;
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07001132
1133 if (clone_flags & CLONE_NEWPID) {
Eric W. Biederman61bce0f2009-01-07 18:08:49 -08001134 retval = pid_ns_prepare_proc(p->nsproxy->pid_ns);
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07001135 if (retval < 0)
1136 goto bad_fork_free_pid;
1137 }
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001138 }
1139
Steven Rostedte8e1abe2008-12-03 11:04:51 -05001140 ftrace_graph_init_task(p);
1141
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001142 p->pid = pid_nr(pid);
1143 p->tgid = p->pid;
1144 if (clone_flags & CLONE_THREAD)
1145 p->tgid = current->tgid;
1146
Serge E. Hallyne885dcd2008-07-25 01:47:06 -07001147 if (current->nsproxy != p->nsproxy) {
1148 retval = ns_cgroup_clone(p, pid);
1149 if (retval)
Steven Rostedte8e1abe2008-12-03 11:04:51 -05001150 goto bad_fork_free_graph;
Serge E. Hallyne885dcd2008-07-25 01:47:06 -07001151 }
1152
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1154 /*
1155 * Clear TID on mm_release()?
1156 */
1157 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001158#ifdef CONFIG_FUTEX
Ingo Molnar8f17d3a2006-03-27 01:16:27 -08001159 p->robust_list = NULL;
1160#ifdef CONFIG_COMPAT
1161 p->compat_robust_list = NULL;
1162#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07001163 INIT_LIST_HEAD(&p->pi_state_list);
1164 p->pi_state_cache = NULL;
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001165#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 /*
GOTO Masanorif9a38792006-03-13 21:20:44 -08001167 * sigaltstack should be cleared when sharing the same VM
1168 */
1169 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1170 p->sas_ss_sp = p->sas_ss_size = 0;
1171
1172 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 * Syscall tracing should be turned off in the child regardless
1174 * of CLONE_PTRACE.
1175 */
1176 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
Laurent Viviered75e8d2005-09-03 15:57:18 -07001177#ifdef TIF_SYSCALL_EMU
1178 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1179#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01001180 clear_all_latency_tracing(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 /* ok, now we should be set up.. */
1183 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1184 p->pdeath_signal = 0;
1185 p->exit_state = 0;
1186
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 /*
1188 * Ok, make it visible to the rest of the system.
1189 * We dont wake it up yet.
1190 */
1191 p->group_leader = p;
Oleg Nesterov47e65322006-03-28 16:11:25 -08001192 INIT_LIST_HEAD(&p->thread_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193
Paul Menageb4f48b62007-10-18 23:39:33 -07001194 /* Now that the task is set up, run cgroup callbacks if
1195 * necessary. We need to run them before the task is visible
1196 * on the tasklist. */
1197 cgroup_fork_callbacks(p);
1198 cgroup_callbacks_done = 1;
1199
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200 /* Need tasklist lock for parent etc handling! */
1201 write_lock_irq(&tasklist_lock);
1202
1203 /*
Nick Piggin476d1392005-06-25 14:57:29 -07001204 * The task hasn't been attached yet, so its cpus_allowed mask will
1205 * not be changed, nor will its assigned CPU.
1206 *
1207 * The cpus_allowed mask of the parent may have changed after it was
1208 * copied first time - so re-copy it here, then check the child's CPU
1209 * to ensure it is on a valid CPU (and if not, just force it back to
1210 * parent's CPU). This avoids alot of nasty races.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 */
1212 p->cpus_allowed = current->cpus_allowed;
Peter Zijlstra6f505b12008-01-25 21:08:30 +01001213 p->rt.nr_cpus_allowed = current->rt.nr_cpus_allowed;
Srivatsa Vaddagiri26ff6ad2005-09-16 19:27:40 -07001214 if (unlikely(!cpu_isset(task_cpu(p), p->cpus_allowed) ||
1215 !cpu_online(task_cpu(p))))
Nick Piggin476d1392005-06-25 14:57:29 -07001216 set_task_cpu(p, smp_processor_id());
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 /* CLONE_PARENT re-uses the old parent */
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01001219 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 p->real_parent = current->real_parent;
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01001221 p->parent_exec_id = current->parent_exec_id;
1222 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223 p->real_parent = current;
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01001224 p->parent_exec_id = current->self_exec_id;
1225 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226
Oleg Nesterov3f17da62006-02-15 22:13:24 +03001227 spin_lock(&current->sighand->siglock);
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001228
1229 /*
1230 * Process group and session signals need to be delivered to just the
1231 * parent before the fork or both the parent and the child after the
1232 * fork. Restart if a signal comes in before we add the new process to
1233 * it's process group.
1234 * A fatal signal pending means that current will exit, so the new
1235 * thread can't slip out of an OOM kill (or normal SIGKILL).
1236 */
Daniel Walker23ff4442007-10-18 03:06:07 -07001237 recalc_sigpending();
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001238 if (signal_pending(current)) {
1239 spin_unlock(&current->sighand->siglock);
1240 write_unlock_irq(&tasklist_lock);
1241 retval = -ERESTARTNOINTR;
Steven Rostedte8e1abe2008-12-03 11:04:51 -05001242 goto bad_fork_free_graph;
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001243 }
1244
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 if (clone_flags & CLONE_THREAD) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 p->group_leader = current->group_leader;
Oleg Nesterov47e65322006-03-28 16:11:25 -08001247 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 }
1249
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001250 if (likely(p->pid)) {
Roland McGrathf4700212008-03-24 18:36:23 -07001251 list_add_tail(&p->sibling, &p->real_parent->children);
Roland McGrath09a05392008-07-25 19:45:47 -07001252 tracehook_finish_clone(p, clone_flags, trace);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001254 if (thread_group_leader(p)) {
Eric W. Biederman5cd17562007-12-04 23:45:04 -08001255 if (clone_flags & CLONE_NEWPID)
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001256 p->nsproxy->pid_ns->child_reaper = p;
Oleg Nesterovc97d9892006-03-28 16:11:06 -08001257
Oleg Nesterovfea9d172008-02-08 04:19:19 -08001258 p->signal->leader_pid = pid;
Alan Cox9c9f4de2008-10-13 10:37:26 +01001259 tty_kref_put(p->signal->tty);
1260 p->signal->tty = tty_kref_get(current->signal->tty);
Eric W. Biederman5cd17562007-12-04 23:45:04 -08001261 set_task_pgrp(p, task_pgrp_nr(current));
1262 set_task_session(p, task_session_nr(current));
1263 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1264 attach_pid(p, PIDTYPE_SID, task_session(current));
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -07001265 list_add_tail_rcu(&p->tasks, &init_task.tasks);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 __get_cpu_var(process_counts)++;
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001267 }
Sukadev Bhattiprolu85868992007-05-10 22:23:03 -07001268 attach_pid(p, PIDTYPE_PID, pid);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001269 nr_threads++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 }
1271
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272 total_forks++;
Oleg Nesterov3f17da62006-02-15 22:13:24 +03001273 spin_unlock(&current->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 write_unlock_irq(&tasklist_lock);
Andrew Mortonc13cf852005-11-28 13:43:48 -08001275 proc_fork_connector(p);
Paul Menage817929e2007-10-18 23:39:36 -07001276 cgroup_post_fork(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 return p;
1278
Steven Rostedte8e1abe2008-12-03 11:04:51 -05001279bad_fork_free_graph:
1280 ftrace_graph_exit_task(p);
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001281bad_fork_free_pid:
1282 if (pid != &init_struct_pid)
1283 free_pid(pid);
Jens Axboefd0928d2008-01-24 08:52:45 +01001284bad_fork_cleanup_io:
1285 put_io_context(p->io_context);
Serge E. Hallynab516012006-10-02 02:18:06 -07001286bad_fork_cleanup_namespaces:
Linus Torvalds444f3782007-01-30 13:35:18 -08001287 exit_task_namespaces(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288bad_fork_cleanup_mm:
1289 if (p->mm)
1290 mmput(p->mm);
1291bad_fork_cleanup_signal:
Oleg Nesterov6b3934ef2006-03-28 16:11:16 -08001292 cleanup_signal(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293bad_fork_cleanup_sighand:
Oleg Nesterova7e53282006-03-28 16:11:27 -08001294 __cleanup_sighand(p->sighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295bad_fork_cleanup_fs:
1296 exit_fs(p); /* blocking */
1297bad_fork_cleanup_files:
1298 exit_files(p); /* blocking */
1299bad_fork_cleanup_semundo:
1300 exit_sem(p);
1301bad_fork_cleanup_audit:
1302 audit_free(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303bad_fork_cleanup_policy:
1304#ifdef CONFIG_NUMA
Lee Schermerhornf0be3d32008-04-28 02:13:08 -07001305 mpol_put(p->mempolicy);
Paul Menageb4f48b62007-10-18 23:39:33 -07001306bad_fork_cleanup_cgroup:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307#endif
Paul Menageb4f48b62007-10-18 23:39:33 -07001308 cgroup_exit(p, cgroup_callbacks_done);
Shailabh Nagar35df17c2006-08-31 21:27:38 -07001309 delayacct_tsk_free(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310 if (p->binfmt)
1311 module_put(p->binfmt->module);
1312bad_fork_cleanup_put_domain:
Al Viroa1261f542005-11-13 16:06:55 -08001313 module_put(task_thread_info(p)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314bad_fork_cleanup_count:
David Howellsd84f4f92008-11-14 10:39:23 +11001315 atomic_dec(&p->cred->user->processes);
David Howells3b11a1d2008-11-14 10:39:26 +11001316 put_cred(p->real_cred);
David Howellsf1752ee2008-11-14 10:39:17 +11001317 put_cred(p->cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318bad_fork_free:
1319 free_task(p);
Oleg Nesterovfe7d37d2006-01-08 01:04:02 -08001320fork_out:
1321 return ERR_PTR(retval);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322}
1323
Adrian Bunk6b2fb3c2008-02-06 01:37:55 -08001324noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325{
1326 memset(regs, 0, sizeof(struct pt_regs));
1327 return regs;
1328}
1329
Al Viro9abcf402007-02-01 13:52:48 +00001330struct task_struct * __cpuinit fork_idle(int cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331{
Ingo Molnar36c8b582006-07-03 00:25:41 -07001332 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333 struct pt_regs regs;
1334
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001335 task = copy_process(CLONE_VM, 0, idle_regs(&regs), 0, NULL,
Roland McGrath09a05392008-07-25 19:45:47 -07001336 &init_struct_pid, 0);
Akinobu Mita753ca4f2006-11-25 11:09:34 -08001337 if (!IS_ERR(task))
1338 init_idle(task, cpu);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001339
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340 return task;
1341}
1342
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343/*
1344 * Ok, this is the main fork-routine.
1345 *
1346 * It copies the process, and if successful kick-starts
1347 * it and waits for it to finish using the VM if required.
1348 */
1349long do_fork(unsigned long clone_flags,
1350 unsigned long stack_start,
1351 struct pt_regs *regs,
1352 unsigned long stack_size,
1353 int __user *parent_tidptr,
1354 int __user *child_tidptr)
1355{
1356 struct task_struct *p;
1357 int trace = 0;
Eric W. Biederman92476d72006-03-31 02:31:42 -08001358 long nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359
Andrew Mortonbdff7462008-02-04 22:27:22 -08001360 /*
Serge Hallyn18b6e042008-10-15 16:38:45 -05001361 * Do some preliminary argument and permissions checking before we
1362 * actually start allocating stuff
1363 */
1364 if (clone_flags & CLONE_NEWUSER) {
1365 if (clone_flags & CLONE_THREAD)
1366 return -EINVAL;
1367 /* hopefully this check will go away when userns support is
1368 * complete
1369 */
Serge E. Hallyn7657d902008-12-03 13:17:33 -06001370 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1371 !capable(CAP_SETGID))
Serge Hallyn18b6e042008-10-15 16:38:45 -05001372 return -EPERM;
1373 }
1374
1375 /*
Andrew Mortonbdff7462008-02-04 22:27:22 -08001376 * We hope to recycle these flags after 2.6.26
1377 */
1378 if (unlikely(clone_flags & CLONE_STOPPED)) {
1379 static int __read_mostly count = 100;
1380
1381 if (count > 0 && printk_ratelimit()) {
1382 char comm[TASK_COMM_LEN];
1383
1384 count--;
1385 printk(KERN_INFO "fork(): process `%s' used deprecated "
1386 "clone flags 0x%lx\n",
1387 get_task_comm(comm, current),
1388 clone_flags & CLONE_STOPPED);
1389 }
1390 }
1391
Roland McGrath09a05392008-07-25 19:45:47 -07001392 /*
1393 * When called from kernel_thread, don't do user tracing stuff.
1394 */
1395 if (likely(user_mode(regs)))
1396 trace = tracehook_prepare_clone(clone_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397
Sukadev Bhattiprolua6f5e062007-10-18 23:39:53 -07001398 p = copy_process(clone_flags, stack_start, regs, stack_size,
Roland McGrath09a05392008-07-25 19:45:47 -07001399 child_tidptr, NULL, trace);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 /*
1401 * Do this prior waking up the new thread - the thread pointer
1402 * might get invalid after that point, if the thread exits quickly.
1403 */
1404 if (!IS_ERR(p)) {
1405 struct completion vfork;
1406
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04001407 trace_sched_process_fork(current, p);
1408
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001409 nr = task_pid_vnr(p);
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001410
1411 if (clone_flags & CLONE_PARENT_SETTID)
1412 put_user(nr, parent_tidptr);
Sukadev Bhattiprolua6f5e062007-10-18 23:39:53 -07001413
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 if (clone_flags & CLONE_VFORK) {
1415 p->vfork_done = &vfork;
1416 init_completion(&vfork);
1417 }
1418
Al Viroa64e6492008-11-12 18:37:41 -05001419 audit_finish_fork(p);
Roland McGrath09a05392008-07-25 19:45:47 -07001420 tracehook_report_clone(trace, regs, clone_flags, nr, p);
1421
1422 /*
1423 * We set PF_STARTING at creation in case tracing wants to
1424 * use this to distinguish a fully live task from one that
1425 * hasn't gotten to tracehook_report_clone() yet. Now we
1426 * clear it and set the child going.
1427 */
1428 p->flags &= ~PF_STARTING;
1429
1430 if (unlikely(clone_flags & CLONE_STOPPED)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431 /*
1432 * We'll start up with an immediate SIGSTOP.
1433 */
1434 sigaddset(&p->pending.signal, SIGSTOP);
1435 set_tsk_thread_flag(p, TIF_SIGPENDING);
Oleg Nesterovd9ae90a2008-02-06 01:36:13 -08001436 __set_task_state(p, TASK_STOPPED);
Roland McGrath09a05392008-07-25 19:45:47 -07001437 } else {
1438 wake_up_new_task(p, clone_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 }
1440
Roland McGrath09a05392008-07-25 19:45:47 -07001441 tracehook_report_clone_complete(trace, regs,
1442 clone_flags, nr, p);
1443
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 if (clone_flags & CLONE_VFORK) {
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -07001445 freezer_do_not_count();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446 wait_for_completion(&vfork);
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -07001447 freezer_count();
Roland McGrathdaded342008-07-25 19:45:47 -07001448 tracehook_report_vfork_done(p, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 }
1450 } else {
Eric W. Biederman92476d72006-03-31 02:31:42 -08001451 nr = PTR_ERR(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452 }
Eric W. Biederman92476d72006-03-31 02:31:42 -08001453 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454}
1455
Ravikiran G Thirumalai5fd63b32006-01-11 22:46:15 +01001456#ifndef ARCH_MIN_MMSTRUCT_ALIGN
1457#define ARCH_MIN_MMSTRUCT_ALIGN 0
1458#endif
1459
Alexey Dobriyan51cc5062008-07-25 19:45:34 -07001460static void sighand_ctor(void *data)
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08001461{
1462 struct sighand_struct *sighand = data;
1463
Christoph Lametera35afb82007-05-16 22:10:57 -07001464 spin_lock_init(&sighand->siglock);
Davide Libenzib8fceee2007-09-20 12:40:16 -07001465 init_waitqueue_head(&sighand->signalfd_wqh);
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08001466}
1467
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468void __init proc_caches_init(void)
1469{
1470 sighand_cachep = kmem_cache_create("sighand_cache",
1471 sizeof(struct sighand_struct), 0,
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08001472 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU,
Paul Mundt20c2df82007-07-20 10:11:58 +09001473 sighand_ctor);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 signal_cachep = kmem_cache_create("signal_cache",
1475 sizeof(struct signal_struct), 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09001476 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
1477 files_cachep = kmem_cache_create("files_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 sizeof(struct files_struct), 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09001479 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
1480 fs_cachep = kmem_cache_create("fs_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 sizeof(struct fs_struct), 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09001482 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 mm_cachep = kmem_cache_create("mm_struct",
Ravikiran G Thirumalai5fd63b32006-01-11 22:46:15 +01001484 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
Paul Mundt20c2df82007-07-20 10:11:58 +09001485 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
David Howells8feae132009-01-08 12:04:47 +00001486 mmap_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487}
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001488
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001489/*
1490 * Check constraints on flags passed to the unshare system call and
1491 * force unsharing of additional process context as appropriate.
1492 */
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001493static void check_unshare_flags(unsigned long *flags_ptr)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001494{
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(&current->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 */
1526static int unshare_thread(unsigned long unshare_flags)
1527{
1528 if (unshare_flags & CLONE_THREAD)
1529 return -EINVAL;
1530
1531 return 0;
1532}
1533
1534/*
JANAK DESAI99d14192006-02-07 12:58:59 -08001535 * Unshare the filesystem structure if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001536 */
1537static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1538{
1539 struct fs_struct *fs = current->fs;
1540
1541 if ((unshare_flags & CLONE_FS) &&
JANAK DESAI99d14192006-02-07 12:58:59 -08001542 (fs && atomic_read(&fs->count) > 1)) {
1543 *new_fsp = __copy_fs_struct(current->fs);
1544 if (!*new_fsp)
1545 return -ENOMEM;
1546 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001547
1548 return 0;
1549}
1550
1551/*
Oleg Nesterovdae3c5a2006-12-08 02:36:09 -08001552 * Unsharing of sighand is not supported yet
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001553 */
1554static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp)
1555{
1556 struct sighand_struct *sigh = current->sighand;
1557
Oleg Nesterovdae3c5a2006-12-08 02:36:09 -08001558 if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001559 return -EINVAL;
1560 else
1561 return 0;
1562}
1563
1564/*
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001565 * Unshare vm if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001566 */
1567static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp)
1568{
1569 struct mm_struct *mm = current->mm;
1570
1571 if ((unshare_flags & CLONE_VM) &&
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001572 (mm && atomic_read(&mm->mm_users) > 1)) {
Oleg Nesterov2d61b8672006-03-18 20:41:10 +03001573 return -EINVAL;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001574 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001575
1576 return 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001577}
1578
1579/*
JANAK DESAIa016f332006-02-07 12:59:02 -08001580 * Unshare file descriptor table if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001581 */
1582static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1583{
1584 struct files_struct *fd = current->files;
JANAK DESAIa016f332006-02-07 12:59:02 -08001585 int error = 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001586
1587 if ((unshare_flags & CLONE_FILES) &&
JANAK DESAIa016f332006-02-07 12:59:02 -08001588 (fd && atomic_read(&fd->count) > 1)) {
1589 *new_fdp = dup_fd(fd, &error);
1590 if (!*new_fdp)
1591 return error;
1592 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001593
1594 return 0;
1595}
1596
1597/*
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001598 * unshare allows a process to 'unshare' part of the process
1599 * context which was originally shared using clone. copy_*
1600 * functions used by do_fork() cannot be used here directly
1601 * because they modify an inactive task_struct that is being
1602 * constructed. Here we are modifying the current, active,
1603 * task_struct.
1604 */
Heiko Carstens6559eed82009-01-14 14:14:32 +01001605SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001606{
1607 int err = 0;
1608 struct fs_struct *fs, *new_fs = NULL;
Oleg Nesterovdae3c5a2006-12-08 02:36:09 -08001609 struct sighand_struct *new_sigh = NULL;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001610 struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL;
1611 struct files_struct *fd, *new_fd = NULL;
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07001612 struct nsproxy *new_nsproxy = NULL;
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001613 int do_sysvsem = 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001614
1615 check_unshare_flags(&unshare_flags);
1616
Eric W. Biederman06f9d4f2006-03-22 00:07:40 -08001617 /* Return -EINVAL for all unsupported flags */
1618 err = -EINVAL;
1619 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
Kirill Korotaev25b21cb2006-10-02 02:18:19 -07001620 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
Serge Hallyn18b6e042008-10-15 16:38:45 -05001621 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
Eric W. Biederman06f9d4f2006-03-22 00:07:40 -08001622 goto bad_unshare_out;
1623
Manfred Spraul6013f672008-04-29 01:00:59 -07001624 /*
1625 * CLONE_NEWIPC must also detach from the undolist: after switching
1626 * to a new ipc namespace, the semaphore arrays from the old
1627 * namespace are unreachable.
1628 */
1629 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001630 do_sysvsem = 1;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001631 if ((err = unshare_thread(unshare_flags)))
1632 goto bad_unshare_out;
1633 if ((err = unshare_fs(unshare_flags, &new_fs)))
1634 goto bad_unshare_cleanup_thread;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001635 if ((err = unshare_sighand(unshare_flags, &new_sigh)))
Badari Pulavartye3222c42007-05-08 00:25:21 -07001636 goto bad_unshare_cleanup_fs;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001637 if ((err = unshare_vm(unshare_flags, &new_mm)))
1638 goto bad_unshare_cleanup_sigh;
1639 if ((err = unshare_fd(unshare_flags, &new_fd)))
1640 goto bad_unshare_cleanup_vm;
Badari Pulavartye3222c42007-05-08 00:25:21 -07001641 if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
1642 new_fs)))
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001643 goto bad_unshare_cleanup_fd;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001644
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001645 if (new_fs || new_mm || new_fd || do_sysvsem || new_nsproxy) {
1646 if (do_sysvsem) {
1647 /*
1648 * CLONE_SYSVSEM is equivalent to sys_exit().
1649 */
1650 exit_sem(current);
1651 }
Serge E. Hallynab516012006-10-02 02:18:06 -07001652
Serge Hallync0b2fc32006-10-02 02:18:18 -07001653 if (new_nsproxy) {
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07001654 switch_task_namespaces(current, new_nsproxy);
1655 new_nsproxy = NULL;
Serge Hallync0b2fc32006-10-02 02:18:18 -07001656 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001657
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07001658 task_lock(current);
1659
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001660 if (new_fs) {
1661 fs = current->fs;
1662 current->fs = new_fs;
1663 new_fs = fs;
1664 }
1665
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001666 if (new_mm) {
1667 mm = current->mm;
1668 active_mm = current->active_mm;
1669 current->mm = new_mm;
1670 current->active_mm = new_mm;
1671 activate_mm(active_mm, new_mm);
1672 new_mm = mm;
1673 }
1674
1675 if (new_fd) {
1676 fd = current->files;
1677 current->files = new_fd;
1678 new_fd = fd;
1679 }
1680
1681 task_unlock(current);
1682 }
1683
Serge Hallync0b2fc32006-10-02 02:18:18 -07001684 if (new_nsproxy)
Linus Torvalds444f3782007-01-30 13:35:18 -08001685 put_nsproxy(new_nsproxy);
Serge Hallync0b2fc32006-10-02 02:18:18 -07001686
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001687bad_unshare_cleanup_fd:
1688 if (new_fd)
1689 put_files_struct(new_fd);
1690
1691bad_unshare_cleanup_vm:
1692 if (new_mm)
1693 mmput(new_mm);
1694
1695bad_unshare_cleanup_sigh:
1696 if (new_sigh)
1697 if (atomic_dec_and_test(&new_sigh->count))
1698 kmem_cache_free(sighand_cachep, new_sigh);
1699
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001700bad_unshare_cleanup_fs:
1701 if (new_fs)
1702 put_fs_struct(new_fs);
1703
1704bad_unshare_cleanup_thread:
1705bad_unshare_out:
1706 return err;
1707}
Al Viro3b125382008-04-22 05:31:30 -04001708
1709/*
1710 * Helper to unshare the files of the current task.
1711 * We don't want to expose copy_files internals to
1712 * the exec layer of the kernel.
1713 */
1714
1715int unshare_files(struct files_struct **displaced)
1716{
1717 struct task_struct *task = current;
Al Viro50704512008-04-26 05:25:00 +01001718 struct files_struct *copy = NULL;
Al Viro3b125382008-04-22 05:31:30 -04001719 int error;
1720
1721 error = unshare_fd(CLONE_FILES, &copy);
1722 if (error || !copy) {
1723 *displaced = NULL;
1724 return error;
1725 }
1726 *displaced = task->files;
1727 task_lock(task);
1728 task->files = copy;
1729 task_unlock(task);
1730 return 0;
1731}