blob: cfaff92f61ff584d72b4d311607dc41bd308c084 [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>
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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070049#include <linux/profile.h>
50#include <linux/rmap.h>
51#include <linux/acct.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070052#include <linux/tsacct_kern.h>
Matt Helsley9f460802005-11-07 00:59:16 -080053#include <linux/cn_proc.h>
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -070054#include <linux/freezer.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070055#include <linux/delayacct.h>
Shailabh Nagarad4ecbc2006-07-14 00:24:44 -070056#include <linux/taskstats_kern.h>
Arjan van de Ven0a425402006-09-26 10:52:38 +020057#include <linux/random.h>
Miloslav Trmac522ed772007-07-15 23:40:56 -070058#include <linux/tty.h>
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -070059#include <linux/proc_fs.h>
Jens Axboefd0928d2008-01-24 08:52:45 +010060#include <linux/blkdev.h>
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -040061#include <trace/sched.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
63#include <asm/pgtable.h>
64#include <asm/pgalloc.h>
65#include <asm/uaccess.h>
66#include <asm/mmu_context.h>
67#include <asm/cacheflush.h>
68#include <asm/tlbflush.h>
69
70/*
71 * Protected counters by write_lock_irq(&tasklist_lock)
72 */
73unsigned long total_forks; /* Handle normal Linux uptimes. */
74int nr_threads; /* The idle threads do not count.. */
75
76int max_threads; /* tunable limit on nr_threads */
77
78DEFINE_PER_CPU(unsigned long, process_counts) = 0;
79
Christoph Hellwigc59923a2006-07-10 04:45:40 -070080__cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
Linus Torvalds1da177e2005-04-16 15:20:36 -070081
82int nr_processes(void)
83{
84 int cpu;
85 int total = 0;
86
87 for_each_online_cpu(cpu)
88 total += per_cpu(process_counts, cpu);
89
90 return total;
91}
92
93#ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
94# define alloc_task_struct() kmem_cache_alloc(task_struct_cachep, GFP_KERNEL)
95# define free_task_struct(tsk) kmem_cache_free(task_struct_cachep, (tsk))
Christoph Lametere18b8902006-12-06 20:33:20 -080096static struct kmem_cache *task_struct_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -070097#endif
98
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -070099#ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
100static inline struct thread_info *alloc_thread_info(struct task_struct *tsk)
101{
102#ifdef CONFIG_DEBUG_STACK_USAGE
103 gfp_t mask = GFP_KERNEL | __GFP_ZERO;
104#else
105 gfp_t mask = GFP_KERNEL;
106#endif
107 return (struct thread_info *)__get_free_pages(mask, THREAD_SIZE_ORDER);
108}
109
110static inline void free_thread_info(struct thread_info *ti)
111{
112 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
113}
114#endif
115
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116/* SLAB cache for signal_struct structures (tsk->signal) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800117static struct kmem_cache *signal_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118
119/* SLAB cache for sighand_struct structures (tsk->sighand) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800120struct kmem_cache *sighand_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121
122/* SLAB cache for files_struct structures (tsk->files) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800123struct kmem_cache *files_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124
125/* SLAB cache for fs_struct structures (tsk->fs) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800126struct kmem_cache *fs_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127
128/* SLAB cache for vm_area_struct structures */
Christoph Lametere18b8902006-12-06 20:33:20 -0800129struct kmem_cache *vm_area_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130
131/* SLAB cache for mm_struct structures (tsk->mm) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800132static struct kmem_cache *mm_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133
134void free_task(struct task_struct *tsk)
135{
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700136 prop_local_destroy_single(&tsk->dirties);
Roman Zippelf7e42172007-05-09 02:35:17 -0700137 free_thread_info(tsk->stack);
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700138 rt_mutex_debug_task_free(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139 free_task_struct(tsk);
140}
141EXPORT_SYMBOL(free_task);
142
Andrew Morton158d9eb2006-03-31 02:31:34 -0800143void __put_task_struct(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144{
Eugene Teo270f7222007-10-18 23:40:38 -0700145 WARN_ON(!tsk->exit_state);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 WARN_ON(atomic_read(&tsk->usage));
147 WARN_ON(tsk == current);
148
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 security_task_free(tsk);
150 free_uid(tsk->user);
151 put_group_info(tsk->group_info);
Shailabh Nagar35df17c2006-08-31 21:27:38 -0700152 delayacct_tsk_free(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153
154 if (!profile_handoff_task(tsk))
155 free_task(tsk);
156}
157
Suresh Siddha2adee9b2008-04-16 10:25:35 +0200158/*
159 * macro override instead of weak attribute alias, to workaround
160 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
161 */
162#ifndef arch_task_cache_init
163#define arch_task_cache_init()
164#endif
Suresh Siddha61c46282008-03-10 15:28:04 -0700165
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166void __init fork_init(unsigned long mempages)
167{
168#ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
169#ifndef ARCH_MIN_TASKALIGN
170#define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
171#endif
172 /* create a slab on which task_structs can be allocated */
173 task_struct_cachep =
174 kmem_cache_create("task_struct", sizeof(struct task_struct),
Paul Mundt20c2df82007-07-20 10:11:58 +0900175 ARCH_MIN_TASKALIGN, SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176#endif
177
Suresh Siddha61c46282008-03-10 15:28:04 -0700178 /* do the arch specific task caches init */
179 arch_task_cache_init();
180
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 /*
182 * The default maximum number of threads is set to a safe
183 * value: the thread structures can take up at most half
184 * of memory.
185 */
186 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
187
188 /*
189 * we need to allow at least 20 threads to boot a system
190 */
191 if(max_threads < 20)
192 max_threads = 20;
193
194 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
195 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
196 init_task.signal->rlim[RLIMIT_SIGPENDING] =
197 init_task.signal->rlim[RLIMIT_NPROC];
198}
199
Suresh Siddha61c46282008-03-10 15:28:04 -0700200int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
201 struct task_struct *src)
202{
203 *dst = *src;
204 return 0;
205}
206
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207static struct task_struct *dup_task_struct(struct task_struct *orig)
208{
209 struct task_struct *tsk;
210 struct thread_info *ti;
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700211 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212
213 prepare_to_copy(orig);
214
215 tsk = alloc_task_struct();
216 if (!tsk)
217 return NULL;
218
219 ti = alloc_thread_info(tsk);
220 if (!ti) {
221 free_task_struct(tsk);
222 return NULL;
223 }
224
Suresh Siddha61c46282008-03-10 15:28:04 -0700225 err = arch_dup_task_struct(tsk, orig);
226 if (err)
227 goto out;
228
Roman Zippelf7e42172007-05-09 02:35:17 -0700229 tsk->stack = ti;
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700230
231 err = prop_local_init_single(&tsk->dirties);
Suresh Siddha61c46282008-03-10 15:28:04 -0700232 if (err)
233 goto out;
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700234
Al Viro10ebffd2005-11-13 16:06:56 -0800235 setup_thread_stack(tsk, orig);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236
Arjan van de Ven0a425402006-09-26 10:52:38 +0200237#ifdef CONFIG_CC_STACKPROTECTOR
238 tsk->stack_canary = get_random_int();
239#endif
240
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 /* One for us, one for whoever does the "release_task()" (usually parent) */
242 atomic_set(&tsk->usage,2);
Giancarlo Formicuccia4b5d37a2005-09-09 13:01:22 -0700243 atomic_set(&tsk->fs_excl, 0);
Alexey Dobriyan6c5c9342006-09-29 01:59:40 -0700244#ifdef CONFIG_BLK_DEV_IO_TRACE
Jens Axboe2056a782006-03-23 20:00:26 +0100245 tsk->btrace_seq = 0;
Alexey Dobriyan6c5c9342006-09-29 01:59:40 -0700246#endif
Jens Axboea0aa7f62006-04-20 13:05:33 +0200247 tsk->splice_pipe = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 return tsk;
Suresh Siddha61c46282008-03-10 15:28:04 -0700249
250out:
251 free_thread_info(ti);
252 free_task_struct(tsk);
253 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254}
255
256#ifdef CONFIG_MMU
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700257static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258{
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700259 struct vm_area_struct *mpnt, *tmp, **pprev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 struct rb_node **rb_link, *rb_parent;
261 int retval;
262 unsigned long charge;
263 struct mempolicy *pol;
264
265 down_write(&oldmm->mmap_sem);
Ralf Baechleec8c0442006-12-12 17:14:57 +0000266 flush_cache_dup_mm(oldmm);
Ingo Molnarad339452006-07-03 00:25:15 -0700267 /*
268 * Not linked in yet - no deadlock potential:
269 */
270 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
Hugh Dickins7ee78232005-10-29 18:16:08 -0700271
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272 mm->locked_vm = 0;
273 mm->mmap = NULL;
274 mm->mmap_cache = NULL;
275 mm->free_area_cache = oldmm->mmap_base;
Wolfgang Wander1363c3c2005-06-21 17:14:49 -0700276 mm->cached_hole_size = ~0UL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 mm->map_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 cpus_clear(mm->cpu_vm_mask);
279 mm->mm_rb = RB_ROOT;
280 rb_link = &mm->mm_rb.rb_node;
281 rb_parent = NULL;
282 pprev = &mm->mmap;
283
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700284 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 struct file *file;
286
287 if (mpnt->vm_flags & VM_DONTCOPY) {
Hugh Dickins3b6bfcd2005-07-12 13:58:09 -0700288 long pages = vma_pages(mpnt);
289 mm->total_vm -= pages;
Hugh Dickinsab50b8e2005-10-29 18:15:56 -0700290 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
Hugh Dickins3b6bfcd2005-07-12 13:58:09 -0700291 -pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 continue;
293 }
294 charge = 0;
295 if (mpnt->vm_flags & VM_ACCOUNT) {
296 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
297 if (security_vm_enough_memory(len))
298 goto fail_nomem;
299 charge = len;
300 }
Christoph Lametere94b1762006-12-06 20:33:17 -0800301 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 if (!tmp)
303 goto fail_nomem;
304 *tmp = *mpnt;
Lee Schermerhorn846a16b2008-04-28 02:13:09 -0700305 pol = mpol_dup(vma_policy(mpnt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306 retval = PTR_ERR(pol);
307 if (IS_ERR(pol))
308 goto fail_nomem_policy;
309 vma_set_policy(tmp, pol);
310 tmp->vm_flags &= ~VM_LOCKED;
311 tmp->vm_mm = mm;
312 tmp->vm_next = NULL;
313 anon_vma_link(tmp);
314 file = tmp->vm_file;
315 if (file) {
Josef "Jeff" Sipekf3a43f32006-12-08 02:36:43 -0800316 struct inode *inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 get_file(file);
318 if (tmp->vm_flags & VM_DENYWRITE)
319 atomic_dec(&inode->i_writecount);
Daniel Walker23ff4442007-10-18 03:06:07 -0700320
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321 /* insert tmp into the share list, just after mpnt */
322 spin_lock(&file->f_mapping->i_mmap_lock);
323 tmp->vm_truncate_count = mpnt->vm_truncate_count;
324 flush_dcache_mmap_lock(file->f_mapping);
325 vma_prio_tree_add(tmp, mpnt);
326 flush_dcache_mmap_unlock(file->f_mapping);
327 spin_unlock(&file->f_mapping->i_mmap_lock);
328 }
329
330 /*
Mel Gormana1e78772008-07-23 21:27:23 -0700331 * Clear hugetlb-related page reserves for children. This only
332 * affects MAP_PRIVATE mappings. Faults generated by the child
333 * are not guaranteed to succeed, even if read-only
334 */
335 if (is_vm_hugetlb_page(tmp))
336 reset_vma_resv_huge_pages(tmp);
337
338 /*
Hugh Dickins7ee78232005-10-29 18:16:08 -0700339 * Link in the new vma and copy the page table entries.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 *pprev = tmp;
342 pprev = &tmp->vm_next;
343
344 __vma_link_rb(mm, tmp, rb_link, rb_parent);
345 rb_link = &tmp->vm_rb.rb_right;
346 rb_parent = &tmp->vm_rb;
347
348 mm->map_count++;
Hugh Dickins0b0db142005-11-21 21:32:20 -0800349 retval = copy_page_range(mm, oldmm, mpnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350
351 if (tmp->vm_ops && tmp->vm_ops->open)
352 tmp->vm_ops->open(tmp);
353
354 if (retval)
355 goto out;
356 }
Jeremy Fitzhardinged6dd61c2007-05-02 19:27:14 +0200357 /* a new mm has just been created */
358 arch_dup_mmap(oldmm, mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360out:
Hugh Dickins7ee78232005-10-29 18:16:08 -0700361 up_write(&mm->mmap_sem);
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700362 flush_tlb_mm(oldmm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 up_write(&oldmm->mmap_sem);
364 return retval;
365fail_nomem_policy:
366 kmem_cache_free(vm_area_cachep, tmp);
367fail_nomem:
368 retval = -ENOMEM;
369 vm_unacct_memory(charge);
370 goto out;
371}
372
373static inline int mm_alloc_pgd(struct mm_struct * mm)
374{
375 mm->pgd = pgd_alloc(mm);
376 if (unlikely(!mm->pgd))
377 return -ENOMEM;
378 return 0;
379}
380
381static inline void mm_free_pgd(struct mm_struct * mm)
382{
Benjamin Herrenschmidt5e541972008-02-04 22:29:14 -0800383 pgd_free(mm, mm->pgd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384}
385#else
386#define dup_mmap(mm, oldmm) (0)
387#define mm_alloc_pgd(mm) (0)
388#define mm_free_pgd(mm)
389#endif /* CONFIG_MMU */
390
Daniel Walker23ff4442007-10-18 03:06:07 -0700391__cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
Christoph Lametere94b1762006-12-06 20:33:17 -0800393#define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394#define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
395
396#include <linux/init_task.h>
397
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800398static struct mm_struct * mm_init(struct mm_struct * mm, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399{
400 atomic_set(&mm->mm_users, 1);
401 atomic_set(&mm->mm_count, 1);
402 init_rwsem(&mm->mmap_sem);
403 INIT_LIST_HEAD(&mm->mmlist);
Kawai, Hidehiro3cb4a0b2007-07-19 01:48:28 -0700404 mm->flags = (current->mm) ? current->mm->flags
405 : MMF_DUMP_FILTER_DEFAULT;
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700406 mm->core_state = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 mm->nr_ptes = 0;
Hugh Dickins42946212005-10-29 18:16:05 -0700408 set_mm_counter(mm, file_rss, 0);
Hugh Dickins404351e2005-10-29 18:16:04 -0700409 set_mm_counter(mm, anon_rss, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 spin_lock_init(&mm->page_table_lock);
411 rwlock_init(&mm->ioctx_list_lock);
412 mm->ioctx_list = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 mm->free_area_cache = TASK_UNMAPPED_BASE;
Wolfgang Wander1363c3c2005-06-21 17:14:49 -0700414 mm->cached_hole_size = ~0UL;
Balbir Singhcf475ad2008-04-29 01:00:16 -0700415 mm_init_owner(mm, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416
417 if (likely(!mm_alloc_pgd(mm))) {
418 mm->def_flags = 0;
Andrea Arcangelicddb8a52008-07-28 15:46:29 -0700419 mmu_notifier_mm_init(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 return mm;
421 }
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800422
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 free_mm(mm);
424 return NULL;
425}
426
427/*
428 * Allocate and initialize an mm_struct.
429 */
430struct mm_struct * mm_alloc(void)
431{
432 struct mm_struct * mm;
433
434 mm = allocate_mm();
435 if (mm) {
436 memset(mm, 0, sizeof(*mm));
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800437 mm = mm_init(mm, current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 }
439 return mm;
440}
441
442/*
443 * Called when the last reference to the mm
444 * is dropped: either by a lazy thread or by
445 * mmput. Free the page directory and the mm.
446 */
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800447void __mmdrop(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448{
449 BUG_ON(mm == &init_mm);
450 mm_free_pgd(mm);
451 destroy_context(mm);
Andrea Arcangelicddb8a52008-07-28 15:46:29 -0700452 mmu_notifier_mm_destroy(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 free_mm(mm);
454}
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200455EXPORT_SYMBOL_GPL(__mmdrop);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456
457/*
458 * Decrement the use count and release all resources for an mm.
459 */
460void mmput(struct mm_struct *mm)
461{
Andrew Morton0ae26f12006-06-23 02:05:15 -0700462 might_sleep();
463
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 if (atomic_dec_and_test(&mm->mm_users)) {
465 exit_aio(mm);
466 exit_mmap(mm);
Matt Helsley925d1c42008-04-29 01:01:36 -0700467 set_mm_exe_file(mm, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 if (!list_empty(&mm->mmlist)) {
469 spin_lock(&mmlist_lock);
470 list_del(&mm->mmlist);
471 spin_unlock(&mmlist_lock);
472 }
473 put_swap_token(mm);
474 mmdrop(mm);
475 }
476}
477EXPORT_SYMBOL_GPL(mmput);
478
479/**
480 * get_task_mm - acquire a reference to the task's mm
481 *
Oleg Nesterov246bb0b2008-07-25 01:47:38 -0700482 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 * this kernel workthread has transiently adopted a user mm with use_mm,
484 * to do its AIO) is not set and if so returns a reference to it, after
485 * bumping up the use count. User must release the mm via mmput()
486 * after use. Typically used by /proc and ptrace.
487 */
488struct mm_struct *get_task_mm(struct task_struct *task)
489{
490 struct mm_struct *mm;
491
492 task_lock(task);
493 mm = task->mm;
494 if (mm) {
Oleg Nesterov246bb0b2008-07-25 01:47:38 -0700495 if (task->flags & PF_KTHREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 mm = NULL;
497 else
498 atomic_inc(&mm->mm_users);
499 }
500 task_unlock(task);
501 return mm;
502}
503EXPORT_SYMBOL_GPL(get_task_mm);
504
505/* Please note the differences between mmput and mm_release.
506 * mmput is called whenever we stop holding onto a mm_struct,
507 * error success whatever.
508 *
509 * mm_release is called after a mm_struct has been removed
510 * from the current process.
511 *
512 * This difference is important for error handling, when we
513 * only half set up a mm_struct for a new process and need to restore
514 * the old one. Because we mmput the new mm_struct before
515 * restoring the old one. . .
516 * Eric Biederman 10 January 1998
517 */
518void mm_release(struct task_struct *tsk, struct mm_struct *mm)
519{
520 struct completion *vfork_done = tsk->vfork_done;
521
522 /* Get rid of any cached register state */
523 deactivate_mm(tsk, mm);
524
525 /* notify parent sleeping on vfork() */
526 if (vfork_done) {
527 tsk->vfork_done = NULL;
528 complete(vfork_done);
529 }
Roland McGrathfec1d012006-12-06 20:36:34 -0800530
531 /*
532 * If we're exiting normally, clear a user-space tid field if
533 * requested. We leave this alone when dying by signal, to leave
534 * the value intact in a core dump, and to save the unnecessary
535 * trouble otherwise. Userland only wants this done for a sys_exit.
536 */
537 if (tsk->clear_child_tid
538 && !(tsk->flags & PF_SIGNALED)
539 && atomic_read(&mm->mm_users) > 1) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 u32 __user * tidptr = tsk->clear_child_tid;
541 tsk->clear_child_tid = NULL;
542
543 /*
544 * We don't check the error code - if userspace has
545 * not set up a proper pointer then tough luck.
546 */
547 put_user(0, tidptr);
548 sys_futex(tidptr, FUTEX_WAKE, 1, NULL, NULL, 0);
549 }
550}
551
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800552/*
553 * Allocate a new mm structure and copy contents from the
554 * mm structure of the passed in task structure.
555 */
Carsten Otte402b0862008-03-25 18:47:10 +0100556struct mm_struct *dup_mm(struct task_struct *tsk)
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800557{
558 struct mm_struct *mm, *oldmm = current->mm;
559 int err;
560
561 if (!oldmm)
562 return NULL;
563
564 mm = allocate_mm();
565 if (!mm)
566 goto fail_nomem;
567
568 memcpy(mm, oldmm, sizeof(*mm));
569
Ashwin Chaugule7602bdf2006-12-06 20:31:57 -0800570 /* Initializing for Swap token stuff */
571 mm->token_priority = 0;
572 mm->last_interval = 0;
573
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800574 if (!mm_init(mm, tsk))
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800575 goto fail_nomem;
576
577 if (init_new_context(tsk, mm))
578 goto fail_nocontext;
579
Matt Helsley925d1c42008-04-29 01:01:36 -0700580 dup_mm_exe_file(oldmm, mm);
581
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800582 err = dup_mmap(mm, oldmm);
583 if (err)
584 goto free_pt;
585
586 mm->hiwater_rss = get_mm_rss(mm);
587 mm->hiwater_vm = mm->total_vm;
588
589 return mm;
590
591free_pt:
592 mmput(mm);
593
594fail_nomem:
595 return NULL;
596
597fail_nocontext:
598 /*
599 * If init_new_context() failed, we cannot use mmput() to free the mm
600 * because it calls destroy_context()
601 */
602 mm_free_pgd(mm);
603 free_mm(mm);
604 return NULL;
605}
606
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607static int copy_mm(unsigned long clone_flags, struct task_struct * tsk)
608{
609 struct mm_struct * mm, *oldmm;
610 int retval;
611
612 tsk->min_flt = tsk->maj_flt = 0;
613 tsk->nvcsw = tsk->nivcsw = 0;
614
615 tsk->mm = NULL;
616 tsk->active_mm = NULL;
617
618 /*
619 * Are we cloning a kernel thread?
620 *
621 * We need to steal a active VM for that..
622 */
623 oldmm = current->mm;
624 if (!oldmm)
625 return 0;
626
627 if (clone_flags & CLONE_VM) {
628 atomic_inc(&oldmm->mm_users);
629 mm = oldmm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 goto good_mm;
631 }
632
633 retval = -ENOMEM;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800634 mm = dup_mm(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 if (!mm)
636 goto fail_nomem;
637
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638good_mm:
Ashwin Chaugule7602bdf2006-12-06 20:31:57 -0800639 /* Initializing for Swap token stuff */
640 mm->token_priority = 0;
641 mm->last_interval = 0;
642
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 tsk->mm = mm;
644 tsk->active_mm = mm;
645 return 0;
646
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647fail_nomem:
648 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649}
650
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700651static struct fs_struct *__copy_fs_struct(struct fs_struct *old)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652{
653 struct fs_struct *fs = kmem_cache_alloc(fs_cachep, GFP_KERNEL);
654 /* We don't need to lock fs - think why ;-) */
655 if (fs) {
656 atomic_set(&fs->count, 1);
657 rwlock_init(&fs->lock);
658 fs->umask = old->umask;
659 read_lock(&old->lock);
Jan Blunck6ac08c32008-02-14 19:34:38 -0800660 fs->root = old->root;
661 path_get(&old->root);
662 fs->pwd = old->pwd;
663 path_get(&old->pwd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 read_unlock(&old->lock);
665 }
666 return fs;
667}
668
669struct fs_struct *copy_fs_struct(struct fs_struct *old)
670{
671 return __copy_fs_struct(old);
672}
673
674EXPORT_SYMBOL_GPL(copy_fs_struct);
675
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700676static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677{
678 if (clone_flags & CLONE_FS) {
679 atomic_inc(&current->fs->count);
680 return 0;
681 }
682 tsk->fs = __copy_fs_struct(current->fs);
683 if (!tsk->fs)
684 return -ENOMEM;
685 return 0;
686}
687
JANAK DESAIa016f332006-02-07 12:59:02 -0800688static int copy_files(unsigned long clone_flags, struct task_struct * tsk)
689{
690 struct files_struct *oldf, *newf;
691 int error = 0;
692
693 /*
694 * A background process may not have any files ...
695 */
696 oldf = current->files;
697 if (!oldf)
698 goto out;
699
700 if (clone_flags & CLONE_FILES) {
701 atomic_inc(&oldf->count);
702 goto out;
703 }
704
JANAK DESAIa016f332006-02-07 12:59:02 -0800705 newf = dup_fd(oldf, &error);
706 if (!newf)
707 goto out;
708
709 tsk->files = newf;
710 error = 0;
711out:
712 return error;
713}
714
Jens Axboefadad8782008-01-24 08:54:47 +0100715static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
Jens Axboefd0928d2008-01-24 08:52:45 +0100716{
717#ifdef CONFIG_BLOCK
718 struct io_context *ioc = current->io_context;
719
720 if (!ioc)
721 return 0;
Jens Axboefadad8782008-01-24 08:54:47 +0100722 /*
723 * Share io context with parent, if CLONE_IO is set
724 */
725 if (clone_flags & CLONE_IO) {
726 tsk->io_context = ioc_task_link(ioc);
727 if (unlikely(!tsk->io_context))
728 return -ENOMEM;
729 } else if (ioprio_valid(ioc->ioprio)) {
Jens Axboefd0928d2008-01-24 08:52:45 +0100730 tsk->io_context = alloc_io_context(GFP_KERNEL, -1);
731 if (unlikely(!tsk->io_context))
732 return -ENOMEM;
733
Jens Axboefd0928d2008-01-24 08:52:45 +0100734 tsk->io_context->ioprio = ioc->ioprio;
735 }
736#endif
737 return 0;
738}
739
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700740static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741{
742 struct sighand_struct *sig;
743
744 if (clone_flags & (CLONE_SIGHAND | CLONE_THREAD)) {
745 atomic_inc(&current->sighand->count);
746 return 0;
747 }
748 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
Ingo Molnare56d0902006-01-08 01:01:37 -0800749 rcu_assign_pointer(tsk->sighand, sig);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 if (!sig)
751 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 atomic_set(&sig->count, 1);
753 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
754 return 0;
755}
756
Oleg Nesterova7e53282006-03-28 16:11:27 -0800757void __cleanup_sighand(struct sighand_struct *sighand)
Oleg Nesterovc81addc2006-03-28 16:11:17 -0800758{
Oleg Nesterovc81addc2006-03-28 16:11:17 -0800759 if (atomic_dec_and_test(&sighand->count))
760 kmem_cache_free(sighand_cachep, sighand);
761}
762
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700763static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764{
765 struct signal_struct *sig;
766 int ret;
767
768 if (clone_flags & CLONE_THREAD) {
769 atomic_inc(&current->signal->count);
770 atomic_inc(&current->signal->live);
771 return 0;
772 }
773 sig = kmem_cache_alloc(signal_cachep, GFP_KERNEL);
774 tsk->signal = sig;
775 if (!sig)
776 return -ENOMEM;
777
778 ret = copy_thread_group_keys(tsk);
779 if (ret < 0) {
780 kmem_cache_free(signal_cachep, sig);
781 return ret;
782 }
783
784 atomic_set(&sig->count, 1);
785 atomic_set(&sig->live, 1);
786 init_waitqueue_head(&sig->wait_chldexit);
787 sig->flags = 0;
788 sig->group_exit_code = 0;
789 sig->group_exit_task = NULL;
790 sig->group_stop_count = 0;
Oleg Nesterovdb51aec2008-04-30 00:52:52 -0700791 sig->curr_target = tsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 init_sigpending(&sig->shared_pending);
793 INIT_LIST_HEAD(&sig->posix_timers);
794
Thomas Gleixnerc9cb2e32007-02-16 01:27:49 -0800795 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
Thomas Gleixner2ff678b2006-01-09 20:52:34 -0800796 sig->it_real_incr.tv64 = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 sig->real_timer.function = it_real_fn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798
799 sig->it_virt_expires = cputime_zero;
800 sig->it_virt_incr = cputime_zero;
801 sig->it_prof_expires = cputime_zero;
802 sig->it_prof_incr = cputime_zero;
803
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 sig->leader = 0; /* session leadership doesn't inherit */
Eric W. Biedermanab521dc2007-02-12 00:53:00 -0800805 sig->tty_old_pgrp = NULL;
Alan Cox9c9f4de2008-10-13 10:37:26 +0100806 sig->tty = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807
808 sig->utime = sig->stime = sig->cutime = sig->cstime = cputime_zero;
Laurent Vivier9ac52312007-10-15 17:00:19 +0200809 sig->gtime = cputime_zero;
810 sig->cgtime = cputime_zero;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811 sig->nvcsw = sig->nivcsw = sig->cnvcsw = sig->cnivcsw = 0;
812 sig->min_flt = sig->maj_flt = sig->cmin_flt = sig->cmaj_flt = 0;
Eric Dumazet6eaeeab2007-05-10 22:22:37 -0700813 sig->inblock = sig->oublock = sig->cinblock = sig->coublock = 0;
Andrea Righi59954772008-07-27 17:29:15 +0200814 task_io_accounting_init(&sig->ioac);
Balbir Singh172ba842007-07-09 18:52:00 +0200815 sig->sum_sched_runtime = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 INIT_LIST_HEAD(&sig->cpu_timers[0]);
817 INIT_LIST_HEAD(&sig->cpu_timers[1]);
818 INIT_LIST_HEAD(&sig->cpu_timers[2]);
Shailabh Nagarad4ecbc2006-07-14 00:24:44 -0700819 taskstats_tgid_init(sig);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820
821 task_lock(current->group_leader);
822 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
823 task_unlock(current->group_leader);
824
825 if (sig->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) {
826 /*
827 * New sole thread in the process gets an expiry time
828 * of the whole CPU time limit.
829 */
830 tsk->it_prof_expires =
831 secs_to_cputime(sig->rlim[RLIMIT_CPU].rlim_cur);
832 }
KaiGai Kohei0e464812006-06-25 05:49:24 -0700833 acct_init_pacct(&sig->pacct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834
Miloslav Trmac522ed772007-07-15 23:40:56 -0700835 tty_audit_fork(sig);
836
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837 return 0;
838}
839
Oleg Nesterov6b3934e2006-03-28 16:11:16 -0800840void __cleanup_signal(struct signal_struct *sig)
841{
842 exit_thread_group_keys(sig);
Alan Cox9c9f4de2008-10-13 10:37:26 +0100843 tty_kref_put(sig->tty);
Oleg Nesterov6b3934e2006-03-28 16:11:16 -0800844 kmem_cache_free(signal_cachep, sig);
845}
846
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700847static void cleanup_signal(struct task_struct *tsk)
Oleg Nesterov6b3934e2006-03-28 16:11:16 -0800848{
849 struct signal_struct *sig = tsk->signal;
850
851 atomic_dec(&sig->live);
852
853 if (atomic_dec_and_test(&sig->count))
854 __cleanup_signal(sig);
855}
856
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700857static void copy_flags(unsigned long clone_flags, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858{
859 unsigned long new_flags = p->flags;
860
Rafael J. Wysocki83144182007-07-17 04:03:35 -0700861 new_flags &= ~PF_SUPERPRIV;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 new_flags |= PF_FORKNOEXEC;
Roland McGrath09a05392008-07-25 19:45:47 -0700863 new_flags |= PF_STARTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 p->flags = new_flags;
Rafael J. Wysocki2e131892007-10-18 03:04:45 -0700865 clear_freeze_flag(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866}
867
868asmlinkage long sys_set_tid_address(int __user *tidptr)
869{
870 current->clear_child_tid = tidptr;
871
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700872 return task_pid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873}
874
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700875static void rt_mutex_init_task(struct task_struct *p)
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700876{
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700877 spin_lock_init(&p->pi_lock);
Zilvinas Valinskase29e1752007-03-16 13:38:34 -0800878#ifdef CONFIG_RT_MUTEXES
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700879 plist_head_init(&p->pi_waiters, &p->pi_lock);
880 p->pi_blocked_on = NULL;
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700881#endif
882}
883
Balbir Singhcf475ad2008-04-29 01:00:16 -0700884#ifdef CONFIG_MM_OWNER
885void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
886{
887 mm->owner = p;
888}
889#endif /* CONFIG_MM_OWNER */
890
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891/*
892 * This creates a new process as a copy of the old one,
893 * but does not actually start it yet.
894 *
895 * It copies the registers, and all the appropriate
896 * parts of the process environment (as per the clone
897 * flags). The actual kick-off is left to the caller.
898 */
Ingo Molnar36c8b582006-07-03 00:25:41 -0700899static struct task_struct *copy_process(unsigned long clone_flags,
900 unsigned long stack_start,
901 struct pt_regs *regs,
902 unsigned long stack_size,
Ingo Molnar36c8b582006-07-03 00:25:41 -0700903 int __user *child_tidptr,
Roland McGrath09a05392008-07-25 19:45:47 -0700904 struct pid *pid,
905 int trace)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906{
907 int retval;
Mariusz Kozlowskia24efe62007-10-18 23:41:09 -0700908 struct task_struct *p;
Paul Menageb4f48b62007-10-18 23:39:33 -0700909 int cgroup_callbacks_done = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910
911 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
912 return ERR_PTR(-EINVAL);
913
914 /*
915 * Thread groups must share signals as well, and detached threads
916 * can only be started up within the thread group.
917 */
918 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
919 return ERR_PTR(-EINVAL);
920
921 /*
922 * Shared signal handlers imply shared VM. By way of the above,
923 * thread groups also imply shared VM. Blocking this case allows
924 * for various simplifications in other code.
925 */
926 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
927 return ERR_PTR(-EINVAL);
928
929 retval = security_task_create(clone_flags);
930 if (retval)
931 goto fork_out;
932
933 retval = -ENOMEM;
934 p = dup_task_struct(current);
935 if (!p)
936 goto fork_out;
937
Peter Zijlstrabea493a2006-10-17 00:10:33 -0700938 rt_mutex_init_task(p);
939
Ingo Molnard12c1a32008-07-14 12:09:28 +0200940#ifdef CONFIG_PROVE_LOCKING
Ingo Molnarde30a2b2006-07-03 00:24:42 -0700941 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
942 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
943#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 retval = -EAGAIN;
945 if (atomic_read(&p->user->processes) >=
946 p->signal->rlim[RLIMIT_NPROC].rlim_cur) {
947 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
Cedric Le Goateracce2922007-07-15 23:40:59 -0700948 p->user != current->nsproxy->user_ns->root_user)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 goto bad_fork_free;
950 }
951
952 atomic_inc(&p->user->__count);
953 atomic_inc(&p->user->processes);
954 get_group_info(p->group_info);
955
956 /*
957 * If multiple threads are within copy_process(), then this check
958 * triggers too late. This doesn't hurt, the check is only there
959 * to stop root fork bombs.
960 */
961 if (nr_threads >= max_threads)
962 goto bad_fork_cleanup_count;
963
Al Viroa1261f52005-11-13 16:06:55 -0800964 if (!try_module_get(task_thread_info(p)->exec_domain->module))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 goto bad_fork_cleanup_count;
966
967 if (p->binfmt && !try_module_get(p->binfmt->module))
968 goto bad_fork_cleanup_put_domain;
969
970 p->did_exec = 0;
Shailabh Nagarca74e922006-07-14 00:24:36 -0700971 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 copy_flags(clone_flags, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 INIT_LIST_HEAD(&p->children);
974 INIT_LIST_HEAD(&p->sibling);
Paul E. McKenneye260be62008-01-25 21:08:24 +0100975#ifdef CONFIG_PREEMPT_RCU
976 p->rcu_read_lock_nesting = 0;
977 p->rcu_flipctr_idx = 0;
978#endif /* #ifdef CONFIG_PREEMPT_RCU */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 p->vfork_done = NULL;
980 spin_lock_init(&p->alloc_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981
982 clear_tsk_thread_flag(p, TIF_SIGPENDING);
983 init_sigpending(&p->pending);
984
985 p->utime = cputime_zero;
986 p->stime = cputime_zero;
Laurent Vivier9ac52312007-10-15 17:00:19 +0200987 p->gtime = cputime_zero;
Michael Neulingc66f08b2007-10-18 03:06:34 -0700988 p->utimescaled = cputime_zero;
989 p->stimescaled = cputime_zero;
Peter Zijlstra73a2bcb2007-10-29 21:18:11 +0100990 p->prev_utime = cputime_zero;
Balbir Singh93018992007-10-30 00:26:32 +0100991 p->prev_stime = cputime_zero;
Balbir Singh172ba842007-07-09 18:52:00 +0200992
Ingo Molnar82a1fcb2008-01-25 21:08:02 +0100993#ifdef CONFIG_DETECT_SOFTLOCKUP
994 p->last_switch_count = 0;
995 p->last_switch_timestamp = 0;
996#endif
997
Andrea Righi59954772008-07-27 17:29:15 +0200998 task_io_accounting_init(&p->ioac);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 acct_clear_integrals(p);
1000
Daniel Walker23ff4442007-10-18 03:06:07 -07001001 p->it_virt_expires = cputime_zero;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 p->it_prof_expires = cputime_zero;
Daniel Walker23ff4442007-10-18 03:06:07 -07001003 p->it_sched_expires = 0;
1004 INIT_LIST_HEAD(&p->cpu_timers[0]);
1005 INIT_LIST_HEAD(&p->cpu_timers[1]);
1006 INIT_LIST_HEAD(&p->cpu_timers[2]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007
1008 p->lock_depth = -1; /* -1 = no lock */
1009 do_posix_clock_monotonic_gettime(&p->start_time);
Tomas Janousek924b42d2007-07-15 23:39:42 -07001010 p->real_start_time = p->start_time;
1011 monotonic_to_bootbased(&p->real_start_time);
Alexey Dobriyan57c521c2007-10-16 23:31:35 -07001012#ifdef CONFIG_SECURITY
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 p->security = NULL;
Alexey Dobriyan57c521c2007-10-16 23:31:35 -07001014#endif
Serge E. Hallyn3b7391d2008-02-04 22:29:45 -08001015 p->cap_bset = current->cap_bset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 p->io_context = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 p->audit_context = NULL;
Paul Menageb4f48b62007-10-18 23:39:33 -07001018 cgroup_fork(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019#ifdef CONFIG_NUMA
Lee Schermerhorn846a16b2008-04-28 02:13:09 -07001020 p->mempolicy = mpol_dup(p->mempolicy);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 if (IS_ERR(p->mempolicy)) {
1022 retval = PTR_ERR(p->mempolicy);
1023 p->mempolicy = NULL;
Paul Menageb4f48b62007-10-18 23:39:33 -07001024 goto bad_fork_cleanup_cgroup;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 }
Paul Jacksonc61afb12006-03-24 03:16:08 -08001026 mpol_fix_fork_child_flag(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027#endif
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001028#ifdef CONFIG_TRACE_IRQFLAGS
1029 p->irq_events = 0;
Russell Kingb36e4752006-08-27 12:26:34 +01001030#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1031 p->hardirqs_enabled = 1;
1032#else
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001033 p->hardirqs_enabled = 0;
Russell Kingb36e4752006-08-27 12:26:34 +01001034#endif
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001035 p->hardirq_enable_ip = 0;
1036 p->hardirq_enable_event = 0;
1037 p->hardirq_disable_ip = _THIS_IP_;
1038 p->hardirq_disable_event = 0;
1039 p->softirqs_enabled = 1;
1040 p->softirq_enable_ip = _THIS_IP_;
1041 p->softirq_enable_event = 0;
1042 p->softirq_disable_ip = 0;
1043 p->softirq_disable_event = 0;
1044 p->hardirq_context = 0;
1045 p->softirq_context = 0;
1046#endif
Ingo Molnarfbb9ce952006-07-03 00:24:50 -07001047#ifdef CONFIG_LOCKDEP
1048 p->lockdep_depth = 0; /* no locks held yet */
1049 p->curr_chain_key = 0;
1050 p->lockdep_recursion = 0;
1051#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052
Ingo Molnar408894e2006-01-09 15:59:20 -08001053#ifdef CONFIG_DEBUG_MUTEXES
1054 p->blocked_on = NULL; /* not blocked yet */
1055#endif
1056
Srivatsa Vaddagiri3c90e6e2007-11-09 22:39:39 +01001057 /* Perform scheduler related setup. Assign this task to a CPU. */
1058 sched_fork(p, clone_flags);
1059
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 if ((retval = security_task_alloc(p)))
1061 goto bad_fork_cleanup_policy;
1062 if ((retval = audit_alloc(p)))
1063 goto bad_fork_cleanup_security;
1064 /* copy all the process information */
1065 if ((retval = copy_semundo(clone_flags, p)))
1066 goto bad_fork_cleanup_audit;
1067 if ((retval = copy_files(clone_flags, p)))
1068 goto bad_fork_cleanup_semundo;
1069 if ((retval = copy_fs(clone_flags, p)))
1070 goto bad_fork_cleanup_files;
1071 if ((retval = copy_sighand(clone_flags, p)))
1072 goto bad_fork_cleanup_fs;
1073 if ((retval = copy_signal(clone_flags, p)))
1074 goto bad_fork_cleanup_sighand;
1075 if ((retval = copy_mm(clone_flags, p)))
1076 goto bad_fork_cleanup_signal;
1077 if ((retval = copy_keys(clone_flags, p)))
1078 goto bad_fork_cleanup_mm;
Serge E. Hallynab516012006-10-02 02:18:06 -07001079 if ((retval = copy_namespaces(clone_flags, p)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 goto bad_fork_cleanup_keys;
Jens Axboefadad8782008-01-24 08:54:47 +01001081 if ((retval = copy_io(clone_flags, p)))
Jens Axboefd0928d2008-01-24 08:52:45 +01001082 goto bad_fork_cleanup_namespaces;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 retval = copy_thread(0, clone_flags, stack_start, stack_size, p, regs);
1084 if (retval)
Jens Axboefd0928d2008-01-24 08:52:45 +01001085 goto bad_fork_cleanup_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001087 if (pid != &init_struct_pid) {
1088 retval = -ENOMEM;
1089 pid = alloc_pid(task_active_pid_ns(p));
1090 if (!pid)
Jens Axboefd0928d2008-01-24 08:52:45 +01001091 goto bad_fork_cleanup_io;
Pavel Emelyanov6f4e6432007-10-18 23:40:11 -07001092
1093 if (clone_flags & CLONE_NEWPID) {
1094 retval = pid_ns_prepare_proc(task_active_pid_ns(p));
1095 if (retval < 0)
1096 goto bad_fork_free_pid;
1097 }
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001098 }
1099
1100 p->pid = pid_nr(pid);
1101 p->tgid = p->pid;
1102 if (clone_flags & CLONE_THREAD)
1103 p->tgid = current->tgid;
1104
Serge E. Hallyne885dcd2008-07-25 01:47:06 -07001105 if (current->nsproxy != p->nsproxy) {
1106 retval = ns_cgroup_clone(p, pid);
1107 if (retval)
1108 goto bad_fork_free_pid;
1109 }
1110
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1112 /*
1113 * Clear TID on mm_release()?
1114 */
1115 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr: NULL;
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001116#ifdef CONFIG_FUTEX
Ingo Molnar8f17d3a2006-03-27 01:16:27 -08001117 p->robust_list = NULL;
1118#ifdef CONFIG_COMPAT
1119 p->compat_robust_list = NULL;
1120#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07001121 INIT_LIST_HEAD(&p->pi_state_list);
1122 p->pi_state_cache = NULL;
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001123#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 /*
GOTO Masanorif9a38792006-03-13 21:20:44 -08001125 * sigaltstack should be cleared when sharing the same VM
1126 */
1127 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1128 p->sas_ss_sp = p->sas_ss_size = 0;
1129
1130 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 * Syscall tracing should be turned off in the child regardless
1132 * of CLONE_PTRACE.
1133 */
1134 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
Laurent Viviered75e8d2005-09-03 15:57:18 -07001135#ifdef TIF_SYSCALL_EMU
1136 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1137#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01001138 clear_all_latency_tracing(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139
1140 /* Our parent execution domain becomes current domain
1141 These must match for thread signalling to apply */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 p->parent_exec_id = p->self_exec_id;
1143
1144 /* ok, now we should be set up.. */
1145 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1146 p->pdeath_signal = 0;
1147 p->exit_state = 0;
1148
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 /*
1150 * Ok, make it visible to the rest of the system.
1151 * We dont wake it up yet.
1152 */
1153 p->group_leader = p;
Oleg Nesterov47e65322006-03-28 16:11:25 -08001154 INIT_LIST_HEAD(&p->thread_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155
Paul Menageb4f48b62007-10-18 23:39:33 -07001156 /* Now that the task is set up, run cgroup callbacks if
1157 * necessary. We need to run them before the task is visible
1158 * on the tasklist. */
1159 cgroup_fork_callbacks(p);
1160 cgroup_callbacks_done = 1;
1161
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 /* Need tasklist lock for parent etc handling! */
1163 write_lock_irq(&tasklist_lock);
1164
1165 /*
Nick Piggin476d1392005-06-25 14:57:29 -07001166 * The task hasn't been attached yet, so its cpus_allowed mask will
1167 * not be changed, nor will its assigned CPU.
1168 *
1169 * The cpus_allowed mask of the parent may have changed after it was
1170 * copied first time - so re-copy it here, then check the child's CPU
1171 * to ensure it is on a valid CPU (and if not, just force it back to
1172 * parent's CPU). This avoids alot of nasty races.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 */
1174 p->cpus_allowed = current->cpus_allowed;
Peter Zijlstra6f505b12008-01-25 21:08:30 +01001175 p->rt.nr_cpus_allowed = current->rt.nr_cpus_allowed;
Srivatsa Vaddagiri26ff6ad2005-09-16 19:27:40 -07001176 if (unlikely(!cpu_isset(task_cpu(p), p->cpus_allowed) ||
1177 !cpu_online(task_cpu(p))))
Nick Piggin476d1392005-06-25 14:57:29 -07001178 set_task_cpu(p, smp_processor_id());
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 /* CLONE_PARENT re-uses the old parent */
1181 if (clone_flags & (CLONE_PARENT|CLONE_THREAD))
1182 p->real_parent = current->real_parent;
1183 else
1184 p->real_parent = current;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185
Oleg Nesterov3f17da62006-02-15 22:13:24 +03001186 spin_lock(&current->sighand->siglock);
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001187
1188 /*
1189 * Process group and session signals need to be delivered to just the
1190 * parent before the fork or both the parent and the child after the
1191 * fork. Restart if a signal comes in before we add the new process to
1192 * it's process group.
1193 * A fatal signal pending means that current will exit, so the new
1194 * thread can't slip out of an OOM kill (or normal SIGKILL).
1195 */
Daniel Walker23ff4442007-10-18 03:06:07 -07001196 recalc_sigpending();
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001197 if (signal_pending(current)) {
1198 spin_unlock(&current->sighand->siglock);
1199 write_unlock_irq(&tasklist_lock);
1200 retval = -ERESTARTNOINTR;
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001201 goto bad_fork_free_pid;
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001202 }
1203
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204 if (clone_flags & CLONE_THREAD) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 p->group_leader = current->group_leader;
Oleg Nesterov47e65322006-03-28 16:11:25 -08001206 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208 if (!cputime_eq(current->signal->it_virt_expires,
1209 cputime_zero) ||
1210 !cputime_eq(current->signal->it_prof_expires,
1211 cputime_zero) ||
1212 current->signal->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY ||
1213 !list_empty(&current->signal->cpu_timers[0]) ||
1214 !list_empty(&current->signal->cpu_timers[1]) ||
1215 !list_empty(&current->signal->cpu_timers[2])) {
1216 /*
1217 * Have child wake up on its first tick to check
1218 * for process CPU timers.
1219 */
1220 p->it_prof_expires = jiffies_to_cputime(1);
1221 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222 }
1223
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001224 if (likely(p->pid)) {
Roland McGrathf4700212008-03-24 18:36:23 -07001225 list_add_tail(&p->sibling, &p->real_parent->children);
Roland McGrath09a05392008-07-25 19:45:47 -07001226 tracehook_finish_clone(p, clone_flags, trace);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001228 if (thread_group_leader(p)) {
Eric W. Biederman5cd17562007-12-04 23:45:04 -08001229 if (clone_flags & CLONE_NEWPID)
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001230 p->nsproxy->pid_ns->child_reaper = p;
Oleg Nesterovc97d9892006-03-28 16:11:06 -08001231
Oleg Nesterovfea9d172008-02-08 04:19:19 -08001232 p->signal->leader_pid = pid;
Alan Cox9c9f4de2008-10-13 10:37:26 +01001233 tty_kref_put(p->signal->tty);
1234 p->signal->tty = tty_kref_get(current->signal->tty);
Eric W. Biederman5cd17562007-12-04 23:45:04 -08001235 set_task_pgrp(p, task_pgrp_nr(current));
1236 set_task_session(p, task_session_nr(current));
1237 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1238 attach_pid(p, PIDTYPE_SID, task_session(current));
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -07001239 list_add_tail_rcu(&p->tasks, &init_task.tasks);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 __get_cpu_var(process_counts)++;
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001241 }
Sukadev Bhattiprolu85868992007-05-10 22:23:03 -07001242 attach_pid(p, PIDTYPE_PID, pid);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001243 nr_threads++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244 }
1245
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 total_forks++;
Oleg Nesterov3f17da62006-02-15 22:13:24 +03001247 spin_unlock(&current->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 write_unlock_irq(&tasklist_lock);
Andrew Mortonc13cf852005-11-28 13:43:48 -08001249 proc_fork_connector(p);
Paul Menage817929e2007-10-18 23:39:36 -07001250 cgroup_post_fork(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 return p;
1252
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001253bad_fork_free_pid:
1254 if (pid != &init_struct_pid)
1255 free_pid(pid);
Jens Axboefd0928d2008-01-24 08:52:45 +01001256bad_fork_cleanup_io:
1257 put_io_context(p->io_context);
Serge E. Hallynab516012006-10-02 02:18:06 -07001258bad_fork_cleanup_namespaces:
Linus Torvalds444f3782007-01-30 13:35:18 -08001259 exit_task_namespaces(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260bad_fork_cleanup_keys:
1261 exit_keys(p);
1262bad_fork_cleanup_mm:
1263 if (p->mm)
1264 mmput(p->mm);
1265bad_fork_cleanup_signal:
Oleg Nesterov6b3934e2006-03-28 16:11:16 -08001266 cleanup_signal(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267bad_fork_cleanup_sighand:
Oleg Nesterova7e53282006-03-28 16:11:27 -08001268 __cleanup_sighand(p->sighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269bad_fork_cleanup_fs:
1270 exit_fs(p); /* blocking */
1271bad_fork_cleanup_files:
1272 exit_files(p); /* blocking */
1273bad_fork_cleanup_semundo:
1274 exit_sem(p);
1275bad_fork_cleanup_audit:
1276 audit_free(p);
1277bad_fork_cleanup_security:
1278 security_task_free(p);
1279bad_fork_cleanup_policy:
1280#ifdef CONFIG_NUMA
Lee Schermerhornf0be3d32008-04-28 02:13:08 -07001281 mpol_put(p->mempolicy);
Paul Menageb4f48b62007-10-18 23:39:33 -07001282bad_fork_cleanup_cgroup:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283#endif
Paul Menageb4f48b62007-10-18 23:39:33 -07001284 cgroup_exit(p, cgroup_callbacks_done);
Shailabh Nagar35df17c2006-08-31 21:27:38 -07001285 delayacct_tsk_free(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286 if (p->binfmt)
1287 module_put(p->binfmt->module);
1288bad_fork_cleanup_put_domain:
Al Viroa1261f52005-11-13 16:06:55 -08001289 module_put(task_thread_info(p)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290bad_fork_cleanup_count:
1291 put_group_info(p->group_info);
1292 atomic_dec(&p->user->processes);
1293 free_uid(p->user);
1294bad_fork_free:
1295 free_task(p);
Oleg Nesterovfe7d37d2006-01-08 01:04:02 -08001296fork_out:
1297 return ERR_PTR(retval);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298}
1299
Adrian Bunk6b2fb3c2008-02-06 01:37:55 -08001300noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301{
1302 memset(regs, 0, sizeof(struct pt_regs));
1303 return regs;
1304}
1305
Al Viro9abcf402007-02-01 13:52:48 +00001306struct task_struct * __cpuinit fork_idle(int cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307{
Ingo Molnar36c8b582006-07-03 00:25:41 -07001308 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309 struct pt_regs regs;
1310
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001311 task = copy_process(CLONE_VM, 0, idle_regs(&regs), 0, NULL,
Roland McGrath09a05392008-07-25 19:45:47 -07001312 &init_struct_pid, 0);
Akinobu Mita753ca4f2006-11-25 11:09:34 -08001313 if (!IS_ERR(task))
1314 init_idle(task, cpu);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001315
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 return task;
1317}
1318
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319/*
1320 * Ok, this is the main fork-routine.
1321 *
1322 * It copies the process, and if successful kick-starts
1323 * it and waits for it to finish using the VM if required.
1324 */
1325long do_fork(unsigned long clone_flags,
1326 unsigned long stack_start,
1327 struct pt_regs *regs,
1328 unsigned long stack_size,
1329 int __user *parent_tidptr,
1330 int __user *child_tidptr)
1331{
1332 struct task_struct *p;
1333 int trace = 0;
Eric W. Biederman92476d72006-03-31 02:31:42 -08001334 long nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335
Andrew Mortonbdff7462008-02-04 22:27:22 -08001336 /*
1337 * We hope to recycle these flags after 2.6.26
1338 */
1339 if (unlikely(clone_flags & CLONE_STOPPED)) {
1340 static int __read_mostly count = 100;
1341
1342 if (count > 0 && printk_ratelimit()) {
1343 char comm[TASK_COMM_LEN];
1344
1345 count--;
1346 printk(KERN_INFO "fork(): process `%s' used deprecated "
1347 "clone flags 0x%lx\n",
1348 get_task_comm(comm, current),
1349 clone_flags & CLONE_STOPPED);
1350 }
1351 }
1352
Roland McGrath09a05392008-07-25 19:45:47 -07001353 /*
1354 * When called from kernel_thread, don't do user tracing stuff.
1355 */
1356 if (likely(user_mode(regs)))
1357 trace = tracehook_prepare_clone(clone_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358
Sukadev Bhattiprolua6f5e062007-10-18 23:39:53 -07001359 p = copy_process(clone_flags, stack_start, regs, stack_size,
Roland McGrath09a05392008-07-25 19:45:47 -07001360 child_tidptr, NULL, trace);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 /*
1362 * Do this prior waking up the new thread - the thread pointer
1363 * might get invalid after that point, if the thread exits quickly.
1364 */
1365 if (!IS_ERR(p)) {
1366 struct completion vfork;
1367
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04001368 trace_sched_process_fork(current, p);
1369
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001370 nr = task_pid_vnr(p);
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001371
1372 if (clone_flags & CLONE_PARENT_SETTID)
1373 put_user(nr, parent_tidptr);
Sukadev Bhattiprolua6f5e062007-10-18 23:39:53 -07001374
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 if (clone_flags & CLONE_VFORK) {
1376 p->vfork_done = &vfork;
1377 init_completion(&vfork);
1378 }
1379
Roland McGrath09a05392008-07-25 19:45:47 -07001380 tracehook_report_clone(trace, regs, clone_flags, nr, p);
1381
1382 /*
1383 * We set PF_STARTING at creation in case tracing wants to
1384 * use this to distinguish a fully live task from one that
1385 * hasn't gotten to tracehook_report_clone() yet. Now we
1386 * clear it and set the child going.
1387 */
1388 p->flags &= ~PF_STARTING;
1389
1390 if (unlikely(clone_flags & CLONE_STOPPED)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 /*
1392 * We'll start up with an immediate SIGSTOP.
1393 */
1394 sigaddset(&p->pending.signal, SIGSTOP);
1395 set_tsk_thread_flag(p, TIF_SIGPENDING);
Oleg Nesterovd9ae90a2008-02-06 01:36:13 -08001396 __set_task_state(p, TASK_STOPPED);
Roland McGrath09a05392008-07-25 19:45:47 -07001397 } else {
1398 wake_up_new_task(p, clone_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 }
1400
Roland McGrath09a05392008-07-25 19:45:47 -07001401 tracehook_report_clone_complete(trace, regs,
1402 clone_flags, nr, p);
1403
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 if (clone_flags & CLONE_VFORK) {
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -07001405 freezer_do_not_count();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 wait_for_completion(&vfork);
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -07001407 freezer_count();
Roland McGrathdaded342008-07-25 19:45:47 -07001408 tracehook_report_vfork_done(p, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 }
1410 } else {
Eric W. Biederman92476d72006-03-31 02:31:42 -08001411 nr = PTR_ERR(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 }
Eric W. Biederman92476d72006-03-31 02:31:42 -08001413 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414}
1415
Ravikiran G Thirumalai5fd63b32006-01-11 22:46:15 +01001416#ifndef ARCH_MIN_MMSTRUCT_ALIGN
1417#define ARCH_MIN_MMSTRUCT_ALIGN 0
1418#endif
1419
Alexey Dobriyan51cc5062008-07-25 19:45:34 -07001420static void sighand_ctor(void *data)
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08001421{
1422 struct sighand_struct *sighand = data;
1423
Christoph Lametera35afb82007-05-16 22:10:57 -07001424 spin_lock_init(&sighand->siglock);
Davide Libenzib8fceee2007-09-20 12:40:16 -07001425 init_waitqueue_head(&sighand->signalfd_wqh);
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08001426}
1427
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428void __init proc_caches_init(void)
1429{
1430 sighand_cachep = kmem_cache_create("sighand_cache",
1431 sizeof(struct sighand_struct), 0,
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08001432 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU,
Paul Mundt20c2df82007-07-20 10:11:58 +09001433 sighand_ctor);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434 signal_cachep = kmem_cache_create("signal_cache",
1435 sizeof(struct signal_struct), 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09001436 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
1437 files_cachep = kmem_cache_create("files_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 sizeof(struct files_struct), 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09001439 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
1440 fs_cachep = kmem_cache_create("fs_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 sizeof(struct fs_struct), 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09001442 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 vm_area_cachep = kmem_cache_create("vm_area_struct",
1444 sizeof(struct vm_area_struct), 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09001445 SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446 mm_cachep = kmem_cache_create("mm_struct",
Ravikiran G Thirumalai5fd63b32006-01-11 22:46:15 +01001447 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
Paul Mundt20c2df82007-07-20 10:11:58 +09001448 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449}
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001450
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001451/*
1452 * Check constraints on flags passed to the unshare system call and
1453 * force unsharing of additional process context as appropriate.
1454 */
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001455static void check_unshare_flags(unsigned long *flags_ptr)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001456{
1457 /*
1458 * If unsharing a thread from a thread group, must also
1459 * unshare vm.
1460 */
1461 if (*flags_ptr & CLONE_THREAD)
1462 *flags_ptr |= CLONE_VM;
1463
1464 /*
1465 * If unsharing vm, must also unshare signal handlers.
1466 */
1467 if (*flags_ptr & CLONE_VM)
1468 *flags_ptr |= CLONE_SIGHAND;
1469
1470 /*
1471 * If unsharing signal handlers and the task was created
1472 * using CLONE_THREAD, then must unshare the thread
1473 */
1474 if ((*flags_ptr & CLONE_SIGHAND) &&
1475 (atomic_read(&current->signal->count) > 1))
1476 *flags_ptr |= CLONE_THREAD;
1477
1478 /*
1479 * If unsharing namespace, must also unshare filesystem information.
1480 */
1481 if (*flags_ptr & CLONE_NEWNS)
1482 *flags_ptr |= CLONE_FS;
1483}
1484
1485/*
1486 * Unsharing of tasks created with CLONE_THREAD is not supported yet
1487 */
1488static int unshare_thread(unsigned long unshare_flags)
1489{
1490 if (unshare_flags & CLONE_THREAD)
1491 return -EINVAL;
1492
1493 return 0;
1494}
1495
1496/*
JANAK DESAI99d14192006-02-07 12:58:59 -08001497 * Unshare the filesystem structure if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001498 */
1499static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1500{
1501 struct fs_struct *fs = current->fs;
1502
1503 if ((unshare_flags & CLONE_FS) &&
JANAK DESAI99d14192006-02-07 12:58:59 -08001504 (fs && atomic_read(&fs->count) > 1)) {
1505 *new_fsp = __copy_fs_struct(current->fs);
1506 if (!*new_fsp)
1507 return -ENOMEM;
1508 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001509
1510 return 0;
1511}
1512
1513/*
Oleg Nesterovdae3c5a2006-12-08 02:36:09 -08001514 * Unsharing of sighand is not supported yet
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001515 */
1516static int unshare_sighand(unsigned long unshare_flags, struct sighand_struct **new_sighp)
1517{
1518 struct sighand_struct *sigh = current->sighand;
1519
Oleg Nesterovdae3c5a2006-12-08 02:36:09 -08001520 if ((unshare_flags & CLONE_SIGHAND) && atomic_read(&sigh->count) > 1)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001521 return -EINVAL;
1522 else
1523 return 0;
1524}
1525
1526/*
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001527 * Unshare vm if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001528 */
1529static int unshare_vm(unsigned long unshare_flags, struct mm_struct **new_mmp)
1530{
1531 struct mm_struct *mm = current->mm;
1532
1533 if ((unshare_flags & CLONE_VM) &&
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001534 (mm && atomic_read(&mm->mm_users) > 1)) {
Oleg Nesterov2d61b862006-03-18 20:41:10 +03001535 return -EINVAL;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001536 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001537
1538 return 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001539}
1540
1541/*
JANAK DESAIa016f332006-02-07 12:59:02 -08001542 * Unshare file descriptor table if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001543 */
1544static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1545{
1546 struct files_struct *fd = current->files;
JANAK DESAIa016f332006-02-07 12:59:02 -08001547 int error = 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001548
1549 if ((unshare_flags & CLONE_FILES) &&
JANAK DESAIa016f332006-02-07 12:59:02 -08001550 (fd && atomic_read(&fd->count) > 1)) {
1551 *new_fdp = dup_fd(fd, &error);
1552 if (!*new_fdp)
1553 return error;
1554 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001555
1556 return 0;
1557}
1558
1559/*
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001560 * unshare allows a process to 'unshare' part of the process
1561 * context which was originally shared using clone. copy_*
1562 * functions used by do_fork() cannot be used here directly
1563 * because they modify an inactive task_struct that is being
1564 * constructed. Here we are modifying the current, active,
1565 * task_struct.
1566 */
1567asmlinkage long sys_unshare(unsigned long unshare_flags)
1568{
1569 int err = 0;
1570 struct fs_struct *fs, *new_fs = NULL;
Oleg Nesterovdae3c5a2006-12-08 02:36:09 -08001571 struct sighand_struct *new_sigh = NULL;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001572 struct mm_struct *mm, *new_mm = NULL, *active_mm = NULL;
1573 struct files_struct *fd, *new_fd = NULL;
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07001574 struct nsproxy *new_nsproxy = NULL;
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001575 int do_sysvsem = 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001576
1577 check_unshare_flags(&unshare_flags);
1578
Eric W. Biederman06f9d4f2006-03-22 00:07:40 -08001579 /* Return -EINVAL for all unsupported flags */
1580 err = -EINVAL;
1581 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
Kirill Korotaev25b21cb2006-10-02 02:18:19 -07001582 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
Eric W. Biederman9dd776b2007-09-26 22:04:26 -07001583 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWUSER|
1584 CLONE_NEWNET))
Eric W. Biederman06f9d4f2006-03-22 00:07:40 -08001585 goto bad_unshare_out;
1586
Manfred Spraul6013f672008-04-29 01:00:59 -07001587 /*
1588 * CLONE_NEWIPC must also detach from the undolist: after switching
1589 * to a new ipc namespace, the semaphore arrays from the old
1590 * namespace are unreachable.
1591 */
1592 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001593 do_sysvsem = 1;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001594 if ((err = unshare_thread(unshare_flags)))
1595 goto bad_unshare_out;
1596 if ((err = unshare_fs(unshare_flags, &new_fs)))
1597 goto bad_unshare_cleanup_thread;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001598 if ((err = unshare_sighand(unshare_flags, &new_sigh)))
Badari Pulavartye3222c42007-05-08 00:25:21 -07001599 goto bad_unshare_cleanup_fs;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001600 if ((err = unshare_vm(unshare_flags, &new_mm)))
1601 goto bad_unshare_cleanup_sigh;
1602 if ((err = unshare_fd(unshare_flags, &new_fd)))
1603 goto bad_unshare_cleanup_vm;
Badari Pulavartye3222c42007-05-08 00:25:21 -07001604 if ((err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
1605 new_fs)))
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001606 goto bad_unshare_cleanup_fd;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001607
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001608 if (new_fs || new_mm || new_fd || do_sysvsem || new_nsproxy) {
1609 if (do_sysvsem) {
1610 /*
1611 * CLONE_SYSVSEM is equivalent to sys_exit().
1612 */
1613 exit_sem(current);
1614 }
Serge E. Hallynab516012006-10-02 02:18:06 -07001615
Serge Hallync0b2fc32006-10-02 02:18:18 -07001616 if (new_nsproxy) {
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07001617 switch_task_namespaces(current, new_nsproxy);
1618 new_nsproxy = NULL;
Serge Hallync0b2fc32006-10-02 02:18:18 -07001619 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001620
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07001621 task_lock(current);
1622
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001623 if (new_fs) {
1624 fs = current->fs;
1625 current->fs = new_fs;
1626 new_fs = fs;
1627 }
1628
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001629 if (new_mm) {
1630 mm = current->mm;
1631 active_mm = current->active_mm;
1632 current->mm = new_mm;
1633 current->active_mm = new_mm;
1634 activate_mm(active_mm, new_mm);
1635 new_mm = mm;
1636 }
1637
1638 if (new_fd) {
1639 fd = current->files;
1640 current->files = new_fd;
1641 new_fd = fd;
1642 }
1643
1644 task_unlock(current);
1645 }
1646
Serge Hallync0b2fc32006-10-02 02:18:18 -07001647 if (new_nsproxy)
Linus Torvalds444f3782007-01-30 13:35:18 -08001648 put_nsproxy(new_nsproxy);
Serge Hallync0b2fc32006-10-02 02:18:18 -07001649
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001650bad_unshare_cleanup_fd:
1651 if (new_fd)
1652 put_files_struct(new_fd);
1653
1654bad_unshare_cleanup_vm:
1655 if (new_mm)
1656 mmput(new_mm);
1657
1658bad_unshare_cleanup_sigh:
1659 if (new_sigh)
1660 if (atomic_dec_and_test(&new_sigh->count))
1661 kmem_cache_free(sighand_cachep, new_sigh);
1662
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001663bad_unshare_cleanup_fs:
1664 if (new_fs)
1665 put_fs_struct(new_fs);
1666
1667bad_unshare_cleanup_thread:
1668bad_unshare_out:
1669 return err;
1670}
Al Viro3b125382008-04-22 05:31:30 -04001671
1672/*
1673 * Helper to unshare the files of the current task.
1674 * We don't want to expose copy_files internals to
1675 * the exec layer of the kernel.
1676 */
1677
1678int unshare_files(struct files_struct **displaced)
1679{
1680 struct task_struct *task = current;
Al Viro50704512008-04-26 05:25:00 +01001681 struct files_struct *copy = NULL;
Al Viro3b125382008-04-22 05:31:30 -04001682 int error;
1683
1684 error = unshare_fd(CLONE_FILES, &copy);
1685 if (error || !copy) {
1686 *displaced = NULL;
1687 return error;
1688 }
1689 *displaced = task->files;
1690 task_lock(task);
1691 task->files = copy;
1692 task_unlock(task);
1693 return 0;
1694}