blob: f0e7781ba9b4f9f7046a9ca608978166de4340c9 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/kernel/fork.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 */
6
7/*
8 * 'fork.c' contains the help-routines for the 'fork' system call
9 * (see also entry.S and others).
10 * Fork is rather simple, once you get the hang of it, but the memory
11 * management can be a bitch. See 'mm/memory.c': 'copy_page_range()'
12 */
13
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/slab.h>
15#include <linux/init.h>
16#include <linux/unistd.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/module.h>
18#include <linux/vmalloc.h>
19#include <linux/completion.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/personality.h>
21#include <linux/mempolicy.h>
22#include <linux/sem.h>
23#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040024#include <linux/fdtable.h>
Jens Axboeda9cbc82008-06-30 20:42:08 +020025#include <linux/iocontext.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/key.h>
27#include <linux/binfmts.h>
28#include <linux/mman.h>
Andrea Arcangelicddb8a52008-07-28 15:46:29 -070029#include <linux/mmu_notifier.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include <linux/fs.h>
Serge E. Hallynab516012006-10-02 02:18:06 -070031#include <linux/nsproxy.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080032#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <linux/cpu.h>
Paul Menageb4f48b62007-10-18 23:39:33 -070034#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035#include <linux/security.h>
Mel Gormana1e78772008-07-23 21:27:23 -070036#include <linux/hugetlb.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#include <linux/swap.h>
38#include <linux/syscalls.h>
39#include <linux/jiffies.h>
40#include <linux/futex.h>
Linus Torvalds8141c7f2008-11-15 10:20:36 -080041#include <linux/compat.h>
Eric Dumazet207205a2011-03-22 16:30:44 -070042#include <linux/kthread.h>
Andrew Morton7c3ab732006-12-10 02:19:19 -080043#include <linux/task_io_accounting_ops.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070044#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070045#include <linux/ptrace.h>
46#include <linux/mount.h>
47#include <linux/audit.h>
Pavel Emelianov78fb7462008-02-07 00:13:51 -080048#include <linux/memcontrol.h>
Frederic Weisbeckerf201ae22008-11-23 06:22:56 +010049#include <linux/ftrace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070050#include <linux/profile.h>
51#include <linux/rmap.h>
Hugh Dickinsf8af4da2009-09-21 17:01:57 -070052#include <linux/ksm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070053#include <linux/acct.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070054#include <linux/tsacct_kern.h>
Matt Helsley9f460802005-11-07 00:59:16 -080055#include <linux/cn_proc.h>
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -070056#include <linux/freezer.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070057#include <linux/delayacct.h>
Shailabh Nagarad4ecbc2006-07-14 00:24:44 -070058#include <linux/taskstats_kern.h>
Arjan van de Ven0a425402006-09-26 10:52:38 +020059#include <linux/random.h>
Miloslav Trmac522ed772007-07-15 23:40:56 -070060#include <linux/tty.h>
Jens Axboefd0928d2008-01-24 08:52:45 +010061#include <linux/blkdev.h>
Al Viro5ad4e532009-03-29 19:50:06 -040062#include <linux/fs_struct.h>
Eric Sandeen7c9f8862008-04-22 16:38:23 -050063#include <linux/magic.h>
Ingo Molnarcdd6c482009-09-21 12:02:48 +020064#include <linux/perf_event.h>
Stanislaw Gruszka42c4ab42009-07-29 12:15:26 +020065#include <linux/posix-timers.h>
Avi Kivity8e7cac72009-11-29 16:34:48 +020066#include <linux/user-return-notifier.h>
Ying Han3d5992d2010-10-26 14:21:23 -070067#include <linux/oom.h>
Andrea Arcangeliba761492011-01-13 15:46:58 -080068#include <linux/khugepaged.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070069
70#include <asm/pgtable.h>
71#include <asm/pgalloc.h>
72#include <asm/uaccess.h>
73#include <asm/mmu_context.h>
74#include <asm/cacheflush.h>
75#include <asm/tlbflush.h>
76
Steven Rostedtad8d75f2009-04-14 19:39:12 -040077#include <trace/events/sched.h>
78
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -080079#define CREATE_TRACE_POINTS
80#include <trace/events/task.h>
81
Linus Torvalds1da177e2005-04-16 15:20:36 -070082/*
83 * Protected counters by write_lock_irq(&tasklist_lock)
84 */
85unsigned long total_forks; /* Handle normal Linux uptimes. */
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -070086int nr_threads; /* The idle threads do not count.. */
Linus Torvalds1da177e2005-04-16 15:20:36 -070087
88int max_threads; /* tunable limit on nr_threads */
89
90DEFINE_PER_CPU(unsigned long, process_counts) = 0;
91
Christoph Hellwigc59923a2006-07-10 04:45:40 -070092__cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
Paul E. McKenneydb1466b2010-03-03 07:46:56 -080093
94#ifdef CONFIG_PROVE_RCU
95int lockdep_tasklist_lock_is_held(void)
96{
97 return lockdep_is_held(&tasklist_lock);
98}
99EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
100#endif /* #ifdef CONFIG_PROVE_RCU */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101
102int nr_processes(void)
103{
104 int cpu;
105 int total = 0;
106
Ian Campbell1d510752009-11-03 10:11:14 +0000107 for_each_possible_cpu(cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108 total += per_cpu(process_counts, cpu);
109
110 return total;
111}
112
113#ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
Eric Dumazet504f52b2011-03-22 16:30:41 -0700114# define alloc_task_struct_node(node) \
115 kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node)
116# define free_task_struct(tsk) \
117 kmem_cache_free(task_struct_cachep, (tsk))
Christoph Lametere18b8902006-12-06 20:33:20 -0800118static struct kmem_cache *task_struct_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119#endif
120
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700121#ifndef __HAVE_ARCH_THREAD_INFO_ALLOCATOR
Eric Dumazetb6a84012011-03-22 16:30:42 -0700122static struct thread_info *alloc_thread_info_node(struct task_struct *tsk,
123 int node)
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700124{
125#ifdef CONFIG_DEBUG_STACK_USAGE
126 gfp_t mask = GFP_KERNEL | __GFP_ZERO;
127#else
128 gfp_t mask = GFP_KERNEL;
129#endif
Eric Dumazetb6a84012011-03-22 16:30:42 -0700130 struct page *page = alloc_pages_node(node, mask, THREAD_SIZE_ORDER);
131
132 return page ? page_address(page) : NULL;
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700133}
134
135static inline void free_thread_info(struct thread_info *ti)
136{
137 free_pages((unsigned long)ti, THREAD_SIZE_ORDER);
138}
139#endif
140
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141/* SLAB cache for signal_struct structures (tsk->signal) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800142static struct kmem_cache *signal_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143
144/* SLAB cache for sighand_struct structures (tsk->sighand) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800145struct kmem_cache *sighand_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146
147/* SLAB cache for files_struct structures (tsk->files) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800148struct kmem_cache *files_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149
150/* SLAB cache for fs_struct structures (tsk->fs) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800151struct kmem_cache *fs_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152
153/* SLAB cache for vm_area_struct structures */
Christoph Lametere18b8902006-12-06 20:33:20 -0800154struct kmem_cache *vm_area_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155
156/* SLAB cache for mm_struct structures (tsk->mm) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800157static struct kmem_cache *mm_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700159static void account_kernel_stack(struct thread_info *ti, int account)
160{
161 struct zone *zone = page_zone(virt_to_page(ti));
162
163 mod_zone_page_state(zone, NR_KERNEL_STACK, account);
164}
165
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166void free_task(struct task_struct *tsk)
167{
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700168 account_kernel_stack(tsk->stack, -1);
Roman Zippelf7e42172007-05-09 02:35:17 -0700169 free_thread_info(tsk->stack);
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700170 rt_mutex_debug_task_free(tsk);
Frederic Weisbeckerfb526072008-11-25 21:07:04 +0100171 ftrace_graph_exit_task(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 free_task_struct(tsk);
173}
174EXPORT_SYMBOL(free_task);
175
Oleg Nesterovea6d2902010-05-26 14:43:16 -0700176static inline void free_signal_struct(struct signal_struct *sig)
177{
Oleg Nesterov97101eb2010-05-26 14:43:20 -0700178 taskstats_tgid_free(sig);
Mike Galbraith1c5354d2011-01-05 11:16:04 +0100179 sched_autogroup_exit(sig);
Oleg Nesterovea6d2902010-05-26 14:43:16 -0700180 kmem_cache_free(signal_cachep, sig);
181}
182
183static inline void put_signal_struct(struct signal_struct *sig)
184{
Mike Galbraith1c5354d2011-01-05 11:16:04 +0100185 if (atomic_dec_and_test(&sig->sigcnt))
Oleg Nesterovea6d2902010-05-26 14:43:16 -0700186 free_signal_struct(sig);
187}
188
Andrew Morton158d9eb2006-03-31 02:31:34 -0800189void __put_task_struct(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190{
Eugene Teo270f7222007-10-18 23:40:38 -0700191 WARN_ON(!tsk->exit_state);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 WARN_ON(atomic_read(&tsk->usage));
193 WARN_ON(tsk == current);
194
Kees Cook1a2a4d02011-12-21 12:17:03 -0800195 security_task_free(tsk);
David Howellse0e81732009-09-02 09:13:40 +0100196 exit_creds(tsk);
Shailabh Nagar35df17c2006-08-31 21:27:38 -0700197 delayacct_tsk_free(tsk);
Oleg Nesterovea6d2902010-05-26 14:43:16 -0700198 put_signal_struct(tsk->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199
200 if (!profile_handoff_task(tsk))
201 free_task(tsk);
202}
Rik van Riel77c100c2011-02-01 09:51:46 -0500203EXPORT_SYMBOL_GPL(__put_task_struct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204
Suresh Siddha2adee9b2008-04-16 10:25:35 +0200205/*
206 * macro override instead of weak attribute alias, to workaround
207 * gcc 4.1.0 and 4.1.1 bugs with weak attribute and empty functions.
208 */
209#ifndef arch_task_cache_init
210#define arch_task_cache_init()
211#endif
Suresh Siddha61c46282008-03-10 15:28:04 -0700212
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213void __init fork_init(unsigned long mempages)
214{
215#ifndef __HAVE_ARCH_TASK_STRUCT_ALLOCATOR
216#ifndef ARCH_MIN_TASKALIGN
217#define ARCH_MIN_TASKALIGN L1_CACHE_BYTES
218#endif
219 /* create a slab on which task_structs can be allocated */
220 task_struct_cachep =
221 kmem_cache_create("task_struct", sizeof(struct task_struct),
Vegard Nossum2dff4402008-05-31 15:56:17 +0200222 ARCH_MIN_TASKALIGN, SLAB_PANIC | SLAB_NOTRACK, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223#endif
224
Suresh Siddha61c46282008-03-10 15:28:04 -0700225 /* do the arch specific task caches init */
226 arch_task_cache_init();
227
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 /*
229 * The default maximum number of threads is set to a safe
230 * value: the thread structures can take up at most half
231 * of memory.
232 */
233 max_threads = mempages / (8 * THREAD_SIZE / PAGE_SIZE);
234
235 /*
236 * we need to allow at least 20 threads to boot a system
237 */
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700238 if (max_threads < 20)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 max_threads = 20;
240
241 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
242 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
243 init_task.signal->rlim[RLIMIT_SIGPENDING] =
244 init_task.signal->rlim[RLIMIT_NPROC];
245}
246
Suresh Siddha61c46282008-03-10 15:28:04 -0700247int __attribute__((weak)) arch_dup_task_struct(struct task_struct *dst,
248 struct task_struct *src)
249{
250 *dst = *src;
251 return 0;
252}
253
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254static struct task_struct *dup_task_struct(struct task_struct *orig)
255{
256 struct task_struct *tsk;
257 struct thread_info *ti;
Eric Sandeen7c9f8862008-04-22 16:38:23 -0500258 unsigned long *stackend;
Eric Dumazet207205a2011-03-22 16:30:44 -0700259 int node = tsk_fork_get_node(orig);
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700260 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261
262 prepare_to_copy(orig);
263
Eric Dumazet504f52b2011-03-22 16:30:41 -0700264 tsk = alloc_task_struct_node(node);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 if (!tsk)
266 return NULL;
267
Eric Dumazetb6a84012011-03-22 16:30:42 -0700268 ti = alloc_thread_info_node(tsk, node);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 if (!ti) {
270 free_task_struct(tsk);
271 return NULL;
272 }
273
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700274 err = arch_dup_task_struct(tsk, orig);
Suresh Siddha61c46282008-03-10 15:28:04 -0700275 if (err)
276 goto out;
277
Roman Zippelf7e42172007-05-09 02:35:17 -0700278 tsk->stack = ti;
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700279
Al Viro10ebffd2005-11-13 16:06:56 -0800280 setup_thread_stack(tsk, orig);
Avi Kivity8e7cac72009-11-29 16:34:48 +0200281 clear_user_return_notifier(tsk);
Mike Galbraithf26f9af2010-12-08 11:05:42 +0100282 clear_tsk_need_resched(tsk);
Eric Sandeen7c9f8862008-04-22 16:38:23 -0500283 stackend = end_of_stack(tsk);
284 *stackend = STACK_END_MAGIC; /* for overflow detection */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285
Arjan van de Ven0a425402006-09-26 10:52:38 +0200286#ifdef CONFIG_CC_STACKPROTECTOR
287 tsk->stack_canary = get_random_int();
288#endif
289
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700290 /*
291 * One for us, one for whoever does the "release_task()" (usually
292 * parent)
293 */
294 atomic_set(&tsk->usage, 2);
Alexey Dobriyan6c5c9342006-09-29 01:59:40 -0700295#ifdef CONFIG_BLK_DEV_IO_TRACE
Jens Axboe2056a782006-03-23 20:00:26 +0100296 tsk->btrace_seq = 0;
Alexey Dobriyan6c5c9342006-09-29 01:59:40 -0700297#endif
Jens Axboea0aa7f62006-04-20 13:05:33 +0200298 tsk->splice_pipe = NULL;
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700299
300 account_kernel_stack(ti, 1);
301
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 return tsk;
Suresh Siddha61c46282008-03-10 15:28:04 -0700303
304out:
305 free_thread_info(ti);
306 free_task_struct(tsk);
307 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308}
309
310#ifdef CONFIG_MMU
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700311static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312{
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700313 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314 struct rb_node **rb_link, *rb_parent;
315 int retval;
316 unsigned long charge;
317 struct mempolicy *pol;
318
319 down_write(&oldmm->mmap_sem);
Ralf Baechleec8c0442006-12-12 17:14:57 +0000320 flush_cache_dup_mm(oldmm);
Ingo Molnarad339452006-07-03 00:25:15 -0700321 /*
322 * Not linked in yet - no deadlock potential:
323 */
324 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
Hugh Dickins7ee78232005-10-29 18:16:08 -0700325
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326 mm->locked_vm = 0;
327 mm->mmap = NULL;
328 mm->mmap_cache = NULL;
329 mm->free_area_cache = oldmm->mmap_base;
Wolfgang Wander1363c3c2005-06-21 17:14:49 -0700330 mm->cached_hole_size = ~0UL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 mm->map_count = 0;
Rusty Russell94894242009-03-30 22:05:12 -0600332 cpumask_clear(mm_cpumask(mm));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 mm->mm_rb = RB_ROOT;
334 rb_link = &mm->mm_rb.rb_node;
335 rb_parent = NULL;
336 pprev = &mm->mmap;
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700337 retval = ksm_fork(mm, oldmm);
338 if (retval)
339 goto out;
Andrea Arcangeliba761492011-01-13 15:46:58 -0800340 retval = khugepaged_fork(mm, oldmm);
341 if (retval)
342 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700344 prev = NULL;
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700345 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 struct file *file;
347
348 if (mpnt->vm_flags & VM_DONTCOPY) {
Hugh Dickins3b6bfcd2005-07-12 13:58:09 -0700349 long pages = vma_pages(mpnt);
350 mm->total_vm -= pages;
Hugh Dickinsab50b8e2005-10-29 18:15:56 -0700351 vm_stat_account(mm, mpnt->vm_flags, mpnt->vm_file,
Hugh Dickins3b6bfcd2005-07-12 13:58:09 -0700352 -pages);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 continue;
354 }
355 charge = 0;
356 if (mpnt->vm_flags & VM_ACCOUNT) {
357 unsigned int len = (mpnt->vm_end - mpnt->vm_start) >> PAGE_SHIFT;
358 if (security_vm_enough_memory(len))
359 goto fail_nomem;
360 charge = len;
361 }
Christoph Lametere94b1762006-12-06 20:33:17 -0800362 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 if (!tmp)
364 goto fail_nomem;
365 *tmp = *mpnt;
Rik van Riel5beb4932010-03-05 13:42:07 -0800366 INIT_LIST_HEAD(&tmp->anon_vma_chain);
Lee Schermerhorn846a16b2008-04-28 02:13:09 -0700367 pol = mpol_dup(vma_policy(mpnt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 retval = PTR_ERR(pol);
369 if (IS_ERR(pol))
370 goto fail_nomem_policy;
371 vma_set_policy(tmp, pol);
Andrea Arcangelia247c3a2010-09-22 13:05:12 -0700372 tmp->vm_mm = mm;
Rik van Riel5beb4932010-03-05 13:42:07 -0800373 if (anon_vma_fork(tmp, mpnt))
374 goto fail_nomem_anon_vma_fork;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375 tmp->vm_flags &= ~VM_LOCKED;
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700376 tmp->vm_next = tmp->vm_prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 file = tmp->vm_file;
378 if (file) {
Josef "Jeff" Sipekf3a43f32006-12-08 02:36:43 -0800379 struct inode *inode = file->f_path.dentry->d_inode;
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000380 struct address_space *mapping = file->f_mapping;
381
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 get_file(file);
383 if (tmp->vm_flags & VM_DENYWRITE)
384 atomic_dec(&inode->i_writecount);
Peter Zijlstra3d48ae42011-05-24 17:12:06 -0700385 mutex_lock(&mapping->i_mmap_mutex);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000386 if (tmp->vm_flags & VM_SHARED)
387 mapping->i_mmap_writable++;
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000388 flush_dcache_mmap_lock(mapping);
389 /* insert tmp into the share list, just after mpnt */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 vma_prio_tree_add(tmp, mpnt);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000391 flush_dcache_mmap_unlock(mapping);
Peter Zijlstra3d48ae42011-05-24 17:12:06 -0700392 mutex_unlock(&mapping->i_mmap_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 }
394
395 /*
Mel Gormana1e78772008-07-23 21:27:23 -0700396 * Clear hugetlb-related page reserves for children. This only
397 * affects MAP_PRIVATE mappings. Faults generated by the child
398 * are not guaranteed to succeed, even if read-only
399 */
400 if (is_vm_hugetlb_page(tmp))
401 reset_vma_resv_huge_pages(tmp);
402
403 /*
Hugh Dickins7ee78232005-10-29 18:16:08 -0700404 * Link in the new vma and copy the page table entries.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 *pprev = tmp;
407 pprev = &tmp->vm_next;
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700408 tmp->vm_prev = prev;
409 prev = tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410
411 __vma_link_rb(mm, tmp, rb_link, rb_parent);
412 rb_link = &tmp->vm_rb.rb_right;
413 rb_parent = &tmp->vm_rb;
414
415 mm->map_count++;
Hugh Dickins0b0db142005-11-21 21:32:20 -0800416 retval = copy_page_range(mm, oldmm, mpnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417
418 if (tmp->vm_ops && tmp->vm_ops->open)
419 tmp->vm_ops->open(tmp);
420
421 if (retval)
422 goto out;
423 }
Jeremy Fitzhardinged6dd61c2007-05-02 19:27:14 +0200424 /* a new mm has just been created */
425 arch_dup_mmap(oldmm, mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427out:
Hugh Dickins7ee78232005-10-29 18:16:08 -0700428 up_write(&mm->mmap_sem);
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700429 flush_tlb_mm(oldmm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 up_write(&oldmm->mmap_sem);
431 return retval;
Rik van Riel5beb4932010-03-05 13:42:07 -0800432fail_nomem_anon_vma_fork:
433 mpol_put(pol);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434fail_nomem_policy:
435 kmem_cache_free(vm_area_cachep, tmp);
436fail_nomem:
437 retval = -ENOMEM;
438 vm_unacct_memory(charge);
439 goto out;
440}
441
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700442static inline int mm_alloc_pgd(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443{
444 mm->pgd = pgd_alloc(mm);
445 if (unlikely(!mm->pgd))
446 return -ENOMEM;
447 return 0;
448}
449
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700450static inline void mm_free_pgd(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451{
Benjamin Herrenschmidt5e541972008-02-04 22:29:14 -0800452 pgd_free(mm, mm->pgd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453}
454#else
455#define dup_mmap(mm, oldmm) (0)
456#define mm_alloc_pgd(mm) (0)
457#define mm_free_pgd(mm)
458#endif /* CONFIG_MMU */
459
Daniel Walker23ff4442007-10-18 03:06:07 -0700460__cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461
Christoph Lametere94b1762006-12-06 20:33:17 -0800462#define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463#define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
464
Hidehiro Kawai4cb0e112009-01-06 14:42:47 -0800465static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
466
467static int __init coredump_filter_setup(char *s)
468{
469 default_dump_filter =
470 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
471 MMF_DUMP_FILTER_MASK;
472 return 1;
473}
474
475__setup("coredump_filter=", coredump_filter_setup);
476
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477#include <linux/init_task.h>
478
Alexey Dobriyan858f0992009-09-23 15:57:32 -0700479static void mm_init_aio(struct mm_struct *mm)
480{
481#ifdef CONFIG_AIO
482 spin_lock_init(&mm->ioctx_lock);
483 INIT_HLIST_HEAD(&mm->ioctx_list);
484#endif
485}
486
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700487static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488{
489 atomic_set(&mm->mm_users, 1);
490 atomic_set(&mm->mm_count, 1);
491 init_rwsem(&mm->mmap_sem);
492 INIT_LIST_HEAD(&mm->mmlist);
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700493 mm->flags = (current->mm) ?
494 (current->mm->flags & MMF_INIT_MASK) : default_dump_filter;
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700495 mm->core_state = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 mm->nr_ptes = 0;
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -0800497 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 spin_lock_init(&mm->page_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 mm->free_area_cache = TASK_UNMAPPED_BASE;
Wolfgang Wander1363c3c2005-06-21 17:14:49 -0700500 mm->cached_hole_size = ~0UL;
Alexey Dobriyan858f0992009-09-23 15:57:32 -0700501 mm_init_aio(mm);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700502 mm_init_owner(mm, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503
504 if (likely(!mm_alloc_pgd(mm))) {
505 mm->def_flags = 0;
Andrea Arcangelicddb8a52008-07-28 15:46:29 -0700506 mmu_notifier_mm_init(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 return mm;
508 }
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800509
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 free_mm(mm);
511 return NULL;
512}
513
514/*
515 * Allocate and initialize an mm_struct.
516 */
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700517struct mm_struct *mm_alloc(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518{
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700519 struct mm_struct *mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520
521 mm = allocate_mm();
KOSAKI Motohirode03c722011-05-24 17:12:15 -0700522 if (!mm)
523 return NULL;
524
525 memset(mm, 0, sizeof(*mm));
Linus Torvalds6345d242011-05-29 11:32:28 -0700526 mm_init_cpumask(mm);
527 return mm_init(mm, current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528}
529
530/*
531 * Called when the last reference to the mm
532 * is dropped: either by a lazy thread or by
533 * mmput. Free the page directory and the mm.
534 */
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800535void __mmdrop(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536{
537 BUG_ON(mm == &init_mm);
538 mm_free_pgd(mm);
539 destroy_context(mm);
Andrea Arcangelicddb8a52008-07-28 15:46:29 -0700540 mmu_notifier_mm_destroy(mm);
Andrea Arcangelie7a00c42011-01-13 15:46:45 -0800541#ifdef CONFIG_TRANSPARENT_HUGEPAGE
542 VM_BUG_ON(mm->pmd_huge_pte);
543#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 free_mm(mm);
545}
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200546EXPORT_SYMBOL_GPL(__mmdrop);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547
548/*
549 * Decrement the use count and release all resources for an mm.
550 */
551void mmput(struct mm_struct *mm)
552{
Andrew Morton0ae26f12006-06-23 02:05:15 -0700553 might_sleep();
554
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 if (atomic_dec_and_test(&mm->mm_users)) {
556 exit_aio(mm);
Andrea Arcangeli1c2fb7a2009-09-21 17:02:22 -0700557 ksm_exit(mm);
Andrea Arcangeliba761492011-01-13 15:46:58 -0800558 khugepaged_exit(mm); /* must run before exit_mmap */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 exit_mmap(mm);
Matt Helsley925d1c42008-04-29 01:01:36 -0700560 set_mm_exe_file(mm, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 if (!list_empty(&mm->mmlist)) {
562 spin_lock(&mmlist_lock);
563 list_del(&mm->mmlist);
564 spin_unlock(&mmlist_lock);
565 }
566 put_swap_token(mm);
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -0700567 if (mm->binfmt)
568 module_put(mm->binfmt->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 mmdrop(mm);
570 }
571}
572EXPORT_SYMBOL_GPL(mmput);
573
Jiri Slaby38646012011-05-26 16:25:46 -0700574/*
575 * We added or removed a vma mapping the executable. The vmas are only mapped
576 * during exec and are not mapped with the mmap system call.
577 * Callers must hold down_write() on the mm's mmap_sem for these
578 */
579void added_exe_file_vma(struct mm_struct *mm)
580{
581 mm->num_exe_file_vmas++;
582}
583
584void removed_exe_file_vma(struct mm_struct *mm)
585{
586 mm->num_exe_file_vmas--;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700587 if ((mm->num_exe_file_vmas == 0) && mm->exe_file) {
Jiri Slaby38646012011-05-26 16:25:46 -0700588 fput(mm->exe_file);
589 mm->exe_file = NULL;
590 }
591
592}
593
594void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
595{
596 if (new_exe_file)
597 get_file(new_exe_file);
598 if (mm->exe_file)
599 fput(mm->exe_file);
600 mm->exe_file = new_exe_file;
601 mm->num_exe_file_vmas = 0;
602}
603
604struct file *get_mm_exe_file(struct mm_struct *mm)
605{
606 struct file *exe_file;
607
608 /* We need mmap_sem to protect against races with removal of
609 * VM_EXECUTABLE vmas */
610 down_read(&mm->mmap_sem);
611 exe_file = mm->exe_file;
612 if (exe_file)
613 get_file(exe_file);
614 up_read(&mm->mmap_sem);
615 return exe_file;
616}
617
618static void dup_mm_exe_file(struct mm_struct *oldmm, struct mm_struct *newmm)
619{
620 /* It's safe to write the exe_file pointer without exe_file_lock because
621 * this is called during fork when the task is not yet in /proc */
622 newmm->exe_file = get_mm_exe_file(oldmm);
623}
624
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625/**
626 * get_task_mm - acquire a reference to the task's mm
627 *
Oleg Nesterov246bb0b2008-07-25 01:47:38 -0700628 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 * this kernel workthread has transiently adopted a user mm with use_mm,
630 * to do its AIO) is not set and if so returns a reference to it, after
631 * bumping up the use count. User must release the mm via mmput()
632 * after use. Typically used by /proc and ptrace.
633 */
634struct mm_struct *get_task_mm(struct task_struct *task)
635{
636 struct mm_struct *mm;
637
638 task_lock(task);
639 mm = task->mm;
640 if (mm) {
Oleg Nesterov246bb0b2008-07-25 01:47:38 -0700641 if (task->flags & PF_KTHREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 mm = NULL;
643 else
644 atomic_inc(&mm->mm_users);
645 }
646 task_unlock(task);
647 return mm;
648}
649EXPORT_SYMBOL_GPL(get_task_mm);
650
Christopher Yeoh8cdb8782012-02-02 11:34:09 +1030651struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
652{
653 struct mm_struct *mm;
654 int err;
655
656 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
657 if (err)
658 return ERR_PTR(err);
659
660 mm = get_task_mm(task);
661 if (mm && mm != current->mm &&
662 !ptrace_may_access(task, mode)) {
663 mmput(mm);
664 mm = ERR_PTR(-EACCES);
665 }
666 mutex_unlock(&task->signal->cred_guard_mutex);
667
668 return mm;
669}
670
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671/* Please note the differences between mmput and mm_release.
672 * mmput is called whenever we stop holding onto a mm_struct,
673 * error success whatever.
674 *
675 * mm_release is called after a mm_struct has been removed
676 * from the current process.
677 *
678 * This difference is important for error handling, when we
679 * only half set up a mm_struct for a new process and need to restore
680 * the old one. Because we mmput the new mm_struct before
681 * restoring the old one. . .
682 * Eric Biederman 10 January 1998
683 */
684void mm_release(struct task_struct *tsk, struct mm_struct *mm)
685{
686 struct completion *vfork_done = tsk->vfork_done;
687
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800688 /* Get rid of any futexes when releasing the mm */
689#ifdef CONFIG_FUTEX
Peter Zijlstrafc6b1772009-10-05 18:17:32 +0200690 if (unlikely(tsk->robust_list)) {
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800691 exit_robust_list(tsk);
Peter Zijlstrafc6b1772009-10-05 18:17:32 +0200692 tsk->robust_list = NULL;
693 }
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800694#ifdef CONFIG_COMPAT
Peter Zijlstrafc6b1772009-10-05 18:17:32 +0200695 if (unlikely(tsk->compat_robust_list)) {
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800696 compat_exit_robust_list(tsk);
Peter Zijlstrafc6b1772009-10-05 18:17:32 +0200697 tsk->compat_robust_list = NULL;
698 }
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800699#endif
Thomas Gleixner322a2c12009-10-05 18:18:03 +0200700 if (unlikely(!list_empty(&tsk->pi_state_list)))
701 exit_pi_state_list(tsk);
Linus Torvalds8141c7f2008-11-15 10:20:36 -0800702#endif
703
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 /* Get rid of any cached register state */
705 deactivate_mm(tsk, mm);
706
707 /* notify parent sleeping on vfork() */
708 if (vfork_done) {
709 tsk->vfork_done = NULL;
710 complete(vfork_done);
711 }
Roland McGrathfec1d012006-12-06 20:36:34 -0800712
713 /*
714 * If we're exiting normally, clear a user-space tid field if
715 * requested. We leave this alone when dying by signal, to leave
716 * the value intact in a core dump, and to save the unnecessary
717 * trouble otherwise. Userland only wants this done for a sys_exit.
718 */
Eric Dumazet9c8a8222009-08-06 15:09:28 -0700719 if (tsk->clear_child_tid) {
720 if (!(tsk->flags & PF_SIGNALED) &&
721 atomic_read(&mm->mm_users) > 1) {
722 /*
723 * We don't check the error code - if userspace has
724 * not set up a proper pointer then tough luck.
725 */
726 put_user(0, tsk->clear_child_tid);
727 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
728 1, NULL, NULL, 0);
729 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 tsk->clear_child_tid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 }
732}
733
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800734/*
735 * Allocate a new mm structure and copy contents from the
736 * mm structure of the passed in task structure.
737 */
Carsten Otte402b0862008-03-25 18:47:10 +0100738struct mm_struct *dup_mm(struct task_struct *tsk)
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800739{
740 struct mm_struct *mm, *oldmm = current->mm;
741 int err;
742
743 if (!oldmm)
744 return NULL;
745
746 mm = allocate_mm();
747 if (!mm)
748 goto fail_nomem;
749
750 memcpy(mm, oldmm, sizeof(*mm));
Linus Torvalds6345d242011-05-29 11:32:28 -0700751 mm_init_cpumask(mm);
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800752
Ashwin Chaugule7602bdf2006-12-06 20:31:57 -0800753 /* Initializing for Swap token stuff */
754 mm->token_priority = 0;
755 mm->last_interval = 0;
756
Andrea Arcangelie7a00c42011-01-13 15:46:45 -0800757#ifdef CONFIG_TRANSPARENT_HUGEPAGE
758 mm->pmd_huge_pte = NULL;
759#endif
760
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800761 if (!mm_init(mm, tsk))
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800762 goto fail_nomem;
763
764 if (init_new_context(tsk, mm))
765 goto fail_nocontext;
766
Matt Helsley925d1c42008-04-29 01:01:36 -0700767 dup_mm_exe_file(oldmm, mm);
768
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800769 err = dup_mmap(mm, oldmm);
770 if (err)
771 goto free_pt;
772
773 mm->hiwater_rss = get_mm_rss(mm);
774 mm->hiwater_vm = mm->total_vm;
775
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -0700776 if (mm->binfmt && !try_module_get(mm->binfmt->module))
777 goto free_pt;
778
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800779 return mm;
780
781free_pt:
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -0700782 /* don't put binfmt in mmput, we haven't got module yet */
783 mm->binfmt = NULL;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800784 mmput(mm);
785
786fail_nomem:
787 return NULL;
788
789fail_nocontext:
790 /*
791 * If init_new_context() failed, we cannot use mmput() to free the mm
792 * because it calls destroy_context()
793 */
794 mm_free_pgd(mm);
795 free_mm(mm);
796 return NULL;
797}
798
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700799static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800{
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700801 struct mm_struct *mm, *oldmm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 int retval;
803
804 tsk->min_flt = tsk->maj_flt = 0;
805 tsk->nvcsw = tsk->nivcsw = 0;
Mandeep Singh Baines17406b82009-02-06 15:37:47 -0800806#ifdef CONFIG_DETECT_HUNG_TASK
807 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
808#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809
810 tsk->mm = NULL;
811 tsk->active_mm = NULL;
812
813 /*
814 * Are we cloning a kernel thread?
815 *
816 * We need to steal a active VM for that..
817 */
818 oldmm = current->mm;
819 if (!oldmm)
820 return 0;
821
822 if (clone_flags & CLONE_VM) {
823 atomic_inc(&oldmm->mm_users);
824 mm = oldmm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 goto good_mm;
826 }
827
828 retval = -ENOMEM;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -0800829 mm = dup_mm(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 if (!mm)
831 goto fail_nomem;
832
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833good_mm:
Ashwin Chaugule7602bdf2006-12-06 20:31:57 -0800834 /* Initializing for Swap token stuff */
835 mm->token_priority = 0;
836 mm->last_interval = 0;
837
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 tsk->mm = mm;
839 tsk->active_mm = mm;
840 return 0;
841
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842fail_nomem:
843 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844}
845
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700846static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847{
Al Viro498052b2009-03-30 07:20:30 -0400848 struct fs_struct *fs = current->fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 if (clone_flags & CLONE_FS) {
Al Viro498052b2009-03-30 07:20:30 -0400850 /* tsk->fs is already what we want */
Nick Piggin2a4419b2010-08-18 04:37:33 +1000851 spin_lock(&fs->lock);
Al Viro498052b2009-03-30 07:20:30 -0400852 if (fs->in_exec) {
Nick Piggin2a4419b2010-08-18 04:37:33 +1000853 spin_unlock(&fs->lock);
Al Viro498052b2009-03-30 07:20:30 -0400854 return -EAGAIN;
855 }
856 fs->users++;
Nick Piggin2a4419b2010-08-18 04:37:33 +1000857 spin_unlock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 return 0;
859 }
Al Viro498052b2009-03-30 07:20:30 -0400860 tsk->fs = copy_fs_struct(fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 if (!tsk->fs)
862 return -ENOMEM;
863 return 0;
864}
865
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700866static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
JANAK DESAIa016f332006-02-07 12:59:02 -0800867{
868 struct files_struct *oldf, *newf;
869 int error = 0;
870
871 /*
872 * A background process may not have any files ...
873 */
874 oldf = current->files;
875 if (!oldf)
876 goto out;
877
878 if (clone_flags & CLONE_FILES) {
879 atomic_inc(&oldf->count);
880 goto out;
881 }
882
JANAK DESAIa016f332006-02-07 12:59:02 -0800883 newf = dup_fd(oldf, &error);
884 if (!newf)
885 goto out;
886
887 tsk->files = newf;
888 error = 0;
889out:
890 return error;
891}
892
Jens Axboefadad8782008-01-24 08:54:47 +0100893static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
Jens Axboefd0928d2008-01-24 08:52:45 +0100894{
895#ifdef CONFIG_BLOCK
896 struct io_context *ioc = current->io_context;
Tejun Heo6e736be2011-12-14 00:33:38 +0100897 struct io_context *new_ioc;
Jens Axboefd0928d2008-01-24 08:52:45 +0100898
899 if (!ioc)
900 return 0;
Jens Axboefadad8782008-01-24 08:54:47 +0100901 /*
902 * Share io context with parent, if CLONE_IO is set
903 */
904 if (clone_flags & CLONE_IO) {
905 tsk->io_context = ioc_task_link(ioc);
906 if (unlikely(!tsk->io_context))
907 return -ENOMEM;
908 } else if (ioprio_valid(ioc->ioprio)) {
Tejun Heo6e736be2011-12-14 00:33:38 +0100909 new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
910 if (unlikely(!new_ioc))
Jens Axboefd0928d2008-01-24 08:52:45 +0100911 return -ENOMEM;
912
Tejun Heo6e736be2011-12-14 00:33:38 +0100913 new_ioc->ioprio = ioc->ioprio;
Tejun Heob2efa052011-12-14 00:33:39 +0100914 put_io_context(new_ioc, NULL);
Jens Axboefd0928d2008-01-24 08:52:45 +0100915 }
916#endif
917 return 0;
918}
919
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700920static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921{
922 struct sighand_struct *sig;
923
Zhaolei60348802009-01-06 14:40:46 -0800924 if (clone_flags & CLONE_SIGHAND) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 atomic_inc(&current->sighand->count);
926 return 0;
927 }
928 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
Ingo Molnare56d0902006-01-08 01:01:37 -0800929 rcu_assign_pointer(tsk->sighand, sig);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 if (!sig)
931 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 atomic_set(&sig->count, 1);
933 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
934 return 0;
935}
936
Oleg Nesterova7e53282006-03-28 16:11:27 -0800937void __cleanup_sighand(struct sighand_struct *sighand)
Oleg Nesterovc81addc2006-03-28 16:11:17 -0800938{
Oleg Nesterovc81addc2006-03-28 16:11:17 -0800939 if (atomic_dec_and_test(&sighand->count))
940 kmem_cache_free(sighand_cachep, sighand);
941}
942
Frank Mayharf06febc2008-09-12 09:54:39 -0700943
944/*
945 * Initialize POSIX timer handling for a thread group.
946 */
947static void posix_cpu_timers_init_group(struct signal_struct *sig)
948{
Jiri Slaby78d7d402010-03-05 13:42:54 -0800949 unsigned long cpu_limit;
950
Frank Mayharf06febc2008-09-12 09:54:39 -0700951 /* Thread group counters. */
952 thread_group_cputime_init(sig);
953
Jiri Slaby78d7d402010-03-05 13:42:54 -0800954 cpu_limit = ACCESS_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
955 if (cpu_limit != RLIM_INFINITY) {
956 sig->cputime_expires.prof_exp = secs_to_cputime(cpu_limit);
Oleg Nesterov6279a7512009-03-27 01:06:07 +0100957 sig->cputimer.running = 1;
958 }
959
Frank Mayharf06febc2008-09-12 09:54:39 -0700960 /* The timer lists. */
961 INIT_LIST_HEAD(&sig->cpu_timers[0]);
962 INIT_LIST_HEAD(&sig->cpu_timers[1]);
963 INIT_LIST_HEAD(&sig->cpu_timers[2]);
964}
965
Alexey Dobriyana39bc512007-10-18 23:41:10 -0700966static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967{
968 struct signal_struct *sig;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969
Oleg Nesterov4ab6c082009-08-26 14:29:24 -0700970 if (clone_flags & CLONE_THREAD)
Peter Zijlstra490dea42008-11-24 17:06:57 +0100971 return 0;
Oleg Nesterov6279a7512009-03-27 01:06:07 +0100972
Veaceslav Falicoa56704e2010-03-10 15:23:01 -0800973 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 tsk->signal = sig;
975 if (!sig)
976 return -ENOMEM;
977
Oleg Nesterovb3ac0222010-05-26 14:43:24 -0700978 sig->nr_threads = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 atomic_set(&sig->live, 1);
Oleg Nesterovb3ac0222010-05-26 14:43:24 -0700980 atomic_set(&sig->sigcnt, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 init_waitqueue_head(&sig->wait_chldexit);
Sukadev Bhattiprolub3bfa0c2009-04-02 16:58:08 -0700982 if (clone_flags & CLONE_NEWPID)
983 sig->flags |= SIGNAL_UNKILLABLE;
Oleg Nesterovdb51aec2008-04-30 00:52:52 -0700984 sig->curr_target = tsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 init_sigpending(&sig->shared_pending);
986 INIT_LIST_HEAD(&sig->posix_timers);
987
Thomas Gleixnerc9cb2e32007-02-16 01:27:49 -0800988 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 sig->real_timer.function = it_real_fn;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 task_lock(current->group_leader);
992 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
993 task_unlock(current->group_leader);
994
Oleg Nesterov6279a7512009-03-27 01:06:07 +0100995 posix_cpu_timers_init_group(sig);
996
Miloslav Trmac522ed772007-07-15 23:40:56 -0700997 tty_audit_fork(sig);
Mike Galbraith5091faa2010-11-30 14:18:03 +0100998 sched_autogroup_fork(sig);
Miloslav Trmac522ed772007-07-15 23:40:56 -0700999
Ben Blum4714d1d2011-05-26 16:25:18 -07001000#ifdef CONFIG_CGROUPS
Tejun Heo257058a2011-12-12 18:12:21 -08001001 init_rwsem(&sig->group_rwsem);
Ben Blum4714d1d2011-05-26 16:25:18 -07001002#endif
1003
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001004 sig->oom_adj = current->signal->oom_adj;
David Rientjesa63d83f2010-08-09 17:19:46 -07001005 sig->oom_score_adj = current->signal->oom_score_adj;
Mandeep Singh Bainesdabb16f2011-01-13 15:46:05 -08001006 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001007
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001008 mutex_init(&sig->cred_guard_mutex);
1009
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 return 0;
1011}
1012
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001013static void copy_flags(unsigned long clone_flags, struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014{
1015 unsigned long new_flags = p->flags;
1016
Tejun Heo21aa9af2010-06-08 21:40:37 +02001017 new_flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 new_flags |= PF_FORKNOEXEC;
Roland McGrath09a05392008-07-25 19:45:47 -07001019 new_flags |= PF_STARTING;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 p->flags = new_flags;
1021}
1022
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001023SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024{
1025 current->clear_child_tid = tidptr;
1026
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001027 return task_pid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028}
1029
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001030static void rt_mutex_init_task(struct task_struct *p)
Ingo Molnar23f78d4a2006-06-27 02:54:53 -07001031{
Thomas Gleixner1d615482009-11-17 14:54:03 +01001032 raw_spin_lock_init(&p->pi_lock);
Zilvinas Valinskase29e1752007-03-16 13:38:34 -08001033#ifdef CONFIG_RT_MUTEXES
Dima Zavin732375c2011-07-07 17:27:59 -07001034 plist_head_init(&p->pi_waiters);
Ingo Molnar23f78d4a2006-06-27 02:54:53 -07001035 p->pi_blocked_on = NULL;
Ingo Molnar23f78d4a2006-06-27 02:54:53 -07001036#endif
1037}
1038
Balbir Singhcf475ad2008-04-29 01:00:16 -07001039#ifdef CONFIG_MM_OWNER
1040void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
1041{
1042 mm->owner = p;
1043}
1044#endif /* CONFIG_MM_OWNER */
1045
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046/*
Frank Mayharf06febc2008-09-12 09:54:39 -07001047 * Initialize POSIX timer handling for a single task.
1048 */
1049static void posix_cpu_timers_init(struct task_struct *tsk)
1050{
Martin Schwidefsky64861632011-12-15 14:56:09 +01001051 tsk->cputime_expires.prof_exp = 0;
1052 tsk->cputime_expires.virt_exp = 0;
Frank Mayharf06febc2008-09-12 09:54:39 -07001053 tsk->cputime_expires.sched_exp = 0;
1054 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1055 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1056 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1057}
1058
1059/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 * This creates a new process as a copy of the old one,
1061 * but does not actually start it yet.
1062 *
1063 * It copies the registers, and all the appropriate
1064 * parts of the process environment (as per the clone
1065 * flags). The actual kick-off is left to the caller.
1066 */
Ingo Molnar36c8b582006-07-03 00:25:41 -07001067static struct task_struct *copy_process(unsigned long clone_flags,
1068 unsigned long stack_start,
1069 struct pt_regs *regs,
1070 unsigned long stack_size,
Ingo Molnar36c8b582006-07-03 00:25:41 -07001071 int __user *child_tidptr,
Roland McGrath09a05392008-07-25 19:45:47 -07001072 struct pid *pid,
1073 int trace)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074{
1075 int retval;
Mariusz Kozlowskia24efe62007-10-18 23:41:09 -07001076 struct task_struct *p;
Paul Menageb4f48b62007-10-18 23:39:33 -07001077 int cgroup_callbacks_done = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078
1079 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1080 return ERR_PTR(-EINVAL);
1081
1082 /*
1083 * Thread groups must share signals as well, and detached threads
1084 * can only be started up within the thread group.
1085 */
1086 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1087 return ERR_PTR(-EINVAL);
1088
1089 /*
1090 * Shared signal handlers imply shared VM. By way of the above,
1091 * thread groups also imply shared VM. Blocking this case allows
1092 * for various simplifications in other code.
1093 */
1094 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1095 return ERR_PTR(-EINVAL);
1096
Sukadev Bhattiprolu123be072009-09-23 15:57:20 -07001097 /*
1098 * Siblings of global init remain as zombies on exit since they are
1099 * not reaped by their parent (swapper). To solve this and to avoid
1100 * multi-rooted process trees, prevent global and container-inits
1101 * from creating siblings.
1102 */
1103 if ((clone_flags & CLONE_PARENT) &&
1104 current->signal->flags & SIGNAL_UNKILLABLE)
1105 return ERR_PTR(-EINVAL);
1106
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 retval = security_task_create(clone_flags);
1108 if (retval)
1109 goto fork_out;
1110
1111 retval = -ENOMEM;
1112 p = dup_task_struct(current);
1113 if (!p)
1114 goto fork_out;
1115
Steven Rostedtf7e8b612009-06-02 16:39:48 -04001116 ftrace_graph_init_task(p);
1117
Peter Zijlstrabea493a2006-10-17 00:10:33 -07001118 rt_mutex_init_task(p);
1119
Ingo Molnard12c1a32008-07-14 12:09:28 +02001120#ifdef CONFIG_PROVE_LOCKING
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001121 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1122 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1123#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 retval = -EAGAIN;
David Howells3b11a1d2008-11-14 10:39:26 +11001125 if (atomic_read(&p->real_cred->user->processes) >=
Jiri Slaby78d7d402010-03-05 13:42:54 -08001126 task_rlimit(p, RLIMIT_NPROC)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 if (!capable(CAP_SYS_ADMIN) && !capable(CAP_SYS_RESOURCE) &&
Serge Hallyn18b6e042008-10-15 16:38:45 -05001128 p->real_cred->user != INIT_USER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 goto bad_fork_free;
1130 }
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001131 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132
David Howellsf1752ee2008-11-14 10:39:17 +11001133 retval = copy_creds(p, clone_flags);
1134 if (retval < 0)
1135 goto bad_fork_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136
1137 /*
1138 * If multiple threads are within copy_process(), then this check
1139 * triggers too late. This doesn't hurt, the check is only there
1140 * to stop root fork bombs.
1141 */
Li Zefan04ec93f2009-02-06 08:17:19 +00001142 retval = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 if (nr_threads >= max_threads)
1144 goto bad_fork_cleanup_count;
1145
Al Viroa1261f52005-11-13 16:06:55 -08001146 if (!try_module_get(task_thread_info(p)->exec_domain->module))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 goto bad_fork_cleanup_count;
1148
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 p->did_exec = 0;
Shailabh Nagarca74e922006-07-14 00:24:36 -07001150 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151 copy_flags(clone_flags, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 INIT_LIST_HEAD(&p->children);
1153 INIT_LIST_HEAD(&p->sibling);
Paul E. McKenneyf41d9112009-08-22 13:56:52 -07001154 rcu_copy_process(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155 p->vfork_done = NULL;
1156 spin_lock_init(&p->alloc_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 init_sigpending(&p->pending);
1159
Martin Schwidefsky64861632011-12-15 14:56:09 +01001160 p->utime = p->stime = p->gtime = 0;
1161 p->utimescaled = p->stimescaled = 0;
Hidetoshi Setod99ca3b2009-12-02 17:26:47 +09001162#ifndef CONFIG_VIRT_CPU_ACCOUNTING
Martin Schwidefsky64861632011-12-15 14:56:09 +01001163 p->prev_utime = p->prev_stime = 0;
Hidetoshi Setod99ca3b2009-12-02 17:26:47 +09001164#endif
KAMEZAWA Hiroyukia3a2e762010-04-06 14:34:42 -07001165#if defined(SPLIT_RSS_COUNTING)
1166 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1167#endif
Balbir Singh172ba842007-07-09 18:52:00 +02001168
Arjan van de Ven69766752008-09-01 15:52:40 -07001169 p->default_timer_slack_ns = current->timer_slack_ns;
1170
Andrea Righi59954772008-07-27 17:29:15 +02001171 task_io_accounting_init(&p->ioac);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 acct_clear_integrals(p);
1173
Frank Mayharf06febc2008-09-12 09:54:39 -07001174 posix_cpu_timers_init(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 do_posix_clock_monotonic_gettime(&p->start_time);
Tomas Janousek924b42d2007-07-15 23:39:42 -07001177 p->real_start_time = p->start_time;
1178 monotonic_to_bootbased(&p->real_start_time);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179 p->io_context = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 p->audit_context = NULL;
Ben Blum4714d1d2011-05-26 16:25:18 -07001181 if (clone_flags & CLONE_THREAD)
Tejun Heo257058a2011-12-12 18:12:21 -08001182 threadgroup_change_begin(current);
Paul Menageb4f48b62007-10-18 23:39:33 -07001183 cgroup_fork(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184#ifdef CONFIG_NUMA
Lee Schermerhorn846a16b2008-04-28 02:13:09 -07001185 p->mempolicy = mpol_dup(p->mempolicy);
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001186 if (IS_ERR(p->mempolicy)) {
1187 retval = PTR_ERR(p->mempolicy);
1188 p->mempolicy = NULL;
1189 goto bad_fork_cleanup_cgroup;
1190 }
Paul Jacksonc61afb12006-03-24 03:16:08 -08001191 mpol_fix_fork_child_flag(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192#endif
Michal Hocko778d3b02011-07-26 16:08:30 -07001193#ifdef CONFIG_CPUSETS
1194 p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1195 p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
1196#endif
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001197#ifdef CONFIG_TRACE_IRQFLAGS
1198 p->irq_events = 0;
Russell Kingb36e4752006-08-27 12:26:34 +01001199#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
1200 p->hardirqs_enabled = 1;
1201#else
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001202 p->hardirqs_enabled = 0;
Russell Kingb36e4752006-08-27 12:26:34 +01001203#endif
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001204 p->hardirq_enable_ip = 0;
1205 p->hardirq_enable_event = 0;
1206 p->hardirq_disable_ip = _THIS_IP_;
1207 p->hardirq_disable_event = 0;
1208 p->softirqs_enabled = 1;
1209 p->softirq_enable_ip = _THIS_IP_;
1210 p->softirq_enable_event = 0;
1211 p->softirq_disable_ip = 0;
1212 p->softirq_disable_event = 0;
1213 p->hardirq_context = 0;
1214 p->softirq_context = 0;
1215#endif
Ingo Molnarfbb9ce952006-07-03 00:24:50 -07001216#ifdef CONFIG_LOCKDEP
1217 p->lockdep_depth = 0; /* no locks held yet */
1218 p->curr_chain_key = 0;
1219 p->lockdep_recursion = 0;
1220#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221
Ingo Molnar408894e2006-01-09 15:59:20 -08001222#ifdef CONFIG_DEBUG_MUTEXES
1223 p->blocked_on = NULL; /* not blocked yet */
1224#endif
KAMEZAWA Hiroyuki569b8462009-12-15 16:47:03 -08001225#ifdef CONFIG_CGROUP_MEM_RES_CTLR
1226 p->memcg_batch.do_batch = 0;
1227 p->memcg_batch.memcg = NULL;
1228#endif
Markus Metzger0f481402009-04-03 16:43:48 +02001229
Srivatsa Vaddagiri3c90e6e2007-11-09 22:39:39 +01001230 /* Perform scheduler related setup. Assign this task to a CPU. */
Samir Bellabes3e51e3e2011-05-11 18:18:05 +02001231 sched_fork(p);
Peter Zijlstra6ab423e2009-05-25 14:45:27 +02001232
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001233 retval = perf_event_init_task(p);
Peter Zijlstra6ab423e2009-05-25 14:45:27 +02001234 if (retval)
1235 goto bad_fork_cleanup_policy;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001236 retval = audit_alloc(p);
1237 if (retval)
David Howellsf1752ee2008-11-14 10:39:17 +11001238 goto bad_fork_cleanup_policy;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 /* copy all the process information */
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001240 retval = copy_semundo(clone_flags, p);
1241 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242 goto bad_fork_cleanup_audit;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001243 retval = copy_files(clone_flags, p);
1244 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 goto bad_fork_cleanup_semundo;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001246 retval = copy_fs(clone_flags, p);
1247 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 goto bad_fork_cleanup_files;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001249 retval = copy_sighand(clone_flags, p);
1250 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 goto bad_fork_cleanup_fs;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001252 retval = copy_signal(clone_flags, p);
1253 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 goto bad_fork_cleanup_sighand;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001255 retval = copy_mm(clone_flags, p);
1256 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257 goto bad_fork_cleanup_signal;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001258 retval = copy_namespaces(clone_flags, p);
1259 if (retval)
David Howellsd84f4f92008-11-14 10:39:23 +11001260 goto bad_fork_cleanup_mm;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001261 retval = copy_io(clone_flags, p);
1262 if (retval)
Jens Axboefd0928d2008-01-24 08:52:45 +01001263 goto bad_fork_cleanup_namespaces;
Alexey Dobriyan6f2c55b2009-04-02 16:56:59 -07001264 retval = copy_thread(clone_flags, stack_start, stack_size, p, regs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265 if (retval)
Jens Axboefd0928d2008-01-24 08:52:45 +01001266 goto bad_fork_cleanup_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001268 if (pid != &init_struct_pid) {
1269 retval = -ENOMEM;
Eric W. Biederman61bce0f2009-01-07 18:08:49 -08001270 pid = alloc_pid(p->nsproxy->pid_ns);
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001271 if (!pid)
Jens Axboefd0928d2008-01-24 08:52:45 +01001272 goto bad_fork_cleanup_io;
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001273 }
1274
1275 p->pid = pid_nr(pid);
1276 p->tgid = p->pid;
1277 if (clone_flags & CLONE_THREAD)
1278 p->tgid = current->tgid;
1279
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1281 /*
1282 * Clear TID on mm_release()?
1283 */
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001284 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
Jens Axboe73c10102011-03-08 13:19:51 +01001285#ifdef CONFIG_BLOCK
1286 p->plug = NULL;
1287#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001288#ifdef CONFIG_FUTEX
Ingo Molnar8f17d3a2006-03-27 01:16:27 -08001289 p->robust_list = NULL;
1290#ifdef CONFIG_COMPAT
1291 p->compat_robust_list = NULL;
1292#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07001293 INIT_LIST_HEAD(&p->pi_state_list);
1294 p->pi_state_cache = NULL;
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001295#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 /*
GOTO Masanorif9a38792006-03-13 21:20:44 -08001297 * sigaltstack should be cleared when sharing the same VM
1298 */
1299 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
1300 p->sas_ss_sp = p->sas_ss_size = 0;
1301
1302 /*
Oleg Nesterov65808072009-12-15 16:47:16 -08001303 * Syscall tracing and stepping should be turned off in the
1304 * child regardless of CLONE_PTRACE.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 */
Oleg Nesterov65808072009-12-15 16:47:16 -08001306 user_disable_single_step(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
Laurent Viviered75e8d2005-09-03 15:57:18 -07001308#ifdef TIF_SYSCALL_EMU
1309 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1310#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01001311 clear_all_latency_tracing(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313 /* ok, now we should be set up.. */
1314 p->exit_signal = (clone_flags & CLONE_THREAD) ? -1 : (clone_flags & CSIGNAL);
1315 p->pdeath_signal = 0;
1316 p->exit_state = 0;
1317
Wu Fengguang9d823e82011-06-11 18:10:12 -06001318 p->nr_dirtied = 0;
1319 p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
Wu Fengguang83712352011-06-11 19:25:42 -06001320 p->dirty_paused_when = 0;
Wu Fengguang9d823e82011-06-11 18:10:12 -06001321
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322 /*
1323 * Ok, make it visible to the rest of the system.
1324 * We dont wake it up yet.
1325 */
1326 p->group_leader = p;
Oleg Nesterov47e65322006-03-28 16:11:25 -08001327 INIT_LIST_HEAD(&p->thread_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328
Paul Menageb4f48b62007-10-18 23:39:33 -07001329 /* Now that the task is set up, run cgroup callbacks if
1330 * necessary. We need to run them before the task is visible
1331 * on the tasklist. */
1332 cgroup_fork_callbacks(p);
1333 cgroup_callbacks_done = 1;
1334
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 /* Need tasklist lock for parent etc handling! */
1336 write_lock_irq(&tasklist_lock);
1337
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 /* CLONE_PARENT re-uses the old parent */
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01001339 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340 p->real_parent = current->real_parent;
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01001341 p->parent_exec_id = current->parent_exec_id;
1342 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 p->real_parent = current;
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01001344 p->parent_exec_id = current->self_exec_id;
1345 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346
Oleg Nesterov3f17da62006-02-15 22:13:24 +03001347 spin_lock(&current->sighand->siglock);
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001348
1349 /*
1350 * Process group and session signals need to be delivered to just the
1351 * parent before the fork or both the parent and the child after the
1352 * fork. Restart if a signal comes in before we add the new process to
1353 * it's process group.
1354 * A fatal signal pending means that current will exit, so the new
1355 * thread can't slip out of an OOM kill (or normal SIGKILL).
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001356 */
Daniel Walker23ff4442007-10-18 03:06:07 -07001357 recalc_sigpending();
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001358 if (signal_pending(current)) {
1359 spin_unlock(&current->sighand->siglock);
1360 write_unlock_irq(&tasklist_lock);
1361 retval = -ERESTARTNOINTR;
Steven Rostedtf7e8b612009-06-02 16:39:48 -04001362 goto bad_fork_free_pid;
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001363 }
1364
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 if (clone_flags & CLONE_THREAD) {
Oleg Nesterovb3ac0222010-05-26 14:43:24 -07001366 current->signal->nr_threads++;
Oleg Nesterov4ab6c082009-08-26 14:29:24 -07001367 atomic_inc(&current->signal->live);
Oleg Nesterovb3ac0222010-05-26 14:43:24 -07001368 atomic_inc(&current->signal->sigcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 p->group_leader = current->group_leader;
Oleg Nesterov47e65322006-03-28 16:11:25 -08001370 list_add_tail_rcu(&p->thread_group, &p->group_leader->thread_group);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 }
1372
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001373 if (likely(p->pid)) {
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001374 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001376 if (thread_group_leader(p)) {
Eric W. Biederman45a68622011-03-23 16:43:12 -07001377 if (is_child_reaper(pid))
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001378 p->nsproxy->pid_ns->child_reaper = p;
Oleg Nesterovc97d9892006-03-28 16:11:06 -08001379
Oleg Nesterovfea9d172008-02-08 04:19:19 -08001380 p->signal->leader_pid = pid;
Alan Cox9c9f4de2008-10-13 10:37:26 +01001381 p->signal->tty = tty_kref_get(current->signal->tty);
Eric W. Biederman5cd17562007-12-04 23:45:04 -08001382 attach_pid(p, PIDTYPE_PGID, task_pgrp(current));
1383 attach_pid(p, PIDTYPE_SID, task_session(current));
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001384 list_add_tail(&p->sibling, &p->real_parent->children);
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -07001385 list_add_tail_rcu(&p->tasks, &init_task.tasks);
Christoph Lameter909ea962010-12-08 16:22:55 +01001386 __this_cpu_inc(process_counts);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001387 }
Sukadev Bhattiprolu85868992007-05-10 22:23:03 -07001388 attach_pid(p, PIDTYPE_PID, pid);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001389 nr_threads++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 }
1391
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 total_forks++;
Oleg Nesterov3f17da62006-02-15 22:13:24 +03001393 spin_unlock(&current->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 write_unlock_irq(&tasklist_lock);
Andrew Mortonc13cf852005-11-28 13:43:48 -08001395 proc_fork_connector(p);
Paul Menage817929e2007-10-18 23:39:36 -07001396 cgroup_post_fork(p);
Ben Blum4714d1d2011-05-26 16:25:18 -07001397 if (clone_flags & CLONE_THREAD)
Tejun Heo257058a2011-12-12 18:12:21 -08001398 threadgroup_change_end(current);
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001399 perf_event_fork(p);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001400
1401 trace_task_newtask(p, clone_flags);
1402
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403 return p;
1404
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001405bad_fork_free_pid:
1406 if (pid != &init_struct_pid)
1407 free_pid(pid);
Jens Axboefd0928d2008-01-24 08:52:45 +01001408bad_fork_cleanup_io:
Louis Rillingb69f2292009-12-04 14:52:42 +01001409 if (p->io_context)
1410 exit_io_context(p);
Serge E. Hallynab516012006-10-02 02:18:06 -07001411bad_fork_cleanup_namespaces:
Linus Torvalds444f3782007-01-30 13:35:18 -08001412 exit_task_namespaces(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413bad_fork_cleanup_mm:
David Rientjesc9f01242011-10-31 17:07:15 -07001414 if (p->mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415 mmput(p->mm);
1416bad_fork_cleanup_signal:
Oleg Nesterov4ab6c082009-08-26 14:29:24 -07001417 if (!(clone_flags & CLONE_THREAD))
Mike Galbraith1c5354d2011-01-05 11:16:04 +01001418 free_signal_struct(p->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419bad_fork_cleanup_sighand:
Oleg Nesterova7e53282006-03-28 16:11:27 -08001420 __cleanup_sighand(p->sighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421bad_fork_cleanup_fs:
1422 exit_fs(p); /* blocking */
1423bad_fork_cleanup_files:
1424 exit_files(p); /* blocking */
1425bad_fork_cleanup_semundo:
1426 exit_sem(p);
1427bad_fork_cleanup_audit:
1428 audit_free(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429bad_fork_cleanup_policy:
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001430 perf_event_free_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431#ifdef CONFIG_NUMA
Lee Schermerhornf0be3d32008-04-28 02:13:08 -07001432 mpol_put(p->mempolicy);
Paul Menageb4f48b62007-10-18 23:39:33 -07001433bad_fork_cleanup_cgroup:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434#endif
Ben Blum4714d1d2011-05-26 16:25:18 -07001435 if (clone_flags & CLONE_THREAD)
Tejun Heo257058a2011-12-12 18:12:21 -08001436 threadgroup_change_end(current);
Paul Menageb4f48b62007-10-18 23:39:33 -07001437 cgroup_exit(p, cgroup_callbacks_done);
Shailabh Nagar35df17c2006-08-31 21:27:38 -07001438 delayacct_tsk_free(p);
Al Viroa1261f52005-11-13 16:06:55 -08001439 module_put(task_thread_info(p)->exec_domain->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440bad_fork_cleanup_count:
David Howellsd84f4f92008-11-14 10:39:23 +11001441 atomic_dec(&p->cred->user->processes);
David Howellse0e81732009-09-02 09:13:40 +01001442 exit_creds(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443bad_fork_free:
1444 free_task(p);
Oleg Nesterovfe7d37d2006-01-08 01:04:02 -08001445fork_out:
1446 return ERR_PTR(retval);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447}
1448
Adrian Bunk6b2fb3c2008-02-06 01:37:55 -08001449noinline struct pt_regs * __cpuinit __attribute__((weak)) idle_regs(struct pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450{
1451 memset(regs, 0, sizeof(struct pt_regs));
1452 return regs;
1453}
1454
Oleg Nesterovf106eee2010-05-26 14:44:11 -07001455static inline void init_idle_pids(struct pid_link *links)
1456{
1457 enum pid_type type;
1458
1459 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1460 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1461 links[type].pid = &init_struct_pid;
1462 }
1463}
1464
Al Viro9abcf402007-02-01 13:52:48 +00001465struct task_struct * __cpuinit fork_idle(int cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466{
Ingo Molnar36c8b582006-07-03 00:25:41 -07001467 struct task_struct *task;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 struct pt_regs regs;
1469
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001470 task = copy_process(CLONE_VM, 0, idle_regs(&regs), 0, NULL,
Roland McGrath09a05392008-07-25 19:45:47 -07001471 &init_struct_pid, 0);
Oleg Nesterovf106eee2010-05-26 14:44:11 -07001472 if (!IS_ERR(task)) {
1473 init_idle_pids(task->pids);
Akinobu Mita753ca4f2006-11-25 11:09:34 -08001474 init_idle(task, cpu);
Oleg Nesterovf106eee2010-05-26 14:44:11 -07001475 }
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001476
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477 return task;
1478}
1479
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480/*
1481 * Ok, this is the main fork-routine.
1482 *
1483 * It copies the process, and if successful kick-starts
1484 * it and waits for it to finish using the VM if required.
1485 */
1486long do_fork(unsigned long clone_flags,
1487 unsigned long stack_start,
1488 struct pt_regs *regs,
1489 unsigned long stack_size,
1490 int __user *parent_tidptr,
1491 int __user *child_tidptr)
1492{
1493 struct task_struct *p;
1494 int trace = 0;
Eric W. Biederman92476d72006-03-31 02:31:42 -08001495 long nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496
Andrew Mortonbdff7462008-02-04 22:27:22 -08001497 /*
Serge Hallyn18b6e042008-10-15 16:38:45 -05001498 * Do some preliminary argument and permissions checking before we
1499 * actually start allocating stuff
1500 */
1501 if (clone_flags & CLONE_NEWUSER) {
1502 if (clone_flags & CLONE_THREAD)
1503 return -EINVAL;
1504 /* hopefully this check will go away when userns support is
1505 * complete
1506 */
Serge E. Hallyn7657d902008-12-03 13:17:33 -06001507 if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SETUID) ||
1508 !capable(CAP_SETGID))
Serge Hallyn18b6e042008-10-15 16:38:45 -05001509 return -EPERM;
1510 }
1511
1512 /*
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001513 * Determine whether and which event to report to ptracer. When
1514 * called from kernel_thread or CLONE_UNTRACED is explicitly
1515 * requested, no event is reported; otherwise, report if the event
1516 * for the type of forking is enabled.
Roland McGrath09a05392008-07-25 19:45:47 -07001517 */
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001518 if (likely(user_mode(regs)) && !(clone_flags & CLONE_UNTRACED)) {
1519 if (clone_flags & CLONE_VFORK)
1520 trace = PTRACE_EVENT_VFORK;
1521 else if ((clone_flags & CSIGNAL) != SIGCHLD)
1522 trace = PTRACE_EVENT_CLONE;
1523 else
1524 trace = PTRACE_EVENT_FORK;
1525
1526 if (likely(!ptrace_event_enabled(current, trace)))
1527 trace = 0;
1528 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529
Sukadev Bhattiprolua6f5e062007-10-18 23:39:53 -07001530 p = copy_process(clone_flags, stack_start, regs, stack_size,
Roland McGrath09a05392008-07-25 19:45:47 -07001531 child_tidptr, NULL, trace);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 /*
1533 * Do this prior waking up the new thread - the thread pointer
1534 * might get invalid after that point, if the thread exits quickly.
1535 */
1536 if (!IS_ERR(p)) {
1537 struct completion vfork;
1538
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04001539 trace_sched_process_fork(current, p);
1540
Pavel Emelyanov6c5f3e72008-02-08 04:19:20 -08001541 nr = task_pid_vnr(p);
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001542
1543 if (clone_flags & CLONE_PARENT_SETTID)
1544 put_user(nr, parent_tidptr);
Sukadev Bhattiprolua6f5e062007-10-18 23:39:53 -07001545
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546 if (clone_flags & CLONE_VFORK) {
1547 p->vfork_done = &vfork;
1548 init_completion(&vfork);
1549 }
1550
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001551 /*
Roland McGrath09a05392008-07-25 19:45:47 -07001552 * We set PF_STARTING at creation in case tracing wants to
1553 * use this to distinguish a fully live task from one that
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001554 * hasn't finished SIGSTOP raising yet. Now we clear it
1555 * and set the child going.
Roland McGrath09a05392008-07-25 19:45:47 -07001556 */
1557 p->flags &= ~PF_STARTING;
1558
Samir Bellabes3e51e3e2011-05-11 18:18:05 +02001559 wake_up_new_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001561 /* forking complete and child started to run, tell ptracer */
1562 if (unlikely(trace))
1563 ptrace_event(trace, nr);
Roland McGrath09a05392008-07-25 19:45:47 -07001564
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 if (clone_flags & CLONE_VFORK) {
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -07001566 freezer_do_not_count();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 wait_for_completion(&vfork);
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -07001568 freezer_count();
Tejun Heoa288eec2011-06-17 16:50:37 +02001569 ptrace_event(PTRACE_EVENT_VFORK_DONE, nr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570 }
1571 } else {
Eric W. Biederman92476d72006-03-31 02:31:42 -08001572 nr = PTR_ERR(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 }
Eric W. Biederman92476d72006-03-31 02:31:42 -08001574 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575}
1576
Ravikiran G Thirumalai5fd63b32006-01-11 22:46:15 +01001577#ifndef ARCH_MIN_MMSTRUCT_ALIGN
1578#define ARCH_MIN_MMSTRUCT_ALIGN 0
1579#endif
1580
Alexey Dobriyan51cc5062008-07-25 19:45:34 -07001581static void sighand_ctor(void *data)
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08001582{
1583 struct sighand_struct *sighand = data;
1584
Christoph Lametera35afb82007-05-16 22:10:57 -07001585 spin_lock_init(&sighand->siglock);
Davide Libenzib8fceee2007-09-20 12:40:16 -07001586 init_waitqueue_head(&sighand->signalfd_wqh);
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08001587}
1588
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589void __init proc_caches_init(void)
1590{
1591 sighand_cachep = kmem_cache_create("sighand_cache",
1592 sizeof(struct sighand_struct), 0,
Vegard Nossum2dff4402008-05-31 15:56:17 +02001593 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
1594 SLAB_NOTRACK, sighand_ctor);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 signal_cachep = kmem_cache_create("signal_cache",
1596 sizeof(struct signal_struct), 0,
Vegard Nossum2dff4402008-05-31 15:56:17 +02001597 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
Paul Mundt20c2df82007-07-20 10:11:58 +09001598 files_cachep = kmem_cache_create("files_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 sizeof(struct files_struct), 0,
Vegard Nossum2dff4402008-05-31 15:56:17 +02001600 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
Paul Mundt20c2df82007-07-20 10:11:58 +09001601 fs_cachep = kmem_cache_create("fs_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 sizeof(struct fs_struct), 0,
Vegard Nossum2dff4402008-05-31 15:56:17 +02001603 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
Linus Torvalds6345d242011-05-29 11:32:28 -07001604 /*
1605 * FIXME! The "sizeof(struct mm_struct)" currently includes the
1606 * whole struct cpumask for the OFFSTACK case. We could change
1607 * this to *only* allocate as much of it as required by the
1608 * maximum number of CPU's we can ever have. The cpumask_allocation
1609 * is at the end of the structure, exactly for that reason.
1610 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 mm_cachep = kmem_cache_create("mm_struct",
Ravikiran G Thirumalai5fd63b32006-01-11 22:46:15 +01001612 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
Vegard Nossum2dff4402008-05-31 15:56:17 +02001613 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK, NULL);
David Howells33e5d7692009-04-02 16:56:32 -07001614 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC);
David Howells8feae132009-01-08 12:04:47 +00001615 mmap_init();
Al Viro66577192011-06-28 15:41:10 -04001616 nsproxy_cache_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617}
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001618
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001619/*
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07001620 * Check constraints on flags passed to the unshare system call.
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001621 */
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07001622static int check_unshare_flags(unsigned long unshare_flags)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001623{
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07001624 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
1625 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
1626 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET))
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001627 return -EINVAL;
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07001628 /*
1629 * Not implemented, but pretend it works if there is nothing to
1630 * unshare. Note that unsharing CLONE_THREAD or CLONE_SIGHAND
1631 * needs to unshare vm.
1632 */
1633 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
1634 /* FIXME: get_task_mm() increments ->mm_users */
1635 if (atomic_read(&current->mm->mm_users) > 1)
1636 return -EINVAL;
1637 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001638
1639 return 0;
1640}
1641
1642/*
JANAK DESAI99d14192006-02-07 12:58:59 -08001643 * Unshare the filesystem structure if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001644 */
1645static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
1646{
1647 struct fs_struct *fs = current->fs;
1648
Al Viro498052b2009-03-30 07:20:30 -04001649 if (!(unshare_flags & CLONE_FS) || !fs)
1650 return 0;
1651
1652 /* don't need lock here; in the worst case we'll do useless copy */
1653 if (fs->users == 1)
1654 return 0;
1655
1656 *new_fsp = copy_fs_struct(fs);
1657 if (!*new_fsp)
1658 return -ENOMEM;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001659
1660 return 0;
1661}
1662
1663/*
JANAK DESAIa016f332006-02-07 12:59:02 -08001664 * Unshare file descriptor table if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001665 */
1666static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
1667{
1668 struct files_struct *fd = current->files;
JANAK DESAIa016f332006-02-07 12:59:02 -08001669 int error = 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001670
1671 if ((unshare_flags & CLONE_FILES) &&
JANAK DESAIa016f332006-02-07 12:59:02 -08001672 (fd && atomic_read(&fd->count) > 1)) {
1673 *new_fdp = dup_fd(fd, &error);
1674 if (!*new_fdp)
1675 return error;
1676 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001677
1678 return 0;
1679}
1680
1681/*
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001682 * unshare allows a process to 'unshare' part of the process
1683 * context which was originally shared using clone. copy_*
1684 * functions used by do_fork() cannot be used here directly
1685 * because they modify an inactive task_struct that is being
1686 * constructed. Here we are modifying the current, active,
1687 * task_struct.
1688 */
Heiko Carstens6559eed82009-01-14 14:14:32 +01001689SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001690{
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001691 struct fs_struct *fs, *new_fs = NULL;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001692 struct files_struct *fd, *new_fd = NULL;
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07001693 struct nsproxy *new_nsproxy = NULL;
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001694 int do_sysvsem = 0;
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07001695 int err;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001696
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07001697 err = check_unshare_flags(unshare_flags);
1698 if (err)
Eric W. Biederman06f9d4f2006-03-22 00:07:40 -08001699 goto bad_unshare_out;
1700
Manfred Spraul6013f672008-04-29 01:00:59 -07001701 /*
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07001702 * If unsharing namespace, must also unshare filesystem information.
1703 */
1704 if (unshare_flags & CLONE_NEWNS)
1705 unshare_flags |= CLONE_FS;
1706 /*
Manfred Spraul6013f672008-04-29 01:00:59 -07001707 * CLONE_NEWIPC must also detach from the undolist: after switching
1708 * to a new ipc namespace, the semaphore arrays from the old
1709 * namespace are unreachable.
1710 */
1711 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001712 do_sysvsem = 1;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001713 err = unshare_fs(unshare_flags, &new_fs);
1714 if (err)
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07001715 goto bad_unshare_out;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001716 err = unshare_fd(unshare_flags, &new_fd);
1717 if (err)
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07001718 goto bad_unshare_cleanup_fs;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001719 err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy, new_fs);
1720 if (err)
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001721 goto bad_unshare_cleanup_fd;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001722
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07001723 if (new_fs || new_fd || do_sysvsem || new_nsproxy) {
Manfred Spraul9edff4a2008-04-29 01:00:57 -07001724 if (do_sysvsem) {
1725 /*
1726 * CLONE_SYSVSEM is equivalent to sys_exit().
1727 */
1728 exit_sem(current);
1729 }
Serge E. Hallynab516012006-10-02 02:18:06 -07001730
Serge Hallync0b2fc32006-10-02 02:18:18 -07001731 if (new_nsproxy) {
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07001732 switch_task_namespaces(current, new_nsproxy);
1733 new_nsproxy = NULL;
Serge Hallync0b2fc32006-10-02 02:18:18 -07001734 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001735
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07001736 task_lock(current);
1737
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001738 if (new_fs) {
1739 fs = current->fs;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001740 spin_lock(&fs->lock);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001741 current->fs = new_fs;
Al Viro498052b2009-03-30 07:20:30 -04001742 if (--fs->users)
1743 new_fs = NULL;
1744 else
1745 new_fs = fs;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001746 spin_unlock(&fs->lock);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001747 }
1748
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001749 if (new_fd) {
1750 fd = current->files;
1751 current->files = new_fd;
1752 new_fd = fd;
1753 }
1754
1755 task_unlock(current);
1756 }
1757
Serge Hallync0b2fc32006-10-02 02:18:18 -07001758 if (new_nsproxy)
Linus Torvalds444f3782007-01-30 13:35:18 -08001759 put_nsproxy(new_nsproxy);
Serge Hallync0b2fc32006-10-02 02:18:18 -07001760
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001761bad_unshare_cleanup_fd:
1762 if (new_fd)
1763 put_files_struct(new_fd);
1764
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001765bad_unshare_cleanup_fs:
1766 if (new_fs)
Al Viro498052b2009-03-30 07:20:30 -04001767 free_fs_struct(new_fs);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001768
JANAK DESAIcf2e3402006-02-07 12:58:58 -08001769bad_unshare_out:
1770 return err;
1771}
Al Viro3b125382008-04-22 05:31:30 -04001772
1773/*
1774 * Helper to unshare the files of the current task.
1775 * We don't want to expose copy_files internals to
1776 * the exec layer of the kernel.
1777 */
1778
1779int unshare_files(struct files_struct **displaced)
1780{
1781 struct task_struct *task = current;
Al Viro50704512008-04-26 05:25:00 +01001782 struct files_struct *copy = NULL;
Al Viro3b125382008-04-22 05:31:30 -04001783 int error;
1784
1785 error = unshare_fd(CLONE_FILES, &copy);
1786 if (error || !copy) {
1787 *displaced = NULL;
1788 return error;
1789 }
1790 *displaced = task->files;
1791 task_lock(task);
1792 task->files = copy;
1793 task_unlock(task);
1794 return 0;
1795}