blob: 234f4bfb8efd03b5eab94d07c15ad07cb82215a3 [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>
Ingo Molnar4eb5aaa2017-02-08 18:51:29 +010015#include <linux/sched/autogroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/init.h>
17#include <linux/unistd.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <linux/module.h>
19#include <linux/vmalloc.h>
20#include <linux/completion.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>
Davidlohr Bueso615d6e82014-04-07 15:37:25 -070032#include <linux/mm.h>
33#include <linux/vmacache.h>
Serge E. Hallynab516012006-10-02 02:18:06 -070034#include <linux/nsproxy.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080035#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#include <linux/cpu.h>
Paul Menageb4f48b62007-10-18 23:39:33 -070037#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038#include <linux/security.h>
Mel Gormana1e78772008-07-23 21:27:23 -070039#include <linux/hugetlb.h>
Will Drewrye2cfabdf2012-04-12 16:47:57 -050040#include <linux/seccomp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#include <linux/swap.h>
42#include <linux/syscalls.h>
43#include <linux/jiffies.h>
44#include <linux/futex.h>
Linus Torvalds8141c7f2008-11-15 10:20:36 -080045#include <linux/compat.h>
Eric Dumazet207205a2011-03-22 16:30:44 -070046#include <linux/kthread.h>
Andrew Morton7c3ab7382006-12-10 02:19:19 -080047#include <linux/task_io_accounting_ops.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070048#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070049#include <linux/ptrace.h>
50#include <linux/mount.h>
51#include <linux/audit.h>
Pavel Emelianov78fb7462008-02-07 00:13:51 -080052#include <linux/memcontrol.h>
Frederic Weisbeckerf201ae22008-11-23 06:22:56 +010053#include <linux/ftrace.h>
Mike Galbraith5e2bf012012-05-10 13:01:45 -070054#include <linux/proc_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070055#include <linux/profile.h>
56#include <linux/rmap.h>
Hugh Dickinsf8af4da2009-09-21 17:01:57 -070057#include <linux/ksm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070058#include <linux/acct.h>
Pavel Emelyanov893e26e2017-02-22 15:42:27 -080059#include <linux/userfaultfd_k.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070060#include <linux/tsacct_kern.h>
Matt Helsley9f460802005-11-07 00:59:16 -080061#include <linux/cn_proc.h>
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -070062#include <linux/freezer.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070063#include <linux/delayacct.h>
Shailabh Nagarad4ecbc2006-07-14 00:24:44 -070064#include <linux/taskstats_kern.h>
Arjan van de Ven0a4254052006-09-26 10:52:38 +020065#include <linux/random.h>
Miloslav Trmac522ed772007-07-15 23:40:56 -070066#include <linux/tty.h>
Jens Axboefd0928d2008-01-24 08:52:45 +010067#include <linux/blkdev.h>
Al Viro5ad4e532009-03-29 19:50:06 -040068#include <linux/fs_struct.h>
Eric Sandeen7c9f8862008-04-22 16:38:23 -050069#include <linux/magic.h>
Ingo Molnarcdd6c482009-09-21 12:02:48 +020070#include <linux/perf_event.h>
Stanislaw Gruszka42c4ab42009-07-29 12:15:26 +020071#include <linux/posix-timers.h>
Avi Kivity8e7cac72009-11-29 16:34:48 +020072#include <linux/user-return-notifier.h>
Ying Han3d5992d2010-10-26 14:21:23 -070073#include <linux/oom.h>
Andrea Arcangeliba761492011-01-13 15:46:58 -080074#include <linux/khugepaged.h>
Oleg Nesterovd80e7312012-02-24 20:07:11 +010075#include <linux/signalfd.h>
Srikar Dronamraju0326f5a2012-03-13 23:30:11 +053076#include <linux/uprobes.h>
Kent Overstreeta27bb332013-05-07 16:19:08 -070077#include <linux/aio.h>
Gideon Israel Dsouza52f5684c2014-04-07 15:39:20 -070078#include <linux/compiler.h>
Heinrich Schuchardt16db3d32015-04-16 12:47:50 -070079#include <linux/sysctl.h>
Dmitry Vyukov5c9a8752016-03-22 14:27:30 -070080#include <linux/kcov.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070081
82#include <asm/pgtable.h>
83#include <asm/pgalloc.h>
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080084#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070085#include <asm/mmu_context.h>
86#include <asm/cacheflush.h>
87#include <asm/tlbflush.h>
88
Steven Rostedtad8d75f2009-04-14 19:39:12 -040089#include <trace/events/sched.h>
90
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -080091#define CREATE_TRACE_POINTS
92#include <trace/events/task.h>
93
Linus Torvalds1da177e2005-04-16 15:20:36 -070094/*
Heinrich Schuchardtac1b3982015-04-16 12:47:47 -070095 * Minimum number of threads to boot the kernel
96 */
97#define MIN_THREADS 20
98
99/*
100 * Maximum number of threads
101 */
102#define MAX_THREADS FUTEX_TID_MASK
103
104/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 * Protected counters by write_lock_irq(&tasklist_lock)
106 */
107unsigned long total_forks; /* Handle normal Linux uptimes. */
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700108int nr_threads; /* The idle threads do not count.. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109
110int max_threads; /* tunable limit on nr_threads */
111
112DEFINE_PER_CPU(unsigned long, process_counts) = 0;
113
Christoph Hellwigc59923a2006-07-10 04:45:40 -0700114__cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
Paul E. McKenneydb1466b2010-03-03 07:46:56 -0800115
116#ifdef CONFIG_PROVE_RCU
117int lockdep_tasklist_lock_is_held(void)
118{
119 return lockdep_is_held(&tasklist_lock);
120}
121EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
122#endif /* #ifdef CONFIG_PROVE_RCU */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123
124int nr_processes(void)
125{
126 int cpu;
127 int total = 0;
128
Ian Campbell1d510752009-11-03 10:11:14 +0000129 for_each_possible_cpu(cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 total += per_cpu(process_counts, cpu);
131
132 return total;
133}
134
Akinobu Mitaf19b9f72012-07-30 14:42:33 -0700135void __weak arch_release_task_struct(struct task_struct *tsk)
136{
137}
138
Thomas Gleixnerf5e10282012-05-05 15:05:48 +0000139#ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
Christoph Lametere18b8902006-12-06 20:33:20 -0800140static struct kmem_cache *task_struct_cachep;
Thomas Gleixner41101802012-05-05 15:05:41 +0000141
142static inline struct task_struct *alloc_task_struct_node(int node)
143{
144 return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node);
145}
146
Thomas Gleixner41101802012-05-05 15:05:41 +0000147static inline void free_task_struct(struct task_struct *tsk)
148{
Thomas Gleixner41101802012-05-05 15:05:41 +0000149 kmem_cache_free(task_struct_cachep, tsk);
150}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151#endif
152
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700153void __weak arch_release_thread_stack(unsigned long *stack)
Akinobu Mitaf19b9f72012-07-30 14:42:33 -0700154{
155}
156
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700157#ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR
Thomas Gleixner41101802012-05-05 15:05:41 +0000158
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000159/*
160 * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
161 * kmemcache based allocator.
162 */
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700163# if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK)
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700164
165#ifdef CONFIG_VMAP_STACK
166/*
167 * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB
168 * flush. Try to minimize the number of calls by caching stacks.
169 */
170#define NR_CACHED_STACKS 2
171static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]);
172#endif
173
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700174static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node)
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700175{
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700176#ifdef CONFIG_VMAP_STACK
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700177 void *stack;
178 int i;
179
180 local_irq_disable();
181 for (i = 0; i < NR_CACHED_STACKS; i++) {
182 struct vm_struct *s = this_cpu_read(cached_stacks[i]);
183
184 if (!s)
185 continue;
186 this_cpu_write(cached_stacks[i], NULL);
187
188 tsk->stack_vm_area = s;
189 local_irq_enable();
190 return s->addr;
191 }
192 local_irq_enable();
193
194 stack = __vmalloc_node_range(THREAD_SIZE, THREAD_SIZE,
195 VMALLOC_START, VMALLOC_END,
196 THREADINFO_GFP | __GFP_HIGHMEM,
197 PAGE_KERNEL,
198 0, node, __builtin_return_address(0));
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700199
200 /*
201 * We can't call find_vm_area() in interrupt context, and
202 * free_thread_stack() can be called in interrupt context,
203 * so cache the vm_struct.
204 */
205 if (stack)
206 tsk->stack_vm_area = find_vm_area(stack);
207 return stack;
208#else
Vladimir Davydov49491482016-07-26 15:24:24 -0700209 struct page *page = alloc_pages_node(node, THREADINFO_GFP,
210 THREAD_SIZE_ORDER);
Eric Dumazetb6a84012011-03-22 16:30:42 -0700211
212 return page ? page_address(page) : NULL;
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700213#endif
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700214}
215
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700216static inline void free_thread_stack(struct task_struct *tsk)
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700217{
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700218#ifdef CONFIG_VMAP_STACK
219 if (task_stack_vm_area(tsk)) {
220 unsigned long flags;
221 int i;
222
223 local_irq_save(flags);
224 for (i = 0; i < NR_CACHED_STACKS; i++) {
225 if (this_cpu_read(cached_stacks[i]))
226 continue;
227
228 this_cpu_write(cached_stacks[i], tsk->stack_vm_area);
229 local_irq_restore(flags);
230 return;
231 }
232 local_irq_restore(flags);
233
Andrey Ryabinin0f110a92016-12-12 16:44:14 -0800234 vfree_atomic(tsk->stack);
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700235 return;
236 }
237#endif
238
239 __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER);
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700240}
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000241# else
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700242static struct kmem_cache *thread_stack_cache;
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000243
Michael Ellerman9521d392016-06-25 21:53:30 +1000244static unsigned long *alloc_thread_stack_node(struct task_struct *tsk,
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000245 int node)
246{
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700247 return kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node);
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000248}
249
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700250static void free_thread_stack(struct task_struct *tsk)
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000251{
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700252 kmem_cache_free(thread_stack_cache, tsk->stack);
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000253}
254
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700255void thread_stack_cache_init(void)
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000256{
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700257 thread_stack_cache = kmem_cache_create("thread_stack", THREAD_SIZE,
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000258 THREAD_SIZE, 0, NULL);
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700259 BUG_ON(thread_stack_cache == NULL);
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000260}
261# endif
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700262#endif
263
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264/* SLAB cache for signal_struct structures (tsk->signal) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800265static struct kmem_cache *signal_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266
267/* SLAB cache for sighand_struct structures (tsk->sighand) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800268struct kmem_cache *sighand_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269
270/* SLAB cache for files_struct structures (tsk->files) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800271struct kmem_cache *files_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272
273/* SLAB cache for fs_struct structures (tsk->fs) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800274struct kmem_cache *fs_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275
276/* SLAB cache for vm_area_struct structures */
Christoph Lametere18b8902006-12-06 20:33:20 -0800277struct kmem_cache *vm_area_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278
279/* SLAB cache for mm_struct structures (tsk->mm) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800280static struct kmem_cache *mm_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700282static void account_kernel_stack(struct task_struct *tsk, int account)
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700283{
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700284 void *stack = task_stack_page(tsk);
285 struct vm_struct *vm = task_stack_vm_area(tsk);
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700286
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700287 BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0);
Andy Lutomirskiefdc9492016-07-28 15:48:17 -0700288
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700289 if (vm) {
290 int i;
291
292 BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE);
293
294 for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
295 mod_zone_page_state(page_zone(vm->pages[i]),
296 NR_KERNEL_STACK_KB,
297 PAGE_SIZE / 1024 * account);
298 }
299
300 /* All stack pages belong to the same memcg. */
301 memcg_kmem_update_page_stat(vm->pages[0], MEMCG_KERNEL_STACK_KB,
302 account * (THREAD_SIZE / 1024));
303 } else {
304 /*
305 * All stack pages are in the same zone and belong to the
306 * same memcg.
307 */
308 struct page *first_page = virt_to_page(stack);
309
310 mod_zone_page_state(page_zone(first_page), NR_KERNEL_STACK_KB,
311 THREAD_SIZE / 1024 * account);
312
313 memcg_kmem_update_page_stat(first_page, MEMCG_KERNEL_STACK_KB,
314 account * (THREAD_SIZE / 1024));
315 }
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700316}
317
Andy Lutomirski68f24b082016-09-15 22:45:48 -0700318static void release_task_stack(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319{
Andy Lutomirski405c0752016-10-31 08:11:43 -0700320 if (WARN_ON(tsk->state != TASK_DEAD))
321 return; /* Better to leak the stack than to free prematurely */
322
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700323 account_kernel_stack(tsk, -1);
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700324 arch_release_thread_stack(tsk->stack);
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700325 free_thread_stack(tsk);
Andy Lutomirski68f24b082016-09-15 22:45:48 -0700326 tsk->stack = NULL;
327#ifdef CONFIG_VMAP_STACK
328 tsk->stack_vm_area = NULL;
329#endif
330}
331
332#ifdef CONFIG_THREAD_INFO_IN_TASK
333void put_task_stack(struct task_struct *tsk)
334{
335 if (atomic_dec_and_test(&tsk->stack_refcount))
336 release_task_stack(tsk);
337}
338#endif
339
340void free_task(struct task_struct *tsk)
341{
342#ifndef CONFIG_THREAD_INFO_IN_TASK
343 /*
344 * The task is finally done with both the stack and thread_info,
345 * so free both.
346 */
347 release_task_stack(tsk);
348#else
349 /*
350 * If the task had a separate stack allocation, it should be gone
351 * by now.
352 */
353 WARN_ON_ONCE(atomic_read(&tsk->stack_refcount) != 0);
354#endif
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700355 rt_mutex_debug_task_free(tsk);
Frederic Weisbeckerfb526072008-11-25 21:07:04 +0100356 ftrace_graph_exit_task(tsk);
Will Drewrye2cfabdf2012-04-12 16:47:57 -0500357 put_seccomp_filter(tsk);
Akinobu Mitaf19b9f72012-07-30 14:42:33 -0700358 arch_release_task_struct(tsk);
Oleg Nesterov1da5c462016-11-29 18:50:57 +0100359 if (tsk->flags & PF_KTHREAD)
360 free_kthread_struct(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 free_task_struct(tsk);
362}
363EXPORT_SYMBOL(free_task);
364
Oleg Nesterovea6d2902010-05-26 14:43:16 -0700365static inline void free_signal_struct(struct signal_struct *sig)
366{
Oleg Nesterov97101eb42010-05-26 14:43:20 -0700367 taskstats_tgid_free(sig);
Mike Galbraith1c5354d2011-01-05 11:16:04 +0100368 sched_autogroup_exit(sig);
Michal Hocko7283094ec2016-10-07 16:58:54 -0700369 /*
370 * __mmdrop is not safe to call from softirq context on x86 due to
371 * pgd_dtor so postpone it to the async context
372 */
Michal Hocko26db62f2016-10-07 16:58:51 -0700373 if (sig->oom_mm)
Michal Hocko7283094ec2016-10-07 16:58:54 -0700374 mmdrop_async(sig->oom_mm);
Oleg Nesterovea6d2902010-05-26 14:43:16 -0700375 kmem_cache_free(signal_cachep, sig);
376}
377
378static inline void put_signal_struct(struct signal_struct *sig)
379{
Mike Galbraith1c5354d2011-01-05 11:16:04 +0100380 if (atomic_dec_and_test(&sig->sigcnt))
Oleg Nesterovea6d2902010-05-26 14:43:16 -0700381 free_signal_struct(sig);
382}
383
Andrew Morton158d9eb2006-03-31 02:31:34 -0800384void __put_task_struct(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385{
Eugene Teo270f7222007-10-18 23:40:38 -0700386 WARN_ON(!tsk->exit_state);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387 WARN_ON(atomic_read(&tsk->usage));
388 WARN_ON(tsk == current);
389
Tejun Heo2e91fa72015-10-15 16:41:53 -0400390 cgroup_free(tsk);
Mike Galbraith156654f2014-02-28 07:23:11 +0100391 task_numa_free(tsk);
Kees Cook1a2a4d02011-12-21 12:17:03 -0800392 security_task_free(tsk);
David Howellse0e81732009-09-02 09:13:40 +0100393 exit_creds(tsk);
Shailabh Nagar35df17c2006-08-31 21:27:38 -0700394 delayacct_tsk_free(tsk);
Oleg Nesterovea6d2902010-05-26 14:43:16 -0700395 put_signal_struct(tsk->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396
397 if (!profile_handoff_task(tsk))
398 free_task(tsk);
399}
Rik van Riel77c100c2011-02-01 09:51:46 -0500400EXPORT_SYMBOL_GPL(__put_task_struct);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401
Thomas Gleixner6c0a9fa2012-05-05 15:05:40 +0000402void __init __weak arch_task_cache_init(void) { }
Suresh Siddha61c46282008-03-10 15:28:04 -0700403
Heinrich Schuchardtff691f62015-04-16 12:47:44 -0700404/*
405 * set_max_threads
406 */
Heinrich Schuchardt16db3d32015-04-16 12:47:50 -0700407static void set_max_threads(unsigned int max_threads_suggested)
Heinrich Schuchardtff691f62015-04-16 12:47:44 -0700408{
Heinrich Schuchardtac1b3982015-04-16 12:47:47 -0700409 u64 threads;
Heinrich Schuchardtff691f62015-04-16 12:47:44 -0700410
411 /*
Heinrich Schuchardtac1b3982015-04-16 12:47:47 -0700412 * The number of threads shall be limited such that the thread
413 * structures may only consume a small part of the available memory.
Heinrich Schuchardtff691f62015-04-16 12:47:44 -0700414 */
Heinrich Schuchardtac1b3982015-04-16 12:47:47 -0700415 if (fls64(totalram_pages) + fls64(PAGE_SIZE) > 64)
416 threads = MAX_THREADS;
417 else
418 threads = div64_u64((u64) totalram_pages * (u64) PAGE_SIZE,
419 (u64) THREAD_SIZE * 8UL);
420
Heinrich Schuchardt16db3d32015-04-16 12:47:50 -0700421 if (threads > max_threads_suggested)
422 threads = max_threads_suggested;
423
Heinrich Schuchardtac1b3982015-04-16 12:47:47 -0700424 max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS);
Heinrich Schuchardtff691f62015-04-16 12:47:44 -0700425}
426
Ingo Molnar5aaeb5c2015-07-17 12:28:12 +0200427#ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
428/* Initialized by the architecture: */
429int arch_task_struct_size __read_mostly;
430#endif
Dave Hansen0c8c0f02015-07-17 12:28:11 +0200431
Heinrich Schuchardtff691f62015-04-16 12:47:44 -0700432void __init fork_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433{
Eric W. Biederman25f9c082016-08-08 14:41:52 -0500434 int i;
Thomas Gleixnerf5e10282012-05-05 15:05:48 +0000435#ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436#ifndef ARCH_MIN_TASKALIGN
Peter Zijlstrae2747952017-01-11 14:17:48 +0100437#define ARCH_MIN_TASKALIGN 0
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438#endif
Peter Zijlstra95cb64c2017-02-18 15:26:45 +0100439 int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN);
Peter Zijlstrae2747952017-01-11 14:17:48 +0100440
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 /* create a slab on which task_structs can be allocated */
Vladimir Davydov5d097052016-01-14 15:18:21 -0800442 task_struct_cachep = kmem_cache_create("task_struct",
Peter Zijlstrae2747952017-01-11 14:17:48 +0100443 arch_task_struct_size, align,
Vladimir Davydov5d097052016-01-14 15:18:21 -0800444 SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445#endif
446
Suresh Siddha61c46282008-03-10 15:28:04 -0700447 /* do the arch specific task caches init */
448 arch_task_cache_init();
449
Heinrich Schuchardt16db3d32015-04-16 12:47:50 -0700450 set_max_threads(MAX_THREADS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451
452 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
453 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
454 init_task.signal->rlim[RLIMIT_SIGPENDING] =
455 init_task.signal->rlim[RLIMIT_NPROC];
Eric W. Biedermanb376c3e2016-08-08 13:41:24 -0500456
Eric W. Biederman25f9c082016-08-08 14:41:52 -0500457 for (i = 0; i < UCOUNT_COUNTS; i++) {
458 init_user_ns.ucount_max[i] = max_threads/2;
459 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460}
461
Gideon Israel Dsouza52f5684c2014-04-07 15:39:20 -0700462int __weak arch_dup_task_struct(struct task_struct *dst,
Suresh Siddha61c46282008-03-10 15:28:04 -0700463 struct task_struct *src)
464{
465 *dst = *src;
466 return 0;
467}
468
Aaron Tomlind4311ff2014-09-12 14:16:17 +0100469void set_task_stack_end_magic(struct task_struct *tsk)
470{
471 unsigned long *stackend;
472
473 stackend = end_of_stack(tsk);
474 *stackend = STACK_END_MAGIC; /* for overflow detection */
475}
476
Andi Kleen725fc622016-05-23 16:24:05 -0700477static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478{
479 struct task_struct *tsk;
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700480 unsigned long *stack;
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700481 struct vm_struct *stack_vm_area;
Peter Zijlstra3e26c142007-10-16 23:25:50 -0700482 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483
Andi Kleen725fc622016-05-23 16:24:05 -0700484 if (node == NUMA_NO_NODE)
485 node = tsk_fork_get_node(orig);
Eric Dumazet504f52b2011-03-22 16:30:41 -0700486 tsk = alloc_task_struct_node(node);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 if (!tsk)
488 return NULL;
489
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700490 stack = alloc_thread_stack_node(tsk, node);
491 if (!stack)
Akinobu Mitaf19b9f72012-07-30 14:42:33 -0700492 goto free_tsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700494 stack_vm_area = task_stack_vm_area(tsk);
495
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700496 err = arch_dup_task_struct(tsk, orig);
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700497
498 /*
499 * arch_dup_task_struct() clobbers the stack-related fields. Make
500 * sure they're properly initialized before using any stack-related
501 * functions again.
502 */
503 tsk->stack = stack;
504#ifdef CONFIG_VMAP_STACK
505 tsk->stack_vm_area = stack_vm_area;
506#endif
Andy Lutomirski68f24b082016-09-15 22:45:48 -0700507#ifdef CONFIG_THREAD_INFO_IN_TASK
508 atomic_set(&tsk->stack_refcount, 1);
509#endif
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700510
Suresh Siddha61c46282008-03-10 15:28:04 -0700511 if (err)
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700512 goto free_stack;
Suresh Siddha61c46282008-03-10 15:28:04 -0700513
Kees Cookdbd952122014-06-27 15:18:48 -0700514#ifdef CONFIG_SECCOMP
515 /*
516 * We must handle setting up seccomp filters once we're under
517 * the sighand lock in case orig has changed between now and
518 * then. Until then, filter must be NULL to avoid messing up
519 * the usage counts on the error path calling free_task.
520 */
521 tsk->seccomp.filter = NULL;
522#endif
Andrew Morton87bec582012-07-30 14:42:31 -0700523
524 setup_thread_stack(tsk, orig);
Avi Kivity8e7cac72009-11-29 16:34:48 +0200525 clear_user_return_notifier(tsk);
Mike Galbraithf26f9af2010-12-08 11:05:42 +0100526 clear_tsk_need_resched(tsk);
Aaron Tomlind4311ff2014-09-12 14:16:17 +0100527 set_task_stack_end_magic(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528
Arjan van de Ven0a4254052006-09-26 10:52:38 +0200529#ifdef CONFIG_CC_STACKPROTECTOR
530 tsk->stack_canary = get_random_int();
531#endif
532
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700533 /*
534 * One for us, one for whoever does the "release_task()" (usually
535 * parent)
536 */
537 atomic_set(&tsk->usage, 2);
Alexey Dobriyan6c5c9342006-09-29 01:59:40 -0700538#ifdef CONFIG_BLK_DEV_IO_TRACE
Jens Axboe2056a782006-03-23 20:00:26 +0100539 tsk->btrace_seq = 0;
Alexey Dobriyan6c5c9342006-09-29 01:59:40 -0700540#endif
Jens Axboea0aa7f62006-04-20 13:05:33 +0200541 tsk->splice_pipe = NULL;
Eric Dumazet5640f762012-09-23 23:04:42 +0000542 tsk->task_frag.page = NULL;
Sebastian Andrzej Siewior093e5842015-12-21 18:17:10 +0100543 tsk->wake_q.next = NULL;
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700544
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700545 account_kernel_stack(tsk, 1);
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700546
Dmitry Vyukov5c9a8752016-03-22 14:27:30 -0700547 kcov_task_init(tsk);
548
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 return tsk;
Suresh Siddha61c46282008-03-10 15:28:04 -0700550
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700551free_stack:
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700552 free_thread_stack(tsk);
Akinobu Mitaf19b9f72012-07-30 14:42:33 -0700553free_tsk:
Suresh Siddha61c46282008-03-10 15:28:04 -0700554 free_task_struct(tsk);
555 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556}
557
558#ifdef CONFIG_MMU
Emese Revfy0766f782016-06-20 20:42:34 +0200559static __latent_entropy int dup_mmap(struct mm_struct *mm,
560 struct mm_struct *oldmm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561{
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700562 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 struct rb_node **rb_link, *rb_parent;
564 int retval;
565 unsigned long charge;
Pavel Emelyanov893e26e2017-02-22 15:42:27 -0800566 LIST_HEAD(uf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567
Oleg Nesterov32cdba12012-11-14 19:03:42 +0100568 uprobe_start_dup_mmap();
Michal Hocko7c051262016-05-23 16:25:48 -0700569 if (down_write_killable(&oldmm->mmap_sem)) {
570 retval = -EINTR;
571 goto fail_uprobe_end;
572 }
Ralf Baechleec8c0442006-12-12 17:14:57 +0000573 flush_cache_dup_mm(oldmm);
Oleg Nesterovf8ac4ec2012-08-08 17:11:42 +0200574 uprobe_dup_mmap(oldmm, mm);
Ingo Molnarad339452006-07-03 00:25:15 -0700575 /*
576 * Not linked in yet - no deadlock potential:
577 */
578 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
Hugh Dickins7ee78232005-10-29 18:16:08 -0700579
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -0700580 /* No ordering required: file already has been exposed. */
581 RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
582
Vladimir Davydov4f7d4612014-08-08 14:22:01 -0700583 mm->total_vm = oldmm->total_vm;
Konstantin Khlebnikov84638332016-01-14 15:22:07 -0800584 mm->data_vm = oldmm->data_vm;
Vladimir Davydov4f7d4612014-08-08 14:22:01 -0700585 mm->exec_vm = oldmm->exec_vm;
586 mm->stack_vm = oldmm->stack_vm;
587
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 rb_link = &mm->mm_rb.rb_node;
589 rb_parent = NULL;
590 pprev = &mm->mmap;
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700591 retval = ksm_fork(mm, oldmm);
592 if (retval)
593 goto out;
Andrea Arcangeliba761492011-01-13 15:46:58 -0800594 retval = khugepaged_fork(mm, oldmm);
595 if (retval)
596 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700598 prev = NULL;
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700599 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 struct file *file;
601
602 if (mpnt->vm_flags & VM_DONTCOPY) {
Konstantin Khlebnikov84638332016-01-14 15:22:07 -0800603 vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 continue;
605 }
606 charge = 0;
607 if (mpnt->vm_flags & VM_ACCOUNT) {
Huang Shijieb2412b72012-07-30 14:42:30 -0700608 unsigned long len = vma_pages(mpnt);
609
Al Viro191c5422012-02-13 03:58:52 +0000610 if (security_vm_enough_memory_mm(oldmm, len)) /* sic */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 goto fail_nomem;
612 charge = len;
613 }
Christoph Lametere94b1762006-12-06 20:33:17 -0800614 tmp = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 if (!tmp)
616 goto fail_nomem;
617 *tmp = *mpnt;
Rik van Riel5beb4932010-03-05 13:42:07 -0800618 INIT_LIST_HEAD(&tmp->anon_vma_chain);
Oleg Nesterovef0855d2013-09-11 14:20:14 -0700619 retval = vma_dup_policy(mpnt, tmp);
620 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 goto fail_nomem_policy;
Andrea Arcangelia247c3a2010-09-22 13:05:12 -0700622 tmp->vm_mm = mm;
Pavel Emelyanov893e26e2017-02-22 15:42:27 -0800623 retval = dup_userfaultfd(tmp, &uf);
624 if (retval)
625 goto fail_nomem_anon_vma_fork;
Rik van Riel5beb4932010-03-05 13:42:07 -0800626 if (anon_vma_fork(tmp, mpnt))
627 goto fail_nomem_anon_vma_fork;
Pavel Emelyanov893e26e2017-02-22 15:42:27 -0800628 tmp->vm_flags &= ~(VM_LOCKED | VM_LOCKONFAULT);
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700629 tmp->vm_next = tmp->vm_prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 file = tmp->vm_file;
631 if (file) {
Al Viro496ad9a2013-01-23 17:07:38 -0500632 struct inode *inode = file_inode(file);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000633 struct address_space *mapping = file->f_mapping;
634
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 get_file(file);
636 if (tmp->vm_flags & VM_DENYWRITE)
637 atomic_dec(&inode->i_writecount);
Davidlohr Bueso83cde9e2014-12-12 16:54:21 -0800638 i_mmap_lock_write(mapping);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000639 if (tmp->vm_flags & VM_SHARED)
David Herrmann4bb5f5d2014-08-08 14:25:25 -0700640 atomic_inc(&mapping->i_mmap_writable);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000641 flush_dcache_mmap_lock(mapping);
642 /* insert tmp into the share list, just after mpnt */
Kirill A. Shutemov27ba0642015-02-10 14:09:59 -0800643 vma_interval_tree_insert_after(tmp, mpnt,
644 &mapping->i_mmap);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000645 flush_dcache_mmap_unlock(mapping);
Davidlohr Bueso83cde9e2014-12-12 16:54:21 -0800646 i_mmap_unlock_write(mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 }
648
649 /*
Mel Gormana1e78772008-07-23 21:27:23 -0700650 * Clear hugetlb-related page reserves for children. This only
651 * affects MAP_PRIVATE mappings. Faults generated by the child
652 * are not guaranteed to succeed, even if read-only
653 */
654 if (is_vm_hugetlb_page(tmp))
655 reset_vma_resv_huge_pages(tmp);
656
657 /*
Hugh Dickins7ee78232005-10-29 18:16:08 -0700658 * Link in the new vma and copy the page table entries.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 *pprev = tmp;
661 pprev = &tmp->vm_next;
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700662 tmp->vm_prev = prev;
663 prev = tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664
665 __vma_link_rb(mm, tmp, rb_link, rb_parent);
666 rb_link = &tmp->vm_rb.rb_right;
667 rb_parent = &tmp->vm_rb;
668
669 mm->map_count++;
Hugh Dickins0b0db142005-11-21 21:32:20 -0800670 retval = copy_page_range(mm, oldmm, mpnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
672 if (tmp->vm_ops && tmp->vm_ops->open)
673 tmp->vm_ops->open(tmp);
674
675 if (retval)
676 goto out;
677 }
Jeremy Fitzhardinged6dd61c2007-05-02 19:27:14 +0200678 /* a new mm has just been created */
679 arch_dup_mmap(oldmm, mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 retval = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681out:
Hugh Dickins7ee78232005-10-29 18:16:08 -0700682 up_write(&mm->mmap_sem);
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700683 flush_tlb_mm(oldmm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 up_write(&oldmm->mmap_sem);
Pavel Emelyanov893e26e2017-02-22 15:42:27 -0800685 dup_userfaultfd_complete(&uf);
Michal Hocko7c051262016-05-23 16:25:48 -0700686fail_uprobe_end:
Oleg Nesterov32cdba12012-11-14 19:03:42 +0100687 uprobe_end_dup_mmap();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 return retval;
Rik van Riel5beb4932010-03-05 13:42:07 -0800689fail_nomem_anon_vma_fork:
Oleg Nesterovef0855d2013-09-11 14:20:14 -0700690 mpol_put(vma_policy(tmp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691fail_nomem_policy:
692 kmem_cache_free(vm_area_cachep, tmp);
693fail_nomem:
694 retval = -ENOMEM;
695 vm_unacct_memory(charge);
696 goto out;
697}
698
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700699static inline int mm_alloc_pgd(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700{
701 mm->pgd = pgd_alloc(mm);
702 if (unlikely(!mm->pgd))
703 return -ENOMEM;
704 return 0;
705}
706
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700707static inline void mm_free_pgd(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708{
Benjamin Herrenschmidt5e541972008-02-04 22:29:14 -0800709 pgd_free(mm, mm->pgd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710}
711#else
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -0700712static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
713{
714 down_write(&oldmm->mmap_sem);
715 RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
716 up_write(&oldmm->mmap_sem);
717 return 0;
718}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719#define mm_alloc_pgd(mm) (0)
720#define mm_free_pgd(mm)
721#endif /* CONFIG_MMU */
722
Daniel Walker23ff4442007-10-18 03:06:07 -0700723__cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724
Christoph Lametere94b1762006-12-06 20:33:17 -0800725#define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726#define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
727
Hidehiro Kawai4cb0e112009-01-06 14:42:47 -0800728static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
729
730static int __init coredump_filter_setup(char *s)
731{
732 default_dump_filter =
733 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
734 MMF_DUMP_FILTER_MASK;
735 return 1;
736}
737
738__setup("coredump_filter=", coredump_filter_setup);
739
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740#include <linux/init_task.h>
741
Alexey Dobriyan858f0992009-09-23 15:57:32 -0700742static void mm_init_aio(struct mm_struct *mm)
743{
744#ifdef CONFIG_AIO
745 spin_lock_init(&mm->ioctx_lock);
Benjamin LaHaisedb446a02013-07-30 12:54:40 -0400746 mm->ioctx_table = NULL;
Alexey Dobriyan858f0992009-09-23 15:57:32 -0700747#endif
748}
749
Vladimir Davydov33144e82014-08-08 14:22:03 -0700750static void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
751{
752#ifdef CONFIG_MEMCG
753 mm->owner = p;
754#endif
755}
756
Eric W. Biedermanbfedb582016-10-13 21:23:16 -0500757static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
758 struct user_namespace *user_ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759{
Vladimir Davydov41f727f2014-08-08 14:21:56 -0700760 mm->mmap = NULL;
761 mm->mm_rb = RB_ROOT;
762 mm->vmacache_seqnum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763 atomic_set(&mm->mm_users, 1);
764 atomic_set(&mm->mm_count, 1);
765 init_rwsem(&mm->mmap_sem);
766 INIT_LIST_HEAD(&mm->mmlist);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700767 mm->core_state = NULL;
Kirill A. Shutemove1f56c82013-11-14 14:30:48 -0800768 atomic_long_set(&mm->nr_ptes, 0);
Kirill A. Shutemov2d2f5112015-02-12 14:59:59 -0800769 mm_nr_pmds_init(mm);
Vladimir Davydov41f727f2014-08-08 14:21:56 -0700770 mm->map_count = 0;
771 mm->locked_vm = 0;
Vladimir Davydovce65cef2014-08-08 14:21:58 -0700772 mm->pinned_vm = 0;
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -0800773 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 spin_lock_init(&mm->page_table_lock);
Vladimir Davydov41f727f2014-08-08 14:21:56 -0700775 mm_init_cpumask(mm);
Alexey Dobriyan858f0992009-09-23 15:57:32 -0700776 mm_init_aio(mm);
Balbir Singhcf475ad22008-04-29 01:00:16 -0700777 mm_init_owner(mm, p);
Vladimir Davydov41f727f2014-08-08 14:21:56 -0700778 mmu_notifier_mm_init(mm);
Rik van Riel20841402013-12-18 17:08:44 -0800779 clear_tlb_flush_pending(mm);
Vladimir Davydov41f727f2014-08-08 14:21:56 -0700780#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
781 mm->pmd_huge_pte = NULL;
782#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783
Alex Thorltona0715cc2014-04-07 15:37:10 -0700784 if (current->mm) {
785 mm->flags = current->mm->flags & MMF_INIT_MASK;
786 mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK;
787 } else {
788 mm->flags = default_dump_filter;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 mm->def_flags = 0;
Alex Thorltona0715cc2014-04-07 15:37:10 -0700790 }
791
Vladimir Davydov41f727f2014-08-08 14:21:56 -0700792 if (mm_alloc_pgd(mm))
793 goto fail_nopgd;
Pavel Emelianov78fb7462008-02-07 00:13:51 -0800794
Vladimir Davydov41f727f2014-08-08 14:21:56 -0700795 if (init_new_context(p, mm))
796 goto fail_nocontext;
797
Eric W. Biedermanbfedb582016-10-13 21:23:16 -0500798 mm->user_ns = get_user_ns(user_ns);
Vladimir Davydov41f727f2014-08-08 14:21:56 -0700799 return mm;
800
801fail_nocontext:
802 mm_free_pgd(mm);
803fail_nopgd:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 free_mm(mm);
805 return NULL;
806}
807
Konstantin Khlebnikovc3f03272012-03-21 16:33:48 -0700808static void check_mm(struct mm_struct *mm)
809{
810 int i;
811
812 for (i = 0; i < NR_MM_COUNTERS; i++) {
813 long x = atomic_long_read(&mm->rss_stat.count[i]);
814
815 if (unlikely(x))
816 printk(KERN_ALERT "BUG: Bad rss-counter state "
817 "mm:%p idx:%d val:%ld\n", mm, i, x);
818 }
Kirill A. Shutemovb30fe6c2015-02-11 15:26:53 -0800819
820 if (atomic_long_read(&mm->nr_ptes))
821 pr_alert("BUG: non-zero nr_ptes on freeing mm: %ld\n",
822 atomic_long_read(&mm->nr_ptes));
823 if (mm_nr_pmds(mm))
824 pr_alert("BUG: non-zero nr_pmds on freeing mm: %ld\n",
825 mm_nr_pmds(mm));
826
Kirill A. Shutemove009bb32013-11-14 14:31:07 -0800827#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
Sasha Levin96dad672014-10-09 15:28:39 -0700828 VM_BUG_ON_MM(mm->pmd_huge_pte, mm);
Konstantin Khlebnikovc3f03272012-03-21 16:33:48 -0700829#endif
830}
831
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832/*
833 * Allocate and initialize an mm_struct.
834 */
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700835struct mm_struct *mm_alloc(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836{
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700837 struct mm_struct *mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838
839 mm = allocate_mm();
KOSAKI Motohirode03c722011-05-24 17:12:15 -0700840 if (!mm)
841 return NULL;
842
843 memset(mm, 0, sizeof(*mm));
Eric W. Biedermanbfedb582016-10-13 21:23:16 -0500844 return mm_init(mm, current, current_user_ns());
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845}
846
847/*
848 * Called when the last reference to the mm
849 * is dropped: either by a lazy thread or by
850 * mmput. Free the page directory and the mm.
851 */
Harvey Harrison7ad5b3a2008-02-08 04:19:53 -0800852void __mmdrop(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853{
854 BUG_ON(mm == &init_mm);
855 mm_free_pgd(mm);
856 destroy_context(mm);
Andrea Arcangelicddb8a52008-07-28 15:46:29 -0700857 mmu_notifier_mm_destroy(mm);
Konstantin Khlebnikovc3f03272012-03-21 16:33:48 -0700858 check_mm(mm);
Eric W. Biedermanbfedb582016-10-13 21:23:16 -0500859 put_user_ns(mm->user_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 free_mm(mm);
861}
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200862EXPORT_SYMBOL_GPL(__mmdrop);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863
Michal Hockoec8d7c12016-05-20 16:57:21 -0700864static inline void __mmput(struct mm_struct *mm)
865{
866 VM_BUG_ON(atomic_read(&mm->mm_users));
867
868 uprobe_clear_state(mm);
869 exit_aio(mm);
870 ksm_exit(mm);
871 khugepaged_exit(mm); /* must run before exit_mmap */
872 exit_mmap(mm);
Aaron Lu6fcb52a2016-10-07 17:00:08 -0700873 mm_put_huge_zero_page(mm);
Michal Hockoec8d7c12016-05-20 16:57:21 -0700874 set_mm_exe_file(mm, NULL);
875 if (!list_empty(&mm->mmlist)) {
876 spin_lock(&mmlist_lock);
877 list_del(&mm->mmlist);
878 spin_unlock(&mmlist_lock);
879 }
880 if (mm->binfmt)
881 module_put(mm->binfmt->module);
Michal Hocko862e3072016-10-07 16:58:57 -0700882 set_bit(MMF_OOM_SKIP, &mm->flags);
Michal Hockoec8d7c12016-05-20 16:57:21 -0700883 mmdrop(mm);
884}
885
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886/*
887 * Decrement the use count and release all resources for an mm.
888 */
889void mmput(struct mm_struct *mm)
890{
Andrew Morton0ae26f12006-06-23 02:05:15 -0700891 might_sleep();
892
Michal Hockoec8d7c12016-05-20 16:57:21 -0700893 if (atomic_dec_and_test(&mm->mm_users))
894 __mmput(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895}
896EXPORT_SYMBOL_GPL(mmput);
897
Michal Hocko7ef949d72016-05-26 15:16:22 -0700898#ifdef CONFIG_MMU
Michal Hockoec8d7c12016-05-20 16:57:21 -0700899static void mmput_async_fn(struct work_struct *work)
900{
901 struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
902 __mmput(mm);
903}
904
905void mmput_async(struct mm_struct *mm)
906{
907 if (atomic_dec_and_test(&mm->mm_users)) {
908 INIT_WORK(&mm->async_put_work, mmput_async_fn);
909 schedule_work(&mm->async_put_work);
910 }
911}
Michal Hocko7ef949d72016-05-26 15:16:22 -0700912#endif
Michal Hockoec8d7c12016-05-20 16:57:21 -0700913
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -0700914/**
915 * set_mm_exe_file - change a reference to the mm's executable file
916 *
917 * This changes mm's executable file (shown as symlink /proc/[pid]/exe).
918 *
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -0700919 * Main users are mmput() and sys_execve(). Callers prevent concurrent
920 * invocations: in mmput() nobody alive left, in execve task is single
921 * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the
922 * mm->exe_file, but does so without using set_mm_exe_file() in order
923 * to do avoid the need for any locks.
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -0700924 */
Jiri Slaby38646012011-05-26 16:25:46 -0700925void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
926{
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -0700927 struct file *old_exe_file;
928
929 /*
930 * It is safe to dereference the exe_file without RCU as
931 * this function is only called if nobody else can access
932 * this mm -- see comment above for justification.
933 */
934 old_exe_file = rcu_dereference_raw(mm->exe_file);
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -0700935
Jiri Slaby38646012011-05-26 16:25:46 -0700936 if (new_exe_file)
937 get_file(new_exe_file);
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -0700938 rcu_assign_pointer(mm->exe_file, new_exe_file);
939 if (old_exe_file)
940 fput(old_exe_file);
Jiri Slaby38646012011-05-26 16:25:46 -0700941}
942
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -0700943/**
944 * get_mm_exe_file - acquire a reference to the mm's executable file
945 *
946 * Returns %NULL if mm has no associated executable file.
947 * User must release file via fput().
948 */
Jiri Slaby38646012011-05-26 16:25:46 -0700949struct file *get_mm_exe_file(struct mm_struct *mm)
950{
951 struct file *exe_file;
952
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -0700953 rcu_read_lock();
954 exe_file = rcu_dereference(mm->exe_file);
955 if (exe_file && !get_file_rcu(exe_file))
956 exe_file = NULL;
957 rcu_read_unlock();
Jiri Slaby38646012011-05-26 16:25:46 -0700958 return exe_file;
959}
Davidlohr Bueso11163342015-04-16 12:49:12 -0700960EXPORT_SYMBOL(get_mm_exe_file);
Jiri Slaby38646012011-05-26 16:25:46 -0700961
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962/**
Mateusz Guzikcd81a9172016-08-23 16:20:38 +0200963 * get_task_exe_file - acquire a reference to the task's executable file
964 *
965 * Returns %NULL if task's mm (if any) has no associated executable file or
966 * this is a kernel thread with borrowed mm (see the comment above get_task_mm).
967 * User must release file via fput().
968 */
969struct file *get_task_exe_file(struct task_struct *task)
970{
971 struct file *exe_file = NULL;
972 struct mm_struct *mm;
973
974 task_lock(task);
975 mm = task->mm;
976 if (mm) {
977 if (!(task->flags & PF_KTHREAD))
978 exe_file = get_mm_exe_file(mm);
979 }
980 task_unlock(task);
981 return exe_file;
982}
983EXPORT_SYMBOL(get_task_exe_file);
984
985/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 * get_task_mm - acquire a reference to the task's mm
987 *
Oleg Nesterov246bb0b2008-07-25 01:47:38 -0700988 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 * this kernel workthread has transiently adopted a user mm with use_mm,
990 * to do its AIO) is not set and if so returns a reference to it, after
991 * bumping up the use count. User must release the mm via mmput()
992 * after use. Typically used by /proc and ptrace.
993 */
994struct mm_struct *get_task_mm(struct task_struct *task)
995{
996 struct mm_struct *mm;
997
998 task_lock(task);
999 mm = task->mm;
1000 if (mm) {
Oleg Nesterov246bb0b2008-07-25 01:47:38 -07001001 if (task->flags & PF_KTHREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 mm = NULL;
1003 else
Vegard Nossum3fce3712017-02-27 14:30:10 -08001004 mmget(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 }
1006 task_unlock(task);
1007 return mm;
1008}
1009EXPORT_SYMBOL_GPL(get_task_mm);
1010
Christopher Yeoh8cdb8782012-02-02 11:34:09 +10301011struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
1012{
1013 struct mm_struct *mm;
1014 int err;
1015
1016 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
1017 if (err)
1018 return ERR_PTR(err);
1019
1020 mm = get_task_mm(task);
1021 if (mm && mm != current->mm &&
1022 !ptrace_may_access(task, mode)) {
1023 mmput(mm);
1024 mm = ERR_PTR(-EACCES);
1025 }
1026 mutex_unlock(&task->signal->cred_guard_mutex);
1027
1028 return mm;
1029}
1030
Oleg Nesterov57b59c42012-03-05 14:59:13 -08001031static void complete_vfork_done(struct task_struct *tsk)
Oleg Nesterovc415c3b2012-03-05 14:59:13 -08001032{
Oleg Nesterovd68b46f2012-03-05 14:59:13 -08001033 struct completion *vfork;
Oleg Nesterovc415c3b2012-03-05 14:59:13 -08001034
Oleg Nesterovd68b46f2012-03-05 14:59:13 -08001035 task_lock(tsk);
1036 vfork = tsk->vfork_done;
1037 if (likely(vfork)) {
1038 tsk->vfork_done = NULL;
1039 complete(vfork);
1040 }
1041 task_unlock(tsk);
1042}
1043
1044static int wait_for_vfork_done(struct task_struct *child,
1045 struct completion *vfork)
1046{
1047 int killed;
1048
1049 freezer_do_not_count();
1050 killed = wait_for_completion_killable(vfork);
1051 freezer_count();
1052
1053 if (killed) {
1054 task_lock(child);
1055 child->vfork_done = NULL;
1056 task_unlock(child);
1057 }
1058
1059 put_task_struct(child);
1060 return killed;
Oleg Nesterovc415c3b2012-03-05 14:59:13 -08001061}
1062
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063/* Please note the differences between mmput and mm_release.
1064 * mmput is called whenever we stop holding onto a mm_struct,
1065 * error success whatever.
1066 *
1067 * mm_release is called after a mm_struct has been removed
1068 * from the current process.
1069 *
1070 * This difference is important for error handling, when we
1071 * only half set up a mm_struct for a new process and need to restore
1072 * the old one. Because we mmput the new mm_struct before
1073 * restoring the old one. . .
1074 * Eric Biederman 10 January 1998
1075 */
1076void mm_release(struct task_struct *tsk, struct mm_struct *mm)
1077{
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001078 /* Get rid of any futexes when releasing the mm */
1079#ifdef CONFIG_FUTEX
Peter Zijlstrafc6b1772009-10-05 18:17:32 +02001080 if (unlikely(tsk->robust_list)) {
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001081 exit_robust_list(tsk);
Peter Zijlstrafc6b1772009-10-05 18:17:32 +02001082 tsk->robust_list = NULL;
1083 }
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001084#ifdef CONFIG_COMPAT
Peter Zijlstrafc6b1772009-10-05 18:17:32 +02001085 if (unlikely(tsk->compat_robust_list)) {
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001086 compat_exit_robust_list(tsk);
Peter Zijlstrafc6b1772009-10-05 18:17:32 +02001087 tsk->compat_robust_list = NULL;
1088 }
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001089#endif
Thomas Gleixner322a2c12009-10-05 18:18:03 +02001090 if (unlikely(!list_empty(&tsk->pi_state_list)))
1091 exit_pi_state_list(tsk);
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001092#endif
1093
Srikar Dronamraju0326f5a2012-03-13 23:30:11 +05301094 uprobe_free_utask(tsk);
1095
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 /* Get rid of any cached register state */
1097 deactivate_mm(tsk, mm);
1098
Roland McGrathfec1d012006-12-06 20:36:34 -08001099 /*
Michal Hocko735f2772016-09-01 16:15:13 -07001100 * Signal userspace if we're not exiting with a core dump
1101 * because we want to leave the value intact for debugging
1102 * purposes.
Roland McGrathfec1d012006-12-06 20:36:34 -08001103 */
Eric Dumazet9c8a8222009-08-06 15:09:28 -07001104 if (tsk->clear_child_tid) {
Michal Hocko735f2772016-09-01 16:15:13 -07001105 if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) &&
Eric Dumazet9c8a8222009-08-06 15:09:28 -07001106 atomic_read(&mm->mm_users) > 1) {
1107 /*
1108 * We don't check the error code - if userspace has
1109 * not set up a proper pointer then tough luck.
1110 */
1111 put_user(0, tsk->clear_child_tid);
1112 sys_futex(tsk->clear_child_tid, FUTEX_WAKE,
1113 1, NULL, NULL, 0);
1114 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 tsk->clear_child_tid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 }
Konstantin Khlebnikovf7505d642012-05-31 16:26:21 -07001117
1118 /*
1119 * All done, finally we can wake up parent and return this mm to him.
1120 * Also kthread_stop() uses this completion for synchronization.
1121 */
1122 if (tsk->vfork_done)
1123 complete_vfork_done(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124}
1125
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001126/*
1127 * Allocate a new mm structure and copy contents from the
1128 * mm structure of the passed in task structure.
1129 */
DaeSeok Younff252c12014-01-23 15:55:46 -08001130static struct mm_struct *dup_mm(struct task_struct *tsk)
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001131{
1132 struct mm_struct *mm, *oldmm = current->mm;
1133 int err;
1134
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001135 mm = allocate_mm();
1136 if (!mm)
1137 goto fail_nomem;
1138
1139 memcpy(mm, oldmm, sizeof(*mm));
1140
Eric W. Biedermanbfedb582016-10-13 21:23:16 -05001141 if (!mm_init(mm, tsk, mm->user_ns))
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001142 goto fail_nomem;
1143
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001144 err = dup_mmap(mm, oldmm);
1145 if (err)
1146 goto free_pt;
1147
1148 mm->hiwater_rss = get_mm_rss(mm);
1149 mm->hiwater_vm = mm->total_vm;
1150
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001151 if (mm->binfmt && !try_module_get(mm->binfmt->module))
1152 goto free_pt;
1153
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001154 return mm;
1155
1156free_pt:
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001157 /* don't put binfmt in mmput, we haven't got module yet */
1158 mm->binfmt = NULL;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001159 mmput(mm);
1160
1161fail_nomem:
1162 return NULL;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001163}
1164
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001165static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166{
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001167 struct mm_struct *mm, *oldmm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 int retval;
1169
1170 tsk->min_flt = tsk->maj_flt = 0;
1171 tsk->nvcsw = tsk->nivcsw = 0;
Mandeep Singh Baines17406b82009-02-06 15:37:47 -08001172#ifdef CONFIG_DETECT_HUNG_TASK
1173 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
1174#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175
1176 tsk->mm = NULL;
1177 tsk->active_mm = NULL;
1178
1179 /*
1180 * Are we cloning a kernel thread?
1181 *
1182 * We need to steal a active VM for that..
1183 */
1184 oldmm = current->mm;
1185 if (!oldmm)
1186 return 0;
1187
Davidlohr Bueso615d6e82014-04-07 15:37:25 -07001188 /* initialize the new vmacache entries */
1189 vmacache_flush(tsk);
1190
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191 if (clone_flags & CLONE_VM) {
Vegard Nossum3fce3712017-02-27 14:30:10 -08001192 mmget(oldmm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193 mm = oldmm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 goto good_mm;
1195 }
1196
1197 retval = -ENOMEM;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001198 mm = dup_mm(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 if (!mm)
1200 goto fail_nomem;
1201
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202good_mm:
1203 tsk->mm = mm;
1204 tsk->active_mm = mm;
1205 return 0;
1206
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207fail_nomem:
1208 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209}
1210
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001211static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212{
Al Viro498052b2009-03-30 07:20:30 -04001213 struct fs_struct *fs = current->fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 if (clone_flags & CLONE_FS) {
Al Viro498052b2009-03-30 07:20:30 -04001215 /* tsk->fs is already what we want */
Nick Piggin2a4419b2010-08-18 04:37:33 +10001216 spin_lock(&fs->lock);
Al Viro498052b2009-03-30 07:20:30 -04001217 if (fs->in_exec) {
Nick Piggin2a4419b2010-08-18 04:37:33 +10001218 spin_unlock(&fs->lock);
Al Viro498052b2009-03-30 07:20:30 -04001219 return -EAGAIN;
1220 }
1221 fs->users++;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001222 spin_unlock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223 return 0;
1224 }
Al Viro498052b2009-03-30 07:20:30 -04001225 tsk->fs = copy_fs_struct(fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226 if (!tsk->fs)
1227 return -ENOMEM;
1228 return 0;
1229}
1230
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001231static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
JANAK DESAIa016f332006-02-07 12:59:02 -08001232{
1233 struct files_struct *oldf, *newf;
1234 int error = 0;
1235
1236 /*
1237 * A background process may not have any files ...
1238 */
1239 oldf = current->files;
1240 if (!oldf)
1241 goto out;
1242
1243 if (clone_flags & CLONE_FILES) {
1244 atomic_inc(&oldf->count);
1245 goto out;
1246 }
1247
JANAK DESAIa016f332006-02-07 12:59:02 -08001248 newf = dup_fd(oldf, &error);
1249 if (!newf)
1250 goto out;
1251
1252 tsk->files = newf;
1253 error = 0;
1254out:
1255 return error;
1256}
1257
Jens Axboefadad8782008-01-24 08:54:47 +01001258static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
Jens Axboefd0928d2008-01-24 08:52:45 +01001259{
1260#ifdef CONFIG_BLOCK
1261 struct io_context *ioc = current->io_context;
Tejun Heo6e736be2011-12-14 00:33:38 +01001262 struct io_context *new_ioc;
Jens Axboefd0928d2008-01-24 08:52:45 +01001263
1264 if (!ioc)
1265 return 0;
Jens Axboefadad8782008-01-24 08:54:47 +01001266 /*
1267 * Share io context with parent, if CLONE_IO is set
1268 */
1269 if (clone_flags & CLONE_IO) {
Tejun Heo3d487492012-03-05 13:15:25 -08001270 ioc_task_link(ioc);
1271 tsk->io_context = ioc;
Jens Axboefadad8782008-01-24 08:54:47 +01001272 } else if (ioprio_valid(ioc->ioprio)) {
Tejun Heo6e736be2011-12-14 00:33:38 +01001273 new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
1274 if (unlikely(!new_ioc))
Jens Axboefd0928d2008-01-24 08:52:45 +01001275 return -ENOMEM;
1276
Tejun Heo6e736be2011-12-14 00:33:38 +01001277 new_ioc->ioprio = ioc->ioprio;
Tejun Heo11a31222012-02-07 07:51:30 +01001278 put_io_context(new_ioc);
Jens Axboefd0928d2008-01-24 08:52:45 +01001279 }
1280#endif
1281 return 0;
1282}
1283
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001284static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285{
1286 struct sighand_struct *sig;
1287
Zhaolei60348802009-01-06 14:40:46 -08001288 if (clone_flags & CLONE_SIGHAND) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 atomic_inc(&current->sighand->count);
1290 return 0;
1291 }
1292 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
Ingo Molnare56d0902006-01-08 01:01:37 -08001293 rcu_assign_pointer(tsk->sighand, sig);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294 if (!sig)
1295 return -ENOMEM;
Peter Zijlstra9d7fb042015-06-30 11:30:54 +02001296
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297 atomic_set(&sig->count, 1);
1298 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
1299 return 0;
1300}
1301
Oleg Nesterova7e53282006-03-28 16:11:27 -08001302void __cleanup_sighand(struct sighand_struct *sighand)
Oleg Nesterovc81addc2006-03-28 16:11:17 -08001303{
Oleg Nesterovd80e7312012-02-24 20:07:11 +01001304 if (atomic_dec_and_test(&sighand->count)) {
1305 signalfd_cleanup(sighand);
Oleg Nesterov392809b2014-09-28 23:44:18 +02001306 /*
1307 * sighand_cachep is SLAB_DESTROY_BY_RCU so we can free it
1308 * without an RCU grace period, see __lock_task_sighand().
1309 */
Oleg Nesterovc81addc2006-03-28 16:11:17 -08001310 kmem_cache_free(sighand_cachep, sighand);
Oleg Nesterovd80e7312012-02-24 20:07:11 +01001311 }
Oleg Nesterovc81addc2006-03-28 16:11:17 -08001312}
1313
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05001314#ifdef CONFIG_POSIX_TIMERS
Frank Mayharf06febc2008-09-12 09:54:39 -07001315/*
1316 * Initialize POSIX timer handling for a thread group.
1317 */
1318static void posix_cpu_timers_init_group(struct signal_struct *sig)
1319{
Jiri Slaby78d7d402010-03-05 13:42:54 -08001320 unsigned long cpu_limit;
1321
Jason Low316c1608d2015-04-28 13:00:20 -07001322 cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
Jiri Slaby78d7d402010-03-05 13:42:54 -08001323 if (cpu_limit != RLIM_INFINITY) {
Frederic Weisbeckerebd7e7f2017-01-31 04:09:34 +01001324 sig->cputime_expires.prof_exp = cpu_limit * NSEC_PER_SEC;
Jason Lowd5c373e2015-10-14 12:07:55 -07001325 sig->cputimer.running = true;
Oleg Nesterov6279a7512009-03-27 01:06:07 +01001326 }
1327
Frank Mayharf06febc2008-09-12 09:54:39 -07001328 /* The timer lists. */
1329 INIT_LIST_HEAD(&sig->cpu_timers[0]);
1330 INIT_LIST_HEAD(&sig->cpu_timers[1]);
1331 INIT_LIST_HEAD(&sig->cpu_timers[2]);
1332}
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05001333#else
1334static inline void posix_cpu_timers_init_group(struct signal_struct *sig) { }
1335#endif
Frank Mayharf06febc2008-09-12 09:54:39 -07001336
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001337static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338{
1339 struct signal_struct *sig;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340
Oleg Nesterov4ab6c082009-08-26 14:29:24 -07001341 if (clone_flags & CLONE_THREAD)
Peter Zijlstra490dea42008-11-24 17:06:57 +01001342 return 0;
Oleg Nesterov6279a7512009-03-27 01:06:07 +01001343
Veaceslav Falicoa56704e2010-03-10 15:23:01 -08001344 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345 tsk->signal = sig;
1346 if (!sig)
1347 return -ENOMEM;
1348
Oleg Nesterovb3ac0222010-05-26 14:43:24 -07001349 sig->nr_threads = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 atomic_set(&sig->live, 1);
Oleg Nesterovb3ac0222010-05-26 14:43:24 -07001351 atomic_set(&sig->sigcnt, 1);
Oleg Nesterov0c740d02014-01-21 15:49:56 -08001352
1353 /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
1354 sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node);
1355 tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head);
1356
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357 init_waitqueue_head(&sig->wait_chldexit);
Oleg Nesterovdb51aec2008-04-30 00:52:52 -07001358 sig->curr_target = tsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 init_sigpending(&sig->shared_pending);
Rik van Riele78c3492014-08-16 13:40:10 -04001360 seqlock_init(&sig->stats_lock);
Peter Zijlstra9d7fb042015-06-30 11:30:54 +02001361 prev_cputime_init(&sig->prev_cputime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362
Nicolas Pitrebaa73d92016-11-11 00:10:10 -05001363#ifdef CONFIG_POSIX_TIMERS
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05001364 INIT_LIST_HEAD(&sig->posix_timers);
Thomas Gleixnerc9cb2e32007-02-16 01:27:49 -08001365 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366 sig->real_timer.function = it_real_fn;
Nicolas Pitrebaa73d92016-11-11 00:10:10 -05001367#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 task_lock(current->group_leader);
1370 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
1371 task_unlock(current->group_leader);
1372
Oleg Nesterov6279a7512009-03-27 01:06:07 +01001373 posix_cpu_timers_init_group(sig);
1374
Miloslav Trmac522ed772007-07-15 23:40:56 -07001375 tty_audit_fork(sig);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001376 sched_autogroup_fork(sig);
Miloslav Trmac522ed772007-07-15 23:40:56 -07001377
David Rientjesa63d83f2010-08-09 17:19:46 -07001378 sig->oom_score_adj = current->signal->oom_score_adj;
Mandeep Singh Bainesdabb16f632011-01-13 15:46:05 -08001379 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001380
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001381 mutex_init(&sig->cred_guard_mutex);
1382
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 return 0;
1384}
1385
Kees Cookdbd952122014-06-27 15:18:48 -07001386static void copy_seccomp(struct task_struct *p)
1387{
1388#ifdef CONFIG_SECCOMP
1389 /*
1390 * Must be called with sighand->lock held, which is common to
1391 * all threads in the group. Holding cred_guard_mutex is not
1392 * needed because this new task is not yet running and cannot
1393 * be racing exec.
1394 */
Guenter Roeck69f6a342014-08-10 20:50:30 -07001395 assert_spin_locked(&current->sighand->siglock);
Kees Cookdbd952122014-06-27 15:18:48 -07001396
1397 /* Ref-count the new filter user, and assign it. */
1398 get_seccomp_filter(current);
1399 p->seccomp = current->seccomp;
1400
1401 /*
1402 * Explicitly enable no_new_privs here in case it got set
1403 * between the task_struct being duplicated and holding the
1404 * sighand lock. The seccomp state and nnp must be in sync.
1405 */
1406 if (task_no_new_privs(current))
1407 task_set_no_new_privs(p);
1408
1409 /*
1410 * If the parent gained a seccomp mode after copying thread
1411 * flags and between before we held the sighand lock, we have
1412 * to manually enable the seccomp thread flag here.
1413 */
1414 if (p->seccomp.mode != SECCOMP_MODE_DISABLED)
1415 set_tsk_thread_flag(p, TIF_SECCOMP);
1416#endif
1417}
1418
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001419SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420{
1421 current->clear_child_tid = tidptr;
1422
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001423 return task_pid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424}
1425
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001426static void rt_mutex_init_task(struct task_struct *p)
Ingo Molnar23f78d4a2006-06-27 02:54:53 -07001427{
Thomas Gleixner1d615482009-11-17 14:54:03 +01001428 raw_spin_lock_init(&p->pi_lock);
Zilvinas Valinskase29e1752007-03-16 13:38:34 -08001429#ifdef CONFIG_RT_MUTEXES
Peter Zijlstrafb00aca2013-11-07 14:43:43 +01001430 p->pi_waiters = RB_ROOT;
1431 p->pi_waiters_leftmost = NULL;
Ingo Molnar23f78d4a2006-06-27 02:54:53 -07001432 p->pi_blocked_on = NULL;
Ingo Molnar23f78d4a2006-06-27 02:54:53 -07001433#endif
1434}
1435
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05001436#ifdef CONFIG_POSIX_TIMERS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437/*
Frank Mayharf06febc2008-09-12 09:54:39 -07001438 * Initialize POSIX timer handling for a single task.
1439 */
1440static void posix_cpu_timers_init(struct task_struct *tsk)
1441{
Martin Schwidefsky64861632011-12-15 14:56:09 +01001442 tsk->cputime_expires.prof_exp = 0;
1443 tsk->cputime_expires.virt_exp = 0;
Frank Mayharf06febc2008-09-12 09:54:39 -07001444 tsk->cputime_expires.sched_exp = 0;
1445 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1446 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1447 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1448}
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05001449#else
1450static inline void posix_cpu_timers_init(struct task_struct *tsk) { }
1451#endif
Frank Mayharf06febc2008-09-12 09:54:39 -07001452
Oleg Nesterov81907732013-07-03 15:08:31 -07001453static inline void
1454init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid)
1455{
1456 task->pids[type].pid = pid;
1457}
1458
Frank Mayharf06febc2008-09-12 09:54:39 -07001459/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460 * This creates a new process as a copy of the old one,
1461 * but does not actually start it yet.
1462 *
1463 * It copies the registers, and all the appropriate
1464 * parts of the process environment (as per the clone
1465 * flags). The actual kick-off is left to the caller.
1466 */
Emese Revfy0766f782016-06-20 20:42:34 +02001467static __latent_entropy struct task_struct *copy_process(
1468 unsigned long clone_flags,
Ingo Molnar36c8b582006-07-03 00:25:41 -07001469 unsigned long stack_start,
Ingo Molnar36c8b582006-07-03 00:25:41 -07001470 unsigned long stack_size,
Ingo Molnar36c8b582006-07-03 00:25:41 -07001471 int __user *child_tidptr,
Roland McGrath09a05392008-07-25 19:45:47 -07001472 struct pid *pid,
Josh Triplett3033f14a2015-06-25 15:01:19 -07001473 int trace,
Andi Kleen725fc622016-05-23 16:24:05 -07001474 unsigned long tls,
1475 int node)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476{
1477 int retval;
Mariusz Kozlowskia24efe62007-10-18 23:41:09 -07001478 struct task_struct *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479
1480 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1481 return ERR_PTR(-EINVAL);
1482
Eric W. Biedermane66eded2013-03-13 11:51:49 -07001483 if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS))
1484 return ERR_PTR(-EINVAL);
1485
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 /*
1487 * Thread groups must share signals as well, and detached threads
1488 * can only be started up within the thread group.
1489 */
1490 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1491 return ERR_PTR(-EINVAL);
1492
1493 /*
1494 * Shared signal handlers imply shared VM. By way of the above,
1495 * thread groups also imply shared VM. Blocking this case allows
1496 * for various simplifications in other code.
1497 */
1498 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1499 return ERR_PTR(-EINVAL);
1500
Sukadev Bhattiprolu123be072009-09-23 15:57:20 -07001501 /*
1502 * Siblings of global init remain as zombies on exit since they are
1503 * not reaped by their parent (swapper). To solve this and to avoid
1504 * multi-rooted process trees, prevent global and container-inits
1505 * from creating siblings.
1506 */
1507 if ((clone_flags & CLONE_PARENT) &&
1508 current->signal->flags & SIGNAL_UNKILLABLE)
1509 return ERR_PTR(-EINVAL);
1510
Eric W. Biederman8382fca2012-12-20 19:26:06 -08001511 /*
Oleg Nesterov40a0d322013-09-11 14:19:41 -07001512 * If the new process will be in a different pid or user namespace
Eric W. Biedermanfaf00da2015-08-10 18:25:44 -05001513 * do not allow it to share a thread group with the forking task.
Eric W. Biederman8382fca2012-12-20 19:26:06 -08001514 */
Eric W. Biedermanfaf00da2015-08-10 18:25:44 -05001515 if (clone_flags & CLONE_THREAD) {
Oleg Nesterov40a0d322013-09-11 14:19:41 -07001516 if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) ||
1517 (task_active_pid_ns(current) !=
1518 current->nsproxy->pid_ns_for_children))
1519 return ERR_PTR(-EINVAL);
1520 }
Eric W. Biederman8382fca2012-12-20 19:26:06 -08001521
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522 retval = security_task_create(clone_flags);
1523 if (retval)
1524 goto fork_out;
1525
1526 retval = -ENOMEM;
Andi Kleen725fc622016-05-23 16:24:05 -07001527 p = dup_task_struct(current, node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 if (!p)
1529 goto fork_out;
1530
Steven Rostedtf7e8b612009-06-02 16:39:48 -04001531 ftrace_graph_init_task(p);
1532
Peter Zijlstrabea493a2006-10-17 00:10:33 -07001533 rt_mutex_init_task(p);
1534
Ingo Molnard12c1a32008-07-14 12:09:28 +02001535#ifdef CONFIG_PROVE_LOCKING
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001536 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1537 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1538#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 retval = -EAGAIN;
David Howells3b11a1d2008-11-14 10:39:26 +11001540 if (atomic_read(&p->real_cred->user->processes) >=
Jiri Slaby78d7d402010-03-05 13:42:54 -08001541 task_rlimit(p, RLIMIT_NPROC)) {
Eric Parisb57922b2013-07-03 15:08:29 -07001542 if (p->real_cred->user != INIT_USER &&
1543 !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 goto bad_fork_free;
1545 }
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001546 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547
David Howellsf1752ee2008-11-14 10:39:17 +11001548 retval = copy_creds(p, clone_flags);
1549 if (retval < 0)
1550 goto bad_fork_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551
1552 /*
1553 * If multiple threads are within copy_process(), then this check
1554 * triggers too late. This doesn't hurt, the check is only there
1555 * to stop root fork bombs.
1556 */
Li Zefan04ec93f2009-02-06 08:17:19 +00001557 retval = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 if (nr_threads >= max_threads)
1559 goto bad_fork_cleanup_count;
1560
Shailabh Nagarca74e922006-07-14 00:24:36 -07001561 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
Peter Zijlstrac1de45c2016-11-28 23:03:05 -08001562 p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE);
David Rientjes514ddb42014-04-07 15:37:27 -07001563 p->flags |= PF_FORKNOEXEC;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564 INIT_LIST_HEAD(&p->children);
1565 INIT_LIST_HEAD(&p->sibling);
Paul E. McKenneyf41d9112009-08-22 13:56:52 -07001566 rcu_copy_process(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 p->vfork_done = NULL;
1568 spin_lock_init(&p->alloc_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570 init_sigpending(&p->pending);
1571
Martin Schwidefsky64861632011-12-15 14:56:09 +01001572 p->utime = p->stime = p->gtime = 0;
Stanislaw Gruszka40565b52016-11-15 03:06:51 +01001573#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
Martin Schwidefsky64861632011-12-15 14:56:09 +01001574 p->utimescaled = p->stimescaled = 0;
Stanislaw Gruszka40565b52016-11-15 03:06:51 +01001575#endif
Peter Zijlstra9d7fb042015-06-30 11:30:54 +02001576 prev_cputime_init(&p->prev_cputime);
1577
Frederic Weisbecker6a616712012-12-16 20:00:34 +01001578#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
Frederic Weisbeckerb7ce2272015-11-19 16:47:34 +01001579 seqcount_init(&p->vtime_seqcount);
Frederic Weisbecker6a616712012-12-16 20:00:34 +01001580 p->vtime_snap = 0;
Frederic Weisbecker7098c1e2015-11-19 16:47:30 +01001581 p->vtime_snap_whence = VTIME_INACTIVE;
Frederic Weisbecker6a616712012-12-16 20:00:34 +01001582#endif
1583
KAMEZAWA Hiroyukia3a2e762010-04-06 14:34:42 -07001584#if defined(SPLIT_RSS_COUNTING)
1585 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1586#endif
Balbir Singh172ba842007-07-09 18:52:00 +02001587
Arjan van de Ven69766752008-09-01 15:52:40 -07001588 p->default_timer_slack_ns = current->timer_slack_ns;
1589
Andrea Righi59954772008-07-27 17:29:15 +02001590 task_io_accounting_init(&p->ioac);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591 acct_clear_integrals(p);
1592
Frank Mayharf06febc2008-09-12 09:54:39 -07001593 posix_cpu_timers_init(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594
Thomas Gleixnerccbf62d2014-07-16 21:04:34 +00001595 p->start_time = ktime_get_ns();
Thomas Gleixner57e0be02014-07-16 21:04:32 +00001596 p->real_start_time = ktime_get_boot_ns();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 p->io_context = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 p->audit_context = NULL;
Paul Menageb4f48b62007-10-18 23:39:33 -07001599 cgroup_fork(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600#ifdef CONFIG_NUMA
Lee Schermerhorn846a16b2008-04-28 02:13:09 -07001601 p->mempolicy = mpol_dup(p->mempolicy);
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001602 if (IS_ERR(p->mempolicy)) {
1603 retval = PTR_ERR(p->mempolicy);
1604 p->mempolicy = NULL;
Li Zefane8604cb2014-03-28 15:18:27 +08001605 goto bad_fork_cleanup_threadgroup_lock;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001606 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607#endif
Michal Hocko778d3b02011-07-26 16:08:30 -07001608#ifdef CONFIG_CPUSETS
1609 p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1610 p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
Mel Gormancc9a6c82012-03-21 16:34:11 -07001611 seqcount_init(&p->mems_allowed_seq);
Michal Hocko778d3b02011-07-26 16:08:30 -07001612#endif
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001613#ifdef CONFIG_TRACE_IRQFLAGS
1614 p->irq_events = 0;
1615 p->hardirqs_enabled = 0;
1616 p->hardirq_enable_ip = 0;
1617 p->hardirq_enable_event = 0;
1618 p->hardirq_disable_ip = _THIS_IP_;
1619 p->hardirq_disable_event = 0;
1620 p->softirqs_enabled = 1;
1621 p->softirq_enable_ip = _THIS_IP_;
1622 p->softirq_enable_event = 0;
1623 p->softirq_disable_ip = 0;
1624 p->softirq_disable_event = 0;
1625 p->hardirq_context = 0;
1626 p->softirq_context = 0;
1627#endif
David Hildenbrand8bcbde52015-05-11 17:52:06 +02001628
1629 p->pagefault_disabled = 0;
1630
Ingo Molnarfbb9ce952006-07-03 00:24:50 -07001631#ifdef CONFIG_LOCKDEP
1632 p->lockdep_depth = 0; /* no locks held yet */
1633 p->curr_chain_key = 0;
1634 p->lockdep_recursion = 0;
1635#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636
Ingo Molnar408894e2006-01-09 15:59:20 -08001637#ifdef CONFIG_DEBUG_MUTEXES
1638 p->blocked_on = NULL; /* not blocked yet */
1639#endif
Kent Overstreetcafe5632013-03-23 16:11:31 -07001640#ifdef CONFIG_BCACHE
1641 p->sequential_io = 0;
1642 p->sequential_io_avg = 0;
1643#endif
Markus Metzger0f481402009-04-03 16:43:48 +02001644
Srivatsa Vaddagiri3c90e6e2007-11-09 22:39:39 +01001645 /* Perform scheduler related setup. Assign this task to a CPU. */
Dario Faggioliaab03e02013-11-28 11:14:43 +01001646 retval = sched_fork(clone_flags, p);
1647 if (retval)
1648 goto bad_fork_cleanup_policy;
Peter Zijlstra6ab423e2009-05-25 14:45:27 +02001649
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001650 retval = perf_event_init_task(p);
Peter Zijlstra6ab423e2009-05-25 14:45:27 +02001651 if (retval)
1652 goto bad_fork_cleanup_policy;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001653 retval = audit_alloc(p);
1654 if (retval)
Peter Zijlstra6c72e3502014-10-02 16:17:02 -07001655 goto bad_fork_cleanup_perf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656 /* copy all the process information */
Jack Millerab602f72014-08-08 14:23:19 -07001657 shm_init_task(p);
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001658 retval = copy_semundo(clone_flags, p);
1659 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660 goto bad_fork_cleanup_audit;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001661 retval = copy_files(clone_flags, p);
1662 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 goto bad_fork_cleanup_semundo;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001664 retval = copy_fs(clone_flags, p);
1665 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 goto bad_fork_cleanup_files;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001667 retval = copy_sighand(clone_flags, p);
1668 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 goto bad_fork_cleanup_fs;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001670 retval = copy_signal(clone_flags, p);
1671 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672 goto bad_fork_cleanup_sighand;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001673 retval = copy_mm(clone_flags, p);
1674 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675 goto bad_fork_cleanup_signal;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001676 retval = copy_namespaces(clone_flags, p);
1677 if (retval)
David Howellsd84f4f92008-11-14 10:39:23 +11001678 goto bad_fork_cleanup_mm;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001679 retval = copy_io(clone_flags, p);
1680 if (retval)
Jens Axboefd0928d2008-01-24 08:52:45 +01001681 goto bad_fork_cleanup_namespaces;
Josh Triplett3033f14a2015-06-25 15:01:19 -07001682 retval = copy_thread_tls(clone_flags, stack_start, stack_size, p, tls);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 if (retval)
Jens Axboefd0928d2008-01-24 08:52:45 +01001684 goto bad_fork_cleanup_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001686 if (pid != &init_struct_pid) {
Andy Lutomirskic2b1df22013-08-22 11:39:16 -07001687 pid = alloc_pid(p->nsproxy->pid_ns_for_children);
Michal Hocko35f71bc2015-04-16 12:47:38 -07001688 if (IS_ERR(pid)) {
1689 retval = PTR_ERR(pid);
Jiri Slaby0740aa52016-05-20 17:00:25 -07001690 goto bad_fork_cleanup_thread;
Michal Hocko35f71bc2015-04-16 12:47:38 -07001691 }
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001692 }
1693
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? child_tidptr : NULL;
1695 /*
1696 * Clear TID on mm_release()?
1697 */
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001698 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? child_tidptr : NULL;
Jens Axboe73c10102011-03-08 13:19:51 +01001699#ifdef CONFIG_BLOCK
1700 p->plug = NULL;
1701#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001702#ifdef CONFIG_FUTEX
Ingo Molnar8f17d3a2006-03-27 01:16:27 -08001703 p->robust_list = NULL;
1704#ifdef CONFIG_COMPAT
1705 p->compat_robust_list = NULL;
1706#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07001707 INIT_LIST_HEAD(&p->pi_state_list);
1708 p->pi_state_cache = NULL;
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07001709#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 /*
GOTO Masanorif9a38792006-03-13 21:20:44 -08001711 * sigaltstack should be cleared when sharing the same VM
1712 */
1713 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
Stas Sergeev2a742132016-04-14 23:20:04 +03001714 sas_ss_reset(p);
GOTO Masanorif9a38792006-03-13 21:20:44 -08001715
1716 /*
Oleg Nesterov65808072009-12-15 16:47:16 -08001717 * Syscall tracing and stepping should be turned off in the
1718 * child regardless of CLONE_PTRACE.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 */
Oleg Nesterov65808072009-12-15 16:47:16 -08001720 user_disable_single_step(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
Laurent Viviered75e8d2005-09-03 15:57:18 -07001722#ifdef TIF_SYSCALL_EMU
1723 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
1724#endif
Arjan van de Ven97455122008-01-25 21:08:34 +01001725 clear_all_latency_tracing(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 /* ok, now we should be set up.. */
Oleg Nesterov18c830d2013-07-03 15:08:32 -07001728 p->pid = pid_nr(pid);
1729 if (clone_flags & CLONE_THREAD) {
Oleg Nesterov5f8aadd2012-03-14 19:55:38 +01001730 p->exit_signal = -1;
Oleg Nesterov18c830d2013-07-03 15:08:32 -07001731 p->group_leader = current->group_leader;
1732 p->tgid = current->tgid;
1733 } else {
1734 if (clone_flags & CLONE_PARENT)
1735 p->exit_signal = current->group_leader->exit_signal;
1736 else
1737 p->exit_signal = (clone_flags & CSIGNAL);
1738 p->group_leader = p;
1739 p->tgid = p->pid;
1740 }
Oleg Nesterov5f8aadd2012-03-14 19:55:38 +01001741
Wu Fengguang9d823e82011-06-11 18:10:12 -06001742 p->nr_dirtied = 0;
1743 p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
Wu Fengguang83712352011-06-11 19:25:42 -06001744 p->dirty_paused_when = 0;
Wu Fengguang9d823e82011-06-11 18:10:12 -06001745
Oleg Nesterovbb8cbbf2013-11-13 15:36:12 +01001746 p->pdeath_signal = 0;
Oleg Nesterov47e65322006-03-28 16:11:25 -08001747 INIT_LIST_HEAD(&p->thread_group);
Al Viro158e1642012-06-27 09:24:13 +04001748 p->task_works = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749
Ingo Molnar780de9d2017-02-02 11:50:56 +01001750 cgroup_threadgroup_change_begin(current);
Oleg Nesterov18c830d2013-07-03 15:08:32 -07001751 /*
Aleksa Sarai7e476822015-06-09 21:32:09 +10001752 * Ensure that the cgroup subsystem policies allow the new process to be
1753 * forked. It should be noted the the new process's css_set can be changed
1754 * between here and cgroup_post_fork() if an organisation operation is in
1755 * progress.
1756 */
Oleg Nesterovb53202e2015-12-03 10:24:08 -05001757 retval = cgroup_can_fork(p);
Aleksa Sarai7e476822015-06-09 21:32:09 +10001758 if (retval)
1759 goto bad_fork_free_pid;
1760
1761 /*
Oleg Nesterov18c830d2013-07-03 15:08:32 -07001762 * Make it visible to the rest of the system, but dont wake it up yet.
1763 * Need tasklist lock for parent etc handling!
1764 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 write_lock_irq(&tasklist_lock);
1766
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767 /* CLONE_PARENT re-uses the old parent */
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01001768 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 p->real_parent = current->real_parent;
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01001770 p->parent_exec_id = current->parent_exec_id;
1771 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772 p->real_parent = current;
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01001773 p->parent_exec_id = current->self_exec_id;
1774 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775
Oleg Nesterov3f17da62006-02-15 22:13:24 +03001776 spin_lock(&current->sighand->siglock);
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001777
1778 /*
Kees Cookdbd952122014-06-27 15:18:48 -07001779 * Copy seccomp details explicitly here, in case they were changed
1780 * before holding sighand lock.
1781 */
1782 copy_seccomp(p);
1783
1784 /*
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001785 * Process group and session signals need to be delivered to just the
1786 * parent before the fork or both the parent and the child after the
1787 * fork. Restart if a signal comes in before we add the new process to
1788 * it's process group.
1789 * A fatal signal pending means that current will exit, so the new
1790 * thread can't slip out of an OOM kill (or normal SIGKILL).
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001791 */
Daniel Walker23ff4442007-10-18 03:06:07 -07001792 recalc_sigpending();
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001793 if (signal_pending(current)) {
1794 spin_unlock(&current->sighand->siglock);
1795 write_unlock_irq(&tasklist_lock);
1796 retval = -ERESTARTNOINTR;
Aleksa Sarai7e476822015-06-09 21:32:09 +10001797 goto bad_fork_cancel_cgroup;
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08001798 }
1799
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001800 if (likely(p->pid)) {
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001801 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802
Oleg Nesterov81907732013-07-03 15:08:31 -07001803 init_task_pid(p, PIDTYPE_PID, pid);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001804 if (thread_group_leader(p)) {
Oleg Nesterov81907732013-07-03 15:08:31 -07001805 init_task_pid(p, PIDTYPE_PGID, task_pgrp(current));
1806 init_task_pid(p, PIDTYPE_SID, task_session(current));
1807
Eric W. Biederman1c4042c2010-07-12 17:10:36 -07001808 if (is_child_reaper(pid)) {
Eric W. Biederman17cf22c2010-03-02 14:51:53 -08001809 ns_of_pid(pid)->child_reaper = p;
Eric W. Biederman1c4042c2010-07-12 17:10:36 -07001810 p->signal->flags |= SIGNAL_UNKILLABLE;
1811 }
Oleg Nesterovc97d9892006-03-28 16:11:06 -08001812
Oleg Nesterovfea9d172008-02-08 04:19:19 -08001813 p->signal->leader_pid = pid;
Alan Cox9c9f4de2008-10-13 10:37:26 +01001814 p->signal->tty = tty_kref_get(current->signal->tty);
Pavel Tikhomirov749860c2017-01-30 18:06:12 +03001815 /*
1816 * Inherit has_child_subreaper flag under the same
1817 * tasklist_lock with adding child to the process tree
1818 * for propagate_has_child_subreaper optimization.
1819 */
1820 p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper ||
1821 p->real_parent->signal->is_child_subreaper;
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001822 list_add_tail(&p->sibling, &p->real_parent->children);
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -07001823 list_add_tail_rcu(&p->tasks, &init_task.tasks);
Oleg Nesterov81907732013-07-03 15:08:31 -07001824 attach_pid(p, PIDTYPE_PGID);
1825 attach_pid(p, PIDTYPE_SID);
Christoph Lameter909ea962010-12-08 16:22:55 +01001826 __this_cpu_inc(process_counts);
Oleg Nesterov80628ca2013-07-03 15:08:30 -07001827 } else {
1828 current->signal->nr_threads++;
1829 atomic_inc(&current->signal->live);
1830 atomic_inc(&current->signal->sigcnt);
Oleg Nesterov80628ca2013-07-03 15:08:30 -07001831 list_add_tail_rcu(&p->thread_group,
1832 &p->group_leader->thread_group);
Oleg Nesterov0c740d02014-01-21 15:49:56 -08001833 list_add_tail_rcu(&p->thread_node,
1834 &p->signal->thread_head);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001835 }
Oleg Nesterov81907732013-07-03 15:08:31 -07001836 attach_pid(p, PIDTYPE_PID);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001837 nr_threads++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838 }
1839
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 total_forks++;
Oleg Nesterov3f17da62006-02-15 22:13:24 +03001841 spin_unlock(&current->sighand->siglock);
Oleg Nesterov4af42062014-04-13 20:58:54 +02001842 syscall_tracepoint_update(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843 write_unlock_irq(&tasklist_lock);
Oleg Nesterov4af42062014-04-13 20:58:54 +02001844
Andrew Mortonc13cf852005-11-28 13:43:48 -08001845 proc_fork_connector(p);
Oleg Nesterovb53202e2015-12-03 10:24:08 -05001846 cgroup_post_fork(p);
Ingo Molnar780de9d2017-02-02 11:50:56 +01001847 cgroup_threadgroup_change_end(current);
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001848 perf_event_fork(p);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001849
1850 trace_task_newtask(p, clone_flags);
Oleg Nesterov3ab67962013-10-16 19:39:37 +02001851 uprobe_copy_process(p, clone_flags);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001852
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853 return p;
1854
Aleksa Sarai7e476822015-06-09 21:32:09 +10001855bad_fork_cancel_cgroup:
Oleg Nesterovb53202e2015-12-03 10:24:08 -05001856 cgroup_cancel_fork(p);
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001857bad_fork_free_pid:
Ingo Molnar780de9d2017-02-02 11:50:56 +01001858 cgroup_threadgroup_change_end(current);
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07001859 if (pid != &init_struct_pid)
1860 free_pid(pid);
Jiri Slaby0740aa52016-05-20 17:00:25 -07001861bad_fork_cleanup_thread:
1862 exit_thread(p);
Jens Axboefd0928d2008-01-24 08:52:45 +01001863bad_fork_cleanup_io:
Louis Rillingb69f2292009-12-04 14:52:42 +01001864 if (p->io_context)
1865 exit_io_context(p);
Serge E. Hallynab516012006-10-02 02:18:06 -07001866bad_fork_cleanup_namespaces:
Linus Torvalds444f3782007-01-30 13:35:18 -08001867 exit_task_namespaces(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868bad_fork_cleanup_mm:
David Rientjesc9f01242011-10-31 17:07:15 -07001869 if (p->mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870 mmput(p->mm);
1871bad_fork_cleanup_signal:
Oleg Nesterov4ab6c082009-08-26 14:29:24 -07001872 if (!(clone_flags & CLONE_THREAD))
Mike Galbraith1c5354d2011-01-05 11:16:04 +01001873 free_signal_struct(p->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874bad_fork_cleanup_sighand:
Oleg Nesterova7e53282006-03-28 16:11:27 -08001875 __cleanup_sighand(p->sighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876bad_fork_cleanup_fs:
1877 exit_fs(p); /* blocking */
1878bad_fork_cleanup_files:
1879 exit_files(p); /* blocking */
1880bad_fork_cleanup_semundo:
1881 exit_sem(p);
1882bad_fork_cleanup_audit:
1883 audit_free(p);
Peter Zijlstra6c72e3502014-10-02 16:17:02 -07001884bad_fork_cleanup_perf:
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001885 perf_event_free_task(p);
Peter Zijlstra6c72e3502014-10-02 16:17:02 -07001886bad_fork_cleanup_policy:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887#ifdef CONFIG_NUMA
Lee Schermerhornf0be3d32008-04-28 02:13:08 -07001888 mpol_put(p->mempolicy);
Li Zefane8604cb2014-03-28 15:18:27 +08001889bad_fork_cleanup_threadgroup_lock:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890#endif
Shailabh Nagar35df17c2006-08-31 21:27:38 -07001891 delayacct_tsk_free(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892bad_fork_cleanup_count:
David Howellsd84f4f92008-11-14 10:39:23 +11001893 atomic_dec(&p->cred->user->processes);
David Howellse0e81732009-09-02 09:13:40 +01001894 exit_creds(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895bad_fork_free:
Andy Lutomirski405c0752016-10-31 08:11:43 -07001896 p->state = TASK_DEAD;
Andy Lutomirski68f24b082016-09-15 22:45:48 -07001897 put_task_stack(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898 free_task(p);
Oleg Nesterovfe7d37d2006-01-08 01:04:02 -08001899fork_out:
1900 return ERR_PTR(retval);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901}
1902
Oleg Nesterovf106eee2010-05-26 14:44:11 -07001903static inline void init_idle_pids(struct pid_link *links)
1904{
1905 enum pid_type type;
1906
1907 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1908 INIT_HLIST_NODE(&links[type].node); /* not really needed */
1909 links[type].pid = &init_struct_pid;
1910 }
1911}
1912
Paul Gortmaker0db06282013-06-19 14:53:51 -04001913struct task_struct *fork_idle(int cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914{
Ingo Molnar36c8b582006-07-03 00:25:41 -07001915 struct task_struct *task;
Andi Kleen725fc622016-05-23 16:24:05 -07001916 task = copy_process(CLONE_VM, 0, 0, NULL, &init_struct_pid, 0, 0,
1917 cpu_to_node(cpu));
Oleg Nesterovf106eee2010-05-26 14:44:11 -07001918 if (!IS_ERR(task)) {
1919 init_idle_pids(task->pids);
Akinobu Mita753ca4f2006-11-25 11:09:34 -08001920 init_idle(task, cpu);
Oleg Nesterovf106eee2010-05-26 14:44:11 -07001921 }
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08001922
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 return task;
1924}
1925
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926/*
1927 * Ok, this is the main fork-routine.
1928 *
1929 * It copies the process, and if successful kick-starts
1930 * it and waits for it to finish using the VM if required.
1931 */
Josh Triplett3033f14a2015-06-25 15:01:19 -07001932long _do_fork(unsigned long clone_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933 unsigned long stack_start,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 unsigned long stack_size,
1935 int __user *parent_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07001936 int __user *child_tidptr,
1937 unsigned long tls)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938{
1939 struct task_struct *p;
1940 int trace = 0;
Eric W. Biederman92476d72006-03-31 02:31:42 -08001941 long nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942
Andrew Mortonbdff7462008-02-04 22:27:22 -08001943 /*
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001944 * Determine whether and which event to report to ptracer. When
1945 * called from kernel_thread or CLONE_UNTRACED is explicitly
1946 * requested, no event is reported; otherwise, report if the event
1947 * for the type of forking is enabled.
Roland McGrath09a05392008-07-25 19:45:47 -07001948 */
Al Viroe80d6662012-10-22 23:10:08 -04001949 if (!(clone_flags & CLONE_UNTRACED)) {
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001950 if (clone_flags & CLONE_VFORK)
1951 trace = PTRACE_EVENT_VFORK;
1952 else if ((clone_flags & CSIGNAL) != SIGCHLD)
1953 trace = PTRACE_EVENT_CLONE;
1954 else
1955 trace = PTRACE_EVENT_FORK;
1956
1957 if (likely(!ptrace_event_enabled(current, trace)))
1958 trace = 0;
1959 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960
Al Viro62e791c2012-10-22 22:52:26 -04001961 p = copy_process(clone_flags, stack_start, stack_size,
Andi Kleen725fc622016-05-23 16:24:05 -07001962 child_tidptr, NULL, trace, tls, NUMA_NO_NODE);
Emese Revfy38addce2016-06-20 20:41:19 +02001963 add_latent_entropy();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964 /*
1965 * Do this prior waking up the new thread - the thread pointer
1966 * might get invalid after that point, if the thread exits quickly.
1967 */
1968 if (!IS_ERR(p)) {
1969 struct completion vfork;
Matthew Dempsky4e523652014-06-06 14:36:42 -07001970 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04001972 trace_sched_process_fork(current, p);
1973
Matthew Dempsky4e523652014-06-06 14:36:42 -07001974 pid = get_task_pid(p, PIDTYPE_PID);
1975 nr = pid_vnr(pid);
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07001976
1977 if (clone_flags & CLONE_PARENT_SETTID)
1978 put_user(nr, parent_tidptr);
Sukadev Bhattiprolua6f5e062007-10-18 23:39:53 -07001979
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980 if (clone_flags & CLONE_VFORK) {
1981 p->vfork_done = &vfork;
1982 init_completion(&vfork);
Oleg Nesterovd68b46f2012-03-05 14:59:13 -08001983 get_task_struct(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 }
1985
Samir Bellabes3e51e3e2011-05-11 18:18:05 +02001986 wake_up_new_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001988 /* forking complete and child started to run, tell ptracer */
1989 if (unlikely(trace))
Matthew Dempsky4e523652014-06-06 14:36:42 -07001990 ptrace_event_pid(trace, pid);
Roland McGrath09a05392008-07-25 19:45:47 -07001991
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 if (clone_flags & CLONE_VFORK) {
Oleg Nesterovd68b46f2012-03-05 14:59:13 -08001993 if (!wait_for_vfork_done(p, &vfork))
Matthew Dempsky4e523652014-06-06 14:36:42 -07001994 ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995 }
Matthew Dempsky4e523652014-06-06 14:36:42 -07001996
1997 put_pid(pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998 } else {
Eric W. Biederman92476d72006-03-31 02:31:42 -08001999 nr = PTR_ERR(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 }
Eric W. Biederman92476d72006-03-31 02:31:42 -08002001 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002}
2003
Josh Triplett3033f14a2015-06-25 15:01:19 -07002004#ifndef CONFIG_HAVE_COPY_THREAD_TLS
2005/* For compatibility with architectures that call do_fork directly rather than
2006 * using the syscall entry points below. */
2007long do_fork(unsigned long clone_flags,
2008 unsigned long stack_start,
2009 unsigned long stack_size,
2010 int __user *parent_tidptr,
2011 int __user *child_tidptr)
2012{
2013 return _do_fork(clone_flags, stack_start, stack_size,
2014 parent_tidptr, child_tidptr, 0);
2015}
2016#endif
2017
Al Viro2aa3a7f2012-09-21 19:55:31 -04002018/*
2019 * Create a kernel thread.
2020 */
2021pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
2022{
Josh Triplett3033f14a2015-06-25 15:01:19 -07002023 return _do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn,
2024 (unsigned long)arg, NULL, NULL, 0);
Al Viro2aa3a7f2012-09-21 19:55:31 -04002025}
Al Viro2aa3a7f2012-09-21 19:55:31 -04002026
Al Virod2125042012-10-23 13:17:59 -04002027#ifdef __ARCH_WANT_SYS_FORK
2028SYSCALL_DEFINE0(fork)
2029{
2030#ifdef CONFIG_MMU
Josh Triplett3033f14a2015-06-25 15:01:19 -07002031 return _do_fork(SIGCHLD, 0, 0, NULL, NULL, 0);
Al Virod2125042012-10-23 13:17:59 -04002032#else
2033 /* can not support in nommu mode */
Daeseok Youn5d59e182014-01-23 15:55:47 -08002034 return -EINVAL;
Al Virod2125042012-10-23 13:17:59 -04002035#endif
2036}
2037#endif
2038
2039#ifdef __ARCH_WANT_SYS_VFORK
2040SYSCALL_DEFINE0(vfork)
2041{
Josh Triplett3033f14a2015-06-25 15:01:19 -07002042 return _do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, 0,
2043 0, NULL, NULL, 0);
Al Virod2125042012-10-23 13:17:59 -04002044}
2045#endif
2046
2047#ifdef __ARCH_WANT_SYS_CLONE
2048#ifdef CONFIG_CLONE_BACKWARDS
2049SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
2050 int __user *, parent_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07002051 unsigned long, tls,
Al Virod2125042012-10-23 13:17:59 -04002052 int __user *, child_tidptr)
2053#elif defined(CONFIG_CLONE_BACKWARDS2)
2054SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags,
2055 int __user *, parent_tidptr,
2056 int __user *, child_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07002057 unsigned long, tls)
Michal Simekdfa97712013-08-13 16:00:53 -07002058#elif defined(CONFIG_CLONE_BACKWARDS3)
2059SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp,
2060 int, stack_size,
2061 int __user *, parent_tidptr,
2062 int __user *, child_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07002063 unsigned long, tls)
Al Virod2125042012-10-23 13:17:59 -04002064#else
2065SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
2066 int __user *, parent_tidptr,
2067 int __user *, child_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07002068 unsigned long, tls)
Al Virod2125042012-10-23 13:17:59 -04002069#endif
2070{
Josh Triplett3033f14a2015-06-25 15:01:19 -07002071 return _do_fork(clone_flags, newsp, 0, parent_tidptr, child_tidptr, tls);
Al Virod2125042012-10-23 13:17:59 -04002072}
2073#endif
2074
Oleg Nesterov0f1b92c2017-01-30 18:06:11 +03002075void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data)
2076{
2077 struct task_struct *leader, *parent, *child;
2078 int res;
2079
2080 read_lock(&tasklist_lock);
2081 leader = top = top->group_leader;
2082down:
2083 for_each_thread(leader, parent) {
2084 list_for_each_entry(child, &parent->children, sibling) {
2085 res = visitor(child, data);
2086 if (res) {
2087 if (res < 0)
2088 goto out;
2089 leader = child;
2090 goto down;
2091 }
2092up:
2093 ;
2094 }
2095 }
2096
2097 if (leader != top) {
2098 child = leader;
2099 parent = child->real_parent;
2100 leader = parent->group_leader;
2101 goto up;
2102 }
2103out:
2104 read_unlock(&tasklist_lock);
2105}
2106
Ravikiran G Thirumalai5fd63b32006-01-11 22:46:15 +01002107#ifndef ARCH_MIN_MMSTRUCT_ALIGN
2108#define ARCH_MIN_MMSTRUCT_ALIGN 0
2109#endif
2110
Alexey Dobriyan51cc5062008-07-25 19:45:34 -07002111static void sighand_ctor(void *data)
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08002112{
2113 struct sighand_struct *sighand = data;
2114
Christoph Lametera35afb82007-05-16 22:10:57 -07002115 spin_lock_init(&sighand->siglock);
Davide Libenzib8fceee2007-09-20 12:40:16 -07002116 init_waitqueue_head(&sighand->signalfd_wqh);
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08002117}
2118
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119void __init proc_caches_init(void)
2120{
2121 sighand_cachep = kmem_cache_create("sighand_cache",
2122 sizeof(struct sighand_struct), 0,
Vegard Nossum2dff4402008-05-31 15:56:17 +02002123 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_DESTROY_BY_RCU|
Vladimir Davydov5d097052016-01-14 15:18:21 -08002124 SLAB_NOTRACK|SLAB_ACCOUNT, sighand_ctor);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125 signal_cachep = kmem_cache_create("signal_cache",
2126 sizeof(struct signal_struct), 0,
Vladimir Davydov5d097052016-01-14 15:18:21 -08002127 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
2128 NULL);
Paul Mundt20c2df82007-07-20 10:11:58 +09002129 files_cachep = kmem_cache_create("files_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130 sizeof(struct files_struct), 0,
Vladimir Davydov5d097052016-01-14 15:18:21 -08002131 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
2132 NULL);
Paul Mundt20c2df82007-07-20 10:11:58 +09002133 fs_cachep = kmem_cache_create("fs_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134 sizeof(struct fs_struct), 0,
Vladimir Davydov5d097052016-01-14 15:18:21 -08002135 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
2136 NULL);
Linus Torvalds6345d242011-05-29 11:32:28 -07002137 /*
2138 * FIXME! The "sizeof(struct mm_struct)" currently includes the
2139 * whole struct cpumask for the OFFSTACK case. We could change
2140 * this to *only* allocate as much of it as required by the
2141 * maximum number of CPU's we can ever have. The cpumask_allocation
2142 * is at the end of the structure, exactly for that reason.
2143 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002144 mm_cachep = kmem_cache_create("mm_struct",
Ravikiran G Thirumalai5fd63b32006-01-11 22:46:15 +01002145 sizeof(struct mm_struct), ARCH_MIN_MMSTRUCT_ALIGN,
Vladimir Davydov5d097052016-01-14 15:18:21 -08002146 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_NOTRACK|SLAB_ACCOUNT,
2147 NULL);
2148 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT);
David Howells8feae132009-01-08 12:04:47 +00002149 mmap_init();
Al Viro66577192011-06-28 15:41:10 -04002150 nsproxy_cache_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151}
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002152
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002153/*
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002154 * Check constraints on flags passed to the unshare system call.
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002155 */
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002156static int check_unshare_flags(unsigned long unshare_flags)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002157{
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002158 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
2159 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
Eric W. Biederman50804fe2010-03-02 15:41:50 -08002160 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET|
Aditya Kalia79a9082016-01-29 02:54:06 -06002161 CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP))
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002162 return -EINVAL;
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002163 /*
Eric W. Biederman12c641a2015-08-10 17:35:07 -05002164 * Not implemented, but pretend it works if there is nothing
2165 * to unshare. Note that unsharing the address space or the
2166 * signal handlers also need to unshare the signal queues (aka
2167 * CLONE_THREAD).
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002168 */
2169 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
Eric W. Biederman12c641a2015-08-10 17:35:07 -05002170 if (!thread_group_empty(current))
2171 return -EINVAL;
2172 }
2173 if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) {
2174 if (atomic_read(&current->sighand->count) > 1)
2175 return -EINVAL;
2176 }
2177 if (unshare_flags & CLONE_VM) {
2178 if (!current_is_single_threaded())
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002179 return -EINVAL;
2180 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002181
2182 return 0;
2183}
2184
2185/*
JANAK DESAI99d14192006-02-07 12:58:59 -08002186 * Unshare the filesystem structure if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002187 */
2188static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
2189{
2190 struct fs_struct *fs = current->fs;
2191
Al Viro498052b2009-03-30 07:20:30 -04002192 if (!(unshare_flags & CLONE_FS) || !fs)
2193 return 0;
2194
2195 /* don't need lock here; in the worst case we'll do useless copy */
2196 if (fs->users == 1)
2197 return 0;
2198
2199 *new_fsp = copy_fs_struct(fs);
2200 if (!*new_fsp)
2201 return -ENOMEM;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002202
2203 return 0;
2204}
2205
2206/*
JANAK DESAIa016f332006-02-07 12:59:02 -08002207 * Unshare file descriptor table if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002208 */
2209static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
2210{
2211 struct files_struct *fd = current->files;
JANAK DESAIa016f332006-02-07 12:59:02 -08002212 int error = 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002213
2214 if ((unshare_flags & CLONE_FILES) &&
JANAK DESAIa016f332006-02-07 12:59:02 -08002215 (fd && atomic_read(&fd->count) > 1)) {
2216 *new_fdp = dup_fd(fd, &error);
2217 if (!*new_fdp)
2218 return error;
2219 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002220
2221 return 0;
2222}
2223
2224/*
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002225 * unshare allows a process to 'unshare' part of the process
2226 * context which was originally shared using clone. copy_*
2227 * functions used by do_fork() cannot be used here directly
2228 * because they modify an inactive task_struct that is being
2229 * constructed. Here we are modifying the current, active,
2230 * task_struct.
2231 */
Heiko Carstens6559eed82009-01-14 14:14:32 +01002232SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002233{
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002234 struct fs_struct *fs, *new_fs = NULL;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002235 struct files_struct *fd, *new_fd = NULL;
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002236 struct cred *new_cred = NULL;
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07002237 struct nsproxy *new_nsproxy = NULL;
Manfred Spraul9edff4a2008-04-29 01:00:57 -07002238 int do_sysvsem = 0;
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002239 int err;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002240
Eric W. Biederman50804fe2010-03-02 15:41:50 -08002241 /*
Eric W. Biedermanfaf00da2015-08-10 18:25:44 -05002242 * If unsharing a user namespace must also unshare the thread group
2243 * and unshare the filesystem root and working directories.
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002244 */
2245 if (unshare_flags & CLONE_NEWUSER)
Eric W. Biedermane66eded2013-03-13 11:51:49 -07002246 unshare_flags |= CLONE_THREAD | CLONE_FS;
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002247 /*
Eric W. Biederman50804fe2010-03-02 15:41:50 -08002248 * If unsharing vm, must also unshare signal handlers.
2249 */
2250 if (unshare_flags & CLONE_VM)
2251 unshare_flags |= CLONE_SIGHAND;
Manfred Spraul6013f672008-04-29 01:00:59 -07002252 /*
Eric W. Biederman12c641a2015-08-10 17:35:07 -05002253 * If unsharing a signal handlers, must also unshare the signal queues.
2254 */
2255 if (unshare_flags & CLONE_SIGHAND)
2256 unshare_flags |= CLONE_THREAD;
2257 /*
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002258 * If unsharing namespace, must also unshare filesystem information.
2259 */
2260 if (unshare_flags & CLONE_NEWNS)
2261 unshare_flags |= CLONE_FS;
Eric W. Biederman50804fe2010-03-02 15:41:50 -08002262
2263 err = check_unshare_flags(unshare_flags);
2264 if (err)
2265 goto bad_unshare_out;
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002266 /*
Manfred Spraul6013f672008-04-29 01:00:59 -07002267 * CLONE_NEWIPC must also detach from the undolist: after switching
2268 * to a new ipc namespace, the semaphore arrays from the old
2269 * namespace are unreachable.
2270 */
2271 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
Manfred Spraul9edff4a2008-04-29 01:00:57 -07002272 do_sysvsem = 1;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002273 err = unshare_fs(unshare_flags, &new_fs);
2274 if (err)
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002275 goto bad_unshare_out;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002276 err = unshare_fd(unshare_flags, &new_fd);
2277 if (err)
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002278 goto bad_unshare_cleanup_fs;
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002279 err = unshare_userns(unshare_flags, &new_cred);
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002280 if (err)
Manfred Spraul9edff4a2008-04-29 01:00:57 -07002281 goto bad_unshare_cleanup_fd;
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002282 err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
2283 new_cred, new_fs);
2284 if (err)
2285 goto bad_unshare_cleanup_cred;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002286
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002287 if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) {
Manfred Spraul9edff4a2008-04-29 01:00:57 -07002288 if (do_sysvsem) {
2289 /*
2290 * CLONE_SYSVSEM is equivalent to sys_exit().
2291 */
2292 exit_sem(current);
2293 }
Jack Millerab602f72014-08-08 14:23:19 -07002294 if (unshare_flags & CLONE_NEWIPC) {
2295 /* Orphan segments in old ns (see sem above). */
2296 exit_shm(current);
2297 shm_init_task(current);
2298 }
Serge E. Hallynab516012006-10-02 02:18:06 -07002299
Alan Cox6f977e62013-02-27 17:03:23 -08002300 if (new_nsproxy)
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07002301 switch_task_namespaces(current, new_nsproxy);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002302
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07002303 task_lock(current);
2304
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002305 if (new_fs) {
2306 fs = current->fs;
Nick Piggin2a4419b2010-08-18 04:37:33 +10002307 spin_lock(&fs->lock);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002308 current->fs = new_fs;
Al Viro498052b2009-03-30 07:20:30 -04002309 if (--fs->users)
2310 new_fs = NULL;
2311 else
2312 new_fs = fs;
Nick Piggin2a4419b2010-08-18 04:37:33 +10002313 spin_unlock(&fs->lock);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002314 }
2315
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002316 if (new_fd) {
2317 fd = current->files;
2318 current->files = new_fd;
2319 new_fd = fd;
2320 }
2321
2322 task_unlock(current);
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002323
2324 if (new_cred) {
2325 /* Install the new user namespace */
2326 commit_creds(new_cred);
2327 new_cred = NULL;
2328 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002329 }
2330
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002331bad_unshare_cleanup_cred:
2332 if (new_cred)
2333 put_cred(new_cred);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002334bad_unshare_cleanup_fd:
2335 if (new_fd)
2336 put_files_struct(new_fd);
2337
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002338bad_unshare_cleanup_fs:
2339 if (new_fs)
Al Viro498052b2009-03-30 07:20:30 -04002340 free_fs_struct(new_fs);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002341
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002342bad_unshare_out:
2343 return err;
2344}
Al Viro3b125382008-04-22 05:31:30 -04002345
2346/*
2347 * Helper to unshare the files of the current task.
2348 * We don't want to expose copy_files internals to
2349 * the exec layer of the kernel.
2350 */
2351
2352int unshare_files(struct files_struct **displaced)
2353{
2354 struct task_struct *task = current;
Al Viro50704512008-04-26 05:25:00 +01002355 struct files_struct *copy = NULL;
Al Viro3b125382008-04-22 05:31:30 -04002356 int error;
2357
2358 error = unshare_fd(CLONE_FILES, &copy);
2359 if (error || !copy) {
2360 *displaced = NULL;
2361 return error;
2362 }
2363 *displaced = task->files;
2364 task_lock(task);
2365 task->files = copy;
2366 task_unlock(task);
2367 return 0;
2368}
Heinrich Schuchardt16db3d32015-04-16 12:47:50 -07002369
2370int sysctl_max_threads(struct ctl_table *table, int write,
2371 void __user *buffer, size_t *lenp, loff_t *ppos)
2372{
2373 struct ctl_table t;
2374 int ret;
2375 int threads = max_threads;
2376 int min = MIN_THREADS;
2377 int max = MAX_THREADS;
2378
2379 t = *table;
2380 t.data = &threads;
2381 t.extra1 = &min;
2382 t.extra2 = &max;
2383
2384 ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos);
2385 if (ret || !write)
2386 return ret;
2387
2388 set_max_threads(threads);
2389
2390 return 0;
2391}