blob: bcdf5312521036486b7613e7499d7174e62526c6 [file] [log] [blame]
Thomas Gleixner457c8992019-05-19 13:08:55 +01001// SPDX-License-Identifier: GPL-2.0-only
Linus Torvalds1da177e2005-04-16 15:20:36 -07002/*
3 * linux/kernel/fork.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 */
7
8/*
9 * 'fork.c' contains the help-routines for the 'fork' system call
10 * (see also entry.S and others).
11 * Fork is rather simple, once you get the hang of it, but the memory
12 * management can be a bitch. See 'mm/memory.c': 'copy_page_range()'
13 */
14
Christian Braunerb3e583822019-03-27 13:04:15 +010015#include <linux/anon_inodes.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/slab.h>
Ingo Molnar4eb5aaa2017-02-08 18:51:29 +010017#include <linux/sched/autogroup.h>
Ingo Molnar6e84f312017-02-08 18:51:29 +010018#include <linux/sched/mm.h>
Ingo Molnarf7ccbae2017-02-08 18:51:30 +010019#include <linux/sched/coredump.h>
Ingo Molnar8703e8a2017-02-08 18:51:30 +010020#include <linux/sched/user.h>
Ingo Molnar6a3827d2017-02-08 18:51:31 +010021#include <linux/sched/numa_balancing.h>
Ingo Molnar03441a32017-02-08 18:51:35 +010022#include <linux/sched/stat.h>
Ingo Molnar29930022017-02-08 18:51:36 +010023#include <linux/sched/task.h>
Ingo Molnar68db0cf2017-02-08 18:51:37 +010024#include <linux/sched/task_stack.h>
Ingo Molnar32ef5512017-02-05 11:48:36 +010025#include <linux/sched/cputime.h>
Christian Braunerb3e583822019-03-27 13:04:15 +010026#include <linux/seq_file.h>
Ingo Molnar037741a2017-02-03 10:08:30 +010027#include <linux/rtmutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028#include <linux/init.h>
29#include <linux/unistd.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include <linux/module.h>
31#include <linux/vmalloc.h>
32#include <linux/completion.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <linux/personality.h>
34#include <linux/mempolicy.h>
35#include <linux/sem.h>
36#include <linux/file.h>
Al Viro9f3acc32008-04-24 07:44:08 -040037#include <linux/fdtable.h>
Jens Axboeda9cbc82008-06-30 20:42:08 +020038#include <linux/iocontext.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070039#include <linux/key.h>
40#include <linux/binfmts.h>
41#include <linux/mman.h>
Andrea Arcangelicddb8a52008-07-28 15:46:29 -070042#include <linux/mmu_notifier.h>
Jérôme Glisse133ff0e2017-09-08 16:11:23 -070043#include <linux/hmm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <linux/fs.h>
Davidlohr Bueso615d6e82014-04-07 15:37:25 -070045#include <linux/mm.h>
46#include <linux/vmacache.h>
Serge E. Hallynab516012006-10-02 02:18:06 -070047#include <linux/nsproxy.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080048#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070049#include <linux/cpu.h>
Paul Menageb4f48b62007-10-18 23:39:33 -070050#include <linux/cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070051#include <linux/security.h>
Mel Gormana1e78772008-07-23 21:27:23 -070052#include <linux/hugetlb.h>
Will Drewrye2cfabdf2012-04-12 16:47:57 -050053#include <linux/seccomp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070054#include <linux/swap.h>
55#include <linux/syscalls.h>
56#include <linux/jiffies.h>
57#include <linux/futex.h>
Linus Torvalds8141c7f2008-11-15 10:20:36 -080058#include <linux/compat.h>
Eric Dumazet207205a2011-03-22 16:30:44 -070059#include <linux/kthread.h>
Andrew Morton7c3ab7382006-12-10 02:19:19 -080060#include <linux/task_io_accounting_ops.h>
Dipankar Sarmaab2af1f2005-09-09 13:04:13 -070061#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070062#include <linux/ptrace.h>
63#include <linux/mount.h>
64#include <linux/audit.h>
Pavel Emelianov78fb7462008-02-07 00:13:51 -080065#include <linux/memcontrol.h>
Frederic Weisbeckerf201ae22008-11-23 06:22:56 +010066#include <linux/ftrace.h>
Mike Galbraith5e2bf012012-05-10 13:01:45 -070067#include <linux/proc_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <linux/profile.h>
69#include <linux/rmap.h>
Hugh Dickinsf8af4da2009-09-21 17:01:57 -070070#include <linux/ksm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070071#include <linux/acct.h>
Pavel Emelyanov893e26e2017-02-22 15:42:27 -080072#include <linux/userfaultfd_k.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070073#include <linux/tsacct_kern.h>
Matt Helsley9f460802005-11-07 00:59:16 -080074#include <linux/cn_proc.h>
Rafael J. Wysockiba96a0c2007-05-23 13:57:25 -070075#include <linux/freezer.h>
Shailabh Nagarca74e922006-07-14 00:24:36 -070076#include <linux/delayacct.h>
Shailabh Nagarad4ecbc2006-07-14 00:24:44 -070077#include <linux/taskstats_kern.h>
Arjan van de Ven0a4254052006-09-26 10:52:38 +020078#include <linux/random.h>
Miloslav Trmac522ed772007-07-15 23:40:56 -070079#include <linux/tty.h>
Jens Axboefd0928d2008-01-24 08:52:45 +010080#include <linux/blkdev.h>
Al Viro5ad4e532009-03-29 19:50:06 -040081#include <linux/fs_struct.h>
Eric Sandeen7c9f8862008-04-22 16:38:23 -050082#include <linux/magic.h>
Ingo Molnarcdd6c482009-09-21 12:02:48 +020083#include <linux/perf_event.h>
Stanislaw Gruszka42c4ab42009-07-29 12:15:26 +020084#include <linux/posix-timers.h>
Avi Kivity8e7cac72009-11-29 16:34:48 +020085#include <linux/user-return-notifier.h>
Ying Han3d5992d2010-10-26 14:21:23 -070086#include <linux/oom.h>
Andrea Arcangeliba761492011-01-13 15:46:58 -080087#include <linux/khugepaged.h>
Oleg Nesterovd80e7312012-02-24 20:07:11 +010088#include <linux/signalfd.h>
Srikar Dronamraju0326f5a92012-03-13 23:30:11 +053089#include <linux/uprobes.h>
Kent Overstreeta27bb332013-05-07 16:19:08 -070090#include <linux/aio.h>
Gideon Israel Dsouza52f5684c2014-04-07 15:39:20 -070091#include <linux/compiler.h>
Heinrich Schuchardt16db3d32015-04-16 12:47:50 -070092#include <linux/sysctl.h>
Dmitry Vyukov5c9a8752016-03-22 14:27:30 -070093#include <linux/kcov.h>
Josh Poimboeufd83a7cb2017-02-13 19:42:40 -060094#include <linux/livepatch.h>
Mark Rutland48ac3c12017-07-14 12:23:09 +010095#include <linux/thread_info.h>
Alexander Popovafaef012018-08-17 01:16:58 +030096#include <linux/stackleak.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070097
98#include <asm/pgtable.h>
99#include <asm/pgalloc.h>
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -0800100#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101#include <asm/mmu_context.h>
102#include <asm/cacheflush.h>
103#include <asm/tlbflush.h>
104
Steven Rostedtad8d75f2009-04-14 19:39:12 -0400105#include <trace/events/sched.h>
106
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -0800107#define CREATE_TRACE_POINTS
108#include <trace/events/task.h>
109
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110/*
Heinrich Schuchardtac1b3982015-04-16 12:47:47 -0700111 * Minimum number of threads to boot the kernel
112 */
113#define MIN_THREADS 20
114
115/*
116 * Maximum number of threads
117 */
118#define MAX_THREADS FUTEX_TID_MASK
119
120/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121 * Protected counters by write_lock_irq(&tasklist_lock)
122 */
123unsigned long total_forks; /* Handle normal Linux uptimes. */
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700124int nr_threads; /* The idle threads do not count.. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125
Kefeng Wang8856ae42019-05-31 22:30:12 -0700126static int max_threads; /* tunable limit on nr_threads */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127
Sai Praneeth Prakhya8495f7e2019-09-25 16:47:27 -0700128#define NAMED_ARRAY_INDEX(x) [x] = __stringify(x)
129
130static const char * const resident_page_types[] = {
131 NAMED_ARRAY_INDEX(MM_FILEPAGES),
132 NAMED_ARRAY_INDEX(MM_ANONPAGES),
133 NAMED_ARRAY_INDEX(MM_SWAPENTS),
134 NAMED_ARRAY_INDEX(MM_SHMEMPAGES),
135};
136
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137DEFINE_PER_CPU(unsigned long, process_counts) = 0;
138
Christoph Hellwigc59923a2006-07-10 04:45:40 -0700139__cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
Paul E. McKenneydb1466b2010-03-03 07:46:56 -0800140
141#ifdef CONFIG_PROVE_RCU
142int lockdep_tasklist_lock_is_held(void)
143{
144 return lockdep_is_held(&tasklist_lock);
145}
146EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
147#endif /* #ifdef CONFIG_PROVE_RCU */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148
149int nr_processes(void)
150{
151 int cpu;
152 int total = 0;
153
Ian Campbell1d510752009-11-03 10:11:14 +0000154 for_each_possible_cpu(cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 total += per_cpu(process_counts, cpu);
156
157 return total;
158}
159
Akinobu Mitaf19b9f72012-07-30 14:42:33 -0700160void __weak arch_release_task_struct(struct task_struct *tsk)
161{
162}
163
Thomas Gleixnerf5e10282012-05-05 15:05:48 +0000164#ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
Christoph Lametere18b8902006-12-06 20:33:20 -0800165static struct kmem_cache *task_struct_cachep;
Thomas Gleixner41101802012-05-05 15:05:41 +0000166
167static inline struct task_struct *alloc_task_struct_node(int node)
168{
169 return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node);
170}
171
Thomas Gleixner41101802012-05-05 15:05:41 +0000172static inline void free_task_struct(struct task_struct *tsk)
173{
Thomas Gleixner41101802012-05-05 15:05:41 +0000174 kmem_cache_free(task_struct_cachep, tsk);
175}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700176#endif
177
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700178#ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR
Thomas Gleixner41101802012-05-05 15:05:41 +0000179
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000180/*
181 * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
182 * kmemcache based allocator.
183 */
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700184# if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK)
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700185
186#ifdef CONFIG_VMAP_STACK
187/*
188 * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB
189 * flush. Try to minimize the number of calls by caching stacks.
190 */
191#define NR_CACHED_STACKS 2
192static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]);
Hoeun Ryu19659c52017-05-08 15:56:11 -0700193
194static int free_vm_stack_cache(unsigned int cpu)
195{
196 struct vm_struct **cached_vm_stacks = per_cpu_ptr(cached_stacks, cpu);
197 int i;
198
199 for (i = 0; i < NR_CACHED_STACKS; i++) {
200 struct vm_struct *vm_stack = cached_vm_stacks[i];
201
202 if (!vm_stack)
203 continue;
204
205 vfree(vm_stack->addr);
206 cached_vm_stacks[i] = NULL;
207 }
208
209 return 0;
210}
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700211#endif
212
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700213static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node)
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700214{
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700215#ifdef CONFIG_VMAP_STACK
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700216 void *stack;
217 int i;
218
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700219 for (i = 0; i < NR_CACHED_STACKS; i++) {
Christoph Lameter112166f2017-07-12 14:33:11 -0700220 struct vm_struct *s;
221
222 s = this_cpu_xchg(cached_stacks[i], NULL);
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700223
224 if (!s)
225 continue;
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700226
Konstantin Khlebnikovca182552017-10-13 15:58:22 -0700227 /* Clear stale pointers from reused stack. */
228 memset(s->addr, 0, THREAD_SIZE);
Kees Cooke01e8062018-04-20 14:55:31 -0700229
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700230 tsk->stack_vm_area = s;
Shakeel Buttba4a4572019-01-08 15:22:57 -0800231 tsk->stack = s->addr;
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700232 return s->addr;
233 }
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700234
Roman Gushchin9b6f7e12018-10-26 15:03:19 -0700235 /*
236 * Allocated stacks are cached and later reused by new threads,
237 * so memcg accounting is performed manually on assigning/releasing
238 * stacks to tasks. Drop __GFP_ACCOUNT.
239 */
Mark Rutland48ac3c12017-07-14 12:23:09 +0100240 stack = __vmalloc_node_range(THREAD_SIZE, THREAD_ALIGN,
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700241 VMALLOC_START, VMALLOC_END,
Roman Gushchin9b6f7e12018-10-26 15:03:19 -0700242 THREADINFO_GFP & ~__GFP_ACCOUNT,
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700243 PAGE_KERNEL,
244 0, node, __builtin_return_address(0));
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700245
246 /*
247 * We can't call find_vm_area() in interrupt context, and
248 * free_thread_stack() can be called in interrupt context,
249 * so cache the vm_struct.
250 */
Rik van Riel5eed6f12018-12-21 14:30:54 -0800251 if (stack) {
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700252 tsk->stack_vm_area = find_vm_area(stack);
Rik van Riel5eed6f12018-12-21 14:30:54 -0800253 tsk->stack = stack;
254 }
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700255 return stack;
256#else
Vladimir Davydov49491482016-07-26 15:24:24 -0700257 struct page *page = alloc_pages_node(node, THREADINFO_GFP,
258 THREAD_SIZE_ORDER);
Eric Dumazetb6a84012011-03-22 16:30:42 -0700259
Andrea Arcangeli1bf45802019-06-28 12:07:14 -0700260 if (likely(page)) {
261 tsk->stack = page_address(page);
262 return tsk->stack;
263 }
264 return NULL;
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700265#endif
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700266}
267
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700268static inline void free_thread_stack(struct task_struct *tsk)
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700269{
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700270#ifdef CONFIG_VMAP_STACK
Roman Gushchin9b6f7e12018-10-26 15:03:19 -0700271 struct vm_struct *vm = task_stack_vm_area(tsk);
272
273 if (vm) {
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700274 int i;
275
Roman Gushchin9b6f7e12018-10-26 15:03:19 -0700276 for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
277 mod_memcg_page_state(vm->pages[i],
278 MEMCG_KERNEL_STACK_KB,
279 -(int)(PAGE_SIZE / 1024));
280
281 memcg_kmem_uncharge(vm->pages[i], 0);
282 }
283
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700284 for (i = 0; i < NR_CACHED_STACKS; i++) {
Christoph Lameter112166f2017-07-12 14:33:11 -0700285 if (this_cpu_cmpxchg(cached_stacks[i],
286 NULL, tsk->stack_vm_area) != NULL)
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700287 continue;
288
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700289 return;
290 }
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700291
Andrey Ryabinin0f110a92016-12-12 16:44:14 -0800292 vfree_atomic(tsk->stack);
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700293 return;
294 }
295#endif
296
297 __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER);
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700298}
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000299# else
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700300static struct kmem_cache *thread_stack_cache;
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000301
Michael Ellerman9521d392016-06-25 21:53:30 +1000302static unsigned long *alloc_thread_stack_node(struct task_struct *tsk,
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000303 int node)
304{
Rik van Riel5eed6f12018-12-21 14:30:54 -0800305 unsigned long *stack;
306 stack = kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node);
307 tsk->stack = stack;
308 return stack;
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000309}
310
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700311static void free_thread_stack(struct task_struct *tsk)
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000312{
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700313 kmem_cache_free(thread_stack_cache, tsk->stack);
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000314}
315
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700316void thread_stack_cache_init(void)
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000317{
David Windsorf9d299462017-06-10 22:50:41 -0400318 thread_stack_cache = kmem_cache_create_usercopy("thread_stack",
319 THREAD_SIZE, THREAD_SIZE, 0, 0,
320 THREAD_SIZE, NULL);
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700321 BUG_ON(thread_stack_cache == NULL);
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000322}
323# endif
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700324#endif
325
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326/* SLAB cache for signal_struct structures (tsk->signal) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800327static struct kmem_cache *signal_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328
329/* SLAB cache for sighand_struct structures (tsk->sighand) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800330struct kmem_cache *sighand_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331
332/* SLAB cache for files_struct structures (tsk->files) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800333struct kmem_cache *files_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334
335/* SLAB cache for fs_struct structures (tsk->fs) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800336struct kmem_cache *fs_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337
338/* SLAB cache for vm_area_struct structures */
Linus Torvalds3928d4f2018-07-21 13:48:51 -0700339static struct kmem_cache *vm_area_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340
341/* SLAB cache for mm_struct structures (tsk->mm) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800342static struct kmem_cache *mm_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343
Linus Torvalds490fc052018-07-21 15:24:03 -0700344struct vm_area_struct *vm_area_alloc(struct mm_struct *mm)
Linus Torvalds3928d4f2018-07-21 13:48:51 -0700345{
Andrew Mortona6704682018-08-21 21:53:06 -0700346 struct vm_area_struct *vma;
Linus Torvalds490fc052018-07-21 15:24:03 -0700347
Andrew Mortona6704682018-08-21 21:53:06 -0700348 vma = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
Kirill A. Shutemov027232d2018-07-26 16:37:25 -0700349 if (vma)
350 vma_init(vma, mm);
Linus Torvalds490fc052018-07-21 15:24:03 -0700351 return vma;
Linus Torvalds3928d4f2018-07-21 13:48:51 -0700352}
353
354struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig)
355{
Linus Torvalds95faf692018-07-21 14:48:45 -0700356 struct vm_area_struct *new = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
357
358 if (new) {
359 *new = *orig;
360 INIT_LIST_HEAD(&new->anon_vma_chain);
361 }
362 return new;
Linus Torvalds3928d4f2018-07-21 13:48:51 -0700363}
364
365void vm_area_free(struct vm_area_struct *vma)
366{
367 kmem_cache_free(vm_area_cachep, vma);
368}
369
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700370static void account_kernel_stack(struct task_struct *tsk, int account)
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700371{
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700372 void *stack = task_stack_page(tsk);
373 struct vm_struct *vm = task_stack_vm_area(tsk);
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700374
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700375 BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0);
Andy Lutomirskiefdc9492016-07-28 15:48:17 -0700376
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700377 if (vm) {
378 int i;
379
380 BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE);
381
382 for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
383 mod_zone_page_state(page_zone(vm->pages[i]),
384 NR_KERNEL_STACK_KB,
385 PAGE_SIZE / 1024 * account);
386 }
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700387 } else {
388 /*
389 * All stack pages are in the same zone and belong to the
390 * same memcg.
391 */
392 struct page *first_page = virt_to_page(stack);
393
394 mod_zone_page_state(page_zone(first_page), NR_KERNEL_STACK_KB,
395 THREAD_SIZE / 1024 * account);
396
Johannes Weinered52be72017-07-06 15:40:49 -0700397 mod_memcg_page_state(first_page, MEMCG_KERNEL_STACK_KB,
398 account * (THREAD_SIZE / 1024));
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700399 }
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700400}
401
Roman Gushchin9b6f7e12018-10-26 15:03:19 -0700402static int memcg_charge_kernel_stack(struct task_struct *tsk)
403{
404#ifdef CONFIG_VMAP_STACK
405 struct vm_struct *vm = task_stack_vm_area(tsk);
406 int ret;
407
408 if (vm) {
409 int i;
410
411 for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
412 /*
413 * If memcg_kmem_charge() fails, page->mem_cgroup
414 * pointer is NULL, and both memcg_kmem_uncharge()
415 * and mod_memcg_page_state() in free_thread_stack()
416 * will ignore this page. So it's safe.
417 */
418 ret = memcg_kmem_charge(vm->pages[i], GFP_KERNEL, 0);
419 if (ret)
420 return ret;
421
422 mod_memcg_page_state(vm->pages[i],
423 MEMCG_KERNEL_STACK_KB,
424 PAGE_SIZE / 1024);
425 }
426 }
427#endif
428 return 0;
429}
430
Andy Lutomirski68f24b082016-09-15 22:45:48 -0700431static void release_task_stack(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432{
Andy Lutomirski405c0752016-10-31 08:11:43 -0700433 if (WARN_ON(tsk->state != TASK_DEAD))
434 return; /* Better to leak the stack than to free prematurely */
435
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700436 account_kernel_stack(tsk, -1);
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700437 free_thread_stack(tsk);
Andy Lutomirski68f24b082016-09-15 22:45:48 -0700438 tsk->stack = NULL;
439#ifdef CONFIG_VMAP_STACK
440 tsk->stack_vm_area = NULL;
441#endif
442}
443
444#ifdef CONFIG_THREAD_INFO_IN_TASK
445void put_task_stack(struct task_struct *tsk)
446{
Elena Reshetovaf0b89d32019-01-18 14:27:30 +0200447 if (refcount_dec_and_test(&tsk->stack_refcount))
Andy Lutomirski68f24b082016-09-15 22:45:48 -0700448 release_task_stack(tsk);
449}
450#endif
451
452void free_task(struct task_struct *tsk)
453{
454#ifndef CONFIG_THREAD_INFO_IN_TASK
455 /*
456 * The task is finally done with both the stack and thread_info,
457 * so free both.
458 */
459 release_task_stack(tsk);
460#else
461 /*
462 * If the task had a separate stack allocation, it should be gone
463 * by now.
464 */
Elena Reshetovaf0b89d32019-01-18 14:27:30 +0200465 WARN_ON_ONCE(refcount_read(&tsk->stack_refcount) != 0);
Andy Lutomirski68f24b082016-09-15 22:45:48 -0700466#endif
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700467 rt_mutex_debug_task_free(tsk);
Frederic Weisbeckerfb526072008-11-25 21:07:04 +0100468 ftrace_graph_exit_task(tsk);
Will Drewrye2cfabdf2012-04-12 16:47:57 -0500469 put_seccomp_filter(tsk);
Akinobu Mitaf19b9f72012-07-30 14:42:33 -0700470 arch_release_task_struct(tsk);
Oleg Nesterov1da5c462016-11-29 18:50:57 +0100471 if (tsk->flags & PF_KTHREAD)
472 free_kthread_struct(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 free_task_struct(tsk);
474}
475EXPORT_SYMBOL(free_task);
476
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477#ifdef CONFIG_MMU
Emese Revfy0766f782016-06-20 20:42:34 +0200478static __latent_entropy int dup_mmap(struct mm_struct *mm,
479 struct mm_struct *oldmm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480{
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700481 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 struct rb_node **rb_link, *rb_parent;
483 int retval;
484 unsigned long charge;
Pavel Emelyanov893e26e2017-02-22 15:42:27 -0800485 LIST_HEAD(uf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486
Oleg Nesterov32cdba12012-11-14 19:03:42 +0100487 uprobe_start_dup_mmap();
Michal Hocko7c051262016-05-23 16:25:48 -0700488 if (down_write_killable(&oldmm->mmap_sem)) {
489 retval = -EINTR;
490 goto fail_uprobe_end;
491 }
Ralf Baechleec8c04462006-12-12 17:14:57 +0000492 flush_cache_dup_mm(oldmm);
Oleg Nesterovf8ac4ec2012-08-08 17:11:42 +0200493 uprobe_dup_mmap(oldmm, mm);
Ingo Molnarad339452006-07-03 00:25:15 -0700494 /*
495 * Not linked in yet - no deadlock potential:
496 */
497 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
Hugh Dickins7ee78232005-10-29 18:16:08 -0700498
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -0700499 /* No ordering required: file already has been exposed. */
500 RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
501
Vladimir Davydov4f7d4612014-08-08 14:22:01 -0700502 mm->total_vm = oldmm->total_vm;
Konstantin Khlebnikov84638332016-01-14 15:22:07 -0800503 mm->data_vm = oldmm->data_vm;
Vladimir Davydov4f7d4612014-08-08 14:22:01 -0700504 mm->exec_vm = oldmm->exec_vm;
505 mm->stack_vm = oldmm->stack_vm;
506
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 rb_link = &mm->mm_rb.rb_node;
508 rb_parent = NULL;
509 pprev = &mm->mmap;
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700510 retval = ksm_fork(mm, oldmm);
511 if (retval)
512 goto out;
Andrea Arcangeliba761492011-01-13 15:46:58 -0800513 retval = khugepaged_fork(mm, oldmm);
514 if (retval)
515 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700517 prev = NULL;
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700518 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 struct file *file;
520
521 if (mpnt->vm_flags & VM_DONTCOPY) {
Konstantin Khlebnikov84638332016-01-14 15:22:07 -0800522 vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 continue;
524 }
525 charge = 0;
Tetsuo Handa655c79b2018-06-14 15:26:34 -0700526 /*
527 * Don't duplicate many vmas if we've been oom-killed (for
528 * example)
529 */
530 if (fatal_signal_pending(current)) {
531 retval = -EINTR;
532 goto out;
533 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 if (mpnt->vm_flags & VM_ACCOUNT) {
Huang Shijieb2412b72012-07-30 14:42:30 -0700535 unsigned long len = vma_pages(mpnt);
536
Al Viro191c5422012-02-13 03:58:52 +0000537 if (security_vm_enough_memory_mm(oldmm, len)) /* sic */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 goto fail_nomem;
539 charge = len;
540 }
Linus Torvalds3928d4f2018-07-21 13:48:51 -0700541 tmp = vm_area_dup(mpnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 if (!tmp)
543 goto fail_nomem;
Oleg Nesterovef0855d2013-09-11 14:20:14 -0700544 retval = vma_dup_policy(mpnt, tmp);
545 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 goto fail_nomem_policy;
Andrea Arcangelia247c3a2010-09-22 13:05:12 -0700547 tmp->vm_mm = mm;
Pavel Emelyanov893e26e2017-02-22 15:42:27 -0800548 retval = dup_userfaultfd(tmp, &uf);
549 if (retval)
550 goto fail_nomem_anon_vma_fork;
Rik van Rield2cd9ed2017-09-06 16:25:15 -0700551 if (tmp->vm_flags & VM_WIPEONFORK) {
552 /* VM_WIPEONFORK gets a clean slate in the child. */
553 tmp->anon_vma = NULL;
554 if (anon_vma_prepare(tmp))
555 goto fail_nomem_anon_vma_fork;
556 } else if (anon_vma_fork(tmp, mpnt))
Rik van Riel5beb4932010-03-05 13:42:07 -0800557 goto fail_nomem_anon_vma_fork;
Pavel Emelyanov893e26e2017-02-22 15:42:27 -0800558 tmp->vm_flags &= ~(VM_LOCKED | VM_LOCKONFAULT);
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700559 tmp->vm_next = tmp->vm_prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 file = tmp->vm_file;
561 if (file) {
Al Viro496ad9a2013-01-23 17:07:38 -0500562 struct inode *inode = file_inode(file);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000563 struct address_space *mapping = file->f_mapping;
564
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 get_file(file);
566 if (tmp->vm_flags & VM_DENYWRITE)
567 atomic_dec(&inode->i_writecount);
Davidlohr Bueso83cde9e2014-12-12 16:54:21 -0800568 i_mmap_lock_write(mapping);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000569 if (tmp->vm_flags & VM_SHARED)
David Herrmann4bb5f5d2014-08-08 14:25:25 -0700570 atomic_inc(&mapping->i_mmap_writable);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000571 flush_dcache_mmap_lock(mapping);
572 /* insert tmp into the share list, just after mpnt */
Kirill A. Shutemov27ba0642015-02-10 14:09:59 -0800573 vma_interval_tree_insert_after(tmp, mpnt,
574 &mapping->i_mmap);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000575 flush_dcache_mmap_unlock(mapping);
Davidlohr Bueso83cde9e2014-12-12 16:54:21 -0800576 i_mmap_unlock_write(mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 }
578
579 /*
Mel Gormana1e78772008-07-23 21:27:23 -0700580 * Clear hugetlb-related page reserves for children. This only
581 * affects MAP_PRIVATE mappings. Faults generated by the child
582 * are not guaranteed to succeed, even if read-only
583 */
584 if (is_vm_hugetlb_page(tmp))
585 reset_vma_resv_huge_pages(tmp);
586
587 /*
Hugh Dickins7ee78232005-10-29 18:16:08 -0700588 * Link in the new vma and copy the page table entries.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 *pprev = tmp;
591 pprev = &tmp->vm_next;
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700592 tmp->vm_prev = prev;
593 prev = tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594
595 __vma_link_rb(mm, tmp, rb_link, rb_parent);
596 rb_link = &tmp->vm_rb.rb_right;
597 rb_parent = &tmp->vm_rb;
598
599 mm->map_count++;
Rik van Rield2cd9ed2017-09-06 16:25:15 -0700600 if (!(tmp->vm_flags & VM_WIPEONFORK))
601 retval = copy_page_range(mm, oldmm, mpnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602
603 if (tmp->vm_ops && tmp->vm_ops->open)
604 tmp->vm_ops->open(tmp);
605
606 if (retval)
607 goto out;
608 }
Jeremy Fitzhardinged6dd61c2007-05-02 19:27:14 +0200609 /* a new mm has just been created */
Nadav Amit1ed0cc52018-09-04 15:45:41 -0700610 retval = arch_dup_mmap(oldmm, mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611out:
Hugh Dickins7ee78232005-10-29 18:16:08 -0700612 up_write(&mm->mmap_sem);
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700613 flush_tlb_mm(oldmm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 up_write(&oldmm->mmap_sem);
Pavel Emelyanov893e26e2017-02-22 15:42:27 -0800615 dup_userfaultfd_complete(&uf);
Michal Hocko7c051262016-05-23 16:25:48 -0700616fail_uprobe_end:
Oleg Nesterov32cdba12012-11-14 19:03:42 +0100617 uprobe_end_dup_mmap();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 return retval;
Rik van Riel5beb4932010-03-05 13:42:07 -0800619fail_nomem_anon_vma_fork:
Oleg Nesterovef0855d2013-09-11 14:20:14 -0700620 mpol_put(vma_policy(tmp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621fail_nomem_policy:
Linus Torvalds3928d4f2018-07-21 13:48:51 -0700622 vm_area_free(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623fail_nomem:
624 retval = -ENOMEM;
625 vm_unacct_memory(charge);
626 goto out;
627}
628
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700629static inline int mm_alloc_pgd(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630{
631 mm->pgd = pgd_alloc(mm);
632 if (unlikely(!mm->pgd))
633 return -ENOMEM;
634 return 0;
635}
636
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700637static inline void mm_free_pgd(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638{
Benjamin Herrenschmidt5e541972008-02-04 22:29:14 -0800639 pgd_free(mm, mm->pgd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640}
641#else
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -0700642static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
643{
644 down_write(&oldmm->mmap_sem);
645 RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
646 up_write(&oldmm->mmap_sem);
647 return 0;
648}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649#define mm_alloc_pgd(mm) (0)
650#define mm_free_pgd(mm)
651#endif /* CONFIG_MMU */
652
Andrew Mortond70f2a12018-01-31 16:15:51 -0800653static void check_mm(struct mm_struct *mm)
654{
655 int i;
656
Sai Praneeth Prakhya8495f7e2019-09-25 16:47:27 -0700657 BUILD_BUG_ON_MSG(ARRAY_SIZE(resident_page_types) != NR_MM_COUNTERS,
658 "Please make sure 'struct resident_page_types[]' is updated as well");
659
Andrew Mortond70f2a12018-01-31 16:15:51 -0800660 for (i = 0; i < NR_MM_COUNTERS; i++) {
661 long x = atomic_long_read(&mm->rss_stat.count[i]);
662
663 if (unlikely(x))
Sai Praneeth Prakhya8495f7e2019-09-25 16:47:27 -0700664 pr_alert("BUG: Bad rss-counter state mm:%p type:%s val:%ld\n",
665 mm, resident_page_types[i], x);
Andrew Mortond70f2a12018-01-31 16:15:51 -0800666 }
667
668 if (mm_pgtables_bytes(mm))
669 pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n",
670 mm_pgtables_bytes(mm));
671
672#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
673 VM_BUG_ON_MM(mm->pmd_huge_pte, mm);
674#endif
675}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676
Christoph Lametere94b1762006-12-06 20:33:17 -0800677#define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678#define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
679
Andrew Mortond70f2a12018-01-31 16:15:51 -0800680/*
681 * Called when the last reference to the mm
682 * is dropped: either by a lazy thread or by
683 * mmput. Free the page directory and the mm.
684 */
Andrew Mortond34bc482018-02-21 14:45:17 -0800685void __mmdrop(struct mm_struct *mm)
Andrew Mortond70f2a12018-01-31 16:15:51 -0800686{
687 BUG_ON(mm == &init_mm);
Mark Rutland3eda69c2018-04-05 16:25:12 -0700688 WARN_ON_ONCE(mm == current->mm);
689 WARN_ON_ONCE(mm == current->active_mm);
Andrew Mortond70f2a12018-01-31 16:15:51 -0800690 mm_free_pgd(mm);
691 destroy_context(mm);
Andrew Mortond70f2a12018-01-31 16:15:51 -0800692 mmu_notifier_mm_destroy(mm);
693 check_mm(mm);
694 put_user_ns(mm->user_ns);
695 free_mm(mm);
696}
Andrew Mortond34bc482018-02-21 14:45:17 -0800697EXPORT_SYMBOL_GPL(__mmdrop);
Andrew Mortond70f2a12018-01-31 16:15:51 -0800698
699static void mmdrop_async_fn(struct work_struct *work)
700{
701 struct mm_struct *mm;
702
703 mm = container_of(work, struct mm_struct, async_put_work);
704 __mmdrop(mm);
705}
706
707static void mmdrop_async(struct mm_struct *mm)
708{
709 if (unlikely(atomic_dec_and_test(&mm->mm_count))) {
710 INIT_WORK(&mm->async_put_work, mmdrop_async_fn);
711 schedule_work(&mm->async_put_work);
712 }
713}
714
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715static inline void free_signal_struct(struct signal_struct *sig)
716{
717 taskstats_tgid_free(sig);
718 sched_autogroup_exit(sig);
719 /*
720 * __mmdrop is not safe to call from softirq context on x86 due to
721 * pgd_dtor so postpone it to the async context
722 */
723 if (sig->oom_mm)
724 mmdrop_async(sig->oom_mm);
725 kmem_cache_free(signal_cachep, sig);
726}
727
728static inline void put_signal_struct(struct signal_struct *sig)
729{
Elena Reshetova60d4de32019-01-18 14:27:27 +0200730 if (refcount_dec_and_test(&sig->sigcnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 free_signal_struct(sig);
732}
733
734void __put_task_struct(struct task_struct *tsk)
735{
736 WARN_ON(!tsk->exit_state);
Elena Reshetovaec1d2812019-01-18 14:27:29 +0200737 WARN_ON(refcount_read(&tsk->usage));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 WARN_ON(tsk == current);
739
740 cgroup_free(tsk);
Jann Horn16d51a52019-07-16 17:20:45 +0200741 task_numa_free(tsk, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 security_task_free(tsk);
743 exit_creds(tsk);
744 delayacct_tsk_free(tsk);
745 put_signal_struct(tsk->signal);
746
747 if (!profile_handoff_task(tsk))
748 free_task(tsk);
749}
750EXPORT_SYMBOL_GPL(__put_task_struct);
751
752void __init __weak arch_task_cache_init(void) { }
753
754/*
755 * set_max_threads
756 */
757static void set_max_threads(unsigned int max_threads_suggested)
758{
759 u64 threads;
Arun KSca79b0c2018-12-28 00:34:29 -0800760 unsigned long nr_pages = totalram_pages();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
762 /*
763 * The number of threads shall be limited such that the thread
764 * structures may only consume a small part of the available memory.
765 */
Arun KS3d6357d2018-12-28 00:34:20 -0800766 if (fls64(nr_pages) + fls64(PAGE_SIZE) > 64)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 threads = MAX_THREADS;
768 else
Arun KS3d6357d2018-12-28 00:34:20 -0800769 threads = div64_u64((u64) nr_pages * (u64) PAGE_SIZE,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 (u64) THREAD_SIZE * 8UL);
771
772 if (threads > max_threads_suggested)
773 threads = max_threads_suggested;
774
775 max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS);
776}
777
778#ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
779/* Initialized by the architecture: */
780int arch_task_struct_size __read_mostly;
781#endif
782
Christoph Hellwig4189ff22019-08-12 08:55:24 +0200783#ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
Kees Cook59054292017-08-16 13:00:58 -0700784static void task_struct_whitelist(unsigned long *offset, unsigned long *size)
785{
786 /* Fetch thread_struct whitelist for the architecture. */
787 arch_thread_struct_whitelist(offset, size);
788
789 /*
790 * Handle zero-sized whitelist or empty thread_struct, otherwise
791 * adjust offset to position of thread_struct in task_struct.
792 */
793 if (unlikely(*size == 0))
794 *offset = 0;
795 else
796 *offset += offsetof(struct task_struct, thread);
797}
Christoph Hellwig4189ff22019-08-12 08:55:24 +0200798#endif /* CONFIG_ARCH_TASK_STRUCT_ALLOCATOR */
Kees Cook59054292017-08-16 13:00:58 -0700799
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800void __init fork_init(void)
801{
802 int i;
803#ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
804#ifndef ARCH_MIN_TASKALIGN
805#define ARCH_MIN_TASKALIGN 0
806#endif
807 int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN);
Kees Cook59054292017-08-16 13:00:58 -0700808 unsigned long useroffset, usersize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809
810 /* create a slab on which task_structs can be allocated */
Kees Cook59054292017-08-16 13:00:58 -0700811 task_struct_whitelist(&useroffset, &usersize);
812 task_struct_cachep = kmem_cache_create_usercopy("task_struct",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 arch_task_struct_size, align,
Kees Cook59054292017-08-16 13:00:58 -0700814 SLAB_PANIC|SLAB_ACCOUNT,
815 useroffset, usersize, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816#endif
817
818 /* do the arch specific task caches init */
819 arch_task_cache_init();
820
821 set_max_threads(MAX_THREADS);
822
823 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
824 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
825 init_task.signal->rlim[RLIMIT_SIGPENDING] =
826 init_task.signal->rlim[RLIMIT_NPROC];
827
828 for (i = 0; i < UCOUNT_COUNTS; i++) {
829 init_user_ns.ucount_max[i] = max_threads/2;
830 }
831
832#ifdef CONFIG_VMAP_STACK
833 cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "fork:vm_stack_cache",
834 NULL, free_vm_stack_cache);
835#endif
836
837 lockdep_init_task(&init_task);
Nadav Amitaad42dd2019-04-26 16:22:44 -0700838 uprobes_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839}
840
841int __weak arch_dup_task_struct(struct task_struct *dst,
842 struct task_struct *src)
843{
844 *dst = *src;
845 return 0;
846}
847
848void set_task_stack_end_magic(struct task_struct *tsk)
849{
850 unsigned long *stackend;
851
852 stackend = end_of_stack(tsk);
853 *stackend = STACK_END_MAGIC; /* for overflow detection */
854}
855
856static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
857{
858 struct task_struct *tsk;
859 unsigned long *stack;
YueHaibing0f4991e2018-12-28 00:40:00 -0800860 struct vm_struct *stack_vm_area __maybe_unused;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 int err;
862
863 if (node == NUMA_NO_NODE)
864 node = tsk_fork_get_node(orig);
865 tsk = alloc_task_struct_node(node);
866 if (!tsk)
867 return NULL;
868
869 stack = alloc_thread_stack_node(tsk, node);
870 if (!stack)
871 goto free_tsk;
872
Roman Gushchin9b6f7e12018-10-26 15:03:19 -0700873 if (memcg_charge_kernel_stack(tsk))
874 goto free_stack;
875
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876 stack_vm_area = task_stack_vm_area(tsk);
877
878 err = arch_dup_task_struct(tsk, orig);
879
880 /*
881 * arch_dup_task_struct() clobbers the stack-related fields. Make
882 * sure they're properly initialized before using any stack-related
883 * functions again.
884 */
885 tsk->stack = stack;
886#ifdef CONFIG_VMAP_STACK
887 tsk->stack_vm_area = stack_vm_area;
888#endif
889#ifdef CONFIG_THREAD_INFO_IN_TASK
Elena Reshetovaf0b89d32019-01-18 14:27:30 +0200890 refcount_set(&tsk->stack_refcount, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891#endif
892
893 if (err)
894 goto free_stack;
895
896#ifdef CONFIG_SECCOMP
897 /*
898 * We must handle setting up seccomp filters once we're under
899 * the sighand lock in case orig has changed between now and
900 * then. Until then, filter must be NULL to avoid messing up
901 * the usage counts on the error path calling free_task.
902 */
903 tsk->seccomp.filter = NULL;
904#endif
905
906 setup_thread_stack(tsk, orig);
907 clear_user_return_notifier(tsk);
908 clear_tsk_need_resched(tsk);
909 set_task_stack_end_magic(tsk);
910
Linus Torvalds050e9ba2018-06-14 12:21:18 +0900911#ifdef CONFIG_STACKPROTECTOR
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 tsk->stack_canary = get_random_canary();
913#endif
Sebastian Andrzej Siewior3bd370622019-04-23 16:26:36 +0200914 if (orig->cpus_ptr == &orig->cpus_mask)
915 tsk->cpus_ptr = &tsk->cpus_mask;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916
917 /*
Eric W. Biederman0ff7b2cf2019-09-14 07:33:58 -0500918 * One for the user space visible state that goes away when reaped.
919 * One for the scheduler.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 */
Eric W. Biederman0ff7b2cf2019-09-14 07:33:58 -0500921 refcount_set(&tsk->rcu_users, 2);
922 /* One for the rcu users */
923 refcount_set(&tsk->usage, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924#ifdef CONFIG_BLK_DEV_IO_TRACE
925 tsk->btrace_seq = 0;
926#endif
927 tsk->splice_pipe = NULL;
928 tsk->task_frag.page = NULL;
929 tsk->wake_q.next = NULL;
930
931 account_kernel_stack(tsk, 1);
932
933 kcov_task_init(tsk);
934
935#ifdef CONFIG_FAULT_INJECTION
936 tsk->fail_nth = 0;
937#endif
938
Josef Bacik2c323012018-07-31 12:39:04 -0400939#ifdef CONFIG_BLK_CGROUP
940 tsk->throttle_queue = NULL;
941 tsk->use_memdelay = 0;
942#endif
943
Shakeel Buttd46eb14b2018-08-17 15:46:39 -0700944#ifdef CONFIG_MEMCG
945 tsk->active_memcg = NULL;
946#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 return tsk;
948
949free_stack:
950 free_thread_stack(tsk);
951free_tsk:
952 free_task_struct(tsk);
953 return NULL;
954}
955
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956__cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
957
Hidehiro Kawai4cb0e112009-01-06 14:42:47 -0800958static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
959
960static int __init coredump_filter_setup(char *s)
961{
962 default_dump_filter =
963 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
964 MMF_DUMP_FILTER_MASK;
965 return 1;
966}
967
968__setup("coredump_filter=", coredump_filter_setup);
969
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970#include <linux/init_task.h>
971
Alexey Dobriyan858f0992009-09-23 15:57:32 -0700972static void mm_init_aio(struct mm_struct *mm)
973{
974#ifdef CONFIG_AIO
975 spin_lock_init(&mm->ioctx_lock);
Benjamin LaHaisedb446a02013-07-30 12:54:40 -0400976 mm->ioctx_table = NULL;
Alexey Dobriyan858f0992009-09-23 15:57:32 -0700977#endif
978}
979
Andrea Arcangelic3f3ce02019-05-14 15:40:46 -0700980static __always_inline void mm_clear_owner(struct mm_struct *mm,
981 struct task_struct *p)
982{
983#ifdef CONFIG_MEMCG
984 if (mm->owner == p)
985 WRITE_ONCE(mm->owner, NULL);
986#endif
987}
988
Vladimir Davydov33144e82014-08-08 14:22:03 -0700989static void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
990{
991#ifdef CONFIG_MEMCG
992 mm->owner = p;
993#endif
994}
995
Eric Biggers355627f2017-08-31 16:15:26 -0700996static void mm_init_uprobes_state(struct mm_struct *mm)
997{
998#ifdef CONFIG_UPROBES
999 mm->uprobes_state.xol_area = NULL;
1000#endif
1001}
1002
Eric W. Biedermanbfedb582016-10-13 21:23:16 -05001003static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
1004 struct user_namespace *user_ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005{
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001006 mm->mmap = NULL;
1007 mm->mm_rb = RB_ROOT;
1008 mm->vmacache_seqnum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 atomic_set(&mm->mm_users, 1);
1010 atomic_set(&mm->mm_count, 1);
1011 init_rwsem(&mm->mmap_sem);
1012 INIT_LIST_HEAD(&mm->mmlist);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001013 mm->core_state = NULL;
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -08001014 mm_pgtables_bytes_init(mm);
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001015 mm->map_count = 0;
1016 mm->locked_vm = 0;
Davidlohr Bueso70f8a3c2019-02-06 09:59:15 -08001017 atomic64_set(&mm->pinned_vm, 0);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001018 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 spin_lock_init(&mm->page_table_lock);
Yang Shi88aa7cc2018-06-07 17:05:28 -07001020 spin_lock_init(&mm->arg_lock);
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001021 mm_init_cpumask(mm);
Alexey Dobriyan858f0992009-09-23 15:57:32 -07001022 mm_init_aio(mm);
Balbir Singhcf475ad2008-04-29 01:00:16 -07001023 mm_init_owner(mm, p);
Eric Biggers2b7e8662017-08-25 15:55:43 -07001024 RCU_INIT_POINTER(mm->exe_file, NULL);
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001025 mmu_notifier_mm_init(mm);
Nadav Amit16af97d2017-08-10 15:23:56 -07001026 init_tlb_flush_pending(mm);
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001027#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
1028 mm->pmd_huge_pte = NULL;
1029#endif
Eric Biggers355627f2017-08-31 16:15:26 -07001030 mm_init_uprobes_state(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031
Alex Thorltona0715cc2014-04-07 15:37:10 -07001032 if (current->mm) {
1033 mm->flags = current->mm->flags & MMF_INIT_MASK;
1034 mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK;
1035 } else {
1036 mm->flags = default_dump_filter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 mm->def_flags = 0;
Alex Thorltona0715cc2014-04-07 15:37:10 -07001038 }
1039
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001040 if (mm_alloc_pgd(mm))
1041 goto fail_nopgd;
Pavel Emelianov78fb7462008-02-07 00:13:51 -08001042
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001043 if (init_new_context(p, mm))
1044 goto fail_nocontext;
1045
Eric W. Biedermanbfedb582016-10-13 21:23:16 -05001046 mm->user_ns = get_user_ns(user_ns);
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001047 return mm;
1048
1049fail_nocontext:
1050 mm_free_pgd(mm);
1051fail_nopgd:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 free_mm(mm);
1053 return NULL;
1054}
1055
1056/*
1057 * Allocate and initialize an mm_struct.
1058 */
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001059struct mm_struct *mm_alloc(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060{
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001061 struct mm_struct *mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062
1063 mm = allocate_mm();
KOSAKI Motohirode03c722011-05-24 17:12:15 -07001064 if (!mm)
1065 return NULL;
1066
1067 memset(mm, 0, sizeof(*mm));
Eric W. Biedermanbfedb582016-10-13 21:23:16 -05001068 return mm_init(mm, current, current_user_ns());
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069}
1070
Michal Hockoec8d7c12016-05-20 16:57:21 -07001071static inline void __mmput(struct mm_struct *mm)
1072{
1073 VM_BUG_ON(atomic_read(&mm->mm_users));
1074
1075 uprobe_clear_state(mm);
1076 exit_aio(mm);
1077 ksm_exit(mm);
1078 khugepaged_exit(mm); /* must run before exit_mmap */
1079 exit_mmap(mm);
Aaron Lu6fcb52a2016-10-07 17:00:08 -07001080 mm_put_huge_zero_page(mm);
Michal Hockoec8d7c12016-05-20 16:57:21 -07001081 set_mm_exe_file(mm, NULL);
1082 if (!list_empty(&mm->mmlist)) {
1083 spin_lock(&mmlist_lock);
1084 list_del(&mm->mmlist);
1085 spin_unlock(&mmlist_lock);
1086 }
1087 if (mm->binfmt)
1088 module_put(mm->binfmt->module);
1089 mmdrop(mm);
1090}
1091
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092/*
1093 * Decrement the use count and release all resources for an mm.
1094 */
1095void mmput(struct mm_struct *mm)
1096{
Andrew Morton0ae26f12006-06-23 02:05:15 -07001097 might_sleep();
1098
Michal Hockoec8d7c12016-05-20 16:57:21 -07001099 if (atomic_dec_and_test(&mm->mm_users))
1100 __mmput(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101}
1102EXPORT_SYMBOL_GPL(mmput);
1103
Sherry Yanga1b22892017-10-03 16:15:00 -07001104#ifdef CONFIG_MMU
1105static void mmput_async_fn(struct work_struct *work)
1106{
1107 struct mm_struct *mm = container_of(work, struct mm_struct,
1108 async_put_work);
1109
1110 __mmput(mm);
1111}
1112
1113void mmput_async(struct mm_struct *mm)
1114{
1115 if (atomic_dec_and_test(&mm->mm_users)) {
1116 INIT_WORK(&mm->async_put_work, mmput_async_fn);
1117 schedule_work(&mm->async_put_work);
1118 }
1119}
1120#endif
1121
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -07001122/**
1123 * set_mm_exe_file - change a reference to the mm's executable file
1124 *
1125 * This changes mm's executable file (shown as symlink /proc/[pid]/exe).
1126 *
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001127 * Main users are mmput() and sys_execve(). Callers prevent concurrent
1128 * invocations: in mmput() nobody alive left, in execve task is single
1129 * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the
1130 * mm->exe_file, but does so without using set_mm_exe_file() in order
1131 * to do avoid the need for any locks.
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -07001132 */
Jiri Slaby38646012011-05-26 16:25:46 -07001133void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1134{
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001135 struct file *old_exe_file;
1136
1137 /*
1138 * It is safe to dereference the exe_file without RCU as
1139 * this function is only called if nobody else can access
1140 * this mm -- see comment above for justification.
1141 */
1142 old_exe_file = rcu_dereference_raw(mm->exe_file);
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -07001143
Jiri Slaby38646012011-05-26 16:25:46 -07001144 if (new_exe_file)
1145 get_file(new_exe_file);
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -07001146 rcu_assign_pointer(mm->exe_file, new_exe_file);
1147 if (old_exe_file)
1148 fput(old_exe_file);
Jiri Slaby38646012011-05-26 16:25:46 -07001149}
1150
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -07001151/**
1152 * get_mm_exe_file - acquire a reference to the mm's executable file
1153 *
1154 * Returns %NULL if mm has no associated executable file.
1155 * User must release file via fput().
1156 */
Jiri Slaby38646012011-05-26 16:25:46 -07001157struct file *get_mm_exe_file(struct mm_struct *mm)
1158{
1159 struct file *exe_file;
1160
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -07001161 rcu_read_lock();
1162 exe_file = rcu_dereference(mm->exe_file);
1163 if (exe_file && !get_file_rcu(exe_file))
1164 exe_file = NULL;
1165 rcu_read_unlock();
Jiri Slaby38646012011-05-26 16:25:46 -07001166 return exe_file;
1167}
Davidlohr Bueso11163342015-04-16 12:49:12 -07001168EXPORT_SYMBOL(get_mm_exe_file);
Jiri Slaby38646012011-05-26 16:25:46 -07001169
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170/**
Mateusz Guzikcd81a9172016-08-23 16:20:38 +02001171 * get_task_exe_file - acquire a reference to the task's executable file
1172 *
1173 * Returns %NULL if task's mm (if any) has no associated executable file or
1174 * this is a kernel thread with borrowed mm (see the comment above get_task_mm).
1175 * User must release file via fput().
1176 */
1177struct file *get_task_exe_file(struct task_struct *task)
1178{
1179 struct file *exe_file = NULL;
1180 struct mm_struct *mm;
1181
1182 task_lock(task);
1183 mm = task->mm;
1184 if (mm) {
1185 if (!(task->flags & PF_KTHREAD))
1186 exe_file = get_mm_exe_file(mm);
1187 }
1188 task_unlock(task);
1189 return exe_file;
1190}
1191EXPORT_SYMBOL(get_task_exe_file);
1192
1193/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 * get_task_mm - acquire a reference to the task's mm
1195 *
Oleg Nesterov246bb0b2008-07-25 01:47:38 -07001196 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197 * this kernel workthread has transiently adopted a user mm with use_mm,
1198 * to do its AIO) is not set and if so returns a reference to it, after
1199 * bumping up the use count. User must release the mm via mmput()
1200 * after use. Typically used by /proc and ptrace.
1201 */
1202struct mm_struct *get_task_mm(struct task_struct *task)
1203{
1204 struct mm_struct *mm;
1205
1206 task_lock(task);
1207 mm = task->mm;
1208 if (mm) {
Oleg Nesterov246bb0b2008-07-25 01:47:38 -07001209 if (task->flags & PF_KTHREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 mm = NULL;
1211 else
Vegard Nossum3fce3712017-02-27 14:30:10 -08001212 mmget(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 }
1214 task_unlock(task);
1215 return mm;
1216}
1217EXPORT_SYMBOL_GPL(get_task_mm);
1218
Christopher Yeoh8cdb8782012-02-02 11:34:09 +10301219struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
1220{
1221 struct mm_struct *mm;
1222 int err;
1223
1224 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
1225 if (err)
1226 return ERR_PTR(err);
1227
1228 mm = get_task_mm(task);
1229 if (mm && mm != current->mm &&
1230 !ptrace_may_access(task, mode)) {
1231 mmput(mm);
1232 mm = ERR_PTR(-EACCES);
1233 }
1234 mutex_unlock(&task->signal->cred_guard_mutex);
1235
1236 return mm;
1237}
1238
Oleg Nesterov57b59c42012-03-05 14:59:13 -08001239static void complete_vfork_done(struct task_struct *tsk)
Oleg Nesterovc415c3b2012-03-05 14:59:13 -08001240{
Oleg Nesterovd68b46f2012-03-05 14:59:13 -08001241 struct completion *vfork;
Oleg Nesterovc415c3b2012-03-05 14:59:13 -08001242
Oleg Nesterovd68b46f2012-03-05 14:59:13 -08001243 task_lock(tsk);
1244 vfork = tsk->vfork_done;
1245 if (likely(vfork)) {
1246 tsk->vfork_done = NULL;
1247 complete(vfork);
1248 }
1249 task_unlock(tsk);
1250}
1251
1252static int wait_for_vfork_done(struct task_struct *child,
1253 struct completion *vfork)
1254{
1255 int killed;
1256
1257 freezer_do_not_count();
Roman Gushchin76f969e2019-04-19 10:03:04 -07001258 cgroup_enter_frozen();
Oleg Nesterovd68b46f2012-03-05 14:59:13 -08001259 killed = wait_for_completion_killable(vfork);
Roman Gushchin76f969e2019-04-19 10:03:04 -07001260 cgroup_leave_frozen(false);
Oleg Nesterovd68b46f2012-03-05 14:59:13 -08001261 freezer_count();
1262
1263 if (killed) {
1264 task_lock(child);
1265 child->vfork_done = NULL;
1266 task_unlock(child);
1267 }
1268
1269 put_task_struct(child);
1270 return killed;
Oleg Nesterovc415c3b2012-03-05 14:59:13 -08001271}
1272
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273/* Please note the differences between mmput and mm_release.
1274 * mmput is called whenever we stop holding onto a mm_struct,
1275 * error success whatever.
1276 *
1277 * mm_release is called after a mm_struct has been removed
1278 * from the current process.
1279 *
1280 * This difference is important for error handling, when we
1281 * only half set up a mm_struct for a new process and need to restore
1282 * the old one. Because we mmput the new mm_struct before
1283 * restoring the old one. . .
1284 * Eric Biederman 10 January 1998
1285 */
1286void mm_release(struct task_struct *tsk, struct mm_struct *mm)
1287{
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001288 /* Get rid of any futexes when releasing the mm */
1289#ifdef CONFIG_FUTEX
Peter Zijlstrafc6b1772009-10-05 18:17:32 +02001290 if (unlikely(tsk->robust_list)) {
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001291 exit_robust_list(tsk);
Peter Zijlstrafc6b1772009-10-05 18:17:32 +02001292 tsk->robust_list = NULL;
1293 }
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001294#ifdef CONFIG_COMPAT
Peter Zijlstrafc6b1772009-10-05 18:17:32 +02001295 if (unlikely(tsk->compat_robust_list)) {
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001296 compat_exit_robust_list(tsk);
Peter Zijlstrafc6b1772009-10-05 18:17:32 +02001297 tsk->compat_robust_list = NULL;
1298 }
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001299#endif
Thomas Gleixner322a2c12009-10-05 18:18:03 +02001300 if (unlikely(!list_empty(&tsk->pi_state_list)))
1301 exit_pi_state_list(tsk);
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001302#endif
1303
Srikar Dronamraju0326f5a92012-03-13 23:30:11 +05301304 uprobe_free_utask(tsk);
1305
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306 /* Get rid of any cached register state */
1307 deactivate_mm(tsk, mm);
1308
Roland McGrathfec1d012006-12-06 20:36:34 -08001309 /*
Michal Hocko735f2772016-09-01 16:15:13 -07001310 * Signal userspace if we're not exiting with a core dump
1311 * because we want to leave the value intact for debugging
1312 * purposes.
Roland McGrathfec1d012006-12-06 20:36:34 -08001313 */
Eric Dumazet9c8a8222009-08-06 15:09:28 -07001314 if (tsk->clear_child_tid) {
Michal Hocko735f2772016-09-01 16:15:13 -07001315 if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) &&
Eric Dumazet9c8a8222009-08-06 15:09:28 -07001316 atomic_read(&mm->mm_users) > 1) {
1317 /*
1318 * We don't check the error code - if userspace has
1319 * not set up a proper pointer then tough luck.
1320 */
1321 put_user(0, tsk->clear_child_tid);
Dominik Brodowski2de0db92018-03-11 11:34:26 +01001322 do_futex(tsk->clear_child_tid, FUTEX_WAKE,
1323 1, NULL, NULL, 0, 0);
Eric Dumazet9c8a8222009-08-06 15:09:28 -07001324 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 tsk->clear_child_tid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 }
Konstantin Khlebnikovf7505d642012-05-31 16:26:21 -07001327
1328 /*
1329 * All done, finally we can wake up parent and return this mm to him.
1330 * Also kthread_stop() uses this completion for synchronization.
1331 */
1332 if (tsk->vfork_done)
1333 complete_vfork_done(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334}
1335
Nadav Amit13585fa2019-04-25 17:11:25 -07001336/**
1337 * dup_mm() - duplicates an existing mm structure
1338 * @tsk: the task_struct with which the new mm will be associated.
1339 * @oldmm: the mm to duplicate.
1340 *
1341 * Allocates a new mm structure and duplicates the provided @oldmm structure
1342 * content into it.
1343 *
1344 * Return: the duplicated mm or NULL on failure.
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001345 */
Nadav Amit13585fa2019-04-25 17:11:25 -07001346static struct mm_struct *dup_mm(struct task_struct *tsk,
1347 struct mm_struct *oldmm)
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001348{
Nadav Amit13585fa2019-04-25 17:11:25 -07001349 struct mm_struct *mm;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001350 int err;
1351
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001352 mm = allocate_mm();
1353 if (!mm)
1354 goto fail_nomem;
1355
1356 memcpy(mm, oldmm, sizeof(*mm));
1357
Eric W. Biedermanbfedb582016-10-13 21:23:16 -05001358 if (!mm_init(mm, tsk, mm->user_ns))
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001359 goto fail_nomem;
1360
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001361 err = dup_mmap(mm, oldmm);
1362 if (err)
1363 goto free_pt;
1364
1365 mm->hiwater_rss = get_mm_rss(mm);
1366 mm->hiwater_vm = mm->total_vm;
1367
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001368 if (mm->binfmt && !try_module_get(mm->binfmt->module))
1369 goto free_pt;
1370
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001371 return mm;
1372
1373free_pt:
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001374 /* don't put binfmt in mmput, we haven't got module yet */
1375 mm->binfmt = NULL;
Andrea Arcangelic3f3ce02019-05-14 15:40:46 -07001376 mm_init_owner(mm, NULL);
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001377 mmput(mm);
1378
1379fail_nomem:
1380 return NULL;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001381}
1382
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001383static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384{
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001385 struct mm_struct *mm, *oldmm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 int retval;
1387
1388 tsk->min_flt = tsk->maj_flt = 0;
1389 tsk->nvcsw = tsk->nivcsw = 0;
Mandeep Singh Baines17406b82009-02-06 15:37:47 -08001390#ifdef CONFIG_DETECT_HUNG_TASK
1391 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
Dmitry Vyukova2e51442018-08-21 21:55:52 -07001392 tsk->last_switch_time = 0;
Mandeep Singh Baines17406b82009-02-06 15:37:47 -08001393#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394
1395 tsk->mm = NULL;
1396 tsk->active_mm = NULL;
1397
1398 /*
1399 * Are we cloning a kernel thread?
1400 *
1401 * We need to steal a active VM for that..
1402 */
1403 oldmm = current->mm;
1404 if (!oldmm)
1405 return 0;
1406
Davidlohr Bueso615d6e82014-04-07 15:37:25 -07001407 /* initialize the new vmacache entries */
1408 vmacache_flush(tsk);
1409
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 if (clone_flags & CLONE_VM) {
Vegard Nossum3fce3712017-02-27 14:30:10 -08001411 mmget(oldmm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 mm = oldmm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413 goto good_mm;
1414 }
1415
1416 retval = -ENOMEM;
Nadav Amit13585fa2019-04-25 17:11:25 -07001417 mm = dup_mm(tsk, current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418 if (!mm)
1419 goto fail_nomem;
1420
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421good_mm:
1422 tsk->mm = mm;
1423 tsk->active_mm = mm;
1424 return 0;
1425
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426fail_nomem:
1427 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428}
1429
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001430static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431{
Al Viro498052b2009-03-30 07:20:30 -04001432 struct fs_struct *fs = current->fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433 if (clone_flags & CLONE_FS) {
Al Viro498052b2009-03-30 07:20:30 -04001434 /* tsk->fs is already what we want */
Nick Piggin2a4419b2010-08-18 04:37:33 +10001435 spin_lock(&fs->lock);
Al Viro498052b2009-03-30 07:20:30 -04001436 if (fs->in_exec) {
Nick Piggin2a4419b2010-08-18 04:37:33 +10001437 spin_unlock(&fs->lock);
Al Viro498052b2009-03-30 07:20:30 -04001438 return -EAGAIN;
1439 }
1440 fs->users++;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001441 spin_unlock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 return 0;
1443 }
Al Viro498052b2009-03-30 07:20:30 -04001444 tsk->fs = copy_fs_struct(fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445 if (!tsk->fs)
1446 return -ENOMEM;
1447 return 0;
1448}
1449
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001450static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
JANAK DESAIa016f332006-02-07 12:59:02 -08001451{
1452 struct files_struct *oldf, *newf;
1453 int error = 0;
1454
1455 /*
1456 * A background process may not have any files ...
1457 */
1458 oldf = current->files;
1459 if (!oldf)
1460 goto out;
1461
1462 if (clone_flags & CLONE_FILES) {
1463 atomic_inc(&oldf->count);
1464 goto out;
1465 }
1466
JANAK DESAIa016f332006-02-07 12:59:02 -08001467 newf = dup_fd(oldf, &error);
1468 if (!newf)
1469 goto out;
1470
1471 tsk->files = newf;
1472 error = 0;
1473out:
1474 return error;
1475}
1476
Jens Axboefadad8782008-01-24 08:54:47 +01001477static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
Jens Axboefd0928d2008-01-24 08:52:45 +01001478{
1479#ifdef CONFIG_BLOCK
1480 struct io_context *ioc = current->io_context;
Tejun Heo6e736be2011-12-14 00:33:38 +01001481 struct io_context *new_ioc;
Jens Axboefd0928d2008-01-24 08:52:45 +01001482
1483 if (!ioc)
1484 return 0;
Jens Axboefadad8782008-01-24 08:54:47 +01001485 /*
1486 * Share io context with parent, if CLONE_IO is set
1487 */
1488 if (clone_flags & CLONE_IO) {
Tejun Heo3d487492012-03-05 13:15:25 -08001489 ioc_task_link(ioc);
1490 tsk->io_context = ioc;
Jens Axboefadad8782008-01-24 08:54:47 +01001491 } else if (ioprio_valid(ioc->ioprio)) {
Tejun Heo6e736be2011-12-14 00:33:38 +01001492 new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
1493 if (unlikely(!new_ioc))
Jens Axboefd0928d2008-01-24 08:52:45 +01001494 return -ENOMEM;
1495
Tejun Heo6e736be2011-12-14 00:33:38 +01001496 new_ioc->ioprio = ioc->ioprio;
Tejun Heo11a31222012-02-07 07:51:30 +01001497 put_io_context(new_ioc);
Jens Axboefd0928d2008-01-24 08:52:45 +01001498 }
1499#endif
1500 return 0;
1501}
1502
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001503static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504{
1505 struct sighand_struct *sig;
1506
Zhaolei60348802009-01-06 14:40:46 -08001507 if (clone_flags & CLONE_SIGHAND) {
Elena Reshetovad036bda2019-01-18 14:27:26 +02001508 refcount_inc(&current->sighand->count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 return 0;
1510 }
1511 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
Ingo Molnare56d0902006-01-08 01:01:37 -08001512 rcu_assign_pointer(tsk->sighand, sig);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 if (!sig)
1514 return -ENOMEM;
Peter Zijlstra9d7fb042015-06-30 11:30:54 +02001515
Elena Reshetovad036bda2019-01-18 14:27:26 +02001516 refcount_set(&sig->count, 1);
Jann Horn06e62a42018-08-21 22:00:58 -07001517 spin_lock_irq(&current->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
Jann Horn06e62a42018-08-21 22:00:58 -07001519 spin_unlock_irq(&current->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 return 0;
1521}
1522
Oleg Nesterova7e53282006-03-28 16:11:27 -08001523void __cleanup_sighand(struct sighand_struct *sighand)
Oleg Nesterovc81addc2006-03-28 16:11:17 -08001524{
Elena Reshetovad036bda2019-01-18 14:27:26 +02001525 if (refcount_dec_and_test(&sighand->count)) {
Oleg Nesterovd80e7312012-02-24 20:07:11 +01001526 signalfd_cleanup(sighand);
Oleg Nesterov392809b2014-09-28 23:44:18 +02001527 /*
Paul E. McKenney5f0d5a32017-01-18 02:53:44 -08001528 * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it
Oleg Nesterov392809b2014-09-28 23:44:18 +02001529 * without an RCU grace period, see __lock_task_sighand().
1530 */
Oleg Nesterovc81addc2006-03-28 16:11:17 -08001531 kmem_cache_free(sighand_cachep, sighand);
Oleg Nesterovd80e7312012-02-24 20:07:11 +01001532 }
Oleg Nesterovc81addc2006-03-28 16:11:17 -08001533}
1534
Frank Mayharf06febc2008-09-12 09:54:39 -07001535/*
1536 * Initialize POSIX timer handling for a thread group.
1537 */
1538static void posix_cpu_timers_init_group(struct signal_struct *sig)
1539{
Thomas Gleixner2b699422019-08-21 21:09:04 +02001540 struct posix_cputimers *pct = &sig->posix_cputimers;
Jiri Slaby78d7d402010-03-05 13:42:54 -08001541 unsigned long cpu_limit;
1542
Jason Low316c1608d2015-04-28 13:00:20 -07001543 cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
Thomas Gleixner3a245c02019-08-21 21:09:06 +02001544 posix_cputimers_group_init(pct, cpu_limit);
Frank Mayharf06febc2008-09-12 09:54:39 -07001545}
1546
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001547static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548{
1549 struct signal_struct *sig;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550
Oleg Nesterov4ab6c082009-08-26 14:29:24 -07001551 if (clone_flags & CLONE_THREAD)
Peter Zijlstra490dea42008-11-24 17:06:57 +01001552 return 0;
Oleg Nesterov6279a7512009-03-27 01:06:07 +01001553
Veaceslav Falicoa56704e2010-03-10 15:23:01 -08001554 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 tsk->signal = sig;
1556 if (!sig)
1557 return -ENOMEM;
1558
Oleg Nesterovb3ac0222010-05-26 14:43:24 -07001559 sig->nr_threads = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 atomic_set(&sig->live, 1);
Elena Reshetova60d4de32019-01-18 14:27:27 +02001561 refcount_set(&sig->sigcnt, 1);
Oleg Nesterov0c740d02014-01-21 15:49:56 -08001562
1563 /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
1564 sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node);
1565 tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head);
1566
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 init_waitqueue_head(&sig->wait_chldexit);
Oleg Nesterovdb51aec2008-04-30 00:52:52 -07001568 sig->curr_target = tsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 init_sigpending(&sig->shared_pending);
Eric W. Biedermanc3ad2c32018-07-23 15:20:37 -05001570 INIT_HLIST_HEAD(&sig->multiprocess);
Rik van Riele78c3492014-08-16 13:40:10 -04001571 seqlock_init(&sig->stats_lock);
Peter Zijlstra9d7fb042015-06-30 11:30:54 +02001572 prev_cputime_init(&sig->prev_cputime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573
Nicolas Pitrebaa73d92016-11-11 00:10:10 -05001574#ifdef CONFIG_POSIX_TIMERS
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05001575 INIT_LIST_HEAD(&sig->posix_timers);
Thomas Gleixnerc9cb2e32007-02-16 01:27:49 -08001576 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 sig->real_timer.function = it_real_fn;
Nicolas Pitrebaa73d92016-11-11 00:10:10 -05001578#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 task_lock(current->group_leader);
1581 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
1582 task_unlock(current->group_leader);
1583
Oleg Nesterov6279a7512009-03-27 01:06:07 +01001584 posix_cpu_timers_init_group(sig);
1585
Miloslav Trmac522ed772007-07-15 23:40:56 -07001586 tty_audit_fork(sig);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001587 sched_autogroup_fork(sig);
Miloslav Trmac522ed772007-07-15 23:40:56 -07001588
David Rientjesa63d83f2010-08-09 17:19:46 -07001589 sig->oom_score_adj = current->signal->oom_score_adj;
Mandeep Singh Bainesdabb16f632011-01-13 15:46:05 -08001590 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001591
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001592 mutex_init(&sig->cred_guard_mutex);
1593
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 return 0;
1595}
1596
Kees Cookdbd952122014-06-27 15:18:48 -07001597static void copy_seccomp(struct task_struct *p)
1598{
1599#ifdef CONFIG_SECCOMP
1600 /*
1601 * Must be called with sighand->lock held, which is common to
1602 * all threads in the group. Holding cred_guard_mutex is not
1603 * needed because this new task is not yet running and cannot
1604 * be racing exec.
1605 */
Guenter Roeck69f6a342014-08-10 20:50:30 -07001606 assert_spin_locked(&current->sighand->siglock);
Kees Cookdbd952122014-06-27 15:18:48 -07001607
1608 /* Ref-count the new filter user, and assign it. */
1609 get_seccomp_filter(current);
1610 p->seccomp = current->seccomp;
1611
1612 /*
1613 * Explicitly enable no_new_privs here in case it got set
1614 * between the task_struct being duplicated and holding the
1615 * sighand lock. The seccomp state and nnp must be in sync.
1616 */
1617 if (task_no_new_privs(current))
1618 task_set_no_new_privs(p);
1619
1620 /*
1621 * If the parent gained a seccomp mode after copying thread
1622 * flags and between before we held the sighand lock, we have
1623 * to manually enable the seccomp thread flag here.
1624 */
1625 if (p->seccomp.mode != SECCOMP_MODE_DISABLED)
1626 set_tsk_thread_flag(p, TIF_SECCOMP);
1627#endif
1628}
1629
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001630SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631{
1632 current->clear_child_tid = tidptr;
1633
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001634 return task_pid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635}
1636
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001637static void rt_mutex_init_task(struct task_struct *p)
Ingo Molnar23f78d4a2006-06-27 02:54:53 -07001638{
Thomas Gleixner1d615482009-11-17 14:54:03 +01001639 raw_spin_lock_init(&p->pi_lock);
Zilvinas Valinskase29e1752007-03-16 13:38:34 -08001640#ifdef CONFIG_RT_MUTEXES
Davidlohr Buesoa23ba902017-09-08 16:15:01 -07001641 p->pi_waiters = RB_ROOT_CACHED;
Xunlei Pange96a77052017-03-23 15:56:08 +01001642 p->pi_top_task = NULL;
Ingo Molnar23f78d4a2006-06-27 02:54:53 -07001643 p->pi_blocked_on = NULL;
Ingo Molnar23f78d4a2006-06-27 02:54:53 -07001644#endif
1645}
1646
Eric W. Biederman2c470472017-09-26 13:06:43 -05001647static inline void init_task_pid_links(struct task_struct *task)
1648{
1649 enum pid_type type;
1650
1651 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1652 INIT_HLIST_NODE(&task->pid_links[type]);
1653 }
1654}
1655
Oleg Nesterov81907732013-07-03 15:08:31 -07001656static inline void
1657init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid)
1658{
Eric W. Biederman2c470472017-09-26 13:06:43 -05001659 if (type == PIDTYPE_PID)
1660 task->thread_pid = pid;
1661 else
1662 task->signal->pids[type] = pid;
Oleg Nesterov81907732013-07-03 15:08:31 -07001663}
1664
Ingo Molnar6bfbaa52017-02-03 21:37:49 +01001665static inline void rcu_copy_process(struct task_struct *p)
1666{
1667#ifdef CONFIG_PREEMPT_RCU
1668 p->rcu_read_lock_nesting = 0;
1669 p->rcu_read_unlock_special.s = 0;
1670 p->rcu_blocked_node = NULL;
1671 INIT_LIST_HEAD(&p->rcu_node_entry);
1672#endif /* #ifdef CONFIG_PREEMPT_RCU */
1673#ifdef CONFIG_TASKS_RCU
1674 p->rcu_tasks_holdout = false;
1675 INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
1676 p->rcu_tasks_idle_cpu = -1;
1677#endif /* #ifdef CONFIG_TASKS_RCU */
1678}
1679
Christian Brauner3695eae2019-07-28 00:22:29 +02001680struct pid *pidfd_pid(const struct file *file)
1681{
1682 if (file->f_op == &pidfd_fops)
1683 return file->private_data;
1684
1685 return ERR_PTR(-EBADF);
1686}
1687
Christian Braunerb3e583822019-03-27 13:04:15 +01001688static int pidfd_release(struct inode *inode, struct file *file)
1689{
1690 struct pid *pid = file->private_data;
1691
1692 file->private_data = NULL;
1693 put_pid(pid);
1694 return 0;
1695}
1696
1697#ifdef CONFIG_PROC_FS
1698static void pidfd_show_fdinfo(struct seq_file *m, struct file *f)
1699{
1700 struct pid_namespace *ns = proc_pid_ns(file_inode(m->file));
1701 struct pid *pid = f->private_data;
1702
1703 seq_put_decimal_ull(m, "Pid:\t", pid_nr_ns(pid, ns));
1704 seq_putc(m, '\n');
1705}
1706#endif
1707
Joel Fernandes (Google)b53b0b92019-04-30 12:21:53 -04001708/*
1709 * Poll support for process exit notification.
1710 */
1711static unsigned int pidfd_poll(struct file *file, struct poll_table_struct *pts)
1712{
1713 struct task_struct *task;
1714 struct pid *pid = file->private_data;
1715 int poll_flags = 0;
1716
1717 poll_wait(file, &pid->wait_pidfd, pts);
1718
1719 rcu_read_lock();
1720 task = pid_task(pid, PIDTYPE_PID);
1721 /*
1722 * Inform pollers only when the whole thread group exits.
1723 * If the thread group leader exits before all other threads in the
1724 * group, then poll(2) should block, similar to the wait(2) family.
1725 */
1726 if (!task || (task->exit_state && thread_group_empty(task)))
1727 poll_flags = POLLIN | POLLRDNORM;
1728 rcu_read_unlock();
1729
1730 return poll_flags;
1731}
1732
Christian Braunerb3e583822019-03-27 13:04:15 +01001733const struct file_operations pidfd_fops = {
1734 .release = pidfd_release,
Joel Fernandes (Google)b53b0b92019-04-30 12:21:53 -04001735 .poll = pidfd_poll,
Christian Braunerb3e583822019-03-27 13:04:15 +01001736#ifdef CONFIG_PROC_FS
1737 .show_fdinfo = pidfd_show_fdinfo,
1738#endif
1739};
1740
Andrea Arcangelic3f3ce02019-05-14 15:40:46 -07001741static void __delayed_free_task(struct rcu_head *rhp)
1742{
1743 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
1744
1745 free_task(tsk);
1746}
1747
1748static __always_inline void delayed_free_task(struct task_struct *tsk)
1749{
1750 if (IS_ENABLED(CONFIG_MEMCG))
1751 call_rcu(&tsk->rcu, __delayed_free_task);
1752 else
1753 free_task(tsk);
1754}
1755
Frank Mayharf06febc2008-09-12 09:54:39 -07001756/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 * This creates a new process as a copy of the old one,
1758 * but does not actually start it yet.
1759 *
1760 * It copies the registers, and all the appropriate
1761 * parts of the process environment (as per the clone
1762 * flags). The actual kick-off is left to the caller.
1763 */
Emese Revfy0766f782016-06-20 20:42:34 +02001764static __latent_entropy struct task_struct *copy_process(
Roland McGrath09a05392008-07-25 19:45:47 -07001765 struct pid *pid,
Josh Triplett3033f14a2015-06-25 15:01:19 -07001766 int trace,
Christian Brauner7f192e32019-05-25 11:36:41 +02001767 int node,
1768 struct kernel_clone_args *args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769{
Christian Braunerb3e583822019-03-27 13:04:15 +01001770 int pidfd = -1, retval;
Mariusz Kozlowskia24efe62007-10-18 23:41:09 -07001771 struct task_struct *p;
Eric W. Biedermanc3ad2c32018-07-23 15:20:37 -05001772 struct multiprocess_signals delayed;
Al Viro6fd2fe42019-06-26 22:22:09 -04001773 struct file *pidfile = NULL;
Christian Brauner7f192e32019-05-25 11:36:41 +02001774 u64 clone_flags = args->flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775
Marcos Paulo de Souza667b6092018-02-06 15:39:34 -08001776 /*
1777 * Don't allow sharing the root directory with processes in a different
1778 * namespace
1779 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1781 return ERR_PTR(-EINVAL);
1782
Eric W. Biedermane66eded2013-03-13 11:51:49 -07001783 if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS))
1784 return ERR_PTR(-EINVAL);
1785
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786 /*
1787 * Thread groups must share signals as well, and detached threads
1788 * can only be started up within the thread group.
1789 */
1790 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1791 return ERR_PTR(-EINVAL);
1792
1793 /*
1794 * Shared signal handlers imply shared VM. By way of the above,
1795 * thread groups also imply shared VM. Blocking this case allows
1796 * for various simplifications in other code.
1797 */
1798 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1799 return ERR_PTR(-EINVAL);
1800
Sukadev Bhattiprolu123be072009-09-23 15:57:20 -07001801 /*
1802 * Siblings of global init remain as zombies on exit since they are
1803 * not reaped by their parent (swapper). To solve this and to avoid
1804 * multi-rooted process trees, prevent global and container-inits
1805 * from creating siblings.
1806 */
1807 if ((clone_flags & CLONE_PARENT) &&
1808 current->signal->flags & SIGNAL_UNKILLABLE)
1809 return ERR_PTR(-EINVAL);
1810
Eric W. Biederman8382fca2012-12-20 19:26:06 -08001811 /*
Oleg Nesterov40a0d322013-09-11 14:19:41 -07001812 * If the new process will be in a different pid or user namespace
Eric W. Biedermanfaf00da2015-08-10 18:25:44 -05001813 * do not allow it to share a thread group with the forking task.
Eric W. Biederman8382fca2012-12-20 19:26:06 -08001814 */
Eric W. Biedermanfaf00da2015-08-10 18:25:44 -05001815 if (clone_flags & CLONE_THREAD) {
Oleg Nesterov40a0d322013-09-11 14:19:41 -07001816 if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) ||
1817 (task_active_pid_ns(current) !=
1818 current->nsproxy->pid_ns_for_children))
1819 return ERR_PTR(-EINVAL);
1820 }
Eric W. Biederman8382fca2012-12-20 19:26:06 -08001821
Christian Braunerb3e583822019-03-27 13:04:15 +01001822 if (clone_flags & CLONE_PIDFD) {
Christian Braunerb3e583822019-03-27 13:04:15 +01001823 /*
Christian Braunerb3e583822019-03-27 13:04:15 +01001824 * - CLONE_DETACHED is blocked so that we can potentially
1825 * reuse it later for CLONE_PIDFD.
1826 * - CLONE_THREAD is blocked until someone really needs it.
1827 */
Christian Brauner7f192e32019-05-25 11:36:41 +02001828 if (clone_flags & (CLONE_DETACHED | CLONE_THREAD))
Christian Braunerb3e583822019-03-27 13:04:15 +01001829 return ERR_PTR(-EINVAL);
Christian Braunerb3e583822019-03-27 13:04:15 +01001830 }
1831
Eric W. Biedermanc3ad2c32018-07-23 15:20:37 -05001832 /*
1833 * Force any signals received before this point to be delivered
1834 * before the fork happens. Collect up signals sent to multiple
1835 * processes that happen during the fork and delay them so that
1836 * they appear to happen after the fork.
1837 */
1838 sigemptyset(&delayed.signal);
1839 INIT_HLIST_NODE(&delayed.node);
1840
1841 spin_lock_irq(&current->sighand->siglock);
1842 if (!(clone_flags & CLONE_THREAD))
1843 hlist_add_head(&delayed.node, &current->signal->multiprocess);
1844 recalc_sigpending();
1845 spin_unlock_irq(&current->sighand->siglock);
1846 retval = -ERESTARTNOINTR;
1847 if (signal_pending(current))
1848 goto fork_out;
1849
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850 retval = -ENOMEM;
Andi Kleen725fc622016-05-23 16:24:05 -07001851 p = dup_task_struct(current, node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 if (!p)
1853 goto fork_out;
1854
Vegard Nossum4d6501d2017-05-09 09:39:59 +02001855 /*
1856 * This _must_ happen before we call free_task(), i.e. before we jump
1857 * to any of the bad_fork_* labels. This is to avoid freeing
1858 * p->set_child_tid which is (ab)used as a kthread's data pointer for
1859 * kernel threads (PF_KTHREAD).
1860 */
Christian Brauner7f192e32019-05-25 11:36:41 +02001861 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? args->child_tid : NULL;
Vegard Nossum4d6501d2017-05-09 09:39:59 +02001862 /*
1863 * Clear TID on mm_release()?
1864 */
Christian Brauner7f192e32019-05-25 11:36:41 +02001865 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? args->child_tid : NULL;
Vegard Nossum4d6501d2017-05-09 09:39:59 +02001866
Steven Rostedtf7e8b612009-06-02 16:39:48 -04001867 ftrace_graph_init_task(p);
1868
Peter Zijlstrabea493a2006-10-17 00:10:33 -07001869 rt_mutex_init_task(p);
1870
Ingo Molnard12c1a32008-07-14 12:09:28 +02001871#ifdef CONFIG_PROVE_LOCKING
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001872 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1873 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1874#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 retval = -EAGAIN;
David Howells3b11a1d2008-11-14 10:39:26 +11001876 if (atomic_read(&p->real_cred->user->processes) >=
Jiri Slaby78d7d402010-03-05 13:42:54 -08001877 task_rlimit(p, RLIMIT_NPROC)) {
Eric Parisb57922b2013-07-03 15:08:29 -07001878 if (p->real_cred->user != INIT_USER &&
1879 !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880 goto bad_fork_free;
1881 }
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001882 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883
David Howellsf1752ee2008-11-14 10:39:17 +11001884 retval = copy_creds(p, clone_flags);
1885 if (retval < 0)
1886 goto bad_fork_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887
1888 /*
1889 * If multiple threads are within copy_process(), then this check
1890 * triggers too late. This doesn't hurt, the check is only there
1891 * to stop root fork bombs.
1892 */
Li Zefan04ec93f2009-02-06 08:17:19 +00001893 retval = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894 if (nr_threads >= max_threads)
1895 goto bad_fork_cleanup_count;
1896
Shailabh Nagarca74e922006-07-14 00:24:36 -07001897 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
Peter Zijlstrac1de45c2016-11-28 23:03:05 -08001898 p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE);
David Rientjes514ddb42014-04-07 15:37:27 -07001899 p->flags |= PF_FORKNOEXEC;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 INIT_LIST_HEAD(&p->children);
1901 INIT_LIST_HEAD(&p->sibling);
Paul E. McKenneyf41d9112009-08-22 13:56:52 -07001902 rcu_copy_process(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 p->vfork_done = NULL;
1904 spin_lock_init(&p->alloc_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906 init_sigpending(&p->pending);
1907
Martin Schwidefsky64861632011-12-15 14:56:09 +01001908 p->utime = p->stime = p->gtime = 0;
Stanislaw Gruszka40565b52016-11-15 03:06:51 +01001909#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
Martin Schwidefsky64861632011-12-15 14:56:09 +01001910 p->utimescaled = p->stimescaled = 0;
Stanislaw Gruszka40565b52016-11-15 03:06:51 +01001911#endif
Peter Zijlstra9d7fb042015-06-30 11:30:54 +02001912 prev_cputime_init(&p->prev_cputime);
1913
Frederic Weisbecker6a616712012-12-16 20:00:34 +01001914#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
Frederic Weisbeckerbac5b6b2017-06-29 19:15:10 +02001915 seqcount_init(&p->vtime.seqcount);
1916 p->vtime.starttime = 0;
1917 p->vtime.state = VTIME_INACTIVE;
Frederic Weisbecker6a616712012-12-16 20:00:34 +01001918#endif
1919
KAMEZAWA Hiroyukia3a2e762010-04-06 14:34:42 -07001920#if defined(SPLIT_RSS_COUNTING)
1921 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1922#endif
Balbir Singh172ba842007-07-09 18:52:00 +02001923
Arjan van de Ven69766752008-09-01 15:52:40 -07001924 p->default_timer_slack_ns = current->timer_slack_ns;
1925
Johannes Weinereb414682018-10-26 15:06:27 -07001926#ifdef CONFIG_PSI
1927 p->psi_flags = 0;
1928#endif
1929
Andrea Righi59954772008-07-27 17:29:15 +02001930 task_io_accounting_init(&p->ioac);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931 acct_clear_integrals(p);
1932
Thomas Gleixner3a245c02019-08-21 21:09:06 +02001933 posix_cputimers_init(&p->posix_cputimers);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 p->io_context = NULL;
Richard Guy Briggsc0b0ae82018-05-12 21:58:21 -04001936 audit_set_context(p, NULL);
Paul Menageb4f48b62007-10-18 23:39:33 -07001937 cgroup_fork(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938#ifdef CONFIG_NUMA
Lee Schermerhorn846a16b2008-04-28 02:13:09 -07001939 p->mempolicy = mpol_dup(p->mempolicy);
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001940 if (IS_ERR(p->mempolicy)) {
1941 retval = PTR_ERR(p->mempolicy);
1942 p->mempolicy = NULL;
Li Zefane8604cb2014-03-28 15:18:27 +08001943 goto bad_fork_cleanup_threadgroup_lock;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001944 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945#endif
Michal Hocko778d3b02011-07-26 16:08:30 -07001946#ifdef CONFIG_CPUSETS
1947 p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1948 p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
Mel Gormancc9a6c82012-03-21 16:34:11 -07001949 seqcount_init(&p->mems_allowed_seq);
Michal Hocko778d3b02011-07-26 16:08:30 -07001950#endif
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001951#ifdef CONFIG_TRACE_IRQFLAGS
1952 p->irq_events = 0;
1953 p->hardirqs_enabled = 0;
1954 p->hardirq_enable_ip = 0;
1955 p->hardirq_enable_event = 0;
1956 p->hardirq_disable_ip = _THIS_IP_;
1957 p->hardirq_disable_event = 0;
1958 p->softirqs_enabled = 1;
1959 p->softirq_enable_ip = _THIS_IP_;
1960 p->softirq_enable_event = 0;
1961 p->softirq_disable_ip = 0;
1962 p->softirq_disable_event = 0;
1963 p->hardirq_context = 0;
1964 p->softirq_context = 0;
1965#endif
David Hildenbrand8bcbde52015-05-11 17:52:06 +02001966
1967 p->pagefault_disabled = 0;
1968
Ingo Molnarfbb9ce952006-07-03 00:24:50 -07001969#ifdef CONFIG_LOCKDEP
Byungchul Parkb09be672017-08-07 16:12:52 +09001970 lockdep_init_task(p);
Ingo Molnarfbb9ce952006-07-03 00:24:50 -07001971#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972
Ingo Molnar408894e2006-01-09 15:59:20 -08001973#ifdef CONFIG_DEBUG_MUTEXES
1974 p->blocked_on = NULL; /* not blocked yet */
1975#endif
Kent Overstreetcafe5632013-03-23 16:11:31 -07001976#ifdef CONFIG_BCACHE
1977 p->sequential_io = 0;
1978 p->sequential_io_avg = 0;
1979#endif
Markus Metzger0f481402009-04-03 16:43:48 +02001980
Srivatsa Vaddagiri3c90e6e2007-11-09 22:39:39 +01001981 /* Perform scheduler related setup. Assign this task to a CPU. */
Dario Faggioliaab03e02013-11-28 11:14:43 +01001982 retval = sched_fork(clone_flags, p);
1983 if (retval)
1984 goto bad_fork_cleanup_policy;
Peter Zijlstra6ab423e2009-05-25 14:45:27 +02001985
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001986 retval = perf_event_init_task(p);
Peter Zijlstra6ab423e2009-05-25 14:45:27 +02001987 if (retval)
1988 goto bad_fork_cleanup_policy;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001989 retval = audit_alloc(p);
1990 if (retval)
Peter Zijlstra6c72e3502014-10-02 16:17:02 -07001991 goto bad_fork_cleanup_perf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 /* copy all the process information */
Jack Millerab602f72014-08-08 14:23:19 -07001993 shm_init_task(p);
Tetsuo Handae4e55b42017-03-24 20:46:33 +09001994 retval = security_task_alloc(p, clone_flags);
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001995 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 goto bad_fork_cleanup_audit;
Tetsuo Handae4e55b42017-03-24 20:46:33 +09001997 retval = copy_semundo(clone_flags, p);
1998 if (retval)
1999 goto bad_fork_cleanup_security;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002000 retval = copy_files(clone_flags, p);
2001 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 goto bad_fork_cleanup_semundo;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002003 retval = copy_fs(clone_flags, p);
2004 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005 goto bad_fork_cleanup_files;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002006 retval = copy_sighand(clone_flags, p);
2007 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002008 goto bad_fork_cleanup_fs;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002009 retval = copy_signal(clone_flags, p);
2010 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 goto bad_fork_cleanup_sighand;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002012 retval = copy_mm(clone_flags, p);
2013 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014 goto bad_fork_cleanup_signal;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002015 retval = copy_namespaces(clone_flags, p);
2016 if (retval)
David Howellsd84f4f92008-11-14 10:39:23 +11002017 goto bad_fork_cleanup_mm;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002018 retval = copy_io(clone_flags, p);
2019 if (retval)
Jens Axboefd0928d2008-01-24 08:52:45 +01002020 goto bad_fork_cleanup_namespaces;
Christian Brauner7f192e32019-05-25 11:36:41 +02002021 retval = copy_thread_tls(clone_flags, args->stack, args->stack_size, p,
2022 args->tls);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023 if (retval)
Jens Axboefd0928d2008-01-24 08:52:45 +01002024 goto bad_fork_cleanup_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025
Alexander Popovafaef012018-08-17 01:16:58 +03002026 stackleak_task_init(p);
2027
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07002028 if (pid != &init_struct_pid) {
Andy Lutomirskic2b1df22013-08-22 11:39:16 -07002029 pid = alloc_pid(p->nsproxy->pid_ns_for_children);
Michal Hocko35f71bc2015-04-16 12:47:38 -07002030 if (IS_ERR(pid)) {
2031 retval = PTR_ERR(pid);
Jiri Slaby0740aa52016-05-20 17:00:25 -07002032 goto bad_fork_cleanup_thread;
Michal Hocko35f71bc2015-04-16 12:47:38 -07002033 }
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07002034 }
2035
Christian Braunerb3e583822019-03-27 13:04:15 +01002036 /*
2037 * This has to happen after we've potentially unshared the file
2038 * descriptor table (so that the pidfd doesn't leak into the child
2039 * if the fd table isn't shared).
2040 */
2041 if (clone_flags & CLONE_PIDFD) {
Al Viro6fd2fe42019-06-26 22:22:09 -04002042 retval = get_unused_fd_flags(O_RDWR | O_CLOEXEC);
Christian Braunerb3e583822019-03-27 13:04:15 +01002043 if (retval < 0)
2044 goto bad_fork_free_pid;
2045
2046 pidfd = retval;
Al Viro6fd2fe42019-06-26 22:22:09 -04002047
2048 pidfile = anon_inode_getfile("[pidfd]", &pidfd_fops, pid,
2049 O_RDWR | O_CLOEXEC);
2050 if (IS_ERR(pidfile)) {
2051 put_unused_fd(pidfd);
Christian Brauner28dd29c2019-07-01 16:01:46 +02002052 retval = PTR_ERR(pidfile);
Al Viro6fd2fe42019-06-26 22:22:09 -04002053 goto bad_fork_free_pid;
2054 }
2055 get_pid(pid); /* held by pidfile now */
2056
Christian Brauner7f192e32019-05-25 11:36:41 +02002057 retval = put_user(pidfd, args->pidfd);
Christian Braunerb3e583822019-03-27 13:04:15 +01002058 if (retval)
2059 goto bad_fork_put_pidfd;
2060 }
2061
Jens Axboe73c10102011-03-08 13:19:51 +01002062#ifdef CONFIG_BLOCK
2063 p->plug = NULL;
2064#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07002065#ifdef CONFIG_FUTEX
Ingo Molnar8f17d3a2006-03-27 01:16:27 -08002066 p->robust_list = NULL;
2067#ifdef CONFIG_COMPAT
2068 p->compat_robust_list = NULL;
2069#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07002070 INIT_LIST_HEAD(&p->pi_state_list);
2071 p->pi_state_cache = NULL;
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07002072#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073 /*
GOTO Masanorif9a38792006-03-13 21:20:44 -08002074 * sigaltstack should be cleared when sharing the same VM
2075 */
2076 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
Stas Sergeev2a742132016-04-14 23:20:04 +03002077 sas_ss_reset(p);
GOTO Masanorif9a38792006-03-13 21:20:44 -08002078
2079 /*
Oleg Nesterov65808072009-12-15 16:47:16 -08002080 * Syscall tracing and stepping should be turned off in the
2081 * child regardless of CLONE_PTRACE.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082 */
Oleg Nesterov65808072009-12-15 16:47:16 -08002083 user_disable_single_step(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
Laurent Viviered75e8d2005-09-03 15:57:18 -07002085#ifdef TIF_SYSCALL_EMU
2086 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
2087#endif
Lin Fenge02c9b02019-05-14 15:42:34 -07002088 clear_tsk_latency_tracing(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090 /* ok, now we should be set up.. */
Oleg Nesterov18c830d2013-07-03 15:08:32 -07002091 p->pid = pid_nr(pid);
2092 if (clone_flags & CLONE_THREAD) {
Oleg Nesterov5f8aadd2012-03-14 19:55:38 +01002093 p->exit_signal = -1;
Oleg Nesterov18c830d2013-07-03 15:08:32 -07002094 p->group_leader = current->group_leader;
2095 p->tgid = current->tgid;
2096 } else {
2097 if (clone_flags & CLONE_PARENT)
2098 p->exit_signal = current->group_leader->exit_signal;
2099 else
Christian Brauner7f192e32019-05-25 11:36:41 +02002100 p->exit_signal = args->exit_signal;
Oleg Nesterov18c830d2013-07-03 15:08:32 -07002101 p->group_leader = p;
2102 p->tgid = p->pid;
2103 }
Oleg Nesterov5f8aadd2012-03-14 19:55:38 +01002104
Wu Fengguang9d823e82011-06-11 18:10:12 -06002105 p->nr_dirtied = 0;
2106 p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
Wu Fengguang83712352011-06-11 19:25:42 -06002107 p->dirty_paused_when = 0;
Wu Fengguang9d823e82011-06-11 18:10:12 -06002108
Oleg Nesterovbb8cbbf2013-11-13 15:36:12 +01002109 p->pdeath_signal = 0;
Oleg Nesterov47e65322006-03-28 16:11:25 -08002110 INIT_LIST_HEAD(&p->thread_group);
Al Viro158e1642012-06-27 09:24:13 +04002111 p->task_works = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112
Ingo Molnar780de9d2017-02-02 11:50:56 +01002113 cgroup_threadgroup_change_begin(current);
Oleg Nesterov18c830d2013-07-03 15:08:32 -07002114 /*
Aleksa Sarai7e476822015-06-09 21:32:09 +10002115 * Ensure that the cgroup subsystem policies allow the new process to be
2116 * forked. It should be noted the the new process's css_set can be changed
2117 * between here and cgroup_post_fork() if an organisation operation is in
2118 * progress.
2119 */
Oleg Nesterovb53202e2015-12-03 10:24:08 -05002120 retval = cgroup_can_fork(p);
Aleksa Sarai7e476822015-06-09 21:32:09 +10002121 if (retval)
Christian Braunerc3b71122019-05-10 11:53:46 +02002122 goto bad_fork_cgroup_threadgroup_change_end;
Aleksa Sarai7e476822015-06-09 21:32:09 +10002123
2124 /*
David Herrmann7b558512019-01-08 13:58:52 +01002125 * From this point on we must avoid any synchronous user-space
2126 * communication until we take the tasklist-lock. In particular, we do
2127 * not want user-space to be able to predict the process start-time by
2128 * stalling fork(2) after we recorded the start_time but before it is
2129 * visible to the system.
2130 */
2131
2132 p->start_time = ktime_get_ns();
Jason A. Donenfeld9285ec42019-06-21 22:32:48 +02002133 p->real_start_time = ktime_get_boottime_ns();
David Herrmann7b558512019-01-08 13:58:52 +01002134
2135 /*
Oleg Nesterov18c830d2013-07-03 15:08:32 -07002136 * Make it visible to the rest of the system, but dont wake it up yet.
2137 * Need tasklist lock for parent etc handling!
2138 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002139 write_lock_irq(&tasklist_lock);
2140
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141 /* CLONE_PARENT re-uses the old parent */
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01002142 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143 p->real_parent = current->real_parent;
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01002144 p->parent_exec_id = current->parent_exec_id;
2145 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146 p->real_parent = current;
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01002147 p->parent_exec_id = current->self_exec_id;
2148 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149
Josh Poimboeufd83a7cb2017-02-13 19:42:40 -06002150 klp_copy_process(p);
2151
Oleg Nesterov3f17da62006-02-15 22:13:24 +03002152 spin_lock(&current->sighand->siglock);
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08002153
2154 /*
Kees Cookdbd952122014-06-27 15:18:48 -07002155 * Copy seccomp details explicitly here, in case they were changed
2156 * before holding sighand lock.
2157 */
2158 copy_seccomp(p);
2159
Mathieu Desnoyersd7822b12018-06-02 08:43:54 -04002160 rseq_fork(p, clone_flags);
2161
Eric W. Biederman4ca1d3e2018-07-13 15:30:33 -05002162 /* Don't start children in a dying pid namespace */
Gargi Sharmae8cfbc22017-11-17 15:30:34 -08002163 if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) {
Kirill Tkhai3fd37222017-05-12 19:11:31 +03002164 retval = -ENOMEM;
2165 goto bad_fork_cancel_cgroup;
2166 }
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08002167
Eric W. Biederman7673bf52018-07-23 08:01:10 -05002168 /* Let kill terminate clone/fork in the middle */
2169 if (fatal_signal_pending(current)) {
2170 retval = -EINTR;
2171 goto bad_fork_cancel_cgroup;
2172 }
2173
Al Viro6fd2fe42019-06-26 22:22:09 -04002174 /* past the last point of failure */
2175 if (pidfile)
2176 fd_install(pidfd, pidfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177
Eric W. Biederman2c470472017-09-26 13:06:43 -05002178 init_task_pid_links(p);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08002179 if (likely(p->pid)) {
Tejun Heo4b9d33e2011-06-17 16:50:38 +02002180 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181
Oleg Nesterov81907732013-07-03 15:08:31 -07002182 init_task_pid(p, PIDTYPE_PID, pid);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08002183 if (thread_group_leader(p)) {
Eric W. Biederman6883f812017-06-04 04:32:13 -05002184 init_task_pid(p, PIDTYPE_TGID, pid);
Oleg Nesterov81907732013-07-03 15:08:31 -07002185 init_task_pid(p, PIDTYPE_PGID, task_pgrp(current));
2186 init_task_pid(p, PIDTYPE_SID, task_session(current));
2187
Eric W. Biederman1c4042c2010-07-12 17:10:36 -07002188 if (is_child_reaper(pid)) {
Eric W. Biederman17cf22c2010-03-02 14:51:53 -08002189 ns_of_pid(pid)->child_reaper = p;
Eric W. Biederman1c4042c2010-07-12 17:10:36 -07002190 p->signal->flags |= SIGNAL_UNKILLABLE;
2191 }
Eric W. Biedermanc3ad2c32018-07-23 15:20:37 -05002192 p->signal->shared_pending.signal = delayed.signal;
Alan Cox9c9f4de2008-10-13 10:37:26 +01002193 p->signal->tty = tty_kref_get(current->signal->tty);
Pavel Tikhomirov749860c2017-01-30 18:06:12 +03002194 /*
2195 * Inherit has_child_subreaper flag under the same
2196 * tasklist_lock with adding child to the process tree
2197 * for propagate_has_child_subreaper optimization.
2198 */
2199 p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper ||
2200 p->real_parent->signal->is_child_subreaper;
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08002201 list_add_tail(&p->sibling, &p->real_parent->children);
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -07002202 list_add_tail_rcu(&p->tasks, &init_task.tasks);
Eric W. Biederman6883f812017-06-04 04:32:13 -05002203 attach_pid(p, PIDTYPE_TGID);
Oleg Nesterov81907732013-07-03 15:08:31 -07002204 attach_pid(p, PIDTYPE_PGID);
2205 attach_pid(p, PIDTYPE_SID);
Christoph Lameter909ea962010-12-08 16:22:55 +01002206 __this_cpu_inc(process_counts);
Oleg Nesterov80628ca2013-07-03 15:08:30 -07002207 } else {
2208 current->signal->nr_threads++;
2209 atomic_inc(&current->signal->live);
Elena Reshetova60d4de32019-01-18 14:27:27 +02002210 refcount_inc(&current->signal->sigcnt);
Eric W. Biederman924de3b2018-07-23 13:38:00 -05002211 task_join_group_stop(p);
Oleg Nesterov80628ca2013-07-03 15:08:30 -07002212 list_add_tail_rcu(&p->thread_group,
2213 &p->group_leader->thread_group);
Oleg Nesterov0c740d02014-01-21 15:49:56 -08002214 list_add_tail_rcu(&p->thread_node,
2215 &p->signal->thread_head);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08002216 }
Oleg Nesterov81907732013-07-03 15:08:31 -07002217 attach_pid(p, PIDTYPE_PID);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08002218 nr_threads++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220 total_forks++;
Eric W. Biedermanc3ad2c32018-07-23 15:20:37 -05002221 hlist_del_init(&delayed.node);
Oleg Nesterov3f17da62006-02-15 22:13:24 +03002222 spin_unlock(&current->sighand->siglock);
Oleg Nesterov4af42062014-04-13 20:58:54 +02002223 syscall_tracepoint_update(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002224 write_unlock_irq(&tasklist_lock);
Oleg Nesterov4af42062014-04-13 20:58:54 +02002225
Andrew Mortonc13cf852005-11-28 13:43:48 -08002226 proc_fork_connector(p);
Oleg Nesterovb53202e2015-12-03 10:24:08 -05002227 cgroup_post_fork(p);
Ingo Molnar780de9d2017-02-02 11:50:56 +01002228 cgroup_threadgroup_change_end(current);
Ingo Molnarcdd6c482009-09-21 12:02:48 +02002229 perf_event_fork(p);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08002230
2231 trace_task_newtask(p, clone_flags);
Oleg Nesterov3ab67962013-10-16 19:39:37 +02002232 uprobe_copy_process(p, clone_flags);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08002233
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234 return p;
2235
Aleksa Sarai7e476822015-06-09 21:32:09 +10002236bad_fork_cancel_cgroup:
Kirill Tkhai3fd37222017-05-12 19:11:31 +03002237 spin_unlock(&current->sighand->siglock);
2238 write_unlock_irq(&tasklist_lock);
Oleg Nesterovb53202e2015-12-03 10:24:08 -05002239 cgroup_cancel_fork(p);
Christian Braunerc3b71122019-05-10 11:53:46 +02002240bad_fork_cgroup_threadgroup_change_end:
2241 cgroup_threadgroup_change_end(current);
Christian Braunerb3e583822019-03-27 13:04:15 +01002242bad_fork_put_pidfd:
Al Viro6fd2fe42019-06-26 22:22:09 -04002243 if (clone_flags & CLONE_PIDFD) {
2244 fput(pidfile);
2245 put_unused_fd(pidfd);
2246 }
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07002247bad_fork_free_pid:
2248 if (pid != &init_struct_pid)
2249 free_pid(pid);
Jiri Slaby0740aa52016-05-20 17:00:25 -07002250bad_fork_cleanup_thread:
2251 exit_thread(p);
Jens Axboefd0928d2008-01-24 08:52:45 +01002252bad_fork_cleanup_io:
Louis Rillingb69f2292009-12-04 14:52:42 +01002253 if (p->io_context)
2254 exit_io_context(p);
Serge E. Hallynab516012006-10-02 02:18:06 -07002255bad_fork_cleanup_namespaces:
Linus Torvalds444f3782007-01-30 13:35:18 -08002256 exit_task_namespaces(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002257bad_fork_cleanup_mm:
Andrea Arcangelic3f3ce02019-05-14 15:40:46 -07002258 if (p->mm) {
2259 mm_clear_owner(p->mm, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002260 mmput(p->mm);
Andrea Arcangelic3f3ce02019-05-14 15:40:46 -07002261 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002262bad_fork_cleanup_signal:
Oleg Nesterov4ab6c082009-08-26 14:29:24 -07002263 if (!(clone_flags & CLONE_THREAD))
Mike Galbraith1c5354d2011-01-05 11:16:04 +01002264 free_signal_struct(p->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002265bad_fork_cleanup_sighand:
Oleg Nesterova7e53282006-03-28 16:11:27 -08002266 __cleanup_sighand(p->sighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002267bad_fork_cleanup_fs:
2268 exit_fs(p); /* blocking */
2269bad_fork_cleanup_files:
2270 exit_files(p); /* blocking */
2271bad_fork_cleanup_semundo:
2272 exit_sem(p);
Tetsuo Handae4e55b42017-03-24 20:46:33 +09002273bad_fork_cleanup_security:
2274 security_task_free(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002275bad_fork_cleanup_audit:
2276 audit_free(p);
Peter Zijlstra6c72e3502014-10-02 16:17:02 -07002277bad_fork_cleanup_perf:
Ingo Molnarcdd6c482009-09-21 12:02:48 +02002278 perf_event_free_task(p);
Peter Zijlstra6c72e3502014-10-02 16:17:02 -07002279bad_fork_cleanup_policy:
Byungchul Parkb09be672017-08-07 16:12:52 +09002280 lockdep_free_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281#ifdef CONFIG_NUMA
Lee Schermerhornf0be3d32008-04-28 02:13:08 -07002282 mpol_put(p->mempolicy);
Li Zefane8604cb2014-03-28 15:18:27 +08002283bad_fork_cleanup_threadgroup_lock:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284#endif
Shailabh Nagar35df17c2006-08-31 21:27:38 -07002285 delayacct_tsk_free(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286bad_fork_cleanup_count:
David Howellsd84f4f92008-11-14 10:39:23 +11002287 atomic_dec(&p->cred->user->processes);
David Howellse0e81732009-09-02 09:13:40 +01002288 exit_creds(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002289bad_fork_free:
Andy Lutomirski405c0752016-10-31 08:11:43 -07002290 p->state = TASK_DEAD;
Andy Lutomirski68f24b082016-09-15 22:45:48 -07002291 put_task_stack(p);
Andrea Arcangelic3f3ce02019-05-14 15:40:46 -07002292 delayed_free_task(p);
Oleg Nesterovfe7d37d2006-01-08 01:04:02 -08002293fork_out:
Eric W. Biedermanc3ad2c32018-07-23 15:20:37 -05002294 spin_lock_irq(&current->sighand->siglock);
2295 hlist_del_init(&delayed.node);
2296 spin_unlock_irq(&current->sighand->siglock);
Oleg Nesterovfe7d37d2006-01-08 01:04:02 -08002297 return ERR_PTR(retval);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298}
2299
Eric W. Biederman2c470472017-09-26 13:06:43 -05002300static inline void init_idle_pids(struct task_struct *idle)
Oleg Nesterovf106eee2010-05-26 14:44:11 -07002301{
2302 enum pid_type type;
2303
2304 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
Eric W. Biederman2c470472017-09-26 13:06:43 -05002305 INIT_HLIST_NODE(&idle->pid_links[type]); /* not really needed */
2306 init_task_pid(idle, type, &init_struct_pid);
Oleg Nesterovf106eee2010-05-26 14:44:11 -07002307 }
2308}
2309
Paul Gortmaker0db06282013-06-19 14:53:51 -04002310struct task_struct *fork_idle(int cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002311{
Ingo Molnar36c8b582006-07-03 00:25:41 -07002312 struct task_struct *task;
Christian Brauner7f192e32019-05-25 11:36:41 +02002313 struct kernel_clone_args args = {
2314 .flags = CLONE_VM,
2315 };
2316
2317 task = copy_process(&init_struct_pid, 0, cpu_to_node(cpu), &args);
Oleg Nesterovf106eee2010-05-26 14:44:11 -07002318 if (!IS_ERR(task)) {
Eric W. Biederman2c470472017-09-26 13:06:43 -05002319 init_idle_pids(task);
Akinobu Mita753ca4f2006-11-25 11:09:34 -08002320 init_idle(task, cpu);
Oleg Nesterovf106eee2010-05-26 14:44:11 -07002321 }
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08002322
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323 return task;
2324}
2325
Nadav Amit13585fa2019-04-25 17:11:25 -07002326struct mm_struct *copy_init_mm(void)
2327{
2328 return dup_mm(NULL, &init_mm);
2329}
2330
Linus Torvalds1da177e2005-04-16 15:20:36 -07002331/*
2332 * Ok, this is the main fork-routine.
2333 *
2334 * It copies the process, and if successful kick-starts
2335 * it and waits for it to finish using the VM if required.
Eugene Syromiatnikova0eb9ab2019-09-11 18:45:40 +01002336 *
2337 * args->exit_signal is expected to be checked for sanity by the caller.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338 */
Christian Brauner7f192e32019-05-25 11:36:41 +02002339long _do_fork(struct kernel_clone_args *args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340{
Christian Brauner7f192e32019-05-25 11:36:41 +02002341 u64 clone_flags = args->flags;
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002342 struct completion vfork;
2343 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344 struct task_struct *p;
2345 int trace = 0;
Eric W. Biederman92476d72006-03-31 02:31:42 -08002346 long nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347
Andrew Mortonbdff7462008-02-04 22:27:22 -08002348 /*
Tejun Heo4b9d33e2011-06-17 16:50:38 +02002349 * Determine whether and which event to report to ptracer. When
2350 * called from kernel_thread or CLONE_UNTRACED is explicitly
2351 * requested, no event is reported; otherwise, report if the event
2352 * for the type of forking is enabled.
Roland McGrath09a05392008-07-25 19:45:47 -07002353 */
Al Viroe80d6662012-10-22 23:10:08 -04002354 if (!(clone_flags & CLONE_UNTRACED)) {
Tejun Heo4b9d33e2011-06-17 16:50:38 +02002355 if (clone_flags & CLONE_VFORK)
2356 trace = PTRACE_EVENT_VFORK;
Christian Brauner7f192e32019-05-25 11:36:41 +02002357 else if (args->exit_signal != SIGCHLD)
Tejun Heo4b9d33e2011-06-17 16:50:38 +02002358 trace = PTRACE_EVENT_CLONE;
2359 else
2360 trace = PTRACE_EVENT_FORK;
2361
2362 if (likely(!ptrace_event_enabled(current, trace)))
2363 trace = 0;
2364 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002365
Christian Brauner7f192e32019-05-25 11:36:41 +02002366 p = copy_process(NULL, trace, NUMA_NO_NODE, args);
Emese Revfy38addce2016-06-20 20:41:19 +02002367 add_latent_entropy();
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002368
2369 if (IS_ERR(p))
2370 return PTR_ERR(p);
2371
Linus Torvalds1da177e2005-04-16 15:20:36 -07002372 /*
2373 * Do this prior waking up the new thread - the thread pointer
2374 * might get invalid after that point, if the thread exits quickly.
2375 */
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002376 trace_sched_process_fork(current, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002378 pid = get_task_pid(p, PIDTYPE_PID);
2379 nr = pid_vnr(pid);
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04002380
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002381 if (clone_flags & CLONE_PARENT_SETTID)
Christian Brauner7f192e32019-05-25 11:36:41 +02002382 put_user(nr, args->parent_tid);
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07002383
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002384 if (clone_flags & CLONE_VFORK) {
2385 p->vfork_done = &vfork;
2386 init_completion(&vfork);
2387 get_task_struct(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002388 }
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002389
2390 wake_up_new_task(p);
2391
2392 /* forking complete and child started to run, tell ptracer */
2393 if (unlikely(trace))
2394 ptrace_event_pid(trace, pid);
2395
2396 if (clone_flags & CLONE_VFORK) {
2397 if (!wait_for_vfork_done(p, &vfork))
2398 ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid);
2399 }
2400
2401 put_pid(pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -08002402 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002403}
2404
Dmitry V. Levin028b6e8a2019-07-14 19:20:47 +03002405bool legacy_clone_args_valid(const struct kernel_clone_args *kargs)
2406{
2407 /* clone(CLONE_PIDFD) uses parent_tidptr to return a pidfd */
2408 if ((kargs->flags & CLONE_PIDFD) &&
2409 (kargs->flags & CLONE_PARENT_SETTID))
2410 return false;
2411
2412 return true;
2413}
2414
Josh Triplett3033f14a2015-06-25 15:01:19 -07002415#ifndef CONFIG_HAVE_COPY_THREAD_TLS
2416/* For compatibility with architectures that call do_fork directly rather than
2417 * using the syscall entry points below. */
2418long do_fork(unsigned long clone_flags,
2419 unsigned long stack_start,
2420 unsigned long stack_size,
2421 int __user *parent_tidptr,
2422 int __user *child_tidptr)
2423{
Christian Brauner7f192e32019-05-25 11:36:41 +02002424 struct kernel_clone_args args = {
2425 .flags = (clone_flags & ~CSIGNAL),
Dmitry V. Levin028b6e8a2019-07-14 19:20:47 +03002426 .pidfd = parent_tidptr,
Christian Brauner7f192e32019-05-25 11:36:41 +02002427 .child_tid = child_tidptr,
2428 .parent_tid = parent_tidptr,
2429 .exit_signal = (clone_flags & CSIGNAL),
2430 .stack = stack_start,
2431 .stack_size = stack_size,
2432 };
2433
Dmitry V. Levin028b6e8a2019-07-14 19:20:47 +03002434 if (!legacy_clone_args_valid(&args))
2435 return -EINVAL;
2436
Christian Brauner7f192e32019-05-25 11:36:41 +02002437 return _do_fork(&args);
Josh Triplett3033f14a2015-06-25 15:01:19 -07002438}
2439#endif
2440
Al Viro2aa3a7f2012-09-21 19:55:31 -04002441/*
2442 * Create a kernel thread.
2443 */
2444pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
2445{
Christian Brauner7f192e32019-05-25 11:36:41 +02002446 struct kernel_clone_args args = {
2447 .flags = ((flags | CLONE_VM | CLONE_UNTRACED) & ~CSIGNAL),
2448 .exit_signal = (flags & CSIGNAL),
2449 .stack = (unsigned long)fn,
2450 .stack_size = (unsigned long)arg,
2451 };
2452
2453 return _do_fork(&args);
Al Viro2aa3a7f2012-09-21 19:55:31 -04002454}
Al Viro2aa3a7f2012-09-21 19:55:31 -04002455
Al Virod2125042012-10-23 13:17:59 -04002456#ifdef __ARCH_WANT_SYS_FORK
2457SYSCALL_DEFINE0(fork)
2458{
2459#ifdef CONFIG_MMU
Christian Brauner7f192e32019-05-25 11:36:41 +02002460 struct kernel_clone_args args = {
2461 .exit_signal = SIGCHLD,
2462 };
2463
2464 return _do_fork(&args);
Al Virod2125042012-10-23 13:17:59 -04002465#else
2466 /* can not support in nommu mode */
Daeseok Youn5d59e182014-01-23 15:55:47 -08002467 return -EINVAL;
Al Virod2125042012-10-23 13:17:59 -04002468#endif
2469}
2470#endif
2471
2472#ifdef __ARCH_WANT_SYS_VFORK
2473SYSCALL_DEFINE0(vfork)
2474{
Christian Brauner7f192e32019-05-25 11:36:41 +02002475 struct kernel_clone_args args = {
2476 .flags = CLONE_VFORK | CLONE_VM,
2477 .exit_signal = SIGCHLD,
2478 };
2479
2480 return _do_fork(&args);
Al Virod2125042012-10-23 13:17:59 -04002481}
2482#endif
2483
2484#ifdef __ARCH_WANT_SYS_CLONE
2485#ifdef CONFIG_CLONE_BACKWARDS
2486SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
2487 int __user *, parent_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07002488 unsigned long, tls,
Al Virod2125042012-10-23 13:17:59 -04002489 int __user *, child_tidptr)
2490#elif defined(CONFIG_CLONE_BACKWARDS2)
2491SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags,
2492 int __user *, parent_tidptr,
2493 int __user *, child_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07002494 unsigned long, tls)
Michal Simekdfa97712013-08-13 16:00:53 -07002495#elif defined(CONFIG_CLONE_BACKWARDS3)
2496SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp,
2497 int, stack_size,
2498 int __user *, parent_tidptr,
2499 int __user *, child_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07002500 unsigned long, tls)
Al Virod2125042012-10-23 13:17:59 -04002501#else
2502SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
2503 int __user *, parent_tidptr,
2504 int __user *, child_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07002505 unsigned long, tls)
Al Virod2125042012-10-23 13:17:59 -04002506#endif
2507{
Christian Brauner7f192e32019-05-25 11:36:41 +02002508 struct kernel_clone_args args = {
2509 .flags = (clone_flags & ~CSIGNAL),
2510 .pidfd = parent_tidptr,
2511 .child_tid = child_tidptr,
2512 .parent_tid = parent_tidptr,
2513 .exit_signal = (clone_flags & CSIGNAL),
2514 .stack = newsp,
2515 .tls = tls,
2516 };
2517
Dmitry V. Levin028b6e8a2019-07-14 19:20:47 +03002518 if (!legacy_clone_args_valid(&args))
Christian Brauner7f192e32019-05-25 11:36:41 +02002519 return -EINVAL;
2520
2521 return _do_fork(&args);
2522}
Christian Braunerd68dbb02019-06-21 01:26:35 +02002523#endif
Christian Brauner7f192e32019-05-25 11:36:41 +02002524
Christian Braunerd68dbb02019-06-21 01:26:35 +02002525#ifdef __ARCH_WANT_SYS_CLONE3
Christian Brauner7f192e32019-05-25 11:36:41 +02002526noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs,
2527 struct clone_args __user *uargs,
Aleksa Saraif14c2342019-10-01 11:10:53 +10002528 size_t usize)
Christian Brauner7f192e32019-05-25 11:36:41 +02002529{
Aleksa Saraif14c2342019-10-01 11:10:53 +10002530 int err;
Christian Brauner7f192e32019-05-25 11:36:41 +02002531 struct clone_args args;
2532
Aleksa Saraif14c2342019-10-01 11:10:53 +10002533 if (unlikely(usize > PAGE_SIZE))
Christian Brauner7f192e32019-05-25 11:36:41 +02002534 return -E2BIG;
Aleksa Saraif14c2342019-10-01 11:10:53 +10002535 if (unlikely(usize < CLONE_ARGS_SIZE_VER0))
Christian Brauner7f192e32019-05-25 11:36:41 +02002536 return -EINVAL;
2537
Aleksa Saraif14c2342019-10-01 11:10:53 +10002538 err = copy_struct_from_user(&args, sizeof(args), uargs, usize);
2539 if (err)
2540 return err;
Christian Brauner7f192e32019-05-25 11:36:41 +02002541
Eugene Syromiatnikova0eb9ab2019-09-11 18:45:40 +01002542 /*
2543 * Verify that higher 32bits of exit_signal are unset and that
2544 * it is a valid signal
2545 */
2546 if (unlikely((args.exit_signal & ~((u64)CSIGNAL)) ||
2547 !valid_signal(args.exit_signal)))
2548 return -EINVAL;
2549
Christian Brauner7f192e32019-05-25 11:36:41 +02002550 *kargs = (struct kernel_clone_args){
2551 .flags = args.flags,
2552 .pidfd = u64_to_user_ptr(args.pidfd),
2553 .child_tid = u64_to_user_ptr(args.child_tid),
2554 .parent_tid = u64_to_user_ptr(args.parent_tid),
2555 .exit_signal = args.exit_signal,
2556 .stack = args.stack,
2557 .stack_size = args.stack_size,
2558 .tls = args.tls,
2559 };
2560
2561 return 0;
2562}
2563
2564static bool clone3_args_valid(const struct kernel_clone_args *kargs)
2565{
2566 /*
2567 * All lower bits of the flag word are taken.
2568 * Verify that no other unknown flags are passed along.
2569 */
2570 if (kargs->flags & ~CLONE_LEGACY_FLAGS)
2571 return false;
2572
2573 /*
2574 * - make the CLONE_DETACHED bit reuseable for clone3
2575 * - make the CSIGNAL bits reuseable for clone3
2576 */
2577 if (kargs->flags & (CLONE_DETACHED | CSIGNAL))
2578 return false;
2579
2580 if ((kargs->flags & (CLONE_THREAD | CLONE_PARENT)) &&
2581 kargs->exit_signal)
2582 return false;
2583
2584 return true;
2585}
2586
Christian Brauner501bd012019-09-27 17:28:42 +02002587/**
2588 * clone3 - create a new process with specific properties
2589 * @uargs: argument structure
2590 * @size: size of @uargs
2591 *
2592 * clone3() is the extensible successor to clone()/clone2().
2593 * It takes a struct as argument that is versioned by its size.
2594 *
2595 * Return: On success, a positive PID for the child process.
2596 * On error, a negative errno number.
2597 */
Christian Brauner7f192e32019-05-25 11:36:41 +02002598SYSCALL_DEFINE2(clone3, struct clone_args __user *, uargs, size_t, size)
2599{
2600 int err;
2601
2602 struct kernel_clone_args kargs;
2603
2604 err = copy_clone_args_from_user(&kargs, uargs, size);
2605 if (err)
2606 return err;
2607
2608 if (!clone3_args_valid(&kargs))
2609 return -EINVAL;
2610
2611 return _do_fork(&kargs);
Al Virod2125042012-10-23 13:17:59 -04002612}
2613#endif
2614
Oleg Nesterov0f1b92c2017-01-30 18:06:11 +03002615void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data)
2616{
2617 struct task_struct *leader, *parent, *child;
2618 int res;
2619
2620 read_lock(&tasklist_lock);
2621 leader = top = top->group_leader;
2622down:
2623 for_each_thread(leader, parent) {
2624 list_for_each_entry(child, &parent->children, sibling) {
2625 res = visitor(child, data);
2626 if (res) {
2627 if (res < 0)
2628 goto out;
2629 leader = child;
2630 goto down;
2631 }
2632up:
2633 ;
2634 }
2635 }
2636
2637 if (leader != top) {
2638 child = leader;
2639 parent = child->real_parent;
2640 leader = parent->group_leader;
2641 goto up;
2642 }
2643out:
2644 read_unlock(&tasklist_lock);
2645}
2646
Ravikiran G Thirumalai5fd63b32006-01-11 22:46:15 +01002647#ifndef ARCH_MIN_MMSTRUCT_ALIGN
2648#define ARCH_MIN_MMSTRUCT_ALIGN 0
2649#endif
2650
Alexey Dobriyan51cc5062008-07-25 19:45:34 -07002651static void sighand_ctor(void *data)
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08002652{
2653 struct sighand_struct *sighand = data;
2654
Christoph Lametera35afb82007-05-16 22:10:57 -07002655 spin_lock_init(&sighand->siglock);
Davide Libenzib8fceee2007-09-20 12:40:16 -07002656 init_waitqueue_head(&sighand->signalfd_wqh);
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08002657}
2658
Linus Torvalds1da177e2005-04-16 15:20:36 -07002659void __init proc_caches_init(void)
2660{
Rik van Rielc1a2f7f2018-07-16 15:03:31 -04002661 unsigned int mm_size;
2662
Linus Torvalds1da177e2005-04-16 15:20:36 -07002663 sighand_cachep = kmem_cache_create("sighand_cache",
2664 sizeof(struct sighand_struct), 0,
Paul E. McKenney5f0d5a32017-01-18 02:53:44 -08002665 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_TYPESAFE_BY_RCU|
Levin, Alexander (Sasha Levin)75f296d2017-11-15 17:35:54 -08002666 SLAB_ACCOUNT, sighand_ctor);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002667 signal_cachep = kmem_cache_create("signal_cache",
2668 sizeof(struct signal_struct), 0,
Levin, Alexander (Sasha Levin)75f296d2017-11-15 17:35:54 -08002669 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
Vladimir Davydov5d097052016-01-14 15:18:21 -08002670 NULL);
Paul Mundt20c2df82007-07-20 10:11:58 +09002671 files_cachep = kmem_cache_create("files_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002672 sizeof(struct files_struct), 0,
Levin, Alexander (Sasha Levin)75f296d2017-11-15 17:35:54 -08002673 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
Vladimir Davydov5d097052016-01-14 15:18:21 -08002674 NULL);
Paul Mundt20c2df82007-07-20 10:11:58 +09002675 fs_cachep = kmem_cache_create("fs_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002676 sizeof(struct fs_struct), 0,
Levin, Alexander (Sasha Levin)75f296d2017-11-15 17:35:54 -08002677 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
Vladimir Davydov5d097052016-01-14 15:18:21 -08002678 NULL);
Rik van Rielc1a2f7f2018-07-16 15:03:31 -04002679
Linus Torvalds6345d242011-05-29 11:32:28 -07002680 /*
Rik van Rielc1a2f7f2018-07-16 15:03:31 -04002681 * The mm_cpumask is located at the end of mm_struct, and is
2682 * dynamically sized based on the maximum CPU number this system
2683 * can have, taking hotplug into account (nr_cpu_ids).
Linus Torvalds6345d242011-05-29 11:32:28 -07002684 */
Rik van Rielc1a2f7f2018-07-16 15:03:31 -04002685 mm_size = sizeof(struct mm_struct) + cpumask_size();
2686
David Windsor07dcd7f2017-08-15 16:45:00 -07002687 mm_cachep = kmem_cache_create_usercopy("mm_struct",
Rik van Rielc1a2f7f2018-07-16 15:03:31 -04002688 mm_size, ARCH_MIN_MMSTRUCT_ALIGN,
Levin, Alexander (Sasha Levin)75f296d2017-11-15 17:35:54 -08002689 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
David Windsor07dcd7f2017-08-15 16:45:00 -07002690 offsetof(struct mm_struct, saved_auxv),
2691 sizeof_field(struct mm_struct, saved_auxv),
Vladimir Davydov5d097052016-01-14 15:18:21 -08002692 NULL);
2693 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT);
David Howells8feae132009-01-08 12:04:47 +00002694 mmap_init();
Al Viro66577192011-06-28 15:41:10 -04002695 nsproxy_cache_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002696}
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002697
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002698/*
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002699 * Check constraints on flags passed to the unshare system call.
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002700 */
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002701static int check_unshare_flags(unsigned long unshare_flags)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002702{
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002703 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
2704 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
Eric W. Biederman50804fe2010-03-02 15:41:50 -08002705 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET|
Aditya Kalia79a9082016-01-29 02:54:06 -06002706 CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP))
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002707 return -EINVAL;
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002708 /*
Eric W. Biederman12c641a2015-08-10 17:35:07 -05002709 * Not implemented, but pretend it works if there is nothing
2710 * to unshare. Note that unsharing the address space or the
2711 * signal handlers also need to unshare the signal queues (aka
2712 * CLONE_THREAD).
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002713 */
2714 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
Eric W. Biederman12c641a2015-08-10 17:35:07 -05002715 if (!thread_group_empty(current))
2716 return -EINVAL;
2717 }
2718 if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) {
Elena Reshetovad036bda2019-01-18 14:27:26 +02002719 if (refcount_read(&current->sighand->count) > 1)
Eric W. Biederman12c641a2015-08-10 17:35:07 -05002720 return -EINVAL;
2721 }
2722 if (unshare_flags & CLONE_VM) {
2723 if (!current_is_single_threaded())
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002724 return -EINVAL;
2725 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002726
2727 return 0;
2728}
2729
2730/*
JANAK DESAI99d14192006-02-07 12:58:59 -08002731 * Unshare the filesystem structure if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002732 */
2733static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
2734{
2735 struct fs_struct *fs = current->fs;
2736
Al Viro498052b2009-03-30 07:20:30 -04002737 if (!(unshare_flags & CLONE_FS) || !fs)
2738 return 0;
2739
2740 /* don't need lock here; in the worst case we'll do useless copy */
2741 if (fs->users == 1)
2742 return 0;
2743
2744 *new_fsp = copy_fs_struct(fs);
2745 if (!*new_fsp)
2746 return -ENOMEM;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002747
2748 return 0;
2749}
2750
2751/*
JANAK DESAIa016f332006-02-07 12:59:02 -08002752 * Unshare file descriptor table if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002753 */
2754static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
2755{
2756 struct files_struct *fd = current->files;
JANAK DESAIa016f332006-02-07 12:59:02 -08002757 int error = 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002758
2759 if ((unshare_flags & CLONE_FILES) &&
JANAK DESAIa016f332006-02-07 12:59:02 -08002760 (fd && atomic_read(&fd->count) > 1)) {
2761 *new_fdp = dup_fd(fd, &error);
2762 if (!*new_fdp)
2763 return error;
2764 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002765
2766 return 0;
2767}
2768
2769/*
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002770 * unshare allows a process to 'unshare' part of the process
2771 * context which was originally shared using clone. copy_*
2772 * functions used by do_fork() cannot be used here directly
2773 * because they modify an inactive task_struct that is being
2774 * constructed. Here we are modifying the current, active,
2775 * task_struct.
2776 */
Dominik Brodowski9b321052018-03-11 11:34:42 +01002777int ksys_unshare(unsigned long unshare_flags)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002778{
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002779 struct fs_struct *fs, *new_fs = NULL;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002780 struct files_struct *fd, *new_fd = NULL;
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002781 struct cred *new_cred = NULL;
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07002782 struct nsproxy *new_nsproxy = NULL;
Manfred Spraul9edff4a2008-04-29 01:00:57 -07002783 int do_sysvsem = 0;
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002784 int err;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002785
Eric W. Biederman50804fe2010-03-02 15:41:50 -08002786 /*
Eric W. Biedermanfaf00da2015-08-10 18:25:44 -05002787 * If unsharing a user namespace must also unshare the thread group
2788 * and unshare the filesystem root and working directories.
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002789 */
2790 if (unshare_flags & CLONE_NEWUSER)
Eric W. Biedermane66eded2013-03-13 11:51:49 -07002791 unshare_flags |= CLONE_THREAD | CLONE_FS;
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002792 /*
Eric W. Biederman50804fe2010-03-02 15:41:50 -08002793 * If unsharing vm, must also unshare signal handlers.
2794 */
2795 if (unshare_flags & CLONE_VM)
2796 unshare_flags |= CLONE_SIGHAND;
Manfred Spraul6013f672008-04-29 01:00:59 -07002797 /*
Eric W. Biederman12c641a2015-08-10 17:35:07 -05002798 * If unsharing a signal handlers, must also unshare the signal queues.
2799 */
2800 if (unshare_flags & CLONE_SIGHAND)
2801 unshare_flags |= CLONE_THREAD;
2802 /*
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002803 * If unsharing namespace, must also unshare filesystem information.
2804 */
2805 if (unshare_flags & CLONE_NEWNS)
2806 unshare_flags |= CLONE_FS;
Eric W. Biederman50804fe2010-03-02 15:41:50 -08002807
2808 err = check_unshare_flags(unshare_flags);
2809 if (err)
2810 goto bad_unshare_out;
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002811 /*
Manfred Spraul6013f672008-04-29 01:00:59 -07002812 * CLONE_NEWIPC must also detach from the undolist: after switching
2813 * to a new ipc namespace, the semaphore arrays from the old
2814 * namespace are unreachable.
2815 */
2816 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
Manfred Spraul9edff4a2008-04-29 01:00:57 -07002817 do_sysvsem = 1;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002818 err = unshare_fs(unshare_flags, &new_fs);
2819 if (err)
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002820 goto bad_unshare_out;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002821 err = unshare_fd(unshare_flags, &new_fd);
2822 if (err)
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002823 goto bad_unshare_cleanup_fs;
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002824 err = unshare_userns(unshare_flags, &new_cred);
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002825 if (err)
Manfred Spraul9edff4a2008-04-29 01:00:57 -07002826 goto bad_unshare_cleanup_fd;
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002827 err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
2828 new_cred, new_fs);
2829 if (err)
2830 goto bad_unshare_cleanup_cred;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002831
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002832 if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) {
Manfred Spraul9edff4a2008-04-29 01:00:57 -07002833 if (do_sysvsem) {
2834 /*
2835 * CLONE_SYSVSEM is equivalent to sys_exit().
2836 */
2837 exit_sem(current);
2838 }
Jack Millerab602f72014-08-08 14:23:19 -07002839 if (unshare_flags & CLONE_NEWIPC) {
2840 /* Orphan segments in old ns (see sem above). */
2841 exit_shm(current);
2842 shm_init_task(current);
2843 }
Serge E. Hallynab516012006-10-02 02:18:06 -07002844
Alan Cox6f977e62013-02-27 17:03:23 -08002845 if (new_nsproxy)
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07002846 switch_task_namespaces(current, new_nsproxy);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002847
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07002848 task_lock(current);
2849
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002850 if (new_fs) {
2851 fs = current->fs;
Nick Piggin2a4419b2010-08-18 04:37:33 +10002852 spin_lock(&fs->lock);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002853 current->fs = new_fs;
Al Viro498052b2009-03-30 07:20:30 -04002854 if (--fs->users)
2855 new_fs = NULL;
2856 else
2857 new_fs = fs;
Nick Piggin2a4419b2010-08-18 04:37:33 +10002858 spin_unlock(&fs->lock);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002859 }
2860
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002861 if (new_fd) {
2862 fd = current->files;
2863 current->files = new_fd;
2864 new_fd = fd;
2865 }
2866
2867 task_unlock(current);
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002868
2869 if (new_cred) {
2870 /* Install the new user namespace */
2871 commit_creds(new_cred);
2872 new_cred = NULL;
2873 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002874 }
2875
Hari Bathinie4222672017-03-08 02:11:36 +05302876 perf_event_namespaces(current);
2877
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002878bad_unshare_cleanup_cred:
2879 if (new_cred)
2880 put_cred(new_cred);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002881bad_unshare_cleanup_fd:
2882 if (new_fd)
2883 put_files_struct(new_fd);
2884
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002885bad_unshare_cleanup_fs:
2886 if (new_fs)
Al Viro498052b2009-03-30 07:20:30 -04002887 free_fs_struct(new_fs);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002888
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002889bad_unshare_out:
2890 return err;
2891}
Al Viro3b125382008-04-22 05:31:30 -04002892
Dominik Brodowski9b321052018-03-11 11:34:42 +01002893SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
2894{
2895 return ksys_unshare(unshare_flags);
2896}
2897
Al Viro3b125382008-04-22 05:31:30 -04002898/*
2899 * Helper to unshare the files of the current task.
2900 * We don't want to expose copy_files internals to
2901 * the exec layer of the kernel.
2902 */
2903
2904int unshare_files(struct files_struct **displaced)
2905{
2906 struct task_struct *task = current;
Al Viro50704512008-04-26 05:25:00 +01002907 struct files_struct *copy = NULL;
Al Viro3b125382008-04-22 05:31:30 -04002908 int error;
2909
2910 error = unshare_fd(CLONE_FILES, &copy);
2911 if (error || !copy) {
2912 *displaced = NULL;
2913 return error;
2914 }
2915 *displaced = task->files;
2916 task_lock(task);
2917 task->files = copy;
2918 task_unlock(task);
2919 return 0;
2920}
Heinrich Schuchardt16db3d32015-04-16 12:47:50 -07002921
2922int sysctl_max_threads(struct ctl_table *table, int write,
2923 void __user *buffer, size_t *lenp, loff_t *ppos)
2924{
2925 struct ctl_table t;
2926 int ret;
2927 int threads = max_threads;
Michal Hockob0f53db2019-10-06 17:58:19 -07002928 int min = 1;
Heinrich Schuchardt16db3d32015-04-16 12:47:50 -07002929 int max = MAX_THREADS;
2930
2931 t = *table;
2932 t.data = &threads;
2933 t.extra1 = &min;
2934 t.extra2 = &max;
2935
2936 ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos);
2937 if (ret || !write)
2938 return ret;
2939
Michal Hockob0f53db2019-10-06 17:58:19 -07002940 max_threads = threads;
Heinrich Schuchardt16db3d32015-04-16 12:47:50 -07002941
2942 return 0;
2943}