blob: 8f3e2d97d77100dd746d98f7410c8601c3c8891b [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 Dronamraju0326f5a2012-03-13 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
128DEFINE_PER_CPU(unsigned long, process_counts) = 0;
129
Christoph Hellwigc59923a2006-07-10 04:45:40 -0700130__cacheline_aligned DEFINE_RWLOCK(tasklist_lock); /* outer */
Paul E. McKenneydb1466b2010-03-03 07:46:56 -0800131
132#ifdef CONFIG_PROVE_RCU
133int lockdep_tasklist_lock_is_held(void)
134{
135 return lockdep_is_held(&tasklist_lock);
136}
137EXPORT_SYMBOL_GPL(lockdep_tasklist_lock_is_held);
138#endif /* #ifdef CONFIG_PROVE_RCU */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139
140int nr_processes(void)
141{
142 int cpu;
143 int total = 0;
144
Ian Campbell1d510752009-11-03 10:11:14 +0000145 for_each_possible_cpu(cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 total += per_cpu(process_counts, cpu);
147
148 return total;
149}
150
Akinobu Mitaf19b9f72012-07-30 14:42:33 -0700151void __weak arch_release_task_struct(struct task_struct *tsk)
152{
153}
154
Thomas Gleixnerf5e10282012-05-05 15:05:48 +0000155#ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
Christoph Lametere18b8902006-12-06 20:33:20 -0800156static struct kmem_cache *task_struct_cachep;
Thomas Gleixner41101802012-05-05 15:05:41 +0000157
158static inline struct task_struct *alloc_task_struct_node(int node)
159{
160 return kmem_cache_alloc_node(task_struct_cachep, GFP_KERNEL, node);
161}
162
Thomas Gleixner41101802012-05-05 15:05:41 +0000163static inline void free_task_struct(struct task_struct *tsk)
164{
Thomas Gleixner41101802012-05-05 15:05:41 +0000165 kmem_cache_free(task_struct_cachep, tsk);
166}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167#endif
168
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700169#ifndef CONFIG_ARCH_THREAD_STACK_ALLOCATOR
Thomas Gleixner41101802012-05-05 15:05:41 +0000170
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000171/*
172 * Allocate pages if THREAD_SIZE is >= PAGE_SIZE, otherwise use a
173 * kmemcache based allocator.
174 */
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700175# if THREAD_SIZE >= PAGE_SIZE || defined(CONFIG_VMAP_STACK)
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700176
177#ifdef CONFIG_VMAP_STACK
178/*
179 * vmalloc() is a bit slow, and calling vfree() enough times will force a TLB
180 * flush. Try to minimize the number of calls by caching stacks.
181 */
182#define NR_CACHED_STACKS 2
183static DEFINE_PER_CPU(struct vm_struct *, cached_stacks[NR_CACHED_STACKS]);
Hoeun Ryu19659c52017-05-08 15:56:11 -0700184
185static int free_vm_stack_cache(unsigned int cpu)
186{
187 struct vm_struct **cached_vm_stacks = per_cpu_ptr(cached_stacks, cpu);
188 int i;
189
190 for (i = 0; i < NR_CACHED_STACKS; i++) {
191 struct vm_struct *vm_stack = cached_vm_stacks[i];
192
193 if (!vm_stack)
194 continue;
195
196 vfree(vm_stack->addr);
197 cached_vm_stacks[i] = NULL;
198 }
199
200 return 0;
201}
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700202#endif
203
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700204static unsigned long *alloc_thread_stack_node(struct task_struct *tsk, int node)
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700205{
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700206#ifdef CONFIG_VMAP_STACK
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700207 void *stack;
208 int i;
209
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700210 for (i = 0; i < NR_CACHED_STACKS; i++) {
Christoph Lameter112166f2017-07-12 14:33:11 -0700211 struct vm_struct *s;
212
213 s = this_cpu_xchg(cached_stacks[i], NULL);
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700214
215 if (!s)
216 continue;
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700217
Konstantin Khlebnikovca182552017-10-13 15:58:22 -0700218 /* Clear stale pointers from reused stack. */
219 memset(s->addr, 0, THREAD_SIZE);
Kees Cooke01e8062018-04-20 14:55:31 -0700220
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700221 tsk->stack_vm_area = s;
Shakeel Buttba4a4572019-01-08 15:22:57 -0800222 tsk->stack = s->addr;
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700223 return s->addr;
224 }
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700225
Roman Gushchin9b6f7e12018-10-26 15:03:19 -0700226 /*
227 * Allocated stacks are cached and later reused by new threads,
228 * so memcg accounting is performed manually on assigning/releasing
229 * stacks to tasks. Drop __GFP_ACCOUNT.
230 */
Mark Rutland48ac3c12017-07-14 12:23:09 +0100231 stack = __vmalloc_node_range(THREAD_SIZE, THREAD_ALIGN,
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700232 VMALLOC_START, VMALLOC_END,
Roman Gushchin9b6f7e12018-10-26 15:03:19 -0700233 THREADINFO_GFP & ~__GFP_ACCOUNT,
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700234 PAGE_KERNEL,
235 0, node, __builtin_return_address(0));
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700236
237 /*
238 * We can't call find_vm_area() in interrupt context, and
239 * free_thread_stack() can be called in interrupt context,
240 * so cache the vm_struct.
241 */
Rik van Riel5eed6f12018-12-21 14:30:54 -0800242 if (stack) {
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700243 tsk->stack_vm_area = find_vm_area(stack);
Rik van Riel5eed6f12018-12-21 14:30:54 -0800244 tsk->stack = stack;
245 }
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700246 return stack;
247#else
Vladimir Davydov49491482016-07-26 15:24:24 -0700248 struct page *page = alloc_pages_node(node, THREADINFO_GFP,
249 THREAD_SIZE_ORDER);
Eric Dumazetb6a84012011-03-22 16:30:42 -0700250
Andrea Arcangeli1bf45802019-06-28 12:07:14 -0700251 if (likely(page)) {
252 tsk->stack = page_address(page);
253 return tsk->stack;
254 }
255 return NULL;
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700256#endif
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700257}
258
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700259static inline void free_thread_stack(struct task_struct *tsk)
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700260{
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700261#ifdef CONFIG_VMAP_STACK
Roman Gushchin9b6f7e12018-10-26 15:03:19 -0700262 struct vm_struct *vm = task_stack_vm_area(tsk);
263
264 if (vm) {
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700265 int i;
266
Roman Gushchin9b6f7e12018-10-26 15:03:19 -0700267 for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
268 mod_memcg_page_state(vm->pages[i],
269 MEMCG_KERNEL_STACK_KB,
270 -(int)(PAGE_SIZE / 1024));
271
272 memcg_kmem_uncharge(vm->pages[i], 0);
273 }
274
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700275 for (i = 0; i < NR_CACHED_STACKS; i++) {
Christoph Lameter112166f2017-07-12 14:33:11 -0700276 if (this_cpu_cmpxchg(cached_stacks[i],
277 NULL, tsk->stack_vm_area) != NULL)
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700278 continue;
279
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700280 return;
281 }
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700282
Andrey Ryabinin0f110a92016-12-12 16:44:14 -0800283 vfree_atomic(tsk->stack);
Andy Lutomirskiac496bf2016-09-15 22:45:49 -0700284 return;
285 }
286#endif
287
288 __free_pages(virt_to_page(tsk->stack), THREAD_SIZE_ORDER);
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700289}
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000290# else
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700291static struct kmem_cache *thread_stack_cache;
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000292
Michael Ellerman9521d392016-06-25 21:53:30 +1000293static unsigned long *alloc_thread_stack_node(struct task_struct *tsk,
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000294 int node)
295{
Rik van Riel5eed6f12018-12-21 14:30:54 -0800296 unsigned long *stack;
297 stack = kmem_cache_alloc_node(thread_stack_cache, THREADINFO_GFP, node);
298 tsk->stack = stack;
299 return stack;
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000300}
301
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700302static void free_thread_stack(struct task_struct *tsk)
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000303{
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700304 kmem_cache_free(thread_stack_cache, tsk->stack);
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000305}
306
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700307void thread_stack_cache_init(void)
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000308{
David Windsorf9d299462017-06-10 22:50:41 -0400309 thread_stack_cache = kmem_cache_create_usercopy("thread_stack",
310 THREAD_SIZE, THREAD_SIZE, 0, 0,
311 THREAD_SIZE, NULL);
Linus Torvaldsb235bee2016-06-24 15:09:37 -0700312 BUG_ON(thread_stack_cache == NULL);
Thomas Gleixner0d15d742012-05-05 15:05:41 +0000313}
314# endif
FUJITA Tomonorib69c49b2008-07-25 01:45:40 -0700315#endif
316
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317/* SLAB cache for signal_struct structures (tsk->signal) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800318static struct kmem_cache *signal_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319
320/* SLAB cache for sighand_struct structures (tsk->sighand) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800321struct kmem_cache *sighand_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322
323/* SLAB cache for files_struct structures (tsk->files) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800324struct kmem_cache *files_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325
326/* SLAB cache for fs_struct structures (tsk->fs) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800327struct kmem_cache *fs_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328
329/* SLAB cache for vm_area_struct structures */
Linus Torvalds3928d4f2018-07-21 13:48:51 -0700330static struct kmem_cache *vm_area_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331
332/* SLAB cache for mm_struct structures (tsk->mm) */
Christoph Lametere18b8902006-12-06 20:33:20 -0800333static struct kmem_cache *mm_cachep;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334
Linus Torvalds490fc052018-07-21 15:24:03 -0700335struct vm_area_struct *vm_area_alloc(struct mm_struct *mm)
Linus Torvalds3928d4f2018-07-21 13:48:51 -0700336{
Andrew Mortona6704682018-08-21 21:53:06 -0700337 struct vm_area_struct *vma;
Linus Torvalds490fc052018-07-21 15:24:03 -0700338
Andrew Mortona6704682018-08-21 21:53:06 -0700339 vma = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
Kirill A. Shutemov027232d2018-07-26 16:37:25 -0700340 if (vma)
341 vma_init(vma, mm);
Linus Torvalds490fc052018-07-21 15:24:03 -0700342 return vma;
Linus Torvalds3928d4f2018-07-21 13:48:51 -0700343}
344
345struct vm_area_struct *vm_area_dup(struct vm_area_struct *orig)
346{
Linus Torvalds95faf692018-07-21 14:48:45 -0700347 struct vm_area_struct *new = kmem_cache_alloc(vm_area_cachep, GFP_KERNEL);
348
349 if (new) {
350 *new = *orig;
351 INIT_LIST_HEAD(&new->anon_vma_chain);
352 }
353 return new;
Linus Torvalds3928d4f2018-07-21 13:48:51 -0700354}
355
356void vm_area_free(struct vm_area_struct *vma)
357{
358 kmem_cache_free(vm_area_cachep, vma);
359}
360
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700361static void account_kernel_stack(struct task_struct *tsk, int account)
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700362{
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700363 void *stack = task_stack_page(tsk);
364 struct vm_struct *vm = task_stack_vm_area(tsk);
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700365
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700366 BUILD_BUG_ON(IS_ENABLED(CONFIG_VMAP_STACK) && PAGE_SIZE % 1024 != 0);
Andy Lutomirskiefdc9492016-07-28 15:48:17 -0700367
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700368 if (vm) {
369 int i;
370
371 BUG_ON(vm->nr_pages != THREAD_SIZE / PAGE_SIZE);
372
373 for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
374 mod_zone_page_state(page_zone(vm->pages[i]),
375 NR_KERNEL_STACK_KB,
376 PAGE_SIZE / 1024 * account);
377 }
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700378 } else {
379 /*
380 * All stack pages are in the same zone and belong to the
381 * same memcg.
382 */
383 struct page *first_page = virt_to_page(stack);
384
385 mod_zone_page_state(page_zone(first_page), NR_KERNEL_STACK_KB,
386 THREAD_SIZE / 1024 * account);
387
Johannes Weinered52be72017-07-06 15:40:49 -0700388 mod_memcg_page_state(first_page, MEMCG_KERNEL_STACK_KB,
389 account * (THREAD_SIZE / 1024));
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700390 }
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700391}
392
Roman Gushchin9b6f7e12018-10-26 15:03:19 -0700393static int memcg_charge_kernel_stack(struct task_struct *tsk)
394{
395#ifdef CONFIG_VMAP_STACK
396 struct vm_struct *vm = task_stack_vm_area(tsk);
397 int ret;
398
399 if (vm) {
400 int i;
401
402 for (i = 0; i < THREAD_SIZE / PAGE_SIZE; i++) {
403 /*
404 * If memcg_kmem_charge() fails, page->mem_cgroup
405 * pointer is NULL, and both memcg_kmem_uncharge()
406 * and mod_memcg_page_state() in free_thread_stack()
407 * will ignore this page. So it's safe.
408 */
409 ret = memcg_kmem_charge(vm->pages[i], GFP_KERNEL, 0);
410 if (ret)
411 return ret;
412
413 mod_memcg_page_state(vm->pages[i],
414 MEMCG_KERNEL_STACK_KB,
415 PAGE_SIZE / 1024);
416 }
417 }
418#endif
419 return 0;
420}
421
Andy Lutomirski68f24b082016-09-15 22:45:48 -0700422static void release_task_stack(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423{
Andy Lutomirski405c0752016-10-31 08:11:43 -0700424 if (WARN_ON(tsk->state != TASK_DEAD))
425 return; /* Better to leak the stack than to free prematurely */
426
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700427 account_kernel_stack(tsk, -1);
Andy Lutomirskiba14a192016-08-11 02:35:21 -0700428 free_thread_stack(tsk);
Andy Lutomirski68f24b082016-09-15 22:45:48 -0700429 tsk->stack = NULL;
430#ifdef CONFIG_VMAP_STACK
431 tsk->stack_vm_area = NULL;
432#endif
433}
434
435#ifdef CONFIG_THREAD_INFO_IN_TASK
436void put_task_stack(struct task_struct *tsk)
437{
Elena Reshetovaf0b89d32019-01-18 14:27:30 +0200438 if (refcount_dec_and_test(&tsk->stack_refcount))
Andy Lutomirski68f24b082016-09-15 22:45:48 -0700439 release_task_stack(tsk);
440}
441#endif
442
443void free_task(struct task_struct *tsk)
444{
445#ifndef CONFIG_THREAD_INFO_IN_TASK
446 /*
447 * The task is finally done with both the stack and thread_info,
448 * so free both.
449 */
450 release_task_stack(tsk);
451#else
452 /*
453 * If the task had a separate stack allocation, it should be gone
454 * by now.
455 */
Elena Reshetovaf0b89d32019-01-18 14:27:30 +0200456 WARN_ON_ONCE(refcount_read(&tsk->stack_refcount) != 0);
Andy Lutomirski68f24b082016-09-15 22:45:48 -0700457#endif
Ingo Molnar23f78d4a2006-06-27 02:54:53 -0700458 rt_mutex_debug_task_free(tsk);
Frederic Weisbeckerfb526072008-11-25 21:07:04 +0100459 ftrace_graph_exit_task(tsk);
Will Drewrye2cfabdf2012-04-12 16:47:57 -0500460 put_seccomp_filter(tsk);
Akinobu Mitaf19b9f72012-07-30 14:42:33 -0700461 arch_release_task_struct(tsk);
Oleg Nesterov1da5c462016-11-29 18:50:57 +0100462 if (tsk->flags & PF_KTHREAD)
463 free_kthread_struct(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 free_task_struct(tsk);
465}
466EXPORT_SYMBOL(free_task);
467
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468#ifdef CONFIG_MMU
Emese Revfy0766f782016-06-20 20:42:34 +0200469static __latent_entropy int dup_mmap(struct mm_struct *mm,
470 struct mm_struct *oldmm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471{
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700472 struct vm_area_struct *mpnt, *tmp, *prev, **pprev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 struct rb_node **rb_link, *rb_parent;
474 int retval;
475 unsigned long charge;
Pavel Emelyanov893e26e2017-02-22 15:42:27 -0800476 LIST_HEAD(uf);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477
Oleg Nesterov32cdba12012-11-14 19:03:42 +0100478 uprobe_start_dup_mmap();
Michal Hocko7c051262016-05-23 16:25:48 -0700479 if (down_write_killable(&oldmm->mmap_sem)) {
480 retval = -EINTR;
481 goto fail_uprobe_end;
482 }
Ralf Baechleec8c0442006-12-12 17:14:57 +0000483 flush_cache_dup_mm(oldmm);
Oleg Nesterovf8ac4ec2012-08-08 17:11:42 +0200484 uprobe_dup_mmap(oldmm, mm);
Ingo Molnarad339452006-07-03 00:25:15 -0700485 /*
486 * Not linked in yet - no deadlock potential:
487 */
488 down_write_nested(&mm->mmap_sem, SINGLE_DEPTH_NESTING);
Hugh Dickins7ee78232005-10-29 18:16:08 -0700489
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -0700490 /* No ordering required: file already has been exposed. */
491 RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
492
Vladimir Davydov4f7d4612014-08-08 14:22:01 -0700493 mm->total_vm = oldmm->total_vm;
Konstantin Khlebnikov84638332016-01-14 15:22:07 -0800494 mm->data_vm = oldmm->data_vm;
Vladimir Davydov4f7d4612014-08-08 14:22:01 -0700495 mm->exec_vm = oldmm->exec_vm;
496 mm->stack_vm = oldmm->stack_vm;
497
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 rb_link = &mm->mm_rb.rb_node;
499 rb_parent = NULL;
500 pprev = &mm->mmap;
Hugh Dickinsf8af4da2009-09-21 17:01:57 -0700501 retval = ksm_fork(mm, oldmm);
502 if (retval)
503 goto out;
Andrea Arcangeliba761492011-01-13 15:46:58 -0800504 retval = khugepaged_fork(mm, oldmm);
505 if (retval)
506 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700508 prev = NULL;
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700509 for (mpnt = oldmm->mmap; mpnt; mpnt = mpnt->vm_next) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 struct file *file;
511
512 if (mpnt->vm_flags & VM_DONTCOPY) {
Konstantin Khlebnikov84638332016-01-14 15:22:07 -0800513 vm_stat_account(mm, mpnt->vm_flags, -vma_pages(mpnt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 continue;
515 }
516 charge = 0;
Tetsuo Handa655c79b2018-06-14 15:26:34 -0700517 /*
518 * Don't duplicate many vmas if we've been oom-killed (for
519 * example)
520 */
521 if (fatal_signal_pending(current)) {
522 retval = -EINTR;
523 goto out;
524 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 if (mpnt->vm_flags & VM_ACCOUNT) {
Huang Shijieb2412b72012-07-30 14:42:30 -0700526 unsigned long len = vma_pages(mpnt);
527
Al Viro191c5422012-02-13 03:58:52 +0000528 if (security_vm_enough_memory_mm(oldmm, len)) /* sic */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 goto fail_nomem;
530 charge = len;
531 }
Linus Torvalds3928d4f2018-07-21 13:48:51 -0700532 tmp = vm_area_dup(mpnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 if (!tmp)
534 goto fail_nomem;
Oleg Nesterovef0855d2013-09-11 14:20:14 -0700535 retval = vma_dup_policy(mpnt, tmp);
536 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 goto fail_nomem_policy;
Andrea Arcangelia247c3a2010-09-22 13:05:12 -0700538 tmp->vm_mm = mm;
Pavel Emelyanov893e26e2017-02-22 15:42:27 -0800539 retval = dup_userfaultfd(tmp, &uf);
540 if (retval)
541 goto fail_nomem_anon_vma_fork;
Rik van Rield2cd9ed2017-09-06 16:25:15 -0700542 if (tmp->vm_flags & VM_WIPEONFORK) {
543 /* VM_WIPEONFORK gets a clean slate in the child. */
544 tmp->anon_vma = NULL;
545 if (anon_vma_prepare(tmp))
546 goto fail_nomem_anon_vma_fork;
547 } else if (anon_vma_fork(tmp, mpnt))
Rik van Riel5beb4932010-03-05 13:42:07 -0800548 goto fail_nomem_anon_vma_fork;
Pavel Emelyanov893e26e2017-02-22 15:42:27 -0800549 tmp->vm_flags &= ~(VM_LOCKED | VM_LOCKONFAULT);
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700550 tmp->vm_next = tmp->vm_prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 file = tmp->vm_file;
552 if (file) {
Al Viro496ad9a2013-01-23 17:07:38 -0500553 struct inode *inode = file_inode(file);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000554 struct address_space *mapping = file->f_mapping;
555
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 get_file(file);
557 if (tmp->vm_flags & VM_DENYWRITE)
558 atomic_dec(&inode->i_writecount);
Davidlohr Bueso83cde9e2014-12-12 16:54:21 -0800559 i_mmap_lock_write(mapping);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000560 if (tmp->vm_flags & VM_SHARED)
David Herrmann4bb5f5d2014-08-08 14:25:25 -0700561 atomic_inc(&mapping->i_mmap_writable);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000562 flush_dcache_mmap_lock(mapping);
563 /* insert tmp into the share list, just after mpnt */
Kirill A. Shutemov27ba0642015-02-10 14:09:59 -0800564 vma_interval_tree_insert_after(tmp, mpnt,
565 &mapping->i_mmap);
Hugh Dickinsb88ed202008-12-10 20:48:52 +0000566 flush_dcache_mmap_unlock(mapping);
Davidlohr Bueso83cde9e2014-12-12 16:54:21 -0800567 i_mmap_unlock_write(mapping);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 }
569
570 /*
Mel Gormana1e78772008-07-23 21:27:23 -0700571 * Clear hugetlb-related page reserves for children. This only
572 * affects MAP_PRIVATE mappings. Faults generated by the child
573 * are not guaranteed to succeed, even if read-only
574 */
575 if (is_vm_hugetlb_page(tmp))
576 reset_vma_resv_huge_pages(tmp);
577
578 /*
Hugh Dickins7ee78232005-10-29 18:16:08 -0700579 * Link in the new vma and copy the page table entries.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 *pprev = tmp;
582 pprev = &tmp->vm_next;
Linus Torvalds297c5ee2010-08-20 16:24:55 -0700583 tmp->vm_prev = prev;
584 prev = tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585
586 __vma_link_rb(mm, tmp, rb_link, rb_parent);
587 rb_link = &tmp->vm_rb.rb_right;
588 rb_parent = &tmp->vm_rb;
589
590 mm->map_count++;
Rik van Rield2cd9ed2017-09-06 16:25:15 -0700591 if (!(tmp->vm_flags & VM_WIPEONFORK))
592 retval = copy_page_range(mm, oldmm, mpnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593
594 if (tmp->vm_ops && tmp->vm_ops->open)
595 tmp->vm_ops->open(tmp);
596
597 if (retval)
598 goto out;
599 }
Jeremy Fitzhardinged6dd61c2007-05-02 19:27:14 +0200600 /* a new mm has just been created */
Nadav Amit1ed0cc52018-09-04 15:45:41 -0700601 retval = arch_dup_mmap(oldmm, mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602out:
Hugh Dickins7ee78232005-10-29 18:16:08 -0700603 up_write(&mm->mmap_sem);
Hugh Dickinsfd3e42f2005-10-29 18:16:06 -0700604 flush_tlb_mm(oldmm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 up_write(&oldmm->mmap_sem);
Pavel Emelyanov893e26e2017-02-22 15:42:27 -0800606 dup_userfaultfd_complete(&uf);
Michal Hocko7c051262016-05-23 16:25:48 -0700607fail_uprobe_end:
Oleg Nesterov32cdba12012-11-14 19:03:42 +0100608 uprobe_end_dup_mmap();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 return retval;
Rik van Riel5beb4932010-03-05 13:42:07 -0800610fail_nomem_anon_vma_fork:
Oleg Nesterovef0855d2013-09-11 14:20:14 -0700611 mpol_put(vma_policy(tmp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612fail_nomem_policy:
Linus Torvalds3928d4f2018-07-21 13:48:51 -0700613 vm_area_free(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614fail_nomem:
615 retval = -ENOMEM;
616 vm_unacct_memory(charge);
617 goto out;
618}
619
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700620static inline int mm_alloc_pgd(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621{
622 mm->pgd = pgd_alloc(mm);
623 if (unlikely(!mm->pgd))
624 return -ENOMEM;
625 return 0;
626}
627
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -0700628static inline void mm_free_pgd(struct mm_struct *mm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629{
Benjamin Herrenschmidt5e541972008-02-04 22:29:14 -0800630 pgd_free(mm, mm->pgd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631}
632#else
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -0700633static int dup_mmap(struct mm_struct *mm, struct mm_struct *oldmm)
634{
635 down_write(&oldmm->mmap_sem);
636 RCU_INIT_POINTER(mm->exe_file, get_mm_exe_file(oldmm));
637 up_write(&oldmm->mmap_sem);
638 return 0;
639}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640#define mm_alloc_pgd(mm) (0)
641#define mm_free_pgd(mm)
642#endif /* CONFIG_MMU */
643
Andrew Mortond70f2a12018-01-31 16:15:51 -0800644static void check_mm(struct mm_struct *mm)
645{
646 int i;
647
648 for (i = 0; i < NR_MM_COUNTERS; i++) {
649 long x = atomic_long_read(&mm->rss_stat.count[i]);
650
651 if (unlikely(x))
652 printk(KERN_ALERT "BUG: Bad rss-counter state "
653 "mm:%p idx:%d val:%ld\n", mm, i, x);
654 }
655
656 if (mm_pgtables_bytes(mm))
657 pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n",
658 mm_pgtables_bytes(mm));
659
660#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
661 VM_BUG_ON_MM(mm->pmd_huge_pte, mm);
662#endif
663}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664
Christoph Lametere94b1762006-12-06 20:33:17 -0800665#define allocate_mm() (kmem_cache_alloc(mm_cachep, GFP_KERNEL))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666#define free_mm(mm) (kmem_cache_free(mm_cachep, (mm)))
667
Andrew Mortond70f2a12018-01-31 16:15:51 -0800668/*
669 * Called when the last reference to the mm
670 * is dropped: either by a lazy thread or by
671 * mmput. Free the page directory and the mm.
672 */
Andrew Mortond34bc482018-02-21 14:45:17 -0800673void __mmdrop(struct mm_struct *mm)
Andrew Mortond70f2a12018-01-31 16:15:51 -0800674{
675 BUG_ON(mm == &init_mm);
Mark Rutland3eda69c2018-04-05 16:25:12 -0700676 WARN_ON_ONCE(mm == current->mm);
677 WARN_ON_ONCE(mm == current->active_mm);
Andrew Mortond70f2a12018-01-31 16:15:51 -0800678 mm_free_pgd(mm);
679 destroy_context(mm);
680 hmm_mm_destroy(mm);
681 mmu_notifier_mm_destroy(mm);
682 check_mm(mm);
683 put_user_ns(mm->user_ns);
684 free_mm(mm);
685}
Andrew Mortond34bc482018-02-21 14:45:17 -0800686EXPORT_SYMBOL_GPL(__mmdrop);
Andrew Mortond70f2a12018-01-31 16:15:51 -0800687
688static void mmdrop_async_fn(struct work_struct *work)
689{
690 struct mm_struct *mm;
691
692 mm = container_of(work, struct mm_struct, async_put_work);
693 __mmdrop(mm);
694}
695
696static void mmdrop_async(struct mm_struct *mm)
697{
698 if (unlikely(atomic_dec_and_test(&mm->mm_count))) {
699 INIT_WORK(&mm->async_put_work, mmdrop_async_fn);
700 schedule_work(&mm->async_put_work);
701 }
702}
703
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704static inline void free_signal_struct(struct signal_struct *sig)
705{
706 taskstats_tgid_free(sig);
707 sched_autogroup_exit(sig);
708 /*
709 * __mmdrop is not safe to call from softirq context on x86 due to
710 * pgd_dtor so postpone it to the async context
711 */
712 if (sig->oom_mm)
713 mmdrop_async(sig->oom_mm);
714 kmem_cache_free(signal_cachep, sig);
715}
716
717static inline void put_signal_struct(struct signal_struct *sig)
718{
Elena Reshetova60d4de32019-01-18 14:27:27 +0200719 if (refcount_dec_and_test(&sig->sigcnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 free_signal_struct(sig);
721}
722
723void __put_task_struct(struct task_struct *tsk)
724{
725 WARN_ON(!tsk->exit_state);
Elena Reshetovaec1d2812019-01-18 14:27:29 +0200726 WARN_ON(refcount_read(&tsk->usage));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 WARN_ON(tsk == current);
728
729 cgroup_free(tsk);
730 task_numa_free(tsk);
731 security_task_free(tsk);
732 exit_creds(tsk);
733 delayacct_tsk_free(tsk);
734 put_signal_struct(tsk->signal);
735
736 if (!profile_handoff_task(tsk))
737 free_task(tsk);
738}
739EXPORT_SYMBOL_GPL(__put_task_struct);
740
741void __init __weak arch_task_cache_init(void) { }
742
743/*
744 * set_max_threads
745 */
746static void set_max_threads(unsigned int max_threads_suggested)
747{
748 u64 threads;
Arun KSca79b0c2018-12-28 00:34:29 -0800749 unsigned long nr_pages = totalram_pages();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
751 /*
752 * The number of threads shall be limited such that the thread
753 * structures may only consume a small part of the available memory.
754 */
Arun KS3d6357d2018-12-28 00:34:20 -0800755 if (fls64(nr_pages) + fls64(PAGE_SIZE) > 64)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 threads = MAX_THREADS;
757 else
Arun KS3d6357d2018-12-28 00:34:20 -0800758 threads = div64_u64((u64) nr_pages * (u64) PAGE_SIZE,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 (u64) THREAD_SIZE * 8UL);
760
761 if (threads > max_threads_suggested)
762 threads = max_threads_suggested;
763
764 max_threads = clamp_t(u64, threads, MIN_THREADS, MAX_THREADS);
765}
766
767#ifdef CONFIG_ARCH_WANTS_DYNAMIC_TASK_STRUCT
768/* Initialized by the architecture: */
769int arch_task_struct_size __read_mostly;
770#endif
771
Kees Cook59054292017-08-16 13:00:58 -0700772static void task_struct_whitelist(unsigned long *offset, unsigned long *size)
773{
774 /* Fetch thread_struct whitelist for the architecture. */
775 arch_thread_struct_whitelist(offset, size);
776
777 /*
778 * Handle zero-sized whitelist or empty thread_struct, otherwise
779 * adjust offset to position of thread_struct in task_struct.
780 */
781 if (unlikely(*size == 0))
782 *offset = 0;
783 else
784 *offset += offsetof(struct task_struct, thread);
785}
786
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787void __init fork_init(void)
788{
789 int i;
790#ifndef CONFIG_ARCH_TASK_STRUCT_ALLOCATOR
791#ifndef ARCH_MIN_TASKALIGN
792#define ARCH_MIN_TASKALIGN 0
793#endif
794 int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN);
Kees Cook59054292017-08-16 13:00:58 -0700795 unsigned long useroffset, usersize;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796
797 /* create a slab on which task_structs can be allocated */
Kees Cook59054292017-08-16 13:00:58 -0700798 task_struct_whitelist(&useroffset, &usersize);
799 task_struct_cachep = kmem_cache_create_usercopy("task_struct",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 arch_task_struct_size, align,
Kees Cook59054292017-08-16 13:00:58 -0700801 SLAB_PANIC|SLAB_ACCOUNT,
802 useroffset, usersize, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803#endif
804
805 /* do the arch specific task caches init */
806 arch_task_cache_init();
807
808 set_max_threads(MAX_THREADS);
809
810 init_task.signal->rlim[RLIMIT_NPROC].rlim_cur = max_threads/2;
811 init_task.signal->rlim[RLIMIT_NPROC].rlim_max = max_threads/2;
812 init_task.signal->rlim[RLIMIT_SIGPENDING] =
813 init_task.signal->rlim[RLIMIT_NPROC];
814
815 for (i = 0; i < UCOUNT_COUNTS; i++) {
816 init_user_ns.ucount_max[i] = max_threads/2;
817 }
818
819#ifdef CONFIG_VMAP_STACK
820 cpuhp_setup_state(CPUHP_BP_PREPARE_DYN, "fork:vm_stack_cache",
821 NULL, free_vm_stack_cache);
822#endif
823
824 lockdep_init_task(&init_task);
Nadav Amitaad42dd2019-04-26 16:22:44 -0700825 uprobes_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826}
827
828int __weak arch_dup_task_struct(struct task_struct *dst,
829 struct task_struct *src)
830{
831 *dst = *src;
832 return 0;
833}
834
835void set_task_stack_end_magic(struct task_struct *tsk)
836{
837 unsigned long *stackend;
838
839 stackend = end_of_stack(tsk);
840 *stackend = STACK_END_MAGIC; /* for overflow detection */
841}
842
843static struct task_struct *dup_task_struct(struct task_struct *orig, int node)
844{
845 struct task_struct *tsk;
846 unsigned long *stack;
YueHaibing0f4991e2018-12-28 00:40:00 -0800847 struct vm_struct *stack_vm_area __maybe_unused;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 int err;
849
850 if (node == NUMA_NO_NODE)
851 node = tsk_fork_get_node(orig);
852 tsk = alloc_task_struct_node(node);
853 if (!tsk)
854 return NULL;
855
856 stack = alloc_thread_stack_node(tsk, node);
857 if (!stack)
858 goto free_tsk;
859
Roman Gushchin9b6f7e12018-10-26 15:03:19 -0700860 if (memcg_charge_kernel_stack(tsk))
861 goto free_stack;
862
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 stack_vm_area = task_stack_vm_area(tsk);
864
865 err = arch_dup_task_struct(tsk, orig);
866
867 /*
868 * arch_dup_task_struct() clobbers the stack-related fields. Make
869 * sure they're properly initialized before using any stack-related
870 * functions again.
871 */
872 tsk->stack = stack;
873#ifdef CONFIG_VMAP_STACK
874 tsk->stack_vm_area = stack_vm_area;
875#endif
876#ifdef CONFIG_THREAD_INFO_IN_TASK
Elena Reshetovaf0b89d32019-01-18 14:27:30 +0200877 refcount_set(&tsk->stack_refcount, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878#endif
879
880 if (err)
881 goto free_stack;
882
883#ifdef CONFIG_SECCOMP
884 /*
885 * We must handle setting up seccomp filters once we're under
886 * the sighand lock in case orig has changed between now and
887 * then. Until then, filter must be NULL to avoid messing up
888 * the usage counts on the error path calling free_task.
889 */
890 tsk->seccomp.filter = NULL;
891#endif
892
893 setup_thread_stack(tsk, orig);
894 clear_user_return_notifier(tsk);
895 clear_tsk_need_resched(tsk);
896 set_task_stack_end_magic(tsk);
897
Linus Torvalds050e9ba2018-06-14 12:21:18 +0900898#ifdef CONFIG_STACKPROTECTOR
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 tsk->stack_canary = get_random_canary();
900#endif
Sebastian Andrzej Siewior3bd370622019-04-23 16:26:36 +0200901 if (orig->cpus_ptr == &orig->cpus_mask)
902 tsk->cpus_ptr = &tsk->cpus_mask;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903
904 /*
905 * One for us, one for whoever does the "release_task()" (usually
906 * parent)
907 */
Elena Reshetovaec1d2812019-01-18 14:27:29 +0200908 refcount_set(&tsk->usage, 2);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909#ifdef CONFIG_BLK_DEV_IO_TRACE
910 tsk->btrace_seq = 0;
911#endif
912 tsk->splice_pipe = NULL;
913 tsk->task_frag.page = NULL;
914 tsk->wake_q.next = NULL;
915
916 account_kernel_stack(tsk, 1);
917
918 kcov_task_init(tsk);
919
920#ifdef CONFIG_FAULT_INJECTION
921 tsk->fail_nth = 0;
922#endif
923
Josef Bacik2c323012018-07-31 12:39:04 -0400924#ifdef CONFIG_BLK_CGROUP
925 tsk->throttle_queue = NULL;
926 tsk->use_memdelay = 0;
927#endif
928
Shakeel Buttd46eb14b2018-08-17 15:46:39 -0700929#ifdef CONFIG_MEMCG
930 tsk->active_memcg = NULL;
931#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 return tsk;
933
934free_stack:
935 free_thread_stack(tsk);
936free_tsk:
937 free_task_struct(tsk);
938 return NULL;
939}
940
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941__cacheline_aligned_in_smp DEFINE_SPINLOCK(mmlist_lock);
942
Hidehiro Kawai4cb0e112009-01-06 14:42:47 -0800943static unsigned long default_dump_filter = MMF_DUMP_FILTER_DEFAULT;
944
945static int __init coredump_filter_setup(char *s)
946{
947 default_dump_filter =
948 (simple_strtoul(s, NULL, 0) << MMF_DUMP_FILTER_SHIFT) &
949 MMF_DUMP_FILTER_MASK;
950 return 1;
951}
952
953__setup("coredump_filter=", coredump_filter_setup);
954
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955#include <linux/init_task.h>
956
Alexey Dobriyan858f0992009-09-23 15:57:32 -0700957static void mm_init_aio(struct mm_struct *mm)
958{
959#ifdef CONFIG_AIO
960 spin_lock_init(&mm->ioctx_lock);
Benjamin LaHaisedb446a02013-07-30 12:54:40 -0400961 mm->ioctx_table = NULL;
Alexey Dobriyan858f0992009-09-23 15:57:32 -0700962#endif
963}
964
Andrea Arcangelic3f3ce02019-05-14 15:40:46 -0700965static __always_inline void mm_clear_owner(struct mm_struct *mm,
966 struct task_struct *p)
967{
968#ifdef CONFIG_MEMCG
969 if (mm->owner == p)
970 WRITE_ONCE(mm->owner, NULL);
971#endif
972}
973
Vladimir Davydov33144e82014-08-08 14:22:03 -0700974static void mm_init_owner(struct mm_struct *mm, struct task_struct *p)
975{
976#ifdef CONFIG_MEMCG
977 mm->owner = p;
978#endif
979}
980
Eric Biggers355627f2017-08-31 16:15:26 -0700981static void mm_init_uprobes_state(struct mm_struct *mm)
982{
983#ifdef CONFIG_UPROBES
984 mm->uprobes_state.xol_area = NULL;
985#endif
986}
987
Eric W. Biedermanbfedb582016-10-13 21:23:16 -0500988static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
989 struct user_namespace *user_ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990{
Vladimir Davydov41f727f2014-08-08 14:21:56 -0700991 mm->mmap = NULL;
992 mm->mm_rb = RB_ROOT;
993 mm->vmacache_seqnum = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 atomic_set(&mm->mm_users, 1);
995 atomic_set(&mm->mm_count, 1);
996 init_rwsem(&mm->mmap_sem);
997 INIT_LIST_HEAD(&mm->mmlist);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700998 mm->core_state = NULL;
Kirill A. Shutemovaf5b0f62017-11-15 17:35:40 -0800999 mm_pgtables_bytes_init(mm);
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001000 mm->map_count = 0;
1001 mm->locked_vm = 0;
Davidlohr Bueso70f8a3c2019-02-06 09:59:15 -08001002 atomic64_set(&mm->pinned_vm, 0);
KAMEZAWA Hiroyukid559db02010-03-05 13:41:39 -08001003 memset(&mm->rss_stat, 0, sizeof(mm->rss_stat));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 spin_lock_init(&mm->page_table_lock);
Yang Shi88aa7cc2018-06-07 17:05:28 -07001005 spin_lock_init(&mm->arg_lock);
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001006 mm_init_cpumask(mm);
Alexey Dobriyan858f0992009-09-23 15:57:32 -07001007 mm_init_aio(mm);
Balbir Singhcf475ad22008-04-29 01:00:16 -07001008 mm_init_owner(mm, p);
Eric Biggers2b7e8662017-08-25 15:55:43 -07001009 RCU_INIT_POINTER(mm->exe_file, NULL);
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001010 mmu_notifier_mm_init(mm);
Jérôme Glisse133ff0e2017-09-08 16:11:23 -07001011 hmm_mm_init(mm);
Nadav Amit16af97d2017-08-10 15:23:56 -07001012 init_tlb_flush_pending(mm);
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001013#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
1014 mm->pmd_huge_pte = NULL;
1015#endif
Eric Biggers355627f2017-08-31 16:15:26 -07001016 mm_init_uprobes_state(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017
Alex Thorltona0715cc2014-04-07 15:37:10 -07001018 if (current->mm) {
1019 mm->flags = current->mm->flags & MMF_INIT_MASK;
1020 mm->def_flags = current->mm->def_flags & VM_INIT_DEF_MASK;
1021 } else {
1022 mm->flags = default_dump_filter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 mm->def_flags = 0;
Alex Thorltona0715cc2014-04-07 15:37:10 -07001024 }
1025
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001026 if (mm_alloc_pgd(mm))
1027 goto fail_nopgd;
Pavel Emelianov78fb7462008-02-07 00:13:51 -08001028
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001029 if (init_new_context(p, mm))
1030 goto fail_nocontext;
1031
Eric W. Biedermanbfedb582016-10-13 21:23:16 -05001032 mm->user_ns = get_user_ns(user_ns);
Vladimir Davydov41f727f2014-08-08 14:21:56 -07001033 return mm;
1034
1035fail_nocontext:
1036 mm_free_pgd(mm);
1037fail_nopgd:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 free_mm(mm);
1039 return NULL;
1040}
1041
1042/*
1043 * Allocate and initialize an mm_struct.
1044 */
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001045struct mm_struct *mm_alloc(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046{
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001047 struct mm_struct *mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048
1049 mm = allocate_mm();
KOSAKI Motohirode03c722011-05-24 17:12:15 -07001050 if (!mm)
1051 return NULL;
1052
1053 memset(mm, 0, sizeof(*mm));
Eric W. Biedermanbfedb582016-10-13 21:23:16 -05001054 return mm_init(mm, current, current_user_ns());
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055}
1056
Michal Hockoec8d7c12016-05-20 16:57:21 -07001057static inline void __mmput(struct mm_struct *mm)
1058{
1059 VM_BUG_ON(atomic_read(&mm->mm_users));
1060
1061 uprobe_clear_state(mm);
1062 exit_aio(mm);
1063 ksm_exit(mm);
1064 khugepaged_exit(mm); /* must run before exit_mmap */
1065 exit_mmap(mm);
Aaron Lu6fcb52a2016-10-07 17:00:08 -07001066 mm_put_huge_zero_page(mm);
Michal Hockoec8d7c12016-05-20 16:57:21 -07001067 set_mm_exe_file(mm, NULL);
1068 if (!list_empty(&mm->mmlist)) {
1069 spin_lock(&mmlist_lock);
1070 list_del(&mm->mmlist);
1071 spin_unlock(&mmlist_lock);
1072 }
1073 if (mm->binfmt)
1074 module_put(mm->binfmt->module);
1075 mmdrop(mm);
1076}
1077
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078/*
1079 * Decrement the use count and release all resources for an mm.
1080 */
1081void mmput(struct mm_struct *mm)
1082{
Andrew Morton0ae26f12006-06-23 02:05:15 -07001083 might_sleep();
1084
Michal Hockoec8d7c12016-05-20 16:57:21 -07001085 if (atomic_dec_and_test(&mm->mm_users))
1086 __mmput(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087}
1088EXPORT_SYMBOL_GPL(mmput);
1089
Sherry Yanga1b22892017-10-03 16:15:00 -07001090#ifdef CONFIG_MMU
1091static void mmput_async_fn(struct work_struct *work)
1092{
1093 struct mm_struct *mm = container_of(work, struct mm_struct,
1094 async_put_work);
1095
1096 __mmput(mm);
1097}
1098
1099void mmput_async(struct mm_struct *mm)
1100{
1101 if (atomic_dec_and_test(&mm->mm_users)) {
1102 INIT_WORK(&mm->async_put_work, mmput_async_fn);
1103 schedule_work(&mm->async_put_work);
1104 }
1105}
1106#endif
1107
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -07001108/**
1109 * set_mm_exe_file - change a reference to the mm's executable file
1110 *
1111 * This changes mm's executable file (shown as symlink /proc/[pid]/exe).
1112 *
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001113 * Main users are mmput() and sys_execve(). Callers prevent concurrent
1114 * invocations: in mmput() nobody alive left, in execve task is single
1115 * threaded. sys_prctl(PR_SET_MM_MAP/EXE_FILE) also needs to set the
1116 * mm->exe_file, but does so without using set_mm_exe_file() in order
1117 * to do avoid the need for any locks.
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -07001118 */
Jiri Slaby38646012011-05-26 16:25:46 -07001119void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file)
1120{
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001121 struct file *old_exe_file;
1122
1123 /*
1124 * It is safe to dereference the exe_file without RCU as
1125 * this function is only called if nobody else can access
1126 * this mm -- see comment above for justification.
1127 */
1128 old_exe_file = rcu_dereference_raw(mm->exe_file);
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -07001129
Jiri Slaby38646012011-05-26 16:25:46 -07001130 if (new_exe_file)
1131 get_file(new_exe_file);
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -07001132 rcu_assign_pointer(mm->exe_file, new_exe_file);
1133 if (old_exe_file)
1134 fput(old_exe_file);
Jiri Slaby38646012011-05-26 16:25:46 -07001135}
1136
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -07001137/**
1138 * get_mm_exe_file - acquire a reference to the mm's executable file
1139 *
1140 * Returns %NULL if mm has no associated executable file.
1141 * User must release file via fput().
1142 */
Jiri Slaby38646012011-05-26 16:25:46 -07001143struct file *get_mm_exe_file(struct mm_struct *mm)
1144{
1145 struct file *exe_file;
1146
Konstantin Khlebnikov90f31d02015-04-16 12:47:56 -07001147 rcu_read_lock();
1148 exe_file = rcu_dereference(mm->exe_file);
1149 if (exe_file && !get_file_rcu(exe_file))
1150 exe_file = NULL;
1151 rcu_read_unlock();
Jiri Slaby38646012011-05-26 16:25:46 -07001152 return exe_file;
1153}
Davidlohr Bueso11163342015-04-16 12:49:12 -07001154EXPORT_SYMBOL(get_mm_exe_file);
Jiri Slaby38646012011-05-26 16:25:46 -07001155
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156/**
Mateusz Guzikcd81a9172016-08-23 16:20:38 +02001157 * get_task_exe_file - acquire a reference to the task's executable file
1158 *
1159 * Returns %NULL if task's mm (if any) has no associated executable file or
1160 * this is a kernel thread with borrowed mm (see the comment above get_task_mm).
1161 * User must release file via fput().
1162 */
1163struct file *get_task_exe_file(struct task_struct *task)
1164{
1165 struct file *exe_file = NULL;
1166 struct mm_struct *mm;
1167
1168 task_lock(task);
1169 mm = task->mm;
1170 if (mm) {
1171 if (!(task->flags & PF_KTHREAD))
1172 exe_file = get_mm_exe_file(mm);
1173 }
1174 task_unlock(task);
1175 return exe_file;
1176}
1177EXPORT_SYMBOL(get_task_exe_file);
1178
1179/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 * get_task_mm - acquire a reference to the task's mm
1181 *
Oleg Nesterov246bb0b2008-07-25 01:47:38 -07001182 * Returns %NULL if the task has no mm. Checks PF_KTHREAD (meaning
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 * this kernel workthread has transiently adopted a user mm with use_mm,
1184 * to do its AIO) is not set and if so returns a reference to it, after
1185 * bumping up the use count. User must release the mm via mmput()
1186 * after use. Typically used by /proc and ptrace.
1187 */
1188struct mm_struct *get_task_mm(struct task_struct *task)
1189{
1190 struct mm_struct *mm;
1191
1192 task_lock(task);
1193 mm = task->mm;
1194 if (mm) {
Oleg Nesterov246bb0b2008-07-25 01:47:38 -07001195 if (task->flags & PF_KTHREAD)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 mm = NULL;
1197 else
Vegard Nossum3fce3712017-02-27 14:30:10 -08001198 mmget(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 }
1200 task_unlock(task);
1201 return mm;
1202}
1203EXPORT_SYMBOL_GPL(get_task_mm);
1204
Christopher Yeoh8cdb8782012-02-02 11:34:09 +10301205struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
1206{
1207 struct mm_struct *mm;
1208 int err;
1209
1210 err = mutex_lock_killable(&task->signal->cred_guard_mutex);
1211 if (err)
1212 return ERR_PTR(err);
1213
1214 mm = get_task_mm(task);
1215 if (mm && mm != current->mm &&
1216 !ptrace_may_access(task, mode)) {
1217 mmput(mm);
1218 mm = ERR_PTR(-EACCES);
1219 }
1220 mutex_unlock(&task->signal->cred_guard_mutex);
1221
1222 return mm;
1223}
1224
Oleg Nesterov57b59c42012-03-05 14:59:13 -08001225static void complete_vfork_done(struct task_struct *tsk)
Oleg Nesterovc415c3b2012-03-05 14:59:13 -08001226{
Oleg Nesterovd68b46f2012-03-05 14:59:13 -08001227 struct completion *vfork;
Oleg Nesterovc415c3b2012-03-05 14:59:13 -08001228
Oleg Nesterovd68b46f2012-03-05 14:59:13 -08001229 task_lock(tsk);
1230 vfork = tsk->vfork_done;
1231 if (likely(vfork)) {
1232 tsk->vfork_done = NULL;
1233 complete(vfork);
1234 }
1235 task_unlock(tsk);
1236}
1237
1238static int wait_for_vfork_done(struct task_struct *child,
1239 struct completion *vfork)
1240{
1241 int killed;
1242
1243 freezer_do_not_count();
Roman Gushchin76f969e2019-04-19 10:03:04 -07001244 cgroup_enter_frozen();
Oleg Nesterovd68b46f2012-03-05 14:59:13 -08001245 killed = wait_for_completion_killable(vfork);
Roman Gushchin76f969e2019-04-19 10:03:04 -07001246 cgroup_leave_frozen(false);
Oleg Nesterovd68b46f2012-03-05 14:59:13 -08001247 freezer_count();
1248
1249 if (killed) {
1250 task_lock(child);
1251 child->vfork_done = NULL;
1252 task_unlock(child);
1253 }
1254
1255 put_task_struct(child);
1256 return killed;
Oleg Nesterovc415c3b2012-03-05 14:59:13 -08001257}
1258
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259/* Please note the differences between mmput and mm_release.
1260 * mmput is called whenever we stop holding onto a mm_struct,
1261 * error success whatever.
1262 *
1263 * mm_release is called after a mm_struct has been removed
1264 * from the current process.
1265 *
1266 * This difference is important for error handling, when we
1267 * only half set up a mm_struct for a new process and need to restore
1268 * the old one. Because we mmput the new mm_struct before
1269 * restoring the old one. . .
1270 * Eric Biederman 10 January 1998
1271 */
1272void mm_release(struct task_struct *tsk, struct mm_struct *mm)
1273{
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001274 /* Get rid of any futexes when releasing the mm */
1275#ifdef CONFIG_FUTEX
Peter Zijlstrafc6b1772009-10-05 18:17:32 +02001276 if (unlikely(tsk->robust_list)) {
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001277 exit_robust_list(tsk);
Peter Zijlstrafc6b1772009-10-05 18:17:32 +02001278 tsk->robust_list = NULL;
1279 }
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001280#ifdef CONFIG_COMPAT
Peter Zijlstrafc6b1772009-10-05 18:17:32 +02001281 if (unlikely(tsk->compat_robust_list)) {
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001282 compat_exit_robust_list(tsk);
Peter Zijlstrafc6b1772009-10-05 18:17:32 +02001283 tsk->compat_robust_list = NULL;
1284 }
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001285#endif
Thomas Gleixner322a2c12009-10-05 18:18:03 +02001286 if (unlikely(!list_empty(&tsk->pi_state_list)))
1287 exit_pi_state_list(tsk);
Linus Torvalds8141c7f2008-11-15 10:20:36 -08001288#endif
1289
Srikar Dronamraju0326f5a2012-03-13 23:30:11 +05301290 uprobe_free_utask(tsk);
1291
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 /* Get rid of any cached register state */
1293 deactivate_mm(tsk, mm);
1294
Roland McGrathfec1d012006-12-06 20:36:34 -08001295 /*
Michal Hocko735f2772016-09-01 16:15:13 -07001296 * Signal userspace if we're not exiting with a core dump
1297 * because we want to leave the value intact for debugging
1298 * purposes.
Roland McGrathfec1d012006-12-06 20:36:34 -08001299 */
Eric Dumazet9c8a8222009-08-06 15:09:28 -07001300 if (tsk->clear_child_tid) {
Michal Hocko735f2772016-09-01 16:15:13 -07001301 if (!(tsk->signal->flags & SIGNAL_GROUP_COREDUMP) &&
Eric Dumazet9c8a8222009-08-06 15:09:28 -07001302 atomic_read(&mm->mm_users) > 1) {
1303 /*
1304 * We don't check the error code - if userspace has
1305 * not set up a proper pointer then tough luck.
1306 */
1307 put_user(0, tsk->clear_child_tid);
Dominik Brodowski2de0db92018-03-11 11:34:26 +01001308 do_futex(tsk->clear_child_tid, FUTEX_WAKE,
1309 1, NULL, NULL, 0, 0);
Eric Dumazet9c8a8222009-08-06 15:09:28 -07001310 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 tsk->clear_child_tid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 }
Konstantin Khlebnikovf7505d642012-05-31 16:26:21 -07001313
1314 /*
1315 * All done, finally we can wake up parent and return this mm to him.
1316 * Also kthread_stop() uses this completion for synchronization.
1317 */
1318 if (tsk->vfork_done)
1319 complete_vfork_done(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320}
1321
Nadav Amit13585fa2019-04-25 17:11:25 -07001322/**
1323 * dup_mm() - duplicates an existing mm structure
1324 * @tsk: the task_struct with which the new mm will be associated.
1325 * @oldmm: the mm to duplicate.
1326 *
1327 * Allocates a new mm structure and duplicates the provided @oldmm structure
1328 * content into it.
1329 *
1330 * Return: the duplicated mm or NULL on failure.
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001331 */
Nadav Amit13585fa2019-04-25 17:11:25 -07001332static struct mm_struct *dup_mm(struct task_struct *tsk,
1333 struct mm_struct *oldmm)
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001334{
Nadav Amit13585fa2019-04-25 17:11:25 -07001335 struct mm_struct *mm;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001336 int err;
1337
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001338 mm = allocate_mm();
1339 if (!mm)
1340 goto fail_nomem;
1341
1342 memcpy(mm, oldmm, sizeof(*mm));
1343
Eric W. Biedermanbfedb582016-10-13 21:23:16 -05001344 if (!mm_init(mm, tsk, mm->user_ns))
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001345 goto fail_nomem;
1346
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001347 err = dup_mmap(mm, oldmm);
1348 if (err)
1349 goto free_pt;
1350
1351 mm->hiwater_rss = get_mm_rss(mm);
1352 mm->hiwater_vm = mm->total_vm;
1353
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001354 if (mm->binfmt && !try_module_get(mm->binfmt->module))
1355 goto free_pt;
1356
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001357 return mm;
1358
1359free_pt:
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001360 /* don't put binfmt in mmput, we haven't got module yet */
1361 mm->binfmt = NULL;
Andrea Arcangelic3f3ce02019-05-14 15:40:46 -07001362 mm_init_owner(mm, NULL);
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001363 mmput(mm);
1364
1365fail_nomem:
1366 return NULL;
JANAK DESAIa0a7ec32006-02-07 12:59:01 -08001367}
1368
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001369static int copy_mm(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370{
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001371 struct mm_struct *mm, *oldmm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 int retval;
1373
1374 tsk->min_flt = tsk->maj_flt = 0;
1375 tsk->nvcsw = tsk->nivcsw = 0;
Mandeep Singh Baines17406b82009-02-06 15:37:47 -08001376#ifdef CONFIG_DETECT_HUNG_TASK
1377 tsk->last_switch_count = tsk->nvcsw + tsk->nivcsw;
Dmitry Vyukova2e51442018-08-21 21:55:52 -07001378 tsk->last_switch_time = 0;
Mandeep Singh Baines17406b82009-02-06 15:37:47 -08001379#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380
1381 tsk->mm = NULL;
1382 tsk->active_mm = NULL;
1383
1384 /*
1385 * Are we cloning a kernel thread?
1386 *
1387 * We need to steal a active VM for that..
1388 */
1389 oldmm = current->mm;
1390 if (!oldmm)
1391 return 0;
1392
Davidlohr Bueso615d6e82014-04-07 15:37:25 -07001393 /* initialize the new vmacache entries */
1394 vmacache_flush(tsk);
1395
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 if (clone_flags & CLONE_VM) {
Vegard Nossum3fce3712017-02-27 14:30:10 -08001397 mmget(oldmm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 mm = oldmm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 goto good_mm;
1400 }
1401
1402 retval = -ENOMEM;
Nadav Amit13585fa2019-04-25 17:11:25 -07001403 mm = dup_mm(tsk, current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 if (!mm)
1405 goto fail_nomem;
1406
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407good_mm:
1408 tsk->mm = mm;
1409 tsk->active_mm = mm;
1410 return 0;
1411
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412fail_nomem:
1413 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414}
1415
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001416static int copy_fs(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417{
Al Viro498052b2009-03-30 07:20:30 -04001418 struct fs_struct *fs = current->fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 if (clone_flags & CLONE_FS) {
Al Viro498052b2009-03-30 07:20:30 -04001420 /* tsk->fs is already what we want */
Nick Piggin2a4419b2010-08-18 04:37:33 +10001421 spin_lock(&fs->lock);
Al Viro498052b2009-03-30 07:20:30 -04001422 if (fs->in_exec) {
Nick Piggin2a4419b2010-08-18 04:37:33 +10001423 spin_unlock(&fs->lock);
Al Viro498052b2009-03-30 07:20:30 -04001424 return -EAGAIN;
1425 }
1426 fs->users++;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001427 spin_unlock(&fs->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 return 0;
1429 }
Al Viro498052b2009-03-30 07:20:30 -04001430 tsk->fs = copy_fs_struct(fs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431 if (!tsk->fs)
1432 return -ENOMEM;
1433 return 0;
1434}
1435
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001436static int copy_files(unsigned long clone_flags, struct task_struct *tsk)
JANAK DESAIa016f332006-02-07 12:59:02 -08001437{
1438 struct files_struct *oldf, *newf;
1439 int error = 0;
1440
1441 /*
1442 * A background process may not have any files ...
1443 */
1444 oldf = current->files;
1445 if (!oldf)
1446 goto out;
1447
1448 if (clone_flags & CLONE_FILES) {
1449 atomic_inc(&oldf->count);
1450 goto out;
1451 }
1452
JANAK DESAIa016f332006-02-07 12:59:02 -08001453 newf = dup_fd(oldf, &error);
1454 if (!newf)
1455 goto out;
1456
1457 tsk->files = newf;
1458 error = 0;
1459out:
1460 return error;
1461}
1462
Jens Axboefadad8782008-01-24 08:54:47 +01001463static int copy_io(unsigned long clone_flags, struct task_struct *tsk)
Jens Axboefd0928d2008-01-24 08:52:45 +01001464{
1465#ifdef CONFIG_BLOCK
1466 struct io_context *ioc = current->io_context;
Tejun Heo6e736be2011-12-14 00:33:38 +01001467 struct io_context *new_ioc;
Jens Axboefd0928d2008-01-24 08:52:45 +01001468
1469 if (!ioc)
1470 return 0;
Jens Axboefadad8782008-01-24 08:54:47 +01001471 /*
1472 * Share io context with parent, if CLONE_IO is set
1473 */
1474 if (clone_flags & CLONE_IO) {
Tejun Heo3d487492012-03-05 13:15:25 -08001475 ioc_task_link(ioc);
1476 tsk->io_context = ioc;
Jens Axboefadad8782008-01-24 08:54:47 +01001477 } else if (ioprio_valid(ioc->ioprio)) {
Tejun Heo6e736be2011-12-14 00:33:38 +01001478 new_ioc = get_task_io_context(tsk, GFP_KERNEL, NUMA_NO_NODE);
1479 if (unlikely(!new_ioc))
Jens Axboefd0928d2008-01-24 08:52:45 +01001480 return -ENOMEM;
1481
Tejun Heo6e736be2011-12-14 00:33:38 +01001482 new_ioc->ioprio = ioc->ioprio;
Tejun Heo11a31222012-02-07 07:51:30 +01001483 put_io_context(new_ioc);
Jens Axboefd0928d2008-01-24 08:52:45 +01001484 }
1485#endif
1486 return 0;
1487}
1488
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001489static int copy_sighand(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490{
1491 struct sighand_struct *sig;
1492
Zhaolei60348802009-01-06 14:40:46 -08001493 if (clone_flags & CLONE_SIGHAND) {
Elena Reshetovad036bda2019-01-18 14:27:26 +02001494 refcount_inc(&current->sighand->count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 return 0;
1496 }
1497 sig = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
Ingo Molnare56d0902006-01-08 01:01:37 -08001498 rcu_assign_pointer(tsk->sighand, sig);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 if (!sig)
1500 return -ENOMEM;
Peter Zijlstra9d7fb042015-06-30 11:30:54 +02001501
Elena Reshetovad036bda2019-01-18 14:27:26 +02001502 refcount_set(&sig->count, 1);
Jann Horn06e62a42018-08-21 22:00:58 -07001503 spin_lock_irq(&current->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 memcpy(sig->action, current->sighand->action, sizeof(sig->action));
Jann Horn06e62a42018-08-21 22:00:58 -07001505 spin_unlock_irq(&current->sighand->siglock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 return 0;
1507}
1508
Oleg Nesterova7e53282006-03-28 16:11:27 -08001509void __cleanup_sighand(struct sighand_struct *sighand)
Oleg Nesterovc81addc2006-03-28 16:11:17 -08001510{
Elena Reshetovad036bda2019-01-18 14:27:26 +02001511 if (refcount_dec_and_test(&sighand->count)) {
Oleg Nesterovd80e7312012-02-24 20:07:11 +01001512 signalfd_cleanup(sighand);
Oleg Nesterov392809b2014-09-28 23:44:18 +02001513 /*
Paul E. McKenney5f0d5a32017-01-18 02:53:44 -08001514 * sighand_cachep is SLAB_TYPESAFE_BY_RCU so we can free it
Oleg Nesterov392809b2014-09-28 23:44:18 +02001515 * without an RCU grace period, see __lock_task_sighand().
1516 */
Oleg Nesterovc81addc2006-03-28 16:11:17 -08001517 kmem_cache_free(sighand_cachep, sighand);
Oleg Nesterovd80e7312012-02-24 20:07:11 +01001518 }
Oleg Nesterovc81addc2006-03-28 16:11:17 -08001519}
1520
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05001521#ifdef CONFIG_POSIX_TIMERS
Frank Mayharf06febc2008-09-12 09:54:39 -07001522/*
1523 * Initialize POSIX timer handling for a thread group.
1524 */
1525static void posix_cpu_timers_init_group(struct signal_struct *sig)
1526{
Jiri Slaby78d7d402010-03-05 13:42:54 -08001527 unsigned long cpu_limit;
1528
Jason Low316c1608d2015-04-28 13:00:20 -07001529 cpu_limit = READ_ONCE(sig->rlim[RLIMIT_CPU].rlim_cur);
Jiri Slaby78d7d402010-03-05 13:42:54 -08001530 if (cpu_limit != RLIM_INFINITY) {
Frederic Weisbeckerebd7e7f2017-01-31 04:09:34 +01001531 sig->cputime_expires.prof_exp = cpu_limit * NSEC_PER_SEC;
Jason Lowd5c373e2015-10-14 12:07:55 -07001532 sig->cputimer.running = true;
Oleg Nesterov6279a7512009-03-27 01:06:07 +01001533 }
1534
Frank Mayharf06febc2008-09-12 09:54:39 -07001535 /* The timer lists. */
1536 INIT_LIST_HEAD(&sig->cpu_timers[0]);
1537 INIT_LIST_HEAD(&sig->cpu_timers[1]);
1538 INIT_LIST_HEAD(&sig->cpu_timers[2]);
1539}
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05001540#else
1541static inline void posix_cpu_timers_init_group(struct signal_struct *sig) { }
1542#endif
Frank Mayharf06febc2008-09-12 09:54:39 -07001543
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001544static int copy_signal(unsigned long clone_flags, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545{
1546 struct signal_struct *sig;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547
Oleg Nesterov4ab6c082009-08-26 14:29:24 -07001548 if (clone_flags & CLONE_THREAD)
Peter Zijlstra490dea42008-11-24 17:06:57 +01001549 return 0;
Oleg Nesterov6279a7512009-03-27 01:06:07 +01001550
Veaceslav Falicoa56704e2010-03-10 15:23:01 -08001551 sig = kmem_cache_zalloc(signal_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 tsk->signal = sig;
1553 if (!sig)
1554 return -ENOMEM;
1555
Oleg Nesterovb3ac0222010-05-26 14:43:24 -07001556 sig->nr_threads = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 atomic_set(&sig->live, 1);
Elena Reshetova60d4de32019-01-18 14:27:27 +02001558 refcount_set(&sig->sigcnt, 1);
Oleg Nesterov0c740d02014-01-21 15:49:56 -08001559
1560 /* list_add(thread_node, thread_head) without INIT_LIST_HEAD() */
1561 sig->thread_head = (struct list_head)LIST_HEAD_INIT(tsk->thread_node);
1562 tsk->thread_node = (struct list_head)LIST_HEAD_INIT(sig->thread_head);
1563
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564 init_waitqueue_head(&sig->wait_chldexit);
Oleg Nesterovdb51aec2008-04-30 00:52:52 -07001565 sig->curr_target = tsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566 init_sigpending(&sig->shared_pending);
Eric W. Biedermanc3ad2c32018-07-23 15:20:37 -05001567 INIT_HLIST_HEAD(&sig->multiprocess);
Rik van Riele78c3492014-08-16 13:40:10 -04001568 seqlock_init(&sig->stats_lock);
Peter Zijlstra9d7fb042015-06-30 11:30:54 +02001569 prev_cputime_init(&sig->prev_cputime);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570
Nicolas Pitrebaa73d92016-11-11 00:10:10 -05001571#ifdef CONFIG_POSIX_TIMERS
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05001572 INIT_LIST_HEAD(&sig->posix_timers);
Thomas Gleixnerc9cb2e32007-02-16 01:27:49 -08001573 hrtimer_init(&sig->real_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 sig->real_timer.function = it_real_fn;
Nicolas Pitrebaa73d92016-11-11 00:10:10 -05001575#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 task_lock(current->group_leader);
1578 memcpy(sig->rlim, current->signal->rlim, sizeof sig->rlim);
1579 task_unlock(current->group_leader);
1580
Oleg Nesterov6279a7512009-03-27 01:06:07 +01001581 posix_cpu_timers_init_group(sig);
1582
Miloslav Trmac522ed772007-07-15 23:40:56 -07001583 tty_audit_fork(sig);
Mike Galbraith5091faa2010-11-30 14:18:03 +01001584 sched_autogroup_fork(sig);
Miloslav Trmac522ed772007-07-15 23:40:56 -07001585
David Rientjesa63d83f2010-08-09 17:19:46 -07001586 sig->oom_score_adj = current->signal->oom_score_adj;
Mandeep Singh Bainesdabb16f632011-01-13 15:46:05 -08001587 sig->oom_score_adj_min = current->signal->oom_score_adj_min;
KOSAKI Motohiro28b83c52009-09-21 17:03:13 -07001588
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001589 mutex_init(&sig->cred_guard_mutex);
1590
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591 return 0;
1592}
1593
Kees Cookdbd952122014-06-27 15:18:48 -07001594static void copy_seccomp(struct task_struct *p)
1595{
1596#ifdef CONFIG_SECCOMP
1597 /*
1598 * Must be called with sighand->lock held, which is common to
1599 * all threads in the group. Holding cred_guard_mutex is not
1600 * needed because this new task is not yet running and cannot
1601 * be racing exec.
1602 */
Guenter Roeck69f6a342014-08-10 20:50:30 -07001603 assert_spin_locked(&current->sighand->siglock);
Kees Cookdbd952122014-06-27 15:18:48 -07001604
1605 /* Ref-count the new filter user, and assign it. */
1606 get_seccomp_filter(current);
1607 p->seccomp = current->seccomp;
1608
1609 /*
1610 * Explicitly enable no_new_privs here in case it got set
1611 * between the task_struct being duplicated and holding the
1612 * sighand lock. The seccomp state and nnp must be in sync.
1613 */
1614 if (task_no_new_privs(current))
1615 task_set_no_new_privs(p);
1616
1617 /*
1618 * If the parent gained a seccomp mode after copying thread
1619 * flags and between before we held the sighand lock, we have
1620 * to manually enable the seccomp thread flag here.
1621 */
1622 if (p->seccomp.mode != SECCOMP_MODE_DISABLED)
1623 set_tsk_thread_flag(p, TIF_SECCOMP);
1624#endif
1625}
1626
Heiko Carstens17da2bd2009-01-14 14:14:10 +01001627SYSCALL_DEFINE1(set_tid_address, int __user *, tidptr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628{
1629 current->clear_child_tid = tidptr;
1630
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001631 return task_pid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632}
1633
Alexey Dobriyana39bc512007-10-18 23:41:10 -07001634static void rt_mutex_init_task(struct task_struct *p)
Ingo Molnar23f78d4a2006-06-27 02:54:53 -07001635{
Thomas Gleixner1d615482009-11-17 14:54:03 +01001636 raw_spin_lock_init(&p->pi_lock);
Zilvinas Valinskase29e1752007-03-16 13:38:34 -08001637#ifdef CONFIG_RT_MUTEXES
Davidlohr Buesoa23ba902017-09-08 16:15:01 -07001638 p->pi_waiters = RB_ROOT_CACHED;
Xunlei Pange96a77052017-03-23 15:56:08 +01001639 p->pi_top_task = NULL;
Ingo Molnar23f78d4a2006-06-27 02:54:53 -07001640 p->pi_blocked_on = NULL;
Ingo Molnar23f78d4a2006-06-27 02:54:53 -07001641#endif
1642}
1643
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05001644#ifdef CONFIG_POSIX_TIMERS
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645/*
Frank Mayharf06febc2008-09-12 09:54:39 -07001646 * Initialize POSIX timer handling for a single task.
1647 */
1648static void posix_cpu_timers_init(struct task_struct *tsk)
1649{
Martin Schwidefsky64861632011-12-15 14:56:09 +01001650 tsk->cputime_expires.prof_exp = 0;
1651 tsk->cputime_expires.virt_exp = 0;
Frank Mayharf06febc2008-09-12 09:54:39 -07001652 tsk->cputime_expires.sched_exp = 0;
1653 INIT_LIST_HEAD(&tsk->cpu_timers[0]);
1654 INIT_LIST_HEAD(&tsk->cpu_timers[1]);
1655 INIT_LIST_HEAD(&tsk->cpu_timers[2]);
1656}
Nicolas Pitreb18b6a92017-01-21 00:09:08 -05001657#else
1658static inline void posix_cpu_timers_init(struct task_struct *tsk) { }
1659#endif
Frank Mayharf06febc2008-09-12 09:54:39 -07001660
Eric W. Biederman2c470472017-09-26 13:06:43 -05001661static inline void init_task_pid_links(struct task_struct *task)
1662{
1663 enum pid_type type;
1664
1665 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
1666 INIT_HLIST_NODE(&task->pid_links[type]);
1667 }
1668}
1669
Oleg Nesterov81907732013-07-03 15:08:31 -07001670static inline void
1671init_task_pid(struct task_struct *task, enum pid_type type, struct pid *pid)
1672{
Eric W. Biederman2c470472017-09-26 13:06:43 -05001673 if (type == PIDTYPE_PID)
1674 task->thread_pid = pid;
1675 else
1676 task->signal->pids[type] = pid;
Oleg Nesterov81907732013-07-03 15:08:31 -07001677}
1678
Ingo Molnar6bfbaa52017-02-03 21:37:49 +01001679static inline void rcu_copy_process(struct task_struct *p)
1680{
1681#ifdef CONFIG_PREEMPT_RCU
1682 p->rcu_read_lock_nesting = 0;
1683 p->rcu_read_unlock_special.s = 0;
1684 p->rcu_blocked_node = NULL;
1685 INIT_LIST_HEAD(&p->rcu_node_entry);
1686#endif /* #ifdef CONFIG_PREEMPT_RCU */
1687#ifdef CONFIG_TASKS_RCU
1688 p->rcu_tasks_holdout = false;
1689 INIT_LIST_HEAD(&p->rcu_tasks_holdout_list);
1690 p->rcu_tasks_idle_cpu = -1;
1691#endif /* #ifdef CONFIG_TASKS_RCU */
1692}
1693
Christian Braunerb3e583822019-03-27 13:04:15 +01001694static int pidfd_release(struct inode *inode, struct file *file)
1695{
1696 struct pid *pid = file->private_data;
1697
1698 file->private_data = NULL;
1699 put_pid(pid);
1700 return 0;
1701}
1702
1703#ifdef CONFIG_PROC_FS
1704static void pidfd_show_fdinfo(struct seq_file *m, struct file *f)
1705{
1706 struct pid_namespace *ns = proc_pid_ns(file_inode(m->file));
1707 struct pid *pid = f->private_data;
1708
1709 seq_put_decimal_ull(m, "Pid:\t", pid_nr_ns(pid, ns));
1710 seq_putc(m, '\n');
1711}
1712#endif
1713
Joel Fernandes (Google)b53b0b92019-04-30 12:21:53 -04001714/*
1715 * Poll support for process exit notification.
1716 */
1717static unsigned int pidfd_poll(struct file *file, struct poll_table_struct *pts)
1718{
1719 struct task_struct *task;
1720 struct pid *pid = file->private_data;
1721 int poll_flags = 0;
1722
1723 poll_wait(file, &pid->wait_pidfd, pts);
1724
1725 rcu_read_lock();
1726 task = pid_task(pid, PIDTYPE_PID);
1727 /*
1728 * Inform pollers only when the whole thread group exits.
1729 * If the thread group leader exits before all other threads in the
1730 * group, then poll(2) should block, similar to the wait(2) family.
1731 */
1732 if (!task || (task->exit_state && thread_group_empty(task)))
1733 poll_flags = POLLIN | POLLRDNORM;
1734 rcu_read_unlock();
1735
1736 return poll_flags;
1737}
1738
Christian Braunerb3e583822019-03-27 13:04:15 +01001739const struct file_operations pidfd_fops = {
1740 .release = pidfd_release,
Joel Fernandes (Google)b53b0b92019-04-30 12:21:53 -04001741 .poll = pidfd_poll,
Christian Braunerb3e583822019-03-27 13:04:15 +01001742#ifdef CONFIG_PROC_FS
1743 .show_fdinfo = pidfd_show_fdinfo,
1744#endif
1745};
1746
Andrea Arcangelic3f3ce02019-05-14 15:40:46 -07001747static void __delayed_free_task(struct rcu_head *rhp)
1748{
1749 struct task_struct *tsk = container_of(rhp, struct task_struct, rcu);
1750
1751 free_task(tsk);
1752}
1753
1754static __always_inline void delayed_free_task(struct task_struct *tsk)
1755{
1756 if (IS_ENABLED(CONFIG_MEMCG))
1757 call_rcu(&tsk->rcu, __delayed_free_task);
1758 else
1759 free_task(tsk);
1760}
1761
Frank Mayharf06febc2008-09-12 09:54:39 -07001762/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 * This creates a new process as a copy of the old one,
1764 * but does not actually start it yet.
1765 *
1766 * It copies the registers, and all the appropriate
1767 * parts of the process environment (as per the clone
1768 * flags). The actual kick-off is left to the caller.
1769 */
Emese Revfy0766f782016-06-20 20:42:34 +02001770static __latent_entropy struct task_struct *copy_process(
Roland McGrath09a05392008-07-25 19:45:47 -07001771 struct pid *pid,
Josh Triplett3033f14a2015-06-25 15:01:19 -07001772 int trace,
Christian Brauner7f192e32019-05-25 11:36:41 +02001773 int node,
1774 struct kernel_clone_args *args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775{
Christian Braunerb3e583822019-03-27 13:04:15 +01001776 int pidfd = -1, retval;
Mariusz Kozlowskia24efe62007-10-18 23:41:09 -07001777 struct task_struct *p;
Eric W. Biedermanc3ad2c32018-07-23 15:20:37 -05001778 struct multiprocess_signals delayed;
Al Viro6fd2fe42019-06-26 22:22:09 -04001779 struct file *pidfile = NULL;
Christian Brauner7f192e32019-05-25 11:36:41 +02001780 u64 clone_flags = args->flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781
Marcos Paulo de Souza667b6092018-02-06 15:39:34 -08001782 /*
1783 * Don't allow sharing the root directory with processes in a different
1784 * namespace
1785 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786 if ((clone_flags & (CLONE_NEWNS|CLONE_FS)) == (CLONE_NEWNS|CLONE_FS))
1787 return ERR_PTR(-EINVAL);
1788
Eric W. Biedermane66eded2013-03-13 11:51:49 -07001789 if ((clone_flags & (CLONE_NEWUSER|CLONE_FS)) == (CLONE_NEWUSER|CLONE_FS))
1790 return ERR_PTR(-EINVAL);
1791
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792 /*
1793 * Thread groups must share signals as well, and detached threads
1794 * can only be started up within the thread group.
1795 */
1796 if ((clone_flags & CLONE_THREAD) && !(clone_flags & CLONE_SIGHAND))
1797 return ERR_PTR(-EINVAL);
1798
1799 /*
1800 * Shared signal handlers imply shared VM. By way of the above,
1801 * thread groups also imply shared VM. Blocking this case allows
1802 * for various simplifications in other code.
1803 */
1804 if ((clone_flags & CLONE_SIGHAND) && !(clone_flags & CLONE_VM))
1805 return ERR_PTR(-EINVAL);
1806
Sukadev Bhattiprolu123be072009-09-23 15:57:20 -07001807 /*
1808 * Siblings of global init remain as zombies on exit since they are
1809 * not reaped by their parent (swapper). To solve this and to avoid
1810 * multi-rooted process trees, prevent global and container-inits
1811 * from creating siblings.
1812 */
1813 if ((clone_flags & CLONE_PARENT) &&
1814 current->signal->flags & SIGNAL_UNKILLABLE)
1815 return ERR_PTR(-EINVAL);
1816
Eric W. Biederman8382fca2012-12-20 19:26:06 -08001817 /*
Oleg Nesterov40a0d322013-09-11 14:19:41 -07001818 * If the new process will be in a different pid or user namespace
Eric W. Biedermanfaf00da2015-08-10 18:25:44 -05001819 * do not allow it to share a thread group with the forking task.
Eric W. Biederman8382fca2012-12-20 19:26:06 -08001820 */
Eric W. Biedermanfaf00da2015-08-10 18:25:44 -05001821 if (clone_flags & CLONE_THREAD) {
Oleg Nesterov40a0d322013-09-11 14:19:41 -07001822 if ((clone_flags & (CLONE_NEWUSER | CLONE_NEWPID)) ||
1823 (task_active_pid_ns(current) !=
1824 current->nsproxy->pid_ns_for_children))
1825 return ERR_PTR(-EINVAL);
1826 }
Eric W. Biederman8382fca2012-12-20 19:26:06 -08001827
Christian Braunerb3e583822019-03-27 13:04:15 +01001828 if (clone_flags & CLONE_PIDFD) {
Christian Braunerb3e583822019-03-27 13:04:15 +01001829 /*
Christian Braunerb3e583822019-03-27 13:04:15 +01001830 * - CLONE_DETACHED is blocked so that we can potentially
1831 * reuse it later for CLONE_PIDFD.
1832 * - CLONE_THREAD is blocked until someone really needs it.
1833 */
Christian Brauner7f192e32019-05-25 11:36:41 +02001834 if (clone_flags & (CLONE_DETACHED | CLONE_THREAD))
Christian Braunerb3e583822019-03-27 13:04:15 +01001835 return ERR_PTR(-EINVAL);
Christian Braunerb3e583822019-03-27 13:04:15 +01001836 }
1837
Eric W. Biedermanc3ad2c32018-07-23 15:20:37 -05001838 /*
1839 * Force any signals received before this point to be delivered
1840 * before the fork happens. Collect up signals sent to multiple
1841 * processes that happen during the fork and delay them so that
1842 * they appear to happen after the fork.
1843 */
1844 sigemptyset(&delayed.signal);
1845 INIT_HLIST_NODE(&delayed.node);
1846
1847 spin_lock_irq(&current->sighand->siglock);
1848 if (!(clone_flags & CLONE_THREAD))
1849 hlist_add_head(&delayed.node, &current->signal->multiprocess);
1850 recalc_sigpending();
1851 spin_unlock_irq(&current->sighand->siglock);
1852 retval = -ERESTARTNOINTR;
1853 if (signal_pending(current))
1854 goto fork_out;
1855
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856 retval = -ENOMEM;
Andi Kleen725fc622016-05-23 16:24:05 -07001857 p = dup_task_struct(current, node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858 if (!p)
1859 goto fork_out;
1860
Vegard Nossum4d6501dc2017-05-09 09:39:59 +02001861 /*
1862 * This _must_ happen before we call free_task(), i.e. before we jump
1863 * to any of the bad_fork_* labels. This is to avoid freeing
1864 * p->set_child_tid which is (ab)used as a kthread's data pointer for
1865 * kernel threads (PF_KTHREAD).
1866 */
Christian Brauner7f192e32019-05-25 11:36:41 +02001867 p->set_child_tid = (clone_flags & CLONE_CHILD_SETTID) ? args->child_tid : NULL;
Vegard Nossum4d6501dc2017-05-09 09:39:59 +02001868 /*
1869 * Clear TID on mm_release()?
1870 */
Christian Brauner7f192e32019-05-25 11:36:41 +02001871 p->clear_child_tid = (clone_flags & CLONE_CHILD_CLEARTID) ? args->child_tid : NULL;
Vegard Nossum4d6501dc2017-05-09 09:39:59 +02001872
Steven Rostedtf7e8b612009-06-02 16:39:48 -04001873 ftrace_graph_init_task(p);
1874
Peter Zijlstrabea493a2006-10-17 00:10:33 -07001875 rt_mutex_init_task(p);
1876
Ingo Molnard12c1a32008-07-14 12:09:28 +02001877#ifdef CONFIG_PROVE_LOCKING
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001878 DEBUG_LOCKS_WARN_ON(!p->hardirqs_enabled);
1879 DEBUG_LOCKS_WARN_ON(!p->softirqs_enabled);
1880#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881 retval = -EAGAIN;
David Howells3b11a1d2008-11-14 10:39:26 +11001882 if (atomic_read(&p->real_cred->user->processes) >=
Jiri Slaby78d7d402010-03-05 13:42:54 -08001883 task_rlimit(p, RLIMIT_NPROC)) {
Eric Parisb57922b2013-07-03 15:08:29 -07001884 if (p->real_cred->user != INIT_USER &&
1885 !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886 goto bad_fork_free;
1887 }
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001888 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889
David Howellsf1752ee2008-11-14 10:39:17 +11001890 retval = copy_creds(p, clone_flags);
1891 if (retval < 0)
1892 goto bad_fork_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893
1894 /*
1895 * If multiple threads are within copy_process(), then this check
1896 * triggers too late. This doesn't hurt, the check is only there
1897 * to stop root fork bombs.
1898 */
Li Zefan04ec93f2009-02-06 08:17:19 +00001899 retval = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 if (nr_threads >= max_threads)
1901 goto bad_fork_cleanup_count;
1902
Shailabh Nagarca74e922006-07-14 00:24:36 -07001903 delayacct_tsk_init(p); /* Must remain after dup_task_struct() */
Peter Zijlstrac1de45c2016-11-28 23:03:05 -08001904 p->flags &= ~(PF_SUPERPRIV | PF_WQ_WORKER | PF_IDLE);
David Rientjes514ddb42014-04-07 15:37:27 -07001905 p->flags |= PF_FORKNOEXEC;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906 INIT_LIST_HEAD(&p->children);
1907 INIT_LIST_HEAD(&p->sibling);
Paul E. McKenneyf41d9112009-08-22 13:56:52 -07001908 rcu_copy_process(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909 p->vfork_done = NULL;
1910 spin_lock_init(&p->alloc_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912 init_sigpending(&p->pending);
1913
Martin Schwidefsky64861632011-12-15 14:56:09 +01001914 p->utime = p->stime = p->gtime = 0;
Stanislaw Gruszka40565b52016-11-15 03:06:51 +01001915#ifdef CONFIG_ARCH_HAS_SCALED_CPUTIME
Martin Schwidefsky64861632011-12-15 14:56:09 +01001916 p->utimescaled = p->stimescaled = 0;
Stanislaw Gruszka40565b52016-11-15 03:06:51 +01001917#endif
Peter Zijlstra9d7fb042015-06-30 11:30:54 +02001918 prev_cputime_init(&p->prev_cputime);
1919
Frederic Weisbecker6a616712012-12-16 20:00:34 +01001920#ifdef CONFIG_VIRT_CPU_ACCOUNTING_GEN
Frederic Weisbeckerbac5b6b2017-06-29 19:15:10 +02001921 seqcount_init(&p->vtime.seqcount);
1922 p->vtime.starttime = 0;
1923 p->vtime.state = VTIME_INACTIVE;
Frederic Weisbecker6a616712012-12-16 20:00:34 +01001924#endif
1925
KAMEZAWA Hiroyukia3a2e762010-04-06 14:34:42 -07001926#if defined(SPLIT_RSS_COUNTING)
1927 memset(&p->rss_stat, 0, sizeof(p->rss_stat));
1928#endif
Balbir Singh172ba842007-07-09 18:52:00 +02001929
Arjan van de Ven69766752008-09-01 15:52:40 -07001930 p->default_timer_slack_ns = current->timer_slack_ns;
1931
Johannes Weinereb414682018-10-26 15:06:27 -07001932#ifdef CONFIG_PSI
1933 p->psi_flags = 0;
1934#endif
1935
Andrea Righi59954772008-07-27 17:29:15 +02001936 task_io_accounting_init(&p->ioac);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937 acct_clear_integrals(p);
1938
Frank Mayharf06febc2008-09-12 09:54:39 -07001939 posix_cpu_timers_init(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 p->io_context = NULL;
Richard Guy Briggsc0b0ae82018-05-12 21:58:21 -04001942 audit_set_context(p, NULL);
Paul Menageb4f48b62007-10-18 23:39:33 -07001943 cgroup_fork(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944#ifdef CONFIG_NUMA
Lee Schermerhorn846a16b2008-04-28 02:13:09 -07001945 p->mempolicy = mpol_dup(p->mempolicy);
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001946 if (IS_ERR(p->mempolicy)) {
1947 retval = PTR_ERR(p->mempolicy);
1948 p->mempolicy = NULL;
Li Zefane8604cb2014-03-28 15:18:27 +08001949 goto bad_fork_cleanup_threadgroup_lock;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001950 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951#endif
Michal Hocko778d3b02011-07-26 16:08:30 -07001952#ifdef CONFIG_CPUSETS
1953 p->cpuset_mem_spread_rotor = NUMA_NO_NODE;
1954 p->cpuset_slab_spread_rotor = NUMA_NO_NODE;
Mel Gormancc9a6c82012-03-21 16:34:11 -07001955 seqcount_init(&p->mems_allowed_seq);
Michal Hocko778d3b02011-07-26 16:08:30 -07001956#endif
Ingo Molnarde30a2b2006-07-03 00:24:42 -07001957#ifdef CONFIG_TRACE_IRQFLAGS
1958 p->irq_events = 0;
1959 p->hardirqs_enabled = 0;
1960 p->hardirq_enable_ip = 0;
1961 p->hardirq_enable_event = 0;
1962 p->hardirq_disable_ip = _THIS_IP_;
1963 p->hardirq_disable_event = 0;
1964 p->softirqs_enabled = 1;
1965 p->softirq_enable_ip = _THIS_IP_;
1966 p->softirq_enable_event = 0;
1967 p->softirq_disable_ip = 0;
1968 p->softirq_disable_event = 0;
1969 p->hardirq_context = 0;
1970 p->softirq_context = 0;
1971#endif
David Hildenbrand8bcbde52015-05-11 17:52:06 +02001972
1973 p->pagefault_disabled = 0;
1974
Ingo Molnarfbb9ce952006-07-03 00:24:50 -07001975#ifdef CONFIG_LOCKDEP
Byungchul Parkb09be672017-08-07 16:12:52 +09001976 lockdep_init_task(p);
Ingo Molnarfbb9ce952006-07-03 00:24:50 -07001977#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978
Ingo Molnar408894e2006-01-09 15:59:20 -08001979#ifdef CONFIG_DEBUG_MUTEXES
1980 p->blocked_on = NULL; /* not blocked yet */
1981#endif
Kent Overstreetcafe5632013-03-23 16:11:31 -07001982#ifdef CONFIG_BCACHE
1983 p->sequential_io = 0;
1984 p->sequential_io_avg = 0;
1985#endif
Markus Metzger0f481402009-04-03 16:43:48 +02001986
Srivatsa Vaddagiri3c90e6e2007-11-09 22:39:39 +01001987 /* Perform scheduler related setup. Assign this task to a CPU. */
Dario Faggioliaab03e02013-11-28 11:14:43 +01001988 retval = sched_fork(clone_flags, p);
1989 if (retval)
1990 goto bad_fork_cleanup_policy;
Peter Zijlstra6ab423e2009-05-25 14:45:27 +02001991
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001992 retval = perf_event_init_task(p);
Peter Zijlstra6ab423e2009-05-25 14:45:27 +02001993 if (retval)
1994 goto bad_fork_cleanup_policy;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07001995 retval = audit_alloc(p);
1996 if (retval)
Peter Zijlstra6c72e3502014-10-02 16:17:02 -07001997 goto bad_fork_cleanup_perf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998 /* copy all the process information */
Jack Millerab602f72014-08-08 14:23:19 -07001999 shm_init_task(p);
Tetsuo Handae4e55b42017-03-24 20:46:33 +09002000 retval = security_task_alloc(p, clone_flags);
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002001 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 goto bad_fork_cleanup_audit;
Tetsuo Handae4e55b42017-03-24 20:46:33 +09002003 retval = copy_semundo(clone_flags, p);
2004 if (retval)
2005 goto bad_fork_cleanup_security;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002006 retval = copy_files(clone_flags, p);
2007 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002008 goto bad_fork_cleanup_semundo;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002009 retval = copy_fs(clone_flags, p);
2010 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 goto bad_fork_cleanup_files;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002012 retval = copy_sighand(clone_flags, p);
2013 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014 goto bad_fork_cleanup_fs;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002015 retval = copy_signal(clone_flags, p);
2016 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017 goto bad_fork_cleanup_sighand;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002018 retval = copy_mm(clone_flags, p);
2019 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002020 goto bad_fork_cleanup_signal;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002021 retval = copy_namespaces(clone_flags, p);
2022 if (retval)
David Howellsd84f4f92008-11-14 10:39:23 +11002023 goto bad_fork_cleanup_mm;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002024 retval = copy_io(clone_flags, p);
2025 if (retval)
Jens Axboefd0928d2008-01-24 08:52:45 +01002026 goto bad_fork_cleanup_namespaces;
Christian Brauner7f192e32019-05-25 11:36:41 +02002027 retval = copy_thread_tls(clone_flags, args->stack, args->stack_size, p,
2028 args->tls);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029 if (retval)
Jens Axboefd0928d2008-01-24 08:52:45 +01002030 goto bad_fork_cleanup_io;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031
Alexander Popovafaef012018-08-17 01:16:58 +03002032 stackleak_task_init(p);
2033
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07002034 if (pid != &init_struct_pid) {
Andy Lutomirskic2b1df22013-08-22 11:39:16 -07002035 pid = alloc_pid(p->nsproxy->pid_ns_for_children);
Michal Hocko35f71bc2015-04-16 12:47:38 -07002036 if (IS_ERR(pid)) {
2037 retval = PTR_ERR(pid);
Jiri Slaby0740aa52016-05-20 17:00:25 -07002038 goto bad_fork_cleanup_thread;
Michal Hocko35f71bc2015-04-16 12:47:38 -07002039 }
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07002040 }
2041
Christian Braunerb3e583822019-03-27 13:04:15 +01002042 /*
2043 * This has to happen after we've potentially unshared the file
2044 * descriptor table (so that the pidfd doesn't leak into the child
2045 * if the fd table isn't shared).
2046 */
2047 if (clone_flags & CLONE_PIDFD) {
Al Viro6fd2fe42019-06-26 22:22:09 -04002048 retval = get_unused_fd_flags(O_RDWR | O_CLOEXEC);
Christian Braunerb3e583822019-03-27 13:04:15 +01002049 if (retval < 0)
2050 goto bad_fork_free_pid;
2051
2052 pidfd = retval;
Al Viro6fd2fe42019-06-26 22:22:09 -04002053
2054 pidfile = anon_inode_getfile("[pidfd]", &pidfd_fops, pid,
2055 O_RDWR | O_CLOEXEC);
2056 if (IS_ERR(pidfile)) {
2057 put_unused_fd(pidfd);
Christian Brauner28dd29c2019-07-01 16:01:46 +02002058 retval = PTR_ERR(pidfile);
Al Viro6fd2fe42019-06-26 22:22:09 -04002059 goto bad_fork_free_pid;
2060 }
2061 get_pid(pid); /* held by pidfile now */
2062
Christian Brauner7f192e32019-05-25 11:36:41 +02002063 retval = put_user(pidfd, args->pidfd);
Christian Braunerb3e583822019-03-27 13:04:15 +01002064 if (retval)
2065 goto bad_fork_put_pidfd;
2066 }
2067
Jens Axboe73c10102011-03-08 13:19:51 +01002068#ifdef CONFIG_BLOCK
2069 p->plug = NULL;
2070#endif
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07002071#ifdef CONFIG_FUTEX
Ingo Molnar8f17d3a2006-03-27 01:16:27 -08002072 p->robust_list = NULL;
2073#ifdef CONFIG_COMPAT
2074 p->compat_robust_list = NULL;
2075#endif
Ingo Molnarc87e2832006-06-27 02:54:58 -07002076 INIT_LIST_HEAD(&p->pi_state_list);
2077 p->pi_state_cache = NULL;
Alexey Dobriyan42b2dd02007-10-16 23:27:30 -07002078#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079 /*
GOTO Masanorif9a38792006-03-13 21:20:44 -08002080 * sigaltstack should be cleared when sharing the same VM
2081 */
2082 if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
Stas Sergeev2a742132016-04-14 23:20:04 +03002083 sas_ss_reset(p);
GOTO Masanorif9a38792006-03-13 21:20:44 -08002084
2085 /*
Oleg Nesterov65808072009-12-15 16:47:16 -08002086 * Syscall tracing and stepping should be turned off in the
2087 * child regardless of CLONE_PTRACE.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088 */
Oleg Nesterov65808072009-12-15 16:47:16 -08002089 user_disable_single_step(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090 clear_tsk_thread_flag(p, TIF_SYSCALL_TRACE);
Laurent Viviered75e8d2005-09-03 15:57:18 -07002091#ifdef TIF_SYSCALL_EMU
2092 clear_tsk_thread_flag(p, TIF_SYSCALL_EMU);
2093#endif
Lin Fenge02c9b02019-05-14 15:42:34 -07002094 clear_tsk_latency_tracing(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 /* ok, now we should be set up.. */
Oleg Nesterov18c830d2013-07-03 15:08:32 -07002097 p->pid = pid_nr(pid);
2098 if (clone_flags & CLONE_THREAD) {
Oleg Nesterov5f8aadd2012-03-14 19:55:38 +01002099 p->exit_signal = -1;
Oleg Nesterov18c830d2013-07-03 15:08:32 -07002100 p->group_leader = current->group_leader;
2101 p->tgid = current->tgid;
2102 } else {
2103 if (clone_flags & CLONE_PARENT)
2104 p->exit_signal = current->group_leader->exit_signal;
2105 else
Christian Brauner7f192e32019-05-25 11:36:41 +02002106 p->exit_signal = args->exit_signal;
Oleg Nesterov18c830d2013-07-03 15:08:32 -07002107 p->group_leader = p;
2108 p->tgid = p->pid;
2109 }
Oleg Nesterov5f8aadd2012-03-14 19:55:38 +01002110
Wu Fengguang9d823e82011-06-11 18:10:12 -06002111 p->nr_dirtied = 0;
2112 p->nr_dirtied_pause = 128 >> (PAGE_SHIFT - 10);
Wu Fengguang83712352011-06-11 19:25:42 -06002113 p->dirty_paused_when = 0;
Wu Fengguang9d823e82011-06-11 18:10:12 -06002114
Oleg Nesterovbb8cbbf2013-11-13 15:36:12 +01002115 p->pdeath_signal = 0;
Oleg Nesterov47e65322006-03-28 16:11:25 -08002116 INIT_LIST_HEAD(&p->thread_group);
Al Viro158e1642012-06-27 09:24:13 +04002117 p->task_works = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118
Ingo Molnar780de9d2017-02-02 11:50:56 +01002119 cgroup_threadgroup_change_begin(current);
Oleg Nesterov18c830d2013-07-03 15:08:32 -07002120 /*
Aleksa Sarai7e476822015-06-09 21:32:09 +10002121 * Ensure that the cgroup subsystem policies allow the new process to be
2122 * forked. It should be noted the the new process's css_set can be changed
2123 * between here and cgroup_post_fork() if an organisation operation is in
2124 * progress.
2125 */
Oleg Nesterovb53202e2015-12-03 10:24:08 -05002126 retval = cgroup_can_fork(p);
Aleksa Sarai7e476822015-06-09 21:32:09 +10002127 if (retval)
Christian Braunerc3b71122019-05-10 11:53:46 +02002128 goto bad_fork_cgroup_threadgroup_change_end;
Aleksa Sarai7e476822015-06-09 21:32:09 +10002129
2130 /*
David Herrmann7b558512019-01-08 13:58:52 +01002131 * From this point on we must avoid any synchronous user-space
2132 * communication until we take the tasklist-lock. In particular, we do
2133 * not want user-space to be able to predict the process start-time by
2134 * stalling fork(2) after we recorded the start_time but before it is
2135 * visible to the system.
2136 */
2137
2138 p->start_time = ktime_get_ns();
Jason A. Donenfeld9285ec42019-06-21 22:32:48 +02002139 p->real_start_time = ktime_get_boottime_ns();
David Herrmann7b558512019-01-08 13:58:52 +01002140
2141 /*
Oleg Nesterov18c830d2013-07-03 15:08:32 -07002142 * Make it visible to the rest of the system, but dont wake it up yet.
2143 * Need tasklist lock for parent etc handling!
2144 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145 write_lock_irq(&tasklist_lock);
2146
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147 /* CLONE_PARENT re-uses the old parent */
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01002148 if (clone_flags & (CLONE_PARENT|CLONE_THREAD)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149 p->real_parent = current->real_parent;
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01002150 p->parent_exec_id = current->parent_exec_id;
2151 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152 p->real_parent = current;
Oleg Nesterov2d5516c2009-03-02 22:58:45 +01002153 p->parent_exec_id = current->self_exec_id;
2154 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155
Josh Poimboeufd83a7cb2017-02-13 19:42:40 -06002156 klp_copy_process(p);
2157
Oleg Nesterov3f17da62006-02-15 22:13:24 +03002158 spin_lock(&current->sighand->siglock);
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08002159
2160 /*
Kees Cookdbd952122014-06-27 15:18:48 -07002161 * Copy seccomp details explicitly here, in case they were changed
2162 * before holding sighand lock.
2163 */
2164 copy_seccomp(p);
2165
Mathieu Desnoyersd7822b12018-06-02 08:43:54 -04002166 rseq_fork(p, clone_flags);
2167
Eric W. Biederman4ca1d3e2018-07-13 15:30:33 -05002168 /* Don't start children in a dying pid namespace */
Gargi Sharmae8cfbc22017-11-17 15:30:34 -08002169 if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) {
Kirill Tkhai3fd37222017-05-12 19:11:31 +03002170 retval = -ENOMEM;
2171 goto bad_fork_cancel_cgroup;
2172 }
Oleg Nesterov4a2c7a72006-03-28 16:11:26 -08002173
Eric W. Biederman7673bf52018-07-23 08:01:10 -05002174 /* Let kill terminate clone/fork in the middle */
2175 if (fatal_signal_pending(current)) {
2176 retval = -EINTR;
2177 goto bad_fork_cancel_cgroup;
2178 }
2179
Al Viro6fd2fe42019-06-26 22:22:09 -04002180 /* past the last point of failure */
2181 if (pidfile)
2182 fd_install(pidfd, pidfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183
Eric W. Biederman2c470472017-09-26 13:06:43 -05002184 init_task_pid_links(p);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08002185 if (likely(p->pid)) {
Tejun Heo4b9d33e2011-06-17 16:50:38 +02002186 ptrace_init_task(p, (clone_flags & CLONE_PTRACE) || trace);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002187
Oleg Nesterov81907732013-07-03 15:08:31 -07002188 init_task_pid(p, PIDTYPE_PID, pid);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08002189 if (thread_group_leader(p)) {
Eric W. Biederman6883f812017-06-04 04:32:13 -05002190 init_task_pid(p, PIDTYPE_TGID, pid);
Oleg Nesterov81907732013-07-03 15:08:31 -07002191 init_task_pid(p, PIDTYPE_PGID, task_pgrp(current));
2192 init_task_pid(p, PIDTYPE_SID, task_session(current));
2193
Eric W. Biederman1c4042c2010-07-12 17:10:36 -07002194 if (is_child_reaper(pid)) {
Eric W. Biederman17cf22c2010-03-02 14:51:53 -08002195 ns_of_pid(pid)->child_reaper = p;
Eric W. Biederman1c4042c2010-07-12 17:10:36 -07002196 p->signal->flags |= SIGNAL_UNKILLABLE;
2197 }
Eric W. Biedermanc3ad2c32018-07-23 15:20:37 -05002198 p->signal->shared_pending.signal = delayed.signal;
Alan Cox9c9f4de2008-10-13 10:37:26 +01002199 p->signal->tty = tty_kref_get(current->signal->tty);
Pavel Tikhomirov749860c2017-01-30 18:06:12 +03002200 /*
2201 * Inherit has_child_subreaper flag under the same
2202 * tasklist_lock with adding child to the process tree
2203 * for propagate_has_child_subreaper optimization.
2204 */
2205 p->signal->has_child_subreaper = p->real_parent->signal->has_child_subreaper ||
2206 p->real_parent->signal->is_child_subreaper;
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08002207 list_add_tail(&p->sibling, &p->real_parent->children);
Eric W. Biederman5e85d4a2006-04-18 22:20:16 -07002208 list_add_tail_rcu(&p->tasks, &init_task.tasks);
Eric W. Biederman6883f812017-06-04 04:32:13 -05002209 attach_pid(p, PIDTYPE_TGID);
Oleg Nesterov81907732013-07-03 15:08:31 -07002210 attach_pid(p, PIDTYPE_PGID);
2211 attach_pid(p, PIDTYPE_SID);
Christoph Lameter909ea962010-12-08 16:22:55 +01002212 __this_cpu_inc(process_counts);
Oleg Nesterov80628ca2013-07-03 15:08:30 -07002213 } else {
2214 current->signal->nr_threads++;
2215 atomic_inc(&current->signal->live);
Elena Reshetova60d4de32019-01-18 14:27:27 +02002216 refcount_inc(&current->signal->sigcnt);
Eric W. Biederman924de3b2018-07-23 13:38:00 -05002217 task_join_group_stop(p);
Oleg Nesterov80628ca2013-07-03 15:08:30 -07002218 list_add_tail_rcu(&p->thread_group,
2219 &p->group_leader->thread_group);
Oleg Nesterov0c740d02014-01-21 15:49:56 -08002220 list_add_tail_rcu(&p->thread_node,
2221 &p->signal->thread_head);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08002222 }
Oleg Nesterov81907732013-07-03 15:08:31 -07002223 attach_pid(p, PIDTYPE_PID);
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08002224 nr_threads++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002225 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002226 total_forks++;
Eric W. Biedermanc3ad2c32018-07-23 15:20:37 -05002227 hlist_del_init(&delayed.node);
Oleg Nesterov3f17da62006-02-15 22:13:24 +03002228 spin_unlock(&current->sighand->siglock);
Oleg Nesterov4af42062014-04-13 20:58:54 +02002229 syscall_tracepoint_update(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230 write_unlock_irq(&tasklist_lock);
Oleg Nesterov4af42062014-04-13 20:58:54 +02002231
Andrew Mortonc13cf852005-11-28 13:43:48 -08002232 proc_fork_connector(p);
Oleg Nesterovb53202e2015-12-03 10:24:08 -05002233 cgroup_post_fork(p);
Ingo Molnar780de9d2017-02-02 11:50:56 +01002234 cgroup_threadgroup_change_end(current);
Ingo Molnarcdd6c482009-09-21 12:02:48 +02002235 perf_event_fork(p);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08002236
2237 trace_task_newtask(p, clone_flags);
Oleg Nesterov3ab67962013-10-16 19:39:37 +02002238 uprobe_copy_process(p, clone_flags);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08002239
Linus Torvalds1da177e2005-04-16 15:20:36 -07002240 return p;
2241
Aleksa Sarai7e476822015-06-09 21:32:09 +10002242bad_fork_cancel_cgroup:
Kirill Tkhai3fd37222017-05-12 19:11:31 +03002243 spin_unlock(&current->sighand->siglock);
2244 write_unlock_irq(&tasklist_lock);
Oleg Nesterovb53202e2015-12-03 10:24:08 -05002245 cgroup_cancel_fork(p);
Christian Braunerc3b71122019-05-10 11:53:46 +02002246bad_fork_cgroup_threadgroup_change_end:
2247 cgroup_threadgroup_change_end(current);
Christian Braunerb3e583822019-03-27 13:04:15 +01002248bad_fork_put_pidfd:
Al Viro6fd2fe42019-06-26 22:22:09 -04002249 if (clone_flags & CLONE_PIDFD) {
2250 fput(pidfile);
2251 put_unused_fd(pidfd);
2252 }
Pavel Emelyanov425fb2b2007-10-18 23:40:07 -07002253bad_fork_free_pid:
2254 if (pid != &init_struct_pid)
2255 free_pid(pid);
Jiri Slaby0740aa52016-05-20 17:00:25 -07002256bad_fork_cleanup_thread:
2257 exit_thread(p);
Jens Axboefd0928d2008-01-24 08:52:45 +01002258bad_fork_cleanup_io:
Louis Rillingb69f2292009-12-04 14:52:42 +01002259 if (p->io_context)
2260 exit_io_context(p);
Serge E. Hallynab516012006-10-02 02:18:06 -07002261bad_fork_cleanup_namespaces:
Linus Torvalds444f3782007-01-30 13:35:18 -08002262 exit_task_namespaces(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263bad_fork_cleanup_mm:
Andrea Arcangelic3f3ce02019-05-14 15:40:46 -07002264 if (p->mm) {
2265 mm_clear_owner(p->mm, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266 mmput(p->mm);
Andrea Arcangelic3f3ce02019-05-14 15:40:46 -07002267 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268bad_fork_cleanup_signal:
Oleg Nesterov4ab6c082009-08-26 14:29:24 -07002269 if (!(clone_flags & CLONE_THREAD))
Mike Galbraith1c5354d2011-01-05 11:16:04 +01002270 free_signal_struct(p->signal);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271bad_fork_cleanup_sighand:
Oleg Nesterova7e53282006-03-28 16:11:27 -08002272 __cleanup_sighand(p->sighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002273bad_fork_cleanup_fs:
2274 exit_fs(p); /* blocking */
2275bad_fork_cleanup_files:
2276 exit_files(p); /* blocking */
2277bad_fork_cleanup_semundo:
2278 exit_sem(p);
Tetsuo Handae4e55b42017-03-24 20:46:33 +09002279bad_fork_cleanup_security:
2280 security_task_free(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281bad_fork_cleanup_audit:
2282 audit_free(p);
Peter Zijlstra6c72e3502014-10-02 16:17:02 -07002283bad_fork_cleanup_perf:
Ingo Molnarcdd6c482009-09-21 12:02:48 +02002284 perf_event_free_task(p);
Peter Zijlstra6c72e3502014-10-02 16:17:02 -07002285bad_fork_cleanup_policy:
Byungchul Parkb09be672017-08-07 16:12:52 +09002286 lockdep_free_task(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002287#ifdef CONFIG_NUMA
Lee Schermerhornf0be3d32008-04-28 02:13:08 -07002288 mpol_put(p->mempolicy);
Li Zefane8604cb2014-03-28 15:18:27 +08002289bad_fork_cleanup_threadgroup_lock:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290#endif
Shailabh Nagar35df17c2006-08-31 21:27:38 -07002291 delayacct_tsk_free(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002292bad_fork_cleanup_count:
David Howellsd84f4f92008-11-14 10:39:23 +11002293 atomic_dec(&p->cred->user->processes);
David Howellse0e81732009-09-02 09:13:40 +01002294 exit_creds(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002295bad_fork_free:
Andy Lutomirski405c0752016-10-31 08:11:43 -07002296 p->state = TASK_DEAD;
Andy Lutomirski68f24b082016-09-15 22:45:48 -07002297 put_task_stack(p);
Andrea Arcangelic3f3ce02019-05-14 15:40:46 -07002298 delayed_free_task(p);
Oleg Nesterovfe7d37d2006-01-08 01:04:02 -08002299fork_out:
Eric W. Biedermanc3ad2c32018-07-23 15:20:37 -05002300 spin_lock_irq(&current->sighand->siglock);
2301 hlist_del_init(&delayed.node);
2302 spin_unlock_irq(&current->sighand->siglock);
Oleg Nesterovfe7d37d2006-01-08 01:04:02 -08002303 return ERR_PTR(retval);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002304}
2305
Eric W. Biederman2c470472017-09-26 13:06:43 -05002306static inline void init_idle_pids(struct task_struct *idle)
Oleg Nesterovf106eee2010-05-26 14:44:11 -07002307{
2308 enum pid_type type;
2309
2310 for (type = PIDTYPE_PID; type < PIDTYPE_MAX; ++type) {
Eric W. Biederman2c470472017-09-26 13:06:43 -05002311 INIT_HLIST_NODE(&idle->pid_links[type]); /* not really needed */
2312 init_task_pid(idle, type, &init_struct_pid);
Oleg Nesterovf106eee2010-05-26 14:44:11 -07002313 }
2314}
2315
Paul Gortmaker0db06282013-06-19 14:53:51 -04002316struct task_struct *fork_idle(int cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002317{
Ingo Molnar36c8b582006-07-03 00:25:41 -07002318 struct task_struct *task;
Christian Brauner7f192e32019-05-25 11:36:41 +02002319 struct kernel_clone_args args = {
2320 .flags = CLONE_VM,
2321 };
2322
2323 task = copy_process(&init_struct_pid, 0, cpu_to_node(cpu), &args);
Oleg Nesterovf106eee2010-05-26 14:44:11 -07002324 if (!IS_ERR(task)) {
Eric W. Biederman2c470472017-09-26 13:06:43 -05002325 init_idle_pids(task);
Akinobu Mita753ca4f2006-11-25 11:09:34 -08002326 init_idle(task, cpu);
Oleg Nesterovf106eee2010-05-26 14:44:11 -07002327 }
Oleg Nesterov73b9ebf2006-03-28 16:11:07 -08002328
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329 return task;
2330}
2331
Nadav Amit13585fa2019-04-25 17:11:25 -07002332struct mm_struct *copy_init_mm(void)
2333{
2334 return dup_mm(NULL, &init_mm);
2335}
2336
Linus Torvalds1da177e2005-04-16 15:20:36 -07002337/*
2338 * Ok, this is the main fork-routine.
2339 *
2340 * It copies the process, and if successful kick-starts
2341 * it and waits for it to finish using the VM if required.
2342 */
Christian Brauner7f192e32019-05-25 11:36:41 +02002343long _do_fork(struct kernel_clone_args *args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344{
Christian Brauner7f192e32019-05-25 11:36:41 +02002345 u64 clone_flags = args->flags;
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002346 struct completion vfork;
2347 struct pid *pid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348 struct task_struct *p;
2349 int trace = 0;
Eric W. Biederman92476d72006-03-31 02:31:42 -08002350 long nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351
Andrew Mortonbdff7462008-02-04 22:27:22 -08002352 /*
Tejun Heo4b9d33e2011-06-17 16:50:38 +02002353 * Determine whether and which event to report to ptracer. When
2354 * called from kernel_thread or CLONE_UNTRACED is explicitly
2355 * requested, no event is reported; otherwise, report if the event
2356 * for the type of forking is enabled.
Roland McGrath09a05392008-07-25 19:45:47 -07002357 */
Al Viroe80d6662012-10-22 23:10:08 -04002358 if (!(clone_flags & CLONE_UNTRACED)) {
Tejun Heo4b9d33e2011-06-17 16:50:38 +02002359 if (clone_flags & CLONE_VFORK)
2360 trace = PTRACE_EVENT_VFORK;
Christian Brauner7f192e32019-05-25 11:36:41 +02002361 else if (args->exit_signal != SIGCHLD)
Tejun Heo4b9d33e2011-06-17 16:50:38 +02002362 trace = PTRACE_EVENT_CLONE;
2363 else
2364 trace = PTRACE_EVENT_FORK;
2365
2366 if (likely(!ptrace_event_enabled(current, trace)))
2367 trace = 0;
2368 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002369
Christian Brauner7f192e32019-05-25 11:36:41 +02002370 p = copy_process(NULL, trace, NUMA_NO_NODE, args);
Emese Revfy38addce2016-06-20 20:41:19 +02002371 add_latent_entropy();
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002372
2373 if (IS_ERR(p))
2374 return PTR_ERR(p);
2375
Linus Torvalds1da177e2005-04-16 15:20:36 -07002376 /*
2377 * Do this prior waking up the new thread - the thread pointer
2378 * might get invalid after that point, if the thread exits quickly.
2379 */
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002380 trace_sched_process_fork(current, p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002382 pid = get_task_pid(p, PIDTYPE_PID);
2383 nr = pid_vnr(pid);
Mathieu Desnoyers0a16b602008-07-18 12:16:17 -04002384
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002385 if (clone_flags & CLONE_PARENT_SETTID)
Christian Brauner7f192e32019-05-25 11:36:41 +02002386 put_user(nr, args->parent_tid);
Pavel Emelyanov30e49c22007-10-18 23:40:10 -07002387
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002388 if (clone_flags & CLONE_VFORK) {
2389 p->vfork_done = &vfork;
2390 init_completion(&vfork);
2391 get_task_struct(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002392 }
Marcos Paulo de Souza9f5325a2018-02-06 15:39:30 -08002393
2394 wake_up_new_task(p);
2395
2396 /* forking complete and child started to run, tell ptracer */
2397 if (unlikely(trace))
2398 ptrace_event_pid(trace, pid);
2399
2400 if (clone_flags & CLONE_VFORK) {
2401 if (!wait_for_vfork_done(p, &vfork))
2402 ptrace_event_pid(PTRACE_EVENT_VFORK_DONE, pid);
2403 }
2404
2405 put_pid(pid);
Eric W. Biederman92476d72006-03-31 02:31:42 -08002406 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002407}
2408
Josh Triplett3033f14a2015-06-25 15:01:19 -07002409#ifndef CONFIG_HAVE_COPY_THREAD_TLS
2410/* For compatibility with architectures that call do_fork directly rather than
2411 * using the syscall entry points below. */
2412long do_fork(unsigned long clone_flags,
2413 unsigned long stack_start,
2414 unsigned long stack_size,
2415 int __user *parent_tidptr,
2416 int __user *child_tidptr)
2417{
Christian Brauner7f192e32019-05-25 11:36:41 +02002418 struct kernel_clone_args args = {
2419 .flags = (clone_flags & ~CSIGNAL),
2420 .child_tid = child_tidptr,
2421 .parent_tid = parent_tidptr,
2422 .exit_signal = (clone_flags & CSIGNAL),
2423 .stack = stack_start,
2424 .stack_size = stack_size,
2425 };
2426
2427 return _do_fork(&args);
Josh Triplett3033f14a2015-06-25 15:01:19 -07002428}
2429#endif
2430
Al Viro2aa3a7f2012-09-21 19:55:31 -04002431/*
2432 * Create a kernel thread.
2433 */
2434pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
2435{
Christian Brauner7f192e32019-05-25 11:36:41 +02002436 struct kernel_clone_args args = {
2437 .flags = ((flags | CLONE_VM | CLONE_UNTRACED) & ~CSIGNAL),
2438 .exit_signal = (flags & CSIGNAL),
2439 .stack = (unsigned long)fn,
2440 .stack_size = (unsigned long)arg,
2441 };
2442
2443 return _do_fork(&args);
Al Viro2aa3a7f2012-09-21 19:55:31 -04002444}
Al Viro2aa3a7f2012-09-21 19:55:31 -04002445
Al Virod2125042012-10-23 13:17:59 -04002446#ifdef __ARCH_WANT_SYS_FORK
2447SYSCALL_DEFINE0(fork)
2448{
2449#ifdef CONFIG_MMU
Christian Brauner7f192e32019-05-25 11:36:41 +02002450 struct kernel_clone_args args = {
2451 .exit_signal = SIGCHLD,
2452 };
2453
2454 return _do_fork(&args);
Al Virod2125042012-10-23 13:17:59 -04002455#else
2456 /* can not support in nommu mode */
Daeseok Youn5d59e182014-01-23 15:55:47 -08002457 return -EINVAL;
Al Virod2125042012-10-23 13:17:59 -04002458#endif
2459}
2460#endif
2461
2462#ifdef __ARCH_WANT_SYS_VFORK
2463SYSCALL_DEFINE0(vfork)
2464{
Christian Brauner7f192e32019-05-25 11:36:41 +02002465 struct kernel_clone_args args = {
2466 .flags = CLONE_VFORK | CLONE_VM,
2467 .exit_signal = SIGCHLD,
2468 };
2469
2470 return _do_fork(&args);
Al Virod2125042012-10-23 13:17:59 -04002471}
2472#endif
2473
2474#ifdef __ARCH_WANT_SYS_CLONE
2475#ifdef CONFIG_CLONE_BACKWARDS
2476SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
2477 int __user *, parent_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07002478 unsigned long, tls,
Al Virod2125042012-10-23 13:17:59 -04002479 int __user *, child_tidptr)
2480#elif defined(CONFIG_CLONE_BACKWARDS2)
2481SYSCALL_DEFINE5(clone, unsigned long, newsp, unsigned long, clone_flags,
2482 int __user *, parent_tidptr,
2483 int __user *, child_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07002484 unsigned long, tls)
Michal Simekdfa97712013-08-13 16:00:53 -07002485#elif defined(CONFIG_CLONE_BACKWARDS3)
2486SYSCALL_DEFINE6(clone, unsigned long, clone_flags, unsigned long, newsp,
2487 int, stack_size,
2488 int __user *, parent_tidptr,
2489 int __user *, child_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07002490 unsigned long, tls)
Al Virod2125042012-10-23 13:17:59 -04002491#else
2492SYSCALL_DEFINE5(clone, unsigned long, clone_flags, unsigned long, newsp,
2493 int __user *, parent_tidptr,
2494 int __user *, child_tidptr,
Josh Triplett3033f14a2015-06-25 15:01:19 -07002495 unsigned long, tls)
Al Virod2125042012-10-23 13:17:59 -04002496#endif
2497{
Christian Brauner7f192e32019-05-25 11:36:41 +02002498 struct kernel_clone_args args = {
2499 .flags = (clone_flags & ~CSIGNAL),
2500 .pidfd = parent_tidptr,
2501 .child_tid = child_tidptr,
2502 .parent_tid = parent_tidptr,
2503 .exit_signal = (clone_flags & CSIGNAL),
2504 .stack = newsp,
2505 .tls = tls,
2506 };
2507
2508 /* clone(CLONE_PIDFD) uses parent_tidptr to return a pidfd */
2509 if ((clone_flags & CLONE_PIDFD) && (clone_flags & CLONE_PARENT_SETTID))
2510 return -EINVAL;
2511
2512 return _do_fork(&args);
2513}
Christian Braunerd68dbb02019-06-21 01:26:35 +02002514#endif
Christian Brauner7f192e32019-05-25 11:36:41 +02002515
Christian Braunerd68dbb02019-06-21 01:26:35 +02002516#ifdef __ARCH_WANT_SYS_CLONE3
Christian Brauner7f192e32019-05-25 11:36:41 +02002517noinline static int copy_clone_args_from_user(struct kernel_clone_args *kargs,
2518 struct clone_args __user *uargs,
2519 size_t size)
2520{
2521 struct clone_args args;
2522
2523 if (unlikely(size > PAGE_SIZE))
2524 return -E2BIG;
2525
2526 if (unlikely(size < sizeof(struct clone_args)))
2527 return -EINVAL;
2528
2529 if (unlikely(!access_ok(uargs, size)))
2530 return -EFAULT;
2531
2532 if (size > sizeof(struct clone_args)) {
2533 unsigned char __user *addr;
2534 unsigned char __user *end;
2535 unsigned char val;
2536
2537 addr = (void __user *)uargs + sizeof(struct clone_args);
2538 end = (void __user *)uargs + size;
2539
2540 for (; addr < end; addr++) {
2541 if (get_user(val, addr))
2542 return -EFAULT;
2543 if (val)
2544 return -E2BIG;
2545 }
2546
2547 size = sizeof(struct clone_args);
2548 }
2549
2550 if (copy_from_user(&args, uargs, size))
2551 return -EFAULT;
2552
2553 *kargs = (struct kernel_clone_args){
2554 .flags = args.flags,
2555 .pidfd = u64_to_user_ptr(args.pidfd),
2556 .child_tid = u64_to_user_ptr(args.child_tid),
2557 .parent_tid = u64_to_user_ptr(args.parent_tid),
2558 .exit_signal = args.exit_signal,
2559 .stack = args.stack,
2560 .stack_size = args.stack_size,
2561 .tls = args.tls,
2562 };
2563
2564 return 0;
2565}
2566
2567static bool clone3_args_valid(const struct kernel_clone_args *kargs)
2568{
2569 /*
2570 * All lower bits of the flag word are taken.
2571 * Verify that no other unknown flags are passed along.
2572 */
2573 if (kargs->flags & ~CLONE_LEGACY_FLAGS)
2574 return false;
2575
2576 /*
2577 * - make the CLONE_DETACHED bit reuseable for clone3
2578 * - make the CSIGNAL bits reuseable for clone3
2579 */
2580 if (kargs->flags & (CLONE_DETACHED | CSIGNAL))
2581 return false;
2582
2583 if ((kargs->flags & (CLONE_THREAD | CLONE_PARENT)) &&
2584 kargs->exit_signal)
2585 return false;
2586
2587 return true;
2588}
2589
2590SYSCALL_DEFINE2(clone3, struct clone_args __user *, uargs, size_t, size)
2591{
2592 int err;
2593
2594 struct kernel_clone_args kargs;
2595
2596 err = copy_clone_args_from_user(&kargs, uargs, size);
2597 if (err)
2598 return err;
2599
2600 if (!clone3_args_valid(&kargs))
2601 return -EINVAL;
2602
2603 return _do_fork(&kargs);
Al Virod2125042012-10-23 13:17:59 -04002604}
2605#endif
2606
Oleg Nesterov0f1b92c2017-01-30 18:06:11 +03002607void walk_process_tree(struct task_struct *top, proc_visitor visitor, void *data)
2608{
2609 struct task_struct *leader, *parent, *child;
2610 int res;
2611
2612 read_lock(&tasklist_lock);
2613 leader = top = top->group_leader;
2614down:
2615 for_each_thread(leader, parent) {
2616 list_for_each_entry(child, &parent->children, sibling) {
2617 res = visitor(child, data);
2618 if (res) {
2619 if (res < 0)
2620 goto out;
2621 leader = child;
2622 goto down;
2623 }
2624up:
2625 ;
2626 }
2627 }
2628
2629 if (leader != top) {
2630 child = leader;
2631 parent = child->real_parent;
2632 leader = parent->group_leader;
2633 goto up;
2634 }
2635out:
2636 read_unlock(&tasklist_lock);
2637}
2638
Ravikiran G Thirumalai5fd63b32006-01-11 22:46:15 +01002639#ifndef ARCH_MIN_MMSTRUCT_ALIGN
2640#define ARCH_MIN_MMSTRUCT_ALIGN 0
2641#endif
2642
Alexey Dobriyan51cc5062008-07-25 19:45:34 -07002643static void sighand_ctor(void *data)
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08002644{
2645 struct sighand_struct *sighand = data;
2646
Christoph Lametera35afb82007-05-16 22:10:57 -07002647 spin_lock_init(&sighand->siglock);
Davide Libenzib8fceee2007-09-20 12:40:16 -07002648 init_waitqueue_head(&sighand->signalfd_wqh);
Oleg Nesterovaa1757f2006-03-28 16:11:12 -08002649}
2650
Linus Torvalds1da177e2005-04-16 15:20:36 -07002651void __init proc_caches_init(void)
2652{
Rik van Rielc1a2f7f2018-07-16 15:03:31 -04002653 unsigned int mm_size;
2654
Linus Torvalds1da177e2005-04-16 15:20:36 -07002655 sighand_cachep = kmem_cache_create("sighand_cache",
2656 sizeof(struct sighand_struct), 0,
Paul E. McKenney5f0d5a32017-01-18 02:53:44 -08002657 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_TYPESAFE_BY_RCU|
Levin, Alexander (Sasha Levin)75f296d2017-11-15 17:35:54 -08002658 SLAB_ACCOUNT, sighand_ctor);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002659 signal_cachep = kmem_cache_create("signal_cache",
2660 sizeof(struct signal_struct), 0,
Levin, Alexander (Sasha Levin)75f296d2017-11-15 17:35:54 -08002661 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
Vladimir Davydov5d097052016-01-14 15:18:21 -08002662 NULL);
Paul Mundt20c2df82007-07-20 10:11:58 +09002663 files_cachep = kmem_cache_create("files_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002664 sizeof(struct files_struct), 0,
Levin, Alexander (Sasha Levin)75f296d2017-11-15 17:35:54 -08002665 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
Vladimir Davydov5d097052016-01-14 15:18:21 -08002666 NULL);
Paul Mundt20c2df82007-07-20 10:11:58 +09002667 fs_cachep = kmem_cache_create("fs_cache",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002668 sizeof(struct fs_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);
Rik van Rielc1a2f7f2018-07-16 15:03:31 -04002671
Linus Torvalds6345d242011-05-29 11:32:28 -07002672 /*
Rik van Rielc1a2f7f2018-07-16 15:03:31 -04002673 * The mm_cpumask is located at the end of mm_struct, and is
2674 * dynamically sized based on the maximum CPU number this system
2675 * can have, taking hotplug into account (nr_cpu_ids).
Linus Torvalds6345d242011-05-29 11:32:28 -07002676 */
Rik van Rielc1a2f7f2018-07-16 15:03:31 -04002677 mm_size = sizeof(struct mm_struct) + cpumask_size();
2678
David Windsor07dcd7f2017-08-15 16:45:00 -07002679 mm_cachep = kmem_cache_create_usercopy("mm_struct",
Rik van Rielc1a2f7f2018-07-16 15:03:31 -04002680 mm_size, ARCH_MIN_MMSTRUCT_ALIGN,
Levin, Alexander (Sasha Levin)75f296d2017-11-15 17:35:54 -08002681 SLAB_HWCACHE_ALIGN|SLAB_PANIC|SLAB_ACCOUNT,
David Windsor07dcd7f2017-08-15 16:45:00 -07002682 offsetof(struct mm_struct, saved_auxv),
2683 sizeof_field(struct mm_struct, saved_auxv),
Vladimir Davydov5d097052016-01-14 15:18:21 -08002684 NULL);
2685 vm_area_cachep = KMEM_CACHE(vm_area_struct, SLAB_PANIC|SLAB_ACCOUNT);
David Howells8feae132009-01-08 12:04:47 +00002686 mmap_init();
Al Viro66577192011-06-28 15:41:10 -04002687 nsproxy_cache_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002688}
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002689
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002690/*
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002691 * Check constraints on flags passed to the unshare system call.
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002692 */
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002693static int check_unshare_flags(unsigned long unshare_flags)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002694{
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002695 if (unshare_flags & ~(CLONE_THREAD|CLONE_FS|CLONE_NEWNS|CLONE_SIGHAND|
2696 CLONE_VM|CLONE_FILES|CLONE_SYSVSEM|
Eric W. Biederman50804fe2010-03-02 15:41:50 -08002697 CLONE_NEWUTS|CLONE_NEWIPC|CLONE_NEWNET|
Aditya Kalia79a9082016-01-29 02:54:06 -06002698 CLONE_NEWUSER|CLONE_NEWPID|CLONE_NEWCGROUP))
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002699 return -EINVAL;
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002700 /*
Eric W. Biederman12c641a2015-08-10 17:35:07 -05002701 * Not implemented, but pretend it works if there is nothing
2702 * to unshare. Note that unsharing the address space or the
2703 * signal handlers also need to unshare the signal queues (aka
2704 * CLONE_THREAD).
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002705 */
2706 if (unshare_flags & (CLONE_THREAD | CLONE_SIGHAND | CLONE_VM)) {
Eric W. Biederman12c641a2015-08-10 17:35:07 -05002707 if (!thread_group_empty(current))
2708 return -EINVAL;
2709 }
2710 if (unshare_flags & (CLONE_SIGHAND | CLONE_VM)) {
Elena Reshetovad036bda2019-01-18 14:27:26 +02002711 if (refcount_read(&current->sighand->count) > 1)
Eric W. Biederman12c641a2015-08-10 17:35:07 -05002712 return -EINVAL;
2713 }
2714 if (unshare_flags & CLONE_VM) {
2715 if (!current_is_single_threaded())
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002716 return -EINVAL;
2717 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002718
2719 return 0;
2720}
2721
2722/*
JANAK DESAI99d14192006-02-07 12:58:59 -08002723 * Unshare the filesystem structure if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002724 */
2725static int unshare_fs(unsigned long unshare_flags, struct fs_struct **new_fsp)
2726{
2727 struct fs_struct *fs = current->fs;
2728
Al Viro498052b2009-03-30 07:20:30 -04002729 if (!(unshare_flags & CLONE_FS) || !fs)
2730 return 0;
2731
2732 /* don't need lock here; in the worst case we'll do useless copy */
2733 if (fs->users == 1)
2734 return 0;
2735
2736 *new_fsp = copy_fs_struct(fs);
2737 if (!*new_fsp)
2738 return -ENOMEM;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002739
2740 return 0;
2741}
2742
2743/*
JANAK DESAIa016f332006-02-07 12:59:02 -08002744 * Unshare file descriptor table if it is being shared
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002745 */
2746static int unshare_fd(unsigned long unshare_flags, struct files_struct **new_fdp)
2747{
2748 struct files_struct *fd = current->files;
JANAK DESAIa016f332006-02-07 12:59:02 -08002749 int error = 0;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002750
2751 if ((unshare_flags & CLONE_FILES) &&
JANAK DESAIa016f332006-02-07 12:59:02 -08002752 (fd && atomic_read(&fd->count) > 1)) {
2753 *new_fdp = dup_fd(fd, &error);
2754 if (!*new_fdp)
2755 return error;
2756 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002757
2758 return 0;
2759}
2760
2761/*
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002762 * unshare allows a process to 'unshare' part of the process
2763 * context which was originally shared using clone. copy_*
2764 * functions used by do_fork() cannot be used here directly
2765 * because they modify an inactive task_struct that is being
2766 * constructed. Here we are modifying the current, active,
2767 * task_struct.
2768 */
Dominik Brodowski9b321052018-03-11 11:34:42 +01002769int ksys_unshare(unsigned long unshare_flags)
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002770{
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002771 struct fs_struct *fs, *new_fs = NULL;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002772 struct files_struct *fd, *new_fd = NULL;
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002773 struct cred *new_cred = NULL;
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07002774 struct nsproxy *new_nsproxy = NULL;
Manfred Spraul9edff4a2008-04-29 01:00:57 -07002775 int do_sysvsem = 0;
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002776 int err;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002777
Eric W. Biederman50804fe2010-03-02 15:41:50 -08002778 /*
Eric W. Biedermanfaf00da2015-08-10 18:25:44 -05002779 * If unsharing a user namespace must also unshare the thread group
2780 * and unshare the filesystem root and working directories.
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002781 */
2782 if (unshare_flags & CLONE_NEWUSER)
Eric W. Biedermane66eded2013-03-13 11:51:49 -07002783 unshare_flags |= CLONE_THREAD | CLONE_FS;
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002784 /*
Eric W. Biederman50804fe2010-03-02 15:41:50 -08002785 * If unsharing vm, must also unshare signal handlers.
2786 */
2787 if (unshare_flags & CLONE_VM)
2788 unshare_flags |= CLONE_SIGHAND;
Manfred Spraul6013f672008-04-29 01:00:59 -07002789 /*
Eric W. Biederman12c641a2015-08-10 17:35:07 -05002790 * If unsharing a signal handlers, must also unshare the signal queues.
2791 */
2792 if (unshare_flags & CLONE_SIGHAND)
2793 unshare_flags |= CLONE_THREAD;
2794 /*
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002795 * If unsharing namespace, must also unshare filesystem information.
2796 */
2797 if (unshare_flags & CLONE_NEWNS)
2798 unshare_flags |= CLONE_FS;
Eric W. Biederman50804fe2010-03-02 15:41:50 -08002799
2800 err = check_unshare_flags(unshare_flags);
2801 if (err)
2802 goto bad_unshare_out;
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002803 /*
Manfred Spraul6013f672008-04-29 01:00:59 -07002804 * CLONE_NEWIPC must also detach from the undolist: after switching
2805 * to a new ipc namespace, the semaphore arrays from the old
2806 * namespace are unreachable.
2807 */
2808 if (unshare_flags & (CLONE_NEWIPC|CLONE_SYSVSEM))
Manfred Spraul9edff4a2008-04-29 01:00:57 -07002809 do_sysvsem = 1;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002810 err = unshare_fs(unshare_flags, &new_fs);
2811 if (err)
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002812 goto bad_unshare_out;
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002813 err = unshare_fd(unshare_flags, &new_fd);
2814 if (err)
Oleg Nesterov9bfb23f2011-03-22 16:34:09 -07002815 goto bad_unshare_cleanup_fs;
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002816 err = unshare_userns(unshare_flags, &new_cred);
Daniel Rebelo de Oliveirafb0a6852011-07-26 16:08:39 -07002817 if (err)
Manfred Spraul9edff4a2008-04-29 01:00:57 -07002818 goto bad_unshare_cleanup_fd;
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002819 err = unshare_nsproxy_namespaces(unshare_flags, &new_nsproxy,
2820 new_cred, new_fs);
2821 if (err)
2822 goto bad_unshare_cleanup_cred;
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002823
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002824 if (new_fs || new_fd || do_sysvsem || new_cred || new_nsproxy) {
Manfred Spraul9edff4a2008-04-29 01:00:57 -07002825 if (do_sysvsem) {
2826 /*
2827 * CLONE_SYSVSEM is equivalent to sys_exit().
2828 */
2829 exit_sem(current);
2830 }
Jack Millerab602f72014-08-08 14:23:19 -07002831 if (unshare_flags & CLONE_NEWIPC) {
2832 /* Orphan segments in old ns (see sem above). */
2833 exit_shm(current);
2834 shm_init_task(current);
2835 }
Serge E. Hallynab516012006-10-02 02:18:06 -07002836
Alan Cox6f977e62013-02-27 17:03:23 -08002837 if (new_nsproxy)
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07002838 switch_task_namespaces(current, new_nsproxy);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002839
Pavel Emelyanovcf7b7082007-10-18 23:39:54 -07002840 task_lock(current);
2841
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002842 if (new_fs) {
2843 fs = current->fs;
Nick Piggin2a4419b2010-08-18 04:37:33 +10002844 spin_lock(&fs->lock);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002845 current->fs = new_fs;
Al Viro498052b2009-03-30 07:20:30 -04002846 if (--fs->users)
2847 new_fs = NULL;
2848 else
2849 new_fs = fs;
Nick Piggin2a4419b2010-08-18 04:37:33 +10002850 spin_unlock(&fs->lock);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002851 }
2852
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002853 if (new_fd) {
2854 fd = current->files;
2855 current->files = new_fd;
2856 new_fd = fd;
2857 }
2858
2859 task_unlock(current);
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002860
2861 if (new_cred) {
2862 /* Install the new user namespace */
2863 commit_creds(new_cred);
2864 new_cred = NULL;
2865 }
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002866 }
2867
Hari Bathinie4222672017-03-08 02:11:36 +05302868 perf_event_namespaces(current);
2869
Eric W. Biedermanb2e0d9872012-07-26 05:15:35 -07002870bad_unshare_cleanup_cred:
2871 if (new_cred)
2872 put_cred(new_cred);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002873bad_unshare_cleanup_fd:
2874 if (new_fd)
2875 put_files_struct(new_fd);
2876
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002877bad_unshare_cleanup_fs:
2878 if (new_fs)
Al Viro498052b2009-03-30 07:20:30 -04002879 free_fs_struct(new_fs);
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002880
JANAK DESAIcf2e3402006-02-07 12:58:58 -08002881bad_unshare_out:
2882 return err;
2883}
Al Viro3b125382008-04-22 05:31:30 -04002884
Dominik Brodowski9b321052018-03-11 11:34:42 +01002885SYSCALL_DEFINE1(unshare, unsigned long, unshare_flags)
2886{
2887 return ksys_unshare(unshare_flags);
2888}
2889
Al Viro3b125382008-04-22 05:31:30 -04002890/*
2891 * Helper to unshare the files of the current task.
2892 * We don't want to expose copy_files internals to
2893 * the exec layer of the kernel.
2894 */
2895
2896int unshare_files(struct files_struct **displaced)
2897{
2898 struct task_struct *task = current;
Al Viro50704512008-04-26 05:25:00 +01002899 struct files_struct *copy = NULL;
Al Viro3b125382008-04-22 05:31:30 -04002900 int error;
2901
2902 error = unshare_fd(CLONE_FILES, &copy);
2903 if (error || !copy) {
2904 *displaced = NULL;
2905 return error;
2906 }
2907 *displaced = task->files;
2908 task_lock(task);
2909 task->files = copy;
2910 task_unlock(task);
2911 return 0;
2912}
Heinrich Schuchardt16db3d32015-04-16 12:47:50 -07002913
2914int sysctl_max_threads(struct ctl_table *table, int write,
2915 void __user *buffer, size_t *lenp, loff_t *ppos)
2916{
2917 struct ctl_table t;
2918 int ret;
2919 int threads = max_threads;
2920 int min = MIN_THREADS;
2921 int max = MAX_THREADS;
2922
2923 t = *table;
2924 t.data = &threads;
2925 t.extra1 = &min;
2926 t.extra2 = &max;
2927
2928 ret = proc_dointvec_minmax(&t, write, buffer, lenp, ppos);
2929 if (ret || !write)
2930 return ret;
2931
2932 set_max_threads(threads);
2933
2934 return 0;
2935}