blob: 13282510bc0da71aa9789ff72ba015bede5c96e7 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/ipc/shm.c
3 * Copyright (C) 1992, 1993 Krishna Balasubramanian
4 * Many improvements/fixes by Bruno Haible.
5 * Replaced `struct shm_desc' by `struct vm_area_struct', July 1994.
6 * Fixed the shm swap deallocation (shm_unuse()), August 1998 Andrea Arcangeli.
7 *
8 * /proc/sysvipc/shm support (c) 1999 Dragos Acostachioaie <dragos@iname.com>
9 * BIGMEM support, Andrea Arcangeli <andrea@suse.de>
10 * SMP thread shm, Jean-Luc Boyard <jean-luc.boyard@siemens.fr>
11 * HIGHMEM support, Ingo Molnar <mingo@redhat.com>
12 * Make shmmax, shmall, shmmni sysctl'able, Christoph Rohland <cr@sap.com>
13 * Shared /dev/zero support, Kanoj Sarcar <kanoj@sgi.com>
14 * Move the mm functionality over to mm/shmem.c, Christoph Rohland <cr@sap.com>
15 *
Steve Grubb073115d2006-04-02 17:07:33 -040016 * support for audit of ipc object properties and permission changes
17 * Dustin Kirkland <dustin.kirkland@us.ibm.com>
Kirill Korotaev4e982312006-10-02 02:18:22 -070018 *
19 * namespaces support
20 * OpenVZ, SWsoft Inc.
21 * Pavel Emelianov <xemul@openvz.org>
Davidlohr Buesoc2c737a2013-09-11 14:26:23 -070022 *
23 * Better ipc lock (kern_ipc_perm.lock) handling
24 * Davidlohr Bueso <davidlohr.bueso@hp.com>, June 2013.
Linus Torvalds1da177e2005-04-16 15:20:36 -070025 */
26
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/slab.h>
28#include <linux/mm.h>
29#include <linux/hugetlb.h>
30#include <linux/shm.h>
31#include <linux/init.h>
32#include <linux/file.h>
33#include <linux/mman.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#include <linux/shmem_fs.h>
35#include <linux/security.h>
36#include <linux/syscalls.h>
37#include <linux/audit.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080038#include <linux/capability.h>
Stephen Rothwell7d87e14c2005-05-01 08:59:12 -070039#include <linux/ptrace.h>
Mike Waychison19b49462005-09-06 15:17:10 -070040#include <linux/seq_file.h>
Nadia Derbey3e148c72007-10-18 23:40:54 -070041#include <linux/rwsem.h>
Kirill Korotaev4e982312006-10-02 02:18:22 -070042#include <linux/nsproxy.h>
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -080043#include <linux/mount.h>
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -080044#include <linux/ipc_namespace.h>
Stephen Rothwell7d87e14c2005-05-01 08:59:12 -070045
Paul McQuade7153e402014-06-06 14:37:37 -070046#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070047
48#include "util.h"
49
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -080050struct shm_file_data {
51 int id;
52 struct ipc_namespace *ns;
53 struct file *file;
54 const struct vm_operations_struct *vm_ops;
55};
56
57#define shm_file_data(file) (*((struct shm_file_data **)&(file)->private_data))
58
Arjan van de Ven9a321442007-02-12 00:55:35 -080059static const struct file_operations shm_file_operations;
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +040060static const struct vm_operations_struct shm_vm_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
Pierre Peiffered2ddbf2008-02-08 04:18:57 -080062#define shm_ids(ns) ((ns)->ids[IPC_SHM_IDS])
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
Kirill Korotaev4e982312006-10-02 02:18:22 -070064#define shm_unlock(shp) \
65 ipc_unlock(&(shp)->shm_perm)
Kirill Korotaev4e982312006-10-02 02:18:22 -070066
Nadia Derbey7748dbf2007-10-18 23:40:49 -070067static int newseg(struct ipc_namespace *, struct ipc_params *);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -080068static void shm_open(struct vm_area_struct *vma);
69static void shm_close(struct vm_area_struct *vma);
Manfred Spraul239521f2014-01-27 17:07:04 -080070static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -070071#ifdef CONFIG_PROC_FS
Mike Waychison19b49462005-09-06 15:17:10 -070072static int sysvipc_shm_proc_show(struct seq_file *s, void *it);
Linus Torvalds1da177e2005-04-16 15:20:36 -070073#endif
74
Pierre Peiffered2ddbf2008-02-08 04:18:57 -080075void shm_init_ns(struct ipc_namespace *ns)
Kirill Korotaev4e982312006-10-02 02:18:22 -070076{
Kirill Korotaev4e982312006-10-02 02:18:22 -070077 ns->shm_ctlmax = SHMMAX;
78 ns->shm_ctlall = SHMALL;
79 ns->shm_ctlmni = SHMMNI;
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -070080 ns->shm_rmid_forced = 0;
Kirill Korotaev4e982312006-10-02 02:18:22 -070081 ns->shm_tot = 0;
WANG Conge8148f72009-01-06 14:42:49 -080082 ipc_init_ids(&shm_ids(ns));
Kirill Korotaev4e982312006-10-02 02:18:22 -070083}
Linus Torvalds1da177e2005-04-16 15:20:36 -070084
Nadia Derbeyf4566f02007-10-18 23:40:53 -070085/*
Davidlohr Buesod9a605e2013-09-11 14:26:24 -070086 * Called with shm_ids.rwsem (writer) and the shp structure locked.
87 * Only shm_ids.rwsem remains locked on exit.
Nadia Derbeyf4566f02007-10-18 23:40:53 -070088 */
Pierre Peiffer01b8b072008-02-08 04:18:57 -080089static void do_shm_rmid(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp)
Kirill Korotaev4e982312006-10-02 02:18:22 -070090{
Pierre Peiffer01b8b072008-02-08 04:18:57 -080091 struct shmid_kernel *shp;
92 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
93
Manfred Spraul239521f2014-01-27 17:07:04 -080094 if (shp->shm_nattch) {
Kirill Korotaev4e982312006-10-02 02:18:22 -070095 shp->shm_perm.mode |= SHM_DEST;
96 /* Do not find it any more */
97 shp->shm_perm.key = IPC_PRIVATE;
98 shm_unlock(shp);
99 } else
100 shm_destroy(ns, shp);
101}
102
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -0800103#ifdef CONFIG_IPC_NS
Kirill Korotaev4e982312006-10-02 02:18:22 -0700104void shm_exit_ns(struct ipc_namespace *ns)
105{
Pierre Peiffer01b8b072008-02-08 04:18:57 -0800106 free_ipcs(ns, &shm_ids(ns), do_shm_rmid);
Serge E. Hallyn7d6feeb2009-12-15 16:47:27 -0800107 idr_destroy(&ns->ids[IPC_SHM_IDS].ipcs_idr);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700108}
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -0800109#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110
Linus Torvalds140d0b22011-08-04 19:35:59 -1000111static int __init ipc_ns_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112{
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800113 shm_init_ns(&init_ipc_ns);
Linus Torvalds140d0b22011-08-04 19:35:59 -1000114 return 0;
115}
116
117pure_initcall(ipc_ns_init);
118
Manfred Spraul239521f2014-01-27 17:07:04 -0800119void __init shm_init(void)
Linus Torvalds140d0b22011-08-04 19:35:59 -1000120{
Mike Waychison19b49462005-09-06 15:17:10 -0700121 ipc_init_proc_interface("sysvipc/shm",
Helge Dellerb7952182010-10-27 15:34:16 -0700122#if BITS_PER_LONG <= 32
123 " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime rss swap\n",
124#else
125 " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime rss swap\n",
126#endif
Kirill Korotaev4e982312006-10-02 02:18:22 -0700127 IPC_SHM_IDS, sysvipc_shm_proc_show);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128}
129
Davidlohr Bueso8b8d52a2013-09-11 14:26:15 -0700130static inline struct shmid_kernel *shm_obtain_object(struct ipc_namespace *ns, int id)
131{
Davidlohr Bueso55b7ae52015-06-30 14:58:42 -0700132 struct kern_ipc_perm *ipcp = ipc_obtain_object_idr(&shm_ids(ns), id);
Davidlohr Bueso8b8d52a2013-09-11 14:26:15 -0700133
134 if (IS_ERR(ipcp))
135 return ERR_CAST(ipcp);
136
137 return container_of(ipcp, struct shmid_kernel, shm_perm);
138}
139
140static inline struct shmid_kernel *shm_obtain_object_check(struct ipc_namespace *ns, int id)
141{
142 struct kern_ipc_perm *ipcp = ipc_obtain_object_check(&shm_ids(ns), id);
143
144 if (IS_ERR(ipcp))
145 return ERR_CAST(ipcp);
146
147 return container_of(ipcp, struct shmid_kernel, shm_perm);
148}
149
Nadia Derbey3e148c72007-10-18 23:40:54 -0700150/*
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700151 * shm_lock_(check_) routines are called in the paths where the rwsem
Nadia Derbey00c2bf82008-07-25 01:48:03 -0700152 * is not necessarily held.
Nadia Derbey3e148c72007-10-18 23:40:54 -0700153 */
Nadia Derbey023a5352007-10-18 23:40:51 -0700154static inline struct shmid_kernel *shm_lock(struct ipc_namespace *ns, int id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700156 struct kern_ipc_perm *ipcp = ipc_lock(&shm_ids(ns), id);
157
Davidlohr Buesoc5c89752015-06-30 14:58:36 -0700158 /*
Kirill A. Shutemov1ac0b6d2016-02-17 13:11:35 -0800159 * Callers of shm_lock() must validate the status of the returned ipc
160 * object pointer (as returned by ipc_lock()), and error out as
161 * appropriate.
Davidlohr Buesoc5c89752015-06-30 14:58:36 -0700162 */
Kirill A. Shutemov1ac0b6d2016-02-17 13:11:35 -0800163 if (IS_ERR(ipcp))
164 return (void *)ipcp;
Nadia Derbey03f02c72007-10-18 23:40:51 -0700165 return container_of(ipcp, struct shmid_kernel, shm_perm);
Nadia Derbey023a5352007-10-18 23:40:51 -0700166}
167
Vasiliy Kulikov4c677e22011-07-29 03:56:40 +0400168static inline void shm_lock_by_ptr(struct shmid_kernel *ipcp)
169{
170 rcu_read_lock();
Davidlohr Buesocf9d5d72013-07-08 16:01:11 -0700171 ipc_lock_object(&ipcp->shm_perm);
Vasiliy Kulikov4c677e22011-07-29 03:56:40 +0400172}
173
Davidlohr Bueso53dad6d2013-09-23 17:04:45 -0700174static void shm_rcu_free(struct rcu_head *head)
175{
176 struct ipc_rcu *p = container_of(head, struct ipc_rcu, rcu);
177 struct shmid_kernel *shp = ipc_rcu_to_struct(p);
178
179 security_shm_free(shp);
180 ipc_rcu_free(head);
181}
182
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700183static inline void shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *s)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184{
Jack Millerab602f72014-08-08 14:23:19 -0700185 list_del(&s->shm_clist);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700186 ipc_rmid(&shm_ids(ns), &s->shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187}
188
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189
Kirill A. Shutemov1ac0b6d2016-02-17 13:11:35 -0800190static int __shm_open(struct vm_area_struct *vma)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700191{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800192 struct file *file = vma->vm_file;
193 struct shm_file_data *sfd = shm_file_data(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 struct shmid_kernel *shp;
195
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800196 shp = shm_lock(sfd->ns, sfd->id);
Kirill A. Shutemov1ac0b6d2016-02-17 13:11:35 -0800197
198 if (IS_ERR(shp))
199 return PTR_ERR(shp);
200
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 shp->shm_atim = get_seconds();
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700202 shp->shm_lprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 shp->shm_nattch++;
204 shm_unlock(shp);
Kirill A. Shutemov1ac0b6d2016-02-17 13:11:35 -0800205 return 0;
206}
207
208/* This is called by fork, once for every shm attach. */
209static void shm_open(struct vm_area_struct *vma)
210{
211 int err = __shm_open(vma);
212 /*
213 * We raced in the idr lookup or with shm_destroy().
214 * Either way, the ID is busted.
215 */
216 WARN_ON_ONCE(err);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217}
218
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219/*
220 * shm_destroy - free the struct shmid_kernel
221 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700222 * @ns: namespace
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 * @shp: struct to free
224 *
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700225 * It has to be called with shp and shm_ids.rwsem (writer) locked,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 * but returns with shp unlocked and freed.
227 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700228static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229{
Greg Thelena399b292013-11-21 14:32:00 -0800230 struct file *shm_file;
231
232 shm_file = shp->shm_file;
233 shp->shm_file = NULL;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700234 ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700235 shm_rmid(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 shm_unlock(shp);
Greg Thelena399b292013-11-21 14:32:00 -0800237 if (!is_file_hugepages(shm_file))
238 shmem_lock(shm_file, 0, shp->mlock_user);
Hugh Dickins353d5c32009-08-24 16:30:28 +0100239 else if (shp->mlock_user)
Dave Hansen07a46ed2014-12-12 16:58:22 -0800240 user_shm_unlock(i_size_read(file_inode(shm_file)),
241 shp->mlock_user);
Greg Thelena399b292013-11-21 14:32:00 -0800242 fput(shm_file);
Davidlohr Bueso53dad6d2013-09-23 17:04:45 -0700243 ipc_rcu_putref(shp, shm_rcu_free);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244}
245
246/*
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700247 * shm_may_destroy - identifies whether shm segment should be destroyed now
248 *
249 * Returns true if and only if there are no active users of the segment and
250 * one of the following is true:
251 *
252 * 1) shmctl(id, IPC_RMID, NULL) was called for this shp
253 *
254 * 2) sysctl kernel.shm_rmid_forced is set to 1.
255 */
256static bool shm_may_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp)
257{
258 return (shp->shm_nattch == 0) &&
259 (ns->shm_rmid_forced ||
260 (shp->shm_perm.mode & SHM_DEST));
261}
262
263/*
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800264 * remove the attach descriptor vma.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 * free memory for segment if it is marked destroyed.
266 * The descriptor has already been removed from the current->mm->mmap list
267 * and will later be kfree()d.
268 */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800269static void shm_close(struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270{
Manfred Spraul239521f2014-01-27 17:07:04 -0800271 struct file *file = vma->vm_file;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800272 struct shm_file_data *sfd = shm_file_data(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 struct shmid_kernel *shp;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800274 struct ipc_namespace *ns = sfd->ns;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700275
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700276 down_write(&shm_ids(ns).rwsem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 /* remove from the list of attaches of the shm segment */
Nadia Derbey00c2bf82008-07-25 01:48:03 -0700278 shp = shm_lock(ns, sfd->id);
Kirill A. Shutemov1ac0b6d2016-02-17 13:11:35 -0800279
280 /*
281 * We raced in the idr lookup or with shm_destroy().
282 * Either way, the ID is busted.
283 */
284 if (WARN_ON_ONCE(IS_ERR(shp)))
285 goto done; /* no-op */
286
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700287 shp->shm_lprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 shp->shm_dtim = get_seconds();
289 shp->shm_nattch--;
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700290 if (shm_may_destroy(ns, shp))
Kirill Korotaev4e982312006-10-02 02:18:22 -0700291 shm_destroy(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292 else
293 shm_unlock(shp);
Kirill A. Shutemov1ac0b6d2016-02-17 13:11:35 -0800294done:
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700295 up_write(&shm_ids(ns).rwsem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296}
297
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700298/* Called with ns->shm_ids(ns).rwsem locked */
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700299static int shm_try_destroy_orphaned(int id, void *p, void *data)
300{
301 struct ipc_namespace *ns = data;
Vasiliy Kulikov4c677e22011-07-29 03:56:40 +0400302 struct kern_ipc_perm *ipcp = p;
303 struct shmid_kernel *shp = container_of(ipcp, struct shmid_kernel, shm_perm);
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700304
305 /*
306 * We want to destroy segments without users and with already
307 * exit'ed originating process.
Vasiliy Kulikov4c677e22011-07-29 03:56:40 +0400308 *
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700309 * As shp->* are changed under rwsem, it's safe to skip shp locking.
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700310 */
Vasiliy Kulikov4c677e22011-07-29 03:56:40 +0400311 if (shp->shm_creator != NULL)
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700312 return 0;
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700313
Vasiliy Kulikov4c677e22011-07-29 03:56:40 +0400314 if (shm_may_destroy(ns, shp)) {
315 shm_lock_by_ptr(shp);
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700316 shm_destroy(ns, shp);
Vasiliy Kulikov4c677e22011-07-29 03:56:40 +0400317 }
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700318 return 0;
319}
320
321void shm_destroy_orphaned(struct ipc_namespace *ns)
322{
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700323 down_write(&shm_ids(ns).rwsem);
Vasiliy Kulikov33a30ed2011-08-03 22:26:55 +0400324 if (shm_ids(ns).in_use)
Vasiliy Kulikov4c677e22011-07-29 03:56:40 +0400325 idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_orphaned, ns);
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700326 up_write(&shm_ids(ns).rwsem);
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700327}
328
Jack Miller83293c02014-08-08 14:23:21 -0700329/* Locking assumes this will only be called with task == current */
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700330void exit_shm(struct task_struct *task)
331{
Vasiliy Kulikov4c677e22011-07-29 03:56:40 +0400332 struct ipc_namespace *ns = task->nsproxy->ipc_ns;
Jack Millerab602f72014-08-08 14:23:19 -0700333 struct shmid_kernel *shp, *n;
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700334
Jack Miller83293c02014-08-08 14:23:21 -0700335 if (list_empty(&task->sysvshm.shm_clist))
Vasiliy Kulikov298507d2011-08-03 22:28:26 +0400336 return;
337
Jack Miller83293c02014-08-08 14:23:21 -0700338 /*
339 * If kernel.shm_rmid_forced is not set then only keep track of
340 * which shmids are orphaned, so that a later set of the sysctl
341 * can clean them up.
342 */
343 if (!ns->shm_rmid_forced) {
344 down_read(&shm_ids(ns).rwsem);
345 list_for_each_entry(shp, &task->sysvshm.shm_clist, shm_clist)
346 shp->shm_creator = NULL;
347 /*
348 * Only under read lock but we are only called on current
349 * so no entry on the list will be shared.
350 */
351 list_del(&task->sysvshm.shm_clist);
352 up_read(&shm_ids(ns).rwsem);
353 return;
354 }
355
356 /*
357 * Destroy all already created segments, that were not yet mapped,
358 * and mark any mapped as orphan to cover the sysctl toggling.
359 * Destroy is skipped if shm_may_destroy() returns false.
360 */
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700361 down_write(&shm_ids(ns).rwsem);
Jack Miller83293c02014-08-08 14:23:21 -0700362 list_for_each_entry_safe(shp, n, &task->sysvshm.shm_clist, shm_clist) {
363 shp->shm_creator = NULL;
364
365 if (shm_may_destroy(ns, shp)) {
366 shm_lock_by_ptr(shp);
367 shm_destroy(ns, shp);
368 }
369 }
370
371 /* Remove the list head from any segments still attached. */
Jack Millerab602f72014-08-08 14:23:19 -0700372 list_del(&task->sysvshm.shm_clist);
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700373 up_write(&shm_ids(ns).rwsem);
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -0700374}
375
Nick Piggind0217ac2007-07-19 01:47:03 -0700376static int shm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800377{
378 struct file *file = vma->vm_file;
379 struct shm_file_data *sfd = shm_file_data(file);
380
Nick Piggind0217ac2007-07-19 01:47:03 -0700381 return sfd->vm_ops->fault(vma, vmf);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800382}
383
384#ifdef CONFIG_NUMA
Adrian Bunkd823e3e2007-10-16 23:26:42 -0700385static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800386{
387 struct file *file = vma->vm_file;
388 struct shm_file_data *sfd = shm_file_data(file);
389 int err = 0;
390 if (sfd->vm_ops->set_policy)
391 err = sfd->vm_ops->set_policy(vma, new);
392 return err;
393}
394
Adrian Bunkd823e3e2007-10-16 23:26:42 -0700395static struct mempolicy *shm_get_policy(struct vm_area_struct *vma,
396 unsigned long addr)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800397{
398 struct file *file = vma->vm_file;
399 struct shm_file_data *sfd = shm_file_data(file);
400 struct mempolicy *pol = NULL;
401
402 if (sfd->vm_ops->get_policy)
403 pol = sfd->vm_ops->get_policy(vma, addr);
Lee Schermerhorn52cd3b02008-04-28 02:13:16 -0700404 else if (vma->vm_policy)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800405 pol = vma->vm_policy;
Lee Schermerhorn52cd3b02008-04-28 02:13:16 -0700406
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800407 return pol;
408}
409#endif
410
Manfred Spraul239521f2014-01-27 17:07:04 -0800411static int shm_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800413 struct shm_file_data *sfd = shm_file_data(file);
David Howellsb0e15192006-01-06 00:11:42 -0800414 int ret;
415
Kirill A. Shutemov1ac0b6d2016-02-17 13:11:35 -0800416 /*
417 * In case of remap_file_pages() emulation, the file can represent
418 * removed IPC ID: propogate shm_lock() error to caller.
419 */
420 ret =__shm_open(vma);
421 if (ret)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800422 return ret;
Kirill A. Shutemov1ac0b6d2016-02-17 13:11:35 -0800423
424 ret = sfd->file->f_op->mmap(sfd->file, vma);
425 if (ret) {
426 shm_close(vma);
427 return ret;
428 }
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800429 sfd->vm_ops = vma->vm_ops;
David Howells2e92a3b2007-07-31 00:37:24 -0700430#ifdef CONFIG_MMU
Davidlohr Buesod0edd852015-09-09 15:39:20 -0700431 WARN_ON(!sfd->vm_ops->fault);
David Howells2e92a3b2007-07-31 00:37:24 -0700432#endif
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800433 vma->vm_ops = &shm_vm_ops;
Kirill A. Shutemov1ac0b6d2016-02-17 13:11:35 -0800434 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435}
436
Kirill Korotaev4e982312006-10-02 02:18:22 -0700437static int shm_release(struct inode *ino, struct file *file)
438{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800439 struct shm_file_data *sfd = shm_file_data(file);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700440
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800441 put_ipc_ns(sfd->ns);
442 shm_file_data(file) = NULL;
443 kfree(sfd);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700444 return 0;
445}
446
Josef Bacik02c24a82011-07-16 20:44:56 -0400447static int shm_fsync(struct file *file, loff_t start, loff_t end, int datasync)
Adam Litke516dffd2007-03-01 15:46:08 -0800448{
Adam Litke516dffd2007-03-01 15:46:08 -0800449 struct shm_file_data *sfd = shm_file_data(file);
Adam Litke516dffd2007-03-01 15:46:08 -0800450
Christoph Hellwig7ea80852010-05-26 17:53:25 +0200451 if (!sfd->file->f_op->fsync)
452 return -EINVAL;
Josef Bacik02c24a82011-07-16 20:44:56 -0400453 return sfd->file->f_op->fsync(sfd->file, start, end, datasync);
Adam Litke516dffd2007-03-01 15:46:08 -0800454}
455
Will Deacon7d8a4562012-06-07 14:21:13 -0700456static long shm_fallocate(struct file *file, int mode, loff_t offset,
457 loff_t len)
458{
459 struct shm_file_data *sfd = shm_file_data(file);
460
461 if (!sfd->file->f_op->fallocate)
462 return -EOPNOTSUPP;
463 return sfd->file->f_op->fallocate(file, mode, offset, len);
464}
465
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800466static unsigned long shm_get_unmapped_area(struct file *file,
467 unsigned long addr, unsigned long len, unsigned long pgoff,
468 unsigned long flags)
469{
470 struct shm_file_data *sfd = shm_file_data(file);
Al Viroc4caa772009-11-30 08:38:43 -0500471 return sfd->file->f_op->get_unmapped_area(sfd->file, addr, len,
472 pgoff, flags);
Adam Litke516dffd2007-03-01 15:46:08 -0800473}
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800474
Arjan van de Ven9a321442007-02-12 00:55:35 -0800475static const struct file_operations shm_file_operations = {
Kirill Korotaev4e982312006-10-02 02:18:22 -0700476 .mmap = shm_mmap,
Adam Litke516dffd2007-03-01 15:46:08 -0800477 .fsync = shm_fsync,
Kirill Korotaev4e982312006-10-02 02:18:22 -0700478 .release = shm_release,
David Howellsed5e5892010-01-15 17:01:32 -0800479#ifndef CONFIG_MMU
480 .get_unmapped_area = shm_get_unmapped_area,
481#endif
Arnd Bergmann6038f372010-08-15 18:52:59 +0200482 .llseek = noop_llseek,
Will Deacon7d8a4562012-06-07 14:21:13 -0700483 .fallocate = shm_fallocate,
Al Viroc4caa772009-11-30 08:38:43 -0500484};
485
486static const struct file_operations shm_file_operations_huge = {
487 .mmap = shm_mmap,
488 .fsync = shm_fsync,
489 .release = shm_release,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800490 .get_unmapped_area = shm_get_unmapped_area,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200491 .llseek = noop_llseek,
Will Deacon7d8a4562012-06-07 14:21:13 -0700492 .fallocate = shm_fallocate,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493};
494
Yaowei Bai2954e4402016-01-20 15:01:11 -0800495bool is_file_shm_hugepages(struct file *file)
Al Viroc4caa772009-11-30 08:38:43 -0500496{
497 return file->f_op == &shm_file_operations_huge;
498}
499
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +0400500static const struct vm_operations_struct shm_vm_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 .open = shm_open, /* callback for a new vm-area open */
502 .close = shm_close, /* callback for when the vm-area is released */
Nick Piggin54cb8822007-07-19 01:46:59 -0700503 .fault = shm_fault,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800504#if defined(CONFIG_NUMA)
505 .set_policy = shm_set_policy,
506 .get_policy = shm_get_policy,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507#endif
508};
509
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700510/**
511 * newseg - Create a new shared memory segment
512 * @ns: namespace
513 * @params: ptr to the structure that contains key, size and shmflg
514 *
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700515 * Called with shm_ids.rwsem held as a writer.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700516 */
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700517static int newseg(struct ipc_namespace *ns, struct ipc_params *params)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518{
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700519 key_t key = params->key;
520 int shmflg = params->flg;
521 size_t size = params->u.size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 int error;
523 struct shmid_kernel *shp;
Robin Holtd69f3ba2013-04-30 19:15:54 -0700524 size_t numpages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
Manfred Spraul239521f2014-01-27 17:07:04 -0800525 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 char name[13];
527 int id;
KOSAKI Motohiroca16d142011-05-26 19:16:19 +0900528 vm_flags_t acctflag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529
Kirill Korotaev4e982312006-10-02 02:18:22 -0700530 if (size < SHMMIN || size > ns->shm_ctlmax)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 return -EINVAL;
532
Manfred Spraul13763272014-06-06 14:37:41 -0700533 if (numpages << PAGE_SHIFT < size)
534 return -ENOSPC;
535
Manfred Spraul09c6eb12014-06-06 14:37:40 -0700536 if (ns->shm_tot + numpages < ns->shm_tot ||
537 ns->shm_tot + numpages > ns->shm_ctlall)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 return -ENOSPC;
539
540 shp = ipc_rcu_alloc(sizeof(*shp));
541 if (!shp)
542 return -ENOMEM;
543
544 shp->shm_perm.key = key;
Andrew Mortonb33291c2006-01-08 01:02:21 -0800545 shp->shm_perm.mode = (shmflg & S_IRWXUGO);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 shp->mlock_user = NULL;
547
548 shp->shm_perm.security = NULL;
549 error = security_shm_alloc(shp);
550 if (error) {
Davidlohr Bueso53dad6d2013-09-23 17:04:45 -0700551 ipc_rcu_putref(shp, ipc_rcu_free);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 return error;
553 }
554
Manfred Spraul239521f2014-01-27 17:07:04 -0800555 sprintf(name, "SYSV%08x", key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 if (shmflg & SHM_HUGETLB) {
Andrew Mortonc103a4d2013-07-08 16:01:08 -0700557 struct hstate *hs;
Li Zefan091d0d52013-05-09 15:08:15 +0800558 size_t hugesize;
559
Andrew Mortonc103a4d2013-07-08 16:01:08 -0700560 hs = hstate_sizelog((shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK);
Li Zefan091d0d52013-05-09 15:08:15 +0800561 if (!hs) {
562 error = -EINVAL;
563 goto no_file;
564 }
565 hugesize = ALIGN(size, huge_page_size(hs));
Naoya Horiguchiaf73e4d2013-05-07 16:18:13 -0700566
Mel Gorman5a6fe122009-02-10 14:02:27 +0000567 /* hugetlb_file_setup applies strict accounting */
568 if (shmflg & SHM_NORESERVE)
569 acctflag = VM_NORESERVE;
Naoya Horiguchiaf73e4d2013-05-07 16:18:13 -0700570 file = hugetlb_file_setup(name, hugesize, acctflag,
Andi Kleen42d73952012-12-11 16:01:34 -0800571 &shp->mlock_user, HUGETLB_SHMFS_INODE,
572 (shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 } else {
Badari Pulavartybf8f9722005-11-07 00:59:27 -0800574 /*
575 * Do not allow no accounting for OVERCOMMIT_NEVER, even
Manfred Spraul239521f2014-01-27 17:07:04 -0800576 * if it's asked for.
Badari Pulavartybf8f9722005-11-07 00:59:27 -0800577 */
578 if ((shmflg & SHM_NORESERVE) &&
579 sysctl_overcommit_memory != OVERCOMMIT_NEVER)
Linus Torvaldsfc8744a2009-01-31 15:08:56 -0800580 acctflag = VM_NORESERVE;
Stephen Smalleye1832f22015-08-06 15:46:55 -0700581 file = shmem_kernel_file_setup(name, size, acctflag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 }
583 error = PTR_ERR(file);
584 if (IS_ERR(file))
585 goto no_file;
586
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700587 shp->shm_cprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 shp->shm_lprid = 0;
589 shp->shm_atim = shp->shm_dtim = 0;
590 shp->shm_ctim = get_seconds();
591 shp->shm_segsz = size;
592 shp->shm_nattch = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 shp->shm_file = file;
Vasiliy Kulikov5774ed02011-07-29 03:55:31 +0400594 shp->shm_creator = current;
Linus Torvaldsb9a53222015-09-30 12:48:40 -0400595
596 id = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni);
597 if (id < 0) {
598 error = id;
599 goto no_id;
600 }
601
Jack Millerab602f72014-08-08 14:23:19 -0700602 list_add(&shp->shm_clist, &current->sysvshm.shm_clist);
Davidlohr Buesodbfcd912013-07-08 16:01:09 -0700603
Badari Pulavarty30475cc2007-06-16 10:15:59 -0700604 /*
605 * shmid gets reported as "inode#" in /proc/pid/maps.
606 * proc-ps tools use this. Changing this will break them.
607 */
Al Viro496ad9a2013-01-23 17:07:38 -0500608 file_inode(file)->i_ino = shp->shm_perm.id;
Krishnakumar R551110a2005-10-29 18:16:45 -0700609
Kirill Korotaev4e982312006-10-02 02:18:22 -0700610 ns->shm_tot += numpages;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700611 error = shp->shm_perm.id;
Davidlohr Buesodbfcd912013-07-08 16:01:09 -0700612
Davidlohr Buesocf9d5d72013-07-08 16:01:11 -0700613 ipc_unlock_object(&shp->shm_perm);
Davidlohr Buesodbfcd912013-07-08 16:01:09 -0700614 rcu_read_unlock();
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700615 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616
617no_id:
Hugh Dickins2195d282009-09-12 12:21:27 +0100618 if (is_file_hugepages(file) && shp->mlock_user)
Hugh Dickins353d5c32009-08-24 16:30:28 +0100619 user_shm_unlock(size, shp->mlock_user);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 fput(file);
621no_file:
Davidlohr Bueso53dad6d2013-09-23 17:04:45 -0700622 ipc_rcu_putref(shp, shm_rcu_free);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 return error;
624}
625
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700626/*
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700627 * Called with shm_ids.rwsem and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700628 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700629static inline int shm_security(struct kern_ipc_perm *ipcp, int shmflg)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700630{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700631 struct shmid_kernel *shp;
632
633 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
634 return security_shm_associate(shp, shmflg);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700635}
636
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700637/*
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700638 * Called with shm_ids.rwsem and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700639 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700640static inline int shm_more_checks(struct kern_ipc_perm *ipcp,
641 struct ipc_params *params)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700642{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700643 struct shmid_kernel *shp;
644
645 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
646 if (shp->shm_segsz < params->u.size)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700647 return -EINVAL;
648
649 return 0;
650}
651
Heiko Carstensd5460c92009-01-14 14:14:27 +0100652SYSCALL_DEFINE3(shmget, key_t, key, size_t, size, int, shmflg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700654 struct ipc_namespace *ns;
Mathias Krauseeb66ec42014-06-06 14:37:36 -0700655 static const struct ipc_ops shm_ops = {
656 .getnew = newseg,
657 .associate = shm_security,
658 .more_checks = shm_more_checks,
659 };
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700660 struct ipc_params shm_params;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661
Kirill Korotaev4e982312006-10-02 02:18:22 -0700662 ns = current->nsproxy->ipc_ns;
663
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700664 shm_params.key = key;
665 shm_params.flg = shmflg;
666 shm_params.u.size = size;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700667
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700668 return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669}
670
671static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version)
672{
Manfred Spraul239521f2014-01-27 17:07:04 -0800673 switch (version) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 case IPC_64:
675 return copy_to_user(buf, in, sizeof(*in));
676 case IPC_OLD:
677 {
678 struct shmid_ds out;
679
Vasiliy Kulikov3af54c92010-10-30 18:22:49 +0400680 memset(&out, 0, sizeof(out));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm);
682 out.shm_segsz = in->shm_segsz;
683 out.shm_atime = in->shm_atime;
684 out.shm_dtime = in->shm_dtime;
685 out.shm_ctime = in->shm_ctime;
686 out.shm_cpid = in->shm_cpid;
687 out.shm_lpid = in->shm_lpid;
688 out.shm_nattch = in->shm_nattch;
689
690 return copy_to_user(buf, &out, sizeof(out));
691 }
692 default:
693 return -EINVAL;
694 }
695}
696
Pierre Peiffer016d7132008-04-29 01:00:50 -0700697static inline unsigned long
698copy_shmid_from_user(struct shmid64_ds *out, void __user *buf, int version)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699{
Manfred Spraul239521f2014-01-27 17:07:04 -0800700 switch (version) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 case IPC_64:
Pierre Peiffer016d7132008-04-29 01:00:50 -0700702 if (copy_from_user(out, buf, sizeof(*out)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 case IPC_OLD:
706 {
707 struct shmid_ds tbuf_old;
708
709 if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old)))
710 return -EFAULT;
711
Pierre Peiffer016d7132008-04-29 01:00:50 -0700712 out->shm_perm.uid = tbuf_old.shm_perm.uid;
713 out->shm_perm.gid = tbuf_old.shm_perm.gid;
714 out->shm_perm.mode = tbuf_old.shm_perm.mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715
716 return 0;
717 }
718 default:
719 return -EINVAL;
720 }
721}
722
723static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version)
724{
Manfred Spraul239521f2014-01-27 17:07:04 -0800725 switch (version) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 case IPC_64:
727 return copy_to_user(buf, in, sizeof(*in));
728 case IPC_OLD:
729 {
730 struct shminfo out;
731
Manfred Spraul239521f2014-01-27 17:07:04 -0800732 if (in->shmmax > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 out.shmmax = INT_MAX;
734 else
735 out.shmmax = (int)in->shmmax;
736
737 out.shmmin = in->shmmin;
738 out.shmmni = in->shmmni;
739 out.shmseg = in->shmseg;
Paul McQuade46c0a8c2014-06-06 14:37:37 -0700740 out.shmall = in->shmall;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741
742 return copy_to_user(buf, &out, sizeof(out));
743 }
744 default:
745 return -EINVAL;
746 }
747}
748
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700749/*
Helge Dellerb7952182010-10-27 15:34:16 -0700750 * Calculate and add used RSS and swap pages of a shm.
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700751 * Called with shm_ids.rwsem held as a reader
Helge Dellerb7952182010-10-27 15:34:16 -0700752 */
753static void shm_add_rss_swap(struct shmid_kernel *shp,
754 unsigned long *rss_add, unsigned long *swp_add)
755{
756 struct inode *inode;
757
Al Viro496ad9a2013-01-23 17:07:38 -0500758 inode = file_inode(shp->shm_file);
Helge Dellerb7952182010-10-27 15:34:16 -0700759
760 if (is_file_hugepages(shp->shm_file)) {
761 struct address_space *mapping = inode->i_mapping;
762 struct hstate *h = hstate_file(shp->shm_file);
763 *rss_add += pages_per_huge_page(h) * mapping->nrpages;
764 } else {
765#ifdef CONFIG_SHMEM
766 struct shmem_inode_info *info = SHMEM_I(inode);
767 spin_lock(&info->lock);
768 *rss_add += inode->i_mapping->nrpages;
769 *swp_add += info->swapped;
770 spin_unlock(&info->lock);
771#else
772 *rss_add += inode->i_mapping->nrpages;
773#endif
774 }
775}
776
777/*
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700778 * Called with shm_ids.rwsem held as a reader
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700779 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700780static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss,
781 unsigned long *swp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700783 int next_id;
784 int total, in_use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785
786 *rss = 0;
787 *swp = 0;
788
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700789 in_use = shm_ids(ns).in_use;
790
791 for (total = 0, next_id = 0; total < in_use; next_id++) {
Tony Battersbye562aeb2009-04-02 16:58:26 -0700792 struct kern_ipc_perm *ipc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 struct shmid_kernel *shp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
Tony Battersbye562aeb2009-04-02 16:58:26 -0700795 ipc = idr_find(&shm_ids(ns).ipcs_idr, next_id);
796 if (ipc == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 continue;
Tony Battersbye562aeb2009-04-02 16:58:26 -0700798 shp = container_of(ipc, struct shmid_kernel, shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799
Helge Dellerb7952182010-10-27 15:34:16 -0700800 shm_add_rss_swap(shp, rss, swp);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700801
802 total++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 }
804}
805
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700806/*
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700807 * This function handles some shmctl commands which require the rwsem
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700808 * to be held in write mode.
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700809 * NOTE: no locks must be held, the rwsem is taken inside this function.
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700810 */
811static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd,
812 struct shmid_ds __user *buf, int version)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813{
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700814 struct kern_ipc_perm *ipcp;
Pierre Peiffer016d7132008-04-29 01:00:50 -0700815 struct shmid64_ds shmid64;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 struct shmid_kernel *shp;
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700817 int err;
818
819 if (cmd == IPC_SET) {
Pierre Peiffer016d7132008-04-29 01:00:50 -0700820 if (copy_shmid_from_user(&shmid64, buf, version))
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700821 return -EFAULT;
822 }
823
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700824 down_write(&shm_ids(ns).rwsem);
Davidlohr Bueso7b4cc5d2013-07-08 16:01:12 -0700825 rcu_read_lock();
826
Davidlohr Bueso79ccf0f2013-09-11 14:26:16 -0700827 ipcp = ipcctl_pre_down_nolock(ns, &shm_ids(ns), shmid, cmd,
828 &shmid64.shm_perm, 0);
Davidlohr Bueso7b4cc5d2013-07-08 16:01:12 -0700829 if (IS_ERR(ipcp)) {
830 err = PTR_ERR(ipcp);
Davidlohr Bueso7b4cc5d2013-07-08 16:01:12 -0700831 goto out_unlock1;
832 }
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700833
Pierre Peiffera5f75e72008-04-29 01:00:54 -0700834 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700835
836 err = security_shm_shmctl(shp, cmd);
837 if (err)
Davidlohr Bueso79ccf0f2013-09-11 14:26:16 -0700838 goto out_unlock1;
Davidlohr Bueso7b4cc5d2013-07-08 16:01:12 -0700839
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700840 switch (cmd) {
841 case IPC_RMID:
Davidlohr Bueso79ccf0f2013-09-11 14:26:16 -0700842 ipc_lock_object(&shp->shm_perm);
Davidlohr Bueso7b4cc5d2013-07-08 16:01:12 -0700843 /* do_shm_rmid unlocks the ipc object and rcu */
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700844 do_shm_rmid(ns, ipcp);
845 goto out_up;
846 case IPC_SET:
Davidlohr Bueso79ccf0f2013-09-11 14:26:16 -0700847 ipc_lock_object(&shp->shm_perm);
Eric W. Biederman1efdb692012-02-07 16:54:11 -0800848 err = ipc_update_perm(&shmid64.shm_perm, ipcp);
849 if (err)
Davidlohr Bueso7b4cc5d2013-07-08 16:01:12 -0700850 goto out_unlock0;
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700851 shp->shm_ctim = get_seconds();
852 break;
853 default:
854 err = -EINVAL;
Davidlohr Bueso79ccf0f2013-09-11 14:26:16 -0700855 goto out_unlock1;
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700856 }
Davidlohr Bueso7b4cc5d2013-07-08 16:01:12 -0700857
858out_unlock0:
859 ipc_unlock_object(&shp->shm_perm);
860out_unlock1:
861 rcu_read_unlock();
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700862out_up:
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700863 up_write(&shm_ids(ns).rwsem);
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700864 return err;
865}
866
Davidlohr Bueso68eccc12013-09-11 14:26:18 -0700867static int shmctl_nolock(struct ipc_namespace *ns, int shmid,
868 int cmd, int version, void __user *buf)
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700869{
Davidlohr Bueso68eccc12013-09-11 14:26:18 -0700870 int err;
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700871 struct shmid_kernel *shp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872
Davidlohr Bueso68eccc12013-09-11 14:26:18 -0700873 /* preliminary security checks for *_INFO */
874 if (cmd == IPC_INFO || cmd == SHM_INFO) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 err = security_shm_shmctl(NULL, cmd);
876 if (err)
877 return err;
Davidlohr Bueso68eccc12013-09-11 14:26:18 -0700878 }
879
880 switch (cmd) {
881 case IPC_INFO:
882 {
883 struct shminfo64 shminfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884
WANG Conge8148f72009-01-06 14:42:49 -0800885 memset(&shminfo, 0, sizeof(shminfo));
Kirill Korotaev4e982312006-10-02 02:18:22 -0700886 shminfo.shmmni = shminfo.shmseg = ns->shm_ctlmni;
887 shminfo.shmmax = ns->shm_ctlmax;
888 shminfo.shmall = ns->shm_ctlall;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889
890 shminfo.shmmin = SHMMIN;
Manfred Spraul239521f2014-01-27 17:07:04 -0800891 if (copy_shminfo_to_user(buf, &shminfo, version))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892 return -EFAULT;
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700893
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700894 down_read(&shm_ids(ns).rwsem);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700895 err = ipc_get_maxid(&shm_ids(ns));
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700896 up_read(&shm_ids(ns).rwsem);
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700897
Manfred Spraul239521f2014-01-27 17:07:04 -0800898 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 err = 0;
900 goto out;
901 }
902 case SHM_INFO:
903 {
904 struct shm_info shm_info;
905
WANG Conge8148f72009-01-06 14:42:49 -0800906 memset(&shm_info, 0, sizeof(shm_info));
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700907 down_read(&shm_ids(ns).rwsem);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700908 shm_info.used_ids = shm_ids(ns).in_use;
Manfred Spraul239521f2014-01-27 17:07:04 -0800909 shm_get_stat(ns, &shm_info.shm_rss, &shm_info.shm_swp);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700910 shm_info.shm_tot = ns->shm_tot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 shm_info.swap_attempts = 0;
912 shm_info.swap_successes = 0;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700913 err = ipc_get_maxid(&shm_ids(ns));
Davidlohr Buesod9a605e2013-09-11 14:26:24 -0700914 up_read(&shm_ids(ns).rwsem);
WANG Conge8148f72009-01-06 14:42:49 -0800915 if (copy_to_user(buf, &shm_info, sizeof(shm_info))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 err = -EFAULT;
917 goto out;
918 }
919
920 err = err < 0 ? 0 : err;
921 goto out;
922 }
923 case SHM_STAT:
924 case IPC_STAT:
925 {
926 struct shmid64_ds tbuf;
927 int result;
Nadia Derbey023a5352007-10-18 23:40:51 -0700928
Davidlohr Buesoc97cb9c2013-09-11 14:26:20 -0700929 rcu_read_lock();
Nadia Derbey023a5352007-10-18 23:40:51 -0700930 if (cmd == SHM_STAT) {
Davidlohr Buesoc97cb9c2013-09-11 14:26:20 -0700931 shp = shm_obtain_object(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -0700932 if (IS_ERR(shp)) {
933 err = PTR_ERR(shp);
Davidlohr Buesoc97cb9c2013-09-11 14:26:20 -0700934 goto out_unlock;
Nadia Derbey023a5352007-10-18 23:40:51 -0700935 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700936 result = shp->shm_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 } else {
Davidlohr Buesoc97cb9c2013-09-11 14:26:20 -0700938 shp = shm_obtain_object_check(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -0700939 if (IS_ERR(shp)) {
940 err = PTR_ERR(shp);
Davidlohr Buesoc97cb9c2013-09-11 14:26:20 -0700941 goto out_unlock;
Nadia Derbey023a5352007-10-18 23:40:51 -0700942 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 result = 0;
944 }
Davidlohr Buesoc97cb9c2013-09-11 14:26:20 -0700945
WANG Conge8148f72009-01-06 14:42:49 -0800946 err = -EACCES;
Serge E. Hallynb0e77592011-03-23 16:43:24 -0700947 if (ipcperms(ns, &shp->shm_perm, S_IRUGO))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 goto out_unlock;
Davidlohr Buesoc97cb9c2013-09-11 14:26:20 -0700949
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 err = security_shm_shmctl(shp, cmd);
951 if (err)
952 goto out_unlock;
Davidlohr Buesoc97cb9c2013-09-11 14:26:20 -0700953
Nadia Derbey023a5352007-10-18 23:40:51 -0700954 memset(&tbuf, 0, sizeof(tbuf));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 kernel_to_ipc64_perm(&shp->shm_perm, &tbuf.shm_perm);
956 tbuf.shm_segsz = shp->shm_segsz;
957 tbuf.shm_atime = shp->shm_atim;
958 tbuf.shm_dtime = shp->shm_dtim;
959 tbuf.shm_ctime = shp->shm_ctim;
960 tbuf.shm_cpid = shp->shm_cprid;
961 tbuf.shm_lpid = shp->shm_lprid;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800962 tbuf.shm_nattch = shp->shm_nattch;
Davidlohr Buesoc97cb9c2013-09-11 14:26:20 -0700963 rcu_read_unlock();
964
965 if (copy_shmid_to_user(buf, &tbuf, version))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 err = -EFAULT;
967 else
968 err = result;
969 goto out;
970 }
Davidlohr Bueso68eccc12013-09-11 14:26:18 -0700971 default:
972 return -EINVAL;
973 }
974
975out_unlock:
Davidlohr Buesoc97cb9c2013-09-11 14:26:20 -0700976 rcu_read_unlock();
Davidlohr Bueso68eccc12013-09-11 14:26:18 -0700977out:
978 return err;
979}
980
981SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, struct shmid_ds __user *, buf)
982{
983 struct shmid_kernel *shp;
984 int err, version;
985 struct ipc_namespace *ns;
986
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -0700987 if (cmd < 0 || shmid < 0)
988 return -EINVAL;
Davidlohr Bueso68eccc12013-09-11 14:26:18 -0700989
990 version = ipc_parse_version(&cmd);
991 ns = current->nsproxy->ipc_ns;
992
993 switch (cmd) {
994 case IPC_INFO:
995 case SHM_INFO:
996 case SHM_STAT:
997 case IPC_STAT:
998 return shmctl_nolock(ns, shmid, cmd, version, buf);
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -0700999 case IPC_RMID:
1000 case IPC_SET:
1001 return shmctl_down(ns, shmid, cmd, buf, version);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 case SHM_LOCK:
1003 case SHM_UNLOCK:
1004 {
Hugh Dickins850465792012-01-20 14:34:19 -08001005 struct file *shm_file;
Lee Schermerhorn89e004ea2008-10-18 20:26:43 -07001006
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001007 rcu_read_lock();
1008 shp = shm_obtain_object_check(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -07001009 if (IS_ERR(shp)) {
1010 err = PTR_ERR(shp);
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001011 goto out_unlock1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013
Al Viroa33e6752008-12-10 03:40:06 -05001014 audit_ipc_obj(&(shp->shm_perm));
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001015 err = security_shm_shmctl(shp, cmd);
1016 if (err)
1017 goto out_unlock1;
Steve Grubb073115d2006-04-02 17:07:33 -04001018
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001019 ipc_lock_object(&shp->shm_perm);
Rafael Aquini0f3d2b02014-01-27 17:07:01 -08001020
1021 /* check if shm_destroy() is tearing down shp */
1022 if (!ipc_valid_object(&shp->shm_perm)) {
1023 err = -EIDRM;
1024 goto out_unlock0;
1025 }
1026
Serge E. Hallynb0e77592011-03-23 16:43:24 -07001027 if (!ns_capable(ns->user_ns, CAP_IPC_LOCK)) {
Eric W. Biederman1efdb692012-02-07 16:54:11 -08001028 kuid_t euid = current_euid();
Eric W. Biederman1efdb692012-02-07 16:54:11 -08001029 if (!uid_eq(euid, shp->shm_perm.uid) &&
Jesper Nilsson3a726602013-11-21 14:32:08 -08001030 !uid_eq(euid, shp->shm_perm.cuid)) {
1031 err = -EPERM;
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001032 goto out_unlock0;
Jesper Nilsson3a726602013-11-21 14:32:08 -08001033 }
1034 if (cmd == SHM_LOCK && !rlimit(RLIMIT_MEMLOCK)) {
1035 err = -EPERM;
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001036 goto out_unlock0;
Jesper Nilsson3a726602013-11-21 14:32:08 -08001037 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 }
1039
Hugh Dickins850465792012-01-20 14:34:19 -08001040 shm_file = shp->shm_file;
1041 if (is_file_hugepages(shm_file))
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001042 goto out_unlock0;
Hugh Dickins850465792012-01-20 14:34:19 -08001043
1044 if (cmd == SHM_LOCK) {
David Howells86a264a2008-11-14 10:39:18 +11001045 struct user_struct *user = current_user();
Hugh Dickins850465792012-01-20 14:34:19 -08001046 err = shmem_lock(shm_file, 1, user);
1047 if (!err && !(shp->shm_perm.mode & SHM_LOCKED)) {
1048 shp->shm_perm.mode |= SHM_LOCKED;
1049 shp->mlock_user = user;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 }
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001051 goto out_unlock0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 }
Hugh Dickins850465792012-01-20 14:34:19 -08001053
1054 /* SHM_UNLOCK */
1055 if (!(shp->shm_perm.mode & SHM_LOCKED))
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001056 goto out_unlock0;
Hugh Dickins850465792012-01-20 14:34:19 -08001057 shmem_lock(shm_file, 0, shp->mlock_user);
1058 shp->shm_perm.mode &= ~SHM_LOCKED;
1059 shp->mlock_user = NULL;
1060 get_file(shm_file);
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001061 ipc_unlock_object(&shp->shm_perm);
1062 rcu_read_unlock();
Hugh Dickins24513262012-01-20 14:34:21 -08001063 shmem_unlock_mapping(shm_file->f_mapping);
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001064
Hugh Dickins850465792012-01-20 14:34:19 -08001065 fput(shm_file);
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -07001066 return err;
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001067 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 default:
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -07001069 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 }
1071
Davidlohr Bueso2caacaa2013-09-11 14:26:21 -07001072out_unlock0:
1073 ipc_unlock_object(&shp->shm_perm);
1074out_unlock1:
1075 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 return err;
1077}
1078
1079/*
1080 * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists.
1081 *
1082 * NOTE! Despite the name, this is NOT a direct system call entrypoint. The
1083 * "raddr" thing points to kernel space, and there has to be a wrapper around
1084 * this.
1085 */
Will Deacon079a96a2012-07-30 14:42:38 -07001086long do_shmat(int shmid, char __user *shmaddr, int shmflg, ulong *raddr,
1087 unsigned long shmlba)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088{
1089 struct shmid_kernel *shp;
1090 unsigned long addr;
1091 unsigned long size;
Manfred Spraul239521f2014-01-27 17:07:04 -08001092 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093 int err;
1094 unsigned long flags;
1095 unsigned long prot;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 int acc_mode;
Kirill Korotaev4e982312006-10-02 02:18:22 -07001097 struct ipc_namespace *ns;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001098 struct shm_file_data *sfd;
1099 struct path path;
Al Viroaeb5d722008-09-02 15:28:45 -04001100 fmode_t f_mode;
Michel Lespinasse41badc12013-02-22 16:32:47 -08001101 unsigned long populate = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001103 err = -EINVAL;
1104 if (shmid < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 goto out;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001106 else if ((addr = (ulong)shmaddr)) {
Will Deacon079a96a2012-07-30 14:42:38 -07001107 if (addr & (shmlba - 1)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 if (shmflg & SHM_RND)
Will Deacon079a96a2012-07-30 14:42:38 -07001109 addr &= ~(shmlba - 1); /* round down */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 else
1111#ifndef __ARCH_FORCE_SHMLBA
1112 if (addr & ~PAGE_MASK)
1113#endif
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001114 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 }
1116 flags = MAP_SHARED | MAP_FIXED;
1117 } else {
1118 if ((shmflg & SHM_REMAP))
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001119 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120
1121 flags = MAP_SHARED;
1122 }
1123
1124 if (shmflg & SHM_RDONLY) {
1125 prot = PROT_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 acc_mode = S_IRUGO;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001127 f_mode = FMODE_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 } else {
1129 prot = PROT_READ | PROT_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 acc_mode = S_IRUGO | S_IWUGO;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001131 f_mode = FMODE_READ | FMODE_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 }
1133 if (shmflg & SHM_EXEC) {
1134 prot |= PROT_EXEC;
1135 acc_mode |= S_IXUGO;
1136 }
1137
1138 /*
1139 * We cannot rely on the fs check since SYSV IPC does have an
1140 * additional creator id...
1141 */
Kirill Korotaev4e982312006-10-02 02:18:22 -07001142 ns = current->nsproxy->ipc_ns;
Davidlohr Buesoc2c737a2013-09-11 14:26:23 -07001143 rcu_read_lock();
1144 shp = shm_obtain_object_check(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -07001145 if (IS_ERR(shp)) {
1146 err = PTR_ERR(shp);
Davidlohr Buesoc2c737a2013-09-11 14:26:23 -07001147 goto out_unlock;
Nadia Derbey023a5352007-10-18 23:40:51 -07001148 }
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001149
1150 err = -EACCES;
Serge E. Hallynb0e77592011-03-23 16:43:24 -07001151 if (ipcperms(ns, &shp->shm_perm, acc_mode))
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001152 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153
1154 err = security_shm_shmat(shp, shmaddr, shmflg);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001155 if (err)
1156 goto out_unlock;
1157
Davidlohr Buesoc2c737a2013-09-11 14:26:23 -07001158 ipc_lock_object(&shp->shm_perm);
Greg Thelena399b292013-11-21 14:32:00 -08001159
1160 /* check if shm_destroy() is tearing down shp */
Rafael Aquini0f3d2b02014-01-27 17:07:01 -08001161 if (!ipc_valid_object(&shp->shm_perm)) {
Greg Thelena399b292013-11-21 14:32:00 -08001162 ipc_unlock_object(&shp->shm_perm);
1163 err = -EIDRM;
1164 goto out_unlock;
1165 }
1166
Al Viro2c48b9c2009-08-09 00:52:35 +04001167 path = shp->shm_file->f_path;
1168 path_get(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 shp->shm_nattch++;
David Howells75c3cfa2015-03-17 22:26:12 +00001170 size = i_size_read(d_inode(path.dentry));
Davidlohr Buesoc2c737a2013-09-11 14:26:23 -07001171 ipc_unlock_object(&shp->shm_perm);
1172 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001174 err = -ENOMEM;
1175 sfd = kzalloc(sizeof(*sfd), GFP_KERNEL);
Davidlohr Buesof42569b2013-09-11 14:26:22 -07001176 if (!sfd) {
1177 path_put(&path);
1178 goto out_nattch;
1179 }
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001180
Al Viro2c48b9c2009-08-09 00:52:35 +04001181 file = alloc_file(&path, f_mode,
1182 is_file_hugepages(shp->shm_file) ?
Al Viroc4caa772009-11-30 08:38:43 -05001183 &shm_file_operations_huge :
1184 &shm_file_operations);
Anatol Pomozov39b65252012-09-12 20:11:55 -07001185 err = PTR_ERR(file);
Davidlohr Buesof42569b2013-09-11 14:26:22 -07001186 if (IS_ERR(file)) {
1187 kfree(sfd);
1188 path_put(&path);
1189 goto out_nattch;
1190 }
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001191
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001192 file->private_data = sfd;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001193 file->f_mapping = shp->shm_file->f_mapping;
Nadia Derbey7ca7e562007-10-18 23:40:48 -07001194 sfd->id = shp->shm_perm.id;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001195 sfd->ns = get_ipc_ns(ns);
1196 sfd->file = shp->shm_file;
1197 sfd->vm_ops = NULL;
1198
Al Viro8b3ec682012-05-30 17:11:23 -04001199 err = security_mmap_file(file, prot, flags);
1200 if (err)
1201 goto out_fput;
1202
Michal Hocko91f4f942016-05-23 16:25:51 -07001203 if (down_write_killable(&current->mm->mmap_sem)) {
1204 err = -EINTR;
1205 goto out_fput;
1206 }
1207
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208 if (addr && !(shmflg & SHM_REMAP)) {
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001209 err = -EINVAL;
Manfred Spraul247a8ce2014-06-06 14:37:38 -07001210 if (addr + size < addr)
1211 goto invalid;
1212
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 if (find_vma_intersection(current->mm, addr, addr + size))
1214 goto invalid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215 }
Davidlohr Buesof42569b2013-09-11 14:26:22 -07001216
Michel Lespinassebebeb3d2013-02-22 16:32:37 -08001217 addr = do_mmap_pgoff(file, addr, size, prot, flags, 0, &populate);
1218 *raddr = addr;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001219 err = 0;
Michel Lespinassebebeb3d2013-02-22 16:32:37 -08001220 if (IS_ERR_VALUE(addr))
1221 err = (long)addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222invalid:
1223 up_write(&current->mm->mmap_sem);
Michel Lespinassebebeb3d2013-02-22 16:32:37 -08001224 if (populate)
Michel Lespinasse41badc12013-02-22 16:32:47 -08001225 mm_populate(addr, populate);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226
Al Viro8b3ec682012-05-30 17:11:23 -04001227out_fput:
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001228 fput(file);
1229
1230out_nattch:
Davidlohr Buesod9a605e2013-09-11 14:26:24 -07001231 down_write(&shm_ids(ns).rwsem);
Nadia Derbey00c2bf82008-07-25 01:48:03 -07001232 shp = shm_lock(ns, shmid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 shp->shm_nattch--;
Vasiliy Kulikovb34a6b12011-07-26 16:08:48 -07001234 if (shm_may_destroy(ns, shp))
Kirill Korotaev4e982312006-10-02 02:18:22 -07001235 shm_destroy(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 else
1237 shm_unlock(shp);
Davidlohr Buesod9a605e2013-09-11 14:26:24 -07001238 up_write(&shm_ids(ns).rwsem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 return err;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001240
1241out_unlock:
Davidlohr Buesoc2c737a2013-09-11 14:26:23 -07001242 rcu_read_unlock();
Davidlohr Buesof42569b2013-09-11 14:26:22 -07001243out:
1244 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245}
1246
Heiko Carstensd5460c92009-01-14 14:14:27 +01001247SYSCALL_DEFINE3(shmat, int, shmid, char __user *, shmaddr, int, shmflg)
Stephen Rothwell7d87e14c2005-05-01 08:59:12 -07001248{
1249 unsigned long ret;
1250 long err;
1251
Will Deacon079a96a2012-07-30 14:42:38 -07001252 err = do_shmat(shmid, shmaddr, shmflg, &ret, SHMLBA);
Stephen Rothwell7d87e14c2005-05-01 08:59:12 -07001253 if (err)
1254 return err;
1255 force_successful_syscall_return();
1256 return (long)ret;
1257}
1258
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259/*
1260 * detach and kill segment if marked destroyed.
1261 * The work is done in shm_close.
1262 */
Heiko Carstensd5460c92009-01-14 14:14:27 +01001263SYSCALL_DEFINE1(shmdt, char __user *, shmaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264{
1265 struct mm_struct *mm = current->mm;
Mike Frysinger586c7e62009-06-09 16:26:23 -07001266 struct vm_area_struct *vma;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 unsigned long addr = (unsigned long)shmaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 int retval = -EINVAL;
Mike Frysinger586c7e62009-06-09 16:26:23 -07001269#ifdef CONFIG_MMU
1270 loff_t size = 0;
Dave Hansend3c97902014-12-12 16:58:19 -08001271 struct file *file;
Mike Frysinger586c7e62009-06-09 16:26:23 -07001272 struct vm_area_struct *next;
1273#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274
Hugh Dickinsdf1e2fb2006-03-24 03:18:06 -08001275 if (addr & ~PAGE_MASK)
1276 return retval;
1277
Michal Hocko91f4f942016-05-23 16:25:51 -07001278 if (down_write_killable(&mm->mmap_sem))
1279 return -EINTR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280
1281 /*
1282 * This function tries to be smart and unmap shm segments that
1283 * were modified by partial mlock or munmap calls:
1284 * - It first determines the size of the shm segment that should be
1285 * unmapped: It searches for a vma that is backed by shm and that
1286 * started at address shmaddr. It records it's size and then unmaps
1287 * it.
1288 * - Then it unmaps all shm vmas that started at shmaddr and that
Dave Hansend3c97902014-12-12 16:58:19 -08001289 * are within the initially determined size and that are from the
1290 * same shm segment from which we determined the size.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 * Errors from do_munmap are ignored: the function only fails if
1292 * it's called with invalid parameters or if it's called to unmap
1293 * a part of a vma. Both calls in this function are for full vmas,
1294 * the parameters are directly copied from the vma itself and always
1295 * valid - therefore do_munmap cannot fail. (famous last words?)
1296 */
1297 /*
1298 * If it had been mremap()'d, the starting address would not
1299 * match the usual checks anyway. So assume all vma's are
1300 * above the starting address given.
1301 */
1302 vma = find_vma(mm, addr);
1303
David Howells8feae132009-01-08 12:04:47 +00001304#ifdef CONFIG_MMU
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 while (vma) {
1306 next = vma->vm_next;
1307
1308 /*
1309 * Check if the starting address would match, i.e. it's
1310 * a fragment created by mprotect() and/or munmap(), or it
1311 * otherwise it starts at this address with no hassles.
1312 */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001313 if ((vma->vm_ops == &shm_vm_ops) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314 (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) {
1315
Dave Hansend3c97902014-12-12 16:58:19 -08001316 /*
1317 * Record the file of the shm segment being
1318 * unmapped. With mremap(), someone could place
1319 * page from another segment but with equal offsets
1320 * in the range we are unmapping.
1321 */
1322 file = vma->vm_file;
Dave Hansen07a46ed2014-12-12 16:58:22 -08001323 size = i_size_read(file_inode(vma->vm_file));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1325 /*
1326 * We discovered the size of the shm segment, so
1327 * break out of here and fall through to the next
1328 * loop that uses the size information to stop
1329 * searching for matching vma's.
1330 */
1331 retval = 0;
1332 vma = next;
1333 break;
1334 }
1335 vma = next;
1336 }
1337
1338 /*
1339 * We need look no further than the maximum address a fragment
1340 * could possibly have landed at. Also cast things to loff_t to
Lucas De Marchi25985ed2011-03-30 22:57:33 -03001341 * prevent overflows and make comparisons vs. equal-width types.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 */
KAMEZAWA Hiroyuki8e367092006-02-10 01:51:12 -08001343 size = PAGE_ALIGN(size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 while (vma && (loff_t)(vma->vm_end - addr) <= size) {
1345 next = vma->vm_next;
1346
1347 /* finding a matching vma now does not alter retval */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001348 if ((vma->vm_ops == &shm_vm_ops) &&
Dave Hansend3c97902014-12-12 16:58:19 -08001349 ((vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) &&
1350 (vma->vm_file == file))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1352 vma = next;
1353 }
1354
David Howells8feae132009-01-08 12:04:47 +00001355#else /* CONFIG_MMU */
1356 /* under NOMMU conditions, the exact address to be destroyed must be
1357 * given */
Davidlohr Bueso530fcd162013-09-11 14:26:28 -07001358 if (vma && vma->vm_start == addr && vma->vm_ops == &shm_vm_ops) {
David Howells8feae132009-01-08 12:04:47 +00001359 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1360 retval = 0;
1361 }
1362
1363#endif
1364
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 up_write(&mm->mmap_sem);
1366 return retval;
1367}
1368
1369#ifdef CONFIG_PROC_FS
Mike Waychison19b49462005-09-06 15:17:10 -07001370static int sysvipc_shm_proc_show(struct seq_file *s, void *it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371{
Eric W. Biederman1efdb692012-02-07 16:54:11 -08001372 struct user_namespace *user_ns = seq_user_ns(s);
Mike Waychison19b49462005-09-06 15:17:10 -07001373 struct shmid_kernel *shp = it;
Helge Dellerb7952182010-10-27 15:34:16 -07001374 unsigned long rss = 0, swp = 0;
1375
1376 shm_add_rss_swap(shp, &rss, &swp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377
Paul Menage6c826812008-06-12 15:21:49 -07001378#if BITS_PER_LONG <= 32
1379#define SIZE_SPEC "%10lu"
1380#else
1381#define SIZE_SPEC "%21lu"
1382#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383
Joe Perches7f032d62015-04-15 16:17:54 -07001384 seq_printf(s,
1385 "%10d %10d %4o " SIZE_SPEC " %5u %5u "
1386 "%5lu %5u %5u %5u %5u %10lu %10lu %10lu "
1387 SIZE_SPEC " " SIZE_SPEC "\n",
1388 shp->shm_perm.key,
1389 shp->shm_perm.id,
1390 shp->shm_perm.mode,
1391 shp->shm_segsz,
1392 shp->shm_cprid,
1393 shp->shm_lprid,
1394 shp->shm_nattch,
1395 from_kuid_munged(user_ns, shp->shm_perm.uid),
1396 from_kgid_munged(user_ns, shp->shm_perm.gid),
1397 from_kuid_munged(user_ns, shp->shm_perm.cuid),
1398 from_kgid_munged(user_ns, shp->shm_perm.cgid),
1399 shp->shm_atim,
1400 shp->shm_dtim,
1401 shp->shm_ctim,
1402 rss * PAGE_SIZE,
1403 swp * PAGE_SIZE);
1404
1405 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406}
1407#endif