blob: 7bc46a9fe1f8709cadae67f329f57b0943403832 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022 */
23
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/slab.h>
25#include <linux/mm.h>
26#include <linux/hugetlb.h>
27#include <linux/shm.h>
28#include <linux/init.h>
29#include <linux/file.h>
30#include <linux/mman.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <linux/shmem_fs.h>
32#include <linux/security.h>
33#include <linux/syscalls.h>
34#include <linux/audit.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080035#include <linux/capability.h>
Stephen Rothwell7d87e142005-05-01 08:59:12 -070036#include <linux/ptrace.h>
Mike Waychison19b49462005-09-06 15:17:10 -070037#include <linux/seq_file.h>
Nadia Derbey3e148c72007-10-18 23:40:54 -070038#include <linux/rwsem.h>
Kirill Korotaev4e982312006-10-02 02:18:22 -070039#include <linux/nsproxy.h>
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -080040#include <linux/mount.h>
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -080041#include <linux/ipc_namespace.h>
Stephen Rothwell7d87e142005-05-01 08:59:12 -070042
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#include <asm/uaccess.h>
44
45#include "util.h"
46
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -080047struct shm_file_data {
48 int id;
49 struct ipc_namespace *ns;
50 struct file *file;
51 const struct vm_operations_struct *vm_ops;
52};
53
54#define shm_file_data(file) (*((struct shm_file_data **)&(file)->private_data))
55
Arjan van de Ven9a321442007-02-12 00:55:35 -080056static const struct file_operations shm_file_operations;
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +040057static const struct vm_operations_struct shm_vm_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -070058
Pierre Peiffered2ddbf2008-02-08 04:18:57 -080059#define shm_ids(ns) ((ns)->ids[IPC_SHM_IDS])
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
Kirill Korotaev4e982312006-10-02 02:18:22 -070061#define shm_unlock(shp) \
62 ipc_unlock(&(shp)->shm_perm)
Kirill Korotaev4e982312006-10-02 02:18:22 -070063
Nadia Derbey7748dbf2007-10-18 23:40:49 -070064static int newseg(struct ipc_namespace *, struct ipc_params *);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -080065static void shm_open(struct vm_area_struct *vma);
66static void shm_close(struct vm_area_struct *vma);
Kirill Korotaev4e982312006-10-02 02:18:22 -070067static void shm_destroy (struct ipc_namespace *ns, struct shmid_kernel *shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#ifdef CONFIG_PROC_FS
Mike Waychison19b49462005-09-06 15:17:10 -070069static int sysvipc_shm_proc_show(struct seq_file *s, void *it);
Linus Torvalds1da177e2005-04-16 15:20:36 -070070#endif
71
Pierre Peiffered2ddbf2008-02-08 04:18:57 -080072void shm_init_ns(struct ipc_namespace *ns)
Kirill Korotaev4e982312006-10-02 02:18:22 -070073{
Kirill Korotaev4e982312006-10-02 02:18:22 -070074 ns->shm_ctlmax = SHMMAX;
75 ns->shm_ctlall = SHMALL;
76 ns->shm_ctlmni = SHMMNI;
77 ns->shm_tot = 0;
WANG Conge8148f72009-01-06 14:42:49 -080078 ipc_init_ids(&shm_ids(ns));
Kirill Korotaev4e982312006-10-02 02:18:22 -070079}
Linus Torvalds1da177e2005-04-16 15:20:36 -070080
Nadia Derbeyf4566f02007-10-18 23:40:53 -070081/*
Nadia Derbey3e148c72007-10-18 23:40:54 -070082 * Called with shm_ids.rw_mutex (writer) and the shp structure locked.
83 * Only shm_ids.rw_mutex remains locked on exit.
Nadia Derbeyf4566f02007-10-18 23:40:53 -070084 */
Pierre Peiffer01b8b072008-02-08 04:18:57 -080085static void do_shm_rmid(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp)
Kirill Korotaev4e982312006-10-02 02:18:22 -070086{
Pierre Peiffer01b8b072008-02-08 04:18:57 -080087 struct shmid_kernel *shp;
88 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
89
Kirill Korotaev4e982312006-10-02 02:18:22 -070090 if (shp->shm_nattch){
91 shp->shm_perm.mode |= SHM_DEST;
92 /* Do not find it any more */
93 shp->shm_perm.key = IPC_PRIVATE;
94 shm_unlock(shp);
95 } else
96 shm_destroy(ns, shp);
97}
98
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -080099#ifdef CONFIG_IPC_NS
Kirill Korotaev4e982312006-10-02 02:18:22 -0700100void shm_exit_ns(struct ipc_namespace *ns)
101{
Pierre Peiffer01b8b072008-02-08 04:18:57 -0800102 free_ipcs(ns, &shm_ids(ns), do_shm_rmid);
Serge E. Hallyn7d6feeb2009-12-15 16:47:27 -0800103 idr_destroy(&ns->ids[IPC_SHM_IDS].ipcs_idr);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700104}
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -0800105#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106
107void __init shm_init (void)
108{
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800109 shm_init_ns(&init_ipc_ns);
Mike Waychison19b49462005-09-06 15:17:10 -0700110 ipc_init_proc_interface("sysvipc/shm",
111 " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime\n",
Kirill Korotaev4e982312006-10-02 02:18:22 -0700112 IPC_SHM_IDS, sysvipc_shm_proc_show);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113}
114
Nadia Derbey3e148c72007-10-18 23:40:54 -0700115/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700116 * shm_lock_(check_) routines are called in the paths where the rw_mutex
Nadia Derbey00c2bf82008-07-25 01:48:03 -0700117 * is not necessarily held.
Nadia Derbey3e148c72007-10-18 23:40:54 -0700118 */
Nadia Derbey023a5352007-10-18 23:40:51 -0700119static inline struct shmid_kernel *shm_lock(struct ipc_namespace *ns, int id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700121 struct kern_ipc_perm *ipcp = ipc_lock(&shm_ids(ns), id);
122
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800123 if (IS_ERR(ipcp))
124 return (struct shmid_kernel *)ipcp;
125
Nadia Derbey03f02c72007-10-18 23:40:51 -0700126 return container_of(ipcp, struct shmid_kernel, shm_perm);
Nadia Derbey023a5352007-10-18 23:40:51 -0700127}
128
129static inline struct shmid_kernel *shm_lock_check(struct ipc_namespace *ns,
130 int id)
131{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700132 struct kern_ipc_perm *ipcp = ipc_lock_check(&shm_ids(ns), id);
133
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800134 if (IS_ERR(ipcp))
135 return (struct shmid_kernel *)ipcp;
136
Nadia Derbey03f02c72007-10-18 23:40:51 -0700137 return container_of(ipcp, struct shmid_kernel, shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138}
139
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700140static inline void shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *s)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700142 ipc_rmid(&shm_ids(ns), &s->shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143}
144
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800146/* This is called by fork, once for every shm attach. */
147static void shm_open(struct vm_area_struct *vma)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700148{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800149 struct file *file = vma->vm_file;
150 struct shm_file_data *sfd = shm_file_data(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 struct shmid_kernel *shp;
152
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800153 shp = shm_lock(sfd->ns, sfd->id);
Nadia Derbey023a5352007-10-18 23:40:51 -0700154 BUG_ON(IS_ERR(shp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 shp->shm_atim = get_seconds();
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700156 shp->shm_lprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 shp->shm_nattch++;
158 shm_unlock(shp);
159}
160
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161/*
162 * shm_destroy - free the struct shmid_kernel
163 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700164 * @ns: namespace
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 * @shp: struct to free
166 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700167 * It has to be called with shp and shm_ids.rw_mutex (writer) locked,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 * but returns with shp unlocked and freed.
169 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700170static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700172 ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700173 shm_rmid(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 shm_unlock(shp);
175 if (!is_file_hugepages(shp->shm_file))
176 shmem_lock(shp->shm_file, 0, shp->mlock_user);
Hugh Dickins353d5c32009-08-24 16:30:28 +0100177 else if (shp->mlock_user)
Josef Sipek6d630792006-12-08 02:37:11 -0800178 user_shm_unlock(shp->shm_file->f_path.dentry->d_inode->i_size,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 shp->mlock_user);
180 fput (shp->shm_file);
181 security_shm_free(shp);
182 ipc_rcu_putref(shp);
183}
184
185/*
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800186 * remove the attach descriptor vma.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 * free memory for segment if it is marked destroyed.
188 * The descriptor has already been removed from the current->mm->mmap list
189 * and will later be kfree()d.
190 */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800191static void shm_close(struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800193 struct file * file = vma->vm_file;
194 struct shm_file_data *sfd = shm_file_data(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 struct shmid_kernel *shp;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800196 struct ipc_namespace *ns = sfd->ns;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700197
Nadia Derbey3e148c72007-10-18 23:40:54 -0700198 down_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 /* remove from the list of attaches of the shm segment */
Nadia Derbey00c2bf82008-07-25 01:48:03 -0700200 shp = shm_lock(ns, sfd->id);
Nadia Derbey023a5352007-10-18 23:40:51 -0700201 BUG_ON(IS_ERR(shp));
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_dtim = get_seconds();
204 shp->shm_nattch--;
205 if(shp->shm_nattch == 0 &&
Andrew Mortonb33291c2006-01-08 01:02:21 -0800206 shp->shm_perm.mode & SHM_DEST)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700207 shm_destroy(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 else
209 shm_unlock(shp);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700210 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211}
212
Nick Piggind0217ac2007-07-19 01:47:03 -0700213static int shm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800214{
215 struct file *file = vma->vm_file;
216 struct shm_file_data *sfd = shm_file_data(file);
217
Nick Piggind0217ac2007-07-19 01:47:03 -0700218 return sfd->vm_ops->fault(vma, vmf);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800219}
220
221#ifdef CONFIG_NUMA
Adrian Bunkd823e3e2007-10-16 23:26:42 -0700222static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800223{
224 struct file *file = vma->vm_file;
225 struct shm_file_data *sfd = shm_file_data(file);
226 int err = 0;
227 if (sfd->vm_ops->set_policy)
228 err = sfd->vm_ops->set_policy(vma, new);
229 return err;
230}
231
Adrian Bunkd823e3e2007-10-16 23:26:42 -0700232static struct mempolicy *shm_get_policy(struct vm_area_struct *vma,
233 unsigned long addr)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800234{
235 struct file *file = vma->vm_file;
236 struct shm_file_data *sfd = shm_file_data(file);
237 struct mempolicy *pol = NULL;
238
239 if (sfd->vm_ops->get_policy)
240 pol = sfd->vm_ops->get_policy(vma, addr);
Lee Schermerhorn52cd3b02008-04-28 02:13:16 -0700241 else if (vma->vm_policy)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800242 pol = vma->vm_policy;
Lee Schermerhorn52cd3b02008-04-28 02:13:16 -0700243
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800244 return pol;
245}
246#endif
247
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248static int shm_mmap(struct file * file, struct vm_area_struct * vma)
249{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800250 struct shm_file_data *sfd = shm_file_data(file);
David Howellsb0e15192006-01-06 00:11:42 -0800251 int ret;
252
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800253 ret = sfd->file->f_op->mmap(sfd->file, vma);
254 if (ret != 0)
255 return ret;
256 sfd->vm_ops = vma->vm_ops;
David Howells2e92a3b2007-07-31 00:37:24 -0700257#ifdef CONFIG_MMU
Nick Piggin54cb8822007-07-19 01:46:59 -0700258 BUG_ON(!sfd->vm_ops->fault);
David Howells2e92a3b2007-07-31 00:37:24 -0700259#endif
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800260 vma->vm_ops = &shm_vm_ops;
261 shm_open(vma);
David Howellsb0e15192006-01-06 00:11:42 -0800262
263 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264}
265
Kirill Korotaev4e982312006-10-02 02:18:22 -0700266static int shm_release(struct inode *ino, struct file *file)
267{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800268 struct shm_file_data *sfd = shm_file_data(file);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700269
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800270 put_ipc_ns(sfd->ns);
271 shm_file_data(file) = NULL;
272 kfree(sfd);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700273 return 0;
274}
275
Christoph Hellwig7ea80852010-05-26 17:53:25 +0200276static int shm_fsync(struct file *file, int datasync)
Adam Litke516dffd2007-03-01 15:46:08 -0800277{
Adam Litke516dffd2007-03-01 15:46:08 -0800278 struct shm_file_data *sfd = shm_file_data(file);
Adam Litke516dffd2007-03-01 15:46:08 -0800279
Christoph Hellwig7ea80852010-05-26 17:53:25 +0200280 if (!sfd->file->f_op->fsync)
281 return -EINVAL;
282 return sfd->file->f_op->fsync(sfd->file, datasync);
Adam Litke516dffd2007-03-01 15:46:08 -0800283}
284
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800285static unsigned long shm_get_unmapped_area(struct file *file,
286 unsigned long addr, unsigned long len, unsigned long pgoff,
287 unsigned long flags)
288{
289 struct shm_file_data *sfd = shm_file_data(file);
Al Viroc4caa772009-11-30 08:38:43 -0500290 return sfd->file->f_op->get_unmapped_area(sfd->file, addr, len,
291 pgoff, flags);
Adam Litke516dffd2007-03-01 15:46:08 -0800292}
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800293
Arjan van de Ven9a321442007-02-12 00:55:35 -0800294static const struct file_operations shm_file_operations = {
Kirill Korotaev4e982312006-10-02 02:18:22 -0700295 .mmap = shm_mmap,
Adam Litke516dffd2007-03-01 15:46:08 -0800296 .fsync = shm_fsync,
Kirill Korotaev4e982312006-10-02 02:18:22 -0700297 .release = shm_release,
David Howellsed5e5892010-01-15 17:01:32 -0800298#ifndef CONFIG_MMU
299 .get_unmapped_area = shm_get_unmapped_area,
300#endif
Arnd Bergmann6038f372010-08-15 18:52:59 +0200301 .llseek = noop_llseek,
Al Viroc4caa772009-11-30 08:38:43 -0500302};
303
304static const struct file_operations shm_file_operations_huge = {
305 .mmap = shm_mmap,
306 .fsync = shm_fsync,
307 .release = shm_release,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800308 .get_unmapped_area = shm_get_unmapped_area,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200309 .llseek = noop_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310};
311
Al Viroc4caa772009-11-30 08:38:43 -0500312int is_file_shm_hugepages(struct file *file)
313{
314 return file->f_op == &shm_file_operations_huge;
315}
316
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +0400317static const struct vm_operations_struct shm_vm_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318 .open = shm_open, /* callback for a new vm-area open */
319 .close = shm_close, /* callback for when the vm-area is released */
Nick Piggin54cb8822007-07-19 01:46:59 -0700320 .fault = shm_fault,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800321#if defined(CONFIG_NUMA)
322 .set_policy = shm_set_policy,
323 .get_policy = shm_get_policy,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324#endif
325};
326
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700327/**
328 * newseg - Create a new shared memory segment
329 * @ns: namespace
330 * @params: ptr to the structure that contains key, size and shmflg
331 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700332 * Called with shm_ids.rw_mutex held as a writer.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700333 */
334
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700335static int newseg(struct ipc_namespace *ns, struct ipc_params *params)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336{
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700337 key_t key = params->key;
338 int shmflg = params->flg;
339 size_t size = params->u.size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 int error;
341 struct shmid_kernel *shp;
342 int numpages = (size + PAGE_SIZE -1) >> PAGE_SHIFT;
343 struct file * file;
344 char name[13];
345 int id;
Mel Gorman5a6fe122009-02-10 14:02:27 +0000346 int acctflag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347
Kirill Korotaev4e982312006-10-02 02:18:22 -0700348 if (size < SHMMIN || size > ns->shm_ctlmax)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 return -EINVAL;
350
Guy Streeterf66d45e2007-01-23 12:20:04 -0600351 if (ns->shm_tot + numpages > ns->shm_ctlall)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 return -ENOSPC;
353
354 shp = ipc_rcu_alloc(sizeof(*shp));
355 if (!shp)
356 return -ENOMEM;
357
358 shp->shm_perm.key = key;
Andrew Mortonb33291c2006-01-08 01:02:21 -0800359 shp->shm_perm.mode = (shmflg & S_IRWXUGO);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 shp->mlock_user = NULL;
361
362 shp->shm_perm.security = NULL;
363 error = security_shm_alloc(shp);
364 if (error) {
365 ipc_rcu_putref(shp);
366 return error;
367 }
368
Eric W. Biederman9d665862007-06-16 10:16:16 -0700369 sprintf (name, "SYSV%08x", key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 if (shmflg & SHM_HUGETLB) {
Mel Gorman5a6fe122009-02-10 14:02:27 +0000371 /* hugetlb_file_setup applies strict accounting */
372 if (shmflg & SHM_NORESERVE)
373 acctflag = VM_NORESERVE;
Hugh Dickins353d5c32009-08-24 16:30:28 +0100374 file = hugetlb_file_setup(name, size, acctflag,
Eric B Munson6bfde052009-09-21 17:03:43 -0700375 &shp->mlock_user, HUGETLB_SHMFS_INODE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 } else {
Badari Pulavartybf8f9722005-11-07 00:59:27 -0800377 /*
378 * Do not allow no accounting for OVERCOMMIT_NEVER, even
379 * if it's asked for.
380 */
381 if ((shmflg & SHM_NORESERVE) &&
382 sysctl_overcommit_memory != OVERCOMMIT_NEVER)
Linus Torvaldsfc8744a2009-01-31 15:08:56 -0800383 acctflag = VM_NORESERVE;
Badari Pulavartybf8f9722005-11-07 00:59:27 -0800384 file = shmem_file_setup(name, size, acctflag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 }
386 error = PTR_ERR(file);
387 if (IS_ERR(file))
388 goto no_file;
389
Pierre Peiffer48dea402008-04-29 01:00:35 -0700390 id = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni);
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700391 if (id < 0) {
392 error = id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 goto no_id;
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700394 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700396 shp->shm_cprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 shp->shm_lprid = 0;
398 shp->shm_atim = shp->shm_dtim = 0;
399 shp->shm_ctim = get_seconds();
400 shp->shm_segsz = size;
401 shp->shm_nattch = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 shp->shm_file = file;
Badari Pulavarty30475cc2007-06-16 10:15:59 -0700403 /*
404 * shmid gets reported as "inode#" in /proc/pid/maps.
405 * proc-ps tools use this. Changing this will break them.
406 */
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700407 file->f_dentry->d_inode->i_ino = shp->shm_perm.id;
Krishnakumar R551110a2005-10-29 18:16:45 -0700408
Kirill Korotaev4e982312006-10-02 02:18:22 -0700409 ns->shm_tot += numpages;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700410 error = shp->shm_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411 shm_unlock(shp);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700412 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413
414no_id:
Hugh Dickins2195d282009-09-12 12:21:27 +0100415 if (is_file_hugepages(file) && shp->mlock_user)
Hugh Dickins353d5c32009-08-24 16:30:28 +0100416 user_shm_unlock(size, shp->mlock_user);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 fput(file);
418no_file:
419 security_shm_free(shp);
420 ipc_rcu_putref(shp);
421 return error;
422}
423
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700424/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700425 * Called with shm_ids.rw_mutex and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700426 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700427static inline int shm_security(struct kern_ipc_perm *ipcp, int shmflg)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700428{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700429 struct shmid_kernel *shp;
430
431 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
432 return security_shm_associate(shp, shmflg);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700433}
434
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700435/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700436 * Called with shm_ids.rw_mutex and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700437 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700438static inline int shm_more_checks(struct kern_ipc_perm *ipcp,
439 struct ipc_params *params)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700440{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700441 struct shmid_kernel *shp;
442
443 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
444 if (shp->shm_segsz < params->u.size)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700445 return -EINVAL;
446
447 return 0;
448}
449
Heiko Carstensd5460c92009-01-14 14:14:27 +0100450SYSCALL_DEFINE3(shmget, key_t, key, size_t, size, int, shmflg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700452 struct ipc_namespace *ns;
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700453 struct ipc_ops shm_ops;
454 struct ipc_params shm_params;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455
Kirill Korotaev4e982312006-10-02 02:18:22 -0700456 ns = current->nsproxy->ipc_ns;
457
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700458 shm_ops.getnew = newseg;
459 shm_ops.associate = shm_security;
460 shm_ops.more_checks = shm_more_checks;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700461
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700462 shm_params.key = key;
463 shm_params.flg = shmflg;
464 shm_params.u.size = size;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700465
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700466 return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467}
468
469static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version)
470{
471 switch(version) {
472 case IPC_64:
473 return copy_to_user(buf, in, sizeof(*in));
474 case IPC_OLD:
475 {
476 struct shmid_ds out;
477
478 ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm);
479 out.shm_segsz = in->shm_segsz;
480 out.shm_atime = in->shm_atime;
481 out.shm_dtime = in->shm_dtime;
482 out.shm_ctime = in->shm_ctime;
483 out.shm_cpid = in->shm_cpid;
484 out.shm_lpid = in->shm_lpid;
485 out.shm_nattch = in->shm_nattch;
486
487 return copy_to_user(buf, &out, sizeof(out));
488 }
489 default:
490 return -EINVAL;
491 }
492}
493
Pierre Peiffer016d7132008-04-29 01:00:50 -0700494static inline unsigned long
495copy_shmid_from_user(struct shmid64_ds *out, void __user *buf, int version)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496{
497 switch(version) {
498 case IPC_64:
Pierre Peiffer016d7132008-04-29 01:00:50 -0700499 if (copy_from_user(out, buf, sizeof(*out)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 case IPC_OLD:
503 {
504 struct shmid_ds tbuf_old;
505
506 if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old)))
507 return -EFAULT;
508
Pierre Peiffer016d7132008-04-29 01:00:50 -0700509 out->shm_perm.uid = tbuf_old.shm_perm.uid;
510 out->shm_perm.gid = tbuf_old.shm_perm.gid;
511 out->shm_perm.mode = tbuf_old.shm_perm.mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512
513 return 0;
514 }
515 default:
516 return -EINVAL;
517 }
518}
519
520static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version)
521{
522 switch(version) {
523 case IPC_64:
524 return copy_to_user(buf, in, sizeof(*in));
525 case IPC_OLD:
526 {
527 struct shminfo out;
528
529 if(in->shmmax > INT_MAX)
530 out.shmmax = INT_MAX;
531 else
532 out.shmmax = (int)in->shmmax;
533
534 out.shmmin = in->shmmin;
535 out.shmmni = in->shmmni;
536 out.shmseg = in->shmseg;
537 out.shmall = in->shmall;
538
539 return copy_to_user(buf, &out, sizeof(out));
540 }
541 default:
542 return -EINVAL;
543 }
544}
545
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700546/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700547 * Called with shm_ids.rw_mutex held as a reader
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700548 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700549static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss,
550 unsigned long *swp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700552 int next_id;
553 int total, in_use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554
555 *rss = 0;
556 *swp = 0;
557
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700558 in_use = shm_ids(ns).in_use;
559
560 for (total = 0, next_id = 0; total < in_use; next_id++) {
Tony Battersbye562aeb2009-04-02 16:58:26 -0700561 struct kern_ipc_perm *ipc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 struct shmid_kernel *shp;
563 struct inode *inode;
564
Tony Battersbye562aeb2009-04-02 16:58:26 -0700565 ipc = idr_find(&shm_ids(ns).ipcs_idr, next_id);
566 if (ipc == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 continue;
Tony Battersbye562aeb2009-04-02 16:58:26 -0700568 shp = container_of(ipc, struct shmid_kernel, shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569
Josef Sipek6d630792006-12-08 02:37:11 -0800570 inode = shp->shm_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571
572 if (is_file_hugepages(shp->shm_file)) {
573 struct address_space *mapping = inode->i_mapping;
Andi Kleena5516432008-07-23 21:27:41 -0700574 struct hstate *h = hstate_file(shp->shm_file);
575 *rss += pages_per_huge_page(h) * mapping->nrpages;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 } else {
Tony Battersbya68e61e2009-02-04 15:12:04 -0800577#ifdef CONFIG_SHMEM
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 struct shmem_inode_info *info = SHMEM_I(inode);
579 spin_lock(&info->lock);
580 *rss += inode->i_mapping->nrpages;
581 *swp += info->swapped;
582 spin_unlock(&info->lock);
Tony Battersbya68e61e2009-02-04 15:12:04 -0800583#else
584 *rss += inode->i_mapping->nrpages;
585#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700587
588 total++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 }
590}
591
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700592/*
593 * This function handles some shmctl commands which require the rw_mutex
594 * to be held in write mode.
595 * NOTE: no locks must be held, the rw_mutex is taken inside this function.
596 */
597static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd,
598 struct shmid_ds __user *buf, int version)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599{
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700600 struct kern_ipc_perm *ipcp;
Pierre Peiffer016d7132008-04-29 01:00:50 -0700601 struct shmid64_ds shmid64;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 struct shmid_kernel *shp;
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700603 int err;
604
605 if (cmd == IPC_SET) {
Pierre Peiffer016d7132008-04-29 01:00:50 -0700606 if (copy_shmid_from_user(&shmid64, buf, version))
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700607 return -EFAULT;
608 }
609
Pierre Peiffera5f75e72008-04-29 01:00:54 -0700610 ipcp = ipcctl_pre_down(&shm_ids(ns), shmid, cmd, &shmid64.shm_perm, 0);
611 if (IS_ERR(ipcp))
612 return PTR_ERR(ipcp);
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700613
Pierre Peiffera5f75e72008-04-29 01:00:54 -0700614 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700615
616 err = security_shm_shmctl(shp, cmd);
617 if (err)
618 goto out_unlock;
619 switch (cmd) {
620 case IPC_RMID:
621 do_shm_rmid(ns, ipcp);
622 goto out_up;
623 case IPC_SET:
Pierre Peiffer8f4a3802008-04-29 01:00:51 -0700624 ipc_update_perm(&shmid64.shm_perm, ipcp);
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700625 shp->shm_ctim = get_seconds();
626 break;
627 default:
628 err = -EINVAL;
629 }
630out_unlock:
631 shm_unlock(shp);
632out_up:
633 up_write(&shm_ids(ns).rw_mutex);
634 return err;
635}
636
Heiko Carstensd5460c92009-01-14 14:14:27 +0100637SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, struct shmid_ds __user *, buf)
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700638{
639 struct shmid_kernel *shp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 int err, version;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700641 struct ipc_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642
643 if (cmd < 0 || shmid < 0) {
644 err = -EINVAL;
645 goto out;
646 }
647
648 version = ipc_parse_version(&cmd);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700649 ns = current->nsproxy->ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
651 switch (cmd) { /* replace with proc interface ? */
652 case IPC_INFO:
653 {
654 struct shminfo64 shminfo;
655
656 err = security_shm_shmctl(NULL, cmd);
657 if (err)
658 return err;
659
WANG Conge8148f72009-01-06 14:42:49 -0800660 memset(&shminfo, 0, sizeof(shminfo));
Kirill Korotaev4e982312006-10-02 02:18:22 -0700661 shminfo.shmmni = shminfo.shmseg = ns->shm_ctlmni;
662 shminfo.shmmax = ns->shm_ctlmax;
663 shminfo.shmall = ns->shm_ctlall;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664
665 shminfo.shmmin = SHMMIN;
666 if(copy_shminfo_to_user (buf, &shminfo, version))
667 return -EFAULT;
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700668
Nadia Derbey3e148c72007-10-18 23:40:54 -0700669 down_read(&shm_ids(ns).rw_mutex);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700670 err = ipc_get_maxid(&shm_ids(ns));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700671 up_read(&shm_ids(ns).rw_mutex);
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700672
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 if(err<0)
674 err = 0;
675 goto out;
676 }
677 case SHM_INFO:
678 {
679 struct shm_info shm_info;
680
681 err = security_shm_shmctl(NULL, cmd);
682 if (err)
683 return err;
684
WANG Conge8148f72009-01-06 14:42:49 -0800685 memset(&shm_info, 0, sizeof(shm_info));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700686 down_read(&shm_ids(ns).rw_mutex);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700687 shm_info.used_ids = shm_ids(ns).in_use;
688 shm_get_stat (ns, &shm_info.shm_rss, &shm_info.shm_swp);
689 shm_info.shm_tot = ns->shm_tot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690 shm_info.swap_attempts = 0;
691 shm_info.swap_successes = 0;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700692 err = ipc_get_maxid(&shm_ids(ns));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700693 up_read(&shm_ids(ns).rw_mutex);
WANG Conge8148f72009-01-06 14:42:49 -0800694 if (copy_to_user(buf, &shm_info, sizeof(shm_info))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 err = -EFAULT;
696 goto out;
697 }
698
699 err = err < 0 ? 0 : err;
700 goto out;
701 }
702 case SHM_STAT:
703 case IPC_STAT:
704 {
705 struct shmid64_ds tbuf;
706 int result;
Nadia Derbey023a5352007-10-18 23:40:51 -0700707
Nadia Derbey023a5352007-10-18 23:40:51 -0700708 if (cmd == SHM_STAT) {
709 shp = shm_lock(ns, shmid);
710 if (IS_ERR(shp)) {
711 err = PTR_ERR(shp);
712 goto out;
713 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700714 result = shp->shm_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 } else {
Nadia Derbey023a5352007-10-18 23:40:51 -0700716 shp = shm_lock_check(ns, shmid);
717 if (IS_ERR(shp)) {
718 err = PTR_ERR(shp);
719 goto out;
720 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 result = 0;
722 }
WANG Conge8148f72009-01-06 14:42:49 -0800723 err = -EACCES;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 if (ipcperms (&shp->shm_perm, S_IRUGO))
725 goto out_unlock;
726 err = security_shm_shmctl(shp, cmd);
727 if (err)
728 goto out_unlock;
Nadia Derbey023a5352007-10-18 23:40:51 -0700729 memset(&tbuf, 0, sizeof(tbuf));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 kernel_to_ipc64_perm(&shp->shm_perm, &tbuf.shm_perm);
731 tbuf.shm_segsz = shp->shm_segsz;
732 tbuf.shm_atime = shp->shm_atim;
733 tbuf.shm_dtime = shp->shm_dtim;
734 tbuf.shm_ctime = shp->shm_ctim;
735 tbuf.shm_cpid = shp->shm_cprid;
736 tbuf.shm_lpid = shp->shm_lprid;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800737 tbuf.shm_nattch = shp->shm_nattch;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 shm_unlock(shp);
739 if(copy_shmid_to_user (buf, &tbuf, version))
740 err = -EFAULT;
741 else
742 err = result;
743 goto out;
744 }
745 case SHM_LOCK:
746 case SHM_UNLOCK:
747 {
Lee Schermerhorn89e004ea2008-10-18 20:26:43 -0700748 struct file *uninitialized_var(shm_file);
749
750 lru_add_drain_all(); /* drain pagevecs to lru lists */
751
Nadia Derbey023a5352007-10-18 23:40:51 -0700752 shp = shm_lock_check(ns, shmid);
753 if (IS_ERR(shp)) {
754 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 goto out;
756 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757
Al Viroa33e6752008-12-10 03:40:06 -0500758 audit_ipc_obj(&(shp->shm_perm));
Steve Grubb073115d2006-04-02 17:07:33 -0400759
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 if (!capable(CAP_IPC_LOCK)) {
David Howells414c0702008-11-14 10:39:06 +1100761 uid_t euid = current_euid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 err = -EPERM;
David Howells414c0702008-11-14 10:39:06 +1100763 if (euid != shp->shm_perm.uid &&
764 euid != shp->shm_perm.cuid)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 goto out_unlock;
Jiri Slabyf1eb1332010-03-10 15:23:05 -0800766 if (cmd == SHM_LOCK && !rlimit(RLIMIT_MEMLOCK))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 goto out_unlock;
768 }
769
770 err = security_shm_shmctl(shp, cmd);
771 if (err)
772 goto out_unlock;
773
774 if(cmd==SHM_LOCK) {
David Howells86a264a2008-11-14 10:39:18 +1100775 struct user_struct *user = current_user();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 if (!is_file_hugepages(shp->shm_file)) {
777 err = shmem_lock(shp->shm_file, 1, user);
Pavel Emelianov7be77e22007-07-31 00:38:48 -0700778 if (!err && !(shp->shm_perm.mode & SHM_LOCKED)){
Andrew Mortonb33291c2006-01-08 01:02:21 -0800779 shp->shm_perm.mode |= SHM_LOCKED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 shp->mlock_user = user;
781 }
782 }
783 } else if (!is_file_hugepages(shp->shm_file)) {
784 shmem_lock(shp->shm_file, 0, shp->mlock_user);
Andrew Mortonb33291c2006-01-08 01:02:21 -0800785 shp->shm_perm.mode &= ~SHM_LOCKED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 shp->mlock_user = NULL;
787 }
788 shm_unlock(shp);
789 goto out;
790 }
791 case IPC_RMID:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 case IPC_SET:
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700793 err = shmctl_down(ns, shmid, cmd, buf, version);
794 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 default:
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700796 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 }
798
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799out_unlock:
800 shm_unlock(shp);
801out:
802 return err;
803}
804
805/*
806 * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists.
807 *
808 * NOTE! Despite the name, this is NOT a direct system call entrypoint. The
809 * "raddr" thing points to kernel space, and there has to be a wrapper around
810 * this.
811 */
812long do_shmat(int shmid, char __user *shmaddr, int shmflg, ulong *raddr)
813{
814 struct shmid_kernel *shp;
815 unsigned long addr;
816 unsigned long size;
817 struct file * file;
818 int err;
819 unsigned long flags;
820 unsigned long prot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 int acc_mode;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800822 unsigned long user_addr;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700823 struct ipc_namespace *ns;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800824 struct shm_file_data *sfd;
825 struct path path;
Al Viroaeb5d722008-09-02 15:28:45 -0400826 fmode_t f_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800828 err = -EINVAL;
829 if (shmid < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 goto out;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800831 else if ((addr = (ulong)shmaddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832 if (addr & (SHMLBA-1)) {
833 if (shmflg & SHM_RND)
834 addr &= ~(SHMLBA-1); /* round down */
835 else
836#ifndef __ARCH_FORCE_SHMLBA
837 if (addr & ~PAGE_MASK)
838#endif
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800839 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 }
841 flags = MAP_SHARED | MAP_FIXED;
842 } else {
843 if ((shmflg & SHM_REMAP))
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800844 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845
846 flags = MAP_SHARED;
847 }
848
849 if (shmflg & SHM_RDONLY) {
850 prot = PROT_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 acc_mode = S_IRUGO;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800852 f_mode = FMODE_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 } else {
854 prot = PROT_READ | PROT_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 acc_mode = S_IRUGO | S_IWUGO;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800856 f_mode = FMODE_READ | FMODE_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 }
858 if (shmflg & SHM_EXEC) {
859 prot |= PROT_EXEC;
860 acc_mode |= S_IXUGO;
861 }
862
863 /*
864 * We cannot rely on the fs check since SYSV IPC does have an
865 * additional creator id...
866 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700867 ns = current->nsproxy->ipc_ns;
Nadia Derbey023a5352007-10-18 23:40:51 -0700868 shp = shm_lock_check(ns, shmid);
869 if (IS_ERR(shp)) {
870 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 goto out;
Nadia Derbey023a5352007-10-18 23:40:51 -0700872 }
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800873
874 err = -EACCES;
875 if (ipcperms(&shp->shm_perm, acc_mode))
876 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877
878 err = security_shm_shmat(shp, shmaddr, shmflg);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800879 if (err)
880 goto out_unlock;
881
Al Viro2c48b9c2009-08-09 00:52:35 +0400882 path = shp->shm_file->f_path;
883 path_get(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 shp->shm_nattch++;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800885 size = i_size_read(path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 shm_unlock(shp);
887
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800888 err = -ENOMEM;
889 sfd = kzalloc(sizeof(*sfd), GFP_KERNEL);
890 if (!sfd)
Dave Hansence8d2cd2007-10-16 23:31:13 -0700891 goto out_put_dentry;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800892
Al Viro2c48b9c2009-08-09 00:52:35 +0400893 file = alloc_file(&path, f_mode,
894 is_file_hugepages(shp->shm_file) ?
Al Viroc4caa772009-11-30 08:38:43 -0500895 &shm_file_operations_huge :
896 &shm_file_operations);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800897 if (!file)
898 goto out_free;
899
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800900 file->private_data = sfd;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800901 file->f_mapping = shp->shm_file->f_mapping;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700902 sfd->id = shp->shm_perm.id;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800903 sfd->ns = get_ipc_ns(ns);
904 sfd->file = shp->shm_file;
905 sfd->vm_ops = NULL;
906
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 down_write(&current->mm->mmap_sem);
908 if (addr && !(shmflg & SHM_REMAP)) {
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800909 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 if (find_vma_intersection(current->mm, addr, addr + size))
911 goto invalid;
912 /*
913 * If shm segment goes below stack, make sure there is some
914 * space left for the stack to grow (at least 4 pages).
915 */
916 if (addr < current->mm->start_stack &&
917 addr > current->mm->start_stack - size - PAGE_SIZE * 5)
918 goto invalid;
919 }
920
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800921 user_addr = do_mmap (file, addr, size, prot, flags, 0);
922 *raddr = user_addr;
923 err = 0;
924 if (IS_ERR_VALUE(user_addr))
925 err = (long)user_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926invalid:
927 up_write(&current->mm->mmap_sem);
928
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800929 fput(file);
930
931out_nattch:
Nadia Derbey3e148c72007-10-18 23:40:54 -0700932 down_write(&shm_ids(ns).rw_mutex);
Nadia Derbey00c2bf82008-07-25 01:48:03 -0700933 shp = shm_lock(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -0700934 BUG_ON(IS_ERR(shp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935 shp->shm_nattch--;
936 if(shp->shm_nattch == 0 &&
Andrew Mortonb33291c2006-01-08 01:02:21 -0800937 shp->shm_perm.mode & SHM_DEST)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700938 shm_destroy(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 else
940 shm_unlock(shp);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700941 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943out:
944 return err;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800945
946out_unlock:
947 shm_unlock(shp);
948 goto out;
949
950out_free:
951 kfree(sfd);
Dave Hansence8d2cd2007-10-16 23:31:13 -0700952out_put_dentry:
Al Viro2c48b9c2009-08-09 00:52:35 +0400953 path_put(&path);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800954 goto out_nattch;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955}
956
Heiko Carstensd5460c92009-01-14 14:14:27 +0100957SYSCALL_DEFINE3(shmat, int, shmid, char __user *, shmaddr, int, shmflg)
Stephen Rothwell7d87e142005-05-01 08:59:12 -0700958{
959 unsigned long ret;
960 long err;
961
962 err = do_shmat(shmid, shmaddr, shmflg, &ret);
963 if (err)
964 return err;
965 force_successful_syscall_return();
966 return (long)ret;
967}
968
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969/*
970 * detach and kill segment if marked destroyed.
971 * The work is done in shm_close.
972 */
Heiko Carstensd5460c92009-01-14 14:14:27 +0100973SYSCALL_DEFINE1(shmdt, char __user *, shmaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974{
975 struct mm_struct *mm = current->mm;
Mike Frysinger586c7e62009-06-09 16:26:23 -0700976 struct vm_area_struct *vma;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 unsigned long addr = (unsigned long)shmaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 int retval = -EINVAL;
Mike Frysinger586c7e62009-06-09 16:26:23 -0700979#ifdef CONFIG_MMU
980 loff_t size = 0;
981 struct vm_area_struct *next;
982#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983
Hugh Dickinsdf1e2fb2006-03-24 03:18:06 -0800984 if (addr & ~PAGE_MASK)
985 return retval;
986
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 down_write(&mm->mmap_sem);
988
989 /*
990 * This function tries to be smart and unmap shm segments that
991 * were modified by partial mlock or munmap calls:
992 * - It first determines the size of the shm segment that should be
993 * unmapped: It searches for a vma that is backed by shm and that
994 * started at address shmaddr. It records it's size and then unmaps
995 * it.
996 * - Then it unmaps all shm vmas that started at shmaddr and that
997 * are within the initially determined size.
998 * Errors from do_munmap are ignored: the function only fails if
999 * it's called with invalid parameters or if it's called to unmap
1000 * a part of a vma. Both calls in this function are for full vmas,
1001 * the parameters are directly copied from the vma itself and always
1002 * valid - therefore do_munmap cannot fail. (famous last words?)
1003 */
1004 /*
1005 * If it had been mremap()'d, the starting address would not
1006 * match the usual checks anyway. So assume all vma's are
1007 * above the starting address given.
1008 */
1009 vma = find_vma(mm, addr);
1010
David Howells8feae132009-01-08 12:04:47 +00001011#ifdef CONFIG_MMU
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 while (vma) {
1013 next = vma->vm_next;
1014
1015 /*
1016 * Check if the starting address would match, i.e. it's
1017 * a fragment created by mprotect() and/or munmap(), or it
1018 * otherwise it starts at this address with no hassles.
1019 */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001020 if ((vma->vm_ops == &shm_vm_ops) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) {
1022
1023
Josef Sipek6d630792006-12-08 02:37:11 -08001024 size = vma->vm_file->f_path.dentry->d_inode->i_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1026 /*
1027 * We discovered the size of the shm segment, so
1028 * break out of here and fall through to the next
1029 * loop that uses the size information to stop
1030 * searching for matching vma's.
1031 */
1032 retval = 0;
1033 vma = next;
1034 break;
1035 }
1036 vma = next;
1037 }
1038
1039 /*
1040 * We need look no further than the maximum address a fragment
1041 * could possibly have landed at. Also cast things to loff_t to
1042 * prevent overflows and make comparisions vs. equal-width types.
1043 */
KAMEZAWA Hiroyuki8e367092006-02-10 01:51:12 -08001044 size = PAGE_ALIGN(size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 while (vma && (loff_t)(vma->vm_end - addr) <= size) {
1046 next = vma->vm_next;
1047
1048 /* finding a matching vma now does not alter retval */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001049 if ((vma->vm_ops == &shm_vm_ops) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff)
1051
1052 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1053 vma = next;
1054 }
1055
David Howells8feae132009-01-08 12:04:47 +00001056#else /* CONFIG_MMU */
1057 /* under NOMMU conditions, the exact address to be destroyed must be
1058 * given */
1059 retval = -EINVAL;
1060 if (vma->vm_start == addr && vma->vm_ops == &shm_vm_ops) {
1061 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1062 retval = 0;
1063 }
1064
1065#endif
1066
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 up_write(&mm->mmap_sem);
1068 return retval;
1069}
1070
1071#ifdef CONFIG_PROC_FS
Mike Waychison19b49462005-09-06 15:17:10 -07001072static int sysvipc_shm_proc_show(struct seq_file *s, void *it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073{
Mike Waychison19b49462005-09-06 15:17:10 -07001074 struct shmid_kernel *shp = it;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075
Paul Menage6c826812008-06-12 15:21:49 -07001076#if BITS_PER_LONG <= 32
1077#define SIZE_SPEC "%10lu"
1078#else
1079#define SIZE_SPEC "%21lu"
1080#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081
Paul Menage6c826812008-06-12 15:21:49 -07001082 return seq_printf(s,
1083 "%10d %10d %4o " SIZE_SPEC " %5u %5u "
1084 "%5lu %5u %5u %5u %5u %10lu %10lu %10lu\n",
Mike Waychison19b49462005-09-06 15:17:10 -07001085 shp->shm_perm.key,
Nadia Derbey7ca7e562007-10-18 23:40:48 -07001086 shp->shm_perm.id,
Andrew Mortonb33291c2006-01-08 01:02:21 -08001087 shp->shm_perm.mode,
Mike Waychison19b49462005-09-06 15:17:10 -07001088 shp->shm_segsz,
1089 shp->shm_cprid,
1090 shp->shm_lprid,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001091 shp->shm_nattch,
Mike Waychison19b49462005-09-06 15:17:10 -07001092 shp->shm_perm.uid,
1093 shp->shm_perm.gid,
1094 shp->shm_perm.cuid,
1095 shp->shm_perm.cgid,
1096 shp->shm_atim,
1097 shp->shm_dtim,
1098 shp->shm_ctim);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099}
1100#endif