blob: 3818fae625c5252363380fa9c7a521e8b5c1a7d1 [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>
Stephen Rothwell7d87e142005-05-01 08:59:12 -070041
Linus Torvalds1da177e2005-04-16 15:20:36 -070042#include <asm/uaccess.h>
43
44#include "util.h"
45
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -080046struct shm_file_data {
47 int id;
48 struct ipc_namespace *ns;
49 struct file *file;
50 const struct vm_operations_struct *vm_ops;
51};
52
53#define shm_file_data(file) (*((struct shm_file_data **)&(file)->private_data))
54
Arjan van de Ven9a321442007-02-12 00:55:35 -080055static const struct file_operations shm_file_operations;
Linus Torvalds1da177e2005-04-16 15:20:36 -070056static struct vm_operations_struct shm_vm_ops;
57
Kirill Korotaev4e982312006-10-02 02:18:22 -070058static struct ipc_ids init_shm_ids;
Linus Torvalds1da177e2005-04-16 15:20:36 -070059
Kirill Korotaev4e982312006-10-02 02:18:22 -070060#define shm_ids(ns) (*((ns)->ids[IPC_SHM_IDS]))
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
Kirill Korotaev4e982312006-10-02 02:18:22 -070062#define shm_unlock(shp) \
63 ipc_unlock(&(shp)->shm_perm)
Nadia Derbey1b531f22007-10-18 23:40:55 -070064#define shm_buildid(id, seq) ipc_buildid(id, seq)
Kirill Korotaev4e982312006-10-02 02:18:22 -070065
Nadia Derbey7748dbf2007-10-18 23:40:49 -070066static int newseg(struct ipc_namespace *, struct ipc_params *);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -080067static void shm_open(struct vm_area_struct *vma);
68static void shm_close(struct vm_area_struct *vma);
Kirill Korotaev4e982312006-10-02 02:18:22 -070069static void shm_destroy (struct ipc_namespace *ns, struct shmid_kernel *shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -070070#ifdef CONFIG_PROC_FS
Mike Waychison19b49462005-09-06 15:17:10 -070071static int sysvipc_shm_proc_show(struct seq_file *s, void *it);
Linus Torvalds1da177e2005-04-16 15:20:36 -070072#endif
73
Cedric Le Goater7d69a1f2007-07-15 23:40:58 -070074static void __shm_init_ns(struct ipc_namespace *ns, struct ipc_ids *ids)
Kirill Korotaev4e982312006-10-02 02:18:22 -070075{
76 ns->ids[IPC_SHM_IDS] = ids;
77 ns->shm_ctlmax = SHMMAX;
78 ns->shm_ctlall = SHMALL;
79 ns->shm_ctlmni = SHMMNI;
80 ns->shm_tot = 0;
Nadia Derbey7ca7e562007-10-18 23:40:48 -070081 ipc_init_ids(ids);
Kirill Korotaev4e982312006-10-02 02:18:22 -070082}
Linus Torvalds1da177e2005-04-16 15:20:36 -070083
Nadia Derbeyf4566f02007-10-18 23:40:53 -070084/*
Nadia Derbey3e148c72007-10-18 23:40:54 -070085 * Called with shm_ids.rw_mutex (writer) and the shp structure locked.
86 * Only shm_ids.rw_mutex remains locked on exit.
Nadia Derbeyf4566f02007-10-18 23:40:53 -070087 */
Kirill Korotaev4e982312006-10-02 02:18:22 -070088static void do_shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *shp)
89{
90 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
Kirill Korotaev4e982312006-10-02 02:18:22 -070099int shm_init_ns(struct ipc_namespace *ns)
100{
101 struct ipc_ids *ids;
102
103 ids = kmalloc(sizeof(struct ipc_ids), GFP_KERNEL);
104 if (ids == NULL)
105 return -ENOMEM;
106
107 __shm_init_ns(ns, ids);
108 return 0;
109}
110
111void shm_exit_ns(struct ipc_namespace *ns)
112{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700113 struct shmid_kernel *shp;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700114 int next_id;
115 int total, in_use;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700116
Nadia Derbey3e148c72007-10-18 23:40:54 -0700117 down_write(&shm_ids(ns).rw_mutex);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700118
119 in_use = shm_ids(ns).in_use;
120
121 for (total = 0, next_id = 0; total < in_use; next_id++) {
122 shp = idr_find(&shm_ids(ns).ipcs_idr, next_id);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700123 if (shp == NULL)
124 continue;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700125 ipc_lock_by_ptr(&shp->shm_perm);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700126 do_shm_rmid(ns, shp);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700127 total++;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700128 }
Nadia Derbey3e148c72007-10-18 23:40:54 -0700129 up_write(&shm_ids(ns).rw_mutex);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700130
131 kfree(ns->ids[IPC_SHM_IDS]);
132 ns->ids[IPC_SHM_IDS] = NULL;
133}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134
135void __init shm_init (void)
136{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700137 __shm_init_ns(&init_ipc_ns, &init_shm_ids);
Mike Waychison19b49462005-09-06 15:17:10 -0700138 ipc_init_proc_interface("sysvipc/shm",
139 " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime\n",
Kirill Korotaev4e982312006-10-02 02:18:22 -0700140 IPC_SHM_IDS, sysvipc_shm_proc_show);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141}
142
Nadia Derbey3e148c72007-10-18 23:40:54 -0700143/*
144 * shm_lock_(check_)down routines are called in the paths where the rw_mutex
145 * is held to protect access to the idr tree.
146 */
147static inline struct shmid_kernel *shm_lock_down(struct ipc_namespace *ns,
148 int id)
149{
150 struct kern_ipc_perm *ipcp = ipc_lock_down(&shm_ids(ns), id);
151
152 return container_of(ipcp, struct shmid_kernel, shm_perm);
153}
154
155static inline struct shmid_kernel *shm_lock_check_down(
156 struct ipc_namespace *ns,
157 int id)
158{
159 struct kern_ipc_perm *ipcp = ipc_lock_check_down(&shm_ids(ns), id);
160
161 return container_of(ipcp, struct shmid_kernel, shm_perm);
162}
163
164/*
165 * shm_lock_(check_) routines are called in the paths where the rw_mutex
166 * is not held.
167 */
Nadia Derbey023a5352007-10-18 23:40:51 -0700168static inline struct shmid_kernel *shm_lock(struct ipc_namespace *ns, int id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700170 struct kern_ipc_perm *ipcp = ipc_lock(&shm_ids(ns), id);
171
172 return container_of(ipcp, struct shmid_kernel, shm_perm);
Nadia Derbey023a5352007-10-18 23:40:51 -0700173}
174
175static inline struct shmid_kernel *shm_lock_check(struct ipc_namespace *ns,
176 int id)
177{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700178 struct kern_ipc_perm *ipcp = ipc_lock_check(&shm_ids(ns), id);
179
180 return container_of(ipcp, struct shmid_kernel, shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181}
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{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700185 ipc_rmid(&shm_ids(ns), &s->shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186}
187
Kirill Korotaev4e982312006-10-02 02:18:22 -0700188static inline int shm_addid(struct ipc_namespace *ns, struct shmid_kernel *shp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700190 return ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191}
192
193
194
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800195/* This is called by fork, once for every shm attach. */
196static void shm_open(struct vm_area_struct *vma)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700197{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800198 struct file *file = vma->vm_file;
199 struct shm_file_data *sfd = shm_file_data(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200 struct shmid_kernel *shp;
201
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800202 shp = shm_lock(sfd->ns, sfd->id);
Nadia Derbey023a5352007-10-18 23:40:51 -0700203 BUG_ON(IS_ERR(shp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 shp->shm_atim = get_seconds();
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700205 shp->shm_lprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206 shp->shm_nattch++;
207 shm_unlock(shp);
208}
209
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210/*
211 * shm_destroy - free the struct shmid_kernel
212 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700213 * @ns: namespace
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 * @shp: struct to free
215 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700216 * It has to be called with shp and shm_ids.rw_mutex (writer) locked,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 * but returns with shp unlocked and freed.
218 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700219static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700221 ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700222 shm_rmid(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 shm_unlock(shp);
224 if (!is_file_hugepages(shp->shm_file))
225 shmem_lock(shp->shm_file, 0, shp->mlock_user);
226 else
Josef Sipek6d630792006-12-08 02:37:11 -0800227 user_shm_unlock(shp->shm_file->f_path.dentry->d_inode->i_size,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 shp->mlock_user);
229 fput (shp->shm_file);
230 security_shm_free(shp);
231 ipc_rcu_putref(shp);
232}
233
234/*
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800235 * remove the attach descriptor vma.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 * free memory for segment if it is marked destroyed.
237 * The descriptor has already been removed from the current->mm->mmap list
238 * and will later be kfree()d.
239 */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800240static void shm_close(struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800242 struct file * file = vma->vm_file;
243 struct shm_file_data *sfd = shm_file_data(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 struct shmid_kernel *shp;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800245 struct ipc_namespace *ns = sfd->ns;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700246
Nadia Derbey3e148c72007-10-18 23:40:54 -0700247 down_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248 /* remove from the list of attaches of the shm segment */
Nadia Derbey3e148c72007-10-18 23:40:54 -0700249 shp = shm_lock_down(ns, sfd->id);
Nadia Derbey023a5352007-10-18 23:40:51 -0700250 BUG_ON(IS_ERR(shp));
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700251 shp->shm_lprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 shp->shm_dtim = get_seconds();
253 shp->shm_nattch--;
254 if(shp->shm_nattch == 0 &&
Andrew Mortonb33291c2006-01-08 01:02:21 -0800255 shp->shm_perm.mode & SHM_DEST)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700256 shm_destroy(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 else
258 shm_unlock(shp);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700259 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260}
261
Nick Piggind0217ac2007-07-19 01:47:03 -0700262static int shm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800263{
264 struct file *file = vma->vm_file;
265 struct shm_file_data *sfd = shm_file_data(file);
266
Nick Piggind0217ac2007-07-19 01:47:03 -0700267 return sfd->vm_ops->fault(vma, vmf);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800268}
269
270#ifdef CONFIG_NUMA
Adrian Bunkd823e3e2007-10-16 23:26:42 -0700271static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800272{
273 struct file *file = vma->vm_file;
274 struct shm_file_data *sfd = shm_file_data(file);
275 int err = 0;
276 if (sfd->vm_ops->set_policy)
277 err = sfd->vm_ops->set_policy(vma, new);
278 return err;
279}
280
Adrian Bunkd823e3e2007-10-16 23:26:42 -0700281static struct mempolicy *shm_get_policy(struct vm_area_struct *vma,
282 unsigned long addr)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800283{
284 struct file *file = vma->vm_file;
285 struct shm_file_data *sfd = shm_file_data(file);
286 struct mempolicy *pol = NULL;
287
288 if (sfd->vm_ops->get_policy)
289 pol = sfd->vm_ops->get_policy(vma, addr);
Adam Litke22741922007-06-16 10:16:15 -0700290 else if (vma->vm_policy)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800291 pol = vma->vm_policy;
Adam Litke22741922007-06-16 10:16:15 -0700292 else
293 pol = current->mempolicy;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800294 return pol;
295}
296#endif
297
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298static int shm_mmap(struct file * file, struct vm_area_struct * vma)
299{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800300 struct shm_file_data *sfd = shm_file_data(file);
David Howellsb0e15192006-01-06 00:11:42 -0800301 int ret;
302
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800303 ret = sfd->file->f_op->mmap(sfd->file, vma);
304 if (ret != 0)
305 return ret;
306 sfd->vm_ops = vma->vm_ops;
David Howells2e92a3b2007-07-31 00:37:24 -0700307#ifdef CONFIG_MMU
Nick Piggin54cb8822007-07-19 01:46:59 -0700308 BUG_ON(!sfd->vm_ops->fault);
David Howells2e92a3b2007-07-31 00:37:24 -0700309#endif
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800310 vma->vm_ops = &shm_vm_ops;
311 shm_open(vma);
David Howellsb0e15192006-01-06 00:11:42 -0800312
313 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314}
315
Kirill Korotaev4e982312006-10-02 02:18:22 -0700316static int shm_release(struct inode *ino, struct file *file)
317{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800318 struct shm_file_data *sfd = shm_file_data(file);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700319
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800320 put_ipc_ns(sfd->ns);
321 shm_file_data(file) = NULL;
322 kfree(sfd);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700323 return 0;
324}
325
Adam Litke516dffd2007-03-01 15:46:08 -0800326static int shm_fsync(struct file *file, struct dentry *dentry, int datasync)
327{
328 int (*fsync) (struct file *, struct dentry *, int datasync);
329 struct shm_file_data *sfd = shm_file_data(file);
330 int ret = -EINVAL;
331
332 fsync = sfd->file->f_op->fsync;
333 if (fsync)
334 ret = fsync(sfd->file, sfd->file->f_path.dentry, datasync);
335 return ret;
336}
337
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800338static unsigned long shm_get_unmapped_area(struct file *file,
339 unsigned long addr, unsigned long len, unsigned long pgoff,
340 unsigned long flags)
341{
342 struct shm_file_data *sfd = shm_file_data(file);
Adam Litke516dffd2007-03-01 15:46:08 -0800343 return get_unmapped_area(sfd->file, addr, len, pgoff, flags);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800344}
Adam Litke516dffd2007-03-01 15:46:08 -0800345
346int is_file_shm_hugepages(struct file *file)
347{
348 int ret = 0;
349
350 if (file->f_op == &shm_file_operations) {
351 struct shm_file_data *sfd;
352 sfd = shm_file_data(file);
353 ret = is_file_hugepages(sfd->file);
354 }
355 return ret;
356}
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800357
Arjan van de Ven9a321442007-02-12 00:55:35 -0800358static const struct file_operations shm_file_operations = {
Kirill Korotaev4e982312006-10-02 02:18:22 -0700359 .mmap = shm_mmap,
Adam Litke516dffd2007-03-01 15:46:08 -0800360 .fsync = shm_fsync,
Kirill Korotaev4e982312006-10-02 02:18:22 -0700361 .release = shm_release,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800362 .get_unmapped_area = shm_get_unmapped_area,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363};
364
365static struct vm_operations_struct shm_vm_ops = {
366 .open = shm_open, /* callback for a new vm-area open */
367 .close = shm_close, /* callback for when the vm-area is released */
Nick Piggin54cb8822007-07-19 01:46:59 -0700368 .fault = shm_fault,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800369#if defined(CONFIG_NUMA)
370 .set_policy = shm_set_policy,
371 .get_policy = shm_get_policy,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372#endif
373};
374
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700375/**
376 * newseg - Create a new shared memory segment
377 * @ns: namespace
378 * @params: ptr to the structure that contains key, size and shmflg
379 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700380 * Called with shm_ids.rw_mutex held as a writer.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700381 */
382
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700383static int newseg(struct ipc_namespace *ns, struct ipc_params *params)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384{
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700385 key_t key = params->key;
386 int shmflg = params->flg;
387 size_t size = params->u.size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 int error;
389 struct shmid_kernel *shp;
390 int numpages = (size + PAGE_SIZE -1) >> PAGE_SHIFT;
391 struct file * file;
392 char name[13];
393 int id;
394
Kirill Korotaev4e982312006-10-02 02:18:22 -0700395 if (size < SHMMIN || size > ns->shm_ctlmax)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 return -EINVAL;
397
Guy Streeterf66d45e2007-01-23 12:20:04 -0600398 if (ns->shm_tot + numpages > ns->shm_ctlall)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 return -ENOSPC;
400
401 shp = ipc_rcu_alloc(sizeof(*shp));
402 if (!shp)
403 return -ENOMEM;
404
405 shp->shm_perm.key = key;
Andrew Mortonb33291c2006-01-08 01:02:21 -0800406 shp->shm_perm.mode = (shmflg & S_IRWXUGO);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 shp->mlock_user = NULL;
408
409 shp->shm_perm.security = NULL;
410 error = security_shm_alloc(shp);
411 if (error) {
412 ipc_rcu_putref(shp);
413 return error;
414 }
415
Eric W. Biederman9d665862007-06-16 10:16:16 -0700416 sprintf (name, "SYSV%08x", key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 if (shmflg & SHM_HUGETLB) {
Eric W. Biederman9d665862007-06-16 10:16:16 -0700418 /* hugetlb_file_setup takes care of mlock user accounting */
419 file = hugetlb_file_setup(name, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 shp->mlock_user = current->user;
421 } else {
Badari Pulavartybf8f9722005-11-07 00:59:27 -0800422 int acctflag = VM_ACCOUNT;
423 /*
424 * Do not allow no accounting for OVERCOMMIT_NEVER, even
425 * if it's asked for.
426 */
427 if ((shmflg & SHM_NORESERVE) &&
428 sysctl_overcommit_memory != OVERCOMMIT_NEVER)
429 acctflag = 0;
Badari Pulavartybf8f9722005-11-07 00:59:27 -0800430 file = shmem_file_setup(name, size, acctflag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 }
432 error = PTR_ERR(file);
433 if (IS_ERR(file))
434 goto no_file;
435
Kirill Korotaev4e982312006-10-02 02:18:22 -0700436 id = shm_addid(ns, shp);
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700437 if (id < 0) {
438 error = id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 goto no_id;
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700440 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700442 shp->shm_cprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 shp->shm_lprid = 0;
444 shp->shm_atim = shp->shm_dtim = 0;
445 shp->shm_ctim = get_seconds();
446 shp->shm_segsz = size;
447 shp->shm_nattch = 0;
Nadia Derbey1b531f22007-10-18 23:40:55 -0700448 shp->shm_perm.id = shm_buildid(id, shp->shm_perm.seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 shp->shm_file = file;
Badari Pulavarty30475cc2007-06-16 10:15:59 -0700450 /*
451 * shmid gets reported as "inode#" in /proc/pid/maps.
452 * proc-ps tools use this. Changing this will break them.
453 */
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700454 file->f_dentry->d_inode->i_ino = shp->shm_perm.id;
Krishnakumar R551110a2005-10-29 18:16:45 -0700455
Kirill Korotaev4e982312006-10-02 02:18:22 -0700456 ns->shm_tot += numpages;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700457 error = shp->shm_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 shm_unlock(shp);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700459 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460
461no_id:
462 fput(file);
463no_file:
464 security_shm_free(shp);
465 ipc_rcu_putref(shp);
466 return error;
467}
468
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700469/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700470 * Called with shm_ids.rw_mutex and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700471 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700472static inline int shm_security(struct kern_ipc_perm *ipcp, int shmflg)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700473{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700474 struct shmid_kernel *shp;
475
476 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
477 return security_shm_associate(shp, shmflg);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700478}
479
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700480/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700481 * Called with shm_ids.rw_mutex and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700482 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700483static inline int shm_more_checks(struct kern_ipc_perm *ipcp,
484 struct ipc_params *params)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700485{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700486 struct shmid_kernel *shp;
487
488 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
489 if (shp->shm_segsz < params->u.size)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700490 return -EINVAL;
491
492 return 0;
493}
494
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495asmlinkage long sys_shmget (key_t key, size_t size, int shmflg)
496{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700497 struct ipc_namespace *ns;
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700498 struct ipc_ops shm_ops;
499 struct ipc_params shm_params;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500
Kirill Korotaev4e982312006-10-02 02:18:22 -0700501 ns = current->nsproxy->ipc_ns;
502
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700503 shm_ops.getnew = newseg;
504 shm_ops.associate = shm_security;
505 shm_ops.more_checks = shm_more_checks;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700506
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700507 shm_params.key = key;
508 shm_params.flg = shmflg;
509 shm_params.u.size = size;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700510
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700511 return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512}
513
514static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version)
515{
516 switch(version) {
517 case IPC_64:
518 return copy_to_user(buf, in, sizeof(*in));
519 case IPC_OLD:
520 {
521 struct shmid_ds out;
522
523 ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm);
524 out.shm_segsz = in->shm_segsz;
525 out.shm_atime = in->shm_atime;
526 out.shm_dtime = in->shm_dtime;
527 out.shm_ctime = in->shm_ctime;
528 out.shm_cpid = in->shm_cpid;
529 out.shm_lpid = in->shm_lpid;
530 out.shm_nattch = in->shm_nattch;
531
532 return copy_to_user(buf, &out, sizeof(out));
533 }
534 default:
535 return -EINVAL;
536 }
537}
538
539struct shm_setbuf {
540 uid_t uid;
541 gid_t gid;
542 mode_t mode;
543};
544
545static inline unsigned long copy_shmid_from_user(struct shm_setbuf *out, void __user *buf, int version)
546{
547 switch(version) {
548 case IPC_64:
549 {
550 struct shmid64_ds tbuf;
551
552 if (copy_from_user(&tbuf, buf, sizeof(tbuf)))
553 return -EFAULT;
554
555 out->uid = tbuf.shm_perm.uid;
556 out->gid = tbuf.shm_perm.gid;
Andrew Mortonb33291c2006-01-08 01:02:21 -0800557 out->mode = tbuf.shm_perm.mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558
559 return 0;
560 }
561 case IPC_OLD:
562 {
563 struct shmid_ds tbuf_old;
564
565 if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old)))
566 return -EFAULT;
567
568 out->uid = tbuf_old.shm_perm.uid;
569 out->gid = tbuf_old.shm_perm.gid;
Andrew Mortonb33291c2006-01-08 01:02:21 -0800570 out->mode = tbuf_old.shm_perm.mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571
572 return 0;
573 }
574 default:
575 return -EINVAL;
576 }
577}
578
579static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version)
580{
581 switch(version) {
582 case IPC_64:
583 return copy_to_user(buf, in, sizeof(*in));
584 case IPC_OLD:
585 {
586 struct shminfo out;
587
588 if(in->shmmax > INT_MAX)
589 out.shmmax = INT_MAX;
590 else
591 out.shmmax = (int)in->shmmax;
592
593 out.shmmin = in->shmmin;
594 out.shmmni = in->shmmni;
595 out.shmseg = in->shmseg;
596 out.shmall = in->shmall;
597
598 return copy_to_user(buf, &out, sizeof(out));
599 }
600 default:
601 return -EINVAL;
602 }
603}
604
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700605/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700606 * Called with shm_ids.rw_mutex held as a reader
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700607 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700608static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss,
609 unsigned long *swp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700611 int next_id;
612 int total, in_use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613
614 *rss = 0;
615 *swp = 0;
616
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700617 in_use = shm_ids(ns).in_use;
618
619 for (total = 0, next_id = 0; total < in_use; next_id++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 struct shmid_kernel *shp;
621 struct inode *inode;
622
Nadia Derbey637c3662007-10-18 23:40:50 -0700623 shp = idr_find(&shm_ids(ns).ipcs_idr, next_id);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700624 if (shp == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 continue;
626
Josef Sipek6d630792006-12-08 02:37:11 -0800627 inode = shp->shm_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628
629 if (is_file_hugepages(shp->shm_file)) {
630 struct address_space *mapping = inode->i_mapping;
631 *rss += (HPAGE_SIZE/PAGE_SIZE)*mapping->nrpages;
632 } else {
633 struct shmem_inode_info *info = SHMEM_I(inode);
634 spin_lock(&info->lock);
635 *rss += inode->i_mapping->nrpages;
636 *swp += info->swapped;
637 spin_unlock(&info->lock);
638 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700639
640 total++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 }
642}
643
644asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf)
645{
646 struct shm_setbuf setbuf;
647 struct shmid_kernel *shp;
648 int err, version;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700649 struct ipc_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
651 if (cmd < 0 || shmid < 0) {
652 err = -EINVAL;
653 goto out;
654 }
655
656 version = ipc_parse_version(&cmd);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700657 ns = current->nsproxy->ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
659 switch (cmd) { /* replace with proc interface ? */
660 case IPC_INFO:
661 {
662 struct shminfo64 shminfo;
663
664 err = security_shm_shmctl(NULL, cmd);
665 if (err)
666 return err;
667
668 memset(&shminfo,0,sizeof(shminfo));
Kirill Korotaev4e982312006-10-02 02:18:22 -0700669 shminfo.shmmni = shminfo.shmseg = ns->shm_ctlmni;
670 shminfo.shmmax = ns->shm_ctlmax;
671 shminfo.shmall = ns->shm_ctlall;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672
673 shminfo.shmmin = SHMMIN;
674 if(copy_shminfo_to_user (buf, &shminfo, version))
675 return -EFAULT;
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700676
Nadia Derbey3e148c72007-10-18 23:40:54 -0700677 down_read(&shm_ids(ns).rw_mutex);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700678 err = ipc_get_maxid(&shm_ids(ns));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700679 up_read(&shm_ids(ns).rw_mutex);
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700680
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 if(err<0)
682 err = 0;
683 goto out;
684 }
685 case SHM_INFO:
686 {
687 struct shm_info shm_info;
688
689 err = security_shm_shmctl(NULL, cmd);
690 if (err)
691 return err;
692
693 memset(&shm_info,0,sizeof(shm_info));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700694 down_read(&shm_ids(ns).rw_mutex);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700695 shm_info.used_ids = shm_ids(ns).in_use;
696 shm_get_stat (ns, &shm_info.shm_rss, &shm_info.shm_swp);
697 shm_info.shm_tot = ns->shm_tot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 shm_info.swap_attempts = 0;
699 shm_info.swap_successes = 0;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700700 err = ipc_get_maxid(&shm_ids(ns));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700701 up_read(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 if(copy_to_user (buf, &shm_info, sizeof(shm_info))) {
703 err = -EFAULT;
704 goto out;
705 }
706
707 err = err < 0 ? 0 : err;
708 goto out;
709 }
710 case SHM_STAT:
711 case IPC_STAT:
712 {
713 struct shmid64_ds tbuf;
714 int result;
Nadia Derbey023a5352007-10-18 23:40:51 -0700715
716 if (!buf) {
717 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 goto out;
Nadia Derbey023a5352007-10-18 23:40:51 -0700719 }
720
721 if (cmd == SHM_STAT) {
722 shp = shm_lock(ns, shmid);
723 if (IS_ERR(shp)) {
724 err = PTR_ERR(shp);
725 goto out;
726 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700727 result = shp->shm_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 } else {
Nadia Derbey023a5352007-10-18 23:40:51 -0700729 shp = shm_lock_check(ns, shmid);
730 if (IS_ERR(shp)) {
731 err = PTR_ERR(shp);
732 goto out;
733 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 result = 0;
735 }
736 err=-EACCES;
737 if (ipcperms (&shp->shm_perm, S_IRUGO))
738 goto out_unlock;
739 err = security_shm_shmctl(shp, cmd);
740 if (err)
741 goto out_unlock;
Nadia Derbey023a5352007-10-18 23:40:51 -0700742 memset(&tbuf, 0, sizeof(tbuf));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 kernel_to_ipc64_perm(&shp->shm_perm, &tbuf.shm_perm);
744 tbuf.shm_segsz = shp->shm_segsz;
745 tbuf.shm_atime = shp->shm_atim;
746 tbuf.shm_dtime = shp->shm_dtim;
747 tbuf.shm_ctime = shp->shm_ctim;
748 tbuf.shm_cpid = shp->shm_cprid;
749 tbuf.shm_lpid = shp->shm_lprid;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800750 tbuf.shm_nattch = shp->shm_nattch;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 shm_unlock(shp);
752 if(copy_shmid_to_user (buf, &tbuf, version))
753 err = -EFAULT;
754 else
755 err = result;
756 goto out;
757 }
758 case SHM_LOCK:
759 case SHM_UNLOCK:
760 {
Nadia Derbey023a5352007-10-18 23:40:51 -0700761 shp = shm_lock_check(ns, shmid);
762 if (IS_ERR(shp)) {
763 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 goto out;
765 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766
Steve Grubb073115d2006-04-02 17:07:33 -0400767 err = audit_ipc_obj(&(shp->shm_perm));
768 if (err)
769 goto out_unlock;
770
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 if (!capable(CAP_IPC_LOCK)) {
772 err = -EPERM;
773 if (current->euid != shp->shm_perm.uid &&
774 current->euid != shp->shm_perm.cuid)
775 goto out_unlock;
776 if (cmd == SHM_LOCK &&
777 !current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur)
778 goto out_unlock;
779 }
780
781 err = security_shm_shmctl(shp, cmd);
782 if (err)
783 goto out_unlock;
784
785 if(cmd==SHM_LOCK) {
786 struct user_struct * user = current->user;
787 if (!is_file_hugepages(shp->shm_file)) {
788 err = shmem_lock(shp->shm_file, 1, user);
Pavel Emelianov7be77e22007-07-31 00:38:48 -0700789 if (!err && !(shp->shm_perm.mode & SHM_LOCKED)){
Andrew Mortonb33291c2006-01-08 01:02:21 -0800790 shp->shm_perm.mode |= SHM_LOCKED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 shp->mlock_user = user;
792 }
793 }
794 } else if (!is_file_hugepages(shp->shm_file)) {
795 shmem_lock(shp->shm_file, 0, shp->mlock_user);
Andrew Mortonb33291c2006-01-08 01:02:21 -0800796 shp->shm_perm.mode &= ~SHM_LOCKED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 shp->mlock_user = NULL;
798 }
799 shm_unlock(shp);
800 goto out;
801 }
802 case IPC_RMID:
803 {
804 /*
805 * We cannot simply remove the file. The SVID states
806 * that the block remains until the last person
807 * detaches from it, then is deleted. A shmat() on
808 * an RMID segment is legal in older Linux and if
809 * we change it apps break...
810 *
811 * Instead we set a destroyed flag, and then blow
812 * the name away when the usage hits zero.
813 */
Nadia Derbey3e148c72007-10-18 23:40:54 -0700814 down_write(&shm_ids(ns).rw_mutex);
815 shp = shm_lock_check_down(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -0700816 if (IS_ERR(shp)) {
817 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 goto out_up;
Nadia Derbey023a5352007-10-18 23:40:51 -0700819 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820
Steve Grubb073115d2006-04-02 17:07:33 -0400821 err = audit_ipc_obj(&(shp->shm_perm));
822 if (err)
823 goto out_unlock_up;
824
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 if (current->euid != shp->shm_perm.uid &&
826 current->euid != shp->shm_perm.cuid &&
827 !capable(CAP_SYS_ADMIN)) {
828 err=-EPERM;
829 goto out_unlock_up;
830 }
831
832 err = security_shm_shmctl(shp, cmd);
833 if (err)
834 goto out_unlock_up;
835
Kirill Korotaev4e982312006-10-02 02:18:22 -0700836 do_shm_rmid(ns, shp);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700837 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 goto out;
839 }
840
841 case IPC_SET:
842 {
Nadia Derbey023a5352007-10-18 23:40:51 -0700843 if (!buf) {
844 err = -EFAULT;
845 goto out;
846 }
847
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 if (copy_shmid_from_user (&setbuf, buf, version)) {
849 err = -EFAULT;
850 goto out;
851 }
Nadia Derbey3e148c72007-10-18 23:40:54 -0700852 down_write(&shm_ids(ns).rw_mutex);
853 shp = shm_lock_check_down(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -0700854 if (IS_ERR(shp)) {
855 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 goto out_up;
Nadia Derbey023a5352007-10-18 23:40:51 -0700857 }
Steve Grubb073115d2006-04-02 17:07:33 -0400858 err = audit_ipc_obj(&(shp->shm_perm));
859 if (err)
860 goto out_unlock_up;
Linda Knippersac032212006-05-16 22:03:48 -0400861 err = audit_ipc_set_perm(0, setbuf.uid, setbuf.gid, setbuf.mode);
Steve Grubb073115d2006-04-02 17:07:33 -0400862 if (err)
863 goto out_unlock_up;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 err=-EPERM;
865 if (current->euid != shp->shm_perm.uid &&
866 current->euid != shp->shm_perm.cuid &&
867 !capable(CAP_SYS_ADMIN)) {
868 goto out_unlock_up;
869 }
870
871 err = security_shm_shmctl(shp, cmd);
872 if (err)
873 goto out_unlock_up;
874
875 shp->shm_perm.uid = setbuf.uid;
876 shp->shm_perm.gid = setbuf.gid;
Andrew Mortonb33291c2006-01-08 01:02:21 -0800877 shp->shm_perm.mode = (shp->shm_perm.mode & ~S_IRWXUGO)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 | (setbuf.mode & S_IRWXUGO);
879 shp->shm_ctim = get_seconds();
880 break;
881 }
882
883 default:
884 err = -EINVAL;
885 goto out;
886 }
887
888 err = 0;
889out_unlock_up:
890 shm_unlock(shp);
891out_up:
Nadia Derbey3e148c72007-10-18 23:40:54 -0700892 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 goto out;
894out_unlock:
895 shm_unlock(shp);
896out:
897 return err;
898}
899
900/*
901 * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists.
902 *
903 * NOTE! Despite the name, this is NOT a direct system call entrypoint. The
904 * "raddr" thing points to kernel space, and there has to be a wrapper around
905 * this.
906 */
907long do_shmat(int shmid, char __user *shmaddr, int shmflg, ulong *raddr)
908{
909 struct shmid_kernel *shp;
910 unsigned long addr;
911 unsigned long size;
912 struct file * file;
913 int err;
914 unsigned long flags;
915 unsigned long prot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 int acc_mode;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800917 unsigned long user_addr;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700918 struct ipc_namespace *ns;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800919 struct shm_file_data *sfd;
920 struct path path;
921 mode_t f_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800923 err = -EINVAL;
924 if (shmid < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 goto out;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800926 else if ((addr = (ulong)shmaddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 if (addr & (SHMLBA-1)) {
928 if (shmflg & SHM_RND)
929 addr &= ~(SHMLBA-1); /* round down */
930 else
931#ifndef __ARCH_FORCE_SHMLBA
932 if (addr & ~PAGE_MASK)
933#endif
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800934 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935 }
936 flags = MAP_SHARED | MAP_FIXED;
937 } else {
938 if ((shmflg & SHM_REMAP))
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800939 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940
941 flags = MAP_SHARED;
942 }
943
944 if (shmflg & SHM_RDONLY) {
945 prot = PROT_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 acc_mode = S_IRUGO;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800947 f_mode = FMODE_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 } else {
949 prot = PROT_READ | PROT_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 acc_mode = S_IRUGO | S_IWUGO;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800951 f_mode = FMODE_READ | FMODE_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 }
953 if (shmflg & SHM_EXEC) {
954 prot |= PROT_EXEC;
955 acc_mode |= S_IXUGO;
956 }
957
958 /*
959 * We cannot rely on the fs check since SYSV IPC does have an
960 * additional creator id...
961 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700962 ns = current->nsproxy->ipc_ns;
Nadia Derbey023a5352007-10-18 23:40:51 -0700963 shp = shm_lock_check(ns, shmid);
964 if (IS_ERR(shp)) {
965 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 goto out;
Nadia Derbey023a5352007-10-18 23:40:51 -0700967 }
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800968
969 err = -EACCES;
970 if (ipcperms(&shp->shm_perm, acc_mode))
971 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972
973 err = security_shm_shmat(shp, shmaddr, shmflg);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800974 if (err)
975 goto out_unlock;
976
977 path.dentry = dget(shp->shm_file->f_path.dentry);
Dave Hansence8d2cd2007-10-16 23:31:13 -0700978 path.mnt = shp->shm_file->f_path.mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 shp->shm_nattch++;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800980 size = i_size_read(path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 shm_unlock(shp);
982
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800983 err = -ENOMEM;
984 sfd = kzalloc(sizeof(*sfd), GFP_KERNEL);
985 if (!sfd)
Dave Hansence8d2cd2007-10-16 23:31:13 -0700986 goto out_put_dentry;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800987
988 err = -ENOMEM;
Dave Hansence8d2cd2007-10-16 23:31:13 -0700989
990 file = alloc_file(path.mnt, path.dentry, f_mode, &shm_file_operations);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800991 if (!file)
992 goto out_free;
993
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800994 file->private_data = sfd;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800995 file->f_mapping = shp->shm_file->f_mapping;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700996 sfd->id = shp->shm_perm.id;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800997 sfd->ns = get_ipc_ns(ns);
998 sfd->file = shp->shm_file;
999 sfd->vm_ops = NULL;
1000
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 down_write(&current->mm->mmap_sem);
1002 if (addr && !(shmflg & SHM_REMAP)) {
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001003 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 if (find_vma_intersection(current->mm, addr, addr + size))
1005 goto invalid;
1006 /*
1007 * If shm segment goes below stack, make sure there is some
1008 * space left for the stack to grow (at least 4 pages).
1009 */
1010 if (addr < current->mm->start_stack &&
1011 addr > current->mm->start_stack - size - PAGE_SIZE * 5)
1012 goto invalid;
1013 }
1014
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001015 user_addr = do_mmap (file, addr, size, prot, flags, 0);
1016 *raddr = user_addr;
1017 err = 0;
1018 if (IS_ERR_VALUE(user_addr))
1019 err = (long)user_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020invalid:
1021 up_write(&current->mm->mmap_sem);
1022
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001023 fput(file);
1024
1025out_nattch:
Nadia Derbey3e148c72007-10-18 23:40:54 -07001026 down_write(&shm_ids(ns).rw_mutex);
1027 shp = shm_lock_down(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -07001028 BUG_ON(IS_ERR(shp));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029 shp->shm_nattch--;
1030 if(shp->shm_nattch == 0 &&
Andrew Mortonb33291c2006-01-08 01:02:21 -08001031 shp->shm_perm.mode & SHM_DEST)
Kirill Korotaev4e982312006-10-02 02:18:22 -07001032 shm_destroy(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 else
1034 shm_unlock(shp);
Nadia Derbey3e148c72007-10-18 23:40:54 -07001035 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037out:
1038 return err;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001039
1040out_unlock:
1041 shm_unlock(shp);
1042 goto out;
1043
1044out_free:
1045 kfree(sfd);
Dave Hansence8d2cd2007-10-16 23:31:13 -07001046out_put_dentry:
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001047 dput(path.dentry);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001048 goto out_nattch;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049}
1050
Stephen Rothwell7d87e142005-05-01 08:59:12 -07001051asmlinkage long sys_shmat(int shmid, char __user *shmaddr, int shmflg)
1052{
1053 unsigned long ret;
1054 long err;
1055
1056 err = do_shmat(shmid, shmaddr, shmflg, &ret);
1057 if (err)
1058 return err;
1059 force_successful_syscall_return();
1060 return (long)ret;
1061}
1062
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063/*
1064 * detach and kill segment if marked destroyed.
1065 * The work is done in shm_close.
1066 */
1067asmlinkage long sys_shmdt(char __user *shmaddr)
1068{
1069 struct mm_struct *mm = current->mm;
1070 struct vm_area_struct *vma, *next;
1071 unsigned long addr = (unsigned long)shmaddr;
1072 loff_t size = 0;
1073 int retval = -EINVAL;
1074
Hugh Dickinsdf1e2fb2006-03-24 03:18:06 -08001075 if (addr & ~PAGE_MASK)
1076 return retval;
1077
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 down_write(&mm->mmap_sem);
1079
1080 /*
1081 * This function tries to be smart and unmap shm segments that
1082 * were modified by partial mlock or munmap calls:
1083 * - It first determines the size of the shm segment that should be
1084 * unmapped: It searches for a vma that is backed by shm and that
1085 * started at address shmaddr. It records it's size and then unmaps
1086 * it.
1087 * - Then it unmaps all shm vmas that started at shmaddr and that
1088 * are within the initially determined size.
1089 * Errors from do_munmap are ignored: the function only fails if
1090 * it's called with invalid parameters or if it's called to unmap
1091 * a part of a vma. Both calls in this function are for full vmas,
1092 * the parameters are directly copied from the vma itself and always
1093 * valid - therefore do_munmap cannot fail. (famous last words?)
1094 */
1095 /*
1096 * If it had been mremap()'d, the starting address would not
1097 * match the usual checks anyway. So assume all vma's are
1098 * above the starting address given.
1099 */
1100 vma = find_vma(mm, addr);
1101
1102 while (vma) {
1103 next = vma->vm_next;
1104
1105 /*
1106 * Check if the starting address would match, i.e. it's
1107 * a fragment created by mprotect() and/or munmap(), or it
1108 * otherwise it starts at this address with no hassles.
1109 */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001110 if ((vma->vm_ops == &shm_vm_ops) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111 (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) {
1112
1113
Josef Sipek6d630792006-12-08 02:37:11 -08001114 size = vma->vm_file->f_path.dentry->d_inode->i_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1116 /*
1117 * We discovered the size of the shm segment, so
1118 * break out of here and fall through to the next
1119 * loop that uses the size information to stop
1120 * searching for matching vma's.
1121 */
1122 retval = 0;
1123 vma = next;
1124 break;
1125 }
1126 vma = next;
1127 }
1128
1129 /*
1130 * We need look no further than the maximum address a fragment
1131 * could possibly have landed at. Also cast things to loff_t to
1132 * prevent overflows and make comparisions vs. equal-width types.
1133 */
KAMEZAWA Hiroyuki8e367092006-02-10 01:51:12 -08001134 size = PAGE_ALIGN(size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 while (vma && (loff_t)(vma->vm_end - addr) <= size) {
1136 next = vma->vm_next;
1137
1138 /* finding a matching vma now does not alter retval */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001139 if ((vma->vm_ops == &shm_vm_ops) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff)
1141
1142 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1143 vma = next;
1144 }
1145
1146 up_write(&mm->mmap_sem);
1147 return retval;
1148}
1149
1150#ifdef CONFIG_PROC_FS
Mike Waychison19b49462005-09-06 15:17:10 -07001151static int sysvipc_shm_proc_show(struct seq_file *s, void *it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152{
Mike Waychison19b49462005-09-06 15:17:10 -07001153 struct shmid_kernel *shp = it;
1154 char *format;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156#define SMALL_STRING "%10d %10d %4o %10u %5u %5u %5d %5u %5u %5u %5u %10lu %10lu %10lu\n"
1157#define BIG_STRING "%10d %10d %4o %21u %5u %5u %5d %5u %5u %5u %5u %10lu %10lu %10lu\n"
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158
Mike Waychison19b49462005-09-06 15:17:10 -07001159 if (sizeof(size_t) <= sizeof(int))
1160 format = SMALL_STRING;
1161 else
1162 format = BIG_STRING;
1163 return seq_printf(s, format,
1164 shp->shm_perm.key,
Nadia Derbey7ca7e562007-10-18 23:40:48 -07001165 shp->shm_perm.id,
Andrew Mortonb33291c2006-01-08 01:02:21 -08001166 shp->shm_perm.mode,
Mike Waychison19b49462005-09-06 15:17:10 -07001167 shp->shm_segsz,
1168 shp->shm_cprid,
1169 shp->shm_lprid,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001170 shp->shm_nattch,
Mike Waychison19b49462005-09-06 15:17:10 -07001171 shp->shm_perm.uid,
1172 shp->shm_perm.gid,
1173 shp->shm_perm.cuid,
1174 shp->shm_perm.cgid,
1175 shp->shm_atim,
1176 shp->shm_dtim,
1177 shp->shm_ctim);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178}
1179#endif