blob: 20e03dfc6adb69647837d15e740baa1d99011005 [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;
Linus Torvalds1da177e2005-04-16 15:20:36 -070057static struct vm_operations_struct shm_vm_ops;
58
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;
Pierre Peiffered2ddbf2008-02-08 04:18:57 -080078 ipc_init_ids(&ns->ids[IPC_SHM_IDS]);
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);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700103}
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -0800104#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105
106void __init shm_init (void)
107{
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800108 shm_init_ns(&init_ipc_ns);
Mike Waychison19b49462005-09-06 15:17:10 -0700109 ipc_init_proc_interface("sysvipc/shm",
110 " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime\n",
Kirill Korotaev4e982312006-10-02 02:18:22 -0700111 IPC_SHM_IDS, sysvipc_shm_proc_show);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112}
113
Nadia Derbey3e148c72007-10-18 23:40:54 -0700114/*
115 * shm_lock_(check_)down routines are called in the paths where the rw_mutex
116 * is held to protect access to the idr tree.
117 */
118static inline struct shmid_kernel *shm_lock_down(struct ipc_namespace *ns,
119 int id)
120{
121 struct kern_ipc_perm *ipcp = ipc_lock_down(&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 Derbey3e148c72007-10-18 23:40:54 -0700126 return container_of(ipcp, struct shmid_kernel, shm_perm);
127}
128
129static inline struct shmid_kernel *shm_lock_check_down(
130 struct ipc_namespace *ns,
131 int id)
132{
133 struct kern_ipc_perm *ipcp = ipc_lock_check_down(&shm_ids(ns), id);
134
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800135 if (IS_ERR(ipcp))
136 return (struct shmid_kernel *)ipcp;
137
Nadia Derbey3e148c72007-10-18 23:40:54 -0700138 return container_of(ipcp, struct shmid_kernel, shm_perm);
139}
140
141/*
142 * shm_lock_(check_) routines are called in the paths where the rw_mutex
143 * is not held.
144 */
Nadia Derbey023a5352007-10-18 23:40:51 -0700145static inline struct shmid_kernel *shm_lock(struct ipc_namespace *ns, int id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700147 struct kern_ipc_perm *ipcp = ipc_lock(&shm_ids(ns), id);
148
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800149 if (IS_ERR(ipcp))
150 return (struct shmid_kernel *)ipcp;
151
Nadia Derbey03f02c72007-10-18 23:40:51 -0700152 return container_of(ipcp, struct shmid_kernel, shm_perm);
Nadia Derbey023a5352007-10-18 23:40:51 -0700153}
154
155static inline struct shmid_kernel *shm_lock_check(struct ipc_namespace *ns,
156 int id)
157{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700158 struct kern_ipc_perm *ipcp = ipc_lock_check(&shm_ids(ns), id);
159
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800160 if (IS_ERR(ipcp))
161 return (struct shmid_kernel *)ipcp;
162
Nadia Derbey03f02c72007-10-18 23:40:51 -0700163 return container_of(ipcp, struct shmid_kernel, shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164}
165
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700166static inline void shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *s)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700168 ipc_rmid(&shm_ids(ns), &s->shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169}
170
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800172/* This is called by fork, once for every shm attach. */
173static void shm_open(struct vm_area_struct *vma)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700174{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800175 struct file *file = vma->vm_file;
176 struct shm_file_data *sfd = shm_file_data(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 struct shmid_kernel *shp;
178
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800179 shp = shm_lock(sfd->ns, sfd->id);
Nadia Derbey023a5352007-10-18 23:40:51 -0700180 BUG_ON(IS_ERR(shp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 shp->shm_atim = get_seconds();
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700182 shp->shm_lprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 shp->shm_nattch++;
184 shm_unlock(shp);
185}
186
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187/*
188 * shm_destroy - free the struct shmid_kernel
189 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700190 * @ns: namespace
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 * @shp: struct to free
192 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700193 * It has to be called with shp and shm_ids.rw_mutex (writer) locked,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 * but returns with shp unlocked and freed.
195 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700196static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700198 ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700199 shm_rmid(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200 shm_unlock(shp);
201 if (!is_file_hugepages(shp->shm_file))
202 shmem_lock(shp->shm_file, 0, shp->mlock_user);
203 else
Josef Sipek6d630792006-12-08 02:37:11 -0800204 user_shm_unlock(shp->shm_file->f_path.dentry->d_inode->i_size,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205 shp->mlock_user);
206 fput (shp->shm_file);
207 security_shm_free(shp);
208 ipc_rcu_putref(shp);
209}
210
211/*
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800212 * remove the attach descriptor vma.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 * free memory for segment if it is marked destroyed.
214 * The descriptor has already been removed from the current->mm->mmap list
215 * and will later be kfree()d.
216 */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800217static void shm_close(struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800219 struct file * file = vma->vm_file;
220 struct shm_file_data *sfd = shm_file_data(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 struct shmid_kernel *shp;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800222 struct ipc_namespace *ns = sfd->ns;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700223
Nadia Derbey3e148c72007-10-18 23:40:54 -0700224 down_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 /* remove from the list of attaches of the shm segment */
Nadia Derbey3e148c72007-10-18 23:40:54 -0700226 shp = shm_lock_down(ns, sfd->id);
Nadia Derbey023a5352007-10-18 23:40:51 -0700227 BUG_ON(IS_ERR(shp));
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700228 shp->shm_lprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 shp->shm_dtim = get_seconds();
230 shp->shm_nattch--;
231 if(shp->shm_nattch == 0 &&
Andrew Mortonb33291c2006-01-08 01:02:21 -0800232 shp->shm_perm.mode & SHM_DEST)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700233 shm_destroy(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 else
235 shm_unlock(shp);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700236 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237}
238
Nick Piggind0217ac2007-07-19 01:47:03 -0700239static int shm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800240{
241 struct file *file = vma->vm_file;
242 struct shm_file_data *sfd = shm_file_data(file);
243
Nick Piggind0217ac2007-07-19 01:47:03 -0700244 return sfd->vm_ops->fault(vma, vmf);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800245}
246
247#ifdef CONFIG_NUMA
Adrian Bunkd823e3e2007-10-16 23:26:42 -0700248static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800249{
250 struct file *file = vma->vm_file;
251 struct shm_file_data *sfd = shm_file_data(file);
252 int err = 0;
253 if (sfd->vm_ops->set_policy)
254 err = sfd->vm_ops->set_policy(vma, new);
255 return err;
256}
257
Adrian Bunkd823e3e2007-10-16 23:26:42 -0700258static struct mempolicy *shm_get_policy(struct vm_area_struct *vma,
259 unsigned long addr)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800260{
261 struct file *file = vma->vm_file;
262 struct shm_file_data *sfd = shm_file_data(file);
263 struct mempolicy *pol = NULL;
264
265 if (sfd->vm_ops->get_policy)
266 pol = sfd->vm_ops->get_policy(vma, addr);
Lee Schermerhorn52cd3b02008-04-28 02:13:16 -0700267 else if (vma->vm_policy)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800268 pol = vma->vm_policy;
Lee Schermerhorn52cd3b02008-04-28 02:13:16 -0700269
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800270 return pol;
271}
272#endif
273
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274static int shm_mmap(struct file * file, struct vm_area_struct * vma)
275{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800276 struct shm_file_data *sfd = shm_file_data(file);
David Howellsb0e15192006-01-06 00:11:42 -0800277 int ret;
278
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800279 ret = sfd->file->f_op->mmap(sfd->file, vma);
280 if (ret != 0)
281 return ret;
282 sfd->vm_ops = vma->vm_ops;
David Howells2e92a3b2007-07-31 00:37:24 -0700283#ifdef CONFIG_MMU
Nick Piggin54cb8822007-07-19 01:46:59 -0700284 BUG_ON(!sfd->vm_ops->fault);
David Howells2e92a3b2007-07-31 00:37:24 -0700285#endif
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800286 vma->vm_ops = &shm_vm_ops;
287 shm_open(vma);
David Howellsb0e15192006-01-06 00:11:42 -0800288
289 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290}
291
Kirill Korotaev4e982312006-10-02 02:18:22 -0700292static int shm_release(struct inode *ino, struct file *file)
293{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800294 struct shm_file_data *sfd = shm_file_data(file);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700295
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800296 put_ipc_ns(sfd->ns);
297 shm_file_data(file) = NULL;
298 kfree(sfd);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700299 return 0;
300}
301
Adam Litke516dffd2007-03-01 15:46:08 -0800302static int shm_fsync(struct file *file, struct dentry *dentry, int datasync)
303{
304 int (*fsync) (struct file *, struct dentry *, int datasync);
305 struct shm_file_data *sfd = shm_file_data(file);
306 int ret = -EINVAL;
307
308 fsync = sfd->file->f_op->fsync;
309 if (fsync)
310 ret = fsync(sfd->file, sfd->file->f_path.dentry, datasync);
311 return ret;
312}
313
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800314static unsigned long shm_get_unmapped_area(struct file *file,
315 unsigned long addr, unsigned long len, unsigned long pgoff,
316 unsigned long flags)
317{
318 struct shm_file_data *sfd = shm_file_data(file);
Adam Litke516dffd2007-03-01 15:46:08 -0800319 return get_unmapped_area(sfd->file, addr, len, pgoff, flags);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800320}
Adam Litke516dffd2007-03-01 15:46:08 -0800321
322int is_file_shm_hugepages(struct file *file)
323{
324 int ret = 0;
325
326 if (file->f_op == &shm_file_operations) {
327 struct shm_file_data *sfd;
328 sfd = shm_file_data(file);
329 ret = is_file_hugepages(sfd->file);
330 }
331 return ret;
332}
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800333
Arjan van de Ven9a321442007-02-12 00:55:35 -0800334static const struct file_operations shm_file_operations = {
Kirill Korotaev4e982312006-10-02 02:18:22 -0700335 .mmap = shm_mmap,
Adam Litke516dffd2007-03-01 15:46:08 -0800336 .fsync = shm_fsync,
Kirill Korotaev4e982312006-10-02 02:18:22 -0700337 .release = shm_release,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800338 .get_unmapped_area = shm_get_unmapped_area,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339};
340
341static struct vm_operations_struct shm_vm_ops = {
342 .open = shm_open, /* callback for a new vm-area open */
343 .close = shm_close, /* callback for when the vm-area is released */
Nick Piggin54cb8822007-07-19 01:46:59 -0700344 .fault = shm_fault,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800345#if defined(CONFIG_NUMA)
346 .set_policy = shm_set_policy,
347 .get_policy = shm_get_policy,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348#endif
349};
350
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700351/**
352 * newseg - Create a new shared memory segment
353 * @ns: namespace
354 * @params: ptr to the structure that contains key, size and shmflg
355 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700356 * Called with shm_ids.rw_mutex held as a writer.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700357 */
358
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700359static int newseg(struct ipc_namespace *ns, struct ipc_params *params)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360{
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700361 key_t key = params->key;
362 int shmflg = params->flg;
363 size_t size = params->u.size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 int error;
365 struct shmid_kernel *shp;
366 int numpages = (size + PAGE_SIZE -1) >> PAGE_SHIFT;
367 struct file * file;
368 char name[13];
369 int id;
370
Kirill Korotaev4e982312006-10-02 02:18:22 -0700371 if (size < SHMMIN || size > ns->shm_ctlmax)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 return -EINVAL;
373
Guy Streeterf66d45e2007-01-23 12:20:04 -0600374 if (ns->shm_tot + numpages > ns->shm_ctlall)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375 return -ENOSPC;
376
377 shp = ipc_rcu_alloc(sizeof(*shp));
378 if (!shp)
379 return -ENOMEM;
380
381 shp->shm_perm.key = key;
Andrew Mortonb33291c2006-01-08 01:02:21 -0800382 shp->shm_perm.mode = (shmflg & S_IRWXUGO);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 shp->mlock_user = NULL;
384
385 shp->shm_perm.security = NULL;
386 error = security_shm_alloc(shp);
387 if (error) {
388 ipc_rcu_putref(shp);
389 return error;
390 }
391
Eric W. Biederman9d665862007-06-16 10:16:16 -0700392 sprintf (name, "SYSV%08x", key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 if (shmflg & SHM_HUGETLB) {
Eric W. Biederman9d665862007-06-16 10:16:16 -0700394 /* hugetlb_file_setup takes care of mlock user accounting */
395 file = hugetlb_file_setup(name, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 shp->mlock_user = current->user;
397 } else {
Badari Pulavartybf8f9722005-11-07 00:59:27 -0800398 int acctflag = VM_ACCOUNT;
399 /*
400 * Do not allow no accounting for OVERCOMMIT_NEVER, even
401 * if it's asked for.
402 */
403 if ((shmflg & SHM_NORESERVE) &&
404 sysctl_overcommit_memory != OVERCOMMIT_NEVER)
405 acctflag = 0;
Badari Pulavartybf8f9722005-11-07 00:59:27 -0800406 file = shmem_file_setup(name, size, acctflag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 }
408 error = PTR_ERR(file);
409 if (IS_ERR(file))
410 goto no_file;
411
Pierre Peiffer48dea402008-04-29 01:00:35 -0700412 id = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni);
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700413 if (id < 0) {
414 error = id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 goto no_id;
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700416 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700418 shp->shm_cprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 shp->shm_lprid = 0;
420 shp->shm_atim = shp->shm_dtim = 0;
421 shp->shm_ctim = get_seconds();
422 shp->shm_segsz = size;
423 shp->shm_nattch = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 shp->shm_file = file;
Badari Pulavarty30475cc2007-06-16 10:15:59 -0700425 /*
426 * shmid gets reported as "inode#" in /proc/pid/maps.
427 * proc-ps tools use this. Changing this will break them.
428 */
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700429 file->f_dentry->d_inode->i_ino = shp->shm_perm.id;
Krishnakumar R551110a2005-10-29 18:16:45 -0700430
Kirill Korotaev4e982312006-10-02 02:18:22 -0700431 ns->shm_tot += numpages;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700432 error = shp->shm_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 shm_unlock(shp);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700434 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435
436no_id:
437 fput(file);
438no_file:
439 security_shm_free(shp);
440 ipc_rcu_putref(shp);
441 return error;
442}
443
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700444/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700445 * Called with shm_ids.rw_mutex and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700446 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700447static inline int shm_security(struct kern_ipc_perm *ipcp, int shmflg)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700448{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700449 struct shmid_kernel *shp;
450
451 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
452 return security_shm_associate(shp, shmflg);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700453}
454
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700455/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700456 * Called with shm_ids.rw_mutex and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700457 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700458static inline int shm_more_checks(struct kern_ipc_perm *ipcp,
459 struct ipc_params *params)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700460{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700461 struct shmid_kernel *shp;
462
463 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
464 if (shp->shm_segsz < params->u.size)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700465 return -EINVAL;
466
467 return 0;
468}
469
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470asmlinkage long sys_shmget (key_t key, size_t size, int shmflg)
471{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700472 struct ipc_namespace *ns;
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700473 struct ipc_ops shm_ops;
474 struct ipc_params shm_params;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475
Kirill Korotaev4e982312006-10-02 02:18:22 -0700476 ns = current->nsproxy->ipc_ns;
477
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700478 shm_ops.getnew = newseg;
479 shm_ops.associate = shm_security;
480 shm_ops.more_checks = shm_more_checks;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700481
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700482 shm_params.key = key;
483 shm_params.flg = shmflg;
484 shm_params.u.size = size;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700485
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700486 return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487}
488
489static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version)
490{
491 switch(version) {
492 case IPC_64:
493 return copy_to_user(buf, in, sizeof(*in));
494 case IPC_OLD:
495 {
496 struct shmid_ds out;
497
498 ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm);
499 out.shm_segsz = in->shm_segsz;
500 out.shm_atime = in->shm_atime;
501 out.shm_dtime = in->shm_dtime;
502 out.shm_ctime = in->shm_ctime;
503 out.shm_cpid = in->shm_cpid;
504 out.shm_lpid = in->shm_lpid;
505 out.shm_nattch = in->shm_nattch;
506
507 return copy_to_user(buf, &out, sizeof(out));
508 }
509 default:
510 return -EINVAL;
511 }
512}
513
Pierre Peiffer016d7132008-04-29 01:00:50 -0700514static inline unsigned long
515copy_shmid_from_user(struct shmid64_ds *out, void __user *buf, int version)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516{
517 switch(version) {
518 case IPC_64:
Pierre Peiffer016d7132008-04-29 01:00:50 -0700519 if (copy_from_user(out, buf, sizeof(*out)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 case IPC_OLD:
523 {
524 struct shmid_ds tbuf_old;
525
526 if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old)))
527 return -EFAULT;
528
Pierre Peiffer016d7132008-04-29 01:00:50 -0700529 out->shm_perm.uid = tbuf_old.shm_perm.uid;
530 out->shm_perm.gid = tbuf_old.shm_perm.gid;
531 out->shm_perm.mode = tbuf_old.shm_perm.mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532
533 return 0;
534 }
535 default:
536 return -EINVAL;
537 }
538}
539
540static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version)
541{
542 switch(version) {
543 case IPC_64:
544 return copy_to_user(buf, in, sizeof(*in));
545 case IPC_OLD:
546 {
547 struct shminfo out;
548
549 if(in->shmmax > INT_MAX)
550 out.shmmax = INT_MAX;
551 else
552 out.shmmax = (int)in->shmmax;
553
554 out.shmmin = in->shmmin;
555 out.shmmni = in->shmmni;
556 out.shmseg = in->shmseg;
557 out.shmall = in->shmall;
558
559 return copy_to_user(buf, &out, sizeof(out));
560 }
561 default:
562 return -EINVAL;
563 }
564}
565
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700566/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700567 * Called with shm_ids.rw_mutex held as a reader
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700568 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700569static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss,
570 unsigned long *swp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700572 int next_id;
573 int total, in_use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574
575 *rss = 0;
576 *swp = 0;
577
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700578 in_use = shm_ids(ns).in_use;
579
580 for (total = 0, next_id = 0; total < in_use; next_id++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 struct shmid_kernel *shp;
582 struct inode *inode;
583
Nadia Derbey637c3662007-10-18 23:40:50 -0700584 shp = idr_find(&shm_ids(ns).ipcs_idr, next_id);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700585 if (shp == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 continue;
587
Josef Sipek6d630792006-12-08 02:37:11 -0800588 inode = shp->shm_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589
590 if (is_file_hugepages(shp->shm_file)) {
591 struct address_space *mapping = inode->i_mapping;
592 *rss += (HPAGE_SIZE/PAGE_SIZE)*mapping->nrpages;
593 } else {
594 struct shmem_inode_info *info = SHMEM_I(inode);
595 spin_lock(&info->lock);
596 *rss += inode->i_mapping->nrpages;
597 *swp += info->swapped;
598 spin_unlock(&info->lock);
599 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700600
601 total++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 }
603}
604
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700605/*
606 * This function handles some shmctl commands which require the rw_mutex
607 * to be held in write mode.
608 * NOTE: no locks must be held, the rw_mutex is taken inside this function.
609 */
610static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd,
611 struct shmid_ds __user *buf, int version)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612{
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700613 struct kern_ipc_perm *ipcp;
Pierre Peiffer016d7132008-04-29 01:00:50 -0700614 struct shmid64_ds shmid64;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 struct shmid_kernel *shp;
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700616 int err;
617
618 if (cmd == IPC_SET) {
Pierre Peiffer016d7132008-04-29 01:00:50 -0700619 if (copy_shmid_from_user(&shmid64, buf, version))
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700620 return -EFAULT;
621 }
622
623 down_write(&shm_ids(ns).rw_mutex);
624 shp = shm_lock_check_down(ns, shmid);
625 if (IS_ERR(shp)) {
626 err = PTR_ERR(shp);
627 goto out_up;
628 }
629
630 ipcp = &shp->shm_perm;
631
632 err = audit_ipc_obj(ipcp);
633 if (err)
634 goto out_unlock;
635
636 if (cmd == IPC_SET) {
Pierre Peiffer016d7132008-04-29 01:00:50 -0700637 err = audit_ipc_set_perm(0, shmid64.shm_perm.uid,
638 shmid64.shm_perm.gid,
639 shmid64.shm_perm.mode);
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700640 if (err)
641 goto out_unlock;
642 }
643
644 if (current->euid != ipcp->uid &&
645 current->euid != ipcp->cuid &&
646 !capable(CAP_SYS_ADMIN)) {
647 err = -EPERM;
648 goto out_unlock;
649 }
650
651 err = security_shm_shmctl(shp, cmd);
652 if (err)
653 goto out_unlock;
654 switch (cmd) {
655 case IPC_RMID:
656 do_shm_rmid(ns, ipcp);
657 goto out_up;
658 case IPC_SET:
Pierre Peiffer8f4a3802008-04-29 01:00:51 -0700659 ipc_update_perm(&shmid64.shm_perm, ipcp);
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700660 shp->shm_ctim = get_seconds();
661 break;
662 default:
663 err = -EINVAL;
664 }
665out_unlock:
666 shm_unlock(shp);
667out_up:
668 up_write(&shm_ids(ns).rw_mutex);
669 return err;
670}
671
672asmlinkage long sys_shmctl(int shmid, int cmd, struct shmid_ds __user *buf)
673{
674 struct shmid_kernel *shp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 int err, version;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700676 struct ipc_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677
678 if (cmd < 0 || shmid < 0) {
679 err = -EINVAL;
680 goto out;
681 }
682
683 version = ipc_parse_version(&cmd);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700684 ns = current->nsproxy->ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685
686 switch (cmd) { /* replace with proc interface ? */
687 case IPC_INFO:
688 {
689 struct shminfo64 shminfo;
690
691 err = security_shm_shmctl(NULL, cmd);
692 if (err)
693 return err;
694
695 memset(&shminfo,0,sizeof(shminfo));
Kirill Korotaev4e982312006-10-02 02:18:22 -0700696 shminfo.shmmni = shminfo.shmseg = ns->shm_ctlmni;
697 shminfo.shmmax = ns->shm_ctlmax;
698 shminfo.shmall = ns->shm_ctlall;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699
700 shminfo.shmmin = SHMMIN;
701 if(copy_shminfo_to_user (buf, &shminfo, version))
702 return -EFAULT;
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700703
Nadia Derbey3e148c72007-10-18 23:40:54 -0700704 down_read(&shm_ids(ns).rw_mutex);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700705 err = ipc_get_maxid(&shm_ids(ns));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700706 up_read(&shm_ids(ns).rw_mutex);
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700707
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 if(err<0)
709 err = 0;
710 goto out;
711 }
712 case SHM_INFO:
713 {
714 struct shm_info shm_info;
715
716 err = security_shm_shmctl(NULL, cmd);
717 if (err)
718 return err;
719
720 memset(&shm_info,0,sizeof(shm_info));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700721 down_read(&shm_ids(ns).rw_mutex);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700722 shm_info.used_ids = shm_ids(ns).in_use;
723 shm_get_stat (ns, &shm_info.shm_rss, &shm_info.shm_swp);
724 shm_info.shm_tot = ns->shm_tot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 shm_info.swap_attempts = 0;
726 shm_info.swap_successes = 0;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700727 err = ipc_get_maxid(&shm_ids(ns));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700728 up_read(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 if(copy_to_user (buf, &shm_info, sizeof(shm_info))) {
730 err = -EFAULT;
731 goto out;
732 }
733
734 err = err < 0 ? 0 : err;
735 goto out;
736 }
737 case SHM_STAT:
738 case IPC_STAT:
739 {
740 struct shmid64_ds tbuf;
741 int result;
Nadia Derbey023a5352007-10-18 23:40:51 -0700742
743 if (!buf) {
744 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 goto out;
Nadia Derbey023a5352007-10-18 23:40:51 -0700746 }
747
748 if (cmd == SHM_STAT) {
749 shp = shm_lock(ns, shmid);
750 if (IS_ERR(shp)) {
751 err = PTR_ERR(shp);
752 goto out;
753 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700754 result = shp->shm_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 } else {
Nadia Derbey023a5352007-10-18 23:40:51 -0700756 shp = shm_lock_check(ns, shmid);
757 if (IS_ERR(shp)) {
758 err = PTR_ERR(shp);
759 goto out;
760 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 result = 0;
762 }
763 err=-EACCES;
764 if (ipcperms (&shp->shm_perm, S_IRUGO))
765 goto out_unlock;
766 err = security_shm_shmctl(shp, cmd);
767 if (err)
768 goto out_unlock;
Nadia Derbey023a5352007-10-18 23:40:51 -0700769 memset(&tbuf, 0, sizeof(tbuf));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 kernel_to_ipc64_perm(&shp->shm_perm, &tbuf.shm_perm);
771 tbuf.shm_segsz = shp->shm_segsz;
772 tbuf.shm_atime = shp->shm_atim;
773 tbuf.shm_dtime = shp->shm_dtim;
774 tbuf.shm_ctime = shp->shm_ctim;
775 tbuf.shm_cpid = shp->shm_cprid;
776 tbuf.shm_lpid = shp->shm_lprid;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800777 tbuf.shm_nattch = shp->shm_nattch;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778 shm_unlock(shp);
779 if(copy_shmid_to_user (buf, &tbuf, version))
780 err = -EFAULT;
781 else
782 err = result;
783 goto out;
784 }
785 case SHM_LOCK:
786 case SHM_UNLOCK:
787 {
Nadia Derbey023a5352007-10-18 23:40:51 -0700788 shp = shm_lock_check(ns, shmid);
789 if (IS_ERR(shp)) {
790 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 goto out;
792 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793
Steve Grubb073115d2006-04-02 17:07:33 -0400794 err = audit_ipc_obj(&(shp->shm_perm));
795 if (err)
796 goto out_unlock;
797
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 if (!capable(CAP_IPC_LOCK)) {
799 err = -EPERM;
800 if (current->euid != shp->shm_perm.uid &&
801 current->euid != shp->shm_perm.cuid)
802 goto out_unlock;
803 if (cmd == SHM_LOCK &&
804 !current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur)
805 goto out_unlock;
806 }
807
808 err = security_shm_shmctl(shp, cmd);
809 if (err)
810 goto out_unlock;
811
812 if(cmd==SHM_LOCK) {
813 struct user_struct * user = current->user;
814 if (!is_file_hugepages(shp->shm_file)) {
815 err = shmem_lock(shp->shm_file, 1, user);
Pavel Emelianov7be77e22007-07-31 00:38:48 -0700816 if (!err && !(shp->shm_perm.mode & SHM_LOCKED)){
Andrew Mortonb33291c2006-01-08 01:02:21 -0800817 shp->shm_perm.mode |= SHM_LOCKED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 shp->mlock_user = user;
819 }
820 }
821 } else if (!is_file_hugepages(shp->shm_file)) {
822 shmem_lock(shp->shm_file, 0, shp->mlock_user);
Andrew Mortonb33291c2006-01-08 01:02:21 -0800823 shp->shm_perm.mode &= ~SHM_LOCKED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 shp->mlock_user = NULL;
825 }
826 shm_unlock(shp);
827 goto out;
828 }
829 case IPC_RMID:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 case IPC_SET:
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700831 err = shmctl_down(ns, shmid, cmd, buf, version);
832 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 default:
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700834 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 }
836
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837out_unlock:
838 shm_unlock(shp);
839out:
840 return err;
841}
842
843/*
844 * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists.
845 *
846 * NOTE! Despite the name, this is NOT a direct system call entrypoint. The
847 * "raddr" thing points to kernel space, and there has to be a wrapper around
848 * this.
849 */
850long do_shmat(int shmid, char __user *shmaddr, int shmflg, ulong *raddr)
851{
852 struct shmid_kernel *shp;
853 unsigned long addr;
854 unsigned long size;
855 struct file * file;
856 int err;
857 unsigned long flags;
858 unsigned long prot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 int acc_mode;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800860 unsigned long user_addr;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700861 struct ipc_namespace *ns;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800862 struct shm_file_data *sfd;
863 struct path path;
864 mode_t f_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800866 err = -EINVAL;
867 if (shmid < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 goto out;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800869 else if ((addr = (ulong)shmaddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870 if (addr & (SHMLBA-1)) {
871 if (shmflg & SHM_RND)
872 addr &= ~(SHMLBA-1); /* round down */
873 else
874#ifndef __ARCH_FORCE_SHMLBA
875 if (addr & ~PAGE_MASK)
876#endif
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800877 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 }
879 flags = MAP_SHARED | MAP_FIXED;
880 } else {
881 if ((shmflg & SHM_REMAP))
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800882 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883
884 flags = MAP_SHARED;
885 }
886
887 if (shmflg & SHM_RDONLY) {
888 prot = PROT_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 acc_mode = S_IRUGO;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800890 f_mode = FMODE_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 } else {
892 prot = PROT_READ | PROT_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 acc_mode = S_IRUGO | S_IWUGO;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800894 f_mode = FMODE_READ | FMODE_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 }
896 if (shmflg & SHM_EXEC) {
897 prot |= PROT_EXEC;
898 acc_mode |= S_IXUGO;
899 }
900
901 /*
902 * We cannot rely on the fs check since SYSV IPC does have an
903 * additional creator id...
904 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700905 ns = current->nsproxy->ipc_ns;
Nadia Derbey023a5352007-10-18 23:40:51 -0700906 shp = shm_lock_check(ns, shmid);
907 if (IS_ERR(shp)) {
908 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 goto out;
Nadia Derbey023a5352007-10-18 23:40:51 -0700910 }
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800911
912 err = -EACCES;
913 if (ipcperms(&shp->shm_perm, acc_mode))
914 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915
916 err = security_shm_shmat(shp, shmaddr, shmflg);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800917 if (err)
918 goto out_unlock;
919
920 path.dentry = dget(shp->shm_file->f_path.dentry);
Dave Hansence8d2cd2007-10-16 23:31:13 -0700921 path.mnt = shp->shm_file->f_path.mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 shp->shm_nattch++;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800923 size = i_size_read(path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 shm_unlock(shp);
925
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800926 err = -ENOMEM;
927 sfd = kzalloc(sizeof(*sfd), GFP_KERNEL);
928 if (!sfd)
Dave Hansence8d2cd2007-10-16 23:31:13 -0700929 goto out_put_dentry;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800930
931 err = -ENOMEM;
Dave Hansence8d2cd2007-10-16 23:31:13 -0700932
933 file = alloc_file(path.mnt, path.dentry, f_mode, &shm_file_operations);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800934 if (!file)
935 goto out_free;
936
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800937 file->private_data = sfd;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800938 file->f_mapping = shp->shm_file->f_mapping;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700939 sfd->id = shp->shm_perm.id;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800940 sfd->ns = get_ipc_ns(ns);
941 sfd->file = shp->shm_file;
942 sfd->vm_ops = NULL;
943
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 down_write(&current->mm->mmap_sem);
945 if (addr && !(shmflg & SHM_REMAP)) {
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800946 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 if (find_vma_intersection(current->mm, addr, addr + size))
948 goto invalid;
949 /*
950 * If shm segment goes below stack, make sure there is some
951 * space left for the stack to grow (at least 4 pages).
952 */
953 if (addr < current->mm->start_stack &&
954 addr > current->mm->start_stack - size - PAGE_SIZE * 5)
955 goto invalid;
956 }
957
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800958 user_addr = do_mmap (file, addr, size, prot, flags, 0);
959 *raddr = user_addr;
960 err = 0;
961 if (IS_ERR_VALUE(user_addr))
962 err = (long)user_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963invalid:
964 up_write(&current->mm->mmap_sem);
965
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800966 fput(file);
967
968out_nattch:
Nadia Derbey3e148c72007-10-18 23:40:54 -0700969 down_write(&shm_ids(ns).rw_mutex);
970 shp = shm_lock_down(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -0700971 BUG_ON(IS_ERR(shp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 shp->shm_nattch--;
973 if(shp->shm_nattch == 0 &&
Andrew Mortonb33291c2006-01-08 01:02:21 -0800974 shp->shm_perm.mode & SHM_DEST)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700975 shm_destroy(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976 else
977 shm_unlock(shp);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700978 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980out:
981 return err;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800982
983out_unlock:
984 shm_unlock(shp);
985 goto out;
986
987out_free:
988 kfree(sfd);
Dave Hansence8d2cd2007-10-16 23:31:13 -0700989out_put_dentry:
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800990 dput(path.dentry);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800991 goto out_nattch;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992}
993
Stephen Rothwell7d87e142005-05-01 08:59:12 -0700994asmlinkage long sys_shmat(int shmid, char __user *shmaddr, int shmflg)
995{
996 unsigned long ret;
997 long err;
998
999 err = do_shmat(shmid, shmaddr, shmflg, &ret);
1000 if (err)
1001 return err;
1002 force_successful_syscall_return();
1003 return (long)ret;
1004}
1005
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006/*
1007 * detach and kill segment if marked destroyed.
1008 * The work is done in shm_close.
1009 */
1010asmlinkage long sys_shmdt(char __user *shmaddr)
1011{
1012 struct mm_struct *mm = current->mm;
1013 struct vm_area_struct *vma, *next;
1014 unsigned long addr = (unsigned long)shmaddr;
1015 loff_t size = 0;
1016 int retval = -EINVAL;
1017
Hugh Dickinsdf1e2fb2006-03-24 03:18:06 -08001018 if (addr & ~PAGE_MASK)
1019 return retval;
1020
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 down_write(&mm->mmap_sem);
1022
1023 /*
1024 * This function tries to be smart and unmap shm segments that
1025 * were modified by partial mlock or munmap calls:
1026 * - It first determines the size of the shm segment that should be
1027 * unmapped: It searches for a vma that is backed by shm and that
1028 * started at address shmaddr. It records it's size and then unmaps
1029 * it.
1030 * - Then it unmaps all shm vmas that started at shmaddr and that
1031 * are within the initially determined size.
1032 * Errors from do_munmap are ignored: the function only fails if
1033 * it's called with invalid parameters or if it's called to unmap
1034 * a part of a vma. Both calls in this function are for full vmas,
1035 * the parameters are directly copied from the vma itself and always
1036 * valid - therefore do_munmap cannot fail. (famous last words?)
1037 */
1038 /*
1039 * If it had been mremap()'d, the starting address would not
1040 * match the usual checks anyway. So assume all vma's are
1041 * above the starting address given.
1042 */
1043 vma = find_vma(mm, addr);
1044
1045 while (vma) {
1046 next = vma->vm_next;
1047
1048 /*
1049 * Check if the starting address would match, i.e. it's
1050 * a fragment created by mprotect() and/or munmap(), or it
1051 * otherwise it starts at this address with no hassles.
1052 */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001053 if ((vma->vm_ops == &shm_vm_ops) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054 (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) {
1055
1056
Josef Sipek6d630792006-12-08 02:37:11 -08001057 size = vma->vm_file->f_path.dentry->d_inode->i_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1059 /*
1060 * We discovered the size of the shm segment, so
1061 * break out of here and fall through to the next
1062 * loop that uses the size information to stop
1063 * searching for matching vma's.
1064 */
1065 retval = 0;
1066 vma = next;
1067 break;
1068 }
1069 vma = next;
1070 }
1071
1072 /*
1073 * We need look no further than the maximum address a fragment
1074 * could possibly have landed at. Also cast things to loff_t to
1075 * prevent overflows and make comparisions vs. equal-width types.
1076 */
KAMEZAWA Hiroyuki8e367092006-02-10 01:51:12 -08001077 size = PAGE_ALIGN(size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 while (vma && (loff_t)(vma->vm_end - addr) <= size) {
1079 next = vma->vm_next;
1080
1081 /* finding a matching vma now does not alter retval */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001082 if ((vma->vm_ops == &shm_vm_ops) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff)
1084
1085 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1086 vma = next;
1087 }
1088
1089 up_write(&mm->mmap_sem);
1090 return retval;
1091}
1092
1093#ifdef CONFIG_PROC_FS
Mike Waychison19b49462005-09-06 15:17:10 -07001094static int sysvipc_shm_proc_show(struct seq_file *s, void *it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095{
Mike Waychison19b49462005-09-06 15:17:10 -07001096 struct shmid_kernel *shp = it;
1097 char *format;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099#define SMALL_STRING "%10d %10d %4o %10u %5u %5u %5d %5u %5u %5u %5u %10lu %10lu %10lu\n"
1100#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 -07001101
Mike Waychison19b49462005-09-06 15:17:10 -07001102 if (sizeof(size_t) <= sizeof(int))
1103 format = SMALL_STRING;
1104 else
1105 format = BIG_STRING;
1106 return seq_printf(s, format,
1107 shp->shm_perm.key,
Nadia Derbey7ca7e562007-10-18 23:40:48 -07001108 shp->shm_perm.id,
Andrew Mortonb33291c2006-01-08 01:02:21 -08001109 shp->shm_perm.mode,
Mike Waychison19b49462005-09-06 15:17:10 -07001110 shp->shm_segsz,
1111 shp->shm_cprid,
1112 shp->shm_lprid,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001113 shp->shm_nattch,
Mike Waychison19b49462005-09-06 15:17:10 -07001114 shp->shm_perm.uid,
1115 shp->shm_perm.gid,
1116 shp->shm_perm.cuid,
1117 shp->shm_perm.cgid,
1118 shp->shm_atim,
1119 shp->shm_dtim,
1120 shp->shm_ctim);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121}
1122#endif