blob: 554429ade079ffbcd8b19fbfd657772db2eb36ec [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
Nadia Derbey3e148c72007-10-18 23:40:54 -0700129/*
130 * shm_lock_(check_) routines are called in the paths where the rw_mutex
131 * is not held.
132 */
Nadia Derbey023a5352007-10-18 23:40:51 -0700133static inline struct shmid_kernel *shm_lock(struct ipc_namespace *ns, int id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700135 struct kern_ipc_perm *ipcp = ipc_lock(&shm_ids(ns), id);
136
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800137 if (IS_ERR(ipcp))
138 return (struct shmid_kernel *)ipcp;
139
Nadia Derbey03f02c72007-10-18 23:40:51 -0700140 return container_of(ipcp, struct shmid_kernel, shm_perm);
Nadia Derbey023a5352007-10-18 23:40:51 -0700141}
142
143static inline struct shmid_kernel *shm_lock_check(struct ipc_namespace *ns,
144 int id)
145{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700146 struct kern_ipc_perm *ipcp = ipc_lock_check(&shm_ids(ns), id);
147
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800148 if (IS_ERR(ipcp))
149 return (struct shmid_kernel *)ipcp;
150
Nadia Derbey03f02c72007-10-18 23:40:51 -0700151 return container_of(ipcp, struct shmid_kernel, shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152}
153
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700154static inline void shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *s)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700156 ipc_rmid(&shm_ids(ns), &s->shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157}
158
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800160/* This is called by fork, once for every shm attach. */
161static void shm_open(struct vm_area_struct *vma)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700162{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800163 struct file *file = vma->vm_file;
164 struct shm_file_data *sfd = shm_file_data(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 struct shmid_kernel *shp;
166
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800167 shp = shm_lock(sfd->ns, sfd->id);
Nadia Derbey023a5352007-10-18 23:40:51 -0700168 BUG_ON(IS_ERR(shp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 shp->shm_atim = get_seconds();
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700170 shp->shm_lprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 shp->shm_nattch++;
172 shm_unlock(shp);
173}
174
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175/*
176 * shm_destroy - free the struct shmid_kernel
177 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700178 * @ns: namespace
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 * @shp: struct to free
180 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700181 * It has to be called with shp and shm_ids.rw_mutex (writer) locked,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 * but returns with shp unlocked and freed.
183 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700184static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700186 ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700187 shm_rmid(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 shm_unlock(shp);
189 if (!is_file_hugepages(shp->shm_file))
190 shmem_lock(shp->shm_file, 0, shp->mlock_user);
191 else
Josef Sipek6d630792006-12-08 02:37:11 -0800192 user_shm_unlock(shp->shm_file->f_path.dentry->d_inode->i_size,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 shp->mlock_user);
194 fput (shp->shm_file);
195 security_shm_free(shp);
196 ipc_rcu_putref(shp);
197}
198
199/*
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800200 * remove the attach descriptor vma.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 * free memory for segment if it is marked destroyed.
202 * The descriptor has already been removed from the current->mm->mmap list
203 * and will later be kfree()d.
204 */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800205static void shm_close(struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800207 struct file * file = vma->vm_file;
208 struct shm_file_data *sfd = shm_file_data(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 struct shmid_kernel *shp;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800210 struct ipc_namespace *ns = sfd->ns;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700211
Nadia Derbey3e148c72007-10-18 23:40:54 -0700212 down_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 /* remove from the list of attaches of the shm segment */
Nadia Derbey3e148c72007-10-18 23:40:54 -0700214 shp = shm_lock_down(ns, sfd->id);
Nadia Derbey023a5352007-10-18 23:40:51 -0700215 BUG_ON(IS_ERR(shp));
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700216 shp->shm_lprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 shp->shm_dtim = get_seconds();
218 shp->shm_nattch--;
219 if(shp->shm_nattch == 0 &&
Andrew Mortonb33291c2006-01-08 01:02:21 -0800220 shp->shm_perm.mode & SHM_DEST)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700221 shm_destroy(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 else
223 shm_unlock(shp);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700224 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225}
226
Nick Piggind0217ac2007-07-19 01:47:03 -0700227static int shm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800228{
229 struct file *file = vma->vm_file;
230 struct shm_file_data *sfd = shm_file_data(file);
231
Nick Piggind0217ac2007-07-19 01:47:03 -0700232 return sfd->vm_ops->fault(vma, vmf);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800233}
234
235#ifdef CONFIG_NUMA
Adrian Bunkd823e3e2007-10-16 23:26:42 -0700236static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800237{
238 struct file *file = vma->vm_file;
239 struct shm_file_data *sfd = shm_file_data(file);
240 int err = 0;
241 if (sfd->vm_ops->set_policy)
242 err = sfd->vm_ops->set_policy(vma, new);
243 return err;
244}
245
Adrian Bunkd823e3e2007-10-16 23:26:42 -0700246static struct mempolicy *shm_get_policy(struct vm_area_struct *vma,
247 unsigned long addr)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800248{
249 struct file *file = vma->vm_file;
250 struct shm_file_data *sfd = shm_file_data(file);
251 struct mempolicy *pol = NULL;
252
253 if (sfd->vm_ops->get_policy)
254 pol = sfd->vm_ops->get_policy(vma, addr);
Lee Schermerhorn52cd3b02008-04-28 02:13:16 -0700255 else if (vma->vm_policy)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800256 pol = vma->vm_policy;
Lee Schermerhorn52cd3b02008-04-28 02:13:16 -0700257
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800258 return pol;
259}
260#endif
261
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262static int shm_mmap(struct file * file, struct vm_area_struct * vma)
263{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800264 struct shm_file_data *sfd = shm_file_data(file);
David Howellsb0e15192006-01-06 00:11:42 -0800265 int ret;
266
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800267 ret = sfd->file->f_op->mmap(sfd->file, vma);
268 if (ret != 0)
269 return ret;
270 sfd->vm_ops = vma->vm_ops;
David Howells2e92a3b2007-07-31 00:37:24 -0700271#ifdef CONFIG_MMU
Nick Piggin54cb8822007-07-19 01:46:59 -0700272 BUG_ON(!sfd->vm_ops->fault);
David Howells2e92a3b2007-07-31 00:37:24 -0700273#endif
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800274 vma->vm_ops = &shm_vm_ops;
275 shm_open(vma);
David Howellsb0e15192006-01-06 00:11:42 -0800276
277 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278}
279
Kirill Korotaev4e982312006-10-02 02:18:22 -0700280static int shm_release(struct inode *ino, struct file *file)
281{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800282 struct shm_file_data *sfd = shm_file_data(file);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700283
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800284 put_ipc_ns(sfd->ns);
285 shm_file_data(file) = NULL;
286 kfree(sfd);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700287 return 0;
288}
289
Adam Litke516dffd2007-03-01 15:46:08 -0800290static int shm_fsync(struct file *file, struct dentry *dentry, int datasync)
291{
292 int (*fsync) (struct file *, struct dentry *, int datasync);
293 struct shm_file_data *sfd = shm_file_data(file);
294 int ret = -EINVAL;
295
296 fsync = sfd->file->f_op->fsync;
297 if (fsync)
298 ret = fsync(sfd->file, sfd->file->f_path.dentry, datasync);
299 return ret;
300}
301
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800302static unsigned long shm_get_unmapped_area(struct file *file,
303 unsigned long addr, unsigned long len, unsigned long pgoff,
304 unsigned long flags)
305{
306 struct shm_file_data *sfd = shm_file_data(file);
Adam Litke516dffd2007-03-01 15:46:08 -0800307 return get_unmapped_area(sfd->file, addr, len, pgoff, flags);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800308}
Adam Litke516dffd2007-03-01 15:46:08 -0800309
310int is_file_shm_hugepages(struct file *file)
311{
312 int ret = 0;
313
314 if (file->f_op == &shm_file_operations) {
315 struct shm_file_data *sfd;
316 sfd = shm_file_data(file);
317 ret = is_file_hugepages(sfd->file);
318 }
319 return ret;
320}
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800321
Arjan van de Ven9a321442007-02-12 00:55:35 -0800322static const struct file_operations shm_file_operations = {
Kirill Korotaev4e982312006-10-02 02:18:22 -0700323 .mmap = shm_mmap,
Adam Litke516dffd2007-03-01 15:46:08 -0800324 .fsync = shm_fsync,
Kirill Korotaev4e982312006-10-02 02:18:22 -0700325 .release = shm_release,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800326 .get_unmapped_area = shm_get_unmapped_area,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327};
328
329static struct vm_operations_struct shm_vm_ops = {
330 .open = shm_open, /* callback for a new vm-area open */
331 .close = shm_close, /* callback for when the vm-area is released */
Nick Piggin54cb8822007-07-19 01:46:59 -0700332 .fault = shm_fault,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800333#if defined(CONFIG_NUMA)
334 .set_policy = shm_set_policy,
335 .get_policy = shm_get_policy,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336#endif
337};
338
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700339/**
340 * newseg - Create a new shared memory segment
341 * @ns: namespace
342 * @params: ptr to the structure that contains key, size and shmflg
343 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700344 * Called with shm_ids.rw_mutex held as a writer.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700345 */
346
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700347static int newseg(struct ipc_namespace *ns, struct ipc_params *params)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348{
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700349 key_t key = params->key;
350 int shmflg = params->flg;
351 size_t size = params->u.size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 int error;
353 struct shmid_kernel *shp;
354 int numpages = (size + PAGE_SIZE -1) >> PAGE_SHIFT;
355 struct file * file;
356 char name[13];
357 int id;
358
Kirill Korotaev4e982312006-10-02 02:18:22 -0700359 if (size < SHMMIN || size > ns->shm_ctlmax)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 return -EINVAL;
361
Guy Streeterf66d45e2007-01-23 12:20:04 -0600362 if (ns->shm_tot + numpages > ns->shm_ctlall)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 return -ENOSPC;
364
365 shp = ipc_rcu_alloc(sizeof(*shp));
366 if (!shp)
367 return -ENOMEM;
368
369 shp->shm_perm.key = key;
Andrew Mortonb33291c2006-01-08 01:02:21 -0800370 shp->shm_perm.mode = (shmflg & S_IRWXUGO);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 shp->mlock_user = NULL;
372
373 shp->shm_perm.security = NULL;
374 error = security_shm_alloc(shp);
375 if (error) {
376 ipc_rcu_putref(shp);
377 return error;
378 }
379
Eric W. Biederman9d665862007-06-16 10:16:16 -0700380 sprintf (name, "SYSV%08x", key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 if (shmflg & SHM_HUGETLB) {
Eric W. Biederman9d665862007-06-16 10:16:16 -0700382 /* hugetlb_file_setup takes care of mlock user accounting */
383 file = hugetlb_file_setup(name, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 shp->mlock_user = current->user;
385 } else {
Badari Pulavartybf8f9722005-11-07 00:59:27 -0800386 int acctflag = VM_ACCOUNT;
387 /*
388 * Do not allow no accounting for OVERCOMMIT_NEVER, even
389 * if it's asked for.
390 */
391 if ((shmflg & SHM_NORESERVE) &&
392 sysctl_overcommit_memory != OVERCOMMIT_NEVER)
393 acctflag = 0;
Badari Pulavartybf8f9722005-11-07 00:59:27 -0800394 file = shmem_file_setup(name, size, acctflag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 }
396 error = PTR_ERR(file);
397 if (IS_ERR(file))
398 goto no_file;
399
Pierre Peiffer48dea402008-04-29 01:00:35 -0700400 id = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni);
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700401 if (id < 0) {
402 error = id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 goto no_id;
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700404 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700406 shp->shm_cprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 shp->shm_lprid = 0;
408 shp->shm_atim = shp->shm_dtim = 0;
409 shp->shm_ctim = get_seconds();
410 shp->shm_segsz = size;
411 shp->shm_nattch = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 shp->shm_file = file;
Badari Pulavarty30475cc2007-06-16 10:15:59 -0700413 /*
414 * shmid gets reported as "inode#" in /proc/pid/maps.
415 * proc-ps tools use this. Changing this will break them.
416 */
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700417 file->f_dentry->d_inode->i_ino = shp->shm_perm.id;
Krishnakumar R551110a2005-10-29 18:16:45 -0700418
Kirill Korotaev4e982312006-10-02 02:18:22 -0700419 ns->shm_tot += numpages;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700420 error = shp->shm_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 shm_unlock(shp);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700422 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423
424no_id:
425 fput(file);
426no_file:
427 security_shm_free(shp);
428 ipc_rcu_putref(shp);
429 return error;
430}
431
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700432/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700433 * Called with shm_ids.rw_mutex and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700434 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700435static inline int shm_security(struct kern_ipc_perm *ipcp, int shmflg)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700436{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700437 struct shmid_kernel *shp;
438
439 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
440 return security_shm_associate(shp, shmflg);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700441}
442
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700443/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700444 * Called with shm_ids.rw_mutex and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700445 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700446static inline int shm_more_checks(struct kern_ipc_perm *ipcp,
447 struct ipc_params *params)
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 if (shp->shm_segsz < params->u.size)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700453 return -EINVAL;
454
455 return 0;
456}
457
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458asmlinkage long sys_shmget (key_t key, size_t size, int shmflg)
459{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700460 struct ipc_namespace *ns;
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700461 struct ipc_ops shm_ops;
462 struct ipc_params shm_params;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463
Kirill Korotaev4e982312006-10-02 02:18:22 -0700464 ns = current->nsproxy->ipc_ns;
465
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700466 shm_ops.getnew = newseg;
467 shm_ops.associate = shm_security;
468 shm_ops.more_checks = shm_more_checks;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700469
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700470 shm_params.key = key;
471 shm_params.flg = shmflg;
472 shm_params.u.size = size;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700473
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700474 return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475}
476
477static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version)
478{
479 switch(version) {
480 case IPC_64:
481 return copy_to_user(buf, in, sizeof(*in));
482 case IPC_OLD:
483 {
484 struct shmid_ds out;
485
486 ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm);
487 out.shm_segsz = in->shm_segsz;
488 out.shm_atime = in->shm_atime;
489 out.shm_dtime = in->shm_dtime;
490 out.shm_ctime = in->shm_ctime;
491 out.shm_cpid = in->shm_cpid;
492 out.shm_lpid = in->shm_lpid;
493 out.shm_nattch = in->shm_nattch;
494
495 return copy_to_user(buf, &out, sizeof(out));
496 }
497 default:
498 return -EINVAL;
499 }
500}
501
Pierre Peiffer016d7132008-04-29 01:00:50 -0700502static inline unsigned long
503copy_shmid_from_user(struct shmid64_ds *out, void __user *buf, int version)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504{
505 switch(version) {
506 case IPC_64:
Pierre Peiffer016d7132008-04-29 01:00:50 -0700507 if (copy_from_user(out, buf, sizeof(*out)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 case IPC_OLD:
511 {
512 struct shmid_ds tbuf_old;
513
514 if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old)))
515 return -EFAULT;
516
Pierre Peiffer016d7132008-04-29 01:00:50 -0700517 out->shm_perm.uid = tbuf_old.shm_perm.uid;
518 out->shm_perm.gid = tbuf_old.shm_perm.gid;
519 out->shm_perm.mode = tbuf_old.shm_perm.mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520
521 return 0;
522 }
523 default:
524 return -EINVAL;
525 }
526}
527
528static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version)
529{
530 switch(version) {
531 case IPC_64:
532 return copy_to_user(buf, in, sizeof(*in));
533 case IPC_OLD:
534 {
535 struct shminfo out;
536
537 if(in->shmmax > INT_MAX)
538 out.shmmax = INT_MAX;
539 else
540 out.shmmax = (int)in->shmmax;
541
542 out.shmmin = in->shmmin;
543 out.shmmni = in->shmmni;
544 out.shmseg = in->shmseg;
545 out.shmall = in->shmall;
546
547 return copy_to_user(buf, &out, sizeof(out));
548 }
549 default:
550 return -EINVAL;
551 }
552}
553
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700554/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700555 * Called with shm_ids.rw_mutex held as a reader
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700556 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700557static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss,
558 unsigned long *swp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700560 int next_id;
561 int total, in_use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562
563 *rss = 0;
564 *swp = 0;
565
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700566 in_use = shm_ids(ns).in_use;
567
568 for (total = 0, next_id = 0; total < in_use; next_id++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 struct shmid_kernel *shp;
570 struct inode *inode;
571
Nadia Derbey637c3662007-10-18 23:40:50 -0700572 shp = idr_find(&shm_ids(ns).ipcs_idr, next_id);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700573 if (shp == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 continue;
575
Josef Sipek6d630792006-12-08 02:37:11 -0800576 inode = shp->shm_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577
578 if (is_file_hugepages(shp->shm_file)) {
579 struct address_space *mapping = inode->i_mapping;
580 *rss += (HPAGE_SIZE/PAGE_SIZE)*mapping->nrpages;
581 } else {
582 struct shmem_inode_info *info = SHMEM_I(inode);
583 spin_lock(&info->lock);
584 *rss += inode->i_mapping->nrpages;
585 *swp += info->swapped;
586 spin_unlock(&info->lock);
587 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700588
589 total++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 }
591}
592
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700593/*
594 * This function handles some shmctl commands which require the rw_mutex
595 * to be held in write mode.
596 * NOTE: no locks must be held, the rw_mutex is taken inside this function.
597 */
598static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd,
599 struct shmid_ds __user *buf, int version)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600{
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700601 struct kern_ipc_perm *ipcp;
Pierre Peiffer016d7132008-04-29 01:00:50 -0700602 struct shmid64_ds shmid64;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 struct shmid_kernel *shp;
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700604 int err;
605
606 if (cmd == IPC_SET) {
Pierre Peiffer016d7132008-04-29 01:00:50 -0700607 if (copy_shmid_from_user(&shmid64, buf, version))
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700608 return -EFAULT;
609 }
610
Pierre Peiffera5f75e72008-04-29 01:00:54 -0700611 ipcp = ipcctl_pre_down(&shm_ids(ns), shmid, cmd, &shmid64.shm_perm, 0);
612 if (IS_ERR(ipcp))
613 return PTR_ERR(ipcp);
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700614
Pierre Peiffera5f75e72008-04-29 01:00:54 -0700615 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700616
617 err = security_shm_shmctl(shp, cmd);
618 if (err)
619 goto out_unlock;
620 switch (cmd) {
621 case IPC_RMID:
622 do_shm_rmid(ns, ipcp);
623 goto out_up;
624 case IPC_SET:
Pierre Peiffer8f4a3802008-04-29 01:00:51 -0700625 ipc_update_perm(&shmid64.shm_perm, ipcp);
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700626 shp->shm_ctim = get_seconds();
627 break;
628 default:
629 err = -EINVAL;
630 }
631out_unlock:
632 shm_unlock(shp);
633out_up:
634 up_write(&shm_ids(ns).rw_mutex);
635 return err;
636}
637
638asmlinkage long sys_shmctl(int shmid, int cmd, struct shmid_ds __user *buf)
639{
640 struct shmid_kernel *shp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 int err, version;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700642 struct ipc_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643
644 if (cmd < 0 || shmid < 0) {
645 err = -EINVAL;
646 goto out;
647 }
648
649 version = ipc_parse_version(&cmd);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700650 ns = current->nsproxy->ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651
652 switch (cmd) { /* replace with proc interface ? */
653 case IPC_INFO:
654 {
655 struct shminfo64 shminfo;
656
657 err = security_shm_shmctl(NULL, cmd);
658 if (err)
659 return err;
660
661 memset(&shminfo,0,sizeof(shminfo));
Kirill Korotaev4e982312006-10-02 02:18:22 -0700662 shminfo.shmmni = shminfo.shmseg = ns->shm_ctlmni;
663 shminfo.shmmax = ns->shm_ctlmax;
664 shminfo.shmall = ns->shm_ctlall;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665
666 shminfo.shmmin = SHMMIN;
667 if(copy_shminfo_to_user (buf, &shminfo, version))
668 return -EFAULT;
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700669
Nadia Derbey3e148c72007-10-18 23:40:54 -0700670 down_read(&shm_ids(ns).rw_mutex);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700671 err = ipc_get_maxid(&shm_ids(ns));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700672 up_read(&shm_ids(ns).rw_mutex);
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700673
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 if(err<0)
675 err = 0;
676 goto out;
677 }
678 case SHM_INFO:
679 {
680 struct shm_info shm_info;
681
682 err = security_shm_shmctl(NULL, cmd);
683 if (err)
684 return err;
685
686 memset(&shm_info,0,sizeof(shm_info));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700687 down_read(&shm_ids(ns).rw_mutex);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700688 shm_info.used_ids = shm_ids(ns).in_use;
689 shm_get_stat (ns, &shm_info.shm_rss, &shm_info.shm_swp);
690 shm_info.shm_tot = ns->shm_tot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 shm_info.swap_attempts = 0;
692 shm_info.swap_successes = 0;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700693 err = ipc_get_maxid(&shm_ids(ns));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700694 up_read(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 if(copy_to_user (buf, &shm_info, sizeof(shm_info))) {
696 err = -EFAULT;
697 goto out;
698 }
699
700 err = err < 0 ? 0 : err;
701 goto out;
702 }
703 case SHM_STAT:
704 case IPC_STAT:
705 {
706 struct shmid64_ds tbuf;
707 int result;
Nadia Derbey023a5352007-10-18 23:40:51 -0700708
709 if (!buf) {
710 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 goto out;
Nadia Derbey023a5352007-10-18 23:40:51 -0700712 }
713
714 if (cmd == SHM_STAT) {
715 shp = shm_lock(ns, shmid);
716 if (IS_ERR(shp)) {
717 err = PTR_ERR(shp);
718 goto out;
719 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700720 result = shp->shm_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 } else {
Nadia Derbey023a5352007-10-18 23:40:51 -0700722 shp = shm_lock_check(ns, shmid);
723 if (IS_ERR(shp)) {
724 err = PTR_ERR(shp);
725 goto out;
726 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 result = 0;
728 }
729 err=-EACCES;
730 if (ipcperms (&shp->shm_perm, S_IRUGO))
731 goto out_unlock;
732 err = security_shm_shmctl(shp, cmd);
733 if (err)
734 goto out_unlock;
Nadia Derbey023a5352007-10-18 23:40:51 -0700735 memset(&tbuf, 0, sizeof(tbuf));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 kernel_to_ipc64_perm(&shp->shm_perm, &tbuf.shm_perm);
737 tbuf.shm_segsz = shp->shm_segsz;
738 tbuf.shm_atime = shp->shm_atim;
739 tbuf.shm_dtime = shp->shm_dtim;
740 tbuf.shm_ctime = shp->shm_ctim;
741 tbuf.shm_cpid = shp->shm_cprid;
742 tbuf.shm_lpid = shp->shm_lprid;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800743 tbuf.shm_nattch = shp->shm_nattch;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 shm_unlock(shp);
745 if(copy_shmid_to_user (buf, &tbuf, version))
746 err = -EFAULT;
747 else
748 err = result;
749 goto out;
750 }
751 case SHM_LOCK:
752 case SHM_UNLOCK:
753 {
Nadia Derbey023a5352007-10-18 23:40:51 -0700754 shp = shm_lock_check(ns, shmid);
755 if (IS_ERR(shp)) {
756 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 goto out;
758 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759
Steve Grubb073115d2006-04-02 17:07:33 -0400760 err = audit_ipc_obj(&(shp->shm_perm));
761 if (err)
762 goto out_unlock;
763
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764 if (!capable(CAP_IPC_LOCK)) {
765 err = -EPERM;
766 if (current->euid != shp->shm_perm.uid &&
767 current->euid != shp->shm_perm.cuid)
768 goto out_unlock;
769 if (cmd == SHM_LOCK &&
770 !current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur)
771 goto out_unlock;
772 }
773
774 err = security_shm_shmctl(shp, cmd);
775 if (err)
776 goto out_unlock;
777
778 if(cmd==SHM_LOCK) {
779 struct user_struct * user = current->user;
780 if (!is_file_hugepages(shp->shm_file)) {
781 err = shmem_lock(shp->shm_file, 1, user);
Pavel Emelianov7be77e22007-07-31 00:38:48 -0700782 if (!err && !(shp->shm_perm.mode & SHM_LOCKED)){
Andrew Mortonb33291c2006-01-08 01:02:21 -0800783 shp->shm_perm.mode |= SHM_LOCKED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 shp->mlock_user = user;
785 }
786 }
787 } else if (!is_file_hugepages(shp->shm_file)) {
788 shmem_lock(shp->shm_file, 0, shp->mlock_user);
Andrew Mortonb33291c2006-01-08 01:02:21 -0800789 shp->shm_perm.mode &= ~SHM_LOCKED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790 shp->mlock_user = NULL;
791 }
792 shm_unlock(shp);
793 goto out;
794 }
795 case IPC_RMID:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 case IPC_SET:
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700797 err = shmctl_down(ns, shmid, cmd, buf, version);
798 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 default:
Pierre Peiffer8d4cc8b2008-04-29 01:00:47 -0700800 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 }
802
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803out_unlock:
804 shm_unlock(shp);
805out:
806 return err;
807}
808
809/*
810 * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists.
811 *
812 * NOTE! Despite the name, this is NOT a direct system call entrypoint. The
813 * "raddr" thing points to kernel space, and there has to be a wrapper around
814 * this.
815 */
816long do_shmat(int shmid, char __user *shmaddr, int shmflg, ulong *raddr)
817{
818 struct shmid_kernel *shp;
819 unsigned long addr;
820 unsigned long size;
821 struct file * file;
822 int err;
823 unsigned long flags;
824 unsigned long prot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825 int acc_mode;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800826 unsigned long user_addr;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700827 struct ipc_namespace *ns;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800828 struct shm_file_data *sfd;
829 struct path path;
830 mode_t f_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800832 err = -EINVAL;
833 if (shmid < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834 goto out;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800835 else if ((addr = (ulong)shmaddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 if (addr & (SHMLBA-1)) {
837 if (shmflg & SHM_RND)
838 addr &= ~(SHMLBA-1); /* round down */
839 else
840#ifndef __ARCH_FORCE_SHMLBA
841 if (addr & ~PAGE_MASK)
842#endif
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800843 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 }
845 flags = MAP_SHARED | MAP_FIXED;
846 } else {
847 if ((shmflg & SHM_REMAP))
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800848 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
850 flags = MAP_SHARED;
851 }
852
853 if (shmflg & SHM_RDONLY) {
854 prot = PROT_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 acc_mode = S_IRUGO;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800856 f_mode = FMODE_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 } else {
858 prot = PROT_READ | PROT_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 acc_mode = S_IRUGO | S_IWUGO;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800860 f_mode = FMODE_READ | FMODE_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861 }
862 if (shmflg & SHM_EXEC) {
863 prot |= PROT_EXEC;
864 acc_mode |= S_IXUGO;
865 }
866
867 /*
868 * We cannot rely on the fs check since SYSV IPC does have an
869 * additional creator id...
870 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700871 ns = current->nsproxy->ipc_ns;
Nadia Derbey023a5352007-10-18 23:40:51 -0700872 shp = shm_lock_check(ns, shmid);
873 if (IS_ERR(shp)) {
874 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 goto out;
Nadia Derbey023a5352007-10-18 23:40:51 -0700876 }
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800877
878 err = -EACCES;
879 if (ipcperms(&shp->shm_perm, acc_mode))
880 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881
882 err = security_shm_shmat(shp, shmaddr, shmflg);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800883 if (err)
884 goto out_unlock;
885
886 path.dentry = dget(shp->shm_file->f_path.dentry);
Dave Hansence8d2cd2007-10-16 23:31:13 -0700887 path.mnt = shp->shm_file->f_path.mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 shp->shm_nattch++;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800889 size = i_size_read(path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 shm_unlock(shp);
891
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800892 err = -ENOMEM;
893 sfd = kzalloc(sizeof(*sfd), GFP_KERNEL);
894 if (!sfd)
Dave Hansence8d2cd2007-10-16 23:31:13 -0700895 goto out_put_dentry;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800896
897 err = -ENOMEM;
Dave Hansence8d2cd2007-10-16 23:31:13 -0700898
899 file = alloc_file(path.mnt, path.dentry, f_mode, &shm_file_operations);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800900 if (!file)
901 goto out_free;
902
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800903 file->private_data = sfd;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800904 file->f_mapping = shp->shm_file->f_mapping;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700905 sfd->id = shp->shm_perm.id;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800906 sfd->ns = get_ipc_ns(ns);
907 sfd->file = shp->shm_file;
908 sfd->vm_ops = NULL;
909
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 down_write(&current->mm->mmap_sem);
911 if (addr && !(shmflg & SHM_REMAP)) {
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800912 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913 if (find_vma_intersection(current->mm, addr, addr + size))
914 goto invalid;
915 /*
916 * If shm segment goes below stack, make sure there is some
917 * space left for the stack to grow (at least 4 pages).
918 */
919 if (addr < current->mm->start_stack &&
920 addr > current->mm->start_stack - size - PAGE_SIZE * 5)
921 goto invalid;
922 }
923
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800924 user_addr = do_mmap (file, addr, size, prot, flags, 0);
925 *raddr = user_addr;
926 err = 0;
927 if (IS_ERR_VALUE(user_addr))
928 err = (long)user_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929invalid:
930 up_write(&current->mm->mmap_sem);
931
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800932 fput(file);
933
934out_nattch:
Nadia Derbey3e148c72007-10-18 23:40:54 -0700935 down_write(&shm_ids(ns).rw_mutex);
936 shp = shm_lock_down(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -0700937 BUG_ON(IS_ERR(shp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 shp->shm_nattch--;
939 if(shp->shm_nattch == 0 &&
Andrew Mortonb33291c2006-01-08 01:02:21 -0800940 shp->shm_perm.mode & SHM_DEST)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700941 shm_destroy(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 else
943 shm_unlock(shp);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700944 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946out:
947 return err;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800948
949out_unlock:
950 shm_unlock(shp);
951 goto out;
952
953out_free:
954 kfree(sfd);
Dave Hansence8d2cd2007-10-16 23:31:13 -0700955out_put_dentry:
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800956 dput(path.dentry);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800957 goto out_nattch;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958}
959
Stephen Rothwell7d87e142005-05-01 08:59:12 -0700960asmlinkage long sys_shmat(int shmid, char __user *shmaddr, int shmflg)
961{
962 unsigned long ret;
963 long err;
964
965 err = do_shmat(shmid, shmaddr, shmflg, &ret);
966 if (err)
967 return err;
968 force_successful_syscall_return();
969 return (long)ret;
970}
971
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972/*
973 * detach and kill segment if marked destroyed.
974 * The work is done in shm_close.
975 */
976asmlinkage long sys_shmdt(char __user *shmaddr)
977{
978 struct mm_struct *mm = current->mm;
979 struct vm_area_struct *vma, *next;
980 unsigned long addr = (unsigned long)shmaddr;
981 loff_t size = 0;
982 int retval = -EINVAL;
983
Hugh Dickinsdf1e2fb2006-03-24 03:18:06 -0800984 if (addr & ~PAGE_MASK)
985 return retval;
986
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 down_write(&mm->mmap_sem);
988
989 /*
990 * This function tries to be smart and unmap shm segments that
991 * were modified by partial mlock or munmap calls:
992 * - It first determines the size of the shm segment that should be
993 * unmapped: It searches for a vma that is backed by shm and that
994 * started at address shmaddr. It records it's size and then unmaps
995 * it.
996 * - Then it unmaps all shm vmas that started at shmaddr and that
997 * are within the initially determined size.
998 * Errors from do_munmap are ignored: the function only fails if
999 * it's called with invalid parameters or if it's called to unmap
1000 * a part of a vma. Both calls in this function are for full vmas,
1001 * the parameters are directly copied from the vma itself and always
1002 * valid - therefore do_munmap cannot fail. (famous last words?)
1003 */
1004 /*
1005 * If it had been mremap()'d, the starting address would not
1006 * match the usual checks anyway. So assume all vma's are
1007 * above the starting address given.
1008 */
1009 vma = find_vma(mm, addr);
1010
1011 while (vma) {
1012 next = vma->vm_next;
1013
1014 /*
1015 * Check if the starting address would match, i.e. it's
1016 * a fragment created by mprotect() and/or munmap(), or it
1017 * otherwise it starts at this address with no hassles.
1018 */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001019 if ((vma->vm_ops == &shm_vm_ops) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) {
1021
1022
Josef Sipek6d630792006-12-08 02:37:11 -08001023 size = vma->vm_file->f_path.dentry->d_inode->i_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1025 /*
1026 * We discovered the size of the shm segment, so
1027 * break out of here and fall through to the next
1028 * loop that uses the size information to stop
1029 * searching for matching vma's.
1030 */
1031 retval = 0;
1032 vma = next;
1033 break;
1034 }
1035 vma = next;
1036 }
1037
1038 /*
1039 * We need look no further than the maximum address a fragment
1040 * could possibly have landed at. Also cast things to loff_t to
1041 * prevent overflows and make comparisions vs. equal-width types.
1042 */
KAMEZAWA Hiroyuki8e367092006-02-10 01:51:12 -08001043 size = PAGE_ALIGN(size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 while (vma && (loff_t)(vma->vm_end - addr) <= size) {
1045 next = vma->vm_next;
1046
1047 /* finding a matching vma now does not alter retval */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001048 if ((vma->vm_ops == &shm_vm_ops) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff)
1050
1051 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1052 vma = next;
1053 }
1054
1055 up_write(&mm->mmap_sem);
1056 return retval;
1057}
1058
1059#ifdef CONFIG_PROC_FS
Mike Waychison19b49462005-09-06 15:17:10 -07001060static int sysvipc_shm_proc_show(struct seq_file *s, void *it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061{
Mike Waychison19b49462005-09-06 15:17:10 -07001062 struct shmid_kernel *shp = it;
1063 char *format;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065#define SMALL_STRING "%10d %10d %4o %10u %5u %5u %5d %5u %5u %5u %5u %10lu %10lu %10lu\n"
1066#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 -07001067
Mike Waychison19b49462005-09-06 15:17:10 -07001068 if (sizeof(size_t) <= sizeof(int))
1069 format = SMALL_STRING;
1070 else
1071 format = BIG_STRING;
1072 return seq_printf(s, format,
1073 shp->shm_perm.key,
Nadia Derbey7ca7e562007-10-18 23:40:48 -07001074 shp->shm_perm.id,
Andrew Mortonb33291c2006-01-08 01:02:21 -08001075 shp->shm_perm.mode,
Mike Waychison19b49462005-09-06 15:17:10 -07001076 shp->shm_segsz,
1077 shp->shm_cprid,
1078 shp->shm_lprid,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001079 shp->shm_nattch,
Mike Waychison19b49462005-09-06 15:17:10 -07001080 shp->shm_perm.uid,
1081 shp->shm_perm.gid,
1082 shp->shm_perm.cuid,
1083 shp->shm_perm.cgid,
1084 shp->shm_atim,
1085 shp->shm_dtim,
1086 shp->shm_ctim);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087}
1088#endif