blob: c47e87278a92d87acf6bdfefbb16ab73186c24b2 [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)
Nadia Derbey1b531f22007-10-18 23:40:55 -070063#define shm_buildid(id, seq) ipc_buildid(id, seq)
Kirill Korotaev4e982312006-10-02 02:18:22 -070064
Nadia Derbey7748dbf2007-10-18 23:40:49 -070065static int newseg(struct ipc_namespace *, struct ipc_params *);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -080066static void shm_open(struct vm_area_struct *vma);
67static void shm_close(struct vm_area_struct *vma);
Kirill Korotaev4e982312006-10-02 02:18:22 -070068static void shm_destroy (struct ipc_namespace *ns, struct shmid_kernel *shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -070069#ifdef CONFIG_PROC_FS
Mike Waychison19b49462005-09-06 15:17:10 -070070static int sysvipc_shm_proc_show(struct seq_file *s, void *it);
Linus Torvalds1da177e2005-04-16 15:20:36 -070071#endif
72
Pierre Peiffered2ddbf2008-02-08 04:18:57 -080073void shm_init_ns(struct ipc_namespace *ns)
Kirill Korotaev4e982312006-10-02 02:18:22 -070074{
Kirill Korotaev4e982312006-10-02 02:18:22 -070075 ns->shm_ctlmax = SHMMAX;
76 ns->shm_ctlall = SHMALL;
77 ns->shm_ctlmni = SHMMNI;
78 ns->shm_tot = 0;
Pierre Peiffered2ddbf2008-02-08 04:18:57 -080079 ipc_init_ids(&ns->ids[IPC_SHM_IDS]);
Kirill Korotaev4e982312006-10-02 02:18:22 -070080}
Linus Torvalds1da177e2005-04-16 15:20:36 -070081
Nadia Derbeyf4566f02007-10-18 23:40:53 -070082/*
Nadia Derbey3e148c72007-10-18 23:40:54 -070083 * Called with shm_ids.rw_mutex (writer) and the shp structure locked.
84 * Only shm_ids.rw_mutex remains locked on exit.
Nadia Derbeyf4566f02007-10-18 23:40:53 -070085 */
Pierre Peiffer01b8b072008-02-08 04:18:57 -080086static void do_shm_rmid(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp)
Kirill Korotaev4e982312006-10-02 02:18:22 -070087{
Pierre Peiffer01b8b072008-02-08 04:18:57 -080088 struct shmid_kernel *shp;
89 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
90
Kirill Korotaev4e982312006-10-02 02:18:22 -070091 if (shp->shm_nattch){
92 shp->shm_perm.mode |= SHM_DEST;
93 /* Do not find it any more */
94 shp->shm_perm.key = IPC_PRIVATE;
95 shm_unlock(shp);
96 } else
97 shm_destroy(ns, shp);
98}
99
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -0800100#ifdef CONFIG_IPC_NS
Kirill Korotaev4e982312006-10-02 02:18:22 -0700101void shm_exit_ns(struct ipc_namespace *ns)
102{
Pierre Peiffer01b8b072008-02-08 04:18:57 -0800103 free_ipcs(ns, &shm_ids(ns), do_shm_rmid);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700104}
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -0800105#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106
107void __init shm_init (void)
108{
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800109 shm_init_ns(&init_ipc_ns);
Mike Waychison19b49462005-09-06 15:17:10 -0700110 ipc_init_proc_interface("sysvipc/shm",
111 " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime\n",
Kirill Korotaev4e982312006-10-02 02:18:22 -0700112 IPC_SHM_IDS, sysvipc_shm_proc_show);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113}
114
Nadia Derbey3e148c72007-10-18 23:40:54 -0700115/*
116 * shm_lock_(check_)down routines are called in the paths where the rw_mutex
117 * is held to protect access to the idr tree.
118 */
119static inline struct shmid_kernel *shm_lock_down(struct ipc_namespace *ns,
120 int id)
121{
122 struct kern_ipc_perm *ipcp = ipc_lock_down(&shm_ids(ns), id);
123
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800124 if (IS_ERR(ipcp))
125 return (struct shmid_kernel *)ipcp;
126
Nadia Derbey3e148c72007-10-18 23:40:54 -0700127 return container_of(ipcp, struct shmid_kernel, shm_perm);
128}
129
130static inline struct shmid_kernel *shm_lock_check_down(
131 struct ipc_namespace *ns,
132 int id)
133{
134 struct kern_ipc_perm *ipcp = ipc_lock_check_down(&shm_ids(ns), id);
135
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800136 if (IS_ERR(ipcp))
137 return (struct shmid_kernel *)ipcp;
138
Nadia Derbey3e148c72007-10-18 23:40:54 -0700139 return container_of(ipcp, struct shmid_kernel, shm_perm);
140}
141
142/*
143 * shm_lock_(check_) routines are called in the paths where the rw_mutex
144 * is not held.
145 */
Nadia Derbey023a5352007-10-18 23:40:51 -0700146static inline struct shmid_kernel *shm_lock(struct ipc_namespace *ns, int id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700148 struct kern_ipc_perm *ipcp = ipc_lock(&shm_ids(ns), id);
149
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800150 if (IS_ERR(ipcp))
151 return (struct shmid_kernel *)ipcp;
152
Nadia Derbey03f02c72007-10-18 23:40:51 -0700153 return container_of(ipcp, struct shmid_kernel, shm_perm);
Nadia Derbey023a5352007-10-18 23:40:51 -0700154}
155
156static inline struct shmid_kernel *shm_lock_check(struct ipc_namespace *ns,
157 int id)
158{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700159 struct kern_ipc_perm *ipcp = ipc_lock_check(&shm_ids(ns), id);
160
Pierre Peifferb1ed88b2008-02-06 01:36:23 -0800161 if (IS_ERR(ipcp))
162 return (struct shmid_kernel *)ipcp;
163
Nadia Derbey03f02c72007-10-18 23:40:51 -0700164 return container_of(ipcp, struct shmid_kernel, shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165}
166
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700167static inline void shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *s)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700169 ipc_rmid(&shm_ids(ns), &s->shm_perm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170}
171
Kirill Korotaev4e982312006-10-02 02:18:22 -0700172static inline int shm_addid(struct ipc_namespace *ns, struct shmid_kernel *shp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700174 return ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175}
176
177
178
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800179/* This is called by fork, once for every shm attach. */
180static void shm_open(struct vm_area_struct *vma)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700181{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800182 struct file *file = vma->vm_file;
183 struct shm_file_data *sfd = shm_file_data(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184 struct shmid_kernel *shp;
185
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800186 shp = shm_lock(sfd->ns, sfd->id);
Nadia Derbey023a5352007-10-18 23:40:51 -0700187 BUG_ON(IS_ERR(shp));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 shp->shm_atim = get_seconds();
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700189 shp->shm_lprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 shp->shm_nattch++;
191 shm_unlock(shp);
192}
193
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194/*
195 * shm_destroy - free the struct shmid_kernel
196 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700197 * @ns: namespace
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 * @shp: struct to free
199 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700200 * It has to be called with shp and shm_ids.rw_mutex (writer) locked,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201 * but returns with shp unlocked and freed.
202 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700203static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700205 ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700206 shm_rmid(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 shm_unlock(shp);
208 if (!is_file_hugepages(shp->shm_file))
209 shmem_lock(shp->shm_file, 0, shp->mlock_user);
210 else
Josef Sipek6d630792006-12-08 02:37:11 -0800211 user_shm_unlock(shp->shm_file->f_path.dentry->d_inode->i_size,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212 shp->mlock_user);
213 fput (shp->shm_file);
214 security_shm_free(shp);
215 ipc_rcu_putref(shp);
216}
217
218/*
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800219 * remove the attach descriptor vma.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 * free memory for segment if it is marked destroyed.
221 * The descriptor has already been removed from the current->mm->mmap list
222 * and will later be kfree()d.
223 */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800224static void shm_close(struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800226 struct file * file = vma->vm_file;
227 struct shm_file_data *sfd = shm_file_data(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 struct shmid_kernel *shp;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800229 struct ipc_namespace *ns = sfd->ns;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700230
Nadia Derbey3e148c72007-10-18 23:40:54 -0700231 down_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 /* remove from the list of attaches of the shm segment */
Nadia Derbey3e148c72007-10-18 23:40:54 -0700233 shp = shm_lock_down(ns, sfd->id);
Nadia Derbey023a5352007-10-18 23:40:51 -0700234 BUG_ON(IS_ERR(shp));
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700235 shp->shm_lprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 shp->shm_dtim = get_seconds();
237 shp->shm_nattch--;
238 if(shp->shm_nattch == 0 &&
Andrew Mortonb33291c2006-01-08 01:02:21 -0800239 shp->shm_perm.mode & SHM_DEST)
Kirill Korotaev4e982312006-10-02 02:18:22 -0700240 shm_destroy(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 else
242 shm_unlock(shp);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700243 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244}
245
Nick Piggind0217ac2007-07-19 01:47:03 -0700246static int shm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800247{
248 struct file *file = vma->vm_file;
249 struct shm_file_data *sfd = shm_file_data(file);
250
Nick Piggind0217ac2007-07-19 01:47:03 -0700251 return sfd->vm_ops->fault(vma, vmf);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800252}
253
254#ifdef CONFIG_NUMA
Adrian Bunkd823e3e2007-10-16 23:26:42 -0700255static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800256{
257 struct file *file = vma->vm_file;
258 struct shm_file_data *sfd = shm_file_data(file);
259 int err = 0;
260 if (sfd->vm_ops->set_policy)
261 err = sfd->vm_ops->set_policy(vma, new);
262 return err;
263}
264
Adrian Bunkd823e3e2007-10-16 23:26:42 -0700265static struct mempolicy *shm_get_policy(struct vm_area_struct *vma,
266 unsigned long addr)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800267{
268 struct file *file = vma->vm_file;
269 struct shm_file_data *sfd = shm_file_data(file);
270 struct mempolicy *pol = NULL;
271
272 if (sfd->vm_ops->get_policy)
273 pol = sfd->vm_ops->get_policy(vma, addr);
Adam Litke22741922007-06-16 10:16:15 -0700274 else if (vma->vm_policy)
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800275 pol = vma->vm_policy;
Adam Litke22741922007-06-16 10:16:15 -0700276 else
277 pol = current->mempolicy;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800278 return pol;
279}
280#endif
281
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282static int shm_mmap(struct file * file, struct vm_area_struct * vma)
283{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800284 struct shm_file_data *sfd = shm_file_data(file);
David Howellsb0e15192006-01-06 00:11:42 -0800285 int ret;
286
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800287 ret = sfd->file->f_op->mmap(sfd->file, vma);
288 if (ret != 0)
289 return ret;
290 sfd->vm_ops = vma->vm_ops;
David Howells2e92a3b2007-07-31 00:37:24 -0700291#ifdef CONFIG_MMU
Nick Piggin54cb8822007-07-19 01:46:59 -0700292 BUG_ON(!sfd->vm_ops->fault);
David Howells2e92a3b2007-07-31 00:37:24 -0700293#endif
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800294 vma->vm_ops = &shm_vm_ops;
295 shm_open(vma);
David Howellsb0e15192006-01-06 00:11:42 -0800296
297 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298}
299
Kirill Korotaev4e982312006-10-02 02:18:22 -0700300static int shm_release(struct inode *ino, struct file *file)
301{
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800302 struct shm_file_data *sfd = shm_file_data(file);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700303
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800304 put_ipc_ns(sfd->ns);
305 shm_file_data(file) = NULL;
306 kfree(sfd);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700307 return 0;
308}
309
Adam Litke516dffd2007-03-01 15:46:08 -0800310static int shm_fsync(struct file *file, struct dentry *dentry, int datasync)
311{
312 int (*fsync) (struct file *, struct dentry *, int datasync);
313 struct shm_file_data *sfd = shm_file_data(file);
314 int ret = -EINVAL;
315
316 fsync = sfd->file->f_op->fsync;
317 if (fsync)
318 ret = fsync(sfd->file, sfd->file->f_path.dentry, datasync);
319 return ret;
320}
321
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800322static unsigned long shm_get_unmapped_area(struct file *file,
323 unsigned long addr, unsigned long len, unsigned long pgoff,
324 unsigned long flags)
325{
326 struct shm_file_data *sfd = shm_file_data(file);
Adam Litke516dffd2007-03-01 15:46:08 -0800327 return get_unmapped_area(sfd->file, addr, len, pgoff, flags);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800328}
Adam Litke516dffd2007-03-01 15:46:08 -0800329
330int is_file_shm_hugepages(struct file *file)
331{
332 int ret = 0;
333
334 if (file->f_op == &shm_file_operations) {
335 struct shm_file_data *sfd;
336 sfd = shm_file_data(file);
337 ret = is_file_hugepages(sfd->file);
338 }
339 return ret;
340}
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800341
Arjan van de Ven9a321442007-02-12 00:55:35 -0800342static const struct file_operations shm_file_operations = {
Kirill Korotaev4e982312006-10-02 02:18:22 -0700343 .mmap = shm_mmap,
Adam Litke516dffd2007-03-01 15:46:08 -0800344 .fsync = shm_fsync,
Kirill Korotaev4e982312006-10-02 02:18:22 -0700345 .release = shm_release,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800346 .get_unmapped_area = shm_get_unmapped_area,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347};
348
349static struct vm_operations_struct shm_vm_ops = {
350 .open = shm_open, /* callback for a new vm-area open */
351 .close = shm_close, /* callback for when the vm-area is released */
Nick Piggin54cb8822007-07-19 01:46:59 -0700352 .fault = shm_fault,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800353#if defined(CONFIG_NUMA)
354 .set_policy = shm_set_policy,
355 .get_policy = shm_get_policy,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356#endif
357};
358
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700359/**
360 * newseg - Create a new shared memory segment
361 * @ns: namespace
362 * @params: ptr to the structure that contains key, size and shmflg
363 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700364 * Called with shm_ids.rw_mutex held as a writer.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700365 */
366
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700367static int newseg(struct ipc_namespace *ns, struct ipc_params *params)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368{
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700369 key_t key = params->key;
370 int shmflg = params->flg;
371 size_t size = params->u.size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 int error;
373 struct shmid_kernel *shp;
374 int numpages = (size + PAGE_SIZE -1) >> PAGE_SHIFT;
375 struct file * file;
376 char name[13];
377 int id;
378
Kirill Korotaev4e982312006-10-02 02:18:22 -0700379 if (size < SHMMIN || size > ns->shm_ctlmax)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 return -EINVAL;
381
Guy Streeterf66d45e2007-01-23 12:20:04 -0600382 if (ns->shm_tot + numpages > ns->shm_ctlall)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 return -ENOSPC;
384
385 shp = ipc_rcu_alloc(sizeof(*shp));
386 if (!shp)
387 return -ENOMEM;
388
389 shp->shm_perm.key = key;
Andrew Mortonb33291c2006-01-08 01:02:21 -0800390 shp->shm_perm.mode = (shmflg & S_IRWXUGO);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 shp->mlock_user = NULL;
392
393 shp->shm_perm.security = NULL;
394 error = security_shm_alloc(shp);
395 if (error) {
396 ipc_rcu_putref(shp);
397 return error;
398 }
399
Eric W. Biederman9d665862007-06-16 10:16:16 -0700400 sprintf (name, "SYSV%08x", key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 if (shmflg & SHM_HUGETLB) {
Eric W. Biederman9d665862007-06-16 10:16:16 -0700402 /* hugetlb_file_setup takes care of mlock user accounting */
403 file = hugetlb_file_setup(name, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 shp->mlock_user = current->user;
405 } else {
Badari Pulavartybf8f9722005-11-07 00:59:27 -0800406 int acctflag = VM_ACCOUNT;
407 /*
408 * Do not allow no accounting for OVERCOMMIT_NEVER, even
409 * if it's asked for.
410 */
411 if ((shmflg & SHM_NORESERVE) &&
412 sysctl_overcommit_memory != OVERCOMMIT_NEVER)
413 acctflag = 0;
Badari Pulavartybf8f9722005-11-07 00:59:27 -0800414 file = shmem_file_setup(name, size, acctflag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 }
416 error = PTR_ERR(file);
417 if (IS_ERR(file))
418 goto no_file;
419
Kirill Korotaev4e982312006-10-02 02:18:22 -0700420 id = shm_addid(ns, shp);
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700421 if (id < 0) {
422 error = id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 goto no_id;
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700424 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425
Pavel Emelyanovb4888932007-10-18 23:40:14 -0700426 shp->shm_cprid = task_tgid_vnr(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 shp->shm_lprid = 0;
428 shp->shm_atim = shp->shm_dtim = 0;
429 shp->shm_ctim = get_seconds();
430 shp->shm_segsz = size;
431 shp->shm_nattch = 0;
Nadia Derbey1b531f22007-10-18 23:40:55 -0700432 shp->shm_perm.id = shm_buildid(id, shp->shm_perm.seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 shp->shm_file = file;
Badari Pulavarty30475cc2007-06-16 10:15:59 -0700434 /*
435 * shmid gets reported as "inode#" in /proc/pid/maps.
436 * proc-ps tools use this. Changing this will break them.
437 */
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700438 file->f_dentry->d_inode->i_ino = shp->shm_perm.id;
Krishnakumar R551110a2005-10-29 18:16:45 -0700439
Kirill Korotaev4e982312006-10-02 02:18:22 -0700440 ns->shm_tot += numpages;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700441 error = shp->shm_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 shm_unlock(shp);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700443 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444
445no_id:
446 fput(file);
447no_file:
448 security_shm_free(shp);
449 ipc_rcu_putref(shp);
450 return error;
451}
452
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700453/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700454 * Called with shm_ids.rw_mutex and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700455 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700456static inline int shm_security(struct kern_ipc_perm *ipcp, int shmflg)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700457{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700458 struct shmid_kernel *shp;
459
460 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
461 return security_shm_associate(shp, shmflg);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700462}
463
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700464/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700465 * Called with shm_ids.rw_mutex and ipcp locked.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700466 */
Nadia Derbey03f02c72007-10-18 23:40:51 -0700467static inline int shm_more_checks(struct kern_ipc_perm *ipcp,
468 struct ipc_params *params)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700469{
Nadia Derbey03f02c72007-10-18 23:40:51 -0700470 struct shmid_kernel *shp;
471
472 shp = container_of(ipcp, struct shmid_kernel, shm_perm);
473 if (shp->shm_segsz < params->u.size)
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700474 return -EINVAL;
475
476 return 0;
477}
478
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479asmlinkage long sys_shmget (key_t key, size_t size, int shmflg)
480{
Kirill Korotaev4e982312006-10-02 02:18:22 -0700481 struct ipc_namespace *ns;
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700482 struct ipc_ops shm_ops;
483 struct ipc_params shm_params;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
Kirill Korotaev4e982312006-10-02 02:18:22 -0700485 ns = current->nsproxy->ipc_ns;
486
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700487 shm_ops.getnew = newseg;
488 shm_ops.associate = shm_security;
489 shm_ops.more_checks = shm_more_checks;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700490
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700491 shm_params.key = key;
492 shm_params.flg = shmflg;
493 shm_params.u.size = size;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700494
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700495 return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496}
497
498static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version)
499{
500 switch(version) {
501 case IPC_64:
502 return copy_to_user(buf, in, sizeof(*in));
503 case IPC_OLD:
504 {
505 struct shmid_ds out;
506
507 ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm);
508 out.shm_segsz = in->shm_segsz;
509 out.shm_atime = in->shm_atime;
510 out.shm_dtime = in->shm_dtime;
511 out.shm_ctime = in->shm_ctime;
512 out.shm_cpid = in->shm_cpid;
513 out.shm_lpid = in->shm_lpid;
514 out.shm_nattch = in->shm_nattch;
515
516 return copy_to_user(buf, &out, sizeof(out));
517 }
518 default:
519 return -EINVAL;
520 }
521}
522
523struct shm_setbuf {
524 uid_t uid;
525 gid_t gid;
526 mode_t mode;
527};
528
529static inline unsigned long copy_shmid_from_user(struct shm_setbuf *out, void __user *buf, int version)
530{
531 switch(version) {
532 case IPC_64:
533 {
534 struct shmid64_ds tbuf;
535
536 if (copy_from_user(&tbuf, buf, sizeof(tbuf)))
537 return -EFAULT;
538
539 out->uid = tbuf.shm_perm.uid;
540 out->gid = tbuf.shm_perm.gid;
Andrew Mortonb33291c2006-01-08 01:02:21 -0800541 out->mode = tbuf.shm_perm.mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542
543 return 0;
544 }
545 case IPC_OLD:
546 {
547 struct shmid_ds tbuf_old;
548
549 if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old)))
550 return -EFAULT;
551
552 out->uid = tbuf_old.shm_perm.uid;
553 out->gid = tbuf_old.shm_perm.gid;
Andrew Mortonb33291c2006-01-08 01:02:21 -0800554 out->mode = tbuf_old.shm_perm.mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555
556 return 0;
557 }
558 default:
559 return -EINVAL;
560 }
561}
562
563static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version)
564{
565 switch(version) {
566 case IPC_64:
567 return copy_to_user(buf, in, sizeof(*in));
568 case IPC_OLD:
569 {
570 struct shminfo out;
571
572 if(in->shmmax > INT_MAX)
573 out.shmmax = INT_MAX;
574 else
575 out.shmmax = (int)in->shmmax;
576
577 out.shmmin = in->shmmin;
578 out.shmmni = in->shmmni;
579 out.shmseg = in->shmseg;
580 out.shmall = in->shmall;
581
582 return copy_to_user(buf, &out, sizeof(out));
583 }
584 default:
585 return -EINVAL;
586 }
587}
588
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700589/*
Nadia Derbey3e148c72007-10-18 23:40:54 -0700590 * Called with shm_ids.rw_mutex held as a reader
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700591 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700592static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss,
593 unsigned long *swp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700595 int next_id;
596 int total, in_use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597
598 *rss = 0;
599 *swp = 0;
600
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700601 in_use = shm_ids(ns).in_use;
602
603 for (total = 0, next_id = 0; total < in_use; next_id++) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 struct shmid_kernel *shp;
605 struct inode *inode;
606
Nadia Derbey637c3662007-10-18 23:40:50 -0700607 shp = idr_find(&shm_ids(ns).ipcs_idr, next_id);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700608 if (shp == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 continue;
610
Josef Sipek6d630792006-12-08 02:37:11 -0800611 inode = shp->shm_file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612
613 if (is_file_hugepages(shp->shm_file)) {
614 struct address_space *mapping = inode->i_mapping;
615 *rss += (HPAGE_SIZE/PAGE_SIZE)*mapping->nrpages;
616 } else {
617 struct shmem_inode_info *info = SHMEM_I(inode);
618 spin_lock(&info->lock);
619 *rss += inode->i_mapping->nrpages;
620 *swp += info->swapped;
621 spin_unlock(&info->lock);
622 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700623
624 total++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 }
626}
627
628asmlinkage long sys_shmctl (int shmid, int cmd, struct shmid_ds __user *buf)
629{
630 struct shm_setbuf setbuf;
631 struct shmid_kernel *shp;
632 int err, version;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700633 struct ipc_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634
635 if (cmd < 0 || shmid < 0) {
636 err = -EINVAL;
637 goto out;
638 }
639
640 version = ipc_parse_version(&cmd);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700641 ns = current->nsproxy->ipc_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642
643 switch (cmd) { /* replace with proc interface ? */
644 case IPC_INFO:
645 {
646 struct shminfo64 shminfo;
647
648 err = security_shm_shmctl(NULL, cmd);
649 if (err)
650 return err;
651
652 memset(&shminfo,0,sizeof(shminfo));
Kirill Korotaev4e982312006-10-02 02:18:22 -0700653 shminfo.shmmni = shminfo.shmseg = ns->shm_ctlmni;
654 shminfo.shmmax = ns->shm_ctlmax;
655 shminfo.shmall = ns->shm_ctlall;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656
657 shminfo.shmmin = SHMMIN;
658 if(copy_shminfo_to_user (buf, &shminfo, version))
659 return -EFAULT;
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700660
Nadia Derbey3e148c72007-10-18 23:40:54 -0700661 down_read(&shm_ids(ns).rw_mutex);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700662 err = ipc_get_maxid(&shm_ids(ns));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700663 up_read(&shm_ids(ns).rw_mutex);
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700664
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 if(err<0)
666 err = 0;
667 goto out;
668 }
669 case SHM_INFO:
670 {
671 struct shm_info shm_info;
672
673 err = security_shm_shmctl(NULL, cmd);
674 if (err)
675 return err;
676
677 memset(&shm_info,0,sizeof(shm_info));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700678 down_read(&shm_ids(ns).rw_mutex);
Kirill Korotaev4e982312006-10-02 02:18:22 -0700679 shm_info.used_ids = shm_ids(ns).in_use;
680 shm_get_stat (ns, &shm_info.shm_rss, &shm_info.shm_swp);
681 shm_info.shm_tot = ns->shm_tot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 shm_info.swap_attempts = 0;
683 shm_info.swap_successes = 0;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700684 err = ipc_get_maxid(&shm_ids(ns));
Nadia Derbey3e148c72007-10-18 23:40:54 -0700685 up_read(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 if(copy_to_user (buf, &shm_info, sizeof(shm_info))) {
687 err = -EFAULT;
688 goto out;
689 }
690
691 err = err < 0 ? 0 : err;
692 goto out;
693 }
694 case SHM_STAT:
695 case IPC_STAT:
696 {
697 struct shmid64_ds tbuf;
698 int result;
Nadia Derbey023a5352007-10-18 23:40:51 -0700699
700 if (!buf) {
701 err = -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 goto out;
Nadia Derbey023a5352007-10-18 23:40:51 -0700703 }
704
705 if (cmd == SHM_STAT) {
706 shp = shm_lock(ns, shmid);
707 if (IS_ERR(shp)) {
708 err = PTR_ERR(shp);
709 goto out;
710 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700711 result = shp->shm_perm.id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 } else {
Nadia Derbey023a5352007-10-18 23:40:51 -0700713 shp = shm_lock_check(ns, shmid);
714 if (IS_ERR(shp)) {
715 err = PTR_ERR(shp);
716 goto out;
717 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 result = 0;
719 }
720 err=-EACCES;
721 if (ipcperms (&shp->shm_perm, S_IRUGO))
722 goto out_unlock;
723 err = security_shm_shmctl(shp, cmd);
724 if (err)
725 goto out_unlock;
Nadia Derbey023a5352007-10-18 23:40:51 -0700726 memset(&tbuf, 0, sizeof(tbuf));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 kernel_to_ipc64_perm(&shp->shm_perm, &tbuf.shm_perm);
728 tbuf.shm_segsz = shp->shm_segsz;
729 tbuf.shm_atime = shp->shm_atim;
730 tbuf.shm_dtime = shp->shm_dtim;
731 tbuf.shm_ctime = shp->shm_ctim;
732 tbuf.shm_cpid = shp->shm_cprid;
733 tbuf.shm_lpid = shp->shm_lprid;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800734 tbuf.shm_nattch = shp->shm_nattch;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 shm_unlock(shp);
736 if(copy_shmid_to_user (buf, &tbuf, version))
737 err = -EFAULT;
738 else
739 err = result;
740 goto out;
741 }
742 case SHM_LOCK:
743 case SHM_UNLOCK:
744 {
Nadia Derbey023a5352007-10-18 23:40:51 -0700745 shp = shm_lock_check(ns, shmid);
746 if (IS_ERR(shp)) {
747 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 goto out;
749 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
Steve Grubb073115d2006-04-02 17:07:33 -0400751 err = audit_ipc_obj(&(shp->shm_perm));
752 if (err)
753 goto out_unlock;
754
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 if (!capable(CAP_IPC_LOCK)) {
756 err = -EPERM;
757 if (current->euid != shp->shm_perm.uid &&
758 current->euid != shp->shm_perm.cuid)
759 goto out_unlock;
760 if (cmd == SHM_LOCK &&
761 !current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur)
762 goto out_unlock;
763 }
764
765 err = security_shm_shmctl(shp, cmd);
766 if (err)
767 goto out_unlock;
768
769 if(cmd==SHM_LOCK) {
770 struct user_struct * user = current->user;
771 if (!is_file_hugepages(shp->shm_file)) {
772 err = shmem_lock(shp->shm_file, 1, user);
Pavel Emelianov7be77e22007-07-31 00:38:48 -0700773 if (!err && !(shp->shm_perm.mode & SHM_LOCKED)){
Andrew Mortonb33291c2006-01-08 01:02:21 -0800774 shp->shm_perm.mode |= SHM_LOCKED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 shp->mlock_user = user;
776 }
777 }
778 } else if (!is_file_hugepages(shp->shm_file)) {
779 shmem_lock(shp->shm_file, 0, shp->mlock_user);
Andrew Mortonb33291c2006-01-08 01:02:21 -0800780 shp->shm_perm.mode &= ~SHM_LOCKED;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 shp->mlock_user = NULL;
782 }
783 shm_unlock(shp);
784 goto out;
785 }
786 case IPC_RMID:
787 {
788 /*
789 * We cannot simply remove the file. The SVID states
790 * that the block remains until the last person
791 * detaches from it, then is deleted. A shmat() on
792 * an RMID segment is legal in older Linux and if
793 * we change it apps break...
794 *
795 * Instead we set a destroyed flag, and then blow
796 * the name away when the usage hits zero.
797 */
Nadia Derbey3e148c72007-10-18 23:40:54 -0700798 down_write(&shm_ids(ns).rw_mutex);
799 shp = shm_lock_check_down(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -0700800 if (IS_ERR(shp)) {
801 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 goto out_up;
Nadia Derbey023a5352007-10-18 23:40:51 -0700803 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804
Steve Grubb073115d2006-04-02 17:07:33 -0400805 err = audit_ipc_obj(&(shp->shm_perm));
806 if (err)
807 goto out_unlock_up;
808
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 if (current->euid != shp->shm_perm.uid &&
810 current->euid != shp->shm_perm.cuid &&
811 !capable(CAP_SYS_ADMIN)) {
812 err=-EPERM;
813 goto out_unlock_up;
814 }
815
816 err = security_shm_shmctl(shp, cmd);
817 if (err)
818 goto out_unlock_up;
819
Pierre Peiffer01b8b072008-02-08 04:18:57 -0800820 do_shm_rmid(ns, &shp->shm_perm);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700821 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 goto out;
823 }
824
825 case IPC_SET:
826 {
Nadia Derbey023a5352007-10-18 23:40:51 -0700827 if (!buf) {
828 err = -EFAULT;
829 goto out;
830 }
831
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832 if (copy_shmid_from_user (&setbuf, buf, version)) {
833 err = -EFAULT;
834 goto out;
835 }
Nadia Derbey3e148c72007-10-18 23:40:54 -0700836 down_write(&shm_ids(ns).rw_mutex);
837 shp = shm_lock_check_down(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -0700838 if (IS_ERR(shp)) {
839 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 goto out_up;
Nadia Derbey023a5352007-10-18 23:40:51 -0700841 }
Steve Grubb073115d2006-04-02 17:07:33 -0400842 err = audit_ipc_obj(&(shp->shm_perm));
843 if (err)
844 goto out_unlock_up;
Linda Knippersac032212006-05-16 22:03:48 -0400845 err = audit_ipc_set_perm(0, setbuf.uid, setbuf.gid, setbuf.mode);
Steve Grubb073115d2006-04-02 17:07:33 -0400846 if (err)
847 goto out_unlock_up;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 err=-EPERM;
849 if (current->euid != shp->shm_perm.uid &&
850 current->euid != shp->shm_perm.cuid &&
851 !capable(CAP_SYS_ADMIN)) {
852 goto out_unlock_up;
853 }
854
855 err = security_shm_shmctl(shp, cmd);
856 if (err)
857 goto out_unlock_up;
858
859 shp->shm_perm.uid = setbuf.uid;
860 shp->shm_perm.gid = setbuf.gid;
Andrew Mortonb33291c2006-01-08 01:02:21 -0800861 shp->shm_perm.mode = (shp->shm_perm.mode & ~S_IRWXUGO)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 | (setbuf.mode & S_IRWXUGO);
863 shp->shm_ctim = get_seconds();
864 break;
865 }
866
867 default:
868 err = -EINVAL;
869 goto out;
870 }
871
872 err = 0;
873out_unlock_up:
874 shm_unlock(shp);
875out_up:
Nadia Derbey3e148c72007-10-18 23:40:54 -0700876 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 goto out;
878out_unlock:
879 shm_unlock(shp);
880out:
881 return err;
882}
883
884/*
885 * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists.
886 *
887 * NOTE! Despite the name, this is NOT a direct system call entrypoint. The
888 * "raddr" thing points to kernel space, and there has to be a wrapper around
889 * this.
890 */
891long do_shmat(int shmid, char __user *shmaddr, int shmflg, ulong *raddr)
892{
893 struct shmid_kernel *shp;
894 unsigned long addr;
895 unsigned long size;
896 struct file * file;
897 int err;
898 unsigned long flags;
899 unsigned long prot;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 int acc_mode;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800901 unsigned long user_addr;
Kirill Korotaev4e982312006-10-02 02:18:22 -0700902 struct ipc_namespace *ns;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800903 struct shm_file_data *sfd;
904 struct path path;
905 mode_t f_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800907 err = -EINVAL;
908 if (shmid < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 goto out;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800910 else if ((addr = (ulong)shmaddr)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 if (addr & (SHMLBA-1)) {
912 if (shmflg & SHM_RND)
913 addr &= ~(SHMLBA-1); /* round down */
914 else
915#ifndef __ARCH_FORCE_SHMLBA
916 if (addr & ~PAGE_MASK)
917#endif
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800918 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 }
920 flags = MAP_SHARED | MAP_FIXED;
921 } else {
922 if ((shmflg & SHM_REMAP))
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800923 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924
925 flags = MAP_SHARED;
926 }
927
928 if (shmflg & SHM_RDONLY) {
929 prot = PROT_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 acc_mode = S_IRUGO;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800931 f_mode = FMODE_READ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 } else {
933 prot = PROT_READ | PROT_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 acc_mode = S_IRUGO | S_IWUGO;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800935 f_mode = FMODE_READ | FMODE_WRITE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 }
937 if (shmflg & SHM_EXEC) {
938 prot |= PROT_EXEC;
939 acc_mode |= S_IXUGO;
940 }
941
942 /*
943 * We cannot rely on the fs check since SYSV IPC does have an
944 * additional creator id...
945 */
Kirill Korotaev4e982312006-10-02 02:18:22 -0700946 ns = current->nsproxy->ipc_ns;
Nadia Derbey023a5352007-10-18 23:40:51 -0700947 shp = shm_lock_check(ns, shmid);
948 if (IS_ERR(shp)) {
949 err = PTR_ERR(shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 goto out;
Nadia Derbey023a5352007-10-18 23:40:51 -0700951 }
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800952
953 err = -EACCES;
954 if (ipcperms(&shp->shm_perm, acc_mode))
955 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956
957 err = security_shm_shmat(shp, shmaddr, shmflg);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800958 if (err)
959 goto out_unlock;
960
961 path.dentry = dget(shp->shm_file->f_path.dentry);
Dave Hansence8d2cd2007-10-16 23:31:13 -0700962 path.mnt = shp->shm_file->f_path.mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 shp->shm_nattch++;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800964 size = i_size_read(path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 shm_unlock(shp);
966
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800967 err = -ENOMEM;
968 sfd = kzalloc(sizeof(*sfd), GFP_KERNEL);
969 if (!sfd)
Dave Hansence8d2cd2007-10-16 23:31:13 -0700970 goto out_put_dentry;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800971
972 err = -ENOMEM;
Dave Hansence8d2cd2007-10-16 23:31:13 -0700973
974 file = alloc_file(path.mnt, path.dentry, f_mode, &shm_file_operations);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800975 if (!file)
976 goto out_free;
977
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800978 file->private_data = sfd;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800979 file->f_mapping = shp->shm_file->f_mapping;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700980 sfd->id = shp->shm_perm.id;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800981 sfd->ns = get_ipc_ns(ns);
982 sfd->file = shp->shm_file;
983 sfd->vm_ops = NULL;
984
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 down_write(&current->mm->mmap_sem);
986 if (addr && !(shmflg & SHM_REMAP)) {
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800987 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 if (find_vma_intersection(current->mm, addr, addr + size))
989 goto invalid;
990 /*
991 * If shm segment goes below stack, make sure there is some
992 * space left for the stack to grow (at least 4 pages).
993 */
994 if (addr < current->mm->start_stack &&
995 addr > current->mm->start_stack - size - PAGE_SIZE * 5)
996 goto invalid;
997 }
998
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -0800999 user_addr = do_mmap (file, addr, size, prot, flags, 0);
1000 *raddr = user_addr;
1001 err = 0;
1002 if (IS_ERR_VALUE(user_addr))
1003 err = (long)user_addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004invalid:
1005 up_write(&current->mm->mmap_sem);
1006
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001007 fput(file);
1008
1009out_nattch:
Nadia Derbey3e148c72007-10-18 23:40:54 -07001010 down_write(&shm_ids(ns).rw_mutex);
1011 shp = shm_lock_down(ns, shmid);
Nadia Derbey023a5352007-10-18 23:40:51 -07001012 BUG_ON(IS_ERR(shp));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 shp->shm_nattch--;
1014 if(shp->shm_nattch == 0 &&
Andrew Mortonb33291c2006-01-08 01:02:21 -08001015 shp->shm_perm.mode & SHM_DEST)
Kirill Korotaev4e982312006-10-02 02:18:22 -07001016 shm_destroy(ns, shp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 else
1018 shm_unlock(shp);
Nadia Derbey3e148c72007-10-18 23:40:54 -07001019 up_write(&shm_ids(ns).rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021out:
1022 return err;
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001023
1024out_unlock:
1025 shm_unlock(shp);
1026 goto out;
1027
1028out_free:
1029 kfree(sfd);
Dave Hansence8d2cd2007-10-16 23:31:13 -07001030out_put_dentry:
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001031 dput(path.dentry);
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001032 goto out_nattch;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033}
1034
Stephen Rothwell7d87e142005-05-01 08:59:12 -07001035asmlinkage long sys_shmat(int shmid, char __user *shmaddr, int shmflg)
1036{
1037 unsigned long ret;
1038 long err;
1039
1040 err = do_shmat(shmid, shmaddr, shmflg, &ret);
1041 if (err)
1042 return err;
1043 force_successful_syscall_return();
1044 return (long)ret;
1045}
1046
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047/*
1048 * detach and kill segment if marked destroyed.
1049 * The work is done in shm_close.
1050 */
1051asmlinkage long sys_shmdt(char __user *shmaddr)
1052{
1053 struct mm_struct *mm = current->mm;
1054 struct vm_area_struct *vma, *next;
1055 unsigned long addr = (unsigned long)shmaddr;
1056 loff_t size = 0;
1057 int retval = -EINVAL;
1058
Hugh Dickinsdf1e2fb2006-03-24 03:18:06 -08001059 if (addr & ~PAGE_MASK)
1060 return retval;
1061
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062 down_write(&mm->mmap_sem);
1063
1064 /*
1065 * This function tries to be smart and unmap shm segments that
1066 * were modified by partial mlock or munmap calls:
1067 * - It first determines the size of the shm segment that should be
1068 * unmapped: It searches for a vma that is backed by shm and that
1069 * started at address shmaddr. It records it's size and then unmaps
1070 * it.
1071 * - Then it unmaps all shm vmas that started at shmaddr and that
1072 * are within the initially determined size.
1073 * Errors from do_munmap are ignored: the function only fails if
1074 * it's called with invalid parameters or if it's called to unmap
1075 * a part of a vma. Both calls in this function are for full vmas,
1076 * the parameters are directly copied from the vma itself and always
1077 * valid - therefore do_munmap cannot fail. (famous last words?)
1078 */
1079 /*
1080 * If it had been mremap()'d, the starting address would not
1081 * match the usual checks anyway. So assume all vma's are
1082 * above the starting address given.
1083 */
1084 vma = find_vma(mm, addr);
1085
1086 while (vma) {
1087 next = vma->vm_next;
1088
1089 /*
1090 * Check if the starting address would match, i.e. it's
1091 * a fragment created by mprotect() and/or munmap(), or it
1092 * otherwise it starts at this address with no hassles.
1093 */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001094 if ((vma->vm_ops == &shm_vm_ops) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) {
1096
1097
Josef Sipek6d630792006-12-08 02:37:11 -08001098 size = vma->vm_file->f_path.dentry->d_inode->i_size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1100 /*
1101 * We discovered the size of the shm segment, so
1102 * break out of here and fall through to the next
1103 * loop that uses the size information to stop
1104 * searching for matching vma's.
1105 */
1106 retval = 0;
1107 vma = next;
1108 break;
1109 }
1110 vma = next;
1111 }
1112
1113 /*
1114 * We need look no further than the maximum address a fragment
1115 * could possibly have landed at. Also cast things to loff_t to
1116 * prevent overflows and make comparisions vs. equal-width types.
1117 */
KAMEZAWA Hiroyuki8e367092006-02-10 01:51:12 -08001118 size = PAGE_ALIGN(size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 while (vma && (loff_t)(vma->vm_end - addr) <= size) {
1120 next = vma->vm_next;
1121
1122 /* finding a matching vma now does not alter retval */
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001123 if ((vma->vm_ops == &shm_vm_ops) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff)
1125
1126 do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start);
1127 vma = next;
1128 }
1129
1130 up_write(&mm->mmap_sem);
1131 return retval;
1132}
1133
1134#ifdef CONFIG_PROC_FS
Mike Waychison19b49462005-09-06 15:17:10 -07001135static int sysvipc_shm_proc_show(struct seq_file *s, void *it)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136{
Mike Waychison19b49462005-09-06 15:17:10 -07001137 struct shmid_kernel *shp = it;
1138 char *format;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140#define SMALL_STRING "%10d %10d %4o %10u %5u %5u %5d %5u %5u %5u %5u %10lu %10lu %10lu\n"
1141#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 -07001142
Mike Waychison19b49462005-09-06 15:17:10 -07001143 if (sizeof(size_t) <= sizeof(int))
1144 format = SMALL_STRING;
1145 else
1146 format = BIG_STRING;
1147 return seq_printf(s, format,
1148 shp->shm_perm.key,
Nadia Derbey7ca7e562007-10-18 23:40:48 -07001149 shp->shm_perm.id,
Andrew Mortonb33291c2006-01-08 01:02:21 -08001150 shp->shm_perm.mode,
Mike Waychison19b49462005-09-06 15:17:10 -07001151 shp->shm_segsz,
1152 shp->shm_cprid,
1153 shp->shm_lprid,
Eric W. Biedermanbc56bba2007-02-20 13:57:53 -08001154 shp->shm_nattch,
Mike Waychison19b49462005-09-06 15:17:10 -07001155 shp->shm_perm.uid,
1156 shp->shm_perm.gid,
1157 shp->shm_perm.cuid,
1158 shp->shm_perm.cgid,
1159 shp->shm_atim,
1160 shp->shm_dtim,
1161 shp->shm_ctim);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162}
1163#endif