]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/ipc/shm.c | |
4 | * Copyright (C) 1992, 1993 Krishna Balasubramanian | |
5 | * Many improvements/fixes by Bruno Haible. | |
6 | * Replaced `struct shm_desc' by `struct vm_area_struct', July 1994. | |
7 | * Fixed the shm swap deallocation (shm_unuse()), August 1998 Andrea Arcangeli. | |
8 | * | |
9 | * /proc/sysvipc/shm support (c) 1999 Dragos Acostachioaie <dragos@iname.com> | |
10 | * BIGMEM support, Andrea Arcangeli <andrea@suse.de> | |
11 | * SMP thread shm, Jean-Luc Boyard <jean-luc.boyard@siemens.fr> | |
12 | * HIGHMEM support, Ingo Molnar <mingo@redhat.com> | |
13 | * Make shmmax, shmall, shmmni sysctl'able, Christoph Rohland <cr@sap.com> | |
14 | * Shared /dev/zero support, Kanoj Sarcar <kanoj@sgi.com> | |
15 | * Move the mm functionality over to mm/shmem.c, Christoph Rohland <cr@sap.com> | |
16 | * | |
073115d6 SG |
17 | * support for audit of ipc object properties and permission changes |
18 | * Dustin Kirkland <dustin.kirkland@us.ibm.com> | |
4e982311 KK |
19 | * |
20 | * namespaces support | |
21 | * OpenVZ, SWsoft Inc. | |
22 | * Pavel Emelianov <xemul@openvz.org> | |
c2c737a0 DB |
23 | * |
24 | * Better ipc lock (kern_ipc_perm.lock) handling | |
25 | * Davidlohr Bueso <davidlohr.bueso@hp.com>, June 2013. | |
1da177e4 LT |
26 | */ |
27 | ||
1da177e4 LT |
28 | #include <linux/slab.h> |
29 | #include <linux/mm.h> | |
30 | #include <linux/hugetlb.h> | |
31 | #include <linux/shm.h> | |
32 | #include <linux/init.h> | |
33 | #include <linux/file.h> | |
34 | #include <linux/mman.h> | |
1da177e4 LT |
35 | #include <linux/shmem_fs.h> |
36 | #include <linux/security.h> | |
37 | #include <linux/syscalls.h> | |
38 | #include <linux/audit.h> | |
c59ede7b | 39 | #include <linux/capability.h> |
7d87e14c | 40 | #include <linux/ptrace.h> |
19b4946c | 41 | #include <linux/seq_file.h> |
3e148c79 | 42 | #include <linux/rwsem.h> |
4e982311 | 43 | #include <linux/nsproxy.h> |
bc56bba8 | 44 | #include <linux/mount.h> |
ae5e1b22 | 45 | #include <linux/ipc_namespace.h> |
7d87e14c | 46 | |
7153e402 | 47 | #include <linux/uaccess.h> |
1da177e4 LT |
48 | |
49 | #include "util.h" | |
50 | ||
bc56bba8 EB |
51 | struct shm_file_data { |
52 | int id; | |
53 | struct ipc_namespace *ns; | |
54 | struct file *file; | |
55 | const struct vm_operations_struct *vm_ops; | |
56 | }; | |
57 | ||
58 | #define shm_file_data(file) (*((struct shm_file_data **)&(file)->private_data)) | |
59 | ||
9a32144e | 60 | static const struct file_operations shm_file_operations; |
f0f37e2f | 61 | static const struct vm_operations_struct shm_vm_ops; |
1da177e4 | 62 | |
ed2ddbf8 | 63 | #define shm_ids(ns) ((ns)->ids[IPC_SHM_IDS]) |
1da177e4 | 64 | |
4e982311 KK |
65 | #define shm_unlock(shp) \ |
66 | ipc_unlock(&(shp)->shm_perm) | |
1da177e4 | 67 | |
7748dbfa | 68 | static int newseg(struct ipc_namespace *, struct ipc_params *); |
bc56bba8 EB |
69 | static void shm_open(struct vm_area_struct *vma); |
70 | static void shm_close(struct vm_area_struct *vma); | |
239521f3 | 71 | static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp); |
1da177e4 | 72 | #ifdef CONFIG_PROC_FS |
19b4946c | 73 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it); |
1da177e4 LT |
74 | #endif |
75 | ||
0cfb6aee | 76 | int shm_init_ns(struct ipc_namespace *ns) |
4e982311 | 77 | { |
4e982311 KK |
78 | ns->shm_ctlmax = SHMMAX; |
79 | ns->shm_ctlall = SHMALL; | |
80 | ns->shm_ctlmni = SHMMNI; | |
b34a6b1d | 81 | ns->shm_rmid_forced = 0; |
4e982311 | 82 | ns->shm_tot = 0; |
0cfb6aee | 83 | return ipc_init_ids(&shm_ids(ns)); |
4e982311 KK |
84 | } |
85 | ||
f4566f04 | 86 | /* |
d9a605e4 DB |
87 | * Called with shm_ids.rwsem (writer) and the shp structure locked. |
88 | * Only shm_ids.rwsem remains locked on exit. | |
f4566f04 | 89 | */ |
01b8b07a | 90 | static void do_shm_rmid(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp) |
4e982311 | 91 | { |
01b8b07a | 92 | struct shmid_kernel *shp; |
63980c80 | 93 | |
01b8b07a PP |
94 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
95 | ||
239521f3 | 96 | if (shp->shm_nattch) { |
4e982311 KK |
97 | shp->shm_perm.mode |= SHM_DEST; |
98 | /* Do not find it any more */ | |
0cfb6aee | 99 | ipc_set_key_private(&shm_ids(ns), &shp->shm_perm); |
4e982311 KK |
100 | shm_unlock(shp); |
101 | } else | |
102 | shm_destroy(ns, shp); | |
103 | } | |
104 | ||
ae5e1b22 | 105 | #ifdef CONFIG_IPC_NS |
4e982311 KK |
106 | void shm_exit_ns(struct ipc_namespace *ns) |
107 | { | |
01b8b07a | 108 | free_ipcs(ns, &shm_ids(ns), do_shm_rmid); |
7d6feeb2 | 109 | idr_destroy(&ns->ids[IPC_SHM_IDS].ipcs_idr); |
0cfb6aee | 110 | rhashtable_destroy(&ns->ids[IPC_SHM_IDS].key_ht); |
4e982311 | 111 | } |
ae5e1b22 | 112 | #endif |
1da177e4 | 113 | |
140d0b21 | 114 | static int __init ipc_ns_init(void) |
1da177e4 | 115 | { |
0cfb6aee GK |
116 | const int err = shm_init_ns(&init_ipc_ns); |
117 | WARN(err, "ipc: sysv shm_init_ns failed: %d\n", err); | |
118 | return err; | |
140d0b21 LT |
119 | } |
120 | ||
121 | pure_initcall(ipc_ns_init); | |
122 | ||
239521f3 | 123 | void __init shm_init(void) |
140d0b21 | 124 | { |
19b4946c | 125 | ipc_init_proc_interface("sysvipc/shm", |
b7952180 HD |
126 | #if BITS_PER_LONG <= 32 |
127 | " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime rss swap\n", | |
128 | #else | |
129 | " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime rss swap\n", | |
130 | #endif | |
4e982311 | 131 | IPC_SHM_IDS, sysvipc_shm_proc_show); |
1da177e4 LT |
132 | } |
133 | ||
8b8d52ac DB |
134 | static inline struct shmid_kernel *shm_obtain_object(struct ipc_namespace *ns, int id) |
135 | { | |
55b7ae50 | 136 | struct kern_ipc_perm *ipcp = ipc_obtain_object_idr(&shm_ids(ns), id); |
8b8d52ac DB |
137 | |
138 | if (IS_ERR(ipcp)) | |
139 | return ERR_CAST(ipcp); | |
140 | ||
141 | return container_of(ipcp, struct shmid_kernel, shm_perm); | |
142 | } | |
143 | ||
144 | static inline struct shmid_kernel *shm_obtain_object_check(struct ipc_namespace *ns, int id) | |
145 | { | |
146 | struct kern_ipc_perm *ipcp = ipc_obtain_object_check(&shm_ids(ns), id); | |
147 | ||
148 | if (IS_ERR(ipcp)) | |
149 | return ERR_CAST(ipcp); | |
150 | ||
151 | return container_of(ipcp, struct shmid_kernel, shm_perm); | |
152 | } | |
153 | ||
3e148c79 | 154 | /* |
d9a605e4 | 155 | * shm_lock_(check_) routines are called in the paths where the rwsem |
00c2bf85 | 156 | * is not necessarily held. |
3e148c79 | 157 | */ |
023a5355 | 158 | static inline struct shmid_kernel *shm_lock(struct ipc_namespace *ns, int id) |
1da177e4 | 159 | { |
03f02c76 ND |
160 | struct kern_ipc_perm *ipcp = ipc_lock(&shm_ids(ns), id); |
161 | ||
c5c8975b | 162 | /* |
1ac0b6de KS |
163 | * Callers of shm_lock() must validate the status of the returned ipc |
164 | * object pointer (as returned by ipc_lock()), and error out as | |
165 | * appropriate. | |
c5c8975b | 166 | */ |
1ac0b6de KS |
167 | if (IS_ERR(ipcp)) |
168 | return (void *)ipcp; | |
03f02c76 | 169 | return container_of(ipcp, struct shmid_kernel, shm_perm); |
023a5355 ND |
170 | } |
171 | ||
4c677e2e VK |
172 | static inline void shm_lock_by_ptr(struct shmid_kernel *ipcp) |
173 | { | |
174 | rcu_read_lock(); | |
cf9d5d78 | 175 | ipc_lock_object(&ipcp->shm_perm); |
4c677e2e VK |
176 | } |
177 | ||
53dad6d3 DB |
178 | static void shm_rcu_free(struct rcu_head *head) |
179 | { | |
dba4cdd3 MS |
180 | struct kern_ipc_perm *ptr = container_of(head, struct kern_ipc_perm, |
181 | rcu); | |
182 | struct shmid_kernel *shp = container_of(ptr, struct shmid_kernel, | |
183 | shm_perm); | |
53dad6d3 | 184 | security_shm_free(shp); |
42e618f7 | 185 | kvfree(shp); |
53dad6d3 DB |
186 | } |
187 | ||
7ca7e564 | 188 | static inline void shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *s) |
1da177e4 | 189 | { |
ab602f79 | 190 | list_del(&s->shm_clist); |
7ca7e564 | 191 | ipc_rmid(&shm_ids(ns), &s->shm_perm); |
1da177e4 LT |
192 | } |
193 | ||
1da177e4 | 194 | |
1ac0b6de | 195 | static int __shm_open(struct vm_area_struct *vma) |
4e982311 | 196 | { |
bc56bba8 EB |
197 | struct file *file = vma->vm_file; |
198 | struct shm_file_data *sfd = shm_file_data(file); | |
1da177e4 LT |
199 | struct shmid_kernel *shp; |
200 | ||
bc56bba8 | 201 | shp = shm_lock(sfd->ns, sfd->id); |
1ac0b6de KS |
202 | |
203 | if (IS_ERR(shp)) | |
204 | return PTR_ERR(shp); | |
205 | ||
e6118fde EB |
206 | if (shp->shm_file != sfd->file) { |
207 | /* ID was reused */ | |
208 | shm_unlock(shp); | |
209 | return -EINVAL; | |
210 | } | |
211 | ||
7ff2819e | 212 | shp->shm_atim = ktime_get_real_seconds(); |
b488893a | 213 | shp->shm_lprid = task_tgid_vnr(current); |
1da177e4 LT |
214 | shp->shm_nattch++; |
215 | shm_unlock(shp); | |
1ac0b6de KS |
216 | return 0; |
217 | } | |
218 | ||
219 | /* This is called by fork, once for every shm attach. */ | |
220 | static void shm_open(struct vm_area_struct *vma) | |
221 | { | |
222 | int err = __shm_open(vma); | |
223 | /* | |
224 | * We raced in the idr lookup or with shm_destroy(). | |
225 | * Either way, the ID is busted. | |
226 | */ | |
227 | WARN_ON_ONCE(err); | |
1da177e4 LT |
228 | } |
229 | ||
1da177e4 LT |
230 | /* |
231 | * shm_destroy - free the struct shmid_kernel | |
232 | * | |
f4566f04 | 233 | * @ns: namespace |
1da177e4 LT |
234 | * @shp: struct to free |
235 | * | |
d9a605e4 | 236 | * It has to be called with shp and shm_ids.rwsem (writer) locked, |
1da177e4 LT |
237 | * but returns with shp unlocked and freed. |
238 | */ | |
4e982311 | 239 | static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) |
1da177e4 | 240 | { |
a399b29d GT |
241 | struct file *shm_file; |
242 | ||
243 | shm_file = shp->shm_file; | |
244 | shp->shm_file = NULL; | |
4e982311 | 245 | ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT; |
7ca7e564 | 246 | shm_rmid(ns, shp); |
1da177e4 | 247 | shm_unlock(shp); |
a399b29d GT |
248 | if (!is_file_hugepages(shm_file)) |
249 | shmem_lock(shm_file, 0, shp->mlock_user); | |
353d5c30 | 250 | else if (shp->mlock_user) |
07a46ed2 DH |
251 | user_shm_unlock(i_size_read(file_inode(shm_file)), |
252 | shp->mlock_user); | |
a399b29d | 253 | fput(shm_file); |
dba4cdd3 | 254 | ipc_rcu_putref(&shp->shm_perm, shm_rcu_free); |
1da177e4 LT |
255 | } |
256 | ||
b34a6b1d VK |
257 | /* |
258 | * shm_may_destroy - identifies whether shm segment should be destroyed now | |
259 | * | |
260 | * Returns true if and only if there are no active users of the segment and | |
261 | * one of the following is true: | |
262 | * | |
263 | * 1) shmctl(id, IPC_RMID, NULL) was called for this shp | |
264 | * | |
265 | * 2) sysctl kernel.shm_rmid_forced is set to 1. | |
266 | */ | |
267 | static bool shm_may_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) | |
268 | { | |
269 | return (shp->shm_nattch == 0) && | |
270 | (ns->shm_rmid_forced || | |
271 | (shp->shm_perm.mode & SHM_DEST)); | |
272 | } | |
273 | ||
1da177e4 | 274 | /* |
bc56bba8 | 275 | * remove the attach descriptor vma. |
1da177e4 LT |
276 | * free memory for segment if it is marked destroyed. |
277 | * The descriptor has already been removed from the current->mm->mmap list | |
278 | * and will later be kfree()d. | |
279 | */ | |
bc56bba8 | 280 | static void shm_close(struct vm_area_struct *vma) |
1da177e4 | 281 | { |
239521f3 | 282 | struct file *file = vma->vm_file; |
bc56bba8 | 283 | struct shm_file_data *sfd = shm_file_data(file); |
1da177e4 | 284 | struct shmid_kernel *shp; |
bc56bba8 | 285 | struct ipc_namespace *ns = sfd->ns; |
4e982311 | 286 | |
d9a605e4 | 287 | down_write(&shm_ids(ns).rwsem); |
1da177e4 | 288 | /* remove from the list of attaches of the shm segment */ |
00c2bf85 | 289 | shp = shm_lock(ns, sfd->id); |
1ac0b6de KS |
290 | |
291 | /* | |
292 | * We raced in the idr lookup or with shm_destroy(). | |
293 | * Either way, the ID is busted. | |
294 | */ | |
295 | if (WARN_ON_ONCE(IS_ERR(shp))) | |
296 | goto done; /* no-op */ | |
297 | ||
b488893a | 298 | shp->shm_lprid = task_tgid_vnr(current); |
7ff2819e | 299 | shp->shm_dtim = ktime_get_real_seconds(); |
1da177e4 | 300 | shp->shm_nattch--; |
b34a6b1d VK |
301 | if (shm_may_destroy(ns, shp)) |
302 | shm_destroy(ns, shp); | |
303 | else | |
304 | shm_unlock(shp); | |
1ac0b6de | 305 | done: |
d9a605e4 | 306 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
307 | } |
308 | ||
d9a605e4 | 309 | /* Called with ns->shm_ids(ns).rwsem locked */ |
b34a6b1d VK |
310 | static int shm_try_destroy_orphaned(int id, void *p, void *data) |
311 | { | |
312 | struct ipc_namespace *ns = data; | |
4c677e2e VK |
313 | struct kern_ipc_perm *ipcp = p; |
314 | struct shmid_kernel *shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
b34a6b1d VK |
315 | |
316 | /* | |
317 | * We want to destroy segments without users and with already | |
318 | * exit'ed originating process. | |
4c677e2e | 319 | * |
d9a605e4 | 320 | * As shp->* are changed under rwsem, it's safe to skip shp locking. |
b34a6b1d | 321 | */ |
4c677e2e | 322 | if (shp->shm_creator != NULL) |
b34a6b1d | 323 | return 0; |
b34a6b1d | 324 | |
4c677e2e VK |
325 | if (shm_may_destroy(ns, shp)) { |
326 | shm_lock_by_ptr(shp); | |
4e982311 | 327 | shm_destroy(ns, shp); |
4c677e2e | 328 | } |
b34a6b1d VK |
329 | return 0; |
330 | } | |
331 | ||
332 | void shm_destroy_orphaned(struct ipc_namespace *ns) | |
333 | { | |
d9a605e4 | 334 | down_write(&shm_ids(ns).rwsem); |
33a30ed4 | 335 | if (shm_ids(ns).in_use) |
4c677e2e | 336 | idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_orphaned, ns); |
d9a605e4 | 337 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
338 | } |
339 | ||
83293c0f | 340 | /* Locking assumes this will only be called with task == current */ |
b34a6b1d VK |
341 | void exit_shm(struct task_struct *task) |
342 | { | |
4c677e2e | 343 | struct ipc_namespace *ns = task->nsproxy->ipc_ns; |
ab602f79 | 344 | struct shmid_kernel *shp, *n; |
b34a6b1d | 345 | |
83293c0f JM |
346 | if (list_empty(&task->sysvshm.shm_clist)) |
347 | return; | |
348 | ||
349 | /* | |
350 | * If kernel.shm_rmid_forced is not set then only keep track of | |
351 | * which shmids are orphaned, so that a later set of the sysctl | |
352 | * can clean them up. | |
353 | */ | |
354 | if (!ns->shm_rmid_forced) { | |
355 | down_read(&shm_ids(ns).rwsem); | |
356 | list_for_each_entry(shp, &task->sysvshm.shm_clist, shm_clist) | |
357 | shp->shm_creator = NULL; | |
358 | /* | |
359 | * Only under read lock but we are only called on current | |
360 | * so no entry on the list will be shared. | |
361 | */ | |
362 | list_del(&task->sysvshm.shm_clist); | |
363 | up_read(&shm_ids(ns).rwsem); | |
298507d4 | 364 | return; |
83293c0f | 365 | } |
298507d4 | 366 | |
83293c0f JM |
367 | /* |
368 | * Destroy all already created segments, that were not yet mapped, | |
369 | * and mark any mapped as orphan to cover the sysctl toggling. | |
370 | * Destroy is skipped if shm_may_destroy() returns false. | |
371 | */ | |
d9a605e4 | 372 | down_write(&shm_ids(ns).rwsem); |
83293c0f JM |
373 | list_for_each_entry_safe(shp, n, &task->sysvshm.shm_clist, shm_clist) { |
374 | shp->shm_creator = NULL; | |
375 | ||
376 | if (shm_may_destroy(ns, shp)) { | |
377 | shm_lock_by_ptr(shp); | |
378 | shm_destroy(ns, shp); | |
379 | } | |
380 | } | |
381 | ||
382 | /* Remove the list head from any segments still attached. */ | |
ab602f79 | 383 | list_del(&task->sysvshm.shm_clist); |
d9a605e4 | 384 | up_write(&shm_ids(ns).rwsem); |
1da177e4 LT |
385 | } |
386 | ||
11bac800 | 387 | static int shm_fault(struct vm_fault *vmf) |
bc56bba8 | 388 | { |
11bac800 | 389 | struct file *file = vmf->vma->vm_file; |
bc56bba8 EB |
390 | struct shm_file_data *sfd = shm_file_data(file); |
391 | ||
11bac800 | 392 | return sfd->vm_ops->fault(vmf); |
bc56bba8 EB |
393 | } |
394 | ||
62853bef MK |
395 | static int shm_split(struct vm_area_struct *vma, unsigned long addr) |
396 | { | |
397 | struct file *file = vma->vm_file; | |
398 | struct shm_file_data *sfd = shm_file_data(file); | |
399 | ||
400 | if (sfd->vm_ops && sfd->vm_ops->split) | |
401 | return sfd->vm_ops->split(vma, addr); | |
402 | ||
403 | return 0; | |
404 | } | |
405 | ||
bc56bba8 | 406 | #ifdef CONFIG_NUMA |
d823e3e7 | 407 | static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new) |
bc56bba8 EB |
408 | { |
409 | struct file *file = vma->vm_file; | |
410 | struct shm_file_data *sfd = shm_file_data(file); | |
411 | int err = 0; | |
63980c80 | 412 | |
bc56bba8 EB |
413 | if (sfd->vm_ops->set_policy) |
414 | err = sfd->vm_ops->set_policy(vma, new); | |
415 | return err; | |
416 | } | |
417 | ||
d823e3e7 AB |
418 | static struct mempolicy *shm_get_policy(struct vm_area_struct *vma, |
419 | unsigned long addr) | |
bc56bba8 EB |
420 | { |
421 | struct file *file = vma->vm_file; | |
422 | struct shm_file_data *sfd = shm_file_data(file); | |
423 | struct mempolicy *pol = NULL; | |
424 | ||
425 | if (sfd->vm_ops->get_policy) | |
426 | pol = sfd->vm_ops->get_policy(vma, addr); | |
52cd3b07 | 427 | else if (vma->vm_policy) |
bc56bba8 | 428 | pol = vma->vm_policy; |
52cd3b07 | 429 | |
bc56bba8 EB |
430 | return pol; |
431 | } | |
432 | #endif | |
433 | ||
239521f3 | 434 | static int shm_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 435 | { |
bc56bba8 | 436 | struct shm_file_data *sfd = shm_file_data(file); |
b0e15190 DH |
437 | int ret; |
438 | ||
1ac0b6de | 439 | /* |
e6118fde EB |
440 | * In case of remap_file_pages() emulation, the file can represent an |
441 | * IPC ID that was removed, and possibly even reused by another shm | |
442 | * segment already. Propagate this case as an error to caller. | |
1ac0b6de | 443 | */ |
63980c80 | 444 | ret = __shm_open(vma); |
1ac0b6de KS |
445 | if (ret) |
446 | return ret; | |
447 | ||
f74ac015 | 448 | ret = call_mmap(sfd->file, vma); |
1ac0b6de KS |
449 | if (ret) { |
450 | shm_close(vma); | |
bc56bba8 | 451 | return ret; |
1ac0b6de | 452 | } |
bc56bba8 | 453 | sfd->vm_ops = vma->vm_ops; |
2e92a3ba | 454 | #ifdef CONFIG_MMU |
d0edd852 | 455 | WARN_ON(!sfd->vm_ops->fault); |
2e92a3ba | 456 | #endif |
bc56bba8 | 457 | vma->vm_ops = &shm_vm_ops; |
1ac0b6de | 458 | return 0; |
1da177e4 LT |
459 | } |
460 | ||
4e982311 KK |
461 | static int shm_release(struct inode *ino, struct file *file) |
462 | { | |
bc56bba8 | 463 | struct shm_file_data *sfd = shm_file_data(file); |
4e982311 | 464 | |
bc56bba8 | 465 | put_ipc_ns(sfd->ns); |
e6118fde | 466 | fput(sfd->file); |
bc56bba8 EB |
467 | shm_file_data(file) = NULL; |
468 | kfree(sfd); | |
4e982311 KK |
469 | return 0; |
470 | } | |
471 | ||
02c24a82 | 472 | static int shm_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
516dffdc | 473 | { |
516dffdc | 474 | struct shm_file_data *sfd = shm_file_data(file); |
516dffdc | 475 | |
7ea80859 CH |
476 | if (!sfd->file->f_op->fsync) |
477 | return -EINVAL; | |
0f41074a | 478 | return sfd->file->f_op->fsync(sfd->file, start, end, datasync); |
516dffdc AL |
479 | } |
480 | ||
7d8a4569 WD |
481 | static long shm_fallocate(struct file *file, int mode, loff_t offset, |
482 | loff_t len) | |
483 | { | |
484 | struct shm_file_data *sfd = shm_file_data(file); | |
485 | ||
486 | if (!sfd->file->f_op->fallocate) | |
487 | return -EOPNOTSUPP; | |
488 | return sfd->file->f_op->fallocate(file, mode, offset, len); | |
489 | } | |
490 | ||
bc56bba8 EB |
491 | static unsigned long shm_get_unmapped_area(struct file *file, |
492 | unsigned long addr, unsigned long len, unsigned long pgoff, | |
493 | unsigned long flags) | |
494 | { | |
495 | struct shm_file_data *sfd = shm_file_data(file); | |
63980c80 | 496 | |
c4caa778 AV |
497 | return sfd->file->f_op->get_unmapped_area(sfd->file, addr, len, |
498 | pgoff, flags); | |
bc56bba8 | 499 | } |
bc56bba8 | 500 | |
9a32144e | 501 | static const struct file_operations shm_file_operations = { |
4e982311 | 502 | .mmap = shm_mmap, |
516dffdc | 503 | .fsync = shm_fsync, |
4e982311 | 504 | .release = shm_release, |
ed5e5894 | 505 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 506 | .llseek = noop_llseek, |
7d8a4569 | 507 | .fallocate = shm_fallocate, |
c4caa778 AV |
508 | }; |
509 | ||
c01d5b30 HD |
510 | /* |
511 | * shm_file_operations_huge is now identical to shm_file_operations, | |
512 | * but we keep it distinct for the sake of is_file_shm_hugepages(). | |
513 | */ | |
c4caa778 AV |
514 | static const struct file_operations shm_file_operations_huge = { |
515 | .mmap = shm_mmap, | |
516 | .fsync = shm_fsync, | |
517 | .release = shm_release, | |
bc56bba8 | 518 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 519 | .llseek = noop_llseek, |
7d8a4569 | 520 | .fallocate = shm_fallocate, |
1da177e4 LT |
521 | }; |
522 | ||
2954e440 | 523 | bool is_file_shm_hugepages(struct file *file) |
c4caa778 AV |
524 | { |
525 | return file->f_op == &shm_file_operations_huge; | |
526 | } | |
527 | ||
f0f37e2f | 528 | static const struct vm_operations_struct shm_vm_ops = { |
1da177e4 LT |
529 | .open = shm_open, /* callback for a new vm-area open */ |
530 | .close = shm_close, /* callback for when the vm-area is released */ | |
54cb8821 | 531 | .fault = shm_fault, |
62853bef | 532 | .split = shm_split, |
bc56bba8 EB |
533 | #if defined(CONFIG_NUMA) |
534 | .set_policy = shm_set_policy, | |
535 | .get_policy = shm_get_policy, | |
1da177e4 LT |
536 | #endif |
537 | }; | |
538 | ||
f4566f04 ND |
539 | /** |
540 | * newseg - Create a new shared memory segment | |
541 | * @ns: namespace | |
542 | * @params: ptr to the structure that contains key, size and shmflg | |
543 | * | |
d9a605e4 | 544 | * Called with shm_ids.rwsem held as a writer. |
f4566f04 | 545 | */ |
7748dbfa | 546 | static int newseg(struct ipc_namespace *ns, struct ipc_params *params) |
1da177e4 | 547 | { |
7748dbfa ND |
548 | key_t key = params->key; |
549 | int shmflg = params->flg; | |
550 | size_t size = params->u.size; | |
1da177e4 LT |
551 | int error; |
552 | struct shmid_kernel *shp; | |
d69f3bad | 553 | size_t numpages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; |
239521f3 | 554 | struct file *file; |
1da177e4 | 555 | char name[13]; |
ca16d140 | 556 | vm_flags_t acctflag = 0; |
1da177e4 | 557 | |
4e982311 | 558 | if (size < SHMMIN || size > ns->shm_ctlmax) |
1da177e4 LT |
559 | return -EINVAL; |
560 | ||
1376327c MS |
561 | if (numpages << PAGE_SHIFT < size) |
562 | return -ENOSPC; | |
563 | ||
09c6eb1f MS |
564 | if (ns->shm_tot + numpages < ns->shm_tot || |
565 | ns->shm_tot + numpages > ns->shm_ctlall) | |
1da177e4 LT |
566 | return -ENOSPC; |
567 | ||
42e618f7 KC |
568 | shp = kvmalloc(sizeof(*shp), GFP_KERNEL); |
569 | if (unlikely(!shp)) | |
1da177e4 LT |
570 | return -ENOMEM; |
571 | ||
572 | shp->shm_perm.key = key; | |
b33291c0 | 573 | shp->shm_perm.mode = (shmflg & S_IRWXUGO); |
1da177e4 LT |
574 | shp->mlock_user = NULL; |
575 | ||
576 | shp->shm_perm.security = NULL; | |
577 | error = security_shm_alloc(shp); | |
578 | if (error) { | |
42e618f7 | 579 | kvfree(shp); |
1da177e4 LT |
580 | return error; |
581 | } | |
582 | ||
239521f3 | 583 | sprintf(name, "SYSV%08x", key); |
1da177e4 | 584 | if (shmflg & SHM_HUGETLB) { |
c103a4dc | 585 | struct hstate *hs; |
091d0d55 LZ |
586 | size_t hugesize; |
587 | ||
c103a4dc | 588 | hs = hstate_sizelog((shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); |
091d0d55 LZ |
589 | if (!hs) { |
590 | error = -EINVAL; | |
591 | goto no_file; | |
592 | } | |
593 | hugesize = ALIGN(size, huge_page_size(hs)); | |
af73e4d9 | 594 | |
5a6fe125 MG |
595 | /* hugetlb_file_setup applies strict accounting */ |
596 | if (shmflg & SHM_NORESERVE) | |
597 | acctflag = VM_NORESERVE; | |
af73e4d9 | 598 | file = hugetlb_file_setup(name, hugesize, acctflag, |
42d7395f AK |
599 | &shp->mlock_user, HUGETLB_SHMFS_INODE, |
600 | (shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); | |
1da177e4 | 601 | } else { |
bf8f972d BP |
602 | /* |
603 | * Do not allow no accounting for OVERCOMMIT_NEVER, even | |
239521f3 | 604 | * if it's asked for. |
bf8f972d BP |
605 | */ |
606 | if ((shmflg & SHM_NORESERVE) && | |
607 | sysctl_overcommit_memory != OVERCOMMIT_NEVER) | |
fc8744ad | 608 | acctflag = VM_NORESERVE; |
e1832f29 | 609 | file = shmem_kernel_file_setup(name, size, acctflag); |
1da177e4 LT |
610 | } |
611 | error = PTR_ERR(file); | |
612 | if (IS_ERR(file)) | |
613 | goto no_file; | |
614 | ||
b488893a | 615 | shp->shm_cprid = task_tgid_vnr(current); |
1da177e4 LT |
616 | shp->shm_lprid = 0; |
617 | shp->shm_atim = shp->shm_dtim = 0; | |
7ff2819e | 618 | shp->shm_ctim = ktime_get_real_seconds(); |
1da177e4 LT |
619 | shp->shm_segsz = size; |
620 | shp->shm_nattch = 0; | |
1da177e4 | 621 | shp->shm_file = file; |
5774ed01 | 622 | shp->shm_creator = current; |
b9a53227 | 623 | |
39c96a1b | 624 | /* ipc_addid() locks shp upon success. */ |
a2642f87 MS |
625 | error = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni); |
626 | if (error < 0) | |
b9a53227 | 627 | goto no_id; |
b9a53227 | 628 | |
ab602f79 | 629 | list_add(&shp->shm_clist, ¤t->sysvshm.shm_clist); |
dbfcd91f | 630 | |
30475cc1 BP |
631 | /* |
632 | * shmid gets reported as "inode#" in /proc/pid/maps. | |
633 | * proc-ps tools use this. Changing this will break them. | |
634 | */ | |
496ad9aa | 635 | file_inode(file)->i_ino = shp->shm_perm.id; |
551110a9 | 636 | |
4e982311 | 637 | ns->shm_tot += numpages; |
7ca7e564 | 638 | error = shp->shm_perm.id; |
dbfcd91f | 639 | |
cf9d5d78 | 640 | ipc_unlock_object(&shp->shm_perm); |
dbfcd91f | 641 | rcu_read_unlock(); |
7ca7e564 | 642 | return error; |
1da177e4 LT |
643 | |
644 | no_id: | |
2195d281 | 645 | if (is_file_hugepages(file) && shp->mlock_user) |
353d5c30 | 646 | user_shm_unlock(size, shp->mlock_user); |
1da177e4 LT |
647 | fput(file); |
648 | no_file: | |
a2642f87 | 649 | call_rcu(&shp->shm_perm.rcu, shm_rcu_free); |
1da177e4 LT |
650 | return error; |
651 | } | |
652 | ||
f4566f04 | 653 | /* |
d9a605e4 | 654 | * Called with shm_ids.rwsem and ipcp locked. |
f4566f04 | 655 | */ |
03f02c76 | 656 | static inline int shm_security(struct kern_ipc_perm *ipcp, int shmflg) |
7748dbfa | 657 | { |
03f02c76 ND |
658 | struct shmid_kernel *shp; |
659 | ||
660 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
661 | return security_shm_associate(shp, shmflg); | |
7748dbfa ND |
662 | } |
663 | ||
f4566f04 | 664 | /* |
d9a605e4 | 665 | * Called with shm_ids.rwsem and ipcp locked. |
f4566f04 | 666 | */ |
03f02c76 ND |
667 | static inline int shm_more_checks(struct kern_ipc_perm *ipcp, |
668 | struct ipc_params *params) | |
7748dbfa | 669 | { |
03f02c76 ND |
670 | struct shmid_kernel *shp; |
671 | ||
672 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
673 | if (shp->shm_segsz < params->u.size) | |
7748dbfa ND |
674 | return -EINVAL; |
675 | ||
676 | return 0; | |
677 | } | |
678 | ||
d5460c99 | 679 | SYSCALL_DEFINE3(shmget, key_t, key, size_t, size, int, shmflg) |
1da177e4 | 680 | { |
4e982311 | 681 | struct ipc_namespace *ns; |
eb66ec44 MK |
682 | static const struct ipc_ops shm_ops = { |
683 | .getnew = newseg, | |
684 | .associate = shm_security, | |
685 | .more_checks = shm_more_checks, | |
686 | }; | |
7748dbfa | 687 | struct ipc_params shm_params; |
4e982311 KK |
688 | |
689 | ns = current->nsproxy->ipc_ns; | |
1da177e4 | 690 | |
7748dbfa ND |
691 | shm_params.key = key; |
692 | shm_params.flg = shmflg; | |
693 | shm_params.u.size = size; | |
1da177e4 | 694 | |
7748dbfa | 695 | return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params); |
1da177e4 LT |
696 | } |
697 | ||
698 | static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version) | |
699 | { | |
239521f3 | 700 | switch (version) { |
1da177e4 LT |
701 | case IPC_64: |
702 | return copy_to_user(buf, in, sizeof(*in)); | |
703 | case IPC_OLD: | |
704 | { | |
705 | struct shmid_ds out; | |
706 | ||
3af54c9b | 707 | memset(&out, 0, sizeof(out)); |
1da177e4 LT |
708 | ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm); |
709 | out.shm_segsz = in->shm_segsz; | |
710 | out.shm_atime = in->shm_atime; | |
711 | out.shm_dtime = in->shm_dtime; | |
712 | out.shm_ctime = in->shm_ctime; | |
713 | out.shm_cpid = in->shm_cpid; | |
714 | out.shm_lpid = in->shm_lpid; | |
715 | out.shm_nattch = in->shm_nattch; | |
716 | ||
717 | return copy_to_user(buf, &out, sizeof(out)); | |
718 | } | |
719 | default: | |
720 | return -EINVAL; | |
721 | } | |
722 | } | |
723 | ||
016d7132 PP |
724 | static inline unsigned long |
725 | copy_shmid_from_user(struct shmid64_ds *out, void __user *buf, int version) | |
1da177e4 | 726 | { |
239521f3 | 727 | switch (version) { |
1da177e4 | 728 | case IPC_64: |
016d7132 | 729 | if (copy_from_user(out, buf, sizeof(*out))) |
1da177e4 | 730 | return -EFAULT; |
1da177e4 | 731 | return 0; |
1da177e4 LT |
732 | case IPC_OLD: |
733 | { | |
734 | struct shmid_ds tbuf_old; | |
735 | ||
736 | if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) | |
737 | return -EFAULT; | |
738 | ||
016d7132 PP |
739 | out->shm_perm.uid = tbuf_old.shm_perm.uid; |
740 | out->shm_perm.gid = tbuf_old.shm_perm.gid; | |
741 | out->shm_perm.mode = tbuf_old.shm_perm.mode; | |
1da177e4 LT |
742 | |
743 | return 0; | |
744 | } | |
745 | default: | |
746 | return -EINVAL; | |
747 | } | |
748 | } | |
749 | ||
750 | static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version) | |
751 | { | |
239521f3 | 752 | switch (version) { |
1da177e4 LT |
753 | case IPC_64: |
754 | return copy_to_user(buf, in, sizeof(*in)); | |
755 | case IPC_OLD: | |
756 | { | |
757 | struct shminfo out; | |
758 | ||
239521f3 | 759 | if (in->shmmax > INT_MAX) |
1da177e4 LT |
760 | out.shmmax = INT_MAX; |
761 | else | |
762 | out.shmmax = (int)in->shmmax; | |
763 | ||
764 | out.shmmin = in->shmmin; | |
765 | out.shmmni = in->shmmni; | |
766 | out.shmseg = in->shmseg; | |
46c0a8ca | 767 | out.shmall = in->shmall; |
1da177e4 LT |
768 | |
769 | return copy_to_user(buf, &out, sizeof(out)); | |
770 | } | |
771 | default: | |
772 | return -EINVAL; | |
773 | } | |
774 | } | |
775 | ||
b7952180 HD |
776 | /* |
777 | * Calculate and add used RSS and swap pages of a shm. | |
d9a605e4 | 778 | * Called with shm_ids.rwsem held as a reader |
b7952180 HD |
779 | */ |
780 | static void shm_add_rss_swap(struct shmid_kernel *shp, | |
781 | unsigned long *rss_add, unsigned long *swp_add) | |
782 | { | |
783 | struct inode *inode; | |
784 | ||
496ad9aa | 785 | inode = file_inode(shp->shm_file); |
b7952180 HD |
786 | |
787 | if (is_file_hugepages(shp->shm_file)) { | |
788 | struct address_space *mapping = inode->i_mapping; | |
789 | struct hstate *h = hstate_file(shp->shm_file); | |
790 | *rss_add += pages_per_huge_page(h) * mapping->nrpages; | |
791 | } else { | |
792 | #ifdef CONFIG_SHMEM | |
793 | struct shmem_inode_info *info = SHMEM_I(inode); | |
63980c80 | 794 | |
4595ef88 | 795 | spin_lock_irq(&info->lock); |
b7952180 HD |
796 | *rss_add += inode->i_mapping->nrpages; |
797 | *swp_add += info->swapped; | |
4595ef88 | 798 | spin_unlock_irq(&info->lock); |
b7952180 HD |
799 | #else |
800 | *rss_add += inode->i_mapping->nrpages; | |
801 | #endif | |
802 | } | |
803 | } | |
804 | ||
f4566f04 | 805 | /* |
d9a605e4 | 806 | * Called with shm_ids.rwsem held as a reader |
f4566f04 | 807 | */ |
4e982311 KK |
808 | static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss, |
809 | unsigned long *swp) | |
1da177e4 | 810 | { |
7ca7e564 ND |
811 | int next_id; |
812 | int total, in_use; | |
1da177e4 LT |
813 | |
814 | *rss = 0; | |
815 | *swp = 0; | |
816 | ||
7ca7e564 ND |
817 | in_use = shm_ids(ns).in_use; |
818 | ||
819 | for (total = 0, next_id = 0; total < in_use; next_id++) { | |
e562aebc | 820 | struct kern_ipc_perm *ipc; |
1da177e4 | 821 | struct shmid_kernel *shp; |
1da177e4 | 822 | |
e562aebc TB |
823 | ipc = idr_find(&shm_ids(ns).ipcs_idr, next_id); |
824 | if (ipc == NULL) | |
1da177e4 | 825 | continue; |
e562aebc | 826 | shp = container_of(ipc, struct shmid_kernel, shm_perm); |
1da177e4 | 827 | |
b7952180 | 828 | shm_add_rss_swap(shp, rss, swp); |
7ca7e564 ND |
829 | |
830 | total++; | |
1da177e4 LT |
831 | } |
832 | } | |
833 | ||
8d4cc8b5 | 834 | /* |
d9a605e4 | 835 | * This function handles some shmctl commands which require the rwsem |
8d4cc8b5 | 836 | * to be held in write mode. |
d9a605e4 | 837 | * NOTE: no locks must be held, the rwsem is taken inside this function. |
8d4cc8b5 PP |
838 | */ |
839 | static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd, | |
9ba720c1 | 840 | struct shmid64_ds *shmid64) |
1da177e4 | 841 | { |
8d4cc8b5 | 842 | struct kern_ipc_perm *ipcp; |
8d4cc8b5 PP |
843 | struct shmid_kernel *shp; |
844 | int err; | |
845 | ||
d9a605e4 | 846 | down_write(&shm_ids(ns).rwsem); |
7b4cc5d8 DB |
847 | rcu_read_lock(); |
848 | ||
79ccf0f8 | 849 | ipcp = ipcctl_pre_down_nolock(ns, &shm_ids(ns), shmid, cmd, |
9ba720c1 | 850 | &shmid64->shm_perm, 0); |
7b4cc5d8 DB |
851 | if (IS_ERR(ipcp)) { |
852 | err = PTR_ERR(ipcp); | |
7b4cc5d8 DB |
853 | goto out_unlock1; |
854 | } | |
8d4cc8b5 | 855 | |
a5f75e7f | 856 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
8d4cc8b5 PP |
857 | |
858 | err = security_shm_shmctl(shp, cmd); | |
859 | if (err) | |
79ccf0f8 | 860 | goto out_unlock1; |
7b4cc5d8 | 861 | |
8d4cc8b5 PP |
862 | switch (cmd) { |
863 | case IPC_RMID: | |
79ccf0f8 | 864 | ipc_lock_object(&shp->shm_perm); |
7b4cc5d8 | 865 | /* do_shm_rmid unlocks the ipc object and rcu */ |
8d4cc8b5 PP |
866 | do_shm_rmid(ns, ipcp); |
867 | goto out_up; | |
868 | case IPC_SET: | |
79ccf0f8 | 869 | ipc_lock_object(&shp->shm_perm); |
9ba720c1 | 870 | err = ipc_update_perm(&shmid64->shm_perm, ipcp); |
1efdb69b | 871 | if (err) |
7b4cc5d8 | 872 | goto out_unlock0; |
7ff2819e | 873 | shp->shm_ctim = ktime_get_real_seconds(); |
8d4cc8b5 PP |
874 | break; |
875 | default: | |
876 | err = -EINVAL; | |
79ccf0f8 | 877 | goto out_unlock1; |
8d4cc8b5 | 878 | } |
7b4cc5d8 DB |
879 | |
880 | out_unlock0: | |
881 | ipc_unlock_object(&shp->shm_perm); | |
882 | out_unlock1: | |
883 | rcu_read_unlock(); | |
8d4cc8b5 | 884 | out_up: |
d9a605e4 | 885 | up_write(&shm_ids(ns).rwsem); |
8d4cc8b5 PP |
886 | return err; |
887 | } | |
888 | ||
9ba720c1 AV |
889 | static int shmctl_ipc_info(struct ipc_namespace *ns, |
890 | struct shminfo64 *shminfo) | |
8d4cc8b5 | 891 | { |
9ba720c1 AV |
892 | int err = security_shm_shmctl(NULL, IPC_INFO); |
893 | if (!err) { | |
894 | memset(shminfo, 0, sizeof(*shminfo)); | |
895 | shminfo->shmmni = shminfo->shmseg = ns->shm_ctlmni; | |
896 | shminfo->shmmax = ns->shm_ctlmax; | |
897 | shminfo->shmall = ns->shm_ctlall; | |
898 | shminfo->shmmin = SHMMIN; | |
d9a605e4 | 899 | down_read(&shm_ids(ns).rwsem); |
7ca7e564 | 900 | err = ipc_get_maxid(&shm_ids(ns)); |
d9a605e4 | 901 | up_read(&shm_ids(ns).rwsem); |
239521f3 | 902 | if (err < 0) |
1da177e4 | 903 | err = 0; |
1da177e4 | 904 | } |
9ba720c1 AV |
905 | return err; |
906 | } | |
1da177e4 | 907 | |
9ba720c1 AV |
908 | static int shmctl_shm_info(struct ipc_namespace *ns, |
909 | struct shm_info *shm_info) | |
910 | { | |
911 | int err = security_shm_shmctl(NULL, SHM_INFO); | |
912 | if (!err) { | |
913 | memset(shm_info, 0, sizeof(*shm_info)); | |
d9a605e4 | 914 | down_read(&shm_ids(ns).rwsem); |
9ba720c1 AV |
915 | shm_info->used_ids = shm_ids(ns).in_use; |
916 | shm_get_stat(ns, &shm_info->shm_rss, &shm_info->shm_swp); | |
917 | shm_info->shm_tot = ns->shm_tot; | |
918 | shm_info->swap_attempts = 0; | |
919 | shm_info->swap_successes = 0; | |
7ca7e564 | 920 | err = ipc_get_maxid(&shm_ids(ns)); |
d9a605e4 | 921 | up_read(&shm_ids(ns).rwsem); |
9ba720c1 AV |
922 | if (err < 0) |
923 | err = 0; | |
1da177e4 | 924 | } |
9ba720c1 AV |
925 | return err; |
926 | } | |
c97cb9cc | 927 | |
9ba720c1 AV |
928 | static int shmctl_stat(struct ipc_namespace *ns, int shmid, |
929 | int cmd, struct shmid64_ds *tbuf) | |
930 | { | |
931 | struct shmid_kernel *shp; | |
932 | int result; | |
933 | int err; | |
c97cb9cc | 934 | |
9ba720c1 AV |
935 | rcu_read_lock(); |
936 | if (cmd == SHM_STAT) { | |
937 | shp = shm_obtain_object(ns, shmid); | |
938 | if (IS_ERR(shp)) { | |
939 | err = PTR_ERR(shp); | |
940 | goto out_unlock; | |
941 | } | |
942 | result = shp->shm_perm.id; | |
943 | } else { | |
944 | shp = shm_obtain_object_check(ns, shmid); | |
945 | if (IS_ERR(shp)) { | |
946 | err = PTR_ERR(shp); | |
1da177e4 | 947 | goto out_unlock; |
1da177e4 | 948 | } |
9ba720c1 AV |
949 | result = 0; |
950 | } | |
1da177e4 | 951 | |
9ba720c1 AV |
952 | err = -EACCES; |
953 | if (ipcperms(ns, &shp->shm_perm, S_IRUGO)) | |
954 | goto out_unlock; | |
c97cb9cc | 955 | |
9ba720c1 AV |
956 | err = security_shm_shmctl(shp, cmd); |
957 | if (err) | |
958 | goto out_unlock; | |
959 | ||
960 | memset(tbuf, 0, sizeof(*tbuf)); | |
961 | kernel_to_ipc64_perm(&shp->shm_perm, &tbuf->shm_perm); | |
962 | tbuf->shm_segsz = shp->shm_segsz; | |
963 | tbuf->shm_atime = shp->shm_atim; | |
964 | tbuf->shm_dtime = shp->shm_dtim; | |
965 | tbuf->shm_ctime = shp->shm_ctim; | |
966 | tbuf->shm_cpid = shp->shm_cprid; | |
967 | tbuf->shm_lpid = shp->shm_lprid; | |
968 | tbuf->shm_nattch = shp->shm_nattch; | |
969 | rcu_read_unlock(); | |
970 | return result; | |
68eccc1d DB |
971 | |
972 | out_unlock: | |
c97cb9cc | 973 | rcu_read_unlock(); |
68eccc1d DB |
974 | return err; |
975 | } | |
976 | ||
9ba720c1 | 977 | static int shmctl_do_lock(struct ipc_namespace *ns, int shmid, int cmd) |
68eccc1d DB |
978 | { |
979 | struct shmid_kernel *shp; | |
9ba720c1 AV |
980 | struct file *shm_file; |
981 | int err; | |
68eccc1d | 982 | |
9ba720c1 AV |
983 | rcu_read_lock(); |
984 | shp = shm_obtain_object_check(ns, shmid); | |
985 | if (IS_ERR(shp)) { | |
986 | err = PTR_ERR(shp); | |
987 | goto out_unlock1; | |
1da177e4 | 988 | } |
c97cb9cc | 989 | |
9ba720c1 AV |
990 | audit_ipc_obj(&(shp->shm_perm)); |
991 | err = security_shm_shmctl(shp, cmd); | |
992 | if (err) | |
993 | goto out_unlock1; | |
c97cb9cc | 994 | |
9ba720c1 | 995 | ipc_lock_object(&shp->shm_perm); |
c97cb9cc | 996 | |
9ba720c1 AV |
997 | /* check if shm_destroy() is tearing down shp */ |
998 | if (!ipc_valid_object(&shp->shm_perm)) { | |
999 | err = -EIDRM; | |
1000 | goto out_unlock0; | |
1da177e4 | 1001 | } |
073115d6 | 1002 | |
9ba720c1 AV |
1003 | if (!ns_capable(ns->user_ns, CAP_IPC_LOCK)) { |
1004 | kuid_t euid = current_euid(); | |
0f3d2b01 | 1005 | |
9ba720c1 AV |
1006 | if (!uid_eq(euid, shp->shm_perm.uid) && |
1007 | !uid_eq(euid, shp->shm_perm.cuid)) { | |
1008 | err = -EPERM; | |
0f3d2b01 RA |
1009 | goto out_unlock0; |
1010 | } | |
9ba720c1 AV |
1011 | if (cmd == SHM_LOCK && !rlimit(RLIMIT_MEMLOCK)) { |
1012 | err = -EPERM; | |
1013 | goto out_unlock0; | |
1da177e4 | 1014 | } |
68eccc1d DB |
1015 | } |
1016 | ||
9ba720c1 AV |
1017 | shm_file = shp->shm_file; |
1018 | if (is_file_hugepages(shm_file)) | |
1019 | goto out_unlock0; | |
85046579 | 1020 | |
9ba720c1 AV |
1021 | if (cmd == SHM_LOCK) { |
1022 | struct user_struct *user = current_user(); | |
63980c80 | 1023 | |
9ba720c1 AV |
1024 | err = shmem_lock(shm_file, 1, user); |
1025 | if (!err && !(shp->shm_perm.mode & SHM_LOCKED)) { | |
1026 | shp->shm_perm.mode |= SHM_LOCKED; | |
1027 | shp->mlock_user = user; | |
1da177e4 | 1028 | } |
9ba720c1 | 1029 | goto out_unlock0; |
1da177e4 LT |
1030 | } |
1031 | ||
9ba720c1 AV |
1032 | /* SHM_UNLOCK */ |
1033 | if (!(shp->shm_perm.mode & SHM_LOCKED)) | |
1034 | goto out_unlock0; | |
1035 | shmem_lock(shm_file, 0, shp->mlock_user); | |
1036 | shp->shm_perm.mode &= ~SHM_LOCKED; | |
1037 | shp->mlock_user = NULL; | |
1038 | get_file(shm_file); | |
1039 | ipc_unlock_object(&shp->shm_perm); | |
1040 | rcu_read_unlock(); | |
1041 | shmem_unlock_mapping(shm_file->f_mapping); | |
1042 | ||
1043 | fput(shm_file); | |
1044 | return err; | |
1045 | ||
2caacaa8 DB |
1046 | out_unlock0: |
1047 | ipc_unlock_object(&shp->shm_perm); | |
1048 | out_unlock1: | |
c97cb9cc | 1049 | rcu_read_unlock(); |
68eccc1d DB |
1050 | return err; |
1051 | } | |
1052 | ||
1053 | SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, struct shmid_ds __user *, buf) | |
1054 | { | |
68eccc1d DB |
1055 | int err, version; |
1056 | struct ipc_namespace *ns; | |
553f770e | 1057 | struct shmid64_ds sem64; |
68eccc1d | 1058 | |
2caacaa8 DB |
1059 | if (cmd < 0 || shmid < 0) |
1060 | return -EINVAL; | |
68eccc1d DB |
1061 | |
1062 | version = ipc_parse_version(&cmd); | |
1063 | ns = current->nsproxy->ipc_ns; | |
1064 | ||
1065 | switch (cmd) { | |
9ba720c1 AV |
1066 | case IPC_INFO: { |
1067 | struct shminfo64 shminfo; | |
1068 | err = shmctl_ipc_info(ns, &shminfo); | |
1069 | if (err < 0) | |
1070 | return err; | |
1071 | if (copy_shminfo_to_user(buf, &shminfo, version)) | |
1072 | err = -EFAULT; | |
1073 | return err; | |
1074 | } | |
1075 | case SHM_INFO: { | |
1076 | struct shm_info shm_info; | |
1077 | err = shmctl_shm_info(ns, &shm_info); | |
1078 | if (err < 0) | |
1079 | return err; | |
1080 | if (copy_to_user(buf, &shm_info, sizeof(shm_info))) | |
1081 | err = -EFAULT; | |
1082 | return err; | |
1083 | } | |
68eccc1d | 1084 | case SHM_STAT: |
9ba720c1 | 1085 | case IPC_STAT: { |
553f770e | 1086 | err = shmctl_stat(ns, shmid, cmd, &sem64); |
9ba720c1 AV |
1087 | if (err < 0) |
1088 | return err; | |
553f770e | 1089 | if (copy_shmid_to_user(buf, &sem64, version)) |
9ba720c1 AV |
1090 | err = -EFAULT; |
1091 | return err; | |
1092 | } | |
2caacaa8 | 1093 | case IPC_SET: |
553f770e | 1094 | if (copy_shmid_from_user(&sem64, buf, version)) |
9ba720c1 | 1095 | return -EFAULT; |
553f770e | 1096 | /* fallthru */ |
9ba720c1 | 1097 | case IPC_RMID: |
553f770e | 1098 | return shmctl_down(ns, shmid, cmd, &sem64); |
1da177e4 LT |
1099 | case SHM_LOCK: |
1100 | case SHM_UNLOCK: | |
9ba720c1 AV |
1101 | return shmctl_do_lock(ns, shmid, cmd); |
1102 | default: | |
1103 | return -EINVAL; | |
1104 | } | |
1105 | } | |
89e004ea | 1106 | |
553f770e AV |
1107 | #ifdef CONFIG_COMPAT |
1108 | ||
1109 | struct compat_shmid_ds { | |
1110 | struct compat_ipc_perm shm_perm; | |
1111 | int shm_segsz; | |
1112 | compat_time_t shm_atime; | |
1113 | compat_time_t shm_dtime; | |
1114 | compat_time_t shm_ctime; | |
1115 | compat_ipc_pid_t shm_cpid; | |
1116 | compat_ipc_pid_t shm_lpid; | |
1117 | unsigned short shm_nattch; | |
1118 | unsigned short shm_unused; | |
1119 | compat_uptr_t shm_unused2; | |
1120 | compat_uptr_t shm_unused3; | |
1121 | }; | |
1da177e4 | 1122 | |
553f770e AV |
1123 | struct compat_shminfo64 { |
1124 | compat_ulong_t shmmax; | |
1125 | compat_ulong_t shmmin; | |
1126 | compat_ulong_t shmmni; | |
1127 | compat_ulong_t shmseg; | |
1128 | compat_ulong_t shmall; | |
1129 | compat_ulong_t __unused1; | |
1130 | compat_ulong_t __unused2; | |
1131 | compat_ulong_t __unused3; | |
1132 | compat_ulong_t __unused4; | |
1133 | }; | |
073115d6 | 1134 | |
553f770e AV |
1135 | struct compat_shm_info { |
1136 | compat_int_t used_ids; | |
1137 | compat_ulong_t shm_tot, shm_rss, shm_swp; | |
1138 | compat_ulong_t swap_attempts, swap_successes; | |
1139 | }; | |
0f3d2b01 | 1140 | |
553f770e AV |
1141 | static int copy_compat_shminfo_to_user(void __user *buf, struct shminfo64 *in, |
1142 | int version) | |
1143 | { | |
1144 | if (in->shmmax > INT_MAX) | |
1145 | in->shmmax = INT_MAX; | |
1146 | if (version == IPC_64) { | |
1147 | struct compat_shminfo64 info; | |
1148 | memset(&info, 0, sizeof(info)); | |
1149 | info.shmmax = in->shmmax; | |
1150 | info.shmmin = in->shmmin; | |
1151 | info.shmmni = in->shmmni; | |
1152 | info.shmseg = in->shmseg; | |
1153 | info.shmall = in->shmall; | |
1154 | return copy_to_user(buf, &info, sizeof(info)); | |
1155 | } else { | |
1156 | struct shminfo info; | |
1157 | memset(&info, 0, sizeof(info)); | |
1158 | info.shmmax = in->shmmax; | |
1159 | info.shmmin = in->shmmin; | |
1160 | info.shmmni = in->shmmni; | |
1161 | info.shmseg = in->shmseg; | |
1162 | info.shmall = in->shmall; | |
1163 | return copy_to_user(buf, &info, sizeof(info)); | |
1164 | } | |
1165 | } | |
0f3d2b01 | 1166 | |
553f770e AV |
1167 | static int put_compat_shm_info(struct shm_info *ip, |
1168 | struct compat_shm_info __user *uip) | |
1169 | { | |
1170 | struct compat_shm_info info; | |
1171 | ||
1172 | memset(&info, 0, sizeof(info)); | |
1173 | info.used_ids = ip->used_ids; | |
1174 | info.shm_tot = ip->shm_tot; | |
1175 | info.shm_rss = ip->shm_rss; | |
1176 | info.shm_swp = ip->shm_swp; | |
1177 | info.swap_attempts = ip->swap_attempts; | |
1178 | info.swap_successes = ip->swap_successes; | |
b776e4b1 | 1179 | return copy_to_user(uip, &info, sizeof(info)); |
553f770e | 1180 | } |
1da177e4 | 1181 | |
553f770e AV |
1182 | static int copy_compat_shmid_to_user(void __user *buf, struct shmid64_ds *in, |
1183 | int version) | |
1184 | { | |
1185 | if (version == IPC_64) { | |
1186 | struct compat_shmid64_ds v; | |
1187 | memset(&v, 0, sizeof(v)); | |
28327fae | 1188 | to_compat_ipc64_perm(&v.shm_perm, &in->shm_perm); |
553f770e AV |
1189 | v.shm_atime = in->shm_atime; |
1190 | v.shm_dtime = in->shm_dtime; | |
1191 | v.shm_ctime = in->shm_ctime; | |
1192 | v.shm_segsz = in->shm_segsz; | |
1193 | v.shm_nattch = in->shm_nattch; | |
1194 | v.shm_cpid = in->shm_cpid; | |
1195 | v.shm_lpid = in->shm_lpid; | |
1196 | return copy_to_user(buf, &v, sizeof(v)); | |
1197 | } else { | |
1198 | struct compat_shmid_ds v; | |
1199 | memset(&v, 0, sizeof(v)); | |
28327fae | 1200 | to_compat_ipc_perm(&v.shm_perm, &in->shm_perm); |
553f770e | 1201 | v.shm_perm.key = in->shm_perm.key; |
553f770e AV |
1202 | v.shm_atime = in->shm_atime; |
1203 | v.shm_dtime = in->shm_dtime; | |
1204 | v.shm_ctime = in->shm_ctime; | |
1205 | v.shm_segsz = in->shm_segsz; | |
1206 | v.shm_nattch = in->shm_nattch; | |
1207 | v.shm_cpid = in->shm_cpid; | |
1208 | v.shm_lpid = in->shm_lpid; | |
1209 | return copy_to_user(buf, &v, sizeof(v)); | |
1210 | } | |
1211 | } | |
85046579 | 1212 | |
553f770e AV |
1213 | static int copy_compat_shmid_from_user(struct shmid64_ds *out, void __user *buf, |
1214 | int version) | |
1215 | { | |
1216 | memset(out, 0, sizeof(*out)); | |
1217 | if (version == IPC_64) { | |
6aa211e8 | 1218 | struct compat_shmid64_ds __user *p = buf; |
28327fae | 1219 | return get_compat_ipc64_perm(&out->shm_perm, &p->shm_perm); |
553f770e | 1220 | } else { |
6aa211e8 | 1221 | struct compat_shmid_ds __user *p = buf; |
28327fae | 1222 | return get_compat_ipc_perm(&out->shm_perm, &p->shm_perm); |
553f770e | 1223 | } |
553f770e | 1224 | } |
63980c80 | 1225 | |
553f770e AV |
1226 | COMPAT_SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, void __user *, uptr) |
1227 | { | |
1228 | struct ipc_namespace *ns; | |
1229 | struct shmid64_ds sem64; | |
1230 | int version = compat_ipc_parse_version(&cmd); | |
1231 | int err; | |
85046579 | 1232 | |
553f770e AV |
1233 | ns = current->nsproxy->ipc_ns; |
1234 | ||
1235 | if (cmd < 0 || shmid < 0) | |
1236 | return -EINVAL; | |
2caacaa8 | 1237 | |
553f770e AV |
1238 | switch (cmd) { |
1239 | case IPC_INFO: { | |
1240 | struct shminfo64 shminfo; | |
1241 | err = shmctl_ipc_info(ns, &shminfo); | |
1242 | if (err < 0) | |
1243 | return err; | |
1244 | if (copy_compat_shminfo_to_user(uptr, &shminfo, version)) | |
1245 | err = -EFAULT; | |
1246 | return err; | |
1247 | } | |
1248 | case SHM_INFO: { | |
1249 | struct shm_info shm_info; | |
1250 | err = shmctl_shm_info(ns, &shm_info); | |
1251 | if (err < 0) | |
1252 | return err; | |
1253 | if (put_compat_shm_info(&shm_info, uptr)) | |
1254 | err = -EFAULT; | |
8d4cc8b5 | 1255 | return err; |
2caacaa8 | 1256 | } |
553f770e AV |
1257 | case IPC_STAT: |
1258 | case SHM_STAT: | |
1259 | err = shmctl_stat(ns, shmid, cmd, &sem64); | |
1260 | if (err < 0) | |
1261 | return err; | |
58aff0af | 1262 | if (copy_compat_shmid_to_user(uptr, &sem64, version)) |
553f770e AV |
1263 | err = -EFAULT; |
1264 | return err; | |
1265 | ||
1266 | case IPC_SET: | |
1267 | if (copy_compat_shmid_from_user(&sem64, uptr, version)) | |
1268 | return -EFAULT; | |
1269 | /* fallthru */ | |
1270 | case IPC_RMID: | |
1271 | return shmctl_down(ns, shmid, cmd, &sem64); | |
1272 | case SHM_LOCK: | |
1273 | case SHM_UNLOCK: | |
1274 | return shmctl_do_lock(ns, shmid, cmd); | |
1275 | break; | |
1da177e4 | 1276 | default: |
8d4cc8b5 | 1277 | return -EINVAL; |
1da177e4 | 1278 | } |
1da177e4 LT |
1279 | return err; |
1280 | } | |
553f770e | 1281 | #endif |
1da177e4 LT |
1282 | |
1283 | /* | |
1284 | * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists. | |
1285 | * | |
1286 | * NOTE! Despite the name, this is NOT a direct system call entrypoint. The | |
1287 | * "raddr" thing points to kernel space, and there has to be a wrapper around | |
1288 | * this. | |
1289 | */ | |
95e91b83 DB |
1290 | long do_shmat(int shmid, char __user *shmaddr, int shmflg, |
1291 | ulong *raddr, unsigned long shmlba) | |
1da177e4 LT |
1292 | { |
1293 | struct shmid_kernel *shp; | |
f0cb8802 | 1294 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1295 | unsigned long size; |
239521f3 | 1296 | struct file *file; |
1da177e4 | 1297 | int err; |
f0cb8802 | 1298 | unsigned long flags = MAP_SHARED; |
1da177e4 | 1299 | unsigned long prot; |
1da177e4 | 1300 | int acc_mode; |
4e982311 | 1301 | struct ipc_namespace *ns; |
bc56bba8 EB |
1302 | struct shm_file_data *sfd; |
1303 | struct path path; | |
aeb5d727 | 1304 | fmode_t f_mode; |
41badc15 | 1305 | unsigned long populate = 0; |
1da177e4 | 1306 | |
bc56bba8 EB |
1307 | err = -EINVAL; |
1308 | if (shmid < 0) | |
1da177e4 | 1309 | goto out; |
f0cb8802 DB |
1310 | |
1311 | if (addr) { | |
079a96ae | 1312 | if (addr & (shmlba - 1)) { |
438b8789 | 1313 | if (shmflg & SHM_RND) { |
74389f5d | 1314 | addr &= ~(shmlba - 1); /* round down */ |
438b8789 DB |
1315 | |
1316 | /* | |
1317 | * Ensure that the round-down is non-nil | |
1318 | * when remapping. This can happen for | |
1319 | * cases when addr < shmlba. | |
1320 | */ | |
1321 | if (!addr && (shmflg & SHM_REMAP)) | |
1322 | goto out; | |
1323 | } else | |
1da177e4 LT |
1324 | #ifndef __ARCH_FORCE_SHMLBA |
1325 | if (addr & ~PAGE_MASK) | |
1326 | #endif | |
bc56bba8 | 1327 | goto out; |
1da177e4 | 1328 | } |
1da177e4 | 1329 | |
f0cb8802 DB |
1330 | flags |= MAP_FIXED; |
1331 | } else if ((shmflg & SHM_REMAP)) | |
1332 | goto out; | |
1da177e4 LT |
1333 | |
1334 | if (shmflg & SHM_RDONLY) { | |
1335 | prot = PROT_READ; | |
1da177e4 | 1336 | acc_mode = S_IRUGO; |
bc56bba8 | 1337 | f_mode = FMODE_READ; |
1da177e4 LT |
1338 | } else { |
1339 | prot = PROT_READ | PROT_WRITE; | |
1da177e4 | 1340 | acc_mode = S_IRUGO | S_IWUGO; |
bc56bba8 | 1341 | f_mode = FMODE_READ | FMODE_WRITE; |
1da177e4 LT |
1342 | } |
1343 | if (shmflg & SHM_EXEC) { | |
1344 | prot |= PROT_EXEC; | |
1345 | acc_mode |= S_IXUGO; | |
1346 | } | |
1347 | ||
1348 | /* | |
1349 | * We cannot rely on the fs check since SYSV IPC does have an | |
1350 | * additional creator id... | |
1351 | */ | |
4e982311 | 1352 | ns = current->nsproxy->ipc_ns; |
c2c737a0 DB |
1353 | rcu_read_lock(); |
1354 | shp = shm_obtain_object_check(ns, shmid); | |
023a5355 ND |
1355 | if (IS_ERR(shp)) { |
1356 | err = PTR_ERR(shp); | |
c2c737a0 | 1357 | goto out_unlock; |
023a5355 | 1358 | } |
bc56bba8 EB |
1359 | |
1360 | err = -EACCES; | |
b0e77598 | 1361 | if (ipcperms(ns, &shp->shm_perm, acc_mode)) |
bc56bba8 | 1362 | goto out_unlock; |
1da177e4 LT |
1363 | |
1364 | err = security_shm_shmat(shp, shmaddr, shmflg); | |
bc56bba8 EB |
1365 | if (err) |
1366 | goto out_unlock; | |
1367 | ||
c2c737a0 | 1368 | ipc_lock_object(&shp->shm_perm); |
a399b29d GT |
1369 | |
1370 | /* check if shm_destroy() is tearing down shp */ | |
0f3d2b01 | 1371 | if (!ipc_valid_object(&shp->shm_perm)) { |
a399b29d GT |
1372 | ipc_unlock_object(&shp->shm_perm); |
1373 | err = -EIDRM; | |
1374 | goto out_unlock; | |
1375 | } | |
1376 | ||
2c48b9c4 AV |
1377 | path = shp->shm_file->f_path; |
1378 | path_get(&path); | |
1da177e4 | 1379 | shp->shm_nattch++; |
75c3cfa8 | 1380 | size = i_size_read(d_inode(path.dentry)); |
c2c737a0 DB |
1381 | ipc_unlock_object(&shp->shm_perm); |
1382 | rcu_read_unlock(); | |
1da177e4 | 1383 | |
bc56bba8 EB |
1384 | err = -ENOMEM; |
1385 | sfd = kzalloc(sizeof(*sfd), GFP_KERNEL); | |
f42569b1 DB |
1386 | if (!sfd) { |
1387 | path_put(&path); | |
1388 | goto out_nattch; | |
1389 | } | |
bc56bba8 | 1390 | |
2c48b9c4 AV |
1391 | file = alloc_file(&path, f_mode, |
1392 | is_file_hugepages(shp->shm_file) ? | |
c4caa778 AV |
1393 | &shm_file_operations_huge : |
1394 | &shm_file_operations); | |
39b65252 | 1395 | err = PTR_ERR(file); |
f42569b1 DB |
1396 | if (IS_ERR(file)) { |
1397 | kfree(sfd); | |
1398 | path_put(&path); | |
1399 | goto out_nattch; | |
1400 | } | |
bc56bba8 | 1401 | |
bc56bba8 | 1402 | file->private_data = sfd; |
bc56bba8 | 1403 | file->f_mapping = shp->shm_file->f_mapping; |
7ca7e564 | 1404 | sfd->id = shp->shm_perm.id; |
bc56bba8 | 1405 | sfd->ns = get_ipc_ns(ns); |
e6118fde EB |
1406 | /* |
1407 | * We need to take a reference to the real shm file to prevent the | |
1408 | * pointer from becoming stale in cases where the lifetime of the outer | |
1409 | * file extends beyond that of the shm segment. It's not usually | |
1410 | * possible, but it can happen during remap_file_pages() emulation as | |
1411 | * that unmaps the memory, then does ->mmap() via file reference only. | |
1412 | * We'll deny the ->mmap() if the shm segment was since removed, but to | |
1413 | * detect shm ID reuse we need to compare the file pointers. | |
1414 | */ | |
1415 | sfd->file = get_file(shp->shm_file); | |
bc56bba8 EB |
1416 | sfd->vm_ops = NULL; |
1417 | ||
8b3ec681 AV |
1418 | err = security_mmap_file(file, prot, flags); |
1419 | if (err) | |
1420 | goto out_fput; | |
1421 | ||
91f4f94e MH |
1422 | if (down_write_killable(¤t->mm->mmap_sem)) { |
1423 | err = -EINTR; | |
1424 | goto out_fput; | |
1425 | } | |
1426 | ||
1da177e4 | 1427 | if (addr && !(shmflg & SHM_REMAP)) { |
bc56bba8 | 1428 | err = -EINVAL; |
247a8ce8 MS |
1429 | if (addr + size < addr) |
1430 | goto invalid; | |
1431 | ||
1da177e4 LT |
1432 | if (find_vma_intersection(current->mm, addr, addr + size)) |
1433 | goto invalid; | |
1da177e4 | 1434 | } |
f42569b1 | 1435 | |
897ab3e0 | 1436 | addr = do_mmap_pgoff(file, addr, size, prot, flags, 0, &populate, NULL); |
bebeb3d6 | 1437 | *raddr = addr; |
bc56bba8 | 1438 | err = 0; |
bebeb3d6 ML |
1439 | if (IS_ERR_VALUE(addr)) |
1440 | err = (long)addr; | |
1da177e4 LT |
1441 | invalid: |
1442 | up_write(¤t->mm->mmap_sem); | |
bebeb3d6 | 1443 | if (populate) |
41badc15 | 1444 | mm_populate(addr, populate); |
1da177e4 | 1445 | |
8b3ec681 | 1446 | out_fput: |
bc56bba8 EB |
1447 | fput(file); |
1448 | ||
1449 | out_nattch: | |
d9a605e4 | 1450 | down_write(&shm_ids(ns).rwsem); |
00c2bf85 | 1451 | shp = shm_lock(ns, shmid); |
1da177e4 | 1452 | shp->shm_nattch--; |
b34a6b1d | 1453 | if (shm_may_destroy(ns, shp)) |
4e982311 | 1454 | shm_destroy(ns, shp); |
1da177e4 LT |
1455 | else |
1456 | shm_unlock(shp); | |
d9a605e4 | 1457 | up_write(&shm_ids(ns).rwsem); |
1da177e4 | 1458 | return err; |
bc56bba8 EB |
1459 | |
1460 | out_unlock: | |
c2c737a0 | 1461 | rcu_read_unlock(); |
f42569b1 DB |
1462 | out: |
1463 | return err; | |
1da177e4 LT |
1464 | } |
1465 | ||
d5460c99 | 1466 | SYSCALL_DEFINE3(shmat, int, shmid, char __user *, shmaddr, int, shmflg) |
7d87e14c SR |
1467 | { |
1468 | unsigned long ret; | |
1469 | long err; | |
1470 | ||
079a96ae | 1471 | err = do_shmat(shmid, shmaddr, shmflg, &ret, SHMLBA); |
7d87e14c SR |
1472 | if (err) |
1473 | return err; | |
1474 | force_successful_syscall_return(); | |
1475 | return (long)ret; | |
1476 | } | |
1477 | ||
a78ee9ed AV |
1478 | #ifdef CONFIG_COMPAT |
1479 | ||
1480 | #ifndef COMPAT_SHMLBA | |
1481 | #define COMPAT_SHMLBA SHMLBA | |
1482 | #endif | |
1483 | ||
1484 | COMPAT_SYSCALL_DEFINE3(shmat, int, shmid, compat_uptr_t, shmaddr, int, shmflg) | |
1485 | { | |
1486 | unsigned long ret; | |
1487 | long err; | |
1488 | ||
1489 | err = do_shmat(shmid, compat_ptr(shmaddr), shmflg, &ret, COMPAT_SHMLBA); | |
1490 | if (err) | |
1491 | return err; | |
1492 | force_successful_syscall_return(); | |
1493 | return (long)ret; | |
1494 | } | |
1495 | #endif | |
1496 | ||
1da177e4 LT |
1497 | /* |
1498 | * detach and kill segment if marked destroyed. | |
1499 | * The work is done in shm_close. | |
1500 | */ | |
d5460c99 | 1501 | SYSCALL_DEFINE1(shmdt, char __user *, shmaddr) |
1da177e4 LT |
1502 | { |
1503 | struct mm_struct *mm = current->mm; | |
586c7e6a | 1504 | struct vm_area_struct *vma; |
1da177e4 | 1505 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1506 | int retval = -EINVAL; |
586c7e6a MF |
1507 | #ifdef CONFIG_MMU |
1508 | loff_t size = 0; | |
d3c97900 | 1509 | struct file *file; |
586c7e6a MF |
1510 | struct vm_area_struct *next; |
1511 | #endif | |
1da177e4 | 1512 | |
df1e2fb5 HD |
1513 | if (addr & ~PAGE_MASK) |
1514 | return retval; | |
1515 | ||
91f4f94e MH |
1516 | if (down_write_killable(&mm->mmap_sem)) |
1517 | return -EINTR; | |
1da177e4 LT |
1518 | |
1519 | /* | |
1520 | * This function tries to be smart and unmap shm segments that | |
1521 | * were modified by partial mlock or munmap calls: | |
1522 | * - It first determines the size of the shm segment that should be | |
1523 | * unmapped: It searches for a vma that is backed by shm and that | |
1524 | * started at address shmaddr. It records it's size and then unmaps | |
1525 | * it. | |
1526 | * - Then it unmaps all shm vmas that started at shmaddr and that | |
d3c97900 DH |
1527 | * are within the initially determined size and that are from the |
1528 | * same shm segment from which we determined the size. | |
1da177e4 LT |
1529 | * Errors from do_munmap are ignored: the function only fails if |
1530 | * it's called with invalid parameters or if it's called to unmap | |
1531 | * a part of a vma. Both calls in this function are for full vmas, | |
1532 | * the parameters are directly copied from the vma itself and always | |
1533 | * valid - therefore do_munmap cannot fail. (famous last words?) | |
1534 | */ | |
1535 | /* | |
1536 | * If it had been mremap()'d, the starting address would not | |
1537 | * match the usual checks anyway. So assume all vma's are | |
1538 | * above the starting address given. | |
1539 | */ | |
1540 | vma = find_vma(mm, addr); | |
1541 | ||
8feae131 | 1542 | #ifdef CONFIG_MMU |
1da177e4 LT |
1543 | while (vma) { |
1544 | next = vma->vm_next; | |
1545 | ||
1546 | /* | |
1547 | * Check if the starting address would match, i.e. it's | |
1548 | * a fragment created by mprotect() and/or munmap(), or it | |
1549 | * otherwise it starts at this address with no hassles. | |
1550 | */ | |
bc56bba8 | 1551 | if ((vma->vm_ops == &shm_vm_ops) && |
1da177e4 LT |
1552 | (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) { |
1553 | ||
d3c97900 DH |
1554 | /* |
1555 | * Record the file of the shm segment being | |
1556 | * unmapped. With mremap(), someone could place | |
1557 | * page from another segment but with equal offsets | |
1558 | * in the range we are unmapping. | |
1559 | */ | |
1560 | file = vma->vm_file; | |
07a46ed2 | 1561 | size = i_size_read(file_inode(vma->vm_file)); |
897ab3e0 | 1562 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1563 | /* |
1564 | * We discovered the size of the shm segment, so | |
1565 | * break out of here and fall through to the next | |
1566 | * loop that uses the size information to stop | |
1567 | * searching for matching vma's. | |
1568 | */ | |
1569 | retval = 0; | |
1570 | vma = next; | |
1571 | break; | |
1572 | } | |
1573 | vma = next; | |
1574 | } | |
1575 | ||
1576 | /* | |
1577 | * We need look no further than the maximum address a fragment | |
1578 | * could possibly have landed at. Also cast things to loff_t to | |
25985edc | 1579 | * prevent overflows and make comparisons vs. equal-width types. |
1da177e4 | 1580 | */ |
8e36709d | 1581 | size = PAGE_ALIGN(size); |
1da177e4 LT |
1582 | while (vma && (loff_t)(vma->vm_end - addr) <= size) { |
1583 | next = vma->vm_next; | |
1584 | ||
1585 | /* finding a matching vma now does not alter retval */ | |
bc56bba8 | 1586 | if ((vma->vm_ops == &shm_vm_ops) && |
d3c97900 DH |
1587 | ((vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) && |
1588 | (vma->vm_file == file)) | |
897ab3e0 | 1589 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1590 | vma = next; |
1591 | } | |
1592 | ||
63980c80 | 1593 | #else /* CONFIG_MMU */ |
8feae131 | 1594 | /* under NOMMU conditions, the exact address to be destroyed must be |
63980c80 SP |
1595 | * given |
1596 | */ | |
530fcd16 | 1597 | if (vma && vma->vm_start == addr && vma->vm_ops == &shm_vm_ops) { |
897ab3e0 | 1598 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
8feae131 DH |
1599 | retval = 0; |
1600 | } | |
1601 | ||
1602 | #endif | |
1603 | ||
1da177e4 LT |
1604 | up_write(&mm->mmap_sem); |
1605 | return retval; | |
1606 | } | |
1607 | ||
1608 | #ifdef CONFIG_PROC_FS | |
19b4946c | 1609 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it) |
1da177e4 | 1610 | { |
1efdb69b | 1611 | struct user_namespace *user_ns = seq_user_ns(s); |
ade9f91b KC |
1612 | struct kern_ipc_perm *ipcp = it; |
1613 | struct shmid_kernel *shp; | |
b7952180 HD |
1614 | unsigned long rss = 0, swp = 0; |
1615 | ||
ade9f91b | 1616 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
b7952180 | 1617 | shm_add_rss_swap(shp, &rss, &swp); |
1da177e4 | 1618 | |
6c826818 PM |
1619 | #if BITS_PER_LONG <= 32 |
1620 | #define SIZE_SPEC "%10lu" | |
1621 | #else | |
1622 | #define SIZE_SPEC "%21lu" | |
1623 | #endif | |
1da177e4 | 1624 | |
7f032d6e JP |
1625 | seq_printf(s, |
1626 | "%10d %10d %4o " SIZE_SPEC " %5u %5u " | |
7ff2819e | 1627 | "%5lu %5u %5u %5u %5u %10llu %10llu %10llu " |
7f032d6e JP |
1628 | SIZE_SPEC " " SIZE_SPEC "\n", |
1629 | shp->shm_perm.key, | |
1630 | shp->shm_perm.id, | |
1631 | shp->shm_perm.mode, | |
1632 | shp->shm_segsz, | |
1633 | shp->shm_cprid, | |
1634 | shp->shm_lprid, | |
1635 | shp->shm_nattch, | |
1636 | from_kuid_munged(user_ns, shp->shm_perm.uid), | |
1637 | from_kgid_munged(user_ns, shp->shm_perm.gid), | |
1638 | from_kuid_munged(user_ns, shp->shm_perm.cuid), | |
1639 | from_kgid_munged(user_ns, shp->shm_perm.cgid), | |
1640 | shp->shm_atim, | |
1641 | shp->shm_dtim, | |
1642 | shp->shm_ctim, | |
1643 | rss * PAGE_SIZE, | |
1644 | swp * PAGE_SIZE); | |
1645 | ||
1646 | return 0; | |
1da177e4 LT |
1647 | } |
1648 | #endif |