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> |
0eb71a9d | 46 | #include <linux/rhashtable.h> |
7d87e14c | 47 | |
7153e402 | 48 | #include <linux/uaccess.h> |
1da177e4 LT |
49 | |
50 | #include "util.h" | |
51 | ||
a2e102cd EB |
52 | struct shmid_kernel /* private to the kernel */ |
53 | { | |
54 | struct kern_ipc_perm shm_perm; | |
55 | struct file *shm_file; | |
56 | unsigned long shm_nattch; | |
57 | unsigned long shm_segsz; | |
58 | time64_t shm_atim; | |
59 | time64_t shm_dtim; | |
60 | time64_t shm_ctim; | |
98f929b1 EB |
61 | struct pid *shm_cprid; |
62 | struct pid *shm_lprid; | |
a2e102cd EB |
63 | struct user_struct *mlock_user; |
64 | ||
65 | /* The task created the shm object. NULL if the task is dead. */ | |
66 | struct task_struct *shm_creator; | |
67 | struct list_head shm_clist; /* list by creator */ | |
68 | } __randomize_layout; | |
69 | ||
70 | /* shm_mode upper byte flags */ | |
71 | #define SHM_DEST 01000 /* segment will be destroyed on last detach */ | |
72 | #define SHM_LOCKED 02000 /* segment will not be swapped */ | |
73 | ||
bc56bba8 EB |
74 | struct shm_file_data { |
75 | int id; | |
76 | struct ipc_namespace *ns; | |
77 | struct file *file; | |
78 | const struct vm_operations_struct *vm_ops; | |
79 | }; | |
80 | ||
81 | #define shm_file_data(file) (*((struct shm_file_data **)&(file)->private_data)) | |
82 | ||
9a32144e | 83 | static const struct file_operations shm_file_operations; |
f0f37e2f | 84 | static const struct vm_operations_struct shm_vm_ops; |
1da177e4 | 85 | |
ed2ddbf8 | 86 | #define shm_ids(ns) ((ns)->ids[IPC_SHM_IDS]) |
1da177e4 | 87 | |
4e982311 KK |
88 | #define shm_unlock(shp) \ |
89 | ipc_unlock(&(shp)->shm_perm) | |
1da177e4 | 90 | |
7748dbfa | 91 | static int newseg(struct ipc_namespace *, struct ipc_params *); |
bc56bba8 EB |
92 | static void shm_open(struct vm_area_struct *vma); |
93 | static void shm_close(struct vm_area_struct *vma); | |
239521f3 | 94 | static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp); |
1da177e4 | 95 | #ifdef CONFIG_PROC_FS |
19b4946c | 96 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it); |
1da177e4 LT |
97 | #endif |
98 | ||
eae04d25 | 99 | void shm_init_ns(struct ipc_namespace *ns) |
4e982311 | 100 | { |
4e982311 KK |
101 | ns->shm_ctlmax = SHMMAX; |
102 | ns->shm_ctlall = SHMALL; | |
103 | ns->shm_ctlmni = SHMMNI; | |
b34a6b1d | 104 | ns->shm_rmid_forced = 0; |
4e982311 | 105 | ns->shm_tot = 0; |
eae04d25 | 106 | ipc_init_ids(&shm_ids(ns)); |
4e982311 KK |
107 | } |
108 | ||
f4566f04 | 109 | /* |
d9a605e4 DB |
110 | * Called with shm_ids.rwsem (writer) and the shp structure locked. |
111 | * Only shm_ids.rwsem remains locked on exit. | |
f4566f04 | 112 | */ |
01b8b07a | 113 | static void do_shm_rmid(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp) |
4e982311 | 114 | { |
01b8b07a | 115 | struct shmid_kernel *shp; |
63980c80 | 116 | |
01b8b07a PP |
117 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
118 | ||
239521f3 | 119 | if (shp->shm_nattch) { |
4e982311 KK |
120 | shp->shm_perm.mode |= SHM_DEST; |
121 | /* Do not find it any more */ | |
0cfb6aee | 122 | ipc_set_key_private(&shm_ids(ns), &shp->shm_perm); |
4e982311 KK |
123 | shm_unlock(shp); |
124 | } else | |
125 | shm_destroy(ns, shp); | |
126 | } | |
127 | ||
ae5e1b22 | 128 | #ifdef CONFIG_IPC_NS |
4e982311 KK |
129 | void shm_exit_ns(struct ipc_namespace *ns) |
130 | { | |
01b8b07a | 131 | free_ipcs(ns, &shm_ids(ns), do_shm_rmid); |
7d6feeb2 | 132 | idr_destroy(&ns->ids[IPC_SHM_IDS].ipcs_idr); |
0cfb6aee | 133 | rhashtable_destroy(&ns->ids[IPC_SHM_IDS].key_ht); |
4e982311 | 134 | } |
ae5e1b22 | 135 | #endif |
1da177e4 | 136 | |
140d0b21 | 137 | static int __init ipc_ns_init(void) |
1da177e4 | 138 | { |
eae04d25 DB |
139 | shm_init_ns(&init_ipc_ns); |
140 | return 0; | |
140d0b21 LT |
141 | } |
142 | ||
143 | pure_initcall(ipc_ns_init); | |
144 | ||
239521f3 | 145 | void __init shm_init(void) |
140d0b21 | 146 | { |
19b4946c | 147 | ipc_init_proc_interface("sysvipc/shm", |
b7952180 HD |
148 | #if BITS_PER_LONG <= 32 |
149 | " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime rss swap\n", | |
150 | #else | |
151 | " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime rss swap\n", | |
152 | #endif | |
4e982311 | 153 | IPC_SHM_IDS, sysvipc_shm_proc_show); |
1da177e4 LT |
154 | } |
155 | ||
8b8d52ac DB |
156 | static inline struct shmid_kernel *shm_obtain_object(struct ipc_namespace *ns, int id) |
157 | { | |
55b7ae50 | 158 | struct kern_ipc_perm *ipcp = ipc_obtain_object_idr(&shm_ids(ns), id); |
8b8d52ac DB |
159 | |
160 | if (IS_ERR(ipcp)) | |
161 | return ERR_CAST(ipcp); | |
162 | ||
163 | return container_of(ipcp, struct shmid_kernel, shm_perm); | |
164 | } | |
165 | ||
166 | static inline struct shmid_kernel *shm_obtain_object_check(struct ipc_namespace *ns, int id) | |
167 | { | |
168 | struct kern_ipc_perm *ipcp = ipc_obtain_object_check(&shm_ids(ns), id); | |
169 | ||
170 | if (IS_ERR(ipcp)) | |
171 | return ERR_CAST(ipcp); | |
172 | ||
173 | return container_of(ipcp, struct shmid_kernel, shm_perm); | |
174 | } | |
175 | ||
3e148c79 | 176 | /* |
d9a605e4 | 177 | * shm_lock_(check_) routines are called in the paths where the rwsem |
00c2bf85 | 178 | * is not necessarily held. |
3e148c79 | 179 | */ |
023a5355 | 180 | static inline struct shmid_kernel *shm_lock(struct ipc_namespace *ns, int id) |
1da177e4 | 181 | { |
82061c57 | 182 | struct kern_ipc_perm *ipcp; |
03f02c76 | 183 | |
82061c57 DB |
184 | rcu_read_lock(); |
185 | ipcp = ipc_obtain_object_idr(&shm_ids(ns), id); | |
186 | if (IS_ERR(ipcp)) | |
187 | goto err; | |
188 | ||
189 | ipc_lock_object(ipcp); | |
190 | /* | |
191 | * ipc_rmid() may have already freed the ID while ipc_lock_object() | |
192 | * was spinning: here verify that the structure is still valid. | |
193 | * Upon races with RMID, return -EIDRM, thus indicating that | |
194 | * the ID points to a removed identifier. | |
195 | */ | |
196 | if (ipc_valid_object(ipcp)) { | |
197 | /* return a locked ipc object upon success */ | |
198 | return container_of(ipcp, struct shmid_kernel, shm_perm); | |
199 | } | |
200 | ||
201 | ipc_unlock_object(ipcp); | |
9c21dae2 | 202 | ipcp = ERR_PTR(-EIDRM); |
82061c57 DB |
203 | err: |
204 | rcu_read_unlock(); | |
c5c8975b | 205 | /* |
1ac0b6de | 206 | * Callers of shm_lock() must validate the status of the returned ipc |
82061c57 | 207 | * object pointer and error out as appropriate. |
c5c8975b | 208 | */ |
59cf0a93 | 209 | return ERR_CAST(ipcp); |
023a5355 ND |
210 | } |
211 | ||
4c677e2e VK |
212 | static inline void shm_lock_by_ptr(struct shmid_kernel *ipcp) |
213 | { | |
214 | rcu_read_lock(); | |
cf9d5d78 | 215 | ipc_lock_object(&ipcp->shm_perm); |
4c677e2e VK |
216 | } |
217 | ||
53dad6d3 DB |
218 | static void shm_rcu_free(struct rcu_head *head) |
219 | { | |
dba4cdd3 MS |
220 | struct kern_ipc_perm *ptr = container_of(head, struct kern_ipc_perm, |
221 | rcu); | |
222 | struct shmid_kernel *shp = container_of(ptr, struct shmid_kernel, | |
223 | shm_perm); | |
7191adff | 224 | security_shm_free(&shp->shm_perm); |
42e618f7 | 225 | kvfree(shp); |
53dad6d3 DB |
226 | } |
227 | ||
7ca7e564 | 228 | static inline void shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *s) |
1da177e4 | 229 | { |
ab602f79 | 230 | list_del(&s->shm_clist); |
7ca7e564 | 231 | ipc_rmid(&shm_ids(ns), &s->shm_perm); |
1da177e4 LT |
232 | } |
233 | ||
1da177e4 | 234 | |
1ac0b6de | 235 | static int __shm_open(struct vm_area_struct *vma) |
4e982311 | 236 | { |
bc56bba8 EB |
237 | struct file *file = vma->vm_file; |
238 | struct shm_file_data *sfd = shm_file_data(file); | |
1da177e4 LT |
239 | struct shmid_kernel *shp; |
240 | ||
bc56bba8 | 241 | shp = shm_lock(sfd->ns, sfd->id); |
1ac0b6de KS |
242 | |
243 | if (IS_ERR(shp)) | |
244 | return PTR_ERR(shp); | |
245 | ||
3f05317d EB |
246 | if (shp->shm_file != sfd->file) { |
247 | /* ID was reused */ | |
248 | shm_unlock(shp); | |
249 | return -EINVAL; | |
250 | } | |
251 | ||
7ff2819e | 252 | shp->shm_atim = ktime_get_real_seconds(); |
98f929b1 | 253 | ipc_update_pid(&shp->shm_lprid, task_tgid(current)); |
1da177e4 LT |
254 | shp->shm_nattch++; |
255 | shm_unlock(shp); | |
1ac0b6de KS |
256 | return 0; |
257 | } | |
258 | ||
259 | /* This is called by fork, once for every shm attach. */ | |
260 | static void shm_open(struct vm_area_struct *vma) | |
261 | { | |
262 | int err = __shm_open(vma); | |
263 | /* | |
264 | * We raced in the idr lookup or with shm_destroy(). | |
265 | * Either way, the ID is busted. | |
266 | */ | |
267 | WARN_ON_ONCE(err); | |
1da177e4 LT |
268 | } |
269 | ||
1da177e4 LT |
270 | /* |
271 | * shm_destroy - free the struct shmid_kernel | |
272 | * | |
f4566f04 | 273 | * @ns: namespace |
1da177e4 LT |
274 | * @shp: struct to free |
275 | * | |
d9a605e4 | 276 | * It has to be called with shp and shm_ids.rwsem (writer) locked, |
1da177e4 LT |
277 | * but returns with shp unlocked and freed. |
278 | */ | |
4e982311 | 279 | static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) |
1da177e4 | 280 | { |
a399b29d GT |
281 | struct file *shm_file; |
282 | ||
283 | shm_file = shp->shm_file; | |
284 | shp->shm_file = NULL; | |
4e982311 | 285 | ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT; |
7ca7e564 | 286 | shm_rmid(ns, shp); |
1da177e4 | 287 | shm_unlock(shp); |
a399b29d GT |
288 | if (!is_file_hugepages(shm_file)) |
289 | shmem_lock(shm_file, 0, shp->mlock_user); | |
353d5c30 | 290 | else if (shp->mlock_user) |
07a46ed2 DH |
291 | user_shm_unlock(i_size_read(file_inode(shm_file)), |
292 | shp->mlock_user); | |
a399b29d | 293 | fput(shm_file); |
98f929b1 EB |
294 | ipc_update_pid(&shp->shm_cprid, NULL); |
295 | ipc_update_pid(&shp->shm_lprid, NULL); | |
dba4cdd3 | 296 | ipc_rcu_putref(&shp->shm_perm, shm_rcu_free); |
1da177e4 LT |
297 | } |
298 | ||
b34a6b1d VK |
299 | /* |
300 | * shm_may_destroy - identifies whether shm segment should be destroyed now | |
301 | * | |
302 | * Returns true if and only if there are no active users of the segment and | |
303 | * one of the following is true: | |
304 | * | |
305 | * 1) shmctl(id, IPC_RMID, NULL) was called for this shp | |
306 | * | |
307 | * 2) sysctl kernel.shm_rmid_forced is set to 1. | |
308 | */ | |
309 | static bool shm_may_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) | |
310 | { | |
311 | return (shp->shm_nattch == 0) && | |
312 | (ns->shm_rmid_forced || | |
313 | (shp->shm_perm.mode & SHM_DEST)); | |
314 | } | |
315 | ||
1da177e4 | 316 | /* |
bc56bba8 | 317 | * remove the attach descriptor vma. |
1da177e4 LT |
318 | * free memory for segment if it is marked destroyed. |
319 | * The descriptor has already been removed from the current->mm->mmap list | |
320 | * and will later be kfree()d. | |
321 | */ | |
bc56bba8 | 322 | static void shm_close(struct vm_area_struct *vma) |
1da177e4 | 323 | { |
239521f3 | 324 | struct file *file = vma->vm_file; |
bc56bba8 | 325 | struct shm_file_data *sfd = shm_file_data(file); |
1da177e4 | 326 | struct shmid_kernel *shp; |
bc56bba8 | 327 | struct ipc_namespace *ns = sfd->ns; |
4e982311 | 328 | |
d9a605e4 | 329 | down_write(&shm_ids(ns).rwsem); |
1da177e4 | 330 | /* remove from the list of attaches of the shm segment */ |
00c2bf85 | 331 | shp = shm_lock(ns, sfd->id); |
1ac0b6de KS |
332 | |
333 | /* | |
334 | * We raced in the idr lookup or with shm_destroy(). | |
335 | * Either way, the ID is busted. | |
336 | */ | |
337 | if (WARN_ON_ONCE(IS_ERR(shp))) | |
338 | goto done; /* no-op */ | |
339 | ||
98f929b1 | 340 | ipc_update_pid(&shp->shm_lprid, task_tgid(current)); |
7ff2819e | 341 | shp->shm_dtim = ktime_get_real_seconds(); |
1da177e4 | 342 | shp->shm_nattch--; |
b34a6b1d VK |
343 | if (shm_may_destroy(ns, shp)) |
344 | shm_destroy(ns, shp); | |
345 | else | |
346 | shm_unlock(shp); | |
1ac0b6de | 347 | done: |
d9a605e4 | 348 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
349 | } |
350 | ||
d9a605e4 | 351 | /* Called with ns->shm_ids(ns).rwsem locked */ |
b34a6b1d VK |
352 | static int shm_try_destroy_orphaned(int id, void *p, void *data) |
353 | { | |
354 | struct ipc_namespace *ns = data; | |
4c677e2e VK |
355 | struct kern_ipc_perm *ipcp = p; |
356 | struct shmid_kernel *shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
b34a6b1d VK |
357 | |
358 | /* | |
359 | * We want to destroy segments without users and with already | |
360 | * exit'ed originating process. | |
4c677e2e | 361 | * |
d9a605e4 | 362 | * As shp->* are changed under rwsem, it's safe to skip shp locking. |
b34a6b1d | 363 | */ |
4c677e2e | 364 | if (shp->shm_creator != NULL) |
b34a6b1d | 365 | return 0; |
b34a6b1d | 366 | |
4c677e2e VK |
367 | if (shm_may_destroy(ns, shp)) { |
368 | shm_lock_by_ptr(shp); | |
4e982311 | 369 | shm_destroy(ns, shp); |
4c677e2e | 370 | } |
b34a6b1d VK |
371 | return 0; |
372 | } | |
373 | ||
374 | void shm_destroy_orphaned(struct ipc_namespace *ns) | |
375 | { | |
d9a605e4 | 376 | down_write(&shm_ids(ns).rwsem); |
33a30ed4 | 377 | if (shm_ids(ns).in_use) |
4c677e2e | 378 | idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_orphaned, ns); |
d9a605e4 | 379 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
380 | } |
381 | ||
83293c0f | 382 | /* Locking assumes this will only be called with task == current */ |
b34a6b1d VK |
383 | void exit_shm(struct task_struct *task) |
384 | { | |
4c677e2e | 385 | struct ipc_namespace *ns = task->nsproxy->ipc_ns; |
ab602f79 | 386 | struct shmid_kernel *shp, *n; |
b34a6b1d | 387 | |
83293c0f JM |
388 | if (list_empty(&task->sysvshm.shm_clist)) |
389 | return; | |
390 | ||
391 | /* | |
392 | * If kernel.shm_rmid_forced is not set then only keep track of | |
393 | * which shmids are orphaned, so that a later set of the sysctl | |
394 | * can clean them up. | |
395 | */ | |
396 | if (!ns->shm_rmid_forced) { | |
397 | down_read(&shm_ids(ns).rwsem); | |
398 | list_for_each_entry(shp, &task->sysvshm.shm_clist, shm_clist) | |
399 | shp->shm_creator = NULL; | |
400 | /* | |
401 | * Only under read lock but we are only called on current | |
402 | * so no entry on the list will be shared. | |
403 | */ | |
404 | list_del(&task->sysvshm.shm_clist); | |
405 | up_read(&shm_ids(ns).rwsem); | |
298507d4 | 406 | return; |
83293c0f | 407 | } |
298507d4 | 408 | |
83293c0f JM |
409 | /* |
410 | * Destroy all already created segments, that were not yet mapped, | |
411 | * and mark any mapped as orphan to cover the sysctl toggling. | |
412 | * Destroy is skipped if shm_may_destroy() returns false. | |
413 | */ | |
d9a605e4 | 414 | down_write(&shm_ids(ns).rwsem); |
83293c0f JM |
415 | list_for_each_entry_safe(shp, n, &task->sysvshm.shm_clist, shm_clist) { |
416 | shp->shm_creator = NULL; | |
417 | ||
418 | if (shm_may_destroy(ns, shp)) { | |
419 | shm_lock_by_ptr(shp); | |
420 | shm_destroy(ns, shp); | |
421 | } | |
422 | } | |
423 | ||
424 | /* Remove the list head from any segments still attached. */ | |
ab602f79 | 425 | list_del(&task->sysvshm.shm_clist); |
d9a605e4 | 426 | up_write(&shm_ids(ns).rwsem); |
1da177e4 LT |
427 | } |
428 | ||
14f28f57 | 429 | static vm_fault_t shm_fault(struct vm_fault *vmf) |
bc56bba8 | 430 | { |
11bac800 | 431 | struct file *file = vmf->vma->vm_file; |
bc56bba8 EB |
432 | struct shm_file_data *sfd = shm_file_data(file); |
433 | ||
11bac800 | 434 | return sfd->vm_ops->fault(vmf); |
bc56bba8 EB |
435 | } |
436 | ||
dd3b614f | 437 | static int shm_may_split(struct vm_area_struct *vma, unsigned long addr) |
3d942ee0 MK |
438 | { |
439 | struct file *file = vma->vm_file; | |
440 | struct shm_file_data *sfd = shm_file_data(file); | |
441 | ||
dd3b614f DS |
442 | if (sfd->vm_ops->may_split) |
443 | return sfd->vm_ops->may_split(vma, addr); | |
3d942ee0 MK |
444 | |
445 | return 0; | |
446 | } | |
447 | ||
eec3636a JC |
448 | static unsigned long shm_pagesize(struct vm_area_struct *vma) |
449 | { | |
450 | struct file *file = vma->vm_file; | |
451 | struct shm_file_data *sfd = shm_file_data(file); | |
452 | ||
453 | if (sfd->vm_ops->pagesize) | |
454 | return sfd->vm_ops->pagesize(vma); | |
455 | ||
456 | return PAGE_SIZE; | |
457 | } | |
458 | ||
bc56bba8 | 459 | #ifdef CONFIG_NUMA |
d823e3e7 | 460 | static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new) |
bc56bba8 EB |
461 | { |
462 | struct file *file = vma->vm_file; | |
463 | struct shm_file_data *sfd = shm_file_data(file); | |
464 | int err = 0; | |
63980c80 | 465 | |
bc56bba8 EB |
466 | if (sfd->vm_ops->set_policy) |
467 | err = sfd->vm_ops->set_policy(vma, new); | |
468 | return err; | |
469 | } | |
470 | ||
d823e3e7 AB |
471 | static struct mempolicy *shm_get_policy(struct vm_area_struct *vma, |
472 | unsigned long addr) | |
bc56bba8 EB |
473 | { |
474 | struct file *file = vma->vm_file; | |
475 | struct shm_file_data *sfd = shm_file_data(file); | |
476 | struct mempolicy *pol = NULL; | |
477 | ||
478 | if (sfd->vm_ops->get_policy) | |
479 | pol = sfd->vm_ops->get_policy(vma, addr); | |
52cd3b07 | 480 | else if (vma->vm_policy) |
bc56bba8 | 481 | pol = vma->vm_policy; |
52cd3b07 | 482 | |
bc56bba8 EB |
483 | return pol; |
484 | } | |
485 | #endif | |
486 | ||
239521f3 | 487 | static int shm_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 488 | { |
bc56bba8 | 489 | struct shm_file_data *sfd = shm_file_data(file); |
b0e15190 DH |
490 | int ret; |
491 | ||
1ac0b6de | 492 | /* |
3f05317d EB |
493 | * In case of remap_file_pages() emulation, the file can represent an |
494 | * IPC ID that was removed, and possibly even reused by another shm | |
495 | * segment already. Propagate this case as an error to caller. | |
1ac0b6de | 496 | */ |
63980c80 | 497 | ret = __shm_open(vma); |
1ac0b6de KS |
498 | if (ret) |
499 | return ret; | |
500 | ||
f74ac015 | 501 | ret = call_mmap(sfd->file, vma); |
1ac0b6de KS |
502 | if (ret) { |
503 | shm_close(vma); | |
bc56bba8 | 504 | return ret; |
1ac0b6de | 505 | } |
bc56bba8 | 506 | sfd->vm_ops = vma->vm_ops; |
2e92a3ba | 507 | #ifdef CONFIG_MMU |
d0edd852 | 508 | WARN_ON(!sfd->vm_ops->fault); |
2e92a3ba | 509 | #endif |
bc56bba8 | 510 | vma->vm_ops = &shm_vm_ops; |
1ac0b6de | 511 | return 0; |
1da177e4 LT |
512 | } |
513 | ||
4e982311 KK |
514 | static int shm_release(struct inode *ino, struct file *file) |
515 | { | |
bc56bba8 | 516 | struct shm_file_data *sfd = shm_file_data(file); |
4e982311 | 517 | |
bc56bba8 | 518 | put_ipc_ns(sfd->ns); |
3f05317d | 519 | fput(sfd->file); |
bc56bba8 EB |
520 | shm_file_data(file) = NULL; |
521 | kfree(sfd); | |
4e982311 KK |
522 | return 0; |
523 | } | |
524 | ||
02c24a82 | 525 | static int shm_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
516dffdc | 526 | { |
516dffdc | 527 | struct shm_file_data *sfd = shm_file_data(file); |
516dffdc | 528 | |
7ea80859 CH |
529 | if (!sfd->file->f_op->fsync) |
530 | return -EINVAL; | |
0f41074a | 531 | return sfd->file->f_op->fsync(sfd->file, start, end, datasync); |
516dffdc AL |
532 | } |
533 | ||
7d8a4569 WD |
534 | static long shm_fallocate(struct file *file, int mode, loff_t offset, |
535 | loff_t len) | |
536 | { | |
537 | struct shm_file_data *sfd = shm_file_data(file); | |
538 | ||
539 | if (!sfd->file->f_op->fallocate) | |
540 | return -EOPNOTSUPP; | |
541 | return sfd->file->f_op->fallocate(file, mode, offset, len); | |
542 | } | |
543 | ||
bc56bba8 EB |
544 | static unsigned long shm_get_unmapped_area(struct file *file, |
545 | unsigned long addr, unsigned long len, unsigned long pgoff, | |
546 | unsigned long flags) | |
547 | { | |
548 | struct shm_file_data *sfd = shm_file_data(file); | |
63980c80 | 549 | |
c4caa778 AV |
550 | return sfd->file->f_op->get_unmapped_area(sfd->file, addr, len, |
551 | pgoff, flags); | |
bc56bba8 | 552 | } |
bc56bba8 | 553 | |
9a32144e | 554 | static const struct file_operations shm_file_operations = { |
4e982311 | 555 | .mmap = shm_mmap, |
516dffdc | 556 | .fsync = shm_fsync, |
4e982311 | 557 | .release = shm_release, |
ed5e5894 | 558 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 559 | .llseek = noop_llseek, |
7d8a4569 | 560 | .fallocate = shm_fallocate, |
c4caa778 AV |
561 | }; |
562 | ||
c01d5b30 HD |
563 | /* |
564 | * shm_file_operations_huge is now identical to shm_file_operations, | |
565 | * but we keep it distinct for the sake of is_file_shm_hugepages(). | |
566 | */ | |
c4caa778 AV |
567 | static const struct file_operations shm_file_operations_huge = { |
568 | .mmap = shm_mmap, | |
569 | .fsync = shm_fsync, | |
570 | .release = shm_release, | |
bc56bba8 | 571 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 572 | .llseek = noop_llseek, |
7d8a4569 | 573 | .fallocate = shm_fallocate, |
1da177e4 LT |
574 | }; |
575 | ||
2954e440 | 576 | bool is_file_shm_hugepages(struct file *file) |
c4caa778 AV |
577 | { |
578 | return file->f_op == &shm_file_operations_huge; | |
579 | } | |
580 | ||
f0f37e2f | 581 | static const struct vm_operations_struct shm_vm_ops = { |
1da177e4 LT |
582 | .open = shm_open, /* callback for a new vm-area open */ |
583 | .close = shm_close, /* callback for when the vm-area is released */ | |
54cb8821 | 584 | .fault = shm_fault, |
dd3b614f | 585 | .may_split = shm_may_split, |
eec3636a | 586 | .pagesize = shm_pagesize, |
bc56bba8 EB |
587 | #if defined(CONFIG_NUMA) |
588 | .set_policy = shm_set_policy, | |
589 | .get_policy = shm_get_policy, | |
1da177e4 LT |
590 | #endif |
591 | }; | |
592 | ||
f4566f04 ND |
593 | /** |
594 | * newseg - Create a new shared memory segment | |
595 | * @ns: namespace | |
596 | * @params: ptr to the structure that contains key, size and shmflg | |
597 | * | |
d9a605e4 | 598 | * Called with shm_ids.rwsem held as a writer. |
f4566f04 | 599 | */ |
7748dbfa | 600 | static int newseg(struct ipc_namespace *ns, struct ipc_params *params) |
1da177e4 | 601 | { |
7748dbfa ND |
602 | key_t key = params->key; |
603 | int shmflg = params->flg; | |
604 | size_t size = params->u.size; | |
1da177e4 LT |
605 | int error; |
606 | struct shmid_kernel *shp; | |
d69f3bad | 607 | size_t numpages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; |
239521f3 | 608 | struct file *file; |
1da177e4 | 609 | char name[13]; |
ca16d140 | 610 | vm_flags_t acctflag = 0; |
1da177e4 | 611 | |
4e982311 | 612 | if (size < SHMMIN || size > ns->shm_ctlmax) |
1da177e4 LT |
613 | return -EINVAL; |
614 | ||
1376327c MS |
615 | if (numpages << PAGE_SHIFT < size) |
616 | return -ENOSPC; | |
617 | ||
09c6eb1f MS |
618 | if (ns->shm_tot + numpages < ns->shm_tot || |
619 | ns->shm_tot + numpages > ns->shm_ctlall) | |
1da177e4 LT |
620 | return -ENOSPC; |
621 | ||
42e618f7 KC |
622 | shp = kvmalloc(sizeof(*shp), GFP_KERNEL); |
623 | if (unlikely(!shp)) | |
1da177e4 LT |
624 | return -ENOMEM; |
625 | ||
626 | shp->shm_perm.key = key; | |
b33291c0 | 627 | shp->shm_perm.mode = (shmflg & S_IRWXUGO); |
1da177e4 LT |
628 | shp->mlock_user = NULL; |
629 | ||
630 | shp->shm_perm.security = NULL; | |
7191adff | 631 | error = security_shm_alloc(&shp->shm_perm); |
1da177e4 | 632 | if (error) { |
42e618f7 | 633 | kvfree(shp); |
1da177e4 LT |
634 | return error; |
635 | } | |
636 | ||
239521f3 | 637 | sprintf(name, "SYSV%08x", key); |
1da177e4 | 638 | if (shmflg & SHM_HUGETLB) { |
c103a4dc | 639 | struct hstate *hs; |
091d0d55 LZ |
640 | size_t hugesize; |
641 | ||
c103a4dc | 642 | hs = hstate_sizelog((shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); |
091d0d55 LZ |
643 | if (!hs) { |
644 | error = -EINVAL; | |
645 | goto no_file; | |
646 | } | |
647 | hugesize = ALIGN(size, huge_page_size(hs)); | |
af73e4d9 | 648 | |
5a6fe125 MG |
649 | /* hugetlb_file_setup applies strict accounting */ |
650 | if (shmflg & SHM_NORESERVE) | |
651 | acctflag = VM_NORESERVE; | |
af73e4d9 | 652 | file = hugetlb_file_setup(name, hugesize, acctflag, |
42d7395f AK |
653 | &shp->mlock_user, HUGETLB_SHMFS_INODE, |
654 | (shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); | |
1da177e4 | 655 | } else { |
bf8f972d BP |
656 | /* |
657 | * Do not allow no accounting for OVERCOMMIT_NEVER, even | |
239521f3 | 658 | * if it's asked for. |
bf8f972d BP |
659 | */ |
660 | if ((shmflg & SHM_NORESERVE) && | |
661 | sysctl_overcommit_memory != OVERCOMMIT_NEVER) | |
fc8744ad | 662 | acctflag = VM_NORESERVE; |
e1832f29 | 663 | file = shmem_kernel_file_setup(name, size, acctflag); |
1da177e4 LT |
664 | } |
665 | error = PTR_ERR(file); | |
666 | if (IS_ERR(file)) | |
667 | goto no_file; | |
668 | ||
98f929b1 EB |
669 | shp->shm_cprid = get_pid(task_tgid(current)); |
670 | shp->shm_lprid = NULL; | |
1da177e4 | 671 | shp->shm_atim = shp->shm_dtim = 0; |
7ff2819e | 672 | shp->shm_ctim = ktime_get_real_seconds(); |
1da177e4 LT |
673 | shp->shm_segsz = size; |
674 | shp->shm_nattch = 0; | |
1da177e4 | 675 | shp->shm_file = file; |
5774ed01 | 676 | shp->shm_creator = current; |
b9a53227 | 677 | |
39c96a1b | 678 | /* ipc_addid() locks shp upon success. */ |
a2642f87 MS |
679 | error = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni); |
680 | if (error < 0) | |
b9a53227 | 681 | goto no_id; |
b9a53227 | 682 | |
ab602f79 | 683 | list_add(&shp->shm_clist, ¤t->sysvshm.shm_clist); |
dbfcd91f | 684 | |
30475cc1 BP |
685 | /* |
686 | * shmid gets reported as "inode#" in /proc/pid/maps. | |
687 | * proc-ps tools use this. Changing this will break them. | |
688 | */ | |
496ad9aa | 689 | file_inode(file)->i_ino = shp->shm_perm.id; |
551110a9 | 690 | |
4e982311 | 691 | ns->shm_tot += numpages; |
7ca7e564 | 692 | error = shp->shm_perm.id; |
dbfcd91f | 693 | |
cf9d5d78 | 694 | ipc_unlock_object(&shp->shm_perm); |
dbfcd91f | 695 | rcu_read_unlock(); |
7ca7e564 | 696 | return error; |
1da177e4 LT |
697 | |
698 | no_id: | |
2236d4d3 EB |
699 | ipc_update_pid(&shp->shm_cprid, NULL); |
700 | ipc_update_pid(&shp->shm_lprid, NULL); | |
2195d281 | 701 | if (is_file_hugepages(file) && shp->mlock_user) |
353d5c30 | 702 | user_shm_unlock(size, shp->mlock_user); |
1da177e4 | 703 | fput(file); |
39cfffd7 MS |
704 | ipc_rcu_putref(&shp->shm_perm, shm_rcu_free); |
705 | return error; | |
1da177e4 | 706 | no_file: |
a2642f87 | 707 | call_rcu(&shp->shm_perm.rcu, shm_rcu_free); |
1da177e4 LT |
708 | return error; |
709 | } | |
710 | ||
f4566f04 | 711 | /* |
d9a605e4 | 712 | * Called with shm_ids.rwsem and ipcp locked. |
f4566f04 | 713 | */ |
00898e85 | 714 | static int shm_more_checks(struct kern_ipc_perm *ipcp, struct ipc_params *params) |
7748dbfa | 715 | { |
03f02c76 ND |
716 | struct shmid_kernel *shp; |
717 | ||
718 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
719 | if (shp->shm_segsz < params->u.size) | |
7748dbfa ND |
720 | return -EINVAL; |
721 | ||
722 | return 0; | |
723 | } | |
724 | ||
65749e0b | 725 | long ksys_shmget(key_t key, size_t size, int shmflg) |
1da177e4 | 726 | { |
4e982311 | 727 | struct ipc_namespace *ns; |
eb66ec44 MK |
728 | static const struct ipc_ops shm_ops = { |
729 | .getnew = newseg, | |
50ab44b1 | 730 | .associate = security_shm_associate, |
eb66ec44 MK |
731 | .more_checks = shm_more_checks, |
732 | }; | |
7748dbfa | 733 | struct ipc_params shm_params; |
4e982311 KK |
734 | |
735 | ns = current->nsproxy->ipc_ns; | |
1da177e4 | 736 | |
7748dbfa ND |
737 | shm_params.key = key; |
738 | shm_params.flg = shmflg; | |
739 | shm_params.u.size = size; | |
1da177e4 | 740 | |
7748dbfa | 741 | return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params); |
1da177e4 LT |
742 | } |
743 | ||
65749e0b DB |
744 | SYSCALL_DEFINE3(shmget, key_t, key, size_t, size, int, shmflg) |
745 | { | |
746 | return ksys_shmget(key, size, shmflg); | |
747 | } | |
748 | ||
1da177e4 LT |
749 | static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version) |
750 | { | |
239521f3 | 751 | switch (version) { |
1da177e4 LT |
752 | case IPC_64: |
753 | return copy_to_user(buf, in, sizeof(*in)); | |
754 | case IPC_OLD: | |
755 | { | |
756 | struct shmid_ds out; | |
757 | ||
3af54c9b | 758 | memset(&out, 0, sizeof(out)); |
1da177e4 LT |
759 | ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm); |
760 | out.shm_segsz = in->shm_segsz; | |
761 | out.shm_atime = in->shm_atime; | |
762 | out.shm_dtime = in->shm_dtime; | |
763 | out.shm_ctime = in->shm_ctime; | |
764 | out.shm_cpid = in->shm_cpid; | |
765 | out.shm_lpid = in->shm_lpid; | |
766 | out.shm_nattch = in->shm_nattch; | |
767 | ||
768 | return copy_to_user(buf, &out, sizeof(out)); | |
769 | } | |
770 | default: | |
771 | return -EINVAL; | |
772 | } | |
773 | } | |
774 | ||
016d7132 PP |
775 | static inline unsigned long |
776 | copy_shmid_from_user(struct shmid64_ds *out, void __user *buf, int version) | |
1da177e4 | 777 | { |
239521f3 | 778 | switch (version) { |
1da177e4 | 779 | case IPC_64: |
016d7132 | 780 | if (copy_from_user(out, buf, sizeof(*out))) |
1da177e4 | 781 | return -EFAULT; |
1da177e4 | 782 | return 0; |
1da177e4 LT |
783 | case IPC_OLD: |
784 | { | |
785 | struct shmid_ds tbuf_old; | |
786 | ||
787 | if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) | |
788 | return -EFAULT; | |
789 | ||
016d7132 PP |
790 | out->shm_perm.uid = tbuf_old.shm_perm.uid; |
791 | out->shm_perm.gid = tbuf_old.shm_perm.gid; | |
792 | out->shm_perm.mode = tbuf_old.shm_perm.mode; | |
1da177e4 LT |
793 | |
794 | return 0; | |
795 | } | |
796 | default: | |
797 | return -EINVAL; | |
798 | } | |
799 | } | |
800 | ||
801 | static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version) | |
802 | { | |
239521f3 | 803 | switch (version) { |
1da177e4 LT |
804 | case IPC_64: |
805 | return copy_to_user(buf, in, sizeof(*in)); | |
806 | case IPC_OLD: | |
807 | { | |
808 | struct shminfo out; | |
809 | ||
239521f3 | 810 | if (in->shmmax > INT_MAX) |
1da177e4 LT |
811 | out.shmmax = INT_MAX; |
812 | else | |
813 | out.shmmax = (int)in->shmmax; | |
814 | ||
815 | out.shmmin = in->shmmin; | |
816 | out.shmmni = in->shmmni; | |
817 | out.shmseg = in->shmseg; | |
46c0a8ca | 818 | out.shmall = in->shmall; |
1da177e4 LT |
819 | |
820 | return copy_to_user(buf, &out, sizeof(out)); | |
821 | } | |
822 | default: | |
823 | return -EINVAL; | |
824 | } | |
825 | } | |
826 | ||
b7952180 HD |
827 | /* |
828 | * Calculate and add used RSS and swap pages of a shm. | |
d9a605e4 | 829 | * Called with shm_ids.rwsem held as a reader |
b7952180 HD |
830 | */ |
831 | static void shm_add_rss_swap(struct shmid_kernel *shp, | |
832 | unsigned long *rss_add, unsigned long *swp_add) | |
833 | { | |
834 | struct inode *inode; | |
835 | ||
496ad9aa | 836 | inode = file_inode(shp->shm_file); |
b7952180 HD |
837 | |
838 | if (is_file_hugepages(shp->shm_file)) { | |
839 | struct address_space *mapping = inode->i_mapping; | |
840 | struct hstate *h = hstate_file(shp->shm_file); | |
841 | *rss_add += pages_per_huge_page(h) * mapping->nrpages; | |
842 | } else { | |
843 | #ifdef CONFIG_SHMEM | |
844 | struct shmem_inode_info *info = SHMEM_I(inode); | |
63980c80 | 845 | |
4595ef88 | 846 | spin_lock_irq(&info->lock); |
b7952180 HD |
847 | *rss_add += inode->i_mapping->nrpages; |
848 | *swp_add += info->swapped; | |
4595ef88 | 849 | spin_unlock_irq(&info->lock); |
b7952180 HD |
850 | #else |
851 | *rss_add += inode->i_mapping->nrpages; | |
852 | #endif | |
853 | } | |
854 | } | |
855 | ||
f4566f04 | 856 | /* |
d9a605e4 | 857 | * Called with shm_ids.rwsem held as a reader |
f4566f04 | 858 | */ |
4e982311 KK |
859 | static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss, |
860 | unsigned long *swp) | |
1da177e4 | 861 | { |
7ca7e564 ND |
862 | int next_id; |
863 | int total, in_use; | |
1da177e4 LT |
864 | |
865 | *rss = 0; | |
866 | *swp = 0; | |
867 | ||
7ca7e564 ND |
868 | in_use = shm_ids(ns).in_use; |
869 | ||
870 | for (total = 0, next_id = 0; total < in_use; next_id++) { | |
e562aebc | 871 | struct kern_ipc_perm *ipc; |
1da177e4 | 872 | struct shmid_kernel *shp; |
1da177e4 | 873 | |
e562aebc TB |
874 | ipc = idr_find(&shm_ids(ns).ipcs_idr, next_id); |
875 | if (ipc == NULL) | |
1da177e4 | 876 | continue; |
e562aebc | 877 | shp = container_of(ipc, struct shmid_kernel, shm_perm); |
1da177e4 | 878 | |
b7952180 | 879 | shm_add_rss_swap(shp, rss, swp); |
7ca7e564 ND |
880 | |
881 | total++; | |
1da177e4 LT |
882 | } |
883 | } | |
884 | ||
8d4cc8b5 | 885 | /* |
d9a605e4 | 886 | * This function handles some shmctl commands which require the rwsem |
8d4cc8b5 | 887 | * to be held in write mode. |
d9a605e4 | 888 | * NOTE: no locks must be held, the rwsem is taken inside this function. |
8d4cc8b5 PP |
889 | */ |
890 | static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd, | |
9ba720c1 | 891 | struct shmid64_ds *shmid64) |
1da177e4 | 892 | { |
8d4cc8b5 | 893 | struct kern_ipc_perm *ipcp; |
8d4cc8b5 PP |
894 | struct shmid_kernel *shp; |
895 | int err; | |
896 | ||
d9a605e4 | 897 | down_write(&shm_ids(ns).rwsem); |
7b4cc5d8 DB |
898 | rcu_read_lock(); |
899 | ||
4241c1a3 | 900 | ipcp = ipcctl_obtain_check(ns, &shm_ids(ns), shmid, cmd, |
9ba720c1 | 901 | &shmid64->shm_perm, 0); |
7b4cc5d8 DB |
902 | if (IS_ERR(ipcp)) { |
903 | err = PTR_ERR(ipcp); | |
7b4cc5d8 DB |
904 | goto out_unlock1; |
905 | } | |
8d4cc8b5 | 906 | |
a5f75e7f | 907 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
8d4cc8b5 | 908 | |
7191adff | 909 | err = security_shm_shmctl(&shp->shm_perm, cmd); |
8d4cc8b5 | 910 | if (err) |
79ccf0f8 | 911 | goto out_unlock1; |
7b4cc5d8 | 912 | |
8d4cc8b5 PP |
913 | switch (cmd) { |
914 | case IPC_RMID: | |
79ccf0f8 | 915 | ipc_lock_object(&shp->shm_perm); |
7b4cc5d8 | 916 | /* do_shm_rmid unlocks the ipc object and rcu */ |
8d4cc8b5 PP |
917 | do_shm_rmid(ns, ipcp); |
918 | goto out_up; | |
919 | case IPC_SET: | |
79ccf0f8 | 920 | ipc_lock_object(&shp->shm_perm); |
9ba720c1 | 921 | err = ipc_update_perm(&shmid64->shm_perm, ipcp); |
1efdb69b | 922 | if (err) |
7b4cc5d8 | 923 | goto out_unlock0; |
7ff2819e | 924 | shp->shm_ctim = ktime_get_real_seconds(); |
8d4cc8b5 PP |
925 | break; |
926 | default: | |
927 | err = -EINVAL; | |
79ccf0f8 | 928 | goto out_unlock1; |
8d4cc8b5 | 929 | } |
7b4cc5d8 DB |
930 | |
931 | out_unlock0: | |
932 | ipc_unlock_object(&shp->shm_perm); | |
933 | out_unlock1: | |
934 | rcu_read_unlock(); | |
8d4cc8b5 | 935 | out_up: |
d9a605e4 | 936 | up_write(&shm_ids(ns).rwsem); |
8d4cc8b5 PP |
937 | return err; |
938 | } | |
939 | ||
9ba720c1 AV |
940 | static int shmctl_ipc_info(struct ipc_namespace *ns, |
941 | struct shminfo64 *shminfo) | |
8d4cc8b5 | 942 | { |
9ba720c1 AV |
943 | int err = security_shm_shmctl(NULL, IPC_INFO); |
944 | if (!err) { | |
945 | memset(shminfo, 0, sizeof(*shminfo)); | |
946 | shminfo->shmmni = shminfo->shmseg = ns->shm_ctlmni; | |
947 | shminfo->shmmax = ns->shm_ctlmax; | |
948 | shminfo->shmall = ns->shm_ctlall; | |
949 | shminfo->shmmin = SHMMIN; | |
d9a605e4 | 950 | down_read(&shm_ids(ns).rwsem); |
27c331a1 | 951 | err = ipc_get_maxidx(&shm_ids(ns)); |
d9a605e4 | 952 | up_read(&shm_ids(ns).rwsem); |
239521f3 | 953 | if (err < 0) |
1da177e4 | 954 | err = 0; |
1da177e4 | 955 | } |
9ba720c1 AV |
956 | return err; |
957 | } | |
1da177e4 | 958 | |
9ba720c1 AV |
959 | static int shmctl_shm_info(struct ipc_namespace *ns, |
960 | struct shm_info *shm_info) | |
961 | { | |
962 | int err = security_shm_shmctl(NULL, SHM_INFO); | |
963 | if (!err) { | |
964 | memset(shm_info, 0, sizeof(*shm_info)); | |
d9a605e4 | 965 | down_read(&shm_ids(ns).rwsem); |
9ba720c1 AV |
966 | shm_info->used_ids = shm_ids(ns).in_use; |
967 | shm_get_stat(ns, &shm_info->shm_rss, &shm_info->shm_swp); | |
968 | shm_info->shm_tot = ns->shm_tot; | |
969 | shm_info->swap_attempts = 0; | |
970 | shm_info->swap_successes = 0; | |
27c331a1 | 971 | err = ipc_get_maxidx(&shm_ids(ns)); |
d9a605e4 | 972 | up_read(&shm_ids(ns).rwsem); |
9ba720c1 AV |
973 | if (err < 0) |
974 | err = 0; | |
1da177e4 | 975 | } |
9ba720c1 AV |
976 | return err; |
977 | } | |
c97cb9cc | 978 | |
9ba720c1 AV |
979 | static int shmctl_stat(struct ipc_namespace *ns, int shmid, |
980 | int cmd, struct shmid64_ds *tbuf) | |
981 | { | |
982 | struct shmid_kernel *shp; | |
9ba720c1 | 983 | int err; |
c97cb9cc | 984 | |
87ad4b0d PM |
985 | memset(tbuf, 0, sizeof(*tbuf)); |
986 | ||
9ba720c1 | 987 | rcu_read_lock(); |
c21a6970 | 988 | if (cmd == SHM_STAT || cmd == SHM_STAT_ANY) { |
9ba720c1 AV |
989 | shp = shm_obtain_object(ns, shmid); |
990 | if (IS_ERR(shp)) { | |
991 | err = PTR_ERR(shp); | |
992 | goto out_unlock; | |
993 | } | |
c21a6970 | 994 | } else { /* IPC_STAT */ |
9ba720c1 AV |
995 | shp = shm_obtain_object_check(ns, shmid); |
996 | if (IS_ERR(shp)) { | |
997 | err = PTR_ERR(shp); | |
1da177e4 | 998 | goto out_unlock; |
1da177e4 | 999 | } |
9ba720c1 | 1000 | } |
1da177e4 | 1001 | |
c21a6970 DB |
1002 | /* |
1003 | * Semantically SHM_STAT_ANY ought to be identical to | |
1004 | * that functionality provided by the /proc/sysvipc/ | |
1005 | * interface. As such, only audit these calls and | |
1006 | * do not do traditional S_IRUGO permission checks on | |
1007 | * the ipc object. | |
1008 | */ | |
1009 | if (cmd == SHM_STAT_ANY) | |
1010 | audit_ipc_obj(&shp->shm_perm); | |
1011 | else { | |
1012 | err = -EACCES; | |
1013 | if (ipcperms(ns, &shp->shm_perm, S_IRUGO)) | |
1014 | goto out_unlock; | |
1015 | } | |
c97cb9cc | 1016 | |
7191adff | 1017 | err = security_shm_shmctl(&shp->shm_perm, cmd); |
9ba720c1 AV |
1018 | if (err) |
1019 | goto out_unlock; | |
1020 | ||
87ad4b0d PM |
1021 | ipc_lock_object(&shp->shm_perm); |
1022 | ||
1023 | if (!ipc_valid_object(&shp->shm_perm)) { | |
1024 | ipc_unlock_object(&shp->shm_perm); | |
1025 | err = -EIDRM; | |
1026 | goto out_unlock; | |
1027 | } | |
1028 | ||
9ba720c1 AV |
1029 | kernel_to_ipc64_perm(&shp->shm_perm, &tbuf->shm_perm); |
1030 | tbuf->shm_segsz = shp->shm_segsz; | |
1031 | tbuf->shm_atime = shp->shm_atim; | |
1032 | tbuf->shm_dtime = shp->shm_dtim; | |
1033 | tbuf->shm_ctime = shp->shm_ctim; | |
c2ab975c AB |
1034 | #ifndef CONFIG_64BIT |
1035 | tbuf->shm_atime_high = shp->shm_atim >> 32; | |
1036 | tbuf->shm_dtime_high = shp->shm_dtim >> 32; | |
1037 | tbuf->shm_ctime_high = shp->shm_ctim >> 32; | |
1038 | #endif | |
98f929b1 EB |
1039 | tbuf->shm_cpid = pid_vnr(shp->shm_cprid); |
1040 | tbuf->shm_lpid = pid_vnr(shp->shm_lprid); | |
9ba720c1 | 1041 | tbuf->shm_nattch = shp->shm_nattch; |
87ad4b0d | 1042 | |
615c999c MS |
1043 | if (cmd == IPC_STAT) { |
1044 | /* | |
1045 | * As defined in SUS: | |
1046 | * Return 0 on success | |
1047 | */ | |
1048 | err = 0; | |
1049 | } else { | |
1050 | /* | |
1051 | * SHM_STAT and SHM_STAT_ANY (both Linux specific) | |
1052 | * Return the full id, including the sequence number | |
1053 | */ | |
1054 | err = shp->shm_perm.id; | |
1055 | } | |
68eccc1d | 1056 | |
615c999c | 1057 | ipc_unlock_object(&shp->shm_perm); |
68eccc1d | 1058 | out_unlock: |
c97cb9cc | 1059 | rcu_read_unlock(); |
68eccc1d DB |
1060 | return err; |
1061 | } | |
1062 | ||
9ba720c1 | 1063 | static int shmctl_do_lock(struct ipc_namespace *ns, int shmid, int cmd) |
68eccc1d DB |
1064 | { |
1065 | struct shmid_kernel *shp; | |
9ba720c1 AV |
1066 | struct file *shm_file; |
1067 | int err; | |
68eccc1d | 1068 | |
9ba720c1 AV |
1069 | rcu_read_lock(); |
1070 | shp = shm_obtain_object_check(ns, shmid); | |
1071 | if (IS_ERR(shp)) { | |
1072 | err = PTR_ERR(shp); | |
1073 | goto out_unlock1; | |
1da177e4 | 1074 | } |
c97cb9cc | 1075 | |
9ba720c1 | 1076 | audit_ipc_obj(&(shp->shm_perm)); |
7191adff | 1077 | err = security_shm_shmctl(&shp->shm_perm, cmd); |
9ba720c1 AV |
1078 | if (err) |
1079 | goto out_unlock1; | |
c97cb9cc | 1080 | |
9ba720c1 | 1081 | ipc_lock_object(&shp->shm_perm); |
c97cb9cc | 1082 | |
9ba720c1 AV |
1083 | /* check if shm_destroy() is tearing down shp */ |
1084 | if (!ipc_valid_object(&shp->shm_perm)) { | |
1085 | err = -EIDRM; | |
1086 | goto out_unlock0; | |
1da177e4 | 1087 | } |
073115d6 | 1088 | |
9ba720c1 AV |
1089 | if (!ns_capable(ns->user_ns, CAP_IPC_LOCK)) { |
1090 | kuid_t euid = current_euid(); | |
0f3d2b01 | 1091 | |
9ba720c1 AV |
1092 | if (!uid_eq(euid, shp->shm_perm.uid) && |
1093 | !uid_eq(euid, shp->shm_perm.cuid)) { | |
1094 | err = -EPERM; | |
0f3d2b01 RA |
1095 | goto out_unlock0; |
1096 | } | |
9ba720c1 AV |
1097 | if (cmd == SHM_LOCK && !rlimit(RLIMIT_MEMLOCK)) { |
1098 | err = -EPERM; | |
1099 | goto out_unlock0; | |
1da177e4 | 1100 | } |
68eccc1d DB |
1101 | } |
1102 | ||
9ba720c1 AV |
1103 | shm_file = shp->shm_file; |
1104 | if (is_file_hugepages(shm_file)) | |
1105 | goto out_unlock0; | |
85046579 | 1106 | |
9ba720c1 AV |
1107 | if (cmd == SHM_LOCK) { |
1108 | struct user_struct *user = current_user(); | |
63980c80 | 1109 | |
9ba720c1 AV |
1110 | err = shmem_lock(shm_file, 1, user); |
1111 | if (!err && !(shp->shm_perm.mode & SHM_LOCKED)) { | |
1112 | shp->shm_perm.mode |= SHM_LOCKED; | |
1113 | shp->mlock_user = user; | |
1da177e4 | 1114 | } |
9ba720c1 | 1115 | goto out_unlock0; |
1da177e4 LT |
1116 | } |
1117 | ||
9ba720c1 AV |
1118 | /* SHM_UNLOCK */ |
1119 | if (!(shp->shm_perm.mode & SHM_LOCKED)) | |
1120 | goto out_unlock0; | |
1121 | shmem_lock(shm_file, 0, shp->mlock_user); | |
1122 | shp->shm_perm.mode &= ~SHM_LOCKED; | |
1123 | shp->mlock_user = NULL; | |
1124 | get_file(shm_file); | |
1125 | ipc_unlock_object(&shp->shm_perm); | |
1126 | rcu_read_unlock(); | |
1127 | shmem_unlock_mapping(shm_file->f_mapping); | |
1128 | ||
1129 | fput(shm_file); | |
1130 | return err; | |
1131 | ||
2caacaa8 DB |
1132 | out_unlock0: |
1133 | ipc_unlock_object(&shp->shm_perm); | |
1134 | out_unlock1: | |
c97cb9cc | 1135 | rcu_read_unlock(); |
68eccc1d DB |
1136 | return err; |
1137 | } | |
1138 | ||
275f2214 | 1139 | static long ksys_shmctl(int shmid, int cmd, struct shmid_ds __user *buf, int version) |
68eccc1d | 1140 | { |
275f2214 | 1141 | int err; |
68eccc1d | 1142 | struct ipc_namespace *ns; |
553f770e | 1143 | struct shmid64_ds sem64; |
68eccc1d | 1144 | |
2caacaa8 DB |
1145 | if (cmd < 0 || shmid < 0) |
1146 | return -EINVAL; | |
68eccc1d | 1147 | |
68eccc1d DB |
1148 | ns = current->nsproxy->ipc_ns; |
1149 | ||
1150 | switch (cmd) { | |
9ba720c1 AV |
1151 | case IPC_INFO: { |
1152 | struct shminfo64 shminfo; | |
1153 | err = shmctl_ipc_info(ns, &shminfo); | |
1154 | if (err < 0) | |
1155 | return err; | |
1156 | if (copy_shminfo_to_user(buf, &shminfo, version)) | |
1157 | err = -EFAULT; | |
1158 | return err; | |
1159 | } | |
1160 | case SHM_INFO: { | |
1161 | struct shm_info shm_info; | |
1162 | err = shmctl_shm_info(ns, &shm_info); | |
1163 | if (err < 0) | |
1164 | return err; | |
1165 | if (copy_to_user(buf, &shm_info, sizeof(shm_info))) | |
1166 | err = -EFAULT; | |
1167 | return err; | |
1168 | } | |
68eccc1d | 1169 | case SHM_STAT: |
c21a6970 | 1170 | case SHM_STAT_ANY: |
9ba720c1 | 1171 | case IPC_STAT: { |
553f770e | 1172 | err = shmctl_stat(ns, shmid, cmd, &sem64); |
9ba720c1 AV |
1173 | if (err < 0) |
1174 | return err; | |
553f770e | 1175 | if (copy_shmid_to_user(buf, &sem64, version)) |
9ba720c1 AV |
1176 | err = -EFAULT; |
1177 | return err; | |
1178 | } | |
2caacaa8 | 1179 | case IPC_SET: |
553f770e | 1180 | if (copy_shmid_from_user(&sem64, buf, version)) |
9ba720c1 | 1181 | return -EFAULT; |
df561f66 | 1182 | fallthrough; |
9ba720c1 | 1183 | case IPC_RMID: |
553f770e | 1184 | return shmctl_down(ns, shmid, cmd, &sem64); |
1da177e4 LT |
1185 | case SHM_LOCK: |
1186 | case SHM_UNLOCK: | |
9ba720c1 AV |
1187 | return shmctl_do_lock(ns, shmid, cmd); |
1188 | default: | |
1189 | return -EINVAL; | |
1190 | } | |
1191 | } | |
89e004ea | 1192 | |
c84d0791 DB |
1193 | SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, struct shmid_ds __user *, buf) |
1194 | { | |
275f2214 AB |
1195 | return ksys_shmctl(shmid, cmd, buf, IPC_64); |
1196 | } | |
1197 | ||
1198 | #ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION | |
1199 | long ksys_old_shmctl(int shmid, int cmd, struct shmid_ds __user *buf) | |
1200 | { | |
1201 | int version = ipc_parse_version(&cmd); | |
1202 | ||
1203 | return ksys_shmctl(shmid, cmd, buf, version); | |
1204 | } | |
1205 | ||
1206 | SYSCALL_DEFINE3(old_shmctl, int, shmid, int, cmd, struct shmid_ds __user *, buf) | |
1207 | { | |
1208 | return ksys_old_shmctl(shmid, cmd, buf); | |
c84d0791 | 1209 | } |
275f2214 | 1210 | #endif |
c84d0791 | 1211 | |
553f770e AV |
1212 | #ifdef CONFIG_COMPAT |
1213 | ||
1214 | struct compat_shmid_ds { | |
1215 | struct compat_ipc_perm shm_perm; | |
1216 | int shm_segsz; | |
9afc5eee AB |
1217 | old_time32_t shm_atime; |
1218 | old_time32_t shm_dtime; | |
1219 | old_time32_t shm_ctime; | |
553f770e AV |
1220 | compat_ipc_pid_t shm_cpid; |
1221 | compat_ipc_pid_t shm_lpid; | |
1222 | unsigned short shm_nattch; | |
1223 | unsigned short shm_unused; | |
1224 | compat_uptr_t shm_unused2; | |
1225 | compat_uptr_t shm_unused3; | |
1226 | }; | |
1da177e4 | 1227 | |
553f770e AV |
1228 | struct compat_shminfo64 { |
1229 | compat_ulong_t shmmax; | |
1230 | compat_ulong_t shmmin; | |
1231 | compat_ulong_t shmmni; | |
1232 | compat_ulong_t shmseg; | |
1233 | compat_ulong_t shmall; | |
1234 | compat_ulong_t __unused1; | |
1235 | compat_ulong_t __unused2; | |
1236 | compat_ulong_t __unused3; | |
1237 | compat_ulong_t __unused4; | |
1238 | }; | |
073115d6 | 1239 | |
553f770e AV |
1240 | struct compat_shm_info { |
1241 | compat_int_t used_ids; | |
1242 | compat_ulong_t shm_tot, shm_rss, shm_swp; | |
1243 | compat_ulong_t swap_attempts, swap_successes; | |
1244 | }; | |
0f3d2b01 | 1245 | |
553f770e AV |
1246 | static int copy_compat_shminfo_to_user(void __user *buf, struct shminfo64 *in, |
1247 | int version) | |
1248 | { | |
1249 | if (in->shmmax > INT_MAX) | |
1250 | in->shmmax = INT_MAX; | |
1251 | if (version == IPC_64) { | |
1252 | struct compat_shminfo64 info; | |
1253 | memset(&info, 0, sizeof(info)); | |
1254 | info.shmmax = in->shmmax; | |
1255 | info.shmmin = in->shmmin; | |
1256 | info.shmmni = in->shmmni; | |
1257 | info.shmseg = in->shmseg; | |
1258 | info.shmall = in->shmall; | |
1259 | return copy_to_user(buf, &info, sizeof(info)); | |
1260 | } else { | |
1261 | struct shminfo info; | |
1262 | memset(&info, 0, sizeof(info)); | |
1263 | info.shmmax = in->shmmax; | |
1264 | info.shmmin = in->shmmin; | |
1265 | info.shmmni = in->shmmni; | |
1266 | info.shmseg = in->shmseg; | |
1267 | info.shmall = in->shmall; | |
1268 | return copy_to_user(buf, &info, sizeof(info)); | |
1269 | } | |
1270 | } | |
0f3d2b01 | 1271 | |
553f770e AV |
1272 | static int put_compat_shm_info(struct shm_info *ip, |
1273 | struct compat_shm_info __user *uip) | |
1274 | { | |
1275 | struct compat_shm_info info; | |
1276 | ||
1277 | memset(&info, 0, sizeof(info)); | |
1278 | info.used_ids = ip->used_ids; | |
1279 | info.shm_tot = ip->shm_tot; | |
1280 | info.shm_rss = ip->shm_rss; | |
1281 | info.shm_swp = ip->shm_swp; | |
1282 | info.swap_attempts = ip->swap_attempts; | |
1283 | info.swap_successes = ip->swap_successes; | |
b776e4b1 | 1284 | return copy_to_user(uip, &info, sizeof(info)); |
553f770e | 1285 | } |
1da177e4 | 1286 | |
553f770e AV |
1287 | static int copy_compat_shmid_to_user(void __user *buf, struct shmid64_ds *in, |
1288 | int version) | |
1289 | { | |
1290 | if (version == IPC_64) { | |
1291 | struct compat_shmid64_ds v; | |
1292 | memset(&v, 0, sizeof(v)); | |
28327fae | 1293 | to_compat_ipc64_perm(&v.shm_perm, &in->shm_perm); |
c2ab975c AB |
1294 | v.shm_atime = lower_32_bits(in->shm_atime); |
1295 | v.shm_atime_high = upper_32_bits(in->shm_atime); | |
1296 | v.shm_dtime = lower_32_bits(in->shm_dtime); | |
1297 | v.shm_dtime_high = upper_32_bits(in->shm_dtime); | |
1298 | v.shm_ctime = lower_32_bits(in->shm_ctime); | |
1299 | v.shm_ctime_high = upper_32_bits(in->shm_ctime); | |
553f770e AV |
1300 | v.shm_segsz = in->shm_segsz; |
1301 | v.shm_nattch = in->shm_nattch; | |
1302 | v.shm_cpid = in->shm_cpid; | |
1303 | v.shm_lpid = in->shm_lpid; | |
1304 | return copy_to_user(buf, &v, sizeof(v)); | |
1305 | } else { | |
1306 | struct compat_shmid_ds v; | |
1307 | memset(&v, 0, sizeof(v)); | |
28327fae | 1308 | to_compat_ipc_perm(&v.shm_perm, &in->shm_perm); |
553f770e | 1309 | v.shm_perm.key = in->shm_perm.key; |
553f770e AV |
1310 | v.shm_atime = in->shm_atime; |
1311 | v.shm_dtime = in->shm_dtime; | |
1312 | v.shm_ctime = in->shm_ctime; | |
1313 | v.shm_segsz = in->shm_segsz; | |
1314 | v.shm_nattch = in->shm_nattch; | |
1315 | v.shm_cpid = in->shm_cpid; | |
1316 | v.shm_lpid = in->shm_lpid; | |
1317 | return copy_to_user(buf, &v, sizeof(v)); | |
1318 | } | |
1319 | } | |
85046579 | 1320 | |
553f770e AV |
1321 | static int copy_compat_shmid_from_user(struct shmid64_ds *out, void __user *buf, |
1322 | int version) | |
1323 | { | |
1324 | memset(out, 0, sizeof(*out)); | |
1325 | if (version == IPC_64) { | |
6aa211e8 | 1326 | struct compat_shmid64_ds __user *p = buf; |
28327fae | 1327 | return get_compat_ipc64_perm(&out->shm_perm, &p->shm_perm); |
553f770e | 1328 | } else { |
6aa211e8 | 1329 | struct compat_shmid_ds __user *p = buf; |
28327fae | 1330 | return get_compat_ipc_perm(&out->shm_perm, &p->shm_perm); |
553f770e | 1331 | } |
553f770e | 1332 | } |
63980c80 | 1333 | |
1cd377ba | 1334 | static long compat_ksys_shmctl(int shmid, int cmd, void __user *uptr, int version) |
553f770e AV |
1335 | { |
1336 | struct ipc_namespace *ns; | |
1337 | struct shmid64_ds sem64; | |
553f770e | 1338 | int err; |
85046579 | 1339 | |
553f770e AV |
1340 | ns = current->nsproxy->ipc_ns; |
1341 | ||
1342 | if (cmd < 0 || shmid < 0) | |
1343 | return -EINVAL; | |
2caacaa8 | 1344 | |
553f770e AV |
1345 | switch (cmd) { |
1346 | case IPC_INFO: { | |
1347 | struct shminfo64 shminfo; | |
1348 | err = shmctl_ipc_info(ns, &shminfo); | |
1349 | if (err < 0) | |
1350 | return err; | |
1351 | if (copy_compat_shminfo_to_user(uptr, &shminfo, version)) | |
1352 | err = -EFAULT; | |
1353 | return err; | |
1354 | } | |
1355 | case SHM_INFO: { | |
1356 | struct shm_info shm_info; | |
1357 | err = shmctl_shm_info(ns, &shm_info); | |
1358 | if (err < 0) | |
1359 | return err; | |
1360 | if (put_compat_shm_info(&shm_info, uptr)) | |
1361 | err = -EFAULT; | |
8d4cc8b5 | 1362 | return err; |
2caacaa8 | 1363 | } |
553f770e | 1364 | case IPC_STAT: |
c21a6970 | 1365 | case SHM_STAT_ANY: |
553f770e AV |
1366 | case SHM_STAT: |
1367 | err = shmctl_stat(ns, shmid, cmd, &sem64); | |
1368 | if (err < 0) | |
1369 | return err; | |
58aff0af | 1370 | if (copy_compat_shmid_to_user(uptr, &sem64, version)) |
553f770e AV |
1371 | err = -EFAULT; |
1372 | return err; | |
1373 | ||
1374 | case IPC_SET: | |
1375 | if (copy_compat_shmid_from_user(&sem64, uptr, version)) | |
1376 | return -EFAULT; | |
df561f66 | 1377 | fallthrough; |
553f770e AV |
1378 | case IPC_RMID: |
1379 | return shmctl_down(ns, shmid, cmd, &sem64); | |
1380 | case SHM_LOCK: | |
1381 | case SHM_UNLOCK: | |
1382 | return shmctl_do_lock(ns, shmid, cmd); | |
1da177e4 | 1383 | default: |
8d4cc8b5 | 1384 | return -EINVAL; |
1da177e4 | 1385 | } |
1da177e4 LT |
1386 | return err; |
1387 | } | |
c84d0791 DB |
1388 | |
1389 | COMPAT_SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, void __user *, uptr) | |
1390 | { | |
275f2214 | 1391 | return compat_ksys_shmctl(shmid, cmd, uptr, IPC_64); |
c84d0791 | 1392 | } |
275f2214 AB |
1393 | |
1394 | #ifdef CONFIG_ARCH_WANT_COMPAT_IPC_PARSE_VERSION | |
1395 | long compat_ksys_old_shmctl(int shmid, int cmd, void __user *uptr) | |
1396 | { | |
1397 | int version = compat_ipc_parse_version(&cmd); | |
1398 | ||
1399 | return compat_ksys_shmctl(shmid, cmd, uptr, version); | |
1400 | } | |
1401 | ||
1402 | COMPAT_SYSCALL_DEFINE3(old_shmctl, int, shmid, int, cmd, void __user *, uptr) | |
1403 | { | |
1404 | return compat_ksys_old_shmctl(shmid, cmd, uptr); | |
1405 | } | |
1406 | #endif | |
553f770e | 1407 | #endif |
1da177e4 LT |
1408 | |
1409 | /* | |
1410 | * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists. | |
1411 | * | |
1412 | * NOTE! Despite the name, this is NOT a direct system call entrypoint. The | |
1413 | * "raddr" thing points to kernel space, and there has to be a wrapper around | |
1414 | * this. | |
1415 | */ | |
95e91b83 DB |
1416 | long do_shmat(int shmid, char __user *shmaddr, int shmflg, |
1417 | ulong *raddr, unsigned long shmlba) | |
1da177e4 LT |
1418 | { |
1419 | struct shmid_kernel *shp; | |
f0cb8802 | 1420 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1421 | unsigned long size; |
4f089acc | 1422 | struct file *file, *base; |
1da177e4 | 1423 | int err; |
f0cb8802 | 1424 | unsigned long flags = MAP_SHARED; |
1da177e4 | 1425 | unsigned long prot; |
1da177e4 | 1426 | int acc_mode; |
4e982311 | 1427 | struct ipc_namespace *ns; |
bc56bba8 | 1428 | struct shm_file_data *sfd; |
c9c554f2 | 1429 | int f_flags; |
41badc15 | 1430 | unsigned long populate = 0; |
1da177e4 | 1431 | |
bc56bba8 EB |
1432 | err = -EINVAL; |
1433 | if (shmid < 0) | |
1da177e4 | 1434 | goto out; |
f0cb8802 DB |
1435 | |
1436 | if (addr) { | |
079a96ae | 1437 | if (addr & (shmlba - 1)) { |
8f89c007 | 1438 | if (shmflg & SHM_RND) { |
a73ab244 | 1439 | addr &= ~(shmlba - 1); /* round down */ |
8f89c007 DB |
1440 | |
1441 | /* | |
1442 | * Ensure that the round-down is non-nil | |
1443 | * when remapping. This can happen for | |
1444 | * cases when addr < shmlba. | |
1445 | */ | |
1446 | if (!addr && (shmflg & SHM_REMAP)) | |
1447 | goto out; | |
1448 | } else | |
1da177e4 LT |
1449 | #ifndef __ARCH_FORCE_SHMLBA |
1450 | if (addr & ~PAGE_MASK) | |
1451 | #endif | |
bc56bba8 | 1452 | goto out; |
1da177e4 | 1453 | } |
1da177e4 | 1454 | |
f0cb8802 DB |
1455 | flags |= MAP_FIXED; |
1456 | } else if ((shmflg & SHM_REMAP)) | |
1457 | goto out; | |
1da177e4 LT |
1458 | |
1459 | if (shmflg & SHM_RDONLY) { | |
1460 | prot = PROT_READ; | |
1da177e4 | 1461 | acc_mode = S_IRUGO; |
c9c554f2 | 1462 | f_flags = O_RDONLY; |
1da177e4 LT |
1463 | } else { |
1464 | prot = PROT_READ | PROT_WRITE; | |
1da177e4 | 1465 | acc_mode = S_IRUGO | S_IWUGO; |
c9c554f2 | 1466 | f_flags = O_RDWR; |
1da177e4 LT |
1467 | } |
1468 | if (shmflg & SHM_EXEC) { | |
1469 | prot |= PROT_EXEC; | |
1470 | acc_mode |= S_IXUGO; | |
1471 | } | |
1472 | ||
1473 | /* | |
1474 | * We cannot rely on the fs check since SYSV IPC does have an | |
1475 | * additional creator id... | |
1476 | */ | |
4e982311 | 1477 | ns = current->nsproxy->ipc_ns; |
c2c737a0 DB |
1478 | rcu_read_lock(); |
1479 | shp = shm_obtain_object_check(ns, shmid); | |
023a5355 ND |
1480 | if (IS_ERR(shp)) { |
1481 | err = PTR_ERR(shp); | |
c2c737a0 | 1482 | goto out_unlock; |
023a5355 | 1483 | } |
bc56bba8 EB |
1484 | |
1485 | err = -EACCES; | |
b0e77598 | 1486 | if (ipcperms(ns, &shp->shm_perm, acc_mode)) |
bc56bba8 | 1487 | goto out_unlock; |
1da177e4 | 1488 | |
7191adff | 1489 | err = security_shm_shmat(&shp->shm_perm, shmaddr, shmflg); |
bc56bba8 EB |
1490 | if (err) |
1491 | goto out_unlock; | |
1492 | ||
c2c737a0 | 1493 | ipc_lock_object(&shp->shm_perm); |
a399b29d GT |
1494 | |
1495 | /* check if shm_destroy() is tearing down shp */ | |
0f3d2b01 | 1496 | if (!ipc_valid_object(&shp->shm_perm)) { |
a399b29d GT |
1497 | ipc_unlock_object(&shp->shm_perm); |
1498 | err = -EIDRM; | |
1499 | goto out_unlock; | |
1500 | } | |
1501 | ||
4f089acc AV |
1502 | /* |
1503 | * We need to take a reference to the real shm file to prevent the | |
1504 | * pointer from becoming stale in cases where the lifetime of the outer | |
1505 | * file extends beyond that of the shm segment. It's not usually | |
1506 | * possible, but it can happen during remap_file_pages() emulation as | |
1507 | * that unmaps the memory, then does ->mmap() via file reference only. | |
1508 | * We'll deny the ->mmap() if the shm segment was since removed, but to | |
1509 | * detect shm ID reuse we need to compare the file pointers. | |
1510 | */ | |
1511 | base = get_file(shp->shm_file); | |
1da177e4 | 1512 | shp->shm_nattch++; |
4f089acc | 1513 | size = i_size_read(file_inode(base)); |
c2c737a0 DB |
1514 | ipc_unlock_object(&shp->shm_perm); |
1515 | rcu_read_unlock(); | |
1da177e4 | 1516 | |
bc56bba8 EB |
1517 | err = -ENOMEM; |
1518 | sfd = kzalloc(sizeof(*sfd), GFP_KERNEL); | |
f42569b1 | 1519 | if (!sfd) { |
4f089acc | 1520 | fput(base); |
f42569b1 DB |
1521 | goto out_nattch; |
1522 | } | |
bc56bba8 | 1523 | |
4f089acc AV |
1524 | file = alloc_file_clone(base, f_flags, |
1525 | is_file_hugepages(base) ? | |
c4caa778 AV |
1526 | &shm_file_operations_huge : |
1527 | &shm_file_operations); | |
39b65252 | 1528 | err = PTR_ERR(file); |
f42569b1 DB |
1529 | if (IS_ERR(file)) { |
1530 | kfree(sfd); | |
4f089acc | 1531 | fput(base); |
f42569b1 DB |
1532 | goto out_nattch; |
1533 | } | |
bc56bba8 | 1534 | |
7ca7e564 | 1535 | sfd->id = shp->shm_perm.id; |
bc56bba8 | 1536 | sfd->ns = get_ipc_ns(ns); |
4f089acc | 1537 | sfd->file = base; |
bc56bba8 | 1538 | sfd->vm_ops = NULL; |
4f089acc | 1539 | file->private_data = sfd; |
bc56bba8 | 1540 | |
8b3ec681 AV |
1541 | err = security_mmap_file(file, prot, flags); |
1542 | if (err) | |
1543 | goto out_fput; | |
1544 | ||
d8ed45c5 | 1545 | if (mmap_write_lock_killable(current->mm)) { |
91f4f94e MH |
1546 | err = -EINTR; |
1547 | goto out_fput; | |
1548 | } | |
1549 | ||
1da177e4 | 1550 | if (addr && !(shmflg & SHM_REMAP)) { |
bc56bba8 | 1551 | err = -EINVAL; |
247a8ce8 MS |
1552 | if (addr + size < addr) |
1553 | goto invalid; | |
1554 | ||
1da177e4 LT |
1555 | if (find_vma_intersection(current->mm, addr, addr + size)) |
1556 | goto invalid; | |
1da177e4 | 1557 | } |
f42569b1 | 1558 | |
45e55300 | 1559 | addr = do_mmap(file, addr, size, prot, flags, 0, &populate, NULL); |
bebeb3d6 | 1560 | *raddr = addr; |
bc56bba8 | 1561 | err = 0; |
bebeb3d6 ML |
1562 | if (IS_ERR_VALUE(addr)) |
1563 | err = (long)addr; | |
1da177e4 | 1564 | invalid: |
d8ed45c5 | 1565 | mmap_write_unlock(current->mm); |
bebeb3d6 | 1566 | if (populate) |
41badc15 | 1567 | mm_populate(addr, populate); |
1da177e4 | 1568 | |
8b3ec681 | 1569 | out_fput: |
bc56bba8 EB |
1570 | fput(file); |
1571 | ||
1572 | out_nattch: | |
d9a605e4 | 1573 | down_write(&shm_ids(ns).rwsem); |
00c2bf85 | 1574 | shp = shm_lock(ns, shmid); |
1da177e4 | 1575 | shp->shm_nattch--; |
b34a6b1d | 1576 | if (shm_may_destroy(ns, shp)) |
4e982311 | 1577 | shm_destroy(ns, shp); |
1da177e4 LT |
1578 | else |
1579 | shm_unlock(shp); | |
d9a605e4 | 1580 | up_write(&shm_ids(ns).rwsem); |
1da177e4 | 1581 | return err; |
bc56bba8 EB |
1582 | |
1583 | out_unlock: | |
c2c737a0 | 1584 | rcu_read_unlock(); |
f42569b1 DB |
1585 | out: |
1586 | return err; | |
1da177e4 LT |
1587 | } |
1588 | ||
d5460c99 | 1589 | SYSCALL_DEFINE3(shmat, int, shmid, char __user *, shmaddr, int, shmflg) |
7d87e14c SR |
1590 | { |
1591 | unsigned long ret; | |
1592 | long err; | |
1593 | ||
079a96ae | 1594 | err = do_shmat(shmid, shmaddr, shmflg, &ret, SHMLBA); |
7d87e14c SR |
1595 | if (err) |
1596 | return err; | |
1597 | force_successful_syscall_return(); | |
1598 | return (long)ret; | |
1599 | } | |
1600 | ||
a78ee9ed AV |
1601 | #ifdef CONFIG_COMPAT |
1602 | ||
1603 | #ifndef COMPAT_SHMLBA | |
1604 | #define COMPAT_SHMLBA SHMLBA | |
1605 | #endif | |
1606 | ||
1607 | COMPAT_SYSCALL_DEFINE3(shmat, int, shmid, compat_uptr_t, shmaddr, int, shmflg) | |
1608 | { | |
1609 | unsigned long ret; | |
1610 | long err; | |
1611 | ||
1612 | err = do_shmat(shmid, compat_ptr(shmaddr), shmflg, &ret, COMPAT_SHMLBA); | |
1613 | if (err) | |
1614 | return err; | |
1615 | force_successful_syscall_return(); | |
1616 | return (long)ret; | |
1617 | } | |
1618 | #endif | |
1619 | ||
1da177e4 LT |
1620 | /* |
1621 | * detach and kill segment if marked destroyed. | |
1622 | * The work is done in shm_close. | |
1623 | */ | |
da1e2744 | 1624 | long ksys_shmdt(char __user *shmaddr) |
1da177e4 LT |
1625 | { |
1626 | struct mm_struct *mm = current->mm; | |
586c7e6a | 1627 | struct vm_area_struct *vma; |
1da177e4 | 1628 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1629 | int retval = -EINVAL; |
586c7e6a MF |
1630 | #ifdef CONFIG_MMU |
1631 | loff_t size = 0; | |
d3c97900 | 1632 | struct file *file; |
586c7e6a MF |
1633 | struct vm_area_struct *next; |
1634 | #endif | |
1da177e4 | 1635 | |
df1e2fb5 HD |
1636 | if (addr & ~PAGE_MASK) |
1637 | return retval; | |
1638 | ||
d8ed45c5 | 1639 | if (mmap_write_lock_killable(mm)) |
91f4f94e | 1640 | return -EINTR; |
1da177e4 LT |
1641 | |
1642 | /* | |
1643 | * This function tries to be smart and unmap shm segments that | |
1644 | * were modified by partial mlock or munmap calls: | |
1645 | * - It first determines the size of the shm segment that should be | |
1646 | * unmapped: It searches for a vma that is backed by shm and that | |
1647 | * started at address shmaddr. It records it's size and then unmaps | |
1648 | * it. | |
1649 | * - Then it unmaps all shm vmas that started at shmaddr and that | |
d3c97900 DH |
1650 | * are within the initially determined size and that are from the |
1651 | * same shm segment from which we determined the size. | |
1da177e4 LT |
1652 | * Errors from do_munmap are ignored: the function only fails if |
1653 | * it's called with invalid parameters or if it's called to unmap | |
1654 | * a part of a vma. Both calls in this function are for full vmas, | |
1655 | * the parameters are directly copied from the vma itself and always | |
1656 | * valid - therefore do_munmap cannot fail. (famous last words?) | |
1657 | */ | |
1658 | /* | |
1659 | * If it had been mremap()'d, the starting address would not | |
1660 | * match the usual checks anyway. So assume all vma's are | |
1661 | * above the starting address given. | |
1662 | */ | |
1663 | vma = find_vma(mm, addr); | |
1664 | ||
8feae131 | 1665 | #ifdef CONFIG_MMU |
1da177e4 LT |
1666 | while (vma) { |
1667 | next = vma->vm_next; | |
1668 | ||
1669 | /* | |
1670 | * Check if the starting address would match, i.e. it's | |
1671 | * a fragment created by mprotect() and/or munmap(), or it | |
1672 | * otherwise it starts at this address with no hassles. | |
1673 | */ | |
bc56bba8 | 1674 | if ((vma->vm_ops == &shm_vm_ops) && |
1da177e4 LT |
1675 | (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) { |
1676 | ||
d3c97900 DH |
1677 | /* |
1678 | * Record the file of the shm segment being | |
1679 | * unmapped. With mremap(), someone could place | |
1680 | * page from another segment but with equal offsets | |
1681 | * in the range we are unmapping. | |
1682 | */ | |
1683 | file = vma->vm_file; | |
07a46ed2 | 1684 | size = i_size_read(file_inode(vma->vm_file)); |
897ab3e0 | 1685 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1686 | /* |
1687 | * We discovered the size of the shm segment, so | |
1688 | * break out of here and fall through to the next | |
1689 | * loop that uses the size information to stop | |
1690 | * searching for matching vma's. | |
1691 | */ | |
1692 | retval = 0; | |
1693 | vma = next; | |
1694 | break; | |
1695 | } | |
1696 | vma = next; | |
1697 | } | |
1698 | ||
1699 | /* | |
1700 | * We need look no further than the maximum address a fragment | |
1701 | * could possibly have landed at. Also cast things to loff_t to | |
25985edc | 1702 | * prevent overflows and make comparisons vs. equal-width types. |
1da177e4 | 1703 | */ |
8e36709d | 1704 | size = PAGE_ALIGN(size); |
1da177e4 LT |
1705 | while (vma && (loff_t)(vma->vm_end - addr) <= size) { |
1706 | next = vma->vm_next; | |
1707 | ||
1708 | /* finding a matching vma now does not alter retval */ | |
bc56bba8 | 1709 | if ((vma->vm_ops == &shm_vm_ops) && |
d3c97900 DH |
1710 | ((vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) && |
1711 | (vma->vm_file == file)) | |
897ab3e0 | 1712 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1713 | vma = next; |
1714 | } | |
1715 | ||
63980c80 | 1716 | #else /* CONFIG_MMU */ |
8feae131 | 1717 | /* under NOMMU conditions, the exact address to be destroyed must be |
63980c80 SP |
1718 | * given |
1719 | */ | |
530fcd16 | 1720 | if (vma && vma->vm_start == addr && vma->vm_ops == &shm_vm_ops) { |
897ab3e0 | 1721 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
8feae131 DH |
1722 | retval = 0; |
1723 | } | |
1724 | ||
1725 | #endif | |
1726 | ||
d8ed45c5 | 1727 | mmap_write_unlock(mm); |
1da177e4 LT |
1728 | return retval; |
1729 | } | |
1730 | ||
da1e2744 DB |
1731 | SYSCALL_DEFINE1(shmdt, char __user *, shmaddr) |
1732 | { | |
1733 | return ksys_shmdt(shmaddr); | |
1734 | } | |
1735 | ||
1da177e4 | 1736 | #ifdef CONFIG_PROC_FS |
19b4946c | 1737 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it) |
1da177e4 | 1738 | { |
98f929b1 | 1739 | struct pid_namespace *pid_ns = ipc_seq_pid_ns(s); |
1efdb69b | 1740 | struct user_namespace *user_ns = seq_user_ns(s); |
ade9f91b KC |
1741 | struct kern_ipc_perm *ipcp = it; |
1742 | struct shmid_kernel *shp; | |
b7952180 HD |
1743 | unsigned long rss = 0, swp = 0; |
1744 | ||
ade9f91b | 1745 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
b7952180 | 1746 | shm_add_rss_swap(shp, &rss, &swp); |
1da177e4 | 1747 | |
6c826818 PM |
1748 | #if BITS_PER_LONG <= 32 |
1749 | #define SIZE_SPEC "%10lu" | |
1750 | #else | |
1751 | #define SIZE_SPEC "%21lu" | |
1752 | #endif | |
1da177e4 | 1753 | |
7f032d6e JP |
1754 | seq_printf(s, |
1755 | "%10d %10d %4o " SIZE_SPEC " %5u %5u " | |
7ff2819e | 1756 | "%5lu %5u %5u %5u %5u %10llu %10llu %10llu " |
7f032d6e JP |
1757 | SIZE_SPEC " " SIZE_SPEC "\n", |
1758 | shp->shm_perm.key, | |
1759 | shp->shm_perm.id, | |
1760 | shp->shm_perm.mode, | |
1761 | shp->shm_segsz, | |
98f929b1 EB |
1762 | pid_nr_ns(shp->shm_cprid, pid_ns), |
1763 | pid_nr_ns(shp->shm_lprid, pid_ns), | |
7f032d6e JP |
1764 | shp->shm_nattch, |
1765 | from_kuid_munged(user_ns, shp->shm_perm.uid), | |
1766 | from_kgid_munged(user_ns, shp->shm_perm.gid), | |
1767 | from_kuid_munged(user_ns, shp->shm_perm.cuid), | |
1768 | from_kgid_munged(user_ns, shp->shm_perm.cgid), | |
1769 | shp->shm_atim, | |
1770 | shp->shm_dtim, | |
1771 | shp->shm_ctim, | |
1772 | rss * PAGE_SIZE, | |
1773 | swp * PAGE_SIZE); | |
1774 | ||
1775 | return 0; | |
1da177e4 LT |
1776 | } |
1777 | #endif |