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); | |
7191adff | 184 | security_shm_free(&shp->shm_perm); |
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 | ||
7ff2819e | 206 | shp->shm_atim = ktime_get_real_seconds(); |
b488893a | 207 | shp->shm_lprid = task_tgid_vnr(current); |
1da177e4 LT |
208 | shp->shm_nattch++; |
209 | shm_unlock(shp); | |
1ac0b6de KS |
210 | return 0; |
211 | } | |
212 | ||
213 | /* This is called by fork, once for every shm attach. */ | |
214 | static void shm_open(struct vm_area_struct *vma) | |
215 | { | |
216 | int err = __shm_open(vma); | |
217 | /* | |
218 | * We raced in the idr lookup or with shm_destroy(). | |
219 | * Either way, the ID is busted. | |
220 | */ | |
221 | WARN_ON_ONCE(err); | |
1da177e4 LT |
222 | } |
223 | ||
1da177e4 LT |
224 | /* |
225 | * shm_destroy - free the struct shmid_kernel | |
226 | * | |
f4566f04 | 227 | * @ns: namespace |
1da177e4 LT |
228 | * @shp: struct to free |
229 | * | |
d9a605e4 | 230 | * It has to be called with shp and shm_ids.rwsem (writer) locked, |
1da177e4 LT |
231 | * but returns with shp unlocked and freed. |
232 | */ | |
4e982311 | 233 | static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) |
1da177e4 | 234 | { |
a399b29d GT |
235 | struct file *shm_file; |
236 | ||
237 | shm_file = shp->shm_file; | |
238 | shp->shm_file = NULL; | |
4e982311 | 239 | ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT; |
7ca7e564 | 240 | shm_rmid(ns, shp); |
1da177e4 | 241 | shm_unlock(shp); |
a399b29d GT |
242 | if (!is_file_hugepages(shm_file)) |
243 | shmem_lock(shm_file, 0, shp->mlock_user); | |
353d5c30 | 244 | else if (shp->mlock_user) |
07a46ed2 DH |
245 | user_shm_unlock(i_size_read(file_inode(shm_file)), |
246 | shp->mlock_user); | |
a399b29d | 247 | fput(shm_file); |
dba4cdd3 | 248 | ipc_rcu_putref(&shp->shm_perm, shm_rcu_free); |
1da177e4 LT |
249 | } |
250 | ||
b34a6b1d VK |
251 | /* |
252 | * shm_may_destroy - identifies whether shm segment should be destroyed now | |
253 | * | |
254 | * Returns true if and only if there are no active users of the segment and | |
255 | * one of the following is true: | |
256 | * | |
257 | * 1) shmctl(id, IPC_RMID, NULL) was called for this shp | |
258 | * | |
259 | * 2) sysctl kernel.shm_rmid_forced is set to 1. | |
260 | */ | |
261 | static bool shm_may_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) | |
262 | { | |
263 | return (shp->shm_nattch == 0) && | |
264 | (ns->shm_rmid_forced || | |
265 | (shp->shm_perm.mode & SHM_DEST)); | |
266 | } | |
267 | ||
1da177e4 | 268 | /* |
bc56bba8 | 269 | * remove the attach descriptor vma. |
1da177e4 LT |
270 | * free memory for segment if it is marked destroyed. |
271 | * The descriptor has already been removed from the current->mm->mmap list | |
272 | * and will later be kfree()d. | |
273 | */ | |
bc56bba8 | 274 | static void shm_close(struct vm_area_struct *vma) |
1da177e4 | 275 | { |
239521f3 | 276 | struct file *file = vma->vm_file; |
bc56bba8 | 277 | struct shm_file_data *sfd = shm_file_data(file); |
1da177e4 | 278 | struct shmid_kernel *shp; |
bc56bba8 | 279 | struct ipc_namespace *ns = sfd->ns; |
4e982311 | 280 | |
d9a605e4 | 281 | down_write(&shm_ids(ns).rwsem); |
1da177e4 | 282 | /* remove from the list of attaches of the shm segment */ |
00c2bf85 | 283 | shp = shm_lock(ns, sfd->id); |
1ac0b6de KS |
284 | |
285 | /* | |
286 | * We raced in the idr lookup or with shm_destroy(). | |
287 | * Either way, the ID is busted. | |
288 | */ | |
289 | if (WARN_ON_ONCE(IS_ERR(shp))) | |
290 | goto done; /* no-op */ | |
291 | ||
b488893a | 292 | shp->shm_lprid = task_tgid_vnr(current); |
7ff2819e | 293 | shp->shm_dtim = ktime_get_real_seconds(); |
1da177e4 | 294 | shp->shm_nattch--; |
b34a6b1d VK |
295 | if (shm_may_destroy(ns, shp)) |
296 | shm_destroy(ns, shp); | |
297 | else | |
298 | shm_unlock(shp); | |
1ac0b6de | 299 | done: |
d9a605e4 | 300 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
301 | } |
302 | ||
d9a605e4 | 303 | /* Called with ns->shm_ids(ns).rwsem locked */ |
b34a6b1d VK |
304 | static int shm_try_destroy_orphaned(int id, void *p, void *data) |
305 | { | |
306 | struct ipc_namespace *ns = data; | |
4c677e2e VK |
307 | struct kern_ipc_perm *ipcp = p; |
308 | struct shmid_kernel *shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
b34a6b1d VK |
309 | |
310 | /* | |
311 | * We want to destroy segments without users and with already | |
312 | * exit'ed originating process. | |
4c677e2e | 313 | * |
d9a605e4 | 314 | * As shp->* are changed under rwsem, it's safe to skip shp locking. |
b34a6b1d | 315 | */ |
4c677e2e | 316 | if (shp->shm_creator != NULL) |
b34a6b1d | 317 | return 0; |
b34a6b1d | 318 | |
4c677e2e VK |
319 | if (shm_may_destroy(ns, shp)) { |
320 | shm_lock_by_ptr(shp); | |
4e982311 | 321 | shm_destroy(ns, shp); |
4c677e2e | 322 | } |
b34a6b1d VK |
323 | return 0; |
324 | } | |
325 | ||
326 | void shm_destroy_orphaned(struct ipc_namespace *ns) | |
327 | { | |
d9a605e4 | 328 | down_write(&shm_ids(ns).rwsem); |
33a30ed4 | 329 | if (shm_ids(ns).in_use) |
4c677e2e | 330 | idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_orphaned, ns); |
d9a605e4 | 331 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
332 | } |
333 | ||
83293c0f | 334 | /* Locking assumes this will only be called with task == current */ |
b34a6b1d VK |
335 | void exit_shm(struct task_struct *task) |
336 | { | |
4c677e2e | 337 | struct ipc_namespace *ns = task->nsproxy->ipc_ns; |
ab602f79 | 338 | struct shmid_kernel *shp, *n; |
b34a6b1d | 339 | |
83293c0f JM |
340 | if (list_empty(&task->sysvshm.shm_clist)) |
341 | return; | |
342 | ||
343 | /* | |
344 | * If kernel.shm_rmid_forced is not set then only keep track of | |
345 | * which shmids are orphaned, so that a later set of the sysctl | |
346 | * can clean them up. | |
347 | */ | |
348 | if (!ns->shm_rmid_forced) { | |
349 | down_read(&shm_ids(ns).rwsem); | |
350 | list_for_each_entry(shp, &task->sysvshm.shm_clist, shm_clist) | |
351 | shp->shm_creator = NULL; | |
352 | /* | |
353 | * Only under read lock but we are only called on current | |
354 | * so no entry on the list will be shared. | |
355 | */ | |
356 | list_del(&task->sysvshm.shm_clist); | |
357 | up_read(&shm_ids(ns).rwsem); | |
298507d4 | 358 | return; |
83293c0f | 359 | } |
298507d4 | 360 | |
83293c0f JM |
361 | /* |
362 | * Destroy all already created segments, that were not yet mapped, | |
363 | * and mark any mapped as orphan to cover the sysctl toggling. | |
364 | * Destroy is skipped if shm_may_destroy() returns false. | |
365 | */ | |
d9a605e4 | 366 | down_write(&shm_ids(ns).rwsem); |
83293c0f JM |
367 | list_for_each_entry_safe(shp, n, &task->sysvshm.shm_clist, shm_clist) { |
368 | shp->shm_creator = NULL; | |
369 | ||
370 | if (shm_may_destroy(ns, shp)) { | |
371 | shm_lock_by_ptr(shp); | |
372 | shm_destroy(ns, shp); | |
373 | } | |
374 | } | |
375 | ||
376 | /* Remove the list head from any segments still attached. */ | |
ab602f79 | 377 | list_del(&task->sysvshm.shm_clist); |
d9a605e4 | 378 | up_write(&shm_ids(ns).rwsem); |
1da177e4 LT |
379 | } |
380 | ||
11bac800 | 381 | static int shm_fault(struct vm_fault *vmf) |
bc56bba8 | 382 | { |
11bac800 | 383 | struct file *file = vmf->vma->vm_file; |
bc56bba8 EB |
384 | struct shm_file_data *sfd = shm_file_data(file); |
385 | ||
11bac800 | 386 | return sfd->vm_ops->fault(vmf); |
bc56bba8 EB |
387 | } |
388 | ||
389 | #ifdef CONFIG_NUMA | |
d823e3e7 | 390 | static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new) |
bc56bba8 EB |
391 | { |
392 | struct file *file = vma->vm_file; | |
393 | struct shm_file_data *sfd = shm_file_data(file); | |
394 | int err = 0; | |
63980c80 | 395 | |
bc56bba8 EB |
396 | if (sfd->vm_ops->set_policy) |
397 | err = sfd->vm_ops->set_policy(vma, new); | |
398 | return err; | |
399 | } | |
400 | ||
d823e3e7 AB |
401 | static struct mempolicy *shm_get_policy(struct vm_area_struct *vma, |
402 | unsigned long addr) | |
bc56bba8 EB |
403 | { |
404 | struct file *file = vma->vm_file; | |
405 | struct shm_file_data *sfd = shm_file_data(file); | |
406 | struct mempolicy *pol = NULL; | |
407 | ||
408 | if (sfd->vm_ops->get_policy) | |
409 | pol = sfd->vm_ops->get_policy(vma, addr); | |
52cd3b07 | 410 | else if (vma->vm_policy) |
bc56bba8 | 411 | pol = vma->vm_policy; |
52cd3b07 | 412 | |
bc56bba8 EB |
413 | return pol; |
414 | } | |
415 | #endif | |
416 | ||
239521f3 | 417 | static int shm_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 418 | { |
bc56bba8 | 419 | struct shm_file_data *sfd = shm_file_data(file); |
b0e15190 DH |
420 | int ret; |
421 | ||
1ac0b6de KS |
422 | /* |
423 | * In case of remap_file_pages() emulation, the file can represent | |
424 | * removed IPC ID: propogate shm_lock() error to caller. | |
425 | */ | |
63980c80 | 426 | ret = __shm_open(vma); |
1ac0b6de KS |
427 | if (ret) |
428 | return ret; | |
429 | ||
f74ac015 | 430 | ret = call_mmap(sfd->file, vma); |
1ac0b6de KS |
431 | if (ret) { |
432 | shm_close(vma); | |
bc56bba8 | 433 | return ret; |
1ac0b6de | 434 | } |
bc56bba8 | 435 | sfd->vm_ops = vma->vm_ops; |
2e92a3ba | 436 | #ifdef CONFIG_MMU |
d0edd852 | 437 | WARN_ON(!sfd->vm_ops->fault); |
2e92a3ba | 438 | #endif |
bc56bba8 | 439 | vma->vm_ops = &shm_vm_ops; |
1ac0b6de | 440 | return 0; |
1da177e4 LT |
441 | } |
442 | ||
4e982311 KK |
443 | static int shm_release(struct inode *ino, struct file *file) |
444 | { | |
bc56bba8 | 445 | struct shm_file_data *sfd = shm_file_data(file); |
4e982311 | 446 | |
bc56bba8 EB |
447 | put_ipc_ns(sfd->ns); |
448 | shm_file_data(file) = NULL; | |
449 | kfree(sfd); | |
4e982311 KK |
450 | return 0; |
451 | } | |
452 | ||
02c24a82 | 453 | static int shm_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
516dffdc | 454 | { |
516dffdc | 455 | struct shm_file_data *sfd = shm_file_data(file); |
516dffdc | 456 | |
7ea80859 CH |
457 | if (!sfd->file->f_op->fsync) |
458 | return -EINVAL; | |
0f41074a | 459 | return sfd->file->f_op->fsync(sfd->file, start, end, datasync); |
516dffdc AL |
460 | } |
461 | ||
7d8a4569 WD |
462 | static long shm_fallocate(struct file *file, int mode, loff_t offset, |
463 | loff_t len) | |
464 | { | |
465 | struct shm_file_data *sfd = shm_file_data(file); | |
466 | ||
467 | if (!sfd->file->f_op->fallocate) | |
468 | return -EOPNOTSUPP; | |
469 | return sfd->file->f_op->fallocate(file, mode, offset, len); | |
470 | } | |
471 | ||
bc56bba8 EB |
472 | static unsigned long shm_get_unmapped_area(struct file *file, |
473 | unsigned long addr, unsigned long len, unsigned long pgoff, | |
474 | unsigned long flags) | |
475 | { | |
476 | struct shm_file_data *sfd = shm_file_data(file); | |
63980c80 | 477 | |
c4caa778 AV |
478 | return sfd->file->f_op->get_unmapped_area(sfd->file, addr, len, |
479 | pgoff, flags); | |
bc56bba8 | 480 | } |
bc56bba8 | 481 | |
9a32144e | 482 | static const struct file_operations shm_file_operations = { |
4e982311 | 483 | .mmap = shm_mmap, |
516dffdc | 484 | .fsync = shm_fsync, |
4e982311 | 485 | .release = shm_release, |
ed5e5894 | 486 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 487 | .llseek = noop_llseek, |
7d8a4569 | 488 | .fallocate = shm_fallocate, |
c4caa778 AV |
489 | }; |
490 | ||
c01d5b30 HD |
491 | /* |
492 | * shm_file_operations_huge is now identical to shm_file_operations, | |
493 | * but we keep it distinct for the sake of is_file_shm_hugepages(). | |
494 | */ | |
c4caa778 AV |
495 | static const struct file_operations shm_file_operations_huge = { |
496 | .mmap = shm_mmap, | |
497 | .fsync = shm_fsync, | |
498 | .release = shm_release, | |
bc56bba8 | 499 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 500 | .llseek = noop_llseek, |
7d8a4569 | 501 | .fallocate = shm_fallocate, |
1da177e4 LT |
502 | }; |
503 | ||
2954e440 | 504 | bool is_file_shm_hugepages(struct file *file) |
c4caa778 AV |
505 | { |
506 | return file->f_op == &shm_file_operations_huge; | |
507 | } | |
508 | ||
f0f37e2f | 509 | static const struct vm_operations_struct shm_vm_ops = { |
1da177e4 LT |
510 | .open = shm_open, /* callback for a new vm-area open */ |
511 | .close = shm_close, /* callback for when the vm-area is released */ | |
54cb8821 | 512 | .fault = shm_fault, |
bc56bba8 EB |
513 | #if defined(CONFIG_NUMA) |
514 | .set_policy = shm_set_policy, | |
515 | .get_policy = shm_get_policy, | |
1da177e4 LT |
516 | #endif |
517 | }; | |
518 | ||
f4566f04 ND |
519 | /** |
520 | * newseg - Create a new shared memory segment | |
521 | * @ns: namespace | |
522 | * @params: ptr to the structure that contains key, size and shmflg | |
523 | * | |
d9a605e4 | 524 | * Called with shm_ids.rwsem held as a writer. |
f4566f04 | 525 | */ |
7748dbfa | 526 | static int newseg(struct ipc_namespace *ns, struct ipc_params *params) |
1da177e4 | 527 | { |
7748dbfa ND |
528 | key_t key = params->key; |
529 | int shmflg = params->flg; | |
530 | size_t size = params->u.size; | |
1da177e4 LT |
531 | int error; |
532 | struct shmid_kernel *shp; | |
d69f3bad | 533 | size_t numpages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; |
239521f3 | 534 | struct file *file; |
1da177e4 | 535 | char name[13]; |
ca16d140 | 536 | vm_flags_t acctflag = 0; |
1da177e4 | 537 | |
4e982311 | 538 | if (size < SHMMIN || size > ns->shm_ctlmax) |
1da177e4 LT |
539 | return -EINVAL; |
540 | ||
1376327c MS |
541 | if (numpages << PAGE_SHIFT < size) |
542 | return -ENOSPC; | |
543 | ||
09c6eb1f MS |
544 | if (ns->shm_tot + numpages < ns->shm_tot || |
545 | ns->shm_tot + numpages > ns->shm_ctlall) | |
1da177e4 LT |
546 | return -ENOSPC; |
547 | ||
42e618f7 KC |
548 | shp = kvmalloc(sizeof(*shp), GFP_KERNEL); |
549 | if (unlikely(!shp)) | |
1da177e4 LT |
550 | return -ENOMEM; |
551 | ||
552 | shp->shm_perm.key = key; | |
b33291c0 | 553 | shp->shm_perm.mode = (shmflg & S_IRWXUGO); |
1da177e4 LT |
554 | shp->mlock_user = NULL; |
555 | ||
556 | shp->shm_perm.security = NULL; | |
7191adff | 557 | error = security_shm_alloc(&shp->shm_perm); |
1da177e4 | 558 | if (error) { |
42e618f7 | 559 | kvfree(shp); |
1da177e4 LT |
560 | return error; |
561 | } | |
562 | ||
239521f3 | 563 | sprintf(name, "SYSV%08x", key); |
1da177e4 | 564 | if (shmflg & SHM_HUGETLB) { |
c103a4dc | 565 | struct hstate *hs; |
091d0d55 LZ |
566 | size_t hugesize; |
567 | ||
c103a4dc | 568 | hs = hstate_sizelog((shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); |
091d0d55 LZ |
569 | if (!hs) { |
570 | error = -EINVAL; | |
571 | goto no_file; | |
572 | } | |
573 | hugesize = ALIGN(size, huge_page_size(hs)); | |
af73e4d9 | 574 | |
5a6fe125 MG |
575 | /* hugetlb_file_setup applies strict accounting */ |
576 | if (shmflg & SHM_NORESERVE) | |
577 | acctflag = VM_NORESERVE; | |
af73e4d9 | 578 | file = hugetlb_file_setup(name, hugesize, acctflag, |
42d7395f AK |
579 | &shp->mlock_user, HUGETLB_SHMFS_INODE, |
580 | (shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); | |
1da177e4 | 581 | } else { |
bf8f972d BP |
582 | /* |
583 | * Do not allow no accounting for OVERCOMMIT_NEVER, even | |
239521f3 | 584 | * if it's asked for. |
bf8f972d BP |
585 | */ |
586 | if ((shmflg & SHM_NORESERVE) && | |
587 | sysctl_overcommit_memory != OVERCOMMIT_NEVER) | |
fc8744ad | 588 | acctflag = VM_NORESERVE; |
e1832f29 | 589 | file = shmem_kernel_file_setup(name, size, acctflag); |
1da177e4 LT |
590 | } |
591 | error = PTR_ERR(file); | |
592 | if (IS_ERR(file)) | |
593 | goto no_file; | |
594 | ||
b488893a | 595 | shp->shm_cprid = task_tgid_vnr(current); |
1da177e4 LT |
596 | shp->shm_lprid = 0; |
597 | shp->shm_atim = shp->shm_dtim = 0; | |
7ff2819e | 598 | shp->shm_ctim = ktime_get_real_seconds(); |
1da177e4 LT |
599 | shp->shm_segsz = size; |
600 | shp->shm_nattch = 0; | |
1da177e4 | 601 | shp->shm_file = file; |
5774ed01 | 602 | shp->shm_creator = current; |
b9a53227 | 603 | |
39c96a1b | 604 | /* ipc_addid() locks shp upon success. */ |
a2642f87 MS |
605 | error = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni); |
606 | if (error < 0) | |
b9a53227 | 607 | goto no_id; |
b9a53227 | 608 | |
ab602f79 | 609 | list_add(&shp->shm_clist, ¤t->sysvshm.shm_clist); |
dbfcd91f | 610 | |
30475cc1 BP |
611 | /* |
612 | * shmid gets reported as "inode#" in /proc/pid/maps. | |
613 | * proc-ps tools use this. Changing this will break them. | |
614 | */ | |
496ad9aa | 615 | file_inode(file)->i_ino = shp->shm_perm.id; |
551110a9 | 616 | |
4e982311 | 617 | ns->shm_tot += numpages; |
7ca7e564 | 618 | error = shp->shm_perm.id; |
dbfcd91f | 619 | |
cf9d5d78 | 620 | ipc_unlock_object(&shp->shm_perm); |
dbfcd91f | 621 | rcu_read_unlock(); |
7ca7e564 | 622 | return error; |
1da177e4 LT |
623 | |
624 | no_id: | |
2195d281 | 625 | if (is_file_hugepages(file) && shp->mlock_user) |
353d5c30 | 626 | user_shm_unlock(size, shp->mlock_user); |
1da177e4 LT |
627 | fput(file); |
628 | no_file: | |
a2642f87 | 629 | call_rcu(&shp->shm_perm.rcu, shm_rcu_free); |
1da177e4 LT |
630 | return error; |
631 | } | |
632 | ||
f4566f04 | 633 | /* |
d9a605e4 | 634 | * Called with shm_ids.rwsem and ipcp locked. |
f4566f04 | 635 | */ |
03f02c76 | 636 | static inline int shm_security(struct kern_ipc_perm *ipcp, int shmflg) |
7748dbfa | 637 | { |
7191adff | 638 | return security_shm_associate(ipcp, shmflg); |
7748dbfa ND |
639 | } |
640 | ||
f4566f04 | 641 | /* |
d9a605e4 | 642 | * Called with shm_ids.rwsem and ipcp locked. |
f4566f04 | 643 | */ |
03f02c76 ND |
644 | static inline int shm_more_checks(struct kern_ipc_perm *ipcp, |
645 | struct ipc_params *params) | |
7748dbfa | 646 | { |
03f02c76 ND |
647 | struct shmid_kernel *shp; |
648 | ||
649 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
650 | if (shp->shm_segsz < params->u.size) | |
7748dbfa ND |
651 | return -EINVAL; |
652 | ||
653 | return 0; | |
654 | } | |
655 | ||
d5460c99 | 656 | SYSCALL_DEFINE3(shmget, key_t, key, size_t, size, int, shmflg) |
1da177e4 | 657 | { |
4e982311 | 658 | struct ipc_namespace *ns; |
eb66ec44 MK |
659 | static const struct ipc_ops shm_ops = { |
660 | .getnew = newseg, | |
661 | .associate = shm_security, | |
662 | .more_checks = shm_more_checks, | |
663 | }; | |
7748dbfa | 664 | struct ipc_params shm_params; |
4e982311 KK |
665 | |
666 | ns = current->nsproxy->ipc_ns; | |
1da177e4 | 667 | |
7748dbfa ND |
668 | shm_params.key = key; |
669 | shm_params.flg = shmflg; | |
670 | shm_params.u.size = size; | |
1da177e4 | 671 | |
7748dbfa | 672 | return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params); |
1da177e4 LT |
673 | } |
674 | ||
675 | static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version) | |
676 | { | |
239521f3 | 677 | switch (version) { |
1da177e4 LT |
678 | case IPC_64: |
679 | return copy_to_user(buf, in, sizeof(*in)); | |
680 | case IPC_OLD: | |
681 | { | |
682 | struct shmid_ds out; | |
683 | ||
3af54c9b | 684 | memset(&out, 0, sizeof(out)); |
1da177e4 LT |
685 | ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm); |
686 | out.shm_segsz = in->shm_segsz; | |
687 | out.shm_atime = in->shm_atime; | |
688 | out.shm_dtime = in->shm_dtime; | |
689 | out.shm_ctime = in->shm_ctime; | |
690 | out.shm_cpid = in->shm_cpid; | |
691 | out.shm_lpid = in->shm_lpid; | |
692 | out.shm_nattch = in->shm_nattch; | |
693 | ||
694 | return copy_to_user(buf, &out, sizeof(out)); | |
695 | } | |
696 | default: | |
697 | return -EINVAL; | |
698 | } | |
699 | } | |
700 | ||
016d7132 PP |
701 | static inline unsigned long |
702 | copy_shmid_from_user(struct shmid64_ds *out, void __user *buf, int version) | |
1da177e4 | 703 | { |
239521f3 | 704 | switch (version) { |
1da177e4 | 705 | case IPC_64: |
016d7132 | 706 | if (copy_from_user(out, buf, sizeof(*out))) |
1da177e4 | 707 | return -EFAULT; |
1da177e4 | 708 | return 0; |
1da177e4 LT |
709 | case IPC_OLD: |
710 | { | |
711 | struct shmid_ds tbuf_old; | |
712 | ||
713 | if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) | |
714 | return -EFAULT; | |
715 | ||
016d7132 PP |
716 | out->shm_perm.uid = tbuf_old.shm_perm.uid; |
717 | out->shm_perm.gid = tbuf_old.shm_perm.gid; | |
718 | out->shm_perm.mode = tbuf_old.shm_perm.mode; | |
1da177e4 LT |
719 | |
720 | return 0; | |
721 | } | |
722 | default: | |
723 | return -EINVAL; | |
724 | } | |
725 | } | |
726 | ||
727 | static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version) | |
728 | { | |
239521f3 | 729 | switch (version) { |
1da177e4 LT |
730 | case IPC_64: |
731 | return copy_to_user(buf, in, sizeof(*in)); | |
732 | case IPC_OLD: | |
733 | { | |
734 | struct shminfo out; | |
735 | ||
239521f3 | 736 | if (in->shmmax > INT_MAX) |
1da177e4 LT |
737 | out.shmmax = INT_MAX; |
738 | else | |
739 | out.shmmax = (int)in->shmmax; | |
740 | ||
741 | out.shmmin = in->shmmin; | |
742 | out.shmmni = in->shmmni; | |
743 | out.shmseg = in->shmseg; | |
46c0a8ca | 744 | out.shmall = in->shmall; |
1da177e4 LT |
745 | |
746 | return copy_to_user(buf, &out, sizeof(out)); | |
747 | } | |
748 | default: | |
749 | return -EINVAL; | |
750 | } | |
751 | } | |
752 | ||
b7952180 HD |
753 | /* |
754 | * Calculate and add used RSS and swap pages of a shm. | |
d9a605e4 | 755 | * Called with shm_ids.rwsem held as a reader |
b7952180 HD |
756 | */ |
757 | static void shm_add_rss_swap(struct shmid_kernel *shp, | |
758 | unsigned long *rss_add, unsigned long *swp_add) | |
759 | { | |
760 | struct inode *inode; | |
761 | ||
496ad9aa | 762 | inode = file_inode(shp->shm_file); |
b7952180 HD |
763 | |
764 | if (is_file_hugepages(shp->shm_file)) { | |
765 | struct address_space *mapping = inode->i_mapping; | |
766 | struct hstate *h = hstate_file(shp->shm_file); | |
767 | *rss_add += pages_per_huge_page(h) * mapping->nrpages; | |
768 | } else { | |
769 | #ifdef CONFIG_SHMEM | |
770 | struct shmem_inode_info *info = SHMEM_I(inode); | |
63980c80 | 771 | |
4595ef88 | 772 | spin_lock_irq(&info->lock); |
b7952180 HD |
773 | *rss_add += inode->i_mapping->nrpages; |
774 | *swp_add += info->swapped; | |
4595ef88 | 775 | spin_unlock_irq(&info->lock); |
b7952180 HD |
776 | #else |
777 | *rss_add += inode->i_mapping->nrpages; | |
778 | #endif | |
779 | } | |
780 | } | |
781 | ||
f4566f04 | 782 | /* |
d9a605e4 | 783 | * Called with shm_ids.rwsem held as a reader |
f4566f04 | 784 | */ |
4e982311 KK |
785 | static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss, |
786 | unsigned long *swp) | |
1da177e4 | 787 | { |
7ca7e564 ND |
788 | int next_id; |
789 | int total, in_use; | |
1da177e4 LT |
790 | |
791 | *rss = 0; | |
792 | *swp = 0; | |
793 | ||
7ca7e564 ND |
794 | in_use = shm_ids(ns).in_use; |
795 | ||
796 | for (total = 0, next_id = 0; total < in_use; next_id++) { | |
e562aebc | 797 | struct kern_ipc_perm *ipc; |
1da177e4 | 798 | struct shmid_kernel *shp; |
1da177e4 | 799 | |
e562aebc TB |
800 | ipc = idr_find(&shm_ids(ns).ipcs_idr, next_id); |
801 | if (ipc == NULL) | |
1da177e4 | 802 | continue; |
e562aebc | 803 | shp = container_of(ipc, struct shmid_kernel, shm_perm); |
1da177e4 | 804 | |
b7952180 | 805 | shm_add_rss_swap(shp, rss, swp); |
7ca7e564 ND |
806 | |
807 | total++; | |
1da177e4 LT |
808 | } |
809 | } | |
810 | ||
8d4cc8b5 | 811 | /* |
d9a605e4 | 812 | * This function handles some shmctl commands which require the rwsem |
8d4cc8b5 | 813 | * to be held in write mode. |
d9a605e4 | 814 | * NOTE: no locks must be held, the rwsem is taken inside this function. |
8d4cc8b5 PP |
815 | */ |
816 | static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd, | |
9ba720c1 | 817 | struct shmid64_ds *shmid64) |
1da177e4 | 818 | { |
8d4cc8b5 | 819 | struct kern_ipc_perm *ipcp; |
8d4cc8b5 PP |
820 | struct shmid_kernel *shp; |
821 | int err; | |
822 | ||
d9a605e4 | 823 | down_write(&shm_ids(ns).rwsem); |
7b4cc5d8 DB |
824 | rcu_read_lock(); |
825 | ||
79ccf0f8 | 826 | ipcp = ipcctl_pre_down_nolock(ns, &shm_ids(ns), shmid, cmd, |
9ba720c1 | 827 | &shmid64->shm_perm, 0); |
7b4cc5d8 DB |
828 | if (IS_ERR(ipcp)) { |
829 | err = PTR_ERR(ipcp); | |
7b4cc5d8 DB |
830 | goto out_unlock1; |
831 | } | |
8d4cc8b5 | 832 | |
a5f75e7f | 833 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
8d4cc8b5 | 834 | |
7191adff | 835 | err = security_shm_shmctl(&shp->shm_perm, cmd); |
8d4cc8b5 | 836 | if (err) |
79ccf0f8 | 837 | goto out_unlock1; |
7b4cc5d8 | 838 | |
8d4cc8b5 PP |
839 | switch (cmd) { |
840 | case IPC_RMID: | |
79ccf0f8 | 841 | ipc_lock_object(&shp->shm_perm); |
7b4cc5d8 | 842 | /* do_shm_rmid unlocks the ipc object and rcu */ |
8d4cc8b5 PP |
843 | do_shm_rmid(ns, ipcp); |
844 | goto out_up; | |
845 | case IPC_SET: | |
79ccf0f8 | 846 | ipc_lock_object(&shp->shm_perm); |
9ba720c1 | 847 | err = ipc_update_perm(&shmid64->shm_perm, ipcp); |
1efdb69b | 848 | if (err) |
7b4cc5d8 | 849 | goto out_unlock0; |
7ff2819e | 850 | shp->shm_ctim = ktime_get_real_seconds(); |
8d4cc8b5 PP |
851 | break; |
852 | default: | |
853 | err = -EINVAL; | |
79ccf0f8 | 854 | goto out_unlock1; |
8d4cc8b5 | 855 | } |
7b4cc5d8 DB |
856 | |
857 | out_unlock0: | |
858 | ipc_unlock_object(&shp->shm_perm); | |
859 | out_unlock1: | |
860 | rcu_read_unlock(); | |
8d4cc8b5 | 861 | out_up: |
d9a605e4 | 862 | up_write(&shm_ids(ns).rwsem); |
8d4cc8b5 PP |
863 | return err; |
864 | } | |
865 | ||
9ba720c1 AV |
866 | static int shmctl_ipc_info(struct ipc_namespace *ns, |
867 | struct shminfo64 *shminfo) | |
8d4cc8b5 | 868 | { |
9ba720c1 AV |
869 | int err = security_shm_shmctl(NULL, IPC_INFO); |
870 | if (!err) { | |
871 | memset(shminfo, 0, sizeof(*shminfo)); | |
872 | shminfo->shmmni = shminfo->shmseg = ns->shm_ctlmni; | |
873 | shminfo->shmmax = ns->shm_ctlmax; | |
874 | shminfo->shmall = ns->shm_ctlall; | |
875 | shminfo->shmmin = SHMMIN; | |
d9a605e4 | 876 | down_read(&shm_ids(ns).rwsem); |
7ca7e564 | 877 | err = ipc_get_maxid(&shm_ids(ns)); |
d9a605e4 | 878 | up_read(&shm_ids(ns).rwsem); |
239521f3 | 879 | if (err < 0) |
1da177e4 | 880 | err = 0; |
1da177e4 | 881 | } |
9ba720c1 AV |
882 | return err; |
883 | } | |
1da177e4 | 884 | |
9ba720c1 AV |
885 | static int shmctl_shm_info(struct ipc_namespace *ns, |
886 | struct shm_info *shm_info) | |
887 | { | |
888 | int err = security_shm_shmctl(NULL, SHM_INFO); | |
889 | if (!err) { | |
890 | memset(shm_info, 0, sizeof(*shm_info)); | |
d9a605e4 | 891 | down_read(&shm_ids(ns).rwsem); |
9ba720c1 AV |
892 | shm_info->used_ids = shm_ids(ns).in_use; |
893 | shm_get_stat(ns, &shm_info->shm_rss, &shm_info->shm_swp); | |
894 | shm_info->shm_tot = ns->shm_tot; | |
895 | shm_info->swap_attempts = 0; | |
896 | shm_info->swap_successes = 0; | |
7ca7e564 | 897 | err = ipc_get_maxid(&shm_ids(ns)); |
d9a605e4 | 898 | up_read(&shm_ids(ns).rwsem); |
9ba720c1 AV |
899 | if (err < 0) |
900 | err = 0; | |
1da177e4 | 901 | } |
9ba720c1 AV |
902 | return err; |
903 | } | |
c97cb9cc | 904 | |
9ba720c1 AV |
905 | static int shmctl_stat(struct ipc_namespace *ns, int shmid, |
906 | int cmd, struct shmid64_ds *tbuf) | |
907 | { | |
908 | struct shmid_kernel *shp; | |
87ad4b0d | 909 | int id = 0; |
9ba720c1 | 910 | int err; |
c97cb9cc | 911 | |
87ad4b0d PM |
912 | memset(tbuf, 0, sizeof(*tbuf)); |
913 | ||
9ba720c1 AV |
914 | rcu_read_lock(); |
915 | if (cmd == SHM_STAT) { | |
916 | shp = shm_obtain_object(ns, shmid); | |
917 | if (IS_ERR(shp)) { | |
918 | err = PTR_ERR(shp); | |
919 | goto out_unlock; | |
920 | } | |
87ad4b0d | 921 | id = shp->shm_perm.id; |
9ba720c1 AV |
922 | } else { |
923 | shp = shm_obtain_object_check(ns, shmid); | |
924 | if (IS_ERR(shp)) { | |
925 | err = PTR_ERR(shp); | |
1da177e4 | 926 | goto out_unlock; |
1da177e4 | 927 | } |
9ba720c1 | 928 | } |
1da177e4 | 929 | |
9ba720c1 AV |
930 | err = -EACCES; |
931 | if (ipcperms(ns, &shp->shm_perm, S_IRUGO)) | |
932 | goto out_unlock; | |
c97cb9cc | 933 | |
7191adff | 934 | err = security_shm_shmctl(&shp->shm_perm, cmd); |
9ba720c1 AV |
935 | if (err) |
936 | goto out_unlock; | |
937 | ||
87ad4b0d PM |
938 | ipc_lock_object(&shp->shm_perm); |
939 | ||
940 | if (!ipc_valid_object(&shp->shm_perm)) { | |
941 | ipc_unlock_object(&shp->shm_perm); | |
942 | err = -EIDRM; | |
943 | goto out_unlock; | |
944 | } | |
945 | ||
9ba720c1 AV |
946 | kernel_to_ipc64_perm(&shp->shm_perm, &tbuf->shm_perm); |
947 | tbuf->shm_segsz = shp->shm_segsz; | |
948 | tbuf->shm_atime = shp->shm_atim; | |
949 | tbuf->shm_dtime = shp->shm_dtim; | |
950 | tbuf->shm_ctime = shp->shm_ctim; | |
951 | tbuf->shm_cpid = shp->shm_cprid; | |
952 | tbuf->shm_lpid = shp->shm_lprid; | |
953 | tbuf->shm_nattch = shp->shm_nattch; | |
87ad4b0d PM |
954 | |
955 | ipc_unlock_object(&shp->shm_perm); | |
9ba720c1 | 956 | rcu_read_unlock(); |
87ad4b0d | 957 | return id; |
68eccc1d DB |
958 | |
959 | out_unlock: | |
c97cb9cc | 960 | rcu_read_unlock(); |
68eccc1d DB |
961 | return err; |
962 | } | |
963 | ||
9ba720c1 | 964 | static int shmctl_do_lock(struct ipc_namespace *ns, int shmid, int cmd) |
68eccc1d DB |
965 | { |
966 | struct shmid_kernel *shp; | |
9ba720c1 AV |
967 | struct file *shm_file; |
968 | int err; | |
68eccc1d | 969 | |
9ba720c1 AV |
970 | rcu_read_lock(); |
971 | shp = shm_obtain_object_check(ns, shmid); | |
972 | if (IS_ERR(shp)) { | |
973 | err = PTR_ERR(shp); | |
974 | goto out_unlock1; | |
1da177e4 | 975 | } |
c97cb9cc | 976 | |
9ba720c1 | 977 | audit_ipc_obj(&(shp->shm_perm)); |
7191adff | 978 | err = security_shm_shmctl(&shp->shm_perm, cmd); |
9ba720c1 AV |
979 | if (err) |
980 | goto out_unlock1; | |
c97cb9cc | 981 | |
9ba720c1 | 982 | ipc_lock_object(&shp->shm_perm); |
c97cb9cc | 983 | |
9ba720c1 AV |
984 | /* check if shm_destroy() is tearing down shp */ |
985 | if (!ipc_valid_object(&shp->shm_perm)) { | |
986 | err = -EIDRM; | |
987 | goto out_unlock0; | |
1da177e4 | 988 | } |
073115d6 | 989 | |
9ba720c1 AV |
990 | if (!ns_capable(ns->user_ns, CAP_IPC_LOCK)) { |
991 | kuid_t euid = current_euid(); | |
0f3d2b01 | 992 | |
9ba720c1 AV |
993 | if (!uid_eq(euid, shp->shm_perm.uid) && |
994 | !uid_eq(euid, shp->shm_perm.cuid)) { | |
995 | err = -EPERM; | |
0f3d2b01 RA |
996 | goto out_unlock0; |
997 | } | |
9ba720c1 AV |
998 | if (cmd == SHM_LOCK && !rlimit(RLIMIT_MEMLOCK)) { |
999 | err = -EPERM; | |
1000 | goto out_unlock0; | |
1da177e4 | 1001 | } |
68eccc1d DB |
1002 | } |
1003 | ||
9ba720c1 AV |
1004 | shm_file = shp->shm_file; |
1005 | if (is_file_hugepages(shm_file)) | |
1006 | goto out_unlock0; | |
85046579 | 1007 | |
9ba720c1 AV |
1008 | if (cmd == SHM_LOCK) { |
1009 | struct user_struct *user = current_user(); | |
63980c80 | 1010 | |
9ba720c1 AV |
1011 | err = shmem_lock(shm_file, 1, user); |
1012 | if (!err && !(shp->shm_perm.mode & SHM_LOCKED)) { | |
1013 | shp->shm_perm.mode |= SHM_LOCKED; | |
1014 | shp->mlock_user = user; | |
1da177e4 | 1015 | } |
9ba720c1 | 1016 | goto out_unlock0; |
1da177e4 LT |
1017 | } |
1018 | ||
9ba720c1 AV |
1019 | /* SHM_UNLOCK */ |
1020 | if (!(shp->shm_perm.mode & SHM_LOCKED)) | |
1021 | goto out_unlock0; | |
1022 | shmem_lock(shm_file, 0, shp->mlock_user); | |
1023 | shp->shm_perm.mode &= ~SHM_LOCKED; | |
1024 | shp->mlock_user = NULL; | |
1025 | get_file(shm_file); | |
1026 | ipc_unlock_object(&shp->shm_perm); | |
1027 | rcu_read_unlock(); | |
1028 | shmem_unlock_mapping(shm_file->f_mapping); | |
1029 | ||
1030 | fput(shm_file); | |
1031 | return err; | |
1032 | ||
2caacaa8 DB |
1033 | out_unlock0: |
1034 | ipc_unlock_object(&shp->shm_perm); | |
1035 | out_unlock1: | |
c97cb9cc | 1036 | rcu_read_unlock(); |
68eccc1d DB |
1037 | return err; |
1038 | } | |
1039 | ||
1040 | SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, struct shmid_ds __user *, buf) | |
1041 | { | |
68eccc1d DB |
1042 | int err, version; |
1043 | struct ipc_namespace *ns; | |
553f770e | 1044 | struct shmid64_ds sem64; |
68eccc1d | 1045 | |
2caacaa8 DB |
1046 | if (cmd < 0 || shmid < 0) |
1047 | return -EINVAL; | |
68eccc1d DB |
1048 | |
1049 | version = ipc_parse_version(&cmd); | |
1050 | ns = current->nsproxy->ipc_ns; | |
1051 | ||
1052 | switch (cmd) { | |
9ba720c1 AV |
1053 | case IPC_INFO: { |
1054 | struct shminfo64 shminfo; | |
1055 | err = shmctl_ipc_info(ns, &shminfo); | |
1056 | if (err < 0) | |
1057 | return err; | |
1058 | if (copy_shminfo_to_user(buf, &shminfo, version)) | |
1059 | err = -EFAULT; | |
1060 | return err; | |
1061 | } | |
1062 | case SHM_INFO: { | |
1063 | struct shm_info shm_info; | |
1064 | err = shmctl_shm_info(ns, &shm_info); | |
1065 | if (err < 0) | |
1066 | return err; | |
1067 | if (copy_to_user(buf, &shm_info, sizeof(shm_info))) | |
1068 | err = -EFAULT; | |
1069 | return err; | |
1070 | } | |
68eccc1d | 1071 | case SHM_STAT: |
9ba720c1 | 1072 | case IPC_STAT: { |
553f770e | 1073 | err = shmctl_stat(ns, shmid, cmd, &sem64); |
9ba720c1 AV |
1074 | if (err < 0) |
1075 | return err; | |
553f770e | 1076 | if (copy_shmid_to_user(buf, &sem64, version)) |
9ba720c1 AV |
1077 | err = -EFAULT; |
1078 | return err; | |
1079 | } | |
2caacaa8 | 1080 | case IPC_SET: |
553f770e | 1081 | if (copy_shmid_from_user(&sem64, buf, version)) |
9ba720c1 | 1082 | return -EFAULT; |
553f770e | 1083 | /* fallthru */ |
9ba720c1 | 1084 | case IPC_RMID: |
553f770e | 1085 | return shmctl_down(ns, shmid, cmd, &sem64); |
1da177e4 LT |
1086 | case SHM_LOCK: |
1087 | case SHM_UNLOCK: | |
9ba720c1 AV |
1088 | return shmctl_do_lock(ns, shmid, cmd); |
1089 | default: | |
1090 | return -EINVAL; | |
1091 | } | |
1092 | } | |
89e004ea | 1093 | |
553f770e AV |
1094 | #ifdef CONFIG_COMPAT |
1095 | ||
1096 | struct compat_shmid_ds { | |
1097 | struct compat_ipc_perm shm_perm; | |
1098 | int shm_segsz; | |
1099 | compat_time_t shm_atime; | |
1100 | compat_time_t shm_dtime; | |
1101 | compat_time_t shm_ctime; | |
1102 | compat_ipc_pid_t shm_cpid; | |
1103 | compat_ipc_pid_t shm_lpid; | |
1104 | unsigned short shm_nattch; | |
1105 | unsigned short shm_unused; | |
1106 | compat_uptr_t shm_unused2; | |
1107 | compat_uptr_t shm_unused3; | |
1108 | }; | |
1da177e4 | 1109 | |
553f770e AV |
1110 | struct compat_shminfo64 { |
1111 | compat_ulong_t shmmax; | |
1112 | compat_ulong_t shmmin; | |
1113 | compat_ulong_t shmmni; | |
1114 | compat_ulong_t shmseg; | |
1115 | compat_ulong_t shmall; | |
1116 | compat_ulong_t __unused1; | |
1117 | compat_ulong_t __unused2; | |
1118 | compat_ulong_t __unused3; | |
1119 | compat_ulong_t __unused4; | |
1120 | }; | |
073115d6 | 1121 | |
553f770e AV |
1122 | struct compat_shm_info { |
1123 | compat_int_t used_ids; | |
1124 | compat_ulong_t shm_tot, shm_rss, shm_swp; | |
1125 | compat_ulong_t swap_attempts, swap_successes; | |
1126 | }; | |
0f3d2b01 | 1127 | |
553f770e AV |
1128 | static int copy_compat_shminfo_to_user(void __user *buf, struct shminfo64 *in, |
1129 | int version) | |
1130 | { | |
1131 | if (in->shmmax > INT_MAX) | |
1132 | in->shmmax = INT_MAX; | |
1133 | if (version == IPC_64) { | |
1134 | struct compat_shminfo64 info; | |
1135 | memset(&info, 0, sizeof(info)); | |
1136 | info.shmmax = in->shmmax; | |
1137 | info.shmmin = in->shmmin; | |
1138 | info.shmmni = in->shmmni; | |
1139 | info.shmseg = in->shmseg; | |
1140 | info.shmall = in->shmall; | |
1141 | return copy_to_user(buf, &info, sizeof(info)); | |
1142 | } else { | |
1143 | struct shminfo info; | |
1144 | memset(&info, 0, sizeof(info)); | |
1145 | info.shmmax = in->shmmax; | |
1146 | info.shmmin = in->shmmin; | |
1147 | info.shmmni = in->shmmni; | |
1148 | info.shmseg = in->shmseg; | |
1149 | info.shmall = in->shmall; | |
1150 | return copy_to_user(buf, &info, sizeof(info)); | |
1151 | } | |
1152 | } | |
0f3d2b01 | 1153 | |
553f770e AV |
1154 | static int put_compat_shm_info(struct shm_info *ip, |
1155 | struct compat_shm_info __user *uip) | |
1156 | { | |
1157 | struct compat_shm_info info; | |
1158 | ||
1159 | memset(&info, 0, sizeof(info)); | |
1160 | info.used_ids = ip->used_ids; | |
1161 | info.shm_tot = ip->shm_tot; | |
1162 | info.shm_rss = ip->shm_rss; | |
1163 | info.shm_swp = ip->shm_swp; | |
1164 | info.swap_attempts = ip->swap_attempts; | |
1165 | info.swap_successes = ip->swap_successes; | |
b776e4b1 | 1166 | return copy_to_user(uip, &info, sizeof(info)); |
553f770e | 1167 | } |
1da177e4 | 1168 | |
553f770e AV |
1169 | static int copy_compat_shmid_to_user(void __user *buf, struct shmid64_ds *in, |
1170 | int version) | |
1171 | { | |
1172 | if (version == IPC_64) { | |
1173 | struct compat_shmid64_ds v; | |
1174 | memset(&v, 0, sizeof(v)); | |
28327fae | 1175 | to_compat_ipc64_perm(&v.shm_perm, &in->shm_perm); |
553f770e AV |
1176 | v.shm_atime = in->shm_atime; |
1177 | v.shm_dtime = in->shm_dtime; | |
1178 | v.shm_ctime = in->shm_ctime; | |
1179 | v.shm_segsz = in->shm_segsz; | |
1180 | v.shm_nattch = in->shm_nattch; | |
1181 | v.shm_cpid = in->shm_cpid; | |
1182 | v.shm_lpid = in->shm_lpid; | |
1183 | return copy_to_user(buf, &v, sizeof(v)); | |
1184 | } else { | |
1185 | struct compat_shmid_ds v; | |
1186 | memset(&v, 0, sizeof(v)); | |
28327fae | 1187 | to_compat_ipc_perm(&v.shm_perm, &in->shm_perm); |
553f770e | 1188 | v.shm_perm.key = in->shm_perm.key; |
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 | } | |
1198 | } | |
85046579 | 1199 | |
553f770e AV |
1200 | static int copy_compat_shmid_from_user(struct shmid64_ds *out, void __user *buf, |
1201 | int version) | |
1202 | { | |
1203 | memset(out, 0, sizeof(*out)); | |
1204 | if (version == IPC_64) { | |
6aa211e8 | 1205 | struct compat_shmid64_ds __user *p = buf; |
28327fae | 1206 | return get_compat_ipc64_perm(&out->shm_perm, &p->shm_perm); |
553f770e | 1207 | } else { |
6aa211e8 | 1208 | struct compat_shmid_ds __user *p = buf; |
28327fae | 1209 | return get_compat_ipc_perm(&out->shm_perm, &p->shm_perm); |
553f770e | 1210 | } |
553f770e | 1211 | } |
63980c80 | 1212 | |
553f770e AV |
1213 | COMPAT_SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, void __user *, uptr) |
1214 | { | |
1215 | struct ipc_namespace *ns; | |
1216 | struct shmid64_ds sem64; | |
1217 | int version = compat_ipc_parse_version(&cmd); | |
1218 | int err; | |
85046579 | 1219 | |
553f770e AV |
1220 | ns = current->nsproxy->ipc_ns; |
1221 | ||
1222 | if (cmd < 0 || shmid < 0) | |
1223 | return -EINVAL; | |
2caacaa8 | 1224 | |
553f770e AV |
1225 | switch (cmd) { |
1226 | case IPC_INFO: { | |
1227 | struct shminfo64 shminfo; | |
1228 | err = shmctl_ipc_info(ns, &shminfo); | |
1229 | if (err < 0) | |
1230 | return err; | |
1231 | if (copy_compat_shminfo_to_user(uptr, &shminfo, version)) | |
1232 | err = -EFAULT; | |
1233 | return err; | |
1234 | } | |
1235 | case SHM_INFO: { | |
1236 | struct shm_info shm_info; | |
1237 | err = shmctl_shm_info(ns, &shm_info); | |
1238 | if (err < 0) | |
1239 | return err; | |
1240 | if (put_compat_shm_info(&shm_info, uptr)) | |
1241 | err = -EFAULT; | |
8d4cc8b5 | 1242 | return err; |
2caacaa8 | 1243 | } |
553f770e AV |
1244 | case IPC_STAT: |
1245 | case SHM_STAT: | |
1246 | err = shmctl_stat(ns, shmid, cmd, &sem64); | |
1247 | if (err < 0) | |
1248 | return err; | |
58aff0af | 1249 | if (copy_compat_shmid_to_user(uptr, &sem64, version)) |
553f770e AV |
1250 | err = -EFAULT; |
1251 | return err; | |
1252 | ||
1253 | case IPC_SET: | |
1254 | if (copy_compat_shmid_from_user(&sem64, uptr, version)) | |
1255 | return -EFAULT; | |
1256 | /* fallthru */ | |
1257 | case IPC_RMID: | |
1258 | return shmctl_down(ns, shmid, cmd, &sem64); | |
1259 | case SHM_LOCK: | |
1260 | case SHM_UNLOCK: | |
1261 | return shmctl_do_lock(ns, shmid, cmd); | |
1262 | break; | |
1da177e4 | 1263 | default: |
8d4cc8b5 | 1264 | return -EINVAL; |
1da177e4 | 1265 | } |
1da177e4 LT |
1266 | return err; |
1267 | } | |
553f770e | 1268 | #endif |
1da177e4 LT |
1269 | |
1270 | /* | |
1271 | * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists. | |
1272 | * | |
1273 | * NOTE! Despite the name, this is NOT a direct system call entrypoint. The | |
1274 | * "raddr" thing points to kernel space, and there has to be a wrapper around | |
1275 | * this. | |
1276 | */ | |
95e91b83 DB |
1277 | long do_shmat(int shmid, char __user *shmaddr, int shmflg, |
1278 | ulong *raddr, unsigned long shmlba) | |
1da177e4 LT |
1279 | { |
1280 | struct shmid_kernel *shp; | |
f0cb8802 | 1281 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1282 | unsigned long size; |
239521f3 | 1283 | struct file *file; |
1da177e4 | 1284 | int err; |
f0cb8802 | 1285 | unsigned long flags = MAP_SHARED; |
1da177e4 | 1286 | unsigned long prot; |
1da177e4 | 1287 | int acc_mode; |
4e982311 | 1288 | struct ipc_namespace *ns; |
bc56bba8 EB |
1289 | struct shm_file_data *sfd; |
1290 | struct path path; | |
aeb5d727 | 1291 | fmode_t f_mode; |
41badc15 | 1292 | unsigned long populate = 0; |
1da177e4 | 1293 | |
bc56bba8 EB |
1294 | err = -EINVAL; |
1295 | if (shmid < 0) | |
1da177e4 | 1296 | goto out; |
f0cb8802 DB |
1297 | |
1298 | if (addr) { | |
079a96ae | 1299 | if (addr & (shmlba - 1)) { |
95e91b83 DB |
1300 | /* |
1301 | * Round down to the nearest multiple of shmlba. | |
1302 | * For sane do_mmap_pgoff() parameters, avoid | |
1303 | * round downs that trigger nil-page and MAP_FIXED. | |
1304 | */ | |
1305 | if ((shmflg & SHM_RND) && addr >= shmlba) | |
1306 | addr &= ~(shmlba - 1); | |
1da177e4 LT |
1307 | else |
1308 | #ifndef __ARCH_FORCE_SHMLBA | |
1309 | if (addr & ~PAGE_MASK) | |
1310 | #endif | |
bc56bba8 | 1311 | goto out; |
1da177e4 | 1312 | } |
1da177e4 | 1313 | |
f0cb8802 DB |
1314 | flags |= MAP_FIXED; |
1315 | } else if ((shmflg & SHM_REMAP)) | |
1316 | goto out; | |
1da177e4 LT |
1317 | |
1318 | if (shmflg & SHM_RDONLY) { | |
1319 | prot = PROT_READ; | |
1da177e4 | 1320 | acc_mode = S_IRUGO; |
bc56bba8 | 1321 | f_mode = FMODE_READ; |
1da177e4 LT |
1322 | } else { |
1323 | prot = PROT_READ | PROT_WRITE; | |
1da177e4 | 1324 | acc_mode = S_IRUGO | S_IWUGO; |
bc56bba8 | 1325 | f_mode = FMODE_READ | FMODE_WRITE; |
1da177e4 LT |
1326 | } |
1327 | if (shmflg & SHM_EXEC) { | |
1328 | prot |= PROT_EXEC; | |
1329 | acc_mode |= S_IXUGO; | |
1330 | } | |
1331 | ||
1332 | /* | |
1333 | * We cannot rely on the fs check since SYSV IPC does have an | |
1334 | * additional creator id... | |
1335 | */ | |
4e982311 | 1336 | ns = current->nsproxy->ipc_ns; |
c2c737a0 DB |
1337 | rcu_read_lock(); |
1338 | shp = shm_obtain_object_check(ns, shmid); | |
023a5355 ND |
1339 | if (IS_ERR(shp)) { |
1340 | err = PTR_ERR(shp); | |
c2c737a0 | 1341 | goto out_unlock; |
023a5355 | 1342 | } |
bc56bba8 EB |
1343 | |
1344 | err = -EACCES; | |
b0e77598 | 1345 | if (ipcperms(ns, &shp->shm_perm, acc_mode)) |
bc56bba8 | 1346 | goto out_unlock; |
1da177e4 | 1347 | |
7191adff | 1348 | err = security_shm_shmat(&shp->shm_perm, shmaddr, shmflg); |
bc56bba8 EB |
1349 | if (err) |
1350 | goto out_unlock; | |
1351 | ||
c2c737a0 | 1352 | ipc_lock_object(&shp->shm_perm); |
a399b29d GT |
1353 | |
1354 | /* check if shm_destroy() is tearing down shp */ | |
0f3d2b01 | 1355 | if (!ipc_valid_object(&shp->shm_perm)) { |
a399b29d GT |
1356 | ipc_unlock_object(&shp->shm_perm); |
1357 | err = -EIDRM; | |
1358 | goto out_unlock; | |
1359 | } | |
1360 | ||
2c48b9c4 AV |
1361 | path = shp->shm_file->f_path; |
1362 | path_get(&path); | |
1da177e4 | 1363 | shp->shm_nattch++; |
75c3cfa8 | 1364 | size = i_size_read(d_inode(path.dentry)); |
c2c737a0 DB |
1365 | ipc_unlock_object(&shp->shm_perm); |
1366 | rcu_read_unlock(); | |
1da177e4 | 1367 | |
bc56bba8 EB |
1368 | err = -ENOMEM; |
1369 | sfd = kzalloc(sizeof(*sfd), GFP_KERNEL); | |
f42569b1 DB |
1370 | if (!sfd) { |
1371 | path_put(&path); | |
1372 | goto out_nattch; | |
1373 | } | |
bc56bba8 | 1374 | |
2c48b9c4 AV |
1375 | file = alloc_file(&path, f_mode, |
1376 | is_file_hugepages(shp->shm_file) ? | |
c4caa778 AV |
1377 | &shm_file_operations_huge : |
1378 | &shm_file_operations); | |
39b65252 | 1379 | err = PTR_ERR(file); |
f42569b1 DB |
1380 | if (IS_ERR(file)) { |
1381 | kfree(sfd); | |
1382 | path_put(&path); | |
1383 | goto out_nattch; | |
1384 | } | |
bc56bba8 | 1385 | |
bc56bba8 | 1386 | file->private_data = sfd; |
bc56bba8 | 1387 | file->f_mapping = shp->shm_file->f_mapping; |
7ca7e564 | 1388 | sfd->id = shp->shm_perm.id; |
bc56bba8 EB |
1389 | sfd->ns = get_ipc_ns(ns); |
1390 | sfd->file = shp->shm_file; | |
1391 | sfd->vm_ops = NULL; | |
1392 | ||
8b3ec681 AV |
1393 | err = security_mmap_file(file, prot, flags); |
1394 | if (err) | |
1395 | goto out_fput; | |
1396 | ||
91f4f94e MH |
1397 | if (down_write_killable(¤t->mm->mmap_sem)) { |
1398 | err = -EINTR; | |
1399 | goto out_fput; | |
1400 | } | |
1401 | ||
1da177e4 | 1402 | if (addr && !(shmflg & SHM_REMAP)) { |
bc56bba8 | 1403 | err = -EINVAL; |
247a8ce8 MS |
1404 | if (addr + size < addr) |
1405 | goto invalid; | |
1406 | ||
1da177e4 LT |
1407 | if (find_vma_intersection(current->mm, addr, addr + size)) |
1408 | goto invalid; | |
1da177e4 | 1409 | } |
f42569b1 | 1410 | |
897ab3e0 | 1411 | addr = do_mmap_pgoff(file, addr, size, prot, flags, 0, &populate, NULL); |
bebeb3d6 | 1412 | *raddr = addr; |
bc56bba8 | 1413 | err = 0; |
bebeb3d6 ML |
1414 | if (IS_ERR_VALUE(addr)) |
1415 | err = (long)addr; | |
1da177e4 LT |
1416 | invalid: |
1417 | up_write(¤t->mm->mmap_sem); | |
bebeb3d6 | 1418 | if (populate) |
41badc15 | 1419 | mm_populate(addr, populate); |
1da177e4 | 1420 | |
8b3ec681 | 1421 | out_fput: |
bc56bba8 EB |
1422 | fput(file); |
1423 | ||
1424 | out_nattch: | |
d9a605e4 | 1425 | down_write(&shm_ids(ns).rwsem); |
00c2bf85 | 1426 | shp = shm_lock(ns, shmid); |
1da177e4 | 1427 | shp->shm_nattch--; |
b34a6b1d | 1428 | if (shm_may_destroy(ns, shp)) |
4e982311 | 1429 | shm_destroy(ns, shp); |
1da177e4 LT |
1430 | else |
1431 | shm_unlock(shp); | |
d9a605e4 | 1432 | up_write(&shm_ids(ns).rwsem); |
1da177e4 | 1433 | return err; |
bc56bba8 EB |
1434 | |
1435 | out_unlock: | |
c2c737a0 | 1436 | rcu_read_unlock(); |
f42569b1 DB |
1437 | out: |
1438 | return err; | |
1da177e4 LT |
1439 | } |
1440 | ||
d5460c99 | 1441 | SYSCALL_DEFINE3(shmat, int, shmid, char __user *, shmaddr, int, shmflg) |
7d87e14c SR |
1442 | { |
1443 | unsigned long ret; | |
1444 | long err; | |
1445 | ||
079a96ae | 1446 | err = do_shmat(shmid, shmaddr, shmflg, &ret, SHMLBA); |
7d87e14c SR |
1447 | if (err) |
1448 | return err; | |
1449 | force_successful_syscall_return(); | |
1450 | return (long)ret; | |
1451 | } | |
1452 | ||
a78ee9ed AV |
1453 | #ifdef CONFIG_COMPAT |
1454 | ||
1455 | #ifndef COMPAT_SHMLBA | |
1456 | #define COMPAT_SHMLBA SHMLBA | |
1457 | #endif | |
1458 | ||
1459 | COMPAT_SYSCALL_DEFINE3(shmat, int, shmid, compat_uptr_t, shmaddr, int, shmflg) | |
1460 | { | |
1461 | unsigned long ret; | |
1462 | long err; | |
1463 | ||
1464 | err = do_shmat(shmid, compat_ptr(shmaddr), shmflg, &ret, COMPAT_SHMLBA); | |
1465 | if (err) | |
1466 | return err; | |
1467 | force_successful_syscall_return(); | |
1468 | return (long)ret; | |
1469 | } | |
1470 | #endif | |
1471 | ||
1da177e4 LT |
1472 | /* |
1473 | * detach and kill segment if marked destroyed. | |
1474 | * The work is done in shm_close. | |
1475 | */ | |
d5460c99 | 1476 | SYSCALL_DEFINE1(shmdt, char __user *, shmaddr) |
1da177e4 LT |
1477 | { |
1478 | struct mm_struct *mm = current->mm; | |
586c7e6a | 1479 | struct vm_area_struct *vma; |
1da177e4 | 1480 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1481 | int retval = -EINVAL; |
586c7e6a MF |
1482 | #ifdef CONFIG_MMU |
1483 | loff_t size = 0; | |
d3c97900 | 1484 | struct file *file; |
586c7e6a MF |
1485 | struct vm_area_struct *next; |
1486 | #endif | |
1da177e4 | 1487 | |
df1e2fb5 HD |
1488 | if (addr & ~PAGE_MASK) |
1489 | return retval; | |
1490 | ||
91f4f94e MH |
1491 | if (down_write_killable(&mm->mmap_sem)) |
1492 | return -EINTR; | |
1da177e4 LT |
1493 | |
1494 | /* | |
1495 | * This function tries to be smart and unmap shm segments that | |
1496 | * were modified by partial mlock or munmap calls: | |
1497 | * - It first determines the size of the shm segment that should be | |
1498 | * unmapped: It searches for a vma that is backed by shm and that | |
1499 | * started at address shmaddr. It records it's size and then unmaps | |
1500 | * it. | |
1501 | * - Then it unmaps all shm vmas that started at shmaddr and that | |
d3c97900 DH |
1502 | * are within the initially determined size and that are from the |
1503 | * same shm segment from which we determined the size. | |
1da177e4 LT |
1504 | * Errors from do_munmap are ignored: the function only fails if |
1505 | * it's called with invalid parameters or if it's called to unmap | |
1506 | * a part of a vma. Both calls in this function are for full vmas, | |
1507 | * the parameters are directly copied from the vma itself and always | |
1508 | * valid - therefore do_munmap cannot fail. (famous last words?) | |
1509 | */ | |
1510 | /* | |
1511 | * If it had been mremap()'d, the starting address would not | |
1512 | * match the usual checks anyway. So assume all vma's are | |
1513 | * above the starting address given. | |
1514 | */ | |
1515 | vma = find_vma(mm, addr); | |
1516 | ||
8feae131 | 1517 | #ifdef CONFIG_MMU |
1da177e4 LT |
1518 | while (vma) { |
1519 | next = vma->vm_next; | |
1520 | ||
1521 | /* | |
1522 | * Check if the starting address would match, i.e. it's | |
1523 | * a fragment created by mprotect() and/or munmap(), or it | |
1524 | * otherwise it starts at this address with no hassles. | |
1525 | */ | |
bc56bba8 | 1526 | if ((vma->vm_ops == &shm_vm_ops) && |
1da177e4 LT |
1527 | (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) { |
1528 | ||
d3c97900 DH |
1529 | /* |
1530 | * Record the file of the shm segment being | |
1531 | * unmapped. With mremap(), someone could place | |
1532 | * page from another segment but with equal offsets | |
1533 | * in the range we are unmapping. | |
1534 | */ | |
1535 | file = vma->vm_file; | |
07a46ed2 | 1536 | size = i_size_read(file_inode(vma->vm_file)); |
897ab3e0 | 1537 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1538 | /* |
1539 | * We discovered the size of the shm segment, so | |
1540 | * break out of here and fall through to the next | |
1541 | * loop that uses the size information to stop | |
1542 | * searching for matching vma's. | |
1543 | */ | |
1544 | retval = 0; | |
1545 | vma = next; | |
1546 | break; | |
1547 | } | |
1548 | vma = next; | |
1549 | } | |
1550 | ||
1551 | /* | |
1552 | * We need look no further than the maximum address a fragment | |
1553 | * could possibly have landed at. Also cast things to loff_t to | |
25985edc | 1554 | * prevent overflows and make comparisons vs. equal-width types. |
1da177e4 | 1555 | */ |
8e36709d | 1556 | size = PAGE_ALIGN(size); |
1da177e4 LT |
1557 | while (vma && (loff_t)(vma->vm_end - addr) <= size) { |
1558 | next = vma->vm_next; | |
1559 | ||
1560 | /* finding a matching vma now does not alter retval */ | |
bc56bba8 | 1561 | if ((vma->vm_ops == &shm_vm_ops) && |
d3c97900 DH |
1562 | ((vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) && |
1563 | (vma->vm_file == file)) | |
897ab3e0 | 1564 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1565 | vma = next; |
1566 | } | |
1567 | ||
63980c80 | 1568 | #else /* CONFIG_MMU */ |
8feae131 | 1569 | /* under NOMMU conditions, the exact address to be destroyed must be |
63980c80 SP |
1570 | * given |
1571 | */ | |
530fcd16 | 1572 | if (vma && vma->vm_start == addr && vma->vm_ops == &shm_vm_ops) { |
897ab3e0 | 1573 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
8feae131 DH |
1574 | retval = 0; |
1575 | } | |
1576 | ||
1577 | #endif | |
1578 | ||
1da177e4 LT |
1579 | up_write(&mm->mmap_sem); |
1580 | return retval; | |
1581 | } | |
1582 | ||
1583 | #ifdef CONFIG_PROC_FS | |
19b4946c | 1584 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it) |
1da177e4 | 1585 | { |
1efdb69b | 1586 | struct user_namespace *user_ns = seq_user_ns(s); |
ade9f91b KC |
1587 | struct kern_ipc_perm *ipcp = it; |
1588 | struct shmid_kernel *shp; | |
b7952180 HD |
1589 | unsigned long rss = 0, swp = 0; |
1590 | ||
ade9f91b | 1591 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
b7952180 | 1592 | shm_add_rss_swap(shp, &rss, &swp); |
1da177e4 | 1593 | |
6c826818 PM |
1594 | #if BITS_PER_LONG <= 32 |
1595 | #define SIZE_SPEC "%10lu" | |
1596 | #else | |
1597 | #define SIZE_SPEC "%21lu" | |
1598 | #endif | |
1da177e4 | 1599 | |
7f032d6e JP |
1600 | seq_printf(s, |
1601 | "%10d %10d %4o " SIZE_SPEC " %5u %5u " | |
7ff2819e | 1602 | "%5lu %5u %5u %5u %5u %10llu %10llu %10llu " |
7f032d6e JP |
1603 | SIZE_SPEC " " SIZE_SPEC "\n", |
1604 | shp->shm_perm.key, | |
1605 | shp->shm_perm.id, | |
1606 | shp->shm_perm.mode, | |
1607 | shp->shm_segsz, | |
1608 | shp->shm_cprid, | |
1609 | shp->shm_lprid, | |
1610 | shp->shm_nattch, | |
1611 | from_kuid_munged(user_ns, shp->shm_perm.uid), | |
1612 | from_kgid_munged(user_ns, shp->shm_perm.gid), | |
1613 | from_kuid_munged(user_ns, shp->shm_perm.cuid), | |
1614 | from_kgid_munged(user_ns, shp->shm_perm.cgid), | |
1615 | shp->shm_atim, | |
1616 | shp->shm_dtim, | |
1617 | shp->shm_ctim, | |
1618 | rss * PAGE_SIZE, | |
1619 | swp * PAGE_SIZE); | |
1620 | ||
1621 | return 0; | |
1da177e4 LT |
1622 | } |
1623 | #endif |