]>
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 | ||
a2e102cd EB |
51 | struct shmid_kernel /* private to the kernel */ |
52 | { | |
53 | struct kern_ipc_perm shm_perm; | |
54 | struct file *shm_file; | |
55 | unsigned long shm_nattch; | |
56 | unsigned long shm_segsz; | |
57 | time64_t shm_atim; | |
58 | time64_t shm_dtim; | |
59 | time64_t shm_ctim; | |
98f929b1 EB |
60 | struct pid *shm_cprid; |
61 | struct pid *shm_lprid; | |
a2e102cd EB |
62 | struct user_struct *mlock_user; |
63 | ||
64 | /* The task created the shm object. NULL if the task is dead. */ | |
65 | struct task_struct *shm_creator; | |
66 | struct list_head shm_clist; /* list by creator */ | |
67 | } __randomize_layout; | |
68 | ||
69 | /* shm_mode upper byte flags */ | |
70 | #define SHM_DEST 01000 /* segment will be destroyed on last detach */ | |
71 | #define SHM_LOCKED 02000 /* segment will not be swapped */ | |
72 | ||
bc56bba8 EB |
73 | struct shm_file_data { |
74 | int id; | |
75 | struct ipc_namespace *ns; | |
76 | struct file *file; | |
77 | const struct vm_operations_struct *vm_ops; | |
78 | }; | |
79 | ||
80 | #define shm_file_data(file) (*((struct shm_file_data **)&(file)->private_data)) | |
81 | ||
9a32144e | 82 | static const struct file_operations shm_file_operations; |
f0f37e2f | 83 | static const struct vm_operations_struct shm_vm_ops; |
1da177e4 | 84 | |
ed2ddbf8 | 85 | #define shm_ids(ns) ((ns)->ids[IPC_SHM_IDS]) |
1da177e4 | 86 | |
4e982311 KK |
87 | #define shm_unlock(shp) \ |
88 | ipc_unlock(&(shp)->shm_perm) | |
1da177e4 | 89 | |
7748dbfa | 90 | static int newseg(struct ipc_namespace *, struct ipc_params *); |
bc56bba8 EB |
91 | static void shm_open(struct vm_area_struct *vma); |
92 | static void shm_close(struct vm_area_struct *vma); | |
239521f3 | 93 | static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp); |
1da177e4 | 94 | #ifdef CONFIG_PROC_FS |
19b4946c | 95 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it); |
1da177e4 LT |
96 | #endif |
97 | ||
0cfb6aee | 98 | int shm_init_ns(struct ipc_namespace *ns) |
4e982311 | 99 | { |
4e982311 KK |
100 | ns->shm_ctlmax = SHMMAX; |
101 | ns->shm_ctlall = SHMALL; | |
102 | ns->shm_ctlmni = SHMMNI; | |
b34a6b1d | 103 | ns->shm_rmid_forced = 0; |
4e982311 | 104 | ns->shm_tot = 0; |
0cfb6aee | 105 | return ipc_init_ids(&shm_ids(ns)); |
4e982311 KK |
106 | } |
107 | ||
f4566f04 | 108 | /* |
d9a605e4 DB |
109 | * Called with shm_ids.rwsem (writer) and the shp structure locked. |
110 | * Only shm_ids.rwsem remains locked on exit. | |
f4566f04 | 111 | */ |
01b8b07a | 112 | static void do_shm_rmid(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp) |
4e982311 | 113 | { |
01b8b07a | 114 | struct shmid_kernel *shp; |
63980c80 | 115 | |
01b8b07a PP |
116 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
117 | ||
239521f3 | 118 | if (shp->shm_nattch) { |
4e982311 KK |
119 | shp->shm_perm.mode |= SHM_DEST; |
120 | /* Do not find it any more */ | |
0cfb6aee | 121 | ipc_set_key_private(&shm_ids(ns), &shp->shm_perm); |
4e982311 KK |
122 | shm_unlock(shp); |
123 | } else | |
124 | shm_destroy(ns, shp); | |
125 | } | |
126 | ||
ae5e1b22 | 127 | #ifdef CONFIG_IPC_NS |
4e982311 KK |
128 | void shm_exit_ns(struct ipc_namespace *ns) |
129 | { | |
01b8b07a | 130 | free_ipcs(ns, &shm_ids(ns), do_shm_rmid); |
7d6feeb2 | 131 | idr_destroy(&ns->ids[IPC_SHM_IDS].ipcs_idr); |
0cfb6aee | 132 | rhashtable_destroy(&ns->ids[IPC_SHM_IDS].key_ht); |
4e982311 | 133 | } |
ae5e1b22 | 134 | #endif |
1da177e4 | 135 | |
140d0b21 | 136 | static int __init ipc_ns_init(void) |
1da177e4 | 137 | { |
0cfb6aee GK |
138 | const int err = shm_init_ns(&init_ipc_ns); |
139 | WARN(err, "ipc: sysv shm_init_ns failed: %d\n", err); | |
140 | return err; | |
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 | { |
03f02c76 ND |
182 | struct kern_ipc_perm *ipcp = ipc_lock(&shm_ids(ns), id); |
183 | ||
c5c8975b | 184 | /* |
1ac0b6de KS |
185 | * Callers of shm_lock() must validate the status of the returned ipc |
186 | * object pointer (as returned by ipc_lock()), and error out as | |
187 | * appropriate. | |
c5c8975b | 188 | */ |
1ac0b6de KS |
189 | if (IS_ERR(ipcp)) |
190 | return (void *)ipcp; | |
03f02c76 | 191 | return container_of(ipcp, struct shmid_kernel, shm_perm); |
023a5355 ND |
192 | } |
193 | ||
4c677e2e VK |
194 | static inline void shm_lock_by_ptr(struct shmid_kernel *ipcp) |
195 | { | |
196 | rcu_read_lock(); | |
cf9d5d78 | 197 | ipc_lock_object(&ipcp->shm_perm); |
4c677e2e VK |
198 | } |
199 | ||
53dad6d3 DB |
200 | static void shm_rcu_free(struct rcu_head *head) |
201 | { | |
dba4cdd3 MS |
202 | struct kern_ipc_perm *ptr = container_of(head, struct kern_ipc_perm, |
203 | rcu); | |
204 | struct shmid_kernel *shp = container_of(ptr, struct shmid_kernel, | |
205 | shm_perm); | |
7191adff | 206 | security_shm_free(&shp->shm_perm); |
42e618f7 | 207 | kvfree(shp); |
53dad6d3 DB |
208 | } |
209 | ||
7ca7e564 | 210 | static inline void shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *s) |
1da177e4 | 211 | { |
ab602f79 | 212 | list_del(&s->shm_clist); |
7ca7e564 | 213 | ipc_rmid(&shm_ids(ns), &s->shm_perm); |
1da177e4 LT |
214 | } |
215 | ||
1da177e4 | 216 | |
1ac0b6de | 217 | static int __shm_open(struct vm_area_struct *vma) |
4e982311 | 218 | { |
bc56bba8 EB |
219 | struct file *file = vma->vm_file; |
220 | struct shm_file_data *sfd = shm_file_data(file); | |
1da177e4 LT |
221 | struct shmid_kernel *shp; |
222 | ||
bc56bba8 | 223 | shp = shm_lock(sfd->ns, sfd->id); |
1ac0b6de KS |
224 | |
225 | if (IS_ERR(shp)) | |
226 | return PTR_ERR(shp); | |
227 | ||
3f05317d EB |
228 | if (shp->shm_file != sfd->file) { |
229 | /* ID was reused */ | |
230 | shm_unlock(shp); | |
231 | return -EINVAL; | |
232 | } | |
233 | ||
7ff2819e | 234 | shp->shm_atim = ktime_get_real_seconds(); |
98f929b1 | 235 | ipc_update_pid(&shp->shm_lprid, task_tgid(current)); |
1da177e4 LT |
236 | shp->shm_nattch++; |
237 | shm_unlock(shp); | |
1ac0b6de KS |
238 | return 0; |
239 | } | |
240 | ||
241 | /* This is called by fork, once for every shm attach. */ | |
242 | static void shm_open(struct vm_area_struct *vma) | |
243 | { | |
244 | int err = __shm_open(vma); | |
245 | /* | |
246 | * We raced in the idr lookup or with shm_destroy(). | |
247 | * Either way, the ID is busted. | |
248 | */ | |
249 | WARN_ON_ONCE(err); | |
1da177e4 LT |
250 | } |
251 | ||
1da177e4 LT |
252 | /* |
253 | * shm_destroy - free the struct shmid_kernel | |
254 | * | |
f4566f04 | 255 | * @ns: namespace |
1da177e4 LT |
256 | * @shp: struct to free |
257 | * | |
d9a605e4 | 258 | * It has to be called with shp and shm_ids.rwsem (writer) locked, |
1da177e4 LT |
259 | * but returns with shp unlocked and freed. |
260 | */ | |
4e982311 | 261 | static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) |
1da177e4 | 262 | { |
a399b29d GT |
263 | struct file *shm_file; |
264 | ||
265 | shm_file = shp->shm_file; | |
266 | shp->shm_file = NULL; | |
4e982311 | 267 | ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT; |
7ca7e564 | 268 | shm_rmid(ns, shp); |
1da177e4 | 269 | shm_unlock(shp); |
a399b29d GT |
270 | if (!is_file_hugepages(shm_file)) |
271 | shmem_lock(shm_file, 0, shp->mlock_user); | |
353d5c30 | 272 | else if (shp->mlock_user) |
07a46ed2 DH |
273 | user_shm_unlock(i_size_read(file_inode(shm_file)), |
274 | shp->mlock_user); | |
a399b29d | 275 | fput(shm_file); |
98f929b1 EB |
276 | ipc_update_pid(&shp->shm_cprid, NULL); |
277 | ipc_update_pid(&shp->shm_lprid, NULL); | |
dba4cdd3 | 278 | ipc_rcu_putref(&shp->shm_perm, shm_rcu_free); |
1da177e4 LT |
279 | } |
280 | ||
b34a6b1d VK |
281 | /* |
282 | * shm_may_destroy - identifies whether shm segment should be destroyed now | |
283 | * | |
284 | * Returns true if and only if there are no active users of the segment and | |
285 | * one of the following is true: | |
286 | * | |
287 | * 1) shmctl(id, IPC_RMID, NULL) was called for this shp | |
288 | * | |
289 | * 2) sysctl kernel.shm_rmid_forced is set to 1. | |
290 | */ | |
291 | static bool shm_may_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) | |
292 | { | |
293 | return (shp->shm_nattch == 0) && | |
294 | (ns->shm_rmid_forced || | |
295 | (shp->shm_perm.mode & SHM_DEST)); | |
296 | } | |
297 | ||
1da177e4 | 298 | /* |
bc56bba8 | 299 | * remove the attach descriptor vma. |
1da177e4 LT |
300 | * free memory for segment if it is marked destroyed. |
301 | * The descriptor has already been removed from the current->mm->mmap list | |
302 | * and will later be kfree()d. | |
303 | */ | |
bc56bba8 | 304 | static void shm_close(struct vm_area_struct *vma) |
1da177e4 | 305 | { |
239521f3 | 306 | struct file *file = vma->vm_file; |
bc56bba8 | 307 | struct shm_file_data *sfd = shm_file_data(file); |
1da177e4 | 308 | struct shmid_kernel *shp; |
bc56bba8 | 309 | struct ipc_namespace *ns = sfd->ns; |
4e982311 | 310 | |
d9a605e4 | 311 | down_write(&shm_ids(ns).rwsem); |
1da177e4 | 312 | /* remove from the list of attaches of the shm segment */ |
00c2bf85 | 313 | shp = shm_lock(ns, sfd->id); |
1ac0b6de KS |
314 | |
315 | /* | |
316 | * We raced in the idr lookup or with shm_destroy(). | |
317 | * Either way, the ID is busted. | |
318 | */ | |
319 | if (WARN_ON_ONCE(IS_ERR(shp))) | |
320 | goto done; /* no-op */ | |
321 | ||
98f929b1 | 322 | ipc_update_pid(&shp->shm_lprid, task_tgid(current)); |
7ff2819e | 323 | shp->shm_dtim = ktime_get_real_seconds(); |
1da177e4 | 324 | shp->shm_nattch--; |
b34a6b1d VK |
325 | if (shm_may_destroy(ns, shp)) |
326 | shm_destroy(ns, shp); | |
327 | else | |
328 | shm_unlock(shp); | |
1ac0b6de | 329 | done: |
d9a605e4 | 330 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
331 | } |
332 | ||
d9a605e4 | 333 | /* Called with ns->shm_ids(ns).rwsem locked */ |
b34a6b1d VK |
334 | static int shm_try_destroy_orphaned(int id, void *p, void *data) |
335 | { | |
336 | struct ipc_namespace *ns = data; | |
4c677e2e VK |
337 | struct kern_ipc_perm *ipcp = p; |
338 | struct shmid_kernel *shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
b34a6b1d VK |
339 | |
340 | /* | |
341 | * We want to destroy segments without users and with already | |
342 | * exit'ed originating process. | |
4c677e2e | 343 | * |
d9a605e4 | 344 | * As shp->* are changed under rwsem, it's safe to skip shp locking. |
b34a6b1d | 345 | */ |
4c677e2e | 346 | if (shp->shm_creator != NULL) |
b34a6b1d | 347 | return 0; |
b34a6b1d | 348 | |
4c677e2e VK |
349 | if (shm_may_destroy(ns, shp)) { |
350 | shm_lock_by_ptr(shp); | |
4e982311 | 351 | shm_destroy(ns, shp); |
4c677e2e | 352 | } |
b34a6b1d VK |
353 | return 0; |
354 | } | |
355 | ||
356 | void shm_destroy_orphaned(struct ipc_namespace *ns) | |
357 | { | |
d9a605e4 | 358 | down_write(&shm_ids(ns).rwsem); |
33a30ed4 | 359 | if (shm_ids(ns).in_use) |
4c677e2e | 360 | idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_orphaned, ns); |
d9a605e4 | 361 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
362 | } |
363 | ||
83293c0f | 364 | /* Locking assumes this will only be called with task == current */ |
b34a6b1d VK |
365 | void exit_shm(struct task_struct *task) |
366 | { | |
4c677e2e | 367 | struct ipc_namespace *ns = task->nsproxy->ipc_ns; |
ab602f79 | 368 | struct shmid_kernel *shp, *n; |
b34a6b1d | 369 | |
83293c0f JM |
370 | if (list_empty(&task->sysvshm.shm_clist)) |
371 | return; | |
372 | ||
373 | /* | |
374 | * If kernel.shm_rmid_forced is not set then only keep track of | |
375 | * which shmids are orphaned, so that a later set of the sysctl | |
376 | * can clean them up. | |
377 | */ | |
378 | if (!ns->shm_rmid_forced) { | |
379 | down_read(&shm_ids(ns).rwsem); | |
380 | list_for_each_entry(shp, &task->sysvshm.shm_clist, shm_clist) | |
381 | shp->shm_creator = NULL; | |
382 | /* | |
383 | * Only under read lock but we are only called on current | |
384 | * so no entry on the list will be shared. | |
385 | */ | |
386 | list_del(&task->sysvshm.shm_clist); | |
387 | up_read(&shm_ids(ns).rwsem); | |
298507d4 | 388 | return; |
83293c0f | 389 | } |
298507d4 | 390 | |
83293c0f JM |
391 | /* |
392 | * Destroy all already created segments, that were not yet mapped, | |
393 | * and mark any mapped as orphan to cover the sysctl toggling. | |
394 | * Destroy is skipped if shm_may_destroy() returns false. | |
395 | */ | |
d9a605e4 | 396 | down_write(&shm_ids(ns).rwsem); |
83293c0f JM |
397 | list_for_each_entry_safe(shp, n, &task->sysvshm.shm_clist, shm_clist) { |
398 | shp->shm_creator = NULL; | |
399 | ||
400 | if (shm_may_destroy(ns, shp)) { | |
401 | shm_lock_by_ptr(shp); | |
402 | shm_destroy(ns, shp); | |
403 | } | |
404 | } | |
405 | ||
406 | /* Remove the list head from any segments still attached. */ | |
ab602f79 | 407 | list_del(&task->sysvshm.shm_clist); |
d9a605e4 | 408 | up_write(&shm_ids(ns).rwsem); |
1da177e4 LT |
409 | } |
410 | ||
14f28f57 | 411 | static vm_fault_t shm_fault(struct vm_fault *vmf) |
bc56bba8 | 412 | { |
11bac800 | 413 | struct file *file = vmf->vma->vm_file; |
bc56bba8 EB |
414 | struct shm_file_data *sfd = shm_file_data(file); |
415 | ||
11bac800 | 416 | return sfd->vm_ops->fault(vmf); |
bc56bba8 EB |
417 | } |
418 | ||
3d942ee0 MK |
419 | static int shm_split(struct vm_area_struct *vma, unsigned long addr) |
420 | { | |
421 | struct file *file = vma->vm_file; | |
422 | struct shm_file_data *sfd = shm_file_data(file); | |
423 | ||
a61fc2cb | 424 | if (sfd->vm_ops->split) |
3d942ee0 MK |
425 | return sfd->vm_ops->split(vma, addr); |
426 | ||
427 | return 0; | |
428 | } | |
429 | ||
eec3636a JC |
430 | static unsigned long shm_pagesize(struct vm_area_struct *vma) |
431 | { | |
432 | struct file *file = vma->vm_file; | |
433 | struct shm_file_data *sfd = shm_file_data(file); | |
434 | ||
435 | if (sfd->vm_ops->pagesize) | |
436 | return sfd->vm_ops->pagesize(vma); | |
437 | ||
438 | return PAGE_SIZE; | |
439 | } | |
440 | ||
bc56bba8 | 441 | #ifdef CONFIG_NUMA |
d823e3e7 | 442 | static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new) |
bc56bba8 EB |
443 | { |
444 | struct file *file = vma->vm_file; | |
445 | struct shm_file_data *sfd = shm_file_data(file); | |
446 | int err = 0; | |
63980c80 | 447 | |
bc56bba8 EB |
448 | if (sfd->vm_ops->set_policy) |
449 | err = sfd->vm_ops->set_policy(vma, new); | |
450 | return err; | |
451 | } | |
452 | ||
d823e3e7 AB |
453 | static struct mempolicy *shm_get_policy(struct vm_area_struct *vma, |
454 | unsigned long addr) | |
bc56bba8 EB |
455 | { |
456 | struct file *file = vma->vm_file; | |
457 | struct shm_file_data *sfd = shm_file_data(file); | |
458 | struct mempolicy *pol = NULL; | |
459 | ||
460 | if (sfd->vm_ops->get_policy) | |
461 | pol = sfd->vm_ops->get_policy(vma, addr); | |
52cd3b07 | 462 | else if (vma->vm_policy) |
bc56bba8 | 463 | pol = vma->vm_policy; |
52cd3b07 | 464 | |
bc56bba8 EB |
465 | return pol; |
466 | } | |
467 | #endif | |
468 | ||
239521f3 | 469 | static int shm_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 470 | { |
bc56bba8 | 471 | struct shm_file_data *sfd = shm_file_data(file); |
b0e15190 DH |
472 | int ret; |
473 | ||
1ac0b6de | 474 | /* |
3f05317d EB |
475 | * In case of remap_file_pages() emulation, the file can represent an |
476 | * IPC ID that was removed, and possibly even reused by another shm | |
477 | * segment already. Propagate this case as an error to caller. | |
1ac0b6de | 478 | */ |
63980c80 | 479 | ret = __shm_open(vma); |
1ac0b6de KS |
480 | if (ret) |
481 | return ret; | |
482 | ||
f74ac015 | 483 | ret = call_mmap(sfd->file, vma); |
1ac0b6de KS |
484 | if (ret) { |
485 | shm_close(vma); | |
bc56bba8 | 486 | return ret; |
1ac0b6de | 487 | } |
bc56bba8 | 488 | sfd->vm_ops = vma->vm_ops; |
2e92a3ba | 489 | #ifdef CONFIG_MMU |
d0edd852 | 490 | WARN_ON(!sfd->vm_ops->fault); |
2e92a3ba | 491 | #endif |
bc56bba8 | 492 | vma->vm_ops = &shm_vm_ops; |
1ac0b6de | 493 | return 0; |
1da177e4 LT |
494 | } |
495 | ||
4e982311 KK |
496 | static int shm_release(struct inode *ino, struct file *file) |
497 | { | |
bc56bba8 | 498 | struct shm_file_data *sfd = shm_file_data(file); |
4e982311 | 499 | |
bc56bba8 | 500 | put_ipc_ns(sfd->ns); |
3f05317d | 501 | fput(sfd->file); |
bc56bba8 EB |
502 | shm_file_data(file) = NULL; |
503 | kfree(sfd); | |
4e982311 KK |
504 | return 0; |
505 | } | |
506 | ||
02c24a82 | 507 | static int shm_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
516dffdc | 508 | { |
516dffdc | 509 | struct shm_file_data *sfd = shm_file_data(file); |
516dffdc | 510 | |
7ea80859 CH |
511 | if (!sfd->file->f_op->fsync) |
512 | return -EINVAL; | |
0f41074a | 513 | return sfd->file->f_op->fsync(sfd->file, start, end, datasync); |
516dffdc AL |
514 | } |
515 | ||
7d8a4569 WD |
516 | static long shm_fallocate(struct file *file, int mode, loff_t offset, |
517 | loff_t len) | |
518 | { | |
519 | struct shm_file_data *sfd = shm_file_data(file); | |
520 | ||
521 | if (!sfd->file->f_op->fallocate) | |
522 | return -EOPNOTSUPP; | |
523 | return sfd->file->f_op->fallocate(file, mode, offset, len); | |
524 | } | |
525 | ||
bc56bba8 EB |
526 | static unsigned long shm_get_unmapped_area(struct file *file, |
527 | unsigned long addr, unsigned long len, unsigned long pgoff, | |
528 | unsigned long flags) | |
529 | { | |
530 | struct shm_file_data *sfd = shm_file_data(file); | |
63980c80 | 531 | |
c4caa778 AV |
532 | return sfd->file->f_op->get_unmapped_area(sfd->file, addr, len, |
533 | pgoff, flags); | |
bc56bba8 | 534 | } |
bc56bba8 | 535 | |
9a32144e | 536 | static const struct file_operations shm_file_operations = { |
4e982311 | 537 | .mmap = shm_mmap, |
516dffdc | 538 | .fsync = shm_fsync, |
4e982311 | 539 | .release = shm_release, |
ed5e5894 | 540 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 541 | .llseek = noop_llseek, |
7d8a4569 | 542 | .fallocate = shm_fallocate, |
c4caa778 AV |
543 | }; |
544 | ||
c01d5b30 HD |
545 | /* |
546 | * shm_file_operations_huge is now identical to shm_file_operations, | |
547 | * but we keep it distinct for the sake of is_file_shm_hugepages(). | |
548 | */ | |
c4caa778 AV |
549 | static const struct file_operations shm_file_operations_huge = { |
550 | .mmap = shm_mmap, | |
551 | .fsync = shm_fsync, | |
552 | .release = shm_release, | |
bc56bba8 | 553 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 554 | .llseek = noop_llseek, |
7d8a4569 | 555 | .fallocate = shm_fallocate, |
1da177e4 LT |
556 | }; |
557 | ||
2954e440 | 558 | bool is_file_shm_hugepages(struct file *file) |
c4caa778 AV |
559 | { |
560 | return file->f_op == &shm_file_operations_huge; | |
561 | } | |
562 | ||
f0f37e2f | 563 | static const struct vm_operations_struct shm_vm_ops = { |
1da177e4 LT |
564 | .open = shm_open, /* callback for a new vm-area open */ |
565 | .close = shm_close, /* callback for when the vm-area is released */ | |
54cb8821 | 566 | .fault = shm_fault, |
3d942ee0 | 567 | .split = shm_split, |
eec3636a | 568 | .pagesize = shm_pagesize, |
bc56bba8 EB |
569 | #if defined(CONFIG_NUMA) |
570 | .set_policy = shm_set_policy, | |
571 | .get_policy = shm_get_policy, | |
1da177e4 LT |
572 | #endif |
573 | }; | |
574 | ||
f4566f04 ND |
575 | /** |
576 | * newseg - Create a new shared memory segment | |
577 | * @ns: namespace | |
578 | * @params: ptr to the structure that contains key, size and shmflg | |
579 | * | |
d9a605e4 | 580 | * Called with shm_ids.rwsem held as a writer. |
f4566f04 | 581 | */ |
7748dbfa | 582 | static int newseg(struct ipc_namespace *ns, struct ipc_params *params) |
1da177e4 | 583 | { |
7748dbfa ND |
584 | key_t key = params->key; |
585 | int shmflg = params->flg; | |
586 | size_t size = params->u.size; | |
1da177e4 LT |
587 | int error; |
588 | struct shmid_kernel *shp; | |
d69f3bad | 589 | size_t numpages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; |
239521f3 | 590 | struct file *file; |
1da177e4 | 591 | char name[13]; |
ca16d140 | 592 | vm_flags_t acctflag = 0; |
1da177e4 | 593 | |
4e982311 | 594 | if (size < SHMMIN || size > ns->shm_ctlmax) |
1da177e4 LT |
595 | return -EINVAL; |
596 | ||
1376327c MS |
597 | if (numpages << PAGE_SHIFT < size) |
598 | return -ENOSPC; | |
599 | ||
09c6eb1f MS |
600 | if (ns->shm_tot + numpages < ns->shm_tot || |
601 | ns->shm_tot + numpages > ns->shm_ctlall) | |
1da177e4 LT |
602 | return -ENOSPC; |
603 | ||
42e618f7 KC |
604 | shp = kvmalloc(sizeof(*shp), GFP_KERNEL); |
605 | if (unlikely(!shp)) | |
1da177e4 LT |
606 | return -ENOMEM; |
607 | ||
608 | shp->shm_perm.key = key; | |
b33291c0 | 609 | shp->shm_perm.mode = (shmflg & S_IRWXUGO); |
1da177e4 LT |
610 | shp->mlock_user = NULL; |
611 | ||
612 | shp->shm_perm.security = NULL; | |
7191adff | 613 | error = security_shm_alloc(&shp->shm_perm); |
1da177e4 | 614 | if (error) { |
42e618f7 | 615 | kvfree(shp); |
1da177e4 LT |
616 | return error; |
617 | } | |
618 | ||
239521f3 | 619 | sprintf(name, "SYSV%08x", key); |
1da177e4 | 620 | if (shmflg & SHM_HUGETLB) { |
c103a4dc | 621 | struct hstate *hs; |
091d0d55 LZ |
622 | size_t hugesize; |
623 | ||
c103a4dc | 624 | hs = hstate_sizelog((shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); |
091d0d55 LZ |
625 | if (!hs) { |
626 | error = -EINVAL; | |
627 | goto no_file; | |
628 | } | |
629 | hugesize = ALIGN(size, huge_page_size(hs)); | |
af73e4d9 | 630 | |
5a6fe125 MG |
631 | /* hugetlb_file_setup applies strict accounting */ |
632 | if (shmflg & SHM_NORESERVE) | |
633 | acctflag = VM_NORESERVE; | |
af73e4d9 | 634 | file = hugetlb_file_setup(name, hugesize, acctflag, |
42d7395f AK |
635 | &shp->mlock_user, HUGETLB_SHMFS_INODE, |
636 | (shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); | |
1da177e4 | 637 | } else { |
bf8f972d BP |
638 | /* |
639 | * Do not allow no accounting for OVERCOMMIT_NEVER, even | |
239521f3 | 640 | * if it's asked for. |
bf8f972d BP |
641 | */ |
642 | if ((shmflg & SHM_NORESERVE) && | |
643 | sysctl_overcommit_memory != OVERCOMMIT_NEVER) | |
fc8744ad | 644 | acctflag = VM_NORESERVE; |
e1832f29 | 645 | file = shmem_kernel_file_setup(name, size, acctflag); |
1da177e4 LT |
646 | } |
647 | error = PTR_ERR(file); | |
648 | if (IS_ERR(file)) | |
649 | goto no_file; | |
650 | ||
98f929b1 EB |
651 | shp->shm_cprid = get_pid(task_tgid(current)); |
652 | shp->shm_lprid = NULL; | |
1da177e4 | 653 | shp->shm_atim = shp->shm_dtim = 0; |
7ff2819e | 654 | shp->shm_ctim = ktime_get_real_seconds(); |
1da177e4 LT |
655 | shp->shm_segsz = size; |
656 | shp->shm_nattch = 0; | |
1da177e4 | 657 | shp->shm_file = file; |
5774ed01 | 658 | shp->shm_creator = current; |
b9a53227 | 659 | |
39c96a1b | 660 | /* ipc_addid() locks shp upon success. */ |
a2642f87 MS |
661 | error = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni); |
662 | if (error < 0) | |
b9a53227 | 663 | goto no_id; |
b9a53227 | 664 | |
ab602f79 | 665 | list_add(&shp->shm_clist, ¤t->sysvshm.shm_clist); |
dbfcd91f | 666 | |
30475cc1 BP |
667 | /* |
668 | * shmid gets reported as "inode#" in /proc/pid/maps. | |
669 | * proc-ps tools use this. Changing this will break them. | |
670 | */ | |
496ad9aa | 671 | file_inode(file)->i_ino = shp->shm_perm.id; |
551110a9 | 672 | |
4e982311 | 673 | ns->shm_tot += numpages; |
7ca7e564 | 674 | error = shp->shm_perm.id; |
dbfcd91f | 675 | |
cf9d5d78 | 676 | ipc_unlock_object(&shp->shm_perm); |
dbfcd91f | 677 | rcu_read_unlock(); |
7ca7e564 | 678 | return error; |
1da177e4 LT |
679 | |
680 | no_id: | |
2236d4d3 EB |
681 | ipc_update_pid(&shp->shm_cprid, NULL); |
682 | ipc_update_pid(&shp->shm_lprid, NULL); | |
2195d281 | 683 | if (is_file_hugepages(file) && shp->mlock_user) |
353d5c30 | 684 | user_shm_unlock(size, shp->mlock_user); |
1da177e4 LT |
685 | fput(file); |
686 | no_file: | |
a2642f87 | 687 | call_rcu(&shp->shm_perm.rcu, shm_rcu_free); |
1da177e4 LT |
688 | return error; |
689 | } | |
690 | ||
f4566f04 | 691 | /* |
d9a605e4 | 692 | * Called with shm_ids.rwsem and ipcp locked. |
f4566f04 | 693 | */ |
03f02c76 ND |
694 | static inline int shm_more_checks(struct kern_ipc_perm *ipcp, |
695 | struct ipc_params *params) | |
7748dbfa | 696 | { |
03f02c76 ND |
697 | struct shmid_kernel *shp; |
698 | ||
699 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
700 | if (shp->shm_segsz < params->u.size) | |
7748dbfa ND |
701 | return -EINVAL; |
702 | ||
703 | return 0; | |
704 | } | |
705 | ||
65749e0b | 706 | long ksys_shmget(key_t key, size_t size, int shmflg) |
1da177e4 | 707 | { |
4e982311 | 708 | struct ipc_namespace *ns; |
eb66ec44 MK |
709 | static const struct ipc_ops shm_ops = { |
710 | .getnew = newseg, | |
50ab44b1 | 711 | .associate = security_shm_associate, |
eb66ec44 MK |
712 | .more_checks = shm_more_checks, |
713 | }; | |
7748dbfa | 714 | struct ipc_params shm_params; |
4e982311 KK |
715 | |
716 | ns = current->nsproxy->ipc_ns; | |
1da177e4 | 717 | |
7748dbfa ND |
718 | shm_params.key = key; |
719 | shm_params.flg = shmflg; | |
720 | shm_params.u.size = size; | |
1da177e4 | 721 | |
7748dbfa | 722 | return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params); |
1da177e4 LT |
723 | } |
724 | ||
65749e0b DB |
725 | SYSCALL_DEFINE3(shmget, key_t, key, size_t, size, int, shmflg) |
726 | { | |
727 | return ksys_shmget(key, size, shmflg); | |
728 | } | |
729 | ||
1da177e4 LT |
730 | static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version) |
731 | { | |
239521f3 | 732 | switch (version) { |
1da177e4 LT |
733 | case IPC_64: |
734 | return copy_to_user(buf, in, sizeof(*in)); | |
735 | case IPC_OLD: | |
736 | { | |
737 | struct shmid_ds out; | |
738 | ||
3af54c9b | 739 | memset(&out, 0, sizeof(out)); |
1da177e4 LT |
740 | ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm); |
741 | out.shm_segsz = in->shm_segsz; | |
742 | out.shm_atime = in->shm_atime; | |
743 | out.shm_dtime = in->shm_dtime; | |
744 | out.shm_ctime = in->shm_ctime; | |
745 | out.shm_cpid = in->shm_cpid; | |
746 | out.shm_lpid = in->shm_lpid; | |
747 | out.shm_nattch = in->shm_nattch; | |
748 | ||
749 | return copy_to_user(buf, &out, sizeof(out)); | |
750 | } | |
751 | default: | |
752 | return -EINVAL; | |
753 | } | |
754 | } | |
755 | ||
016d7132 PP |
756 | static inline unsigned long |
757 | copy_shmid_from_user(struct shmid64_ds *out, void __user *buf, int version) | |
1da177e4 | 758 | { |
239521f3 | 759 | switch (version) { |
1da177e4 | 760 | case IPC_64: |
016d7132 | 761 | if (copy_from_user(out, buf, sizeof(*out))) |
1da177e4 | 762 | return -EFAULT; |
1da177e4 | 763 | return 0; |
1da177e4 LT |
764 | case IPC_OLD: |
765 | { | |
766 | struct shmid_ds tbuf_old; | |
767 | ||
768 | if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) | |
769 | return -EFAULT; | |
770 | ||
016d7132 PP |
771 | out->shm_perm.uid = tbuf_old.shm_perm.uid; |
772 | out->shm_perm.gid = tbuf_old.shm_perm.gid; | |
773 | out->shm_perm.mode = tbuf_old.shm_perm.mode; | |
1da177e4 LT |
774 | |
775 | return 0; | |
776 | } | |
777 | default: | |
778 | return -EINVAL; | |
779 | } | |
780 | } | |
781 | ||
782 | static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version) | |
783 | { | |
239521f3 | 784 | switch (version) { |
1da177e4 LT |
785 | case IPC_64: |
786 | return copy_to_user(buf, in, sizeof(*in)); | |
787 | case IPC_OLD: | |
788 | { | |
789 | struct shminfo out; | |
790 | ||
239521f3 | 791 | if (in->shmmax > INT_MAX) |
1da177e4 LT |
792 | out.shmmax = INT_MAX; |
793 | else | |
794 | out.shmmax = (int)in->shmmax; | |
795 | ||
796 | out.shmmin = in->shmmin; | |
797 | out.shmmni = in->shmmni; | |
798 | out.shmseg = in->shmseg; | |
46c0a8ca | 799 | out.shmall = in->shmall; |
1da177e4 LT |
800 | |
801 | return copy_to_user(buf, &out, sizeof(out)); | |
802 | } | |
803 | default: | |
804 | return -EINVAL; | |
805 | } | |
806 | } | |
807 | ||
b7952180 HD |
808 | /* |
809 | * Calculate and add used RSS and swap pages of a shm. | |
d9a605e4 | 810 | * Called with shm_ids.rwsem held as a reader |
b7952180 HD |
811 | */ |
812 | static void shm_add_rss_swap(struct shmid_kernel *shp, | |
813 | unsigned long *rss_add, unsigned long *swp_add) | |
814 | { | |
815 | struct inode *inode; | |
816 | ||
496ad9aa | 817 | inode = file_inode(shp->shm_file); |
b7952180 HD |
818 | |
819 | if (is_file_hugepages(shp->shm_file)) { | |
820 | struct address_space *mapping = inode->i_mapping; | |
821 | struct hstate *h = hstate_file(shp->shm_file); | |
822 | *rss_add += pages_per_huge_page(h) * mapping->nrpages; | |
823 | } else { | |
824 | #ifdef CONFIG_SHMEM | |
825 | struct shmem_inode_info *info = SHMEM_I(inode); | |
63980c80 | 826 | |
4595ef88 | 827 | spin_lock_irq(&info->lock); |
b7952180 HD |
828 | *rss_add += inode->i_mapping->nrpages; |
829 | *swp_add += info->swapped; | |
4595ef88 | 830 | spin_unlock_irq(&info->lock); |
b7952180 HD |
831 | #else |
832 | *rss_add += inode->i_mapping->nrpages; | |
833 | #endif | |
834 | } | |
835 | } | |
836 | ||
f4566f04 | 837 | /* |
d9a605e4 | 838 | * Called with shm_ids.rwsem held as a reader |
f4566f04 | 839 | */ |
4e982311 KK |
840 | static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss, |
841 | unsigned long *swp) | |
1da177e4 | 842 | { |
7ca7e564 ND |
843 | int next_id; |
844 | int total, in_use; | |
1da177e4 LT |
845 | |
846 | *rss = 0; | |
847 | *swp = 0; | |
848 | ||
7ca7e564 ND |
849 | in_use = shm_ids(ns).in_use; |
850 | ||
851 | for (total = 0, next_id = 0; total < in_use; next_id++) { | |
e562aebc | 852 | struct kern_ipc_perm *ipc; |
1da177e4 | 853 | struct shmid_kernel *shp; |
1da177e4 | 854 | |
e562aebc TB |
855 | ipc = idr_find(&shm_ids(ns).ipcs_idr, next_id); |
856 | if (ipc == NULL) | |
1da177e4 | 857 | continue; |
e562aebc | 858 | shp = container_of(ipc, struct shmid_kernel, shm_perm); |
1da177e4 | 859 | |
b7952180 | 860 | shm_add_rss_swap(shp, rss, swp); |
7ca7e564 ND |
861 | |
862 | total++; | |
1da177e4 LT |
863 | } |
864 | } | |
865 | ||
8d4cc8b5 | 866 | /* |
d9a605e4 | 867 | * This function handles some shmctl commands which require the rwsem |
8d4cc8b5 | 868 | * to be held in write mode. |
d9a605e4 | 869 | * NOTE: no locks must be held, the rwsem is taken inside this function. |
8d4cc8b5 PP |
870 | */ |
871 | static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd, | |
9ba720c1 | 872 | struct shmid64_ds *shmid64) |
1da177e4 | 873 | { |
8d4cc8b5 | 874 | struct kern_ipc_perm *ipcp; |
8d4cc8b5 PP |
875 | struct shmid_kernel *shp; |
876 | int err; | |
877 | ||
d9a605e4 | 878 | down_write(&shm_ids(ns).rwsem); |
7b4cc5d8 DB |
879 | rcu_read_lock(); |
880 | ||
79ccf0f8 | 881 | ipcp = ipcctl_pre_down_nolock(ns, &shm_ids(ns), shmid, cmd, |
9ba720c1 | 882 | &shmid64->shm_perm, 0); |
7b4cc5d8 DB |
883 | if (IS_ERR(ipcp)) { |
884 | err = PTR_ERR(ipcp); | |
7b4cc5d8 DB |
885 | goto out_unlock1; |
886 | } | |
8d4cc8b5 | 887 | |
a5f75e7f | 888 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
8d4cc8b5 | 889 | |
7191adff | 890 | err = security_shm_shmctl(&shp->shm_perm, cmd); |
8d4cc8b5 | 891 | if (err) |
79ccf0f8 | 892 | goto out_unlock1; |
7b4cc5d8 | 893 | |
8d4cc8b5 PP |
894 | switch (cmd) { |
895 | case IPC_RMID: | |
79ccf0f8 | 896 | ipc_lock_object(&shp->shm_perm); |
7b4cc5d8 | 897 | /* do_shm_rmid unlocks the ipc object and rcu */ |
8d4cc8b5 PP |
898 | do_shm_rmid(ns, ipcp); |
899 | goto out_up; | |
900 | case IPC_SET: | |
79ccf0f8 | 901 | ipc_lock_object(&shp->shm_perm); |
9ba720c1 | 902 | err = ipc_update_perm(&shmid64->shm_perm, ipcp); |
1efdb69b | 903 | if (err) |
7b4cc5d8 | 904 | goto out_unlock0; |
7ff2819e | 905 | shp->shm_ctim = ktime_get_real_seconds(); |
8d4cc8b5 PP |
906 | break; |
907 | default: | |
908 | err = -EINVAL; | |
79ccf0f8 | 909 | goto out_unlock1; |
8d4cc8b5 | 910 | } |
7b4cc5d8 DB |
911 | |
912 | out_unlock0: | |
913 | ipc_unlock_object(&shp->shm_perm); | |
914 | out_unlock1: | |
915 | rcu_read_unlock(); | |
8d4cc8b5 | 916 | out_up: |
d9a605e4 | 917 | up_write(&shm_ids(ns).rwsem); |
8d4cc8b5 PP |
918 | return err; |
919 | } | |
920 | ||
9ba720c1 AV |
921 | static int shmctl_ipc_info(struct ipc_namespace *ns, |
922 | struct shminfo64 *shminfo) | |
8d4cc8b5 | 923 | { |
9ba720c1 AV |
924 | int err = security_shm_shmctl(NULL, IPC_INFO); |
925 | if (!err) { | |
926 | memset(shminfo, 0, sizeof(*shminfo)); | |
927 | shminfo->shmmni = shminfo->shmseg = ns->shm_ctlmni; | |
928 | shminfo->shmmax = ns->shm_ctlmax; | |
929 | shminfo->shmall = ns->shm_ctlall; | |
930 | shminfo->shmmin = SHMMIN; | |
d9a605e4 | 931 | down_read(&shm_ids(ns).rwsem); |
7ca7e564 | 932 | err = ipc_get_maxid(&shm_ids(ns)); |
d9a605e4 | 933 | up_read(&shm_ids(ns).rwsem); |
239521f3 | 934 | if (err < 0) |
1da177e4 | 935 | err = 0; |
1da177e4 | 936 | } |
9ba720c1 AV |
937 | return err; |
938 | } | |
1da177e4 | 939 | |
9ba720c1 AV |
940 | static int shmctl_shm_info(struct ipc_namespace *ns, |
941 | struct shm_info *shm_info) | |
942 | { | |
943 | int err = security_shm_shmctl(NULL, SHM_INFO); | |
944 | if (!err) { | |
945 | memset(shm_info, 0, sizeof(*shm_info)); | |
d9a605e4 | 946 | down_read(&shm_ids(ns).rwsem); |
9ba720c1 AV |
947 | shm_info->used_ids = shm_ids(ns).in_use; |
948 | shm_get_stat(ns, &shm_info->shm_rss, &shm_info->shm_swp); | |
949 | shm_info->shm_tot = ns->shm_tot; | |
950 | shm_info->swap_attempts = 0; | |
951 | shm_info->swap_successes = 0; | |
7ca7e564 | 952 | err = ipc_get_maxid(&shm_ids(ns)); |
d9a605e4 | 953 | up_read(&shm_ids(ns).rwsem); |
9ba720c1 AV |
954 | if (err < 0) |
955 | err = 0; | |
1da177e4 | 956 | } |
9ba720c1 AV |
957 | return err; |
958 | } | |
c97cb9cc | 959 | |
9ba720c1 AV |
960 | static int shmctl_stat(struct ipc_namespace *ns, int shmid, |
961 | int cmd, struct shmid64_ds *tbuf) | |
962 | { | |
963 | struct shmid_kernel *shp; | |
87ad4b0d | 964 | int id = 0; |
9ba720c1 | 965 | int err; |
c97cb9cc | 966 | |
87ad4b0d PM |
967 | memset(tbuf, 0, sizeof(*tbuf)); |
968 | ||
9ba720c1 | 969 | rcu_read_lock(); |
c21a6970 | 970 | if (cmd == SHM_STAT || cmd == SHM_STAT_ANY) { |
9ba720c1 AV |
971 | shp = shm_obtain_object(ns, shmid); |
972 | if (IS_ERR(shp)) { | |
973 | err = PTR_ERR(shp); | |
974 | goto out_unlock; | |
975 | } | |
87ad4b0d | 976 | id = shp->shm_perm.id; |
c21a6970 | 977 | } else { /* IPC_STAT */ |
9ba720c1 AV |
978 | shp = shm_obtain_object_check(ns, shmid); |
979 | if (IS_ERR(shp)) { | |
980 | err = PTR_ERR(shp); | |
1da177e4 | 981 | goto out_unlock; |
1da177e4 | 982 | } |
9ba720c1 | 983 | } |
1da177e4 | 984 | |
c21a6970 DB |
985 | /* |
986 | * Semantically SHM_STAT_ANY ought to be identical to | |
987 | * that functionality provided by the /proc/sysvipc/ | |
988 | * interface. As such, only audit these calls and | |
989 | * do not do traditional S_IRUGO permission checks on | |
990 | * the ipc object. | |
991 | */ | |
992 | if (cmd == SHM_STAT_ANY) | |
993 | audit_ipc_obj(&shp->shm_perm); | |
994 | else { | |
995 | err = -EACCES; | |
996 | if (ipcperms(ns, &shp->shm_perm, S_IRUGO)) | |
997 | goto out_unlock; | |
998 | } | |
c97cb9cc | 999 | |
7191adff | 1000 | err = security_shm_shmctl(&shp->shm_perm, cmd); |
9ba720c1 AV |
1001 | if (err) |
1002 | goto out_unlock; | |
1003 | ||
87ad4b0d PM |
1004 | ipc_lock_object(&shp->shm_perm); |
1005 | ||
1006 | if (!ipc_valid_object(&shp->shm_perm)) { | |
1007 | ipc_unlock_object(&shp->shm_perm); | |
1008 | err = -EIDRM; | |
1009 | goto out_unlock; | |
1010 | } | |
1011 | ||
9ba720c1 AV |
1012 | kernel_to_ipc64_perm(&shp->shm_perm, &tbuf->shm_perm); |
1013 | tbuf->shm_segsz = shp->shm_segsz; | |
1014 | tbuf->shm_atime = shp->shm_atim; | |
1015 | tbuf->shm_dtime = shp->shm_dtim; | |
1016 | tbuf->shm_ctime = shp->shm_ctim; | |
c2ab975c AB |
1017 | #ifndef CONFIG_64BIT |
1018 | tbuf->shm_atime_high = shp->shm_atim >> 32; | |
1019 | tbuf->shm_dtime_high = shp->shm_dtim >> 32; | |
1020 | tbuf->shm_ctime_high = shp->shm_ctim >> 32; | |
1021 | #endif | |
98f929b1 EB |
1022 | tbuf->shm_cpid = pid_vnr(shp->shm_cprid); |
1023 | tbuf->shm_lpid = pid_vnr(shp->shm_lprid); | |
9ba720c1 | 1024 | tbuf->shm_nattch = shp->shm_nattch; |
87ad4b0d PM |
1025 | |
1026 | ipc_unlock_object(&shp->shm_perm); | |
9ba720c1 | 1027 | rcu_read_unlock(); |
87ad4b0d | 1028 | return id; |
68eccc1d DB |
1029 | |
1030 | out_unlock: | |
c97cb9cc | 1031 | rcu_read_unlock(); |
68eccc1d DB |
1032 | return err; |
1033 | } | |
1034 | ||
9ba720c1 | 1035 | static int shmctl_do_lock(struct ipc_namespace *ns, int shmid, int cmd) |
68eccc1d DB |
1036 | { |
1037 | struct shmid_kernel *shp; | |
9ba720c1 AV |
1038 | struct file *shm_file; |
1039 | int err; | |
68eccc1d | 1040 | |
9ba720c1 AV |
1041 | rcu_read_lock(); |
1042 | shp = shm_obtain_object_check(ns, shmid); | |
1043 | if (IS_ERR(shp)) { | |
1044 | err = PTR_ERR(shp); | |
1045 | goto out_unlock1; | |
1da177e4 | 1046 | } |
c97cb9cc | 1047 | |
9ba720c1 | 1048 | audit_ipc_obj(&(shp->shm_perm)); |
7191adff | 1049 | err = security_shm_shmctl(&shp->shm_perm, cmd); |
9ba720c1 AV |
1050 | if (err) |
1051 | goto out_unlock1; | |
c97cb9cc | 1052 | |
9ba720c1 | 1053 | ipc_lock_object(&shp->shm_perm); |
c97cb9cc | 1054 | |
9ba720c1 AV |
1055 | /* check if shm_destroy() is tearing down shp */ |
1056 | if (!ipc_valid_object(&shp->shm_perm)) { | |
1057 | err = -EIDRM; | |
1058 | goto out_unlock0; | |
1da177e4 | 1059 | } |
073115d6 | 1060 | |
9ba720c1 AV |
1061 | if (!ns_capable(ns->user_ns, CAP_IPC_LOCK)) { |
1062 | kuid_t euid = current_euid(); | |
0f3d2b01 | 1063 | |
9ba720c1 AV |
1064 | if (!uid_eq(euid, shp->shm_perm.uid) && |
1065 | !uid_eq(euid, shp->shm_perm.cuid)) { | |
1066 | err = -EPERM; | |
0f3d2b01 RA |
1067 | goto out_unlock0; |
1068 | } | |
9ba720c1 AV |
1069 | if (cmd == SHM_LOCK && !rlimit(RLIMIT_MEMLOCK)) { |
1070 | err = -EPERM; | |
1071 | goto out_unlock0; | |
1da177e4 | 1072 | } |
68eccc1d DB |
1073 | } |
1074 | ||
9ba720c1 AV |
1075 | shm_file = shp->shm_file; |
1076 | if (is_file_hugepages(shm_file)) | |
1077 | goto out_unlock0; | |
85046579 | 1078 | |
9ba720c1 AV |
1079 | if (cmd == SHM_LOCK) { |
1080 | struct user_struct *user = current_user(); | |
63980c80 | 1081 | |
9ba720c1 AV |
1082 | err = shmem_lock(shm_file, 1, user); |
1083 | if (!err && !(shp->shm_perm.mode & SHM_LOCKED)) { | |
1084 | shp->shm_perm.mode |= SHM_LOCKED; | |
1085 | shp->mlock_user = user; | |
1da177e4 | 1086 | } |
9ba720c1 | 1087 | goto out_unlock0; |
1da177e4 LT |
1088 | } |
1089 | ||
9ba720c1 AV |
1090 | /* SHM_UNLOCK */ |
1091 | if (!(shp->shm_perm.mode & SHM_LOCKED)) | |
1092 | goto out_unlock0; | |
1093 | shmem_lock(shm_file, 0, shp->mlock_user); | |
1094 | shp->shm_perm.mode &= ~SHM_LOCKED; | |
1095 | shp->mlock_user = NULL; | |
1096 | get_file(shm_file); | |
1097 | ipc_unlock_object(&shp->shm_perm); | |
1098 | rcu_read_unlock(); | |
1099 | shmem_unlock_mapping(shm_file->f_mapping); | |
1100 | ||
1101 | fput(shm_file); | |
1102 | return err; | |
1103 | ||
2caacaa8 DB |
1104 | out_unlock0: |
1105 | ipc_unlock_object(&shp->shm_perm); | |
1106 | out_unlock1: | |
c97cb9cc | 1107 | rcu_read_unlock(); |
68eccc1d DB |
1108 | return err; |
1109 | } | |
1110 | ||
c84d0791 | 1111 | long ksys_shmctl(int shmid, int cmd, struct shmid_ds __user *buf) |
68eccc1d | 1112 | { |
68eccc1d DB |
1113 | int err, version; |
1114 | struct ipc_namespace *ns; | |
553f770e | 1115 | struct shmid64_ds sem64; |
68eccc1d | 1116 | |
2caacaa8 DB |
1117 | if (cmd < 0 || shmid < 0) |
1118 | return -EINVAL; | |
68eccc1d DB |
1119 | |
1120 | version = ipc_parse_version(&cmd); | |
1121 | ns = current->nsproxy->ipc_ns; | |
1122 | ||
1123 | switch (cmd) { | |
9ba720c1 AV |
1124 | case IPC_INFO: { |
1125 | struct shminfo64 shminfo; | |
1126 | err = shmctl_ipc_info(ns, &shminfo); | |
1127 | if (err < 0) | |
1128 | return err; | |
1129 | if (copy_shminfo_to_user(buf, &shminfo, version)) | |
1130 | err = -EFAULT; | |
1131 | return err; | |
1132 | } | |
1133 | case SHM_INFO: { | |
1134 | struct shm_info shm_info; | |
1135 | err = shmctl_shm_info(ns, &shm_info); | |
1136 | if (err < 0) | |
1137 | return err; | |
1138 | if (copy_to_user(buf, &shm_info, sizeof(shm_info))) | |
1139 | err = -EFAULT; | |
1140 | return err; | |
1141 | } | |
68eccc1d | 1142 | case SHM_STAT: |
c21a6970 | 1143 | case SHM_STAT_ANY: |
9ba720c1 | 1144 | case IPC_STAT: { |
553f770e | 1145 | err = shmctl_stat(ns, shmid, cmd, &sem64); |
9ba720c1 AV |
1146 | if (err < 0) |
1147 | return err; | |
553f770e | 1148 | if (copy_shmid_to_user(buf, &sem64, version)) |
9ba720c1 AV |
1149 | err = -EFAULT; |
1150 | return err; | |
1151 | } | |
2caacaa8 | 1152 | case IPC_SET: |
553f770e | 1153 | if (copy_shmid_from_user(&sem64, buf, version)) |
9ba720c1 | 1154 | return -EFAULT; |
553f770e | 1155 | /* fallthru */ |
9ba720c1 | 1156 | case IPC_RMID: |
553f770e | 1157 | return shmctl_down(ns, shmid, cmd, &sem64); |
1da177e4 LT |
1158 | case SHM_LOCK: |
1159 | case SHM_UNLOCK: | |
9ba720c1 AV |
1160 | return shmctl_do_lock(ns, shmid, cmd); |
1161 | default: | |
1162 | return -EINVAL; | |
1163 | } | |
1164 | } | |
89e004ea | 1165 | |
c84d0791 DB |
1166 | SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, struct shmid_ds __user *, buf) |
1167 | { | |
1168 | return ksys_shmctl(shmid, cmd, buf); | |
1169 | } | |
1170 | ||
553f770e AV |
1171 | #ifdef CONFIG_COMPAT |
1172 | ||
1173 | struct compat_shmid_ds { | |
1174 | struct compat_ipc_perm shm_perm; | |
1175 | int shm_segsz; | |
1176 | compat_time_t shm_atime; | |
1177 | compat_time_t shm_dtime; | |
1178 | compat_time_t shm_ctime; | |
1179 | compat_ipc_pid_t shm_cpid; | |
1180 | compat_ipc_pid_t shm_lpid; | |
1181 | unsigned short shm_nattch; | |
1182 | unsigned short shm_unused; | |
1183 | compat_uptr_t shm_unused2; | |
1184 | compat_uptr_t shm_unused3; | |
1185 | }; | |
1da177e4 | 1186 | |
553f770e AV |
1187 | struct compat_shminfo64 { |
1188 | compat_ulong_t shmmax; | |
1189 | compat_ulong_t shmmin; | |
1190 | compat_ulong_t shmmni; | |
1191 | compat_ulong_t shmseg; | |
1192 | compat_ulong_t shmall; | |
1193 | compat_ulong_t __unused1; | |
1194 | compat_ulong_t __unused2; | |
1195 | compat_ulong_t __unused3; | |
1196 | compat_ulong_t __unused4; | |
1197 | }; | |
073115d6 | 1198 | |
553f770e AV |
1199 | struct compat_shm_info { |
1200 | compat_int_t used_ids; | |
1201 | compat_ulong_t shm_tot, shm_rss, shm_swp; | |
1202 | compat_ulong_t swap_attempts, swap_successes; | |
1203 | }; | |
0f3d2b01 | 1204 | |
553f770e AV |
1205 | static int copy_compat_shminfo_to_user(void __user *buf, struct shminfo64 *in, |
1206 | int version) | |
1207 | { | |
1208 | if (in->shmmax > INT_MAX) | |
1209 | in->shmmax = INT_MAX; | |
1210 | if (version == IPC_64) { | |
1211 | struct compat_shminfo64 info; | |
1212 | memset(&info, 0, sizeof(info)); | |
1213 | info.shmmax = in->shmmax; | |
1214 | info.shmmin = in->shmmin; | |
1215 | info.shmmni = in->shmmni; | |
1216 | info.shmseg = in->shmseg; | |
1217 | info.shmall = in->shmall; | |
1218 | return copy_to_user(buf, &info, sizeof(info)); | |
1219 | } else { | |
1220 | struct shminfo info; | |
1221 | memset(&info, 0, sizeof(info)); | |
1222 | info.shmmax = in->shmmax; | |
1223 | info.shmmin = in->shmmin; | |
1224 | info.shmmni = in->shmmni; | |
1225 | info.shmseg = in->shmseg; | |
1226 | info.shmall = in->shmall; | |
1227 | return copy_to_user(buf, &info, sizeof(info)); | |
1228 | } | |
1229 | } | |
0f3d2b01 | 1230 | |
553f770e AV |
1231 | static int put_compat_shm_info(struct shm_info *ip, |
1232 | struct compat_shm_info __user *uip) | |
1233 | { | |
1234 | struct compat_shm_info info; | |
1235 | ||
1236 | memset(&info, 0, sizeof(info)); | |
1237 | info.used_ids = ip->used_ids; | |
1238 | info.shm_tot = ip->shm_tot; | |
1239 | info.shm_rss = ip->shm_rss; | |
1240 | info.shm_swp = ip->shm_swp; | |
1241 | info.swap_attempts = ip->swap_attempts; | |
1242 | info.swap_successes = ip->swap_successes; | |
b776e4b1 | 1243 | return copy_to_user(uip, &info, sizeof(info)); |
553f770e | 1244 | } |
1da177e4 | 1245 | |
553f770e AV |
1246 | static int copy_compat_shmid_to_user(void __user *buf, struct shmid64_ds *in, |
1247 | int version) | |
1248 | { | |
1249 | if (version == IPC_64) { | |
1250 | struct compat_shmid64_ds v; | |
1251 | memset(&v, 0, sizeof(v)); | |
28327fae | 1252 | to_compat_ipc64_perm(&v.shm_perm, &in->shm_perm); |
c2ab975c AB |
1253 | v.shm_atime = lower_32_bits(in->shm_atime); |
1254 | v.shm_atime_high = upper_32_bits(in->shm_atime); | |
1255 | v.shm_dtime = lower_32_bits(in->shm_dtime); | |
1256 | v.shm_dtime_high = upper_32_bits(in->shm_dtime); | |
1257 | v.shm_ctime = lower_32_bits(in->shm_ctime); | |
1258 | v.shm_ctime_high = upper_32_bits(in->shm_ctime); | |
553f770e AV |
1259 | v.shm_segsz = in->shm_segsz; |
1260 | v.shm_nattch = in->shm_nattch; | |
1261 | v.shm_cpid = in->shm_cpid; | |
1262 | v.shm_lpid = in->shm_lpid; | |
1263 | return copy_to_user(buf, &v, sizeof(v)); | |
1264 | } else { | |
1265 | struct compat_shmid_ds v; | |
1266 | memset(&v, 0, sizeof(v)); | |
28327fae | 1267 | to_compat_ipc_perm(&v.shm_perm, &in->shm_perm); |
553f770e | 1268 | v.shm_perm.key = in->shm_perm.key; |
553f770e AV |
1269 | v.shm_atime = in->shm_atime; |
1270 | v.shm_dtime = in->shm_dtime; | |
1271 | v.shm_ctime = in->shm_ctime; | |
1272 | v.shm_segsz = in->shm_segsz; | |
1273 | v.shm_nattch = in->shm_nattch; | |
1274 | v.shm_cpid = in->shm_cpid; | |
1275 | v.shm_lpid = in->shm_lpid; | |
1276 | return copy_to_user(buf, &v, sizeof(v)); | |
1277 | } | |
1278 | } | |
85046579 | 1279 | |
553f770e AV |
1280 | static int copy_compat_shmid_from_user(struct shmid64_ds *out, void __user *buf, |
1281 | int version) | |
1282 | { | |
1283 | memset(out, 0, sizeof(*out)); | |
1284 | if (version == IPC_64) { | |
6aa211e8 | 1285 | struct compat_shmid64_ds __user *p = buf; |
28327fae | 1286 | return get_compat_ipc64_perm(&out->shm_perm, &p->shm_perm); |
553f770e | 1287 | } else { |
6aa211e8 | 1288 | struct compat_shmid_ds __user *p = buf; |
28327fae | 1289 | return get_compat_ipc_perm(&out->shm_perm, &p->shm_perm); |
553f770e | 1290 | } |
553f770e | 1291 | } |
63980c80 | 1292 | |
c84d0791 | 1293 | long compat_ksys_shmctl(int shmid, int cmd, void __user *uptr) |
553f770e AV |
1294 | { |
1295 | struct ipc_namespace *ns; | |
1296 | struct shmid64_ds sem64; | |
1297 | int version = compat_ipc_parse_version(&cmd); | |
1298 | int err; | |
85046579 | 1299 | |
553f770e AV |
1300 | ns = current->nsproxy->ipc_ns; |
1301 | ||
1302 | if (cmd < 0 || shmid < 0) | |
1303 | return -EINVAL; | |
2caacaa8 | 1304 | |
553f770e AV |
1305 | switch (cmd) { |
1306 | case IPC_INFO: { | |
1307 | struct shminfo64 shminfo; | |
1308 | err = shmctl_ipc_info(ns, &shminfo); | |
1309 | if (err < 0) | |
1310 | return err; | |
1311 | if (copy_compat_shminfo_to_user(uptr, &shminfo, version)) | |
1312 | err = -EFAULT; | |
1313 | return err; | |
1314 | } | |
1315 | case SHM_INFO: { | |
1316 | struct shm_info shm_info; | |
1317 | err = shmctl_shm_info(ns, &shm_info); | |
1318 | if (err < 0) | |
1319 | return err; | |
1320 | if (put_compat_shm_info(&shm_info, uptr)) | |
1321 | err = -EFAULT; | |
8d4cc8b5 | 1322 | return err; |
2caacaa8 | 1323 | } |
553f770e | 1324 | case IPC_STAT: |
c21a6970 | 1325 | case SHM_STAT_ANY: |
553f770e AV |
1326 | case SHM_STAT: |
1327 | err = shmctl_stat(ns, shmid, cmd, &sem64); | |
1328 | if (err < 0) | |
1329 | return err; | |
58aff0af | 1330 | if (copy_compat_shmid_to_user(uptr, &sem64, version)) |
553f770e AV |
1331 | err = -EFAULT; |
1332 | return err; | |
1333 | ||
1334 | case IPC_SET: | |
1335 | if (copy_compat_shmid_from_user(&sem64, uptr, version)) | |
1336 | return -EFAULT; | |
1337 | /* fallthru */ | |
1338 | case IPC_RMID: | |
1339 | return shmctl_down(ns, shmid, cmd, &sem64); | |
1340 | case SHM_LOCK: | |
1341 | case SHM_UNLOCK: | |
1342 | return shmctl_do_lock(ns, shmid, cmd); | |
1343 | break; | |
1da177e4 | 1344 | default: |
8d4cc8b5 | 1345 | return -EINVAL; |
1da177e4 | 1346 | } |
1da177e4 LT |
1347 | return err; |
1348 | } | |
c84d0791 DB |
1349 | |
1350 | COMPAT_SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, void __user *, uptr) | |
1351 | { | |
1352 | return compat_ksys_shmctl(shmid, cmd, uptr); | |
1353 | } | |
553f770e | 1354 | #endif |
1da177e4 LT |
1355 | |
1356 | /* | |
1357 | * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists. | |
1358 | * | |
1359 | * NOTE! Despite the name, this is NOT a direct system call entrypoint. The | |
1360 | * "raddr" thing points to kernel space, and there has to be a wrapper around | |
1361 | * this. | |
1362 | */ | |
95e91b83 DB |
1363 | long do_shmat(int shmid, char __user *shmaddr, int shmflg, |
1364 | ulong *raddr, unsigned long shmlba) | |
1da177e4 LT |
1365 | { |
1366 | struct shmid_kernel *shp; | |
f0cb8802 | 1367 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1368 | unsigned long size; |
239521f3 | 1369 | struct file *file; |
1da177e4 | 1370 | int err; |
f0cb8802 | 1371 | unsigned long flags = MAP_SHARED; |
1da177e4 | 1372 | unsigned long prot; |
1da177e4 | 1373 | int acc_mode; |
4e982311 | 1374 | struct ipc_namespace *ns; |
bc56bba8 EB |
1375 | struct shm_file_data *sfd; |
1376 | struct path path; | |
aeb5d727 | 1377 | fmode_t f_mode; |
41badc15 | 1378 | unsigned long populate = 0; |
1da177e4 | 1379 | |
bc56bba8 EB |
1380 | err = -EINVAL; |
1381 | if (shmid < 0) | |
1da177e4 | 1382 | goto out; |
f0cb8802 DB |
1383 | |
1384 | if (addr) { | |
079a96ae | 1385 | if (addr & (shmlba - 1)) { |
8f89c007 | 1386 | if (shmflg & SHM_RND) { |
a73ab244 | 1387 | addr &= ~(shmlba - 1); /* round down */ |
8f89c007 DB |
1388 | |
1389 | /* | |
1390 | * Ensure that the round-down is non-nil | |
1391 | * when remapping. This can happen for | |
1392 | * cases when addr < shmlba. | |
1393 | */ | |
1394 | if (!addr && (shmflg & SHM_REMAP)) | |
1395 | goto out; | |
1396 | } else | |
1da177e4 LT |
1397 | #ifndef __ARCH_FORCE_SHMLBA |
1398 | if (addr & ~PAGE_MASK) | |
1399 | #endif | |
bc56bba8 | 1400 | goto out; |
1da177e4 | 1401 | } |
1da177e4 | 1402 | |
f0cb8802 DB |
1403 | flags |= MAP_FIXED; |
1404 | } else if ((shmflg & SHM_REMAP)) | |
1405 | goto out; | |
1da177e4 LT |
1406 | |
1407 | if (shmflg & SHM_RDONLY) { | |
1408 | prot = PROT_READ; | |
1da177e4 | 1409 | acc_mode = S_IRUGO; |
bc56bba8 | 1410 | f_mode = FMODE_READ; |
1da177e4 LT |
1411 | } else { |
1412 | prot = PROT_READ | PROT_WRITE; | |
1da177e4 | 1413 | acc_mode = S_IRUGO | S_IWUGO; |
bc56bba8 | 1414 | f_mode = FMODE_READ | FMODE_WRITE; |
1da177e4 LT |
1415 | } |
1416 | if (shmflg & SHM_EXEC) { | |
1417 | prot |= PROT_EXEC; | |
1418 | acc_mode |= S_IXUGO; | |
1419 | } | |
1420 | ||
1421 | /* | |
1422 | * We cannot rely on the fs check since SYSV IPC does have an | |
1423 | * additional creator id... | |
1424 | */ | |
4e982311 | 1425 | ns = current->nsproxy->ipc_ns; |
c2c737a0 DB |
1426 | rcu_read_lock(); |
1427 | shp = shm_obtain_object_check(ns, shmid); | |
023a5355 ND |
1428 | if (IS_ERR(shp)) { |
1429 | err = PTR_ERR(shp); | |
c2c737a0 | 1430 | goto out_unlock; |
023a5355 | 1431 | } |
bc56bba8 EB |
1432 | |
1433 | err = -EACCES; | |
b0e77598 | 1434 | if (ipcperms(ns, &shp->shm_perm, acc_mode)) |
bc56bba8 | 1435 | goto out_unlock; |
1da177e4 | 1436 | |
7191adff | 1437 | err = security_shm_shmat(&shp->shm_perm, shmaddr, shmflg); |
bc56bba8 EB |
1438 | if (err) |
1439 | goto out_unlock; | |
1440 | ||
c2c737a0 | 1441 | ipc_lock_object(&shp->shm_perm); |
a399b29d GT |
1442 | |
1443 | /* check if shm_destroy() is tearing down shp */ | |
0f3d2b01 | 1444 | if (!ipc_valid_object(&shp->shm_perm)) { |
a399b29d GT |
1445 | ipc_unlock_object(&shp->shm_perm); |
1446 | err = -EIDRM; | |
1447 | goto out_unlock; | |
1448 | } | |
1449 | ||
2c48b9c4 AV |
1450 | path = shp->shm_file->f_path; |
1451 | path_get(&path); | |
1da177e4 | 1452 | shp->shm_nattch++; |
75c3cfa8 | 1453 | size = i_size_read(d_inode(path.dentry)); |
c2c737a0 DB |
1454 | ipc_unlock_object(&shp->shm_perm); |
1455 | rcu_read_unlock(); | |
1da177e4 | 1456 | |
bc56bba8 EB |
1457 | err = -ENOMEM; |
1458 | sfd = kzalloc(sizeof(*sfd), GFP_KERNEL); | |
f42569b1 DB |
1459 | if (!sfd) { |
1460 | path_put(&path); | |
1461 | goto out_nattch; | |
1462 | } | |
bc56bba8 | 1463 | |
2c48b9c4 AV |
1464 | file = alloc_file(&path, f_mode, |
1465 | is_file_hugepages(shp->shm_file) ? | |
c4caa778 AV |
1466 | &shm_file_operations_huge : |
1467 | &shm_file_operations); | |
39b65252 | 1468 | err = PTR_ERR(file); |
f42569b1 DB |
1469 | if (IS_ERR(file)) { |
1470 | kfree(sfd); | |
1471 | path_put(&path); | |
1472 | goto out_nattch; | |
1473 | } | |
bc56bba8 | 1474 | |
bc56bba8 | 1475 | file->private_data = sfd; |
bc56bba8 | 1476 | file->f_mapping = shp->shm_file->f_mapping; |
7ca7e564 | 1477 | sfd->id = shp->shm_perm.id; |
bc56bba8 | 1478 | sfd->ns = get_ipc_ns(ns); |
3f05317d EB |
1479 | /* |
1480 | * We need to take a reference to the real shm file to prevent the | |
1481 | * pointer from becoming stale in cases where the lifetime of the outer | |
1482 | * file extends beyond that of the shm segment. It's not usually | |
1483 | * possible, but it can happen during remap_file_pages() emulation as | |
1484 | * that unmaps the memory, then does ->mmap() via file reference only. | |
1485 | * We'll deny the ->mmap() if the shm segment was since removed, but to | |
1486 | * detect shm ID reuse we need to compare the file pointers. | |
1487 | */ | |
1488 | sfd->file = get_file(shp->shm_file); | |
bc56bba8 EB |
1489 | sfd->vm_ops = NULL; |
1490 | ||
8b3ec681 AV |
1491 | err = security_mmap_file(file, prot, flags); |
1492 | if (err) | |
1493 | goto out_fput; | |
1494 | ||
91f4f94e MH |
1495 | if (down_write_killable(¤t->mm->mmap_sem)) { |
1496 | err = -EINTR; | |
1497 | goto out_fput; | |
1498 | } | |
1499 | ||
1da177e4 | 1500 | if (addr && !(shmflg & SHM_REMAP)) { |
bc56bba8 | 1501 | err = -EINVAL; |
247a8ce8 MS |
1502 | if (addr + size < addr) |
1503 | goto invalid; | |
1504 | ||
1da177e4 LT |
1505 | if (find_vma_intersection(current->mm, addr, addr + size)) |
1506 | goto invalid; | |
1da177e4 | 1507 | } |
f42569b1 | 1508 | |
897ab3e0 | 1509 | addr = do_mmap_pgoff(file, addr, size, prot, flags, 0, &populate, NULL); |
bebeb3d6 | 1510 | *raddr = addr; |
bc56bba8 | 1511 | err = 0; |
bebeb3d6 ML |
1512 | if (IS_ERR_VALUE(addr)) |
1513 | err = (long)addr; | |
1da177e4 LT |
1514 | invalid: |
1515 | up_write(¤t->mm->mmap_sem); | |
bebeb3d6 | 1516 | if (populate) |
41badc15 | 1517 | mm_populate(addr, populate); |
1da177e4 | 1518 | |
8b3ec681 | 1519 | out_fput: |
bc56bba8 EB |
1520 | fput(file); |
1521 | ||
1522 | out_nattch: | |
d9a605e4 | 1523 | down_write(&shm_ids(ns).rwsem); |
00c2bf85 | 1524 | shp = shm_lock(ns, shmid); |
1da177e4 | 1525 | shp->shm_nattch--; |
b34a6b1d | 1526 | if (shm_may_destroy(ns, shp)) |
4e982311 | 1527 | shm_destroy(ns, shp); |
1da177e4 LT |
1528 | else |
1529 | shm_unlock(shp); | |
d9a605e4 | 1530 | up_write(&shm_ids(ns).rwsem); |
1da177e4 | 1531 | return err; |
bc56bba8 EB |
1532 | |
1533 | out_unlock: | |
c2c737a0 | 1534 | rcu_read_unlock(); |
f42569b1 DB |
1535 | out: |
1536 | return err; | |
1da177e4 LT |
1537 | } |
1538 | ||
d5460c99 | 1539 | SYSCALL_DEFINE3(shmat, int, shmid, char __user *, shmaddr, int, shmflg) |
7d87e14c SR |
1540 | { |
1541 | unsigned long ret; | |
1542 | long err; | |
1543 | ||
079a96ae | 1544 | err = do_shmat(shmid, shmaddr, shmflg, &ret, SHMLBA); |
7d87e14c SR |
1545 | if (err) |
1546 | return err; | |
1547 | force_successful_syscall_return(); | |
1548 | return (long)ret; | |
1549 | } | |
1550 | ||
a78ee9ed AV |
1551 | #ifdef CONFIG_COMPAT |
1552 | ||
1553 | #ifndef COMPAT_SHMLBA | |
1554 | #define COMPAT_SHMLBA SHMLBA | |
1555 | #endif | |
1556 | ||
1557 | COMPAT_SYSCALL_DEFINE3(shmat, int, shmid, compat_uptr_t, shmaddr, int, shmflg) | |
1558 | { | |
1559 | unsigned long ret; | |
1560 | long err; | |
1561 | ||
1562 | err = do_shmat(shmid, compat_ptr(shmaddr), shmflg, &ret, COMPAT_SHMLBA); | |
1563 | if (err) | |
1564 | return err; | |
1565 | force_successful_syscall_return(); | |
1566 | return (long)ret; | |
1567 | } | |
1568 | #endif | |
1569 | ||
1da177e4 LT |
1570 | /* |
1571 | * detach and kill segment if marked destroyed. | |
1572 | * The work is done in shm_close. | |
1573 | */ | |
da1e2744 | 1574 | long ksys_shmdt(char __user *shmaddr) |
1da177e4 LT |
1575 | { |
1576 | struct mm_struct *mm = current->mm; | |
586c7e6a | 1577 | struct vm_area_struct *vma; |
1da177e4 | 1578 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1579 | int retval = -EINVAL; |
586c7e6a MF |
1580 | #ifdef CONFIG_MMU |
1581 | loff_t size = 0; | |
d3c97900 | 1582 | struct file *file; |
586c7e6a MF |
1583 | struct vm_area_struct *next; |
1584 | #endif | |
1da177e4 | 1585 | |
df1e2fb5 HD |
1586 | if (addr & ~PAGE_MASK) |
1587 | return retval; | |
1588 | ||
91f4f94e MH |
1589 | if (down_write_killable(&mm->mmap_sem)) |
1590 | return -EINTR; | |
1da177e4 LT |
1591 | |
1592 | /* | |
1593 | * This function tries to be smart and unmap shm segments that | |
1594 | * were modified by partial mlock or munmap calls: | |
1595 | * - It first determines the size of the shm segment that should be | |
1596 | * unmapped: It searches for a vma that is backed by shm and that | |
1597 | * started at address shmaddr. It records it's size and then unmaps | |
1598 | * it. | |
1599 | * - Then it unmaps all shm vmas that started at shmaddr and that | |
d3c97900 DH |
1600 | * are within the initially determined size and that are from the |
1601 | * same shm segment from which we determined the size. | |
1da177e4 LT |
1602 | * Errors from do_munmap are ignored: the function only fails if |
1603 | * it's called with invalid parameters or if it's called to unmap | |
1604 | * a part of a vma. Both calls in this function are for full vmas, | |
1605 | * the parameters are directly copied from the vma itself and always | |
1606 | * valid - therefore do_munmap cannot fail. (famous last words?) | |
1607 | */ | |
1608 | /* | |
1609 | * If it had been mremap()'d, the starting address would not | |
1610 | * match the usual checks anyway. So assume all vma's are | |
1611 | * above the starting address given. | |
1612 | */ | |
1613 | vma = find_vma(mm, addr); | |
1614 | ||
8feae131 | 1615 | #ifdef CONFIG_MMU |
1da177e4 LT |
1616 | while (vma) { |
1617 | next = vma->vm_next; | |
1618 | ||
1619 | /* | |
1620 | * Check if the starting address would match, i.e. it's | |
1621 | * a fragment created by mprotect() and/or munmap(), or it | |
1622 | * otherwise it starts at this address with no hassles. | |
1623 | */ | |
bc56bba8 | 1624 | if ((vma->vm_ops == &shm_vm_ops) && |
1da177e4 LT |
1625 | (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) { |
1626 | ||
d3c97900 DH |
1627 | /* |
1628 | * Record the file of the shm segment being | |
1629 | * unmapped. With mremap(), someone could place | |
1630 | * page from another segment but with equal offsets | |
1631 | * in the range we are unmapping. | |
1632 | */ | |
1633 | file = vma->vm_file; | |
07a46ed2 | 1634 | size = i_size_read(file_inode(vma->vm_file)); |
897ab3e0 | 1635 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1636 | /* |
1637 | * We discovered the size of the shm segment, so | |
1638 | * break out of here and fall through to the next | |
1639 | * loop that uses the size information to stop | |
1640 | * searching for matching vma's. | |
1641 | */ | |
1642 | retval = 0; | |
1643 | vma = next; | |
1644 | break; | |
1645 | } | |
1646 | vma = next; | |
1647 | } | |
1648 | ||
1649 | /* | |
1650 | * We need look no further than the maximum address a fragment | |
1651 | * could possibly have landed at. Also cast things to loff_t to | |
25985edc | 1652 | * prevent overflows and make comparisons vs. equal-width types. |
1da177e4 | 1653 | */ |
8e36709d | 1654 | size = PAGE_ALIGN(size); |
1da177e4 LT |
1655 | while (vma && (loff_t)(vma->vm_end - addr) <= size) { |
1656 | next = vma->vm_next; | |
1657 | ||
1658 | /* finding a matching vma now does not alter retval */ | |
bc56bba8 | 1659 | if ((vma->vm_ops == &shm_vm_ops) && |
d3c97900 DH |
1660 | ((vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) && |
1661 | (vma->vm_file == file)) | |
897ab3e0 | 1662 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1663 | vma = next; |
1664 | } | |
1665 | ||
63980c80 | 1666 | #else /* CONFIG_MMU */ |
8feae131 | 1667 | /* under NOMMU conditions, the exact address to be destroyed must be |
63980c80 SP |
1668 | * given |
1669 | */ | |
530fcd16 | 1670 | if (vma && vma->vm_start == addr && vma->vm_ops == &shm_vm_ops) { |
897ab3e0 | 1671 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
8feae131 DH |
1672 | retval = 0; |
1673 | } | |
1674 | ||
1675 | #endif | |
1676 | ||
1da177e4 LT |
1677 | up_write(&mm->mmap_sem); |
1678 | return retval; | |
1679 | } | |
1680 | ||
da1e2744 DB |
1681 | SYSCALL_DEFINE1(shmdt, char __user *, shmaddr) |
1682 | { | |
1683 | return ksys_shmdt(shmaddr); | |
1684 | } | |
1685 | ||
1da177e4 | 1686 | #ifdef CONFIG_PROC_FS |
19b4946c | 1687 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it) |
1da177e4 | 1688 | { |
98f929b1 | 1689 | struct pid_namespace *pid_ns = ipc_seq_pid_ns(s); |
1efdb69b | 1690 | struct user_namespace *user_ns = seq_user_ns(s); |
ade9f91b KC |
1691 | struct kern_ipc_perm *ipcp = it; |
1692 | struct shmid_kernel *shp; | |
b7952180 HD |
1693 | unsigned long rss = 0, swp = 0; |
1694 | ||
ade9f91b | 1695 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
b7952180 | 1696 | shm_add_rss_swap(shp, &rss, &swp); |
1da177e4 | 1697 | |
6c826818 PM |
1698 | #if BITS_PER_LONG <= 32 |
1699 | #define SIZE_SPEC "%10lu" | |
1700 | #else | |
1701 | #define SIZE_SPEC "%21lu" | |
1702 | #endif | |
1da177e4 | 1703 | |
7f032d6e JP |
1704 | seq_printf(s, |
1705 | "%10d %10d %4o " SIZE_SPEC " %5u %5u " | |
7ff2819e | 1706 | "%5lu %5u %5u %5u %5u %10llu %10llu %10llu " |
7f032d6e JP |
1707 | SIZE_SPEC " " SIZE_SPEC "\n", |
1708 | shp->shm_perm.key, | |
1709 | shp->shm_perm.id, | |
1710 | shp->shm_perm.mode, | |
1711 | shp->shm_segsz, | |
98f929b1 EB |
1712 | pid_nr_ns(shp->shm_cprid, pid_ns), |
1713 | pid_nr_ns(shp->shm_lprid, pid_ns), | |
7f032d6e JP |
1714 | shp->shm_nattch, |
1715 | from_kuid_munged(user_ns, shp->shm_perm.uid), | |
1716 | from_kgid_munged(user_ns, shp->shm_perm.gid), | |
1717 | from_kuid_munged(user_ns, shp->shm_perm.cuid), | |
1718 | from_kgid_munged(user_ns, shp->shm_perm.cgid), | |
1719 | shp->shm_atim, | |
1720 | shp->shm_dtim, | |
1721 | shp->shm_ctim, | |
1722 | rss * PAGE_SIZE, | |
1723 | swp * PAGE_SIZE); | |
1724 | ||
1725 | return 0; | |
1da177e4 LT |
1726 | } |
1727 | #endif |