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; | |
60 | pid_t shm_cprid; | |
61 | pid_t shm_lprid; | |
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 | ||
7ff2819e | 228 | shp->shm_atim = ktime_get_real_seconds(); |
b488893a | 229 | shp->shm_lprid = task_tgid_vnr(current); |
1da177e4 LT |
230 | shp->shm_nattch++; |
231 | shm_unlock(shp); | |
1ac0b6de KS |
232 | return 0; |
233 | } | |
234 | ||
235 | /* This is called by fork, once for every shm attach. */ | |
236 | static void shm_open(struct vm_area_struct *vma) | |
237 | { | |
238 | int err = __shm_open(vma); | |
239 | /* | |
240 | * We raced in the idr lookup or with shm_destroy(). | |
241 | * Either way, the ID is busted. | |
242 | */ | |
243 | WARN_ON_ONCE(err); | |
1da177e4 LT |
244 | } |
245 | ||
1da177e4 LT |
246 | /* |
247 | * shm_destroy - free the struct shmid_kernel | |
248 | * | |
f4566f04 | 249 | * @ns: namespace |
1da177e4 LT |
250 | * @shp: struct to free |
251 | * | |
d9a605e4 | 252 | * It has to be called with shp and shm_ids.rwsem (writer) locked, |
1da177e4 LT |
253 | * but returns with shp unlocked and freed. |
254 | */ | |
4e982311 | 255 | static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) |
1da177e4 | 256 | { |
a399b29d GT |
257 | struct file *shm_file; |
258 | ||
259 | shm_file = shp->shm_file; | |
260 | shp->shm_file = NULL; | |
4e982311 | 261 | ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT; |
7ca7e564 | 262 | shm_rmid(ns, shp); |
1da177e4 | 263 | shm_unlock(shp); |
a399b29d GT |
264 | if (!is_file_hugepages(shm_file)) |
265 | shmem_lock(shm_file, 0, shp->mlock_user); | |
353d5c30 | 266 | else if (shp->mlock_user) |
07a46ed2 DH |
267 | user_shm_unlock(i_size_read(file_inode(shm_file)), |
268 | shp->mlock_user); | |
a399b29d | 269 | fput(shm_file); |
dba4cdd3 | 270 | ipc_rcu_putref(&shp->shm_perm, shm_rcu_free); |
1da177e4 LT |
271 | } |
272 | ||
b34a6b1d VK |
273 | /* |
274 | * shm_may_destroy - identifies whether shm segment should be destroyed now | |
275 | * | |
276 | * Returns true if and only if there are no active users of the segment and | |
277 | * one of the following is true: | |
278 | * | |
279 | * 1) shmctl(id, IPC_RMID, NULL) was called for this shp | |
280 | * | |
281 | * 2) sysctl kernel.shm_rmid_forced is set to 1. | |
282 | */ | |
283 | static bool shm_may_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) | |
284 | { | |
285 | return (shp->shm_nattch == 0) && | |
286 | (ns->shm_rmid_forced || | |
287 | (shp->shm_perm.mode & SHM_DEST)); | |
288 | } | |
289 | ||
1da177e4 | 290 | /* |
bc56bba8 | 291 | * remove the attach descriptor vma. |
1da177e4 LT |
292 | * free memory for segment if it is marked destroyed. |
293 | * The descriptor has already been removed from the current->mm->mmap list | |
294 | * and will later be kfree()d. | |
295 | */ | |
bc56bba8 | 296 | static void shm_close(struct vm_area_struct *vma) |
1da177e4 | 297 | { |
239521f3 | 298 | struct file *file = vma->vm_file; |
bc56bba8 | 299 | struct shm_file_data *sfd = shm_file_data(file); |
1da177e4 | 300 | struct shmid_kernel *shp; |
bc56bba8 | 301 | struct ipc_namespace *ns = sfd->ns; |
4e982311 | 302 | |
d9a605e4 | 303 | down_write(&shm_ids(ns).rwsem); |
1da177e4 | 304 | /* remove from the list of attaches of the shm segment */ |
00c2bf85 | 305 | shp = shm_lock(ns, sfd->id); |
1ac0b6de KS |
306 | |
307 | /* | |
308 | * We raced in the idr lookup or with shm_destroy(). | |
309 | * Either way, the ID is busted. | |
310 | */ | |
311 | if (WARN_ON_ONCE(IS_ERR(shp))) | |
312 | goto done; /* no-op */ | |
313 | ||
b488893a | 314 | shp->shm_lprid = task_tgid_vnr(current); |
7ff2819e | 315 | shp->shm_dtim = ktime_get_real_seconds(); |
1da177e4 | 316 | shp->shm_nattch--; |
b34a6b1d VK |
317 | if (shm_may_destroy(ns, shp)) |
318 | shm_destroy(ns, shp); | |
319 | else | |
320 | shm_unlock(shp); | |
1ac0b6de | 321 | done: |
d9a605e4 | 322 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
323 | } |
324 | ||
d9a605e4 | 325 | /* Called with ns->shm_ids(ns).rwsem locked */ |
b34a6b1d VK |
326 | static int shm_try_destroy_orphaned(int id, void *p, void *data) |
327 | { | |
328 | struct ipc_namespace *ns = data; | |
4c677e2e VK |
329 | struct kern_ipc_perm *ipcp = p; |
330 | struct shmid_kernel *shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
b34a6b1d VK |
331 | |
332 | /* | |
333 | * We want to destroy segments without users and with already | |
334 | * exit'ed originating process. | |
4c677e2e | 335 | * |
d9a605e4 | 336 | * As shp->* are changed under rwsem, it's safe to skip shp locking. |
b34a6b1d | 337 | */ |
4c677e2e | 338 | if (shp->shm_creator != NULL) |
b34a6b1d | 339 | return 0; |
b34a6b1d | 340 | |
4c677e2e VK |
341 | if (shm_may_destroy(ns, shp)) { |
342 | shm_lock_by_ptr(shp); | |
4e982311 | 343 | shm_destroy(ns, shp); |
4c677e2e | 344 | } |
b34a6b1d VK |
345 | return 0; |
346 | } | |
347 | ||
348 | void shm_destroy_orphaned(struct ipc_namespace *ns) | |
349 | { | |
d9a605e4 | 350 | down_write(&shm_ids(ns).rwsem); |
33a30ed4 | 351 | if (shm_ids(ns).in_use) |
4c677e2e | 352 | idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_orphaned, ns); |
d9a605e4 | 353 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
354 | } |
355 | ||
83293c0f | 356 | /* Locking assumes this will only be called with task == current */ |
b34a6b1d VK |
357 | void exit_shm(struct task_struct *task) |
358 | { | |
4c677e2e | 359 | struct ipc_namespace *ns = task->nsproxy->ipc_ns; |
ab602f79 | 360 | struct shmid_kernel *shp, *n; |
b34a6b1d | 361 | |
83293c0f JM |
362 | if (list_empty(&task->sysvshm.shm_clist)) |
363 | return; | |
364 | ||
365 | /* | |
366 | * If kernel.shm_rmid_forced is not set then only keep track of | |
367 | * which shmids are orphaned, so that a later set of the sysctl | |
368 | * can clean them up. | |
369 | */ | |
370 | if (!ns->shm_rmid_forced) { | |
371 | down_read(&shm_ids(ns).rwsem); | |
372 | list_for_each_entry(shp, &task->sysvshm.shm_clist, shm_clist) | |
373 | shp->shm_creator = NULL; | |
374 | /* | |
375 | * Only under read lock but we are only called on current | |
376 | * so no entry on the list will be shared. | |
377 | */ | |
378 | list_del(&task->sysvshm.shm_clist); | |
379 | up_read(&shm_ids(ns).rwsem); | |
298507d4 | 380 | return; |
83293c0f | 381 | } |
298507d4 | 382 | |
83293c0f JM |
383 | /* |
384 | * Destroy all already created segments, that were not yet mapped, | |
385 | * and mark any mapped as orphan to cover the sysctl toggling. | |
386 | * Destroy is skipped if shm_may_destroy() returns false. | |
387 | */ | |
d9a605e4 | 388 | down_write(&shm_ids(ns).rwsem); |
83293c0f JM |
389 | list_for_each_entry_safe(shp, n, &task->sysvshm.shm_clist, shm_clist) { |
390 | shp->shm_creator = NULL; | |
391 | ||
392 | if (shm_may_destroy(ns, shp)) { | |
393 | shm_lock_by_ptr(shp); | |
394 | shm_destroy(ns, shp); | |
395 | } | |
396 | } | |
397 | ||
398 | /* Remove the list head from any segments still attached. */ | |
ab602f79 | 399 | list_del(&task->sysvshm.shm_clist); |
d9a605e4 | 400 | up_write(&shm_ids(ns).rwsem); |
1da177e4 LT |
401 | } |
402 | ||
11bac800 | 403 | static int shm_fault(struct vm_fault *vmf) |
bc56bba8 | 404 | { |
11bac800 | 405 | struct file *file = vmf->vma->vm_file; |
bc56bba8 EB |
406 | struct shm_file_data *sfd = shm_file_data(file); |
407 | ||
11bac800 | 408 | return sfd->vm_ops->fault(vmf); |
bc56bba8 EB |
409 | } |
410 | ||
411 | #ifdef CONFIG_NUMA | |
d823e3e7 | 412 | static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new) |
bc56bba8 EB |
413 | { |
414 | struct file *file = vma->vm_file; | |
415 | struct shm_file_data *sfd = shm_file_data(file); | |
416 | int err = 0; | |
63980c80 | 417 | |
bc56bba8 EB |
418 | if (sfd->vm_ops->set_policy) |
419 | err = sfd->vm_ops->set_policy(vma, new); | |
420 | return err; | |
421 | } | |
422 | ||
d823e3e7 AB |
423 | static struct mempolicy *shm_get_policy(struct vm_area_struct *vma, |
424 | unsigned long addr) | |
bc56bba8 EB |
425 | { |
426 | struct file *file = vma->vm_file; | |
427 | struct shm_file_data *sfd = shm_file_data(file); | |
428 | struct mempolicy *pol = NULL; | |
429 | ||
430 | if (sfd->vm_ops->get_policy) | |
431 | pol = sfd->vm_ops->get_policy(vma, addr); | |
52cd3b07 | 432 | else if (vma->vm_policy) |
bc56bba8 | 433 | pol = vma->vm_policy; |
52cd3b07 | 434 | |
bc56bba8 EB |
435 | return pol; |
436 | } | |
437 | #endif | |
438 | ||
239521f3 | 439 | static int shm_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 440 | { |
bc56bba8 | 441 | struct shm_file_data *sfd = shm_file_data(file); |
b0e15190 DH |
442 | int ret; |
443 | ||
1ac0b6de KS |
444 | /* |
445 | * In case of remap_file_pages() emulation, the file can represent | |
446 | * removed IPC ID: propogate shm_lock() error to caller. | |
447 | */ | |
63980c80 | 448 | ret = __shm_open(vma); |
1ac0b6de KS |
449 | if (ret) |
450 | return ret; | |
451 | ||
f74ac015 | 452 | ret = call_mmap(sfd->file, vma); |
1ac0b6de KS |
453 | if (ret) { |
454 | shm_close(vma); | |
bc56bba8 | 455 | return ret; |
1ac0b6de | 456 | } |
bc56bba8 | 457 | sfd->vm_ops = vma->vm_ops; |
2e92a3ba | 458 | #ifdef CONFIG_MMU |
d0edd852 | 459 | WARN_ON(!sfd->vm_ops->fault); |
2e92a3ba | 460 | #endif |
bc56bba8 | 461 | vma->vm_ops = &shm_vm_ops; |
1ac0b6de | 462 | return 0; |
1da177e4 LT |
463 | } |
464 | ||
4e982311 KK |
465 | static int shm_release(struct inode *ino, struct file *file) |
466 | { | |
bc56bba8 | 467 | struct shm_file_data *sfd = shm_file_data(file); |
4e982311 | 468 | |
bc56bba8 EB |
469 | put_ipc_ns(sfd->ns); |
470 | shm_file_data(file) = NULL; | |
471 | kfree(sfd); | |
4e982311 KK |
472 | return 0; |
473 | } | |
474 | ||
02c24a82 | 475 | static int shm_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
516dffdc | 476 | { |
516dffdc | 477 | struct shm_file_data *sfd = shm_file_data(file); |
516dffdc | 478 | |
7ea80859 CH |
479 | if (!sfd->file->f_op->fsync) |
480 | return -EINVAL; | |
0f41074a | 481 | return sfd->file->f_op->fsync(sfd->file, start, end, datasync); |
516dffdc AL |
482 | } |
483 | ||
7d8a4569 WD |
484 | static long shm_fallocate(struct file *file, int mode, loff_t offset, |
485 | loff_t len) | |
486 | { | |
487 | struct shm_file_data *sfd = shm_file_data(file); | |
488 | ||
489 | if (!sfd->file->f_op->fallocate) | |
490 | return -EOPNOTSUPP; | |
491 | return sfd->file->f_op->fallocate(file, mode, offset, len); | |
492 | } | |
493 | ||
bc56bba8 EB |
494 | static unsigned long shm_get_unmapped_area(struct file *file, |
495 | unsigned long addr, unsigned long len, unsigned long pgoff, | |
496 | unsigned long flags) | |
497 | { | |
498 | struct shm_file_data *sfd = shm_file_data(file); | |
63980c80 | 499 | |
c4caa778 AV |
500 | return sfd->file->f_op->get_unmapped_area(sfd->file, addr, len, |
501 | pgoff, flags); | |
bc56bba8 | 502 | } |
bc56bba8 | 503 | |
9a32144e | 504 | static const struct file_operations shm_file_operations = { |
4e982311 | 505 | .mmap = shm_mmap, |
516dffdc | 506 | .fsync = shm_fsync, |
4e982311 | 507 | .release = shm_release, |
ed5e5894 | 508 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 509 | .llseek = noop_llseek, |
7d8a4569 | 510 | .fallocate = shm_fallocate, |
c4caa778 AV |
511 | }; |
512 | ||
c01d5b30 HD |
513 | /* |
514 | * shm_file_operations_huge is now identical to shm_file_operations, | |
515 | * but we keep it distinct for the sake of is_file_shm_hugepages(). | |
516 | */ | |
c4caa778 AV |
517 | static const struct file_operations shm_file_operations_huge = { |
518 | .mmap = shm_mmap, | |
519 | .fsync = shm_fsync, | |
520 | .release = shm_release, | |
bc56bba8 | 521 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 522 | .llseek = noop_llseek, |
7d8a4569 | 523 | .fallocate = shm_fallocate, |
1da177e4 LT |
524 | }; |
525 | ||
2954e440 | 526 | bool is_file_shm_hugepages(struct file *file) |
c4caa778 AV |
527 | { |
528 | return file->f_op == &shm_file_operations_huge; | |
529 | } | |
530 | ||
f0f37e2f | 531 | static const struct vm_operations_struct shm_vm_ops = { |
1da177e4 LT |
532 | .open = shm_open, /* callback for a new vm-area open */ |
533 | .close = shm_close, /* callback for when the vm-area is released */ | |
54cb8821 | 534 | .fault = shm_fault, |
bc56bba8 EB |
535 | #if defined(CONFIG_NUMA) |
536 | .set_policy = shm_set_policy, | |
537 | .get_policy = shm_get_policy, | |
1da177e4 LT |
538 | #endif |
539 | }; | |
540 | ||
f4566f04 ND |
541 | /** |
542 | * newseg - Create a new shared memory segment | |
543 | * @ns: namespace | |
544 | * @params: ptr to the structure that contains key, size and shmflg | |
545 | * | |
d9a605e4 | 546 | * Called with shm_ids.rwsem held as a writer. |
f4566f04 | 547 | */ |
7748dbfa | 548 | static int newseg(struct ipc_namespace *ns, struct ipc_params *params) |
1da177e4 | 549 | { |
7748dbfa ND |
550 | key_t key = params->key; |
551 | int shmflg = params->flg; | |
552 | size_t size = params->u.size; | |
1da177e4 LT |
553 | int error; |
554 | struct shmid_kernel *shp; | |
d69f3bad | 555 | size_t numpages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; |
239521f3 | 556 | struct file *file; |
1da177e4 | 557 | char name[13]; |
ca16d140 | 558 | vm_flags_t acctflag = 0; |
1da177e4 | 559 | |
4e982311 | 560 | if (size < SHMMIN || size > ns->shm_ctlmax) |
1da177e4 LT |
561 | return -EINVAL; |
562 | ||
1376327c MS |
563 | if (numpages << PAGE_SHIFT < size) |
564 | return -ENOSPC; | |
565 | ||
09c6eb1f MS |
566 | if (ns->shm_tot + numpages < ns->shm_tot || |
567 | ns->shm_tot + numpages > ns->shm_ctlall) | |
1da177e4 LT |
568 | return -ENOSPC; |
569 | ||
42e618f7 KC |
570 | shp = kvmalloc(sizeof(*shp), GFP_KERNEL); |
571 | if (unlikely(!shp)) | |
1da177e4 LT |
572 | return -ENOMEM; |
573 | ||
574 | shp->shm_perm.key = key; | |
b33291c0 | 575 | shp->shm_perm.mode = (shmflg & S_IRWXUGO); |
1da177e4 LT |
576 | shp->mlock_user = NULL; |
577 | ||
578 | shp->shm_perm.security = NULL; | |
7191adff | 579 | error = security_shm_alloc(&shp->shm_perm); |
1da177e4 | 580 | if (error) { |
42e618f7 | 581 | kvfree(shp); |
1da177e4 LT |
582 | return error; |
583 | } | |
584 | ||
239521f3 | 585 | sprintf(name, "SYSV%08x", key); |
1da177e4 | 586 | if (shmflg & SHM_HUGETLB) { |
c103a4dc | 587 | struct hstate *hs; |
091d0d55 LZ |
588 | size_t hugesize; |
589 | ||
c103a4dc | 590 | hs = hstate_sizelog((shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); |
091d0d55 LZ |
591 | if (!hs) { |
592 | error = -EINVAL; | |
593 | goto no_file; | |
594 | } | |
595 | hugesize = ALIGN(size, huge_page_size(hs)); | |
af73e4d9 | 596 | |
5a6fe125 MG |
597 | /* hugetlb_file_setup applies strict accounting */ |
598 | if (shmflg & SHM_NORESERVE) | |
599 | acctflag = VM_NORESERVE; | |
af73e4d9 | 600 | file = hugetlb_file_setup(name, hugesize, acctflag, |
42d7395f AK |
601 | &shp->mlock_user, HUGETLB_SHMFS_INODE, |
602 | (shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); | |
1da177e4 | 603 | } else { |
bf8f972d BP |
604 | /* |
605 | * Do not allow no accounting for OVERCOMMIT_NEVER, even | |
239521f3 | 606 | * if it's asked for. |
bf8f972d BP |
607 | */ |
608 | if ((shmflg & SHM_NORESERVE) && | |
609 | sysctl_overcommit_memory != OVERCOMMIT_NEVER) | |
fc8744ad | 610 | acctflag = VM_NORESERVE; |
e1832f29 | 611 | file = shmem_kernel_file_setup(name, size, acctflag); |
1da177e4 LT |
612 | } |
613 | error = PTR_ERR(file); | |
614 | if (IS_ERR(file)) | |
615 | goto no_file; | |
616 | ||
b488893a | 617 | shp->shm_cprid = task_tgid_vnr(current); |
1da177e4 LT |
618 | shp->shm_lprid = 0; |
619 | shp->shm_atim = shp->shm_dtim = 0; | |
7ff2819e | 620 | shp->shm_ctim = ktime_get_real_seconds(); |
1da177e4 LT |
621 | shp->shm_segsz = size; |
622 | shp->shm_nattch = 0; | |
1da177e4 | 623 | shp->shm_file = file; |
5774ed01 | 624 | shp->shm_creator = current; |
b9a53227 | 625 | |
39c96a1b | 626 | /* ipc_addid() locks shp upon success. */ |
a2642f87 MS |
627 | error = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni); |
628 | if (error < 0) | |
b9a53227 | 629 | goto no_id; |
b9a53227 | 630 | |
ab602f79 | 631 | list_add(&shp->shm_clist, ¤t->sysvshm.shm_clist); |
dbfcd91f | 632 | |
30475cc1 BP |
633 | /* |
634 | * shmid gets reported as "inode#" in /proc/pid/maps. | |
635 | * proc-ps tools use this. Changing this will break them. | |
636 | */ | |
496ad9aa | 637 | file_inode(file)->i_ino = shp->shm_perm.id; |
551110a9 | 638 | |
4e982311 | 639 | ns->shm_tot += numpages; |
7ca7e564 | 640 | error = shp->shm_perm.id; |
dbfcd91f | 641 | |
cf9d5d78 | 642 | ipc_unlock_object(&shp->shm_perm); |
dbfcd91f | 643 | rcu_read_unlock(); |
7ca7e564 | 644 | return error; |
1da177e4 LT |
645 | |
646 | no_id: | |
2195d281 | 647 | if (is_file_hugepages(file) && shp->mlock_user) |
353d5c30 | 648 | user_shm_unlock(size, shp->mlock_user); |
1da177e4 LT |
649 | fput(file); |
650 | no_file: | |
a2642f87 | 651 | call_rcu(&shp->shm_perm.rcu, shm_rcu_free); |
1da177e4 LT |
652 | return error; |
653 | } | |
654 | ||
f4566f04 | 655 | /* |
d9a605e4 | 656 | * Called with shm_ids.rwsem and ipcp locked. |
f4566f04 | 657 | */ |
03f02c76 | 658 | static inline int shm_security(struct kern_ipc_perm *ipcp, int shmflg) |
7748dbfa | 659 | { |
7191adff | 660 | return security_shm_associate(ipcp, shmflg); |
7748dbfa ND |
661 | } |
662 | ||
f4566f04 | 663 | /* |
d9a605e4 | 664 | * Called with shm_ids.rwsem and ipcp locked. |
f4566f04 | 665 | */ |
03f02c76 ND |
666 | static inline int shm_more_checks(struct kern_ipc_perm *ipcp, |
667 | struct ipc_params *params) | |
7748dbfa | 668 | { |
03f02c76 ND |
669 | struct shmid_kernel *shp; |
670 | ||
671 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
672 | if (shp->shm_segsz < params->u.size) | |
7748dbfa ND |
673 | return -EINVAL; |
674 | ||
675 | return 0; | |
676 | } | |
677 | ||
d5460c99 | 678 | SYSCALL_DEFINE3(shmget, key_t, key, size_t, size, int, shmflg) |
1da177e4 | 679 | { |
4e982311 | 680 | struct ipc_namespace *ns; |
eb66ec44 MK |
681 | static const struct ipc_ops shm_ops = { |
682 | .getnew = newseg, | |
683 | .associate = shm_security, | |
684 | .more_checks = shm_more_checks, | |
685 | }; | |
7748dbfa | 686 | struct ipc_params shm_params; |
4e982311 KK |
687 | |
688 | ns = current->nsproxy->ipc_ns; | |
1da177e4 | 689 | |
7748dbfa ND |
690 | shm_params.key = key; |
691 | shm_params.flg = shmflg; | |
692 | shm_params.u.size = size; | |
1da177e4 | 693 | |
7748dbfa | 694 | return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params); |
1da177e4 LT |
695 | } |
696 | ||
697 | static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version) | |
698 | { | |
239521f3 | 699 | switch (version) { |
1da177e4 LT |
700 | case IPC_64: |
701 | return copy_to_user(buf, in, sizeof(*in)); | |
702 | case IPC_OLD: | |
703 | { | |
704 | struct shmid_ds out; | |
705 | ||
3af54c9b | 706 | memset(&out, 0, sizeof(out)); |
1da177e4 LT |
707 | ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm); |
708 | out.shm_segsz = in->shm_segsz; | |
709 | out.shm_atime = in->shm_atime; | |
710 | out.shm_dtime = in->shm_dtime; | |
711 | out.shm_ctime = in->shm_ctime; | |
712 | out.shm_cpid = in->shm_cpid; | |
713 | out.shm_lpid = in->shm_lpid; | |
714 | out.shm_nattch = in->shm_nattch; | |
715 | ||
716 | return copy_to_user(buf, &out, sizeof(out)); | |
717 | } | |
718 | default: | |
719 | return -EINVAL; | |
720 | } | |
721 | } | |
722 | ||
016d7132 PP |
723 | static inline unsigned long |
724 | copy_shmid_from_user(struct shmid64_ds *out, void __user *buf, int version) | |
1da177e4 | 725 | { |
239521f3 | 726 | switch (version) { |
1da177e4 | 727 | case IPC_64: |
016d7132 | 728 | if (copy_from_user(out, buf, sizeof(*out))) |
1da177e4 | 729 | return -EFAULT; |
1da177e4 | 730 | return 0; |
1da177e4 LT |
731 | case IPC_OLD: |
732 | { | |
733 | struct shmid_ds tbuf_old; | |
734 | ||
735 | if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) | |
736 | return -EFAULT; | |
737 | ||
016d7132 PP |
738 | out->shm_perm.uid = tbuf_old.shm_perm.uid; |
739 | out->shm_perm.gid = tbuf_old.shm_perm.gid; | |
740 | out->shm_perm.mode = tbuf_old.shm_perm.mode; | |
1da177e4 LT |
741 | |
742 | return 0; | |
743 | } | |
744 | default: | |
745 | return -EINVAL; | |
746 | } | |
747 | } | |
748 | ||
749 | static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version) | |
750 | { | |
239521f3 | 751 | switch (version) { |
1da177e4 LT |
752 | case IPC_64: |
753 | return copy_to_user(buf, in, sizeof(*in)); | |
754 | case IPC_OLD: | |
755 | { | |
756 | struct shminfo out; | |
757 | ||
239521f3 | 758 | if (in->shmmax > INT_MAX) |
1da177e4 LT |
759 | out.shmmax = INT_MAX; |
760 | else | |
761 | out.shmmax = (int)in->shmmax; | |
762 | ||
763 | out.shmmin = in->shmmin; | |
764 | out.shmmni = in->shmmni; | |
765 | out.shmseg = in->shmseg; | |
46c0a8ca | 766 | out.shmall = in->shmall; |
1da177e4 LT |
767 | |
768 | return copy_to_user(buf, &out, sizeof(out)); | |
769 | } | |
770 | default: | |
771 | return -EINVAL; | |
772 | } | |
773 | } | |
774 | ||
b7952180 HD |
775 | /* |
776 | * Calculate and add used RSS and swap pages of a shm. | |
d9a605e4 | 777 | * Called with shm_ids.rwsem held as a reader |
b7952180 HD |
778 | */ |
779 | static void shm_add_rss_swap(struct shmid_kernel *shp, | |
780 | unsigned long *rss_add, unsigned long *swp_add) | |
781 | { | |
782 | struct inode *inode; | |
783 | ||
496ad9aa | 784 | inode = file_inode(shp->shm_file); |
b7952180 HD |
785 | |
786 | if (is_file_hugepages(shp->shm_file)) { | |
787 | struct address_space *mapping = inode->i_mapping; | |
788 | struct hstate *h = hstate_file(shp->shm_file); | |
789 | *rss_add += pages_per_huge_page(h) * mapping->nrpages; | |
790 | } else { | |
791 | #ifdef CONFIG_SHMEM | |
792 | struct shmem_inode_info *info = SHMEM_I(inode); | |
63980c80 | 793 | |
4595ef88 | 794 | spin_lock_irq(&info->lock); |
b7952180 HD |
795 | *rss_add += inode->i_mapping->nrpages; |
796 | *swp_add += info->swapped; | |
4595ef88 | 797 | spin_unlock_irq(&info->lock); |
b7952180 HD |
798 | #else |
799 | *rss_add += inode->i_mapping->nrpages; | |
800 | #endif | |
801 | } | |
802 | } | |
803 | ||
f4566f04 | 804 | /* |
d9a605e4 | 805 | * Called with shm_ids.rwsem held as a reader |
f4566f04 | 806 | */ |
4e982311 KK |
807 | static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss, |
808 | unsigned long *swp) | |
1da177e4 | 809 | { |
7ca7e564 ND |
810 | int next_id; |
811 | int total, in_use; | |
1da177e4 LT |
812 | |
813 | *rss = 0; | |
814 | *swp = 0; | |
815 | ||
7ca7e564 ND |
816 | in_use = shm_ids(ns).in_use; |
817 | ||
818 | for (total = 0, next_id = 0; total < in_use; next_id++) { | |
e562aebc | 819 | struct kern_ipc_perm *ipc; |
1da177e4 | 820 | struct shmid_kernel *shp; |
1da177e4 | 821 | |
e562aebc TB |
822 | ipc = idr_find(&shm_ids(ns).ipcs_idr, next_id); |
823 | if (ipc == NULL) | |
1da177e4 | 824 | continue; |
e562aebc | 825 | shp = container_of(ipc, struct shmid_kernel, shm_perm); |
1da177e4 | 826 | |
b7952180 | 827 | shm_add_rss_swap(shp, rss, swp); |
7ca7e564 ND |
828 | |
829 | total++; | |
1da177e4 LT |
830 | } |
831 | } | |
832 | ||
8d4cc8b5 | 833 | /* |
d9a605e4 | 834 | * This function handles some shmctl commands which require the rwsem |
8d4cc8b5 | 835 | * to be held in write mode. |
d9a605e4 | 836 | * NOTE: no locks must be held, the rwsem is taken inside this function. |
8d4cc8b5 PP |
837 | */ |
838 | static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd, | |
9ba720c1 | 839 | struct shmid64_ds *shmid64) |
1da177e4 | 840 | { |
8d4cc8b5 | 841 | struct kern_ipc_perm *ipcp; |
8d4cc8b5 PP |
842 | struct shmid_kernel *shp; |
843 | int err; | |
844 | ||
d9a605e4 | 845 | down_write(&shm_ids(ns).rwsem); |
7b4cc5d8 DB |
846 | rcu_read_lock(); |
847 | ||
79ccf0f8 | 848 | ipcp = ipcctl_pre_down_nolock(ns, &shm_ids(ns), shmid, cmd, |
9ba720c1 | 849 | &shmid64->shm_perm, 0); |
7b4cc5d8 DB |
850 | if (IS_ERR(ipcp)) { |
851 | err = PTR_ERR(ipcp); | |
7b4cc5d8 DB |
852 | goto out_unlock1; |
853 | } | |
8d4cc8b5 | 854 | |
a5f75e7f | 855 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
8d4cc8b5 | 856 | |
7191adff | 857 | err = security_shm_shmctl(&shp->shm_perm, cmd); |
8d4cc8b5 | 858 | if (err) |
79ccf0f8 | 859 | goto out_unlock1; |
7b4cc5d8 | 860 | |
8d4cc8b5 PP |
861 | switch (cmd) { |
862 | case IPC_RMID: | |
79ccf0f8 | 863 | ipc_lock_object(&shp->shm_perm); |
7b4cc5d8 | 864 | /* do_shm_rmid unlocks the ipc object and rcu */ |
8d4cc8b5 PP |
865 | do_shm_rmid(ns, ipcp); |
866 | goto out_up; | |
867 | case IPC_SET: | |
79ccf0f8 | 868 | ipc_lock_object(&shp->shm_perm); |
9ba720c1 | 869 | err = ipc_update_perm(&shmid64->shm_perm, ipcp); |
1efdb69b | 870 | if (err) |
7b4cc5d8 | 871 | goto out_unlock0; |
7ff2819e | 872 | shp->shm_ctim = ktime_get_real_seconds(); |
8d4cc8b5 PP |
873 | break; |
874 | default: | |
875 | err = -EINVAL; | |
79ccf0f8 | 876 | goto out_unlock1; |
8d4cc8b5 | 877 | } |
7b4cc5d8 DB |
878 | |
879 | out_unlock0: | |
880 | ipc_unlock_object(&shp->shm_perm); | |
881 | out_unlock1: | |
882 | rcu_read_unlock(); | |
8d4cc8b5 | 883 | out_up: |
d9a605e4 | 884 | up_write(&shm_ids(ns).rwsem); |
8d4cc8b5 PP |
885 | return err; |
886 | } | |
887 | ||
9ba720c1 AV |
888 | static int shmctl_ipc_info(struct ipc_namespace *ns, |
889 | struct shminfo64 *shminfo) | |
8d4cc8b5 | 890 | { |
9ba720c1 AV |
891 | int err = security_shm_shmctl(NULL, IPC_INFO); |
892 | if (!err) { | |
893 | memset(shminfo, 0, sizeof(*shminfo)); | |
894 | shminfo->shmmni = shminfo->shmseg = ns->shm_ctlmni; | |
895 | shminfo->shmmax = ns->shm_ctlmax; | |
896 | shminfo->shmall = ns->shm_ctlall; | |
897 | shminfo->shmmin = SHMMIN; | |
d9a605e4 | 898 | down_read(&shm_ids(ns).rwsem); |
7ca7e564 | 899 | err = ipc_get_maxid(&shm_ids(ns)); |
d9a605e4 | 900 | up_read(&shm_ids(ns).rwsem); |
239521f3 | 901 | if (err < 0) |
1da177e4 | 902 | err = 0; |
1da177e4 | 903 | } |
9ba720c1 AV |
904 | return err; |
905 | } | |
1da177e4 | 906 | |
9ba720c1 AV |
907 | static int shmctl_shm_info(struct ipc_namespace *ns, |
908 | struct shm_info *shm_info) | |
909 | { | |
910 | int err = security_shm_shmctl(NULL, SHM_INFO); | |
911 | if (!err) { | |
912 | memset(shm_info, 0, sizeof(*shm_info)); | |
d9a605e4 | 913 | down_read(&shm_ids(ns).rwsem); |
9ba720c1 AV |
914 | shm_info->used_ids = shm_ids(ns).in_use; |
915 | shm_get_stat(ns, &shm_info->shm_rss, &shm_info->shm_swp); | |
916 | shm_info->shm_tot = ns->shm_tot; | |
917 | shm_info->swap_attempts = 0; | |
918 | shm_info->swap_successes = 0; | |
7ca7e564 | 919 | err = ipc_get_maxid(&shm_ids(ns)); |
d9a605e4 | 920 | up_read(&shm_ids(ns).rwsem); |
9ba720c1 AV |
921 | if (err < 0) |
922 | err = 0; | |
1da177e4 | 923 | } |
9ba720c1 AV |
924 | return err; |
925 | } | |
c97cb9cc | 926 | |
9ba720c1 AV |
927 | static int shmctl_stat(struct ipc_namespace *ns, int shmid, |
928 | int cmd, struct shmid64_ds *tbuf) | |
929 | { | |
930 | struct shmid_kernel *shp; | |
87ad4b0d | 931 | int id = 0; |
9ba720c1 | 932 | int err; |
c97cb9cc | 933 | |
87ad4b0d PM |
934 | memset(tbuf, 0, sizeof(*tbuf)); |
935 | ||
9ba720c1 AV |
936 | rcu_read_lock(); |
937 | if (cmd == SHM_STAT) { | |
938 | shp = shm_obtain_object(ns, shmid); | |
939 | if (IS_ERR(shp)) { | |
940 | err = PTR_ERR(shp); | |
941 | goto out_unlock; | |
942 | } | |
87ad4b0d | 943 | id = shp->shm_perm.id; |
9ba720c1 AV |
944 | } else { |
945 | shp = shm_obtain_object_check(ns, shmid); | |
946 | if (IS_ERR(shp)) { | |
947 | err = PTR_ERR(shp); | |
1da177e4 | 948 | goto out_unlock; |
1da177e4 | 949 | } |
9ba720c1 | 950 | } |
1da177e4 | 951 | |
9ba720c1 AV |
952 | err = -EACCES; |
953 | if (ipcperms(ns, &shp->shm_perm, S_IRUGO)) | |
954 | goto out_unlock; | |
c97cb9cc | 955 | |
7191adff | 956 | err = security_shm_shmctl(&shp->shm_perm, cmd); |
9ba720c1 AV |
957 | if (err) |
958 | goto out_unlock; | |
959 | ||
87ad4b0d PM |
960 | ipc_lock_object(&shp->shm_perm); |
961 | ||
962 | if (!ipc_valid_object(&shp->shm_perm)) { | |
963 | ipc_unlock_object(&shp->shm_perm); | |
964 | err = -EIDRM; | |
965 | goto out_unlock; | |
966 | } | |
967 | ||
9ba720c1 AV |
968 | kernel_to_ipc64_perm(&shp->shm_perm, &tbuf->shm_perm); |
969 | tbuf->shm_segsz = shp->shm_segsz; | |
970 | tbuf->shm_atime = shp->shm_atim; | |
971 | tbuf->shm_dtime = shp->shm_dtim; | |
972 | tbuf->shm_ctime = shp->shm_ctim; | |
973 | tbuf->shm_cpid = shp->shm_cprid; | |
974 | tbuf->shm_lpid = shp->shm_lprid; | |
975 | tbuf->shm_nattch = shp->shm_nattch; | |
87ad4b0d PM |
976 | |
977 | ipc_unlock_object(&shp->shm_perm); | |
9ba720c1 | 978 | rcu_read_unlock(); |
87ad4b0d | 979 | return id; |
68eccc1d DB |
980 | |
981 | out_unlock: | |
c97cb9cc | 982 | rcu_read_unlock(); |
68eccc1d DB |
983 | return err; |
984 | } | |
985 | ||
9ba720c1 | 986 | static int shmctl_do_lock(struct ipc_namespace *ns, int shmid, int cmd) |
68eccc1d DB |
987 | { |
988 | struct shmid_kernel *shp; | |
9ba720c1 AV |
989 | struct file *shm_file; |
990 | int err; | |
68eccc1d | 991 | |
9ba720c1 AV |
992 | rcu_read_lock(); |
993 | shp = shm_obtain_object_check(ns, shmid); | |
994 | if (IS_ERR(shp)) { | |
995 | err = PTR_ERR(shp); | |
996 | goto out_unlock1; | |
1da177e4 | 997 | } |
c97cb9cc | 998 | |
9ba720c1 | 999 | audit_ipc_obj(&(shp->shm_perm)); |
7191adff | 1000 | err = security_shm_shmctl(&shp->shm_perm, cmd); |
9ba720c1 AV |
1001 | if (err) |
1002 | goto out_unlock1; | |
c97cb9cc | 1003 | |
9ba720c1 | 1004 | ipc_lock_object(&shp->shm_perm); |
c97cb9cc | 1005 | |
9ba720c1 AV |
1006 | /* check if shm_destroy() is tearing down shp */ |
1007 | if (!ipc_valid_object(&shp->shm_perm)) { | |
1008 | err = -EIDRM; | |
1009 | goto out_unlock0; | |
1da177e4 | 1010 | } |
073115d6 | 1011 | |
9ba720c1 AV |
1012 | if (!ns_capable(ns->user_ns, CAP_IPC_LOCK)) { |
1013 | kuid_t euid = current_euid(); | |
0f3d2b01 | 1014 | |
9ba720c1 AV |
1015 | if (!uid_eq(euid, shp->shm_perm.uid) && |
1016 | !uid_eq(euid, shp->shm_perm.cuid)) { | |
1017 | err = -EPERM; | |
0f3d2b01 RA |
1018 | goto out_unlock0; |
1019 | } | |
9ba720c1 AV |
1020 | if (cmd == SHM_LOCK && !rlimit(RLIMIT_MEMLOCK)) { |
1021 | err = -EPERM; | |
1022 | goto out_unlock0; | |
1da177e4 | 1023 | } |
68eccc1d DB |
1024 | } |
1025 | ||
9ba720c1 AV |
1026 | shm_file = shp->shm_file; |
1027 | if (is_file_hugepages(shm_file)) | |
1028 | goto out_unlock0; | |
85046579 | 1029 | |
9ba720c1 AV |
1030 | if (cmd == SHM_LOCK) { |
1031 | struct user_struct *user = current_user(); | |
63980c80 | 1032 | |
9ba720c1 AV |
1033 | err = shmem_lock(shm_file, 1, user); |
1034 | if (!err && !(shp->shm_perm.mode & SHM_LOCKED)) { | |
1035 | shp->shm_perm.mode |= SHM_LOCKED; | |
1036 | shp->mlock_user = user; | |
1da177e4 | 1037 | } |
9ba720c1 | 1038 | goto out_unlock0; |
1da177e4 LT |
1039 | } |
1040 | ||
9ba720c1 AV |
1041 | /* SHM_UNLOCK */ |
1042 | if (!(shp->shm_perm.mode & SHM_LOCKED)) | |
1043 | goto out_unlock0; | |
1044 | shmem_lock(shm_file, 0, shp->mlock_user); | |
1045 | shp->shm_perm.mode &= ~SHM_LOCKED; | |
1046 | shp->mlock_user = NULL; | |
1047 | get_file(shm_file); | |
1048 | ipc_unlock_object(&shp->shm_perm); | |
1049 | rcu_read_unlock(); | |
1050 | shmem_unlock_mapping(shm_file->f_mapping); | |
1051 | ||
1052 | fput(shm_file); | |
1053 | return err; | |
1054 | ||
2caacaa8 DB |
1055 | out_unlock0: |
1056 | ipc_unlock_object(&shp->shm_perm); | |
1057 | out_unlock1: | |
c97cb9cc | 1058 | rcu_read_unlock(); |
68eccc1d DB |
1059 | return err; |
1060 | } | |
1061 | ||
1062 | SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, struct shmid_ds __user *, buf) | |
1063 | { | |
68eccc1d DB |
1064 | int err, version; |
1065 | struct ipc_namespace *ns; | |
553f770e | 1066 | struct shmid64_ds sem64; |
68eccc1d | 1067 | |
2caacaa8 DB |
1068 | if (cmd < 0 || shmid < 0) |
1069 | return -EINVAL; | |
68eccc1d DB |
1070 | |
1071 | version = ipc_parse_version(&cmd); | |
1072 | ns = current->nsproxy->ipc_ns; | |
1073 | ||
1074 | switch (cmd) { | |
9ba720c1 AV |
1075 | case IPC_INFO: { |
1076 | struct shminfo64 shminfo; | |
1077 | err = shmctl_ipc_info(ns, &shminfo); | |
1078 | if (err < 0) | |
1079 | return err; | |
1080 | if (copy_shminfo_to_user(buf, &shminfo, version)) | |
1081 | err = -EFAULT; | |
1082 | return err; | |
1083 | } | |
1084 | case SHM_INFO: { | |
1085 | struct shm_info shm_info; | |
1086 | err = shmctl_shm_info(ns, &shm_info); | |
1087 | if (err < 0) | |
1088 | return err; | |
1089 | if (copy_to_user(buf, &shm_info, sizeof(shm_info))) | |
1090 | err = -EFAULT; | |
1091 | return err; | |
1092 | } | |
68eccc1d | 1093 | case SHM_STAT: |
9ba720c1 | 1094 | case IPC_STAT: { |
553f770e | 1095 | err = shmctl_stat(ns, shmid, cmd, &sem64); |
9ba720c1 AV |
1096 | if (err < 0) |
1097 | return err; | |
553f770e | 1098 | if (copy_shmid_to_user(buf, &sem64, version)) |
9ba720c1 AV |
1099 | err = -EFAULT; |
1100 | return err; | |
1101 | } | |
2caacaa8 | 1102 | case IPC_SET: |
553f770e | 1103 | if (copy_shmid_from_user(&sem64, buf, version)) |
9ba720c1 | 1104 | return -EFAULT; |
553f770e | 1105 | /* fallthru */ |
9ba720c1 | 1106 | case IPC_RMID: |
553f770e | 1107 | return shmctl_down(ns, shmid, cmd, &sem64); |
1da177e4 LT |
1108 | case SHM_LOCK: |
1109 | case SHM_UNLOCK: | |
9ba720c1 AV |
1110 | return shmctl_do_lock(ns, shmid, cmd); |
1111 | default: | |
1112 | return -EINVAL; | |
1113 | } | |
1114 | } | |
89e004ea | 1115 | |
553f770e AV |
1116 | #ifdef CONFIG_COMPAT |
1117 | ||
1118 | struct compat_shmid_ds { | |
1119 | struct compat_ipc_perm shm_perm; | |
1120 | int shm_segsz; | |
1121 | compat_time_t shm_atime; | |
1122 | compat_time_t shm_dtime; | |
1123 | compat_time_t shm_ctime; | |
1124 | compat_ipc_pid_t shm_cpid; | |
1125 | compat_ipc_pid_t shm_lpid; | |
1126 | unsigned short shm_nattch; | |
1127 | unsigned short shm_unused; | |
1128 | compat_uptr_t shm_unused2; | |
1129 | compat_uptr_t shm_unused3; | |
1130 | }; | |
1da177e4 | 1131 | |
553f770e AV |
1132 | struct compat_shminfo64 { |
1133 | compat_ulong_t shmmax; | |
1134 | compat_ulong_t shmmin; | |
1135 | compat_ulong_t shmmni; | |
1136 | compat_ulong_t shmseg; | |
1137 | compat_ulong_t shmall; | |
1138 | compat_ulong_t __unused1; | |
1139 | compat_ulong_t __unused2; | |
1140 | compat_ulong_t __unused3; | |
1141 | compat_ulong_t __unused4; | |
1142 | }; | |
073115d6 | 1143 | |
553f770e AV |
1144 | struct compat_shm_info { |
1145 | compat_int_t used_ids; | |
1146 | compat_ulong_t shm_tot, shm_rss, shm_swp; | |
1147 | compat_ulong_t swap_attempts, swap_successes; | |
1148 | }; | |
0f3d2b01 | 1149 | |
553f770e AV |
1150 | static int copy_compat_shminfo_to_user(void __user *buf, struct shminfo64 *in, |
1151 | int version) | |
1152 | { | |
1153 | if (in->shmmax > INT_MAX) | |
1154 | in->shmmax = INT_MAX; | |
1155 | if (version == IPC_64) { | |
1156 | struct compat_shminfo64 info; | |
1157 | memset(&info, 0, sizeof(info)); | |
1158 | info.shmmax = in->shmmax; | |
1159 | info.shmmin = in->shmmin; | |
1160 | info.shmmni = in->shmmni; | |
1161 | info.shmseg = in->shmseg; | |
1162 | info.shmall = in->shmall; | |
1163 | return copy_to_user(buf, &info, sizeof(info)); | |
1164 | } else { | |
1165 | struct shminfo info; | |
1166 | memset(&info, 0, sizeof(info)); | |
1167 | info.shmmax = in->shmmax; | |
1168 | info.shmmin = in->shmmin; | |
1169 | info.shmmni = in->shmmni; | |
1170 | info.shmseg = in->shmseg; | |
1171 | info.shmall = in->shmall; | |
1172 | return copy_to_user(buf, &info, sizeof(info)); | |
1173 | } | |
1174 | } | |
0f3d2b01 | 1175 | |
553f770e AV |
1176 | static int put_compat_shm_info(struct shm_info *ip, |
1177 | struct compat_shm_info __user *uip) | |
1178 | { | |
1179 | struct compat_shm_info info; | |
1180 | ||
1181 | memset(&info, 0, sizeof(info)); | |
1182 | info.used_ids = ip->used_ids; | |
1183 | info.shm_tot = ip->shm_tot; | |
1184 | info.shm_rss = ip->shm_rss; | |
1185 | info.shm_swp = ip->shm_swp; | |
1186 | info.swap_attempts = ip->swap_attempts; | |
1187 | info.swap_successes = ip->swap_successes; | |
b776e4b1 | 1188 | return copy_to_user(uip, &info, sizeof(info)); |
553f770e | 1189 | } |
1da177e4 | 1190 | |
553f770e AV |
1191 | static int copy_compat_shmid_to_user(void __user *buf, struct shmid64_ds *in, |
1192 | int version) | |
1193 | { | |
1194 | if (version == IPC_64) { | |
1195 | struct compat_shmid64_ds v; | |
1196 | memset(&v, 0, sizeof(v)); | |
28327fae | 1197 | to_compat_ipc64_perm(&v.shm_perm, &in->shm_perm); |
553f770e AV |
1198 | v.shm_atime = in->shm_atime; |
1199 | v.shm_dtime = in->shm_dtime; | |
1200 | v.shm_ctime = in->shm_ctime; | |
1201 | v.shm_segsz = in->shm_segsz; | |
1202 | v.shm_nattch = in->shm_nattch; | |
1203 | v.shm_cpid = in->shm_cpid; | |
1204 | v.shm_lpid = in->shm_lpid; | |
1205 | return copy_to_user(buf, &v, sizeof(v)); | |
1206 | } else { | |
1207 | struct compat_shmid_ds v; | |
1208 | memset(&v, 0, sizeof(v)); | |
28327fae | 1209 | to_compat_ipc_perm(&v.shm_perm, &in->shm_perm); |
553f770e | 1210 | v.shm_perm.key = in->shm_perm.key; |
553f770e AV |
1211 | v.shm_atime = in->shm_atime; |
1212 | v.shm_dtime = in->shm_dtime; | |
1213 | v.shm_ctime = in->shm_ctime; | |
1214 | v.shm_segsz = in->shm_segsz; | |
1215 | v.shm_nattch = in->shm_nattch; | |
1216 | v.shm_cpid = in->shm_cpid; | |
1217 | v.shm_lpid = in->shm_lpid; | |
1218 | return copy_to_user(buf, &v, sizeof(v)); | |
1219 | } | |
1220 | } | |
85046579 | 1221 | |
553f770e AV |
1222 | static int copy_compat_shmid_from_user(struct shmid64_ds *out, void __user *buf, |
1223 | int version) | |
1224 | { | |
1225 | memset(out, 0, sizeof(*out)); | |
1226 | if (version == IPC_64) { | |
6aa211e8 | 1227 | struct compat_shmid64_ds __user *p = buf; |
28327fae | 1228 | return get_compat_ipc64_perm(&out->shm_perm, &p->shm_perm); |
553f770e | 1229 | } else { |
6aa211e8 | 1230 | struct compat_shmid_ds __user *p = buf; |
28327fae | 1231 | return get_compat_ipc_perm(&out->shm_perm, &p->shm_perm); |
553f770e | 1232 | } |
553f770e | 1233 | } |
63980c80 | 1234 | |
553f770e AV |
1235 | COMPAT_SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, void __user *, uptr) |
1236 | { | |
1237 | struct ipc_namespace *ns; | |
1238 | struct shmid64_ds sem64; | |
1239 | int version = compat_ipc_parse_version(&cmd); | |
1240 | int err; | |
85046579 | 1241 | |
553f770e AV |
1242 | ns = current->nsproxy->ipc_ns; |
1243 | ||
1244 | if (cmd < 0 || shmid < 0) | |
1245 | return -EINVAL; | |
2caacaa8 | 1246 | |
553f770e AV |
1247 | switch (cmd) { |
1248 | case IPC_INFO: { | |
1249 | struct shminfo64 shminfo; | |
1250 | err = shmctl_ipc_info(ns, &shminfo); | |
1251 | if (err < 0) | |
1252 | return err; | |
1253 | if (copy_compat_shminfo_to_user(uptr, &shminfo, version)) | |
1254 | err = -EFAULT; | |
1255 | return err; | |
1256 | } | |
1257 | case SHM_INFO: { | |
1258 | struct shm_info shm_info; | |
1259 | err = shmctl_shm_info(ns, &shm_info); | |
1260 | if (err < 0) | |
1261 | return err; | |
1262 | if (put_compat_shm_info(&shm_info, uptr)) | |
1263 | err = -EFAULT; | |
8d4cc8b5 | 1264 | return err; |
2caacaa8 | 1265 | } |
553f770e AV |
1266 | case IPC_STAT: |
1267 | case SHM_STAT: | |
1268 | err = shmctl_stat(ns, shmid, cmd, &sem64); | |
1269 | if (err < 0) | |
1270 | return err; | |
58aff0af | 1271 | if (copy_compat_shmid_to_user(uptr, &sem64, version)) |
553f770e AV |
1272 | err = -EFAULT; |
1273 | return err; | |
1274 | ||
1275 | case IPC_SET: | |
1276 | if (copy_compat_shmid_from_user(&sem64, uptr, version)) | |
1277 | return -EFAULT; | |
1278 | /* fallthru */ | |
1279 | case IPC_RMID: | |
1280 | return shmctl_down(ns, shmid, cmd, &sem64); | |
1281 | case SHM_LOCK: | |
1282 | case SHM_UNLOCK: | |
1283 | return shmctl_do_lock(ns, shmid, cmd); | |
1284 | break; | |
1da177e4 | 1285 | default: |
8d4cc8b5 | 1286 | return -EINVAL; |
1da177e4 | 1287 | } |
1da177e4 LT |
1288 | return err; |
1289 | } | |
553f770e | 1290 | #endif |
1da177e4 LT |
1291 | |
1292 | /* | |
1293 | * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists. | |
1294 | * | |
1295 | * NOTE! Despite the name, this is NOT a direct system call entrypoint. The | |
1296 | * "raddr" thing points to kernel space, and there has to be a wrapper around | |
1297 | * this. | |
1298 | */ | |
95e91b83 DB |
1299 | long do_shmat(int shmid, char __user *shmaddr, int shmflg, |
1300 | ulong *raddr, unsigned long shmlba) | |
1da177e4 LT |
1301 | { |
1302 | struct shmid_kernel *shp; | |
f0cb8802 | 1303 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1304 | unsigned long size; |
239521f3 | 1305 | struct file *file; |
1da177e4 | 1306 | int err; |
f0cb8802 | 1307 | unsigned long flags = MAP_SHARED; |
1da177e4 | 1308 | unsigned long prot; |
1da177e4 | 1309 | int acc_mode; |
4e982311 | 1310 | struct ipc_namespace *ns; |
bc56bba8 EB |
1311 | struct shm_file_data *sfd; |
1312 | struct path path; | |
aeb5d727 | 1313 | fmode_t f_mode; |
41badc15 | 1314 | unsigned long populate = 0; |
1da177e4 | 1315 | |
bc56bba8 EB |
1316 | err = -EINVAL; |
1317 | if (shmid < 0) | |
1da177e4 | 1318 | goto out; |
f0cb8802 DB |
1319 | |
1320 | if (addr) { | |
079a96ae | 1321 | if (addr & (shmlba - 1)) { |
95e91b83 DB |
1322 | /* |
1323 | * Round down to the nearest multiple of shmlba. | |
1324 | * For sane do_mmap_pgoff() parameters, avoid | |
1325 | * round downs that trigger nil-page and MAP_FIXED. | |
1326 | */ | |
1327 | if ((shmflg & SHM_RND) && addr >= shmlba) | |
1328 | addr &= ~(shmlba - 1); | |
1da177e4 LT |
1329 | else |
1330 | #ifndef __ARCH_FORCE_SHMLBA | |
1331 | if (addr & ~PAGE_MASK) | |
1332 | #endif | |
bc56bba8 | 1333 | goto out; |
1da177e4 | 1334 | } |
1da177e4 | 1335 | |
f0cb8802 DB |
1336 | flags |= MAP_FIXED; |
1337 | } else if ((shmflg & SHM_REMAP)) | |
1338 | goto out; | |
1da177e4 LT |
1339 | |
1340 | if (shmflg & SHM_RDONLY) { | |
1341 | prot = PROT_READ; | |
1da177e4 | 1342 | acc_mode = S_IRUGO; |
bc56bba8 | 1343 | f_mode = FMODE_READ; |
1da177e4 LT |
1344 | } else { |
1345 | prot = PROT_READ | PROT_WRITE; | |
1da177e4 | 1346 | acc_mode = S_IRUGO | S_IWUGO; |
bc56bba8 | 1347 | f_mode = FMODE_READ | FMODE_WRITE; |
1da177e4 LT |
1348 | } |
1349 | if (shmflg & SHM_EXEC) { | |
1350 | prot |= PROT_EXEC; | |
1351 | acc_mode |= S_IXUGO; | |
1352 | } | |
1353 | ||
1354 | /* | |
1355 | * We cannot rely on the fs check since SYSV IPC does have an | |
1356 | * additional creator id... | |
1357 | */ | |
4e982311 | 1358 | ns = current->nsproxy->ipc_ns; |
c2c737a0 DB |
1359 | rcu_read_lock(); |
1360 | shp = shm_obtain_object_check(ns, shmid); | |
023a5355 ND |
1361 | if (IS_ERR(shp)) { |
1362 | err = PTR_ERR(shp); | |
c2c737a0 | 1363 | goto out_unlock; |
023a5355 | 1364 | } |
bc56bba8 EB |
1365 | |
1366 | err = -EACCES; | |
b0e77598 | 1367 | if (ipcperms(ns, &shp->shm_perm, acc_mode)) |
bc56bba8 | 1368 | goto out_unlock; |
1da177e4 | 1369 | |
7191adff | 1370 | err = security_shm_shmat(&shp->shm_perm, shmaddr, shmflg); |
bc56bba8 EB |
1371 | if (err) |
1372 | goto out_unlock; | |
1373 | ||
c2c737a0 | 1374 | ipc_lock_object(&shp->shm_perm); |
a399b29d GT |
1375 | |
1376 | /* check if shm_destroy() is tearing down shp */ | |
0f3d2b01 | 1377 | if (!ipc_valid_object(&shp->shm_perm)) { |
a399b29d GT |
1378 | ipc_unlock_object(&shp->shm_perm); |
1379 | err = -EIDRM; | |
1380 | goto out_unlock; | |
1381 | } | |
1382 | ||
2c48b9c4 AV |
1383 | path = shp->shm_file->f_path; |
1384 | path_get(&path); | |
1da177e4 | 1385 | shp->shm_nattch++; |
75c3cfa8 | 1386 | size = i_size_read(d_inode(path.dentry)); |
c2c737a0 DB |
1387 | ipc_unlock_object(&shp->shm_perm); |
1388 | rcu_read_unlock(); | |
1da177e4 | 1389 | |
bc56bba8 EB |
1390 | err = -ENOMEM; |
1391 | sfd = kzalloc(sizeof(*sfd), GFP_KERNEL); | |
f42569b1 DB |
1392 | if (!sfd) { |
1393 | path_put(&path); | |
1394 | goto out_nattch; | |
1395 | } | |
bc56bba8 | 1396 | |
2c48b9c4 AV |
1397 | file = alloc_file(&path, f_mode, |
1398 | is_file_hugepages(shp->shm_file) ? | |
c4caa778 AV |
1399 | &shm_file_operations_huge : |
1400 | &shm_file_operations); | |
39b65252 | 1401 | err = PTR_ERR(file); |
f42569b1 DB |
1402 | if (IS_ERR(file)) { |
1403 | kfree(sfd); | |
1404 | path_put(&path); | |
1405 | goto out_nattch; | |
1406 | } | |
bc56bba8 | 1407 | |
bc56bba8 | 1408 | file->private_data = sfd; |
bc56bba8 | 1409 | file->f_mapping = shp->shm_file->f_mapping; |
7ca7e564 | 1410 | sfd->id = shp->shm_perm.id; |
bc56bba8 EB |
1411 | sfd->ns = get_ipc_ns(ns); |
1412 | sfd->file = shp->shm_file; | |
1413 | sfd->vm_ops = NULL; | |
1414 | ||
8b3ec681 AV |
1415 | err = security_mmap_file(file, prot, flags); |
1416 | if (err) | |
1417 | goto out_fput; | |
1418 | ||
91f4f94e MH |
1419 | if (down_write_killable(¤t->mm->mmap_sem)) { |
1420 | err = -EINTR; | |
1421 | goto out_fput; | |
1422 | } | |
1423 | ||
1da177e4 | 1424 | if (addr && !(shmflg & SHM_REMAP)) { |
bc56bba8 | 1425 | err = -EINVAL; |
247a8ce8 MS |
1426 | if (addr + size < addr) |
1427 | goto invalid; | |
1428 | ||
1da177e4 LT |
1429 | if (find_vma_intersection(current->mm, addr, addr + size)) |
1430 | goto invalid; | |
1da177e4 | 1431 | } |
f42569b1 | 1432 | |
897ab3e0 | 1433 | addr = do_mmap_pgoff(file, addr, size, prot, flags, 0, &populate, NULL); |
bebeb3d6 | 1434 | *raddr = addr; |
bc56bba8 | 1435 | err = 0; |
bebeb3d6 ML |
1436 | if (IS_ERR_VALUE(addr)) |
1437 | err = (long)addr; | |
1da177e4 LT |
1438 | invalid: |
1439 | up_write(¤t->mm->mmap_sem); | |
bebeb3d6 | 1440 | if (populate) |
41badc15 | 1441 | mm_populate(addr, populate); |
1da177e4 | 1442 | |
8b3ec681 | 1443 | out_fput: |
bc56bba8 EB |
1444 | fput(file); |
1445 | ||
1446 | out_nattch: | |
d9a605e4 | 1447 | down_write(&shm_ids(ns).rwsem); |
00c2bf85 | 1448 | shp = shm_lock(ns, shmid); |
1da177e4 | 1449 | shp->shm_nattch--; |
b34a6b1d | 1450 | if (shm_may_destroy(ns, shp)) |
4e982311 | 1451 | shm_destroy(ns, shp); |
1da177e4 LT |
1452 | else |
1453 | shm_unlock(shp); | |
d9a605e4 | 1454 | up_write(&shm_ids(ns).rwsem); |
1da177e4 | 1455 | return err; |
bc56bba8 EB |
1456 | |
1457 | out_unlock: | |
c2c737a0 | 1458 | rcu_read_unlock(); |
f42569b1 DB |
1459 | out: |
1460 | return err; | |
1da177e4 LT |
1461 | } |
1462 | ||
d5460c99 | 1463 | SYSCALL_DEFINE3(shmat, int, shmid, char __user *, shmaddr, int, shmflg) |
7d87e14c SR |
1464 | { |
1465 | unsigned long ret; | |
1466 | long err; | |
1467 | ||
079a96ae | 1468 | err = do_shmat(shmid, shmaddr, shmflg, &ret, SHMLBA); |
7d87e14c SR |
1469 | if (err) |
1470 | return err; | |
1471 | force_successful_syscall_return(); | |
1472 | return (long)ret; | |
1473 | } | |
1474 | ||
a78ee9ed AV |
1475 | #ifdef CONFIG_COMPAT |
1476 | ||
1477 | #ifndef COMPAT_SHMLBA | |
1478 | #define COMPAT_SHMLBA SHMLBA | |
1479 | #endif | |
1480 | ||
1481 | COMPAT_SYSCALL_DEFINE3(shmat, int, shmid, compat_uptr_t, shmaddr, int, shmflg) | |
1482 | { | |
1483 | unsigned long ret; | |
1484 | long err; | |
1485 | ||
1486 | err = do_shmat(shmid, compat_ptr(shmaddr), shmflg, &ret, COMPAT_SHMLBA); | |
1487 | if (err) | |
1488 | return err; | |
1489 | force_successful_syscall_return(); | |
1490 | return (long)ret; | |
1491 | } | |
1492 | #endif | |
1493 | ||
1da177e4 LT |
1494 | /* |
1495 | * detach and kill segment if marked destroyed. | |
1496 | * The work is done in shm_close. | |
1497 | */ | |
d5460c99 | 1498 | SYSCALL_DEFINE1(shmdt, char __user *, shmaddr) |
1da177e4 LT |
1499 | { |
1500 | struct mm_struct *mm = current->mm; | |
586c7e6a | 1501 | struct vm_area_struct *vma; |
1da177e4 | 1502 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1503 | int retval = -EINVAL; |
586c7e6a MF |
1504 | #ifdef CONFIG_MMU |
1505 | loff_t size = 0; | |
d3c97900 | 1506 | struct file *file; |
586c7e6a MF |
1507 | struct vm_area_struct *next; |
1508 | #endif | |
1da177e4 | 1509 | |
df1e2fb5 HD |
1510 | if (addr & ~PAGE_MASK) |
1511 | return retval; | |
1512 | ||
91f4f94e MH |
1513 | if (down_write_killable(&mm->mmap_sem)) |
1514 | return -EINTR; | |
1da177e4 LT |
1515 | |
1516 | /* | |
1517 | * This function tries to be smart and unmap shm segments that | |
1518 | * were modified by partial mlock or munmap calls: | |
1519 | * - It first determines the size of the shm segment that should be | |
1520 | * unmapped: It searches for a vma that is backed by shm and that | |
1521 | * started at address shmaddr. It records it's size and then unmaps | |
1522 | * it. | |
1523 | * - Then it unmaps all shm vmas that started at shmaddr and that | |
d3c97900 DH |
1524 | * are within the initially determined size and that are from the |
1525 | * same shm segment from which we determined the size. | |
1da177e4 LT |
1526 | * Errors from do_munmap are ignored: the function only fails if |
1527 | * it's called with invalid parameters or if it's called to unmap | |
1528 | * a part of a vma. Both calls in this function are for full vmas, | |
1529 | * the parameters are directly copied from the vma itself and always | |
1530 | * valid - therefore do_munmap cannot fail. (famous last words?) | |
1531 | */ | |
1532 | /* | |
1533 | * If it had been mremap()'d, the starting address would not | |
1534 | * match the usual checks anyway. So assume all vma's are | |
1535 | * above the starting address given. | |
1536 | */ | |
1537 | vma = find_vma(mm, addr); | |
1538 | ||
8feae131 | 1539 | #ifdef CONFIG_MMU |
1da177e4 LT |
1540 | while (vma) { |
1541 | next = vma->vm_next; | |
1542 | ||
1543 | /* | |
1544 | * Check if the starting address would match, i.e. it's | |
1545 | * a fragment created by mprotect() and/or munmap(), or it | |
1546 | * otherwise it starts at this address with no hassles. | |
1547 | */ | |
bc56bba8 | 1548 | if ((vma->vm_ops == &shm_vm_ops) && |
1da177e4 LT |
1549 | (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) { |
1550 | ||
d3c97900 DH |
1551 | /* |
1552 | * Record the file of the shm segment being | |
1553 | * unmapped. With mremap(), someone could place | |
1554 | * page from another segment but with equal offsets | |
1555 | * in the range we are unmapping. | |
1556 | */ | |
1557 | file = vma->vm_file; | |
07a46ed2 | 1558 | size = i_size_read(file_inode(vma->vm_file)); |
897ab3e0 | 1559 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1560 | /* |
1561 | * We discovered the size of the shm segment, so | |
1562 | * break out of here and fall through to the next | |
1563 | * loop that uses the size information to stop | |
1564 | * searching for matching vma's. | |
1565 | */ | |
1566 | retval = 0; | |
1567 | vma = next; | |
1568 | break; | |
1569 | } | |
1570 | vma = next; | |
1571 | } | |
1572 | ||
1573 | /* | |
1574 | * We need look no further than the maximum address a fragment | |
1575 | * could possibly have landed at. Also cast things to loff_t to | |
25985edc | 1576 | * prevent overflows and make comparisons vs. equal-width types. |
1da177e4 | 1577 | */ |
8e36709d | 1578 | size = PAGE_ALIGN(size); |
1da177e4 LT |
1579 | while (vma && (loff_t)(vma->vm_end - addr) <= size) { |
1580 | next = vma->vm_next; | |
1581 | ||
1582 | /* finding a matching vma now does not alter retval */ | |
bc56bba8 | 1583 | if ((vma->vm_ops == &shm_vm_ops) && |
d3c97900 DH |
1584 | ((vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) && |
1585 | (vma->vm_file == file)) | |
897ab3e0 | 1586 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1587 | vma = next; |
1588 | } | |
1589 | ||
63980c80 | 1590 | #else /* CONFIG_MMU */ |
8feae131 | 1591 | /* under NOMMU conditions, the exact address to be destroyed must be |
63980c80 SP |
1592 | * given |
1593 | */ | |
530fcd16 | 1594 | if (vma && vma->vm_start == addr && vma->vm_ops == &shm_vm_ops) { |
897ab3e0 | 1595 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
8feae131 DH |
1596 | retval = 0; |
1597 | } | |
1598 | ||
1599 | #endif | |
1600 | ||
1da177e4 LT |
1601 | up_write(&mm->mmap_sem); |
1602 | return retval; | |
1603 | } | |
1604 | ||
1605 | #ifdef CONFIG_PROC_FS | |
19b4946c | 1606 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it) |
1da177e4 | 1607 | { |
1efdb69b | 1608 | struct user_namespace *user_ns = seq_user_ns(s); |
ade9f91b KC |
1609 | struct kern_ipc_perm *ipcp = it; |
1610 | struct shmid_kernel *shp; | |
b7952180 HD |
1611 | unsigned long rss = 0, swp = 0; |
1612 | ||
ade9f91b | 1613 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
b7952180 | 1614 | shm_add_rss_swap(shp, &rss, &swp); |
1da177e4 | 1615 | |
6c826818 PM |
1616 | #if BITS_PER_LONG <= 32 |
1617 | #define SIZE_SPEC "%10lu" | |
1618 | #else | |
1619 | #define SIZE_SPEC "%21lu" | |
1620 | #endif | |
1da177e4 | 1621 | |
7f032d6e JP |
1622 | seq_printf(s, |
1623 | "%10d %10d %4o " SIZE_SPEC " %5u %5u " | |
7ff2819e | 1624 | "%5lu %5u %5u %5u %5u %10llu %10llu %10llu " |
7f032d6e JP |
1625 | SIZE_SPEC " " SIZE_SPEC "\n", |
1626 | shp->shm_perm.key, | |
1627 | shp->shm_perm.id, | |
1628 | shp->shm_perm.mode, | |
1629 | shp->shm_segsz, | |
1630 | shp->shm_cprid, | |
1631 | shp->shm_lprid, | |
1632 | shp->shm_nattch, | |
1633 | from_kuid_munged(user_ns, shp->shm_perm.uid), | |
1634 | from_kgid_munged(user_ns, shp->shm_perm.gid), | |
1635 | from_kuid_munged(user_ns, shp->shm_perm.cuid), | |
1636 | from_kgid_munged(user_ns, shp->shm_perm.cgid), | |
1637 | shp->shm_atim, | |
1638 | shp->shm_dtim, | |
1639 | shp->shm_ctim, | |
1640 | rss * PAGE_SIZE, | |
1641 | swp * PAGE_SIZE); | |
1642 | ||
1643 | return 0; | |
1da177e4 LT |
1644 | } |
1645 | #endif |