]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/ipc/shm.c | |
3 | * Copyright (C) 1992, 1993 Krishna Balasubramanian | |
4 | * Many improvements/fixes by Bruno Haible. | |
5 | * Replaced `struct shm_desc' by `struct vm_area_struct', July 1994. | |
6 | * Fixed the shm swap deallocation (shm_unuse()), August 1998 Andrea Arcangeli. | |
7 | * | |
8 | * /proc/sysvipc/shm support (c) 1999 Dragos Acostachioaie <[email protected]> | |
9 | * BIGMEM support, Andrea Arcangeli <[email protected]> | |
10 | * SMP thread shm, Jean-Luc Boyard <[email protected]> | |
11 | * HIGHMEM support, Ingo Molnar <[email protected]> | |
12 | * Make shmmax, shmall, shmmni sysctl'able, Christoph Rohland <[email protected]> | |
13 | * Shared /dev/zero support, Kanoj Sarcar <[email protected]> | |
14 | * Move the mm functionality over to mm/shmem.c, Christoph Rohland <[email protected]> | |
15 | * | |
073115d6 SG |
16 | * support for audit of ipc object properties and permission changes |
17 | * Dustin Kirkland <[email protected]> | |
4e982311 KK |
18 | * |
19 | * namespaces support | |
20 | * OpenVZ, SWsoft Inc. | |
21 | * Pavel Emelianov <[email protected]> | |
c2c737a0 DB |
22 | * |
23 | * Better ipc lock (kern_ipc_perm.lock) handling | |
24 | * Davidlohr Bueso <[email protected]>, June 2013. | |
1da177e4 LT |
25 | */ |
26 | ||
1da177e4 LT |
27 | #include <linux/slab.h> |
28 | #include <linux/mm.h> | |
29 | #include <linux/hugetlb.h> | |
30 | #include <linux/shm.h> | |
31 | #include <linux/init.h> | |
32 | #include <linux/file.h> | |
33 | #include <linux/mman.h> | |
1da177e4 LT |
34 | #include <linux/shmem_fs.h> |
35 | #include <linux/security.h> | |
36 | #include <linux/syscalls.h> | |
37 | #include <linux/audit.h> | |
c59ede7b | 38 | #include <linux/capability.h> |
7d87e14c | 39 | #include <linux/ptrace.h> |
19b4946c | 40 | #include <linux/seq_file.h> |
3e148c79 | 41 | #include <linux/rwsem.h> |
4e982311 | 42 | #include <linux/nsproxy.h> |
bc56bba8 | 43 | #include <linux/mount.h> |
ae5e1b22 | 44 | #include <linux/ipc_namespace.h> |
7d87e14c | 45 | |
7153e402 | 46 | #include <linux/uaccess.h> |
1da177e4 LT |
47 | |
48 | #include "util.h" | |
49 | ||
bc56bba8 EB |
50 | struct shm_file_data { |
51 | int id; | |
52 | struct ipc_namespace *ns; | |
53 | struct file *file; | |
54 | const struct vm_operations_struct *vm_ops; | |
55 | }; | |
56 | ||
57 | #define shm_file_data(file) (*((struct shm_file_data **)&(file)->private_data)) | |
58 | ||
9a32144e | 59 | static const struct file_operations shm_file_operations; |
f0f37e2f | 60 | static const struct vm_operations_struct shm_vm_ops; |
1da177e4 | 61 | |
ed2ddbf8 | 62 | #define shm_ids(ns) ((ns)->ids[IPC_SHM_IDS]) |
1da177e4 | 63 | |
4e982311 KK |
64 | #define shm_unlock(shp) \ |
65 | ipc_unlock(&(shp)->shm_perm) | |
1da177e4 | 66 | |
7748dbfa | 67 | static int newseg(struct ipc_namespace *, struct ipc_params *); |
bc56bba8 EB |
68 | static void shm_open(struct vm_area_struct *vma); |
69 | static void shm_close(struct vm_area_struct *vma); | |
239521f3 | 70 | static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp); |
1da177e4 | 71 | #ifdef CONFIG_PROC_FS |
19b4946c | 72 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it); |
1da177e4 LT |
73 | #endif |
74 | ||
ed2ddbf8 | 75 | void shm_init_ns(struct ipc_namespace *ns) |
4e982311 | 76 | { |
4e982311 KK |
77 | ns->shm_ctlmax = SHMMAX; |
78 | ns->shm_ctlall = SHMALL; | |
79 | ns->shm_ctlmni = SHMMNI; | |
b34a6b1d | 80 | ns->shm_rmid_forced = 0; |
4e982311 | 81 | ns->shm_tot = 0; |
e8148f75 | 82 | ipc_init_ids(&shm_ids(ns)); |
4e982311 KK |
83 | } |
84 | ||
f4566f04 | 85 | /* |
d9a605e4 DB |
86 | * Called with shm_ids.rwsem (writer) and the shp structure locked. |
87 | * Only shm_ids.rwsem remains locked on exit. | |
f4566f04 | 88 | */ |
01b8b07a | 89 | static void do_shm_rmid(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp) |
4e982311 | 90 | { |
01b8b07a | 91 | struct shmid_kernel *shp; |
63980c80 | 92 | |
01b8b07a PP |
93 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
94 | ||
239521f3 | 95 | if (shp->shm_nattch) { |
4e982311 KK |
96 | shp->shm_perm.mode |= SHM_DEST; |
97 | /* Do not find it any more */ | |
98 | shp->shm_perm.key = IPC_PRIVATE; | |
99 | shm_unlock(shp); | |
100 | } else | |
101 | shm_destroy(ns, shp); | |
102 | } | |
103 | ||
ae5e1b22 | 104 | #ifdef CONFIG_IPC_NS |
4e982311 KK |
105 | void shm_exit_ns(struct ipc_namespace *ns) |
106 | { | |
01b8b07a | 107 | free_ipcs(ns, &shm_ids(ns), do_shm_rmid); |
7d6feeb2 | 108 | idr_destroy(&ns->ids[IPC_SHM_IDS].ipcs_idr); |
4e982311 | 109 | } |
ae5e1b22 | 110 | #endif |
1da177e4 | 111 | |
140d0b21 | 112 | static int __init ipc_ns_init(void) |
1da177e4 | 113 | { |
ed2ddbf8 | 114 | shm_init_ns(&init_ipc_ns); |
140d0b21 LT |
115 | return 0; |
116 | } | |
117 | ||
118 | pure_initcall(ipc_ns_init); | |
119 | ||
239521f3 | 120 | void __init shm_init(void) |
140d0b21 | 121 | { |
19b4946c | 122 | ipc_init_proc_interface("sysvipc/shm", |
b7952180 HD |
123 | #if BITS_PER_LONG <= 32 |
124 | " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime rss swap\n", | |
125 | #else | |
126 | " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime rss swap\n", | |
127 | #endif | |
4e982311 | 128 | IPC_SHM_IDS, sysvipc_shm_proc_show); |
1da177e4 LT |
129 | } |
130 | ||
8b8d52ac DB |
131 | static inline struct shmid_kernel *shm_obtain_object(struct ipc_namespace *ns, int id) |
132 | { | |
55b7ae50 | 133 | struct kern_ipc_perm *ipcp = ipc_obtain_object_idr(&shm_ids(ns), id); |
8b8d52ac DB |
134 | |
135 | if (IS_ERR(ipcp)) | |
136 | return ERR_CAST(ipcp); | |
137 | ||
138 | return container_of(ipcp, struct shmid_kernel, shm_perm); | |
139 | } | |
140 | ||
141 | static inline struct shmid_kernel *shm_obtain_object_check(struct ipc_namespace *ns, int id) | |
142 | { | |
143 | struct kern_ipc_perm *ipcp = ipc_obtain_object_check(&shm_ids(ns), id); | |
144 | ||
145 | if (IS_ERR(ipcp)) | |
146 | return ERR_CAST(ipcp); | |
147 | ||
148 | return container_of(ipcp, struct shmid_kernel, shm_perm); | |
149 | } | |
150 | ||
3e148c79 | 151 | /* |
d9a605e4 | 152 | * shm_lock_(check_) routines are called in the paths where the rwsem |
00c2bf85 | 153 | * is not necessarily held. |
3e148c79 | 154 | */ |
023a5355 | 155 | static inline struct shmid_kernel *shm_lock(struct ipc_namespace *ns, int id) |
1da177e4 | 156 | { |
03f02c76 ND |
157 | struct kern_ipc_perm *ipcp = ipc_lock(&shm_ids(ns), id); |
158 | ||
c5c8975b | 159 | /* |
1ac0b6de KS |
160 | * Callers of shm_lock() must validate the status of the returned ipc |
161 | * object pointer (as returned by ipc_lock()), and error out as | |
162 | * appropriate. | |
c5c8975b | 163 | */ |
1ac0b6de KS |
164 | if (IS_ERR(ipcp)) |
165 | return (void *)ipcp; | |
03f02c76 | 166 | return container_of(ipcp, struct shmid_kernel, shm_perm); |
023a5355 ND |
167 | } |
168 | ||
4c677e2e VK |
169 | static inline void shm_lock_by_ptr(struct shmid_kernel *ipcp) |
170 | { | |
171 | rcu_read_lock(); | |
cf9d5d78 | 172 | ipc_lock_object(&ipcp->shm_perm); |
4c677e2e VK |
173 | } |
174 | ||
53dad6d3 DB |
175 | static void shm_rcu_free(struct rcu_head *head) |
176 | { | |
dba4cdd3 MS |
177 | struct kern_ipc_perm *ptr = container_of(head, struct kern_ipc_perm, |
178 | rcu); | |
179 | struct shmid_kernel *shp = container_of(ptr, struct shmid_kernel, | |
180 | shm_perm); | |
53dad6d3 | 181 | security_shm_free(shp); |
42e618f7 | 182 | kvfree(shp); |
53dad6d3 DB |
183 | } |
184 | ||
7ca7e564 | 185 | static inline void shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *s) |
1da177e4 | 186 | { |
ab602f79 | 187 | list_del(&s->shm_clist); |
7ca7e564 | 188 | ipc_rmid(&shm_ids(ns), &s->shm_perm); |
1da177e4 LT |
189 | } |
190 | ||
1da177e4 | 191 | |
1ac0b6de | 192 | static int __shm_open(struct vm_area_struct *vma) |
4e982311 | 193 | { |
bc56bba8 EB |
194 | struct file *file = vma->vm_file; |
195 | struct shm_file_data *sfd = shm_file_data(file); | |
1da177e4 LT |
196 | struct shmid_kernel *shp; |
197 | ||
bc56bba8 | 198 | shp = shm_lock(sfd->ns, sfd->id); |
1ac0b6de KS |
199 | |
200 | if (IS_ERR(shp)) | |
201 | return PTR_ERR(shp); | |
202 | ||
1da177e4 | 203 | shp->shm_atim = get_seconds(); |
b488893a | 204 | shp->shm_lprid = task_tgid_vnr(current); |
1da177e4 LT |
205 | shp->shm_nattch++; |
206 | shm_unlock(shp); | |
1ac0b6de KS |
207 | return 0; |
208 | } | |
209 | ||
210 | /* This is called by fork, once for every shm attach. */ | |
211 | static void shm_open(struct vm_area_struct *vma) | |
212 | { | |
213 | int err = __shm_open(vma); | |
214 | /* | |
215 | * We raced in the idr lookup or with shm_destroy(). | |
216 | * Either way, the ID is busted. | |
217 | */ | |
218 | WARN_ON_ONCE(err); | |
1da177e4 LT |
219 | } |
220 | ||
1da177e4 LT |
221 | /* |
222 | * shm_destroy - free the struct shmid_kernel | |
223 | * | |
f4566f04 | 224 | * @ns: namespace |
1da177e4 LT |
225 | * @shp: struct to free |
226 | * | |
d9a605e4 | 227 | * It has to be called with shp and shm_ids.rwsem (writer) locked, |
1da177e4 LT |
228 | * but returns with shp unlocked and freed. |
229 | */ | |
4e982311 | 230 | static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) |
1da177e4 | 231 | { |
a399b29d GT |
232 | struct file *shm_file; |
233 | ||
234 | shm_file = shp->shm_file; | |
235 | shp->shm_file = NULL; | |
4e982311 | 236 | ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT; |
7ca7e564 | 237 | shm_rmid(ns, shp); |
1da177e4 | 238 | shm_unlock(shp); |
a399b29d GT |
239 | if (!is_file_hugepages(shm_file)) |
240 | shmem_lock(shm_file, 0, shp->mlock_user); | |
353d5c30 | 241 | else if (shp->mlock_user) |
07a46ed2 DH |
242 | user_shm_unlock(i_size_read(file_inode(shm_file)), |
243 | shp->mlock_user); | |
a399b29d | 244 | fput(shm_file); |
dba4cdd3 | 245 | ipc_rcu_putref(&shp->shm_perm, shm_rcu_free); |
1da177e4 LT |
246 | } |
247 | ||
b34a6b1d VK |
248 | /* |
249 | * shm_may_destroy - identifies whether shm segment should be destroyed now | |
250 | * | |
251 | * Returns true if and only if there are no active users of the segment and | |
252 | * one of the following is true: | |
253 | * | |
254 | * 1) shmctl(id, IPC_RMID, NULL) was called for this shp | |
255 | * | |
256 | * 2) sysctl kernel.shm_rmid_forced is set to 1. | |
257 | */ | |
258 | static bool shm_may_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) | |
259 | { | |
260 | return (shp->shm_nattch == 0) && | |
261 | (ns->shm_rmid_forced || | |
262 | (shp->shm_perm.mode & SHM_DEST)); | |
263 | } | |
264 | ||
1da177e4 | 265 | /* |
bc56bba8 | 266 | * remove the attach descriptor vma. |
1da177e4 LT |
267 | * free memory for segment if it is marked destroyed. |
268 | * The descriptor has already been removed from the current->mm->mmap list | |
269 | * and will later be kfree()d. | |
270 | */ | |
bc56bba8 | 271 | static void shm_close(struct vm_area_struct *vma) |
1da177e4 | 272 | { |
239521f3 | 273 | struct file *file = vma->vm_file; |
bc56bba8 | 274 | struct shm_file_data *sfd = shm_file_data(file); |
1da177e4 | 275 | struct shmid_kernel *shp; |
bc56bba8 | 276 | struct ipc_namespace *ns = sfd->ns; |
4e982311 | 277 | |
d9a605e4 | 278 | down_write(&shm_ids(ns).rwsem); |
1da177e4 | 279 | /* remove from the list of attaches of the shm segment */ |
00c2bf85 | 280 | shp = shm_lock(ns, sfd->id); |
1ac0b6de KS |
281 | |
282 | /* | |
283 | * We raced in the idr lookup or with shm_destroy(). | |
284 | * Either way, the ID is busted. | |
285 | */ | |
286 | if (WARN_ON_ONCE(IS_ERR(shp))) | |
287 | goto done; /* no-op */ | |
288 | ||
b488893a | 289 | shp->shm_lprid = task_tgid_vnr(current); |
1da177e4 LT |
290 | shp->shm_dtim = get_seconds(); |
291 | shp->shm_nattch--; | |
b34a6b1d VK |
292 | if (shm_may_destroy(ns, shp)) |
293 | shm_destroy(ns, shp); | |
294 | else | |
295 | shm_unlock(shp); | |
1ac0b6de | 296 | done: |
d9a605e4 | 297 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
298 | } |
299 | ||
d9a605e4 | 300 | /* Called with ns->shm_ids(ns).rwsem locked */ |
b34a6b1d VK |
301 | static int shm_try_destroy_orphaned(int id, void *p, void *data) |
302 | { | |
303 | struct ipc_namespace *ns = data; | |
4c677e2e VK |
304 | struct kern_ipc_perm *ipcp = p; |
305 | struct shmid_kernel *shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
b34a6b1d VK |
306 | |
307 | /* | |
308 | * We want to destroy segments without users and with already | |
309 | * exit'ed originating process. | |
4c677e2e | 310 | * |
d9a605e4 | 311 | * As shp->* are changed under rwsem, it's safe to skip shp locking. |
b34a6b1d | 312 | */ |
4c677e2e | 313 | if (shp->shm_creator != NULL) |
b34a6b1d | 314 | return 0; |
b34a6b1d | 315 | |
4c677e2e VK |
316 | if (shm_may_destroy(ns, shp)) { |
317 | shm_lock_by_ptr(shp); | |
4e982311 | 318 | shm_destroy(ns, shp); |
4c677e2e | 319 | } |
b34a6b1d VK |
320 | return 0; |
321 | } | |
322 | ||
323 | void shm_destroy_orphaned(struct ipc_namespace *ns) | |
324 | { | |
d9a605e4 | 325 | down_write(&shm_ids(ns).rwsem); |
33a30ed4 | 326 | if (shm_ids(ns).in_use) |
4c677e2e | 327 | idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_orphaned, ns); |
d9a605e4 | 328 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
329 | } |
330 | ||
83293c0f | 331 | /* Locking assumes this will only be called with task == current */ |
b34a6b1d VK |
332 | void exit_shm(struct task_struct *task) |
333 | { | |
4c677e2e | 334 | struct ipc_namespace *ns = task->nsproxy->ipc_ns; |
ab602f79 | 335 | struct shmid_kernel *shp, *n; |
b34a6b1d | 336 | |
83293c0f JM |
337 | if (list_empty(&task->sysvshm.shm_clist)) |
338 | return; | |
339 | ||
340 | /* | |
341 | * If kernel.shm_rmid_forced is not set then only keep track of | |
342 | * which shmids are orphaned, so that a later set of the sysctl | |
343 | * can clean them up. | |
344 | */ | |
345 | if (!ns->shm_rmid_forced) { | |
346 | down_read(&shm_ids(ns).rwsem); | |
347 | list_for_each_entry(shp, &task->sysvshm.shm_clist, shm_clist) | |
348 | shp->shm_creator = NULL; | |
349 | /* | |
350 | * Only under read lock but we are only called on current | |
351 | * so no entry on the list will be shared. | |
352 | */ | |
353 | list_del(&task->sysvshm.shm_clist); | |
354 | up_read(&shm_ids(ns).rwsem); | |
298507d4 | 355 | return; |
83293c0f | 356 | } |
298507d4 | 357 | |
83293c0f JM |
358 | /* |
359 | * Destroy all already created segments, that were not yet mapped, | |
360 | * and mark any mapped as orphan to cover the sysctl toggling. | |
361 | * Destroy is skipped if shm_may_destroy() returns false. | |
362 | */ | |
d9a605e4 | 363 | down_write(&shm_ids(ns).rwsem); |
83293c0f JM |
364 | list_for_each_entry_safe(shp, n, &task->sysvshm.shm_clist, shm_clist) { |
365 | shp->shm_creator = NULL; | |
366 | ||
367 | if (shm_may_destroy(ns, shp)) { | |
368 | shm_lock_by_ptr(shp); | |
369 | shm_destroy(ns, shp); | |
370 | } | |
371 | } | |
372 | ||
373 | /* Remove the list head from any segments still attached. */ | |
ab602f79 | 374 | list_del(&task->sysvshm.shm_clist); |
d9a605e4 | 375 | up_write(&shm_ids(ns).rwsem); |
1da177e4 LT |
376 | } |
377 | ||
11bac800 | 378 | static int shm_fault(struct vm_fault *vmf) |
bc56bba8 | 379 | { |
11bac800 | 380 | struct file *file = vmf->vma->vm_file; |
bc56bba8 EB |
381 | struct shm_file_data *sfd = shm_file_data(file); |
382 | ||
11bac800 | 383 | return sfd->vm_ops->fault(vmf); |
bc56bba8 EB |
384 | } |
385 | ||
386 | #ifdef CONFIG_NUMA | |
d823e3e7 | 387 | static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new) |
bc56bba8 EB |
388 | { |
389 | struct file *file = vma->vm_file; | |
390 | struct shm_file_data *sfd = shm_file_data(file); | |
391 | int err = 0; | |
63980c80 | 392 | |
bc56bba8 EB |
393 | if (sfd->vm_ops->set_policy) |
394 | err = sfd->vm_ops->set_policy(vma, new); | |
395 | return err; | |
396 | } | |
397 | ||
d823e3e7 AB |
398 | static struct mempolicy *shm_get_policy(struct vm_area_struct *vma, |
399 | unsigned long addr) | |
bc56bba8 EB |
400 | { |
401 | struct file *file = vma->vm_file; | |
402 | struct shm_file_data *sfd = shm_file_data(file); | |
403 | struct mempolicy *pol = NULL; | |
404 | ||
405 | if (sfd->vm_ops->get_policy) | |
406 | pol = sfd->vm_ops->get_policy(vma, addr); | |
52cd3b07 | 407 | else if (vma->vm_policy) |
bc56bba8 | 408 | pol = vma->vm_policy; |
52cd3b07 | 409 | |
bc56bba8 EB |
410 | return pol; |
411 | } | |
412 | #endif | |
413 | ||
239521f3 | 414 | static int shm_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 415 | { |
bc56bba8 | 416 | struct shm_file_data *sfd = shm_file_data(file); |
b0e15190 DH |
417 | int ret; |
418 | ||
1ac0b6de KS |
419 | /* |
420 | * In case of remap_file_pages() emulation, the file can represent | |
421 | * removed IPC ID: propogate shm_lock() error to caller. | |
422 | */ | |
63980c80 | 423 | ret = __shm_open(vma); |
1ac0b6de KS |
424 | if (ret) |
425 | return ret; | |
426 | ||
f74ac015 | 427 | ret = call_mmap(sfd->file, vma); |
1ac0b6de KS |
428 | if (ret) { |
429 | shm_close(vma); | |
bc56bba8 | 430 | return ret; |
1ac0b6de | 431 | } |
bc56bba8 | 432 | sfd->vm_ops = vma->vm_ops; |
2e92a3ba | 433 | #ifdef CONFIG_MMU |
d0edd852 | 434 | WARN_ON(!sfd->vm_ops->fault); |
2e92a3ba | 435 | #endif |
bc56bba8 | 436 | vma->vm_ops = &shm_vm_ops; |
1ac0b6de | 437 | return 0; |
1da177e4 LT |
438 | } |
439 | ||
4e982311 KK |
440 | static int shm_release(struct inode *ino, struct file *file) |
441 | { | |
bc56bba8 | 442 | struct shm_file_data *sfd = shm_file_data(file); |
4e982311 | 443 | |
bc56bba8 EB |
444 | put_ipc_ns(sfd->ns); |
445 | shm_file_data(file) = NULL; | |
446 | kfree(sfd); | |
4e982311 KK |
447 | return 0; |
448 | } | |
449 | ||
02c24a82 | 450 | static int shm_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
516dffdc | 451 | { |
516dffdc | 452 | struct shm_file_data *sfd = shm_file_data(file); |
516dffdc | 453 | |
7ea80859 CH |
454 | if (!sfd->file->f_op->fsync) |
455 | return -EINVAL; | |
0f41074a | 456 | return sfd->file->f_op->fsync(sfd->file, start, end, datasync); |
516dffdc AL |
457 | } |
458 | ||
7d8a4569 WD |
459 | static long shm_fallocate(struct file *file, int mode, loff_t offset, |
460 | loff_t len) | |
461 | { | |
462 | struct shm_file_data *sfd = shm_file_data(file); | |
463 | ||
464 | if (!sfd->file->f_op->fallocate) | |
465 | return -EOPNOTSUPP; | |
466 | return sfd->file->f_op->fallocate(file, mode, offset, len); | |
467 | } | |
468 | ||
bc56bba8 EB |
469 | static unsigned long shm_get_unmapped_area(struct file *file, |
470 | unsigned long addr, unsigned long len, unsigned long pgoff, | |
471 | unsigned long flags) | |
472 | { | |
473 | struct shm_file_data *sfd = shm_file_data(file); | |
63980c80 | 474 | |
c4caa778 AV |
475 | return sfd->file->f_op->get_unmapped_area(sfd->file, addr, len, |
476 | pgoff, flags); | |
bc56bba8 | 477 | } |
bc56bba8 | 478 | |
9a32144e | 479 | static const struct file_operations shm_file_operations = { |
4e982311 | 480 | .mmap = shm_mmap, |
516dffdc | 481 | .fsync = shm_fsync, |
4e982311 | 482 | .release = shm_release, |
ed5e5894 | 483 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 484 | .llseek = noop_llseek, |
7d8a4569 | 485 | .fallocate = shm_fallocate, |
c4caa778 AV |
486 | }; |
487 | ||
c01d5b30 HD |
488 | /* |
489 | * shm_file_operations_huge is now identical to shm_file_operations, | |
490 | * but we keep it distinct for the sake of is_file_shm_hugepages(). | |
491 | */ | |
c4caa778 AV |
492 | static const struct file_operations shm_file_operations_huge = { |
493 | .mmap = shm_mmap, | |
494 | .fsync = shm_fsync, | |
495 | .release = shm_release, | |
bc56bba8 | 496 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 497 | .llseek = noop_llseek, |
7d8a4569 | 498 | .fallocate = shm_fallocate, |
1da177e4 LT |
499 | }; |
500 | ||
2954e440 | 501 | bool is_file_shm_hugepages(struct file *file) |
c4caa778 AV |
502 | { |
503 | return file->f_op == &shm_file_operations_huge; | |
504 | } | |
505 | ||
f0f37e2f | 506 | static const struct vm_operations_struct shm_vm_ops = { |
1da177e4 LT |
507 | .open = shm_open, /* callback for a new vm-area open */ |
508 | .close = shm_close, /* callback for when the vm-area is released */ | |
54cb8821 | 509 | .fault = shm_fault, |
bc56bba8 EB |
510 | #if defined(CONFIG_NUMA) |
511 | .set_policy = shm_set_policy, | |
512 | .get_policy = shm_get_policy, | |
1da177e4 LT |
513 | #endif |
514 | }; | |
515 | ||
f4566f04 ND |
516 | /** |
517 | * newseg - Create a new shared memory segment | |
518 | * @ns: namespace | |
519 | * @params: ptr to the structure that contains key, size and shmflg | |
520 | * | |
d9a605e4 | 521 | * Called with shm_ids.rwsem held as a writer. |
f4566f04 | 522 | */ |
7748dbfa | 523 | static int newseg(struct ipc_namespace *ns, struct ipc_params *params) |
1da177e4 | 524 | { |
7748dbfa ND |
525 | key_t key = params->key; |
526 | int shmflg = params->flg; | |
527 | size_t size = params->u.size; | |
1da177e4 LT |
528 | int error; |
529 | struct shmid_kernel *shp; | |
d69f3bad | 530 | size_t numpages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; |
239521f3 | 531 | struct file *file; |
1da177e4 | 532 | char name[13]; |
ca16d140 | 533 | vm_flags_t acctflag = 0; |
1da177e4 | 534 | |
4e982311 | 535 | if (size < SHMMIN || size > ns->shm_ctlmax) |
1da177e4 LT |
536 | return -EINVAL; |
537 | ||
1376327c MS |
538 | if (numpages << PAGE_SHIFT < size) |
539 | return -ENOSPC; | |
540 | ||
09c6eb1f MS |
541 | if (ns->shm_tot + numpages < ns->shm_tot || |
542 | ns->shm_tot + numpages > ns->shm_ctlall) | |
1da177e4 LT |
543 | return -ENOSPC; |
544 | ||
42e618f7 KC |
545 | shp = kvmalloc(sizeof(*shp), GFP_KERNEL); |
546 | if (unlikely(!shp)) | |
1da177e4 LT |
547 | return -ENOMEM; |
548 | ||
549 | shp->shm_perm.key = key; | |
b33291c0 | 550 | shp->shm_perm.mode = (shmflg & S_IRWXUGO); |
1da177e4 LT |
551 | shp->mlock_user = NULL; |
552 | ||
553 | shp->shm_perm.security = NULL; | |
554 | error = security_shm_alloc(shp); | |
555 | if (error) { | |
42e618f7 | 556 | kvfree(shp); |
1da177e4 LT |
557 | return error; |
558 | } | |
559 | ||
239521f3 | 560 | sprintf(name, "SYSV%08x", key); |
1da177e4 | 561 | if (shmflg & SHM_HUGETLB) { |
c103a4dc | 562 | struct hstate *hs; |
091d0d55 LZ |
563 | size_t hugesize; |
564 | ||
c103a4dc | 565 | hs = hstate_sizelog((shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); |
091d0d55 LZ |
566 | if (!hs) { |
567 | error = -EINVAL; | |
568 | goto no_file; | |
569 | } | |
570 | hugesize = ALIGN(size, huge_page_size(hs)); | |
af73e4d9 | 571 | |
5a6fe125 MG |
572 | /* hugetlb_file_setup applies strict accounting */ |
573 | if (shmflg & SHM_NORESERVE) | |
574 | acctflag = VM_NORESERVE; | |
af73e4d9 | 575 | file = hugetlb_file_setup(name, hugesize, acctflag, |
42d7395f AK |
576 | &shp->mlock_user, HUGETLB_SHMFS_INODE, |
577 | (shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); | |
1da177e4 | 578 | } else { |
bf8f972d BP |
579 | /* |
580 | * Do not allow no accounting for OVERCOMMIT_NEVER, even | |
239521f3 | 581 | * if it's asked for. |
bf8f972d BP |
582 | */ |
583 | if ((shmflg & SHM_NORESERVE) && | |
584 | sysctl_overcommit_memory != OVERCOMMIT_NEVER) | |
fc8744ad | 585 | acctflag = VM_NORESERVE; |
e1832f29 | 586 | file = shmem_kernel_file_setup(name, size, acctflag); |
1da177e4 LT |
587 | } |
588 | error = PTR_ERR(file); | |
589 | if (IS_ERR(file)) | |
590 | goto no_file; | |
591 | ||
b488893a | 592 | shp->shm_cprid = task_tgid_vnr(current); |
1da177e4 LT |
593 | shp->shm_lprid = 0; |
594 | shp->shm_atim = shp->shm_dtim = 0; | |
595 | shp->shm_ctim = get_seconds(); | |
596 | shp->shm_segsz = size; | |
597 | shp->shm_nattch = 0; | |
1da177e4 | 598 | shp->shm_file = file; |
5774ed01 | 599 | shp->shm_creator = current; |
b9a53227 | 600 | |
a2642f87 MS |
601 | error = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni); |
602 | if (error < 0) | |
b9a53227 | 603 | goto no_id; |
b9a53227 | 604 | |
ab602f79 | 605 | list_add(&shp->shm_clist, ¤t->sysvshm.shm_clist); |
dbfcd91f | 606 | |
30475cc1 BP |
607 | /* |
608 | * shmid gets reported as "inode#" in /proc/pid/maps. | |
609 | * proc-ps tools use this. Changing this will break them. | |
610 | */ | |
496ad9aa | 611 | file_inode(file)->i_ino = shp->shm_perm.id; |
551110a9 | 612 | |
4e982311 | 613 | ns->shm_tot += numpages; |
7ca7e564 | 614 | error = shp->shm_perm.id; |
dbfcd91f | 615 | |
cf9d5d78 | 616 | ipc_unlock_object(&shp->shm_perm); |
dbfcd91f | 617 | rcu_read_unlock(); |
7ca7e564 | 618 | return error; |
1da177e4 LT |
619 | |
620 | no_id: | |
2195d281 | 621 | if (is_file_hugepages(file) && shp->mlock_user) |
353d5c30 | 622 | user_shm_unlock(size, shp->mlock_user); |
1da177e4 LT |
623 | fput(file); |
624 | no_file: | |
a2642f87 | 625 | call_rcu(&shp->shm_perm.rcu, shm_rcu_free); |
1da177e4 LT |
626 | return error; |
627 | } | |
628 | ||
f4566f04 | 629 | /* |
d9a605e4 | 630 | * Called with shm_ids.rwsem and ipcp locked. |
f4566f04 | 631 | */ |
03f02c76 | 632 | static inline int shm_security(struct kern_ipc_perm *ipcp, int shmflg) |
7748dbfa | 633 | { |
03f02c76 ND |
634 | struct shmid_kernel *shp; |
635 | ||
636 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
637 | return security_shm_associate(shp, shmflg); | |
7748dbfa ND |
638 | } |
639 | ||
f4566f04 | 640 | /* |
d9a605e4 | 641 | * Called with shm_ids.rwsem and ipcp locked. |
f4566f04 | 642 | */ |
03f02c76 ND |
643 | static inline int shm_more_checks(struct kern_ipc_perm *ipcp, |
644 | struct ipc_params *params) | |
7748dbfa | 645 | { |
03f02c76 ND |
646 | struct shmid_kernel *shp; |
647 | ||
648 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
649 | if (shp->shm_segsz < params->u.size) | |
7748dbfa ND |
650 | return -EINVAL; |
651 | ||
652 | return 0; | |
653 | } | |
654 | ||
d5460c99 | 655 | SYSCALL_DEFINE3(shmget, key_t, key, size_t, size, int, shmflg) |
1da177e4 | 656 | { |
4e982311 | 657 | struct ipc_namespace *ns; |
eb66ec44 MK |
658 | static const struct ipc_ops shm_ops = { |
659 | .getnew = newseg, | |
660 | .associate = shm_security, | |
661 | .more_checks = shm_more_checks, | |
662 | }; | |
7748dbfa | 663 | struct ipc_params shm_params; |
4e982311 KK |
664 | |
665 | ns = current->nsproxy->ipc_ns; | |
1da177e4 | 666 | |
7748dbfa ND |
667 | shm_params.key = key; |
668 | shm_params.flg = shmflg; | |
669 | shm_params.u.size = size; | |
1da177e4 | 670 | |
7748dbfa | 671 | return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params); |
1da177e4 LT |
672 | } |
673 | ||
674 | static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version) | |
675 | { | |
239521f3 | 676 | switch (version) { |
1da177e4 LT |
677 | case IPC_64: |
678 | return copy_to_user(buf, in, sizeof(*in)); | |
679 | case IPC_OLD: | |
680 | { | |
681 | struct shmid_ds out; | |
682 | ||
3af54c9b | 683 | memset(&out, 0, sizeof(out)); |
1da177e4 LT |
684 | ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm); |
685 | out.shm_segsz = in->shm_segsz; | |
686 | out.shm_atime = in->shm_atime; | |
687 | out.shm_dtime = in->shm_dtime; | |
688 | out.shm_ctime = in->shm_ctime; | |
689 | out.shm_cpid = in->shm_cpid; | |
690 | out.shm_lpid = in->shm_lpid; | |
691 | out.shm_nattch = in->shm_nattch; | |
692 | ||
693 | return copy_to_user(buf, &out, sizeof(out)); | |
694 | } | |
695 | default: | |
696 | return -EINVAL; | |
697 | } | |
698 | } | |
699 | ||
016d7132 PP |
700 | static inline unsigned long |
701 | copy_shmid_from_user(struct shmid64_ds *out, void __user *buf, int version) | |
1da177e4 | 702 | { |
239521f3 | 703 | switch (version) { |
1da177e4 | 704 | case IPC_64: |
016d7132 | 705 | if (copy_from_user(out, buf, sizeof(*out))) |
1da177e4 | 706 | return -EFAULT; |
1da177e4 | 707 | return 0; |
1da177e4 LT |
708 | case IPC_OLD: |
709 | { | |
710 | struct shmid_ds tbuf_old; | |
711 | ||
712 | if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) | |
713 | return -EFAULT; | |
714 | ||
016d7132 PP |
715 | out->shm_perm.uid = tbuf_old.shm_perm.uid; |
716 | out->shm_perm.gid = tbuf_old.shm_perm.gid; | |
717 | out->shm_perm.mode = tbuf_old.shm_perm.mode; | |
1da177e4 LT |
718 | |
719 | return 0; | |
720 | } | |
721 | default: | |
722 | return -EINVAL; | |
723 | } | |
724 | } | |
725 | ||
726 | static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version) | |
727 | { | |
239521f3 | 728 | switch (version) { |
1da177e4 LT |
729 | case IPC_64: |
730 | return copy_to_user(buf, in, sizeof(*in)); | |
731 | case IPC_OLD: | |
732 | { | |
733 | struct shminfo out; | |
734 | ||
239521f3 | 735 | if (in->shmmax > INT_MAX) |
1da177e4 LT |
736 | out.shmmax = INT_MAX; |
737 | else | |
738 | out.shmmax = (int)in->shmmax; | |
739 | ||
740 | out.shmmin = in->shmmin; | |
741 | out.shmmni = in->shmmni; | |
742 | out.shmseg = in->shmseg; | |
46c0a8ca | 743 | out.shmall = in->shmall; |
1da177e4 LT |
744 | |
745 | return copy_to_user(buf, &out, sizeof(out)); | |
746 | } | |
747 | default: | |
748 | return -EINVAL; | |
749 | } | |
750 | } | |
751 | ||
b7952180 HD |
752 | /* |
753 | * Calculate and add used RSS and swap pages of a shm. | |
d9a605e4 | 754 | * Called with shm_ids.rwsem held as a reader |
b7952180 HD |
755 | */ |
756 | static void shm_add_rss_swap(struct shmid_kernel *shp, | |
757 | unsigned long *rss_add, unsigned long *swp_add) | |
758 | { | |
759 | struct inode *inode; | |
760 | ||
496ad9aa | 761 | inode = file_inode(shp->shm_file); |
b7952180 HD |
762 | |
763 | if (is_file_hugepages(shp->shm_file)) { | |
764 | struct address_space *mapping = inode->i_mapping; | |
765 | struct hstate *h = hstate_file(shp->shm_file); | |
766 | *rss_add += pages_per_huge_page(h) * mapping->nrpages; | |
767 | } else { | |
768 | #ifdef CONFIG_SHMEM | |
769 | struct shmem_inode_info *info = SHMEM_I(inode); | |
63980c80 | 770 | |
4595ef88 | 771 | spin_lock_irq(&info->lock); |
b7952180 HD |
772 | *rss_add += inode->i_mapping->nrpages; |
773 | *swp_add += info->swapped; | |
4595ef88 | 774 | spin_unlock_irq(&info->lock); |
b7952180 HD |
775 | #else |
776 | *rss_add += inode->i_mapping->nrpages; | |
777 | #endif | |
778 | } | |
779 | } | |
780 | ||
f4566f04 | 781 | /* |
d9a605e4 | 782 | * Called with shm_ids.rwsem held as a reader |
f4566f04 | 783 | */ |
4e982311 KK |
784 | static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss, |
785 | unsigned long *swp) | |
1da177e4 | 786 | { |
7ca7e564 ND |
787 | int next_id; |
788 | int total, in_use; | |
1da177e4 LT |
789 | |
790 | *rss = 0; | |
791 | *swp = 0; | |
792 | ||
7ca7e564 ND |
793 | in_use = shm_ids(ns).in_use; |
794 | ||
795 | for (total = 0, next_id = 0; total < in_use; next_id++) { | |
e562aebc | 796 | struct kern_ipc_perm *ipc; |
1da177e4 | 797 | struct shmid_kernel *shp; |
1da177e4 | 798 | |
e562aebc TB |
799 | ipc = idr_find(&shm_ids(ns).ipcs_idr, next_id); |
800 | if (ipc == NULL) | |
1da177e4 | 801 | continue; |
e562aebc | 802 | shp = container_of(ipc, struct shmid_kernel, shm_perm); |
1da177e4 | 803 | |
b7952180 | 804 | shm_add_rss_swap(shp, rss, swp); |
7ca7e564 ND |
805 | |
806 | total++; | |
1da177e4 LT |
807 | } |
808 | } | |
809 | ||
8d4cc8b5 | 810 | /* |
d9a605e4 | 811 | * This function handles some shmctl commands which require the rwsem |
8d4cc8b5 | 812 | * to be held in write mode. |
d9a605e4 | 813 | * NOTE: no locks must be held, the rwsem is taken inside this function. |
8d4cc8b5 PP |
814 | */ |
815 | static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd, | |
9ba720c1 | 816 | struct shmid64_ds *shmid64) |
1da177e4 | 817 | { |
8d4cc8b5 | 818 | struct kern_ipc_perm *ipcp; |
8d4cc8b5 PP |
819 | struct shmid_kernel *shp; |
820 | int err; | |
821 | ||
d9a605e4 | 822 | down_write(&shm_ids(ns).rwsem); |
7b4cc5d8 DB |
823 | rcu_read_lock(); |
824 | ||
79ccf0f8 | 825 | ipcp = ipcctl_pre_down_nolock(ns, &shm_ids(ns), shmid, cmd, |
9ba720c1 | 826 | &shmid64->shm_perm, 0); |
7b4cc5d8 DB |
827 | if (IS_ERR(ipcp)) { |
828 | err = PTR_ERR(ipcp); | |
7b4cc5d8 DB |
829 | goto out_unlock1; |
830 | } | |
8d4cc8b5 | 831 | |
a5f75e7f | 832 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
8d4cc8b5 PP |
833 | |
834 | err = security_shm_shmctl(shp, cmd); | |
835 | if (err) | |
79ccf0f8 | 836 | goto out_unlock1; |
7b4cc5d8 | 837 | |
8d4cc8b5 PP |
838 | switch (cmd) { |
839 | case IPC_RMID: | |
79ccf0f8 | 840 | ipc_lock_object(&shp->shm_perm); |
7b4cc5d8 | 841 | /* do_shm_rmid unlocks the ipc object and rcu */ |
8d4cc8b5 PP |
842 | do_shm_rmid(ns, ipcp); |
843 | goto out_up; | |
844 | case IPC_SET: | |
79ccf0f8 | 845 | ipc_lock_object(&shp->shm_perm); |
9ba720c1 | 846 | err = ipc_update_perm(&shmid64->shm_perm, ipcp); |
1efdb69b | 847 | if (err) |
7b4cc5d8 | 848 | goto out_unlock0; |
8d4cc8b5 PP |
849 | shp->shm_ctim = get_seconds(); |
850 | break; | |
851 | default: | |
852 | err = -EINVAL; | |
79ccf0f8 | 853 | goto out_unlock1; |
8d4cc8b5 | 854 | } |
7b4cc5d8 DB |
855 | |
856 | out_unlock0: | |
857 | ipc_unlock_object(&shp->shm_perm); | |
858 | out_unlock1: | |
859 | rcu_read_unlock(); | |
8d4cc8b5 | 860 | out_up: |
d9a605e4 | 861 | up_write(&shm_ids(ns).rwsem); |
8d4cc8b5 PP |
862 | return err; |
863 | } | |
864 | ||
9ba720c1 AV |
865 | static int shmctl_ipc_info(struct ipc_namespace *ns, |
866 | struct shminfo64 *shminfo) | |
8d4cc8b5 | 867 | { |
9ba720c1 AV |
868 | int err = security_shm_shmctl(NULL, IPC_INFO); |
869 | if (!err) { | |
870 | memset(shminfo, 0, sizeof(*shminfo)); | |
871 | shminfo->shmmni = shminfo->shmseg = ns->shm_ctlmni; | |
872 | shminfo->shmmax = ns->shm_ctlmax; | |
873 | shminfo->shmall = ns->shm_ctlall; | |
874 | shminfo->shmmin = SHMMIN; | |
d9a605e4 | 875 | down_read(&shm_ids(ns).rwsem); |
7ca7e564 | 876 | err = ipc_get_maxid(&shm_ids(ns)); |
d9a605e4 | 877 | up_read(&shm_ids(ns).rwsem); |
239521f3 | 878 | if (err < 0) |
1da177e4 | 879 | err = 0; |
1da177e4 | 880 | } |
9ba720c1 AV |
881 | return err; |
882 | } | |
1da177e4 | 883 | |
9ba720c1 AV |
884 | static int shmctl_shm_info(struct ipc_namespace *ns, |
885 | struct shm_info *shm_info) | |
886 | { | |
887 | int err = security_shm_shmctl(NULL, SHM_INFO); | |
888 | if (!err) { | |
889 | memset(shm_info, 0, sizeof(*shm_info)); | |
d9a605e4 | 890 | down_read(&shm_ids(ns).rwsem); |
9ba720c1 AV |
891 | shm_info->used_ids = shm_ids(ns).in_use; |
892 | shm_get_stat(ns, &shm_info->shm_rss, &shm_info->shm_swp); | |
893 | shm_info->shm_tot = ns->shm_tot; | |
894 | shm_info->swap_attempts = 0; | |
895 | shm_info->swap_successes = 0; | |
7ca7e564 | 896 | err = ipc_get_maxid(&shm_ids(ns)); |
d9a605e4 | 897 | up_read(&shm_ids(ns).rwsem); |
9ba720c1 AV |
898 | if (err < 0) |
899 | err = 0; | |
1da177e4 | 900 | } |
9ba720c1 AV |
901 | return err; |
902 | } | |
c97cb9cc | 903 | |
9ba720c1 AV |
904 | static int shmctl_stat(struct ipc_namespace *ns, int shmid, |
905 | int cmd, struct shmid64_ds *tbuf) | |
906 | { | |
907 | struct shmid_kernel *shp; | |
908 | int result; | |
909 | int err; | |
c97cb9cc | 910 | |
9ba720c1 AV |
911 | rcu_read_lock(); |
912 | if (cmd == SHM_STAT) { | |
913 | shp = shm_obtain_object(ns, shmid); | |
914 | if (IS_ERR(shp)) { | |
915 | err = PTR_ERR(shp); | |
916 | goto out_unlock; | |
917 | } | |
918 | result = shp->shm_perm.id; | |
919 | } else { | |
920 | shp = shm_obtain_object_check(ns, shmid); | |
921 | if (IS_ERR(shp)) { | |
922 | err = PTR_ERR(shp); | |
1da177e4 | 923 | goto out_unlock; |
9ba720c1 AV |
924 | } |
925 | result = 0; | |
926 | } | |
c97cb9cc | 927 | |
9ba720c1 AV |
928 | err = -EACCES; |
929 | if (ipcperms(ns, &shp->shm_perm, S_IRUGO)) | |
930 | goto out_unlock; | |
c97cb9cc | 931 | |
9ba720c1 AV |
932 | err = security_shm_shmctl(shp, cmd); |
933 | if (err) | |
934 | goto out_unlock; | |
935 | ||
936 | memset(tbuf, 0, sizeof(*tbuf)); | |
937 | kernel_to_ipc64_perm(&shp->shm_perm, &tbuf->shm_perm); | |
938 | tbuf->shm_segsz = shp->shm_segsz; | |
939 | tbuf->shm_atime = shp->shm_atim; | |
940 | tbuf->shm_dtime = shp->shm_dtim; | |
941 | tbuf->shm_ctime = shp->shm_ctim; | |
942 | tbuf->shm_cpid = shp->shm_cprid; | |
943 | tbuf->shm_lpid = shp->shm_lprid; | |
944 | tbuf->shm_nattch = shp->shm_nattch; | |
945 | rcu_read_unlock(); | |
946 | return result; | |
68eccc1d DB |
947 | |
948 | out_unlock: | |
c97cb9cc | 949 | rcu_read_unlock(); |
68eccc1d DB |
950 | return err; |
951 | } | |
952 | ||
9ba720c1 | 953 | static int shmctl_do_lock(struct ipc_namespace *ns, int shmid, int cmd) |
68eccc1d DB |
954 | { |
955 | struct shmid_kernel *shp; | |
9ba720c1 AV |
956 | struct file *shm_file; |
957 | int err; | |
68eccc1d | 958 | |
9ba720c1 AV |
959 | rcu_read_lock(); |
960 | shp = shm_obtain_object_check(ns, shmid); | |
961 | if (IS_ERR(shp)) { | |
962 | err = PTR_ERR(shp); | |
963 | goto out_unlock1; | |
964 | } | |
68eccc1d | 965 | |
9ba720c1 AV |
966 | audit_ipc_obj(&(shp->shm_perm)); |
967 | err = security_shm_shmctl(shp, cmd); | |
968 | if (err) | |
969 | goto out_unlock1; | |
89e004ea | 970 | |
9ba720c1 | 971 | ipc_lock_object(&shp->shm_perm); |
1da177e4 | 972 | |
9ba720c1 AV |
973 | /* check if shm_destroy() is tearing down shp */ |
974 | if (!ipc_valid_object(&shp->shm_perm)) { | |
975 | err = -EIDRM; | |
976 | goto out_unlock0; | |
977 | } | |
073115d6 | 978 | |
9ba720c1 AV |
979 | if (!ns_capable(ns->user_ns, CAP_IPC_LOCK)) { |
980 | kuid_t euid = current_euid(); | |
0f3d2b01 | 981 | |
9ba720c1 AV |
982 | if (!uid_eq(euid, shp->shm_perm.uid) && |
983 | !uid_eq(euid, shp->shm_perm.cuid)) { | |
984 | err = -EPERM; | |
0f3d2b01 RA |
985 | goto out_unlock0; |
986 | } | |
9ba720c1 AV |
987 | if (cmd == SHM_LOCK && !rlimit(RLIMIT_MEMLOCK)) { |
988 | err = -EPERM; | |
989 | goto out_unlock0; | |
1da177e4 | 990 | } |
9ba720c1 | 991 | } |
1da177e4 | 992 | |
9ba720c1 AV |
993 | shm_file = shp->shm_file; |
994 | if (is_file_hugepages(shm_file)) | |
995 | goto out_unlock0; | |
85046579 | 996 | |
9ba720c1 AV |
997 | if (cmd == SHM_LOCK) { |
998 | struct user_struct *user = current_user(); | |
63980c80 | 999 | |
9ba720c1 AV |
1000 | err = shmem_lock(shm_file, 1, user); |
1001 | if (!err && !(shp->shm_perm.mode & SHM_LOCKED)) { | |
1002 | shp->shm_perm.mode |= SHM_LOCKED; | |
1003 | shp->mlock_user = user; | |
1da177e4 | 1004 | } |
9ba720c1 | 1005 | goto out_unlock0; |
1da177e4 LT |
1006 | } |
1007 | ||
9ba720c1 AV |
1008 | /* SHM_UNLOCK */ |
1009 | if (!(shp->shm_perm.mode & SHM_LOCKED)) | |
1010 | goto out_unlock0; | |
1011 | shmem_lock(shm_file, 0, shp->mlock_user); | |
1012 | shp->shm_perm.mode &= ~SHM_LOCKED; | |
1013 | shp->mlock_user = NULL; | |
1014 | get_file(shm_file); | |
1015 | ipc_unlock_object(&shp->shm_perm); | |
1016 | rcu_read_unlock(); | |
1017 | shmem_unlock_mapping(shm_file->f_mapping); | |
1018 | ||
1019 | fput(shm_file); | |
1020 | return err; | |
1021 | ||
2caacaa8 DB |
1022 | out_unlock0: |
1023 | ipc_unlock_object(&shp->shm_perm); | |
1024 | out_unlock1: | |
1025 | rcu_read_unlock(); | |
1da177e4 LT |
1026 | return err; |
1027 | } | |
1028 | ||
9ba720c1 AV |
1029 | SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, struct shmid_ds __user *, buf) |
1030 | { | |
1031 | int err, version; | |
1032 | struct ipc_namespace *ns; | |
553f770e | 1033 | struct shmid64_ds sem64; |
9ba720c1 AV |
1034 | |
1035 | if (cmd < 0 || shmid < 0) | |
1036 | return -EINVAL; | |
1037 | ||
1038 | version = ipc_parse_version(&cmd); | |
1039 | ns = current->nsproxy->ipc_ns; | |
1040 | ||
1041 | switch (cmd) { | |
1042 | case IPC_INFO: { | |
1043 | struct shminfo64 shminfo; | |
1044 | err = shmctl_ipc_info(ns, &shminfo); | |
1045 | if (err < 0) | |
1046 | return err; | |
1047 | if (copy_shminfo_to_user(buf, &shminfo, version)) | |
1048 | err = -EFAULT; | |
1049 | return err; | |
1050 | } | |
1051 | case SHM_INFO: { | |
1052 | struct shm_info shm_info; | |
1053 | err = shmctl_shm_info(ns, &shm_info); | |
1054 | if (err < 0) | |
1055 | return err; | |
1056 | if (copy_to_user(buf, &shm_info, sizeof(shm_info))) | |
1057 | err = -EFAULT; | |
1058 | return err; | |
1059 | } | |
1060 | case SHM_STAT: | |
1061 | case IPC_STAT: { | |
553f770e | 1062 | err = shmctl_stat(ns, shmid, cmd, &sem64); |
9ba720c1 AV |
1063 | if (err < 0) |
1064 | return err; | |
553f770e | 1065 | if (copy_shmid_to_user(buf, &sem64, version)) |
9ba720c1 AV |
1066 | err = -EFAULT; |
1067 | return err; | |
1068 | } | |
1069 | case IPC_SET: | |
553f770e | 1070 | if (copy_shmid_from_user(&sem64, buf, version)) |
9ba720c1 | 1071 | return -EFAULT; |
553f770e | 1072 | /* fallthru */ |
9ba720c1 | 1073 | case IPC_RMID: |
553f770e | 1074 | return shmctl_down(ns, shmid, cmd, &sem64); |
9ba720c1 AV |
1075 | case SHM_LOCK: |
1076 | case SHM_UNLOCK: | |
1077 | return shmctl_do_lock(ns, shmid, cmd); | |
1078 | default: | |
1079 | return -EINVAL; | |
1080 | } | |
1081 | } | |
1082 | ||
553f770e AV |
1083 | #ifdef CONFIG_COMPAT |
1084 | ||
1085 | struct compat_shmid_ds { | |
1086 | struct compat_ipc_perm shm_perm; | |
1087 | int shm_segsz; | |
1088 | compat_time_t shm_atime; | |
1089 | compat_time_t shm_dtime; | |
1090 | compat_time_t shm_ctime; | |
1091 | compat_ipc_pid_t shm_cpid; | |
1092 | compat_ipc_pid_t shm_lpid; | |
1093 | unsigned short shm_nattch; | |
1094 | unsigned short shm_unused; | |
1095 | compat_uptr_t shm_unused2; | |
1096 | compat_uptr_t shm_unused3; | |
1097 | }; | |
1098 | ||
1099 | struct compat_shminfo64 { | |
1100 | compat_ulong_t shmmax; | |
1101 | compat_ulong_t shmmin; | |
1102 | compat_ulong_t shmmni; | |
1103 | compat_ulong_t shmseg; | |
1104 | compat_ulong_t shmall; | |
1105 | compat_ulong_t __unused1; | |
1106 | compat_ulong_t __unused2; | |
1107 | compat_ulong_t __unused3; | |
1108 | compat_ulong_t __unused4; | |
1109 | }; | |
1110 | ||
1111 | struct compat_shm_info { | |
1112 | compat_int_t used_ids; | |
1113 | compat_ulong_t shm_tot, shm_rss, shm_swp; | |
1114 | compat_ulong_t swap_attempts, swap_successes; | |
1115 | }; | |
1116 | ||
1117 | static int copy_compat_shminfo_to_user(void __user *buf, struct shminfo64 *in, | |
1118 | int version) | |
1119 | { | |
1120 | if (in->shmmax > INT_MAX) | |
1121 | in->shmmax = INT_MAX; | |
1122 | if (version == IPC_64) { | |
1123 | struct compat_shminfo64 info; | |
1124 | memset(&info, 0, sizeof(info)); | |
1125 | info.shmmax = in->shmmax; | |
1126 | info.shmmin = in->shmmin; | |
1127 | info.shmmni = in->shmmni; | |
1128 | info.shmseg = in->shmseg; | |
1129 | info.shmall = in->shmall; | |
1130 | return copy_to_user(buf, &info, sizeof(info)); | |
1131 | } else { | |
1132 | struct shminfo info; | |
1133 | memset(&info, 0, sizeof(info)); | |
1134 | info.shmmax = in->shmmax; | |
1135 | info.shmmin = in->shmmin; | |
1136 | info.shmmni = in->shmmni; | |
1137 | info.shmseg = in->shmseg; | |
1138 | info.shmall = in->shmall; | |
1139 | return copy_to_user(buf, &info, sizeof(info)); | |
1140 | } | |
1141 | } | |
1142 | ||
1143 | static int put_compat_shm_info(struct shm_info *ip, | |
1144 | struct compat_shm_info __user *uip) | |
1145 | { | |
1146 | struct compat_shm_info info; | |
1147 | ||
1148 | memset(&info, 0, sizeof(info)); | |
1149 | info.used_ids = ip->used_ids; | |
1150 | info.shm_tot = ip->shm_tot; | |
1151 | info.shm_rss = ip->shm_rss; | |
1152 | info.shm_swp = ip->shm_swp; | |
1153 | info.swap_attempts = ip->swap_attempts; | |
1154 | info.swap_successes = ip->swap_successes; | |
1155 | return copy_to_user(up, &info, sizeof(info)); | |
1156 | } | |
1157 | ||
1158 | static int copy_compat_shmid_to_user(void __user *buf, struct shmid64_ds *in, | |
1159 | int version) | |
1160 | { | |
1161 | if (version == IPC_64) { | |
1162 | struct compat_shmid64_ds v; | |
1163 | memset(&v, 0, sizeof(v)); | |
28327fae | 1164 | to_compat_ipc64_perm(&v.shm_perm, &in->shm_perm); |
553f770e AV |
1165 | v.shm_atime = in->shm_atime; |
1166 | v.shm_dtime = in->shm_dtime; | |
1167 | v.shm_ctime = in->shm_ctime; | |
1168 | v.shm_segsz = in->shm_segsz; | |
1169 | v.shm_nattch = in->shm_nattch; | |
1170 | v.shm_cpid = in->shm_cpid; | |
1171 | v.shm_lpid = in->shm_lpid; | |
1172 | return copy_to_user(buf, &v, sizeof(v)); | |
1173 | } else { | |
1174 | struct compat_shmid_ds v; | |
1175 | memset(&v, 0, sizeof(v)); | |
28327fae | 1176 | to_compat_ipc_perm(&v.shm_perm, &in->shm_perm); |
553f770e | 1177 | v.shm_perm.key = in->shm_perm.key; |
553f770e AV |
1178 | v.shm_atime = in->shm_atime; |
1179 | v.shm_dtime = in->shm_dtime; | |
1180 | v.shm_ctime = in->shm_ctime; | |
1181 | v.shm_segsz = in->shm_segsz; | |
1182 | v.shm_nattch = in->shm_nattch; | |
1183 | v.shm_cpid = in->shm_cpid; | |
1184 | v.shm_lpid = in->shm_lpid; | |
1185 | return copy_to_user(buf, &v, sizeof(v)); | |
1186 | } | |
1187 | } | |
1188 | ||
1189 | static int copy_compat_shmid_from_user(struct shmid64_ds *out, void __user *buf, | |
1190 | int version) | |
1191 | { | |
1192 | memset(out, 0, sizeof(*out)); | |
1193 | if (version == IPC_64) { | |
1194 | struct compat_shmid64_ds *p = buf; | |
28327fae | 1195 | return get_compat_ipc64_perm(&out->shm_perm, &p->shm_perm); |
553f770e AV |
1196 | } else { |
1197 | struct compat_shmid_ds *p = buf; | |
28327fae | 1198 | return get_compat_ipc_perm(&out->shm_perm, &p->shm_perm); |
553f770e | 1199 | } |
553f770e AV |
1200 | } |
1201 | ||
1202 | COMPAT_SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, void __user *, uptr) | |
1203 | { | |
1204 | struct ipc_namespace *ns; | |
1205 | struct shmid64_ds sem64; | |
1206 | int version = compat_ipc_parse_version(&cmd); | |
1207 | int err; | |
1208 | ||
1209 | ns = current->nsproxy->ipc_ns; | |
1210 | ||
1211 | if (cmd < 0 || shmid < 0) | |
1212 | return -EINVAL; | |
1213 | ||
1214 | switch (cmd) { | |
1215 | case IPC_INFO: { | |
1216 | struct shminfo64 shminfo; | |
1217 | err = shmctl_ipc_info(ns, &shminfo); | |
1218 | if (err < 0) | |
1219 | return err; | |
1220 | if (copy_compat_shminfo_to_user(uptr, &shminfo, version)) | |
1221 | err = -EFAULT; | |
1222 | return err; | |
1223 | } | |
1224 | case SHM_INFO: { | |
1225 | struct shm_info shm_info; | |
1226 | err = shmctl_shm_info(ns, &shm_info); | |
1227 | if (err < 0) | |
1228 | return err; | |
1229 | if (put_compat_shm_info(&shm_info, uptr)) | |
1230 | err = -EFAULT; | |
1231 | return err; | |
1232 | } | |
1233 | case IPC_STAT: | |
1234 | case SHM_STAT: | |
1235 | err = shmctl_stat(ns, shmid, cmd, &sem64); | |
1236 | if (err < 0) | |
1237 | return err; | |
1238 | if (copy_compat_shmid_to_user(&sem64, uptr, version)) | |
1239 | err = -EFAULT; | |
1240 | return err; | |
1241 | ||
1242 | case IPC_SET: | |
1243 | if (copy_compat_shmid_from_user(&sem64, uptr, version)) | |
1244 | return -EFAULT; | |
1245 | /* fallthru */ | |
1246 | case IPC_RMID: | |
1247 | return shmctl_down(ns, shmid, cmd, &sem64); | |
1248 | case SHM_LOCK: | |
1249 | case SHM_UNLOCK: | |
1250 | return shmctl_do_lock(ns, shmid, cmd); | |
1251 | break; | |
1252 | default: | |
1253 | return -EINVAL; | |
1254 | } | |
1255 | return err; | |
1256 | } | |
1257 | #endif | |
1258 | ||
1da177e4 LT |
1259 | /* |
1260 | * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists. | |
1261 | * | |
1262 | * NOTE! Despite the name, this is NOT a direct system call entrypoint. The | |
1263 | * "raddr" thing points to kernel space, and there has to be a wrapper around | |
1264 | * this. | |
1265 | */ | |
95e91b83 DB |
1266 | long do_shmat(int shmid, char __user *shmaddr, int shmflg, |
1267 | ulong *raddr, unsigned long shmlba) | |
1da177e4 LT |
1268 | { |
1269 | struct shmid_kernel *shp; | |
f0cb8802 | 1270 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1271 | unsigned long size; |
239521f3 | 1272 | struct file *file; |
1da177e4 | 1273 | int err; |
f0cb8802 | 1274 | unsigned long flags = MAP_SHARED; |
1da177e4 | 1275 | unsigned long prot; |
1da177e4 | 1276 | int acc_mode; |
4e982311 | 1277 | struct ipc_namespace *ns; |
bc56bba8 EB |
1278 | struct shm_file_data *sfd; |
1279 | struct path path; | |
aeb5d727 | 1280 | fmode_t f_mode; |
41badc15 | 1281 | unsigned long populate = 0; |
1da177e4 | 1282 | |
bc56bba8 EB |
1283 | err = -EINVAL; |
1284 | if (shmid < 0) | |
1da177e4 | 1285 | goto out; |
f0cb8802 DB |
1286 | |
1287 | if (addr) { | |
079a96ae | 1288 | if (addr & (shmlba - 1)) { |
95e91b83 DB |
1289 | /* |
1290 | * Round down to the nearest multiple of shmlba. | |
1291 | * For sane do_mmap_pgoff() parameters, avoid | |
1292 | * round downs that trigger nil-page and MAP_FIXED. | |
1293 | */ | |
1294 | if ((shmflg & SHM_RND) && addr >= shmlba) | |
1295 | addr &= ~(shmlba - 1); | |
1da177e4 LT |
1296 | else |
1297 | #ifndef __ARCH_FORCE_SHMLBA | |
1298 | if (addr & ~PAGE_MASK) | |
1299 | #endif | |
bc56bba8 | 1300 | goto out; |
1da177e4 | 1301 | } |
1da177e4 | 1302 | |
f0cb8802 DB |
1303 | flags |= MAP_FIXED; |
1304 | } else if ((shmflg & SHM_REMAP)) | |
1305 | goto out; | |
1da177e4 LT |
1306 | |
1307 | if (shmflg & SHM_RDONLY) { | |
1308 | prot = PROT_READ; | |
1da177e4 | 1309 | acc_mode = S_IRUGO; |
bc56bba8 | 1310 | f_mode = FMODE_READ; |
1da177e4 LT |
1311 | } else { |
1312 | prot = PROT_READ | PROT_WRITE; | |
1da177e4 | 1313 | acc_mode = S_IRUGO | S_IWUGO; |
bc56bba8 | 1314 | f_mode = FMODE_READ | FMODE_WRITE; |
1da177e4 LT |
1315 | } |
1316 | if (shmflg & SHM_EXEC) { | |
1317 | prot |= PROT_EXEC; | |
1318 | acc_mode |= S_IXUGO; | |
1319 | } | |
1320 | ||
1321 | /* | |
1322 | * We cannot rely on the fs check since SYSV IPC does have an | |
1323 | * additional creator id... | |
1324 | */ | |
4e982311 | 1325 | ns = current->nsproxy->ipc_ns; |
c2c737a0 DB |
1326 | rcu_read_lock(); |
1327 | shp = shm_obtain_object_check(ns, shmid); | |
023a5355 ND |
1328 | if (IS_ERR(shp)) { |
1329 | err = PTR_ERR(shp); | |
c2c737a0 | 1330 | goto out_unlock; |
023a5355 | 1331 | } |
bc56bba8 EB |
1332 | |
1333 | err = -EACCES; | |
b0e77598 | 1334 | if (ipcperms(ns, &shp->shm_perm, acc_mode)) |
bc56bba8 | 1335 | goto out_unlock; |
1da177e4 LT |
1336 | |
1337 | err = security_shm_shmat(shp, shmaddr, shmflg); | |
bc56bba8 EB |
1338 | if (err) |
1339 | goto out_unlock; | |
1340 | ||
c2c737a0 | 1341 | ipc_lock_object(&shp->shm_perm); |
a399b29d GT |
1342 | |
1343 | /* check if shm_destroy() is tearing down shp */ | |
0f3d2b01 | 1344 | if (!ipc_valid_object(&shp->shm_perm)) { |
a399b29d GT |
1345 | ipc_unlock_object(&shp->shm_perm); |
1346 | err = -EIDRM; | |
1347 | goto out_unlock; | |
1348 | } | |
1349 | ||
2c48b9c4 AV |
1350 | path = shp->shm_file->f_path; |
1351 | path_get(&path); | |
1da177e4 | 1352 | shp->shm_nattch++; |
75c3cfa8 | 1353 | size = i_size_read(d_inode(path.dentry)); |
c2c737a0 DB |
1354 | ipc_unlock_object(&shp->shm_perm); |
1355 | rcu_read_unlock(); | |
1da177e4 | 1356 | |
bc56bba8 EB |
1357 | err = -ENOMEM; |
1358 | sfd = kzalloc(sizeof(*sfd), GFP_KERNEL); | |
f42569b1 DB |
1359 | if (!sfd) { |
1360 | path_put(&path); | |
1361 | goto out_nattch; | |
1362 | } | |
bc56bba8 | 1363 | |
2c48b9c4 AV |
1364 | file = alloc_file(&path, f_mode, |
1365 | is_file_hugepages(shp->shm_file) ? | |
c4caa778 AV |
1366 | &shm_file_operations_huge : |
1367 | &shm_file_operations); | |
39b65252 | 1368 | err = PTR_ERR(file); |
f42569b1 DB |
1369 | if (IS_ERR(file)) { |
1370 | kfree(sfd); | |
1371 | path_put(&path); | |
1372 | goto out_nattch; | |
1373 | } | |
bc56bba8 | 1374 | |
bc56bba8 | 1375 | file->private_data = sfd; |
bc56bba8 | 1376 | file->f_mapping = shp->shm_file->f_mapping; |
7ca7e564 | 1377 | sfd->id = shp->shm_perm.id; |
bc56bba8 EB |
1378 | sfd->ns = get_ipc_ns(ns); |
1379 | sfd->file = shp->shm_file; | |
1380 | sfd->vm_ops = NULL; | |
1381 | ||
8b3ec681 AV |
1382 | err = security_mmap_file(file, prot, flags); |
1383 | if (err) | |
1384 | goto out_fput; | |
1385 | ||
91f4f94e MH |
1386 | if (down_write_killable(¤t->mm->mmap_sem)) { |
1387 | err = -EINTR; | |
1388 | goto out_fput; | |
1389 | } | |
1390 | ||
1da177e4 | 1391 | if (addr && !(shmflg & SHM_REMAP)) { |
bc56bba8 | 1392 | err = -EINVAL; |
247a8ce8 MS |
1393 | if (addr + size < addr) |
1394 | goto invalid; | |
1395 | ||
1da177e4 LT |
1396 | if (find_vma_intersection(current->mm, addr, addr + size)) |
1397 | goto invalid; | |
1da177e4 | 1398 | } |
f42569b1 | 1399 | |
897ab3e0 | 1400 | addr = do_mmap_pgoff(file, addr, size, prot, flags, 0, &populate, NULL); |
bebeb3d6 | 1401 | *raddr = addr; |
bc56bba8 | 1402 | err = 0; |
bebeb3d6 ML |
1403 | if (IS_ERR_VALUE(addr)) |
1404 | err = (long)addr; | |
1da177e4 LT |
1405 | invalid: |
1406 | up_write(¤t->mm->mmap_sem); | |
bebeb3d6 | 1407 | if (populate) |
41badc15 | 1408 | mm_populate(addr, populate); |
1da177e4 | 1409 | |
8b3ec681 | 1410 | out_fput: |
bc56bba8 EB |
1411 | fput(file); |
1412 | ||
1413 | out_nattch: | |
d9a605e4 | 1414 | down_write(&shm_ids(ns).rwsem); |
00c2bf85 | 1415 | shp = shm_lock(ns, shmid); |
1da177e4 | 1416 | shp->shm_nattch--; |
b34a6b1d | 1417 | if (shm_may_destroy(ns, shp)) |
4e982311 | 1418 | shm_destroy(ns, shp); |
1da177e4 LT |
1419 | else |
1420 | shm_unlock(shp); | |
d9a605e4 | 1421 | up_write(&shm_ids(ns).rwsem); |
1da177e4 | 1422 | return err; |
bc56bba8 EB |
1423 | |
1424 | out_unlock: | |
c2c737a0 | 1425 | rcu_read_unlock(); |
f42569b1 DB |
1426 | out: |
1427 | return err; | |
1da177e4 LT |
1428 | } |
1429 | ||
d5460c99 | 1430 | SYSCALL_DEFINE3(shmat, int, shmid, char __user *, shmaddr, int, shmflg) |
7d87e14c SR |
1431 | { |
1432 | unsigned long ret; | |
1433 | long err; | |
1434 | ||
079a96ae | 1435 | err = do_shmat(shmid, shmaddr, shmflg, &ret, SHMLBA); |
7d87e14c SR |
1436 | if (err) |
1437 | return err; | |
1438 | force_successful_syscall_return(); | |
1439 | return (long)ret; | |
1440 | } | |
1441 | ||
1da177e4 LT |
1442 | /* |
1443 | * detach and kill segment if marked destroyed. | |
1444 | * The work is done in shm_close. | |
1445 | */ | |
d5460c99 | 1446 | SYSCALL_DEFINE1(shmdt, char __user *, shmaddr) |
1da177e4 LT |
1447 | { |
1448 | struct mm_struct *mm = current->mm; | |
586c7e6a | 1449 | struct vm_area_struct *vma; |
1da177e4 | 1450 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1451 | int retval = -EINVAL; |
586c7e6a MF |
1452 | #ifdef CONFIG_MMU |
1453 | loff_t size = 0; | |
d3c97900 | 1454 | struct file *file; |
586c7e6a MF |
1455 | struct vm_area_struct *next; |
1456 | #endif | |
1da177e4 | 1457 | |
df1e2fb5 HD |
1458 | if (addr & ~PAGE_MASK) |
1459 | return retval; | |
1460 | ||
91f4f94e MH |
1461 | if (down_write_killable(&mm->mmap_sem)) |
1462 | return -EINTR; | |
1da177e4 LT |
1463 | |
1464 | /* | |
1465 | * This function tries to be smart and unmap shm segments that | |
1466 | * were modified by partial mlock or munmap calls: | |
1467 | * - It first determines the size of the shm segment that should be | |
1468 | * unmapped: It searches for a vma that is backed by shm and that | |
1469 | * started at address shmaddr. It records it's size and then unmaps | |
1470 | * it. | |
1471 | * - Then it unmaps all shm vmas that started at shmaddr and that | |
d3c97900 DH |
1472 | * are within the initially determined size and that are from the |
1473 | * same shm segment from which we determined the size. | |
1da177e4 LT |
1474 | * Errors from do_munmap are ignored: the function only fails if |
1475 | * it's called with invalid parameters or if it's called to unmap | |
1476 | * a part of a vma. Both calls in this function are for full vmas, | |
1477 | * the parameters are directly copied from the vma itself and always | |
1478 | * valid - therefore do_munmap cannot fail. (famous last words?) | |
1479 | */ | |
1480 | /* | |
1481 | * If it had been mremap()'d, the starting address would not | |
1482 | * match the usual checks anyway. So assume all vma's are | |
1483 | * above the starting address given. | |
1484 | */ | |
1485 | vma = find_vma(mm, addr); | |
1486 | ||
8feae131 | 1487 | #ifdef CONFIG_MMU |
1da177e4 LT |
1488 | while (vma) { |
1489 | next = vma->vm_next; | |
1490 | ||
1491 | /* | |
1492 | * Check if the starting address would match, i.e. it's | |
1493 | * a fragment created by mprotect() and/or munmap(), or it | |
1494 | * otherwise it starts at this address with no hassles. | |
1495 | */ | |
bc56bba8 | 1496 | if ((vma->vm_ops == &shm_vm_ops) && |
1da177e4 LT |
1497 | (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) { |
1498 | ||
d3c97900 DH |
1499 | /* |
1500 | * Record the file of the shm segment being | |
1501 | * unmapped. With mremap(), someone could place | |
1502 | * page from another segment but with equal offsets | |
1503 | * in the range we are unmapping. | |
1504 | */ | |
1505 | file = vma->vm_file; | |
07a46ed2 | 1506 | size = i_size_read(file_inode(vma->vm_file)); |
897ab3e0 | 1507 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1508 | /* |
1509 | * We discovered the size of the shm segment, so | |
1510 | * break out of here and fall through to the next | |
1511 | * loop that uses the size information to stop | |
1512 | * searching for matching vma's. | |
1513 | */ | |
1514 | retval = 0; | |
1515 | vma = next; | |
1516 | break; | |
1517 | } | |
1518 | vma = next; | |
1519 | } | |
1520 | ||
1521 | /* | |
1522 | * We need look no further than the maximum address a fragment | |
1523 | * could possibly have landed at. Also cast things to loff_t to | |
25985edc | 1524 | * prevent overflows and make comparisons vs. equal-width types. |
1da177e4 | 1525 | */ |
8e36709d | 1526 | size = PAGE_ALIGN(size); |
1da177e4 LT |
1527 | while (vma && (loff_t)(vma->vm_end - addr) <= size) { |
1528 | next = vma->vm_next; | |
1529 | ||
1530 | /* finding a matching vma now does not alter retval */ | |
bc56bba8 | 1531 | if ((vma->vm_ops == &shm_vm_ops) && |
d3c97900 DH |
1532 | ((vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) && |
1533 | (vma->vm_file == file)) | |
897ab3e0 | 1534 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1535 | vma = next; |
1536 | } | |
1537 | ||
63980c80 | 1538 | #else /* CONFIG_MMU */ |
8feae131 | 1539 | /* under NOMMU conditions, the exact address to be destroyed must be |
63980c80 SP |
1540 | * given |
1541 | */ | |
530fcd16 | 1542 | if (vma && vma->vm_start == addr && vma->vm_ops == &shm_vm_ops) { |
897ab3e0 | 1543 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
8feae131 DH |
1544 | retval = 0; |
1545 | } | |
1546 | ||
1547 | #endif | |
1548 | ||
1da177e4 LT |
1549 | up_write(&mm->mmap_sem); |
1550 | return retval; | |
1551 | } | |
1552 | ||
1553 | #ifdef CONFIG_PROC_FS | |
19b4946c | 1554 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it) |
1da177e4 | 1555 | { |
1efdb69b | 1556 | struct user_namespace *user_ns = seq_user_ns(s); |
19b4946c | 1557 | struct shmid_kernel *shp = it; |
b7952180 HD |
1558 | unsigned long rss = 0, swp = 0; |
1559 | ||
1560 | shm_add_rss_swap(shp, &rss, &swp); | |
1da177e4 | 1561 | |
6c826818 PM |
1562 | #if BITS_PER_LONG <= 32 |
1563 | #define SIZE_SPEC "%10lu" | |
1564 | #else | |
1565 | #define SIZE_SPEC "%21lu" | |
1566 | #endif | |
1da177e4 | 1567 | |
7f032d6e JP |
1568 | seq_printf(s, |
1569 | "%10d %10d %4o " SIZE_SPEC " %5u %5u " | |
1570 | "%5lu %5u %5u %5u %5u %10lu %10lu %10lu " | |
1571 | SIZE_SPEC " " SIZE_SPEC "\n", | |
1572 | shp->shm_perm.key, | |
1573 | shp->shm_perm.id, | |
1574 | shp->shm_perm.mode, | |
1575 | shp->shm_segsz, | |
1576 | shp->shm_cprid, | |
1577 | shp->shm_lprid, | |
1578 | shp->shm_nattch, | |
1579 | from_kuid_munged(user_ns, shp->shm_perm.uid), | |
1580 | from_kgid_munged(user_ns, shp->shm_perm.gid), | |
1581 | from_kuid_munged(user_ns, shp->shm_perm.cuid), | |
1582 | from_kgid_munged(user_ns, shp->shm_perm.cgid), | |
1583 | shp->shm_atim, | |
1584 | shp->shm_dtim, | |
1585 | shp->shm_ctim, | |
1586 | rss * PAGE_SIZE, | |
1587 | swp * PAGE_SIZE); | |
1588 | ||
1589 | return 0; | |
1da177e4 LT |
1590 | } |
1591 | #endif |