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 <dragos@iname.com> | |
9 | * BIGMEM support, Andrea Arcangeli <andrea@suse.de> | |
10 | * SMP thread shm, Jean-Luc Boyard <jean-luc.boyard@siemens.fr> | |
11 | * HIGHMEM support, Ingo Molnar <mingo@redhat.com> | |
12 | * Make shmmax, shmall, shmmni sysctl'able, Christoph Rohland <cr@sap.com> | |
13 | * Shared /dev/zero support, Kanoj Sarcar <kanoj@sgi.com> | |
14 | * Move the mm functionality over to mm/shmem.c, Christoph Rohland <cr@sap.com> | |
15 | * | |
073115d6 SG |
16 | * support for audit of ipc object properties and permission changes |
17 | * Dustin Kirkland <dustin.kirkland@us.ibm.com> | |
4e982311 KK |
18 | * |
19 | * namespaces support | |
20 | * OpenVZ, SWsoft Inc. | |
21 | * Pavel Emelianov <xemul@openvz.org> | |
c2c737a0 DB |
22 | * |
23 | * Better ipc lock (kern_ipc_perm.lock) handling | |
24 | * Davidlohr Bueso <davidlohr.bueso@hp.com>, 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 | ||
66470b18 KC |
175 | static void __shm_free(struct shmid_kernel *shp) |
176 | { | |
177 | kvfree(shp); | |
178 | } | |
179 | ||
53dad6d3 DB |
180 | static void shm_rcu_free(struct rcu_head *head) |
181 | { | |
dba4cdd3 MS |
182 | struct kern_ipc_perm *ptr = container_of(head, struct kern_ipc_perm, |
183 | rcu); | |
184 | struct shmid_kernel *shp = container_of(ptr, struct shmid_kernel, | |
185 | shm_perm); | |
53dad6d3 | 186 | security_shm_free(shp); |
66470b18 | 187 | __shm_free(shp); |
53dad6d3 DB |
188 | } |
189 | ||
7ca7e564 | 190 | static inline void shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *s) |
1da177e4 | 191 | { |
ab602f79 | 192 | list_del(&s->shm_clist); |
7ca7e564 | 193 | ipc_rmid(&shm_ids(ns), &s->shm_perm); |
1da177e4 LT |
194 | } |
195 | ||
1da177e4 | 196 | |
1ac0b6de | 197 | static int __shm_open(struct vm_area_struct *vma) |
4e982311 | 198 | { |
bc56bba8 EB |
199 | struct file *file = vma->vm_file; |
200 | struct shm_file_data *sfd = shm_file_data(file); | |
1da177e4 LT |
201 | struct shmid_kernel *shp; |
202 | ||
bc56bba8 | 203 | shp = shm_lock(sfd->ns, sfd->id); |
1ac0b6de KS |
204 | |
205 | if (IS_ERR(shp)) | |
206 | return PTR_ERR(shp); | |
207 | ||
1da177e4 | 208 | shp->shm_atim = get_seconds(); |
b488893a | 209 | shp->shm_lprid = task_tgid_vnr(current); |
1da177e4 LT |
210 | shp->shm_nattch++; |
211 | shm_unlock(shp); | |
1ac0b6de KS |
212 | return 0; |
213 | } | |
214 | ||
215 | /* This is called by fork, once for every shm attach. */ | |
216 | static void shm_open(struct vm_area_struct *vma) | |
217 | { | |
218 | int err = __shm_open(vma); | |
219 | /* | |
220 | * We raced in the idr lookup or with shm_destroy(). | |
221 | * Either way, the ID is busted. | |
222 | */ | |
223 | WARN_ON_ONCE(err); | |
1da177e4 LT |
224 | } |
225 | ||
1da177e4 LT |
226 | /* |
227 | * shm_destroy - free the struct shmid_kernel | |
228 | * | |
f4566f04 | 229 | * @ns: namespace |
1da177e4 LT |
230 | * @shp: struct to free |
231 | * | |
d9a605e4 | 232 | * It has to be called with shp and shm_ids.rwsem (writer) locked, |
1da177e4 LT |
233 | * but returns with shp unlocked and freed. |
234 | */ | |
4e982311 | 235 | static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) |
1da177e4 | 236 | { |
a399b29d GT |
237 | struct file *shm_file; |
238 | ||
239 | shm_file = shp->shm_file; | |
240 | shp->shm_file = NULL; | |
4e982311 | 241 | ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT; |
7ca7e564 | 242 | shm_rmid(ns, shp); |
1da177e4 | 243 | shm_unlock(shp); |
a399b29d GT |
244 | if (!is_file_hugepages(shm_file)) |
245 | shmem_lock(shm_file, 0, shp->mlock_user); | |
353d5c30 | 246 | else if (shp->mlock_user) |
07a46ed2 DH |
247 | user_shm_unlock(i_size_read(file_inode(shm_file)), |
248 | shp->mlock_user); | |
a399b29d | 249 | fput(shm_file); |
dba4cdd3 | 250 | ipc_rcu_putref(&shp->shm_perm, shm_rcu_free); |
1da177e4 LT |
251 | } |
252 | ||
b34a6b1d VK |
253 | /* |
254 | * shm_may_destroy - identifies whether shm segment should be destroyed now | |
255 | * | |
256 | * Returns true if and only if there are no active users of the segment and | |
257 | * one of the following is true: | |
258 | * | |
259 | * 1) shmctl(id, IPC_RMID, NULL) was called for this shp | |
260 | * | |
261 | * 2) sysctl kernel.shm_rmid_forced is set to 1. | |
262 | */ | |
263 | static bool shm_may_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) | |
264 | { | |
265 | return (shp->shm_nattch == 0) && | |
266 | (ns->shm_rmid_forced || | |
267 | (shp->shm_perm.mode & SHM_DEST)); | |
268 | } | |
269 | ||
1da177e4 | 270 | /* |
bc56bba8 | 271 | * remove the attach descriptor vma. |
1da177e4 LT |
272 | * free memory for segment if it is marked destroyed. |
273 | * The descriptor has already been removed from the current->mm->mmap list | |
274 | * and will later be kfree()d. | |
275 | */ | |
bc56bba8 | 276 | static void shm_close(struct vm_area_struct *vma) |
1da177e4 | 277 | { |
239521f3 | 278 | struct file *file = vma->vm_file; |
bc56bba8 | 279 | struct shm_file_data *sfd = shm_file_data(file); |
1da177e4 | 280 | struct shmid_kernel *shp; |
bc56bba8 | 281 | struct ipc_namespace *ns = sfd->ns; |
4e982311 | 282 | |
d9a605e4 | 283 | down_write(&shm_ids(ns).rwsem); |
1da177e4 | 284 | /* remove from the list of attaches of the shm segment */ |
00c2bf85 | 285 | shp = shm_lock(ns, sfd->id); |
1ac0b6de KS |
286 | |
287 | /* | |
288 | * We raced in the idr lookup or with shm_destroy(). | |
289 | * Either way, the ID is busted. | |
290 | */ | |
291 | if (WARN_ON_ONCE(IS_ERR(shp))) | |
292 | goto done; /* no-op */ | |
293 | ||
b488893a | 294 | shp->shm_lprid = task_tgid_vnr(current); |
1da177e4 LT |
295 | shp->shm_dtim = get_seconds(); |
296 | shp->shm_nattch--; | |
b34a6b1d VK |
297 | if (shm_may_destroy(ns, shp)) |
298 | shm_destroy(ns, shp); | |
299 | else | |
300 | shm_unlock(shp); | |
1ac0b6de | 301 | done: |
d9a605e4 | 302 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
303 | } |
304 | ||
d9a605e4 | 305 | /* Called with ns->shm_ids(ns).rwsem locked */ |
b34a6b1d VK |
306 | static int shm_try_destroy_orphaned(int id, void *p, void *data) |
307 | { | |
308 | struct ipc_namespace *ns = data; | |
4c677e2e VK |
309 | struct kern_ipc_perm *ipcp = p; |
310 | struct shmid_kernel *shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
b34a6b1d VK |
311 | |
312 | /* | |
313 | * We want to destroy segments without users and with already | |
314 | * exit'ed originating process. | |
4c677e2e | 315 | * |
d9a605e4 | 316 | * As shp->* are changed under rwsem, it's safe to skip shp locking. |
b34a6b1d | 317 | */ |
4c677e2e | 318 | if (shp->shm_creator != NULL) |
b34a6b1d | 319 | return 0; |
b34a6b1d | 320 | |
4c677e2e VK |
321 | if (shm_may_destroy(ns, shp)) { |
322 | shm_lock_by_ptr(shp); | |
4e982311 | 323 | shm_destroy(ns, shp); |
4c677e2e | 324 | } |
b34a6b1d VK |
325 | return 0; |
326 | } | |
327 | ||
328 | void shm_destroy_orphaned(struct ipc_namespace *ns) | |
329 | { | |
d9a605e4 | 330 | down_write(&shm_ids(ns).rwsem); |
33a30ed4 | 331 | if (shm_ids(ns).in_use) |
4c677e2e | 332 | idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_orphaned, ns); |
d9a605e4 | 333 | up_write(&shm_ids(ns).rwsem); |
b34a6b1d VK |
334 | } |
335 | ||
83293c0f | 336 | /* Locking assumes this will only be called with task == current */ |
b34a6b1d VK |
337 | void exit_shm(struct task_struct *task) |
338 | { | |
4c677e2e | 339 | struct ipc_namespace *ns = task->nsproxy->ipc_ns; |
ab602f79 | 340 | struct shmid_kernel *shp, *n; |
b34a6b1d | 341 | |
83293c0f JM |
342 | if (list_empty(&task->sysvshm.shm_clist)) |
343 | return; | |
344 | ||
345 | /* | |
346 | * If kernel.shm_rmid_forced is not set then only keep track of | |
347 | * which shmids are orphaned, so that a later set of the sysctl | |
348 | * can clean them up. | |
349 | */ | |
350 | if (!ns->shm_rmid_forced) { | |
351 | down_read(&shm_ids(ns).rwsem); | |
352 | list_for_each_entry(shp, &task->sysvshm.shm_clist, shm_clist) | |
353 | shp->shm_creator = NULL; | |
354 | /* | |
355 | * Only under read lock but we are only called on current | |
356 | * so no entry on the list will be shared. | |
357 | */ | |
358 | list_del(&task->sysvshm.shm_clist); | |
359 | up_read(&shm_ids(ns).rwsem); | |
298507d4 | 360 | return; |
83293c0f | 361 | } |
298507d4 | 362 | |
83293c0f JM |
363 | /* |
364 | * Destroy all already created segments, that were not yet mapped, | |
365 | * and mark any mapped as orphan to cover the sysctl toggling. | |
366 | * Destroy is skipped if shm_may_destroy() returns false. | |
367 | */ | |
d9a605e4 | 368 | down_write(&shm_ids(ns).rwsem); |
83293c0f JM |
369 | list_for_each_entry_safe(shp, n, &task->sysvshm.shm_clist, shm_clist) { |
370 | shp->shm_creator = NULL; | |
371 | ||
372 | if (shm_may_destroy(ns, shp)) { | |
373 | shm_lock_by_ptr(shp); | |
374 | shm_destroy(ns, shp); | |
375 | } | |
376 | } | |
377 | ||
378 | /* Remove the list head from any segments still attached. */ | |
ab602f79 | 379 | list_del(&task->sysvshm.shm_clist); |
d9a605e4 | 380 | up_write(&shm_ids(ns).rwsem); |
1da177e4 LT |
381 | } |
382 | ||
11bac800 | 383 | static int shm_fault(struct vm_fault *vmf) |
bc56bba8 | 384 | { |
11bac800 | 385 | struct file *file = vmf->vma->vm_file; |
bc56bba8 EB |
386 | struct shm_file_data *sfd = shm_file_data(file); |
387 | ||
11bac800 | 388 | return sfd->vm_ops->fault(vmf); |
bc56bba8 EB |
389 | } |
390 | ||
391 | #ifdef CONFIG_NUMA | |
d823e3e7 | 392 | static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new) |
bc56bba8 EB |
393 | { |
394 | struct file *file = vma->vm_file; | |
395 | struct shm_file_data *sfd = shm_file_data(file); | |
396 | int err = 0; | |
63980c80 | 397 | |
bc56bba8 EB |
398 | if (sfd->vm_ops->set_policy) |
399 | err = sfd->vm_ops->set_policy(vma, new); | |
400 | return err; | |
401 | } | |
402 | ||
d823e3e7 AB |
403 | static struct mempolicy *shm_get_policy(struct vm_area_struct *vma, |
404 | unsigned long addr) | |
bc56bba8 EB |
405 | { |
406 | struct file *file = vma->vm_file; | |
407 | struct shm_file_data *sfd = shm_file_data(file); | |
408 | struct mempolicy *pol = NULL; | |
409 | ||
410 | if (sfd->vm_ops->get_policy) | |
411 | pol = sfd->vm_ops->get_policy(vma, addr); | |
52cd3b07 | 412 | else if (vma->vm_policy) |
bc56bba8 | 413 | pol = vma->vm_policy; |
52cd3b07 | 414 | |
bc56bba8 EB |
415 | return pol; |
416 | } | |
417 | #endif | |
418 | ||
239521f3 | 419 | static int shm_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 420 | { |
bc56bba8 | 421 | struct shm_file_data *sfd = shm_file_data(file); |
b0e15190 DH |
422 | int ret; |
423 | ||
1ac0b6de KS |
424 | /* |
425 | * In case of remap_file_pages() emulation, the file can represent | |
426 | * removed IPC ID: propogate shm_lock() error to caller. | |
427 | */ | |
63980c80 | 428 | ret = __shm_open(vma); |
1ac0b6de KS |
429 | if (ret) |
430 | return ret; | |
431 | ||
f74ac015 | 432 | ret = call_mmap(sfd->file, vma); |
1ac0b6de KS |
433 | if (ret) { |
434 | shm_close(vma); | |
bc56bba8 | 435 | return ret; |
1ac0b6de | 436 | } |
bc56bba8 | 437 | sfd->vm_ops = vma->vm_ops; |
2e92a3ba | 438 | #ifdef CONFIG_MMU |
d0edd852 | 439 | WARN_ON(!sfd->vm_ops->fault); |
2e92a3ba | 440 | #endif |
bc56bba8 | 441 | vma->vm_ops = &shm_vm_ops; |
1ac0b6de | 442 | return 0; |
1da177e4 LT |
443 | } |
444 | ||
4e982311 KK |
445 | static int shm_release(struct inode *ino, struct file *file) |
446 | { | |
bc56bba8 | 447 | struct shm_file_data *sfd = shm_file_data(file); |
4e982311 | 448 | |
bc56bba8 EB |
449 | put_ipc_ns(sfd->ns); |
450 | shm_file_data(file) = NULL; | |
451 | kfree(sfd); | |
4e982311 KK |
452 | return 0; |
453 | } | |
454 | ||
02c24a82 | 455 | static int shm_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
516dffdc | 456 | { |
516dffdc | 457 | struct shm_file_data *sfd = shm_file_data(file); |
516dffdc | 458 | |
7ea80859 CH |
459 | if (!sfd->file->f_op->fsync) |
460 | return -EINVAL; | |
0f41074a | 461 | return sfd->file->f_op->fsync(sfd->file, start, end, datasync); |
516dffdc AL |
462 | } |
463 | ||
7d8a4569 WD |
464 | static long shm_fallocate(struct file *file, int mode, loff_t offset, |
465 | loff_t len) | |
466 | { | |
467 | struct shm_file_data *sfd = shm_file_data(file); | |
468 | ||
469 | if (!sfd->file->f_op->fallocate) | |
470 | return -EOPNOTSUPP; | |
471 | return sfd->file->f_op->fallocate(file, mode, offset, len); | |
472 | } | |
473 | ||
bc56bba8 EB |
474 | static unsigned long shm_get_unmapped_area(struct file *file, |
475 | unsigned long addr, unsigned long len, unsigned long pgoff, | |
476 | unsigned long flags) | |
477 | { | |
478 | struct shm_file_data *sfd = shm_file_data(file); | |
63980c80 | 479 | |
c4caa778 AV |
480 | return sfd->file->f_op->get_unmapped_area(sfd->file, addr, len, |
481 | pgoff, flags); | |
bc56bba8 | 482 | } |
bc56bba8 | 483 | |
9a32144e | 484 | static const struct file_operations shm_file_operations = { |
4e982311 | 485 | .mmap = shm_mmap, |
516dffdc | 486 | .fsync = shm_fsync, |
4e982311 | 487 | .release = shm_release, |
ed5e5894 | 488 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 489 | .llseek = noop_llseek, |
7d8a4569 | 490 | .fallocate = shm_fallocate, |
c4caa778 AV |
491 | }; |
492 | ||
c01d5b30 HD |
493 | /* |
494 | * shm_file_operations_huge is now identical to shm_file_operations, | |
495 | * but we keep it distinct for the sake of is_file_shm_hugepages(). | |
496 | */ | |
c4caa778 AV |
497 | static const struct file_operations shm_file_operations_huge = { |
498 | .mmap = shm_mmap, | |
499 | .fsync = shm_fsync, | |
500 | .release = shm_release, | |
bc56bba8 | 501 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 502 | .llseek = noop_llseek, |
7d8a4569 | 503 | .fallocate = shm_fallocate, |
1da177e4 LT |
504 | }; |
505 | ||
2954e440 | 506 | bool is_file_shm_hugepages(struct file *file) |
c4caa778 AV |
507 | { |
508 | return file->f_op == &shm_file_operations_huge; | |
509 | } | |
510 | ||
f0f37e2f | 511 | static const struct vm_operations_struct shm_vm_ops = { |
1da177e4 LT |
512 | .open = shm_open, /* callback for a new vm-area open */ |
513 | .close = shm_close, /* callback for when the vm-area is released */ | |
54cb8821 | 514 | .fault = shm_fault, |
bc56bba8 EB |
515 | #if defined(CONFIG_NUMA) |
516 | .set_policy = shm_set_policy, | |
517 | .get_policy = shm_get_policy, | |
1da177e4 LT |
518 | #endif |
519 | }; | |
520 | ||
3e0c2404 KC |
521 | static struct shmid_kernel *shm_alloc(void) |
522 | { | |
523 | struct shmid_kernel *shp; | |
524 | ||
525 | shp = kvmalloc(sizeof(*shp), GFP_KERNEL); | |
526 | if (unlikely(!shp)) | |
527 | return NULL; | |
528 | ||
529 | atomic_set(&shp->shm_perm.refcount, 1); | |
530 | ||
531 | return shp; | |
532 | } | |
533 | ||
f4566f04 ND |
534 | /** |
535 | * newseg - Create a new shared memory segment | |
536 | * @ns: namespace | |
537 | * @params: ptr to the structure that contains key, size and shmflg | |
538 | * | |
d9a605e4 | 539 | * Called with shm_ids.rwsem held as a writer. |
f4566f04 | 540 | */ |
7748dbfa | 541 | static int newseg(struct ipc_namespace *ns, struct ipc_params *params) |
1da177e4 | 542 | { |
7748dbfa ND |
543 | key_t key = params->key; |
544 | int shmflg = params->flg; | |
545 | size_t size = params->u.size; | |
1da177e4 LT |
546 | int error; |
547 | struct shmid_kernel *shp; | |
d69f3bad | 548 | size_t numpages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; |
239521f3 | 549 | struct file *file; |
1da177e4 LT |
550 | char name[13]; |
551 | int id; | |
ca16d140 | 552 | vm_flags_t acctflag = 0; |
1da177e4 | 553 | |
4e982311 | 554 | if (size < SHMMIN || size > ns->shm_ctlmax) |
1da177e4 LT |
555 | return -EINVAL; |
556 | ||
1376327c MS |
557 | if (numpages << PAGE_SHIFT < size) |
558 | return -ENOSPC; | |
559 | ||
09c6eb1f MS |
560 | if (ns->shm_tot + numpages < ns->shm_tot || |
561 | ns->shm_tot + numpages > ns->shm_ctlall) | |
1da177e4 LT |
562 | return -ENOSPC; |
563 | ||
3e0c2404 | 564 | shp = shm_alloc(); |
1da177e4 LT |
565 | if (!shp) |
566 | return -ENOMEM; | |
567 | ||
568 | shp->shm_perm.key = key; | |
b33291c0 | 569 | shp->shm_perm.mode = (shmflg & S_IRWXUGO); |
1da177e4 LT |
570 | shp->mlock_user = NULL; |
571 | ||
572 | shp->shm_perm.security = NULL; | |
573 | error = security_shm_alloc(shp); | |
574 | if (error) { | |
66470b18 | 575 | __shm_free(shp); |
1da177e4 LT |
576 | return error; |
577 | } | |
578 | ||
239521f3 | 579 | sprintf(name, "SYSV%08x", key); |
1da177e4 | 580 | if (shmflg & SHM_HUGETLB) { |
c103a4dc | 581 | struct hstate *hs; |
091d0d55 LZ |
582 | size_t hugesize; |
583 | ||
c103a4dc | 584 | hs = hstate_sizelog((shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); |
091d0d55 LZ |
585 | if (!hs) { |
586 | error = -EINVAL; | |
587 | goto no_file; | |
588 | } | |
589 | hugesize = ALIGN(size, huge_page_size(hs)); | |
af73e4d9 | 590 | |
5a6fe125 MG |
591 | /* hugetlb_file_setup applies strict accounting */ |
592 | if (shmflg & SHM_NORESERVE) | |
593 | acctflag = VM_NORESERVE; | |
af73e4d9 | 594 | file = hugetlb_file_setup(name, hugesize, acctflag, |
42d7395f AK |
595 | &shp->mlock_user, HUGETLB_SHMFS_INODE, |
596 | (shmflg >> SHM_HUGE_SHIFT) & SHM_HUGE_MASK); | |
1da177e4 | 597 | } else { |
bf8f972d BP |
598 | /* |
599 | * Do not allow no accounting for OVERCOMMIT_NEVER, even | |
239521f3 | 600 | * if it's asked for. |
bf8f972d BP |
601 | */ |
602 | if ((shmflg & SHM_NORESERVE) && | |
603 | sysctl_overcommit_memory != OVERCOMMIT_NEVER) | |
fc8744ad | 604 | acctflag = VM_NORESERVE; |
e1832f29 | 605 | file = shmem_kernel_file_setup(name, size, acctflag); |
1da177e4 LT |
606 | } |
607 | error = PTR_ERR(file); | |
608 | if (IS_ERR(file)) | |
609 | goto no_file; | |
610 | ||
b488893a | 611 | shp->shm_cprid = task_tgid_vnr(current); |
1da177e4 LT |
612 | shp->shm_lprid = 0; |
613 | shp->shm_atim = shp->shm_dtim = 0; | |
614 | shp->shm_ctim = get_seconds(); | |
615 | shp->shm_segsz = size; | |
616 | shp->shm_nattch = 0; | |
1da177e4 | 617 | shp->shm_file = file; |
5774ed01 | 618 | shp->shm_creator = current; |
b9a53227 LT |
619 | |
620 | id = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni); | |
621 | if (id < 0) { | |
622 | error = id; | |
623 | goto no_id; | |
624 | } | |
625 | ||
ab602f79 | 626 | list_add(&shp->shm_clist, ¤t->sysvshm.shm_clist); |
dbfcd91f | 627 | |
30475cc1 BP |
628 | /* |
629 | * shmid gets reported as "inode#" in /proc/pid/maps. | |
630 | * proc-ps tools use this. Changing this will break them. | |
631 | */ | |
496ad9aa | 632 | file_inode(file)->i_ino = shp->shm_perm.id; |
551110a9 | 633 | |
4e982311 | 634 | ns->shm_tot += numpages; |
7ca7e564 | 635 | error = shp->shm_perm.id; |
dbfcd91f | 636 | |
cf9d5d78 | 637 | ipc_unlock_object(&shp->shm_perm); |
dbfcd91f | 638 | rcu_read_unlock(); |
7ca7e564 | 639 | return error; |
1da177e4 LT |
640 | |
641 | no_id: | |
2195d281 | 642 | if (is_file_hugepages(file) && shp->mlock_user) |
353d5c30 | 643 | user_shm_unlock(size, shp->mlock_user); |
1da177e4 LT |
644 | fput(file); |
645 | no_file: | |
dba4cdd3 | 646 | ipc_rcu_putref(&shp->shm_perm, shm_rcu_free); |
1da177e4 LT |
647 | return error; |
648 | } | |
649 | ||
f4566f04 | 650 | /* |
d9a605e4 | 651 | * Called with shm_ids.rwsem and ipcp locked. |
f4566f04 | 652 | */ |
03f02c76 | 653 | static inline int shm_security(struct kern_ipc_perm *ipcp, int shmflg) |
7748dbfa | 654 | { |
03f02c76 ND |
655 | struct shmid_kernel *shp; |
656 | ||
657 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
658 | return security_shm_associate(shp, shmflg); | |
7748dbfa ND |
659 | } |
660 | ||
f4566f04 | 661 | /* |
d9a605e4 | 662 | * Called with shm_ids.rwsem and ipcp locked. |
f4566f04 | 663 | */ |
03f02c76 ND |
664 | static inline int shm_more_checks(struct kern_ipc_perm *ipcp, |
665 | struct ipc_params *params) | |
7748dbfa | 666 | { |
03f02c76 ND |
667 | struct shmid_kernel *shp; |
668 | ||
669 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
670 | if (shp->shm_segsz < params->u.size) | |
7748dbfa ND |
671 | return -EINVAL; |
672 | ||
673 | return 0; | |
674 | } | |
675 | ||
d5460c99 | 676 | SYSCALL_DEFINE3(shmget, key_t, key, size_t, size, int, shmflg) |
1da177e4 | 677 | { |
4e982311 | 678 | struct ipc_namespace *ns; |
eb66ec44 MK |
679 | static const struct ipc_ops shm_ops = { |
680 | .getnew = newseg, | |
681 | .associate = shm_security, | |
682 | .more_checks = shm_more_checks, | |
683 | }; | |
7748dbfa | 684 | struct ipc_params shm_params; |
4e982311 KK |
685 | |
686 | ns = current->nsproxy->ipc_ns; | |
1da177e4 | 687 | |
7748dbfa ND |
688 | shm_params.key = key; |
689 | shm_params.flg = shmflg; | |
690 | shm_params.u.size = size; | |
1da177e4 | 691 | |
7748dbfa | 692 | return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params); |
1da177e4 LT |
693 | } |
694 | ||
695 | static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version) | |
696 | { | |
239521f3 | 697 | switch (version) { |
1da177e4 LT |
698 | case IPC_64: |
699 | return copy_to_user(buf, in, sizeof(*in)); | |
700 | case IPC_OLD: | |
701 | { | |
702 | struct shmid_ds out; | |
703 | ||
3af54c9b | 704 | memset(&out, 0, sizeof(out)); |
1da177e4 LT |
705 | ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm); |
706 | out.shm_segsz = in->shm_segsz; | |
707 | out.shm_atime = in->shm_atime; | |
708 | out.shm_dtime = in->shm_dtime; | |
709 | out.shm_ctime = in->shm_ctime; | |
710 | out.shm_cpid = in->shm_cpid; | |
711 | out.shm_lpid = in->shm_lpid; | |
712 | out.shm_nattch = in->shm_nattch; | |
713 | ||
714 | return copy_to_user(buf, &out, sizeof(out)); | |
715 | } | |
716 | default: | |
717 | return -EINVAL; | |
718 | } | |
719 | } | |
720 | ||
016d7132 PP |
721 | static inline unsigned long |
722 | copy_shmid_from_user(struct shmid64_ds *out, void __user *buf, int version) | |
1da177e4 | 723 | { |
239521f3 | 724 | switch (version) { |
1da177e4 | 725 | case IPC_64: |
016d7132 | 726 | if (copy_from_user(out, buf, sizeof(*out))) |
1da177e4 | 727 | return -EFAULT; |
1da177e4 | 728 | return 0; |
1da177e4 LT |
729 | case IPC_OLD: |
730 | { | |
731 | struct shmid_ds tbuf_old; | |
732 | ||
733 | if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) | |
734 | return -EFAULT; | |
735 | ||
016d7132 PP |
736 | out->shm_perm.uid = tbuf_old.shm_perm.uid; |
737 | out->shm_perm.gid = tbuf_old.shm_perm.gid; | |
738 | out->shm_perm.mode = tbuf_old.shm_perm.mode; | |
1da177e4 LT |
739 | |
740 | return 0; | |
741 | } | |
742 | default: | |
743 | return -EINVAL; | |
744 | } | |
745 | } | |
746 | ||
747 | static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version) | |
748 | { | |
239521f3 | 749 | switch (version) { |
1da177e4 LT |
750 | case IPC_64: |
751 | return copy_to_user(buf, in, sizeof(*in)); | |
752 | case IPC_OLD: | |
753 | { | |
754 | struct shminfo out; | |
755 | ||
239521f3 | 756 | if (in->shmmax > INT_MAX) |
1da177e4 LT |
757 | out.shmmax = INT_MAX; |
758 | else | |
759 | out.shmmax = (int)in->shmmax; | |
760 | ||
761 | out.shmmin = in->shmmin; | |
762 | out.shmmni = in->shmmni; | |
763 | out.shmseg = in->shmseg; | |
46c0a8ca | 764 | out.shmall = in->shmall; |
1da177e4 LT |
765 | |
766 | return copy_to_user(buf, &out, sizeof(out)); | |
767 | } | |
768 | default: | |
769 | return -EINVAL; | |
770 | } | |
771 | } | |
772 | ||
b7952180 HD |
773 | /* |
774 | * Calculate and add used RSS and swap pages of a shm. | |
d9a605e4 | 775 | * Called with shm_ids.rwsem held as a reader |
b7952180 HD |
776 | */ |
777 | static void shm_add_rss_swap(struct shmid_kernel *shp, | |
778 | unsigned long *rss_add, unsigned long *swp_add) | |
779 | { | |
780 | struct inode *inode; | |
781 | ||
496ad9aa | 782 | inode = file_inode(shp->shm_file); |
b7952180 HD |
783 | |
784 | if (is_file_hugepages(shp->shm_file)) { | |
785 | struct address_space *mapping = inode->i_mapping; | |
786 | struct hstate *h = hstate_file(shp->shm_file); | |
787 | *rss_add += pages_per_huge_page(h) * mapping->nrpages; | |
788 | } else { | |
789 | #ifdef CONFIG_SHMEM | |
790 | struct shmem_inode_info *info = SHMEM_I(inode); | |
63980c80 | 791 | |
4595ef88 | 792 | spin_lock_irq(&info->lock); |
b7952180 HD |
793 | *rss_add += inode->i_mapping->nrpages; |
794 | *swp_add += info->swapped; | |
4595ef88 | 795 | spin_unlock_irq(&info->lock); |
b7952180 HD |
796 | #else |
797 | *rss_add += inode->i_mapping->nrpages; | |
798 | #endif | |
799 | } | |
800 | } | |
801 | ||
f4566f04 | 802 | /* |
d9a605e4 | 803 | * Called with shm_ids.rwsem held as a reader |
f4566f04 | 804 | */ |
4e982311 KK |
805 | static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss, |
806 | unsigned long *swp) | |
1da177e4 | 807 | { |
7ca7e564 ND |
808 | int next_id; |
809 | int total, in_use; | |
1da177e4 LT |
810 | |
811 | *rss = 0; | |
812 | *swp = 0; | |
813 | ||
7ca7e564 ND |
814 | in_use = shm_ids(ns).in_use; |
815 | ||
816 | for (total = 0, next_id = 0; total < in_use; next_id++) { | |
e562aebc | 817 | struct kern_ipc_perm *ipc; |
1da177e4 | 818 | struct shmid_kernel *shp; |
1da177e4 | 819 | |
e562aebc TB |
820 | ipc = idr_find(&shm_ids(ns).ipcs_idr, next_id); |
821 | if (ipc == NULL) | |
1da177e4 | 822 | continue; |
e562aebc | 823 | shp = container_of(ipc, struct shmid_kernel, shm_perm); |
1da177e4 | 824 | |
b7952180 | 825 | shm_add_rss_swap(shp, rss, swp); |
7ca7e564 ND |
826 | |
827 | total++; | |
1da177e4 LT |
828 | } |
829 | } | |
830 | ||
8d4cc8b5 | 831 | /* |
d9a605e4 | 832 | * This function handles some shmctl commands which require the rwsem |
8d4cc8b5 | 833 | * to be held in write mode. |
d9a605e4 | 834 | * NOTE: no locks must be held, the rwsem is taken inside this function. |
8d4cc8b5 PP |
835 | */ |
836 | static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd, | |
837 | struct shmid_ds __user *buf, int version) | |
1da177e4 | 838 | { |
8d4cc8b5 | 839 | struct kern_ipc_perm *ipcp; |
016d7132 | 840 | struct shmid64_ds shmid64; |
8d4cc8b5 PP |
841 | struct shmid_kernel *shp; |
842 | int err; | |
843 | ||
844 | if (cmd == IPC_SET) { | |
016d7132 | 845 | if (copy_shmid_from_user(&shmid64, buf, version)) |
8d4cc8b5 PP |
846 | return -EFAULT; |
847 | } | |
848 | ||
d9a605e4 | 849 | down_write(&shm_ids(ns).rwsem); |
7b4cc5d8 DB |
850 | rcu_read_lock(); |
851 | ||
79ccf0f8 DB |
852 | ipcp = ipcctl_pre_down_nolock(ns, &shm_ids(ns), shmid, cmd, |
853 | &shmid64.shm_perm, 0); | |
7b4cc5d8 DB |
854 | if (IS_ERR(ipcp)) { |
855 | err = PTR_ERR(ipcp); | |
7b4cc5d8 DB |
856 | goto out_unlock1; |
857 | } | |
8d4cc8b5 | 858 | |
a5f75e7f | 859 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
8d4cc8b5 PP |
860 | |
861 | err = security_shm_shmctl(shp, cmd); | |
862 | if (err) | |
79ccf0f8 | 863 | goto out_unlock1; |
7b4cc5d8 | 864 | |
8d4cc8b5 PP |
865 | switch (cmd) { |
866 | case IPC_RMID: | |
79ccf0f8 | 867 | ipc_lock_object(&shp->shm_perm); |
7b4cc5d8 | 868 | /* do_shm_rmid unlocks the ipc object and rcu */ |
8d4cc8b5 PP |
869 | do_shm_rmid(ns, ipcp); |
870 | goto out_up; | |
871 | case IPC_SET: | |
79ccf0f8 | 872 | ipc_lock_object(&shp->shm_perm); |
1efdb69b EB |
873 | err = ipc_update_perm(&shmid64.shm_perm, ipcp); |
874 | if (err) | |
7b4cc5d8 | 875 | goto out_unlock0; |
8d4cc8b5 PP |
876 | shp->shm_ctim = get_seconds(); |
877 | break; | |
878 | default: | |
879 | err = -EINVAL; | |
79ccf0f8 | 880 | goto out_unlock1; |
8d4cc8b5 | 881 | } |
7b4cc5d8 DB |
882 | |
883 | out_unlock0: | |
884 | ipc_unlock_object(&shp->shm_perm); | |
885 | out_unlock1: | |
886 | rcu_read_unlock(); | |
8d4cc8b5 | 887 | out_up: |
d9a605e4 | 888 | up_write(&shm_ids(ns).rwsem); |
8d4cc8b5 PP |
889 | return err; |
890 | } | |
891 | ||
68eccc1d DB |
892 | static int shmctl_nolock(struct ipc_namespace *ns, int shmid, |
893 | int cmd, int version, void __user *buf) | |
8d4cc8b5 | 894 | { |
68eccc1d | 895 | int err; |
1da177e4 | 896 | struct shmid_kernel *shp; |
1da177e4 | 897 | |
68eccc1d DB |
898 | /* preliminary security checks for *_INFO */ |
899 | if (cmd == IPC_INFO || cmd == SHM_INFO) { | |
900 | err = security_shm_shmctl(NULL, cmd); | |
901 | if (err) | |
902 | return err; | |
1da177e4 LT |
903 | } |
904 | ||
68eccc1d | 905 | switch (cmd) { |
1da177e4 LT |
906 | case IPC_INFO: |
907 | { | |
908 | struct shminfo64 shminfo; | |
909 | ||
e8148f75 | 910 | memset(&shminfo, 0, sizeof(shminfo)); |
4e982311 KK |
911 | shminfo.shmmni = shminfo.shmseg = ns->shm_ctlmni; |
912 | shminfo.shmmax = ns->shm_ctlmax; | |
913 | shminfo.shmall = ns->shm_ctlall; | |
1da177e4 LT |
914 | |
915 | shminfo.shmmin = SHMMIN; | |
239521f3 | 916 | if (copy_shminfo_to_user(buf, &shminfo, version)) |
1da177e4 | 917 | return -EFAULT; |
f4566f04 | 918 | |
d9a605e4 | 919 | down_read(&shm_ids(ns).rwsem); |
7ca7e564 | 920 | err = ipc_get_maxid(&shm_ids(ns)); |
d9a605e4 | 921 | up_read(&shm_ids(ns).rwsem); |
f4566f04 | 922 | |
239521f3 | 923 | if (err < 0) |
1da177e4 LT |
924 | err = 0; |
925 | goto out; | |
926 | } | |
927 | case SHM_INFO: | |
928 | { | |
929 | struct shm_info shm_info; | |
930 | ||
e8148f75 | 931 | memset(&shm_info, 0, sizeof(shm_info)); |
d9a605e4 | 932 | down_read(&shm_ids(ns).rwsem); |
4e982311 | 933 | shm_info.used_ids = shm_ids(ns).in_use; |
239521f3 | 934 | shm_get_stat(ns, &shm_info.shm_rss, &shm_info.shm_swp); |
4e982311 | 935 | shm_info.shm_tot = ns->shm_tot; |
1da177e4 LT |
936 | shm_info.swap_attempts = 0; |
937 | shm_info.swap_successes = 0; | |
7ca7e564 | 938 | err = ipc_get_maxid(&shm_ids(ns)); |
d9a605e4 | 939 | up_read(&shm_ids(ns).rwsem); |
e8148f75 | 940 | if (copy_to_user(buf, &shm_info, sizeof(shm_info))) { |
1da177e4 LT |
941 | err = -EFAULT; |
942 | goto out; | |
943 | } | |
944 | ||
945 | err = err < 0 ? 0 : err; | |
946 | goto out; | |
947 | } | |
948 | case SHM_STAT: | |
949 | case IPC_STAT: | |
950 | { | |
951 | struct shmid64_ds tbuf; | |
952 | int result; | |
023a5355 | 953 | |
c97cb9cc | 954 | rcu_read_lock(); |
023a5355 | 955 | if (cmd == SHM_STAT) { |
c97cb9cc | 956 | shp = shm_obtain_object(ns, shmid); |
023a5355 ND |
957 | if (IS_ERR(shp)) { |
958 | err = PTR_ERR(shp); | |
c97cb9cc | 959 | goto out_unlock; |
023a5355 | 960 | } |
7ca7e564 | 961 | result = shp->shm_perm.id; |
1da177e4 | 962 | } else { |
c97cb9cc | 963 | shp = shm_obtain_object_check(ns, shmid); |
023a5355 ND |
964 | if (IS_ERR(shp)) { |
965 | err = PTR_ERR(shp); | |
c97cb9cc | 966 | goto out_unlock; |
023a5355 | 967 | } |
1da177e4 LT |
968 | result = 0; |
969 | } | |
c97cb9cc | 970 | |
e8148f75 | 971 | err = -EACCES; |
b0e77598 | 972 | if (ipcperms(ns, &shp->shm_perm, S_IRUGO)) |
1da177e4 | 973 | goto out_unlock; |
c97cb9cc | 974 | |
1da177e4 LT |
975 | err = security_shm_shmctl(shp, cmd); |
976 | if (err) | |
977 | goto out_unlock; | |
c97cb9cc | 978 | |
023a5355 | 979 | memset(&tbuf, 0, sizeof(tbuf)); |
1da177e4 LT |
980 | kernel_to_ipc64_perm(&shp->shm_perm, &tbuf.shm_perm); |
981 | tbuf.shm_segsz = shp->shm_segsz; | |
982 | tbuf.shm_atime = shp->shm_atim; | |
983 | tbuf.shm_dtime = shp->shm_dtim; | |
984 | tbuf.shm_ctime = shp->shm_ctim; | |
985 | tbuf.shm_cpid = shp->shm_cprid; | |
986 | tbuf.shm_lpid = shp->shm_lprid; | |
bc56bba8 | 987 | tbuf.shm_nattch = shp->shm_nattch; |
c97cb9cc DB |
988 | rcu_read_unlock(); |
989 | ||
990 | if (copy_shmid_to_user(buf, &tbuf, version)) | |
1da177e4 LT |
991 | err = -EFAULT; |
992 | else | |
993 | err = result; | |
994 | goto out; | |
995 | } | |
68eccc1d DB |
996 | default: |
997 | return -EINVAL; | |
998 | } | |
999 | ||
1000 | out_unlock: | |
c97cb9cc | 1001 | rcu_read_unlock(); |
68eccc1d DB |
1002 | out: |
1003 | return err; | |
1004 | } | |
1005 | ||
1006 | SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, struct shmid_ds __user *, buf) | |
1007 | { | |
1008 | struct shmid_kernel *shp; | |
1009 | int err, version; | |
1010 | struct ipc_namespace *ns; | |
1011 | ||
2caacaa8 DB |
1012 | if (cmd < 0 || shmid < 0) |
1013 | return -EINVAL; | |
68eccc1d DB |
1014 | |
1015 | version = ipc_parse_version(&cmd); | |
1016 | ns = current->nsproxy->ipc_ns; | |
1017 | ||
1018 | switch (cmd) { | |
1019 | case IPC_INFO: | |
1020 | case SHM_INFO: | |
1021 | case SHM_STAT: | |
1022 | case IPC_STAT: | |
1023 | return shmctl_nolock(ns, shmid, cmd, version, buf); | |
2caacaa8 DB |
1024 | case IPC_RMID: |
1025 | case IPC_SET: | |
1026 | return shmctl_down(ns, shmid, cmd, buf, version); | |
1da177e4 LT |
1027 | case SHM_LOCK: |
1028 | case SHM_UNLOCK: | |
1029 | { | |
85046579 | 1030 | struct file *shm_file; |
89e004ea | 1031 | |
2caacaa8 DB |
1032 | rcu_read_lock(); |
1033 | shp = shm_obtain_object_check(ns, shmid); | |
023a5355 ND |
1034 | if (IS_ERR(shp)) { |
1035 | err = PTR_ERR(shp); | |
2caacaa8 | 1036 | goto out_unlock1; |
1da177e4 | 1037 | } |
1da177e4 | 1038 | |
a33e6751 | 1039 | audit_ipc_obj(&(shp->shm_perm)); |
2caacaa8 DB |
1040 | err = security_shm_shmctl(shp, cmd); |
1041 | if (err) | |
1042 | goto out_unlock1; | |
073115d6 | 1043 | |
2caacaa8 | 1044 | ipc_lock_object(&shp->shm_perm); |
0f3d2b01 RA |
1045 | |
1046 | /* check if shm_destroy() is tearing down shp */ | |
1047 | if (!ipc_valid_object(&shp->shm_perm)) { | |
1048 | err = -EIDRM; | |
1049 | goto out_unlock0; | |
1050 | } | |
1051 | ||
b0e77598 | 1052 | if (!ns_capable(ns->user_ns, CAP_IPC_LOCK)) { |
1efdb69b | 1053 | kuid_t euid = current_euid(); |
63980c80 | 1054 | |
1efdb69b | 1055 | if (!uid_eq(euid, shp->shm_perm.uid) && |
3a72660b JN |
1056 | !uid_eq(euid, shp->shm_perm.cuid)) { |
1057 | err = -EPERM; | |
2caacaa8 | 1058 | goto out_unlock0; |
3a72660b JN |
1059 | } |
1060 | if (cmd == SHM_LOCK && !rlimit(RLIMIT_MEMLOCK)) { | |
1061 | err = -EPERM; | |
2caacaa8 | 1062 | goto out_unlock0; |
3a72660b | 1063 | } |
1da177e4 LT |
1064 | } |
1065 | ||
85046579 HD |
1066 | shm_file = shp->shm_file; |
1067 | if (is_file_hugepages(shm_file)) | |
2caacaa8 | 1068 | goto out_unlock0; |
85046579 HD |
1069 | |
1070 | if (cmd == SHM_LOCK) { | |
86a264ab | 1071 | struct user_struct *user = current_user(); |
63980c80 | 1072 | |
85046579 HD |
1073 | err = shmem_lock(shm_file, 1, user); |
1074 | if (!err && !(shp->shm_perm.mode & SHM_LOCKED)) { | |
1075 | shp->shm_perm.mode |= SHM_LOCKED; | |
1076 | shp->mlock_user = user; | |
1da177e4 | 1077 | } |
2caacaa8 | 1078 | goto out_unlock0; |
1da177e4 | 1079 | } |
85046579 HD |
1080 | |
1081 | /* SHM_UNLOCK */ | |
1082 | if (!(shp->shm_perm.mode & SHM_LOCKED)) | |
2caacaa8 | 1083 | goto out_unlock0; |
85046579 HD |
1084 | shmem_lock(shm_file, 0, shp->mlock_user); |
1085 | shp->shm_perm.mode &= ~SHM_LOCKED; | |
1086 | shp->mlock_user = NULL; | |
1087 | get_file(shm_file); | |
2caacaa8 DB |
1088 | ipc_unlock_object(&shp->shm_perm); |
1089 | rcu_read_unlock(); | |
24513264 | 1090 | shmem_unlock_mapping(shm_file->f_mapping); |
2caacaa8 | 1091 | |
85046579 | 1092 | fput(shm_file); |
8d4cc8b5 | 1093 | return err; |
2caacaa8 | 1094 | } |
1da177e4 | 1095 | default: |
8d4cc8b5 | 1096 | return -EINVAL; |
1da177e4 LT |
1097 | } |
1098 | ||
2caacaa8 DB |
1099 | out_unlock0: |
1100 | ipc_unlock_object(&shp->shm_perm); | |
1101 | out_unlock1: | |
1102 | rcu_read_unlock(); | |
1da177e4 LT |
1103 | return err; |
1104 | } | |
1105 | ||
1106 | /* | |
1107 | * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists. | |
1108 | * | |
1109 | * NOTE! Despite the name, this is NOT a direct system call entrypoint. The | |
1110 | * "raddr" thing points to kernel space, and there has to be a wrapper around | |
1111 | * this. | |
1112 | */ | |
95e91b83 DB |
1113 | long do_shmat(int shmid, char __user *shmaddr, int shmflg, |
1114 | ulong *raddr, unsigned long shmlba) | |
1da177e4 LT |
1115 | { |
1116 | struct shmid_kernel *shp; | |
f0cb8802 | 1117 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1118 | unsigned long size; |
239521f3 | 1119 | struct file *file; |
1da177e4 | 1120 | int err; |
f0cb8802 | 1121 | unsigned long flags = MAP_SHARED; |
1da177e4 | 1122 | unsigned long prot; |
1da177e4 | 1123 | int acc_mode; |
4e982311 | 1124 | struct ipc_namespace *ns; |
bc56bba8 EB |
1125 | struct shm_file_data *sfd; |
1126 | struct path path; | |
aeb5d727 | 1127 | fmode_t f_mode; |
41badc15 | 1128 | unsigned long populate = 0; |
1da177e4 | 1129 | |
bc56bba8 EB |
1130 | err = -EINVAL; |
1131 | if (shmid < 0) | |
1da177e4 | 1132 | goto out; |
f0cb8802 DB |
1133 | |
1134 | if (addr) { | |
079a96ae | 1135 | if (addr & (shmlba - 1)) { |
95e91b83 DB |
1136 | /* |
1137 | * Round down to the nearest multiple of shmlba. | |
1138 | * For sane do_mmap_pgoff() parameters, avoid | |
1139 | * round downs that trigger nil-page and MAP_FIXED. | |
1140 | */ | |
1141 | if ((shmflg & SHM_RND) && addr >= shmlba) | |
1142 | addr &= ~(shmlba - 1); | |
1da177e4 LT |
1143 | else |
1144 | #ifndef __ARCH_FORCE_SHMLBA | |
1145 | if (addr & ~PAGE_MASK) | |
1146 | #endif | |
bc56bba8 | 1147 | goto out; |
1da177e4 | 1148 | } |
1da177e4 | 1149 | |
f0cb8802 DB |
1150 | flags |= MAP_FIXED; |
1151 | } else if ((shmflg & SHM_REMAP)) | |
1152 | goto out; | |
1da177e4 LT |
1153 | |
1154 | if (shmflg & SHM_RDONLY) { | |
1155 | prot = PROT_READ; | |
1da177e4 | 1156 | acc_mode = S_IRUGO; |
bc56bba8 | 1157 | f_mode = FMODE_READ; |
1da177e4 LT |
1158 | } else { |
1159 | prot = PROT_READ | PROT_WRITE; | |
1da177e4 | 1160 | acc_mode = S_IRUGO | S_IWUGO; |
bc56bba8 | 1161 | f_mode = FMODE_READ | FMODE_WRITE; |
1da177e4 LT |
1162 | } |
1163 | if (shmflg & SHM_EXEC) { | |
1164 | prot |= PROT_EXEC; | |
1165 | acc_mode |= S_IXUGO; | |
1166 | } | |
1167 | ||
1168 | /* | |
1169 | * We cannot rely on the fs check since SYSV IPC does have an | |
1170 | * additional creator id... | |
1171 | */ | |
4e982311 | 1172 | ns = current->nsproxy->ipc_ns; |
c2c737a0 DB |
1173 | rcu_read_lock(); |
1174 | shp = shm_obtain_object_check(ns, shmid); | |
023a5355 ND |
1175 | if (IS_ERR(shp)) { |
1176 | err = PTR_ERR(shp); | |
c2c737a0 | 1177 | goto out_unlock; |
023a5355 | 1178 | } |
bc56bba8 EB |
1179 | |
1180 | err = -EACCES; | |
b0e77598 | 1181 | if (ipcperms(ns, &shp->shm_perm, acc_mode)) |
bc56bba8 | 1182 | goto out_unlock; |
1da177e4 LT |
1183 | |
1184 | err = security_shm_shmat(shp, shmaddr, shmflg); | |
bc56bba8 EB |
1185 | if (err) |
1186 | goto out_unlock; | |
1187 | ||
c2c737a0 | 1188 | ipc_lock_object(&shp->shm_perm); |
a399b29d GT |
1189 | |
1190 | /* check if shm_destroy() is tearing down shp */ | |
0f3d2b01 | 1191 | if (!ipc_valid_object(&shp->shm_perm)) { |
a399b29d GT |
1192 | ipc_unlock_object(&shp->shm_perm); |
1193 | err = -EIDRM; | |
1194 | goto out_unlock; | |
1195 | } | |
1196 | ||
2c48b9c4 AV |
1197 | path = shp->shm_file->f_path; |
1198 | path_get(&path); | |
1da177e4 | 1199 | shp->shm_nattch++; |
75c3cfa8 | 1200 | size = i_size_read(d_inode(path.dentry)); |
c2c737a0 DB |
1201 | ipc_unlock_object(&shp->shm_perm); |
1202 | rcu_read_unlock(); | |
1da177e4 | 1203 | |
bc56bba8 EB |
1204 | err = -ENOMEM; |
1205 | sfd = kzalloc(sizeof(*sfd), GFP_KERNEL); | |
f42569b1 DB |
1206 | if (!sfd) { |
1207 | path_put(&path); | |
1208 | goto out_nattch; | |
1209 | } | |
bc56bba8 | 1210 | |
2c48b9c4 AV |
1211 | file = alloc_file(&path, f_mode, |
1212 | is_file_hugepages(shp->shm_file) ? | |
c4caa778 AV |
1213 | &shm_file_operations_huge : |
1214 | &shm_file_operations); | |
39b65252 | 1215 | err = PTR_ERR(file); |
f42569b1 DB |
1216 | if (IS_ERR(file)) { |
1217 | kfree(sfd); | |
1218 | path_put(&path); | |
1219 | goto out_nattch; | |
1220 | } | |
bc56bba8 | 1221 | |
bc56bba8 | 1222 | file->private_data = sfd; |
bc56bba8 | 1223 | file->f_mapping = shp->shm_file->f_mapping; |
7ca7e564 | 1224 | sfd->id = shp->shm_perm.id; |
bc56bba8 EB |
1225 | sfd->ns = get_ipc_ns(ns); |
1226 | sfd->file = shp->shm_file; | |
1227 | sfd->vm_ops = NULL; | |
1228 | ||
8b3ec681 AV |
1229 | err = security_mmap_file(file, prot, flags); |
1230 | if (err) | |
1231 | goto out_fput; | |
1232 | ||
91f4f94e MH |
1233 | if (down_write_killable(¤t->mm->mmap_sem)) { |
1234 | err = -EINTR; | |
1235 | goto out_fput; | |
1236 | } | |
1237 | ||
1da177e4 | 1238 | if (addr && !(shmflg & SHM_REMAP)) { |
bc56bba8 | 1239 | err = -EINVAL; |
247a8ce8 MS |
1240 | if (addr + size < addr) |
1241 | goto invalid; | |
1242 | ||
1da177e4 LT |
1243 | if (find_vma_intersection(current->mm, addr, addr + size)) |
1244 | goto invalid; | |
1da177e4 | 1245 | } |
f42569b1 | 1246 | |
897ab3e0 | 1247 | addr = do_mmap_pgoff(file, addr, size, prot, flags, 0, &populate, NULL); |
bebeb3d6 | 1248 | *raddr = addr; |
bc56bba8 | 1249 | err = 0; |
bebeb3d6 ML |
1250 | if (IS_ERR_VALUE(addr)) |
1251 | err = (long)addr; | |
1da177e4 LT |
1252 | invalid: |
1253 | up_write(¤t->mm->mmap_sem); | |
bebeb3d6 | 1254 | if (populate) |
41badc15 | 1255 | mm_populate(addr, populate); |
1da177e4 | 1256 | |
8b3ec681 | 1257 | out_fput: |
bc56bba8 EB |
1258 | fput(file); |
1259 | ||
1260 | out_nattch: | |
d9a605e4 | 1261 | down_write(&shm_ids(ns).rwsem); |
00c2bf85 | 1262 | shp = shm_lock(ns, shmid); |
1da177e4 | 1263 | shp->shm_nattch--; |
b34a6b1d | 1264 | if (shm_may_destroy(ns, shp)) |
4e982311 | 1265 | shm_destroy(ns, shp); |
1da177e4 LT |
1266 | else |
1267 | shm_unlock(shp); | |
d9a605e4 | 1268 | up_write(&shm_ids(ns).rwsem); |
1da177e4 | 1269 | return err; |
bc56bba8 EB |
1270 | |
1271 | out_unlock: | |
c2c737a0 | 1272 | rcu_read_unlock(); |
f42569b1 DB |
1273 | out: |
1274 | return err; | |
1da177e4 LT |
1275 | } |
1276 | ||
d5460c99 | 1277 | SYSCALL_DEFINE3(shmat, int, shmid, char __user *, shmaddr, int, shmflg) |
7d87e14c SR |
1278 | { |
1279 | unsigned long ret; | |
1280 | long err; | |
1281 | ||
079a96ae | 1282 | err = do_shmat(shmid, shmaddr, shmflg, &ret, SHMLBA); |
7d87e14c SR |
1283 | if (err) |
1284 | return err; | |
1285 | force_successful_syscall_return(); | |
1286 | return (long)ret; | |
1287 | } | |
1288 | ||
1da177e4 LT |
1289 | /* |
1290 | * detach and kill segment if marked destroyed. | |
1291 | * The work is done in shm_close. | |
1292 | */ | |
d5460c99 | 1293 | SYSCALL_DEFINE1(shmdt, char __user *, shmaddr) |
1da177e4 LT |
1294 | { |
1295 | struct mm_struct *mm = current->mm; | |
586c7e6a | 1296 | struct vm_area_struct *vma; |
1da177e4 | 1297 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1298 | int retval = -EINVAL; |
586c7e6a MF |
1299 | #ifdef CONFIG_MMU |
1300 | loff_t size = 0; | |
d3c97900 | 1301 | struct file *file; |
586c7e6a MF |
1302 | struct vm_area_struct *next; |
1303 | #endif | |
1da177e4 | 1304 | |
df1e2fb5 HD |
1305 | if (addr & ~PAGE_MASK) |
1306 | return retval; | |
1307 | ||
91f4f94e MH |
1308 | if (down_write_killable(&mm->mmap_sem)) |
1309 | return -EINTR; | |
1da177e4 LT |
1310 | |
1311 | /* | |
1312 | * This function tries to be smart and unmap shm segments that | |
1313 | * were modified by partial mlock or munmap calls: | |
1314 | * - It first determines the size of the shm segment that should be | |
1315 | * unmapped: It searches for a vma that is backed by shm and that | |
1316 | * started at address shmaddr. It records it's size and then unmaps | |
1317 | * it. | |
1318 | * - Then it unmaps all shm vmas that started at shmaddr and that | |
d3c97900 DH |
1319 | * are within the initially determined size and that are from the |
1320 | * same shm segment from which we determined the size. | |
1da177e4 LT |
1321 | * Errors from do_munmap are ignored: the function only fails if |
1322 | * it's called with invalid parameters or if it's called to unmap | |
1323 | * a part of a vma. Both calls in this function are for full vmas, | |
1324 | * the parameters are directly copied from the vma itself and always | |
1325 | * valid - therefore do_munmap cannot fail. (famous last words?) | |
1326 | */ | |
1327 | /* | |
1328 | * If it had been mremap()'d, the starting address would not | |
1329 | * match the usual checks anyway. So assume all vma's are | |
1330 | * above the starting address given. | |
1331 | */ | |
1332 | vma = find_vma(mm, addr); | |
1333 | ||
8feae131 | 1334 | #ifdef CONFIG_MMU |
1da177e4 LT |
1335 | while (vma) { |
1336 | next = vma->vm_next; | |
1337 | ||
1338 | /* | |
1339 | * Check if the starting address would match, i.e. it's | |
1340 | * a fragment created by mprotect() and/or munmap(), or it | |
1341 | * otherwise it starts at this address with no hassles. | |
1342 | */ | |
bc56bba8 | 1343 | if ((vma->vm_ops == &shm_vm_ops) && |
1da177e4 LT |
1344 | (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) { |
1345 | ||
d3c97900 DH |
1346 | /* |
1347 | * Record the file of the shm segment being | |
1348 | * unmapped. With mremap(), someone could place | |
1349 | * page from another segment but with equal offsets | |
1350 | * in the range we are unmapping. | |
1351 | */ | |
1352 | file = vma->vm_file; | |
07a46ed2 | 1353 | size = i_size_read(file_inode(vma->vm_file)); |
897ab3e0 | 1354 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1355 | /* |
1356 | * We discovered the size of the shm segment, so | |
1357 | * break out of here and fall through to the next | |
1358 | * loop that uses the size information to stop | |
1359 | * searching for matching vma's. | |
1360 | */ | |
1361 | retval = 0; | |
1362 | vma = next; | |
1363 | break; | |
1364 | } | |
1365 | vma = next; | |
1366 | } | |
1367 | ||
1368 | /* | |
1369 | * We need look no further than the maximum address a fragment | |
1370 | * could possibly have landed at. Also cast things to loff_t to | |
25985edc | 1371 | * prevent overflows and make comparisons vs. equal-width types. |
1da177e4 | 1372 | */ |
8e36709d | 1373 | size = PAGE_ALIGN(size); |
1da177e4 LT |
1374 | while (vma && (loff_t)(vma->vm_end - addr) <= size) { |
1375 | next = vma->vm_next; | |
1376 | ||
1377 | /* finding a matching vma now does not alter retval */ | |
bc56bba8 | 1378 | if ((vma->vm_ops == &shm_vm_ops) && |
d3c97900 DH |
1379 | ((vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) && |
1380 | (vma->vm_file == file)) | |
897ab3e0 | 1381 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
1da177e4 LT |
1382 | vma = next; |
1383 | } | |
1384 | ||
63980c80 | 1385 | #else /* CONFIG_MMU */ |
8feae131 | 1386 | /* under NOMMU conditions, the exact address to be destroyed must be |
63980c80 SP |
1387 | * given |
1388 | */ | |
530fcd16 | 1389 | if (vma && vma->vm_start == addr && vma->vm_ops == &shm_vm_ops) { |
897ab3e0 | 1390 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start, NULL); |
8feae131 DH |
1391 | retval = 0; |
1392 | } | |
1393 | ||
1394 | #endif | |
1395 | ||
1da177e4 LT |
1396 | up_write(&mm->mmap_sem); |
1397 | return retval; | |
1398 | } | |
1399 | ||
1400 | #ifdef CONFIG_PROC_FS | |
19b4946c | 1401 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it) |
1da177e4 | 1402 | { |
1efdb69b | 1403 | struct user_namespace *user_ns = seq_user_ns(s); |
19b4946c | 1404 | struct shmid_kernel *shp = it; |
b7952180 HD |
1405 | unsigned long rss = 0, swp = 0; |
1406 | ||
1407 | shm_add_rss_swap(shp, &rss, &swp); | |
1da177e4 | 1408 | |
6c826818 PM |
1409 | #if BITS_PER_LONG <= 32 |
1410 | #define SIZE_SPEC "%10lu" | |
1411 | #else | |
1412 | #define SIZE_SPEC "%21lu" | |
1413 | #endif | |
1da177e4 | 1414 | |
7f032d6e JP |
1415 | seq_printf(s, |
1416 | "%10d %10d %4o " SIZE_SPEC " %5u %5u " | |
1417 | "%5lu %5u %5u %5u %5u %10lu %10lu %10lu " | |
1418 | SIZE_SPEC " " SIZE_SPEC "\n", | |
1419 | shp->shm_perm.key, | |
1420 | shp->shm_perm.id, | |
1421 | shp->shm_perm.mode, | |
1422 | shp->shm_segsz, | |
1423 | shp->shm_cprid, | |
1424 | shp->shm_lprid, | |
1425 | shp->shm_nattch, | |
1426 | from_kuid_munged(user_ns, shp->shm_perm.uid), | |
1427 | from_kgid_munged(user_ns, shp->shm_perm.gid), | |
1428 | from_kuid_munged(user_ns, shp->shm_perm.cuid), | |
1429 | from_kgid_munged(user_ns, shp->shm_perm.cgid), | |
1430 | shp->shm_atim, | |
1431 | shp->shm_dtim, | |
1432 | shp->shm_ctim, | |
1433 | rss * PAGE_SIZE, | |
1434 | swp * PAGE_SIZE); | |
1435 | ||
1436 | return 0; | |
1da177e4 LT |
1437 | } |
1438 | #endif |