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> | |
1da177e4 LT |
22 | */ |
23 | ||
1da177e4 LT |
24 | #include <linux/slab.h> |
25 | #include <linux/mm.h> | |
26 | #include <linux/hugetlb.h> | |
27 | #include <linux/shm.h> | |
28 | #include <linux/init.h> | |
29 | #include <linux/file.h> | |
30 | #include <linux/mman.h> | |
1da177e4 LT |
31 | #include <linux/shmem_fs.h> |
32 | #include <linux/security.h> | |
33 | #include <linux/syscalls.h> | |
34 | #include <linux/audit.h> | |
c59ede7b | 35 | #include <linux/capability.h> |
7d87e14c | 36 | #include <linux/ptrace.h> |
19b4946c | 37 | #include <linux/seq_file.h> |
3e148c79 | 38 | #include <linux/rwsem.h> |
4e982311 | 39 | #include <linux/nsproxy.h> |
bc56bba8 | 40 | #include <linux/mount.h> |
ae5e1b22 | 41 | #include <linux/ipc_namespace.h> |
7d87e14c | 42 | |
1da177e4 LT |
43 | #include <asm/uaccess.h> |
44 | ||
45 | #include "util.h" | |
46 | ||
bc56bba8 EB |
47 | struct shm_file_data { |
48 | int id; | |
49 | struct ipc_namespace *ns; | |
50 | struct file *file; | |
51 | const struct vm_operations_struct *vm_ops; | |
52 | }; | |
53 | ||
54 | #define shm_file_data(file) (*((struct shm_file_data **)&(file)->private_data)) | |
55 | ||
9a32144e | 56 | static const struct file_operations shm_file_operations; |
f0f37e2f | 57 | static const struct vm_operations_struct shm_vm_ops; |
1da177e4 | 58 | |
ed2ddbf8 | 59 | #define shm_ids(ns) ((ns)->ids[IPC_SHM_IDS]) |
1da177e4 | 60 | |
4e982311 KK |
61 | #define shm_unlock(shp) \ |
62 | ipc_unlock(&(shp)->shm_perm) | |
1da177e4 | 63 | |
7748dbfa | 64 | static int newseg(struct ipc_namespace *, struct ipc_params *); |
bc56bba8 EB |
65 | static void shm_open(struct vm_area_struct *vma); |
66 | static void shm_close(struct vm_area_struct *vma); | |
4e982311 | 67 | static void shm_destroy (struct ipc_namespace *ns, struct shmid_kernel *shp); |
1da177e4 | 68 | #ifdef CONFIG_PROC_FS |
19b4946c | 69 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it); |
1da177e4 LT |
70 | #endif |
71 | ||
ed2ddbf8 | 72 | void shm_init_ns(struct ipc_namespace *ns) |
4e982311 | 73 | { |
4e982311 KK |
74 | ns->shm_ctlmax = SHMMAX; |
75 | ns->shm_ctlall = SHMALL; | |
76 | ns->shm_ctlmni = SHMMNI; | |
b34a6b1d | 77 | ns->shm_rmid_forced = 0; |
4e982311 | 78 | ns->shm_tot = 0; |
e8148f75 | 79 | ipc_init_ids(&shm_ids(ns)); |
4e982311 KK |
80 | } |
81 | ||
f4566f04 | 82 | /* |
3e148c79 ND |
83 | * Called with shm_ids.rw_mutex (writer) and the shp structure locked. |
84 | * Only shm_ids.rw_mutex remains locked on exit. | |
f4566f04 | 85 | */ |
01b8b07a | 86 | static void do_shm_rmid(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp) |
4e982311 | 87 | { |
01b8b07a PP |
88 | struct shmid_kernel *shp; |
89 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
90 | ||
4e982311 KK |
91 | if (shp->shm_nattch){ |
92 | shp->shm_perm.mode |= SHM_DEST; | |
93 | /* Do not find it any more */ | |
94 | shp->shm_perm.key = IPC_PRIVATE; | |
95 | shm_unlock(shp); | |
96 | } else | |
97 | shm_destroy(ns, shp); | |
98 | } | |
99 | ||
ae5e1b22 | 100 | #ifdef CONFIG_IPC_NS |
4e982311 KK |
101 | void shm_exit_ns(struct ipc_namespace *ns) |
102 | { | |
01b8b07a | 103 | free_ipcs(ns, &shm_ids(ns), do_shm_rmid); |
7d6feeb2 | 104 | idr_destroy(&ns->ids[IPC_SHM_IDS].ipcs_idr); |
4e982311 | 105 | } |
ae5e1b22 | 106 | #endif |
1da177e4 LT |
107 | |
108 | void __init shm_init (void) | |
109 | { | |
ed2ddbf8 | 110 | shm_init_ns(&init_ipc_ns); |
19b4946c | 111 | ipc_init_proc_interface("sysvipc/shm", |
b7952180 HD |
112 | #if BITS_PER_LONG <= 32 |
113 | " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime rss swap\n", | |
114 | #else | |
115 | " key shmid perms size cpid lpid nattch uid gid cuid cgid atime dtime ctime rss swap\n", | |
116 | #endif | |
4e982311 | 117 | IPC_SHM_IDS, sysvipc_shm_proc_show); |
1da177e4 LT |
118 | } |
119 | ||
3e148c79 ND |
120 | /* |
121 | * shm_lock_(check_) routines are called in the paths where the rw_mutex | |
00c2bf85 | 122 | * is not necessarily held. |
3e148c79 | 123 | */ |
023a5355 | 124 | static inline struct shmid_kernel *shm_lock(struct ipc_namespace *ns, int id) |
1da177e4 | 125 | { |
03f02c76 ND |
126 | struct kern_ipc_perm *ipcp = ipc_lock(&shm_ids(ns), id); |
127 | ||
b1ed88b4 PP |
128 | if (IS_ERR(ipcp)) |
129 | return (struct shmid_kernel *)ipcp; | |
130 | ||
03f02c76 | 131 | return container_of(ipcp, struct shmid_kernel, shm_perm); |
023a5355 ND |
132 | } |
133 | ||
4c677e2e VK |
134 | static inline void shm_lock_by_ptr(struct shmid_kernel *ipcp) |
135 | { | |
136 | rcu_read_lock(); | |
137 | spin_lock(&ipcp->shm_perm.lock); | |
138 | } | |
139 | ||
023a5355 ND |
140 | static inline struct shmid_kernel *shm_lock_check(struct ipc_namespace *ns, |
141 | int id) | |
142 | { | |
03f02c76 ND |
143 | struct kern_ipc_perm *ipcp = ipc_lock_check(&shm_ids(ns), id); |
144 | ||
b1ed88b4 PP |
145 | if (IS_ERR(ipcp)) |
146 | return (struct shmid_kernel *)ipcp; | |
147 | ||
03f02c76 | 148 | return container_of(ipcp, struct shmid_kernel, shm_perm); |
1da177e4 LT |
149 | } |
150 | ||
7ca7e564 | 151 | static inline void shm_rmid(struct ipc_namespace *ns, struct shmid_kernel *s) |
1da177e4 | 152 | { |
7ca7e564 | 153 | ipc_rmid(&shm_ids(ns), &s->shm_perm); |
1da177e4 LT |
154 | } |
155 | ||
1da177e4 | 156 | |
bc56bba8 EB |
157 | /* This is called by fork, once for every shm attach. */ |
158 | static void shm_open(struct vm_area_struct *vma) | |
4e982311 | 159 | { |
bc56bba8 EB |
160 | struct file *file = vma->vm_file; |
161 | struct shm_file_data *sfd = shm_file_data(file); | |
1da177e4 LT |
162 | struct shmid_kernel *shp; |
163 | ||
bc56bba8 | 164 | shp = shm_lock(sfd->ns, sfd->id); |
023a5355 | 165 | BUG_ON(IS_ERR(shp)); |
1da177e4 | 166 | shp->shm_atim = get_seconds(); |
b488893a | 167 | shp->shm_lprid = task_tgid_vnr(current); |
1da177e4 LT |
168 | shp->shm_nattch++; |
169 | shm_unlock(shp); | |
170 | } | |
171 | ||
1da177e4 LT |
172 | /* |
173 | * shm_destroy - free the struct shmid_kernel | |
174 | * | |
f4566f04 | 175 | * @ns: namespace |
1da177e4 LT |
176 | * @shp: struct to free |
177 | * | |
3e148c79 | 178 | * It has to be called with shp and shm_ids.rw_mutex (writer) locked, |
1da177e4 LT |
179 | * but returns with shp unlocked and freed. |
180 | */ | |
4e982311 | 181 | static void shm_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) |
1da177e4 | 182 | { |
4e982311 | 183 | ns->shm_tot -= (shp->shm_segsz + PAGE_SIZE - 1) >> PAGE_SHIFT; |
7ca7e564 | 184 | shm_rmid(ns, shp); |
1da177e4 LT |
185 | shm_unlock(shp); |
186 | if (!is_file_hugepages(shp->shm_file)) | |
187 | shmem_lock(shp->shm_file, 0, shp->mlock_user); | |
353d5c30 | 188 | else if (shp->mlock_user) |
6d63079a | 189 | user_shm_unlock(shp->shm_file->f_path.dentry->d_inode->i_size, |
1da177e4 LT |
190 | shp->mlock_user); |
191 | fput (shp->shm_file); | |
192 | security_shm_free(shp); | |
193 | ipc_rcu_putref(shp); | |
194 | } | |
195 | ||
b34a6b1d VK |
196 | /* |
197 | * shm_may_destroy - identifies whether shm segment should be destroyed now | |
198 | * | |
199 | * Returns true if and only if there are no active users of the segment and | |
200 | * one of the following is true: | |
201 | * | |
202 | * 1) shmctl(id, IPC_RMID, NULL) was called for this shp | |
203 | * | |
204 | * 2) sysctl kernel.shm_rmid_forced is set to 1. | |
205 | */ | |
206 | static bool shm_may_destroy(struct ipc_namespace *ns, struct shmid_kernel *shp) | |
207 | { | |
208 | return (shp->shm_nattch == 0) && | |
209 | (ns->shm_rmid_forced || | |
210 | (shp->shm_perm.mode & SHM_DEST)); | |
211 | } | |
212 | ||
1da177e4 | 213 | /* |
bc56bba8 | 214 | * remove the attach descriptor vma. |
1da177e4 LT |
215 | * free memory for segment if it is marked destroyed. |
216 | * The descriptor has already been removed from the current->mm->mmap list | |
217 | * and will later be kfree()d. | |
218 | */ | |
bc56bba8 | 219 | static void shm_close(struct vm_area_struct *vma) |
1da177e4 | 220 | { |
bc56bba8 EB |
221 | struct file * file = vma->vm_file; |
222 | struct shm_file_data *sfd = shm_file_data(file); | |
1da177e4 | 223 | struct shmid_kernel *shp; |
bc56bba8 | 224 | struct ipc_namespace *ns = sfd->ns; |
4e982311 | 225 | |
3e148c79 | 226 | down_write(&shm_ids(ns).rw_mutex); |
1da177e4 | 227 | /* remove from the list of attaches of the shm segment */ |
00c2bf85 | 228 | shp = shm_lock(ns, sfd->id); |
023a5355 | 229 | BUG_ON(IS_ERR(shp)); |
b488893a | 230 | shp->shm_lprid = task_tgid_vnr(current); |
1da177e4 LT |
231 | shp->shm_dtim = get_seconds(); |
232 | shp->shm_nattch--; | |
b34a6b1d VK |
233 | if (shm_may_destroy(ns, shp)) |
234 | shm_destroy(ns, shp); | |
235 | else | |
236 | shm_unlock(shp); | |
237 | up_write(&shm_ids(ns).rw_mutex); | |
238 | } | |
239 | ||
4c677e2e | 240 | /* Called with ns->shm_ids(ns).rw_mutex locked */ |
b34a6b1d VK |
241 | static int shm_try_destroy_current(int id, void *p, void *data) |
242 | { | |
243 | struct ipc_namespace *ns = data; | |
4c677e2e VK |
244 | struct kern_ipc_perm *ipcp = p; |
245 | struct shmid_kernel *shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
b34a6b1d | 246 | |
4c677e2e | 247 | if (shp->shm_creator != current) |
5774ed01 | 248 | return 0; |
5774ed01 VK |
249 | |
250 | /* | |
251 | * Mark it as orphaned to destroy the segment when | |
252 | * kernel.shm_rmid_forced is changed. | |
253 | * It is noop if the following shm_may_destroy() returns true. | |
254 | */ | |
255 | shp->shm_creator = NULL; | |
256 | ||
257 | /* | |
258 | * Don't even try to destroy it. If shm_rmid_forced=0 and IPC_RMID | |
259 | * is not set, it shouldn't be deleted here. | |
260 | */ | |
4c677e2e | 261 | if (!ns->shm_rmid_forced) |
b34a6b1d | 262 | return 0; |
b34a6b1d | 263 | |
4c677e2e VK |
264 | if (shm_may_destroy(ns, shp)) { |
265 | shm_lock_by_ptr(shp); | |
b34a6b1d | 266 | shm_destroy(ns, shp); |
4c677e2e | 267 | } |
b34a6b1d VK |
268 | return 0; |
269 | } | |
270 | ||
4c677e2e | 271 | /* Called with ns->shm_ids(ns).rw_mutex locked */ |
b34a6b1d VK |
272 | static int shm_try_destroy_orphaned(int id, void *p, void *data) |
273 | { | |
274 | struct ipc_namespace *ns = data; | |
4c677e2e VK |
275 | struct kern_ipc_perm *ipcp = p; |
276 | struct shmid_kernel *shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
b34a6b1d VK |
277 | |
278 | /* | |
279 | * We want to destroy segments without users and with already | |
280 | * exit'ed originating process. | |
4c677e2e VK |
281 | * |
282 | * As shp->* are changed under rw_mutex, it's safe to skip shp locking. | |
b34a6b1d | 283 | */ |
4c677e2e | 284 | if (shp->shm_creator != NULL) |
b34a6b1d | 285 | return 0; |
b34a6b1d | 286 | |
4c677e2e VK |
287 | if (shm_may_destroy(ns, shp)) { |
288 | shm_lock_by_ptr(shp); | |
4e982311 | 289 | shm_destroy(ns, shp); |
4c677e2e | 290 | } |
b34a6b1d VK |
291 | return 0; |
292 | } | |
293 | ||
294 | void shm_destroy_orphaned(struct ipc_namespace *ns) | |
295 | { | |
296 | down_write(&shm_ids(ns).rw_mutex); | |
33a30ed4 | 297 | if (shm_ids(ns).in_use) |
4c677e2e | 298 | idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_orphaned, ns); |
b34a6b1d VK |
299 | up_write(&shm_ids(ns).rw_mutex); |
300 | } | |
301 | ||
302 | ||
303 | void exit_shm(struct task_struct *task) | |
304 | { | |
4c677e2e | 305 | struct ipc_namespace *ns = task->nsproxy->ipc_ns; |
b34a6b1d | 306 | |
298507d4 VK |
307 | if (shm_ids(ns).in_use == 0) |
308 | return; | |
309 | ||
b34a6b1d VK |
310 | /* Destroy all already created segments, but not mapped yet */ |
311 | down_write(&shm_ids(ns).rw_mutex); | |
33a30ed4 | 312 | if (shm_ids(ns).in_use) |
4c677e2e | 313 | idr_for_each(&shm_ids(ns).ipcs_idr, &shm_try_destroy_current, ns); |
3e148c79 | 314 | up_write(&shm_ids(ns).rw_mutex); |
1da177e4 LT |
315 | } |
316 | ||
d0217ac0 | 317 | static int shm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
bc56bba8 EB |
318 | { |
319 | struct file *file = vma->vm_file; | |
320 | struct shm_file_data *sfd = shm_file_data(file); | |
321 | ||
d0217ac0 | 322 | return sfd->vm_ops->fault(vma, vmf); |
bc56bba8 EB |
323 | } |
324 | ||
325 | #ifdef CONFIG_NUMA | |
d823e3e7 | 326 | static int shm_set_policy(struct vm_area_struct *vma, struct mempolicy *new) |
bc56bba8 EB |
327 | { |
328 | struct file *file = vma->vm_file; | |
329 | struct shm_file_data *sfd = shm_file_data(file); | |
330 | int err = 0; | |
331 | if (sfd->vm_ops->set_policy) | |
332 | err = sfd->vm_ops->set_policy(vma, new); | |
333 | return err; | |
334 | } | |
335 | ||
d823e3e7 AB |
336 | static struct mempolicy *shm_get_policy(struct vm_area_struct *vma, |
337 | unsigned long addr) | |
bc56bba8 EB |
338 | { |
339 | struct file *file = vma->vm_file; | |
340 | struct shm_file_data *sfd = shm_file_data(file); | |
341 | struct mempolicy *pol = NULL; | |
342 | ||
343 | if (sfd->vm_ops->get_policy) | |
344 | pol = sfd->vm_ops->get_policy(vma, addr); | |
52cd3b07 | 345 | else if (vma->vm_policy) |
bc56bba8 | 346 | pol = vma->vm_policy; |
52cd3b07 | 347 | |
bc56bba8 EB |
348 | return pol; |
349 | } | |
350 | #endif | |
351 | ||
1da177e4 LT |
352 | static int shm_mmap(struct file * file, struct vm_area_struct * vma) |
353 | { | |
bc56bba8 | 354 | struct shm_file_data *sfd = shm_file_data(file); |
b0e15190 DH |
355 | int ret; |
356 | ||
bc56bba8 EB |
357 | ret = sfd->file->f_op->mmap(sfd->file, vma); |
358 | if (ret != 0) | |
359 | return ret; | |
360 | sfd->vm_ops = vma->vm_ops; | |
2e92a3ba | 361 | #ifdef CONFIG_MMU |
54cb8821 | 362 | BUG_ON(!sfd->vm_ops->fault); |
2e92a3ba | 363 | #endif |
bc56bba8 EB |
364 | vma->vm_ops = &shm_vm_ops; |
365 | shm_open(vma); | |
b0e15190 DH |
366 | |
367 | return ret; | |
1da177e4 LT |
368 | } |
369 | ||
4e982311 KK |
370 | static int shm_release(struct inode *ino, struct file *file) |
371 | { | |
bc56bba8 | 372 | struct shm_file_data *sfd = shm_file_data(file); |
4e982311 | 373 | |
bc56bba8 EB |
374 | put_ipc_ns(sfd->ns); |
375 | shm_file_data(file) = NULL; | |
376 | kfree(sfd); | |
4e982311 KK |
377 | return 0; |
378 | } | |
379 | ||
02c24a82 | 380 | static int shm_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
516dffdc | 381 | { |
516dffdc | 382 | struct shm_file_data *sfd = shm_file_data(file); |
516dffdc | 383 | |
7ea80859 CH |
384 | if (!sfd->file->f_op->fsync) |
385 | return -EINVAL; | |
02c24a82 | 386 | return sfd->file->f_op->fsync(sfd->file, start, end, datasync); |
516dffdc AL |
387 | } |
388 | ||
bc56bba8 EB |
389 | static unsigned long shm_get_unmapped_area(struct file *file, |
390 | unsigned long addr, unsigned long len, unsigned long pgoff, | |
391 | unsigned long flags) | |
392 | { | |
393 | struct shm_file_data *sfd = shm_file_data(file); | |
c4caa778 AV |
394 | return sfd->file->f_op->get_unmapped_area(sfd->file, addr, len, |
395 | pgoff, flags); | |
bc56bba8 | 396 | } |
bc56bba8 | 397 | |
9a32144e | 398 | static const struct file_operations shm_file_operations = { |
4e982311 | 399 | .mmap = shm_mmap, |
516dffdc | 400 | .fsync = shm_fsync, |
4e982311 | 401 | .release = shm_release, |
ed5e5894 DH |
402 | #ifndef CONFIG_MMU |
403 | .get_unmapped_area = shm_get_unmapped_area, | |
404 | #endif | |
6038f373 | 405 | .llseek = noop_llseek, |
c4caa778 AV |
406 | }; |
407 | ||
408 | static const struct file_operations shm_file_operations_huge = { | |
409 | .mmap = shm_mmap, | |
410 | .fsync = shm_fsync, | |
411 | .release = shm_release, | |
bc56bba8 | 412 | .get_unmapped_area = shm_get_unmapped_area, |
6038f373 | 413 | .llseek = noop_llseek, |
1da177e4 LT |
414 | }; |
415 | ||
c4caa778 AV |
416 | int is_file_shm_hugepages(struct file *file) |
417 | { | |
418 | return file->f_op == &shm_file_operations_huge; | |
419 | } | |
420 | ||
f0f37e2f | 421 | static const struct vm_operations_struct shm_vm_ops = { |
1da177e4 LT |
422 | .open = shm_open, /* callback for a new vm-area open */ |
423 | .close = shm_close, /* callback for when the vm-area is released */ | |
54cb8821 | 424 | .fault = shm_fault, |
bc56bba8 EB |
425 | #if defined(CONFIG_NUMA) |
426 | .set_policy = shm_set_policy, | |
427 | .get_policy = shm_get_policy, | |
1da177e4 LT |
428 | #endif |
429 | }; | |
430 | ||
f4566f04 ND |
431 | /** |
432 | * newseg - Create a new shared memory segment | |
433 | * @ns: namespace | |
434 | * @params: ptr to the structure that contains key, size and shmflg | |
435 | * | |
3e148c79 | 436 | * Called with shm_ids.rw_mutex held as a writer. |
f4566f04 ND |
437 | */ |
438 | ||
7748dbfa | 439 | static int newseg(struct ipc_namespace *ns, struct ipc_params *params) |
1da177e4 | 440 | { |
7748dbfa ND |
441 | key_t key = params->key; |
442 | int shmflg = params->flg; | |
443 | size_t size = params->u.size; | |
1da177e4 LT |
444 | int error; |
445 | struct shmid_kernel *shp; | |
446 | int numpages = (size + PAGE_SIZE -1) >> PAGE_SHIFT; | |
447 | struct file * file; | |
448 | char name[13]; | |
449 | int id; | |
ca16d140 | 450 | vm_flags_t acctflag = 0; |
1da177e4 | 451 | |
4e982311 | 452 | if (size < SHMMIN || size > ns->shm_ctlmax) |
1da177e4 LT |
453 | return -EINVAL; |
454 | ||
f66d45e9 | 455 | if (ns->shm_tot + numpages > ns->shm_ctlall) |
1da177e4 LT |
456 | return -ENOSPC; |
457 | ||
458 | shp = ipc_rcu_alloc(sizeof(*shp)); | |
459 | if (!shp) | |
460 | return -ENOMEM; | |
461 | ||
462 | shp->shm_perm.key = key; | |
b33291c0 | 463 | shp->shm_perm.mode = (shmflg & S_IRWXUGO); |
1da177e4 LT |
464 | shp->mlock_user = NULL; |
465 | ||
466 | shp->shm_perm.security = NULL; | |
467 | error = security_shm_alloc(shp); | |
468 | if (error) { | |
469 | ipc_rcu_putref(shp); | |
470 | return error; | |
471 | } | |
472 | ||
9d66586f | 473 | sprintf (name, "SYSV%08x", key); |
1da177e4 | 474 | if (shmflg & SHM_HUGETLB) { |
5a6fe125 MG |
475 | /* hugetlb_file_setup applies strict accounting */ |
476 | if (shmflg & SHM_NORESERVE) | |
477 | acctflag = VM_NORESERVE; | |
353d5c30 | 478 | file = hugetlb_file_setup(name, size, acctflag, |
6bfde05b | 479 | &shp->mlock_user, HUGETLB_SHMFS_INODE); |
1da177e4 | 480 | } else { |
bf8f972d BP |
481 | /* |
482 | * Do not allow no accounting for OVERCOMMIT_NEVER, even | |
483 | * if it's asked for. | |
484 | */ | |
485 | if ((shmflg & SHM_NORESERVE) && | |
486 | sysctl_overcommit_memory != OVERCOMMIT_NEVER) | |
fc8744ad | 487 | acctflag = VM_NORESERVE; |
bf8f972d | 488 | file = shmem_file_setup(name, size, acctflag); |
1da177e4 LT |
489 | } |
490 | error = PTR_ERR(file); | |
491 | if (IS_ERR(file)) | |
492 | goto no_file; | |
493 | ||
48dea404 | 494 | id = ipc_addid(&shm_ids(ns), &shp->shm_perm, ns->shm_ctlmni); |
283bb7fa PP |
495 | if (id < 0) { |
496 | error = id; | |
1da177e4 | 497 | goto no_id; |
283bb7fa | 498 | } |
1da177e4 | 499 | |
b488893a | 500 | shp->shm_cprid = task_tgid_vnr(current); |
1da177e4 LT |
501 | shp->shm_lprid = 0; |
502 | shp->shm_atim = shp->shm_dtim = 0; | |
503 | shp->shm_ctim = get_seconds(); | |
504 | shp->shm_segsz = size; | |
505 | shp->shm_nattch = 0; | |
1da177e4 | 506 | shp->shm_file = file; |
5774ed01 | 507 | shp->shm_creator = current; |
30475cc1 BP |
508 | /* |
509 | * shmid gets reported as "inode#" in /proc/pid/maps. | |
510 | * proc-ps tools use this. Changing this will break them. | |
511 | */ | |
7ca7e564 | 512 | file->f_dentry->d_inode->i_ino = shp->shm_perm.id; |
551110a9 | 513 | |
4e982311 | 514 | ns->shm_tot += numpages; |
7ca7e564 | 515 | error = shp->shm_perm.id; |
1da177e4 | 516 | shm_unlock(shp); |
7ca7e564 | 517 | return error; |
1da177e4 LT |
518 | |
519 | no_id: | |
2195d281 | 520 | if (is_file_hugepages(file) && shp->mlock_user) |
353d5c30 | 521 | user_shm_unlock(size, shp->mlock_user); |
1da177e4 LT |
522 | fput(file); |
523 | no_file: | |
524 | security_shm_free(shp); | |
525 | ipc_rcu_putref(shp); | |
526 | return error; | |
527 | } | |
528 | ||
f4566f04 | 529 | /* |
3e148c79 | 530 | * Called with shm_ids.rw_mutex and ipcp locked. |
f4566f04 | 531 | */ |
03f02c76 | 532 | static inline int shm_security(struct kern_ipc_perm *ipcp, int shmflg) |
7748dbfa | 533 | { |
03f02c76 ND |
534 | struct shmid_kernel *shp; |
535 | ||
536 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
537 | return security_shm_associate(shp, shmflg); | |
7748dbfa ND |
538 | } |
539 | ||
f4566f04 | 540 | /* |
3e148c79 | 541 | * Called with shm_ids.rw_mutex and ipcp locked. |
f4566f04 | 542 | */ |
03f02c76 ND |
543 | static inline int shm_more_checks(struct kern_ipc_perm *ipcp, |
544 | struct ipc_params *params) | |
7748dbfa | 545 | { |
03f02c76 ND |
546 | struct shmid_kernel *shp; |
547 | ||
548 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); | |
549 | if (shp->shm_segsz < params->u.size) | |
7748dbfa ND |
550 | return -EINVAL; |
551 | ||
552 | return 0; | |
553 | } | |
554 | ||
d5460c99 | 555 | SYSCALL_DEFINE3(shmget, key_t, key, size_t, size, int, shmflg) |
1da177e4 | 556 | { |
4e982311 | 557 | struct ipc_namespace *ns; |
7748dbfa ND |
558 | struct ipc_ops shm_ops; |
559 | struct ipc_params shm_params; | |
4e982311 KK |
560 | |
561 | ns = current->nsproxy->ipc_ns; | |
1da177e4 | 562 | |
7748dbfa ND |
563 | shm_ops.getnew = newseg; |
564 | shm_ops.associate = shm_security; | |
565 | shm_ops.more_checks = shm_more_checks; | |
7ca7e564 | 566 | |
7748dbfa ND |
567 | shm_params.key = key; |
568 | shm_params.flg = shmflg; | |
569 | shm_params.u.size = size; | |
1da177e4 | 570 | |
7748dbfa | 571 | return ipcget(ns, &shm_ids(ns), &shm_ops, &shm_params); |
1da177e4 LT |
572 | } |
573 | ||
574 | static inline unsigned long copy_shmid_to_user(void __user *buf, struct shmid64_ds *in, int version) | |
575 | { | |
576 | switch(version) { | |
577 | case IPC_64: | |
578 | return copy_to_user(buf, in, sizeof(*in)); | |
579 | case IPC_OLD: | |
580 | { | |
581 | struct shmid_ds out; | |
582 | ||
3af54c9b | 583 | memset(&out, 0, sizeof(out)); |
1da177e4 LT |
584 | ipc64_perm_to_ipc_perm(&in->shm_perm, &out.shm_perm); |
585 | out.shm_segsz = in->shm_segsz; | |
586 | out.shm_atime = in->shm_atime; | |
587 | out.shm_dtime = in->shm_dtime; | |
588 | out.shm_ctime = in->shm_ctime; | |
589 | out.shm_cpid = in->shm_cpid; | |
590 | out.shm_lpid = in->shm_lpid; | |
591 | out.shm_nattch = in->shm_nattch; | |
592 | ||
593 | return copy_to_user(buf, &out, sizeof(out)); | |
594 | } | |
595 | default: | |
596 | return -EINVAL; | |
597 | } | |
598 | } | |
599 | ||
016d7132 PP |
600 | static inline unsigned long |
601 | copy_shmid_from_user(struct shmid64_ds *out, void __user *buf, int version) | |
1da177e4 LT |
602 | { |
603 | switch(version) { | |
604 | case IPC_64: | |
016d7132 | 605 | if (copy_from_user(out, buf, sizeof(*out))) |
1da177e4 | 606 | return -EFAULT; |
1da177e4 | 607 | return 0; |
1da177e4 LT |
608 | case IPC_OLD: |
609 | { | |
610 | struct shmid_ds tbuf_old; | |
611 | ||
612 | if (copy_from_user(&tbuf_old, buf, sizeof(tbuf_old))) | |
613 | return -EFAULT; | |
614 | ||
016d7132 PP |
615 | out->shm_perm.uid = tbuf_old.shm_perm.uid; |
616 | out->shm_perm.gid = tbuf_old.shm_perm.gid; | |
617 | out->shm_perm.mode = tbuf_old.shm_perm.mode; | |
1da177e4 LT |
618 | |
619 | return 0; | |
620 | } | |
621 | default: | |
622 | return -EINVAL; | |
623 | } | |
624 | } | |
625 | ||
626 | static inline unsigned long copy_shminfo_to_user(void __user *buf, struct shminfo64 *in, int version) | |
627 | { | |
628 | switch(version) { | |
629 | case IPC_64: | |
630 | return copy_to_user(buf, in, sizeof(*in)); | |
631 | case IPC_OLD: | |
632 | { | |
633 | struct shminfo out; | |
634 | ||
635 | if(in->shmmax > INT_MAX) | |
636 | out.shmmax = INT_MAX; | |
637 | else | |
638 | out.shmmax = (int)in->shmmax; | |
639 | ||
640 | out.shmmin = in->shmmin; | |
641 | out.shmmni = in->shmmni; | |
642 | out.shmseg = in->shmseg; | |
643 | out.shmall = in->shmall; | |
644 | ||
645 | return copy_to_user(buf, &out, sizeof(out)); | |
646 | } | |
647 | default: | |
648 | return -EINVAL; | |
649 | } | |
650 | } | |
651 | ||
b7952180 HD |
652 | /* |
653 | * Calculate and add used RSS and swap pages of a shm. | |
654 | * Called with shm_ids.rw_mutex held as a reader | |
655 | */ | |
656 | static void shm_add_rss_swap(struct shmid_kernel *shp, | |
657 | unsigned long *rss_add, unsigned long *swp_add) | |
658 | { | |
659 | struct inode *inode; | |
660 | ||
661 | inode = shp->shm_file->f_path.dentry->d_inode; | |
662 | ||
663 | if (is_file_hugepages(shp->shm_file)) { | |
664 | struct address_space *mapping = inode->i_mapping; | |
665 | struct hstate *h = hstate_file(shp->shm_file); | |
666 | *rss_add += pages_per_huge_page(h) * mapping->nrpages; | |
667 | } else { | |
668 | #ifdef CONFIG_SHMEM | |
669 | struct shmem_inode_info *info = SHMEM_I(inode); | |
670 | spin_lock(&info->lock); | |
671 | *rss_add += inode->i_mapping->nrpages; | |
672 | *swp_add += info->swapped; | |
673 | spin_unlock(&info->lock); | |
674 | #else | |
675 | *rss_add += inode->i_mapping->nrpages; | |
676 | #endif | |
677 | } | |
678 | } | |
679 | ||
f4566f04 | 680 | /* |
3e148c79 | 681 | * Called with shm_ids.rw_mutex held as a reader |
f4566f04 | 682 | */ |
4e982311 KK |
683 | static void shm_get_stat(struct ipc_namespace *ns, unsigned long *rss, |
684 | unsigned long *swp) | |
1da177e4 | 685 | { |
7ca7e564 ND |
686 | int next_id; |
687 | int total, in_use; | |
1da177e4 LT |
688 | |
689 | *rss = 0; | |
690 | *swp = 0; | |
691 | ||
7ca7e564 ND |
692 | in_use = shm_ids(ns).in_use; |
693 | ||
694 | for (total = 0, next_id = 0; total < in_use; next_id++) { | |
e562aebc | 695 | struct kern_ipc_perm *ipc; |
1da177e4 | 696 | struct shmid_kernel *shp; |
1da177e4 | 697 | |
e562aebc TB |
698 | ipc = idr_find(&shm_ids(ns).ipcs_idr, next_id); |
699 | if (ipc == NULL) | |
1da177e4 | 700 | continue; |
e562aebc | 701 | shp = container_of(ipc, struct shmid_kernel, shm_perm); |
1da177e4 | 702 | |
b7952180 | 703 | shm_add_rss_swap(shp, rss, swp); |
7ca7e564 ND |
704 | |
705 | total++; | |
1da177e4 LT |
706 | } |
707 | } | |
708 | ||
8d4cc8b5 PP |
709 | /* |
710 | * This function handles some shmctl commands which require the rw_mutex | |
711 | * to be held in write mode. | |
712 | * NOTE: no locks must be held, the rw_mutex is taken inside this function. | |
713 | */ | |
714 | static int shmctl_down(struct ipc_namespace *ns, int shmid, int cmd, | |
715 | struct shmid_ds __user *buf, int version) | |
1da177e4 | 716 | { |
8d4cc8b5 | 717 | struct kern_ipc_perm *ipcp; |
016d7132 | 718 | struct shmid64_ds shmid64; |
8d4cc8b5 PP |
719 | struct shmid_kernel *shp; |
720 | int err; | |
721 | ||
722 | if (cmd == IPC_SET) { | |
016d7132 | 723 | if (copy_shmid_from_user(&shmid64, buf, version)) |
8d4cc8b5 PP |
724 | return -EFAULT; |
725 | } | |
726 | ||
b0e77598 SH |
727 | ipcp = ipcctl_pre_down(ns, &shm_ids(ns), shmid, cmd, |
728 | &shmid64.shm_perm, 0); | |
a5f75e7f PP |
729 | if (IS_ERR(ipcp)) |
730 | return PTR_ERR(ipcp); | |
8d4cc8b5 | 731 | |
a5f75e7f | 732 | shp = container_of(ipcp, struct shmid_kernel, shm_perm); |
8d4cc8b5 PP |
733 | |
734 | err = security_shm_shmctl(shp, cmd); | |
735 | if (err) | |
736 | goto out_unlock; | |
737 | switch (cmd) { | |
738 | case IPC_RMID: | |
739 | do_shm_rmid(ns, ipcp); | |
740 | goto out_up; | |
741 | case IPC_SET: | |
8f4a3809 | 742 | ipc_update_perm(&shmid64.shm_perm, ipcp); |
8d4cc8b5 PP |
743 | shp->shm_ctim = get_seconds(); |
744 | break; | |
745 | default: | |
746 | err = -EINVAL; | |
747 | } | |
748 | out_unlock: | |
749 | shm_unlock(shp); | |
750 | out_up: | |
751 | up_write(&shm_ids(ns).rw_mutex); | |
752 | return err; | |
753 | } | |
754 | ||
d5460c99 | 755 | SYSCALL_DEFINE3(shmctl, int, shmid, int, cmd, struct shmid_ds __user *, buf) |
8d4cc8b5 | 756 | { |
1da177e4 LT |
757 | struct shmid_kernel *shp; |
758 | int err, version; | |
4e982311 | 759 | struct ipc_namespace *ns; |
1da177e4 LT |
760 | |
761 | if (cmd < 0 || shmid < 0) { | |
762 | err = -EINVAL; | |
763 | goto out; | |
764 | } | |
765 | ||
766 | version = ipc_parse_version(&cmd); | |
4e982311 | 767 | ns = current->nsproxy->ipc_ns; |
1da177e4 LT |
768 | |
769 | switch (cmd) { /* replace with proc interface ? */ | |
770 | case IPC_INFO: | |
771 | { | |
772 | struct shminfo64 shminfo; | |
773 | ||
774 | err = security_shm_shmctl(NULL, cmd); | |
775 | if (err) | |
776 | return err; | |
777 | ||
e8148f75 | 778 | memset(&shminfo, 0, sizeof(shminfo)); |
4e982311 KK |
779 | shminfo.shmmni = shminfo.shmseg = ns->shm_ctlmni; |
780 | shminfo.shmmax = ns->shm_ctlmax; | |
781 | shminfo.shmall = ns->shm_ctlall; | |
1da177e4 LT |
782 | |
783 | shminfo.shmmin = SHMMIN; | |
784 | if(copy_shminfo_to_user (buf, &shminfo, version)) | |
785 | return -EFAULT; | |
f4566f04 | 786 | |
3e148c79 | 787 | down_read(&shm_ids(ns).rw_mutex); |
7ca7e564 | 788 | err = ipc_get_maxid(&shm_ids(ns)); |
3e148c79 | 789 | up_read(&shm_ids(ns).rw_mutex); |
f4566f04 | 790 | |
1da177e4 LT |
791 | if(err<0) |
792 | err = 0; | |
793 | goto out; | |
794 | } | |
795 | case SHM_INFO: | |
796 | { | |
797 | struct shm_info shm_info; | |
798 | ||
799 | err = security_shm_shmctl(NULL, cmd); | |
800 | if (err) | |
801 | return err; | |
802 | ||
e8148f75 | 803 | memset(&shm_info, 0, sizeof(shm_info)); |
3e148c79 | 804 | down_read(&shm_ids(ns).rw_mutex); |
4e982311 KK |
805 | shm_info.used_ids = shm_ids(ns).in_use; |
806 | shm_get_stat (ns, &shm_info.shm_rss, &shm_info.shm_swp); | |
807 | shm_info.shm_tot = ns->shm_tot; | |
1da177e4 LT |
808 | shm_info.swap_attempts = 0; |
809 | shm_info.swap_successes = 0; | |
7ca7e564 | 810 | err = ipc_get_maxid(&shm_ids(ns)); |
3e148c79 | 811 | up_read(&shm_ids(ns).rw_mutex); |
e8148f75 | 812 | if (copy_to_user(buf, &shm_info, sizeof(shm_info))) { |
1da177e4 LT |
813 | err = -EFAULT; |
814 | goto out; | |
815 | } | |
816 | ||
817 | err = err < 0 ? 0 : err; | |
818 | goto out; | |
819 | } | |
820 | case SHM_STAT: | |
821 | case IPC_STAT: | |
822 | { | |
823 | struct shmid64_ds tbuf; | |
824 | int result; | |
023a5355 | 825 | |
023a5355 ND |
826 | if (cmd == SHM_STAT) { |
827 | shp = shm_lock(ns, shmid); | |
828 | if (IS_ERR(shp)) { | |
829 | err = PTR_ERR(shp); | |
830 | goto out; | |
831 | } | |
7ca7e564 | 832 | result = shp->shm_perm.id; |
1da177e4 | 833 | } else { |
023a5355 ND |
834 | shp = shm_lock_check(ns, shmid); |
835 | if (IS_ERR(shp)) { | |
836 | err = PTR_ERR(shp); | |
837 | goto out; | |
838 | } | |
1da177e4 LT |
839 | result = 0; |
840 | } | |
e8148f75 | 841 | err = -EACCES; |
b0e77598 | 842 | if (ipcperms(ns, &shp->shm_perm, S_IRUGO)) |
1da177e4 LT |
843 | goto out_unlock; |
844 | err = security_shm_shmctl(shp, cmd); | |
845 | if (err) | |
846 | goto out_unlock; | |
023a5355 | 847 | memset(&tbuf, 0, sizeof(tbuf)); |
1da177e4 LT |
848 | kernel_to_ipc64_perm(&shp->shm_perm, &tbuf.shm_perm); |
849 | tbuf.shm_segsz = shp->shm_segsz; | |
850 | tbuf.shm_atime = shp->shm_atim; | |
851 | tbuf.shm_dtime = shp->shm_dtim; | |
852 | tbuf.shm_ctime = shp->shm_ctim; | |
853 | tbuf.shm_cpid = shp->shm_cprid; | |
854 | tbuf.shm_lpid = shp->shm_lprid; | |
bc56bba8 | 855 | tbuf.shm_nattch = shp->shm_nattch; |
1da177e4 LT |
856 | shm_unlock(shp); |
857 | if(copy_shmid_to_user (buf, &tbuf, version)) | |
858 | err = -EFAULT; | |
859 | else | |
860 | err = result; | |
861 | goto out; | |
862 | } | |
863 | case SHM_LOCK: | |
864 | case SHM_UNLOCK: | |
865 | { | |
89e004ea LS |
866 | struct file *uninitialized_var(shm_file); |
867 | ||
868 | lru_add_drain_all(); /* drain pagevecs to lru lists */ | |
869 | ||
023a5355 ND |
870 | shp = shm_lock_check(ns, shmid); |
871 | if (IS_ERR(shp)) { | |
872 | err = PTR_ERR(shp); | |
1da177e4 LT |
873 | goto out; |
874 | } | |
1da177e4 | 875 | |
a33e6751 | 876 | audit_ipc_obj(&(shp->shm_perm)); |
073115d6 | 877 | |
b0e77598 | 878 | if (!ns_capable(ns->user_ns, CAP_IPC_LOCK)) { |
414c0708 | 879 | uid_t euid = current_euid(); |
1da177e4 | 880 | err = -EPERM; |
414c0708 DH |
881 | if (euid != shp->shm_perm.uid && |
882 | euid != shp->shm_perm.cuid) | |
1da177e4 | 883 | goto out_unlock; |
f1eb1332 | 884 | if (cmd == SHM_LOCK && !rlimit(RLIMIT_MEMLOCK)) |
1da177e4 LT |
885 | goto out_unlock; |
886 | } | |
887 | ||
888 | err = security_shm_shmctl(shp, cmd); | |
889 | if (err) | |
890 | goto out_unlock; | |
891 | ||
892 | if(cmd==SHM_LOCK) { | |
86a264ab | 893 | struct user_struct *user = current_user(); |
1da177e4 LT |
894 | if (!is_file_hugepages(shp->shm_file)) { |
895 | err = shmem_lock(shp->shm_file, 1, user); | |
7be77e20 | 896 | if (!err && !(shp->shm_perm.mode & SHM_LOCKED)){ |
b33291c0 | 897 | shp->shm_perm.mode |= SHM_LOCKED; |
1da177e4 LT |
898 | shp->mlock_user = user; |
899 | } | |
900 | } | |
901 | } else if (!is_file_hugepages(shp->shm_file)) { | |
902 | shmem_lock(shp->shm_file, 0, shp->mlock_user); | |
b33291c0 | 903 | shp->shm_perm.mode &= ~SHM_LOCKED; |
1da177e4 LT |
904 | shp->mlock_user = NULL; |
905 | } | |
906 | shm_unlock(shp); | |
907 | goto out; | |
908 | } | |
909 | case IPC_RMID: | |
1da177e4 | 910 | case IPC_SET: |
8d4cc8b5 PP |
911 | err = shmctl_down(ns, shmid, cmd, buf, version); |
912 | return err; | |
1da177e4 | 913 | default: |
8d4cc8b5 | 914 | return -EINVAL; |
1da177e4 LT |
915 | } |
916 | ||
1da177e4 LT |
917 | out_unlock: |
918 | shm_unlock(shp); | |
919 | out: | |
920 | return err; | |
921 | } | |
922 | ||
923 | /* | |
924 | * Fix shmaddr, allocate descriptor, map shm, add attach descriptor to lists. | |
925 | * | |
926 | * NOTE! Despite the name, this is NOT a direct system call entrypoint. The | |
927 | * "raddr" thing points to kernel space, and there has to be a wrapper around | |
928 | * this. | |
929 | */ | |
930 | long do_shmat(int shmid, char __user *shmaddr, int shmflg, ulong *raddr) | |
931 | { | |
932 | struct shmid_kernel *shp; | |
933 | unsigned long addr; | |
934 | unsigned long size; | |
935 | struct file * file; | |
936 | int err; | |
937 | unsigned long flags; | |
938 | unsigned long prot; | |
1da177e4 | 939 | int acc_mode; |
bc56bba8 | 940 | unsigned long user_addr; |
4e982311 | 941 | struct ipc_namespace *ns; |
bc56bba8 EB |
942 | struct shm_file_data *sfd; |
943 | struct path path; | |
aeb5d727 | 944 | fmode_t f_mode; |
1da177e4 | 945 | |
bc56bba8 EB |
946 | err = -EINVAL; |
947 | if (shmid < 0) | |
1da177e4 | 948 | goto out; |
bc56bba8 | 949 | else if ((addr = (ulong)shmaddr)) { |
1da177e4 LT |
950 | if (addr & (SHMLBA-1)) { |
951 | if (shmflg & SHM_RND) | |
952 | addr &= ~(SHMLBA-1); /* round down */ | |
953 | else | |
954 | #ifndef __ARCH_FORCE_SHMLBA | |
955 | if (addr & ~PAGE_MASK) | |
956 | #endif | |
bc56bba8 | 957 | goto out; |
1da177e4 LT |
958 | } |
959 | flags = MAP_SHARED | MAP_FIXED; | |
960 | } else { | |
961 | if ((shmflg & SHM_REMAP)) | |
bc56bba8 | 962 | goto out; |
1da177e4 LT |
963 | |
964 | flags = MAP_SHARED; | |
965 | } | |
966 | ||
967 | if (shmflg & SHM_RDONLY) { | |
968 | prot = PROT_READ; | |
1da177e4 | 969 | acc_mode = S_IRUGO; |
bc56bba8 | 970 | f_mode = FMODE_READ; |
1da177e4 LT |
971 | } else { |
972 | prot = PROT_READ | PROT_WRITE; | |
1da177e4 | 973 | acc_mode = S_IRUGO | S_IWUGO; |
bc56bba8 | 974 | f_mode = FMODE_READ | FMODE_WRITE; |
1da177e4 LT |
975 | } |
976 | if (shmflg & SHM_EXEC) { | |
977 | prot |= PROT_EXEC; | |
978 | acc_mode |= S_IXUGO; | |
979 | } | |
980 | ||
981 | /* | |
982 | * We cannot rely on the fs check since SYSV IPC does have an | |
983 | * additional creator id... | |
984 | */ | |
4e982311 | 985 | ns = current->nsproxy->ipc_ns; |
023a5355 ND |
986 | shp = shm_lock_check(ns, shmid); |
987 | if (IS_ERR(shp)) { | |
988 | err = PTR_ERR(shp); | |
1da177e4 | 989 | goto out; |
023a5355 | 990 | } |
bc56bba8 EB |
991 | |
992 | err = -EACCES; | |
b0e77598 | 993 | if (ipcperms(ns, &shp->shm_perm, acc_mode)) |
bc56bba8 | 994 | goto out_unlock; |
1da177e4 LT |
995 | |
996 | err = security_shm_shmat(shp, shmaddr, shmflg); | |
bc56bba8 EB |
997 | if (err) |
998 | goto out_unlock; | |
999 | ||
2c48b9c4 AV |
1000 | path = shp->shm_file->f_path; |
1001 | path_get(&path); | |
1da177e4 | 1002 | shp->shm_nattch++; |
bc56bba8 | 1003 | size = i_size_read(path.dentry->d_inode); |
1da177e4 LT |
1004 | shm_unlock(shp); |
1005 | ||
bc56bba8 EB |
1006 | err = -ENOMEM; |
1007 | sfd = kzalloc(sizeof(*sfd), GFP_KERNEL); | |
1008 | if (!sfd) | |
ce8d2cdf | 1009 | goto out_put_dentry; |
bc56bba8 | 1010 | |
2c48b9c4 AV |
1011 | file = alloc_file(&path, f_mode, |
1012 | is_file_hugepages(shp->shm_file) ? | |
c4caa778 AV |
1013 | &shm_file_operations_huge : |
1014 | &shm_file_operations); | |
bc56bba8 EB |
1015 | if (!file) |
1016 | goto out_free; | |
1017 | ||
bc56bba8 | 1018 | file->private_data = sfd; |
bc56bba8 | 1019 | file->f_mapping = shp->shm_file->f_mapping; |
7ca7e564 | 1020 | sfd->id = shp->shm_perm.id; |
bc56bba8 EB |
1021 | sfd->ns = get_ipc_ns(ns); |
1022 | sfd->file = shp->shm_file; | |
1023 | sfd->vm_ops = NULL; | |
1024 | ||
1da177e4 LT |
1025 | down_write(¤t->mm->mmap_sem); |
1026 | if (addr && !(shmflg & SHM_REMAP)) { | |
bc56bba8 | 1027 | err = -EINVAL; |
1da177e4 LT |
1028 | if (find_vma_intersection(current->mm, addr, addr + size)) |
1029 | goto invalid; | |
1030 | /* | |
1031 | * If shm segment goes below stack, make sure there is some | |
1032 | * space left for the stack to grow (at least 4 pages). | |
1033 | */ | |
1034 | if (addr < current->mm->start_stack && | |
1035 | addr > current->mm->start_stack - size - PAGE_SIZE * 5) | |
1036 | goto invalid; | |
1037 | } | |
1038 | ||
bc56bba8 EB |
1039 | user_addr = do_mmap (file, addr, size, prot, flags, 0); |
1040 | *raddr = user_addr; | |
1041 | err = 0; | |
1042 | if (IS_ERR_VALUE(user_addr)) | |
1043 | err = (long)user_addr; | |
1da177e4 LT |
1044 | invalid: |
1045 | up_write(¤t->mm->mmap_sem); | |
1046 | ||
bc56bba8 EB |
1047 | fput(file); |
1048 | ||
1049 | out_nattch: | |
3e148c79 | 1050 | down_write(&shm_ids(ns).rw_mutex); |
00c2bf85 | 1051 | shp = shm_lock(ns, shmid); |
023a5355 | 1052 | BUG_ON(IS_ERR(shp)); |
1da177e4 | 1053 | shp->shm_nattch--; |
b34a6b1d | 1054 | if (shm_may_destroy(ns, shp)) |
4e982311 | 1055 | shm_destroy(ns, shp); |
1da177e4 LT |
1056 | else |
1057 | shm_unlock(shp); | |
3e148c79 | 1058 | up_write(&shm_ids(ns).rw_mutex); |
1da177e4 | 1059 | |
1da177e4 LT |
1060 | out: |
1061 | return err; | |
bc56bba8 EB |
1062 | |
1063 | out_unlock: | |
1064 | shm_unlock(shp); | |
1065 | goto out; | |
1066 | ||
1067 | out_free: | |
1068 | kfree(sfd); | |
ce8d2cdf | 1069 | out_put_dentry: |
2c48b9c4 | 1070 | path_put(&path); |
bc56bba8 | 1071 | goto out_nattch; |
1da177e4 LT |
1072 | } |
1073 | ||
d5460c99 | 1074 | SYSCALL_DEFINE3(shmat, int, shmid, char __user *, shmaddr, int, shmflg) |
7d87e14c SR |
1075 | { |
1076 | unsigned long ret; | |
1077 | long err; | |
1078 | ||
1079 | err = do_shmat(shmid, shmaddr, shmflg, &ret); | |
1080 | if (err) | |
1081 | return err; | |
1082 | force_successful_syscall_return(); | |
1083 | return (long)ret; | |
1084 | } | |
1085 | ||
1da177e4 LT |
1086 | /* |
1087 | * detach and kill segment if marked destroyed. | |
1088 | * The work is done in shm_close. | |
1089 | */ | |
d5460c99 | 1090 | SYSCALL_DEFINE1(shmdt, char __user *, shmaddr) |
1da177e4 LT |
1091 | { |
1092 | struct mm_struct *mm = current->mm; | |
586c7e6a | 1093 | struct vm_area_struct *vma; |
1da177e4 | 1094 | unsigned long addr = (unsigned long)shmaddr; |
1da177e4 | 1095 | int retval = -EINVAL; |
586c7e6a MF |
1096 | #ifdef CONFIG_MMU |
1097 | loff_t size = 0; | |
1098 | struct vm_area_struct *next; | |
1099 | #endif | |
1da177e4 | 1100 | |
df1e2fb5 HD |
1101 | if (addr & ~PAGE_MASK) |
1102 | return retval; | |
1103 | ||
1da177e4 LT |
1104 | down_write(&mm->mmap_sem); |
1105 | ||
1106 | /* | |
1107 | * This function tries to be smart and unmap shm segments that | |
1108 | * were modified by partial mlock or munmap calls: | |
1109 | * - It first determines the size of the shm segment that should be | |
1110 | * unmapped: It searches for a vma that is backed by shm and that | |
1111 | * started at address shmaddr. It records it's size and then unmaps | |
1112 | * it. | |
1113 | * - Then it unmaps all shm vmas that started at shmaddr and that | |
1114 | * are within the initially determined size. | |
1115 | * Errors from do_munmap are ignored: the function only fails if | |
1116 | * it's called with invalid parameters or if it's called to unmap | |
1117 | * a part of a vma. Both calls in this function are for full vmas, | |
1118 | * the parameters are directly copied from the vma itself and always | |
1119 | * valid - therefore do_munmap cannot fail. (famous last words?) | |
1120 | */ | |
1121 | /* | |
1122 | * If it had been mremap()'d, the starting address would not | |
1123 | * match the usual checks anyway. So assume all vma's are | |
1124 | * above the starting address given. | |
1125 | */ | |
1126 | vma = find_vma(mm, addr); | |
1127 | ||
8feae131 | 1128 | #ifdef CONFIG_MMU |
1da177e4 LT |
1129 | while (vma) { |
1130 | next = vma->vm_next; | |
1131 | ||
1132 | /* | |
1133 | * Check if the starting address would match, i.e. it's | |
1134 | * a fragment created by mprotect() and/or munmap(), or it | |
1135 | * otherwise it starts at this address with no hassles. | |
1136 | */ | |
bc56bba8 | 1137 | if ((vma->vm_ops == &shm_vm_ops) && |
1da177e4 LT |
1138 | (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) { |
1139 | ||
1140 | ||
6d63079a | 1141 | size = vma->vm_file->f_path.dentry->d_inode->i_size; |
1da177e4 LT |
1142 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start); |
1143 | /* | |
1144 | * We discovered the size of the shm segment, so | |
1145 | * break out of here and fall through to the next | |
1146 | * loop that uses the size information to stop | |
1147 | * searching for matching vma's. | |
1148 | */ | |
1149 | retval = 0; | |
1150 | vma = next; | |
1151 | break; | |
1152 | } | |
1153 | vma = next; | |
1154 | } | |
1155 | ||
1156 | /* | |
1157 | * We need look no further than the maximum address a fragment | |
1158 | * could possibly have landed at. Also cast things to loff_t to | |
25985edc | 1159 | * prevent overflows and make comparisons vs. equal-width types. |
1da177e4 | 1160 | */ |
8e36709d | 1161 | size = PAGE_ALIGN(size); |
1da177e4 LT |
1162 | while (vma && (loff_t)(vma->vm_end - addr) <= size) { |
1163 | next = vma->vm_next; | |
1164 | ||
1165 | /* finding a matching vma now does not alter retval */ | |
bc56bba8 | 1166 | if ((vma->vm_ops == &shm_vm_ops) && |
1da177e4 LT |
1167 | (vma->vm_start - addr)/PAGE_SIZE == vma->vm_pgoff) |
1168 | ||
1169 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start); | |
1170 | vma = next; | |
1171 | } | |
1172 | ||
8feae131 DH |
1173 | #else /* CONFIG_MMU */ |
1174 | /* under NOMMU conditions, the exact address to be destroyed must be | |
1175 | * given */ | |
1176 | retval = -EINVAL; | |
1177 | if (vma->vm_start == addr && vma->vm_ops == &shm_vm_ops) { | |
1178 | do_munmap(mm, vma->vm_start, vma->vm_end - vma->vm_start); | |
1179 | retval = 0; | |
1180 | } | |
1181 | ||
1182 | #endif | |
1183 | ||
1da177e4 LT |
1184 | up_write(&mm->mmap_sem); |
1185 | return retval; | |
1186 | } | |
1187 | ||
1188 | #ifdef CONFIG_PROC_FS | |
19b4946c | 1189 | static int sysvipc_shm_proc_show(struct seq_file *s, void *it) |
1da177e4 | 1190 | { |
19b4946c | 1191 | struct shmid_kernel *shp = it; |
b7952180 HD |
1192 | unsigned long rss = 0, swp = 0; |
1193 | ||
1194 | shm_add_rss_swap(shp, &rss, &swp); | |
1da177e4 | 1195 | |
6c826818 PM |
1196 | #if BITS_PER_LONG <= 32 |
1197 | #define SIZE_SPEC "%10lu" | |
1198 | #else | |
1199 | #define SIZE_SPEC "%21lu" | |
1200 | #endif | |
1da177e4 | 1201 | |
6c826818 PM |
1202 | return seq_printf(s, |
1203 | "%10d %10d %4o " SIZE_SPEC " %5u %5u " | |
b7952180 HD |
1204 | "%5lu %5u %5u %5u %5u %10lu %10lu %10lu " |
1205 | SIZE_SPEC " " SIZE_SPEC "\n", | |
19b4946c | 1206 | shp->shm_perm.key, |
7ca7e564 | 1207 | shp->shm_perm.id, |
b33291c0 | 1208 | shp->shm_perm.mode, |
19b4946c MW |
1209 | shp->shm_segsz, |
1210 | shp->shm_cprid, | |
1211 | shp->shm_lprid, | |
bc56bba8 | 1212 | shp->shm_nattch, |
19b4946c MW |
1213 | shp->shm_perm.uid, |
1214 | shp->shm_perm.gid, | |
1215 | shp->shm_perm.cuid, | |
1216 | shp->shm_perm.cgid, | |
1217 | shp->shm_atim, | |
1218 | shp->shm_dtim, | |
b7952180 HD |
1219 | shp->shm_ctim, |
1220 | rss * PAGE_SIZE, | |
1221 | swp * PAGE_SIZE); | |
1da177e4 LT |
1222 | } |
1223 | #endif |