1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20#include <linux/mm.h>
21#include <linux/shm.h>
22#include <linux/init.h>
23#include <linux/msg.h>
24#include <linux/vmalloc.h>
25#include <linux/slab.h>
26#include <linux/capability.h>
27#include <linux/highuid.h>
28#include <linux/security.h>
29#include <linux/rcupdate.h>
30#include <linux/workqueue.h>
31#include <linux/seq_file.h>
32#include <linux/proc_fs.h>
33#include <linux/audit.h>
34#include <linux/nsproxy.h>
35#include <linux/rwsem.h>
36#include <linux/memory.h>
37#include <linux/ipc_namespace.h>
38
39#include <asm/unistd.h>
40
41#include "util.h"
42
43struct ipc_proc_iface {
44 const char *path;
45 const char *header;
46 int ids;
47 int (*show)(struct seq_file *, void *);
48};
49
50#ifdef CONFIG_MEMORY_HOTPLUG
51
52static void ipc_memory_notifier(struct work_struct *work)
53{
54 ipcns_notify(IPCNS_MEMCHANGED);
55}
56
57static DECLARE_WORK(ipc_memory_wq, ipc_memory_notifier);
58
59
60static int ipc_memory_callback(struct notifier_block *self,
61 unsigned long action, void *arg)
62{
63 switch (action) {
64 case MEM_ONLINE:
65 case MEM_OFFLINE:
66
67
68
69
70
71
72
73
74 if (!work_pending(&ipc_memory_wq))
75 schedule_work(&ipc_memory_wq);
76 break;
77 case MEM_GOING_ONLINE:
78 case MEM_GOING_OFFLINE:
79 case MEM_CANCEL_ONLINE:
80 case MEM_CANCEL_OFFLINE:
81 default:
82 break;
83 }
84
85 return NOTIFY_OK;
86}
87
88#endif
89
90
91
92
93
94
95
96
97
98
99
100static int __init ipc_init(void)
101{
102 sem_init();
103 msg_init();
104 shm_init();
105 hotplug_memory_notifier(ipc_memory_callback, IPC_CALLBACK_PRI);
106 register_ipcns_notifier(&init_ipc_ns);
107 return 0;
108}
109__initcall(ipc_init);
110
111
112
113
114
115
116
117
118
119void ipc_init_ids(struct ipc_ids *ids)
120{
121 init_rwsem(&ids->rw_mutex);
122
123 ids->in_use = 0;
124 ids->seq = 0;
125 ids->next_id = -1;
126 {
127 int seq_limit = INT_MAX/SEQ_MULTIPLIER;
128 if (seq_limit > USHRT_MAX)
129 ids->seq_max = USHRT_MAX;
130 else
131 ids->seq_max = seq_limit;
132 }
133
134 idr_init(&ids->ipcs_idr);
135}
136
137#ifdef CONFIG_PROC_FS
138static const struct file_operations sysvipc_proc_fops;
139
140
141
142
143
144
145
146void __init ipc_init_proc_interface(const char *path, const char *header,
147 int ids, int (*show)(struct seq_file *, void *))
148{
149 struct proc_dir_entry *pde;
150 struct ipc_proc_iface *iface;
151
152 iface = kmalloc(sizeof(*iface), GFP_KERNEL);
153 if (!iface)
154 return;
155 iface->path = path;
156 iface->header = header;
157 iface->ids = ids;
158 iface->show = show;
159
160 pde = proc_create_data(path,
161 S_IRUGO,
162 NULL,
163 &sysvipc_proc_fops,
164 iface);
165 if (!pde) {
166 kfree(iface);
167 }
168}
169#endif
170
171
172
173
174
175
176
177
178
179
180
181
182static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
183{
184 struct kern_ipc_perm *ipc;
185 int next_id;
186 int total;
187
188 for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
189 ipc = idr_find(&ids->ipcs_idr, next_id);
190
191 if (ipc == NULL)
192 continue;
193
194 if (ipc->key != key) {
195 total++;
196 continue;
197 }
198
199 ipc_lock_by_ptr(ipc);
200 return ipc;
201 }
202
203 return NULL;
204}
205
206
207
208
209
210
211
212
213int ipc_get_maxid(struct ipc_ids *ids)
214{
215 struct kern_ipc_perm *ipc;
216 int max_id = -1;
217 int total, id;
218
219 if (ids->in_use == 0)
220 return -1;
221
222 if (ids->in_use == IPCMNI)
223 return IPCMNI - 1;
224
225
226 total = 0;
227 for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
228 ipc = idr_find(&ids->ipcs_idr, id);
229 if (ipc != NULL) {
230 max_id = id;
231 total++;
232 }
233 }
234 return max_id;
235}
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)
252{
253 kuid_t euid;
254 kgid_t egid;
255 int id, err;
256 int next_id = ids->next_id;
257
258 if (size > IPCMNI)
259 size = IPCMNI;
260
261 if (ids->in_use >= size)
262 return -ENOSPC;
263
264 spin_lock_init(&new->lock);
265 new->deleted = 0;
266 rcu_read_lock();
267 spin_lock(&new->lock);
268
269 err = idr_get_new_above(&ids->ipcs_idr, new,
270 (next_id < 0) ? 0 : ipcid_to_idx(next_id), &id);
271 if (err) {
272 spin_unlock(&new->lock);
273 rcu_read_unlock();
274 return err;
275 }
276
277 ids->in_use++;
278
279 current_euid_egid(&euid, &egid);
280 new->cuid = new->uid = euid;
281 new->gid = new->cgid = egid;
282
283 if (next_id < 0) {
284 new->seq = ids->seq++;
285 if (ids->seq > ids->seq_max)
286 ids->seq = 0;
287 } else {
288 new->seq = ipcid_to_seqx(next_id);
289 ids->next_id = -1;
290 }
291
292 new->id = ipc_buildid(id, new->seq);
293 return id;
294}
295
296
297
298
299
300
301
302
303
304
305
306static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
307 struct ipc_ops *ops, struct ipc_params *params)
308{
309 int err;
310retry:
311 err = idr_pre_get(&ids->ipcs_idr, GFP_KERNEL);
312
313 if (!err)
314 return -ENOMEM;
315
316 down_write(&ids->rw_mutex);
317 err = ops->getnew(ns, params);
318 up_write(&ids->rw_mutex);
319
320 if (err == -EAGAIN)
321 goto retry;
322
323 return err;
324}
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341static int ipc_check_perms(struct ipc_namespace *ns,
342 struct kern_ipc_perm *ipcp,
343 struct ipc_ops *ops,
344 struct ipc_params *params)
345{
346 int err;
347
348 if (ipcperms(ns, ipcp, params->flg))
349 err = -EACCES;
350 else {
351 err = ops->associate(ipcp, params->flg);
352 if (!err)
353 err = ipcp->id;
354 }
355
356 return err;
357}
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
374 struct ipc_ops *ops, struct ipc_params *params)
375{
376 struct kern_ipc_perm *ipcp;
377 int flg = params->flg;
378 int err;
379retry:
380 err = idr_pre_get(&ids->ipcs_idr, GFP_KERNEL);
381
382
383
384
385
386 down_write(&ids->rw_mutex);
387 ipcp = ipc_findkey(ids, params->key);
388 if (ipcp == NULL) {
389
390 if (!(flg & IPC_CREAT))
391 err = -ENOENT;
392 else if (!err)
393 err = -ENOMEM;
394 else
395 err = ops->getnew(ns, params);
396 } else {
397
398
399 if (flg & IPC_CREAT && flg & IPC_EXCL)
400 err = -EEXIST;
401 else {
402 err = 0;
403 if (ops->more_checks)
404 err = ops->more_checks(ipcp, params);
405 if (!err)
406
407
408
409
410 err = ipc_check_perms(ns, ipcp, ops, params);
411 }
412 ipc_unlock(ipcp);
413 }
414 up_write(&ids->rw_mutex);
415
416 if (err == -EAGAIN)
417 goto retry;
418
419 return err;
420}
421
422
423
424
425
426
427
428
429
430
431
432void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
433{
434 int lid = ipcid_to_idx(ipcp->id);
435
436 idr_remove(&ids->ipcs_idr, lid);
437
438 ids->in_use--;
439
440 ipcp->deleted = 1;
441
442 return;
443}
444
445
446
447
448
449
450
451
452
453void* ipc_alloc(int size)
454{
455 void* out;
456 if(size > PAGE_SIZE)
457 out = vmalloc(size);
458 else
459 out = kmalloc(size, GFP_KERNEL);
460 return out;
461}
462
463
464
465
466
467
468
469
470
471
472void ipc_free(void* ptr, int size)
473{
474 if(size > PAGE_SIZE)
475 vfree(ptr);
476 else
477 kfree(ptr);
478}
479
480
481
482
483
484
485
486
487
488
489
490struct ipc_rcu_hdr
491{
492 int refcount;
493 int is_vmalloc;
494 void *data[0];
495};
496
497
498struct ipc_rcu_grace
499{
500 struct rcu_head rcu;
501
502 void *data[0];
503};
504
505struct ipc_rcu_sched
506{
507 struct work_struct work;
508
509 void *data[0];
510};
511
512#define HDRLEN_KMALLOC (sizeof(struct ipc_rcu_grace) > sizeof(struct ipc_rcu_hdr) ? \
513 sizeof(struct ipc_rcu_grace) : sizeof(struct ipc_rcu_hdr))
514#define HDRLEN_VMALLOC (sizeof(struct ipc_rcu_sched) > HDRLEN_KMALLOC ? \
515 sizeof(struct ipc_rcu_sched) : HDRLEN_KMALLOC)
516
517static inline int rcu_use_vmalloc(int size)
518{
519
520 if (HDRLEN_KMALLOC + size > PAGE_SIZE)
521 return 1;
522 return 0;
523}
524
525
526
527
528
529
530
531
532
533
534void* ipc_rcu_alloc(int size)
535{
536 void* out;
537
538
539
540
541 if (rcu_use_vmalloc(size)) {
542 out = vmalloc(HDRLEN_VMALLOC + size);
543 if (out) {
544 out += HDRLEN_VMALLOC;
545 container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 1;
546 container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
547 }
548 } else {
549 out = kmalloc(HDRLEN_KMALLOC + size, GFP_KERNEL);
550 if (out) {
551 out += HDRLEN_KMALLOC;
552 container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 0;
553 container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
554 }
555 }
556
557 return out;
558}
559
560void ipc_rcu_getref(void *ptr)
561{
562 container_of(ptr, struct ipc_rcu_hdr, data)->refcount++;
563}
564
565static void ipc_do_vfree(struct work_struct *work)
566{
567 vfree(container_of(work, struct ipc_rcu_sched, work));
568}
569
570
571
572
573
574
575
576
577static void ipc_schedule_free(struct rcu_head *head)
578{
579 struct ipc_rcu_grace *grace;
580 struct ipc_rcu_sched *sched;
581
582 grace = container_of(head, struct ipc_rcu_grace, rcu);
583 sched = container_of(&(grace->data[0]), struct ipc_rcu_sched,
584 data[0]);
585
586 INIT_WORK(&sched->work, ipc_do_vfree);
587 schedule_work(&sched->work);
588}
589
590void ipc_rcu_putref(void *ptr)
591{
592 if (--container_of(ptr, struct ipc_rcu_hdr, data)->refcount > 0)
593 return;
594
595 if (container_of(ptr, struct ipc_rcu_hdr, data)->is_vmalloc) {
596 call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
597 ipc_schedule_free);
598 } else {
599 kfree_rcu(container_of(ptr, struct ipc_rcu_grace, data), rcu);
600 }
601}
602
603
604
605
606
607
608
609
610
611
612
613
614
615int ipcperms(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp, short flag)
616{
617 kuid_t euid = current_euid();
618 int requested_mode, granted_mode;
619
620 audit_ipc_obj(ipcp);
621 requested_mode = (flag >> 6) | (flag >> 3) | flag;
622 granted_mode = ipcp->mode;
623 if (uid_eq(euid, ipcp->cuid) ||
624 uid_eq(euid, ipcp->uid))
625 granted_mode >>= 6;
626 else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
627 granted_mode >>= 3;
628
629 if ((requested_mode & ~granted_mode & 0007) &&
630 !ns_capable(ns->user_ns, CAP_IPC_OWNER))
631 return -1;
632
633 return security_ipc_permission(ipcp, flag);
634}
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651void kernel_to_ipc64_perm (struct kern_ipc_perm *in, struct ipc64_perm *out)
652{
653 out->key = in->key;
654 out->uid = from_kuid_munged(current_user_ns(), in->uid);
655 out->gid = from_kgid_munged(current_user_ns(), in->gid);
656 out->cuid = from_kuid_munged(current_user_ns(), in->cuid);
657 out->cgid = from_kgid_munged(current_user_ns(), in->cgid);
658 out->mode = in->mode;
659 out->seq = in->seq;
660}
661
662
663
664
665
666
667
668
669
670
671void ipc64_perm_to_ipc_perm (struct ipc64_perm *in, struct ipc_perm *out)
672{
673 out->key = in->key;
674 SET_UID(out->uid, in->uid);
675 SET_GID(out->gid, in->gid);
676 SET_UID(out->cuid, in->cuid);
677 SET_GID(out->cgid, in->cgid);
678 out->mode = in->mode;
679 out->seq = in->seq;
680}
681
682
683
684
685
686
687
688
689
690
691
692struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
693{
694 struct kern_ipc_perm *out;
695 int lid = ipcid_to_idx(id);
696
697 rcu_read_lock();
698 out = idr_find(&ids->ipcs_idr, lid);
699 if (out == NULL) {
700 rcu_read_unlock();
701 return ERR_PTR(-EINVAL);
702 }
703
704 spin_lock(&out->lock);
705
706
707
708
709 if (out->deleted) {
710 spin_unlock(&out->lock);
711 rcu_read_unlock();
712 return ERR_PTR(-EINVAL);
713 }
714
715 return out;
716}
717
718struct kern_ipc_perm *ipc_lock_check(struct ipc_ids *ids, int id)
719{
720 struct kern_ipc_perm *out;
721
722 out = ipc_lock(ids, id);
723 if (IS_ERR(out))
724 return out;
725
726 if (ipc_checkid(out, id)) {
727 ipc_unlock(out);
728 return ERR_PTR(-EIDRM);
729 }
730
731 return out;
732}
733
734
735
736
737
738
739
740
741
742
743
744int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
745 struct ipc_ops *ops, struct ipc_params *params)
746{
747 if (params->key == IPC_PRIVATE)
748 return ipcget_new(ns, ids, ops, params);
749 else
750 return ipcget_public(ns, ids, ops, params);
751}
752
753
754
755
756
757
758int ipc_update_perm(struct ipc64_perm *in, struct kern_ipc_perm *out)
759{
760 kuid_t uid = make_kuid(current_user_ns(), in->uid);
761 kgid_t gid = make_kgid(current_user_ns(), in->gid);
762 if (!uid_valid(uid) || !gid_valid(gid))
763 return -EINVAL;
764
765 out->uid = uid;
766 out->gid = gid;
767 out->mode = (out->mode & ~S_IRWXUGO)
768 | (in->mode & S_IRWXUGO);
769
770 return 0;
771}
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790struct kern_ipc_perm *ipcctl_pre_down(struct ipc_namespace *ns,
791 struct ipc_ids *ids, int id, int cmd,
792 struct ipc64_perm *perm, int extra_perm)
793{
794 struct kern_ipc_perm *ipcp;
795 kuid_t euid;
796 int err;
797
798 down_write(&ids->rw_mutex);
799 ipcp = ipc_lock_check(ids, id);
800 if (IS_ERR(ipcp)) {
801 err = PTR_ERR(ipcp);
802 goto out_up;
803 }
804
805 audit_ipc_obj(ipcp);
806 if (cmd == IPC_SET)
807 audit_ipc_set_perm(extra_perm, perm->uid,
808 perm->gid, perm->mode);
809
810 euid = current_euid();
811 if (uid_eq(euid, ipcp->cuid) || uid_eq(euid, ipcp->uid) ||
812 ns_capable(ns->user_ns, CAP_SYS_ADMIN))
813 return ipcp;
814
815 err = -EPERM;
816 ipc_unlock(ipcp);
817out_up:
818 up_write(&ids->rw_mutex);
819 return ERR_PTR(err);
820}
821
822#ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION
823
824
825
826
827
828
829
830
831
832
833
834int ipc_parse_version (int *cmd)
835{
836 if (*cmd & IPC_64) {
837 *cmd ^= IPC_64;
838 return IPC_64;
839 } else {
840 return IPC_OLD;
841 }
842}
843
844#endif
845
846#ifdef CONFIG_PROC_FS
847struct ipc_proc_iter {
848 struct ipc_namespace *ns;
849 struct ipc_proc_iface *iface;
850};
851
852
853
854
855static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
856 loff_t *new_pos)
857{
858 struct kern_ipc_perm *ipc;
859 int total, id;
860
861 total = 0;
862 for (id = 0; id < pos && total < ids->in_use; id++) {
863 ipc = idr_find(&ids->ipcs_idr, id);
864 if (ipc != NULL)
865 total++;
866 }
867
868 if (total >= ids->in_use)
869 return NULL;
870
871 for ( ; pos < IPCMNI; pos++) {
872 ipc = idr_find(&ids->ipcs_idr, pos);
873 if (ipc != NULL) {
874 *new_pos = pos + 1;
875 ipc_lock_by_ptr(ipc);
876 return ipc;
877 }
878 }
879
880
881 return NULL;
882}
883
884static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
885{
886 struct ipc_proc_iter *iter = s->private;
887 struct ipc_proc_iface *iface = iter->iface;
888 struct kern_ipc_perm *ipc = it;
889
890
891 if (ipc && ipc != SEQ_START_TOKEN)
892 ipc_unlock(ipc);
893
894 return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
895}
896
897
898
899
900
901static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
902{
903 struct ipc_proc_iter *iter = s->private;
904 struct ipc_proc_iface *iface = iter->iface;
905 struct ipc_ids *ids;
906
907 ids = &iter->ns->ids[iface->ids];
908
909
910
911
912
913 down_read(&ids->rw_mutex);
914
915
916 if (*pos < 0)
917 return NULL;
918
919
920 if (*pos == 0)
921 return SEQ_START_TOKEN;
922
923
924 return sysvipc_find_ipc(ids, *pos - 1, pos);
925}
926
927static void sysvipc_proc_stop(struct seq_file *s, void *it)
928{
929 struct kern_ipc_perm *ipc = it;
930 struct ipc_proc_iter *iter = s->private;
931 struct ipc_proc_iface *iface = iter->iface;
932 struct ipc_ids *ids;
933
934
935 if (ipc && ipc != SEQ_START_TOKEN)
936 ipc_unlock(ipc);
937
938 ids = &iter->ns->ids[iface->ids];
939
940 up_read(&ids->rw_mutex);
941}
942
943static int sysvipc_proc_show(struct seq_file *s, void *it)
944{
945 struct ipc_proc_iter *iter = s->private;
946 struct ipc_proc_iface *iface = iter->iface;
947
948 if (it == SEQ_START_TOKEN)
949 return seq_puts(s, iface->header);
950
951 return iface->show(s, it);
952}
953
954static const struct seq_operations sysvipc_proc_seqops = {
955 .start = sysvipc_proc_start,
956 .stop = sysvipc_proc_stop,
957 .next = sysvipc_proc_next,
958 .show = sysvipc_proc_show,
959};
960
961static int sysvipc_proc_open(struct inode *inode, struct file *file)
962{
963 int ret;
964 struct seq_file *seq;
965 struct ipc_proc_iter *iter;
966
967 ret = -ENOMEM;
968 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
969 if (!iter)
970 goto out;
971
972 ret = seq_open(file, &sysvipc_proc_seqops);
973 if (ret)
974 goto out_kfree;
975
976 seq = file->private_data;
977 seq->private = iter;
978
979 iter->iface = PDE(inode)->data;
980 iter->ns = get_ipc_ns(current->nsproxy->ipc_ns);
981out:
982 return ret;
983out_kfree:
984 kfree(iter);
985 goto out;
986}
987
988static int sysvipc_proc_release(struct inode *inode, struct file *file)
989{
990 struct seq_file *seq = file->private_data;
991 struct ipc_proc_iter *iter = seq->private;
992 put_ipc_ns(iter->ns);
993 return seq_release_private(inode, file);
994}
995
996static const struct file_operations sysvipc_proc_fops = {
997 .open = sysvipc_proc_open,
998 .read = seq_read,
999 .llseek = seq_lseek,
1000 .release = sysvipc_proc_release,
1001};
1002#endif
1003