Loading...
1/*
2 * Copyright (C) 2007
3 *
4 * Author: Eric Biederman <ebiederm@xmision.com>
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License as
8 * published by the Free Software Foundation, version 2 of the
9 * License.
10 */
11
12#include <linux/module.h>
13#include <linux/ipc.h>
14#include <linux/nsproxy.h>
15#include <linux/sysctl.h>
16#include <linux/uaccess.h>
17#include <linux/ipc_namespace.h>
18#include <linux/msg.h>
19#include "util.h"
20
21static void *get_ipc(ctl_table *table)
22{
23 char *which = table->data;
24 struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns;
25 which = (which - (char *)&init_ipc_ns) + (char *)ipc_ns;
26 return which;
27}
28
29#ifdef CONFIG_PROC_SYSCTL
30static int proc_ipc_dointvec(ctl_table *table, int write,
31 void __user *buffer, size_t *lenp, loff_t *ppos)
32{
33 struct ctl_table ipc_table;
34
35 memcpy(&ipc_table, table, sizeof(ipc_table));
36 ipc_table.data = get_ipc(table);
37
38 return proc_dointvec(&ipc_table, write, buffer, lenp, ppos);
39}
40
41static int proc_ipc_dointvec_minmax(ctl_table *table, int write,
42 void __user *buffer, size_t *lenp, loff_t *ppos)
43{
44 struct ctl_table ipc_table;
45
46 memcpy(&ipc_table, table, sizeof(ipc_table));
47 ipc_table.data = get_ipc(table);
48
49 return proc_dointvec_minmax(&ipc_table, write, buffer, lenp, ppos);
50}
51
52static int proc_ipc_dointvec_minmax_orphans(ctl_table *table, int write,
53 void __user *buffer, size_t *lenp, loff_t *ppos)
54{
55 struct ipc_namespace *ns = current->nsproxy->ipc_ns;
56 int err = proc_ipc_dointvec_minmax(table, write, buffer, lenp, ppos);
57
58 if (err < 0)
59 return err;
60 if (ns->shm_rmid_forced)
61 shm_destroy_orphaned(ns);
62 return err;
63}
64
65static int proc_ipc_callback_dointvec_minmax(ctl_table *table, int write,
66 void __user *buffer, size_t *lenp, loff_t *ppos)
67{
68 struct ctl_table ipc_table;
69 size_t lenp_bef = *lenp;
70 int rc;
71
72 memcpy(&ipc_table, table, sizeof(ipc_table));
73 ipc_table.data = get_ipc(table);
74
75 rc = proc_dointvec_minmax(&ipc_table, write, buffer, lenp, ppos);
76
77 if (write && !rc && lenp_bef == *lenp)
78 /*
79 * Tunable has successfully been changed by hand. Disable its
80 * automatic adjustment. This simply requires unregistering
81 * the notifiers that trigger recalculation.
82 */
83 unregister_ipcns_notifier(current->nsproxy->ipc_ns);
84
85 return rc;
86}
87
88static int proc_ipc_doulongvec_minmax(ctl_table *table, int write,
89 void __user *buffer, size_t *lenp, loff_t *ppos)
90{
91 struct ctl_table ipc_table;
92 memcpy(&ipc_table, table, sizeof(ipc_table));
93 ipc_table.data = get_ipc(table);
94
95 return proc_doulongvec_minmax(&ipc_table, write, buffer,
96 lenp, ppos);
97}
98
99/*
100 * Routine that is called when the file "auto_msgmni" has successfully been
101 * written.
102 * Two values are allowed:
103 * 0: unregister msgmni's callback routine from the ipc namespace notifier
104 * chain. This means that msgmni won't be recomputed anymore upon memory
105 * add/remove or ipc namespace creation/removal.
106 * 1: register back the callback routine.
107 */
108static void ipc_auto_callback(int val)
109{
110 if (!val)
111 unregister_ipcns_notifier(current->nsproxy->ipc_ns);
112 else {
113 /*
114 * Re-enable automatic recomputing only if not already
115 * enabled.
116 */
117 recompute_msgmni(current->nsproxy->ipc_ns);
118 cond_register_ipcns_notifier(current->nsproxy->ipc_ns);
119 }
120}
121
122static int proc_ipcauto_dointvec_minmax(ctl_table *table, int write,
123 void __user *buffer, size_t *lenp, loff_t *ppos)
124{
125 struct ctl_table ipc_table;
126 size_t lenp_bef = *lenp;
127 int oldval;
128 int rc;
129
130 memcpy(&ipc_table, table, sizeof(ipc_table));
131 ipc_table.data = get_ipc(table);
132 oldval = *((int *)(ipc_table.data));
133
134 rc = proc_dointvec_minmax(&ipc_table, write, buffer, lenp, ppos);
135
136 if (write && !rc && lenp_bef == *lenp) {
137 int newval = *((int *)(ipc_table.data));
138 /*
139 * The file "auto_msgmni" has correctly been set.
140 * React by (un)registering the corresponding tunable, if the
141 * value has changed.
142 */
143 if (newval != oldval)
144 ipc_auto_callback(newval);
145 }
146
147 return rc;
148}
149
150#else
151#define proc_ipc_doulongvec_minmax NULL
152#define proc_ipc_dointvec NULL
153#define proc_ipc_dointvec_minmax NULL
154#define proc_ipc_dointvec_minmax_orphans NULL
155#define proc_ipc_callback_dointvec_minmax NULL
156#define proc_ipcauto_dointvec_minmax NULL
157#endif
158
159static int zero;
160static int one = 1;
161static int int_max = INT_MAX;
162
163static struct ctl_table ipc_kern_table[] = {
164 {
165 .procname = "shmmax",
166 .data = &init_ipc_ns.shm_ctlmax,
167 .maxlen = sizeof(init_ipc_ns.shm_ctlmax),
168 .mode = 0644,
169 .proc_handler = proc_ipc_doulongvec_minmax,
170 },
171 {
172 .procname = "shmall",
173 .data = &init_ipc_ns.shm_ctlall,
174 .maxlen = sizeof(init_ipc_ns.shm_ctlall),
175 .mode = 0644,
176 .proc_handler = proc_ipc_doulongvec_minmax,
177 },
178 {
179 .procname = "shmmni",
180 .data = &init_ipc_ns.shm_ctlmni,
181 .maxlen = sizeof(init_ipc_ns.shm_ctlmni),
182 .mode = 0644,
183 .proc_handler = proc_ipc_dointvec,
184 },
185 {
186 .procname = "shm_rmid_forced",
187 .data = &init_ipc_ns.shm_rmid_forced,
188 .maxlen = sizeof(init_ipc_ns.shm_rmid_forced),
189 .mode = 0644,
190 .proc_handler = proc_ipc_dointvec_minmax_orphans,
191 .extra1 = &zero,
192 .extra2 = &one,
193 },
194 {
195 .procname = "msgmax",
196 .data = &init_ipc_ns.msg_ctlmax,
197 .maxlen = sizeof(init_ipc_ns.msg_ctlmax),
198 .mode = 0644,
199 .proc_handler = proc_ipc_dointvec_minmax,
200 .extra1 = &zero,
201 .extra2 = &int_max,
202 },
203 {
204 .procname = "msgmni",
205 .data = &init_ipc_ns.msg_ctlmni,
206 .maxlen = sizeof(init_ipc_ns.msg_ctlmni),
207 .mode = 0644,
208 .proc_handler = proc_ipc_callback_dointvec_minmax,
209 .extra1 = &zero,
210 .extra2 = &int_max,
211 },
212 {
213 .procname = "msgmnb",
214 .data = &init_ipc_ns.msg_ctlmnb,
215 .maxlen = sizeof(init_ipc_ns.msg_ctlmnb),
216 .mode = 0644,
217 .proc_handler = proc_ipc_dointvec_minmax,
218 .extra1 = &zero,
219 .extra2 = &int_max,
220 },
221 {
222 .procname = "sem",
223 .data = &init_ipc_ns.sem_ctls,
224 .maxlen = 4*sizeof(int),
225 .mode = 0644,
226 .proc_handler = proc_ipc_dointvec,
227 },
228 {
229 .procname = "auto_msgmni",
230 .data = &init_ipc_ns.auto_msgmni,
231 .maxlen = sizeof(int),
232 .mode = 0644,
233 .proc_handler = proc_ipcauto_dointvec_minmax,
234 .extra1 = &zero,
235 .extra2 = &one,
236 },
237#ifdef CONFIG_CHECKPOINT_RESTORE
238 {
239 .procname = "sem_next_id",
240 .data = &init_ipc_ns.ids[IPC_SEM_IDS].next_id,
241 .maxlen = sizeof(init_ipc_ns.ids[IPC_SEM_IDS].next_id),
242 .mode = 0644,
243 .proc_handler = proc_ipc_dointvec_minmax,
244 .extra1 = &zero,
245 .extra2 = &int_max,
246 },
247 {
248 .procname = "msg_next_id",
249 .data = &init_ipc_ns.ids[IPC_MSG_IDS].next_id,
250 .maxlen = sizeof(init_ipc_ns.ids[IPC_MSG_IDS].next_id),
251 .mode = 0644,
252 .proc_handler = proc_ipc_dointvec_minmax,
253 .extra1 = &zero,
254 .extra2 = &int_max,
255 },
256 {
257 .procname = "shm_next_id",
258 .data = &init_ipc_ns.ids[IPC_SHM_IDS].next_id,
259 .maxlen = sizeof(init_ipc_ns.ids[IPC_SHM_IDS].next_id),
260 .mode = 0644,
261 .proc_handler = proc_ipc_dointvec_minmax,
262 .extra1 = &zero,
263 .extra2 = &int_max,
264 },
265#endif
266 {}
267};
268
269static struct ctl_table ipc_root_table[] = {
270 {
271 .procname = "kernel",
272 .mode = 0555,
273 .child = ipc_kern_table,
274 },
275 {}
276};
277
278static int __init ipc_sysctl_init(void)
279{
280 register_sysctl_table(ipc_root_table);
281 return 0;
282}
283
284device_initcall(ipc_sysctl_init);
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * Copyright (C) 2007
4 *
5 * Author: Eric Biederman <ebiederm@xmision.com>
6 */
7
8#include <linux/module.h>
9#include <linux/ipc.h>
10#include <linux/nsproxy.h>
11#include <linux/sysctl.h>
12#include <linux/uaccess.h>
13#include <linux/capability.h>
14#include <linux/ipc_namespace.h>
15#include <linux/msg.h>
16#include <linux/slab.h>
17#include <linux/cred.h>
18#include "util.h"
19
20static int proc_ipc_dointvec_minmax_orphans(const struct ctl_table *table, int write,
21 void *buffer, size_t *lenp, loff_t *ppos)
22{
23 struct ipc_namespace *ns =
24 container_of(table->data, struct ipc_namespace, shm_rmid_forced);
25 int err;
26
27 err = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
28
29 if (err < 0)
30 return err;
31 if (ns->shm_rmid_forced)
32 shm_destroy_orphaned(ns);
33 return err;
34}
35
36static int proc_ipc_auto_msgmni(const struct ctl_table *table, int write,
37 void *buffer, size_t *lenp, loff_t *ppos)
38{
39 struct ctl_table ipc_table;
40 int dummy = 0;
41
42 memcpy(&ipc_table, table, sizeof(ipc_table));
43 ipc_table.data = &dummy;
44
45 if (write)
46 pr_info_once("writing to auto_msgmni has no effect");
47
48 return proc_dointvec_minmax(&ipc_table, write, buffer, lenp, ppos);
49}
50
51static int proc_ipc_sem_dointvec(const struct ctl_table *table, int write,
52 void *buffer, size_t *lenp, loff_t *ppos)
53{
54 struct ipc_namespace *ns =
55 container_of(table->data, struct ipc_namespace, sem_ctls);
56 int ret, semmni;
57
58 semmni = ns->sem_ctls[3];
59 ret = proc_dointvec(table, write, buffer, lenp, ppos);
60
61 if (!ret)
62 ret = sem_check_semmni(ns);
63
64 /*
65 * Reset the semmni value if an error happens.
66 */
67 if (ret)
68 ns->sem_ctls[3] = semmni;
69 return ret;
70}
71
72int ipc_mni = IPCMNI;
73int ipc_mni_shift = IPCMNI_SHIFT;
74int ipc_min_cycle = RADIX_TREE_MAP_SIZE;
75
76static struct ctl_table ipc_sysctls[] = {
77 {
78 .procname = "shmmax",
79 .data = &init_ipc_ns.shm_ctlmax,
80 .maxlen = sizeof(init_ipc_ns.shm_ctlmax),
81 .mode = 0644,
82 .proc_handler = proc_doulongvec_minmax,
83 },
84 {
85 .procname = "shmall",
86 .data = &init_ipc_ns.shm_ctlall,
87 .maxlen = sizeof(init_ipc_ns.shm_ctlall),
88 .mode = 0644,
89 .proc_handler = proc_doulongvec_minmax,
90 },
91 {
92 .procname = "shmmni",
93 .data = &init_ipc_ns.shm_ctlmni,
94 .maxlen = sizeof(init_ipc_ns.shm_ctlmni),
95 .mode = 0644,
96 .proc_handler = proc_dointvec_minmax,
97 .extra1 = SYSCTL_ZERO,
98 .extra2 = &ipc_mni,
99 },
100 {
101 .procname = "shm_rmid_forced",
102 .data = &init_ipc_ns.shm_rmid_forced,
103 .maxlen = sizeof(init_ipc_ns.shm_rmid_forced),
104 .mode = 0644,
105 .proc_handler = proc_ipc_dointvec_minmax_orphans,
106 .extra1 = SYSCTL_ZERO,
107 .extra2 = SYSCTL_ONE,
108 },
109 {
110 .procname = "msgmax",
111 .data = &init_ipc_ns.msg_ctlmax,
112 .maxlen = sizeof(init_ipc_ns.msg_ctlmax),
113 .mode = 0644,
114 .proc_handler = proc_dointvec_minmax,
115 .extra1 = SYSCTL_ZERO,
116 .extra2 = SYSCTL_INT_MAX,
117 },
118 {
119 .procname = "msgmni",
120 .data = &init_ipc_ns.msg_ctlmni,
121 .maxlen = sizeof(init_ipc_ns.msg_ctlmni),
122 .mode = 0644,
123 .proc_handler = proc_dointvec_minmax,
124 .extra1 = SYSCTL_ZERO,
125 .extra2 = &ipc_mni,
126 },
127 {
128 .procname = "auto_msgmni",
129 .data = NULL,
130 .maxlen = sizeof(int),
131 .mode = 0644,
132 .proc_handler = proc_ipc_auto_msgmni,
133 .extra1 = SYSCTL_ZERO,
134 .extra2 = SYSCTL_ONE,
135 },
136 {
137 .procname = "msgmnb",
138 .data = &init_ipc_ns.msg_ctlmnb,
139 .maxlen = sizeof(init_ipc_ns.msg_ctlmnb),
140 .mode = 0644,
141 .proc_handler = proc_dointvec_minmax,
142 .extra1 = SYSCTL_ZERO,
143 .extra2 = SYSCTL_INT_MAX,
144 },
145 {
146 .procname = "sem",
147 .data = &init_ipc_ns.sem_ctls,
148 .maxlen = 4*sizeof(int),
149 .mode = 0644,
150 .proc_handler = proc_ipc_sem_dointvec,
151 },
152#ifdef CONFIG_CHECKPOINT_RESTORE
153 {
154 .procname = "sem_next_id",
155 .data = &init_ipc_ns.ids[IPC_SEM_IDS].next_id,
156 .maxlen = sizeof(init_ipc_ns.ids[IPC_SEM_IDS].next_id),
157 .mode = 0444,
158 .proc_handler = proc_dointvec_minmax,
159 .extra1 = SYSCTL_ZERO,
160 .extra2 = SYSCTL_INT_MAX,
161 },
162 {
163 .procname = "msg_next_id",
164 .data = &init_ipc_ns.ids[IPC_MSG_IDS].next_id,
165 .maxlen = sizeof(init_ipc_ns.ids[IPC_MSG_IDS].next_id),
166 .mode = 0444,
167 .proc_handler = proc_dointvec_minmax,
168 .extra1 = SYSCTL_ZERO,
169 .extra2 = SYSCTL_INT_MAX,
170 },
171 {
172 .procname = "shm_next_id",
173 .data = &init_ipc_ns.ids[IPC_SHM_IDS].next_id,
174 .maxlen = sizeof(init_ipc_ns.ids[IPC_SHM_IDS].next_id),
175 .mode = 0444,
176 .proc_handler = proc_dointvec_minmax,
177 .extra1 = SYSCTL_ZERO,
178 .extra2 = SYSCTL_INT_MAX,
179 },
180#endif
181};
182
183static struct ctl_table_set *set_lookup(struct ctl_table_root *root)
184{
185 return ¤t->nsproxy->ipc_ns->ipc_set;
186}
187
188static int set_is_seen(struct ctl_table_set *set)
189{
190 return ¤t->nsproxy->ipc_ns->ipc_set == set;
191}
192
193static void ipc_set_ownership(struct ctl_table_header *head,
194 kuid_t *uid, kgid_t *gid)
195{
196 struct ipc_namespace *ns =
197 container_of(head->set, struct ipc_namespace, ipc_set);
198
199 kuid_t ns_root_uid = make_kuid(ns->user_ns, 0);
200 kgid_t ns_root_gid = make_kgid(ns->user_ns, 0);
201
202 *uid = uid_valid(ns_root_uid) ? ns_root_uid : GLOBAL_ROOT_UID;
203 *gid = gid_valid(ns_root_gid) ? ns_root_gid : GLOBAL_ROOT_GID;
204}
205
206static int ipc_permissions(struct ctl_table_header *head, const struct ctl_table *table)
207{
208 int mode = table->mode;
209
210#ifdef CONFIG_CHECKPOINT_RESTORE
211 struct ipc_namespace *ns =
212 container_of(head->set, struct ipc_namespace, ipc_set);
213
214 if (((table->data == &ns->ids[IPC_SEM_IDS].next_id) ||
215 (table->data == &ns->ids[IPC_MSG_IDS].next_id) ||
216 (table->data == &ns->ids[IPC_SHM_IDS].next_id)) &&
217 checkpoint_restore_ns_capable(ns->user_ns))
218 mode = 0666;
219 else
220#endif
221 {
222 kuid_t ns_root_uid;
223 kgid_t ns_root_gid;
224
225 ipc_set_ownership(head, &ns_root_uid, &ns_root_gid);
226
227 if (uid_eq(current_euid(), ns_root_uid))
228 mode >>= 6;
229
230 else if (in_egroup_p(ns_root_gid))
231 mode >>= 3;
232 }
233
234 mode &= 7;
235
236 return (mode << 6) | (mode << 3) | mode;
237}
238
239static struct ctl_table_root set_root = {
240 .lookup = set_lookup,
241 .permissions = ipc_permissions,
242 .set_ownership = ipc_set_ownership,
243};
244
245bool setup_ipc_sysctls(struct ipc_namespace *ns)
246{
247 struct ctl_table *tbl;
248
249 setup_sysctl_set(&ns->ipc_set, &set_root, set_is_seen);
250
251 tbl = kmemdup(ipc_sysctls, sizeof(ipc_sysctls), GFP_KERNEL);
252 if (tbl) {
253 int i;
254
255 for (i = 0; i < ARRAY_SIZE(ipc_sysctls); i++) {
256 if (tbl[i].data == &init_ipc_ns.shm_ctlmax)
257 tbl[i].data = &ns->shm_ctlmax;
258
259 else if (tbl[i].data == &init_ipc_ns.shm_ctlall)
260 tbl[i].data = &ns->shm_ctlall;
261
262 else if (tbl[i].data == &init_ipc_ns.shm_ctlmni)
263 tbl[i].data = &ns->shm_ctlmni;
264
265 else if (tbl[i].data == &init_ipc_ns.shm_rmid_forced)
266 tbl[i].data = &ns->shm_rmid_forced;
267
268 else if (tbl[i].data == &init_ipc_ns.msg_ctlmax)
269 tbl[i].data = &ns->msg_ctlmax;
270
271 else if (tbl[i].data == &init_ipc_ns.msg_ctlmni)
272 tbl[i].data = &ns->msg_ctlmni;
273
274 else if (tbl[i].data == &init_ipc_ns.msg_ctlmnb)
275 tbl[i].data = &ns->msg_ctlmnb;
276
277 else if (tbl[i].data == &init_ipc_ns.sem_ctls)
278 tbl[i].data = &ns->sem_ctls;
279#ifdef CONFIG_CHECKPOINT_RESTORE
280 else if (tbl[i].data == &init_ipc_ns.ids[IPC_SEM_IDS].next_id)
281 tbl[i].data = &ns->ids[IPC_SEM_IDS].next_id;
282
283 else if (tbl[i].data == &init_ipc_ns.ids[IPC_MSG_IDS].next_id)
284 tbl[i].data = &ns->ids[IPC_MSG_IDS].next_id;
285
286 else if (tbl[i].data == &init_ipc_ns.ids[IPC_SHM_IDS].next_id)
287 tbl[i].data = &ns->ids[IPC_SHM_IDS].next_id;
288#endif
289 else
290 tbl[i].data = NULL;
291 }
292
293 ns->ipc_sysctls = __register_sysctl_table(&ns->ipc_set, "kernel", tbl,
294 ARRAY_SIZE(ipc_sysctls));
295 }
296 if (!ns->ipc_sysctls) {
297 kfree(tbl);
298 retire_sysctl_set(&ns->ipc_set);
299 return false;
300 }
301
302 return true;
303}
304
305void retire_ipc_sysctls(struct ipc_namespace *ns)
306{
307 const struct ctl_table *tbl;
308
309 tbl = ns->ipc_sysctls->ctl_table_arg;
310 unregister_sysctl_table(ns->ipc_sysctls);
311 retire_sysctl_set(&ns->ipc_set);
312 kfree(tbl);
313}
314
315static int __init ipc_sysctl_init(void)
316{
317 if (!setup_ipc_sysctls(&init_ipc_ns)) {
318 pr_warn("ipc sysctl registration failed\n");
319 return -ENOMEM;
320 }
321 return 0;
322}
323
324device_initcall(ipc_sysctl_init);
325
326static int __init ipc_mni_extend(char *str)
327{
328 ipc_mni = IPCMNI_EXTEND;
329 ipc_mni_shift = IPCMNI_EXTEND_SHIFT;
330 ipc_min_cycle = IPCMNI_EXTEND_MIN_CYCLE;
331 pr_info("IPCMNI extended to %d.\n", ipc_mni);
332 return 0;
333}
334early_param("ipcmni_extend", ipc_mni_extend);