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(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(&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 NULL
156#define proc_ipcauto_dointvec_minmax NULL
157#endif
158
159static int zero;
160static int one = 1;
161
162static struct ctl_table ipc_kern_table[] = {
163 {
164 .procname = "shmmax",
165 .data = &init_ipc_ns.shm_ctlmax,
166 .maxlen = sizeof (init_ipc_ns.shm_ctlmax),
167 .mode = 0644,
168 .proc_handler = proc_ipc_doulongvec_minmax,
169 },
170 {
171 .procname = "shmall",
172 .data = &init_ipc_ns.shm_ctlall,
173 .maxlen = sizeof (init_ipc_ns.shm_ctlall),
174 .mode = 0644,
175 .proc_handler = proc_ipc_doulongvec_minmax,
176 },
177 {
178 .procname = "shmmni",
179 .data = &init_ipc_ns.shm_ctlmni,
180 .maxlen = sizeof (init_ipc_ns.shm_ctlmni),
181 .mode = 0644,
182 .proc_handler = proc_ipc_dointvec,
183 },
184 {
185 .procname = "shm_rmid_forced",
186 .data = &init_ipc_ns.shm_rmid_forced,
187 .maxlen = sizeof(init_ipc_ns.shm_rmid_forced),
188 .mode = 0644,
189 .proc_handler = proc_ipc_dointvec_minmax_orphans,
190 .extra1 = &zero,
191 .extra2 = &one,
192 },
193 {
194 .procname = "msgmax",
195 .data = &init_ipc_ns.msg_ctlmax,
196 .maxlen = sizeof (init_ipc_ns.msg_ctlmax),
197 .mode = 0644,
198 .proc_handler = proc_ipc_dointvec,
199 },
200 {
201 .procname = "msgmni",
202 .data = &init_ipc_ns.msg_ctlmni,
203 .maxlen = sizeof (init_ipc_ns.msg_ctlmni),
204 .mode = 0644,
205 .proc_handler = proc_ipc_callback_dointvec,
206 },
207 {
208 .procname = "msgmnb",
209 .data = &init_ipc_ns.msg_ctlmnb,
210 .maxlen = sizeof (init_ipc_ns.msg_ctlmnb),
211 .mode = 0644,
212 .proc_handler = proc_ipc_dointvec,
213 },
214 {
215 .procname = "sem",
216 .data = &init_ipc_ns.sem_ctls,
217 .maxlen = 4*sizeof (int),
218 .mode = 0644,
219 .proc_handler = proc_ipc_dointvec,
220 },
221 {
222 .procname = "auto_msgmni",
223 .data = &init_ipc_ns.auto_msgmni,
224 .maxlen = sizeof(int),
225 .mode = 0644,
226 .proc_handler = proc_ipcauto_dointvec_minmax,
227 .extra1 = &zero,
228 .extra2 = &one,
229 },
230 {}
231};
232
233static struct ctl_table ipc_root_table[] = {
234 {
235 .procname = "kernel",
236 .mode = 0555,
237 .child = ipc_kern_table,
238 },
239 {}
240};
241
242static int __init ipc_sysctl_init(void)
243{
244 register_sysctl_table(ipc_root_table);
245 return 0;
246}
247
248__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(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(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(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};
183
184static struct ctl_table_set *set_lookup(struct ctl_table_root *root)
185{
186 return ¤t->nsproxy->ipc_ns->ipc_set;
187}
188
189static int set_is_seen(struct ctl_table_set *set)
190{
191 return ¤t->nsproxy->ipc_ns->ipc_set == set;
192}
193
194static void ipc_set_ownership(struct ctl_table_header *head,
195 struct ctl_table *table,
196 kuid_t *uid, kgid_t *gid)
197{
198 struct ipc_namespace *ns =
199 container_of(head->set, struct ipc_namespace, ipc_set);
200
201 kuid_t ns_root_uid = make_kuid(ns->user_ns, 0);
202 kgid_t ns_root_gid = make_kgid(ns->user_ns, 0);
203
204 *uid = uid_valid(ns_root_uid) ? ns_root_uid : GLOBAL_ROOT_UID;
205 *gid = gid_valid(ns_root_gid) ? ns_root_gid : GLOBAL_ROOT_GID;
206}
207
208static int ipc_permissions(struct ctl_table_header *head, struct ctl_table *table)
209{
210 int mode = table->mode;
211
212#ifdef CONFIG_CHECKPOINT_RESTORE
213 struct ipc_namespace *ns =
214 container_of(head->set, struct ipc_namespace, ipc_set);
215
216 if (((table->data == &ns->ids[IPC_SEM_IDS].next_id) ||
217 (table->data == &ns->ids[IPC_MSG_IDS].next_id) ||
218 (table->data == &ns->ids[IPC_SHM_IDS].next_id)) &&
219 checkpoint_restore_ns_capable(ns->user_ns))
220 mode = 0666;
221 else
222#endif
223 {
224 kuid_t ns_root_uid;
225 kgid_t ns_root_gid;
226
227 ipc_set_ownership(head, table, &ns_root_uid, &ns_root_gid);
228
229 if (uid_eq(current_euid(), ns_root_uid))
230 mode >>= 6;
231
232 else if (in_egroup_p(ns_root_gid))
233 mode >>= 3;
234 }
235
236 mode &= 7;
237
238 return (mode << 6) | (mode << 3) | mode;
239}
240
241static struct ctl_table_root set_root = {
242 .lookup = set_lookup,
243 .permissions = ipc_permissions,
244 .set_ownership = ipc_set_ownership,
245};
246
247bool setup_ipc_sysctls(struct ipc_namespace *ns)
248{
249 struct ctl_table *tbl;
250
251 setup_sysctl_set(&ns->ipc_set, &set_root, set_is_seen);
252
253 tbl = kmemdup(ipc_sysctls, sizeof(ipc_sysctls), GFP_KERNEL);
254 if (tbl) {
255 int i;
256
257 for (i = 0; i < ARRAY_SIZE(ipc_sysctls); i++) {
258 if (tbl[i].data == &init_ipc_ns.shm_ctlmax)
259 tbl[i].data = &ns->shm_ctlmax;
260
261 else if (tbl[i].data == &init_ipc_ns.shm_ctlall)
262 tbl[i].data = &ns->shm_ctlall;
263
264 else if (tbl[i].data == &init_ipc_ns.shm_ctlmni)
265 tbl[i].data = &ns->shm_ctlmni;
266
267 else if (tbl[i].data == &init_ipc_ns.shm_rmid_forced)
268 tbl[i].data = &ns->shm_rmid_forced;
269
270 else if (tbl[i].data == &init_ipc_ns.msg_ctlmax)
271 tbl[i].data = &ns->msg_ctlmax;
272
273 else if (tbl[i].data == &init_ipc_ns.msg_ctlmni)
274 tbl[i].data = &ns->msg_ctlmni;
275
276 else if (tbl[i].data == &init_ipc_ns.msg_ctlmnb)
277 tbl[i].data = &ns->msg_ctlmnb;
278
279 else if (tbl[i].data == &init_ipc_ns.sem_ctls)
280 tbl[i].data = &ns->sem_ctls;
281#ifdef CONFIG_CHECKPOINT_RESTORE
282 else if (tbl[i].data == &init_ipc_ns.ids[IPC_SEM_IDS].next_id)
283 tbl[i].data = &ns->ids[IPC_SEM_IDS].next_id;
284
285 else if (tbl[i].data == &init_ipc_ns.ids[IPC_MSG_IDS].next_id)
286 tbl[i].data = &ns->ids[IPC_MSG_IDS].next_id;
287
288 else if (tbl[i].data == &init_ipc_ns.ids[IPC_SHM_IDS].next_id)
289 tbl[i].data = &ns->ids[IPC_SHM_IDS].next_id;
290#endif
291 else
292 tbl[i].data = NULL;
293 }
294
295 ns->ipc_sysctls = __register_sysctl_table(&ns->ipc_set, "kernel", tbl,
296 ARRAY_SIZE(ipc_sysctls));
297 }
298 if (!ns->ipc_sysctls) {
299 kfree(tbl);
300 retire_sysctl_set(&ns->ipc_set);
301 return false;
302 }
303
304 return true;
305}
306
307void retire_ipc_sysctls(struct ipc_namespace *ns)
308{
309 struct ctl_table *tbl;
310
311 tbl = ns->ipc_sysctls->ctl_table_arg;
312 unregister_sysctl_table(ns->ipc_sysctls);
313 retire_sysctl_set(&ns->ipc_set);
314 kfree(tbl);
315}
316
317static int __init ipc_sysctl_init(void)
318{
319 if (!setup_ipc_sysctls(&init_ipc_ns)) {
320 pr_warn("ipc sysctl registration failed\n");
321 return -ENOMEM;
322 }
323 return 0;
324}
325
326device_initcall(ipc_sysctl_init);
327
328static int __init ipc_mni_extend(char *str)
329{
330 ipc_mni = IPCMNI_EXTEND;
331 ipc_mni_shift = IPCMNI_EXTEND_SHIFT;
332 ipc_min_cycle = IPCMNI_EXTEND_MIN_CYCLE;
333 pr_info("IPCMNI extended to %d.\n", ipc_mni);
334 return 0;
335}
336early_param("ipcmni_extend", ipc_mni_extend);