]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/kernel/seccomp.c | |
3 | * | |
4 | * Copyright 2004-2005 Andrea Arcangeli <andrea@cpushare.com> | |
5 | * | |
e2cfabdf WD |
6 | * Copyright (C) 2012 Google, Inc. |
7 | * Will Drewry <wad@chromium.org> | |
8 | * | |
9 | * This defines a simple but solid secure-computing facility. | |
10 | * | |
11 | * Mode 1 uses a fixed list of allowed system calls. | |
12 | * Mode 2 allows user-defined system call filters in the form | |
13 | * of Berkeley Packet Filters/Linux Socket Filters. | |
1da177e4 LT |
14 | */ |
15 | ||
0b5fa229 | 16 | #include <linux/refcount.h> |
85e7bac3 | 17 | #include <linux/audit.h> |
5b101740 | 18 | #include <linux/compat.h> |
b25e6716 | 19 | #include <linux/coredump.h> |
9ca58825 | 20 | #include <linux/kmemleak.h> |
e2cfabdf | 21 | #include <linux/sched.h> |
68db0cf1 | 22 | #include <linux/sched/task_stack.h> |
e2cfabdf | 23 | #include <linux/seccomp.h> |
c8bee430 | 24 | #include <linux/slab.h> |
48dc92b9 | 25 | #include <linux/syscalls.h> |
9ca58825 | 26 | #include <linux/sysctl.h> |
1da177e4 | 27 | |
a4412fc9 | 28 | #ifdef CONFIG_HAVE_ARCH_SECCOMP_FILTER |
e2cfabdf | 29 | #include <asm/syscall.h> |
a4412fc9 | 30 | #endif |
e2cfabdf WD |
31 | |
32 | #ifdef CONFIG_SECCOMP_FILTER | |
e2cfabdf | 33 | #include <linux/filter.h> |
c2e1f2e3 | 34 | #include <linux/pid.h> |
fb0fadf9 | 35 | #include <linux/ptrace.h> |
e2cfabdf | 36 | #include <linux/security.h> |
e2cfabdf WD |
37 | #include <linux/tracehook.h> |
38 | #include <linux/uaccess.h> | |
39 | ||
40 | /** | |
41 | * struct seccomp_filter - container for seccomp BPF programs | |
42 | * | |
43 | * @usage: reference count to manage the object lifetime. | |
44 | * get/put helpers should be used when accessing an instance | |
45 | * outside of a lifetime-guarded section. In general, this | |
46 | * is only needed for handling filters shared across tasks. | |
8ac8bbe1 | 47 | * @log: true if all actions except for SECCOMP_RET_ALLOW should be logged |
e2cfabdf | 48 | * @prev: points to a previously installed, or inherited, filter |
285fdfc5 | 49 | * @prog: the BPF program to evaluate |
e2cfabdf WD |
50 | * |
51 | * seccomp_filter objects are organized in a tree linked via the @prev | |
52 | * pointer. For any task, it appears to be a singly-linked list starting | |
53 | * with current->seccomp.filter, the most recently attached or inherited filter. | |
54 | * However, multiple filters may share a @prev node, by way of fork(), which | |
55 | * results in a unidirectional tree existing in memory. This is similar to | |
56 | * how namespaces work. | |
57 | * | |
58 | * seccomp_filter objects should never be modified after being attached | |
59 | * to a task_struct (other than @usage). | |
60 | */ | |
61 | struct seccomp_filter { | |
0b5fa229 | 62 | refcount_t usage; |
8ac8bbe1 | 63 | bool log; |
e2cfabdf | 64 | struct seccomp_filter *prev; |
7ae457c1 | 65 | struct bpf_prog *prog; |
e2cfabdf WD |
66 | }; |
67 | ||
68 | /* Limit any path through the tree to 256KB worth of instructions. */ | |
69 | #define MAX_INSNS_PER_PATH ((1 << 18) / sizeof(struct sock_filter)) | |
70 | ||
bd4cf0ed | 71 | /* |
e2cfabdf WD |
72 | * Endianness is explicitly ignored and left for BPF program authors to manage |
73 | * as per the specific architecture. | |
74 | */ | |
bd4cf0ed | 75 | static void populate_seccomp_data(struct seccomp_data *sd) |
e2cfabdf | 76 | { |
bd4cf0ed AS |
77 | struct task_struct *task = current; |
78 | struct pt_regs *regs = task_pt_regs(task); | |
2eac7648 | 79 | unsigned long args[6]; |
e2cfabdf | 80 | |
bd4cf0ed | 81 | sd->nr = syscall_get_nr(task, regs); |
0b747172 | 82 | sd->arch = syscall_get_arch(); |
2eac7648 DB |
83 | syscall_get_arguments(task, regs, 0, 6, args); |
84 | sd->args[0] = args[0]; | |
85 | sd->args[1] = args[1]; | |
86 | sd->args[2] = args[2]; | |
87 | sd->args[3] = args[3]; | |
88 | sd->args[4] = args[4]; | |
89 | sd->args[5] = args[5]; | |
bd4cf0ed | 90 | sd->instruction_pointer = KSTK_EIP(task); |
e2cfabdf WD |
91 | } |
92 | ||
93 | /** | |
94 | * seccomp_check_filter - verify seccomp filter code | |
95 | * @filter: filter to verify | |
96 | * @flen: length of filter | |
97 | * | |
4df95ff4 | 98 | * Takes a previously checked filter (by bpf_check_classic) and |
e2cfabdf WD |
99 | * redirects all filter code that loads struct sk_buff data |
100 | * and related data through seccomp_bpf_load. It also | |
101 | * enforces length and alignment checking of those loads. | |
102 | * | |
103 | * Returns 0 if the rule set is legal or -EINVAL if not. | |
104 | */ | |
105 | static int seccomp_check_filter(struct sock_filter *filter, unsigned int flen) | |
106 | { | |
107 | int pc; | |
108 | for (pc = 0; pc < flen; pc++) { | |
109 | struct sock_filter *ftest = &filter[pc]; | |
110 | u16 code = ftest->code; | |
111 | u32 k = ftest->k; | |
112 | ||
113 | switch (code) { | |
34805931 | 114 | case BPF_LD | BPF_W | BPF_ABS: |
bd4cf0ed | 115 | ftest->code = BPF_LDX | BPF_W | BPF_ABS; |
e2cfabdf WD |
116 | /* 32-bit aligned and not out of bounds. */ |
117 | if (k >= sizeof(struct seccomp_data) || k & 3) | |
118 | return -EINVAL; | |
119 | continue; | |
34805931 | 120 | case BPF_LD | BPF_W | BPF_LEN: |
bd4cf0ed | 121 | ftest->code = BPF_LD | BPF_IMM; |
e2cfabdf WD |
122 | ftest->k = sizeof(struct seccomp_data); |
123 | continue; | |
34805931 | 124 | case BPF_LDX | BPF_W | BPF_LEN: |
bd4cf0ed | 125 | ftest->code = BPF_LDX | BPF_IMM; |
e2cfabdf WD |
126 | ftest->k = sizeof(struct seccomp_data); |
127 | continue; | |
128 | /* Explicitly include allowed calls. */ | |
34805931 DB |
129 | case BPF_RET | BPF_K: |
130 | case BPF_RET | BPF_A: | |
131 | case BPF_ALU | BPF_ADD | BPF_K: | |
132 | case BPF_ALU | BPF_ADD | BPF_X: | |
133 | case BPF_ALU | BPF_SUB | BPF_K: | |
134 | case BPF_ALU | BPF_SUB | BPF_X: | |
135 | case BPF_ALU | BPF_MUL | BPF_K: | |
136 | case BPF_ALU | BPF_MUL | BPF_X: | |
137 | case BPF_ALU | BPF_DIV | BPF_K: | |
138 | case BPF_ALU | BPF_DIV | BPF_X: | |
139 | case BPF_ALU | BPF_AND | BPF_K: | |
140 | case BPF_ALU | BPF_AND | BPF_X: | |
141 | case BPF_ALU | BPF_OR | BPF_K: | |
142 | case BPF_ALU | BPF_OR | BPF_X: | |
143 | case BPF_ALU | BPF_XOR | BPF_K: | |
144 | case BPF_ALU | BPF_XOR | BPF_X: | |
145 | case BPF_ALU | BPF_LSH | BPF_K: | |
146 | case BPF_ALU | BPF_LSH | BPF_X: | |
147 | case BPF_ALU | BPF_RSH | BPF_K: | |
148 | case BPF_ALU | BPF_RSH | BPF_X: | |
149 | case BPF_ALU | BPF_NEG: | |
150 | case BPF_LD | BPF_IMM: | |
151 | case BPF_LDX | BPF_IMM: | |
152 | case BPF_MISC | BPF_TAX: | |
153 | case BPF_MISC | BPF_TXA: | |
154 | case BPF_LD | BPF_MEM: | |
155 | case BPF_LDX | BPF_MEM: | |
156 | case BPF_ST: | |
157 | case BPF_STX: | |
158 | case BPF_JMP | BPF_JA: | |
159 | case BPF_JMP | BPF_JEQ | BPF_K: | |
160 | case BPF_JMP | BPF_JEQ | BPF_X: | |
161 | case BPF_JMP | BPF_JGE | BPF_K: | |
162 | case BPF_JMP | BPF_JGE | BPF_X: | |
163 | case BPF_JMP | BPF_JGT | BPF_K: | |
164 | case BPF_JMP | BPF_JGT | BPF_X: | |
165 | case BPF_JMP | BPF_JSET | BPF_K: | |
166 | case BPF_JMP | BPF_JSET | BPF_X: | |
e2cfabdf WD |
167 | continue; |
168 | default: | |
169 | return -EINVAL; | |
170 | } | |
171 | } | |
172 | return 0; | |
173 | } | |
174 | ||
175 | /** | |
285fdfc5 MS |
176 | * seccomp_run_filters - evaluates all seccomp filters against @sd |
177 | * @sd: optional seccomp data to be passed to filters | |
e95596f2 KC |
178 | * @match: stores struct seccomp_filter that resulted in the return value, |
179 | * unless filter returned SECCOMP_RET_ALLOW, in which case it will | |
180 | * be unchanged. | |
e2cfabdf WD |
181 | * |
182 | * Returns valid seccomp BPF response codes. | |
183 | */ | |
e95596f2 KC |
184 | static u32 seccomp_run_filters(const struct seccomp_data *sd, |
185 | struct seccomp_filter **match) | |
e2cfabdf | 186 | { |
d39bd00d | 187 | struct seccomp_data sd_local; |
acf3b2c7 | 188 | u32 ret = SECCOMP_RET_ALLOW; |
8225d385 PK |
189 | /* Make sure cross-thread synced filter points somewhere sane. */ |
190 | struct seccomp_filter *f = | |
191 | lockless_dereference(current->seccomp.filter); | |
acf3b2c7 WD |
192 | |
193 | /* Ensure unexpected behavior doesn't result in failing open. */ | |
3ba2530c | 194 | if (unlikely(WARN_ON(f == NULL))) |
acf3b2c7 WD |
195 | return SECCOMP_RET_KILL; |
196 | ||
d39bd00d AL |
197 | if (!sd) { |
198 | populate_seccomp_data(&sd_local); | |
199 | sd = &sd_local; | |
200 | } | |
bd4cf0ed | 201 | |
e2cfabdf WD |
202 | /* |
203 | * All filters in the list are evaluated and the lowest BPF return | |
acf3b2c7 | 204 | * value always takes priority (ignoring the DATA). |
e2cfabdf | 205 | */ |
3ba2530c | 206 | for (; f; f = f->prev) { |
88575199 | 207 | u32 cur_ret = BPF_PROG_RUN(f->prog, sd); |
8f577cad | 208 | |
e95596f2 | 209 | if ((cur_ret & SECCOMP_RET_ACTION) < (ret & SECCOMP_RET_ACTION)) { |
acf3b2c7 | 210 | ret = cur_ret; |
e95596f2 KC |
211 | *match = f; |
212 | } | |
e2cfabdf WD |
213 | } |
214 | return ret; | |
215 | } | |
1f41b450 | 216 | #endif /* CONFIG_SECCOMP_FILTER */ |
e2cfabdf | 217 | |
1f41b450 KC |
218 | static inline bool seccomp_may_assign_mode(unsigned long seccomp_mode) |
219 | { | |
69f6a34b | 220 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 | 221 | |
1f41b450 KC |
222 | if (current->seccomp.mode && current->seccomp.mode != seccomp_mode) |
223 | return false; | |
224 | ||
225 | return true; | |
226 | } | |
227 | ||
3ba2530c KC |
228 | static inline void seccomp_assign_mode(struct task_struct *task, |
229 | unsigned long seccomp_mode) | |
1f41b450 | 230 | { |
69f6a34b | 231 | assert_spin_locked(&task->sighand->siglock); |
dbd95212 | 232 | |
3ba2530c KC |
233 | task->seccomp.mode = seccomp_mode; |
234 | /* | |
235 | * Make sure TIF_SECCOMP cannot be set before the mode (and | |
236 | * filter) is set. | |
237 | */ | |
238 | smp_mb__before_atomic(); | |
239 | set_tsk_thread_flag(task, TIF_SECCOMP); | |
1f41b450 KC |
240 | } |
241 | ||
242 | #ifdef CONFIG_SECCOMP_FILTER | |
c2e1f2e3 KC |
243 | /* Returns 1 if the parent is an ancestor of the child. */ |
244 | static int is_ancestor(struct seccomp_filter *parent, | |
245 | struct seccomp_filter *child) | |
246 | { | |
247 | /* NULL is the root ancestor. */ | |
248 | if (parent == NULL) | |
249 | return 1; | |
250 | for (; child; child = child->prev) | |
251 | if (child == parent) | |
252 | return 1; | |
253 | return 0; | |
254 | } | |
255 | ||
256 | /** | |
257 | * seccomp_can_sync_threads: checks if all threads can be synchronized | |
258 | * | |
259 | * Expects sighand and cred_guard_mutex locks to be held. | |
260 | * | |
261 | * Returns 0 on success, -ve on error, or the pid of a thread which was | |
262 | * either not in the correct seccomp mode or it did not have an ancestral | |
263 | * seccomp filter. | |
264 | */ | |
265 | static inline pid_t seccomp_can_sync_threads(void) | |
266 | { | |
267 | struct task_struct *thread, *caller; | |
268 | ||
269 | BUG_ON(!mutex_is_locked(¤t->signal->cred_guard_mutex)); | |
69f6a34b | 270 | assert_spin_locked(¤t->sighand->siglock); |
c2e1f2e3 KC |
271 | |
272 | /* Validate all threads being eligible for synchronization. */ | |
273 | caller = current; | |
274 | for_each_thread(caller, thread) { | |
275 | pid_t failed; | |
276 | ||
277 | /* Skip current, since it is initiating the sync. */ | |
278 | if (thread == caller) | |
279 | continue; | |
280 | ||
281 | if (thread->seccomp.mode == SECCOMP_MODE_DISABLED || | |
282 | (thread->seccomp.mode == SECCOMP_MODE_FILTER && | |
283 | is_ancestor(thread->seccomp.filter, | |
284 | caller->seccomp.filter))) | |
285 | continue; | |
286 | ||
287 | /* Return the first thread that cannot be synchronized. */ | |
288 | failed = task_pid_vnr(thread); | |
289 | /* If the pid cannot be resolved, then return -ESRCH */ | |
290 | if (unlikely(WARN_ON(failed == 0))) | |
291 | failed = -ESRCH; | |
292 | return failed; | |
293 | } | |
294 | ||
295 | return 0; | |
296 | } | |
297 | ||
298 | /** | |
299 | * seccomp_sync_threads: sets all threads to use current's filter | |
300 | * | |
301 | * Expects sighand and cred_guard_mutex locks to be held, and for | |
302 | * seccomp_can_sync_threads() to have returned success already | |
303 | * without dropping the locks. | |
304 | * | |
305 | */ | |
306 | static inline void seccomp_sync_threads(void) | |
307 | { | |
308 | struct task_struct *thread, *caller; | |
309 | ||
310 | BUG_ON(!mutex_is_locked(¤t->signal->cred_guard_mutex)); | |
69f6a34b | 311 | assert_spin_locked(¤t->sighand->siglock); |
c2e1f2e3 KC |
312 | |
313 | /* Synchronize all threads. */ | |
314 | caller = current; | |
315 | for_each_thread(caller, thread) { | |
316 | /* Skip current, since it needs no changes. */ | |
317 | if (thread == caller) | |
318 | continue; | |
319 | ||
320 | /* Get a task reference for the new leaf node. */ | |
321 | get_seccomp_filter(caller); | |
322 | /* | |
323 | * Drop the task reference to the shared ancestor since | |
324 | * current's path will hold a reference. (This also | |
325 | * allows a put before the assignment.) | |
326 | */ | |
327 | put_seccomp_filter(thread); | |
328 | smp_store_release(&thread->seccomp.filter, | |
329 | caller->seccomp.filter); | |
103502a3 JH |
330 | |
331 | /* | |
332 | * Don't let an unprivileged task work around | |
333 | * the no_new_privs restriction by creating | |
334 | * a thread that sets it up, enters seccomp, | |
335 | * then dies. | |
336 | */ | |
337 | if (task_no_new_privs(caller)) | |
338 | task_set_no_new_privs(thread); | |
339 | ||
c2e1f2e3 KC |
340 | /* |
341 | * Opt the other thread into seccomp if needed. | |
342 | * As threads are considered to be trust-realm | |
343 | * equivalent (see ptrace_may_access), it is safe to | |
344 | * allow one thread to transition the other. | |
345 | */ | |
103502a3 | 346 | if (thread->seccomp.mode == SECCOMP_MODE_DISABLED) |
c2e1f2e3 | 347 | seccomp_assign_mode(thread, SECCOMP_MODE_FILTER); |
c2e1f2e3 KC |
348 | } |
349 | } | |
350 | ||
e2cfabdf | 351 | /** |
c8bee430 | 352 | * seccomp_prepare_filter: Prepares a seccomp filter for use. |
e2cfabdf WD |
353 | * @fprog: BPF program to install |
354 | * | |
c8bee430 | 355 | * Returns filter on success or an ERR_PTR on failure. |
e2cfabdf | 356 | */ |
c8bee430 | 357 | static struct seccomp_filter *seccomp_prepare_filter(struct sock_fprog *fprog) |
e2cfabdf | 358 | { |
ac67eb2c DB |
359 | struct seccomp_filter *sfilter; |
360 | int ret; | |
97f2645f | 361 | const bool save_orig = IS_ENABLED(CONFIG_CHECKPOINT_RESTORE); |
e2cfabdf WD |
362 | |
363 | if (fprog->len == 0 || fprog->len > BPF_MAXINSNS) | |
c8bee430 | 364 | return ERR_PTR(-EINVAL); |
d9e12f42 | 365 | |
c8bee430 | 366 | BUG_ON(INT_MAX / fprog->len < sizeof(struct sock_filter)); |
e2cfabdf WD |
367 | |
368 | /* | |
119ce5c8 | 369 | * Installing a seccomp filter requires that the task has |
e2cfabdf WD |
370 | * CAP_SYS_ADMIN in its namespace or be running with no_new_privs. |
371 | * This avoids scenarios where unprivileged tasks can affect the | |
372 | * behavior of privileged children. | |
373 | */ | |
1d4457f9 | 374 | if (!task_no_new_privs(current) && |
e2cfabdf WD |
375 | security_capable_noaudit(current_cred(), current_user_ns(), |
376 | CAP_SYS_ADMIN) != 0) | |
c8bee430 | 377 | return ERR_PTR(-EACCES); |
e2cfabdf | 378 | |
bd4cf0ed | 379 | /* Allocate a new seccomp_filter */ |
ac67eb2c DB |
380 | sfilter = kzalloc(sizeof(*sfilter), GFP_KERNEL | __GFP_NOWARN); |
381 | if (!sfilter) | |
d9e12f42 | 382 | return ERR_PTR(-ENOMEM); |
ac67eb2c DB |
383 | |
384 | ret = bpf_prog_create_from_user(&sfilter->prog, fprog, | |
f8e529ed | 385 | seccomp_check_filter, save_orig); |
ac67eb2c DB |
386 | if (ret < 0) { |
387 | kfree(sfilter); | |
388 | return ERR_PTR(ret); | |
d9e12f42 | 389 | } |
bd4cf0ed | 390 | |
0b5fa229 | 391 | refcount_set(&sfilter->usage, 1); |
e2cfabdf | 392 | |
ac67eb2c | 393 | return sfilter; |
e2cfabdf WD |
394 | } |
395 | ||
396 | /** | |
c8bee430 | 397 | * seccomp_prepare_user_filter - prepares a user-supplied sock_fprog |
e2cfabdf WD |
398 | * @user_filter: pointer to the user data containing a sock_fprog. |
399 | * | |
400 | * Returns 0 on success and non-zero otherwise. | |
401 | */ | |
c8bee430 KC |
402 | static struct seccomp_filter * |
403 | seccomp_prepare_user_filter(const char __user *user_filter) | |
e2cfabdf WD |
404 | { |
405 | struct sock_fprog fprog; | |
c8bee430 | 406 | struct seccomp_filter *filter = ERR_PTR(-EFAULT); |
e2cfabdf WD |
407 | |
408 | #ifdef CONFIG_COMPAT | |
5c38065e | 409 | if (in_compat_syscall()) { |
e2cfabdf WD |
410 | struct compat_sock_fprog fprog32; |
411 | if (copy_from_user(&fprog32, user_filter, sizeof(fprog32))) | |
412 | goto out; | |
413 | fprog.len = fprog32.len; | |
414 | fprog.filter = compat_ptr(fprog32.filter); | |
415 | } else /* falls through to the if below. */ | |
416 | #endif | |
417 | if (copy_from_user(&fprog, user_filter, sizeof(fprog))) | |
418 | goto out; | |
c8bee430 | 419 | filter = seccomp_prepare_filter(&fprog); |
e2cfabdf | 420 | out: |
c8bee430 KC |
421 | return filter; |
422 | } | |
423 | ||
424 | /** | |
425 | * seccomp_attach_filter: validate and attach filter | |
426 | * @flags: flags to change filter behavior | |
427 | * @filter: seccomp filter to add to the current process | |
428 | * | |
dbd95212 KC |
429 | * Caller must be holding current->sighand->siglock lock. |
430 | * | |
c8bee430 KC |
431 | * Returns 0 on success, -ve on error. |
432 | */ | |
433 | static long seccomp_attach_filter(unsigned int flags, | |
434 | struct seccomp_filter *filter) | |
435 | { | |
436 | unsigned long total_insns; | |
437 | struct seccomp_filter *walker; | |
438 | ||
69f6a34b | 439 | assert_spin_locked(¤t->sighand->siglock); |
dbd95212 | 440 | |
c8bee430 KC |
441 | /* Validate resulting filter length. */ |
442 | total_insns = filter->prog->len; | |
443 | for (walker = current->seccomp.filter; walker; walker = walker->prev) | |
444 | total_insns += walker->prog->len + 4; /* 4 instr penalty */ | |
445 | if (total_insns > MAX_INSNS_PER_PATH) | |
446 | return -ENOMEM; | |
447 | ||
c2e1f2e3 KC |
448 | /* If thread sync has been requested, check that it is possible. */ |
449 | if (flags & SECCOMP_FILTER_FLAG_TSYNC) { | |
450 | int ret; | |
451 | ||
452 | ret = seccomp_can_sync_threads(); | |
453 | if (ret) | |
454 | return ret; | |
455 | } | |
456 | ||
8ac8bbe1 TH |
457 | /* Set log flag, if present. */ |
458 | if (flags & SECCOMP_FILTER_FLAG_LOG) | |
459 | filter->log = true; | |
460 | ||
c8bee430 KC |
461 | /* |
462 | * If there is an existing filter, make it the prev and don't drop its | |
463 | * task reference. | |
464 | */ | |
465 | filter->prev = current->seccomp.filter; | |
466 | current->seccomp.filter = filter; | |
467 | ||
c2e1f2e3 KC |
468 | /* Now that the new filter is in place, synchronize to all threads. */ |
469 | if (flags & SECCOMP_FILTER_FLAG_TSYNC) | |
470 | seccomp_sync_threads(); | |
471 | ||
c8bee430 | 472 | return 0; |
e2cfabdf WD |
473 | } |
474 | ||
475 | /* get_seccomp_filter - increments the reference count of the filter on @tsk */ | |
476 | void get_seccomp_filter(struct task_struct *tsk) | |
477 | { | |
478 | struct seccomp_filter *orig = tsk->seccomp.filter; | |
479 | if (!orig) | |
480 | return; | |
481 | /* Reference count is bounded by the number of total processes. */ | |
0b5fa229 | 482 | refcount_inc(&orig->usage); |
e2cfabdf WD |
483 | } |
484 | ||
c8bee430 KC |
485 | static inline void seccomp_filter_free(struct seccomp_filter *filter) |
486 | { | |
487 | if (filter) { | |
bab18991 | 488 | bpf_prog_destroy(filter->prog); |
c8bee430 KC |
489 | kfree(filter); |
490 | } | |
491 | } | |
492 | ||
e2cfabdf WD |
493 | /* put_seccomp_filter - decrements the ref count of tsk->seccomp.filter */ |
494 | void put_seccomp_filter(struct task_struct *tsk) | |
495 | { | |
496 | struct seccomp_filter *orig = tsk->seccomp.filter; | |
497 | /* Clean up single-reference branches iteratively. */ | |
0b5fa229 | 498 | while (orig && refcount_dec_and_test(&orig->usage)) { |
e2cfabdf WD |
499 | struct seccomp_filter *freeme = orig; |
500 | orig = orig->prev; | |
c8bee430 | 501 | seccomp_filter_free(freeme); |
e2cfabdf WD |
502 | } |
503 | } | |
bb6ea430 | 504 | |
b25e6716 MF |
505 | static void seccomp_init_siginfo(siginfo_t *info, int syscall, int reason) |
506 | { | |
507 | memset(info, 0, sizeof(*info)); | |
508 | info->si_signo = SIGSYS; | |
509 | info->si_code = SYS_SECCOMP; | |
510 | info->si_call_addr = (void __user *)KSTK_EIP(current); | |
511 | info->si_errno = reason; | |
512 | info->si_arch = syscall_get_arch(); | |
513 | info->si_syscall = syscall; | |
514 | } | |
515 | ||
bb6ea430 WD |
516 | /** |
517 | * seccomp_send_sigsys - signals the task to allow in-process syscall emulation | |
518 | * @syscall: syscall number to send to userland | |
519 | * @reason: filter-supplied reason code to send to userland (via si_errno) | |
520 | * | |
521 | * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info. | |
522 | */ | |
523 | static void seccomp_send_sigsys(int syscall, int reason) | |
524 | { | |
525 | struct siginfo info; | |
b25e6716 | 526 | seccomp_init_siginfo(&info, syscall, reason); |
bb6ea430 WD |
527 | force_sig_info(SIGSYS, &info, current); |
528 | } | |
e2cfabdf | 529 | #endif /* CONFIG_SECCOMP_FILTER */ |
1da177e4 | 530 | |
56eb631b TH |
531 | /* For use with seccomp_actions_logged */ |
532 | #define SECCOMP_LOG_KILL (1 << 0) | |
533 | #define SECCOMP_LOG_TRAP (1 << 2) | |
534 | #define SECCOMP_LOG_ERRNO (1 << 3) | |
535 | #define SECCOMP_LOG_TRACE (1 << 4) | |
4267083e TH |
536 | #define SECCOMP_LOG_LOG (1 << 5) |
537 | #define SECCOMP_LOG_ALLOW (1 << 6) | |
56eb631b TH |
538 | |
539 | static u32 seccomp_actions_logged = SECCOMP_LOG_KILL | SECCOMP_LOG_TRAP | | |
4267083e TH |
540 | SECCOMP_LOG_ERRNO | SECCOMP_LOG_TRACE | |
541 | SECCOMP_LOG_LOG; | |
56eb631b | 542 | |
8ac8bbe1 TH |
543 | static inline void seccomp_log(unsigned long syscall, long signr, u32 action, |
544 | bool requested) | |
56eb631b TH |
545 | { |
546 | bool log = false; | |
547 | ||
548 | switch (action) { | |
549 | case SECCOMP_RET_ALLOW: | |
8ac8bbe1 | 550 | break; |
56eb631b | 551 | case SECCOMP_RET_TRAP: |
8ac8bbe1 TH |
552 | log = requested && seccomp_actions_logged & SECCOMP_LOG_TRAP; |
553 | break; | |
56eb631b | 554 | case SECCOMP_RET_ERRNO: |
8ac8bbe1 TH |
555 | log = requested && seccomp_actions_logged & SECCOMP_LOG_ERRNO; |
556 | break; | |
56eb631b | 557 | case SECCOMP_RET_TRACE: |
8ac8bbe1 | 558 | log = requested && seccomp_actions_logged & SECCOMP_LOG_TRACE; |
56eb631b | 559 | break; |
4267083e TH |
560 | case SECCOMP_RET_LOG: |
561 | log = seccomp_actions_logged & SECCOMP_LOG_LOG; | |
562 | break; | |
56eb631b TH |
563 | case SECCOMP_RET_KILL: |
564 | default: | |
565 | log = seccomp_actions_logged & SECCOMP_LOG_KILL; | |
566 | } | |
567 | ||
568 | /* | |
4267083e TH |
569 | * Force an audit message to be emitted when the action is RET_KILL, |
570 | * RET_LOG, or the FILTER_FLAG_LOG bit was set and the action is | |
571 | * allowed to be logged by the admin. | |
56eb631b TH |
572 | */ |
573 | if (log) | |
574 | return __audit_seccomp(syscall, signr, action); | |
575 | ||
576 | /* | |
577 | * Let the audit subsystem decide if the action should be audited based | |
578 | * on whether the current task itself is being audited. | |
579 | */ | |
580 | return audit_seccomp(syscall, signr, action); | |
581 | } | |
582 | ||
1da177e4 LT |
583 | /* |
584 | * Secure computing mode 1 allows only read/write/exit/sigreturn. | |
585 | * To be fully secure this must be combined with rlimit | |
586 | * to limit the stack allocations too. | |
587 | */ | |
cb4253aa | 588 | static const int mode1_syscalls[] = { |
1da177e4 LT |
589 | __NR_seccomp_read, __NR_seccomp_write, __NR_seccomp_exit, __NR_seccomp_sigreturn, |
590 | 0, /* null terminated */ | |
591 | }; | |
592 | ||
a4412fc9 | 593 | static void __secure_computing_strict(int this_syscall) |
1da177e4 | 594 | { |
cb4253aa | 595 | const int *syscall_whitelist = mode1_syscalls; |
a4412fc9 | 596 | #ifdef CONFIG_COMPAT |
5c38065e | 597 | if (in_compat_syscall()) |
c983f0e8 | 598 | syscall_whitelist = get_compat_mode1_syscalls(); |
a4412fc9 AL |
599 | #endif |
600 | do { | |
601 | if (*syscall_whitelist == this_syscall) | |
602 | return; | |
603 | } while (*++syscall_whitelist); | |
604 | ||
605 | #ifdef SECCOMP_DEBUG | |
606 | dump_stack(); | |
607 | #endif | |
8ac8bbe1 | 608 | seccomp_log(this_syscall, SIGKILL, SECCOMP_RET_KILL, true); |
a4412fc9 AL |
609 | do_exit(SIGKILL); |
610 | } | |
611 | ||
612 | #ifndef CONFIG_HAVE_ARCH_SECCOMP_FILTER | |
613 | void secure_computing_strict(int this_syscall) | |
614 | { | |
615 | int mode = current->seccomp.mode; | |
616 | ||
97f2645f | 617 | if (IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) && |
13c4a901 TA |
618 | unlikely(current->ptrace & PT_SUSPEND_SECCOMP)) |
619 | return; | |
620 | ||
221272f9 | 621 | if (mode == SECCOMP_MODE_DISABLED) |
a4412fc9 AL |
622 | return; |
623 | else if (mode == SECCOMP_MODE_STRICT) | |
624 | __secure_computing_strict(this_syscall); | |
625 | else | |
626 | BUG(); | |
627 | } | |
628 | #else | |
13aa72f0 AL |
629 | |
630 | #ifdef CONFIG_SECCOMP_FILTER | |
ce6526e8 KC |
631 | static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd, |
632 | const bool recheck_after_trace) | |
13aa72f0 AL |
633 | { |
634 | u32 filter_ret, action; | |
e95596f2 | 635 | struct seccomp_filter *match = NULL; |
13aa72f0 | 636 | int data; |
1da177e4 | 637 | |
3ba2530c KC |
638 | /* |
639 | * Make sure that any changes to mode from another thread have | |
640 | * been seen after TIF_SECCOMP was seen. | |
641 | */ | |
642 | rmb(); | |
643 | ||
e95596f2 | 644 | filter_ret = seccomp_run_filters(sd, &match); |
13aa72f0 AL |
645 | data = filter_ret & SECCOMP_RET_DATA; |
646 | action = filter_ret & SECCOMP_RET_ACTION; | |
647 | ||
648 | switch (action) { | |
649 | case SECCOMP_RET_ERRNO: | |
580c57f1 KC |
650 | /* Set low-order bits as an errno, capped at MAX_ERRNO. */ |
651 | if (data > MAX_ERRNO) | |
652 | data = MAX_ERRNO; | |
d39bd00d | 653 | syscall_set_return_value(current, task_pt_regs(current), |
13aa72f0 AL |
654 | -data, 0); |
655 | goto skip; | |
656 | ||
657 | case SECCOMP_RET_TRAP: | |
658 | /* Show the handler the original registers. */ | |
d39bd00d | 659 | syscall_rollback(current, task_pt_regs(current)); |
13aa72f0 AL |
660 | /* Let the filter pass back 16 bits of data. */ |
661 | seccomp_send_sigsys(this_syscall, data); | |
662 | goto skip; | |
663 | ||
664 | case SECCOMP_RET_TRACE: | |
ce6526e8 KC |
665 | /* We've been put in this state by the ptracer already. */ |
666 | if (recheck_after_trace) | |
667 | return 0; | |
668 | ||
8112c4f1 KC |
669 | /* ENOSYS these calls if there is no tracer attached. */ |
670 | if (!ptrace_event_enabled(current, PTRACE_EVENT_SECCOMP)) { | |
671 | syscall_set_return_value(current, | |
672 | task_pt_regs(current), | |
673 | -ENOSYS, 0); | |
674 | goto skip; | |
675 | } | |
676 | ||
677 | /* Allow the BPF to provide the event message */ | |
678 | ptrace_event(PTRACE_EVENT_SECCOMP, data); | |
679 | /* | |
680 | * The delivery of a fatal signal during event | |
485a252a KC |
681 | * notification may silently skip tracer notification, |
682 | * which could leave us with a potentially unmodified | |
683 | * syscall that the tracer would have liked to have | |
684 | * changed. Since the process is about to die, we just | |
685 | * force the syscall to be skipped and let the signal | |
686 | * kill the process and correctly handle any tracer exit | |
687 | * notifications. | |
8112c4f1 KC |
688 | */ |
689 | if (fatal_signal_pending(current)) | |
485a252a | 690 | goto skip; |
8112c4f1 KC |
691 | /* Check if the tracer forced the syscall to be skipped. */ |
692 | this_syscall = syscall_get_nr(current, task_pt_regs(current)); | |
693 | if (this_syscall < 0) | |
694 | goto skip; | |
695 | ||
ce6526e8 KC |
696 | /* |
697 | * Recheck the syscall, since it may have changed. This | |
698 | * intentionally uses a NULL struct seccomp_data to force | |
699 | * a reload of all registers. This does not goto skip since | |
700 | * a skip would have already been reported. | |
701 | */ | |
702 | if (__seccomp_filter(this_syscall, NULL, true)) | |
703 | return -1; | |
704 | ||
8112c4f1 | 705 | return 0; |
13aa72f0 | 706 | |
4267083e TH |
707 | case SECCOMP_RET_LOG: |
708 | seccomp_log(this_syscall, 0, action, true); | |
709 | return 0; | |
710 | ||
13aa72f0 | 711 | case SECCOMP_RET_ALLOW: |
e95596f2 KC |
712 | /* |
713 | * Note that the "match" filter will always be NULL for | |
714 | * this action since SECCOMP_RET_ALLOW is the starting | |
715 | * state in seccomp_run_filters(). | |
716 | */ | |
8112c4f1 | 717 | return 0; |
13aa72f0 AL |
718 | |
719 | case SECCOMP_RET_KILL: | |
131b6351 | 720 | default: |
8ac8bbe1 | 721 | seccomp_log(this_syscall, SIGSYS, action, true); |
d7276e32 KC |
722 | /* Dump core only if this is the last remaining thread. */ |
723 | if (get_nr_threads(current) == 1) { | |
131b6351 KC |
724 | siginfo_t info; |
725 | ||
d7276e32 KC |
726 | /* Show the original registers in the dump. */ |
727 | syscall_rollback(current, task_pt_regs(current)); | |
728 | /* Trigger a manual coredump since do_exit skips it. */ | |
729 | seccomp_init_siginfo(&info, this_syscall, data); | |
730 | do_coredump(&info); | |
731 | } | |
13aa72f0 AL |
732 | do_exit(SIGSYS); |
733 | } | |
734 | ||
735 | unreachable(); | |
736 | ||
737 | skip: | |
8ac8bbe1 | 738 | seccomp_log(this_syscall, 0, action, match ? match->log : false); |
8112c4f1 KC |
739 | return -1; |
740 | } | |
741 | #else | |
ce6526e8 KC |
742 | static int __seccomp_filter(int this_syscall, const struct seccomp_data *sd, |
743 | const bool recheck_after_trace) | |
8112c4f1 KC |
744 | { |
745 | BUG(); | |
13aa72f0 | 746 | } |
1da177e4 | 747 | #endif |
13aa72f0 | 748 | |
8112c4f1 | 749 | int __secure_computing(const struct seccomp_data *sd) |
13aa72f0 AL |
750 | { |
751 | int mode = current->seccomp.mode; | |
8112c4f1 | 752 | int this_syscall; |
13aa72f0 | 753 | |
97f2645f | 754 | if (IS_ENABLED(CONFIG_CHECKPOINT_RESTORE) && |
13c4a901 | 755 | unlikely(current->ptrace & PT_SUSPEND_SECCOMP)) |
8112c4f1 KC |
756 | return 0; |
757 | ||
758 | this_syscall = sd ? sd->nr : | |
759 | syscall_get_nr(current, task_pt_regs(current)); | |
13c4a901 | 760 | |
13aa72f0 | 761 | switch (mode) { |
e2cfabdf | 762 | case SECCOMP_MODE_STRICT: |
13aa72f0 | 763 | __secure_computing_strict(this_syscall); /* may call do_exit */ |
8112c4f1 | 764 | return 0; |
13aa72f0 | 765 | case SECCOMP_MODE_FILTER: |
ce6526e8 | 766 | return __seccomp_filter(this_syscall, sd, false); |
1da177e4 LT |
767 | default: |
768 | BUG(); | |
769 | } | |
13aa72f0 | 770 | } |
a4412fc9 | 771 | #endif /* CONFIG_HAVE_ARCH_SECCOMP_FILTER */ |
1d9d02fe AA |
772 | |
773 | long prctl_get_seccomp(void) | |
774 | { | |
775 | return current->seccomp.mode; | |
776 | } | |
777 | ||
e2cfabdf | 778 | /** |
3b23dd12 | 779 | * seccomp_set_mode_strict: internal function for setting strict seccomp |
e2cfabdf WD |
780 | * |
781 | * Once current->seccomp.mode is non-zero, it may not be changed. | |
782 | * | |
783 | * Returns 0 on success or -EINVAL on failure. | |
784 | */ | |
3b23dd12 | 785 | static long seccomp_set_mode_strict(void) |
1d9d02fe | 786 | { |
3b23dd12 | 787 | const unsigned long seccomp_mode = SECCOMP_MODE_STRICT; |
e2cfabdf | 788 | long ret = -EINVAL; |
1d9d02fe | 789 | |
dbd95212 KC |
790 | spin_lock_irq(¤t->sighand->siglock); |
791 | ||
1f41b450 | 792 | if (!seccomp_may_assign_mode(seccomp_mode)) |
1d9d02fe AA |
793 | goto out; |
794 | ||
cf99abac | 795 | #ifdef TIF_NOTSC |
3b23dd12 | 796 | disable_TSC(); |
cf99abac | 797 | #endif |
3ba2530c | 798 | seccomp_assign_mode(current, seccomp_mode); |
3b23dd12 KC |
799 | ret = 0; |
800 | ||
801 | out: | |
dbd95212 | 802 | spin_unlock_irq(¤t->sighand->siglock); |
3b23dd12 KC |
803 | |
804 | return ret; | |
805 | } | |
806 | ||
e2cfabdf | 807 | #ifdef CONFIG_SECCOMP_FILTER |
3b23dd12 KC |
808 | /** |
809 | * seccomp_set_mode_filter: internal function for setting seccomp filter | |
48dc92b9 | 810 | * @flags: flags to change filter behavior |
3b23dd12 KC |
811 | * @filter: struct sock_fprog containing filter |
812 | * | |
813 | * This function may be called repeatedly to install additional filters. | |
814 | * Every filter successfully installed will be evaluated (in reverse order) | |
815 | * for each system call the task makes. | |
816 | * | |
817 | * Once current->seccomp.mode is non-zero, it may not be changed. | |
818 | * | |
819 | * Returns 0 on success or -EINVAL on failure. | |
820 | */ | |
48dc92b9 KC |
821 | static long seccomp_set_mode_filter(unsigned int flags, |
822 | const char __user *filter) | |
3b23dd12 KC |
823 | { |
824 | const unsigned long seccomp_mode = SECCOMP_MODE_FILTER; | |
c8bee430 | 825 | struct seccomp_filter *prepared = NULL; |
3b23dd12 KC |
826 | long ret = -EINVAL; |
827 | ||
48dc92b9 | 828 | /* Validate flags. */ |
c2e1f2e3 | 829 | if (flags & ~SECCOMP_FILTER_FLAG_MASK) |
dbd95212 | 830 | return -EINVAL; |
48dc92b9 | 831 | |
c8bee430 KC |
832 | /* Prepare the new filter before holding any locks. */ |
833 | prepared = seccomp_prepare_user_filter(filter); | |
834 | if (IS_ERR(prepared)) | |
835 | return PTR_ERR(prepared); | |
836 | ||
c2e1f2e3 KC |
837 | /* |
838 | * Make sure we cannot change seccomp or nnp state via TSYNC | |
839 | * while another thread is in the middle of calling exec. | |
840 | */ | |
841 | if (flags & SECCOMP_FILTER_FLAG_TSYNC && | |
842 | mutex_lock_killable(¤t->signal->cred_guard_mutex)) | |
843 | goto out_free; | |
844 | ||
dbd95212 KC |
845 | spin_lock_irq(¤t->sighand->siglock); |
846 | ||
3b23dd12 KC |
847 | if (!seccomp_may_assign_mode(seccomp_mode)) |
848 | goto out; | |
849 | ||
c8bee430 | 850 | ret = seccomp_attach_filter(flags, prepared); |
3b23dd12 | 851 | if (ret) |
e2cfabdf | 852 | goto out; |
c8bee430 KC |
853 | /* Do not free the successfully attached filter. */ |
854 | prepared = NULL; | |
1d9d02fe | 855 | |
3ba2530c | 856 | seccomp_assign_mode(current, seccomp_mode); |
e2cfabdf | 857 | out: |
dbd95212 | 858 | spin_unlock_irq(¤t->sighand->siglock); |
c2e1f2e3 KC |
859 | if (flags & SECCOMP_FILTER_FLAG_TSYNC) |
860 | mutex_unlock(¤t->signal->cred_guard_mutex); | |
861 | out_free: | |
c8bee430 | 862 | seccomp_filter_free(prepared); |
1d9d02fe AA |
863 | return ret; |
864 | } | |
3b23dd12 | 865 | #else |
48dc92b9 KC |
866 | static inline long seccomp_set_mode_filter(unsigned int flags, |
867 | const char __user *filter) | |
3b23dd12 KC |
868 | { |
869 | return -EINVAL; | |
870 | } | |
871 | #endif | |
d78ab02c | 872 | |
4f65cecf TH |
873 | static long seccomp_get_action_avail(const char __user *uaction) |
874 | { | |
875 | u32 action; | |
876 | ||
877 | if (copy_from_user(&action, uaction, sizeof(action))) | |
878 | return -EFAULT; | |
879 | ||
880 | switch (action) { | |
881 | case SECCOMP_RET_KILL: | |
882 | case SECCOMP_RET_TRAP: | |
883 | case SECCOMP_RET_ERRNO: | |
884 | case SECCOMP_RET_TRACE: | |
4267083e | 885 | case SECCOMP_RET_LOG: |
4f65cecf TH |
886 | case SECCOMP_RET_ALLOW: |
887 | break; | |
888 | default: | |
889 | return -EOPNOTSUPP; | |
890 | } | |
891 | ||
892 | return 0; | |
893 | } | |
894 | ||
48dc92b9 KC |
895 | /* Common entry point for both prctl and syscall. */ |
896 | static long do_seccomp(unsigned int op, unsigned int flags, | |
897 | const char __user *uargs) | |
898 | { | |
899 | switch (op) { | |
900 | case SECCOMP_SET_MODE_STRICT: | |
901 | if (flags != 0 || uargs != NULL) | |
902 | return -EINVAL; | |
903 | return seccomp_set_mode_strict(); | |
904 | case SECCOMP_SET_MODE_FILTER: | |
905 | return seccomp_set_mode_filter(flags, uargs); | |
4f65cecf TH |
906 | case SECCOMP_GET_ACTION_AVAIL: |
907 | if (flags != 0) | |
908 | return -EINVAL; | |
909 | ||
910 | return seccomp_get_action_avail(uargs); | |
48dc92b9 KC |
911 | default: |
912 | return -EINVAL; | |
913 | } | |
914 | } | |
915 | ||
916 | SYSCALL_DEFINE3(seccomp, unsigned int, op, unsigned int, flags, | |
917 | const char __user *, uargs) | |
918 | { | |
919 | return do_seccomp(op, flags, uargs); | |
920 | } | |
921 | ||
d78ab02c KC |
922 | /** |
923 | * prctl_set_seccomp: configures current->seccomp.mode | |
924 | * @seccomp_mode: requested mode to use | |
925 | * @filter: optional struct sock_fprog for use with SECCOMP_MODE_FILTER | |
926 | * | |
927 | * Returns 0 on success or -EINVAL on failure. | |
928 | */ | |
929 | long prctl_set_seccomp(unsigned long seccomp_mode, char __user *filter) | |
930 | { | |
48dc92b9 KC |
931 | unsigned int op; |
932 | char __user *uargs; | |
933 | ||
3b23dd12 KC |
934 | switch (seccomp_mode) { |
935 | case SECCOMP_MODE_STRICT: | |
48dc92b9 KC |
936 | op = SECCOMP_SET_MODE_STRICT; |
937 | /* | |
938 | * Setting strict mode through prctl always ignored filter, | |
939 | * so make sure it is always NULL here to pass the internal | |
940 | * check in do_seccomp(). | |
941 | */ | |
942 | uargs = NULL; | |
943 | break; | |
3b23dd12 | 944 | case SECCOMP_MODE_FILTER: |
48dc92b9 KC |
945 | op = SECCOMP_SET_MODE_FILTER; |
946 | uargs = filter; | |
947 | break; | |
3b23dd12 KC |
948 | default: |
949 | return -EINVAL; | |
950 | } | |
48dc92b9 KC |
951 | |
952 | /* prctl interface doesn't have flags, so they are always zero. */ | |
953 | return do_seccomp(op, 0, uargs); | |
d78ab02c | 954 | } |
f8e529ed TA |
955 | |
956 | #if defined(CONFIG_SECCOMP_FILTER) && defined(CONFIG_CHECKPOINT_RESTORE) | |
957 | long seccomp_get_filter(struct task_struct *task, unsigned long filter_off, | |
958 | void __user *data) | |
959 | { | |
960 | struct seccomp_filter *filter; | |
961 | struct sock_fprog_kern *fprog; | |
962 | long ret; | |
963 | unsigned long count = 0; | |
964 | ||
965 | if (!capable(CAP_SYS_ADMIN) || | |
966 | current->seccomp.mode != SECCOMP_MODE_DISABLED) { | |
967 | return -EACCES; | |
968 | } | |
969 | ||
970 | spin_lock_irq(&task->sighand->siglock); | |
971 | if (task->seccomp.mode != SECCOMP_MODE_FILTER) { | |
972 | ret = -EINVAL; | |
973 | goto out; | |
974 | } | |
975 | ||
976 | filter = task->seccomp.filter; | |
977 | while (filter) { | |
978 | filter = filter->prev; | |
979 | count++; | |
980 | } | |
981 | ||
982 | if (filter_off >= count) { | |
983 | ret = -ENOENT; | |
984 | goto out; | |
985 | } | |
986 | count -= filter_off; | |
987 | ||
988 | filter = task->seccomp.filter; | |
989 | while (filter && count > 1) { | |
990 | filter = filter->prev; | |
991 | count--; | |
992 | } | |
993 | ||
994 | if (WARN_ON(count != 1 || !filter)) { | |
995 | /* The filter tree shouldn't shrink while we're using it. */ | |
996 | ret = -ENOENT; | |
997 | goto out; | |
998 | } | |
999 | ||
1000 | fprog = filter->prog->orig_prog; | |
1001 | if (!fprog) { | |
470bf1f2 | 1002 | /* This must be a new non-cBPF filter, since we save |
f8e529ed TA |
1003 | * every cBPF filter's orig_prog above when |
1004 | * CONFIG_CHECKPOINT_RESTORE is enabled. | |
1005 | */ | |
1006 | ret = -EMEDIUMTYPE; | |
1007 | goto out; | |
1008 | } | |
1009 | ||
1010 | ret = fprog->len; | |
1011 | if (!data) | |
1012 | goto out; | |
1013 | ||
1014 | get_seccomp_filter(task); | |
1015 | spin_unlock_irq(&task->sighand->siglock); | |
1016 | ||
1017 | if (copy_to_user(data, fprog->filter, bpf_classic_proglen(fprog))) | |
1018 | ret = -EFAULT; | |
1019 | ||
1020 | put_seccomp_filter(task); | |
1021 | return ret; | |
1022 | ||
1023 | out: | |
1024 | spin_unlock_irq(&task->sighand->siglock); | |
1025 | return ret; | |
1026 | } | |
1027 | #endif | |
9ca58825 TH |
1028 | |
1029 | #ifdef CONFIG_SYSCTL | |
1030 | ||
1031 | /* Human readable action names for friendly sysctl interaction */ | |
1032 | #define SECCOMP_RET_KILL_NAME "kill" | |
1033 | #define SECCOMP_RET_TRAP_NAME "trap" | |
1034 | #define SECCOMP_RET_ERRNO_NAME "errno" | |
1035 | #define SECCOMP_RET_TRACE_NAME "trace" | |
4267083e | 1036 | #define SECCOMP_RET_LOG_NAME "log" |
9ca58825 TH |
1037 | #define SECCOMP_RET_ALLOW_NAME "allow" |
1038 | ||
1039 | static const char seccomp_actions_avail[] = SECCOMP_RET_KILL_NAME " " | |
1040 | SECCOMP_RET_TRAP_NAME " " | |
1041 | SECCOMP_RET_ERRNO_NAME " " | |
1042 | SECCOMP_RET_TRACE_NAME " " | |
4267083e | 1043 | SECCOMP_RET_LOG_NAME " " |
9ca58825 TH |
1044 | SECCOMP_RET_ALLOW_NAME; |
1045 | ||
56eb631b TH |
1046 | struct seccomp_log_name { |
1047 | u32 log; | |
1048 | const char *name; | |
1049 | }; | |
1050 | ||
1051 | static const struct seccomp_log_name seccomp_log_names[] = { | |
1052 | { SECCOMP_LOG_KILL, SECCOMP_RET_KILL_NAME }, | |
1053 | { SECCOMP_LOG_TRAP, SECCOMP_RET_TRAP_NAME }, | |
1054 | { SECCOMP_LOG_ERRNO, SECCOMP_RET_ERRNO_NAME }, | |
1055 | { SECCOMP_LOG_TRACE, SECCOMP_RET_TRACE_NAME }, | |
4267083e | 1056 | { SECCOMP_LOG_LOG, SECCOMP_RET_LOG_NAME }, |
56eb631b TH |
1057 | { SECCOMP_LOG_ALLOW, SECCOMP_RET_ALLOW_NAME }, |
1058 | { } | |
1059 | }; | |
1060 | ||
1061 | static bool seccomp_names_from_actions_logged(char *names, size_t size, | |
1062 | u32 actions_logged) | |
1063 | { | |
1064 | const struct seccomp_log_name *cur; | |
1065 | bool append_space = false; | |
1066 | ||
1067 | for (cur = seccomp_log_names; cur->name && size; cur++) { | |
1068 | ssize_t ret; | |
1069 | ||
1070 | if (!(actions_logged & cur->log)) | |
1071 | continue; | |
1072 | ||
1073 | if (append_space) { | |
1074 | ret = strscpy(names, " ", size); | |
1075 | if (ret < 0) | |
1076 | return false; | |
1077 | ||
1078 | names += ret; | |
1079 | size -= ret; | |
1080 | } else | |
1081 | append_space = true; | |
1082 | ||
1083 | ret = strscpy(names, cur->name, size); | |
1084 | if (ret < 0) | |
1085 | return false; | |
1086 | ||
1087 | names += ret; | |
1088 | size -= ret; | |
1089 | } | |
1090 | ||
1091 | return true; | |
1092 | } | |
1093 | ||
1094 | static bool seccomp_action_logged_from_name(u32 *action_logged, | |
1095 | const char *name) | |
1096 | { | |
1097 | const struct seccomp_log_name *cur; | |
1098 | ||
1099 | for (cur = seccomp_log_names; cur->name; cur++) { | |
1100 | if (!strcmp(cur->name, name)) { | |
1101 | *action_logged = cur->log; | |
1102 | return true; | |
1103 | } | |
1104 | } | |
1105 | ||
1106 | return false; | |
1107 | } | |
1108 | ||
1109 | static bool seccomp_actions_logged_from_names(u32 *actions_logged, char *names) | |
1110 | { | |
1111 | char *name; | |
1112 | ||
1113 | *actions_logged = 0; | |
1114 | while ((name = strsep(&names, " ")) && *name) { | |
1115 | u32 action_logged = 0; | |
1116 | ||
1117 | if (!seccomp_action_logged_from_name(&action_logged, name)) | |
1118 | return false; | |
1119 | ||
1120 | *actions_logged |= action_logged; | |
1121 | } | |
1122 | ||
1123 | return true; | |
1124 | } | |
1125 | ||
1126 | static int seccomp_actions_logged_handler(struct ctl_table *ro_table, int write, | |
1127 | void __user *buffer, size_t *lenp, | |
1128 | loff_t *ppos) | |
1129 | { | |
1130 | char names[sizeof(seccomp_actions_avail)]; | |
1131 | struct ctl_table table; | |
1132 | int ret; | |
1133 | ||
1134 | if (write && !capable(CAP_SYS_ADMIN)) | |
1135 | return -EPERM; | |
1136 | ||
1137 | memset(names, 0, sizeof(names)); | |
1138 | ||
1139 | if (!write) { | |
1140 | if (!seccomp_names_from_actions_logged(names, sizeof(names), | |
1141 | seccomp_actions_logged)) | |
1142 | return -EINVAL; | |
1143 | } | |
1144 | ||
1145 | table = *ro_table; | |
1146 | table.data = names; | |
1147 | table.maxlen = sizeof(names); | |
1148 | ret = proc_dostring(&table, write, buffer, lenp, ppos); | |
1149 | if (ret) | |
1150 | return ret; | |
1151 | ||
1152 | if (write) { | |
1153 | u32 actions_logged; | |
1154 | ||
1155 | if (!seccomp_actions_logged_from_names(&actions_logged, | |
1156 | table.data)) | |
1157 | return -EINVAL; | |
1158 | ||
1159 | if (actions_logged & SECCOMP_LOG_ALLOW) | |
1160 | return -EINVAL; | |
1161 | ||
1162 | seccomp_actions_logged = actions_logged; | |
1163 | } | |
1164 | ||
1165 | return 0; | |
1166 | } | |
1167 | ||
9ca58825 TH |
1168 | static struct ctl_path seccomp_sysctl_path[] = { |
1169 | { .procname = "kernel", }, | |
1170 | { .procname = "seccomp", }, | |
1171 | { } | |
1172 | }; | |
1173 | ||
1174 | static struct ctl_table seccomp_sysctl_table[] = { | |
1175 | { | |
1176 | .procname = "actions_avail", | |
1177 | .data = (void *) &seccomp_actions_avail, | |
1178 | .maxlen = sizeof(seccomp_actions_avail), | |
1179 | .mode = 0444, | |
1180 | .proc_handler = proc_dostring, | |
1181 | }, | |
56eb631b TH |
1182 | { |
1183 | .procname = "actions_logged", | |
1184 | .mode = 0644, | |
1185 | .proc_handler = seccomp_actions_logged_handler, | |
1186 | }, | |
9ca58825 TH |
1187 | { } |
1188 | }; | |
1189 | ||
1190 | static int __init seccomp_sysctl_init(void) | |
1191 | { | |
1192 | struct ctl_table_header *hdr; | |
1193 | ||
1194 | hdr = register_sysctl_paths(seccomp_sysctl_path, seccomp_sysctl_table); | |
1195 | if (!hdr) | |
1196 | pr_warn("seccomp: sysctl registration failed\n"); | |
1197 | else | |
1198 | kmemleak_not_leak(hdr); | |
1199 | ||
1200 | return 0; | |
1201 | } | |
1202 | ||
1203 | device_initcall(seccomp_sysctl_init) | |
1204 | ||
1205 | #endif /* CONFIG_SYSCTL */ |