| 1 | /* |
| 2 | * linux/kernel/seccomp.c |
| 3 | * |
| 4 | * Copyright 2004-2005 Andrea Arcangeli <andrea@cpushare.com> |
| 5 | * |
| 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. |
| 14 | */ |
| 15 | |
| 16 | #include <linux/atomic.h> |
| 17 | #include <linux/audit.h> |
| 18 | #include <linux/compat.h> |
| 19 | #include <linux/sched.h> |
| 20 | #include <linux/seccomp.h> |
| 21 | #include <linux/slab.h> |
| 22 | #include <linux/syscalls.h> |
| 23 | |
| 24 | /* #define SECCOMP_DEBUG 1 */ |
| 25 | |
| 26 | #ifdef CONFIG_SECCOMP_FILTER |
| 27 | #include <asm/syscall.h> |
| 28 | #include <linux/filter.h> |
| 29 | #include <linux/ptrace.h> |
| 30 | #include <linux/security.h> |
| 31 | #include <linux/tracehook.h> |
| 32 | #include <linux/uaccess.h> |
| 33 | |
| 34 | /** |
| 35 | * struct seccomp_filter - container for seccomp BPF programs |
| 36 | * |
| 37 | * @usage: reference count to manage the object lifetime. |
| 38 | * get/put helpers should be used when accessing an instance |
| 39 | * outside of a lifetime-guarded section. In general, this |
| 40 | * is only needed for handling filters shared across tasks. |
| 41 | * @prev: points to a previously installed, or inherited, filter |
| 42 | * @len: the number of instructions in the program |
| 43 | * @insnsi: the BPF program instructions to evaluate |
| 44 | * |
| 45 | * seccomp_filter objects are organized in a tree linked via the @prev |
| 46 | * pointer. For any task, it appears to be a singly-linked list starting |
| 47 | * with current->seccomp.filter, the most recently attached or inherited filter. |
| 48 | * However, multiple filters may share a @prev node, by way of fork(), which |
| 49 | * results in a unidirectional tree existing in memory. This is similar to |
| 50 | * how namespaces work. |
| 51 | * |
| 52 | * seccomp_filter objects should never be modified after being attached |
| 53 | * to a task_struct (other than @usage). |
| 54 | */ |
| 55 | struct seccomp_filter { |
| 56 | atomic_t usage; |
| 57 | struct seccomp_filter *prev; |
| 58 | struct sk_filter *prog; |
| 59 | }; |
| 60 | |
| 61 | /* Limit any path through the tree to 256KB worth of instructions. */ |
| 62 | #define MAX_INSNS_PER_PATH ((1 << 18) / sizeof(struct sock_filter)) |
| 63 | |
| 64 | /* |
| 65 | * Endianness is explicitly ignored and left for BPF program authors to manage |
| 66 | * as per the specific architecture. |
| 67 | */ |
| 68 | static void populate_seccomp_data(struct seccomp_data *sd) |
| 69 | { |
| 70 | struct task_struct *task = current; |
| 71 | struct pt_regs *regs = task_pt_regs(task); |
| 72 | unsigned long args[6]; |
| 73 | |
| 74 | sd->nr = syscall_get_nr(task, regs); |
| 75 | sd->arch = syscall_get_arch(); |
| 76 | syscall_get_arguments(task, regs, 0, 6, args); |
| 77 | sd->args[0] = args[0]; |
| 78 | sd->args[1] = args[1]; |
| 79 | sd->args[2] = args[2]; |
| 80 | sd->args[3] = args[3]; |
| 81 | sd->args[4] = args[4]; |
| 82 | sd->args[5] = args[5]; |
| 83 | sd->instruction_pointer = KSTK_EIP(task); |
| 84 | } |
| 85 | |
| 86 | /** |
| 87 | * seccomp_check_filter - verify seccomp filter code |
| 88 | * @filter: filter to verify |
| 89 | * @flen: length of filter |
| 90 | * |
| 91 | * Takes a previously checked filter (by sk_chk_filter) and |
| 92 | * redirects all filter code that loads struct sk_buff data |
| 93 | * and related data through seccomp_bpf_load. It also |
| 94 | * enforces length and alignment checking of those loads. |
| 95 | * |
| 96 | * Returns 0 if the rule set is legal or -EINVAL if not. |
| 97 | */ |
| 98 | static int seccomp_check_filter(struct sock_filter *filter, unsigned int flen) |
| 99 | { |
| 100 | int pc; |
| 101 | for (pc = 0; pc < flen; pc++) { |
| 102 | struct sock_filter *ftest = &filter[pc]; |
| 103 | u16 code = ftest->code; |
| 104 | u32 k = ftest->k; |
| 105 | |
| 106 | switch (code) { |
| 107 | case BPF_LD | BPF_W | BPF_ABS: |
| 108 | ftest->code = BPF_LDX | BPF_W | BPF_ABS; |
| 109 | /* 32-bit aligned and not out of bounds. */ |
| 110 | if (k >= sizeof(struct seccomp_data) || k & 3) |
| 111 | return -EINVAL; |
| 112 | continue; |
| 113 | case BPF_LD | BPF_W | BPF_LEN: |
| 114 | ftest->code = BPF_LD | BPF_IMM; |
| 115 | ftest->k = sizeof(struct seccomp_data); |
| 116 | continue; |
| 117 | case BPF_LDX | BPF_W | BPF_LEN: |
| 118 | ftest->code = BPF_LDX | BPF_IMM; |
| 119 | ftest->k = sizeof(struct seccomp_data); |
| 120 | continue; |
| 121 | /* Explicitly include allowed calls. */ |
| 122 | case BPF_RET | BPF_K: |
| 123 | case BPF_RET | BPF_A: |
| 124 | case BPF_ALU | BPF_ADD | BPF_K: |
| 125 | case BPF_ALU | BPF_ADD | BPF_X: |
| 126 | case BPF_ALU | BPF_SUB | BPF_K: |
| 127 | case BPF_ALU | BPF_SUB | BPF_X: |
| 128 | case BPF_ALU | BPF_MUL | BPF_K: |
| 129 | case BPF_ALU | BPF_MUL | BPF_X: |
| 130 | case BPF_ALU | BPF_DIV | BPF_K: |
| 131 | case BPF_ALU | BPF_DIV | BPF_X: |
| 132 | case BPF_ALU | BPF_AND | BPF_K: |
| 133 | case BPF_ALU | BPF_AND | BPF_X: |
| 134 | case BPF_ALU | BPF_OR | BPF_K: |
| 135 | case BPF_ALU | BPF_OR | BPF_X: |
| 136 | case BPF_ALU | BPF_XOR | BPF_K: |
| 137 | case BPF_ALU | BPF_XOR | BPF_X: |
| 138 | case BPF_ALU | BPF_LSH | BPF_K: |
| 139 | case BPF_ALU | BPF_LSH | BPF_X: |
| 140 | case BPF_ALU | BPF_RSH | BPF_K: |
| 141 | case BPF_ALU | BPF_RSH | BPF_X: |
| 142 | case BPF_ALU | BPF_NEG: |
| 143 | case BPF_LD | BPF_IMM: |
| 144 | case BPF_LDX | BPF_IMM: |
| 145 | case BPF_MISC | BPF_TAX: |
| 146 | case BPF_MISC | BPF_TXA: |
| 147 | case BPF_LD | BPF_MEM: |
| 148 | case BPF_LDX | BPF_MEM: |
| 149 | case BPF_ST: |
| 150 | case BPF_STX: |
| 151 | case BPF_JMP | BPF_JA: |
| 152 | case BPF_JMP | BPF_JEQ | BPF_K: |
| 153 | case BPF_JMP | BPF_JEQ | BPF_X: |
| 154 | case BPF_JMP | BPF_JGE | BPF_K: |
| 155 | case BPF_JMP | BPF_JGE | BPF_X: |
| 156 | case BPF_JMP | BPF_JGT | BPF_K: |
| 157 | case BPF_JMP | BPF_JGT | BPF_X: |
| 158 | case BPF_JMP | BPF_JSET | BPF_K: |
| 159 | case BPF_JMP | BPF_JSET | BPF_X: |
| 160 | continue; |
| 161 | default: |
| 162 | return -EINVAL; |
| 163 | } |
| 164 | } |
| 165 | return 0; |
| 166 | } |
| 167 | |
| 168 | /** |
| 169 | * seccomp_run_filters - evaluates all seccomp filters against @syscall |
| 170 | * @syscall: number of the current system call |
| 171 | * |
| 172 | * Returns valid seccomp BPF response codes. |
| 173 | */ |
| 174 | static u32 seccomp_run_filters(int syscall) |
| 175 | { |
| 176 | struct seccomp_filter *f = ACCESS_ONCE(current->seccomp.filter); |
| 177 | struct seccomp_data sd; |
| 178 | u32 ret = SECCOMP_RET_ALLOW; |
| 179 | |
| 180 | /* Ensure unexpected behavior doesn't result in failing open. */ |
| 181 | if (unlikely(WARN_ON(f == NULL))) |
| 182 | return SECCOMP_RET_KILL; |
| 183 | |
| 184 | /* Make sure cross-thread synced filter points somewhere sane. */ |
| 185 | smp_read_barrier_depends(); |
| 186 | |
| 187 | populate_seccomp_data(&sd); |
| 188 | |
| 189 | /* |
| 190 | * All filters in the list are evaluated and the lowest BPF return |
| 191 | * value always takes priority (ignoring the DATA). |
| 192 | */ |
| 193 | for (; f; f = f->prev) { |
| 194 | u32 cur_ret = SK_RUN_FILTER(f->prog, (void *)&sd); |
| 195 | |
| 196 | if ((cur_ret & SECCOMP_RET_ACTION) < (ret & SECCOMP_RET_ACTION)) |
| 197 | ret = cur_ret; |
| 198 | } |
| 199 | return ret; |
| 200 | } |
| 201 | #endif /* CONFIG_SECCOMP_FILTER */ |
| 202 | |
| 203 | static inline bool seccomp_may_assign_mode(unsigned long seccomp_mode) |
| 204 | { |
| 205 | BUG_ON(!spin_is_locked(¤t->sighand->siglock)); |
| 206 | |
| 207 | if (current->seccomp.mode && current->seccomp.mode != seccomp_mode) |
| 208 | return false; |
| 209 | |
| 210 | return true; |
| 211 | } |
| 212 | |
| 213 | static inline void seccomp_assign_mode(struct task_struct *task, |
| 214 | unsigned long seccomp_mode) |
| 215 | { |
| 216 | BUG_ON(!spin_is_locked(&task->sighand->siglock)); |
| 217 | |
| 218 | task->seccomp.mode = seccomp_mode; |
| 219 | /* |
| 220 | * Make sure TIF_SECCOMP cannot be set before the mode (and |
| 221 | * filter) is set. |
| 222 | */ |
| 223 | smp_mb__before_atomic(); |
| 224 | set_tsk_thread_flag(task, TIF_SECCOMP); |
| 225 | } |
| 226 | |
| 227 | #ifdef CONFIG_SECCOMP_FILTER |
| 228 | /** |
| 229 | * seccomp_prepare_filter: Prepares a seccomp filter for use. |
| 230 | * @fprog: BPF program to install |
| 231 | * |
| 232 | * Returns filter on success or an ERR_PTR on failure. |
| 233 | */ |
| 234 | static struct seccomp_filter *seccomp_prepare_filter(struct sock_fprog *fprog) |
| 235 | { |
| 236 | struct seccomp_filter *filter; |
| 237 | unsigned long fp_size; |
| 238 | struct sock_filter *fp; |
| 239 | int new_len; |
| 240 | long ret; |
| 241 | |
| 242 | if (fprog->len == 0 || fprog->len > BPF_MAXINSNS) |
| 243 | return ERR_PTR(-EINVAL); |
| 244 | BUG_ON(INT_MAX / fprog->len < sizeof(struct sock_filter)); |
| 245 | fp_size = fprog->len * sizeof(struct sock_filter); |
| 246 | |
| 247 | /* |
| 248 | * Installing a seccomp filter requires that the task has |
| 249 | * CAP_SYS_ADMIN in its namespace or be running with no_new_privs. |
| 250 | * This avoids scenarios where unprivileged tasks can affect the |
| 251 | * behavior of privileged children. |
| 252 | */ |
| 253 | if (!task_no_new_privs(current) && |
| 254 | security_capable_noaudit(current_cred(), current_user_ns(), |
| 255 | CAP_SYS_ADMIN) != 0) |
| 256 | return ERR_PTR(-EACCES); |
| 257 | |
| 258 | fp = kzalloc(fp_size, GFP_KERNEL|__GFP_NOWARN); |
| 259 | if (!fp) |
| 260 | return ERR_PTR(-ENOMEM); |
| 261 | |
| 262 | /* Copy the instructions from fprog. */ |
| 263 | ret = -EFAULT; |
| 264 | if (copy_from_user(fp, fprog->filter, fp_size)) |
| 265 | goto free_prog; |
| 266 | |
| 267 | /* Check and rewrite the fprog via the skb checker */ |
| 268 | ret = sk_chk_filter(fp, fprog->len); |
| 269 | if (ret) |
| 270 | goto free_prog; |
| 271 | |
| 272 | /* Check and rewrite the fprog for seccomp use */ |
| 273 | ret = seccomp_check_filter(fp, fprog->len); |
| 274 | if (ret) |
| 275 | goto free_prog; |
| 276 | |
| 277 | /* Convert 'sock_filter' insns to 'sock_filter_int' insns */ |
| 278 | ret = sk_convert_filter(fp, fprog->len, NULL, &new_len); |
| 279 | if (ret) |
| 280 | goto free_prog; |
| 281 | |
| 282 | /* Allocate a new seccomp_filter */ |
| 283 | ret = -ENOMEM; |
| 284 | filter = kzalloc(sizeof(struct seccomp_filter), |
| 285 | GFP_KERNEL|__GFP_NOWARN); |
| 286 | if (!filter) |
| 287 | goto free_prog; |
| 288 | |
| 289 | filter->prog = kzalloc(sk_filter_size(new_len), |
| 290 | GFP_KERNEL|__GFP_NOWARN); |
| 291 | if (!filter->prog) |
| 292 | goto free_filter; |
| 293 | |
| 294 | ret = sk_convert_filter(fp, fprog->len, filter->prog->insnsi, &new_len); |
| 295 | if (ret) |
| 296 | goto free_filter_prog; |
| 297 | kfree(fp); |
| 298 | |
| 299 | atomic_set(&filter->usage, 1); |
| 300 | filter->prog->len = new_len; |
| 301 | |
| 302 | sk_filter_select_runtime(filter->prog); |
| 303 | |
| 304 | return filter; |
| 305 | |
| 306 | free_filter_prog: |
| 307 | kfree(filter->prog); |
| 308 | free_filter: |
| 309 | kfree(filter); |
| 310 | free_prog: |
| 311 | kfree(fp); |
| 312 | return ERR_PTR(ret); |
| 313 | } |
| 314 | |
| 315 | /** |
| 316 | * seccomp_prepare_user_filter - prepares a user-supplied sock_fprog |
| 317 | * @user_filter: pointer to the user data containing a sock_fprog. |
| 318 | * |
| 319 | * Returns 0 on success and non-zero otherwise. |
| 320 | */ |
| 321 | static struct seccomp_filter * |
| 322 | seccomp_prepare_user_filter(const char __user *user_filter) |
| 323 | { |
| 324 | struct sock_fprog fprog; |
| 325 | struct seccomp_filter *filter = ERR_PTR(-EFAULT); |
| 326 | |
| 327 | #ifdef CONFIG_COMPAT |
| 328 | if (is_compat_task()) { |
| 329 | struct compat_sock_fprog fprog32; |
| 330 | if (copy_from_user(&fprog32, user_filter, sizeof(fprog32))) |
| 331 | goto out; |
| 332 | fprog.len = fprog32.len; |
| 333 | fprog.filter = compat_ptr(fprog32.filter); |
| 334 | } else /* falls through to the if below. */ |
| 335 | #endif |
| 336 | if (copy_from_user(&fprog, user_filter, sizeof(fprog))) |
| 337 | goto out; |
| 338 | filter = seccomp_prepare_filter(&fprog); |
| 339 | out: |
| 340 | return filter; |
| 341 | } |
| 342 | |
| 343 | /** |
| 344 | * seccomp_attach_filter: validate and attach filter |
| 345 | * @flags: flags to change filter behavior |
| 346 | * @filter: seccomp filter to add to the current process |
| 347 | * |
| 348 | * Caller must be holding current->sighand->siglock lock. |
| 349 | * |
| 350 | * Returns 0 on success, -ve on error. |
| 351 | */ |
| 352 | static long seccomp_attach_filter(unsigned int flags, |
| 353 | struct seccomp_filter *filter) |
| 354 | { |
| 355 | unsigned long total_insns; |
| 356 | struct seccomp_filter *walker; |
| 357 | |
| 358 | BUG_ON(!spin_is_locked(¤t->sighand->siglock)); |
| 359 | |
| 360 | /* Validate resulting filter length. */ |
| 361 | total_insns = filter->prog->len; |
| 362 | for (walker = current->seccomp.filter; walker; walker = walker->prev) |
| 363 | total_insns += walker->prog->len + 4; /* 4 instr penalty */ |
| 364 | if (total_insns > MAX_INSNS_PER_PATH) |
| 365 | return -ENOMEM; |
| 366 | |
| 367 | /* |
| 368 | * If there is an existing filter, make it the prev and don't drop its |
| 369 | * task reference. |
| 370 | */ |
| 371 | filter->prev = current->seccomp.filter; |
| 372 | current->seccomp.filter = filter; |
| 373 | |
| 374 | return 0; |
| 375 | } |
| 376 | |
| 377 | /* get_seccomp_filter - increments the reference count of the filter on @tsk */ |
| 378 | void get_seccomp_filter(struct task_struct *tsk) |
| 379 | { |
| 380 | struct seccomp_filter *orig = tsk->seccomp.filter; |
| 381 | if (!orig) |
| 382 | return; |
| 383 | /* Reference count is bounded by the number of total processes. */ |
| 384 | atomic_inc(&orig->usage); |
| 385 | } |
| 386 | |
| 387 | static inline void seccomp_filter_free(struct seccomp_filter *filter) |
| 388 | { |
| 389 | if (filter) { |
| 390 | sk_filter_free(filter->prog); |
| 391 | kfree(filter); |
| 392 | } |
| 393 | } |
| 394 | |
| 395 | /* put_seccomp_filter - decrements the ref count of tsk->seccomp.filter */ |
| 396 | void put_seccomp_filter(struct task_struct *tsk) |
| 397 | { |
| 398 | struct seccomp_filter *orig = tsk->seccomp.filter; |
| 399 | /* Clean up single-reference branches iteratively. */ |
| 400 | while (orig && atomic_dec_and_test(&orig->usage)) { |
| 401 | struct seccomp_filter *freeme = orig; |
| 402 | orig = orig->prev; |
| 403 | seccomp_filter_free(freeme); |
| 404 | } |
| 405 | } |
| 406 | |
| 407 | /** |
| 408 | * seccomp_send_sigsys - signals the task to allow in-process syscall emulation |
| 409 | * @syscall: syscall number to send to userland |
| 410 | * @reason: filter-supplied reason code to send to userland (via si_errno) |
| 411 | * |
| 412 | * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info. |
| 413 | */ |
| 414 | static void seccomp_send_sigsys(int syscall, int reason) |
| 415 | { |
| 416 | struct siginfo info; |
| 417 | memset(&info, 0, sizeof(info)); |
| 418 | info.si_signo = SIGSYS; |
| 419 | info.si_code = SYS_SECCOMP; |
| 420 | info.si_call_addr = (void __user *)KSTK_EIP(current); |
| 421 | info.si_errno = reason; |
| 422 | info.si_arch = syscall_get_arch(); |
| 423 | info.si_syscall = syscall; |
| 424 | force_sig_info(SIGSYS, &info, current); |
| 425 | } |
| 426 | #endif /* CONFIG_SECCOMP_FILTER */ |
| 427 | |
| 428 | /* |
| 429 | * Secure computing mode 1 allows only read/write/exit/sigreturn. |
| 430 | * To be fully secure this must be combined with rlimit |
| 431 | * to limit the stack allocations too. |
| 432 | */ |
| 433 | static int mode1_syscalls[] = { |
| 434 | __NR_seccomp_read, __NR_seccomp_write, __NR_seccomp_exit, __NR_seccomp_sigreturn, |
| 435 | 0, /* null terminated */ |
| 436 | }; |
| 437 | |
| 438 | #ifdef CONFIG_COMPAT |
| 439 | static int mode1_syscalls_32[] = { |
| 440 | __NR_seccomp_read_32, __NR_seccomp_write_32, __NR_seccomp_exit_32, __NR_seccomp_sigreturn_32, |
| 441 | 0, /* null terminated */ |
| 442 | }; |
| 443 | #endif |
| 444 | |
| 445 | int __secure_computing(int this_syscall) |
| 446 | { |
| 447 | int exit_sig = 0; |
| 448 | int *syscall; |
| 449 | u32 ret; |
| 450 | |
| 451 | /* |
| 452 | * Make sure that any changes to mode from another thread have |
| 453 | * been seen after TIF_SECCOMP was seen. |
| 454 | */ |
| 455 | rmb(); |
| 456 | |
| 457 | switch (current->seccomp.mode) { |
| 458 | case SECCOMP_MODE_STRICT: |
| 459 | syscall = mode1_syscalls; |
| 460 | #ifdef CONFIG_COMPAT |
| 461 | if (is_compat_task()) |
| 462 | syscall = mode1_syscalls_32; |
| 463 | #endif |
| 464 | do { |
| 465 | if (*syscall == this_syscall) |
| 466 | return 0; |
| 467 | } while (*++syscall); |
| 468 | exit_sig = SIGKILL; |
| 469 | ret = SECCOMP_RET_KILL; |
| 470 | break; |
| 471 | #ifdef CONFIG_SECCOMP_FILTER |
| 472 | case SECCOMP_MODE_FILTER: { |
| 473 | int data; |
| 474 | struct pt_regs *regs = task_pt_regs(current); |
| 475 | ret = seccomp_run_filters(this_syscall); |
| 476 | data = ret & SECCOMP_RET_DATA; |
| 477 | ret &= SECCOMP_RET_ACTION; |
| 478 | switch (ret) { |
| 479 | case SECCOMP_RET_ERRNO: |
| 480 | /* Set the low-order 16-bits as a errno. */ |
| 481 | syscall_set_return_value(current, regs, |
| 482 | -data, 0); |
| 483 | goto skip; |
| 484 | case SECCOMP_RET_TRAP: |
| 485 | /* Show the handler the original registers. */ |
| 486 | syscall_rollback(current, regs); |
| 487 | /* Let the filter pass back 16 bits of data. */ |
| 488 | seccomp_send_sigsys(this_syscall, data); |
| 489 | goto skip; |
| 490 | case SECCOMP_RET_TRACE: |
| 491 | /* Skip these calls if there is no tracer. */ |
| 492 | if (!ptrace_event_enabled(current, PTRACE_EVENT_SECCOMP)) { |
| 493 | syscall_set_return_value(current, regs, |
| 494 | -ENOSYS, 0); |
| 495 | goto skip; |
| 496 | } |
| 497 | /* Allow the BPF to provide the event message */ |
| 498 | ptrace_event(PTRACE_EVENT_SECCOMP, data); |
| 499 | /* |
| 500 | * The delivery of a fatal signal during event |
| 501 | * notification may silently skip tracer notification. |
| 502 | * Terminating the task now avoids executing a system |
| 503 | * call that may not be intended. |
| 504 | */ |
| 505 | if (fatal_signal_pending(current)) |
| 506 | break; |
| 507 | if (syscall_get_nr(current, regs) < 0) |
| 508 | goto skip; /* Explicit request to skip. */ |
| 509 | |
| 510 | return 0; |
| 511 | case SECCOMP_RET_ALLOW: |
| 512 | return 0; |
| 513 | case SECCOMP_RET_KILL: |
| 514 | default: |
| 515 | break; |
| 516 | } |
| 517 | exit_sig = SIGSYS; |
| 518 | break; |
| 519 | } |
| 520 | #endif |
| 521 | default: |
| 522 | BUG(); |
| 523 | } |
| 524 | |
| 525 | #ifdef SECCOMP_DEBUG |
| 526 | dump_stack(); |
| 527 | #endif |
| 528 | audit_seccomp(this_syscall, exit_sig, ret); |
| 529 | do_exit(exit_sig); |
| 530 | #ifdef CONFIG_SECCOMP_FILTER |
| 531 | skip: |
| 532 | audit_seccomp(this_syscall, exit_sig, ret); |
| 533 | #endif |
| 534 | return -1; |
| 535 | } |
| 536 | |
| 537 | long prctl_get_seccomp(void) |
| 538 | { |
| 539 | return current->seccomp.mode; |
| 540 | } |
| 541 | |
| 542 | /** |
| 543 | * seccomp_set_mode_strict: internal function for setting strict seccomp |
| 544 | * |
| 545 | * Once current->seccomp.mode is non-zero, it may not be changed. |
| 546 | * |
| 547 | * Returns 0 on success or -EINVAL on failure. |
| 548 | */ |
| 549 | static long seccomp_set_mode_strict(void) |
| 550 | { |
| 551 | const unsigned long seccomp_mode = SECCOMP_MODE_STRICT; |
| 552 | long ret = -EINVAL; |
| 553 | |
| 554 | spin_lock_irq(¤t->sighand->siglock); |
| 555 | |
| 556 | if (!seccomp_may_assign_mode(seccomp_mode)) |
| 557 | goto out; |
| 558 | |
| 559 | #ifdef TIF_NOTSC |
| 560 | disable_TSC(); |
| 561 | #endif |
| 562 | seccomp_assign_mode(current, seccomp_mode); |
| 563 | ret = 0; |
| 564 | |
| 565 | out: |
| 566 | spin_unlock_irq(¤t->sighand->siglock); |
| 567 | |
| 568 | return ret; |
| 569 | } |
| 570 | |
| 571 | #ifdef CONFIG_SECCOMP_FILTER |
| 572 | /** |
| 573 | * seccomp_set_mode_filter: internal function for setting seccomp filter |
| 574 | * @flags: flags to change filter behavior |
| 575 | * @filter: struct sock_fprog containing filter |
| 576 | * |
| 577 | * This function may be called repeatedly to install additional filters. |
| 578 | * Every filter successfully installed will be evaluated (in reverse order) |
| 579 | * for each system call the task makes. |
| 580 | * |
| 581 | * Once current->seccomp.mode is non-zero, it may not be changed. |
| 582 | * |
| 583 | * Returns 0 on success or -EINVAL on failure. |
| 584 | */ |
| 585 | static long seccomp_set_mode_filter(unsigned int flags, |
| 586 | const char __user *filter) |
| 587 | { |
| 588 | const unsigned long seccomp_mode = SECCOMP_MODE_FILTER; |
| 589 | struct seccomp_filter *prepared = NULL; |
| 590 | long ret = -EINVAL; |
| 591 | |
| 592 | /* Validate flags. */ |
| 593 | if (flags != 0) |
| 594 | return -EINVAL; |
| 595 | |
| 596 | /* Prepare the new filter before holding any locks. */ |
| 597 | prepared = seccomp_prepare_user_filter(filter); |
| 598 | if (IS_ERR(prepared)) |
| 599 | return PTR_ERR(prepared); |
| 600 | |
| 601 | spin_lock_irq(¤t->sighand->siglock); |
| 602 | |
| 603 | if (!seccomp_may_assign_mode(seccomp_mode)) |
| 604 | goto out; |
| 605 | |
| 606 | ret = seccomp_attach_filter(flags, prepared); |
| 607 | if (ret) |
| 608 | goto out; |
| 609 | /* Do not free the successfully attached filter. */ |
| 610 | prepared = NULL; |
| 611 | |
| 612 | seccomp_assign_mode(current, seccomp_mode); |
| 613 | out: |
| 614 | spin_unlock_irq(¤t->sighand->siglock); |
| 615 | seccomp_filter_free(prepared); |
| 616 | return ret; |
| 617 | } |
| 618 | #else |
| 619 | static inline long seccomp_set_mode_filter(unsigned int flags, |
| 620 | const char __user *filter) |
| 621 | { |
| 622 | return -EINVAL; |
| 623 | } |
| 624 | #endif |
| 625 | |
| 626 | /* Common entry point for both prctl and syscall. */ |
| 627 | static long do_seccomp(unsigned int op, unsigned int flags, |
| 628 | const char __user *uargs) |
| 629 | { |
| 630 | switch (op) { |
| 631 | case SECCOMP_SET_MODE_STRICT: |
| 632 | if (flags != 0 || uargs != NULL) |
| 633 | return -EINVAL; |
| 634 | return seccomp_set_mode_strict(); |
| 635 | case SECCOMP_SET_MODE_FILTER: |
| 636 | return seccomp_set_mode_filter(flags, uargs); |
| 637 | default: |
| 638 | return -EINVAL; |
| 639 | } |
| 640 | } |
| 641 | |
| 642 | SYSCALL_DEFINE3(seccomp, unsigned int, op, unsigned int, flags, |
| 643 | const char __user *, uargs) |
| 644 | { |
| 645 | return do_seccomp(op, flags, uargs); |
| 646 | } |
| 647 | |
| 648 | /** |
| 649 | * prctl_set_seccomp: configures current->seccomp.mode |
| 650 | * @seccomp_mode: requested mode to use |
| 651 | * @filter: optional struct sock_fprog for use with SECCOMP_MODE_FILTER |
| 652 | * |
| 653 | * Returns 0 on success or -EINVAL on failure. |
| 654 | */ |
| 655 | long prctl_set_seccomp(unsigned long seccomp_mode, char __user *filter) |
| 656 | { |
| 657 | unsigned int op; |
| 658 | char __user *uargs; |
| 659 | |
| 660 | switch (seccomp_mode) { |
| 661 | case SECCOMP_MODE_STRICT: |
| 662 | op = SECCOMP_SET_MODE_STRICT; |
| 663 | /* |
| 664 | * Setting strict mode through prctl always ignored filter, |
| 665 | * so make sure it is always NULL here to pass the internal |
| 666 | * check in do_seccomp(). |
| 667 | */ |
| 668 | uargs = NULL; |
| 669 | break; |
| 670 | case SECCOMP_MODE_FILTER: |
| 671 | op = SECCOMP_SET_MODE_FILTER; |
| 672 | uargs = filter; |
| 673 | break; |
| 674 | default: |
| 675 | return -EINVAL; |
| 676 | } |
| 677 | |
| 678 | /* prctl interface doesn't have flags, so they are always zero. */ |
| 679 | return do_seccomp(op, 0, uargs); |
| 680 | } |