2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2008-2011 Robert N. M. Watson
5 * Copyright (c) 2010-2011 Jonathan Anderson
6 * Copyright (c) 2012 FreeBSD Foundation
9 * This software was developed at the University of Cambridge Computer
10 * Laboratory with support from a grant from Google, Inc.
12 * Portions of this software were developed by Pawel Jakub Dawidek under
13 * sponsorship from the FreeBSD Foundation.
15 * Redistribution and use in source and binary forms, with or without
16 * modification, are permitted provided that the following conditions
18 * 1. Redistributions of source code must retain the above copyright
19 * notice, this list of conditions and the following disclaimer.
20 * 2. Redistributions in binary form must reproduce the above copyright
21 * notice, this list of conditions and the following disclaimer in the
22 * documentation and/or other materials provided with the distribution.
24 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * FreeBSD kernel capability facility.
40 * Two kernel features are implemented here: capability mode, a sandboxed mode
41 * of execution for processes, and capabilities, a refinement on file
42 * descriptors that allows fine-grained control over operations on the file
43 * descriptor. Collectively, these allow processes to run in the style of a
44 * historic "capability system" in which they can use only resources
45 * explicitly delegated to them. This model is enforced by restricting access
46 * to global namespaces in capability mode.
48 * Capabilities wrap other file descriptor types, binding them to a constant
49 * rights mask set when the capability is created. New capabilities may be
50 * derived from existing capabilities, but only if they have the same or a
51 * strict subset of the rights on the original capability.
53 * System calls permitted in capability mode are defined in capabilities.conf;
54 * calls must be carefully audited for safety to ensure that they don't allow
55 * escape from a sandbox. Some calls permit only a subset of operations in
56 * capability mode -- for example, shm_open(2) is limited to creating
57 * anonymous, rather than named, POSIX shared memory objects.
60 #include <sys/cdefs.h>
61 __FBSDID("$FreeBSD$");
63 #include "opt_capsicum.h"
64 #include "opt_ktrace.h"
66 #include <sys/param.h>
67 #include <sys/capsicum.h>
69 #include <sys/filedesc.h>
70 #include <sys/kernel.h>
71 #include <sys/limits.h>
73 #include <sys/mutex.h>
75 #include <sys/syscallsubr.h>
76 #include <sys/sysproto.h>
77 #include <sys/sysctl.h>
78 #include <sys/systm.h>
79 #include <sys/ucred.h>
81 #include <sys/ktrace.h>
83 #include <security/audit/audit.h>
89 SYSCTL_INT(_kern, OID_AUTO, trap_enotcap, CTLFLAG_RW, &trap_enotcap, 0,
90 "Deliver SIGTRAP on ENOTCAPABLE");
92 #ifdef CAPABILITY_MODE
94 #define IOCTLS_MAX_COUNT 256 /* XXX: Is 256 sane? */
96 FEATURE(security_capability_mode, "Capsicum Capability Mode");
99 * System call to enter capability mode for the process.
102 sys_cap_enter(struct thread *td, struct cap_enter_args *uap)
104 struct ucred *newcred, *oldcred;
107 if (IN_CAPABILITY_MODE(td))
113 oldcred = crcopysafe(p, newcred);
114 newcred->cr_flags |= CRED_FLAG_CAPMODE;
115 proc_set_cred(p, newcred);
122 * System call to query whether the process is in capability mode.
125 sys_cap_getmode(struct thread *td, struct cap_getmode_args *uap)
129 i = IN_CAPABILITY_MODE(td) ? 1 : 0;
130 return (copyout(&i, uap->modep, sizeof(i)));
133 #else /* !CAPABILITY_MODE */
136 sys_cap_enter(struct thread *td, struct cap_enter_args *uap)
143 sys_cap_getmode(struct thread *td, struct cap_getmode_args *uap)
149 #endif /* CAPABILITY_MODE */
153 FEATURE(security_capabilities, "Capsicum Capabilities");
155 MALLOC_DECLARE(M_FILECAPS);
158 _cap_check(const cap_rights_t *havep, const cap_rights_t *needp,
159 enum ktr_cap_fail_type type)
162 if (!cap_rights_contains(havep, needp)) {
164 if (KTRPOINT(curthread, KTR_CAPFAIL))
165 ktrcapfail(type, needp, havep);
167 return (ENOTCAPABLE);
173 * Test whether a capability grants the requested rights.
176 cap_check(const cap_rights_t *havep, const cap_rights_t *needp)
179 return (_cap_check(havep, needp, CAPFAIL_NOTCAPABLE));
183 * Convert capability rights into VM access flags.
186 cap_rights_to_vmprot(cap_rights_t *havep)
190 maxprot = VM_PROT_NONE;
191 if (cap_rights_is_set(havep, CAP_MMAP_R))
192 maxprot |= VM_PROT_READ;
193 if (cap_rights_is_set(havep, CAP_MMAP_W))
194 maxprot |= VM_PROT_WRITE;
195 if (cap_rights_is_set(havep, CAP_MMAP_X))
196 maxprot |= VM_PROT_EXECUTE;
202 * Extract rights from a capability for monitoring purposes -- not for use in
203 * any other way, as we want to keep all capability permission evaluation in
208 cap_rights_fde(struct filedescent *fdep)
211 return (&fdep->fde_rights);
215 cap_rights(struct filedesc *fdp, int fd)
218 return (cap_rights_fde(&fdp->fd_ofiles[fd]));
222 kern_cap_rights_limit(struct thread *td, int fd, cap_rights_t *rights)
224 struct filedesc *fdp;
225 struct filedescent *fdep;
228 fdp = td->td_proc->p_fd;
230 fdep = fdeget_locked(fdp, fd);
232 FILEDESC_XUNLOCK(fdp);
235 error = _cap_check(cap_rights(fdp, fd), rights, CAPFAIL_INCREASE);
237 fdep->fde_rights = *rights;
238 if (!cap_rights_is_set(rights, CAP_IOCTL)) {
239 free(fdep->fde_ioctls, M_FILECAPS);
240 fdep->fde_ioctls = NULL;
241 fdep->fde_nioctls = 0;
243 if (!cap_rights_is_set(rights, CAP_FCNTL))
244 fdep->fde_fcntls = 0;
246 FILEDESC_XUNLOCK(fdp);
251 * System call to limit rights of the given capability.
254 sys_cap_rights_limit(struct thread *td, struct cap_rights_limit_args *uap)
259 cap_rights_init(&rights);
261 error = copyin(uap->rightsp, &rights, sizeof(rights.cr_rights[0]));
264 version = CAPVER(&rights);
265 if (version != CAP_RIGHTS_VERSION_00)
268 error = copyin(uap->rightsp, &rights,
269 sizeof(rights.cr_rights[0]) * CAPARSIZE(&rights));
272 /* Check for race. */
273 if (CAPVER(&rights) != version)
276 if (!cap_rights_is_valid(&rights))
279 if (version != CAP_RIGHTS_VERSION) {
280 rights.cr_rights[0] &= ~(0x3ULL << 62);
281 rights.cr_rights[0] |= ((uint64_t)CAP_RIGHTS_VERSION << 62);
284 if (KTRPOINT(td, KTR_STRUCT))
285 ktrcaprights(&rights);
288 AUDIT_ARG_FD(uap->fd);
289 AUDIT_ARG_RIGHTS(&rights);
290 return (kern_cap_rights_limit(td, uap->fd, &rights));
294 * System call to query the rights mask associated with a capability.
297 sys___cap_rights_get(struct thread *td, struct __cap_rights_get_args *uap)
299 struct filedesc *fdp;
303 if (uap->version != CAP_RIGHTS_VERSION_00)
310 fdp = td->td_proc->p_fd;
312 if (fget_locked(fdp, fd) == NULL) {
313 FILEDESC_SUNLOCK(fdp);
316 rights = *cap_rights(fdp, fd);
317 FILEDESC_SUNLOCK(fdp);
318 n = uap->version + 2;
319 if (uap->version != CAPVER(&rights)) {
321 * For older versions we need to check if the descriptor
322 * doesn't contain rights not understood by the caller.
323 * If it does, we have to return an error.
325 for (i = n; i < CAPARSIZE(&rights); i++) {
326 if ((rights.cr_rights[i] & ~(0x7FULL << 57)) != 0)
330 error = copyout(&rights, uap->rightsp, sizeof(rights.cr_rights[0]) * n);
332 if (error == 0 && KTRPOINT(td, KTR_STRUCT))
333 ktrcaprights(&rights);
339 * Test whether a capability grants the given ioctl command.
340 * If descriptor doesn't have CAP_IOCTL, then ioctls list is empty and
341 * ENOTCAPABLE will be returned.
344 cap_ioctl_check(struct filedesc *fdp, int fd, u_long cmd)
346 struct filedescent *fdep;
351 KASSERT(fd >= 0 && fd < fdp->fd_nfiles,
352 ("%s: invalid fd=%d", __func__, fd));
354 fdep = fdeget_locked(fdp, fd);
355 KASSERT(fdep != NULL,
356 ("%s: invalid fd=%d", __func__, fd));
358 ncmds = fdep->fde_nioctls;
362 cmds = fdep->fde_ioctls;
363 for (i = 0; i < ncmds; i++) {
368 return (ENOTCAPABLE);
372 * Check if the current ioctls list can be replaced by the new one.
375 cap_ioctl_limit_check(struct filedescent *fdep, const u_long *cmds,
383 oncmds = fdep->fde_nioctls;
386 if (oncmds < (ssize_t)ncmds)
387 return (ENOTCAPABLE);
389 ocmds = fdep->fde_ioctls;
390 for (i = 0; i < ncmds; i++) {
391 for (j = 0; j < oncmds; j++) {
392 if (cmds[i] == ocmds[j])
396 return (ENOTCAPABLE);
403 kern_cap_ioctls_limit(struct thread *td, int fd, u_long *cmds, size_t ncmds)
405 struct filedesc *fdp;
406 struct filedescent *fdep;
412 if (ncmds > IOCTLS_MAX_COUNT) {
417 fdp = td->td_proc->p_fd;
420 fdep = fdeget_locked(fdp, fd);
426 error = cap_ioctl_limit_check(fdep, cmds, ncmds);
430 ocmds = fdep->fde_ioctls;
431 fdep->fde_ioctls = cmds;
432 fdep->fde_nioctls = ncmds;
437 FILEDESC_XUNLOCK(fdp);
439 free(cmds, M_FILECAPS);
444 sys_cap_ioctls_limit(struct thread *td, struct cap_ioctls_limit_args *uap)
452 if (ncmds > IOCTLS_MAX_COUNT)
458 cmds = malloc(sizeof(cmds[0]) * ncmds, M_FILECAPS, M_WAITOK);
459 error = copyin(uap->cmds, cmds, sizeof(cmds[0]) * ncmds);
461 free(cmds, M_FILECAPS);
466 return (kern_cap_ioctls_limit(td, uap->fd, cmds, ncmds));
470 sys_cap_ioctls_get(struct thread *td, struct cap_ioctls_get_args *uap)
472 struct filedesc *fdp;
473 struct filedescent *fdep;
474 u_long *cmdsp, *dstcmds;
475 size_t maxcmds, ncmds;
481 maxcmds = uap->maxcmds;
485 fdp = td->td_proc->p_fd;
488 if (dstcmds != NULL) {
489 cmdsp = malloc(sizeof(cmdsp[0]) * IOCTLS_MAX_COUNT, M_FILECAPS,
494 fdep = fdeget_locked(fdp, fd);
497 FILEDESC_SUNLOCK(fdp);
500 count = fdep->fde_nioctls;
501 if (count != -1 && cmdsp != NULL) {
502 ncmds = MIN(count, maxcmds);
503 memcpy(cmdsp, fdep->fde_ioctls, sizeof(cmdsp[0]) * ncmds);
505 FILEDESC_SUNLOCK(fdp);
508 * If all ioctls are allowed (fde_nioctls == -1 && fde_ioctls == NULL)
509 * the only sane thing we can do is to not populate the given array and
510 * return CAP_IOCTLS_ALL.
514 error = copyout(cmdsp, dstcmds,
515 sizeof(cmdsp[0]) * ncmds);
519 td->td_retval[0] = count;
521 td->td_retval[0] = CAP_IOCTLS_ALL;
526 free(cmdsp, M_FILECAPS);
531 * Test whether a capability grants the given fcntl command.
534 cap_fcntl_check_fde(struct filedescent *fdep, int cmd)
538 fcntlcap = (1 << cmd);
539 KASSERT((CAP_FCNTL_ALL & fcntlcap) != 0,
540 ("Unsupported fcntl=%d.", cmd));
542 if ((fdep->fde_fcntls & fcntlcap) != 0)
545 return (ENOTCAPABLE);
549 cap_fcntl_check(struct filedesc *fdp, int fd, int cmd)
552 KASSERT(fd >= 0 && fd < fdp->fd_nfiles,
553 ("%s: invalid fd=%d", __func__, fd));
555 return (cap_fcntl_check_fde(&fdp->fd_ofiles[fd], cmd));
559 sys_cap_fcntls_limit(struct thread *td, struct cap_fcntls_limit_args *uap)
561 struct filedesc *fdp;
562 struct filedescent *fdep;
563 uint32_t fcntlrights;
567 fcntlrights = uap->fcntlrights;
570 AUDIT_ARG_FCNTL_RIGHTS(fcntlrights);
572 if ((fcntlrights & ~CAP_FCNTL_ALL) != 0)
575 fdp = td->td_proc->p_fd;
578 fdep = fdeget_locked(fdp, fd);
580 FILEDESC_XUNLOCK(fdp);
584 if ((fcntlrights & ~fdep->fde_fcntls) != 0) {
585 FILEDESC_XUNLOCK(fdp);
586 return (ENOTCAPABLE);
589 fdep->fde_fcntls = fcntlrights;
590 FILEDESC_XUNLOCK(fdp);
596 sys_cap_fcntls_get(struct thread *td, struct cap_fcntls_get_args *uap)
598 struct filedesc *fdp;
599 struct filedescent *fdep;
607 fdp = td->td_proc->p_fd;
609 fdep = fdeget_locked(fdp, fd);
611 FILEDESC_SUNLOCK(fdp);
614 rights = fdep->fde_fcntls;
615 FILEDESC_SUNLOCK(fdp);
617 return (copyout(&rights, uap->fcntlrightsp, sizeof(rights)));
620 #else /* !CAPABILITIES */
623 * Stub Capability functions for when options CAPABILITIES isn't compiled
628 sys_cap_rights_limit(struct thread *td, struct cap_rights_limit_args *uap)
635 sys___cap_rights_get(struct thread *td, struct __cap_rights_get_args *uap)
642 sys_cap_ioctls_limit(struct thread *td, struct cap_ioctls_limit_args *uap)
649 sys_cap_ioctls_get(struct thread *td, struct cap_ioctls_get_args *uap)
656 sys_cap_fcntls_limit(struct thread *td, struct cap_fcntls_limit_args *uap)
663 sys_cap_fcntls_get(struct thread *td, struct cap_fcntls_get_args *uap)
669 #endif /* CAPABILITIES */