[qemu.git] / qemu-seccomp.c

/*
 * QEMU seccomp mode 2 support with libseccomp
 *
 * Copyright IBM, Corp. 2012
 *
 * Authors:
 *  Eduardo Otubo    <[email protected]>
 *
 * This work is licensed under the terms of the GNU GPL, version 2.  See
 * the COPYING file in the top-level directory.
 *
 * Contributions after 2012-01-13 are licensed under the terms of the
 * GNU GPL, version 2 or (at your option) any later version.
 */
#include "qemu/osdep.h"
#include "qemu/config-file.h"
#include "qemu/option.h"
#include "qemu/module.h"
#include "qemu/error-report.h"
#include <sys/prctl.h>
#include <seccomp.h>
#include "sysemu/seccomp.h"

/* For some architectures (notably ARM) cacheflush is not supported until
 * libseccomp 2.2.3, but configure enforces that we are using a more recent
 * version on those hosts, so it is OK for this check to be less strict.
 */
#if SCMP_VER_MAJOR >= 3
  #define HAVE_CACHEFLUSH
#elif SCMP_VER_MAJOR == 2 && SCMP_VER_MINOR >= 2
  #define HAVE_CACHEFLUSH
#endif

struct QemuSeccompSyscall {
    int32_t num;
    uint8_t set;
};

static const struct QemuSeccompSyscall blacklist[] = {
    /* default set of syscalls to blacklist */
    { SCMP_SYS(reboot),                 QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(swapon),                 QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(swapoff),                QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(syslog),                 QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(mount),                  QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(umount),                 QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(kexec_load),             QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(afs_syscall),            QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(break),                  QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(ftime),                  QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(getpmsg),                QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(gtty),                   QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(lock),                   QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(mpx),                    QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(prof),                   QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(profil),                 QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(putpmsg),                QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(security),               QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(stty),                   QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(tuxcall),                QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(ulimit),                 QEMU_SECCOMP_SET_DEFAULT },
    { SCMP_SYS(vserver),                QEMU_SECCOMP_SET_DEFAULT },
    /* obsolete */
    { SCMP_SYS(readdir),                QEMU_SECCOMP_SET_OBSOLETE },
    { SCMP_SYS(_sysctl),                QEMU_SECCOMP_SET_OBSOLETE },
    { SCMP_SYS(bdflush),                QEMU_SECCOMP_SET_OBSOLETE },
    { SCMP_SYS(create_module),          QEMU_SECCOMP_SET_OBSOLETE },
    { SCMP_SYS(get_kernel_syms),        QEMU_SECCOMP_SET_OBSOLETE },
    { SCMP_SYS(query_module),           QEMU_SECCOMP_SET_OBSOLETE },
    { SCMP_SYS(sgetmask),               QEMU_SECCOMP_SET_OBSOLETE },
    { SCMP_SYS(ssetmask),               QEMU_SECCOMP_SET_OBSOLETE },
    { SCMP_SYS(sysfs),                  QEMU_SECCOMP_SET_OBSOLETE },
    { SCMP_SYS(uselib),                 QEMU_SECCOMP_SET_OBSOLETE },
    { SCMP_SYS(ustat),                  QEMU_SECCOMP_SET_OBSOLETE },
    /* privileged */
    { SCMP_SYS(setuid),                 QEMU_SECCOMP_SET_PRIVILEGED },
    { SCMP_SYS(setgid),                 QEMU_SECCOMP_SET_PRIVILEGED },
    { SCMP_SYS(setpgid),                QEMU_SECCOMP_SET_PRIVILEGED },
    { SCMP_SYS(setsid),                 QEMU_SECCOMP_SET_PRIVILEGED },
    { SCMP_SYS(setreuid),               QEMU_SECCOMP_SET_PRIVILEGED },
    { SCMP_SYS(setregid),               QEMU_SECCOMP_SET_PRIVILEGED },
    { SCMP_SYS(setresuid),              QEMU_SECCOMP_SET_PRIVILEGED },
    { SCMP_SYS(setresgid),              QEMU_SECCOMP_SET_PRIVILEGED },
    { SCMP_SYS(setfsuid),               QEMU_SECCOMP_SET_PRIVILEGED },
    { SCMP_SYS(setfsgid),               QEMU_SECCOMP_SET_PRIVILEGED },
    /* spawn */
    { SCMP_SYS(fork),                   QEMU_SECCOMP_SET_SPAWN },
    { SCMP_SYS(vfork),                  QEMU_SECCOMP_SET_SPAWN },
    { SCMP_SYS(execve),                 QEMU_SECCOMP_SET_SPAWN },
    /* resource control */
    { SCMP_SYS(getpriority),            QEMU_SECCOMP_SET_RESOURCECTL },
    { SCMP_SYS(setpriority),            QEMU_SECCOMP_SET_RESOURCECTL },
    { SCMP_SYS(sched_setparam),         QEMU_SECCOMP_SET_RESOURCECTL },
    { SCMP_SYS(sched_getparam),         QEMU_SECCOMP_SET_RESOURCECTL },
    { SCMP_SYS(sched_setscheduler),     QEMU_SECCOMP_SET_RESOURCECTL },
    { SCMP_SYS(sched_getscheduler),     QEMU_SECCOMP_SET_RESOURCECTL },
    { SCMP_SYS(sched_setaffinity),      QEMU_SECCOMP_SET_RESOURCECTL },
    { SCMP_SYS(sched_getaffinity),      QEMU_SECCOMP_SET_RESOURCECTL },
    { SCMP_SYS(sched_get_priority_max), QEMU_SECCOMP_SET_RESOURCECTL },
    { SCMP_SYS(sched_get_priority_min), QEMU_SECCOMP_SET_RESOURCECTL },
};


static int seccomp_start(uint32_t seccomp_opts)
{
    int rc = 0;
    unsigned int i = 0;
    scmp_filter_ctx ctx;

    ctx = seccomp_init(SCMP_ACT_ALLOW);
    if (ctx == NULL) {
        rc = -1;
        goto seccomp_return;
    }

    for (i = 0; i < ARRAY_SIZE(blacklist); i++) {
        if (!(seccomp_opts & blacklist[i].set)) {
            continue;
        }

        rc = seccomp_rule_add(ctx, SCMP_ACT_KILL, blacklist[i].num, 0);
        if (rc < 0) {
            goto seccomp_return;
        }
    }

    rc = seccomp_load(ctx);

  seccomp_return:
    seccomp_release(ctx);
    return rc;
}

#ifdef CONFIG_SECCOMP
int parse_sandbox(void *opaque, QemuOpts *opts, Error **errp)
{
    if (qemu_opt_get_bool(opts, "enable", false)) {
        uint32_t seccomp_opts = QEMU_SECCOMP_SET_DEFAULT
                | QEMU_SECCOMP_SET_OBSOLETE;
        const char *value = NULL;

        value = qemu_opt_get(opts, "obsolete");
        if (value) {
            if (g_str_equal(value, "allow")) {
                seccomp_opts &= ~QEMU_SECCOMP_SET_OBSOLETE;
            } else if (g_str_equal(value, "deny")) {
                /* this is the default option, this if is here
                 * to provide a little bit of consistency for
                 * the command line */
            } else {
                error_report("invalid argument for obsolete");
                return -1;
            }
        }

        value = qemu_opt_get(opts, "elevateprivileges");
        if (value) {
            if (g_str_equal(value, "deny")) {
                seccomp_opts |= QEMU_SECCOMP_SET_PRIVILEGED;
            } else if (g_str_equal(value, "children")) {
                seccomp_opts |= QEMU_SECCOMP_SET_PRIVILEGED;

                /* calling prctl directly because we're
                 * not sure if host has CAP_SYS_ADMIN set*/
                if (prctl(PR_SET_NO_NEW_PRIVS, 1)) {
                    error_report("failed to set no_new_privs "
                                 "aborting");
                    return -1;
                }
            } else if (g_str_equal(value, "allow")) {
                /* default value */
            } else {
                error_report("invalid argument for elevateprivileges");
                return -1;
            }
        }

        value = qemu_opt_get(opts, "spawn");
        if (value) {
            if (g_str_equal(value, "deny")) {
                seccomp_opts |= QEMU_SECCOMP_SET_SPAWN;
            } else if (g_str_equal(value, "allow")) {
                /* default value */
            } else {
                error_report("invalid argument for spawn");
                return -1;
            }
        }

        value = qemu_opt_get(opts, "resourcecontrol");
        if (value) {
            if (g_str_equal(value, "deny")) {
                seccomp_opts |= QEMU_SECCOMP_SET_RESOURCECTL;
            } else if (g_str_equal(value, "allow")) {
                /* default value */
            } else {
                error_report("invalid argument for resourcecontrol");
                return -1;
            }
        }

        if (seccomp_start(seccomp_opts) < 0) {
            error_report("failed to install seccomp syscall filter "
                         "in the kernel");
            return -1;
        }
    }

    return 0;
}

static QemuOptsList qemu_sandbox_opts = {
    .name = "sandbox",
    .implied_opt_name = "enable",
    .head = QTAILQ_HEAD_INITIALIZER(qemu_sandbox_opts.head),
    .desc = {
        {
            .name = "enable",
            .type = QEMU_OPT_BOOL,
        },
        {
            .name = "obsolete",
            .type = QEMU_OPT_STRING,
        },
        {
            .name = "elevateprivileges",
            .type = QEMU_OPT_STRING,
        },
        {
            .name = "spawn",
            .type = QEMU_OPT_STRING,
        },
        {
            .name = "resourcecontrol",
            .type = QEMU_OPT_STRING,
        },
        { /* end of list */ }
    },
};

static void seccomp_register(void)
{
    qemu_add_opts(&qemu_sandbox_opts);
}
opts_init(seccomp_register);
#endif
Commit	Line	Data
2f668be7 EO	1	/*
	2	* QEMU seccomp mode 2 support with libseccomp
	3	*
	4	* Copyright IBM, Corp. 2012
	5	*
	6	* Authors:
	7	* Eduardo Otubo <[email protected]>
	8	*
	9	* This work is licensed under the terms of the GNU GPL, version 2. See
	10	* the COPYING file in the top-level directory.
	11	*
	12	* Contributions after 2012-01-13 are licensed under the terms of the
	13	* GNU GPL, version 2 or (at your option) any later version.
	14	*/
d38ea87a	15	#include "qemu/osdep.h"
9d0fdecb YMZ	16	#include "qemu/config-file.h"
	17	#include "qemu/option.h"
	18	#include "qemu/module.h"
	19	#include "qemu/error-report.h"
	20	#include <sys/prctl.h>
2f668be7	21	#include <seccomp.h>
9c17d615	22	#include "sysemu/seccomp.h"
2f668be7	23
81bed73b JH	24	/* For some architectures (notably ARM) cacheflush is not supported until
	25	* libseccomp 2.2.3, but configure enforces that we are using a more recent
	26	* version on those hosts, so it is OK for this check to be less strict.
	27	*/
47d2067a AJ	28	#if SCMP_VER_MAJOR >= 3
47d2067a AJ	29	#define HAVE_CACHEFLUSH
81bed73b	30	#elif SCMP_VER_MAJOR == 2 && SCMP_VER_MINOR >= 2
47d2067a AJ	31	#define HAVE_CACHEFLUSH
	32	#endif
	33
2f668be7 EO	34	struct QemuSeccompSyscall {
2f668be7 EO	35	int32_t num;
1bd6152a	36	uint8_t set;
2f668be7 EO	37	};
2f668be7 EO	38
1bd6152a EO	39	static const struct QemuSeccompSyscall blacklist[] = {
	40	/* default set of syscalls to blacklist */
	41	{ SCMP_SYS(reboot), QEMU_SECCOMP_SET_DEFAULT },
	42	{ SCMP_SYS(swapon), QEMU_SECCOMP_SET_DEFAULT },
	43	{ SCMP_SYS(swapoff), QEMU_SECCOMP_SET_DEFAULT },
	44	{ SCMP_SYS(syslog), QEMU_SECCOMP_SET_DEFAULT },
	45	{ SCMP_SYS(mount), QEMU_SECCOMP_SET_DEFAULT },
	46	{ SCMP_SYS(umount), QEMU_SECCOMP_SET_DEFAULT },
	47	{ SCMP_SYS(kexec_load), QEMU_SECCOMP_SET_DEFAULT },
	48	{ SCMP_SYS(afs_syscall), QEMU_SECCOMP_SET_DEFAULT },
	49	{ SCMP_SYS(break), QEMU_SECCOMP_SET_DEFAULT },
	50	{ SCMP_SYS(ftime), QEMU_SECCOMP_SET_DEFAULT },
	51	{ SCMP_SYS(getpmsg), QEMU_SECCOMP_SET_DEFAULT },
	52	{ SCMP_SYS(gtty), QEMU_SECCOMP_SET_DEFAULT },
	53	{ SCMP_SYS(lock), QEMU_SECCOMP_SET_DEFAULT },
	54	{ SCMP_SYS(mpx), QEMU_SECCOMP_SET_DEFAULT },
	55	{ SCMP_SYS(prof), QEMU_SECCOMP_SET_DEFAULT },
	56	{ SCMP_SYS(profil), QEMU_SECCOMP_SET_DEFAULT },
	57	{ SCMP_SYS(putpmsg), QEMU_SECCOMP_SET_DEFAULT },
	58	{ SCMP_SYS(security), QEMU_SECCOMP_SET_DEFAULT },
	59	{ SCMP_SYS(stty), QEMU_SECCOMP_SET_DEFAULT },
	60	{ SCMP_SYS(tuxcall), QEMU_SECCOMP_SET_DEFAULT },
	61	{ SCMP_SYS(ulimit), QEMU_SECCOMP_SET_DEFAULT },
	62	{ SCMP_SYS(vserver), QEMU_SECCOMP_SET_DEFAULT },
2b716fa6 EO	63	/* obsolete */
	64	{ SCMP_SYS(readdir), QEMU_SECCOMP_SET_OBSOLETE },
	65	{ SCMP_SYS(_sysctl), QEMU_SECCOMP_SET_OBSOLETE },
	66	{ SCMP_SYS(bdflush), QEMU_SECCOMP_SET_OBSOLETE },
	67	{ SCMP_SYS(create_module), QEMU_SECCOMP_SET_OBSOLETE },
	68	{ SCMP_SYS(get_kernel_syms), QEMU_SECCOMP_SET_OBSOLETE },
	69	{ SCMP_SYS(query_module), QEMU_SECCOMP_SET_OBSOLETE },
	70	{ SCMP_SYS(sgetmask), QEMU_SECCOMP_SET_OBSOLETE },
	71	{ SCMP_SYS(ssetmask), QEMU_SECCOMP_SET_OBSOLETE },
	72	{ SCMP_SYS(sysfs), QEMU_SECCOMP_SET_OBSOLETE },
	73	{ SCMP_SYS(uselib), QEMU_SECCOMP_SET_OBSOLETE },
	74	{ SCMP_SYS(ustat), QEMU_SECCOMP_SET_OBSOLETE },
73a1e647 EO	75	/* privileged */
	76	{ SCMP_SYS(setuid), QEMU_SECCOMP_SET_PRIVILEGED },
	77	{ SCMP_SYS(setgid), QEMU_SECCOMP_SET_PRIVILEGED },
	78	{ SCMP_SYS(setpgid), QEMU_SECCOMP_SET_PRIVILEGED },
	79	{ SCMP_SYS(setsid), QEMU_SECCOMP_SET_PRIVILEGED },
	80	{ SCMP_SYS(setreuid), QEMU_SECCOMP_SET_PRIVILEGED },
	81	{ SCMP_SYS(setregid), QEMU_SECCOMP_SET_PRIVILEGED },
	82	{ SCMP_SYS(setresuid), QEMU_SECCOMP_SET_PRIVILEGED },
	83	{ SCMP_SYS(setresgid), QEMU_SECCOMP_SET_PRIVILEGED },
	84	{ SCMP_SYS(setfsuid), QEMU_SECCOMP_SET_PRIVILEGED },
	85	{ SCMP_SYS(setfsgid), QEMU_SECCOMP_SET_PRIVILEGED },
995a226f EO	86	/* spawn */
	87	{ SCMP_SYS(fork), QEMU_SECCOMP_SET_SPAWN },
	88	{ SCMP_SYS(vfork), QEMU_SECCOMP_SET_SPAWN },
	89	{ SCMP_SYS(execve), QEMU_SECCOMP_SET_SPAWN },
24f8cdc5 EO	90	/* resource control */
	91	{ SCMP_SYS(getpriority), QEMU_SECCOMP_SET_RESOURCECTL },
	92	{ SCMP_SYS(setpriority), QEMU_SECCOMP_SET_RESOURCECTL },
	93	{ SCMP_SYS(sched_setparam), QEMU_SECCOMP_SET_RESOURCECTL },
	94	{ SCMP_SYS(sched_getparam), QEMU_SECCOMP_SET_RESOURCECTL },
	95	{ SCMP_SYS(sched_setscheduler), QEMU_SECCOMP_SET_RESOURCECTL },
	96	{ SCMP_SYS(sched_getscheduler), QEMU_SECCOMP_SET_RESOURCECTL },
	97	{ SCMP_SYS(sched_setaffinity), QEMU_SECCOMP_SET_RESOURCECTL },
	98	{ SCMP_SYS(sched_getaffinity), QEMU_SECCOMP_SET_RESOURCECTL },
	99	{ SCMP_SYS(sched_get_priority_max), QEMU_SECCOMP_SET_RESOURCECTL },
	100	{ SCMP_SYS(sched_get_priority_min), QEMU_SECCOMP_SET_RESOURCECTL },
2f668be7 EO	101	};
2f668be7 EO	102
2b716fa6	103
9d0fdecb	104	static int seccomp_start(uint32_t seccomp_opts)
2f668be7 EO	105	{
	106	int rc = 0;
	107	unsigned int i = 0;
	108	scmp_filter_ctx ctx;
	109
1bd6152a	110	ctx = seccomp_init(SCMP_ACT_ALLOW);
2f668be7	111	if (ctx == NULL) {
2a13f991	112	rc = -1;
2f668be7 EO	113	goto seccomp_return;
	114	}
	115
1bd6152a	116	for (i = 0; i < ARRAY_SIZE(blacklist); i++) {
2b716fa6 EO	117	if (!(seccomp_opts & blacklist[i].set)) {
	118	continue;
	119	}
	120
1bd6152a	121	rc = seccomp_rule_add(ctx, SCMP_ACT_KILL, blacklist[i].num, 0);
2f668be7 EO	122	if (rc < 0) {
	123	goto seccomp_return;
	124	}
	125	}
	126
	127	rc = seccomp_load(ctx);
	128
	129	seccomp_return:
	130	seccomp_release(ctx);
	131	return rc;
	132	}
9d0fdecb YMZ	133
	134	#ifdef CONFIG_SECCOMP
	135	int parse_sandbox(void opaque, QemuOpts opts, Error **errp)
	136	{
	137	if (qemu_opt_get_bool(opts, "enable", false)) {
	138	uint32_t seccomp_opts = QEMU_SECCOMP_SET_DEFAULT
	139	\| QEMU_SECCOMP_SET_OBSOLETE;
	140	const char *value = NULL;
	141
	142	value = qemu_opt_get(opts, "obsolete");
	143	if (value) {
	144	if (g_str_equal(value, "allow")) {
	145	seccomp_opts &= ~QEMU_SECCOMP_SET_OBSOLETE;
	146	} else if (g_str_equal(value, "deny")) {
	147	/* this is the default option, this if is here
	148	* to provide a little bit of consistency for
	149	* the command line */
	150	} else {
	151	error_report("invalid argument for obsolete");
	152	return -1;
	153	}
	154	}
	155
	156	value = qemu_opt_get(opts, "elevateprivileges");
	157	if (value) {
	158	if (g_str_equal(value, "deny")) {
	159	seccomp_opts \|= QEMU_SECCOMP_SET_PRIVILEGED;
	160	} else if (g_str_equal(value, "children")) {
	161	seccomp_opts \|= QEMU_SECCOMP_SET_PRIVILEGED;
	162
	163	/* calling prctl directly because we're
	164	* not sure if host has CAP_SYS_ADMIN set*/
	165	if (prctl(PR_SET_NO_NEW_PRIVS, 1)) {
	166	error_report("failed to set no_new_privs "
	167	"aborting");
	168	return -1;
	169	}
	170	} else if (g_str_equal(value, "allow")) {
	171	/* default value */
	172	} else {
	173	error_report("invalid argument for elevateprivileges");
	174	return -1;
	175	}
	176	}
	177
	178	value = qemu_opt_get(opts, "spawn");
	179	if (value) {
	180	if (g_str_equal(value, "deny")) {
	181	seccomp_opts \|= QEMU_SECCOMP_SET_SPAWN;
	182	} else if (g_str_equal(value, "allow")) {
	183	/* default value */
	184	} else {
	185	error_report("invalid argument for spawn");
	186	return -1;
	187	}
	188	}
	189
	190	value = qemu_opt_get(opts, "resourcecontrol");
	191	if (value) {
	192	if (g_str_equal(value, "deny")) {
	193	seccomp_opts \|= QEMU_SECCOMP_SET_RESOURCECTL;
	194	} else if (g_str_equal(value, "allow")) {
	195	/* default value */
	196	} else {
197	error_report("invalid argument for resourcecontrol");
198	return -1;
199	}
200	}
201
202	if (seccomp_start(seccomp_opts) < 0) {
203	error_report("failed to install seccomp syscall filter "
204	"in the kernel");
205	return -1;
206	}
207	}
208
209	return 0;
210	}
211
212	static QemuOptsList qemu_sandbox_opts = {
213	.name = "sandbox",
214	.implied_opt_name = "enable",
215	.head = QTAILQ_HEAD_INITIALIZER(qemu_sandbox_opts.head),
216	.desc = {
217	{
218	.name = "enable",
219	.type = QEMU_OPT_BOOL,
220	},
221	{
222	.name = "obsolete",
223	.type = QEMU_OPT_STRING,
224	},
225	{
226	.name = "elevateprivileges",
227	.type = QEMU_OPT_STRING,
228	},
229	{
230	.name = "spawn",
231	.type = QEMU_OPT_STRING,
232	},
233	{
234	.name = "resourcecontrol",
235	.type = QEMU_OPT_STRING,
236	},
237	{ /* end of list */ }
238	},
239	};
240
241	static void seccomp_register(void)
242	{
243	qemu_add_opts(&qemu_sandbox_opts);
244	}
245	opts_init(seccomp_register);
246	#endif