]>
git.ipfire.org Git - thirdparty/systemd.git/blob - src/nspawn/nspawn-seccomp.c
1 /* SPDX-License-Identifier: LGPL-2.1+ */
3 Copyright 2016 Lennart Poettering
7 #include <linux/netlink.h>
8 #include <sys/capability.h>
15 #include "alloc-util.h"
17 #include "nspawn-seccomp.h"
19 #include "seccomp-util.h"
21 #include "string-util.h"
26 static int seccomp_add_default_syscall_filter(
29 uint64_t cap_list_retain
,
30 char **syscall_whitelist
,
31 char **syscall_blacklist
) {
37 /* Let's use set names where we can */
42 { 0, "@file-system" },
54 /* The following four are sets we optionally enable, in case the caps have been configured for it */
55 { CAP_SYS_TIME
, "@clock" },
56 { CAP_SYS_MODULE
, "@module" },
57 { CAP_SYS_RAWIO
, "@raw-io" },
58 { CAP_IPC_LOCK
, "@memlock" },
60 /* Plus a good set of additional syscalls which are not part of any of the groups above */
64 { 0, "copy_file_range" },
66 { 0, "fadvise64_64" },
68 { 0, "get_mempolicy" },
79 { 0, "name_to_handle_at" },
85 { 0, "remap_file_pages" },
86 { 0, "sched_get_priority_max" },
87 { 0, "sched_get_priority_min" },
88 { 0, "sched_getaffinity" },
89 { 0, "sched_getattr" },
90 { 0, "sched_getparam" },
91 { 0, "sched_getscheduler" },
92 { 0, "sched_rr_get_interval" },
97 { 0, "setdomainname" },
102 { 0, "sethostname" },
110 { 0, "userfaultfd" },
113 /* The following individual syscalls are added depending on specified caps */
114 { CAP_SYS_PACCT
, "acct" },
115 { CAP_SYS_PTRACE
, "process_vm_readv" },
116 { CAP_SYS_PTRACE
, "process_vm_writev" },
117 { CAP_SYS_PTRACE
, "ptrace" },
118 { CAP_SYS_BOOT
, "reboot" },
119 { CAP_SYSLOG
, "syslog" },
120 { CAP_SYS_TTY_CONFIG
, "vhangup" },
123 * The following syscalls and groups are knowingly excluded:
126 * @keyring (NB: keyring is not namespaced!)
130 * bpf (NB: bpffs is not namespaced!)
150 for (i
= 0; i
< ELEMENTSOF(whitelist
); i
++) {
151 if (whitelist
[i
].capability
!= 0 && (cap_list_retain
& (1ULL << whitelist
[i
].capability
)) == 0)
154 r
= seccomp_add_syscall_filter_item(ctx
, whitelist
[i
].name
, SCMP_ACT_ALLOW
, syscall_blacklist
);
156 /* If the system call is not known on this architecture, then that's fine, let's ignore it */
157 log_debug_errno(r
, "Failed to add rule for system call %s on %s, ignoring: %m", whitelist
[i
].name
, seccomp_arch_to_string(arch
));
162 STRV_FOREACH(p
, syscall_whitelist
) {
163 r
= seccomp_add_syscall_filter_item(ctx
, *p
, SCMP_ACT_ALLOW
, syscall_blacklist
);
165 log_debug_errno(r
, "Failed to add rule for system call %s on %s, ignoring: %m", *p
, seccomp_arch_to_string(arch
));
173 int setup_seccomp(uint64_t cap_list_retain
, char **syscall_whitelist
, char **syscall_blacklist
) {
177 if (!is_seccomp_available()) {
178 log_debug("SECCOMP features not detected in the kernel, disabling SECCOMP filterering");
182 SECCOMP_FOREACH_LOCAL_ARCH(arch
) {
183 _cleanup_(seccomp_releasep
) scmp_filter_ctx seccomp
= NULL
;
185 log_debug("Applying whitelist on architecture: %s", seccomp_arch_to_string(arch
));
187 r
= seccomp_init_for_arch(&seccomp
, arch
, SCMP_ACT_ERRNO(EPERM
));
189 return log_error_errno(r
, "Failed to allocate seccomp object: %m");
191 r
= seccomp_add_default_syscall_filter(seccomp
, arch
, cap_list_retain
, syscall_whitelist
, syscall_blacklist
);
195 r
= seccomp_load(seccomp
);
196 if (IN_SET(r
, -EPERM
, -EACCES
))
197 return log_error_errno(r
, "Failed to install seccomp filter: %m");
199 log_debug_errno(r
, "Failed to install filter set for architecture %s, skipping: %m", seccomp_arch_to_string(arch
));
202 SECCOMP_FOREACH_LOCAL_ARCH(arch
) {
203 _cleanup_(seccomp_releasep
) scmp_filter_ctx seccomp
= NULL
;
205 log_debug("Applying NETLINK_AUDIT mask on architecture: %s", seccomp_arch_to_string(arch
));
207 r
= seccomp_init_for_arch(&seccomp
, arch
, SCMP_ACT_ALLOW
);
209 return log_error_errno(r
, "Failed to allocate seccomp object: %m");
212 Audit is broken in containers, much of the userspace audit hookup will fail if running inside a
213 container. We don't care and just turn off creation of audit sockets.
215 This will make socket(AF_NETLINK, *, NETLINK_AUDIT) fail with EAFNOSUPPORT which audit userspace uses
216 as indication that audit is disabled in the kernel.
219 r
= seccomp_rule_add_exact(
221 SCMP_ACT_ERRNO(EAFNOSUPPORT
),
224 SCMP_A0(SCMP_CMP_EQ
, AF_NETLINK
),
225 SCMP_A2(SCMP_CMP_EQ
, NETLINK_AUDIT
));
227 log_debug_errno(r
, "Failed to add audit seccomp rule, ignoring: %m");
231 r
= seccomp_load(seccomp
);
232 if (IN_SET(r
, -EPERM
, -EACCES
))
233 return log_error_errno(r
, "Failed to install seccomp audit filter: %m");
235 log_debug_errno(r
, "Failed to install filter set for architecture %s, skipping: %m", seccomp_arch_to_string(arch
));
243 int setup_seccomp(uint64_t cap_list_retain
, char **syscall_whitelist
, char **syscall_blacklist
) {