1 /* SPDX-License-Identifier: LGPL-2.1+ */
9 #include <linux/magic.h>
11 #include <netinet/in.h>
14 #include <sys/param.h>
15 #include <sys/types.h>
18 #include "attach_options.h"
24 #include "lxcseccomp.h"
25 #include "memory_utils.h"
26 #include "namespace.h"
30 #include "storage/storage.h"
31 #include "string_utils.h"
32 #include "syscall_wrappers.h"
35 #if HAVE_SYS_RESOURCE_H
36 #include <sys/resource.h>
39 #if HAVE_SCMP_FILTER_CTX
40 typedef void * scmp_filter_ctx
;
43 typedef signed long personality_t
;
45 /* worth moving to configure.ac? */
46 #define subuidfile "/etc/subuid"
47 #define subgidfile "/etc/subgid"
50 * Defines a generic struct to configure the control group. It is up to the
51 * programmer to specify the right subsystem.
52 * @subsystem : the targeted subsystem
53 * @value : the value to set
54 * @version : The version of the cgroup filesystem on which the controller
57 * @controllers : The controllers to use for this container.
58 * @dir : The name of the directory containing the container's cgroup.
59 * Not that this is a per-container setting.
63 /* information about a specific controller */
64 struct /* controller */ {
70 /* meta information about cgroup configuration */
75 char *monitor_pivot_dir
;
82 struct list_head head
;
85 static void free_lxc_cgroup(struct lxc_cgroup
*ptr
)
93 define_cleanup_function(struct lxc_cgroup
*, free_lxc_cgroup
);
95 #if !HAVE_SYS_RESOURCE_H
96 #define RLIM_INFINITY ((unsigned long)-1)
98 unsigned long rlim_cur
;
99 unsigned long rlim_max
;
100 struct list_head head
;
105 * Defines a structure to configure resource limits to set via setrlimit().
106 * @resource : the resource name in lowercase without the RLIMIT_ prefix
107 * @limit : the limit to set
112 struct list_head head
;
115 static void free_lxc_limit(struct lxc_limit
*ptr
)
118 free_disarm(ptr
->resource
);
122 define_cleanup_function(struct lxc_limit
*, free_lxc_limit
);
130 * Defines a structure to configure kernel parameters at runtime.
131 * @key : the kernel parameters will be configured without the "lxc.sysctl" prefix
132 * @value : the value to set
137 struct list_head head
;
140 static void free_lxc_sysctl(struct lxc_sysctl
*ptr
)
148 define_cleanup_function(struct lxc_sysctl
*, free_lxc_sysctl
);
151 * Defines a structure to configure proc filesystem at runtime.
152 * @filename : the proc filesystem will be configured without the "lxc.proc" prefix
153 * @value : the value to set
158 struct list_head head
;
161 static void free_lxc_proc(struct lxc_proc
*ptr
)
169 define_cleanup_function(struct lxc_proc
*, free_lxc_proc
);
172 * id_map is an id map entry. Form in confile is:
173 * lxc.idmap = u 0 9800 100
174 * lxc.idmap = u 1000 9900 100
175 * lxc.idmap = g 0 9800 100
176 * lxc.idmap = g 1000 9900 100
177 * meaning the container can use uids and gids 0-99 and 1000-1099,
178 * with [ug]id 0 mapping to [ug]id 9800 on the host, and [ug]id 1000 to
179 * [ug]id 9900 on the host.
183 unsigned long hostid
, nsid
, range
;
184 struct list_head head
;
187 /* Defines the number of tty configured and contains the
189 * @max = number of configured ttys
191 struct lxc_tty_info
{
195 struct lxc_terminal_info
*tty
;
198 typedef enum lxc_mount_options_t
{
199 LXC_MOUNT_CREATE_DIR
= 0,
200 LXC_MOUNT_CREATE_FILE
= 1,
201 LXC_MOUNT_OPTIONAL
= 2,
202 LXC_MOUNT_RELATIVE
= 3,
205 } lxc_mount_options_t
;
207 __hidden
extern const char *lxc_mount_options_info
[LXC_MOUNT_MAX
];
209 struct lxc_mount_options
{
210 unsigned int create_dir
: 1;
211 unsigned int create_file
: 1;
212 unsigned int optional
: 1;
213 unsigned int relative
: 1;
214 unsigned int bind_recursively
: 1;
215 unsigned int propagate_recursively
: 1;
216 unsigned int bind
: 1;
217 char userns_path
[PATH_MAX
];
218 unsigned long mnt_flags
;
219 unsigned long prop_flags
;
221 struct lxc_mount_attr attr
;
225 /* Defines a structure to store the rootfs location, the
226 * optionals pivot_root, rootfs mount paths
227 * @path : the rootfs source (directory or device)
228 * @mount : where it is mounted
229 * @buf : static buffer to construct paths
230 * @bev_type : optional backing store type
231 * @managed : whether it is managed by LXC
232 * @dfd_mnt : fd for @mount
233 * @dfd_dev : fd for /dev of the container
250 struct lxc_mount_options mnt_opts
;
251 struct lxc_storage
*storage
;
255 * Automatic mounts for LXC to perform inside the container
258 LXC_AUTO_PROC_RW
= 0x001, /* /proc read-write */
259 LXC_AUTO_PROC_MIXED
= 0x002, /* /proc/sys and /proc/sysrq-trigger read-only */
260 LXC_AUTO_PROC_MASK
= 0x003,
262 LXC_AUTO_SYS_RW
= 0x004, /* /sys */
263 LXC_AUTO_SYS_RO
= 0x008, /* /sys read-only */
264 LXC_AUTO_SYS_MIXED
= 0x00C, /* /sys read-only and /sys/class/net read-write */
265 LXC_AUTO_SYS_MASK
= 0x00C,
267 LXC_AUTO_CGROUP_RO
= 0x010, /* /sys/fs/cgroup (partial mount, read-only) */
268 LXC_AUTO_CGROUP_RW
= 0x020, /* /sys/fs/cgroup (partial mount, read-write) */
269 LXC_AUTO_CGROUP_MIXED
= 0x030, /* /sys/fs/cgroup (partial mount, paths r/o, cgroup r/w) */
270 LXC_AUTO_CGROUP_FULL_RO
= 0x040, /* /sys/fs/cgroup (full mount, read-only) */
271 LXC_AUTO_CGROUP_FULL_RW
= 0x050, /* /sys/fs/cgroup (full mount, read-write) */
272 LXC_AUTO_CGROUP_FULL_MIXED
= 0x060, /* /sys/fs/cgroup (full mount, parent r/o, own r/w) */
274 * These are defined in such a way as to retain binary compatibility
275 * with earlier versions of this code. If the previous mask is applied,
276 * both of these will default back to the _MIXED variants, which is
279 LXC_AUTO_CGROUP_NOSPEC
= 0x0B0, /* /sys/fs/cgroup (partial mount, r/w or mixed, depending on caps) */
280 LXC_AUTO_CGROUP_FULL_NOSPEC
= 0x0E0, /* /sys/fs/cgroup (full mount, r/w or mixed, depending on caps) */
281 LXC_AUTO_CGROUP_FORCE
= 0x100, /* mount cgroups even when cgroup namespaces are supported */
282 LXC_AUTO_CGROUP_MASK
= 0x1F0, /* all known cgroup options */
284 LXC_AUTO_SHMOUNTS
= 0x200, /* shared mount point */
285 LXC_AUTO_SHMOUNTS_MASK
= 0x200, /* shared mount point mask */
286 LXC_AUTO_ALL_MASK
= 0x1FF, /* all known settings */
303 __hidden
extern char *lxchook_names
[NUM_LXC_HOOKS
];
305 struct lxc_state_client
{
307 lxc_state_t states
[MAX_STATE
];
308 struct list_head head
;
311 typedef enum lxc_bpf_devices_rule_t
{
312 LXC_BPF_DEVICE_CGROUP_ALLOWLIST
= 0,
313 LXC_BPF_DEVICE_CGROUP_DENYLIST
= 1,
314 } lxc_bpf_devices_rule_t
;
322 struct list_head head
;
326 lxc_bpf_devices_rule_t list_type
;
327 struct list_head devices
;
330 struct timens_offsets
{
331 /* Currently, either s_boot or ns_boot is set, but not both. */
335 /* Currently, either s_monotonic or ns_monotonic is set, but not both. */
337 int64_t ns_monotonic
;
340 struct environment_entry
{
343 struct list_head head
;
349 struct list_head head
;
354 struct list_head list
;
357 struct string_entry
{
359 struct list_head head
;
363 /* Pointer to the name of the container. Do not free! */
367 personality_t personality
;
368 struct utsname
*utsname
;
371 struct list_head cgroup
;
372 struct list_head cgroup2
;
373 struct bpf_devices bpf_devices
;
377 struct list_head id_map
;
380 * Pointer to the idmap entry for the container's root uid in
381 * the id_map list. Do not free!
383 const struct id_map
*root_nsuid_map
;
386 * Pointer to the idmap entry for the container's root gid in
387 * the id_map list. Do not free!
389 const struct id_map
*root_nsgid_map
;
392 struct list_head netdevs
;
397 struct list_head mount_entries
;
402 /* /dev/tty<idx> devices */
403 struct lxc_tty_info ttys
;
404 /* /dev/console device */
405 struct lxc_terminal console
;
406 /* maximum pty devices allowed by devpts mount */
408 /* file descriptor for the container's /dev/pts mount */
411 /* set to true when rootfs has been setup */
413 struct lxc_rootfs rootfs
;
418 unsigned int hooks_version
;
419 struct lxc_list hooks
[NUM_LXC_HOOKS
];
422 char *lsm_aa_profile
;
423 char *lsm_aa_profile_computed
;
424 bool lsm_aa_profile_created
;
425 unsigned int lsm_aa_allow_nesting
;
426 unsigned int lsm_aa_allow_incomplete
;
427 struct lxc_list lsm_aa_raw
;
428 char *lsm_se_context
;
429 char *lsm_se_keyring_context
;
430 bool keyring_disable_session
;
431 bool transient_procfs_mnt
;
432 struct lxc_seccomp seccomp
;
434 unsigned int autodev
; /* if 1, mount and fill a /dev at start */
435 int autodevtmpfssize
; /* size of the /dev tmpfs */
436 int haltsignal
; /* signal used to halt container */
437 int rebootsignal
; /* signal used to reboot container */
438 int stopsignal
; /* signal used to hard stop container */
439 char *rcfile
; /* Copy of the top level rcfile we read */
441 /* Logfile and loglevel can be set in a container config file. Those
442 * function as defaults. The defaults can be overridden by command line.
443 * However we don't want the command line specified values to be saved
444 * on c->save_config(). So we store the config file specified values
446 char *logfile
; /* the logfile as specified in config */
447 int loglevel
; /* loglevel as specified in config (if any) */
450 unsigned int start_auto
;
451 unsigned int start_delay
;
453 struct lxc_list groups
;
456 /* unshare the mount namespace in the monitor */
457 unsigned int monitor_unshare
;
458 unsigned int monitor_signal_pdeath
;
460 /* list of environment variables we'll add to the container when
462 struct list_head environment
;
464 /* text representation of the config file */
465 char *unexpanded_config
;
466 size_t unexpanded_len
;
467 size_t unexpanded_alloced
;
469 /* default command for lxc-execute */
475 /* The uid to use for the container. */
477 /* The gid to use for the container. */
479 /* The groups to use for the container. */
480 lxc_groups_t init_groups
;
482 /* indicator if the container will be destroyed on shutdown */
483 unsigned int ephemeral
;
485 /* The facility to pass to syslog. Let's users establish as what type of
486 * program liblxc is supposed to write to the syslog. */
489 /* Whether PR_SET_NO_NEW_PRIVS will be set for the container. */
492 /* RLIMIT_* limits */
493 struct list_head limits
;
495 /* Contains generic info about the cgroup configuration for this
496 * container. Note that struct lxc_cgroup contains a union. It is only
497 * valid to access the members of the anonymous "meta" struct within
500 struct lxc_cgroup cgroup_meta
;
505 char *ns_share
[LXC_NS_MAX
];
508 /* init working directory */
511 /* A list of clients registered to be informed about a container state. */
512 struct list_head state_clients
;
515 struct list_head sysctls
;
518 struct list_head procs
;
521 /* Absolute path to the shared mount point on the host */
523 /* Absolute path (in the container) to the shared mount point */
527 struct timens_offsets timens
;
530 __hidden
extern int write_id_mapping(enum idtype idtype
, pid_t pid
, const char *buf
, size_t buf_size
)
533 extern thread_local
struct lxc_conf
*current_config
;
535 __hidden
extern int run_lxc_hooks(const char *name
, char *hook
, struct lxc_conf
*conf
, char *argv
[]);
536 __hidden
extern struct lxc_conf
*lxc_conf_init(void);
537 __hidden
extern void lxc_conf_free(struct lxc_conf
*conf
);
538 __hidden
extern int lxc_storage_prepare(struct lxc_conf
*conf
);
539 __hidden
extern int lxc_rootfs_prepare(struct lxc_conf
*conf
, bool userns
);
540 __hidden
extern void lxc_storage_put(struct lxc_conf
*conf
);
541 __hidden
extern int lxc_rootfs_init(struct lxc_conf
*conf
, bool userns
);
542 __hidden
extern int lxc_rootfs_prepare_parent(struct lxc_handler
*handler
);
543 __hidden
extern int lxc_idmapped_mounts_parent(struct lxc_handler
*handler
);
544 __hidden
extern int lxc_map_ids(struct list_head
*idmap
, pid_t pid
);
545 __hidden
extern int lxc_create_tty(const char *name
, struct lxc_conf
*conf
);
546 __hidden
extern void lxc_delete_tty(struct lxc_tty_info
*ttys
);
547 __hidden
extern int lxc_clear_config_caps(struct lxc_conf
*c
);
548 __hidden
extern int lxc_clear_cgroups(struct lxc_conf
*c
, const char *key
, int version
);
549 __hidden
extern int lxc_clear_mount_entries(struct lxc_conf
*c
);
550 __hidden
extern int lxc_clear_automounts(struct lxc_conf
*c
);
551 __hidden
extern int lxc_clear_hooks(struct lxc_conf
*c
, const char *key
);
552 __hidden
extern int lxc_clear_idmaps(struct lxc_conf
*c
);
553 __hidden
extern int lxc_clear_groups(struct lxc_conf
*c
);
554 __hidden
extern int lxc_clear_environment(struct lxc_conf
*c
);
555 __hidden
extern int lxc_clear_limits(struct lxc_conf
*c
, const char *key
);
556 __hidden
extern int lxc_delete_autodev(struct lxc_handler
*handler
);
557 __hidden
extern int lxc_clear_autodev_tmpfs_size(struct lxc_conf
*c
);
558 __hidden
extern int lxc_setup_rootfs_prepare_root(struct lxc_conf
*conf
, const char *name
,
559 const char *lxcpath
);
560 __hidden
extern int lxc_setup(struct lxc_handler
*handler
);
561 __hidden
extern int lxc_setup_parent(struct lxc_handler
*handler
);
562 __hidden
extern int setup_resource_limits(struct lxc_conf
*conf
, pid_t pid
);
563 __hidden
extern int find_unmapped_nsid(const struct lxc_conf
*conf
, enum idtype idtype
);
564 __hidden
extern int mapped_hostid(unsigned id
, const struct lxc_conf
*conf
, enum idtype idtype
);
565 __hidden
extern int userns_exec_1(const struct lxc_conf
*conf
, int (*fn
)(void *), void *data
,
566 const char *fn_name
);
567 __hidden
extern int userns_exec_full(struct lxc_conf
*conf
, int (*fn
)(void *), void *data
,
568 const char *fn_name
);
569 __hidden
extern int parse_mntopts_legacy(const char *mntopts
, unsigned long *mntflags
, char **mntdata
);
570 __hidden
extern int parse_propagationopts(const char *mntopts
, unsigned long *pflags
);
571 __hidden
extern int parse_lxc_mount_attrs(struct lxc_mount_options
*opts
, char *mnt_opts
);
572 __hidden
extern int parse_mount_attrs(struct lxc_mount_options
*opts
, const char *mntopts
);
573 __hidden
extern void tmp_proc_unmount(struct lxc_conf
*lxc_conf
);
574 __hidden
extern void suggest_default_idmap(void);
575 __hidden
extern FILE *make_anonymous_mount_file(const struct list_head
*mount
,
576 bool include_nesting_helpers
);
577 __hidden
extern int run_script(const char *name
, const char *section
, const char *script
, ...);
578 __hidden
extern int run_script_argv(const char *name
, unsigned int hook_version
, const char *section
,
579 const char *script
, const char *hookname
, char **argsin
);
581 __hidden
extern bool has_cap(int cap
, struct lxc_conf
*conf
);
582 static inline bool lxc_wants_cap(int cap
, struct lxc_conf
*conf
)
584 if (lxc_caps_last_cap() < cap
)
587 return has_cap(cap
, conf
);
590 __hidden
extern int setup_sysctl_parameters(struct lxc_conf
*conf
);
591 __hidden
extern int lxc_clear_sysctls(struct lxc_conf
*c
, const char *key
);
592 __hidden
extern int setup_proc_filesystem(struct lxc_conf
*conf
, pid_t pid
);
593 __hidden
extern int lxc_clear_procs(struct lxc_conf
*c
, const char *key
);
594 __hidden
extern int lxc_clear_apparmor_raw(struct lxc_conf
*c
);
595 __hidden
extern int lxc_clear_namespace(struct lxc_conf
*c
);
596 __hidden
extern int userns_exec_minimal(const struct lxc_conf
*conf
, int (*fn_parent
)(void *),
597 void *fn_parent_data
, int (*fn_child
)(void *),
598 void *fn_child_data
);
599 __hidden
extern int userns_exec_mapped_root(const char *path
, int path_fd
,
600 const struct lxc_conf
*conf
);
601 static inline int chown_mapped_root(const char *path
, const struct lxc_conf
*conf
)
603 return userns_exec_mapped_root(path
, -EBADF
, conf
);
606 __hidden
extern int lxc_sync_fds_parent(struct lxc_handler
*handler
);
607 __hidden
extern int lxc_sync_fds_child(struct lxc_handler
*handler
);
609 static inline const char *get_rootfs_mnt(const struct lxc_rootfs
*rootfs
)
611 static const char *s
= "/";
613 return !is_empty_string(rootfs
->path
) ? rootfs
->mount
: s
;
616 static inline void put_lxc_mount_options(struct lxc_mount_options
*mnt_opts
)
618 mnt_opts
->create_dir
= 0;
619 mnt_opts
->create_file
= 0;
620 mnt_opts
->optional
= 0;
621 mnt_opts
->relative
= 0;
622 mnt_opts
->userns_path
[0] = '\0';
623 mnt_opts
->mnt_flags
= 0;
624 mnt_opts
->prop_flags
= 0;
626 free_disarm(mnt_opts
->data
);
627 free_disarm(mnt_opts
->raw_options
);
630 static inline void put_lxc_rootfs(struct lxc_rootfs
*rootfs
, bool unpin
)
633 close_prot_errno_disarm(rootfs
->dfd_host
);
634 close_prot_errno_disarm(rootfs
->dfd_mnt
);
635 close_prot_errno_disarm(rootfs
->dfd_dev
);
637 close_prot_errno_disarm(rootfs
->fd_path_pin
);
638 close_prot_errno_disarm(rootfs
->dfd_idmapped
);
639 put_lxc_mount_options(&rootfs
->mnt_opts
);
640 storage_put(rootfs
->storage
);
641 rootfs
->storage
= NULL
;
645 static inline void lxc_clear_cgroup2_devices(struct bpf_devices
*bpf_devices
)
647 struct device_item
*device
, *n
;
649 list_for_each_entry_safe(device
, n
, &bpf_devices
->devices
, head
)
650 list_del(&device
->head
);
652 INIT_LIST_HEAD(&bpf_devices
->devices
);
655 static inline int lxc_personality(personality_t persona
)
658 return ret_errno(EINVAL
);
660 return personality(persona
);
663 __hidden
extern int lxc_set_environment(const struct lxc_conf
*conf
);
664 __hidden
extern int parse_cap(const char *cap
);
666 #endif /* __LXC_CONF_H */