1 /*#############################################################################
3 # Pakfire - The IPFire package management system #
4 # Copyright (C) 2022 Pakfire development team #
6 # This program is free software: you can redistribute it and/or modify #
7 # it under the terms of the GNU General Public License as published by #
8 # the Free Software Foundation, either version 3 of the License, or #
9 # (at your option) any later version. #
11 # This program is distributed in the hope that it will be useful, #
12 # but WITHOUT ANY WARRANTY; without even the implied warranty of #
13 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the #
14 # GNU General Public License for more details. #
16 # You should have received a copy of the GNU General Public License #
17 # along with this program. If not, see <http://www.gnu.org/licenses/>. #
19 #############################################################################*/
23 #include <linux/capability.h>
24 #include <linux/sched.h>
26 #include <linux/wait.h>
31 #include <sys/capability.h>
32 #include <sys/epoll.h>
33 #include <sys/eventfd.h>
34 #include <sys/mount.h>
35 #include <sys/personality.h>
36 #include <sys/prctl.h>
37 #include <sys/resource.h>
38 #include <sys/timerfd.h>
39 #include <sys/types.h>
45 #include <netlink/route/link.h>
53 #include <pakfire/arch.h>
54 #include <pakfire/cgroup.h>
55 #include <pakfire/jail.h>
56 #include <pakfire/logging.h>
57 #include <pakfire/mount.h>
58 #include <pakfire/pakfire.h>
59 #include <pakfire/path.h>
60 #include <pakfire/private.h>
61 #include <pakfire/pwd.h>
62 #include <pakfire/string.h>
63 #include <pakfire/util.h>
65 #define BUFFER_SIZE 1024 * 64
66 #define ENVIRON_SIZE 128
67 #define EPOLL_MAX_EVENTS 2
68 #define MAX_MOUNTPOINTS 8
70 // The default environment that will be set for every command
71 static const struct environ
{
76 { "LANG", "C.utf-8" },
77 { "PATH", "/usr/local/sbin:/usr/sbin:/sbin:/usr/local/bin:/usr/bin:/bin", },
80 // Tell everything that it is running inside a Pakfire container
81 { "container", "pakfire" },
85 struct pakfire_jail_mountpoint
{
86 char source
[PATH_MAX
];
87 char target
[PATH_MAX
];
92 struct pakfire_ctx
* ctx
;
93 struct pakfire
* pakfire
;
96 // A unique ID for each jail
98 char __uuid
[UUID_STR_LEN
];
104 struct itimerspec timeout
;
107 struct pakfire_cgroup
* cgroup
;
110 char* env
[ENVIRON_SIZE
];
113 struct pakfire_jail_mountpoint mountpoints
[MAX_MOUNTPOINTS
];
114 unsigned int num_mountpoints
;
117 struct pakfire_jail_callbacks
{
119 pakfire_jail_log_callback log
;
124 struct pakfire_log_buffer
{
125 char data
[BUFFER_SIZE
];
129 struct pakfire_jail_exec
{
132 // PID (of the child)
136 // Socket to pass FDs
139 // Process status (from waitid)
142 // FD to notify the client that the parent has finished initialization
146 struct pakfire_jail_pipes
{
152 #endif /* ENABLE_DEBUG */
156 struct pakfire_jail_communicate
{
157 pakfire_jail_communicate_in in
;
158 pakfire_jail_communicate_out out
;
163 struct pakfire_jail_buffers
{
164 struct pakfire_log_buffer stdout
;
165 struct pakfire_log_buffer stderr
;
168 struct pakfire_log_buffer log_INFO
;
169 struct pakfire_log_buffer log_ERROR
;
171 struct pakfire_log_buffer log_DEBUG
;
172 #endif /* ENABLE_DEBUG */
175 struct pakfire_cgroup
* cgroup
;
176 struct pakfire_cgroup_stats cgroup_stats
;
179 struct pakfire_jail_pty
{
180 // The path to the console
181 char console
[PATH_MAX
];
184 struct pakfire_jail_pty_master
{
187 enum pakfire_jail_pty_flags
{
188 PAKFIRE_JAIL_PTY_READY_TO_READ
= (1 << 0),
189 PAKFIRE_JAIL_PTY_READY_TO_WRITE
= (1 << 1),
194 struct pakfire_jail_pty_stdio
{
196 struct pakfire_log_buffer buffer
;
197 struct termios attrs
;
199 enum pakfire_jail_pty_flags flags
;
203 struct pakfire_jail_pty_stdio stdout
;
207 static int clone3(struct clone_args
* args
, size_t size
) {
208 return syscall(__NR_clone3
, args
, size
);
211 static int pidfd_send_signal(int pidfd
, int sig
, siginfo_t
* info
, unsigned int flags
) {
212 return syscall(SYS_pidfd_send_signal
, pidfd
, sig
, info
, flags
);
215 static int pivot_root(const char* new_root
, const char* old_root
) {
216 return syscall(SYS_pivot_root
, new_root
, old_root
);
219 static int pakfire_jail_exec_has_flag(
220 const struct pakfire_jail_exec
* ctx
, const enum pakfire_jail_exec_flags flag
) {
221 return ctx
->flags
& flag
;
224 static void pakfire_jail_free(struct pakfire_jail
* jail
) {
225 DEBUG(jail
->pakfire
, "Freeing jail at %p\n", jail
);
228 for (unsigned int i
= 0; jail
->env
[i
]; i
++)
232 pakfire_cgroup_unref(jail
->cgroup
);
234 pakfire_unref(jail
->pakfire
);
236 pakfire_ctx_unref(jail
->ctx
);
241 Passes any log messages on to the default pakfire log callback
243 static int pakfire_jail_default_log_callback(struct pakfire
* pakfire
, void* data
,
244 int priority
, const char* line
, size_t length
) {
247 INFO(pakfire
, "%s", line
);
251 ERROR(pakfire
, "%s", line
);
256 DEBUG(pakfire
, "%s", line
);
264 static const char* pakfire_jail_uuid(struct pakfire_jail
* jail
) {
266 uuid_unparse_lower(jail
->uuid
, jail
->__uuid
);
271 static int pakfire_jail_setup_interactive_env(struct pakfire_jail
* jail
) {
273 int r
= pakfire_jail_set_env(jail
, "PS1", "pakfire-jail \\w> ");
278 char* TERM
= secure_getenv("TERM");
280 r
= pakfire_jail_set_env(jail
, "TERM", TERM
);
286 char* LANG
= secure_getenv("LANG");
288 r
= pakfire_jail_set_env(jail
, "LANG", LANG
);
296 PAKFIRE_EXPORT
int pakfire_jail_create(struct pakfire_jail
** jail
, struct pakfire
* pakfire
) {
299 const char* arch
= pakfire_get_effective_arch(pakfire
);
301 // Allocate a new jail
302 struct pakfire_jail
* j
= calloc(1, sizeof(*j
));
307 j
->ctx
= pakfire_ctx(pakfire
);
310 j
->pakfire
= pakfire_ref(pakfire
);
312 // Initialize reference counter
315 // Generate a random UUID
316 uuid_generate_random(j
->uuid
);
318 DEBUG(j
->pakfire
, "Allocated new jail at %p\n", j
);
320 // Set the default logging callback
321 pakfire_jail_set_log_callback(j
, pakfire_jail_default_log_callback
, NULL
);
323 // Set default environment
324 for (const struct environ
* e
= ENV
; e
->key
; e
++) {
325 r
= pakfire_jail_set_env(j
, e
->key
, e
->val
);
330 // Enable all CPU features that CPU has to offer
331 if (!pakfire_arch_is_supported_by_host(arch
)) {
332 r
= pakfire_jail_set_env(j
, "QEMU_CPU", "max");
337 // Set container UUID
338 r
= pakfire_jail_set_env(j
, "container_uuid", pakfire_jail_uuid(j
));
342 // Disable systemctl to talk to systemd
343 if (!pakfire_on_root(j
->pakfire
)) {
344 r
= pakfire_jail_set_env(j
, "SYSTEMD_OFFLINE", "1");
354 pakfire_jail_free(j
);
359 PAKFIRE_EXPORT
struct pakfire_jail
* pakfire_jail_ref(struct pakfire_jail
* jail
) {
365 PAKFIRE_EXPORT
struct pakfire_jail
* pakfire_jail_unref(struct pakfire_jail
* jail
) {
366 if (--jail
->nrefs
> 0)
369 pakfire_jail_free(jail
);
375 PAKFIRE_EXPORT
void pakfire_jail_set_log_callback(struct pakfire_jail
* jail
,
376 pakfire_jail_log_callback callback
, void* data
) {
377 jail
->callbacks
.log
= callback
;
378 jail
->callbacks
.log_data
= data
;
383 PAKFIRE_EXPORT
int pakfire_jail_nice(struct pakfire_jail
* jail
, int nice
) {
384 // Check if nice level is in range
385 if (nice
< -19 || nice
> 20) {
396 int pakfire_jail_set_cgroup(struct pakfire_jail
* jail
, struct pakfire_cgroup
* cgroup
) {
397 // Free any previous cgroup
399 pakfire_cgroup_unref(jail
->cgroup
);
403 // Set any new cgroup
405 DEBUG(jail
->pakfire
, "Setting cgroup %p\n", cgroup
);
407 jail
->cgroup
= pakfire_cgroup_ref(cgroup
);
416 // Returns the length of the environment
417 static unsigned int pakfire_jail_env_length(struct pakfire_jail
* jail
) {
420 // Count everything in the environment
421 for (char** e
= jail
->env
; *e
; e
++)
427 // Finds an existing environment variable and returns its index or -1 if not found
428 static int pakfire_jail_find_env(struct pakfire_jail
* jail
, const char* key
) {
434 const size_t length
= strlen(key
);
436 for (unsigned int i
= 0; jail
->env
[i
]; i
++) {
437 if ((pakfire_string_startswith(jail
->env
[i
], key
)
438 && *(jail
->env
[i
] + length
) == '=')) {
447 // Returns the value of an environment variable or NULL
448 PAKFIRE_EXPORT
const char* pakfire_jail_get_env(struct pakfire_jail
* jail
,
450 int i
= pakfire_jail_find_env(jail
, key
);
454 return jail
->env
[i
] + strlen(key
) + 1;
457 // Sets an environment variable
458 PAKFIRE_EXPORT
int pakfire_jail_set_env(struct pakfire_jail
* jail
,
459 const char* key
, const char* value
) {
460 // Find the index where to write this value to
461 int i
= pakfire_jail_find_env(jail
, key
);
463 i
= pakfire_jail_env_length(jail
);
465 // Return -ENOSPC when the environment is full
466 if (i
>= ENVIRON_SIZE
) {
471 // Free any previous value
475 // Format and set environment variable
476 asprintf(&jail
->env
[i
], "%s=%s", key
, value
);
478 DEBUG(jail
->pakfire
, "Set environment variable: %s\n", jail
->env
[i
]);
483 // Imports an environment
484 PAKFIRE_EXPORT
int pakfire_jail_import_env(struct pakfire_jail
* jail
, const char* env
[]) {
492 // Copy environment variables
493 for (unsigned int i
= 0; env
[i
]; i
++) {
494 r
= pakfire_string_partition(env
[i
], "=", &key
, &val
);
499 r
= pakfire_jail_set_env(jail
, key
, val
);
516 PAKFIRE_EXPORT
int pakfire_jail_set_timeout(
517 struct pakfire_jail
* jail
, unsigned int timeout
) {
519 jail
->timeout
.it_value
.tv_sec
= timeout
;
522 DEBUG(jail
->pakfire
, "Timeout set to %u second(s)\n", timeout
);
524 DEBUG(jail
->pakfire
, "Timeout disabled\n");
529 static int pakfire_jail_create_timer(struct pakfire_jail
* jail
) {
532 // Nothing to do if no timeout has been set
533 if (!jail
->timeout
.it_value
.tv_sec
)
536 // Create a new timer
537 const int fd
= timerfd_create(CLOCK_MONOTONIC
, 0);
539 ERROR(jail
->pakfire
, "Could not create timer: %m\n");
544 r
= timerfd_settime(fd
, 0, &jail
->timeout
, NULL
);
546 ERROR(jail
->pakfire
, "Could not arm timer: %m\n");
560 This function replaces any logging in the child process.
562 All log messages will be sent to the parent process through their respective pipes.
564 static void pakfire_jail_log_redirect(void* data
, int priority
, const char* file
,
565 int line
, const char* fn
, const char* format
, va_list args
) {
566 struct pakfire_jail_pipes
* pipes
= (struct pakfire_jail_pipes
*)data
;
571 fd
= pipes
->log_INFO
[1];
575 fd
= pipes
->log_ERROR
[1];
580 fd
= pipes
->log_DEBUG
[1];
582 #endif /* ENABLE_DEBUG */
584 // Ignore any messages of an unknown priority
589 // Send the log message
591 vdprintf(fd
, format
, args
);
594 static int pakfire_jail_log_buffer_is_full(const struct pakfire_log_buffer
* buffer
) {
595 return (sizeof(buffer
->data
) == buffer
->used
);
598 static int pakfire_jail_fill_buffer(struct pakfire_jail
* jail
, int fd
, struct pakfire_log_buffer
* buffer
) {
601 // Skip this if there is not space left in the buffer
602 if (buffer
->used
>= sizeof(buffer
->data
))
606 r
= read(fd
, buffer
->data
+ buffer
->used
, sizeof(buffer
->data
) - buffer
->used
);
621 // XXX What to do here?
631 static int pakfire_jail_drain_buffer_with_callback(struct pakfire_jail
* jail
,
632 struct pakfire_log_buffer
* buffer
, int priority
, pakfire_jail_communicate_out callback
, void* data
) {
633 const char* eol
= NULL
;
636 while (buffer
->used
) {
637 // Search for the end of the first line
638 eol
= memchr(buffer
->data
, '\n', buffer
->used
);
642 // If the buffer is full, we send the entire content to make space.
643 if (pakfire_jail_log_buffer_is_full(buffer
)) {
644 CTX_DEBUG(jail
->ctx
, "Buffer is full. Sending all content\n");
646 eol
= buffer
->data
+ buffer
->used
- 1;
648 // Otherwise we might have only read parts of the output...
654 // Find the length of the string
655 const size_t length
= eol
- buffer
->data
+ 1;
658 r
= callback(jail
->pakfire
, data
, priority
, buffer
->data
, length
);
660 CTX_ERROR(jail
->ctx
, "The logging callback returned an error: %d\n", r
);
664 // Remove line from buffer
665 memmove(buffer
->data
, buffer
->data
+ length
, buffer
->used
- length
);
666 buffer
->used
-= length
;
672 static int pakfire_jail_drain_buffer(struct pakfire_jail
* jail
, int fd
, struct pakfire_log_buffer
* buffer
) {
675 // Nothing to do if the buffer is empty
680 r
= write(fd
, buffer
->data
, buffer
->used
);
695 memmove(buffer
->data
, buffer
->data
+ r
, buffer
->used
- r
);
704 This function reads as much data as it can from the file descriptor.
705 If it finds a whole line in it, it will send it to the logger and repeat the process.
706 If not newline character is found, it will try to read more data until it finds one.
708 static int pakfire_jail_handle_log(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
,
709 int priority
, int fd
, struct pakfire_log_buffer
* buffer
,
710 pakfire_jail_communicate_out callback
, void* data
) {
713 // Fill up buffer from fd
714 r
= pakfire_jail_fill_buffer(jail
, fd
, buffer
);
719 r
= pakfire_jail_drain_buffer_with_callback(jail
, buffer
, priority
, callback
, data
);
727 static int pakfire_jail_stream_stdin(struct pakfire_jail
* jail
,
728 struct pakfire_jail_exec
* ctx
, const int fd
) {
731 // Nothing to do if there is no stdin callback set
732 if (!ctx
->communicate
.in
) {
733 DEBUG(jail
->pakfire
, "Callback for standard input is not set\n");
737 // Skip if the writing pipe has already been closed
738 if (ctx
->pipes
.stdin
[1] < 0)
741 DEBUG(jail
->pakfire
, "Streaming standard input...\n");
743 // Calling the callback
744 r
= ctx
->communicate
.in(jail
->pakfire
, ctx
->communicate
.data
, fd
);
746 DEBUG(jail
->pakfire
, "Standard input callback finished: %d\n", r
);
748 // The callback signaled that it has written everything
750 DEBUG(jail
->pakfire
, "Closing standard input pipe\n");
752 // Close the file-descriptor
755 // Reset the file-descriptor so it won't be closed again later
756 ctx
->pipes
.stdin
[1] = -1;
766 static int pakfire_jail_recv_fd(struct pakfire_jail
* jail
, int socket
, int* fd
) {
767 const size_t payload_length
= sizeof(fd
);
768 char buffer
[CMSG_SPACE(payload_length
)];
771 struct msghdr msg
= {
772 .msg_control
= buffer
,
773 .msg_controllen
= sizeof(buffer
),
776 // Receive the message
777 r
= recvmsg(socket
, &msg
, 0);
779 CTX_ERROR(jail
->ctx
, "Could not receive file descriptor: %s\n", strerror(errno
));
784 struct cmsghdr
* cmsg
= CMSG_FIRSTHDR(&msg
);
788 *fd
= *((int*)CMSG_DATA(cmsg
));
790 CTX_DEBUG(jail
->ctx
, "Received fd %d from socket %d\n", *fd
, socket
);
795 static int pakfire_jail_send_fd(struct pakfire_jail
* jail
, int socket
, int fd
) {
796 const size_t payload_length
= sizeof(fd
);
797 char buffer
[CMSG_SPACE(payload_length
)];
800 CTX_DEBUG(jail
->ctx
, "Sending fd %d to socket %d\n", fd
, socket
);
803 struct msghdr msg
= {
804 .msg_control
= buffer
,
805 .msg_controllen
= sizeof(buffer
),
809 struct cmsghdr
* cmsg
= CMSG_FIRSTHDR(&msg
);
810 cmsg
->cmsg_level
= SOL_SOCKET
;
811 cmsg
->cmsg_type
= SCM_RIGHTS
;
812 cmsg
->cmsg_len
= CMSG_LEN(payload_length
);
815 *((int*)CMSG_DATA(cmsg
)) = fd
;
818 r
= sendmsg(socket
, &msg
, 0);
820 CTX_ERROR(jail
->ctx
, "Could not send file descriptor: %s\n", strerror(errno
));
827 static int pakfire_jail_setup_pipe(struct pakfire_jail
* jail
, int (*fds
)[2], const int flags
) {
828 int r
= pipe2(*fds
, flags
);
830 ERROR(jail
->pakfire
, "Could not setup pipe: %m\n");
837 static void pakfire_jail_close_pipe(struct pakfire_jail
* jail
, int fds
[2]) {
838 for (unsigned int i
= 0; i
< 2; i
++)
844 This is a convenience function to fetch the reading end of a pipe and
845 closes the write end.
847 static int pakfire_jail_get_pipe_to_read(struct pakfire_jail
* jail
, int (*fds
)[2]) {
848 // Give the variables easier names to avoid confusion
849 int* fd_read
= &(*fds
)[0];
850 int* fd_write
= &(*fds
)[1];
852 // Close the write end of the pipe
853 if (*fd_write
>= 0) {
858 // Return the read end
865 static int pakfire_jail_get_pipe_to_write(struct pakfire_jail
* jail
, int (*fds
)[2]) {
866 // Give the variables easier names to avoid confusion
867 int* fd_read
= &(*fds
)[0];
868 int* fd_write
= &(*fds
)[1];
870 // Close the read end of the pipe
876 // Return the write end
883 static int pakfire_jail_log(struct pakfire
* pakfire
, void* data
, int priority
,
884 const char* line
, const size_t length
) {
885 // Pass everything to the parent logger
886 pakfire_log_condition(pakfire
, priority
, 0, "%.*s", (int)length
, line
);
891 static int pakfire_jail_epoll_add_fd(struct pakfire_jail
* jail
, int epollfd
, int fd
, int events
) {
892 struct epoll_event event
= {
893 .events
= events
|EPOLLHUP
,
901 int flags
= fcntl(fd
, F_GETFL
, 0);
903 // Set modified flags
904 r
= fcntl(fd
, F_SETFL
, flags
|O_NONBLOCK
);
906 CTX_ERROR(jail
->ctx
, "Could not set file descriptor %d into non-blocking mode: %s\n",
907 fd
, strerror(errno
));
911 // Add the file descriptor to the loop
912 r
= epoll_ctl(epollfd
, EPOLL_CTL_ADD
, fd
, &event
);
914 ERROR(jail
->pakfire
, "Could not add file descriptor %d to epoll(): %s\n",
915 fd
, strerror(errno
));
924 static int pakfire_jail_enable_raw_mode(struct pakfire_jail
* jail
,
925 struct pakfire_jail_pty_stdio
* stdio
) {
926 struct termios raw_attrs
;
929 // Skip everything if fd is not a TTY
930 if (!isatty(stdio
->fd
))
934 stdio
->fdflags
= fcntl(stdio
->fd
, F_GETFL
);
935 if (stdio
->fdflags
< 0) {
936 CTX_ERROR(jail
->ctx
, "Could not fetch flags from fd %d: %s\n",
937 stdio
->fd
, strerror(errno
));
941 // Fetch all attributes
942 r
= tcgetattr(stdio
->fd
, &stdio
->attrs
);
944 CTX_ERROR(jail
->ctx
, "Could not fetch terminal attributes from fd %d: %s\n",
945 stdio
->fd
, strerror(errno
));
949 // Copy all attributes
950 raw_attrs
= stdio
->attrs
;
953 cfmakeraw(&raw_attrs
);
957 raw_attrs
.c_oflag
= stdio
->attrs
.c_oflag
;
961 raw_attrs
.c_iflag
= stdio
->attrs
.c_iflag
;
962 raw_attrs
.c_lflag
= stdio
->attrs
.c_lflag
;
966 // Restore the attributes
967 r
= tcsetattr(stdio
->fd
, TCSANOW
, &raw_attrs
);
969 CTX_ERROR(jail
->ctx
, "Could not restore terminal attributes for fd %d: %s\n",
970 stdio
->fd
, strerror(errno
));
977 static int pakfire_jail_restore_attrs(struct pakfire_jail
* jail
,
978 const struct pakfire_jail_pty_stdio
* stdio
) {
981 // Skip everything if fd is not a TTY
982 if (!isatty(stdio
->fd
))
986 r
= fcntl(stdio
->fd
, F_SETFL
, stdio
->fdflags
);
988 CTX_ERROR(jail
->ctx
, "Could not set flags for file descriptor %d: %s\n",
989 stdio
->fd
, strerror(errno
));
993 // Restore the attributes
994 r
= tcsetattr(stdio
->fd
, TCSANOW
, &stdio
->attrs
);
996 CTX_ERROR(jail
->ctx
, "Could not restore terminal attributes for %d, ignoring: %s\n",
997 stdio
->fd
, strerror(errno
));
1004 static int pakfire_jail_setup_pty_forwarding(struct pakfire_jail
* jail
,
1005 struct pakfire_jail_exec
* ctx
, const int epollfd
, const int fd
) {
1006 struct winsize size
;
1009 CTX_DEBUG(jail
->ctx
, "Setting up PTY forwarding on fd %d\n", fd
);
1011 // Store the file descriptor
1012 ctx
->pty
.master
.fd
= fd
;
1014 // Configure stdin/stdout
1015 ctx
->pty
.stdin
.fd
= STDIN_FILENO
;
1016 ctx
->pty
.stdout
.fd
= STDOUT_FILENO
;
1019 if (isatty(ctx
->pty
.stdout
.fd
)) {
1020 r
= ioctl(ctx
->pty
.stdout
.fd
, TIOCGWINSZ
, &size
);
1022 CTX_ERROR(jail
->ctx
, "Failed to determine terminal dimensions: %s\n", strerror(errno
));
1027 r
= ioctl(ctx
->pty
.master
.fd
, TIOCSWINSZ
, &size
);
1029 CTX_ERROR(jail
->ctx
, "Failed setting dimensions: %s\n", strerror(errno
));
1034 // Enable RAW mode on standard input
1035 r
= pakfire_jail_enable_raw_mode(jail
, &ctx
->pty
.stdin
);
1039 // Enable RAW mode on standard output
1040 r
= pakfire_jail_enable_raw_mode(jail
, &ctx
->pty
.stdout
);
1044 // Add the master to the event loop
1045 r
= pakfire_jail_epoll_add_fd(jail
, epollfd
, ctx
->pty
.master
.fd
, EPOLLIN
|EPOLLOUT
|EPOLLET
);
1049 // Add standard input to the event loop
1050 r
= pakfire_jail_epoll_add_fd(jail
, epollfd
, ctx
->pty
.stdin
.fd
, EPOLLIN
|EPOLLET
);
1054 // Add standard output to the event loop
1055 r
= pakfire_jail_epoll_add_fd(jail
, epollfd
, ctx
->pty
.stdout
.fd
, EPOLLOUT
|EPOLLET
);
1062 static int pakfire_jail_forward_pty(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
) {
1065 // Read from standard input
1066 if (ctx
->pty
.stdin
.flags
& PAKFIRE_JAIL_PTY_READY_TO_READ
) {
1067 r
= pakfire_jail_fill_buffer(jail
, ctx
->pty
.stdin
.fd
, &ctx
->pty
.stdin
.buffer
);
1069 CTX_ERROR(jail
->ctx
, "Failed reading from standard input: %s\n", strerror(-r
));
1073 // We are done reading for now
1074 ctx
->pty
.stdin
.flags
&= ~PAKFIRE_JAIL_PTY_READY_TO_READ
;
1076 // But we may have data to write
1077 if (ctx
->pty
.stdin
.buffer
.used
)
1078 ctx
->pty
.master
.flags
|= PAKFIRE_JAIL_PTY_READY_TO_WRITE
;
1081 // Write to the master
1082 if (ctx
->pty
.master
.flags
& PAKFIRE_JAIL_PTY_READY_TO_WRITE
) {
1083 r
= pakfire_jail_drain_buffer(jail
, ctx
->pty
.master
.fd
, &ctx
->pty
.stdin
.buffer
);
1085 CTX_ERROR(jail
->ctx
, "Failed writing to the PTY: %s\n", strerror(-r
));
1089 // We are done writing for now
1090 ctx
->pty
.master
.flags
&= ~PAKFIRE_JAIL_PTY_READY_TO_WRITE
;
1093 // Read from the master
1094 if (ctx
->pty
.master
.flags
& PAKFIRE_JAIL_PTY_READY_TO_READ
) {
1095 r
= pakfire_jail_fill_buffer(jail
, ctx
->pty
.master
.fd
, &ctx
->pty
.stdout
.buffer
);
1097 CTX_ERROR(jail
->ctx
, "Failed reading from the PTY: %s\n", strerror(-r
));
1101 // We are done reading for now
1102 ctx
->pty
.master
.flags
&= ~PAKFIRE_JAIL_PTY_READY_TO_READ
;
1104 // But we may have data to write
1105 if (ctx
->pty
.stdout
.buffer
.used
)
1106 ctx
->pty
.stdout
.flags
|= PAKFIRE_JAIL_PTY_READY_TO_WRITE
;
1109 // Write to standard output
1110 if (ctx
->pty
.stdout
.flags
& PAKFIRE_JAIL_PTY_READY_TO_WRITE
) {
1111 r
= pakfire_jail_drain_buffer(jail
, ctx
->pty
.stdout
.fd
, &ctx
->pty
.stdout
.buffer
);
1113 CTX_ERROR(jail
->ctx
, "Failed writing to standard output: %s\n", strerror(-r
));
1117 // We are done writing for now
1118 ctx
->pty
.stdout
.flags
&= ~PAKFIRE_JAIL_PTY_READY_TO_WRITE
;
1124 static int pakfire_jail_wait(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
) {
1126 struct epoll_event events
[EPOLL_MAX_EVENTS
];
1130 // Fetch file descriptors from context
1131 const int pidfd
= ctx
->pidfd
;
1133 // Fetch the UNIX domain socket
1134 const int socket_recv
= pakfire_jail_get_pipe_to_read(jail
, &ctx
->socket
);
1137 const int timerfd
= pakfire_jail_create_timer(jail
);
1140 const int log_INFO
= pakfire_jail_get_pipe_to_read(jail
, &ctx
->pipes
.log_INFO
);
1141 const int log_ERROR
= pakfire_jail_get_pipe_to_read(jail
, &ctx
->pipes
.log_ERROR
);
1143 const int log_DEBUG
= pakfire_jail_get_pipe_to_read(jail
, &ctx
->pipes
.log_DEBUG
);
1144 #endif /* ENABLE_DEBUG */
1146 // Make a list of all file descriptors we are interested in
1147 const struct pakfire_wait_fds
{
1152 { timerfd
, EPOLLIN
},
1155 { ctx
->pidfd
, EPOLLIN
},
1158 { log_INFO
, EPOLLIN
},
1159 { log_ERROR
, EPOLLIN
},
1161 { log_DEBUG
, EPOLLIN
},
1162 #endif /* ENABLE_DEBUG */
1164 // UNIX Domain Socket
1165 { socket_recv
, EPOLLIN
},
1172 epollfd
= epoll_create1(0);
1174 ERROR(jail
->pakfire
, "Could not initialize epoll(): %m\n");
1179 // Turn file descriptors into non-blocking mode and add them to epoll()
1180 for (const struct pakfire_wait_fds
* fd
= fds
; fd
->events
; fd
++) {
1181 // Skip fds which were not initialized
1185 // Add the FD to the event loop
1186 r
= pakfire_jail_epoll_add_fd(jail
, epollfd
, fd
->fd
, fd
->events
);
1193 // Loop for as long as the process is alive
1195 int num
= epoll_wait(epollfd
, events
, EPOLL_MAX_EVENTS
, -1);
1197 // Ignore if epoll_wait() has been interrupted
1201 ERROR(jail
->pakfire
, "epoll_wait() failed: %m\n");
1207 for (int i
= 0; i
< num
; i
++) {
1208 int e
= events
[i
].events
;
1209 int fd
= events
[i
].data
.fd
;
1211 // Handle PTY forwarding events
1212 if (ctx
->pty
.master
.fd
== fd
) {
1213 if (e
& (EPOLLIN
|EPOLLHUP
))
1214 ctx
->pty
.master
.flags
|= PAKFIRE_JAIL_PTY_READY_TO_READ
;
1216 if (e
& (EPOLLOUT
|EPOLLHUP
))
1217 ctx
->pty
.master
.flags
|= PAKFIRE_JAIL_PTY_READY_TO_WRITE
;
1220 r
= pakfire_jail_forward_pty(jail
, ctx
);
1222 CTX_ERROR(jail
->ctx
, "Failed forwarding the PTY: %s\n", strerror(-r
));
1226 // Handle standard input
1227 } else if (ctx
->pty
.stdin
.fd
== fd
) {
1228 if (e
& (EPOLLIN
|EPOLLHUP
))
1229 ctx
->pty
.stdin
.flags
|= PAKFIRE_JAIL_PTY_READY_TO_READ
;
1232 r
= pakfire_jail_forward_pty(jail
, ctx
);
1234 CTX_ERROR(jail
->ctx
, "Failed forwarding the PTY: %s\n", strerror(-r
));
1238 // Handle standard output
1239 } else if (ctx
->pty
.stdout
.fd
== fd
) {
1240 if (e
& (EPOLLOUT
|EPOLLHUP
))
1241 ctx
->pty
.stdout
.flags
|= PAKFIRE_JAIL_PTY_READY_TO_WRITE
;
1244 r
= pakfire_jail_forward_pty(jail
, ctx
);
1246 CTX_ERROR(jail
->ctx
, "Failed forwarding the PTY: %s\n", strerror(-r
));
1250 // Handle any changes to the PIDFD
1251 } else if (pidfd
== fd
) {
1253 // Call waidid() and store the result
1254 r
= waitid(P_PIDFD
, ctx
->pidfd
, &ctx
->status
, WEXITED
);
1256 ERROR(jail
->pakfire
, "waitid() failed: %m\n");
1260 // Mark that we have ended so that we will process the remaining
1261 // events from epoll() now, but won't restart the outer loop.
1265 // Handle timer events
1266 } else if (timerfd
== fd
) {
1268 DEBUG(jail
->pakfire
, "Timer event received\n");
1271 r
= read(timerfd
, garbage
, sizeof(garbage
));
1273 ERROR(jail
->pakfire
, "Could not disarm timer: %m\n");
1278 // Terminate the process if it hasn't already ended
1280 DEBUG(jail
->pakfire
, "Terminating process...\n");
1282 // Send SIGTERM to the process
1283 r
= pidfd_send_signal(pidfd
, SIGKILL
, NULL
, 0);
1285 ERROR(jail
->pakfire
, "Could not kill process: %m\n");
1291 // Handle socket messages
1292 } else if (socket_recv
== fd
) {
1294 // Receive the passed FD
1295 r
= pakfire_jail_recv_fd(jail
, socket_recv
, &fd
);
1299 // Setup PTY forwarding
1300 if (ctx
->pty
.master
.fd
< 0) {
1301 r
= pakfire_jail_setup_pty_forwarding(jail
, ctx
, epollfd
, fd
);
1303 CTX_ERROR(jail
->ctx
, "Failed setting up PTY forwarding: %s\n", strerror(-r
));
1309 // Handle log INFO messages
1310 } else if (log_INFO
== fd
) {
1312 r
= pakfire_jail_handle_log(jail
, ctx
, LOG_INFO
, fd
,
1313 &ctx
->buffers
.log_INFO
, pakfire_jail_log
, NULL
);
1318 // Handle log ERROR messages
1319 } else if (log_ERROR
== fd
) {
1321 r
= pakfire_jail_handle_log(jail
, ctx
, LOG_ERR
, fd
,
1322 &ctx
->buffers
.log_ERROR
, pakfire_jail_log
, NULL
);
1328 // Handle log DEBUG messages
1329 } else if (log_DEBUG
== fd
) {
1331 r
= pakfire_jail_handle_log(jail
, ctx
, LOG_DEBUG
, fd
,
1332 &ctx
->buffers
.log_DEBUG
, pakfire_jail_log
, NULL
);
1336 #endif /* ENABLE_DEBUG */
1338 // Log a message for anything else
1340 DEBUG(jail
->pakfire
, "Received invalid file descriptor %d\n", fd
);
1344 // Check if any file descriptors have been closed
1346 // Remove the file descriptor
1347 r
= epoll_ctl(epollfd
, EPOLL_CTL_DEL
, fd
, NULL
);
1349 ERROR(jail
->pakfire
, "Could not remove closed file-descriptor %d: %m\n", fd
);
1362 // Restore any changed terminal attributes
1363 if (ctx
->pty
.stdin
.fd
>= 0)
1364 pakfire_jail_restore_attrs(jail
, &ctx
->pty
.stdin
);
1365 if (ctx
->pty
.stdout
.fd
>= 0)
1366 pakfire_jail_restore_attrs(jail
, &ctx
->pty
.stdout
);
1371 int pakfire_jail_capture_stdout(struct pakfire
* pakfire
, void* data
,
1372 int priority
, const char* line
, size_t length
) {
1373 char** output
= (char**)data
;
1376 // Append everything from stdout to a buffer
1377 if (output
&& priority
== LOG_INFO
) {
1378 r
= asprintf(output
, "%s%s", (output
&& *output
) ? *output
: "", line
);
1384 // Send everything else to the default logger
1385 return pakfire_jail_default_log_callback(pakfire
, NULL
, priority
, line
, length
);
1390 // Logs all capabilities of the current process
1391 static int pakfire_jail_show_capabilities(struct pakfire_jail
* jail
) {
1394 cap_flag_value_t value_e
;
1395 cap_flag_value_t value_i
;
1396 cap_flag_value_t value_p
;
1400 pid_t pid
= getpid();
1402 // Fetch all capabilities
1403 caps
= cap_get_proc();
1405 ERROR(jail
->pakfire
, "Could not fetch capabilities: %m\n");
1410 DEBUG(jail
->pakfire
, "Capabilities of PID %d:\n", pid
);
1412 // Iterate over all capabilities
1413 for (unsigned int cap
= 0; cap_valid(cap
); cap
++) {
1414 name
= cap_to_name(cap
);
1416 // Fetch effective value
1417 r
= cap_get_flag(caps
, cap
, CAP_EFFECTIVE
, &value_e
);
1421 // Fetch inheritable value
1422 r
= cap_get_flag(caps
, cap
, CAP_INHERITABLE
, &value_i
);
1426 // Fetch permitted value
1427 r
= cap_get_flag(caps
, cap
, CAP_PERMITTED
, &value_p
);
1431 DEBUG(jail
->pakfire
,
1432 " %-24s : %c%c%c\n",
1434 (value_e
== CAP_SET
) ? 'e' : '-',
1435 (value_i
== CAP_SET
) ? 'i' : '-',
1436 (value_p
== CAP_SET
) ? 'p' : '-'
1456 static int pakfire_jail_set_capabilities(struct pakfire_jail
* jail
) {
1461 // Fetch capabilities
1462 caps
= cap_get_proc();
1464 ERROR(jail
->pakfire
, "Could not read capabilities: %m\n");
1469 // Walk through all capabilities
1470 for (cap_value_t cap
= 0; cap_valid(cap
); cap
++) {
1471 cap_value_t _caps
[] = { cap
};
1473 // Fetch the name of the capability
1474 name
= cap_to_name(cap
);
1476 r
= cap_set_flag(caps
, CAP_EFFECTIVE
, 1, _caps
, CAP_SET
);
1478 ERROR(jail
->pakfire
, "Could not set %s: %m\n", name
);
1482 r
= cap_set_flag(caps
, CAP_INHERITABLE
, 1, _caps
, CAP_SET
);
1484 ERROR(jail
->pakfire
, "Could not set %s: %m\n", name
);
1488 r
= cap_set_flag(caps
, CAP_PERMITTED
, 1, _caps
, CAP_SET
);
1490 ERROR(jail
->pakfire
, "Could not set %s: %m\n", name
);
1499 // Restore all capabilities
1500 r
= cap_set_proc(caps
);
1502 ERROR(jail
->pakfire
, "Restoring capabilities failed: %m\n");
1506 // Add all capabilities to the ambient set
1507 for (unsigned int cap
= 0; cap_valid(cap
); cap
++) {
1508 name
= cap_to_name(cap
);
1510 // Raise the capability
1511 r
= prctl(PR_CAP_AMBIENT
, PR_CAP_AMBIENT_RAISE
, cap
, 0, 0);
1513 ERROR(jail
->pakfire
, "Could not set ambient capability %s: %m\n", name
);
1536 static int pakfire_jail_limit_syscalls(struct pakfire_jail
* jail
) {
1537 const int syscalls
[] = {
1538 // The kernel's keyring isn't namespaced
1541 SCMP_SYS(request_key
),
1543 // Disable userfaultfd
1544 SCMP_SYS(userfaultfd
),
1546 // Disable perf which could leak a lot of information about the host
1547 SCMP_SYS(perf_event_open
),
1553 DEBUG(jail
->pakfire
, "Applying syscall filter...\n");
1555 // Setup a syscall filter which allows everything by default
1556 scmp_filter_ctx ctx
= seccomp_init(SCMP_ACT_ALLOW
);
1558 ERROR(jail
->pakfire
, "Could not setup seccomp filter: %m\n");
1563 for (const int* syscall
= syscalls
; *syscall
; syscall
++) {
1564 r
= seccomp_rule_add(ctx
, SCMP_ACT_ERRNO(EPERM
), *syscall
, 0);
1566 ERROR(jail
->pakfire
, "Could not configure syscall %d: %m\n", *syscall
);
1571 // Load syscall filter into the kernel
1572 r
= seccomp_load(ctx
);
1574 ERROR(jail
->pakfire
, "Could not load syscall filter into the kernel: %m\n");
1580 seccomp_release(ctx
);
1587 PAKFIRE_EXPORT
int pakfire_jail_bind(struct pakfire_jail
* jail
,
1588 const char* source
, const char* target
, int flags
) {
1589 struct pakfire_jail_mountpoint
* mp
= NULL
;
1592 // Check if there is any space left
1593 if (jail
->num_mountpoints
>= MAX_MOUNTPOINTS
) {
1598 // Check for valid inputs
1599 if (!source
|| !target
) {
1604 // Select the next free slot
1605 mp
= &jail
->mountpoints
[jail
->num_mountpoints
];
1608 r
= pakfire_string_set(mp
->source
, source
);
1610 ERROR(jail
->pakfire
, "Could not copy source: %m\n");
1615 r
= pakfire_string_set(mp
->target
, target
);
1617 ERROR(jail
->pakfire
, "Could not copy target: %m\n");
1624 // Increment counter
1625 jail
->num_mountpoints
++;
1630 static int pakfire_jail_mount_networking(struct pakfire_jail
* jail
) {
1633 const char* paths
[] = {
1639 // Bind-mount all paths read-only
1640 for (const char** path
= paths
; *path
; path
++) {
1641 r
= pakfire_bind(jail
->pakfire
, *path
, NULL
, MS_RDONLY
);
1644 // Ignore if we don't have permission
1659 Mounts everything that we require in the new namespace
1661 static int pakfire_jail_mount(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
) {
1662 struct pakfire_jail_mountpoint
* mp
= NULL
;
1666 // Enable loop devices
1667 if (pakfire_jail_exec_has_flag(ctx
, PAKFIRE_JAIL_HAS_LOOP_DEVICES
))
1668 flags
|= PAKFIRE_MOUNT_LOOP_DEVICES
;
1670 // Mount all default stuff
1671 r
= pakfire_mount_all(jail
->pakfire
, PAKFIRE_MNTNS_OUTER
, flags
);
1676 r
= pakfire_populate_dev(jail
->pakfire
, flags
);
1680 // Mount the interpreter (if needed)
1681 r
= pakfire_mount_interpreter(jail
->pakfire
);
1685 // Mount networking stuff
1686 if (pakfire_jail_exec_has_flag(ctx
, PAKFIRE_JAIL_HAS_NETWORKING
)) {
1687 r
= pakfire_jail_mount_networking(jail
);
1692 // Mount all custom stuff
1693 for (unsigned int i
= 0; i
< jail
->num_mountpoints
; i
++) {
1695 mp
= &jail
->mountpoints
[i
];
1698 r
= pakfire_bind(jail
->pakfire
, mp
->source
, mp
->target
, mp
->flags
);
1708 static int pakfire_jail_setup_loopback(struct pakfire_jail
* jail
) {
1709 struct nl_sock
* nl
= NULL
;
1710 struct nl_cache
* cache
= NULL
;
1711 struct rtnl_link
* link
= NULL
;
1712 struct rtnl_link
* change
= NULL
;
1715 DEBUG(jail
->pakfire
, "Setting up loopback...\n");
1717 // Allocate a netlink socket
1718 nl
= nl_socket_alloc();
1720 ERROR(jail
->pakfire
, "Could not allocate a netlink socket: %m\n");
1725 // Connect the socket
1726 r
= nl_connect(nl
, NETLINK_ROUTE
);
1728 ERROR(jail
->pakfire
, "Could not connect netlink socket: %s\n", nl_geterror(r
));
1732 // Allocate the netlink cache
1733 r
= rtnl_link_alloc_cache(nl
, AF_UNSPEC
, &cache
);
1735 ERROR(jail
->pakfire
, "Unable to allocate netlink cache: %s\n", nl_geterror(r
));
1739 // Fetch loopback interface
1740 link
= rtnl_link_get_by_name(cache
, "lo");
1742 ERROR(jail
->pakfire
, "Could not find lo interface. Ignoring.\n");
1747 // Allocate a new link
1748 change
= rtnl_link_alloc();
1750 ERROR(jail
->pakfire
, "Could not allocate change link\n");
1755 // Set the link to UP
1756 rtnl_link_set_flags(change
, IFF_UP
);
1758 // Apply any changes
1759 r
= rtnl_link_change(nl
, link
, change
, 0);
1761 ERROR(jail
->pakfire
, "Unable to activate loopback: %s\n", nl_geterror(r
));
1777 static int pakfire_jail_setup_uid_mapping(struct pakfire_jail
* jail
, pid_t pid
) {
1778 char path
[PATH_MAX
];
1781 // Skip mapping anything when running on /
1782 if (pakfire_on_root(jail
->pakfire
))
1786 r
= pakfire_string_format(path
, "/proc/%d/uid_map", pid
);
1791 const uid_t uid
= pakfire_uid(jail
->pakfire
);
1794 const struct pakfire_subid
* subuid
= pakfire_subuid(jail
->pakfire
);
1798 /* When running as root, we will map the entire range.
1800 When running as a non-privileged user, we will map the root user inside the jail
1801 to the user's UID outside of the jail, and we will map the rest starting from one.
1806 r
= pakfire_file_write(jail
->pakfire
, path
, 0, 0, 0,
1807 "0 %lu %lu\n", subuid
->id
, subuid
->length
);
1809 r
= pakfire_file_write(jail
->pakfire
, path
, 0, 0, 0,
1810 "0 %lu 1\n1 %lu %lu\n", uid
, subuid
->id
, subuid
->length
);
1814 ERROR(jail
->pakfire
, "Could not map UIDs: %m\n");
1821 static int pakfire_jail_setup_gid_mapping(struct pakfire_jail
* jail
, pid_t pid
) {
1822 char path
[PATH_MAX
];
1825 // Skip mapping anything when running on /
1826 if (pakfire_on_root(jail
->pakfire
))
1830 const gid_t gid
= pakfire_gid(jail
->pakfire
);
1833 const struct pakfire_subid
* subgid
= pakfire_subgid(jail
->pakfire
);
1838 r
= pakfire_string_format(path
, "/proc/%d/gid_map", pid
);
1844 r
= pakfire_file_write(jail
->pakfire
, path
, 0, 0, 0,
1845 "0 %lu %lu\n", subgid
->id
, subgid
->length
);
1847 r
= pakfire_file_write(jail
->pakfire
, path
, 0, 0, 0,
1848 "0 %lu 1\n1 %lu %lu\n", gid
, subgid
->id
, subgid
->length
);
1852 ERROR(jail
->pakfire
, "Could not map GIDs: %m\n");
1859 static int pakfire_jail_setgroups(struct pakfire_jail
* jail
, pid_t pid
) {
1860 char path
[PATH_MAX
];
1864 r
= pakfire_string_format(path
, "/proc/%d/setgroups", pid
);
1868 r
= pakfire_file_write(jail
->pakfire
, path
, 0, 0, 0, "deny\n");
1870 CTX_ERROR(jail
->ctx
, "Could not set setgroups to deny: %s\n", strerror(errno
));
1877 static int pakfire_jail_send_signal(struct pakfire_jail
* jail
, int fd
) {
1878 const uint64_t val
= 1;
1881 DEBUG(jail
->pakfire
, "Sending signal...\n");
1883 // Write to the file descriptor
1884 r
= eventfd_write(fd
, val
);
1886 ERROR(jail
->pakfire
, "Could not send signal: %s\n", strerror(errno
));
1890 // Close the file descriptor
1896 static int pakfire_jail_wait_for_signal(struct pakfire_jail
* jail
, int fd
) {
1900 DEBUG(jail
->pakfire
, "Waiting for signal...\n");
1902 r
= eventfd_read(fd
, &val
);
1904 ERROR(jail
->pakfire
, "Error waiting for signal: %s\n", strerror(errno
));
1908 // Close the file descriptor
1915 Performs the initialisation that needs to happen in the parent part
1917 static int pakfire_jail_parent(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
) {
1920 // Setup UID mapping
1921 r
= pakfire_jail_setup_uid_mapping(jail
, ctx
->pid
);
1925 // Write "deny" to /proc/PID/setgroups
1926 r
= pakfire_jail_setgroups(jail
, ctx
->pid
);
1930 // Setup GID mapping
1931 r
= pakfire_jail_setup_gid_mapping(jail
, ctx
->pid
);
1935 // Parent has finished initialisation
1936 DEBUG(jail
->pakfire
, "Parent has finished initialization\n");
1938 // Send signal to client
1939 r
= pakfire_jail_send_signal(jail
, ctx
->completed_fd
);
1946 static int pakfire_jail_switch_root(struct pakfire_jail
* jail
, const char* root
) {
1949 // Change to the new root
1952 ERROR(jail
->pakfire
, "chdir(%s) failed: %m\n", root
);
1957 r
= pivot_root(".", ".");
1959 ERROR(jail
->pakfire
, "Failed changing into the new root directory %s: %m\n", root
);
1963 // Umount the old root
1964 r
= umount2(".", MNT_DETACH
);
1966 ERROR(jail
->pakfire
, "Could not umount the old root filesystem: %m\n");
1973 static int pakfire_jail_open_pty(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
) {
1976 // Allocate a new PTY
1977 ctx
->pty
.master
.fd
= posix_openpt(O_RDWR
|O_NONBLOCK
|O_NOCTTY
|O_CLOEXEC
);
1978 if (ctx
->pty
.master
.fd
< 0)
1982 r
= ptsname_r(ctx
->pty
.master
.fd
, ctx
->pty
.console
, sizeof(ctx
->pty
.console
));
1986 CTX_DEBUG(jail
->ctx
, "Allocated console at %s (%d)\n", ctx
->pty
.console
, ctx
->pty
.master
.fd
);
1988 // Unlock the master device
1989 r
= unlockpt(ctx
->pty
.master
.fd
);
1991 CTX_ERROR(jail
->ctx
, "Could not unlock the PTY: %s\n", strerror(errno
));
1996 r
= pakfire_symlink(jail
->ctx
, ctx
->pty
.console
, "/dev/console");
2003 static int pakfire_jail_setup_terminal(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
) {
2007 // Open a new terminal
2008 fd
= open("/dev/console", O_RDWR
|O_NOCTTY
);
2010 CTX_ERROR(jail
->ctx
, "Failed to open a new terminal: %s\n", strerror(errno
));
2014 CTX_DEBUG(jail
->ctx
, "Opened a new terminal %d\n", fd
);
2016 // Connect the new terminal to standard input
2017 r
= dup2(fd
, STDIN_FILENO
);
2019 CTX_ERROR(jail
->ctx
, "Failed to open standard input: %s\n", strerror(errno
));
2023 // Connect the new terminal to standard output
2024 r
= dup2(fd
, STDOUT_FILENO
);
2026 CTX_ERROR(jail
->ctx
, "Failed to open standard output: %s\n", strerror(errno
));
2030 // Connect the new terminal to standard error
2031 r
= dup2(fd
, STDERR_FILENO
);
2033 CTX_ERROR(jail
->ctx
, "Failed to open standard error: %s\n", strerror(errno
));
2040 static int pakfire_jail_child(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
,
2041 const char* argv
[]) {
2044 // Redirect any logging to our log pipe
2045 pakfire_ctx_set_log_callback(jail
->ctx
, pakfire_jail_log_redirect
, &ctx
->pipes
);
2048 pid_t pid
= getpid();
2050 DEBUG(jail
->pakfire
, "Launched child process in jail with PID %d\n", pid
);
2052 // Wait for the parent to finish initialization
2053 r
= pakfire_jail_wait_for_signal(jail
, ctx
->completed_fd
);
2058 r
= prctl(PR_SET_PDEATHSIG
, SIGKILL
, 0, 0, 0);
2060 ERROR(jail
->pakfire
, "Could not configure to die with parent: %m\n");
2064 // Make this process dumpable
2065 r
= prctl (PR_SET_DUMPABLE
, 1, 0, 0, 0);
2067 ERROR(jail
->pakfire
, "Could not make the process dumpable: %m\n");
2071 // Don't drop any capabilities on setuid()
2072 r
= prctl(PR_SET_KEEPCAPS
, 1);
2074 ERROR(jail
->pakfire
, "Could not set PR_SET_KEEPCAPS: %m\n");
2079 uid_t uid
= getuid();
2080 gid_t gid
= getgid();
2083 uid_t euid
= geteuid();
2084 gid_t egid
= getegid();
2086 DEBUG(jail
->pakfire
, " UID: %u (effective %u)\n", uid
, euid
);
2087 DEBUG(jail
->pakfire
, " GID: %u (effective %u)\n", gid
, egid
);
2089 // Log all mountpoints
2090 pakfire_mount_list(jail
->ctx
);
2092 // Fail if we are not PID 1
2094 CTX_ERROR(jail
->ctx
, "Child process is not PID 1\n");
2098 // Fail if we are not running as root
2099 if (uid
|| gid
|| euid
|| egid
) {
2100 ERROR(jail
->pakfire
, "Child process is not running as root\n");
2104 const int socket_send
= pakfire_jail_get_pipe_to_write(jail
, &ctx
->socket
);
2106 // Mount all default stuff
2107 r
= pakfire_mount_all(jail
->pakfire
, PAKFIRE_MNTNS_INNER
, 0);
2111 const char* root
= pakfire_get_path(jail
->pakfire
);
2112 const char* arch
= pakfire_get_effective_arch(jail
->pakfire
);
2114 // Change mount propagation to slave to receive anything from the parent namespace
2115 r
= pakfire_mount_change_propagation(jail
->ctx
, "/", MS_SLAVE
);
2119 // Make root a mountpoint in the new mount namespace
2120 r
= pakfire_mount_make_mounpoint(jail
->pakfire
, root
);
2124 // Change mount propagation to private
2125 r
= pakfire_mount_change_propagation(jail
->ctx
, root
, MS_PRIVATE
);
2129 // Change root (unless root is /)
2130 if (!pakfire_on_root(jail
->pakfire
)) {
2132 r
= pakfire_jail_mount(jail
, ctx
);
2137 r
= pakfire_jail_switch_root(jail
, root
);
2143 unsigned long persona
= pakfire_arch_personality(arch
);
2145 r
= personality(persona
);
2147 ERROR(jail
->pakfire
, "Could not set personality (%x)\n", (unsigned int)persona
);
2153 if (!pakfire_jail_exec_has_flag(ctx
, PAKFIRE_JAIL_HAS_NETWORKING
)) {
2154 r
= pakfire_jail_setup_loopback(jail
);
2161 DEBUG(jail
->pakfire
, "Setting nice level to %d\n", jail
->nice
);
2163 r
= setpriority(PRIO_PROCESS
, pid
, jail
->nice
);
2165 ERROR(jail
->pakfire
, "Could not set nice level: %m\n");
2170 // Create a new session
2173 CTX_ERROR(jail
->ctx
, "Could not create a new session: %s\n", strerror(errno
));
2177 // Allocate a new PTY
2178 r
= pakfire_jail_open_pty(jail
, ctx
);
2180 CTX_ERROR(jail
->ctx
, "Could not allocate a new PTY: %s\n", strerror(-r
));
2184 // Send the PTY master to the parent process
2185 r
= pakfire_jail_send_fd(jail
, socket_send
, ctx
->pty
.master
.fd
);
2187 CTX_ERROR(jail
->ctx
, "Failed sending the PTY master to the parent: %s\n", strerror(-r
));
2191 // Setup the terminal
2192 r
= pakfire_jail_setup_terminal(jail
, ctx
);
2196 // Close the master of the PTY
2197 close(ctx
->pty
.master
.fd
);
2198 ctx
->pty
.master
.fd
= -1;
2203 // Close other end of log pipes
2204 close(ctx
->pipes
.log_INFO
[0]);
2205 close(ctx
->pipes
.log_ERROR
[0]);
2207 close(ctx
->pipes
.log_DEBUG
[0]);
2208 #endif /* ENABLE_DEBUG */
2210 // Reset open file limit (http://0pointer.net/blog/file-descriptor-limits.html)
2211 r
= pakfire_rlimit_reset_nofile(jail
->pakfire
);
2216 r
= pakfire_jail_set_capabilities(jail
);
2220 // Show capabilities
2221 r
= pakfire_jail_show_capabilities(jail
);
2226 r
= pakfire_jail_limit_syscalls(jail
);
2230 DEBUG(jail
->pakfire
, "Child process initialization done\n");
2231 DEBUG(jail
->pakfire
, "Launching command:\n");
2234 for (unsigned int i
= 0; argv
[i
]; i
++)
2235 DEBUG(jail
->pakfire
, " argv[%u] = %s\n", i
, argv
[i
]);
2238 r
= execvpe(argv
[0], (char**)argv
, jail
->env
);
2240 // Translate errno into regular exit code
2243 // Ignore if the command doesn't exist
2244 if (ctx
->flags
& PAKFIRE_JAIL_NOENT_OK
)
2255 ERROR(jail
->pakfire
, "Could not execve(%s): %m\n", argv
[0]);
2258 // We should not get here
2262 // Run a command in the jail
2263 static int __pakfire_jail_exec(struct pakfire_jail
* jail
, const char* argv
[],
2264 pakfire_jail_communicate_in communicate_in
,
2265 pakfire_jail_communicate_out communicate_out
,
2266 void* data
, int flags
) {
2270 // Check if argv is valid
2271 if (!argv
|| !argv
[0]) {
2276 // Initialize context for this call
2277 struct pakfire_jail_exec ctx
= {
2280 .socket
= { -1, -1 },
2283 .log_INFO
= { -1, -1 },
2284 .log_ERROR
= { -1, -1 },
2286 .log_DEBUG
= { -1, -1 },
2287 #endif /* ENABLE_DEBUG */
2291 .in
= communicate_in
,
2292 .out
= communicate_out
,
2312 DEBUG(jail
->pakfire
, "Executing jail...\n");
2314 // Enable networking in interactive mode
2315 if (ctx
.flags
& PAKFIRE_JAIL_PTY_FORWARDING
)
2316 ctx
.flags
|= PAKFIRE_JAIL_HAS_NETWORKING
;
2319 Setup a file descriptor which can be used to notify the client that the parent
2320 has completed configuration.
2322 ctx
.completed_fd
= eventfd(0, EFD_CLOEXEC
);
2323 if (ctx
.completed_fd
< 0) {
2324 ERROR(jail
->pakfire
, "eventfd() failed: %m\n");
2328 // Create a UNIX domain socket
2329 r
= socketpair(AF_UNIX
, SOCK_DGRAM
|SOCK_CLOEXEC
, 0, ctx
.socket
);
2331 CTX_ERROR(jail
->ctx
, "Could not create UNIX socket: %s\n", strerror(errno
));
2336 // Setup pipes for logging
2338 r
= pakfire_jail_setup_pipe(jail
, &ctx
.pipes
.log_INFO
, O_CLOEXEC
);
2343 r
= pakfire_jail_setup_pipe(jail
, &ctx
.pipes
.log_ERROR
, O_CLOEXEC
);
2349 r
= pakfire_jail_setup_pipe(jail
, &ctx
.pipes
.log_DEBUG
, O_CLOEXEC
);
2352 #endif /* ENABLE_DEBUG */
2354 // Configure child process
2355 struct clone_args args
= {
2365 .exit_signal
= SIGCHLD
,
2366 .pidfd
= (long long unsigned int)&ctx
.pidfd
,
2369 // Launch the process in a cgroup that is a leaf of the configured cgroup
2371 args
.flags
|= CLONE_INTO_CGROUP
;
2374 const char* uuid
= pakfire_jail_uuid(jail
);
2376 // Create a temporary cgroup
2377 r
= pakfire_cgroup_child(&ctx
.cgroup
, jail
->cgroup
, uuid
, 0);
2379 ERROR(jail
->pakfire
, "Could not create cgroup for jail: %m\n");
2383 // Clone into this cgroup
2384 args
.cgroup
= pakfire_cgroup_fd(ctx
.cgroup
);
2388 if (!pakfire_jail_exec_has_flag(&ctx
, PAKFIRE_JAIL_HAS_NETWORKING
)) {
2389 args
.flags
|= CLONE_NEWNET
;
2392 // Fork this process
2393 ctx
.pid
= clone3(&args
, sizeof(args
));
2395 ERROR(jail
->pakfire
, "Could not clone: %m\n");
2399 } else if (ctx
.pid
== 0) {
2400 r
= pakfire_jail_child(jail
, &ctx
, argv
);
2405 r
= pakfire_jail_parent(jail
, &ctx
);
2409 DEBUG(jail
->pakfire
, "Waiting for PID %d to finish its work\n", ctx
.pid
);
2411 // Read output of the child process
2412 r
= pakfire_jail_wait(jail
, &ctx
);
2416 // Handle exit status
2417 switch (ctx
.status
.si_code
) {
2419 DEBUG(jail
->pakfire
, "The child process exited with code %d\n",
2420 ctx
.status
.si_status
);
2423 exit
= ctx
.status
.si_status
;
2427 ERROR(jail
->pakfire
, "The child process was killed\n");
2432 ERROR(jail
->pakfire
, "The child process terminated abnormally\n");
2435 // Log anything else
2437 ERROR(jail
->pakfire
, "Unknown child exit code: %d\n", ctx
.status
.si_code
);
2442 // Destroy the temporary cgroup (if any)
2444 // Read cgroup stats
2445 pakfire_cgroup_stat(ctx
.cgroup
, &ctx
.cgroup_stats
);
2446 pakfire_cgroup_stat_dump(ctx
.cgroup
, &ctx
.cgroup_stats
);
2447 pakfire_cgroup_destroy(ctx
.cgroup
);
2448 pakfire_cgroup_unref(ctx
.cgroup
);
2451 // Close any file descriptors
2454 if (ctx
.pty
.master
.fd
>= 0)
2455 close(ctx
.pty
.master
.fd
);
2456 pakfire_jail_close_pipe(jail
, ctx
.pipes
.log_INFO
);
2457 pakfire_jail_close_pipe(jail
, ctx
.pipes
.log_ERROR
);
2459 pakfire_jail_close_pipe(jail
, ctx
.pipes
.log_DEBUG
);
2460 #endif /* ENABLE_DEBUG */
2461 pakfire_jail_close_pipe(jail
, ctx
.socket
);
2466 PAKFIRE_EXPORT
int pakfire_jail_exec(
2467 struct pakfire_jail
* jail
,
2469 pakfire_jail_communicate_in callback_in
,
2470 pakfire_jail_communicate_out callback_out
,
2471 void* data
, int flags
) {
2472 return __pakfire_jail_exec(jail
, argv
, callback_in
, callback_out
, data
, flags
);
2475 static int pakfire_jail_exec_interactive(
2476 struct pakfire_jail
* jail
, const char* argv
[], int flags
) {
2479 flags
|= PAKFIRE_JAIL_PTY_FORWARDING
;
2481 // Setup interactive stuff
2482 r
= pakfire_jail_setup_interactive_env(jail
);
2486 return __pakfire_jail_exec(jail
, argv
, NULL
, NULL
, NULL
, flags
);
2489 int pakfire_jail_exec_script(struct pakfire_jail
* jail
,
2493 pakfire_jail_communicate_in callback_in
,
2494 pakfire_jail_communicate_out callback_out
,
2496 char path
[PATH_MAX
];
2497 const char** argv
= NULL
;
2501 const char* root
= pakfire_get_path(jail
->pakfire
);
2503 // Write the scriptlet to disk
2504 r
= pakfire_path_append(path
, root
, PAKFIRE_TMP_DIR
"/pakfire-script.XXXXXX");
2508 // Create a temporary file
2509 f
= pakfire_mktemp(path
, 0700);
2511 ERROR(jail
->pakfire
, "Could not create temporary file: %m\n");
2515 DEBUG(jail
->pakfire
, "Writing script to %s:\n%.*s\n", path
, (int)size
, script
);
2518 r
= fprintf(f
, "%s", script
);
2520 ERROR(jail
->pakfire
, "Could not write script to file %s: %m\n", path
);
2527 ERROR(jail
->pakfire
, "Could not close script file %s: %m\n", path
);
2533 // Count how many arguments were passed
2534 unsigned int argc
= 1;
2536 for (const char** arg
= args
; *arg
; arg
++)
2540 argv
= calloc(argc
+ 1, sizeof(*argv
));
2542 ERROR(jail
->pakfire
, "Could not allocate argv: %m\n");
2547 argv
[0] = (root
) ? pakfire_path_relpath(root
, path
) : path
;
2550 for (unsigned int i
= 1; i
< argc
; i
++)
2551 argv
[i
] = args
[i
-1];
2554 r
= pakfire_jail_exec(jail
, argv
, callback_in
, callback_out
, data
, 0);
2562 // Remove script from disk
2570 A convenience function that creates a new jail, runs the given command and destroys
2573 int pakfire_jail_run(struct pakfire
* pakfire
, const char* argv
[], int flags
, char** output
) {
2574 struct pakfire_jail
* jail
= NULL
;
2577 // Create a new jail
2578 r
= pakfire_jail_create(&jail
, pakfire
);
2582 // Execute the command
2583 r
= pakfire_jail_exec(jail
, argv
, NULL
, pakfire_jail_capture_stdout
, output
, 0);
2587 pakfire_jail_unref(jail
);
2592 int pakfire_jail_run_script(struct pakfire
* pakfire
,
2593 const char* script
, const size_t length
, const char* argv
[], int flags
) {
2594 struct pakfire_jail
* jail
= NULL
;
2597 // Create a new jail
2598 r
= pakfire_jail_create(&jail
, pakfire
);
2602 // Execute the command
2603 r
= pakfire_jail_exec_script(jail
, script
, length
, argv
, NULL
, NULL
, NULL
);
2607 pakfire_jail_unref(jail
);
2612 int pakfire_jail_shell(struct pakfire_jail
* jail
) {
2615 const char* argv
[] = {
2616 "/bin/bash", "--login", NULL
,
2619 // Execute /bin/bash
2620 r
= pakfire_jail_exec_interactive(jail
, argv
, 0);
2626 // Ignore any return codes from the shell
2630 static int pakfire_jail_run_if_possible(struct pakfire
* pakfire
, const char** argv
) {
2631 char path
[PATH_MAX
];
2634 r
= pakfire_path(pakfire
, path
, "%s", *argv
);
2638 // Check if the file is executable
2639 r
= access(path
, X_OK
);
2641 DEBUG(pakfire
, "%s is not executable. Skipping...\n", *argv
);
2645 return pakfire_jail_run(pakfire
, argv
, 0, NULL
);
2648 int pakfire_jail_ldconfig(struct pakfire
* pakfire
) {
2649 const char* argv
[] = {
2654 return pakfire_jail_run_if_possible(pakfire
, argv
);
2657 int pakfire_jail_run_systemd_tmpfiles(struct pakfire
* pakfire
) {
2658 const char* argv
[] = {
2659 "/usr/bin/systemd-tmpfiles",
2664 return pakfire_jail_run_if_possible(pakfire
, argv
);