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
{
156 #endif /* ENABLE_DEBUG */
160 struct pakfire_jail_communicate
{
161 pakfire_jail_communicate_in in
;
162 pakfire_jail_communicate_out out
;
167 struct pakfire_jail_buffers
{
168 struct pakfire_log_buffer stdout
;
169 struct pakfire_log_buffer stderr
;
172 struct pakfire_log_buffer log_INFO
;
173 struct pakfire_log_buffer log_ERROR
;
175 struct pakfire_log_buffer log_DEBUG
;
176 #endif /* ENABLE_DEBUG */
179 struct pakfire_cgroup
* cgroup
;
180 struct pakfire_cgroup_stats cgroup_stats
;
183 struct pakfire_jail_pty
{
184 // The path to the console
185 char console
[PATH_MAX
];
188 struct pakfire_jail_pty_master
{
191 enum pakfire_jail_pty_flags
{
192 PAKFIRE_JAIL_PTY_READY_TO_READ
= (1 << 0),
193 PAKFIRE_JAIL_PTY_READY_TO_WRITE
= (1 << 1),
198 struct pakfire_jail_pty_stdio
{
200 char buffer
[LINE_MAX
];
201 struct termios attrs
;
203 enum pakfire_jail_pty_flags flags
;
207 struct pakfire_jail_pty_stdio stdout
;
211 static int clone3(struct clone_args
* args
, size_t size
) {
212 return syscall(__NR_clone3
, args
, size
);
215 static int pidfd_send_signal(int pidfd
, int sig
, siginfo_t
* info
, unsigned int flags
) {
216 return syscall(SYS_pidfd_send_signal
, pidfd
, sig
, info
, flags
);
219 static int pivot_root(const char* new_root
, const char* old_root
) {
220 return syscall(SYS_pivot_root
, new_root
, old_root
);
223 static int pakfire_jail_exec_has_flag(
224 const struct pakfire_jail_exec
* ctx
, const enum pakfire_jail_exec_flags flag
) {
225 return ctx
->flags
& flag
;
228 static void pakfire_jail_free(struct pakfire_jail
* jail
) {
229 DEBUG(jail
->pakfire
, "Freeing jail at %p\n", jail
);
232 for (unsigned int i
= 0; jail
->env
[i
]; i
++)
236 pakfire_cgroup_unref(jail
->cgroup
);
238 pakfire_unref(jail
->pakfire
);
240 pakfire_ctx_unref(jail
->ctx
);
245 Passes any log messages on to the default pakfire log callback
247 static int pakfire_jail_default_log_callback(struct pakfire
* pakfire
, void* data
,
248 int priority
, const char* line
, size_t length
) {
251 INFO(pakfire
, "%s", line
);
255 ERROR(pakfire
, "%s", line
);
260 DEBUG(pakfire
, "%s", line
);
268 static const char* pakfire_jail_uuid(struct pakfire_jail
* jail
) {
270 uuid_unparse_lower(jail
->uuid
, jail
->__uuid
);
275 static int pakfire_jail_setup_interactive_env(struct pakfire_jail
* jail
) {
277 int r
= pakfire_jail_set_env(jail
, "PS1", "pakfire-jail \\w> ");
282 char* TERM
= secure_getenv("TERM");
284 r
= pakfire_jail_set_env(jail
, "TERM", TERM
);
290 char* LANG
= secure_getenv("LANG");
292 r
= pakfire_jail_set_env(jail
, "LANG", LANG
);
300 PAKFIRE_EXPORT
int pakfire_jail_create(struct pakfire_jail
** jail
, struct pakfire
* pakfire
) {
303 const char* arch
= pakfire_get_effective_arch(pakfire
);
305 // Allocate a new jail
306 struct pakfire_jail
* j
= calloc(1, sizeof(*j
));
311 j
->ctx
= pakfire_ctx(pakfire
);
314 j
->pakfire
= pakfire_ref(pakfire
);
316 // Initialize reference counter
319 // Generate a random UUID
320 uuid_generate_random(j
->uuid
);
322 DEBUG(j
->pakfire
, "Allocated new jail at %p\n", j
);
324 // Set the default logging callback
325 pakfire_jail_set_log_callback(j
, pakfire_jail_default_log_callback
, NULL
);
327 // Set default environment
328 for (const struct environ
* e
= ENV
; e
->key
; e
++) {
329 r
= pakfire_jail_set_env(j
, e
->key
, e
->val
);
334 // Enable all CPU features that CPU has to offer
335 if (!pakfire_arch_is_supported_by_host(arch
)) {
336 r
= pakfire_jail_set_env(j
, "QEMU_CPU", "max");
341 // Set container UUID
342 r
= pakfire_jail_set_env(j
, "container_uuid", pakfire_jail_uuid(j
));
346 // Disable systemctl to talk to systemd
347 if (!pakfire_on_root(j
->pakfire
)) {
348 r
= pakfire_jail_set_env(j
, "SYSTEMD_OFFLINE", "1");
358 pakfire_jail_free(j
);
363 PAKFIRE_EXPORT
struct pakfire_jail
* pakfire_jail_ref(struct pakfire_jail
* jail
) {
369 PAKFIRE_EXPORT
struct pakfire_jail
* pakfire_jail_unref(struct pakfire_jail
* jail
) {
370 if (--jail
->nrefs
> 0)
373 pakfire_jail_free(jail
);
379 PAKFIRE_EXPORT
void pakfire_jail_set_log_callback(struct pakfire_jail
* jail
,
380 pakfire_jail_log_callback callback
, void* data
) {
381 jail
->callbacks
.log
= callback
;
382 jail
->callbacks
.log_data
= data
;
387 PAKFIRE_EXPORT
int pakfire_jail_nice(struct pakfire_jail
* jail
, int nice
) {
388 // Check if nice level is in range
389 if (nice
< -19 || nice
> 20) {
400 int pakfire_jail_set_cgroup(struct pakfire_jail
* jail
, struct pakfire_cgroup
* cgroup
) {
401 // Free any previous cgroup
403 pakfire_cgroup_unref(jail
->cgroup
);
407 // Set any new cgroup
409 DEBUG(jail
->pakfire
, "Setting cgroup %p\n", cgroup
);
411 jail
->cgroup
= pakfire_cgroup_ref(cgroup
);
420 // Returns the length of the environment
421 static unsigned int pakfire_jail_env_length(struct pakfire_jail
* jail
) {
424 // Count everything in the environment
425 for (char** e
= jail
->env
; *e
; e
++)
431 // Finds an existing environment variable and returns its index or -1 if not found
432 static int pakfire_jail_find_env(struct pakfire_jail
* jail
, const char* key
) {
438 const size_t length
= strlen(key
);
440 for (unsigned int i
= 0; jail
->env
[i
]; i
++) {
441 if ((pakfire_string_startswith(jail
->env
[i
], key
)
442 && *(jail
->env
[i
] + length
) == '=')) {
451 // Returns the value of an environment variable or NULL
452 PAKFIRE_EXPORT
const char* pakfire_jail_get_env(struct pakfire_jail
* jail
,
454 int i
= pakfire_jail_find_env(jail
, key
);
458 return jail
->env
[i
] + strlen(key
) + 1;
461 // Sets an environment variable
462 PAKFIRE_EXPORT
int pakfire_jail_set_env(struct pakfire_jail
* jail
,
463 const char* key
, const char* value
) {
464 // Find the index where to write this value to
465 int i
= pakfire_jail_find_env(jail
, key
);
467 i
= pakfire_jail_env_length(jail
);
469 // Return -ENOSPC when the environment is full
470 if (i
>= ENVIRON_SIZE
) {
475 // Free any previous value
479 // Format and set environment variable
480 asprintf(&jail
->env
[i
], "%s=%s", key
, value
);
482 DEBUG(jail
->pakfire
, "Set environment variable: %s\n", jail
->env
[i
]);
487 // Imports an environment
488 PAKFIRE_EXPORT
int pakfire_jail_import_env(struct pakfire_jail
* jail
, const char* env
[]) {
496 // Copy environment variables
497 for (unsigned int i
= 0; env
[i
]; i
++) {
498 r
= pakfire_string_partition(env
[i
], "=", &key
, &val
);
503 r
= pakfire_jail_set_env(jail
, key
, val
);
520 PAKFIRE_EXPORT
int pakfire_jail_set_timeout(
521 struct pakfire_jail
* jail
, unsigned int timeout
) {
523 jail
->timeout
.it_value
.tv_sec
= timeout
;
526 DEBUG(jail
->pakfire
, "Timeout set to %u second(s)\n", timeout
);
528 DEBUG(jail
->pakfire
, "Timeout disabled\n");
533 static int pakfire_jail_create_timer(struct pakfire_jail
* jail
) {
536 // Nothing to do if no timeout has been set
537 if (!jail
->timeout
.it_value
.tv_sec
)
540 // Create a new timer
541 const int fd
= timerfd_create(CLOCK_MONOTONIC
, 0);
543 ERROR(jail
->pakfire
, "Could not create timer: %m\n");
548 r
= timerfd_settime(fd
, 0, &jail
->timeout
, NULL
);
550 ERROR(jail
->pakfire
, "Could not arm timer: %m\n");
564 This function replaces any logging in the child process.
566 All log messages will be sent to the parent process through their respective pipes.
568 static void pakfire_jail_log_redirect(void* data
, int priority
, const char* file
,
569 int line
, const char* fn
, const char* format
, va_list args
) {
570 struct pakfire_jail_pipes
* pipes
= (struct pakfire_jail_pipes
*)data
;
575 fd
= pipes
->log_INFO
[1];
579 fd
= pipes
->log_ERROR
[1];
584 fd
= pipes
->log_DEBUG
[1];
586 #endif /* ENABLE_DEBUG */
588 // Ignore any messages of an unknown priority
593 // Send the log message
595 vdprintf(fd
, format
, args
);
598 static int pakfire_jail_log_buffer_is_full(const struct pakfire_log_buffer
* buffer
) {
599 return (sizeof(buffer
->data
) == buffer
->used
);
603 This function reads as much data as it can from the file descriptor.
604 If it finds a whole line in it, it will send it to the logger and repeat the process.
605 If not newline character is found, it will try to read more data until it finds one.
607 static int pakfire_jail_handle_log(struct pakfire_jail
* jail
,
608 struct pakfire_jail_exec
* ctx
, int priority
, int fd
,
609 struct pakfire_log_buffer
* buffer
, pakfire_jail_communicate_out callback
, void* data
) {
610 char line
[BUFFER_SIZE
+ 1];
612 // Fill up buffer from fd
613 if (buffer
->used
< sizeof(buffer
->data
)) {
614 ssize_t bytes_read
= read(fd
, buffer
->data
+ buffer
->used
,
615 sizeof(buffer
->data
) - buffer
->used
);
618 if (bytes_read
< 0) {
619 ERROR(jail
->pakfire
, "Could not read from fd %d: %m\n", fd
);
623 // Update buffer size
624 buffer
->used
+= bytes_read
;
627 // See if we have any lines that we can write
628 while (buffer
->used
) {
629 // Search for the end of the first line
630 char* eol
= memchr(buffer
->data
, '\n', buffer
->used
);
634 // If the buffer is full, we send the content to the logger and try again
635 // This should not happen in practise
636 if (pakfire_jail_log_buffer_is_full(buffer
)) {
637 DEBUG(jail
->pakfire
, "Logging buffer is full. Sending all content\n");
639 eol
= buffer
->data
+ sizeof(buffer
->data
) - 1;
641 // Otherwise we might have only read parts of the output
646 // Find the length of the string
647 size_t length
= eol
- buffer
->data
+ 1;
649 // Copy the line into the buffer
650 memcpy(line
, buffer
->data
, length
);
652 // Terminate the string
657 int r
= callback(jail
->pakfire
, data
, priority
, line
, length
);
659 ERROR(jail
->pakfire
, "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_stream_stdin(struct pakfire_jail
* jail
,
673 struct pakfire_jail_exec
* ctx
, const int fd
) {
676 // Nothing to do if there is no stdin callback set
677 if (!ctx
->communicate
.in
) {
678 DEBUG(jail
->pakfire
, "Callback for standard input is not set\n");
682 // Skip if the writing pipe has already been closed
683 if (ctx
->pipes
.stdin
[1] < 0)
686 DEBUG(jail
->pakfire
, "Streaming standard input...\n");
688 // Calling the callback
689 r
= ctx
->communicate
.in(jail
->pakfire
, ctx
->communicate
.data
, fd
);
691 DEBUG(jail
->pakfire
, "Standard input callback finished: %d\n", r
);
693 // The callback signaled that it has written everything
695 DEBUG(jail
->pakfire
, "Closing standard input pipe\n");
697 // Close the file-descriptor
700 // Reset the file-descriptor so it won't be closed again later
701 ctx
->pipes
.stdin
[1] = -1;
710 static int pakfire_jail_recv_fd(struct pakfire_jail
* jail
, int socket
, int* fd
) {
711 const size_t payload_length
= sizeof(fd
);
712 char buffer
[CMSG_SPACE(payload_length
)];
715 struct msghdr msg
= {
716 .msg_control
= buffer
,
717 .msg_controllen
= sizeof(buffer
),
720 // Receive the message
721 r
= recvmsg(socket
, &msg
, 0);
723 CTX_ERROR(jail
->ctx
, "Could not receive file descriptor: %s\n", strerror(errno
));
728 struct cmsghdr
* cmsg
= CMSG_FIRSTHDR(&msg
);
732 *fd
= *((int*)CMSG_DATA(cmsg
));
734 CTX_DEBUG(jail
->ctx
, "Received fd %d from socket %d\n", *fd
, socket
);
739 static int pakfire_jail_send_fd(struct pakfire_jail
* jail
, int socket
, int fd
) {
740 const size_t payload_length
= sizeof(fd
);
741 char buffer
[CMSG_SPACE(payload_length
)];
744 CTX_DEBUG(jail
->ctx
, "Sending fd %d to socket %d\n", fd
, socket
);
747 struct msghdr msg
= {
748 .msg_control
= buffer
,
749 .msg_controllen
= sizeof(buffer
),
753 struct cmsghdr
* cmsg
= CMSG_FIRSTHDR(&msg
);
754 cmsg
->cmsg_level
= SOL_SOCKET
;
755 cmsg
->cmsg_type
= SCM_RIGHTS
;
756 cmsg
->cmsg_len
= CMSG_LEN(payload_length
);
759 *((int*)CMSG_DATA(cmsg
)) = fd
;
762 r
= sendmsg(socket
, &msg
, 0);
764 CTX_ERROR(jail
->ctx
, "Could not send file descriptor: %s\n", strerror(errno
));
771 static int pakfire_jail_setup_pipe(struct pakfire_jail
* jail
, int (*fds
)[2], const int flags
) {
772 int r
= pipe2(*fds
, flags
);
774 ERROR(jail
->pakfire
, "Could not setup pipe: %m\n");
781 static void pakfire_jail_close_pipe(struct pakfire_jail
* jail
, int fds
[2]) {
782 for (unsigned int i
= 0; i
< 2; i
++)
788 This is a convenience function to fetch the reading end of a pipe and
789 closes the write end.
791 static int pakfire_jail_get_pipe_to_read(struct pakfire_jail
* jail
, int (*fds
)[2]) {
792 // Give the variables easier names to avoid confusion
793 int* fd_read
= &(*fds
)[0];
794 int* fd_write
= &(*fds
)[1];
796 // Close the write end of the pipe
797 if (*fd_write
>= 0) {
802 // Return the read end
809 static int pakfire_jail_get_pipe_to_write(struct pakfire_jail
* jail
, int (*fds
)[2]) {
810 // Give the variables easier names to avoid confusion
811 int* fd_read
= &(*fds
)[0];
812 int* fd_write
= &(*fds
)[1];
814 // Close the read end of the pipe
820 // Return the write end
827 static int pakfire_jail_log(struct pakfire
* pakfire
, void* data
, int priority
,
828 const char* line
, const size_t length
) {
829 // Pass everything to the parent logger
830 pakfire_log_condition(pakfire
, priority
, 0, "%.*s", (int)length
, line
);
835 static int pakfire_jail_epoll_add_fd(struct pakfire_jail
* jail
, int epollfd
, int fd
, int events
) {
836 struct epoll_event event
= {
837 .events
= events
|EPOLLHUP
,
845 int flags
= fcntl(fd
, F_GETFL
, 0);
847 // Set modified flags
848 r
= fcntl(fd
, F_SETFL
, flags
|O_NONBLOCK
);
850 CTX_ERROR(jail
->ctx
, "Could not set file descriptor %d into non-blocking mode: %s\n",
851 fd
, strerror(errno
));
855 // Add the file descriptor to the loop
856 r
= epoll_ctl(epollfd
, EPOLL_CTL_ADD
, fd
, &event
);
858 ERROR(jail
->pakfire
, "Could not add file descriptor %d to epoll(): %s\n",
859 fd
, strerror(errno
));
868 static int pakfire_jail_enable_raw_mode(struct pakfire_jail
* jail
,
869 struct pakfire_jail_pty_stdio
* stdio
) {
870 struct termios raw_attrs
;
874 stdio
->fdflags
= fcntl(stdio
->fd
, F_GETFL
);
875 if (stdio
->fdflags
< 0) {
876 CTX_ERROR(jail
->ctx
, "Could not fetch flags from fd %d: %s\n",
877 stdio
->fd
, strerror(errno
));
881 // Fetch all attributes
882 r
= tcgetattr(stdio
->fd
, &stdio
->attrs
);
884 CTX_ERROR(jail
->ctx
, "Could not fetch terminal attributes from fd %d: %s\n",
885 stdio
->fd
, strerror(errno
));
889 // Copy all attributes
890 raw_attrs
= stdio
->attrs
;
893 cfmakeraw(&raw_attrs
);
895 // Restore the attributes
896 r
= tcsetattr(stdio
->fd
, TCSANOW
, &raw_attrs
);
898 CTX_ERROR(jail
->ctx
, "Could not restore terminal attributes for fd %d: %s\n",
899 stdio
->fd
, strerror(errno
));
906 static int pakfire_jail_restore_attrs(struct pakfire_jail
* jail
,
907 const struct pakfire_jail_pty_stdio
* stdio
) {
911 r
= fcntl(stdio
->fd
, F_SETFL
, stdio
->fdflags
);
913 CTX_ERROR(jail
->ctx
, "Could not set flags for file descriptor %d: %s\n",
914 stdio
->fd
, strerror(errno
));
918 // Restore the attributes
919 r
= tcsetattr(stdio
->fd
, TCSANOW
, &stdio
->attrs
);
921 CTX_ERROR(jail
->ctx
, "Could not restore terminal attributes for %d, ignoring: %s\n",
922 stdio
->fd
, strerror(errno
));
929 static int pakfire_jail_setup_pty_forwarding(struct pakfire_jail
* jail
,
930 struct pakfire_jail_exec
* ctx
, const int epollfd
, const int fd
) {
934 CTX_DEBUG(jail
->ctx
, "Setting up PTY forwarding on fd %d\n", fd
);
936 // Store the file descriptor
937 ctx
->pty
.master
.fd
= fd
;
939 // Configure stdin/stdout
940 ctx
->pty
.stdin
.fd
= STDIN_FILENO
;
941 ctx
->pty
.stdout
.fd
= STDOUT_FILENO
;
944 r
= ioctl(ctx
->pty
.stdout
.fd
, TIOCGWINSZ
, &size
);
946 CTX_ERROR(jail
->ctx
, "Failed to determine terminal dimensions: %s\n", strerror(errno
));
951 r
= ioctl(ctx
->pty
.master
.fd
, TIOCSWINSZ
, &size
);
953 CTX_ERROR(jail
->ctx
, "Failed setting dimensions: %s\n", strerror(errno
));
957 // Enable RAW mode on standard input
958 r
= pakfire_jail_enable_raw_mode(jail
, &ctx
->pty
.stdin
);
962 // Enable RAW mode on standard output
963 r
= pakfire_jail_enable_raw_mode(jail
, &ctx
->pty
.stdout
);
967 // Add the master to the event loop
968 r
= pakfire_jail_epoll_add_fd(jail
, epollfd
, ctx
->pty
.master
.fd
, EPOLLIN
|EPOLLOUT
|EPOLLET
);
972 // Add standard input to the event loop
973 r
= pakfire_jail_epoll_add_fd(jail
, epollfd
, ctx
->pty
.stdin
.fd
, EPOLLIN
|EPOLLET
);
977 // Add standard output to the event loop
978 r
= pakfire_jail_epoll_add_fd(jail
, epollfd
, ctx
->pty
.stdout
.fd
, EPOLLOUT
|EPOLLET
);
985 static int pakfire_jail_wait(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
) {
987 struct epoll_event events
[EPOLL_MAX_EVENTS
];
991 // Fetch file descriptors from context
992 const int stdin
= pakfire_jail_get_pipe_to_write(jail
, &ctx
->pipes
.stdin
);
993 const int stdout
= pakfire_jail_get_pipe_to_read(jail
, &ctx
->pipes
.stdout
);
994 const int stderr
= pakfire_jail_get_pipe_to_read(jail
, &ctx
->pipes
.stderr
);
995 const int pidfd
= ctx
->pidfd
;
997 // Fetch the UNIX domain socket
998 const int socket_recv
= pakfire_jail_get_pipe_to_read(jail
, &ctx
->socket
);
1001 const int timerfd
= pakfire_jail_create_timer(jail
);
1004 const int log_INFO
= pakfire_jail_get_pipe_to_read(jail
, &ctx
->pipes
.log_INFO
);
1005 const int log_ERROR
= pakfire_jail_get_pipe_to_read(jail
, &ctx
->pipes
.log_ERROR
);
1007 const int log_DEBUG
= pakfire_jail_get_pipe_to_read(jail
, &ctx
->pipes
.log_DEBUG
);
1008 #endif /* ENABLE_DEBUG */
1010 // Make a list of all file descriptors we are interested in
1011 const struct pakfire_wait_fds
{
1015 // Standard input/output
1016 { stdin
, EPOLLOUT
},
1017 { stdout
, EPOLLIN
},
1018 { stderr
, EPOLLIN
},
1021 { timerfd
, EPOLLIN
},
1024 { ctx
->pidfd
, EPOLLIN
},
1027 { log_INFO
, EPOLLIN
},
1028 { log_ERROR
, EPOLLIN
},
1030 { log_DEBUG
, EPOLLIN
},
1031 #endif /* ENABLE_DEBUG */
1033 // UNIX Domain Socket
1034 { socket_recv
, EPOLLIN
},
1041 epollfd
= epoll_create1(0);
1043 ERROR(jail
->pakfire
, "Could not initialize epoll(): %m\n");
1048 // Turn file descriptors into non-blocking mode and add them to epoll()
1049 for (const struct pakfire_wait_fds
* fd
= fds
; fd
->events
; fd
++) {
1050 // Skip fds which were not initialized
1054 // Add the FD to the event loop
1055 r
= pakfire_jail_epoll_add_fd(jail
, epollfd
, fd
->fd
, fd
->events
);
1062 // Loop for as long as the process is alive
1064 int num
= epoll_wait(epollfd
, events
, EPOLL_MAX_EVENTS
, -1);
1066 // Ignore if epoll_wait() has been interrupted
1070 ERROR(jail
->pakfire
, "epoll_wait() failed: %m\n");
1076 for (int i
= 0; i
< num
; i
++) {
1077 int e
= events
[i
].events
;
1078 int fd
= events
[i
].data
.fd
;
1080 struct pakfire_log_buffer
* buffer
= NULL
;
1081 pakfire_jail_communicate_out callback
= NULL
;
1085 // Check if there is any data to be read
1087 // Handle any changes to the PIDFD
1089 // Call waidid() and store the result
1090 r
= waitid(P_PIDFD
, ctx
->pidfd
, &ctx
->status
, WEXITED
);
1092 ERROR(jail
->pakfire
, "waitid() failed: %m\n");
1096 // Mark that we have ended so that we will process the remaining
1097 // events from epoll() now, but won't restart the outer loop.
1101 // Handle timer events
1102 } else if (fd
== timerfd
) {
1103 DEBUG(jail
->pakfire
, "Timer event received\n");
1106 r
= read(timerfd
, garbage
, sizeof(garbage
));
1108 ERROR(jail
->pakfire
, "Could not disarm timer: %m\n");
1113 // Terminate the process if it hasn't already ended
1115 DEBUG(jail
->pakfire
, "Terminating process...\n");
1117 // Send SIGTERM to the process
1118 r
= pidfd_send_signal(pidfd
, SIGKILL
, NULL
, 0);
1120 ERROR(jail
->pakfire
, "Could not kill process: %m\n");
1125 // Don't fall through to log processing
1128 // Handle logging messages
1129 } else if (fd
== log_INFO
) {
1130 buffer
= &ctx
->buffers
.log_INFO
;
1131 priority
= LOG_INFO
;
1133 callback
= pakfire_jail_log
;
1135 } else if (fd
== log_ERROR
) {
1136 buffer
= &ctx
->buffers
.log_ERROR
;
1139 callback
= pakfire_jail_log
;
1142 } else if (fd
== log_DEBUG
) {
1143 buffer
= &ctx
->buffers
.log_DEBUG
;
1144 priority
= LOG_DEBUG
;
1146 callback
= pakfire_jail_log
;
1147 #endif /* ENABLE_DEBUG */
1149 // Handle anything from the log pipes
1150 } else if (fd
== stdout
) {
1151 buffer
= &ctx
->buffers
.stdout
;
1152 priority
= LOG_INFO
;
1154 // Send any output to the default logger if no callback is set
1155 if (ctx
->communicate
.out
) {
1156 callback
= ctx
->communicate
.out
;
1157 data
= ctx
->communicate
.data
;
1159 callback
= jail
->callbacks
.log
;
1160 data
= jail
->callbacks
.log_data
;
1163 } else if (fd
== stderr
) {
1164 buffer
= &ctx
->buffers
.stderr
;
1167 // Send any output to the default logger if no callback is set
1168 if (ctx
->communicate
.out
) {
1169 callback
= ctx
->communicate
.out
;
1170 data
= ctx
->communicate
.data
;
1172 callback
= jail
->callbacks
.log
;
1173 data
= jail
->callbacks
.log_data
;
1176 // Handle socket messages
1177 } else if (fd
== socket_recv
) {
1178 // Receive the passed FD
1179 r
= pakfire_jail_recv_fd(jail
, socket_recv
, &fd
);
1183 // Setup PTY forwarding
1184 if (ctx
->pty
.master
.fd
< 0) {
1185 r
= pakfire_jail_setup_pty_forwarding(jail
, ctx
, epollfd
, fd
);
1187 CTX_ERROR(jail
->ctx
, "Failed setting up PTY forwarding: %s\n", strerror(-r
));
1192 // Don't fall through to log processing
1196 DEBUG(jail
->pakfire
, "Received invalid file descriptor %d\n", fd
);
1201 r
= pakfire_jail_handle_log(jail
, ctx
, priority
, fd
, buffer
, callback
, data
);
1207 // Handle standard input
1209 r
= pakfire_jail_stream_stdin(jail
, ctx
, fd
);
1212 // Ignore if we filled up the buffer
1217 ERROR(jail
->pakfire
, "Could not write to stdin: %m\n");
1224 // Check if any file descriptors have been closed
1226 // Remove the file descriptor
1227 r
= epoll_ctl(epollfd
, EPOLL_CTL_DEL
, fd
, NULL
);
1229 ERROR(jail
->pakfire
, "Could not remove closed file-descriptor %d: %m\n", fd
);
1242 // Restore any changed terminal attributes
1243 pakfire_jail_restore_attrs(jail
, &ctx
->pty
.stdin
);
1244 pakfire_jail_restore_attrs(jail
, &ctx
->pty
.stdout
);
1249 int pakfire_jail_capture_stdout(struct pakfire
* pakfire
, void* data
,
1250 int priority
, const char* line
, size_t length
) {
1251 char** output
= (char**)data
;
1254 // Append everything from stdout to a buffer
1255 if (output
&& priority
== LOG_INFO
) {
1256 r
= asprintf(output
, "%s%s", (output
&& *output
) ? *output
: "", line
);
1262 // Send everything else to the default logger
1263 return pakfire_jail_default_log_callback(pakfire
, NULL
, priority
, line
, length
);
1268 // Logs all capabilities of the current process
1269 static int pakfire_jail_show_capabilities(struct pakfire_jail
* jail
) {
1272 cap_flag_value_t value_e
;
1273 cap_flag_value_t value_i
;
1274 cap_flag_value_t value_p
;
1278 pid_t pid
= getpid();
1280 // Fetch all capabilities
1281 caps
= cap_get_proc();
1283 ERROR(jail
->pakfire
, "Could not fetch capabilities: %m\n");
1288 DEBUG(jail
->pakfire
, "Capabilities of PID %d:\n", pid
);
1290 // Iterate over all capabilities
1291 for (unsigned int cap
= 0; cap_valid(cap
); cap
++) {
1292 name
= cap_to_name(cap
);
1294 // Fetch effective value
1295 r
= cap_get_flag(caps
, cap
, CAP_EFFECTIVE
, &value_e
);
1299 // Fetch inheritable value
1300 r
= cap_get_flag(caps
, cap
, CAP_INHERITABLE
, &value_i
);
1304 // Fetch permitted value
1305 r
= cap_get_flag(caps
, cap
, CAP_PERMITTED
, &value_p
);
1309 DEBUG(jail
->pakfire
,
1310 " %-24s : %c%c%c\n",
1312 (value_e
== CAP_SET
) ? 'e' : '-',
1313 (value_i
== CAP_SET
) ? 'i' : '-',
1314 (value_p
== CAP_SET
) ? 'p' : '-'
1334 static int pakfire_jail_set_capabilities(struct pakfire_jail
* jail
) {
1339 // Fetch capabilities
1340 caps
= cap_get_proc();
1342 ERROR(jail
->pakfire
, "Could not read capabilities: %m\n");
1347 // Walk through all capabilities
1348 for (cap_value_t cap
= 0; cap_valid(cap
); cap
++) {
1349 cap_value_t _caps
[] = { cap
};
1351 // Fetch the name of the capability
1352 name
= cap_to_name(cap
);
1354 r
= cap_set_flag(caps
, CAP_EFFECTIVE
, 1, _caps
, CAP_SET
);
1356 ERROR(jail
->pakfire
, "Could not set %s: %m\n", name
);
1360 r
= cap_set_flag(caps
, CAP_INHERITABLE
, 1, _caps
, CAP_SET
);
1362 ERROR(jail
->pakfire
, "Could not set %s: %m\n", name
);
1366 r
= cap_set_flag(caps
, CAP_PERMITTED
, 1, _caps
, CAP_SET
);
1368 ERROR(jail
->pakfire
, "Could not set %s: %m\n", name
);
1377 // Restore all capabilities
1378 r
= cap_set_proc(caps
);
1380 ERROR(jail
->pakfire
, "Restoring capabilities failed: %m\n");
1384 // Add all capabilities to the ambient set
1385 for (unsigned int cap
= 0; cap_valid(cap
); cap
++) {
1386 name
= cap_to_name(cap
);
1388 // Raise the capability
1389 r
= prctl(PR_CAP_AMBIENT
, PR_CAP_AMBIENT_RAISE
, cap
, 0, 0);
1391 ERROR(jail
->pakfire
, "Could not set ambient capability %s: %m\n", name
);
1414 static int pakfire_jail_limit_syscalls(struct pakfire_jail
* jail
) {
1415 const int syscalls
[] = {
1416 // The kernel's keyring isn't namespaced
1419 SCMP_SYS(request_key
),
1421 // Disable userfaultfd
1422 SCMP_SYS(userfaultfd
),
1424 // Disable perf which could leak a lot of information about the host
1425 SCMP_SYS(perf_event_open
),
1431 DEBUG(jail
->pakfire
, "Applying syscall filter...\n");
1433 // Setup a syscall filter which allows everything by default
1434 scmp_filter_ctx ctx
= seccomp_init(SCMP_ACT_ALLOW
);
1436 ERROR(jail
->pakfire
, "Could not setup seccomp filter: %m\n");
1441 for (const int* syscall
= syscalls
; *syscall
; syscall
++) {
1442 r
= seccomp_rule_add(ctx
, SCMP_ACT_ERRNO(EPERM
), *syscall
, 0);
1444 ERROR(jail
->pakfire
, "Could not configure syscall %d: %m\n", *syscall
);
1449 // Load syscall filter into the kernel
1450 r
= seccomp_load(ctx
);
1452 ERROR(jail
->pakfire
, "Could not load syscall filter into the kernel: %m\n");
1458 seccomp_release(ctx
);
1465 PAKFIRE_EXPORT
int pakfire_jail_bind(struct pakfire_jail
* jail
,
1466 const char* source
, const char* target
, int flags
) {
1467 struct pakfire_jail_mountpoint
* mp
= NULL
;
1470 // Check if there is any space left
1471 if (jail
->num_mountpoints
>= MAX_MOUNTPOINTS
) {
1476 // Check for valid inputs
1477 if (!source
|| !target
) {
1482 // Select the next free slot
1483 mp
= &jail
->mountpoints
[jail
->num_mountpoints
];
1486 r
= pakfire_string_set(mp
->source
, source
);
1488 ERROR(jail
->pakfire
, "Could not copy source: %m\n");
1493 r
= pakfire_string_set(mp
->target
, target
);
1495 ERROR(jail
->pakfire
, "Could not copy target: %m\n");
1502 // Increment counter
1503 jail
->num_mountpoints
++;
1508 static int pakfire_jail_mount_networking(struct pakfire_jail
* jail
) {
1511 const char* paths
[] = {
1517 // Bind-mount all paths read-only
1518 for (const char** path
= paths
; *path
; path
++) {
1519 r
= pakfire_bind(jail
->pakfire
, *path
, NULL
, MS_RDONLY
);
1522 // Ignore if we don't have permission
1537 Mounts everything that we require in the new namespace
1539 static int pakfire_jail_mount(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
) {
1540 struct pakfire_jail_mountpoint
* mp
= NULL
;
1544 // Enable loop devices
1545 if (pakfire_jail_exec_has_flag(ctx
, PAKFIRE_JAIL_HAS_LOOP_DEVICES
))
1546 flags
|= PAKFIRE_MOUNT_LOOP_DEVICES
;
1548 // Mount all default stuff
1549 r
= pakfire_mount_all(jail
->pakfire
, PAKFIRE_MNTNS_OUTER
, flags
);
1554 r
= pakfire_populate_dev(jail
->pakfire
, flags
);
1558 // Mount the interpreter (if needed)
1559 r
= pakfire_mount_interpreter(jail
->pakfire
);
1563 // Mount networking stuff
1564 if (pakfire_jail_exec_has_flag(ctx
, PAKFIRE_JAIL_HAS_NETWORKING
)) {
1565 r
= pakfire_jail_mount_networking(jail
);
1570 // Mount all custom stuff
1571 for (unsigned int i
= 0; i
< jail
->num_mountpoints
; i
++) {
1573 mp
= &jail
->mountpoints
[i
];
1576 r
= pakfire_bind(jail
->pakfire
, mp
->source
, mp
->target
, mp
->flags
);
1586 static int pakfire_jail_setup_loopback(struct pakfire_jail
* jail
) {
1587 struct nl_sock
* nl
= NULL
;
1588 struct nl_cache
* cache
= NULL
;
1589 struct rtnl_link
* link
= NULL
;
1590 struct rtnl_link
* change
= NULL
;
1593 DEBUG(jail
->pakfire
, "Setting up loopback...\n");
1595 // Allocate a netlink socket
1596 nl
= nl_socket_alloc();
1598 ERROR(jail
->pakfire
, "Could not allocate a netlink socket: %m\n");
1603 // Connect the socket
1604 r
= nl_connect(nl
, NETLINK_ROUTE
);
1606 ERROR(jail
->pakfire
, "Could not connect netlink socket: %s\n", nl_geterror(r
));
1610 // Allocate the netlink cache
1611 r
= rtnl_link_alloc_cache(nl
, AF_UNSPEC
, &cache
);
1613 ERROR(jail
->pakfire
, "Unable to allocate netlink cache: %s\n", nl_geterror(r
));
1617 // Fetch loopback interface
1618 link
= rtnl_link_get_by_name(cache
, "lo");
1620 ERROR(jail
->pakfire
, "Could not find lo interface. Ignoring.\n");
1625 // Allocate a new link
1626 change
= rtnl_link_alloc();
1628 ERROR(jail
->pakfire
, "Could not allocate change link\n");
1633 // Set the link to UP
1634 rtnl_link_set_flags(change
, IFF_UP
);
1636 // Apply any changes
1637 r
= rtnl_link_change(nl
, link
, change
, 0);
1639 ERROR(jail
->pakfire
, "Unable to activate loopback: %s\n", nl_geterror(r
));
1655 static int pakfire_jail_setup_uid_mapping(struct pakfire_jail
* jail
, pid_t pid
) {
1656 char path
[PATH_MAX
];
1659 // Skip mapping anything when running on /
1660 if (pakfire_on_root(jail
->pakfire
))
1664 r
= pakfire_string_format(path
, "/proc/%d/uid_map", pid
);
1669 const uid_t uid
= pakfire_uid(jail
->pakfire
);
1672 const struct pakfire_subid
* subuid
= pakfire_subuid(jail
->pakfire
);
1676 /* When running as root, we will map the entire range.
1678 When running as a non-privileged user, we will map the root user inside the jail
1679 to the user's UID outside of the jail, and we will map the rest starting from one.
1684 r
= pakfire_file_write(jail
->pakfire
, path
, 0, 0, 0,
1685 "0 %lu %lu\n", subuid
->id
, subuid
->length
);
1687 r
= pakfire_file_write(jail
->pakfire
, path
, 0, 0, 0,
1688 "0 %lu 1\n1 %lu %lu\n", uid
, subuid
->id
, subuid
->length
);
1692 ERROR(jail
->pakfire
, "Could not map UIDs: %m\n");
1699 static int pakfire_jail_setup_gid_mapping(struct pakfire_jail
* jail
, pid_t pid
) {
1700 char path
[PATH_MAX
];
1703 // Skip mapping anything when running on /
1704 if (pakfire_on_root(jail
->pakfire
))
1708 const gid_t gid
= pakfire_gid(jail
->pakfire
);
1711 const struct pakfire_subid
* subgid
= pakfire_subgid(jail
->pakfire
);
1716 r
= pakfire_string_format(path
, "/proc/%d/gid_map", pid
);
1722 r
= pakfire_file_write(jail
->pakfire
, path
, 0, 0, 0,
1723 "0 %lu %lu\n", subgid
->id
, subgid
->length
);
1725 r
= pakfire_file_write(jail
->pakfire
, path
, 0, 0, 0,
1726 "0 %lu 1\n1 %lu %lu\n", gid
, subgid
->id
, subgid
->length
);
1730 ERROR(jail
->pakfire
, "Could not map GIDs: %m\n");
1737 static int pakfire_jail_setgroups(struct pakfire_jail
* jail
, pid_t pid
) {
1738 char path
[PATH_MAX
];
1742 r
= pakfire_string_format(path
, "/proc/%d/setgroups", pid
);
1746 r
= pakfire_file_write(jail
->pakfire
, path
, 0, 0, 0, "deny\n");
1748 CTX_ERROR(jail
->ctx
, "Could not set setgroups to deny: %s\n", strerror(errno
));
1755 static int pakfire_jail_send_signal(struct pakfire_jail
* jail
, int fd
) {
1756 const uint64_t val
= 1;
1759 DEBUG(jail
->pakfire
, "Sending signal...\n");
1761 // Write to the file descriptor
1762 r
= eventfd_write(fd
, val
);
1764 ERROR(jail
->pakfire
, "Could not send signal: %s\n", strerror(errno
));
1768 // Close the file descriptor
1774 static int pakfire_jail_wait_for_signal(struct pakfire_jail
* jail
, int fd
) {
1778 DEBUG(jail
->pakfire
, "Waiting for signal...\n");
1780 r
= eventfd_read(fd
, &val
);
1782 ERROR(jail
->pakfire
, "Error waiting for signal: %s\n", strerror(errno
));
1786 // Close the file descriptor
1793 Performs the initialisation that needs to happen in the parent part
1795 static int pakfire_jail_parent(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
) {
1798 // Setup UID mapping
1799 r
= pakfire_jail_setup_uid_mapping(jail
, ctx
->pid
);
1803 // Write "deny" to /proc/PID/setgroups
1804 r
= pakfire_jail_setgroups(jail
, ctx
->pid
);
1808 // Setup GID mapping
1809 r
= pakfire_jail_setup_gid_mapping(jail
, ctx
->pid
);
1813 // Parent has finished initialisation
1814 DEBUG(jail
->pakfire
, "Parent has finished initialization\n");
1816 // Send signal to client
1817 r
= pakfire_jail_send_signal(jail
, ctx
->completed_fd
);
1824 static int pakfire_jail_switch_root(struct pakfire_jail
* jail
, const char* root
) {
1827 // Change to the new root
1830 ERROR(jail
->pakfire
, "chdir(%s) failed: %m\n", root
);
1835 r
= pivot_root(".", ".");
1837 ERROR(jail
->pakfire
, "Failed changing into the new root directory %s: %m\n", root
);
1841 // Umount the old root
1842 r
= umount2(".", MNT_DETACH
);
1844 ERROR(jail
->pakfire
, "Could not umount the old root filesystem: %m\n");
1851 static int pakfire_jail_open_pty(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
) {
1854 // Allocate a new PTY
1855 ctx
->pty
.master
.fd
= posix_openpt(O_RDWR
|O_NONBLOCK
|O_NOCTTY
|O_CLOEXEC
);
1856 if (ctx
->pty
.master
.fd
< 0)
1860 r
= ptsname_r(ctx
->pty
.master
.fd
, ctx
->pty
.console
, sizeof(ctx
->pty
.console
));
1864 CTX_DEBUG(jail
->ctx
, "Allocated console at %s (%d)\n", ctx
->pty
.console
, ctx
->pty
.master
.fd
);
1868 r
= pakfire_symlink(jail
->ctx
, "/dev/console", ctx
->pty
.console
);
1876 static int pakfire_jail_child(struct pakfire_jail
* jail
, struct pakfire_jail_exec
* ctx
,
1877 const char* argv
[]) {
1880 // Redirect any logging to our log pipe
1881 pakfire_ctx_set_log_callback(jail
->ctx
, pakfire_jail_log_redirect
, &ctx
->pipes
);
1884 pid_t pid
= getpid();
1886 DEBUG(jail
->pakfire
, "Launched child process in jail with PID %d\n", pid
);
1888 // Wait for the parent to finish initialization
1889 r
= pakfire_jail_wait_for_signal(jail
, ctx
->completed_fd
);
1894 r
= prctl(PR_SET_PDEATHSIG
, SIGKILL
, 0, 0, 0);
1896 ERROR(jail
->pakfire
, "Could not configure to die with parent: %m\n");
1900 // Make this process dumpable
1901 r
= prctl (PR_SET_DUMPABLE
, 1, 0, 0, 0);
1903 ERROR(jail
->pakfire
, "Could not make the process dumpable: %m\n");
1907 // Don't drop any capabilities on setuid()
1908 r
= prctl(PR_SET_KEEPCAPS
, 1);
1910 ERROR(jail
->pakfire
, "Could not set PR_SET_KEEPCAPS: %m\n");
1915 uid_t uid
= getuid();
1916 gid_t gid
= getgid();
1919 uid_t euid
= geteuid();
1920 gid_t egid
= getegid();
1922 DEBUG(jail
->pakfire
, " UID: %u (effective %u)\n", uid
, euid
);
1923 DEBUG(jail
->pakfire
, " GID: %u (effective %u)\n", gid
, egid
);
1925 // Log all mountpoints
1926 pakfire_mount_list(jail
->ctx
);
1928 // Fail if we are not PID 1
1930 CTX_ERROR(jail
->ctx
, "Child process is not PID 1\n");
1934 // Fail if we are not running as root
1935 if (uid
|| gid
|| euid
|| egid
) {
1936 ERROR(jail
->pakfire
, "Child process is not running as root\n");
1940 const int socket_send
= pakfire_jail_get_pipe_to_write(jail
, &ctx
->socket
);
1942 // Mount all default stuff
1943 r
= pakfire_mount_all(jail
->pakfire
, PAKFIRE_MNTNS_INNER
, 0);
1947 const char* root
= pakfire_get_path(jail
->pakfire
);
1948 const char* arch
= pakfire_get_effective_arch(jail
->pakfire
);
1950 // Change mount propagation to slave to receive anything from the parent namespace
1951 r
= pakfire_mount_change_propagation(jail
->ctx
, "/", MS_SLAVE
);
1955 // Make root a mountpoint in the new mount namespace
1956 r
= pakfire_mount_make_mounpoint(jail
->pakfire
, root
);
1960 // Change mount propagation to private
1961 r
= pakfire_mount_change_propagation(jail
->ctx
, root
, MS_PRIVATE
);
1965 // Change root (unless root is /)
1966 if (!pakfire_on_root(jail
->pakfire
)) {
1968 r
= pakfire_jail_mount(jail
, ctx
);
1973 r
= pakfire_jail_switch_root(jail
, root
);
1979 unsigned long persona
= pakfire_arch_personality(arch
);
1981 r
= personality(persona
);
1983 ERROR(jail
->pakfire
, "Could not set personality (%x)\n", (unsigned int)persona
);
1989 if (!pakfire_jail_exec_has_flag(ctx
, PAKFIRE_JAIL_HAS_NETWORKING
)) {
1990 r
= pakfire_jail_setup_loopback(jail
);
1997 DEBUG(jail
->pakfire
, "Setting nice level to %d\n", jail
->nice
);
1999 r
= setpriority(PRIO_PROCESS
, pid
, jail
->nice
);
2001 ERROR(jail
->pakfire
, "Could not set nice level: %m\n");
2007 // Create a new session
2010 CTX_ERROR(jail
->ctx
, "Could not create a new session: %s\n", strerror(errno
));
2015 // Allocate a new PTY
2016 r
= pakfire_jail_open_pty(jail
, ctx
);
2018 CTX_ERROR(jail
->ctx
, "Could not allocate a new PTY: %s\n", strerror(-r
));
2022 // Send the PTY master to the parent process
2023 r
= pakfire_jail_send_fd(jail
, socket_send
, ctx
->pty
.master
.fd
);
2025 CTX_ERROR(jail
->ctx
, "Failed sending the PTY master to the parent: %s\n", strerror(-r
));
2029 // Close the master of the PTY
2030 close(ctx
->pty
.master
.fd
);
2031 ctx
->pty
.master
.fd
= -1;
2036 // Close other end of log pipes
2037 close(ctx
->pipes
.log_INFO
[0]);
2038 close(ctx
->pipes
.log_ERROR
[0]);
2040 close(ctx
->pipes
.log_DEBUG
[0]);
2041 #endif /* ENABLE_DEBUG */
2043 // Connect standard input
2044 if (ctx
->pipes
.stdin
[0] >= 0) {
2045 r
= dup2(ctx
->pipes
.stdin
[0], STDIN_FILENO
);
2047 ERROR(jail
->pakfire
, "Could not connect fd %d to stdin: %m\n",
2048 ctx
->pipes
.stdin
[0]);
2054 // Connect standard output and error
2055 if (ctx
->pipes
.stdout
[1] >= 0 && ctx
->pipes
.stderr
[1] >= 0) {
2056 r
= dup2(ctx
->pipes
.stdout
[1], STDOUT_FILENO
);
2058 ERROR(jail
->pakfire
, "Could not connect fd %d to stdout: %m\n",
2059 ctx
->pipes
.stdout
[1]);
2064 r
= dup2(ctx
->pipes
.stderr
[1], STDERR_FILENO
);
2066 ERROR(jail
->pakfire
, "Could not connect fd %d to stderr: %m\n",
2067 ctx
->pipes
.stderr
[1]);
2072 // Close the pipe (as we have moved the original file descriptors)
2073 pakfire_jail_close_pipe(jail
, ctx
->pipes
.stdin
);
2074 pakfire_jail_close_pipe(jail
, ctx
->pipes
.stdout
);
2075 pakfire_jail_close_pipe(jail
, ctx
->pipes
.stderr
);
2078 // Reset open file limit (http://0pointer.net/blog/file-descriptor-limits.html)
2079 r
= pakfire_rlimit_reset_nofile(jail
->pakfire
);
2084 r
= pakfire_jail_set_capabilities(jail
);
2088 // Show capabilities
2089 r
= pakfire_jail_show_capabilities(jail
);
2094 r
= pakfire_jail_limit_syscalls(jail
);
2098 DEBUG(jail
->pakfire
, "Child process initialization done\n");
2099 DEBUG(jail
->pakfire
, "Launching command:\n");
2102 for (unsigned int i
= 0; argv
[i
]; i
++)
2103 DEBUG(jail
->pakfire
, " argv[%u] = %s\n", i
, argv
[i
]);
2106 r
= execvpe(argv
[0], (char**)argv
, jail
->env
);
2108 // Translate errno into regular exit code
2111 // Ignore if the command doesn't exist
2112 if (ctx
->flags
& PAKFIRE_JAIL_NOENT_OK
)
2123 ERROR(jail
->pakfire
, "Could not execve(%s): %m\n", argv
[0]);
2126 // We should not get here
2130 // Run a command in the jail
2131 static int __pakfire_jail_exec(struct pakfire_jail
* jail
, const char* argv
[],
2132 const int interactive
,
2133 pakfire_jail_communicate_in communicate_in
,
2134 pakfire_jail_communicate_out communicate_out
,
2135 void* data
, int flags
) {
2139 // Check if argv is valid
2140 if (!argv
|| !argv
[0]) {
2145 // Initialize context for this call
2146 struct pakfire_jail_exec ctx
= {
2149 .socket
= { -1, -1 },
2152 .stdin
= { -1, -1 },
2153 .stdout
= { -1, -1 },
2154 .stderr
= { -1, -1 },
2155 .log_INFO
= { -1, -1 },
2156 .log_ERROR
= { -1, -1 },
2158 .log_DEBUG
= { -1, -1 },
2159 #endif /* ENABLE_DEBUG */
2163 .in
= communicate_in
,
2164 .out
= communicate_out
,
2184 DEBUG(jail
->pakfire
, "Executing jail...\n");
2186 // Enable networking in interactive mode
2188 ctx
.flags
|= PAKFIRE_JAIL_HAS_NETWORKING
;
2191 Setup a file descriptor which can be used to notify the client that the parent
2192 has completed configuration.
2194 ctx
.completed_fd
= eventfd(0, EFD_CLOEXEC
);
2195 if (ctx
.completed_fd
< 0) {
2196 ERROR(jail
->pakfire
, "eventfd() failed: %m\n");
2200 // Create a UNIX domain socket
2201 r
= socketpair(AF_UNIX
, SOCK_DGRAM
|SOCK_CLOEXEC
, 0, ctx
.socket
);
2203 CTX_ERROR(jail
->ctx
, "Could not create UNIX socket: %s\n", strerror(errno
));
2208 // Create pipes to communicate with child process if we are not running interactively
2210 // stdin (only if callback is set)
2211 if (ctx
.communicate
.in
) {
2212 r
= pakfire_jail_setup_pipe(jail
, &ctx
.pipes
.stdin
, 0);
2218 r
= pakfire_jail_setup_pipe(jail
, &ctx
.pipes
.stdout
, 0);
2223 r
= pakfire_jail_setup_pipe(jail
, &ctx
.pipes
.stderr
, 0);
2228 // Setup pipes for logging
2230 r
= pakfire_jail_setup_pipe(jail
, &ctx
.pipes
.log_INFO
, O_CLOEXEC
);
2235 r
= pakfire_jail_setup_pipe(jail
, &ctx
.pipes
.log_ERROR
, O_CLOEXEC
);
2241 r
= pakfire_jail_setup_pipe(jail
, &ctx
.pipes
.log_DEBUG
, O_CLOEXEC
);
2244 #endif /* ENABLE_DEBUG */
2246 // Configure child process
2247 struct clone_args args
= {
2257 .exit_signal
= SIGCHLD
,
2258 .pidfd
= (long long unsigned int)&ctx
.pidfd
,
2261 // Launch the process in a cgroup that is a leaf of the configured cgroup
2263 args
.flags
|= CLONE_INTO_CGROUP
;
2266 const char* uuid
= pakfire_jail_uuid(jail
);
2268 // Create a temporary cgroup
2269 r
= pakfire_cgroup_child(&ctx
.cgroup
, jail
->cgroup
, uuid
, 0);
2271 ERROR(jail
->pakfire
, "Could not create cgroup for jail: %m\n");
2275 // Clone into this cgroup
2276 args
.cgroup
= pakfire_cgroup_fd(ctx
.cgroup
);
2280 if (!pakfire_jail_exec_has_flag(&ctx
, PAKFIRE_JAIL_HAS_NETWORKING
)) {
2281 args
.flags
|= CLONE_NEWNET
;
2284 // Fork this process
2285 ctx
.pid
= clone3(&args
, sizeof(args
));
2287 ERROR(jail
->pakfire
, "Could not clone: %m\n");
2291 } else if (ctx
.pid
== 0) {
2292 r
= pakfire_jail_child(jail
, &ctx
, argv
);
2297 r
= pakfire_jail_parent(jail
, &ctx
);
2301 DEBUG(jail
->pakfire
, "Waiting for PID %d to finish its work\n", ctx
.pid
);
2303 // Read output of the child process
2304 r
= pakfire_jail_wait(jail
, &ctx
);
2308 // Handle exit status
2309 switch (ctx
.status
.si_code
) {
2311 DEBUG(jail
->pakfire
, "The child process exited with code %d\n",
2312 ctx
.status
.si_status
);
2315 exit
= ctx
.status
.si_status
;
2319 ERROR(jail
->pakfire
, "The child process was killed\n");
2324 ERROR(jail
->pakfire
, "The child process terminated abnormally\n");
2327 // Log anything else
2329 ERROR(jail
->pakfire
, "Unknown child exit code: %d\n", ctx
.status
.si_code
);
2334 // Destroy the temporary cgroup (if any)
2336 // Read cgroup stats
2337 pakfire_cgroup_stat(ctx
.cgroup
, &ctx
.cgroup_stats
);
2338 pakfire_cgroup_stat_dump(ctx
.cgroup
, &ctx
.cgroup_stats
);
2339 pakfire_cgroup_destroy(ctx
.cgroup
);
2340 pakfire_cgroup_unref(ctx
.cgroup
);
2343 // Close any file descriptors
2344 pakfire_jail_close_pipe(jail
, ctx
.pipes
.stdin
);
2345 pakfire_jail_close_pipe(jail
, ctx
.pipes
.stdout
);
2346 pakfire_jail_close_pipe(jail
, ctx
.pipes
.stderr
);
2349 if (ctx
.pty
.master
.fd
>= 0)
2350 close(ctx
.pty
.master
.fd
);
2351 pakfire_jail_close_pipe(jail
, ctx
.pipes
.log_INFO
);
2352 pakfire_jail_close_pipe(jail
, ctx
.pipes
.log_ERROR
);
2354 pakfire_jail_close_pipe(jail
, ctx
.pipes
.log_DEBUG
);
2355 #endif /* ENABLE_DEBUG */
2356 pakfire_jail_close_pipe(jail
, ctx
.socket
);
2361 PAKFIRE_EXPORT
int pakfire_jail_exec(
2362 struct pakfire_jail
* jail
,
2364 pakfire_jail_communicate_in callback_in
,
2365 pakfire_jail_communicate_out callback_out
,
2366 void* data
, int flags
) {
2367 return __pakfire_jail_exec(jail
, argv
, 0, callback_in
, callback_out
, data
, flags
);
2370 static int pakfire_jail_exec_interactive(
2371 struct pakfire_jail
* jail
, const char* argv
[], int flags
) {
2374 // Setup interactive stuff
2375 r
= pakfire_jail_setup_interactive_env(jail
);
2379 return __pakfire_jail_exec(jail
, argv
, 1, NULL
, NULL
, NULL
, flags
);
2382 int pakfire_jail_exec_script(struct pakfire_jail
* jail
,
2386 pakfire_jail_communicate_in callback_in
,
2387 pakfire_jail_communicate_out callback_out
,
2389 char path
[PATH_MAX
];
2390 const char** argv
= NULL
;
2394 const char* root
= pakfire_get_path(jail
->pakfire
);
2396 // Write the scriptlet to disk
2397 r
= pakfire_path_append(path
, root
, PAKFIRE_TMP_DIR
"/pakfire-script.XXXXXX");
2401 // Create a temporary file
2402 f
= pakfire_mktemp(path
, 0700);
2404 ERROR(jail
->pakfire
, "Could not create temporary file: %m\n");
2408 DEBUG(jail
->pakfire
, "Writing script to %s:\n%.*s\n", path
, (int)size
, script
);
2411 r
= fprintf(f
, "%s", script
);
2413 ERROR(jail
->pakfire
, "Could not write script to file %s: %m\n", path
);
2420 ERROR(jail
->pakfire
, "Could not close script file %s: %m\n", path
);
2426 // Count how many arguments were passed
2427 unsigned int argc
= 1;
2429 for (const char** arg
= args
; *arg
; arg
++)
2433 argv
= calloc(argc
+ 1, sizeof(*argv
));
2435 ERROR(jail
->pakfire
, "Could not allocate argv: %m\n");
2440 argv
[0] = (root
) ? pakfire_path_relpath(root
, path
) : path
;
2443 for (unsigned int i
= 1; i
< argc
; i
++)
2444 argv
[i
] = args
[i
-1];
2447 r
= pakfire_jail_exec(jail
, argv
, callback_in
, callback_out
, data
, 0);
2455 // Remove script from disk
2463 A convenience function that creates a new jail, runs the given command and destroys
2466 int pakfire_jail_run(struct pakfire
* pakfire
, const char* argv
[], int flags
, char** output
) {
2467 struct pakfire_jail
* jail
= NULL
;
2470 // Create a new jail
2471 r
= pakfire_jail_create(&jail
, pakfire
);
2475 // Execute the command
2476 r
= pakfire_jail_exec(jail
, argv
, NULL
, pakfire_jail_capture_stdout
, output
, 0);
2480 pakfire_jail_unref(jail
);
2485 int pakfire_jail_run_script(struct pakfire
* pakfire
,
2486 const char* script
, const size_t length
, const char* argv
[], int flags
) {
2487 struct pakfire_jail
* jail
= NULL
;
2490 // Create a new jail
2491 r
= pakfire_jail_create(&jail
, pakfire
);
2495 // Execute the command
2496 r
= pakfire_jail_exec_script(jail
, script
, length
, argv
, NULL
, NULL
, NULL
);
2500 pakfire_jail_unref(jail
);
2505 int pakfire_jail_shell(struct pakfire_jail
* jail
) {
2508 const char* argv
[] = {
2509 "/bin/bash", "--login", NULL
,
2512 // Execute /bin/bash
2513 r
= pakfire_jail_exec_interactive(jail
, argv
, 0);
2519 // Ignore any return codes from the shell
2523 static int pakfire_jail_run_if_possible(struct pakfire
* pakfire
, const char** argv
) {
2524 char path
[PATH_MAX
];
2527 r
= pakfire_path(pakfire
, path
, "%s", *argv
);
2531 // Check if the file is executable
2532 r
= access(path
, X_OK
);
2534 DEBUG(pakfire
, "%s is not executable. Skipping...\n", *argv
);
2538 return pakfire_jail_run(pakfire
, argv
, 0, NULL
);
2541 int pakfire_jail_ldconfig(struct pakfire
* pakfire
) {
2542 const char* argv
[] = {
2547 return pakfire_jail_run_if_possible(pakfire
, argv
);
2550 int pakfire_jail_run_systemd_tmpfiles(struct pakfire
* pakfire
) {
2551 const char* argv
[] = {
2552 "/usr/bin/systemd-tmpfiles",
2557 return pakfire_jail_run_if_possible(pakfire
, argv
);