#############################################################################*/
#include <errno.h>
+#include <fcntl.h>
#include <linux/capability.h>
-#include <linux/fcntl.h>
#include <linux/sched.h>
#include <sys/wait.h>
#include <linux/wait.h>
#include <sys/personality.h>
#include <sys/prctl.h>
#include <sys/resource.h>
+#include <sys/timerfd.h>
#include <sys/types.h>
#include <sys/wait.h>
+#include <termios.h>
+
+// libnl3
+#include <net/if.h>
+#include <netlink/route/link.h>
// libseccomp
#include <seccomp.h>
#include <pakfire/logging.h>
#include <pakfire/mount.h>
#include <pakfire/pakfire.h>
+#include <pakfire/path.h>
#include <pakfire/private.h>
#include <pakfire/pwd.h>
#include <pakfire/string.h>
const char* val;
} ENV[] = {
{ "HOME", "/root" },
- { "LANG", "en_US.utf-8" },
+ { "LANG", "C.utf-8" },
{ "PATH", "/usr/local/sbin:/usr/sbin:/sbin:/usr/local/bin:/usr/bin:/bin", },
{ "TERM", "vt100" },
+
+ // Tell everything that it is running inside a Pakfire container
+ { "container", "pakfire" },
{ NULL, NULL },
};
};
struct pakfire_jail {
+ struct pakfire_ctx* ctx;
struct pakfire* pakfire;
int nrefs;
uuid_t uuid;
char __uuid[UUID_STR_LEN];
- // Flags
- int flags;
-
// Resource Limits
int nice;
+ // Timeout
+ struct itimerspec timeout;
+
// CGroup
struct pakfire_cgroup* cgroup;
// Mountpoints
struct pakfire_jail_mountpoint mountpoints[MAX_MOUNTPOINTS];
unsigned int num_mountpoints;
+
+ // Callbacks
+ struct pakfire_jail_callbacks {
+ // Log
+ pakfire_jail_log_callback log;
+ void* log_data;
+ } callbacks;
};
struct pakfire_log_buffer {
size_t used;
};
-enum pakfire_jail_exec_flags {
- PAKFIRE_JAIL_HAS_NETWORKING = (1 << 0),
-};
-
struct pakfire_jail_exec {
int flags;
pid_t pid;
int pidfd;
+ // Socket to pass FDs
+ int socket[2];
+
// Process status (from waitid)
siginfo_t status;
// Log pipes
struct pakfire_jail_pipes {
- int stdin[2];
- int stdout[2];
- int stderr[2];
-
// Logging
int log_INFO[2];
int log_ERROR[2];
+#ifdef ENABLE_DEBUG
int log_DEBUG[2];
+#endif /* ENABLE_DEBUG */
} pipes;
// Communicate
// Logging
struct pakfire_log_buffer log_INFO;
struct pakfire_log_buffer log_ERROR;
+#ifdef ENABLE_DEBUG
struct pakfire_log_buffer log_DEBUG;
+#endif /* ENABLE_DEBUG */
} buffers;
struct pakfire_cgroup* cgroup;
struct pakfire_cgroup_stats cgroup_stats;
+
+ // PTY
+ struct pakfire_jail_pty {
+ // The path to the console
+ char console[PATH_MAX];
+
+ // The master fd
+ struct pakfire_jail_pty_master {
+ int fd;
+
+ enum pakfire_jail_pty_flags {
+ PAKFIRE_JAIL_PTY_READY_TO_READ = (1 << 0),
+ PAKFIRE_JAIL_PTY_READY_TO_WRITE = (1 << 1),
+ } flags;
+ } master;
+
+ // Standard Input
+ struct pakfire_jail_pty_stdio {
+ int fd;
+ struct pakfire_log_buffer buffer;
+ struct termios attrs;
+ int fdflags;
+ enum pakfire_jail_pty_flags flags;
+ } stdin;
+
+ // Standard Output
+ struct pakfire_jail_pty_stdio stdout;
+ } pty;
};
static int clone3(struct clone_args* args, size_t size) {
return syscall(__NR_clone3, args, size);
}
+static int pidfd_send_signal(int pidfd, int sig, siginfo_t* info, unsigned int flags) {
+ return syscall(SYS_pidfd_send_signal, pidfd, sig, info, flags);
+}
+
+static int pivot_root(const char* new_root, const char* old_root) {
+ return syscall(SYS_pivot_root, new_root, old_root);
+}
+
static int pakfire_jail_exec_has_flag(
const struct pakfire_jail_exec* ctx, const enum pakfire_jail_exec_flags flag) {
return ctx->flags & flag;
if (jail->cgroup)
pakfire_cgroup_unref(jail->cgroup);
-
- pakfire_unref(jail->pakfire);
+ if (jail->pakfire)
+ pakfire_unref(jail->pakfire);
+ if (jail->ctx)
+ pakfire_ctx_unref(jail->ctx);
free(jail);
}
return 0;
}
+static const char* pakfire_jail_uuid(struct pakfire_jail* jail) {
+ if (!*jail->__uuid)
+ uuid_unparse_lower(jail->uuid, jail->__uuid);
+
+ return jail->__uuid;
+}
+
static int pakfire_jail_setup_interactive_env(struct pakfire_jail* jail) {
// Set PS1
int r = pakfire_jail_set_env(jail, "PS1", "pakfire-jail \\w> ");
return 0;
}
-PAKFIRE_EXPORT int pakfire_jail_create(struct pakfire_jail** jail,
- struct pakfire* pakfire, int flags) {
+PAKFIRE_EXPORT int pakfire_jail_create(struct pakfire_jail** jail, struct pakfire* pakfire) {
int r;
+ const char* arch = pakfire_get_effective_arch(pakfire);
+
// Allocate a new jail
struct pakfire_jail* j = calloc(1, sizeof(*j));
if (!j)
return 1;
+ // Reference context
+ j->ctx = pakfire_ctx(pakfire);
+
// Reference Pakfire
j->pakfire = pakfire_ref(pakfire);
// Initialize reference counter
j->nrefs = 1;
- // Store flags
- j->flags = flags;
-
// Generate a random UUID
uuid_generate_random(j->uuid);
DEBUG(j->pakfire, "Allocated new jail at %p\n", j);
+ // Set the default logging callback
+ pakfire_jail_set_log_callback(j, pakfire_jail_default_log_callback, NULL);
+
// Set default environment
for (const struct environ* e = ENV; e->key; e++) {
r = pakfire_jail_set_env(j, e->key, e->val);
goto ERROR;
}
+ // Enable all CPU features that CPU has to offer
+ if (!pakfire_arch_is_supported_by_host(arch)) {
+ r = pakfire_jail_set_env(j, "QEMU_CPU", "max");
+ if (r)
+ goto ERROR;
+ }
+
+ // Set container UUID
+ r = pakfire_jail_set_env(j, "container_uuid", pakfire_jail_uuid(j));
+ if (r)
+ goto ERROR;
+
+ // Disable systemctl to talk to systemd
+ if (!pakfire_on_root(j->pakfire)) {
+ r = pakfire_jail_set_env(j, "SYSTEMD_OFFLINE", "1");
+ if (r)
+ goto ERROR;
+ }
+
// Done
*jail = j;
return 0;
return NULL;
}
-static const char* pakfire_jail_uuid(struct pakfire_jail* jail) {
- if (!*jail->__uuid)
- uuid_unparse_lower(jail->uuid, jail->__uuid);
+// Logging Callback
- return jail->__uuid;
+PAKFIRE_EXPORT void pakfire_jail_set_log_callback(struct pakfire_jail* jail,
+ pakfire_jail_log_callback callback, void* data) {
+ jail->callbacks.log = callback;
+ jail->callbacks.log_data = data;
}
// Resource Limits
return -1;
}
- char buffer[strlen(key) + 2];
- pakfire_string_format(buffer, "%s=", key);
+ const size_t length = strlen(key);
for (unsigned int i = 0; jail->env[i]; i++) {
- if (pakfire_string_startswith(jail->env[i], buffer))
+ if ((pakfire_string_startswith(jail->env[i], key)
+ && *(jail->env[i] + length) == '=')) {
return i;
+ }
}
// Nothing found
return 0;
}
+// Timeout
+
+PAKFIRE_EXPORT int pakfire_jail_set_timeout(
+ struct pakfire_jail* jail, unsigned int timeout) {
+ // Store value
+ jail->timeout.it_value.tv_sec = timeout;
+
+ if (timeout > 0)
+ DEBUG(jail->pakfire, "Timeout set to %u second(s)\n", timeout);
+ else
+ DEBUG(jail->pakfire, "Timeout disabled\n");
+
+ return 0;
+}
+
+static int pakfire_jail_create_timer(struct pakfire_jail* jail) {
+ int r;
+
+ // Nothing to do if no timeout has been set
+ if (!jail->timeout.it_value.tv_sec)
+ return -1;
+
+ // Create a new timer
+ const int fd = timerfd_create(CLOCK_MONOTONIC, 0);
+ if (fd < 0) {
+ ERROR(jail->pakfire, "Could not create timer: %m\n");
+ goto ERROR;
+ }
+
+ // Arm timer
+ r = timerfd_settime(fd, 0, &jail->timeout, NULL);
+ if (r) {
+ ERROR(jail->pakfire, "Could not arm timer: %m\n");
+ goto ERROR;
+ }
+
+ return fd;
+
+ERROR:
+ if (fd >= 0)
+ close(fd);
+
+ return -1;
+}
+
/*
This function replaces any logging in the child process.
All log messages will be sent to the parent process through their respective pipes.
*/
-static void pakfire_jail_log(void* data, int priority, const char* file,
+static void pakfire_jail_log_redirect(void* data, int priority, const char* file,
int line, const char* fn, const char* format, va_list args) {
struct pakfire_jail_pipes* pipes = (struct pakfire_jail_pipes*)data;
int fd;
}
// Send the log message
- if (fd)
+ if (fd >= 0)
vdprintf(fd, format, args);
}
return (sizeof(buffer->data) == buffer->used);
}
-/*
- This function reads as much data as it can from the file descriptor.
- If it finds a whole line in it, it will send it to the logger and repeat the process.
- If not newline character is found, it will try to read more data until it finds one.
-*/
-static int pakfire_jail_handle_log(struct pakfire_jail* jail,
- struct pakfire_jail_exec* ctx, int priority, int fd,
- struct pakfire_log_buffer* buffer, pakfire_jail_communicate_out callback, void* data) {
- char line[BUFFER_SIZE + 1];
+static int pakfire_jail_fill_buffer(struct pakfire_jail* jail, int fd, struct pakfire_log_buffer* buffer) {
+ int r;
- // Fill up buffer from fd
- if (buffer->used < sizeof(buffer->data)) {
- ssize_t bytes_read = read(fd, buffer->data + buffer->used,
- sizeof(buffer->data) - buffer->used);
-
- // Handle errors
- if (bytes_read < 0) {
- ERROR(jail->pakfire, "Could not read from fd %d: %m\n", fd);
- return -1;
+ // Skip this if there is not space left in the buffer
+ if (buffer->used >= sizeof(buffer->data))
+ return 0;
+
+ // Fill the buffer
+ r = read(fd, buffer->data + buffer->used, sizeof(buffer->data) - buffer->used);
+
+ // Handle errors
+ if (r < 0) {
+ switch (errno) {
+ case EAGAIN:
+ case EIO:
+ break;
+
+ default:
+ return -errno;
}
- // Update buffer size
- buffer->used += bytes_read;
+ // EOF
+ } else if (r == 0) {
+ // XXX What to do here?
+
+ // Successful read
+ } else {
+ buffer->used += r;
}
- // See if we have any lines that we can write
+ return 0;
+}
+
+static int pakfire_jail_drain_buffer_with_callback(struct pakfire_jail* jail,
+ struct pakfire_log_buffer* buffer, int priority, pakfire_jail_communicate_out callback, void* data) {
+ const char* eol = NULL;
+ int r;
+
while (buffer->used) {
// Search for the end of the first line
- char* eol = memchr(buffer->data, '\n', buffer->used);
+ eol = memchr(buffer->data, '\n', buffer->used);
// No newline found
if (!eol) {
- // If the buffer is full, we send the content to the logger and try again
- // This should not happen in practise
+ // If the buffer is full, we send the entire content to make space.
if (pakfire_jail_log_buffer_is_full(buffer)) {
- DEBUG(jail->pakfire, "Logging buffer is full. Sending all content\n");
+ CTX_DEBUG(jail->ctx, "Buffer is full. Sending all content\n");
- eol = buffer->data + sizeof(buffer->data) - 1;
+ eol = buffer->data + buffer->used - 1;
- // Otherwise we might have only read parts of the output
- } else
+ // Otherwise we might have only read parts of the output...
+ } else {
break;
+ }
}
// Find the length of the string
- size_t length = eol - buffer->data + 1;
-
- // Copy the line into the buffer
- memcpy(line, buffer->data, length);
+ const size_t length = eol - buffer->data + 1;
- // Terminate the string
- line[length] = '\0';
-
- // Log the line
- if (callback) {
- int r = callback(jail->pakfire, data, priority, line, length);
- if (r) {
- ERROR(jail->pakfire, "The logging callback returned an error: %d\n", r);
- return r;
- }
+ // Call the callback
+ r = callback(jail->pakfire, data, priority, buffer->data, length);
+ if (r) {
+ CTX_ERROR(jail->ctx, "The logging callback returned an error: %d\n", r);
+ return r;
}
// Remove line from buffer
return 0;
}
+static int pakfire_jail_drain_buffer(struct pakfire_jail* jail, int fd, struct pakfire_log_buffer* buffer) {
+ int r;
+
+ // Nothing to do if the buffer is empty
+ if (!buffer->used)
+ return 0;
+
+ // Drain the buffer
+ r = write(fd, buffer->data, buffer->used);
+
+ // Handle errors
+ if (r < 0) {
+ switch (errno) {
+ case EAGAIN:
+ case EIO:
+ break;
+
+ default:
+ return -errno;
+ }
+
+ // Successful write
+ } else {
+ memmove(buffer->data, buffer->data + r, buffer->used - r);
+
+ buffer->used -= r;
+ }
+
+ return 0;
+}
+
+/*
+ This function reads as much data as it can from the file descriptor.
+ If it finds a whole line in it, it will send it to the logger and repeat the process.
+ If not newline character is found, it will try to read more data until it finds one.
+*/
+static int pakfire_jail_handle_log(struct pakfire_jail* jail, struct pakfire_jail_exec* ctx,
+ int priority, int fd, struct pakfire_log_buffer* buffer,
+ pakfire_jail_communicate_out callback, void* data) {
+ int r;
+
+ // Fill up buffer from fd
+ r = pakfire_jail_fill_buffer(jail, fd, buffer);
+ if (r)
+ return r;
+
+ // Drain the buffer
+ r = pakfire_jail_drain_buffer_with_callback(jail, buffer, priority, callback, data);
+ if (r)
+ return r;
+
+ return 0;
+}
+
+#if 0
+static int pakfire_jail_stream_stdin(struct pakfire_jail* jail,
+ struct pakfire_jail_exec* ctx, const int fd) {
+ int r;
+
+ // Nothing to do if there is no stdin callback set
+ if (!ctx->communicate.in) {
+ DEBUG(jail->pakfire, "Callback for standard input is not set\n");
+ return 0;
+ }
+
+ // Skip if the writing pipe has already been closed
+ if (ctx->pipes.stdin[1] < 0)
+ return 0;
+
+ DEBUG(jail->pakfire, "Streaming standard input...\n");
+
+ // Calling the callback
+ r = ctx->communicate.in(jail->pakfire, ctx->communicate.data, fd);
+
+ DEBUG(jail->pakfire, "Standard input callback finished: %d\n", r);
+
+ // The callback signaled that it has written everything
+ if (r == EOF) {
+ DEBUG(jail->pakfire, "Closing standard input pipe\n");
+
+ // Close the file-descriptor
+ close(fd);
+
+ // Reset the file-descriptor so it won't be closed again later
+ ctx->pipes.stdin[1] = -1;
+
+ // Report success
+ r = 0;
+ }
+
+ return r;
+}
+#endif
+
+static int pakfire_jail_recv_fd(struct pakfire_jail* jail, int socket, int* fd) {
+ const size_t payload_length = sizeof(fd);
+ char buffer[CMSG_SPACE(payload_length)];
+ int r;
+
+ struct msghdr msg = {
+ .msg_control = buffer,
+ .msg_controllen = sizeof(buffer),
+ };
+
+ // Receive the message
+ r = recvmsg(socket, &msg, 0);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Could not receive file descriptor: %s\n", strerror(errno));
+ return -errno;
+ }
+
+ // Fetch the payload
+ struct cmsghdr* cmsg = CMSG_FIRSTHDR(&msg);
+ if (!cmsg)
+ return -EBADMSG;
+
+ *fd = *((int*)CMSG_DATA(cmsg));
+
+ CTX_DEBUG(jail->ctx, "Received fd %d from socket %d\n", *fd, socket);
+
+ return 0;
+}
+
+static int pakfire_jail_send_fd(struct pakfire_jail* jail, int socket, int fd) {
+ const size_t payload_length = sizeof(fd);
+ char buffer[CMSG_SPACE(payload_length)];
+ int r;
+
+ CTX_DEBUG(jail->ctx, "Sending fd %d to socket %d\n", fd, socket);
+
+ // Header
+ struct msghdr msg = {
+ .msg_control = buffer,
+ .msg_controllen = sizeof(buffer),
+ };
+
+ // Payload
+ struct cmsghdr* cmsg = CMSG_FIRSTHDR(&msg);
+ cmsg->cmsg_level = SOL_SOCKET;
+ cmsg->cmsg_type = SCM_RIGHTS;
+ cmsg->cmsg_len = CMSG_LEN(payload_length);
+
+ // Set payload
+ *((int*)CMSG_DATA(cmsg)) = fd;
+
+ // Send the message
+ r = sendmsg(socket, &msg, 0);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Could not send file descriptor: %s\n", strerror(errno));
+ return -errno;
+ }
+
+ return 0;
+}
+
static int pakfire_jail_setup_pipe(struct pakfire_jail* jail, int (*fds)[2], const int flags) {
int r = pipe2(*fds, flags);
if (r < 0) {
static void pakfire_jail_close_pipe(struct pakfire_jail* jail, int fds[2]) {
for (unsigned int i = 0; i < 2; i++)
- if (fds[i])
+ if (fds[i] >= 0)
close(fds[i]);
}
This is a convenience function to fetch the reading end of a pipe and
closes the write end.
*/
-static int pakfire_jail_get_pipe(struct pakfire_jail* jail, int (*fds)[2]) {
+static int pakfire_jail_get_pipe_to_read(struct pakfire_jail* jail, int (*fds)[2]) {
// Give the variables easier names to avoid confusion
int* fd_read = &(*fds)[0];
int* fd_write = &(*fds)[1];
// Close the write end of the pipe
- if (*fd_write) {
+ if (*fd_write >= 0) {
close(*fd_write);
- *fd_write = 0;
+ *fd_write = -1;
}
// Return the read end
- return *fd_read;
+ if (*fd_read >= 0)
+ return *fd_read;
+
+ return -1;
+}
+
+static int pakfire_jail_get_pipe_to_write(struct pakfire_jail* jail, int (*fds)[2]) {
+ // Give the variables easier names to avoid confusion
+ int* fd_read = &(*fds)[0];
+ int* fd_write = &(*fds)[1];
+
+ // Close the read end of the pipe
+ if (*fd_read >= 0) {
+ close(*fd_read);
+ *fd_read = -1;
+ }
+
+ // Return the write end
+ if (*fd_write >= 0)
+ return *fd_write;
+
+ return -1;
+}
+
+static int pakfire_jail_log(struct pakfire* pakfire, void* data, int priority,
+ const char* line, const size_t length) {
+ // Pass everything to the parent logger
+ pakfire_log_condition(pakfire, priority, 0, "%.*s", (int)length, line);
+
+ return 0;
+}
+
+static int pakfire_jail_epoll_add_fd(struct pakfire_jail* jail, int epollfd, int fd, int events) {
+ struct epoll_event event = {
+ .events = events|EPOLLHUP,
+ .data = {
+ .fd = fd,
+ },
+ };
+ int r;
+
+ // Read flags
+ int flags = fcntl(fd, F_GETFL, 0);
+
+ // Set modified flags
+ r = fcntl(fd, F_SETFL, flags|O_NONBLOCK);
+ if (r < 0) {
+ CTX_ERROR(jail->ctx, "Could not set file descriptor %d into non-blocking mode: %s\n",
+ fd, strerror(errno));
+ return -errno;
+ }
+
+ // Add the file descriptor to the loop
+ r = epoll_ctl(epollfd, EPOLL_CTL_ADD, fd, &event);
+ if (r < 0) {
+ ERROR(jail->pakfire, "Could not add file descriptor %d to epoll(): %s\n",
+ fd, strerror(errno));
+ return -errno;
+ }
+
+ return 0;
+}
+
+// PTY Forwarding
+
+static int pakfire_jail_enable_raw_mode(struct pakfire_jail* jail,
+ struct pakfire_jail_pty_stdio* stdio) {
+ struct termios raw_attrs;
+ int r;
+
+ // Store flags
+ stdio->fdflags = fcntl(stdio->fd, F_GETFL);
+ if (stdio->fdflags < 0) {
+ CTX_ERROR(jail->ctx, "Could not fetch flags from fd %d: %s\n",
+ stdio->fd, strerror(errno));
+ return -errno;
+ }
+
+ // Fetch all attributes
+ r = tcgetattr(stdio->fd, &stdio->attrs);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Could not fetch terminal attributes from fd %d: %s\n",
+ stdio->fd, strerror(errno));
+ return -errno;
+ }
+
+ // Copy all attributes
+ raw_attrs = stdio->attrs;
+
+ // Make it RAW
+ cfmakeraw(&raw_attrs);
+
+ switch (stdio->fd) {
+ case STDIN_FILENO:
+ raw_attrs.c_oflag = stdio->attrs.c_oflag;
+ break;
+
+ case STDOUT_FILENO:
+ raw_attrs.c_iflag = stdio->attrs.c_iflag;
+ raw_attrs.c_lflag = stdio->attrs.c_lflag;
+ break;
+ }
+
+ // Restore the attributes
+ r = tcsetattr(stdio->fd, TCSANOW, &raw_attrs);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Could not restore terminal attributes for fd %d: %s\n",
+ stdio->fd, strerror(errno));
+ return -errno;
+ }
+
+ return 0;
+}
+
+static int pakfire_jail_restore_attrs(struct pakfire_jail* jail,
+ const struct pakfire_jail_pty_stdio* stdio) {
+ int r;
+
+ // Restore the flags
+ r = fcntl(stdio->fd, F_SETFL, stdio->fdflags);
+ if (r < 0) {
+ CTX_ERROR(jail->ctx, "Could not set flags for file descriptor %d: %s\n",
+ stdio->fd, strerror(errno));
+ return -errno;
+ }
+
+ // Restore the attributes
+ r = tcsetattr(stdio->fd, TCSANOW, &stdio->attrs);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Could not restore terminal attributes for %d, ignoring: %s\n",
+ stdio->fd, strerror(errno));
+ return -errno;
+ }
+
+ return 0;
+}
+
+static int pakfire_jail_setup_pty_forwarding(struct pakfire_jail* jail,
+ struct pakfire_jail_exec* ctx, const int epollfd, const int fd) {
+ struct winsize size;
+ int r;
+
+ CTX_DEBUG(jail->ctx, "Setting up PTY forwarding on fd %d\n", fd);
+
+ // Store the file descriptor
+ ctx->pty.master.fd = fd;
+
+ // Configure stdin/stdout
+ ctx->pty.stdin.fd = STDIN_FILENO;
+ ctx->pty.stdout.fd = STDOUT_FILENO;
+
+ // Fetch dimensions
+ r = ioctl(ctx->pty.stdout.fd, TIOCGWINSZ, &size);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Failed to determine terminal dimensions: %s\n", strerror(errno));
+ return -errno;
+ }
+
+ // Set dimensions
+ r = ioctl(ctx->pty.master.fd, TIOCSWINSZ, &size);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Failed setting dimensions: %s\n", strerror(errno));
+ return -errno;
+ }
+
+ // Enable RAW mode on standard input
+ r = pakfire_jail_enable_raw_mode(jail, &ctx->pty.stdin);
+ if (r)
+ return r;
+
+ // Enable RAW mode on standard output
+ r = pakfire_jail_enable_raw_mode(jail, &ctx->pty.stdout);
+ if (r)
+ return r;
+
+ // Add the master to the event loop
+ r = pakfire_jail_epoll_add_fd(jail, epollfd, ctx->pty.master.fd, EPOLLIN|EPOLLOUT|EPOLLET);
+ if (r)
+ return r;
+
+ // Add standard input to the event loop
+ r = pakfire_jail_epoll_add_fd(jail, epollfd, ctx->pty.stdin.fd, EPOLLIN|EPOLLET);
+ if (r)
+ return r;
+
+ // Add standard output to the event loop
+ r = pakfire_jail_epoll_add_fd(jail, epollfd, ctx->pty.stdout.fd, EPOLLOUT|EPOLLET);
+ if (r)
+ return r;
+
+ return 0;
+}
+
+static int pakfire_jail_forward_pty(struct pakfire_jail* jail, struct pakfire_jail_exec* ctx) {
+ int r;
+
+ // Read from standard input
+ if (ctx->pty.stdin.flags & PAKFIRE_JAIL_PTY_READY_TO_READ) {
+ r = pakfire_jail_fill_buffer(jail, ctx->pty.stdin.fd, &ctx->pty.stdin.buffer);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Failed reading from standard input: %s\n", strerror(-r));
+ return r;
+ }
+
+ // We are done reading for now
+ ctx->pty.stdin.flags &= ~PAKFIRE_JAIL_PTY_READY_TO_READ;
+
+ // But we may have data to write
+ if (ctx->pty.stdin.buffer.used)
+ ctx->pty.master.flags |= PAKFIRE_JAIL_PTY_READY_TO_WRITE;
+ }
+
+ // Write to the master
+ if (ctx->pty.master.flags & PAKFIRE_JAIL_PTY_READY_TO_WRITE) {
+ r = pakfire_jail_drain_buffer(jail, ctx->pty.master.fd, &ctx->pty.stdin.buffer);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Failed writing to the PTY: %s\n", strerror(-r));
+ return r;
+ }
+
+ // We are done writing for now
+ ctx->pty.master.flags &= ~PAKFIRE_JAIL_PTY_READY_TO_WRITE;
+ }
+
+ // Read from the master
+ if (ctx->pty.master.flags & PAKFIRE_JAIL_PTY_READY_TO_READ) {
+ r = pakfire_jail_fill_buffer(jail, ctx->pty.master.fd, &ctx->pty.stdout.buffer);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Failed reading from the PTY: %s\n", strerror(-r));
+ return r;
+ }
+
+ // We are done reading for now
+ ctx->pty.master.flags &= ~PAKFIRE_JAIL_PTY_READY_TO_READ;
+
+ // But we may have data to write
+ if (ctx->pty.stdout.buffer.used)
+ ctx->pty.stdout.flags |= PAKFIRE_JAIL_PTY_READY_TO_WRITE;
+ }
+
+ // Write to standard output
+ if (ctx->pty.stdout.flags & PAKFIRE_JAIL_PTY_READY_TO_WRITE) {
+ r = pakfire_jail_drain_buffer(jail, ctx->pty.stdout.fd, &ctx->pty.stdout.buffer);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Failed writing to standard output: %s\n", strerror(-r));
+ return r;
+ }
+
+ // We are done writing for now
+ ctx->pty.stdout.flags &= ~PAKFIRE_JAIL_PTY_READY_TO_WRITE;
+ }
+
+ return 0;
}
static int pakfire_jail_wait(struct pakfire_jail* jail, struct pakfire_jail_exec* ctx) {
int epollfd = -1;
- struct epoll_event ev;
struct epoll_event events[EPOLL_MAX_EVENTS];
+ char garbage[8];
int r = 0;
// Fetch file descriptors from context
- const int stdout = pakfire_jail_get_pipe(jail, &ctx->pipes.stdout);
- const int stderr = pakfire_jail_get_pipe(jail, &ctx->pipes.stderr);
const int pidfd = ctx->pidfd;
+ // Fetch the UNIX domain socket
+ const int socket_recv = pakfire_jail_get_pipe_to_read(jail, &ctx->socket);
+
+ // Timer
+ const int timerfd = pakfire_jail_create_timer(jail);
+
// Logging
- const int log_INFO = pakfire_jail_get_pipe(jail, &ctx->pipes.log_INFO);
- const int log_ERROR = pakfire_jail_get_pipe(jail, &ctx->pipes.log_ERROR);
- const int log_DEBUG = pakfire_jail_get_pipe(jail, &ctx->pipes.log_DEBUG);
+ const int log_INFO = pakfire_jail_get_pipe_to_read(jail, &ctx->pipes.log_INFO);
+ const int log_ERROR = pakfire_jail_get_pipe_to_read(jail, &ctx->pipes.log_ERROR);
+#ifdef ENABLE_DEBUG
+ const int log_DEBUG = pakfire_jail_get_pipe_to_read(jail, &ctx->pipes.log_DEBUG);
+#endif /* ENABLE_DEBUG */
// Make a list of all file descriptors we are interested in
- int fds[] = {
- stdout, stderr, pidfd, log_INFO, log_ERROR, log_DEBUG,
+ const struct pakfire_wait_fds {
+ const int fd;
+ const int events;
+ } fds[] = {
+ // Timer
+ { timerfd, EPOLLIN },
+
+ // Child Process
+ { ctx->pidfd, EPOLLIN },
+
+ // Log Pipes
+ { log_INFO, EPOLLIN },
+ { log_ERROR, EPOLLIN },
+#ifdef ENABLE_DEBUG
+ { log_DEBUG, EPOLLIN },
+#endif /* ENABLE_DEBUG */
+
+ // UNIX Domain Socket
+ { socket_recv, EPOLLIN },
+
+ // Sentinel
+ { -1, 0 },
};
// Setup epoll
goto ERROR;
}
- ev.events = EPOLLIN|EPOLLHUP;
-
// Turn file descriptors into non-blocking mode and add them to epoll()
- for (unsigned int i = 0; i < sizeof(fds) / sizeof(*fds); i++) {
- int fd = fds[i];
-
+ for (const struct pakfire_wait_fds* fd = fds; fd->events; fd++) {
// Skip fds which were not initialized
- if (fd <= 0)
+ if (fd->fd < 0)
continue;
- ev.data.fd = fd;
-
- if (epoll_ctl(epollfd, EPOLL_CTL_ADD, fd, &ev) < 0) {
- ERROR(jail->pakfire, "Could not add file descriptor %d to epoll(): %m\n", fd);
- r = 1;
+ // Add the FD to the event loop
+ r = pakfire_jail_epoll_add_fd(jail, epollfd, fd->fd, fd->events);
+ if (r)
goto ERROR;
- }
}
int ended = 0;
int e = events[i].events;
int fd = events[i].data.fd;
- struct pakfire_log_buffer* buffer = NULL;
- pakfire_jail_communicate_out callback = NULL;
- void* data = NULL;
- int priority;
+ // Handle PTY forwarding events
+ if (ctx->pty.master.fd == fd) {
+ if (e & (EPOLLIN|EPOLLHUP))
+ ctx->pty.master.flags |= PAKFIRE_JAIL_PTY_READY_TO_READ;
- // Check if there is any data to be read
- if (e & EPOLLIN) {
- // Handle any changes to the PIDFD
- if (fd == pidfd) {
+ if (e & (EPOLLOUT|EPOLLHUP))
+ ctx->pty.master.flags |= PAKFIRE_JAIL_PTY_READY_TO_WRITE;
+
+ // Perform the work
+ r = pakfire_jail_forward_pty(jail, ctx);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Failed forwarding the PTY: %s\n", strerror(-r));
+ goto ERROR;
+ }
+
+ // Handle standard input
+ } else if (ctx->pty.stdin.fd == fd) {
+ if (e & (EPOLLIN|EPOLLHUP))
+ ctx->pty.stdin.flags |= PAKFIRE_JAIL_PTY_READY_TO_READ;
+
+ // Perform the work
+ r = pakfire_jail_forward_pty(jail, ctx);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Failed forwarding the PTY: %s\n", strerror(-r));
+ goto ERROR;
+ }
+
+ // Handle standard output
+ } else if (ctx->pty.stdout.fd == fd) {
+ if (e & (EPOLLOUT|EPOLLHUP))
+ ctx->pty.stdout.flags |= PAKFIRE_JAIL_PTY_READY_TO_WRITE;
+
+ // Perform the work
+ r = pakfire_jail_forward_pty(jail, ctx);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Failed forwarding the PTY: %s\n", strerror(-r));
+ goto ERROR;
+ }
+
+ // Handle any changes to the PIDFD
+ } else if (pidfd == fd) {
+ if (e & EPOLLIN) {
// Call waidid() and store the result
r = waitid(P_PIDFD, ctx->pidfd, &ctx->status, WEXITED);
if (r) {
// Mark that we have ended so that we will process the remaining
// events from epoll() now, but won't restart the outer loop.
ended = 1;
- continue;
-
- // Handle logging messages
- } else if (fd == log_INFO) {
- buffer = &ctx->buffers.log_INFO;
- priority = LOG_INFO;
-
- callback = pakfire_jail_default_log_callback;
+ }
- } else if (fd == log_ERROR) {
- buffer = &ctx->buffers.log_ERROR;
- priority = LOG_ERR;
+ // Handle timer events
+ } else if (timerfd == fd) {
+ if (e & EPOLLIN) {
+ DEBUG(jail->pakfire, "Timer event received\n");
- callback = pakfire_jail_default_log_callback;
+ // Disarm the timer
+ r = read(timerfd, garbage, sizeof(garbage));
+ if (r < 1) {
+ ERROR(jail->pakfire, "Could not disarm timer: %m\n");
+ r = 1;
+ goto ERROR;
+ }
- } else if (fd == log_DEBUG) {
- buffer = &ctx->buffers.log_DEBUG;
- priority = LOG_DEBUG;
+ // Terminate the process if it hasn't already ended
+ if (!ended) {
+ DEBUG(jail->pakfire, "Terminating process...\n");
- callback = pakfire_jail_default_log_callback;
+ // Send SIGTERM to the process
+ r = pidfd_send_signal(pidfd, SIGKILL, NULL, 0);
+ if (r) {
+ ERROR(jail->pakfire, "Could not kill process: %m\n");
+ goto ERROR;
+ }
+ }
+ }
- // Handle anything from the log pipes
- } else if (fd == stdout) {
- buffer = &ctx->buffers.stdout;
- priority = LOG_INFO;
+ // Handle socket messages
+ } else if (socket_recv == fd) {
+ if (e & EPOLLIN) {
+ // Receive the passed FD
+ r = pakfire_jail_recv_fd(jail, socket_recv, &fd);
+ if (r)
+ goto ERROR;
- callback = ctx->communicate.out;
- data = ctx->communicate.data;
+ // Setup PTY forwarding
+ if (ctx->pty.master.fd < 0) {
+ r = pakfire_jail_setup_pty_forwarding(jail, ctx, epollfd, fd);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Failed setting up PTY forwarding: %s\n", strerror(-r));
+ goto ERROR;
+ }
+ }
+ }
- } else if (fd == stderr) {
- buffer = &ctx->buffers.stderr;
- priority = LOG_ERR;
+ // Handle log INFO messages
+ } else if (log_INFO == fd) {
+ if (e & EPOLLIN) {
+ r = pakfire_jail_handle_log(jail, ctx, LOG_INFO, fd,
+ &ctx->buffers.log_INFO, pakfire_jail_log, NULL);
+ if (r)
+ goto ERROR;
+ }
- callback = ctx->communicate.out;
- data = ctx->communicate.data;
+ // Handle log ERROR messages
+ } else if (log_ERROR == fd) {
+ if (e & EPOLLIN) {
+ r = pakfire_jail_handle_log(jail, ctx, LOG_ERR, fd,
+ &ctx->buffers.log_ERROR, pakfire_jail_log, NULL);
+ if (r)
+ goto ERROR;
+ }
- } else {
- DEBUG(jail->pakfire, "Received invalid file descriptor %d\n", fd);
- continue;
+#ifdef ENABLE_DEBUG
+ // Handle log DEBUG messages
+ } else if (log_DEBUG == fd) {
+ if (e & EPOLLIN) {
+ r = pakfire_jail_handle_log(jail, ctx, LOG_DEBUG, fd,
+ &ctx->buffers.log_DEBUG, pakfire_jail_log, NULL);
+ if (r)
+ goto ERROR;
}
+#endif /* ENABLE_DEBUG */
- // Handle log event
- r = pakfire_jail_handle_log(jail, ctx, priority, fd, buffer, callback, data);
- if (r)
- goto ERROR;
+ // Log a message for anything else
+ } else {
+ DEBUG(jail->pakfire, "Received invalid file descriptor %d\n", fd);
+ continue;
}
// Check if any file descriptors have been closed
}
ERROR:
- if (epollfd > 0)
+ if (epollfd >= 0)
close(epollfd);
+ if (timerfd >= 0)
+ close(timerfd);
+
+ // Restore any changed terminal attributes
+ if (ctx->pty.stdin.fd >= 0)
+ pakfire_jail_restore_attrs(jail, &ctx->pty.stdin);
+ if (ctx->pty.stdout.fd >= 0)
+ pakfire_jail_restore_attrs(jail, &ctx->pty.stdout);
return r;
}
// Capabilities
-static int pakfire_jail_drop_capabilities(struct pakfire_jail* jail) {
- const int capabilities[] = {
- // Deny access to the kernel's audit system
- CAP_AUDIT_CONTROL,
- CAP_AUDIT_READ,
- CAP_AUDIT_WRITE,
-
- // Deny suspending block devices
- CAP_BLOCK_SUSPEND,
-
- // Deny any stuff with BPF
- CAP_BPF,
-
- // Deny checkpoint restore
- CAP_CHECKPOINT_RESTORE,
-
- // Deny opening files by inode number (open_by_handle_at)
- CAP_DAC_READ_SEARCH,
+// Logs all capabilities of the current process
+static int pakfire_jail_show_capabilities(struct pakfire_jail* jail) {
+ cap_t caps = NULL;
+ char* name = NULL;
+ cap_flag_value_t value_e;
+ cap_flag_value_t value_i;
+ cap_flag_value_t value_p;
+ int r;
- // Deny setting SUID bits
- CAP_FSETID,
+ // Fetch PID
+ pid_t pid = getpid();
- // Deny locking more memory
- CAP_IPC_LOCK,
+ // Fetch all capabilities
+ caps = cap_get_proc();
+ if (!caps) {
+ ERROR(jail->pakfire, "Could not fetch capabilities: %m\n");
+ r = 1;
+ goto ERROR;
+ }
- // Deny modifying any Apparmor/SELinux/SMACK configuration
- CAP_MAC_ADMIN,
- CAP_MAC_OVERRIDE,
+ DEBUG(jail->pakfire, "Capabilities of PID %d:\n", pid);
- // Deny creating any special devices
- CAP_MKNOD,
+ // Iterate over all capabilities
+ for (unsigned int cap = 0; cap_valid(cap); cap++) {
+ name = cap_to_name(cap);
- // Deny setting any capabilities
- CAP_SETFCAP,
+ // Fetch effective value
+ r = cap_get_flag(caps, cap, CAP_EFFECTIVE, &value_e);
+ if (r)
+ goto ERROR;
- // Deny reading from syslog
- CAP_SYSLOG,
+ // Fetch inheritable value
+ r = cap_get_flag(caps, cap, CAP_INHERITABLE, &value_i);
+ if (r)
+ goto ERROR;
- // Deny any admin actions (mount, sethostname, ...)
- CAP_SYS_ADMIN,
+ // Fetch permitted value
+ r = cap_get_flag(caps, cap, CAP_PERMITTED, &value_p);
+ if (r)
+ goto ERROR;
- // Deny rebooting the system
- CAP_SYS_BOOT,
+ DEBUG(jail->pakfire,
+ " %-24s : %c%c%c\n",
+ name,
+ (value_e == CAP_SET) ? 'e' : '-',
+ (value_i == CAP_SET) ? 'i' : '-',
+ (value_p == CAP_SET) ? 'p' : '-'
+ );
- // Deny loading kernel modules
- CAP_SYS_MODULE,
+ // Free name
+ cap_free(name);
+ name = NULL;
+ }
- // Deny setting nice level
- CAP_SYS_NICE,
+ // Success
+ r = 0;
- // Deny access to /proc/kcore, /dev/mem, /dev/kmem
- CAP_SYS_RAWIO,
+ERROR:
+ if (name)
+ cap_free(name);
+ if (caps)
+ cap_free(caps);
- // Deny circumventing any resource limits
- CAP_SYS_RESOURCE,
+ return r;
+}
- // Deny setting the system time
- CAP_SYS_TIME,
+static int pakfire_jail_set_capabilities(struct pakfire_jail* jail) {
+ cap_t caps = NULL;
+ char* name = NULL;
+ int r;
- // Deny playing with suspend
- CAP_WAKE_ALARM,
+ // Fetch capabilities
+ caps = cap_get_proc();
+ if (!caps) {
+ ERROR(jail->pakfire, "Could not read capabilities: %m\n");
+ r = 1;
+ goto ERROR;
+ }
- 0,
- };
+ // Walk through all capabilities
+ for (cap_value_t cap = 0; cap_valid(cap); cap++) {
+ cap_value_t _caps[] = { cap };
- DEBUG(jail->pakfire, "Dropping capabilities...\n");
+ // Fetch the name of the capability
+ name = cap_to_name(cap);
- size_t num_caps = 0;
- int r;
+ r = cap_set_flag(caps, CAP_EFFECTIVE, 1, _caps, CAP_SET);
+ if (r) {
+ ERROR(jail->pakfire, "Could not set %s: %m\n", name);
+ goto ERROR;
+ }
- // Drop any capabilities
- for (const int* cap = capabilities; *cap; cap++) {
- r = prctl(PR_CAPBSET_DROP, *cap, 0, 0, 0);
+ r = cap_set_flag(caps, CAP_INHERITABLE, 1, _caps, CAP_SET);
if (r) {
- ERROR(jail->pakfire, "Could not drop capability %d: %m\n", *cap);
- return r;
+ ERROR(jail->pakfire, "Could not set %s: %m\n", name);
+ goto ERROR;
}
- num_caps++;
- }
+ r = cap_set_flag(caps, CAP_PERMITTED, 1, _caps, CAP_SET);
+ if (r) {
+ ERROR(jail->pakfire, "Could not set %s: %m\n", name);
+ goto ERROR;
+ }
- // Fetch any capabilities
- cap_t caps = cap_get_proc();
- if (!caps) {
- ERROR(jail->pakfire, "Could not read capabilities: %m\n");
- return 1;
+ // Free name
+ cap_free(name);
+ name = NULL;
}
- /*
- Set inheritable capabilities
-
- This ensures that no processes will be able to gain any of the listed
- capabilities again.
- */
- r = cap_set_flag(caps, CAP_INHERITABLE, num_caps, capabilities, CAP_CLEAR);
+ // Restore all capabilities
+ r = cap_set_proc(caps);
if (r) {
- ERROR(jail->pakfire, "cap_set_flag() failed: %m\n");
+ ERROR(jail->pakfire, "Restoring capabilities failed: %m\n");
goto ERROR;
}
- // Restore capabilities
- r = cap_set_proc(caps);
- if (r) {
- ERROR(jail->pakfire, "Could not restore capabilities: %m\n");
- goto ERROR;
+ // Add all capabilities to the ambient set
+ for (unsigned int cap = 0; cap_valid(cap); cap++) {
+ name = cap_to_name(cap);
+
+ // Raise the capability
+ r = prctl(PR_CAP_AMBIENT, PR_CAP_AMBIENT_RAISE, cap, 0, 0);
+ if (r) {
+ ERROR(jail->pakfire, "Could not set ambient capability %s: %m\n", name);
+ goto ERROR;
+ }
+
+ // Free name
+ cap_free(name);
+ name = NULL;
}
+ // Success
+ r = 0;
+
ERROR:
+ if (name)
+ cap_free(name);
if (caps)
cap_free(caps);
// Bind-mount all paths read-only
for (const char** path = paths; *path; path++) {
r = pakfire_bind(jail->pakfire, *path, NULL, MS_RDONLY);
- if (r)
+ if (r) {
+ switch (errno) {
+ // Ignore if we don't have permission
+ case EPERM:
+ continue;
+
+ default:
+ break;
+ }
return r;
+ }
}
return 0;
*/
static int pakfire_jail_mount(struct pakfire_jail* jail, struct pakfire_jail_exec* ctx) {
struct pakfire_jail_mountpoint* mp = NULL;
+ int flags = 0;
int r;
+ // Enable loop devices
+ if (pakfire_jail_exec_has_flag(ctx, PAKFIRE_JAIL_HAS_LOOP_DEVICES))
+ flags |= PAKFIRE_MOUNT_LOOP_DEVICES;
+
// Mount all default stuff
- r = pakfire_mount_all(jail->pakfire);
+ r = pakfire_mount_all(jail->pakfire, PAKFIRE_MNTNS_OUTER, flags);
if (r)
return r;
- // Mount networking stuff
- if (pakfire_jail_exec_has_flag(ctx, PAKFIRE_JAIL_HAS_NETWORKING)) {
- r = pakfire_jail_mount_networking(jail);
- if (r)
- return r;
+ // Populate /dev
+ r = pakfire_populate_dev(jail->pakfire, flags);
+ if (r)
+ return r;
+
+ // Mount the interpreter (if needed)
+ r = pakfire_mount_interpreter(jail->pakfire);
+ if (r)
+ return r;
+
+ // Mount networking stuff
+ if (pakfire_jail_exec_has_flag(ctx, PAKFIRE_JAIL_HAS_NETWORKING)) {
+ r = pakfire_jail_mount_networking(jail);
+ if (r)
+ return r;
+ }
+
+ // Mount all custom stuff
+ for (unsigned int i = 0; i < jail->num_mountpoints; i++) {
+ // Fetch mountpoint
+ mp = &jail->mountpoints[i];
+
+ // Mount it
+ r = pakfire_bind(jail->pakfire, mp->source, mp->target, mp->flags);
+ if (r)
+ return r;
+ }
+
+ return 0;
+}
+
+// Networking
+
+static int pakfire_jail_setup_loopback(struct pakfire_jail* jail) {
+ struct nl_sock* nl = NULL;
+ struct nl_cache* cache = NULL;
+ struct rtnl_link* link = NULL;
+ struct rtnl_link* change = NULL;
+ int r;
+
+ DEBUG(jail->pakfire, "Setting up loopback...\n");
+
+ // Allocate a netlink socket
+ nl = nl_socket_alloc();
+ if (!nl) {
+ ERROR(jail->pakfire, "Could not allocate a netlink socket: %m\n");
+ r = 1;
+ goto ERROR;
+ }
+
+ // Connect the socket
+ r = nl_connect(nl, NETLINK_ROUTE);
+ if (r) {
+ ERROR(jail->pakfire, "Could not connect netlink socket: %s\n", nl_geterror(r));
+ goto ERROR;
+ }
+
+ // Allocate the netlink cache
+ r = rtnl_link_alloc_cache(nl, AF_UNSPEC, &cache);
+ if (r < 0) {
+ ERROR(jail->pakfire, "Unable to allocate netlink cache: %s\n", nl_geterror(r));
+ goto ERROR;
+ }
+
+ // Fetch loopback interface
+ link = rtnl_link_get_by_name(cache, "lo");
+ if (!link) {
+ ERROR(jail->pakfire, "Could not find lo interface. Ignoring.\n");
+ r = 0;
+ goto ERROR;
+ }
+
+ // Allocate a new link
+ change = rtnl_link_alloc();
+ if (!change) {
+ ERROR(jail->pakfire, "Could not allocate change link\n");
+ r = 1;
+ goto ERROR;
}
- // Mount all custom stuff
- for (unsigned int i = 0; i < jail->num_mountpoints; i++) {
- // Fetch mountpoint
- mp = &jail->mountpoints[i];
+ // Set the link to UP
+ rtnl_link_set_flags(change, IFF_UP);
- // Mount it
- r = pakfire_bind(jail->pakfire, mp->source, mp->target, mp->flags);
- if (r)
- return r;
+ // Apply any changes
+ r = rtnl_link_change(nl, link, change, 0);
+ if (r) {
+ ERROR(jail->pakfire, "Unable to activate loopback: %s\n", nl_geterror(r));
+ goto ERROR;
}
- // Log all mountpoints
- pakfire_mount_list(jail->pakfire);
+ // Success
+ r = 0;
- return 0;
+ERROR:
+ if (nl)
+ nl_socket_free(nl);
+
+ return r;
}
// UID/GID Mapping
"0 %lu %lu\n", subgid->id, subgid->length);
} else {
r = pakfire_file_write(jail->pakfire, path, 0, 0, 0,
- "0 %lu 1\n%1 %lu %lu\n", gid, subgid->id, subgid->length);
+ "0 %lu 1\n1 %lu %lu\n", gid, subgid->id, subgid->length);
}
if (r) {
static int pakfire_jail_setgroups(struct pakfire_jail* jail, pid_t pid) {
char path[PATH_MAX];
- int r = 1;
+ int r;
// Make path
r = pakfire_string_format(path, "/proc/%d/setgroups", pid);
if (r)
return r;
- // Open file for writing
- FILE* f = fopen(path, "w");
- if (!f) {
- ERROR(jail->pakfire, "Could not open %s for writing: %m\n", path);
- goto ERROR;
- }
-
- // Write content
- int bytes_written = fprintf(f, "deny\n");
- if (bytes_written <= 0) {
- ERROR(jail->pakfire, "Could not write to %s: %m\n", path);
- goto ERROR;
- }
-
- r = fclose(f);
- f = NULL;
+ r = pakfire_file_write(jail->pakfire, path, 0, 0, 0, "deny\n");
if (r) {
- ERROR(jail->pakfire, "Could not close %s: %m\n", path);
- goto ERROR;
+ CTX_ERROR(jail->ctx, "Could not set setgroups to deny: %s\n", strerror(errno));
+ r = -errno;
}
-ERROR:
- if (f)
- fclose(f);
-
return r;
}
DEBUG(jail->pakfire, "Sending signal...\n");
// Write to the file descriptor
- ssize_t bytes_written = write(fd, &val, sizeof(val));
- if (bytes_written < 0 || (size_t)bytes_written < sizeof(val)) {
- ERROR(jail->pakfire, "Could not send signal: %m\n");
- r = 1;
+ r = eventfd_write(fd, val);
+ if (r < 0) {
+ ERROR(jail->pakfire, "Could not send signal: %s\n", strerror(errno));
+ r = -errno;
}
// Close the file descriptor
DEBUG(jail->pakfire, "Waiting for signal...\n");
- ssize_t bytes_read = read(fd, &val, sizeof(val));
- if (bytes_read < 0 || (size_t)bytes_read < sizeof(val)) {
- ERROR(jail->pakfire, "Error waiting for signal: %m\n");
- r = 1;
+ r = eventfd_read(fd, &val);
+ if (r < 0) {
+ ERROR(jail->pakfire, "Error waiting for signal: %s\n", strerror(errno));
+ r = -errno;
}
// Close the file descriptor
return r;
}
-static int pakfire_jail_stream_stdin(struct pakfire_jail* jail, struct pakfire_jail_exec* ctx) {
- int r;
-
- // Nothing to do if there is no stdin callback set
- if (!ctx->communicate.in) {
- DEBUG(jail->pakfire, "Callback for standard input is not set\n");
- return 0;
- }
-
- int* fd = &ctx->pipes.stdin[1];
-
- DEBUG(jail->pakfire, "Streaming standard input...\n");
-
- // Calling the callback
- r = ctx->communicate.in(jail->pakfire, ctx->communicate.data, *fd);
-
- DEBUG(jail->pakfire, "Standard input callback finished: %d\n", r);
-
- // Close the file descriptor when we are done
- close(*fd);
- *fd = 0;
-
- return r;
-}
-
/*
Performs the initialisation that needs to happen in the parent part
*/
if (r)
return r;
- // Stream standard input
- r = pakfire_jail_stream_stdin(jail, ctx);
+ return 0;
+}
+
+static int pakfire_jail_switch_root(struct pakfire_jail* jail, const char* root) {
+ int r;
+
+ // Change to the new root
+ r = chdir(root);
+ if (r) {
+ ERROR(jail->pakfire, "chdir(%s) failed: %m\n", root);
+ return r;
+ }
+
+ // Switch Root!
+ r = pivot_root(".", ".");
+ if (r) {
+ ERROR(jail->pakfire, "Failed changing into the new root directory %s: %m\n", root);
+ return r;
+ }
+
+ // Umount the old root
+ r = umount2(".", MNT_DETACH);
+ if (r) {
+ ERROR(jail->pakfire, "Could not umount the old root filesystem: %m\n");
+ return r;
+ }
+
+ return 0;
+}
+
+static int pakfire_jail_open_pty(struct pakfire_jail* jail, struct pakfire_jail_exec* ctx) {
+ int r;
+
+ // Allocate a new PTY
+ ctx->pty.master.fd = posix_openpt(O_RDWR|O_NONBLOCK|O_NOCTTY|O_CLOEXEC);
+ if (ctx->pty.master.fd < 0)
+ return -errno;
+
+ // Fetch the path
+ r = ptsname_r(ctx->pty.master.fd, ctx->pty.console, sizeof(ctx->pty.console));
+ if (r)
+ return -r;
+
+ CTX_DEBUG(jail->ctx, "Allocated console at %s (%d)\n", ctx->pty.console, ctx->pty.master.fd);
+
+ // Unlock the master device
+ r = unlockpt(ctx->pty.master.fd);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Could not unlock the PTY: %s\n", strerror(errno));
+ return -errno;
+ }
+
+ // Create a symlink
+ r = pakfire_symlink(jail->ctx, ctx->pty.console, "/dev/console");
if (r)
return r;
+ return r;
+}
+
+static int pakfire_jail_setup_terminal(struct pakfire_jail* jail, struct pakfire_jail_exec* ctx) {
+ int fd;
+ int r;
+
+ // Open a new terminal
+ fd = open("/dev/console", O_RDWR|O_NOCTTY);
+ if (fd < 0) {
+ CTX_ERROR(jail->ctx, "Failed to open a new terminal: %s\n", strerror(errno));
+ return -errno;
+ }
+
+ CTX_DEBUG(jail->ctx, "Opened a new terminal %d\n", fd);
+
+ // Connect the new terminal to standard input
+ r = dup2(fd, STDIN_FILENO);
+ if (r < 0) {
+ CTX_ERROR(jail->ctx, "Failed to open standard input: %s\n", strerror(errno));
+ return -errno;
+ }
+
+ // Connect the new terminal to standard output
+ r = dup2(fd, STDOUT_FILENO);
+ if (r < 0) {
+ CTX_ERROR(jail->ctx, "Failed to open standard output: %s\n", strerror(errno));
+ return -errno;
+ }
+
+ // Connect the new terminal to standard error
+ r = dup2(fd, STDERR_FILENO);
+ if (r < 0) {
+ CTX_ERROR(jail->ctx, "Failed to open standard error: %s\n", strerror(errno));
+ return -errno;
+ }
+
return 0;
}
int r;
// Redirect any logging to our log pipe
- pakfire_set_log_callback(jail->pakfire, pakfire_jail_log, &ctx->pipes);
-
- // Die with parent
- r = prctl(PR_SET_PDEATHSIG, SIGKILL, 0, 0, 0);
- if (r) {
- ERROR(jail->pakfire, "Could not configure to die with parent: %m\n");
- return 126;
- }
+ pakfire_ctx_set_log_callback(jail->ctx, pakfire_jail_log_redirect, &ctx->pipes);
// Fetch my own PID
pid_t pid = getpid();
if (r)
return r;
- // Perform further initialization
+ // Die with parent
+ r = prctl(PR_SET_PDEATHSIG, SIGKILL, 0, 0, 0);
+ if (r) {
+ ERROR(jail->pakfire, "Could not configure to die with parent: %m\n");
+ return 126;
+ }
+
+ // Make this process dumpable
+ r = prctl (PR_SET_DUMPABLE, 1, 0, 0, 0);
+ if (r) {
+ ERROR(jail->pakfire, "Could not make the process dumpable: %m\n");
+ return 126;
+ }
+
+ // Don't drop any capabilities on setuid()
+ r = prctl(PR_SET_KEEPCAPS, 1);
+ if (r) {
+ ERROR(jail->pakfire, "Could not set PR_SET_KEEPCAPS: %m\n");
+ return 126;
+ }
// Fetch UID/GID
uid_t uid = getuid();
uid_t euid = geteuid();
gid_t egid = getegid();
- DEBUG(jail->pakfire, " UID: %d (effective %d)\n", uid, euid);
- DEBUG(jail->pakfire, " GID: %d (effective %d)\n", gid, egid);
+ DEBUG(jail->pakfire, " UID: %u (effective %u)\n", uid, euid);
+ DEBUG(jail->pakfire, " GID: %u (effective %u)\n", gid, egid);
+
+ // Log all mountpoints
+ pakfire_mount_list(jail->ctx);
+
+ // Fail if we are not PID 1
+ if (pid != 1) {
+ CTX_ERROR(jail->ctx, "Child process is not PID 1\n");
+ return 126;
+ }
- // Check if we are (effectively running as root)
+ // Fail if we are not running as root
if (uid || gid || euid || egid) {
ERROR(jail->pakfire, "Child process is not running as root\n");
return 126;
}
+ const int socket_send = pakfire_jail_get_pipe_to_write(jail, &ctx->socket);
+
+ // Mount all default stuff
+ r = pakfire_mount_all(jail->pakfire, PAKFIRE_MNTNS_INNER, 0);
+ if (r)
+ return 126;
+
const char* root = pakfire_get_path(jail->pakfire);
- const char* arch = pakfire_get_arch(jail->pakfire);
+ const char* arch = pakfire_get_effective_arch(jail->pakfire);
+
+ // Change mount propagation to slave to receive anything from the parent namespace
+ r = pakfire_mount_change_propagation(jail->ctx, "/", MS_SLAVE);
+ if (r)
+ return r;
+
+ // Make root a mountpoint in the new mount namespace
+ r = pakfire_mount_make_mounpoint(jail->pakfire, root);
+ if (r)
+ return r;
+
+ // Change mount propagation to private
+ r = pakfire_mount_change_propagation(jail->ctx, root, MS_PRIVATE);
+ if (r)
+ return r;
// Change root (unless root is /)
if (!pakfire_on_root(jail->pakfire)) {
if (r)
return r;
- // Call chroot()
- r = chroot(root);
- if (r) {
- ERROR(jail->pakfire, "chroot() to %s failed: %m\n", root);
- return 1;
- }
-
- // Change directory to /
- r = chdir("/");
- if (r) {
- ERROR(jail->pakfire, "chdir() after chroot() failed: %m\n");
- return 1;
- }
+ // chroot()
+ r = pakfire_jail_switch_root(jail, root);
+ if (r)
+ return r;
}
// Set personality
}
}
+ // Setup networking
+ if (!pakfire_jail_exec_has_flag(ctx, PAKFIRE_JAIL_HAS_NETWORKING)) {
+ r = pakfire_jail_setup_loopback(jail);
+ if (r)
+ return 1;
+ }
+
// Set nice level
if (jail->nice) {
DEBUG(jail->pakfire, "Setting nice level to %d\n", jail->nice);
}
}
- // Close other end of log pipes
- close(ctx->pipes.log_INFO[0]);
- close(ctx->pipes.log_ERROR[0]);
-#ifdef ENABLE_DEBUG
- close(ctx->pipes.log_DEBUG[0]);
-#endif /* ENABLE_DEBUG */
-
- // Connect standard input
- if (ctx->pipes.stdin[0]) {
- r = dup2(ctx->pipes.stdin[0], STDIN_FILENO);
- if (r < 0) {
- ERROR(jail->pakfire, "Could not connect fd %d to stdin: %m\n",
- ctx->pipes.stdin[0]);
+ // Create a new session
+ r = setsid();
+ if (r < 0) {
+ CTX_ERROR(jail->ctx, "Could not create a new session: %s\n", strerror(errno));
+ return r;
+ }
- return 1;
- }
+ // Allocate a new PTY
+ r = pakfire_jail_open_pty(jail, ctx);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Could not allocate a new PTY: %s\n", strerror(-r));
+ return r;
}
- // Connect standard output and error
- if (ctx->pipes.stdout[1] && ctx->pipes.stderr[1]) {
- r = dup2(ctx->pipes.stdout[1], STDOUT_FILENO);
- if (r < 0) {
- ERROR(jail->pakfire, "Could not connect fd %d to stdout: %m\n",
- ctx->pipes.stdout[1]);
+ // Send the PTY master to the parent process
+ r = pakfire_jail_send_fd(jail, socket_send, ctx->pty.master.fd);
+ if (r) {
+ CTX_ERROR(jail->ctx, "Failed sending the PTY master to the parent: %s\n", strerror(-r));
+ return r;
+ }
- return 1;
- }
+ // Setup the terminal
+ r = pakfire_jail_setup_terminal(jail, ctx);
+ if (r)
+ return r;
- r = dup2(ctx->pipes.stderr[1], STDERR_FILENO);
- if (r < 0) {
- ERROR(jail->pakfire, "Could not connect fd %d to stderr: %m\n",
- ctx->pipes.stderr[1]);
+ // Close the master of the PTY
+ close(ctx->pty.master.fd);
+ ctx->pty.master.fd = -1;
- return 1;
- }
+ // Close the socket
+ close(socket_send);
- // Close the pipe (as we have moved the original file descriptors)
- pakfire_jail_close_pipe(jail, ctx->pipes.stdin);
- pakfire_jail_close_pipe(jail, ctx->pipes.stdout);
- pakfire_jail_close_pipe(jail, ctx->pipes.stderr);
- }
+ // Close other end of log pipes
+ close(ctx->pipes.log_INFO[0]);
+ close(ctx->pipes.log_ERROR[0]);
+#ifdef ENABLE_DEBUG
+ close(ctx->pipes.log_DEBUG[0]);
+#endif /* ENABLE_DEBUG */
// Reset open file limit (http://0pointer.net/blog/file-descriptor-limits.html)
r = pakfire_rlimit_reset_nofile(jail->pakfire);
if (r)
return r;
- // Drop capabilities
- r = pakfire_jail_drop_capabilities(jail);
+ // Set capabilities
+ r = pakfire_jail_set_capabilities(jail);
+ if (r)
+ return r;
+
+ // Show capabilities
+ r = pakfire_jail_show_capabilities(jail);
if (r)
return r;
// Log argv
for (unsigned int i = 0; argv[i]; i++)
- DEBUG(jail->pakfire, " argv[%d] = %s\n", i, argv[i]);
+ DEBUG(jail->pakfire, " argv[%u] = %s\n", i, argv[i]);
// exec() command
r = execvpe(argv[0], (char**)argv, jail->env);
- if (r < 0)
- ERROR(jail->pakfire, "Could not execve(): %m\n");
+ if (r < 0) {
+ // Translate errno into regular exit code
+ switch (errno) {
+ case ENOENT:
+ // Ignore if the command doesn't exist
+ if (ctx->flags & PAKFIRE_JAIL_NOENT_OK)
+ r = 0;
+ else
+ r = 127;
- // Translate errno into regular exit code
- switch (errno) {
- case ENOENT:
- r = 127;
- break;
+ break;
- default:
- r = 1;
+ default:
+ r = 1;
+ }
+
+ ERROR(jail->pakfire, "Could not execve(%s): %m\n", argv[0]);
}
// We should not get here
// Run a command in the jail
static int __pakfire_jail_exec(struct pakfire_jail* jail, const char* argv[],
- const int interactive,
pakfire_jail_communicate_in communicate_in,
pakfire_jail_communicate_out communicate_out,
- void* data) {
+ void* data, int flags) {
int exit = -1;
int r;
return -1;
}
- // Send any output to the default logger if no callback is set
- if (!communicate_out)
- communicate_out = pakfire_jail_default_log_callback;
-
// Initialize context for this call
struct pakfire_jail_exec ctx = {
- .flags = 0,
+ .flags = flags,
+
+ .socket = { -1, -1 },
.pipes = {
- .stdin = { 0, 0 },
- .stdout = { 0, 0 },
- .stderr = { 0, 0 },
+ .log_INFO = { -1, -1 },
+ .log_ERROR = { -1, -1 },
+#ifdef ENABLE_DEBUG
+ .log_DEBUG = { -1, -1 },
+#endif /* ENABLE_DEBUG */
},
.communicate = {
.out = communicate_out,
.data = data,
},
+
+ .pidfd = -1,
+
+ // PTY
+ .pty = {
+ .master = {
+ .fd = -1,
+ },
+ .stdin = {
+ .fd = -1,
+ },
+ .stdout = {
+ .fd = -1,
+ },
+ },
};
DEBUG(jail->pakfire, "Executing jail...\n");
// Enable networking in interactive mode
- if (interactive)
+ if (ctx.flags & PAKFIRE_JAIL_PTY_FORWARDING)
ctx.flags |= PAKFIRE_JAIL_HAS_NETWORKING;
/*
return -1;
}
- // Create pipes to communicate with child process if we are not running interactively
- if (!interactive) {
- // stdin (only if callback is set)
- if (ctx.communicate.in) {
- r = pakfire_jail_setup_pipe(jail, &ctx.pipes.stdin, 0);
- if (r)
- goto ERROR;
- }
-
- // stdout
- r = pakfire_jail_setup_pipe(jail, &ctx.pipes.stdout, 0);
- if (r)
- goto ERROR;
-
- // stderr
- r = pakfire_jail_setup_pipe(jail, &ctx.pipes.stderr, 0);
- if (r)
- goto ERROR;
+ // Create a UNIX domain socket
+ r = socketpair(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0, ctx.socket);
+ if (r < 0) {
+ CTX_ERROR(jail->ctx, "Could not create UNIX socket: %s\n", strerror(errno));
+ r = -errno;
+ goto ERROR;
}
// Setup pipes for logging
CLONE_NEWIPC |
CLONE_NEWNS |
CLONE_NEWPID |
+ CLONE_NEWTIME |
CLONE_NEWUSER |
CLONE_NEWUTS |
CLONE_PIDFD,
break;
case CLD_KILLED:
- case CLD_DUMPED:
ERROR(jail->pakfire, "The child process was killed\n");
+ exit = 139;
+ break;
+
+ case CLD_DUMPED:
+ ERROR(jail->pakfire, "The child process terminated abnormally\n");
break;
// Log anything else
// Destroy the temporary cgroup (if any)
if (ctx.cgroup) {
// Read cgroup stats
- r = pakfire_cgroup_stat(ctx.cgroup, &ctx.cgroup_stats);
- if (r) {
- ERROR(jail->pakfire, "Could not read cgroup stats: %m\n");
- } else {
- pakfire_cgroup_stat_dump(ctx.cgroup, &ctx.cgroup_stats);
- }
-
+ pakfire_cgroup_stat(ctx.cgroup, &ctx.cgroup_stats);
+ pakfire_cgroup_stat_dump(ctx.cgroup, &ctx.cgroup_stats);
pakfire_cgroup_destroy(ctx.cgroup);
pakfire_cgroup_unref(ctx.cgroup);
}
// Close any file descriptors
- pakfire_jail_close_pipe(jail, ctx.pipes.stdin);
- pakfire_jail_close_pipe(jail, ctx.pipes.stdout);
- pakfire_jail_close_pipe(jail, ctx.pipes.stderr);
- if (ctx.pidfd)
+ if (ctx.pidfd >= 0)
close(ctx.pidfd);
+ if (ctx.pty.master.fd >= 0)
+ close(ctx.pty.master.fd);
pakfire_jail_close_pipe(jail, ctx.pipes.log_INFO);
pakfire_jail_close_pipe(jail, ctx.pipes.log_ERROR);
+#ifdef ENABLE_DEBUG
pakfire_jail_close_pipe(jail, ctx.pipes.log_DEBUG);
+#endif /* ENABLE_DEBUG */
+ pakfire_jail_close_pipe(jail, ctx.socket);
return exit;
}
const char* argv[],
pakfire_jail_communicate_in callback_in,
pakfire_jail_communicate_out callback_out,
- void* data) {
- return __pakfire_jail_exec(jail, argv, 0, callback_in, callback_out, data);
+ void* data, int flags) {
+ return __pakfire_jail_exec(jail, argv, callback_in, callback_out, data, flags);
}
static int pakfire_jail_exec_interactive(
- struct pakfire_jail* jail, const char* argv[]) {
+ struct pakfire_jail* jail, const char* argv[], int flags) {
int r;
+ flags |= PAKFIRE_JAIL_PTY_FORWARDING;
+
// Setup interactive stuff
r = pakfire_jail_setup_interactive_env(jail);
if (r)
return r;
- return __pakfire_jail_exec(jail, argv, 1, NULL, NULL, NULL);
+ return __pakfire_jail_exec(jail, argv, NULL, NULL, NULL, flags);
}
int pakfire_jail_exec_script(struct pakfire_jail* jail,
const char* root = pakfire_get_path(jail->pakfire);
// Write the scriptlet to disk
- r = pakfire_path_join(path, root, PAKFIRE_TMP_DIR "/pakfire-script.XXXXXX");
+ r = pakfire_path_append(path, root, PAKFIRE_TMP_DIR "/pakfire-script.XXXXXX");
if (r)
goto ERROR;
argv[i] = args[i-1];
// Run the script
- r = pakfire_jail_exec(jail, argv, callback_in, callback_out, data);
+ r = pakfire_jail_exec(jail, argv, callback_in, callback_out, data, 0);
ERROR:
if (argv)
int r;
// Create a new jail
- r = pakfire_jail_create(&jail, pakfire, flags);
+ r = pakfire_jail_create(&jail, pakfire);
if (r)
goto ERROR;
// Execute the command
- r = pakfire_jail_exec(jail, argv, NULL, pakfire_jail_capture_stdout, output);
+ r = pakfire_jail_exec(jail, argv, NULL, pakfire_jail_capture_stdout, output, 0);
ERROR:
if (jail)
int r;
// Create a new jail
- r = pakfire_jail_create(&jail, pakfire, flags);
+ r = pakfire_jail_create(&jail, pakfire);
if (r)
goto ERROR;
}
int pakfire_jail_shell(struct pakfire_jail* jail) {
+ int r;
+
const char* argv[] = {
"/bin/bash", "--login", NULL,
};
// Execute /bin/bash
- return pakfire_jail_exec_interactive(jail, argv);
+ r = pakfire_jail_exec_interactive(jail, argv, 0);
+
+ // Raise any errors
+ if (r < 0)
+ return r;
+
+ // Ignore any return codes from the shell
+ return 0;
}
-int pakfire_jail_ldconfig(struct pakfire* pakfire) {
+static int pakfire_jail_run_if_possible(struct pakfire* pakfire, const char** argv) {
char path[PATH_MAX];
+ int r;
- const char* ldconfig = "/sbin/ldconfig";
-
- // Check if ldconfig exists before calling it to avoid overhead
- int r = pakfire_path(pakfire, path, "%s", ldconfig);
+ r = pakfire_path(pakfire, path, "%s", *argv);
if (r)
return r;
- // Check if ldconfig is executable
+ // Check if the file is executable
r = access(path, X_OK);
if (r) {
- DEBUG(pakfire, "%s is not executable. Skipping...\n", ldconfig);
+ DEBUG(pakfire, "%s is not executable. Skipping...\n", *argv);
return 0;
}
+ return pakfire_jail_run(pakfire, argv, 0, NULL);
+}
+
+int pakfire_jail_ldconfig(struct pakfire* pakfire) {
+ const char* argv[] = {
+ "/sbin/ldconfig",
+ NULL,
+ };
+
+ return pakfire_jail_run_if_possible(pakfire, argv);
+}
+
+int pakfire_jail_run_systemd_tmpfiles(struct pakfire* pakfire) {
const char* argv[] = {
- ldconfig, NULL,
+ "/usr/bin/systemd-tmpfiles",
+ "--create",
+ NULL,
};
- // Run ldconfig
- return pakfire_jail_run(pakfire, argv, 0, NULL);
+ return pakfire_jail_run_if_possible(pakfire, argv);
}