chiark / gitweb /
util: replace close_pipe() with new safe_close_pair()
[elogind.git] / src / nspawn / nspawn.c
index b2c974d97016b8334b31a9799468fa273daf041d..9a9ed9dc6e0f7e3e6d2759b6ce50711697c194d9 100644 (file)
@@ -86,6 +86,7 @@
 #include "udev-util.h"
 #include "blkid-util.h"
 #include "gpt.h"
+#include "siphash24.h"
 
 #ifdef HAVE_SECCOMP
 #include "seccomp-util.h"
@@ -985,7 +986,7 @@ static int setup_kmsg(const char *dest, int kmsg_socket) {
         /* Store away the fd in the socket, so that it stays open as
          * long as we run the child */
         k = sendmsg(kmsg_socket, &mh, MSG_DONTWAIT|MSG_NOSIGNAL);
-        close_nointr_nofail(fd);
+        safe_close(fd);
 
         if (k < 0) {
                 log_error("Failed to send FIFO fd: %m");
@@ -1399,9 +1400,46 @@ static int reset_audit_loginuid(void) {
         return 0;
 }
 
+#define HASH_KEY SD_ID128_MAKE(c3,c4,f9,19,b5,57,b2,1c,e6,cf,14,27,03,9c,ee,a2)
+
+static int get_mac(struct ether_addr *mac) {
+        int r;
+
+        uint8_t result[8];
+        size_t l, sz;
+        uint8_t *v;
+
+        l = strlen(arg_machine);
+        sz = sizeof(sd_id128_t) + l;
+        v = alloca(sz);
+
+        /* fetch some persistent data unique to the host */
+        r = sd_id128_get_machine((sd_id128_t*) v);
+        if (r < 0)
+                return r;
+
+        /* combine with some data unique (on this host) to this
+         * container instance */
+        memcpy(v + sizeof(sd_id128_t), arg_machine, l);
+
+        /* Let's hash the host machine ID plus the container name. We
+         * use a fixed, but originally randomly created hash key here. */
+        siphash24(result, v, sz, HASH_KEY.bytes);
+
+        assert_cc(ETH_ALEN <= sizeof(result));
+        memcpy(mac->ether_addr_octet, result, ETH_ALEN);
+
+        /* see eth_random_addr in the kernel */
+        mac->ether_addr_octet[0] &= 0xfe;        /* clear multicast bit */
+        mac->ether_addr_octet[0] |= 0x02;        /* set local assignment bit (IEEE802) */
+
+        return 0;
+}
+
 static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) {
         _cleanup_rtnl_message_unref_ sd_rtnl_message *m = NULL;
         _cleanup_rtnl_unref_ sd_rtnl *rtnl = NULL;
+        struct ether_addr mac;
         int r;
 
         if (!arg_private_network)
@@ -1416,9 +1454,14 @@ static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) {
                 memcpy(iface_name, "vb-", 3);
         else
                 memcpy(iface_name, "ve-", 3);
-
         strncpy(iface_name+3, arg_machine, IFNAMSIZ - 3);
 
+        r = get_mac(&mac);
+        if (r < 0) {
+                log_error("Failed to generate predictable MAC address for host0");
+                return r;
+        }
+
         r = sd_rtnl_open(&rtnl, 0);
         if (r < 0) {
                 log_error("Failed to connect to netlink: %s", strerror(-r));
@@ -1467,6 +1510,12 @@ static int setup_veth(pid_t pid, char iface_name[IFNAMSIZ]) {
                 return r;
         }
 
+        r = sd_rtnl_message_append_ether_addr(m, IFLA_ADDRESS, &mac);
+        if (r < 0) {
+                log_error("Failed to add netlink MAC address: %s", strerror(-r));
+                return r;
+        }
+
         r = sd_rtnl_message_append_u32(m, IFLA_NET_NS_PID, pid);
         if (r < 0) {
                 log_error("Failed to add netlink namespace field: %s", strerror(-r));
@@ -1532,6 +1581,12 @@ static int setup_bridge(const char veth_name[]) {
                 return r;
         }
 
+        r = sd_rtnl_message_link_set_flags(m, IFF_UP, IFF_UP);
+        if (r < 0) {
+                log_error("Failed to set IFF_UP flag: %s", strerror(-r));
+                return r;
+        }
+
         r = sd_rtnl_message_append_string(m, IFLA_IFNAME, veth_name);
         if (r < 0) {
                 log_error("Failed to add netlink interface name field: %s", strerror(-r));
@@ -1891,9 +1946,9 @@ static int setup_image(char **device_path, int *loop_nr) {
 
 static int dissect_image(
                 int fd,
-                char **root_device,
-                char **home_device,
-                char **srv_device,
+                char **root_device, bool *root_device_rw,
+                char **home_device, bool *home_device_rw,
+                char **srv_device, bool *srv_device_rw,
                 bool *secondary) {
 
 #ifdef HAVE_BLKID
@@ -1904,6 +1959,7 @@ static int dissect_image(
         _cleanup_blkid_free_probe_ blkid_probe b = NULL;
         _cleanup_udev_unref_ struct udev *udev = NULL;
         struct udev_list_entry *first, *item;
+        bool home_rw = true, root_rw = true, secondary_root_rw = true, srv_rw = true;
         const char *pttype = NULL;
         blkid_partlist pl;
         struct stat st;
@@ -1993,6 +2049,7 @@ static int dissect_image(
         udev_list_entry_foreach(item, first) {
                 _cleanup_udev_device_unref_ struct udev_device *q;
                 const char *stype, *node;
+                unsigned long long flags;
                 sd_id128_t type_id;
                 blkid_partition pp;
                 dev_t qn;
@@ -2023,6 +2080,10 @@ static int dissect_image(
                 if (!pp)
                         continue;
 
+                flags = blkid_partition_get_flags(pp);
+                if (flags & GPT_FLAG_NO_AUTO)
+                        continue;
+
                 nr = blkid_partition_get_partno(pp);
                 if (nr < 0)
                         continue;
@@ -2040,6 +2101,8 @@ static int dissect_image(
                                 continue;
 
                         home_nr = nr;
+                        home_rw = !(flags & GPT_FLAG_READ_ONLY);
+
                         free(home);
                         home = strdup(node);
                         if (!home)
@@ -2050,6 +2113,8 @@ static int dissect_image(
                                 continue;
 
                         srv_nr = nr;
+                        srv_rw = !(flags & GPT_FLAG_READ_ONLY);
+
                         free(srv);
                         srv = strdup(node);
                         if (!srv)
@@ -2062,6 +2127,8 @@ static int dissect_image(
                                 continue;
 
                         root_nr = nr;
+                        root_rw = !(flags & GPT_FLAG_READ_ONLY);
+
                         free(root);
                         root = strdup(node);
                         if (!root)
@@ -2075,6 +2142,9 @@ static int dissect_image(
                                 continue;
 
                         secondary_root_nr = nr;
+                        secondary_root_rw = !(flags & GPT_FLAG_READ_ONLY);
+
+
                         free(secondary_root);
                         secondary_root = strdup(node);
                         if (!secondary_root)
@@ -2092,21 +2162,29 @@ static int dissect_image(
         if (root) {
                 *root_device = root;
                 root = NULL;
+
+                *root_device_rw = root_rw;
                 *secondary = false;
         } else if (secondary_root) {
                 *root_device = secondary_root;
                 secondary_root = NULL;
+
+                *root_device_rw = secondary_root_rw;
                 *secondary = true;
         }
 
         if (home) {
                 *home_device = home;
                 home = NULL;
+
+                *home_device_rw = home_rw;
         }
 
         if (srv) {
                 *srv_device = srv;
                 srv = NULL;
+
+                *srv_device_rw = srv_rw;
         }
 
         return 0;
@@ -2116,7 +2194,7 @@ static int dissect_image(
 #endif
 }
 
-static int mount_device(const char *what, const char *where, const char *directory) {
+static int mount_device(const char *what, const char *where, const char *directory, bool rw) {
 #ifdef HAVE_BLKID
         _cleanup_blkid_free_probe_ blkid_probe b = NULL;
         const char *fstype, *p;
@@ -2125,6 +2203,9 @@ static int mount_device(const char *what, const char *where, const char *directo
         assert(what);
         assert(where);
 
+        if (arg_read_only)
+                rw = false;
+
         if (directory)
                 p = strappenda(where, directory);
         else
@@ -2167,7 +2248,7 @@ static int mount_device(const char *what, const char *where, const char *directo
                 return -ENOTSUP;
         }
 
-        if (mount(what, p, fstype, arg_read_only ? MS_NODEV|MS_RDONLY : 0, NULL) < 0) {
+        if (mount(what, p, fstype, MS_NODEV|(rw ? 0 : MS_RDONLY), NULL) < 0) {
                 log_error("Failed to mount %s: %m", what);
                 return -errno;
         }
@@ -2179,13 +2260,17 @@ static int mount_device(const char *what, const char *where, const char *directo
 #endif
 }
 
-static int mount_devices(const char *where, const char *root_device, const char *home_device, const char *srv_device) {
+static int mount_devices(
+                const char *where,
+                const char *root_device, bool root_device_rw,
+                const char *home_device, bool home_device_rw,
+                const char *srv_device, bool srv_device_rw) {
         int r;
 
         assert(where);
 
         if (root_device) {
-                r = mount_device(root_device, arg_directory, NULL);
+                r = mount_device(root_device, arg_directory, NULL, root_device_rw);
                 if (r < 0) {
                         log_error("Failed to mount root directory: %s", strerror(-r));
                         return r;
@@ -2193,7 +2278,7 @@ static int mount_devices(const char *where, const char *root_device, const char
         }
 
         if (home_device) {
-                r = mount_device(home_device, arg_directory, "/home");
+                r = mount_device(home_device, arg_directory, "/home", home_device_rw);
                 if (r < 0) {
                         log_error("Failed to mount home directory: %s", strerror(-r));
                         return r;
@@ -2201,7 +2286,7 @@ static int mount_devices(const char *where, const char *root_device, const char
         }
 
         if (srv_device) {
-                r = mount_device(srv_device, arg_directory, "/srv");
+                r = mount_device(srv_device, arg_directory, "/srv", srv_device_rw);
                 if (r < 0) {
                         log_error("Failed to mount server data directory: %s", strerror(-r));
                         return r;
@@ -2219,8 +2304,7 @@ static void loop_remove(int nr, int *image_fd) {
 
         if (image_fd && *image_fd >= 0) {
                 ioctl(*image_fd, LOOP_CLR_FD);
-                close_nointr_nofail(*image_fd);
-                *image_fd = -1;
+                *image_fd = safe_close(*image_fd);
         }
 
         control = open("/dev/loop-control", O_RDWR|O_CLOEXEC|O_NOCTTY|O_NONBLOCK);
@@ -2255,9 +2339,9 @@ static int spawn_getent(const char *database, const char *key, pid_t *rpid) {
                         _exit(EXIT_FAILURE);
 
                 if (pipe_fds[0] > 2)
-                        close_nointr_nofail(pipe_fds[0]);
+                        safe_close(pipe_fds[0]);
                 if (pipe_fds[1] > 2)
-                        close_nointr_nofail(pipe_fds[1]);
+                        safe_close(pipe_fds[1]);
 
                 nullfd = open("/dev/null", O_RDWR);
                 if (nullfd < 0)
@@ -2270,7 +2354,7 @@ static int spawn_getent(const char *database, const char *key, pid_t *rpid) {
                         _exit(EXIT_FAILURE);
 
                 if (nullfd > 2)
-                        close_nointr_nofail(nullfd);
+                        safe_close(nullfd);
 
                 reset_all_signal_handlers();
                 close_all_fds(NULL, 0);
@@ -2280,8 +2364,7 @@ static int spawn_getent(const char *database, const char *key, pid_t *rpid) {
                 _exit(EXIT_FAILURE);
         }
 
-        close_nointr_nofail(pipe_fds[1]);
-        pipe_fds[1] = -1;
+        pipe_fds[1] = safe_close(pipe_fds[1]);
 
         *rpid = pid;
 
@@ -2289,8 +2372,6 @@ static int spawn_getent(const char *database, const char *key, pid_t *rpid) {
 }
 
 static int change_uid_gid(char **_home) {
-
-        _cleanup_strv_free_ char **passwd = NULL;
         char line[LINE_MAX], *w, *x, *state, *u, *g, *h;
         _cleanup_free_ uid_t *uids = NULL;
         _cleanup_free_ char *home = NULL;
@@ -2464,7 +2545,7 @@ static int change_uid_gid(char **_home) {
         }
 
         r = mkdir_safe(home, 0755, uid, gid);
-        if (r < 0) {
+        if (r < 0 && r != -EEXIST) {
                 log_error("Failed to make home directory: %s", strerror(-r));
                 return r;
         }
@@ -2499,8 +2580,9 @@ static int change_uid_gid(char **_home) {
 int main(int argc, char *argv[]) {
 
         _cleanup_free_ char *kdbus_domain = NULL, *device_path = NULL, *root_device = NULL, *home_device = NULL, *srv_device = NULL;
+        bool root_device_rw = true, home_device_rw = true, srv_device_rw = true;
         _cleanup_close_ int master = -1, kdbus_fd = -1, image_fd = -1;
-        _cleanup_close_pipe_ int kmsg_socket_pair[2] = { -1, -1 };
+        _cleanup_close_pair_ int kmsg_socket_pair[2] = { -1, -1 };
         _cleanup_fdset_free_ FDSet *fds = NULL;
         int r = EXIT_FAILURE, k, n_fd_passed, loop_nr = -1;
         const char *console = NULL;
@@ -2616,7 +2698,7 @@ int main(int argc, char *argv[]) {
                         goto finish;
                 }
 
-                r = dissect_image(image_fd, &root_device, &home_device, &srv_device, &secondary);
+                r = dissect_image(image_fd, &root_device, &root_device_rw, &home_device, &home_device_rw, &srv_device, &srv_device_rw, &secondary);
                 if (r < 0)
                         goto finish;
         }
@@ -2724,15 +2806,13 @@ int main(int argc, char *argv[]) {
                         if (envp[n_env])
                                 n_env ++;
 
-                        close_nointr_nofail(master);
-                        master = -1;
+                        master = safe_close(master);
 
                         close_nointr(STDIN_FILENO);
                         close_nointr(STDOUT_FILENO);
                         close_nointr(STDERR_FILENO);
 
-                        close_nointr_nofail(kmsg_socket_pair[0]);
-                        kmsg_socket_pair[0] = -1;
+                        kmsg_socket_pair[0] = safe_close(kmsg_socket_pair[0]);
 
                         reset_all_signal_handlers();
 
@@ -2742,7 +2822,7 @@ int main(int argc, char *argv[]) {
                         k = open_terminal(console, O_RDWR);
                         if (k != STDIN_FILENO) {
                                 if (k >= 0) {
-                                        close_nointr_nofail(k);
+                                        safe_close(k);
                                         k = -EINVAL;
                                 }
 
@@ -2777,7 +2857,10 @@ int main(int argc, char *argv[]) {
                                 goto child_fail;
                         }
 
-                        if (mount_devices(arg_directory, root_device, home_device, srv_device) < 0)
+                        if (mount_devices(arg_directory,
+                                          root_device, root_device_rw,
+                                          home_device, home_device_rw,
+                                          srv_device, srv_device_rw) < 0)
                                 goto child_fail;
 
                         /* Turn directory into bind mount */
@@ -2812,8 +2895,7 @@ int main(int argc, char *argv[]) {
                         if (setup_kmsg(arg_directory, kmsg_socket_pair[1]) < 0)
                                 goto child_fail;
 
-                        close_nointr_nofail(kmsg_socket_pair[1]);
-                        kmsg_socket_pair[1] = -1;
+                        kmsg_socket_pair[1] = safe_close(kmsg_socket_pair[1]);
 
                         if (setup_boot_id(arg_directory) < 0)
                                 goto child_fail;
@@ -2840,8 +2922,7 @@ int main(int argc, char *argv[]) {
                          * it can cgroupify us to that we lack access
                          * to certain devices and resources. */
                         eventfd_write(child_ready_fd, 1);
-                        close_nointr_nofail(child_ready_fd);
-                        child_ready_fd = -1;
+                        child_ready_fd = safe_close(child_ready_fd);
 
                         if (chdir(arg_directory) < 0) {
                                 log_error("chdir(%s) failed: %m", arg_directory);
@@ -2942,8 +3023,7 @@ int main(int argc, char *argv[]) {
 
                         /* Wait until the parent is ready with the setup, too... */
                         eventfd_read(parent_ready_fd, &x);
-                        close_nointr_nofail(parent_ready_fd);
-                        parent_ready_fd = -1;
+                        parent_ready_fd = safe_close(parent_ready_fd);
 
                         if (arg_boot) {
                                 char **a;