diff --git a/man/systemd-vmspawn.xml b/man/systemd-vmspawn.xml
index 0446d27630..cfc543a88a 100644
--- a/man/systemd-vmspawn.xml
+++ b/man/systemd-vmspawn.xml
@@ -183,6 +183,7 @@
Set the initrd to use for direct kernel boot.
If the linux kernel supplied is a UKI then this argument is not required.
+ If the option is specified multiple times vmspawn will merge the initrds together.
If no initrd was installed into the image then the image will fail to boot.
diff --git a/src/vmspawn/vmspawn.c b/src/vmspawn/vmspawn.c
index d5de11bfe0..116f0143cd 100644
--- a/src/vmspawn/vmspawn.c
+++ b/src/vmspawn/vmspawn.c
@@ -2,10 +2,17 @@
#include
#include
+#include
#include
#include
+#include
#include
+#include "bootspec.h"
+#include "chase.h"
+#include "dirent-util.h"
+#include "fd-util.h"
+#include "discover-image.h"
#include "sd-daemon.h"
#include "sd-event.h"
#include "sd-id128.h"
@@ -43,6 +50,7 @@
#include "rm-rf.h"
#include "signal-util.h"
#include "socket-util.h"
+#include "stat-util.h"
#include "string-util.h"
#include "strv.h"
#include "tmpfile-util.h"
@@ -64,7 +72,7 @@ static int arg_qemu_vsock = -1;
static unsigned arg_vsock_cid = VMADDR_CID_ANY;
static int arg_tpm = -1;
static char *arg_linux = NULL;
-static char *arg_initrd = NULL;
+static char **arg_initrds = NULL;
static bool arg_qemu_gui = false;
static QemuNetworkStack arg_network_stack = QEMU_NET_NONE;
static int arg_secure_boot = -1;
@@ -86,7 +94,7 @@ STATIC_DESTRUCTOR_REGISTER(arg_runtime_directory, freep);
STATIC_DESTRUCTOR_REGISTER(arg_credentials, machine_credential_context_done);
STATIC_DESTRUCTOR_REGISTER(arg_firmware, freep);
STATIC_DESTRUCTOR_REGISTER(arg_linux, freep);
-STATIC_DESTRUCTOR_REGISTER(arg_initrd, freep);
+STATIC_DESTRUCTOR_REGISTER(arg_initrds, strv_freep);
STATIC_DESTRUCTOR_REGISTER(arg_runtime_mounts, runtime_mount_context_done);
STATIC_DESTRUCTOR_REGISTER(arg_kernel_cmdline_extra, strv_freep);
@@ -312,9 +320,15 @@ static int parse_argv(int argc, char *argv[]) {
break;
case ARG_INITRD: {
- r = parse_path_argument(optarg, /* suppress_root= */ false, &arg_initrd);
+ _cleanup_free_ char *initrd_path = NULL;
+ r = parse_path_argument(optarg, /* suppress_root= */ false, &initrd_path);
if (r < 0)
return r;
+
+ r = strv_consume(&arg_initrds, TAKE_PTR(initrd_path));
+ if (r < 0)
+ return log_oom();
+
break;
}
@@ -810,6 +824,124 @@ static int kernel_cmdline_maybe_append_root(void) {
return 0;
}
+static int discover_boot_entry(const char *root, char **ret_linux, char ***ret_initrds) {
+ _cleanup_(boot_config_free) BootConfig config = BOOT_CONFIG_NULL;
+ _cleanup_free_ char *esp_path = NULL, *xbootldr_path = NULL;
+ int r;
+
+ assert(root);
+ assert(ret_linux);
+ assert(ret_initrds);
+
+ esp_path = path_join(root, "efi");
+ if (!esp_path)
+ return log_oom();
+
+ xbootldr_path = path_join(root, "boot");
+ if (!xbootldr_path)
+ return log_oom();
+
+ r = boot_config_load(&config, esp_path, xbootldr_path);
+ if (r < 0)
+ return r;
+
+ r = boot_config_select_special_entries(&config, /* skip_efivars= */ true);
+ if (r < 0)
+ return log_error_errno(r, "Failed to find special boot config entries: %m");
+
+ const BootEntry *boot_entry = boot_config_default_entry(&config);
+
+ if (!IN_SET(boot_entry->type, BOOT_ENTRY_UNIFIED, BOOT_ENTRY_CONF))
+ boot_entry = NULL;
+
+ /* If we cannot determine a default entry search for UKIs (Type #2 EFI Unified Kernel Images)
+ * then .conf files (Type #1 Boot Loader Specification Entries).
+ * https://uapi-group.org/specifications/specs/boot_loader_specification */
+ if (!boot_entry)
+ FOREACH_ARRAY(entry, config.entries, config.n_entries)
+ if (entry->type == BOOT_ENTRY_UNIFIED) {
+ boot_entry = entry;
+ break;
+ }
+
+ if (!boot_entry)
+ FOREACH_ARRAY(entry, config.entries, config.n_entries)
+ if (entry->type == BOOT_ENTRY_CONF) {
+ boot_entry = entry;
+ break;
+ }
+
+ if (!boot_entry)
+ return log_error_errno(SYNTHETIC_ERRNO(ENOENT), "Failed to discover any boot entries.");
+
+ log_debug("Discovered boot entry %s (%s)", boot_entry->id, boot_entry_type_to_string(boot_entry->type));
+
+ _cleanup_free_ char *linux_kernel = NULL;
+ _cleanup_strv_free_ char **initrds = NULL;
+ if (boot_entry->type == BOOT_ENTRY_UNIFIED) {
+ linux_kernel = path_join(boot_entry->root, boot_entry->kernel);
+ if (!linux_kernel)
+ return log_oom();
+ } else if (boot_entry->type == BOOT_ENTRY_CONF) {
+ linux_kernel = path_join(boot_entry->root, boot_entry->kernel);
+ if (!linux_kernel)
+ return log_oom();
+
+ STRV_FOREACH(initrd, boot_entry->initrd) {
+ _cleanup_free_ char *initrd_path = path_join(boot_entry->root, *initrd);
+ if (!initrd_path)
+ return log_oom();
+
+ r = strv_consume(&initrds, TAKE_PTR(initrd_path));
+ if (r < 0)
+ return log_oom();
+ }
+ } else
+ assert_not_reached();
+
+ *ret_linux = TAKE_PTR(linux_kernel);
+ *ret_initrds = TAKE_PTR(initrds);
+
+ return 0;
+}
+
+static int merge_initrds(char **ret) {
+ _cleanup_(rm_rf_physical_and_freep) char *merged_initrd = NULL;
+ _cleanup_close_ int ofd = -EBADF;
+ int r;
+
+ assert(ret);
+
+ r = tempfn_random_child(NULL, "vmspawn-initrd-", &merged_initrd);
+ if (r < 0)
+ return log_error_errno(r, "Failed to create temporary file: %m");
+
+ ofd = open(merged_initrd, O_WRONLY|O_CREAT|O_EXCL|O_CLOEXEC, 0600);
+ if (ofd < 0)
+ return log_error_errno(errno, "Failed to create regular file %s: %m", merged_initrd);
+
+ size_t total_size = 0;
+ STRV_FOREACH(i, arg_initrds) {
+ _cleanup_close_ int ifd = -EBADF;
+ size_t to_seek = (4 - (total_size % 4)) % 4;
+
+ /* seek to assure 4 byte alignment for each initrd */
+ if (to_seek != 0 && lseek(ofd, to_seek, SEEK_CUR) < 0)
+ return log_error_errno(errno, "Failed to seek %s: %m", merged_initrd);
+
+ ifd = open(*i, O_RDONLY|O_CLOEXEC);
+ if (ifd < 0)
+ return log_error_errno(errno, "Failed to open %s: %m", *i);
+
+ r = copy_bytes(ifd, ofd, UINT64_MAX, COPY_REFLINK);
+ if (r < 0)
+ return log_error_errno(r, "Failed to copy bytes from %s to %s: %m", *i, merged_initrd);
+ }
+
+ *ret = TAKE_PTR(merged_initrd);
+ return 0;
+}
+
static int run_virtual_machine(int kvm_device_fd, int vhost_device_fd) {
_cleanup_(ovmf_config_freep) OvmfConfig *ovmf_config = NULL;
_cleanup_(sd_bus_flush_close_unrefp) sd_bus *bus = NULL;
@@ -864,8 +996,14 @@ static int run_virtual_machine(int kvm_device_fd, int vhost_device_fd) {
kernel = strdup(arg_linux);
if (!kernel)
return log_oom();
- } else if (arg_directory)
- return log_error_errno(SYNTHETIC_ERRNO(ENOENT), "Please specify a kernel (--linux=) when using -D/--directory=, refusing.");
+ } else if (arg_directory) {
+ /* a kernel is required for directory type images so attempt to locate a UKI under /boot and /efi */
+ r = discover_boot_entry(arg_directory, &kernel, &arg_initrds);
+ if (r < 0)
+ return log_error_errno(r, "Failed to locate UKI in directory type image, please specify one with --linux=.");
+
+ log_debug("Discovered UKI image at %s", kernel);
+ }
r = find_qemu_binary(&qemu_binary);
if (r == -EOPNOTSUPP)
@@ -1248,8 +1386,22 @@ static int run_virtual_machine(int kvm_device_fd, int vhost_device_fd) {
return log_oom();
}
- if (arg_initrd) {
- r = strv_extend_many(&cmdline, "-initrd", arg_initrd);
+ char *initrd = NULL;
+ _cleanup_(rm_rf_physical_and_freep) char *merged_initrd = NULL;
+ size_t n_initrds = strv_length(arg_initrds);
+
+ if (n_initrds == 1)
+ initrd = arg_initrds[0];
+ else if (n_initrds > 1) {
+ r = merge_initrds(&merged_initrd);
+ if (r < 0)
+ return r;
+
+ initrd = merged_initrd;
+ }
+
+ if (initrd) {
+ r = strv_extend_many(&cmdline, "-initrd", initrd);
if (r < 0)
return log_oom();
}
@@ -1346,8 +1498,30 @@ static int run_virtual_machine(int kvm_device_fd, int vhost_device_fd) {
static int determine_names(void) {
int r;
- if (!arg_directory && !arg_image)
- return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Failed to determine path, please use -D or -i.");
+ if (!arg_directory && !arg_image) {
+ if (arg_machine) {
+ _cleanup_(image_unrefp) Image *i = NULL;
+
+ r = image_find(IMAGE_MACHINE, arg_machine, NULL, &i);
+ if (r == -ENOENT)
+ return log_error_errno(r, "No image for machine '%s'.", arg_machine);
+ if (r < 0)
+ return log_error_errno(r, "Failed to find image for machine '%s': %m", arg_machine);
+
+ if (IN_SET(i->type, IMAGE_RAW, IMAGE_BLOCK))
+ r = free_and_strdup(&arg_image, i->path);
+ else if (IN_SET(i->type, IMAGE_DIRECTORY, IMAGE_SUBVOLUME))
+ r = free_and_strdup(&arg_directory, i->path);
+ else
+ assert_not_reached();
+ if (r < 0)
+ return log_oom();
+ } else {
+ r = safe_getcwd(&arg_directory);
+ if (r < 0)
+ return log_error_errno(r, "Failed to determine current directory: %m");
+ }
+ }
if (!arg_machine) {
if (arg_directory && path_equal(arg_directory, "/")) {
@@ -1383,6 +1557,9 @@ static int verify_arguments(void) {
if (arg_network_stack == QEMU_NET_TAP && !arg_privileged)
return log_error_errno(SYNTHETIC_ERRNO(EPERM), "--network-tap requires root privileges, refusing.");
+ if (!strv_isempty(arg_initrds) && !arg_linux)
+ return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Option --initrd= cannot be used without --linux=.");
+
return 0;
}
diff --git a/units/meson.build b/units/meson.build
index acfd8d1dcb..0c971ef0bc 100644
--- a/units/meson.build
+++ b/units/meson.build
@@ -436,6 +436,10 @@ units = [
'conditions' : ['ENABLE_NETWORKD'],
},
{ 'file' : 'systemd-nspawn@.service.in' },
+ {
+ 'file' : 'systemd-vmspawn@.service.in',
+ 'conditions' : ['ENABLE_VMSPAWN'],
+ },
{
'file' : 'systemd-oomd.service.in',
'conditions' : ['ENABLE_OOMD'],
diff --git a/units/systemd-vmspawn@.service.in b/units/systemd-vmspawn@.service.in
new file mode 100644
index 0000000000..608002040c
--- /dev/null
+++ b/units/systemd-vmspawn@.service.in
@@ -0,0 +1,34 @@
+# SPDX-License-Identifier: LGPL-2.1-or-later
+#
+# This file is part of systemd.
+#
+# systemd is free software; you can redistribute it and/or modify it
+# under the terms of the GNU Lesser General Public License as published by
+# the Free Software Foundation; either version 2.1 of the License, or
+# (at your option) any later version.
+
+[Unit]
+Description=Virtual Machine %i
+Documentation=man:systemd-vmspawn(1)
+PartOf=machines.target
+Before=machines.target
+After=network.target modprobe@tun.service
+RequiresMountsFor=/var/lib/machines/%i
+
+[Service]
+ExecStart=systemd-vmspawn --quiet --network-tap --machine=%i
+KillMode=mixed
+Type=notify
+Slice=machine.slice
+
+{# Enforce a strict device policy. Make sure to keep these policies in sync if you change them! #}
+DevicePolicy=closed
+DeviceAllow=/dev/net/tun rwm
+DeviceAllow=char-pts rw
+
+# vmspawn itself needs access to /dev/kvm and /dev/vhost-vsock
+DeviceAllow=/dev/kvm rw
+DeviceAllow=/dev/vhost-vsock rw
+
+[Install]
+WantedBy=machines.target