tpm2: re-manufacture software TPM when state dir is incomplete

setup_swtpm() decided whether a software TPM had already been
manufactured by checking whether the state directory was empty. But
manufacture_swtpm() writes swtpm's config files before forking
swtpm_setup, so an interrupted manufacture leaves the directory
non-empty yet without a usable TPM. The next boot then mistook it for a
complete TPM and started swtpm against a broken state directory.

Keying off a swtpm state file like tpm2-00.permall is no better, as
swtpm_setup gives no guarantee any single one is written atomically or
last. Instead, have manufacture_swtpm() write a marker (.manufactured)
as its very last step, once swtpm_setup has exited successfully, and
gate on it: re-manufacture when it is missing in the initrd, and refuse
rather than start a broken TPM outside it.

Signed-off-by: Paul Meyer <katexochen0@gmail.com>

src/shared/swtpm-util.c

@@ -1,5 +1,6 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 
+#include <fcntl.h>
 #include <unistd.h>
 
 #include "sd-json.h"
@@ -8,6 +9,7 @@
 #include "escape.h"
 #include "fd-util.h"
 #include "fileio.h"
+#include "fs-util.h"
 #include "json-util.h"
 #include "log.h"
 #include "memfd-util.h"
@@ -17,6 +19,7 @@
 #include "string-util.h"
 #include "strv.h"
 #include "swtpm-util.h"
+#include "sync-util.h"
 
 static int swtpm_find_best_profile(const char *swtpm_setup, char **ret) {
         int r;
@@ -135,9 +138,9 @@ int manufacture_swtpm(const char *state_dir, const char *secret) {
 
         assert(state_dir);
 
-        _cleanup_close_ int state_dir_fd = xopenat(AT_FDCWD, state_dir, O_RDONLY|O_DIRECTORY|O_CLOEXEC);
+        _cleanup_close_ int state_dir_fd = open(state_dir, O_RDONLY|O_DIRECTORY|O_CLOEXEC);
         if (state_dir_fd < 0)
-                return log_error_errno(state_dir_fd, "Failed to open TPM state directory '%s': %m", state_dir);
+                return log_error_errno(errno, "Failed to open TPM state directory '%s': %m", state_dir);
 
         _cleanup_free_ char *swtpm_setup = NULL;
         r = find_executable("swtpm_setup", &swtpm_setup);
@@ -238,5 +241,15 @@ int manufacture_swtpm(const char *state_dir, const char *secret) {
                 _exit(EXIT_FAILURE);
         }
 
+        /* Persist swtpm_setup's freshly created TPM state before writing the completion marker. */
+        r = syncfs_path(state_dir_fd, NULL);
+        if (r < 0)
+                return log_error_errno(r, "Failed to sync TPM state directory: %m");
+
+        /* Marker, written last, signals that manufacturing completed successfully. */
+        _cleanup_close_ int marker_fd = xopenat(state_dir_fd, SWTPM_MANUFACTURED_MARKER, O_WRONLY|O_CREAT|O_CLOEXEC|O_NOFOLLOW);
+        if (marker_fd < 0)
+                return log_error_errno(marker_fd, "Failed to write '%s' marker: %m", SWTPM_MANUFACTURED_MARKER);
+
         return 0;
 }

src/shared/swtpm-util.h

@@ -1,4 +1,9 @@
 /* SPDX-License-Identifier: LGPL-2.1-or-later */
 #pragma once
 
+/* Marker file written into the TPM state directory once swtpm_setup has successfully created the TPM state.
+ * It is written last, so its presence reliably means a complete TPM was manufactured, rather than a manufacture
+ * that was interrupted halfway through. */
+#define SWTPM_MANUFACTURED_MARKER ".manufactured"
+
 int manufacture_swtpm(const char *state_dir, const char *secret);

src/tpm2-setup/tpm2-swtpm.c

@@ -20,8 +20,8 @@
 #include "main-func.h"
 #include "path-lookup.h"
 #include "path-util.h"
+#include "rm-rf.h"
 #include "sha256.h"
-#include "stat-util.h"
 #include "string-util.h"
 #include "strv.h"
 #include "swtpm-util.h"
@@ -116,18 +116,28 @@ static int setup_swtpm(const char *state_dir, int state_fd, const char *secret)
                         return log_error_errno(r, "Failed to remove 'tpm2-00.volatilestate': %m");
         }
 
-        r = dir_is_empty_at(state_fd, /* path= */ NULL, /* ignore_hidden_or_backup= */ false);
-        if (r < 0)
-                return log_error_errno(r, "Failed to check if TPM state directory is empty: %m");
-        if (r == 0) {
-                log_debug("TPM state directory is already populated, not manufacturing a TPM.");
+        /* manufacture_swtpm() writes its marker only after swtpm_setup has fully created the TPM state.
+         * If the marker file is missing, the existing state is incomplete and recreation is needed. */
+        r = RET_NERRNO(faccessat(state_fd, SWTPM_MANUFACTURED_MARKER, F_OK, AT_SYMLINK_NOFOLLOW));
+        if (r >= 0) {
+                log_debug("TPM state directory holds a fully manufactured TPM, not manufacturing a TPM.");
                 return 0;
         }
+        if (r != -ENOENT)
+                return log_error_errno(r, "Failed to check for TPM manufacture marker: %m");
 
         if (!in_initrd())
                 return log_error_errno(SYNTHETIC_ERRNO(ESTALE), "swtpm TPM state directory has not been initialized in the initrd, refusing.");
 
-        log_debug("TPM state directory is unpopulated, manufacturing a TPM.");
+        /* Cleanup incomplete state before recreating. */
+        _cleanup_close_ int wipe_fd = fd_reopen(state_fd, O_RDONLY|O_DIRECTORY|O_CLOEXEC);
+        if (wipe_fd < 0)
+                return log_error_errno(wipe_fd, "Failed to reopen swtpm state directory: %m");
+        r = rm_rf_children(TAKE_FD(wipe_fd), REMOVE_PHYSICAL, /* root_dev= */ NULL);
+        if (r < 0)
+                return log_error_errno(r, "Failed to clear incomplete swtpm state directory: %m");
+
+        log_debug("TPM state directory holds no fully manufactured TPM, manufacturing a TPM.");
 
         return manufacture_swtpm(state_dir, secret);
 }