[PATCH] isolation: Add --use-chroot-fallback option

[Mateusz Andrzejewski <mandrzejewski06@gmail.com>]

For integrations, which use rootfs (tmpfs, initramfs), it is not
allowed to use pivot_root(). It results with invalid argument (EINVAL)
error. Introduce --use-chroot-fallback option as a workaround with
MS_MOVE + chroot().

Due to weaker isolation of chroot() method (we don't umount old root),
user has tu explicitly enable fallback with the new option. First,
always try to sandbox with pivot_root(). In both cases the new root is
placed into an empty tmpfs.

For the solution to work we keep CAP_SYS_CHROOT capability, which is
dropped at the end of the isolate_prefork() function.

Signed-off-by: Mateusz Andrzejewski <mateusz.andrzejewski@mikronika.com.pl>
---
 conf.c      |  9 ++++++--
 isolation.c | 63 ++++++++++++++++++++++++++++++++++++++++++++---------
 isolation.h |  2 +-
 passt.h     |  2 ++
 4 files changed, 63 insertions(+), 13 deletions(-)

diff --git a/conf.c b/conf.c
index 4755a9f..fe4d5e9 100644
--- a/conf.c
+++ b/conf.c
@@ -654,6 +654,7 @@ static void usage(const char *name, FILE *f, int status)
 		"  --no-ra		Disable router advertisements\n"
 		"  --freebind		Bind to any address for forwarding\n"
 		"  --no-map-gw		Don't map gateway address to host\n"
+		"  --use-chroot-fallback		Use chroot in case pivot_root fails\n"
 		"  -4, --ipv4-only	Enable IPv4 operation only\n"
 		"  -6, --ipv6-only	Enable IPv6 operation only\n"
 		"  -t, --tcp-ports SPEC	TCP port forwarding to %s\n"
@@ -1158,7 +1159,7 @@ static void conf_sock_listen(const struct ctx *c)
  */
 void conf(struct ctx *c, int argc, char **argv)
 {
-	int netns_only = 0, no_map_gw = 0;
+	int netns_only = 0, no_map_gw = 0, use_chroot_fallback = 0;
 	const struct option options[] = {
 		{"debug",	no_argument,		NULL,		'd' },
 		{"quiet",	no_argument,		NULL,		'q' },
@@ -1192,6 +1193,7 @@ void conf(struct ctx *c, int argc, char **argv)
 		{"splice-only",	no_argument,		&c->splice_only, 1 },
 		{"freebind",	no_argument,		&c->freebind,	1 },
 		{"no-map-gw",	no_argument,		&no_map_gw,	1 },
+		{"use-chroot-fallback",	no_argument,	&use_chroot_fallback,	1 },
 		{"ipv4-only",	no_argument,		NULL,		'4' },
 		{"ipv6-only",	no_argument,		NULL,		'6' },
 		{"one-off",	no_argument,		NULL,		'1' },
@@ -1777,6 +1779,8 @@ void conf(struct ctx *c, int argc, char **argv)
 		c->no_dns_search = 1;
 	}
 
+	c->use_chroot_fallback = use_chroot_fallback;
+
 	if (!ifi4 && *c->ip4.ifname_out)
 		ifi4 = if_nametoindex(c->ip4.ifname_out);
 
@@ -1879,7 +1883,8 @@ void conf(struct ctx *c, int argc, char **argv)
 
 	conf_open_files(c);	/* Before any possible setuid() / setgid() */
 
-	isolate_user(uid, gid, !netns_only, userns, c->mode);
+	isolate_user(uid, gid, !netns_only, userns, c->mode,
+		c->use_chroot_fallback);
 
 	if (c->no_icmp)
 		c->no_ndp = 1;
diff --git a/isolation.c b/isolation.c
index 7e6225d..4e6fd32 100644
--- a/isolation.c
+++ b/isolation.c
@@ -166,6 +166,31 @@ static void clamp_caps(void)
 		die_perror("Couldn't drop inheritable capabilities");
 }
 
+/**
+ * move_root() - Use chroot instead of pivot_root for sandboxing.
+ *
+ * Return: negative error code on failure, zero on success.
+ */
+static int move_root(void)
+{
+	if (mount(TMPDIR, "/", "", MS_MOVE, "")) {
+		err_perror("Failed to move root into empty tmpfs");
+		return -errno;
+	}
+
+	if (chroot(".")) {
+		err_perror("Failed to chroot into empty tmpfs");
+		return -errno;
+	}
+
+	if (chdir("/")) {
+		err_perror("Failed to change directory into new root");
+		return -errno;
+	}
+
+	return 0;
+}
+
 /**
  * isolate_initial() - Early, mostly config independent self isolation
  * @argc:	Argument count
@@ -201,8 +226,8 @@ void isolate_initial(int argc, char **argv)
 	 * isolate_prefork().
 	 */
 	keep = BIT(CAP_NET_BIND_SERVICE) | BIT(CAP_SETUID) | BIT(CAP_SETGID) |
-	       BIT(CAP_SYS_ADMIN) | BIT(CAP_NET_ADMIN) | BIT(CAP_DAC_OVERRIDE);
-
+	       BIT(CAP_SYS_ADMIN) | BIT(CAP_NET_ADMIN) | BIT(CAP_DAC_OVERRIDE) |
+		   BIT(CAP_SYS_CHROOT);
 	/* Since Linux 5.12, if we want to update /proc/self/uid_map to create
 	 * a mapping from UID 0, which only happens with pasta spawning a child
 	 * from a non-init user namespace (pasta can't run as root), we need to
@@ -225,6 +250,7 @@ void isolate_initial(int argc, char **argv)
  * @use_userns:	Whether to join or create a userns
  * @userns:	userns path to enter, may be empty
  * @mode:	Mode (passt or pasta)
+ * @use_chroot_fallback: Whether to use chroot() as fallback for sandboxing.
  *
  * Should:
  *  - set our final UID and GID
@@ -233,7 +259,7 @@ void isolate_initial(int argc, char **argv)
  *  - remove filesystem access (we need that for further setup)
  */
 void isolate_user(uid_t uid, gid_t gid, bool use_userns, const char *userns,
-		  enum passt_modes mode)
+		  enum passt_modes mode, bool use_chroot_fallback)
 {
 	uint64_t ns_caps = 0;
 
@@ -277,6 +303,13 @@ void isolate_user(uid_t uid, gid_t gid, bool use_userns, const char *userns,
 	 * netns
 	 */
 	ns_caps |= BIT(CAP_SYS_ADMIN);
+
+	/* Keep CAP_SYS_CHROOT, so we can fallback to chroot
+	* in case of use-chroot-fallback option being enabled
+	*/
+	if (use_chroot_fallback)
+		ns_caps |= BIT(CAP_SYS_CHROOT);
+
 	if (mode == MODE_PASTA) {
 		/* Keep CAP_NET_ADMIN, so we can configure the if */
 		ns_caps |= BIT(CAP_NET_ADMIN);
@@ -310,6 +343,7 @@ int isolate_prefork(const struct ctx *c)
 {
 	int flags = CLONE_NEWIPC | CLONE_NEWNS | CLONE_NEWUTS;
 	uint64_t ns_caps = 0;
+	int result = -1;
 
 	/* If we run in foreground, we have no chance to actually move to a new
 	 * PID namespace. For passt, use CLONE_NEWPID anyway, in case somebody
@@ -331,7 +365,7 @@ int isolate_prefork(const struct ctx *c)
 	if (mount("", TMPDIR, "tmpfs",
 		  MS_NODEV | MS_NOEXEC | MS_NOSUID | MS_RDONLY,
 		  "nr_inodes=2,nr_blocks=0")) {
-		err_perror("Failed to mount empty tmpfs for pivot_root()");
+		err_perror("Failed to mount empty tmpfs for sandboxing");
 		return -errno;
 	}
 
@@ -341,13 +375,22 @@ int isolate_prefork(const struct ctx *c)
 	}
 
 	if (syscall(SYS_pivot_root, ".", ".")) {
-		err_perror("Failed to pivot_root() into empty tmpfs");
-		return -errno;
+		if (c->use_chroot_fallback) {
+			info("Failed to pivot_root(), fallback to chroot()...");
+			result = move_root();
+			if (result)
+				return result;
+		}
+		else {
+			err_perror("Failed to pivot_root() into empty tmpfs");
+			return -errno;
+		}
 	}
-
-	if (umount2(".", MNT_DETACH | UMOUNT_NOFOLLOW)) {
-		err_perror("Failed to unmount original root filesystem");
-		return -errno;
+	else {
+		if (umount2(".", MNT_DETACH | UMOUNT_NOFOLLOW)) {
+			err_perror("Failed to unmount original root filesystem");
+			return -errno;
+		}
 	}
 
 	/* Now that initialization is more-or-less complete, we can
diff --git a/isolation.h b/isolation.h
index 0576168..ea8dacb 100644
--- a/isolation.h
+++ b/isolation.h
@@ -12,7 +12,7 @@
 
 void isolate_initial(int argc, char **argv);
 void isolate_user(uid_t uid, gid_t gid, bool use_userns, const char *userns,
-		  enum passt_modes mode);
+		  enum passt_modes mode, bool use_chroot_fallback);
 int isolate_prefork(const struct ctx *c);
 void isolate_postfork(const struct ctx *c);
 
diff --git a/passt.h b/passt.h
index 16506dc..2085cca 100644
--- a/passt.h
+++ b/passt.h
@@ -214,6 +214,7 @@ struct ip6_ctx {
  * @splice_only:	Only enable loopback forwarding
  * @host_lo_to_ns_lo:	Map host loopback addresses to ns loopback addresses
  * @freebind:		Allow binding of non-local addresses for forwarding
+ * @use_chroot_fallback:		Use chroot in case pivot_root fails
  * @low_wmem:		Low probed net.core.wmem_max
  * @low_rmem:		Low probed net.core.rmem_max
  * @no_bindtodevice:	Unprivileged SO_BINDTODEVICE not available
@@ -299,6 +300,7 @@ struct ctx {
 	int splice_only;
 	int host_lo_to_ns_lo;
 	int freebind;
+	bool use_chroot_fallback;
 
 	int low_wmem;
 	int low_rmem;
-- 
2.43.7