fs: handle circular mappings correctly

From: Christian Brauner <[email protected]>

This test makes sure that setattr behaves correctly on idmapped mounts that make use of circular mappings. Such mappings may e.g. be used to allow two users to share home directories through the same idmapped mount. The tests are explained in detail in code comments.

Cc Seth Forshee <[email protected]>

Cc Eryu Guan <[email protected]>

Cc Christoph Hellwig <[email protected]>

Cc [email protected]

CC [email protected]

Signed-off-by Christian Brauner <[email protected]>

---
 src/idmapped-mounts/idmapped-mounts.c | 434 +++++++++++++++++++++++++-
 src/idmapped-mounts/utils.c           |  43 ++-
 tests/generic/651                     |  27 ++
 tests/generic/651.out                 |   2 +
 4 files changed, 503 insertions(+), 3 deletions(-)
 create mode 100755 tests/generic/651
 create mode 100644 tests/generic/651.out

src/idmapped-mounts/idmapped-mounts.c +433 -1

index 83b7c89a..b803d171 100644
@@ -14,6 +14,7 @@
 #include <linux/limits.h>
 #include <linux/types.h>
 #include <pthread.h>
+#include <pwd.h>
 #include <sched.h>
 #include <stdbool.h>
 #include <sys/fsuid.h>

@@ -409,6 +410,23 @@ static inline bool switch_fsids(uid_t fsuid, gid_t fsgid)
 	return true;
 }
 
+static inline bool switch_resids(uid_t uid, gid_t gid)
+{
+	if (setresgid(gid, gid, gid))
+		return log_errno(false, "failure: setregid");
+
+	if (setresuid(uid, uid, uid))
+		return log_errno(false, "failure: setresuid");
+
+	if (setfsgid(-1) != gid)
+		return log_errno(false, "failure: setfsgid(-1)");
+
+	if (setfsuid(-1) != uid)
+		return log_errno(false, "failure: setfsuid(-1)");
+
+	return true;
+}
+
 static inline bool switch_userns(int fd, uid_t uid, gid_t gid, bool drop_caps)
 {
 	if (setns(fd, CLONE_NEWUSER))

@@ -636,6 +654,11 @@ __attribute__((unused)) static int print_r(int fd, const char *path)
 
 	return ret;
 }
+#else
+__attribute__((unused)) static int print_r(int fd, const char *path)
+{
+	return 0;
+}
 #endif
 
 /* fd_to_fd - transfer data from one fd to another */

@@ -13325,6 +13348,401 @@ out:
 	return fret;
 }
 
+#define USER1 "fsgqa"
+#define USER2 "fsgqa2"
+
+/**
+ * lookup_ids - lookup uid and gid for a username
+ * @name: [in]  name of the user
+ * @uid:  [out] pointer to the user-ID
+ * @gid:  [out] pointer to the group-ID
+ *
+ * Lookup the uid and gid of a user.
+ *
+ * Return: On success, true is returned.
+ *         On error, false is returned.
+ */
+static bool lookup_ids(const char *name, uid_t *uid, gid_t *gid)
+{
+	bool bret = false;
+	struct passwd *pwentp = NULL;
+	struct passwd pwent;
+	char *buf;
+	ssize_t bufsize;
+	int ret;
+
+	bufsize = sysconf(_SC_GETPW_R_SIZE_MAX);
+	if (bufsize < 0)
+		bufsize = 1024;
+
+	buf = malloc(bufsize);
+	if (!buf)
+		return bret;
+
+	ret = getpwnam_r(name, &pwent, buf, bufsize, &pwentp);
+	if (!ret && pwentp) {
+		*uid = pwent.pw_uid;
+		*gid = pwent.pw_gid;
+		bret = true;
+	}
+
+	free(buf);
+	return bret;
+}
+
+/**
+ * setattr_fix - test that setattr works correctly
+ *
+ * Test that ->setattr() works correctly for idmapped mounts with circular
+ * idmappings such as:
+ *
+ * b:1000:1001:1
+ * b:1001:1000:1
+ *
+ * Assume a directory /source with two files:
+ *
+ * /source/file1 | 1000:1000
+ * /source/file2 | 1001:1001
+ *
+ * and we create an idmapped mount of /source at /target with an idmapped of:
+ *
+ * mnt_userns:        1000:1001:1
+ *                    1001:1000:1
+ *
+ * In the idmapped mount file1 will be owned by uid 1001 and file2 by uid 1000:
+ *
+ * /target/file1 | 1001:1001
+ * /target/file2 | 1000:1000
+ *
+ * Because in essence the idmapped mount switches ownership for {g,u}id 1000
+ * and {g,u}id 1001.
+ *
+ * 1. A user with fs{g,u}id 1000 must be allowed to setattr /target/file2 from
+ *    {g,u}id 1000 in the idmapped mount to {g,u}id 1000.
+ * 2. A user with fs{g,u}id 1001 must be allowed to setattr /target/file1 from
+ *    {g,u}id 1001 in the idmapped mount to {g,u}id 1001.
+ * 3. A user with fs{g,u}id 1000 must fail to setattr /target/file1 from
+ *    {g,u}id 1001 in the idmapped mount to {g,u}id 1000.
+ *    This must fail with EPERM. The caller's fs{g,u}id doesn't match the
+ *    {g,u}id of the file.
+ * 4. A user with fs{g,u}id 1001 must fail to setattr /target/file2 from
+ *    {g,u}id 1000 in the idmapped mount to {g,u}id 1000.
+ *    This must fail with EPERM. The caller's fs{g,u}id doesn't match the
+ *    {g,u}id of the file.
+ * 5. Both, a user with fs{g,u}id 1000 and a user with fs{g,u}id 1001, must
+ *    fail to setattr /target/file1 owned by {g,u}id 1001 in the idmapped mount
+ *    and /target/file2 owned by {g,u}id 1000 in the idmapped mount to any
+ *    {g,u}id apart from {g,u}id 1000 or 1001 with EINVAL.
+ *    Only {g,u}id 1000 and 1001 have a mapping in the idmapped mount. Other
+ *    {g,u}id are unmapped.
+ */
+static int setattr_fix(void)
+{
+	int fret = -1;
+	int open_tree_fd = -EBADF;
+	struct mount_attr attr = {
+		.attr_set	= MOUNT_ATTR_IDMAP,
+		.userns_fd	= -EBADF,
+	};
+	int ret;
+	uid_t user1_uid, user2_uid;
+	gid_t user1_gid, user2_gid;
+	pid_t pid;
+	struct list idmap;
+	struct list *it_cur, *it_next;
+
+	if (!caps_supported())
+		return 0;
+
+	list_init(&idmap);
+
+	if (!lookup_ids(USER1, &user1_uid, &user1_gid)) {
+		log_stderr("failure: lookup_user");
+		goto out;
+	}
+
+	if (!lookup_ids(USER2, &user2_uid, &user2_gid)) {
+		log_stderr("failure: lookup_user");
+		goto out;
+	}
+
+	log_debug("Found " USER1 " with uid(%d) and gid(%d) and " USER2 " with uid(%d) and gid(%d)",
+		  user1_uid, user1_gid, user2_uid, user2_gid);
+
+	if (mkdirat(t_dir1_fd, DIR1, 0777)) {
+		log_stderr("failure: mkdirat");
+		goto out;
+	}
+
+	if (mknodat(t_dir1_fd, DIR1 "/" FILE1, S_IFREG | 0644, 0)) {
+		log_stderr("failure: mknodat");
+		goto out;
+	}
+
+	if (chown_r(t_mnt_fd, T_DIR1, user1_uid, user1_gid)) {
+		log_stderr("failure: chown_r");
+		goto out;
+	}
+
+	if (mknodat(t_dir1_fd, DIR1 "/" FILE2, S_IFREG | 0644, 0)) {
+		log_stderr("failure: mknodat");
+		goto out;
+	}
+
+	if (fchownat(t_dir1_fd, DIR1 "/" FILE2, user2_uid, user2_gid, AT_SYMLINK_NOFOLLOW)) {
+		log_stderr("failure: fchownat");
+		goto out;
+	}
+
+	print_r(t_mnt_fd, T_DIR1);
+
+	/* u:1000:1001:1 */
+	ret = add_map_entry(&idmap, user1_uid, user2_uid, 1, ID_TYPE_UID);
+	if (ret) {
+		log_stderr("failure: add_map_entry");
+		goto out;
+	}
+
+	/* u:1001:1000:1 */
+	ret = add_map_entry(&idmap, user2_uid, user1_uid, 1, ID_TYPE_UID);
+	if (ret) {
+		log_stderr("failure: add_map_entry");
+		goto out;
+	}
+
+	/* g:1000:1001:1 */
+	ret = add_map_entry(&idmap, user1_gid, user2_gid, 1, ID_TYPE_GID);
+	if (ret) {
+		log_stderr("failure: add_map_entry");
+		goto out;
+	}
+
+	/* g:1001:1000:1 */
+	ret = add_map_entry(&idmap, user2_gid, user1_gid, 1, ID_TYPE_GID);
+	if (ret) {
+		log_stderr("failure: add_map_entry");
+		goto out;
+	}
+
+	attr.userns_fd = get_userns_fd_from_idmap(&idmap);
+	if (attr.userns_fd < 0) {
+		log_stderr("failure: get_userns_fd");
+		goto out;
+	}
+
+	open_tree_fd = sys_open_tree(t_dir1_fd, DIR1,
+				     AT_NO_AUTOMOUNT |
+				     AT_SYMLINK_NOFOLLOW |
+				     OPEN_TREE_CLOEXEC |
+				     OPEN_TREE_CLONE |
+				     AT_RECURSIVE);
+	if (open_tree_fd < 0) {
+		log_stderr("failure: sys_open_tree");
+		goto out;
+	}
+
+	if (sys_mount_setattr(open_tree_fd, "", AT_EMPTY_PATH, &attr, sizeof(attr))) {
+		log_stderr("failure: sys_mount_setattr");
+		goto out;
+	}
+
+	print_r(open_tree_fd, "");
+
+	pid = fork();
+	if (pid < 0) {
+		log_stderr("failure: fork");
+		goto out;
+	}
+	if (pid == 0) {
+		/* switch to {g,u}id 1001 */
+		if (!switch_resids(user2_uid, user2_gid))
+			die("failure: switch_resids");
+
+		/* drop all capabilities */
+		if (!caps_down())
+			die("failure: caps_down");
+
+		/*
+		 * The {g,u}id 0 is not mapped in this idmapped mount so this
+		 * needs to fail with EINVAL.
+		 */
+		if (!fchownat(open_tree_fd, FILE1, 0, 0, AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EINVAL)
+			die("failure: errno");
+
+		/*
+		 * A user with fs{g,u}id 1001 must be allowed to change
+		 * ownership of /target/file1 owned by {g,u}id 1001 in this
+		 * idmapped mount to {g,u}id 1001.
+		 */
+		if (fchownat(open_tree_fd, FILE1, user2_uid, user2_gid,
+			     AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+
+		/* Verify that the ownership is still {g,u}id 1001. */
+		if (!expected_uid_gid(open_tree_fd, FILE1, AT_SYMLINK_NOFOLLOW,
+				      user2_uid, user2_gid))
+			die("failure: check ownership");
+
+		/*
+		 * A user with fs{g,u}id 1001 must not be allowed to change
+		 * ownership of /target/file1 owned by {g,u}id 1001 in this
+		 * idmapped mount to {g,u}id 1000.
+		 */
+		if (!fchownat(open_tree_fd, FILE1, user1_uid, user1_gid,
+			      AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EPERM)
+			die("failure: errno");
+
+		/* Verify that the ownership is still {g,u}id 1001. */
+		if (!expected_uid_gid(open_tree_fd, FILE1, AT_SYMLINK_NOFOLLOW,
+				      user2_uid, user2_gid))
+			die("failure: check ownership");
+
+		/*
+		 * A user with fs{g,u}id 1001 must not be allowed to change
+		 * ownership of /target/file2 owned by {g,u}id 1000 in this
+		 * idmapped mount to {g,u}id 1000.
+		 */
+		if (!fchownat(open_tree_fd, FILE2, user1_uid, user1_gid,
+			      AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EPERM)
+			die("failure: errno");
+
+		/* Verify that the ownership is still {g,u}id 1000. */
+		if (!expected_uid_gid(open_tree_fd, FILE2, AT_SYMLINK_NOFOLLOW,
+				      user1_uid, user1_gid))
+			die("failure: check ownership");
+
+		/*
+		 * A user with fs{g,u}id 1001 must not be allowed to change
+		 * ownership of /target/file2 owned by {g,u}id 1000 in this
+		 * idmapped mount to {g,u}id 1001.
+		 */
+		if (!fchownat(open_tree_fd, FILE2, user2_uid, user2_gid,
+			      AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EPERM)
+			die("failure: errno");
+
+		/* Verify that the ownership is still {g,u}id 1000. */
+		if (!expected_uid_gid(open_tree_fd, FILE2, AT_SYMLINK_NOFOLLOW,
+				      user1_uid, user1_gid))
+			die("failure: check ownership");
+
+		exit(EXIT_SUCCESS);
+	}
+	if (wait_for_pid(pid))
+		goto out;
+
+	pid = fork();
+	if (pid < 0) {
+		log_stderr("failure: fork");
+		goto out;
+	}
+	if (pid == 0) {
+		/* switch to {g,u}id 1000 */
+		if (!switch_resids(user1_uid, user1_gid))
+			die("failure: switch_resids");
+
+		/* drop all capabilities */
+		if (!caps_down())
+			die("failure: caps_down");
+
+		/*
+		 * The {g,u}id 0 is not mapped in this idmapped mount so this
+		 * needs to fail with EINVAL.
+		 */
+		if (!fchownat(open_tree_fd, FILE1, 0, 0, AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EINVAL)
+			die("failure: errno");
+
+		/*
+		 * A user with fs{g,u}id 1000 must be allowed to change
+		 * ownership of /target/file2 owned by {g,u}id 1000 in this
+		 * idmapped mount to {g,u}id 1000.
+		 */
+		if (fchownat(open_tree_fd, FILE2, user1_uid, user1_gid,
+			     AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+
+		/* Verify that the ownership is still {g,u}id 1000. */
+		if (!expected_uid_gid(open_tree_fd, FILE2, AT_SYMLINK_NOFOLLOW,
+				      user1_uid, user1_gid))
+			die("failure: check ownership");
+
+		/*
+		 * A user with fs{g,u}id 1000 must not be allowed to change
+		 * ownership of /target/file2 owned by {g,u}id 1000 in this
+		 * idmapped mount to {g,u}id 1001.
+		 */
+		if (!fchownat(open_tree_fd, FILE2, user2_uid, user2_gid,
+			      AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EPERM)
+			die("failure: errno");
+
+		/* Verify that the ownership is still {g,u}id 1000. */
+		if (!expected_uid_gid(open_tree_fd, FILE2, AT_SYMLINK_NOFOLLOW,
+				      user1_uid, user1_gid))
+			die("failure: check ownership");
+
+		/*
+		 * A user with fs{g,u}id 1000 must not be allowed to change
+		 * ownership of /target/file1 owned by {g,u}id 1001 in this
+		 * idmapped mount to {g,u}id 1000.
+		 */
+		if (!fchownat(open_tree_fd, FILE1, user1_uid, user1_gid,
+			     AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EPERM)
+			die("failure: errno");
+
+		/* Verify that the ownership is still {g,u}id 1001. */
+		if (!expected_uid_gid(open_tree_fd, FILE1, AT_SYMLINK_NOFOLLOW,
+				      user2_uid, user2_gid))
+			die("failure: check ownership");
+
+		/*
+		 * A user with fs{g,u}id 1000 must not be allowed to change
+		 * ownership of /target/file1 owned by {g,u}id 1001 in this
+		 * idmapped mount to {g,u}id 1001.
+		 */
+		if (!fchownat(open_tree_fd, FILE1, user2_uid, user2_gid,
+			      AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EPERM)
+			die("failure: errno");
+
+		/* Verify that the ownership is still {g,u}id 1001. */
+		if (!expected_uid_gid(open_tree_fd, FILE1, AT_SYMLINK_NOFOLLOW,
+				      user2_uid, user2_gid))
+			die("failure: check ownership");
+
+		exit(EXIT_SUCCESS);
+	}
+	if (wait_for_pid(pid))
+		goto out;
+
+	fret = 0;
+	log_debug("Ran test");
+out:
+	safe_close(attr.userns_fd);
+	safe_close(open_tree_fd);
+
+	list_for_each_safe(it_cur, &idmap, it_next) {
+		list_del(it_cur);
+		free(it_cur->elem);
+		free(it_cur);
+	}
+
+	return fret;
+}
+
 static void usage(void)
 {
 	fprintf(stderr, "Description:\n");

@@ -13342,6 +13760,7 @@ static void usage(void)
 	fprintf(stderr, "--test-fscaps-regression     Run fscap regression tests\n");
 	fprintf(stderr, "--test-nested-userns         Run nested userns idmapped mount testsuite\n");
 	fprintf(stderr, "--test-btrfs                 Run btrfs specific idmapped mount testsuite\n");
+	fprintf(stderr, "--test-setattr-fix           Run setattr regression tests\n");
 
 	_exit(EXIT_SUCCESS);
 }

@@ -13358,6 +13777,7 @@ static const struct option longopts[] = {
 	{"test-fscaps-regression",	no_argument,		0,	'g'},
 	{"test-nested-userns",		no_argument,		0,	'n'},
 	{"test-btrfs",			no_argument,		0,	'b'},
+	{"test-setattr-fix",		no_argument,		0,	'i'},
 	{NULL,				0,			0,	  0},
 };
 

@@ -13449,6 +13869,10 @@ struct t_idmapped_mounts t_btrfs[] = {
 	{ btrfs_subvolume_lookup_user,					"test unprivileged subvolume lookup",								},
 };
 
+struct t_idmapped_mounts t_setattr_fix[] = {
+	{ setattr_fix,							"test that setattr works correctly",								},
+};
+
 static bool run_test(struct t_idmapped_mounts suite[], size_t suite_size)
 {
 	int i;

@@ -13487,7 +13911,8 @@ int main(int argc, char *argv[])
 	int fret, ret;
 	int index = 0;
 	bool supported = false, test_btrfs = false, test_core = false,
-	     test_fscaps_regression = false, test_nested_userns = false;
+	     test_fscaps_regression = false, test_nested_userns = false,
+	     test_setattr_fix = false;
 
 	while ((ret = getopt_long_only(argc, argv, "", longopts, &index)) != -1) {
 		switch (ret) {

@@ -13521,6 +13946,9 @@ int main(int argc, char *argv[])
 		case 'e':
 			t_device_scratch = optarg;
 			break;
+		case 'i':
+			test_setattr_fix = true;
+			break;
 		case 'h':
 			/* fallthrough */
 		default:

@@ -13609,6 +14037,10 @@ int main(int argc, char *argv[])
 	if (test_btrfs && !run_test(t_btrfs, ARRAY_SIZE(t_btrfs)))
 		goto out;
 
+	if (test_setattr_fix &&
+	    !run_test(t_setattr_fix, ARRAY_SIZE(t_setattr_fix)))
+		goto out;
+
 	fret = EXIT_SUCCESS;
 
 out:

src/idmapped-mounts/utils.c +41 -2

index c2afa8dc..faf06fcd 100644
@@ -183,6 +183,43 @@ static int map_ids_from_idmap(struct list *idmap, pid_t pid)
 	return 0;
 }
 
+#ifdef DEBUG_TRACE
+static void __print_idmaps(pid_t pid, bool gid)
+{
+	char path_mapping[STRLITERALLEN("/proc/") + INTTYPE_TO_STRLEN(pid_t) +
+			  STRLITERALLEN("/_id_map") + 1];
+	char *line = NULL;
+	size_t len = 0;
+	int ret;
+	FILE *f;
+
+	ret = snprintf(path_mapping, sizeof(path_mapping), "/proc/%d/%cid_map",
+		       pid, gid ? 'g' : 'u');
+	if (ret < 0 || (size_t)ret >= sizeof(path_mapping))
+		return;
+
+	f = fopen(path_mapping, "r");
+	if (!f)
+		return;
+
+	while ((ret = getline(&line, &len, f)) > 0)
+		fprintf(stderr, "%s", line);
+
+	fclose(f);
+	free(line);
+}
+
+static void print_idmaps(pid_t pid)
+{
+	__print_idmaps(pid, false);
+	__print_idmaps(pid, true);
+}
+#else
+static void print_idmaps(pid_t pid)
+{
+}
+#endif
+
 int get_userns_fd_from_idmap(struct list *idmap)
 {
 	int ret;

@@ -199,10 +236,12 @@ int get_userns_fd_from_idmap(struct list *idmap)
 		return ret;
 
 	ret = snprintf(path_ns, sizeof(path_ns), "/proc/%d/ns/user", pid);
-	if (ret < 0 || (size_t)ret >= sizeof(path_ns))
+	if (ret < 0 || (size_t)ret >= sizeof(path_ns)) {
 		ret = -EIO;
-	else
+	} else {
 		ret = open(path_ns, O_RDONLY | O_CLOEXEC | O_NOCTTY);
+		print_idmaps(pid);
+	}
 
 	(void)kill(pid, SIGKILL);
 	(void)wait_for_pid(pid);

tests/generic/651 new file +9 -0

new file mode 100755
index 00000000..49b288d0
@@ -0,0 +1,27 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2021 Christian Brauner.  All Rights Reserved.
+#
+# FS QA Test 651
+#
+# Test that setattr works correctly.
+#
+. ./common/preamble

tests/generic/651 new file +8 -0

+
+echo "Silence is golden"
+
+$here/src/idmapped-mounts/idmapped-mounts --test-setattr-fix \
+	--device "$TEST_DEV" --mount "$TEST_DIR" --fstype "$FSTYP"
+
+status=$?
+exit

tests/generic/651.out new file +2 -0

new file mode 100644
index 00000000..51d73665
@@ -0,0 +1,2 @@
+QA output created by 651
+Silence is golden

base-commit: bae1d15f6421cbe99b3e2e134c39d50248e7c261

Signature · 1 line

2.30.2

Quoted 7 lines · Christian Brauner

From: Christian Brauner <[email protected]>

This test makes sure that setattr behaves correctly on idmapped mounts that make use of circular mappings. Such mappings may e.g. be used to allow two users to share home directories through the same idmapped mount. The tests are explained in detail in code comments.

Quoted 6 lines

Cc Seth Forshee <[email protected]>

Cc Eryu Guan <[email protected]>

Cc Christoph Hellwig <[email protected]>

Cc [email protected]

CC [email protected]

Signed-off-by Christian Brauner <[email protected]>

Is this test known to fail with v5.15-rc7 kernel? I hit the following failure:

If the bug has been fixed, it'd be better to document the commit that fixed the issue in both commit log and test description.

And is there a "1/2" patch? I only see this "2/2" in mailing list.

Quoted 596 lines

---
 src/idmapped-mounts/idmapped-mounts.c | 434 +++++++++++++++++++++++++-
 src/idmapped-mounts/utils.c           |  43 ++-
 tests/generic/651                     |  27 ++
 tests/generic/651.out                 |   2 +
 4 files changed, 503 insertions(+), 3 deletions(-)
 create mode 100755 tests/generic/651
 create mode 100644 tests/generic/651.out

src/idmapped-mounts/idmapped-mounts.c +433 -1

index 83b7c89a..b803d171 100644
@@ -14,6 +14,7 @@
 #include <linux/limits.h>
 #include <linux/types.h>
 #include <pthread.h>
+#include <pwd.h>
 #include <sched.h>
 #include <stdbool.h>
 #include <sys/fsuid.h>

@@ -409,6 +410,23 @@ static inline bool switch_fsids(uid_t fsuid, gid_t fsgid)
 	return true;
 }
 
+static inline bool switch_resids(uid_t uid, gid_t gid)
+{
+	if (setresgid(gid, gid, gid))
+		return log_errno(false, "failure: setregid");
+
+	if (setresuid(uid, uid, uid))
+		return log_errno(false, "failure: setresuid");
+
+	if (setfsgid(-1) != gid)
+		return log_errno(false, "failure: setfsgid(-1)");
+
+	if (setfsuid(-1) != uid)
+		return log_errno(false, "failure: setfsuid(-1)");
+
+	return true;
+}
+
 static inline bool switch_userns(int fd, uid_t uid, gid_t gid, bool drop_caps)
 {
 	if (setns(fd, CLONE_NEWUSER))

@@ -636,6 +654,11 @@ __attribute__((unused)) static int print_r(int fd, const char *path)
 
 	return ret;
 }
+#else
+__attribute__((unused)) static int print_r(int fd, const char *path)
+{
+	return 0;
+}
 #endif
 
 /* fd_to_fd - transfer data from one fd to another */

@@ -13325,6 +13348,401 @@ out:
 	return fret;
 }
 
+#define USER1 "fsgqa"
+#define USER2 "fsgqa2"
+
+/**
+ * lookup_ids - lookup uid and gid for a username
+ * @name: [in]  name of the user
+ * @uid:  [out] pointer to the user-ID
+ * @gid:  [out] pointer to the group-ID
+ *
+ * Lookup the uid and gid of a user.
+ *
+ * Return: On success, true is returned.
+ *         On error, false is returned.
+ */
+static bool lookup_ids(const char *name, uid_t *uid, gid_t *gid)
+{
+	bool bret = false;
+	struct passwd *pwentp = NULL;
+	struct passwd pwent;
+	char *buf;
+	ssize_t bufsize;
+	int ret;
+
+	bufsize = sysconf(_SC_GETPW_R_SIZE_MAX);
+	if (bufsize < 0)
+		bufsize = 1024;
+
+	buf = malloc(bufsize);
+	if (!buf)
+		return bret;
+
+	ret = getpwnam_r(name, &pwent, buf, bufsize, &pwentp);
+	if (!ret && pwentp) {
+		*uid = pwent.pw_uid;
+		*gid = pwent.pw_gid;
+		bret = true;
+	}
+
+	free(buf);
+	return bret;
+}
+
+/**
+ * setattr_fix - test that setattr works correctly
+ *
+ * Test that ->setattr() works correctly for idmapped mounts with circular
+ * idmappings such as:
+ *
+ * b:1000:1001:1
+ * b:1001:1000:1
+ *
+ * Assume a directory /source with two files:
+ *
+ * /source/file1 | 1000:1000
+ * /source/file2 | 1001:1001
+ *
+ * and we create an idmapped mount of /source at /target with an idmapped of:
+ *
+ * mnt_userns:        1000:1001:1
+ *                    1001:1000:1
+ *
+ * In the idmapped mount file1 will be owned by uid 1001 and file2 by uid 1000:
+ *
+ * /target/file1 | 1001:1001
+ * /target/file2 | 1000:1000
+ *
+ * Because in essence the idmapped mount switches ownership for {g,u}id 1000
+ * and {g,u}id 1001.
+ *
+ * 1. A user with fs{g,u}id 1000 must be allowed to setattr /target/file2 from
+ *    {g,u}id 1000 in the idmapped mount to {g,u}id 1000.
+ * 2. A user with fs{g,u}id 1001 must be allowed to setattr /target/file1 from
+ *    {g,u}id 1001 in the idmapped mount to {g,u}id 1001.
+ * 3. A user with fs{g,u}id 1000 must fail to setattr /target/file1 from
+ *    {g,u}id 1001 in the idmapped mount to {g,u}id 1000.
+ *    This must fail with EPERM. The caller's fs{g,u}id doesn't match the
+ *    {g,u}id of the file.
+ * 4. A user with fs{g,u}id 1001 must fail to setattr /target/file2 from
+ *    {g,u}id 1000 in the idmapped mount to {g,u}id 1000.
+ *    This must fail with EPERM. The caller's fs{g,u}id doesn't match the
+ *    {g,u}id of the file.
+ * 5. Both, a user with fs{g,u}id 1000 and a user with fs{g,u}id 1001, must
+ *    fail to setattr /target/file1 owned by {g,u}id 1001 in the idmapped mount
+ *    and /target/file2 owned by {g,u}id 1000 in the idmapped mount to any
+ *    {g,u}id apart from {g,u}id 1000 or 1001 with EINVAL.
+ *    Only {g,u}id 1000 and 1001 have a mapping in the idmapped mount. Other
+ *    {g,u}id are unmapped.
+ */
+static int setattr_fix(void)
+{
+	int fret = -1;
+	int open_tree_fd = -EBADF;
+	struct mount_attr attr = {
+		.attr_set	= MOUNT_ATTR_IDMAP,
+		.userns_fd	= -EBADF,
+	};
+	int ret;
+	uid_t user1_uid, user2_uid;
+	gid_t user1_gid, user2_gid;
+	pid_t pid;
+	struct list idmap;
+	struct list *it_cur, *it_next;
+
+	if (!caps_supported())
+		return 0;
+
+	list_init(&idmap);
+
+	if (!lookup_ids(USER1, &user1_uid, &user1_gid)) {
+		log_stderr("failure: lookup_user");
+		goto out;
+	}
+
+	if (!lookup_ids(USER2, &user2_uid, &user2_gid)) {
+		log_stderr("failure: lookup_user");
+		goto out;
+	}
+
+	log_debug("Found " USER1 " with uid(%d) and gid(%d) and " USER2 " with uid(%d) and gid(%d)",
+		  user1_uid, user1_gid, user2_uid, user2_gid);
+
+	if (mkdirat(t_dir1_fd, DIR1, 0777)) {
+		log_stderr("failure: mkdirat");
+		goto out;
+	}
+
+	if (mknodat(t_dir1_fd, DIR1 "/" FILE1, S_IFREG | 0644, 0)) {
+		log_stderr("failure: mknodat");
+		goto out;
+	}
+
+	if (chown_r(t_mnt_fd, T_DIR1, user1_uid, user1_gid)) {
+		log_stderr("failure: chown_r");
+		goto out;
+	}
+
+	if (mknodat(t_dir1_fd, DIR1 "/" FILE2, S_IFREG | 0644, 0)) {
+		log_stderr("failure: mknodat");
+		goto out;
+	}
+
+	if (fchownat(t_dir1_fd, DIR1 "/" FILE2, user2_uid, user2_gid, AT_SYMLINK_NOFOLLOW)) {
+		log_stderr("failure: fchownat");
+		goto out;
+	}
+
+	print_r(t_mnt_fd, T_DIR1);
+
+	/* u:1000:1001:1 */
+	ret = add_map_entry(&idmap, user1_uid, user2_uid, 1, ID_TYPE_UID);
+	if (ret) {
+		log_stderr("failure: add_map_entry");
+		goto out;
+	}
+
+	/* u:1001:1000:1 */
+	ret = add_map_entry(&idmap, user2_uid, user1_uid, 1, ID_TYPE_UID);
+	if (ret) {
+		log_stderr("failure: add_map_entry");
+		goto out;
+	}
+
+	/* g:1000:1001:1 */
+	ret = add_map_entry(&idmap, user1_gid, user2_gid, 1, ID_TYPE_GID);
+	if (ret) {
+		log_stderr("failure: add_map_entry");
+		goto out;
+	}
+
+	/* g:1001:1000:1 */
+	ret = add_map_entry(&idmap, user2_gid, user1_gid, 1, ID_TYPE_GID);
+	if (ret) {
+		log_stderr("failure: add_map_entry");
+		goto out;
+	}
+
+	attr.userns_fd = get_userns_fd_from_idmap(&idmap);
+	if (attr.userns_fd < 0) {
+		log_stderr("failure: get_userns_fd");
+		goto out;
+	}
+
+	open_tree_fd = sys_open_tree(t_dir1_fd, DIR1,
+				     AT_NO_AUTOMOUNT |
+				     AT_SYMLINK_NOFOLLOW |
+				     OPEN_TREE_CLOEXEC |
+				     OPEN_TREE_CLONE |
+				     AT_RECURSIVE);
+	if (open_tree_fd < 0) {
+		log_stderr("failure: sys_open_tree");
+		goto out;
+	}
+
+	if (sys_mount_setattr(open_tree_fd, "", AT_EMPTY_PATH, &attr, sizeof(attr))) {
+		log_stderr("failure: sys_mount_setattr");
+		goto out;
+	}
+
+	print_r(open_tree_fd, "");
+
+	pid = fork();
+	if (pid < 0) {
+		log_stderr("failure: fork");
+		goto out;
+	}
+	if (pid == 0) {
+		/* switch to {g,u}id 1001 */
+		if (!switch_resids(user2_uid, user2_gid))
+			die("failure: switch_resids");
+
+		/* drop all capabilities */
+		if (!caps_down())
+			die("failure: caps_down");
+
+		/*
+		 * The {g,u}id 0 is not mapped in this idmapped mount so this
+		 * needs to fail with EINVAL.
+		 */
+		if (!fchownat(open_tree_fd, FILE1, 0, 0, AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EINVAL)
+			die("failure: errno");
+
+		/*
+		 * A user with fs{g,u}id 1001 must be allowed to change
+		 * ownership of /target/file1 owned by {g,u}id 1001 in this
+		 * idmapped mount to {g,u}id 1001.
+		 */
+		if (fchownat(open_tree_fd, FILE1, user2_uid, user2_gid,
+			     AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+
+		/* Verify that the ownership is still {g,u}id 1001. */
+		if (!expected_uid_gid(open_tree_fd, FILE1, AT_SYMLINK_NOFOLLOW,
+				      user2_uid, user2_gid))
+			die("failure: check ownership");
+
+		/*
+		 * A user with fs{g,u}id 1001 must not be allowed to change
+		 * ownership of /target/file1 owned by {g,u}id 1001 in this
+		 * idmapped mount to {g,u}id 1000.
+		 */
+		if (!fchownat(open_tree_fd, FILE1, user1_uid, user1_gid,
+			      AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EPERM)
+			die("failure: errno");
+
+		/* Verify that the ownership is still {g,u}id 1001. */
+		if (!expected_uid_gid(open_tree_fd, FILE1, AT_SYMLINK_NOFOLLOW,
+				      user2_uid, user2_gid))
+			die("failure: check ownership");
+
+		/*
+		 * A user with fs{g,u}id 1001 must not be allowed to change
+		 * ownership of /target/file2 owned by {g,u}id 1000 in this
+		 * idmapped mount to {g,u}id 1000.
+		 */
+		if (!fchownat(open_tree_fd, FILE2, user1_uid, user1_gid,
+			      AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EPERM)
+			die("failure: errno");
+
+		/* Verify that the ownership is still {g,u}id 1000. */
+		if (!expected_uid_gid(open_tree_fd, FILE2, AT_SYMLINK_NOFOLLOW,
+				      user1_uid, user1_gid))
+			die("failure: check ownership");
+
+		/*
+		 * A user with fs{g,u}id 1001 must not be allowed to change
+		 * ownership of /target/file2 owned by {g,u}id 1000 in this
+		 * idmapped mount to {g,u}id 1001.
+		 */
+		if (!fchownat(open_tree_fd, FILE2, user2_uid, user2_gid,
+			      AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EPERM)
+			die("failure: errno");
+
+		/* Verify that the ownership is still {g,u}id 1000. */
+		if (!expected_uid_gid(open_tree_fd, FILE2, AT_SYMLINK_NOFOLLOW,
+				      user1_uid, user1_gid))
+			die("failure: check ownership");
+
+		exit(EXIT_SUCCESS);
+	}
+	if (wait_for_pid(pid))
+		goto out;
+
+	pid = fork();
+	if (pid < 0) {
+		log_stderr("failure: fork");
+		goto out;
+	}
+	if (pid == 0) {
+		/* switch to {g,u}id 1000 */
+		if (!switch_resids(user1_uid, user1_gid))
+			die("failure: switch_resids");
+
+		/* drop all capabilities */
+		if (!caps_down())
+			die("failure: caps_down");
+
+		/*
+		 * The {g,u}id 0 is not mapped in this idmapped mount so this
+		 * needs to fail with EINVAL.
+		 */
+		if (!fchownat(open_tree_fd, FILE1, 0, 0, AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EINVAL)
+			die("failure: errno");
+
+		/*
+		 * A user with fs{g,u}id 1000 must be allowed to change
+		 * ownership of /target/file2 owned by {g,u}id 1000 in this
+		 * idmapped mount to {g,u}id 1000.
+		 */
+		if (fchownat(open_tree_fd, FILE2, user1_uid, user1_gid,
+			     AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+
+		/* Verify that the ownership is still {g,u}id 1000. */
+		if (!expected_uid_gid(open_tree_fd, FILE2, AT_SYMLINK_NOFOLLOW,
+				      user1_uid, user1_gid))
+			die("failure: check ownership");
+
+		/*
+		 * A user with fs{g,u}id 1000 must not be allowed to change
+		 * ownership of /target/file2 owned by {g,u}id 1000 in this
+		 * idmapped mount to {g,u}id 1001.
+		 */
+		if (!fchownat(open_tree_fd, FILE2, user2_uid, user2_gid,
+			      AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EPERM)
+			die("failure: errno");
+
+		/* Verify that the ownership is still {g,u}id 1000. */
+		if (!expected_uid_gid(open_tree_fd, FILE2, AT_SYMLINK_NOFOLLOW,
+				      user1_uid, user1_gid))
+			die("failure: check ownership");
+
+		/*
+		 * A user with fs{g,u}id 1000 must not be allowed to change
+		 * ownership of /target/file1 owned by {g,u}id 1001 in this
+		 * idmapped mount to {g,u}id 1000.
+		 */
+		if (!fchownat(open_tree_fd, FILE1, user1_uid, user1_gid,
+			     AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EPERM)
+			die("failure: errno");
+
+		/* Verify that the ownership is still {g,u}id 1001. */
+		if (!expected_uid_gid(open_tree_fd, FILE1, AT_SYMLINK_NOFOLLOW,
+				      user2_uid, user2_gid))
+			die("failure: check ownership");
+
+		/*
+		 * A user with fs{g,u}id 1000 must not be allowed to change
+		 * ownership of /target/file1 owned by {g,u}id 1001 in this
+		 * idmapped mount to {g,u}id 1001.
+		 */
+		if (!fchownat(open_tree_fd, FILE1, user2_uid, user2_gid,
+			      AT_SYMLINK_NOFOLLOW))
+			die("failure: change ownership");
+		if (errno != EPERM)
+			die("failure: errno");
+
+		/* Verify that the ownership is still {g,u}id 1001. */
+		if (!expected_uid_gid(open_tree_fd, FILE1, AT_SYMLINK_NOFOLLOW,
+				      user2_uid, user2_gid))
+			die("failure: check ownership");
+
+		exit(EXIT_SUCCESS);
+	}
+	if (wait_for_pid(pid))
+		goto out;
+
+	fret = 0;
+	log_debug("Ran test");
+out:
+	safe_close(attr.userns_fd);
+	safe_close(open_tree_fd);
+
+	list_for_each_safe(it_cur, &idmap, it_next) {
+		list_del(it_cur);
+		free(it_cur->elem);
+		free(it_cur);
+	}
+
+	return fret;
+}
+
 static void usage(void)
 {
 	fprintf(stderr, "Description:\n");

@@ -13342,6 +13760,7 @@ static void usage(void)
 	fprintf(stderr, "--test-fscaps-regression     Run fscap regression tests\n");
 	fprintf(stderr, "--test-nested-userns         Run nested userns idmapped mount testsuite\n");
 	fprintf(stderr, "--test-btrfs                 Run btrfs specific idmapped mount testsuite\n");
+	fprintf(stderr, "--test-setattr-fix           Run setattr regression tests\n");
 
 	_exit(EXIT_SUCCESS);
 }

@@ -13358,6 +13777,7 @@ static const struct option longopts[] = {
 	{"test-fscaps-regression",	no_argument,		0,	'g'},
 	{"test-nested-userns",		no_argument,		0,	'n'},
 	{"test-btrfs",			no_argument,		0,	'b'},
+	{"test-setattr-fix",		no_argument,		0,	'i'},
 	{NULL,				0,			0,	  0},
 };
 

@@ -13449,6 +13869,10 @@ struct t_idmapped_mounts t_btrfs[] = {
 	{ btrfs_subvolume_lookup_user,					"test unprivileged subvolume lookup",								},
 };
 
+struct t_idmapped_mounts t_setattr_fix[] = {
+	{ setattr_fix,							"test that setattr works correctly",								},
+};
+
 static bool run_test(struct t_idmapped_mounts suite[], size_t suite_size)
 {
 	int i;

@@ -13487,7 +13911,8 @@ int main(int argc, char *argv[])
 	int fret, ret;
 	int index = 0;
 	bool supported = false, test_btrfs = false, test_core = false,
-	     test_fscaps_regression = false, test_nested_userns = false;
+	     test_fscaps_regression = false, test_nested_userns = false,
+	     test_setattr_fix = false;
 
 	while ((ret = getopt_long_only(argc, argv, "", longopts, &index)) != -1) {
 		switch (ret) {

@@ -13521,6 +13946,9 @@ int main(int argc, char *argv[])
 		case 'e':
 			t_device_scratch = optarg;
 			break;
+		case 'i':
+			test_setattr_fix = true;
+			break;
 		case 'h':
 			/* fallthrough */
 		default:

@@ -13609,6 +14037,10 @@ int main(int argc, char *argv[])
 	if (test_btrfs && !run_test(t_btrfs, ARRAY_SIZE(t_btrfs)))
 		goto out;
 
+	if (test_setattr_fix &&
+	    !run_test(t_setattr_fix, ARRAY_SIZE(t_setattr_fix)))
+		goto out;
+
 	fret = EXIT_SUCCESS;
 
 out:

src/idmapped-mounts/utils.c +41 -2

index c2afa8dc..faf06fcd 100644
@@ -183,6 +183,43 @@ static int map_ids_from_idmap(struct list *idmap, pid_t pid)
 	return 0;
 }
 
+#ifdef DEBUG_TRACE
+static void __print_idmaps(pid_t pid, bool gid)
+{
+	char path_mapping[STRLITERALLEN("/proc/") + INTTYPE_TO_STRLEN(pid_t) +
+			  STRLITERALLEN("/_id_map") + 1];
+	char *line = NULL;
+	size_t len = 0;
+	int ret;
+	FILE *f;
+
+	ret = snprintf(path_mapping, sizeof(path_mapping), "/proc/%d/%cid_map",
+		       pid, gid ? 'g' : 'u');
+	if (ret < 0 || (size_t)ret >= sizeof(path_mapping))
+		return;
+
+	f = fopen(path_mapping, "r");
+	if (!f)
+		return;
+
+	while ((ret = getline(&line, &len, f)) > 0)
+		fprintf(stderr, "%s", line);
+
+	fclose(f);
+	free(line);
+}
+
+static void print_idmaps(pid_t pid)
+{
+	__print_idmaps(pid, false);
+	__print_idmaps(pid, true);
+}
+#else
+static void print_idmaps(pid_t pid)
+{
+}
+#endif
+
 int get_userns_fd_from_idmap(struct list *idmap)
 {
 	int ret;

@@ -199,10 +236,12 @@ int get_userns_fd_from_idmap(struct list *idmap)
 		return ret;
 
 	ret = snprintf(path_ns, sizeof(path_ns), "/proc/%d/ns/user", pid);
-	if (ret < 0 || (size_t)ret >= sizeof(path_ns))
+	if (ret < 0 || (size_t)ret >= sizeof(path_ns)) {
 		ret = -EIO;
-	else
+	} else {
 		ret = open(path_ns, O_RDONLY | O_CLOEXEC | O_NOCTTY);
+		print_idmaps(pid);
+	}
 
 	(void)kill(pid, SIGKILL);
 	(void)wait_for_pid(pid);

tests/generic/651 new file +9 -0

new file mode 100755
index 00000000..49b288d0
@@ -0,0 +1,27 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2021 Christian Brauner.  All Rights Reserved.
+#
+# FS QA Test 651
+#
+# Test that setattr works correctly.
+#
+. ./common/preamble

tests/generic/651 line 19

Quoted 10 lines

tests/generic/651 new file +10 -0

+_begin_fstest auto attr cap idmapped mount perms
+
+# Import common functions.
+. ./common/filter
+
+# real QA test starts here
+
+_supported_fs generic
+_require_idmapped_mounts
+_require_test

This test also requires user 'fsgqa' and 'fsgqa2'

_require_user fsgqa
_require_user fsgqa2

Thanks,
Eryu

Quoted 20 lines

tests/generic/651 new file +8 -0

+
+echo "Silence is golden"
+
+$here/src/idmapped-mounts/idmapped-mounts --test-setattr-fix \
+	--device "$TEST_DEV" --mount "$TEST_DIR" --fstype "$FSTYP"
+
+status=$?
+exit

tests/generic/651.out new file +2 -0

new file mode 100644
index 00000000..51d73665
@@ -0,0 +1,2 @@
+QA output created by 651
+Silence is golden

base-commit: bae1d15f6421cbe99b3e2e134c39d50248e7c261

Signature · 1 line

2.30.2

From: Christian Brauner <[email protected]>

When calling setattr_prepare() to determine the validity of the attributes the ia_{g,u}id fields contain the value that will be written to inode->i_{g,u}id. When the {g,u}id attribute of the file isn't altered and the caller's fs{g,u}id matches the current {g,u}id attribute the attribute change is allowed.

The value in ia_{g,u}id does already account for idmapped mounts and will have taken the relevant idmapping into account. So in order to verify that the {g,u}id attribute isn't changed we simple need to compare the ia_{g,u}id value against the inode's i_{g,u}id value.

This only has any meaning for idmapped mounts as idmapping helpers are idempotent without them. And for idmapped mounts this really only has a meaning when circular idmappings are used, i.e. mappings where e.g. id 1000 is mapped to id 1001 and id 1001 is mapped to id 1000. Such ciruclar mappings can e.g. be useful when sharing the same home directory between multiple users at the same time.

As an example consider a directory with two files: /source/file1 owned by {g,u}id 1000 and /source/file2 owned by {g,u}id 1001. Assume we create an idmapped mount at /target with an idmapping that maps files owned by {g,u}id 1000 to being owned by {g,u}id 1001 and files owned by {g,u}id 1001 to being owned by {g,u}id 1000. In effect, the idmapped mount at /target switches the ownership of /source/file1 and source/file2, i.e. /target/file1 will be owned by {g,u}id 1001 and /target/file2 will be owned by {g,u}id 1000.

This means that a user with fs{g,u}id 1000 must be allowed to setattr /target/file2 from {g,u}id 1000 to {g,u}id 1000. Similar, a user with fs{g,u}id 1001 must be allowed to setattr /target/file1 from {g,u}id 1001 to {g,u}id 1001. Conversely, a user with fs{g,u}id 1000 must fail to setattr /target/file1 from {g,u}id 1001 to {g,u}id 1000. And a user with fs{g,u}id 1001 must fail to setattr /target/file2 from {g,u}id 1000 to {g,u}id 1000. Both cases must fail with EPERM for non-capable callers.

Before this patch we could end up denying legitimate attribute changes and allowing invalid attribute changes when circular mappings are used. To even get into this situation the caller must've been privileged both to create that mapping and to create that idmapped mount.

This hasn't been seen in the wild anywhere but came up when expanding the testsuite during work on a series of hardening patches. All idmapped fstests pass without any regressions and we add new tests to verify the behavior of circular mappings.

Fixes 2f221d6f7b88 ("attr: handle idmapped mounts")

Cc Seth Forshee <[email protected]>

Cc Christoph Hellwig <[email protected]>

Cc Al Viro <[email protected]>

Cc [email protected]

CC [email protected]

Signed-off-by Christian Brauner <[email protected]>

---

 fs/attr.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

fs/attr.c +2 -2

index 473d21b3a86d..66899b6e9bd8 100644
@@ -35,7 +35,7 @@ static bool chown_ok(struct user_namespace *mnt_userns,
 		     kuid_t uid)
 {
 	kuid_t kuid = i_uid_into_mnt(mnt_userns, inode);
-	if (uid_eq(current_fsuid(), kuid) && uid_eq(uid, kuid))
+	if (uid_eq(current_fsuid(), kuid) && uid_eq(uid, inode->i_uid))
 		return true;
 	if (capable_wrt_inode_uidgid(mnt_userns, inode, CAP_CHOWN))
 		return true;

@@ -62,7 +62,7 @@ static bool chgrp_ok(struct user_namespace *mnt_userns,
 {
 	kgid_t kgid = i_gid_into_mnt(mnt_userns, inode);
 	if (uid_eq(current_fsuid(), i_uid_into_mnt(mnt_userns, inode)) &&
-	    (in_group_p(gid) || gid_eq(gid, kgid)))
+	    (in_group_p(gid) || gid_eq(gid, inode->i_gid)))
 		return true;
 	if (capable_wrt_inode_uidgid(mnt_userns, inode, CAP_CHOWN))
 		return true;

base-commit: 8bb7eca972ad531c9b149c0a51ab43a417385813

Signature · 1 line

2.30.2

From: Christian Brauner <[email protected]>

When calling setattr_prepare() to determine the validity of the attributes the ia_{g,u}id fields contain the value that will be written to inode->i_{g,u}id. When the {g,u}id attribute of the file isn't altered and the caller's fs{g,u}id matches the current {g,u}id attribute the attribute change is allowed.

The value in ia_{g,u}id does already account for idmapped mounts and will have taken the relevant idmapping into account. So in order to verify that the {g,u}id attribute isn't changed we simple need to compare the ia_{g,u}id value against the inode's i_{g,u}id value.

This only has any meaning for idmapped mounts as idmapping helpers are idempotent without them. And for idmapped mounts this really only has a meaning when circular idmappings are used, i.e. mappings where e.g. id 1000 is mapped to id 1001 and id 1001 is mapped to id 1000. Such ciruclar mappings can e.g. be useful when sharing the same home directory between multiple users at the same time.

As an example consider a directory with two files: /source/file1 owned by {g,u}id 1000 and /source/file2 owned by {g,u}id 1001. Assume we create an idmapped mount at /target with an idmapping that maps files owned by {g,u}id 1000 to being owned by {g,u}id 1001 and files owned by {g,u}id 1001 to being owned by {g,u}id 1000. In effect, the idmapped mount at /target switches the ownership of /source/file1 and source/file2, i.e. /target/file1 will be owned by {g,u}id 1001 and /target/file2 will be owned by {g,u}id 1000.

This means that a user with fs{g,u}id 1000 must be allowed to setattr /target/file2 from {g,u}id 1000 to {g,u}id 1000. Similar, a user with fs{g,u}id 1001 must be allowed to setattr /target/file1 from {g,u}id 1001 to {g,u}id 1001. Conversely, a user with fs{g,u}id 1000 must fail to setattr /target/file1 from {g,u}id 1001 to {g,u}id 1000. And a user with fs{g,u}id 1001 must fail to setattr /target/file2 from {g,u}id 1000 to {g,u}id 1000. Both cases must fail with EPERM for non-capable callers.

Before this patch we could end up denying legitimate attribute changes and allowing invalid attribute changes when circular mappings are used. To even get into this situation the caller must've been privileged both to create that mapping and to create that idmapped mount.

This hasn't been seen in the wild anywhere but came up when expanding the testsuite during work on a series of hardening patches. All idmapped fstests pass without any regressions and we add new tests to verify the behavior of circular mappings.

Fixes 2f221d6f7b88 ("attr: handle idmapped mounts")

Cc Seth Forshee <[email protected]>

Cc Christoph Hellwig <[email protected]>

Cc Al Viro <[email protected]>

Cc [email protected]

CC [email protected]

Signed-off-by Christian Brauner <[email protected]>