fs-util: tweak how openat_report_new() operates when O_CREAT is used on a dangling...

One of the big mistakes of Linux is that when you create a file with
open() and O_CREAT and the file already exists as dangling symlink that
the symlink will be followed and the file created that it points to.
This has resulted in many vulnerabilities, and triggered the creation of
the O_MOFOLLOW flag, addressing the problem.

O_NOFOLLOW is less than ideal in many ways, but in particular one: when
actually creating a file it makes sense to set, because it is a problem
to follow final symlinks in that case. But if the file is already
existing, it actually does make sense to follow the symlinks. With
openat_report_new() we distinguish these two cases anyway (the whole
function exists only to distinguish the create and the exists-already
case after all), hence let's do something about this: let's simply never
create files "through symlinks".

This can be implemented very easily: just pass O_NOFOLLOW to the 2nd
openat() call, where we actually create files.

And then basically remove 0dd82dab91eaac5e7b17bd5e9a1e07c6d2b78dca
again, because we don't need to care anymore, we already will see ELOOP
when we touch a symlink.

Note that this change means that openat_report_new() will thus start to
deviate from plain openat() behaviour in this one small detail: when
actually creating files we will *never* follow the symlink. That should
be a systematic improvement of security.

Fixes: #34088

diff --git a/src/basic/fs-util.c b/src/basic/fs-util.c
index 1d0533e4f07e91f490171326376fea9863a84bbf..9397c4b38497b8c888abffe8c278af8fc918f718 100644 (file)
--- a/src/basic/fs-util.c
+++ b/src/basic/fs-util.c
@@ -1082,17 +1082,20 @@ int openat_report_new(int dirfd, const char *pathname, int flags, mode_t mode, b
 
         /* Just like openat(), but adds one thing: optionally returns whether we created the file anew or if
          * it already existed before. This is only relevant if O_CREAT is set without O_EXCL, and thus will
-         * shortcut to openat() otherwise */
-
-        if (!ret_newly_created)
-                return RET_NERRNO(openat(dirfd, pathname, flags, mode));
+         * shortcut to openat() otherwise.
+         *
+         * Note that this routine is a bit more strict with symlinks than regular openat() is. If O_NOFOLLOW
+         * is not specified, then we'll follow the symlink when opening an existing file but we will *not*
+         * follow it when creating a new one (because that's a terrible UNIX misfeature and generally a
+         * security hole). */
 
         if (!FLAGS_SET(flags, O_CREAT) || FLAGS_SET(flags, O_EXCL)) {
                 fd = openat(dirfd, pathname, flags, mode);
                 if (fd < 0)
                         return -errno;
 
-                *ret_newly_created = FLAGS_SET(flags, O_CREAT);
+                if (ret_newly_created)
+                        *ret_newly_created = FLAGS_SET(flags, O_CREAT);
                 return fd;
         }
 
@@ -1100,42 +1103,25 @@ int openat_report_new(int dirfd, const char *pathname, int flags, mode_t mode, b
                 /* First, attempt to open without O_CREAT/O_EXCL, i.e. open existing file */
                 fd = openat(dirfd, pathname, flags & ~(O_CREAT | O_EXCL), mode);
                 if (fd >= 0) {
-                        *ret_newly_created = false;
+                        if (ret_newly_created)
+                                *ret_newly_created = false;
                         return fd;
                 }
                 if (errno != ENOENT)
                         return -errno;
 
-                /* So the file didn't exist yet, hence create it with O_CREAT/O_EXCL. */
-                fd = openat(dirfd, pathname, flags | O_CREAT | O_EXCL, mode);
+                /* So the file didn't exist yet, hence create it with O_CREAT/O_EXCL/O_NOFOLLOW. */
+                fd = openat(dirfd, pathname, flags | O_CREAT | O_EXCL | O_NOFOLLOW, mode);
                 if (fd >= 0) {
-                        *ret_newly_created = true;
+                        if (ret_newly_created)
+                                *ret_newly_created = true;
                         return fd;
                 }
                 if (errno != EEXIST)
                         return -errno;
 
-                /* Hmm, so now we got EEXIST? This can indicate two things. First, if the path points to a
-                 * dangling symlink, the first openat() will fail with ENOENT because the symlink is resolved
-                 * and the second openat() will fail with EEXIST because symlinks are not followed when
-                 * O_CREAT|O_EXCL is specified. Let's check for this explicitly and fall back to opening with
-                 * just O_CREAT and assume we're the ones that created the file. */
-
-                struct stat st;
-                if (fstatat(dirfd, pathname, &st, AT_SYMLINK_NOFOLLOW) < 0)
-                        return -errno;
-
-                if (S_ISLNK(st.st_mode)) {
-                        fd = openat(dirfd, pathname, flags | O_CREAT, mode);
-                        if (fd < 0)
-                                return -errno;
-
-                        *ret_newly_created = true;
-                        return fd;
-                }
-
-                /* If we're not operating on a symlink, someone might have created the file between the first
-                 * and second call to openat(). Let's try again but with a limit so we don't spin forever. */
+                /* Hmm, so now we got EEXIST? Then someone might have created the file between the first and
+                 * second call to openat(). Let's try again but with a limit so we don't spin forever. */
 
                 if (--attempts == 0) /* Give up eventually, somebody is playing with us */
                         return -EEXIST;

diff --git a/src/test/test-fs-util.c b/src/test/test-fs-util.c
index 3da3caf4ab9b4029d824647d1b565c32e0dea762..55202240d3fc058514d691152b7f5eab6338471e 100644 (file)
--- a/src/test/test-fs-util.c
+++ b/src/test/test-fs-util.c
@@ -670,6 +670,9 @@ TEST(openat_report_new) {
 
         ASSERT_OK_ERRNO(symlinkat("target", tfd, "link"));
         fd = openat_report_new(tfd, "link", O_RDWR|O_CREAT, 0666, &b);
+        ASSERT_ERROR(fd, EEXIST);
+
+        fd = openat_report_new(tfd, "target", O_RDWR|O_CREAT, 0666, &b);
         ASSERT_OK(fd);
         fd = safe_close(fd);
         ASSERT_TRUE(b);

diff --git a/src/test/test-id128.c b/src/test/test-id128.c
index a6ed640bd6e53401c0f46e5da67497cbc008394e..eb3bfec2bb3294032d2631128f6206416d86d3cb 100644 (file)
--- a/src/test/test-id128.c
+++ b/src/test/test-id128.c
@@ -286,9 +286,8 @@ TEST(id128_at) {
         ASSERT_OK_ERRNO(unlinkat(tfd, "etc/machine-id", 0));
         ASSERT_OK(id128_write_at(tfd, "etc2/machine-id", ID128_FORMAT_PLAIN, id));
         ASSERT_OK_ERRNO(unlinkat(tfd, "etc/machine-id", 0));
-        ASSERT_OK(id128_write_at(tfd, "etc/hoge-id", ID128_FORMAT_PLAIN, id));
-        ASSERT_OK_ERRNO(unlinkat(tfd, "etc/machine-id", 0));
-        ASSERT_OK(id128_write_at(tfd, "etc2/hoge-id", ID128_FORMAT_PLAIN, id));
+        ASSERT_ERROR(id128_write_at(tfd, "etc/hoge-id", ID128_FORMAT_PLAIN, id), EEXIST);
+        ASSERT_OK(id128_write_at(tfd, "etc2/machine-id", ID128_FORMAT_PLAIN, id));
 
         /* id128_read_at() */
         i = SD_ID128_NULL; /* Not necessary in real code, but for testing that the id is really assigned. */