git: 2bd9826995ca - main - vfs: Permit unix sockets to be opened with O_PATH

[Mark Johnston]

The branch main has been updated by markj:

URL: https://cgit.FreeBSD.org/src/commit/?id=2bd9826995ca6b23f8b088cfa035c0ad1c578ac3

commit 2bd9826995ca6b23f8b088cfa035c0ad1c578ac3
Author:     Mark Johnston <markj at FreeBSD.org>
AuthorDate: 2021-09-17 16:34:21 +0000
Commit:     Mark Johnston <markj at FreeBSD.org>
CommitDate: 2021-09-17 18:19:06 +0000

    vfs: Permit unix sockets to be opened with O_PATH
    
    As with FIFOs, a path descriptor for a unix socket cannot be used with
    kevent().
    
    In principle connectat(2) and bindat(2) could be modified to support an
    AT_EMPTY_PATH-like mode which operates on the socket referenced by an
    O_PATH fd referencing a unix socket.  That would eliminate the path
    length limit imposed by sockaddr_un.
    
    Update O_PATH tests.
    
    Reviewed by:    kib
    MFC after:      1 month
    Sponsored by:   The FreeBSD Foundation
    Differential Revision:  https://reviews.freebsd.org/D31970
---
 sys/kern/vfs_vnops.c       | 11 ++++-------
 tests/sys/file/path_test.c | 32 ++++++++++++++++++++++++++++----
 2 files changed, 32 insertions(+), 11 deletions(-)

diff --git a/sys/kern/vfs_vnops.c b/sys/kern/vfs_vnops.c
index 93d5a9e6b127..d4229160f11e 100644
--- a/sys/kern/vfs_vnops.c
+++ b/sys/kern/vfs_vnops.c
@@ -399,13 +399,13 @@ vn_open_vnode(struct vnode *vp, int fmode, struct ucred *cred,
 		if ((fmode & O_PATH) == 0 || (fmode & FEXEC) != 0)
 			return (EMLINK);
 	}
-	if (vp->v_type == VSOCK)
-		return (EOPNOTSUPP);
 	if (vp->v_type != VDIR && fmode & O_DIRECTORY)
 		return (ENOTDIR);
 
 	accmode = 0;
 	if ((fmode & O_PATH) == 0) {
+		if (vp->v_type == VSOCK)
+			return (EOPNOTSUPP);
 		if ((fmode & (FWRITE | O_TRUNC)) != 0) {
 			if (vp->v_type == VDIR)
 				return (EISDIR);
@@ -437,11 +437,8 @@ vn_open_vnode(struct vnode *vp, int fmode, struct ucred *cred,
 			return (error);
 	}
 	if ((fmode & O_PATH) != 0) {
-		if (vp->v_type == VFIFO)
-			error = EPIPE;
-		else
-			error = VOP_ACCESS(vp, VREAD, cred, td);
-		if (error == 0)
+		if (vp->v_type != VFIFO && vp->v_type != VSOCK &&
+		    VOP_ACCESS(vp, VREAD, cred, td) == 0)
 			fp->f_flag |= FKQALLOWED;
 		return (0);
 	}
diff --git a/tests/sys/file/path_test.c b/tests/sys/file/path_test.c
index ad88c691a914..a39862cc78d6 100644
--- a/tests/sys/file/path_test.c
+++ b/tests/sys/file/path_test.c
@@ -845,13 +845,15 @@ ATF_TC_BODY(path_rights, tc)
 	CHECKED_CLOSE(sd[1]);
 }
 
-/* Verify that a local socket can't be opened with O_PATH. */
+/* Verify that a local socket can be opened with O_PATH. */
 ATF_TC_WITHOUT_HEAD(path_unix);
 ATF_TC_BODY(path_unix, tc)
 {
-	char path[PATH_MAX];
+	char buf[BUFSIZ], path[PATH_MAX];
+	struct kevent ev;
 	struct sockaddr_un sun;
-	int pathfd, sd;
+	struct stat sb;
+	int kq, pathfd, sd;
 
 	snprintf(path, sizeof(path), "path_unix.XXXXXX");
 	ATF_REQUIRE_MSG(mktemp(path) == path, FMT_ERR("mktemp"));
@@ -866,9 +868,31 @@ ATF_TC_BODY(path_unix, tc)
 	    FMT_ERR("bind"));
 
 	pathfd = open(path, O_PATH);
-	ATF_REQUIRE_ERRNO(EOPNOTSUPP, pathfd < 0);
+	ATF_REQUIRE_MSG(pathfd >= 0, FMT_ERR("open"));
+
+	ATF_REQUIRE_MSG(fstatat(pathfd, "", &sb, AT_EMPTY_PATH) == 0,
+	    FMT_ERR("fstatat"));
+	ATF_REQUIRE_MSG(sb.st_mode & S_IFSOCK, "socket mode %#x", sb.st_mode);
+	ATF_REQUIRE_MSG(sb.st_ino != 0, "socket has inode number 0");
+
+	memset(buf, 0, sizeof(buf));
+	ATF_REQUIRE_ERRNO(EBADF, write(pathfd, buf, sizeof(buf)));
+	ATF_REQUIRE_ERRNO(EBADF, read(pathfd, buf, sizeof(buf)));
+
+	/* kevent() is disallowed with sockets. */
+	kq = kqueue();
+	ATF_REQUIRE_MSG(kq >= 0, FMT_ERR("kqueue"));
+	EV_SET(&ev, pathfd, EVFILT_READ, EV_ADD | EV_ENABLE, 0, 0, 0);
+	ATF_REQUIRE_ERRNO(EBADF, kevent(kq, &ev, 1, NULL, 0, NULL) == -1);
+
+	/* Should not be able to open a socket without O_PATH. */
+	ATF_REQUIRE_ERRNO(EOPNOTSUPP, openat(pathfd, "", O_EMPTY_PATH) == -1);
+
+	ATF_REQUIRE_MSG(funlinkat(AT_FDCWD, path, pathfd, 0) == 0,
+	    FMT_ERR("funlinkat"));
 
 	CHECKED_CLOSE(sd);
+	CHECKED_CLOSE(pathfd);
 }
 
 ATF_TP_ADD_TCS(tp)