Thread (26 messages) 26 messages, 3 authors, 2020-07-09

Re: [PATCH v6 1/7] net/scm: Regularize compat handling of scm_detach_fds()

From: Kees Cook <hidden>
Date: 2020-07-08 23:46:16
Also in: linux-fsdevel, linux-kselftest, lkml, netdev

On Tue, Jul 07, 2020 at 01:41:03PM +0200, Christian Brauner wrote:
On Mon, Jul 06, 2020 at 01:17:14PM -0700, Kees Cook wrote:
quoted
Duplicate the cleanups from commit 2618d530dd8b ("net/scm: cleanup
scm_detach_fds") into the compat code.

Move the check added in commit 1f466e1f15cf ("net: cleanly handle kernel
vs user buffers for ->msg_control") to before the compat call, even
though it should be impossible for an in-kernel call to also be compat.

Correct the int "flags" argument to unsigned int to match fd_install()
and similar APIs.

Regularize any remaining differences, including a whitespace issue,
a checkpatch warning, and add the check from commit 6900317f5eff ("net,
scm: fix PaX detected msg_controllen overflow in scm_detach_fds") which
fixed an overflow unique to 64-bit. To avoid confusion when comparing
the compat handler to the native handler, just include the same check
in the compat handler.

Fixes: 48a87cc26c13 ("net: netprio: fd passed in SCM_RIGHTS datagram not set correctly")
Fixes: d84295067fc7 ("net: net_cls: fd passed in SCM_RIGHTS datagram not set correctly")
Signed-off-by: Kees Cook <redacted>
---
Thanks. Just a comment below.
Acked-by: Christian Brauner <redacted>
Thanks!
quoted
 include/net/scm.h |  1 +
 net/compat.c      | 55 +++++++++++++++++++++--------------------------
 net/core/scm.c    | 18 ++++++++--------
 3 files changed, 35 insertions(+), 39 deletions(-)
diff --git a/include/net/scm.h b/include/net/scm.h
index 1ce365f4c256..581a94d6c613 100644
--- a/include/net/scm.h
+++ b/include/net/scm.h
@@ -37,6 +37,7 @@ struct scm_cookie {
 #endif
 };
 
+int __scm_install_fd(struct file *file, int __user *ufd, unsigned int o_flags);
 void scm_detach_fds(struct msghdr *msg, struct scm_cookie *scm);
 void scm_detach_fds_compat(struct msghdr *msg, struct scm_cookie *scm);
 int __scm_send(struct socket *sock, struct msghdr *msg, struct scm_cookie *scm);
diff --git a/net/compat.c b/net/compat.c
index 5e3041a2c37d..27d477fdcaa0 100644
--- a/net/compat.c
+++ b/net/compat.c
@@ -281,39 +281,31 @@ int put_cmsg_compat(struct msghdr *kmsg, int level, int type, int len, void *dat
 	return 0;
 }
 
-void scm_detach_fds_compat(struct msghdr *kmsg, struct scm_cookie *scm)
+static int scm_max_fds_compat(struct msghdr *msg)
 {
-	struct compat_cmsghdr __user *cm = (struct compat_cmsghdr __user *) kmsg->msg_control;
-	int fdmax = (kmsg->msg_controllen - sizeof(struct compat_cmsghdr)) / sizeof(int);
-	int fdnum = scm->fp->count;
-	struct file **fp = scm->fp->fp;
-	int __user *cmfptr;
-	int err = 0, i;
+	if (msg->msg_controllen <= sizeof(struct compat_cmsghdr))
+		return 0;
+	return (msg->msg_controllen - sizeof(struct compat_cmsghdr)) / sizeof(int);
+}
 
-	if (fdnum < fdmax)
-		fdmax = fdnum;
+void scm_detach_fds_compat(struct msghdr *msg, struct scm_cookie *scm)
+{
+	struct compat_cmsghdr __user *cm =
+		(struct compat_cmsghdr __user *)msg->msg_control;
+	unsigned int o_flags = (msg->msg_flags & MSG_CMSG_CLOEXEC) ? O_CLOEXEC : 0;
+	int fdmax = min_t(int, scm_max_fds_compat(msg), scm->fp->count);
Just a note that SCM_RIGHTS fd-sending is limited to 253 (SCM_MAX_FD)
fds so min_t should never ouput > SCM_MAX_FD here afaict.
quoted
+	int __user *cmsg_data = CMSG_USER_DATA(cm);
+	int err = 0, i;
 
-	for (i = 0, cmfptr = (int __user *) CMSG_COMPAT_DATA(cm); i < fdmax; i++, cmfptr++) {
-		int new_fd;
-		err = security_file_receive(fp[i]);
+	for (i = 0; i < fdmax; i++) {
+		err = __scm_install_fd(scm->fp->fp[i], cmsg_data + i, o_flags);
 		if (err)
 			break;
-		err = get_unused_fd_flags(MSG_CMSG_CLOEXEC & kmsg->msg_flags
-					  ? O_CLOEXEC : 0);
-		if (err < 0)
-			break;
-		new_fd = err;
-		err = put_user(new_fd, cmfptr);
-		if (err) {
-			put_unused_fd(new_fd);
-			break;
-		}
-		/* Bump the usage count and install the file. */
-		fd_install(new_fd, get_file(fp[i]));
 	}
 
 	if (i > 0) {
 		int cmlen = CMSG_COMPAT_LEN(i * sizeof(int));
+
 		err = put_user(SOL_SOCKET, &cm->cmsg_level);
 		if (!err)
 			err = put_user(SCM_RIGHTS, &cm->cmsg_type);
@@ -321,16 +313,19 @@ void scm_detach_fds_compat(struct msghdr *kmsg, struct scm_cookie *scm)
 			err = put_user(cmlen, &cm->cmsg_len);
 		if (!err) {
 			cmlen = CMSG_COMPAT_SPACE(i * sizeof(int));
-			kmsg->msg_control += cmlen;
-			kmsg->msg_controllen -= cmlen;
+			if (msg->msg_controllen < cmlen)
+				cmlen = msg->msg_controllen;
+			msg->msg_control += cmlen;
+			msg->msg_controllen -= cmlen;
 		}
 	}
-	if (i < fdnum)
-		kmsg->msg_flags |= MSG_CTRUNC;
+
+	if (i < scm->fp->count || (scm->fp->count && fdmax <= 0))
I think fdmax can't be < 0 after your changes? scm_max_fds() guarantees
that fdmax is always >= 0 and min_t() guarantees that fdmax <= scm->fp->count.
So the check should technically be :)
You left our your suggestion! :) But, I think you mean "== 0" ?

The check actually comes from the refactoring from commit 2618d530dd8b
("net/scm: cleanup scm_detach_fds") which I mostly copy/pasted into
compat. However, fdmax is an int, and scm->fp->count is signed so it's
possible fdmax is < 0 (but I don't think count can actually ever be <
0), but I don't want to refactor all the types just to fix this boundary
condition. :)

-- 
Kees Cook
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help