[PATCH] smb: client: validate absolute native symlink targets before NT fixups

[PATCH] smb: client: validate absolute native symlink targets before NT fixups

[Jérémy Jean]

With symlinkroot unset, an absolute target is copied without conversion
to an NT drive path. Later code still assumes an NT prefix is present
when modifying the target and calculating the print name length.

For "/ab", this causes two failures: sym[5] and path[5] are written
past their allocations, and plen -= 2 * poff subtracts an assumed
8-byte prefix from a 6-byte UTF-16 target, wrapping u16 plen to 65534.
That underflow causes another overflow: memcpy() copies 65534 bytes
into a 24-byte buffer. A user with write access to a mounted share
can trigger these bugs with default settings.

Validate the NT drive prefix, including an ASCII drive letter, before
accessing fixed offsets or subtracting the prefix length.

Fixes: 3363da82e02f ("smb: client: fix native SMB symlink traversal")
Assisted-by: Codex:gpt-5
Signed-off-by: Jérémy Jean <Jeremy.Jean@oss.cyber.gouv.fr>
---
 fs/smb/client/reparse.c | 24 +++++++++++++++++-------
 1 file changed, 17 insertions(+), 7 deletions(-)

diff --git a/fs/smb/client/reparse.c b/fs/smb/client/reparse.c
index 5cc5b04..acd5f51 100644
--- a/fs/smb/client/reparse.c
+++ b/fs/smb/client/reparse.c
@@ -3,6 +3,7 @@
  * Copyright (c) 2024 Paulo Alcantara <pc@manguebit.com>
  */
 
+#include <linux/ctype.h>
 #include <linux/fs.h>
 #include <linux/stat.h>
 #include <linux/slab.h>
@@ -159,15 +160,24 @@ static int create_native_symlink(const unsigned int xid, struct inode *inode,
 		convert_delimiter(sym, sep);
 
 	/*
-	 * For absolute NT symlinks it is required to pass also leading
-	 * backslash and to not mangle NT object prefix "\\??\\" and not to
-	 * mangle colon in drive letter. But cifs_convert_path_to_utf16()
-	 * removes leading backslash and replaces '?' and ':'. So temporary
-	 * mask these characters in NT object prefix by '_' and then change
-	 * them back.
+	 * Absolute NT symlinks must retain the leading backslash, "\\??\\"
+	 * prefix and drive-letter colon. cifs_convert_path_to_utf16() strips
+	 * the leading backslash and maps '?' and ':', so temporarily mask
+	 * these characters with '_' and restore them after conversion.
+	 *
+	 * When symlinkroot is unset, sym comes directly from the caller.
+	 * Validate the complete "\\??\\X:" prefix before using fixed offsets
+	 * or subtracting the NT prefix length below. Require an ASCII drive
+	 * letter so the prefix occupies six characters in UTF-16 too.
 	 */
-	if (!(sbflags & CIFS_MOUNT_POSIX_PATHS) && symname[0] == '/')
+	if (!(sbflags & CIFS_MOUNT_POSIX_PATHS) && symname[0] == '/') {
+		if (!strstarts(sym, "\\??\\") || !isascii(sym[4]) ||
+		    !isalpha(sym[4]) || sym[5] != ':') {
+			rc = -EINVAL;
+			goto out;
+		}
 		sym[0] = sym[1] = sym[2] = sym[5] = '_';
+	}
 
 	/*
 	 * On a POSIX paths mount the symlink target is stored verbatim, so
-- 
2.47.3

Re: [PATCH] smb: client: validate absolute native symlink targets before NT fixups

[Namjae Jeon]

On Fri, Sep 11, 2026 at 3:41 AM Jérémy Jean
<Jeremy.Jean@oss.cyber.gouv.fr> wrote:
>
> With symlinkroot unset, an absolute target is copied without conversion
> to an NT drive path. Later code still assumes an NT prefix is present
> when modifying the target and calculating the print name length.
>
> For "/ab", this causes two failures: sym[5] and path[5] are written
> past their allocations, and plen -= 2 * poff subtracts an assumed
> 8-byte prefix from a 6-byte UTF-16 target, wrapping u16 plen to 65534.
> That underflow causes another overflow: memcpy() copies 65534 bytes
> into a 24-byte buffer. A user with write access to a mounted share
> can trigger these bugs with default settings.
>
> Validate the NT drive prefix, including an ASCII drive letter, before
> accessing fixed offsets or subtracting the prefix length.
>
> Fixes: 3363da82e02f ("smb: client: fix native SMB symlink traversal")
> Assisted-by: Codex:gpt-5
> Signed-off-by: Jérémy Jean <Jeremy.Jean@oss.cyber.gouv.fr>
Reviewed-by: Namjae Jeon <linkinjeon@kernel.org>
Thanks.

Re: [PATCH] smb: client: validate absolute native symlink targets before NT fixups

[Paulo Alcantara]

Jérémy Jean <Jeremy.Jean@oss.cyber.gouv.fr> writes:

> With symlinkroot unset, an absolute target is copied without conversion
> to an NT drive path. Later code still assumes an NT prefix is present
> when modifying the target and calculating the print name length.
>
> For "/ab", this causes two failures: sym[5] and path[5] are written
> past their allocations, and plen -= 2 * poff subtracts an assumed
> 8-byte prefix from a 6-byte UTF-16 target, wrapping u16 plen to 65534.
> That underflow causes another overflow: memcpy() copies 65534 bytes
> into a 24-byte buffer. A user with write access to a mounted share
> can trigger these bugs with default settings.
>
> Validate the NT drive prefix, including an ASCII drive letter, before
> accessing fixed offsets or subtracting the prefix length.
> ...

Applied.