[PATCH RFC 3/4] coredump: mitigate privilege escalation of process coredump

[Wander Lairson Costa <wander@redhat.com>]

A set-uid executable might be a vector to a privilege escalation if the
system configures the coredump file name pattern as a relative
directory destiny. The full description of the vulnerability and
a demonstration of how we can exploit it can be found at [1].

We now check if the core dump pattern is relative. If it is, then we
verify if root owns the current directory and if it does, we deny
writing the core file unless the directory is universally writable.

[1] https://www.openwall.com/lists/oss-security/2021/10/20/2

Signed-off-by: Wander Lairson Costa <wander@redhat.com>
---
 fs/coredump.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/fs/coredump.c b/fs/coredump.c
index 07afb5ddb1c4..74eae7bd144d 100644
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -580,6 +580,7 @@ void do_coredump(const kernel_siginfo_t *siginfo)
 	struct core_name cn;
 	struct mm_struct *mm = current->mm;
 	struct linux_binfmt * binfmt;
+	struct inode *pwd_inode;
 	const struct cred *old_cred;
 	struct cred *cred;
 	int retval = 0;
@@ -625,6 +626,20 @@ void do_coredump(const kernel_siginfo_t *siginfo)
 		need_suid_safe = true;
 	}
 
+	/*
+	 * If we are a set-uid/gid root process and the current directory is
+	 * owned by root but not universally writable, prohibit dumps under
+	 * this path.
+	 *
+	 * Mitigate https://www.openwall.com/lists/oss-security/2021/10/20/2
+	 */
+	pwd_inode = current->fs->pwd.dentry->d_inode;
+	if (current->flags & PF_SUID &&
+	    capable(CAP_SYS_ADMIN) &&
+	    uid_eq(pwd_inode->i_uid, GLOBAL_ROOT_UID) &&
+	    !(pwd_inode->i_mode & 0002))
+		need_suid_safe = true;
+
 	retval = coredump_wait(siginfo->si_signo, &core_state);
 	if (retval < 0)
 		goto fail_creds;
-- 
2.27.0