This came in yesterday, and I have verified our regression tests
were missing this and it can cause an oops. Please apply.

There is a an off-by-one comparision on sig against MAXMAPPED_SIG
that can lead to a read outside the sig_map array if sig
is MAXMAPPED_SIG. Fix this.

Verified that the check is an out of bounds case that can cause an oops.

Revised: add comparison fix to second case
Fixes: cd1dbf76b23d ("apparmor: add the ability to mediate signals")
Signed-off-by: Colin Ian King <[email protected]>
Signed-off-by: John Johansen <[email protected]>
---
 security/apparmor/ipc.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/security/apparmor/ipc.c b/security/apparmor/ipc.c
index 66fb9ede9447..7ca0032e7ba9 100644
--- a/security/apparmor/ipc.c
+++ b/security/apparmor/ipc.c
@@ -128,7 +128,7 @@ static inline int map_signal_num(int sig)
                return SIGUNKNOWN;
        else if (sig >= SIGRTMIN)
                return sig - SIGRTMIN + 128;    /* rt sigs mapped to 128 */
-       else if (sig <= MAXMAPPED_SIG)
+       else if (sig < MAXMAPPED_SIG)
                return sig_map[sig];
        return SIGUNKNOWN;
 }
@@ -163,7 +163,7 @@ static void audit_signal_cb(struct audit_buffer *ab, void 
*va)
                        audit_signal_mask(ab, aad(sa)->denied);
                }
        }
-       if (aad(sa)->signal <= MAXMAPPED_SIG)
+       if (aad(sa)->signal < MAXMAPPED_SIG)
                audit_log_format(ab, " signal=%s", sig_names[aad(sa)->signal]);
        else
                audit_log_format(ab, " signal=rtmin+%d",
-- 
2.11.0

Reply via email to