x86_32, entry: Do syscall exit work on badsys (CVE-2014-4508)
authorAndy Lutomirski <[email protected]>
Mon, 23 Jun 2014 21:22:15 +0000 (14:22 -0700)
committerH. Peter Anvin <[email protected]>
Mon, 23 Jun 2014 21:59:26 +0000 (14:59 -0700)
commit554086d85e71f30abe46fc014fea31929a7c6a8a
tree0bc7acf1c743afa0bf086920d66349e5b8e7d04d
parentdda1e95cee38b416b23f751cac65421d781e3c10
x86_32, entry: Do syscall exit work on badsys (CVE-2014-4508)

The bad syscall nr paths are their own incomprehensible route
through the entry control flow.  Rearrange them to work just like
syscalls that return -ENOSYS.

This fixes an OOPS in the audit code when fast-path auditing is
enabled and sysenter gets a bad syscall nr (CVE-2014-4508).

This has probably been broken since Linux 2.6.27:
af0575bba0 i386 syscall audit fast-path

Cc: [email protected]
Cc: Roland McGrath <[email protected]>
Reported-by: Toralf Förster <[email protected]>
Signed-off-by: Andy Lutomirski <[email protected]>
Link: http://lkml.kernel.org/r/e09c499eade6fc321266dd6b54da7beb28d6991c.1403558229.git.luto@amacapital.net
Signed-off-by: H. Peter Anvin <[email protected]>
arch/x86/kernel/entry_32.S