[RFC PATCH 2/3] arch/x86: move secure_computing after ptrace

[Will Drewry <wad@chromium.org>]

At present, seccomp modes 1 and 2 may have their
behavior changed by a ptrace()ing task.  The ptracer
cannot change blocked/disallowed system calls, but it can
change allowed system calls to calls that would otherwise
not be allowed by the seccomp policy.

Signed-off-by: Will Drewry <wad@chromium.org>
---
 arch/x86/kernel/ptrace.c |   13 ++++++-------
 1 file changed, 6 insertions(+), 7 deletions(-)

diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c
index 13b1990..ad649a6 100644
--- a/arch/x86/kernel/ptrace.c
+++ b/arch/x86/kernel/ptrace.c
@@ -1479,13 +1479,6 @@ long syscall_trace_enter(struct pt_regs *regs)
 	if (test_thread_flag(TIF_SINGLESTEP))
 		regs->flags |= X86_EFLAGS_TF;
 
-	/* do the secure computing check first */
-	if (secure_computing(regs->orig_ax)) {
-		/* seccomp failures shouldn't expose any additional code. */
-		ret = -1L;
-		goto out;
-	}
-
 	if (unlikely(test_thread_flag(TIF_SYSCALL_EMU)))
 		ret = -1L;
 
@@ -1493,6 +1486,12 @@ long syscall_trace_enter(struct pt_regs *regs)
 	    tracehook_report_syscall_entry(regs))
 		ret = -1L;
 
+	/* check secure computing after userspace can't change the syscall. */
+	if (!ret && secure_computing(regs->orig_ax)) {
+		ret = -1L;
+		goto out;
+	}
+
 	if (unlikely(test_thread_flag(TIF_SYSCALL_TRACEPOINT)))
 		trace_sys_enter(regs, regs->orig_ax);
 
-- 
1.7.9.5