ARM: Call syscall_trace_exit even when system call skipped

On Tue, Mar 13, 2018 at 4:58 PM, Dmitry V. Levin [off-list ref] wrote:

Hi Kees,

As you probably know, ptracing of processes affected by SECCOMP_RET_TRAP
is broken on ARM since your commit v3.7-rc1-11-gad75b51459ae.

Hi!

Well that's no good. I hadn't seen this problem; the seccomp selftests
don't trip over this.

Could you review the proposed fix, please?

The seccomp selftests still pass with the proposed fix, so that's all
fine by me. :)

P.S. There is a test for this kernel bug in strace test suite,
you might find it useful:
https://github.com/strace/strace/compare/ldv/SECCOMP_RET_TRAP

On Sat, Feb 03, 2018 at 03:21:12PM +0000, Timothy E Baldwin wrote:

quoted

On at least x86 and ARM64, and as documented in the ptrace man page
a skipped system call will still cause a syscall exit ptrace stop.

Previous to this commit 32-bit ARM did not, resulting in strace
being confused when seccomp skips system calls.

This change also impacts programs that use ptrace to skip system calls.

Fixes: ad75b51459ae ("ARM: 7579/1: arch/allow a scno of -1 to not cause a SIGILL")
Signed-off-by: Timothy E Baldwin <redacted>
---
 arch/arm/kernel/entry-common.S | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/arch/arm/kernel/entry-common.S b/arch/arm/kernel/entry-common.S
index 99c908226065..88a65157307d 100644
--- a/arch/arm/kernel/entry-common.S
+++ b/arch/arm/kernel/entry-common.S

@@ -288,16 +288,15 @@ __sys_trace:
      cmp     scno, #-1                       @ skip the syscall?
      bne     2b
      add     sp, sp, #S_OFF                  @ restore stack
-     b       ret_slow_syscall

-__sys_trace_return:
-     str     r0, [sp, #S_R0 + S_OFF]!        @ save returned r0
+__sys_trace_return_nosave:
+     enable_irq_notrace

Why is __sys_trace_return_nosave the correct destination here? The
original handle set up for lr a few lines above is for
__sys_trace_return. It's not clear to me why this change is made?

quoted

      mov     r0, sp
      bl      syscall_trace_exit
      b       ret_slow_syscall

-__sys_trace_return_nosave:
-     enable_irq_notrace
+__sys_trace_return:
+     str     r0, [sp, #S_R0 + S_OFF]!        @ save returned r0
      mov     r0, sp
      bl      syscall_trace_exit
      b       ret_slow_syscall

--
ldv

Otherwise, I think this looks good.

-Kees

-- 
Kees Cook
Pixel Security