]> Git Repo - linux.git/commitdiff
kprobes/x86: Clear TF bit in fault on single-stepping
authorMasami Hiramatsu <[email protected]>
Sat, 11 Jun 2016 14:06:53 +0000 (23:06 +0900)
committerIngo Molnar <[email protected]>
Tue, 14 Jun 2016 10:00:54 +0000 (12:00 +0200)
Fix kprobe_fault_handler() to clear the TF (trap flag) bit of
the flags register in the case of a fault fixup on single-stepping.

If we put a kprobe on the instruction which caused a
page fault (e.g. actual mov instructions in copy_user_*),
that fault happens on the single-stepping buffer. In this
case, kprobes resets running instance so that the CPU can
retry execution on the original ip address.

However, current code forgets to reset the TF bit. Since this
fault happens with TF bit set for enabling single-stepping,
when it retries, it causes a debug exception and kprobes
can not handle it because it already reset itself.

On the most of x86-64 platform, it can be easily reproduced
by using kprobe tracer. E.g.

  # cd /sys/kernel/debug/tracing
  # echo p copy_user_enhanced_fast_string+5 > kprobe_events
  # echo 1 > events/kprobes/enable

And you'll see a kernel panic on do_debug(), since the debug
trap is not handled by kprobes.

To fix this problem, we just need to clear the TF bit when
resetting running kprobe.

Signed-off-by: Masami Hiramatsu <[email protected]>
Reviewed-by: Ananth N Mavinakayanahalli <[email protected]>
Acked-by: Steven Rostedt <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Brian Gerst <[email protected]>
Cc: Denys Vlasenko <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Stephane Eranian <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Vince Weaver <[email protected]>
Cc: [email protected]
Cc: [email protected] # All the way back to ancient kernels
Link: http://lkml.kernel.org/r/20160611140648.25885.37482.stgit@devbox
[ Updated the comments. ]
Signed-off-by: Ingo Molnar <[email protected]>
arch/x86/kernel/kprobes/core.c

index 38cf7a7412503f513bb1aae1c1acc5b76e7d2e1d..7847e5c0e0b5d0ce9af2cd47274b52bab7f2597a 100644 (file)
@@ -961,7 +961,19 @@ int kprobe_fault_handler(struct pt_regs *regs, int trapnr)
                 * normal page fault.
                 */
                regs->ip = (unsigned long)cur->addr;
+               /*
+                * Trap flag (TF) has been set here because this fault
+                * happened where the single stepping will be done.
+                * So clear it by resetting the current kprobe:
+                */
+               regs->flags &= ~X86_EFLAGS_TF;
+
+               /*
+                * If the TF flag was set before the kprobe hit,
+                * don't touch it:
+                */
                regs->flags |= kcb->kprobe_old_flags;
+
                if (kcb->kprobe_status == KPROBE_REENTER)
                        restore_previous_kprobe(kcb);
                else
This page took 0.062659 seconds and 4 git commands to generate.