Kprobe smoke test lockdep warning

Peter Zijlstra · Ingo Molnar · commit d54191b85e29 · 2008-07-15T11:18:03.000+02:00
On Mon, 2008-04-21 at 18:54 -0400, Masami Hiramatsu wrote:
&gt; Thank you for reporting.
&gt;
&gt; Actually, kprobes tries to fixup thread's flags in post_kprobe_handler
&gt; (which is called from kprobe_exceptions_notify) by
&gt; trace_hardirqs_fixup_flags(pt_regs-&gt;flags). However, even the irq flag
&gt; is set in pt_regs-&gt;flags, true hardirq is still off until returning
&gt; from do_debug. Thus, lockdep assumes that hardirq is off without annotation.
&gt;
&gt; IMHO, one possible solution is that fixing hardirq flags right after
&gt; notify_die in do_debug instead of in post_kprobe_handler.

My reply to BZ 10489:

&gt; [    2.707509] Kprobe smoke test started
&gt; [    2.709300] ------------[ cut here ]------------
&gt; [    2.709420] WARNING: at kernel/lockdep.c:2658 check_flags+0x4d/0x12c()
&gt; [    2.709541] Modules linked in:
&gt; [    2.709588] Pid: 1, comm: swapper Not tainted 2.6.25.jml.057 #1
&gt; [    2.709588]  [&lt;c0126acc&gt;] warn_on_slowpath+0x41/0x51
&gt; [    2.709588]  [&lt;c010bafc&gt;] ? save_stack_trace+0x1d/0x3b
&gt; [    2.709588]  [&lt;c0140a83&gt;] ? save_trace+0x37/0x89
&gt; [    2.709588]  [&lt;c011987d&gt;] ? kernel_map_pages+0x103/0x11c
&gt; [    2.709588]  [&lt;c0109803&gt;] ? native_sched_clock+0xca/0xea
&gt; [    2.709588]  [&lt;c0142958&gt;] ? mark_held_locks+0x41/0x5c
&gt; [    2.709588]  [&lt;c0382580&gt;] ? kprobe_exceptions_notify+0x322/0x3af
&gt; [    2.709588]  [&lt;c0142aff&gt;] ? trace_hardirqs_on+0xf1/0x119
&gt; [    2.709588]  [&lt;c03825b3&gt;] ? kprobe_exceptions_notify+0x355/0x3af
&gt; [    2.709588]  [&lt;c0140823&gt;] check_flags+0x4d/0x12c
&gt; [    2.709588]  [&lt;c0143c9d&gt;] lock_release+0x58/0x195
&gt; [    2.709588]  [&lt;c038347c&gt;] ? __atomic_notifier_call_chain+0x0/0x80
&gt; [    2.709588]  [&lt;c03834d6&gt;] __atomic_notifier_call_chain+0x5a/0x80
&gt; [    2.709588]  [&lt;c0383508&gt;] atomic_notifier_call_chain+0xc/0xe
&gt; [    2.709588]  [&lt;c013b6d4&gt;] notify_die+0x2d/0x2f
&gt; [    2.709588]  [&lt;c038168a&gt;] do_debug+0x67/0xfe
&gt; [    2.709588]  [&lt;c0381287&gt;] debug_stack_correct+0x27/0x30
&gt; [    2.709588]  [&lt;c01564c0&gt;] ? kprobe_target+0x1/0x34
&gt; [    2.709588]  [&lt;c0156572&gt;] ? init_test_probes+0x50/0x186
&gt; [    2.709588]  [&lt;c04fae48&gt;] init_kprobes+0x85/0x8c
&gt; [    2.709588]  [&lt;c04e947b&gt;] kernel_init+0x13d/0x298
&gt; [    2.709588]  [&lt;c04e933e&gt;] ? kernel_init+0x0/0x298
&gt; [    2.709588]  [&lt;c04e933e&gt;] ? kernel_init+0x0/0x298
&gt; [    2.709588]  [&lt;c0105ef7&gt;] kernel_thread_helper+0x7/0x10
&gt; [    2.709588]  =======================
&gt; [    2.709588] ---[ end trace 778e504de7e3b1e3 ]---
&gt; [    2.709588] possible reason: unannotated irqs-off.
&gt; [    2.709588] irq event stamp: 370065
&gt; [    2.709588] hardirqs last  enabled at (370065): [&lt;c0382580&gt;] kprobe_exceptions_notify+0x322/0x3af
&gt; [    2.709588] hardirqs last disabled at (370064): [&lt;c0381bb7&gt;] do_int3+0x1d/0x7d
&gt; [    2.709588] softirqs last  enabled at (370050): [&lt;c012b464&gt;] __do_softirq+0xfa/0x100
&gt; [    2.709588] softirqs last disabled at (370045): [&lt;c0107438&gt;] do_softirq+0x74/0xd9
&gt; [    2.714751] Kprobe smoke test passed successfully

how I love this stuff...

Ok, do_debug() is a trap, this can happen at any time regardless of the
machine's IRQ state. So the first thing we do is fix up the IRQ state.
Then we call this die notifier stuff; and return with messed up IRQ
state... YAY.

So, kprobes fudges it..

  notify_die(DIE_DEBUG)
    kprobe_exceptions_notify()
      post_kprobe_handler()
        modify regs-&gt;flags
        trace_hardirqs_fixup_flags(regs-&gt;flags);  &lt;--- must be it

So what's the use of modifying flags if they're not meant to take effect
at some point.

/me tries to reproduce issue; enable kprobes test thingy &amp;&amp; boot

OK, that reproduces..

So the below makes it work - but I'm not getting this code; at the time
I wrote that stuff I CC'ed each and every kprobe maintainer listed in
the usual places but got no reposonse - can some please explain this
stuff to me?

Are the saved flags only for the TF bit or are they made in full effect
later (and if so, where) ?

Signed-off-by: Peter Zijlstra &lt;a.p.zijlstra@chello.nl&gt;
Acked-by: Masami Hiramatsu &lt;mhiramat@redhat.com&gt;
Signed-off-by: Ingo Molnar &lt;mingo@elte.hu&gt;
diff --git a/arch/x86/kernel/kprobes.c b/arch/x86/kernel/kprobes.c
@@ -860,7 +860,6 @@ static int __kprobes post_kprobe_handler(struct pt_regs *regs)
 
 	resume_execution(cur, regs, kcb);
 	regs->flags |= kcb->kprobe_saved_flags;
-	trace_hardirqs_fixup_flags(regs->flags);
 
 	if ((kcb->kprobe_status != KPROBE_REENTER) && cur->post_handler) {
 		kcb->kprobe_status = KPROBE_HIT_SSDONE;
