On Thu, Nov 30, 2017 at 1:35 PM, Andrey Ryabinin <aryabi...@virtuozzo.com> wrote: > Stackdepot doesn't work well with CONFIG_UNWINDER_GUESS=y. > The 'guess' unwinder generate awfully large and inaccurate stacktraces, > thus stackdepot can't deduplicate stacktraces because they all look like > unique. Eventually stackdepot reaches its capacity limit: > > WARNING: CPU: 0 PID: 545 at lib/stackdepot.c:119 depot_save_stack+0x28e/0x550 > Call Trace: > ? kasan_kmalloc+0x144/0x160 > ? depot_save_stack+0x1f5/0x550 > ? do_raw_spin_unlock+0xda/0xf0 > ? preempt_count_sub+0x13/0xc0 > > <...90 lines...> > > ? do_raw_spin_unlock+0xda/0xf0 > > Add STACKDEPOT=n dependency to UNWINDER_GUESS to avoid the problem. > > Signed-off-by: Andrey Ryabinin <aryabi...@virtuozzo.com> > Reported-by: kernel test robot <xiaolong...@intel.com> > Reported-by: Fengguang Wu <fengguang...@intel.com> > --- > arch/x86/Kconfig.debug | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/arch/x86/Kconfig.debug b/arch/x86/Kconfig.debug > index 6293a8768a91..672441c008c7 100644 > --- a/arch/x86/Kconfig.debug > +++ b/arch/x86/Kconfig.debug > @@ -400,6 +400,7 @@ config UNWINDER_FRAME_POINTER > config UNWINDER_GUESS > bool "Guess unwinder" > depends on EXPERT > + depends on !STACKDEPOT > ---help--- > This option enables the "guess" unwinder for unwinding kernel stack > traces. It scans the stack and reports every kernel text address it > -- > 2.13.6 >
Acked-by: Dmitry Vyukov <dvyu...@google.com>
