Re: [PATCH] PM: Fix crash on graph trace through x86 suspend

[Rafael J. Wysocki]

Well, fix up the Steven's address.

On Thu, Mar 3, 2016 at 2:34 AM, Rafael J. Wysocki <rafael@xxxxxxxxxx> wrote:
> On Thu, Mar 3, 2016 at 1:43 AM, Todd Brandt
> <todd.e.brandt@xxxxxxxxxxxxxxx> wrote:
>> On Thu, 2016-03-03 at 01:19 +0100, Rafael J. Wysocki wrote:
>>> On Thu, Mar 3, 2016 at 1:05 AM, Todd Brandt
>>> <todd.e.brandt@xxxxxxxxxxxxxxx> wrote:
>>> > Pause/unpause graph tracing around do_suspend_lowlevel as it has
>>> > inconsistent call/return info after it jumps to the wakeup vector.
>>> > The graph trace buffer will otherwise become misaligned and
>>> > may eventually crash and hang on suspend.
>>> >
>>> > To reproduce the issue and test the fix:
>>> > Run a function_graph trace over suspend/resume and set the graph
>>> > function to suspend_devices_and_enter. This consistently hangs the
>>> > system without this fix.
>>> >
>>> > Signed-off-by: Todd Brandt <todd.e.brandt@xxxxxxxxxxxxxxx>
>>>
>>> OK
>>>
>>> I guess we'll need that in -stable, right?
>>
>> Sure, it's pretty important. The new analyze_suspend features are going
>> to rely pretty heavily on full graph trace of suspend (all we can do is
>> standby and freeze now because of the bug). Thanks!
>
> OK, applied, but let Steven and the x86 folks see it.
>
>>>
>>> > ---
>>> >  arch/x86/kernel/acpi/sleep.c | 7 +++++++
>>> >  1 file changed, 7 insertions(+)
>>> >
>>> > diff --git a/arch/x86/kernel/acpi/sleep.c b/arch/x86/kernel/acpi/sleep.c
>>> > index d1daead..311360e 100644
>>> > --- a/arch/x86/kernel/acpi/sleep.c
>>> > +++ b/arch/x86/kernel/acpi/sleep.c
>>> > @@ -16,6 +16,7 @@
>>> >  #include <asm/cacheflush.h>
>>> >  #include <asm/realmode.h>
>>> >
>>> > +#include <linux/ftrace.h>
>>> >  #include "../../realmode/rm/wakeup.h"
>>> >  #include "sleep.h"
>>> >
>>> > @@ -107,7 +108,13 @@ int x86_acpi_suspend_lowlevel(void)
>>> >         saved_magic = 0x123456789abcdef0L;
>>> >  #endif /* CONFIG_64BIT */
>>> >
>>> > +       /*
>>> > +        * Pause/unpause graph tracing around do_suspend_lowlevel as it has
>>> > +        * inconsistent call/return info after it jumps to the wakeup vector
>>> > +        */
>>> > +       pause_graph_tracing();
>>> >         do_suspend_lowlevel();
>>> > +       unpause_graph_tracing();
>>> >         return 0;
>>> >  }
>>> >
>>> > --
>>> > 2.1.4
>>> >
>>> > --
>
>
> Thanks,
> Rafael