Re: [PATCH][v2] timekeeping: Fix memory overwrite of sleep_time_bin array

From: Chen Yu
Date: Wed Jul 20 2016 - 06:59:30 EST


Hi Thomas,
On Tue, Jul 19, 2016 at 12:40:14PM +0200, Thomas Gleixner wrote:
> On Tue, 19 Jul 2016, Chen Yu wrote:
> > On 2016å07æ19æ 16:36, Thomas Gleixner wrote:
> > > On Tue, 19 Jul 2016, Chen Yu wrote:
> > > > Further investigation shows that, the problem is caused by setting
> > > > /sys/power/pm_trace to 1 before the 1st hibernation, since once
> > > > pm_trace is enabled, the rtc becomes an unmeaningful value after resumed,
> > >
> > > So why is the RTC value useless if pm_trace is enabled? I really have a hard
> > > time to understand why pm_trace would affect the sleep time readout from
> > > RTC.
> >
> > After pm_trace is enabled, during system suspend/hibernate, the hash name of
> > each devices will be written to rtc, so the rtc value depends on what we
> > write in last suspend round, thus pm_trace can be used for diagnose which
> > device failed to suspend(eg, the suspending on this device hang the system,
> > we reboot the system , and check rtc hash value).
> >
> > In our case, after first hibernate/resume round, we found our current system
> > time is at 2117, so syscore_resume -> timekeeping_resume :
> > __timekeeping_inject_sleeptime(tk, &ts_delta) would inject a quite large
> > delta : 2117 - 2017 year, thus the sleep_time_bin is overflow.
>
> While the range check is certainly correct and a good thing to have it's wrong
> in the first place to call __timekeeping_inject_sleeptime() in case that
> pm_trace is enabled simply because that "hash" time value will also wreckage
> timekeeping. Your patch is just curing the symptom in the debug code but not
> fixing the root cause.
>
OK. I've modified the patch.
In case I break any other stuff :p, could you help check
if this patch is in the right direction, thanks:

1. There are two places would invoke __timekeeping_inject_sleeptime(),
they are timekeeping_resume and rtc_resume, so we need to deal with
them respctively.

2. for rtc_resume, if the pm_trace has once been enabled,
we bypass the injection of sleep time.

3. for timekeeping_resume,
Currently we either use nonstop clock source, or use persistent
clock to get the sleep time. As pm_trace breaks systems who use rtc
as a persistent clock, x86 is affected. So we add a
check for x86 that, if the pm_trace has been enabled, we can not
trust the persistent clock delta read from rtc, thus bypass
the injection of sleep time in this case.

4. Why we checked the history of pm_trace: once pm_trace
has been enabled, the delta of rtc would not be reliable anymore.
For example, if we only check current pm_trace, we might still get
memory overwrite:

4.1 echo 1 > /sys/power/pm_trace
4.2 hibernate/resume (rtc is broken, do not add delta from rtc because pm_trace is 1)
4.3 echo 0 > /sys/power/pm_trace
4.4 hibernate/resume (rtc is still broken, but add delta from rtc because pm_trace is 0)

so we have to check if pm_trace has once been enabled, if it is, we will not add
any delta from tsc until system reboots.

Thanks,
Yu

Index: linux/kernel/time/timekeeping.c
===================================================================
--- linux.orig/kernel/time/timekeeping.c
+++ linux/kernel/time/timekeeping.c
@@ -1448,6 +1448,11 @@ void __weak read_boot_clock64(struct tim
ts->tv_nsec = 0;
}

+bool __weak persistent_clock_is_usable(void)
+{
+ return true;
+}
+
/* Flag for if timekeeping_resume() has injected sleeptime */
static bool sleeptime_injected;

@@ -1609,6 +1614,7 @@ void timekeeping_resume(void)
unsigned long flags;
struct timespec64 ts_new, ts_delta;
cycle_t cycle_now, cycle_delta;
+ bool persist_clock_usable = true;

sleeptime_injected = false;
read_persistent_clock64(&ts_new);
@@ -1660,9 +1666,11 @@ void timekeeping_resume(void)
} else if (timespec64_compare(&ts_new, &timekeeping_suspend_time) > 0) {
ts_delta = timespec64_sub(ts_new, timekeeping_suspend_time);
sleeptime_injected = true;
+ if (!persistent_clock_is_usable())
+ persist_clock_usable = false;
}

- if (sleeptime_injected)
+ if (sleeptime_injected && persist_clock_usable)
__timekeeping_inject_sleeptime(tk, &ts_delta);

/* Re-base the last cycle value */
Index: linux/arch/x86/kernel/rtc.c
===================================================================
--- linux.orig/arch/x86/kernel/rtc.c
+++ linux/arch/x86/kernel/rtc.c
@@ -8,6 +8,7 @@
#include <linux/export.h>
#include <linux/pnp.h>
#include <linux/of.h>
+#include <linux/pm-trace.h>

#include <asm/vsyscall.h>
#include <asm/x86_init.h>
@@ -147,6 +148,10 @@ void read_persistent_clock(struct timesp
x86_platform.get_wallclock(ts);
}

+bool persistent_clock_is_usable(void)
+{
+ return !pm_trace_once_enabled();
+}

static struct resource rtc_resources[] = {
[0] = {
Index: linux/drivers/rtc/class.c
===================================================================
--- linux.orig/drivers/rtc/class.c
+++ linux/drivers/rtc/class.c
@@ -20,6 +20,7 @@
#include <linux/idr.h>
#include <linux/slab.h>
#include <linux/workqueue.h>
+#include <linux/pm-trace.h>

#include "rtc-core.h"

@@ -138,7 +139,7 @@ static int rtc_resume(struct device *dev
sleep_time = timespec64_sub(sleep_time,
timespec64_sub(new_system, old_system));

- if (sleep_time.tv_sec >= 0)
+ if ((sleep_time.tv_sec >= 0) && (!pm_trace_once_enabled()) )
timekeeping_inject_sleeptime64(&sleep_time);
rtc_hctosys_ret = 0;
return 0;
Index: linux/kernel/power/main.c
===================================================================
--- linux.orig/kernel/power/main.c
+++ linux/kernel/power/main.c
@@ -532,6 +532,7 @@ power_attr(wake_unlock);

#ifdef CONFIG_PM_TRACE
int pm_trace_enabled;
+bool pm_trace_been_enabled;

static ssize_t pm_trace_show(struct kobject *kobj, struct kobj_attribute *attr,
char *buf)
@@ -548,6 +549,7 @@ pm_trace_store(struct kobject *kobj, str
if (sscanf(buf, "%d", &val) == 1) {
pm_trace_enabled = !!val;
if (pm_trace_enabled) {
+ pm_trace_been_enabled = true;
pr_warn("PM: Enabling pm_trace changes system date and time during resume.\n"
"PM: Correct system time has to be restored manually after resume.\n");
}
Index: linux/include/linux/pm-trace.h
===================================================================
--- linux.orig/include/linux/pm-trace.h
+++ linux/include/linux/pm-trace.h
@@ -6,12 +6,18 @@
#include <linux/types.h>

extern int pm_trace_enabled;
+extern bool pm_trace_been_enabled;

static inline int pm_trace_is_enabled(void)
{
return pm_trace_enabled;
}

+static inline bool pm_trace_once_enabled(void)
+{
+ return pm_trace_been_enabled;
+}
+
struct device;
extern void set_trace_device(struct device *);
extern void generate_pm_trace(const void *tracedata, unsigned int user);
@@ -25,6 +31,7 @@ extern int show_trace_dev_match(char *bu
#else

static inline int pm_trace_is_enabled(void) { return 0; }
+static inline bool pm_trace_once_enabled(void) { return false; }

#define TRACE_DEVICE(dev) do { } while (0)
#define TRACE_RESUME(dev) do { } while (0)