[PATCH] arch: x86: take precautions against spurious timer interrupts

[PATCH] arch: x86: take precautions against spurious timer interrupts

[Vladimir Davydov]

If hpet is enabled by hpet_late_init() - this usually occurs on systems
with buggy BIOS, which does not report about hpet presence through ACPI,
hpet_clockevent's event_handler can be left uninitialized by
clockevents_register_device() because of hpet_clockevent low rating (by
the time hpet_late_init() is called, high prio apic timers have already
been setup). The event_handler is then initialized a bit later by the
clocksource_done_booting() procedure.

Normally, timer interrupts should not be delivered between these two
calls, but if e.g. the kernel is booted using kexec, there might be some
pending interrupts from the previous kernel's context, which can lead to
a NULL pointer dereference.

So, take precautions against spurious timer interrupts by checking the
event_handler value before calling it.
---
 arch/x86/kernel/time.c |   18 +++++++++++++++++-
 1 files changed, 17 insertions(+), 1 deletions(-)

diff --git a/arch/x86/kernel/time.c b/arch/x86/kernel/time.c
index c6eba2b..43bdd3a 100644
--- a/arch/x86/kernel/time.c
+++ b/arch/x86/kernel/time.c
@@ -57,7 +57,23 @@ EXPORT_SYMBOL(profile_pc);
  */
 static irqreturn_t timer_interrupt(int irq, void *dev_id)
 {
-	global_clock_event->event_handler(global_clock_event);
+	/*
+	 * If hpet is enabled by hpet_late_init(), event_handler can be left
+	 * uninitialized by clockevents_register_device() because of
+	 * hpet_clockevent low rating (by the time hpet_late_init() is called,
+	 * high prio apic timers have already been setup). The event_handler is
+	 * then initialized a bit later by the clocksource_done_booting()
+	 * procedure.
+	 *
+	 * Normally, timer interrupts should not be delivered between these two
+	 * calls, but if e.g. the kernel is booted using kexec, there might be
+	 * some pending interrupts from the previous kernel's context, which
+	 * can lead to a NULL pointer dereference.
+	 *
+	 * So, take precautions against spurious timer interrupts.
+	 */
+	if (global_clock_event->event_handler)
+		global_clock_event->event_handler(global_clock_event);
 
 	/* MCA bus quirk: Acknowledge irq0 by setting bit 7 in port 0x61 */
 	if (MCA_bus)
-- 
1.7.1

Re: [PATCH] arch: x86: take precautions against spurious timer interrupts

[Paul Gortmaker]

On 12-04-23 08:18 AM, Vladimir Davydov wrote:
> If hpet is enabled by hpet_late_init() - this usually occurs on systems
> with buggy BIOS, which does not report about hpet presence through ACPI,
> hpet_clockevent's event_handler can be left uninitialized by
> clockevents_register_device() because of hpet_clockevent low rating (by
> the time hpet_late_init() is called, high prio apic timers have already
> been setup). The event_handler is then initialized a bit later by the
> clocksource_done_booting() procedure.
> 
> Normally, timer interrupts should not be delivered between these two
> calls, but if e.g. the kernel is booted using kexec, there might be some
> pending interrupts from the previous kernel's context, which can lead to
> a NULL pointer dereference.

Reading between the lines here, I'm guessing that this is specific
to the kexec use case, and never seen anywhere else?  In which case,
it seems a shame to add another conditional to the main timer_interrupt
for an event that may only happen once at boot, and even then, only
in a corner use-case.  Can you deal with the invalid state somewhere
in an _init section instead, perhaps even within CONFIG_KEXEC?  Or
at least ensure a dummy no-op handler is attached early enough?

Paul.
--

> 
> So, take precautions against spurious timer interrupts by checking the
> event_handler value before calling it.
> ---
>  arch/x86/kernel/time.c |   18 +++++++++++++++++-
>  1 files changed, 17 insertions(+), 1 deletions(-)
> 
> diff --git a/arch/x86/kernel/time.c b/arch/x86/kernel/time.c
> index c6eba2b..43bdd3a 100644
> --- a/arch/x86/kernel/time.c
> +++ b/arch/x86/kernel/time.c
> @@ -57,7 +57,23 @@ EXPORT_SYMBOL(profile_pc);
>   */
>  static irqreturn_t timer_interrupt(int irq, void *dev_id)
>  {
> -	global_clock_event->event_handler(global_clock_event);
> +	/*
> +	 * If hpet is enabled by hpet_late_init(), event_handler can be left
> +	 * uninitialized by clockevents_register_device() because of
> +	 * hpet_clockevent low rating (by the time hpet_late_init() is called,
> +	 * high prio apic timers have already been setup). The event_handler is
> +	 * then initialized a bit later by the clocksource_done_booting()
> +	 * procedure.
> +	 *
> +	 * Normally, timer interrupts should not be delivered between these two
> +	 * calls, but if e.g. the kernel is booted using kexec, there might be
> +	 * some pending interrupts from the previous kernel's context, which
> +	 * can lead to a NULL pointer dereference.
> +	 *
> +	 * So, take precautions against spurious timer interrupts.
> +	 */
> +	if (global_clock_event->event_handler)
> +		global_clock_event->event_handler(global_clock_event);
>  
>  	/* MCA bus quirk: Acknowledge irq0 by setting bit 7 in port 0x61 */
>  	if (MCA_bus)

Re: [PATCH] arch: x86: take precautions against spurious timer interrupts

[Vladimir Davydov]

On 04/24/2012 05:55 PM, Paul Gortmaker wrote:
> On 12-04-23 08:18 AM, Vladimir Davydov wrote:
>> If hpet is enabled by hpet_late_init() - this usually occurs on systems
>> with buggy BIOS, which does not report about hpet presence through ACPI,
>> hpet_clockevent's event_handler can be left uninitialized by
>> clockevents_register_device() because of hpet_clockevent low rating (by
>> the time hpet_late_init() is called, high prio apic timers have already
>> been setup). The event_handler is then initialized a bit later by the
>> clocksource_done_booting() procedure.
>>
>> Normally, timer interrupts should not be delivered between these two
>> calls, but if e.g. the kernel is booted using kexec, there might be some
>> pending interrupts from the previous kernel's context, which can lead to
>> a NULL pointer dereference.
> Reading between the lines here, I'm guessing that this is specific
> to the kexec use case, and never seen anywhere else?

Yes, you're right. I've caught it several times while using kexec.

> In which case,
> it seems a shame to add another conditional to the main timer_interrupt
> for an event that may only happen once at boot, and even then, only
> in a corner use-case.  Can you deal with the invalid state somewhere
> in an _init section instead, perhaps even within CONFIG_KEXEC?  Or
> at least ensure a dummy no-op handler is attached early enough?
>
> Paul.
> --

Agree.

I can init hpet_clockevent.event_handler with clockevents_handle_noop() 
in its definition.

Will rework and resend.

Thanks for replying.

>
>> So, take precautions against spurious timer interrupts by checking the
>> event_handler value before calling it.
>> ---
>>   arch/x86/kernel/time.c |   18 +++++++++++++++++-
>>   1 files changed, 17 insertions(+), 1 deletions(-)
>>
>> diff --git a/arch/x86/kernel/time.c b/arch/x86/kernel/time.c
>> index c6eba2b..43bdd3a 100644
>> --- a/arch/x86/kernel/time.c
>> +++ b/arch/x86/kernel/time.c
>> @@ -57,7 +57,23 @@ EXPORT_SYMBOL(profile_pc);
>>    */
>>   static irqreturn_t timer_interrupt(int irq, void *dev_id)
>>   {
>> -	global_clock_event->event_handler(global_clock_event);
>> +	/*
>> +	 * If hpet is enabled by hpet_late_init(), event_handler can be left
>> +	 * uninitialized by clockevents_register_device() because of
>> +	 * hpet_clockevent low rating (by the time hpet_late_init() is called,
>> +	 * high prio apic timers have already been setup). The event_handler is
>> +	 * then initialized a bit later by the clocksource_done_booting()
>> +	 * procedure.
>> +	 *
>> +	 * Normally, timer interrupts should not be delivered between these two
>> +	 * calls, but if e.g. the kernel is booted using kexec, there might be
>> +	 * some pending interrupts from the previous kernel's context, which
>> +	 * can lead to a NULL pointer dereference.
>> +	 *
>> +	 * So, take precautions against spurious timer interrupts.
>> +	 */
>> +	if (global_clock_event->event_handler)
>> +		global_clock_event->event_handler(global_clock_event);
>>
>>   	/* MCA bus quirk: Acknowledge irq0 by setting bit 7 in port 0x61 */
>>   	if (MCA_bus)