Re: [BUGFIX][PATCH -rc/-mm] prevent kprobes from catching spurious page faults

[Masami Hiramatsu <mhiramat at redhat dot com>]

Mathieu Desnoyers wrote:
> * Ingo Molnar (mingo@elte.hu) wrote:
>> * Masami Hiramatsu <mhiramat@redhat.com> wrote:
>>
>>> -	if (notify_page_fault(regs))
>>> -		return;
>>>  	if (unlikely(kmmio_fault(regs, address)))
>>>  		return;
>>>
>>> @@ -634,6 +632,9 @@ void __kprobes do_page_fault(struct pt_r
>>>  		if (spurious_fault(address, error_code))
>>>  			return;
>>>
>>> +		/* kprobes don't want to hook the spurious faults. */
>>> +		if (notify_page_fault(regs))
>>> +			return;
>>>  		/*
>>>  		 * Don't take the mm semaphore here. If we fixup a prefetch
>>>  		 * fault we could otherwise deadlock.
>>> @@ -641,6 +642,9 @@ void __kprobes do_page_fault(struct pt_r
>>>  		goto bad_area_nosemaphore;
>>>  	}
>>>
>>> +	/* kprobes don't want to hook the spurious faults. */
>>> +	if (notify_page_fault(regs))
>>> +		return;
>> I dont know - this spreads that callback to two places now. Any
>> reason why kprobes cannot call spurious_fault(), if there's a
>> probe active?
>>
>> Also, moving that would remove the planned cleanup of merging these
>> two into one call:
>>
>>  	if (notify_page_fault(regs))
>>  		return;
>>   	if (unlikely(kmmio_fault(regs, address)))
>>   		return;
>>
>> We should reduce the probing cross section, not increase it,
>> especially in such a critical codepath as the pagefault handler.
>>
>> Btw., why cannot kprobes install a dynamic probe to the fault
>> handler itself? That way the default path would have no such
>> callbacks and checks at all.
>>
> 
> Or we could simply merge my 2 LTTng page fault handler tracepoints per
> architecture and be done with it ?

As you can see, these functions are a kind of fixup code.
If it succeed fixup a fault, do_page_fault() has to return because
the fault is fixed.

Since tracepoint itself is just a watchpoint, it should not
change code path. So, I think just moving kmmio_fault() to
notify_page_fault() is enough.

> I'd need to clean up the patchset a little bit to fold a few patches,
> but that would be straightforward enough.

Anyway, I agree with the idea to push tracepoint in the pagefault.
It is very useful for watching system behavior.

Thanks!


> 
> Mathieu
> 

-- 
Masami Hiramatsu

Software Engineer
Hitachi Computer Products (America) Inc.
Software Solutions Division

e-mail: mhiramat@redhat.com