All of lore.kernel.org
 help / color / mirror / Atom feed
* Is there a way to interrupt MMIO with kprobes/ltt/etc...
@ 2002-11-08  3:37 Rusty Lynch
  2002-11-08  8:23 ` Vamsi Krishna S .
  0 siblings, 1 reply; 2+ messages in thread
From: Rusty Lynch @ 2002-11-08  3:37 UTC (permalink / raw)
  To: linux-kernel

I have been looking into the possible ways a fault injection tool could be
implemented on the available tools/hooks in the 2.5 kernel.  I can see how
kprobes would help by allowing me to setup handlers when a specific address
is executed, but what about when a specific memory mapped IO address is
touched or looked at?

I know there has been a lot of activity on kprobes, LTT, and others (isn't
there something else?).  Do any of these patches allow a handler to be
called just before some MMIO is accessed?  Messing with architecture
specific debug registers seems problematic since it makes the solution
architecture specific and the number of watch points is pretty limited.

    -rustyl


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: Is there a way to interrupt MMIO with kprobes/ltt/etc...
  2002-11-08  3:37 Is there a way to interrupt MMIO with kprobes/ltt/etc Rusty Lynch
@ 2002-11-08  8:23 ` Vamsi Krishna S .
  0 siblings, 0 replies; 2+ messages in thread
From: Vamsi Krishna S . @ 2002-11-08  8:23 UTC (permalink / raw)
  To: Rusty Lynch; +Cc: linux-kernel

On Fri, Nov 08, 2002 at 03:40:58AM +0000, Rusty Lynch wrote:
> I have been looking into the possible ways a fault injection tool could be
> implemented on the available tools/hooks in the 2.5 kernel.  I can see how
> kprobes would help by allowing me to setup handlers when a specific address
> is executed, but what about when a specific memory mapped IO address is
> touched or looked at?
> 
> I know there has been a lot of activity on kprobes, LTT, and others (isn't
> there something else?).  Do any of these patches allow a handler to be
> called just before some MMIO is accessed?  Messing with architecture
> specific debug registers seems problematic since it makes the solution
> architecture specific and the number of watch points is pretty limited.
> 
You could do this with the interface provided by kwatchpoints patch [1]
without directly mucking with debug registers. The interface is simple:

int register_kwatch(unsigned long addr, u8 length, u8 type,
                kwatch_handler_t handler)

If you don't want to use debug registers or if they are not enough,
only other possibility I can think of is to find all code locations 
where the MMIO space of interest is touched and put execution 
probes there.

[1] You will need these two patches:
http://marc.theaimsgroup.com/?l=linux-kernel&m=103528454215523&w=2
http://marc.theaimsgroup.com/?l=linux-kernel&m=103528454015520&w=2

Cheers,
Vamsi.
-- 
Vamsi Krishna S.
Linux Technology Center,
IBM Software Lab, Bangalore.
Ph: +91 80 5044959
Internet: vamsi@in.ibm.com

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2002-11-08  8:03 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2002-11-08  3:37 Is there a way to interrupt MMIO with kprobes/ltt/etc Rusty Lynch
2002-11-08  8:23 ` Vamsi Krishna S .

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.