Hi... On Tue, May 26, 2009 at 3:57 PM, pradeep singh <[email protected]> wrote: > Hi All, > > I am having a tough time debugging concurrency issues it seems. > I find myself flabbergasted at debugging panic because of race issues. > > I would like to ask people here, if they have any advice on how to > properly approach these specific set of panics? > How do they do it? > Are there any rules of thumbs which can help me?
OK, first thing that I could recall quickly is: do you really understand the flow of the code? do you know the possibilities of what might happen if several code path happen at the same time? In debugging race condition, I think it's important either by trying to reduce the threads number (if it's solely a multi threading program), or...in a complex situation like in Linux kernel.... try to intersect with as least triggers as possible i.e is it possible to trigger the panic with just interrupt handler code triggered by 2 CPUs instead of 4 or 8 CPUs regards, Mulyadi. -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to [email protected] Please read the FAQ at http://kernelnewbies.org/FAQ
