[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [dm-devel] curious why waitevent sets event_nr field of event_thread struct from kernel every time.



On mar, 2005-03-29 at 17:15 -0500, goggin, edward wrote:
> The multipathd from 0.4.4-pre4 has died a 3 more times.
> I am still not sure what's up here.
> 
Certainly in the logger code, which is very fresh and hard to debug.
Help appreciated here, as my brain kind of melted during the
code-give-birth :/

> Also, curious why waitevent() always sets the event_nr
> field of the event_thread structure in multipathd to
> the event number value pulled from the kernel for the
> mapped device instead of using the event number
> currently saved in the event_thread structure?
> 
> While I can see getting the number from the kernel the
> first time through the code for each mapped device, I
> think the way it is coded it is possible to lose events.
> 
> I think events can get lost if they occur between the
> time multipath pulls out the kernel path state for the
> affected mapped device for the previously detected event
> and the time that the new waitevent pthread calls waitevent
> for the same mapped device.  Granted, the pthread will
> likely make it first -- but there is nothing in place which
> makes this so.
> 
Have you tried your suggested design ?
Is there a regression at first glance ?

> --
> dm-devel mailing list
> dm-devel redhat com
> https://www.redhat.com/mailman/listinfo/dm-devel
-- 
christophe varoqui <christophe varoqui free fr>



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]