2004-11-29 21:12:19

by Darren Hart

[permalink] [raw]
Subject: scheduler BUGON lifespan

I submitted a patch to active_load_balance() that was accepted into mm
and then mainline. The patch included a fix to prevent the system
entering what should have been an impossible state. The previous code
tested for it and then continued, rather than crashing or complaining.
My patch added a BUGON(impossible state) just in case by some fluke it
still happened. How long should this BUGON remain in the kernel? A
month, a year? Is there an accepted duration for such safety nets?

Thanks,

--
Darren Hart <[email protected]>
IBM Linux Technology Center


2004-11-29 21:25:29

by Andrew Morton

[permalink] [raw]
Subject: Re: scheduler BUGON lifespan

Darren Hart <[email protected]> wrote:
>
> How long should this BUGON remain in the kernel?

Until someone thinks to remove it, it seems. There is no established
protocol or period. Often someone will say "hey, this is silly" and will
remove it - usually as part of some wider work.

If you think a BUG or BUG_ON doesn't need to be there any more, feel free
to send a patch..

2004-11-29 22:39:16

by Ingo Molnar

[permalink] [raw]
Subject: Re: scheduler BUGON lifespan


* Darren Hart <[email protected]> wrote:

> I submitted a patch to active_load_balance() that was accepted into mm
> and then mainline. The patch included a fix to prevent the system
> entering what should have been an impossible state. The previous code
> tested for it and then continued, rather than crashing or complaining.
> My patch added a BUGON(impossible state) just in case by some fluke it
> still happened. How long should this BUGON remain in the kernel? A
> month, a year? Is there an accepted duration for such safety nets?

it's pretty random how long it survives. Sometimes i remove one after
never having seen it for months/years and grepping lkml and googling
around for the file & line. The BUG_ON()s are pretty cheap, as they
often check what is being fetched anyway.

Ingo