Subject: Re: kernel stack overflow detection
To: None <itojun@iijlab.net, yamt@mwd.biglobe.ne.jp>
From: None <eeh@netbsd.org>
List: tech-kern
Date: 06/26/2002 16:05:23
| From: itojun@iijlab.net
| Subject: Re: kernel stack overflow detection 
| Date: Wed, 26 Jun 2002 20:21:12 +0900
| > >i'll commit this if no one objects.
| > >http://www2q.biglobe.ne.jp/~yama/file/kstack/kstack4.diff
| > 
| > 	i bet you will make it on by default in the future (after testing)
| > 	i guess?  am i right?
| > 	(there are way too many options, so i would like to see fewer options)
|
| it's one of debug/diagnostic options.
| (eg. LOCKDEBUG, MALLOC_DEBUG, etc...)
| do you mean they should be merged into one, DIAGNOSTIC?
|
| i separated options from DIAGNOSTIC because of
| following reasons:
|
| KSTACK_CHECK_MAGIC scans kernel stack on each context switch.
| so it affects performance too much.
|
| KSTACK_CHECK_DR0 uses a debug register exclusively, so
| people that want to use it for other purposes will not want to
| enable it.

Why are you changing MI code?  Can't it be done all in MD code?

If you really need it in MI code, does this work on anything other
than x86?  Have you tested it on anything else?

Eduardo