Subject: Re: lfs panic on NetBSD-4.0_BETA2
To: None <tech-kern@netbsd.org>
From: Thor Lancelot Simon <tls@rek.tjls.com>
List: tech-kern
Date: 12/19/2006 13:03:49
On Tue, Dec 19, 2006 at 06:22:09PM +0100, Joerg Sonnenberger wrote:
> On Tue, Dec 19, 2006 at 11:58:10AM -0500, Blair Sadewitz wrote:
> > Oh?  Simply by using it, or is there a specific operation?
> 
> Simply by using it. It should be noted that it is heavily using mmap and
> msyncing ever chunk once it is done, so with enough bandwidth it can
> easily pressure LFS a lot.

I will note once again that this is severely buggy application behavior,
and that LFS will never -- and can never -- comfortably accomodate an
application which insists on abusing msync in this way: every single
call to msync will cause a partial-segment write.

And, of course, there is _no reason_ for a BitTorrent client to do anything
like this: if it ever happens to be the case that a chunk is not fully
written to disk, the standard startup check will reveal that that chunk
has an incorrect checksum, and it will simply be re-fetched.

LFS should not crash the system, but the application should also simply
not do this, or should not be used on LFS.

-- 
Thor Lancelot Simon	                                     tls@rek.tjls.com
  "The liberties...lose much of their value whenever those who have greater
   private means are permitted to use their advantages to control the course
   of public debate."					-John Rawls