Re: sync(1) not really syncing??

[Antti Kantee <pooka%cs.hut.fi@localhost>]

On Thu Sep 11 2008 at 15:37:41 +0300, Antti Kantee wrote:
> On Thu Sep 11 2008 at 14:18:45 +0200, Jonathan Schleifer wrote:
> > Am 11.09.2008 um 14:15 schrieb Ephaeton%gmx.net@localhost:
> > 
> > >Yeah I mean having sync(2) block until it's "really" done. Which
> > >could stall the system for quite a bit with big / many buffers around.
> > 
> > That would be a DoS as a user could write big files and then sync to  
> > stall the whole system. It may block sync(1), but *NEVER* the system  
> > as a whole.
> 
> Please, actually study the kernel code before making incorrect claims.
> There is no such possibility for a large file DoS in NetBSD since files
> are flushed as they are written (this, however, may be argued to be
> suboptimal ...).

Not to mention that a single system call blocking does not freeze the
system ... anyway.