>From: Wojciech Puchar <wojtek@chylonia.3miasto.net>
>To: Gregory McGarry <g.mcgarry@ieee.org>
>CC: tech-kern@netbsd.org
>Subject: Re: More on sorted I/O requests
>Date: Fri, 28 Jun 2002 12:05:45 +0200 (CEST)
>
> > /storage/scratch
> > $ dd if=/dev/zero of=bigfile bs=1024k count=200
> > 200+0 records in
> > 200+0 records out
> > 209715200 bytes transferred in 8.326 secs (25187989 bytes/sec)
> > $ rm bigfile
> > $ df
> > Filesystem  1K-blocks     Used     Avail Capacity  Mounted on
> > /dev/wd0a       63248    31634     28451    52%    /
> > /dev/wd0e      253615    83672    157262    34%    /var
> > /dev/wd0f     1465278   372914   1019100    26%    /usr
> > mfs:123         31727     2175     27965     7%    /tmp
> > /dev/wd0g     1018329    56420    909297     5%    /home
> > /dev/wd0h    12766536 11862460    265748    97%    /storage
>
>doing sync few times after rm solves the problem.
>
>it's not I/O request queue problem, just filesystem. delayed free speeds
>up things

I *think* FreeBSD also suffers from this "feature". Is is possible for the 
trickle sync to be activated by not only whatever activates it now 
(timeouts, size of outstanding requests, or whatever...) but also by 
requests for space? I.e. could a request to allocate some space to file X 
would automatically sync delayed deletes if it appears no space is 
available?

_________________________________________________________________
Send and receive Hotmail on your mobile device: http://mobile.msn.com