Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 19 Jun 2001 23:25:14 -0400 (EDT)
From:      Mikhail Teterin <mi@aldan.algebra.com>
To:        dima@unixfreak.org
Cc:        cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org
Subject:   Re: cvs commit: src/sbin/mdmfs Makefile mdmfs.8 mdmfs.c pathnames.h 
Message-ID:  <200106200325.f5K3PF628641@aldan.algebra.com>
In-Reply-To: <20010620030407.C8A433E2F@bazooka.unixfreak.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On 19 Jun, Dima Dorfman wrote:
>> >> So, will
>> >> 
>> >> 	/dev/da0b		/tmp		mfs	rw,async
>> >> 
>> >> Just Work (TM) now?
>> > 
>> > Besides the fact that mdmfs(8) isn't installed as /sbin/mount_mfs,
>> > there are two things wrong (from mdmfs's point of view) with that
>> > line:
>> > 
>> >  *  /dev/da0b --  this  should either  be  "/dev/md" for  autounit,
>> > or  "/dev/mdX" where  X  is  the desired  unit  number. This  small
>> > incompatibility is intentional: I don't  think it's a good idea for
>> > users to blindly assume that md(4) is mfs; at least if they have to
>> > change that, they'll know that something might be different.
>>
>> Ok...  For being  entirely  compatibile, mdmfs  should  check if  the
>> specified partition is being used for paging. If it is, md should use
>> the /dev/md  and use swap  as backing for the  memory disk. If  it is
>> not, it should use /dev/md and the specified disk as vnode-backing.
>
> See above. I don't want people  to just symlink mdmfs to mount_mfs and
> off they go;  that can lead to foot shooting  if they make assumptions
> about mfs that don't  work for md. I don't think it's  too much to ask
> to change /dev/da0b to /dev/md.

I don't know... There is a lot of value in being backwards compatible...
May be with a big warning, but  after an upgrade, the system should just
keep on working. IMHO, that is.
 
>> >  *  There's no  size  specification.  The only  way  I've seen  MFS
>> > used  is  with   an  -s  argument  in  the   options  field;  e.g.,
>> > "rw,async,-s<size>".  If  your  example  works  without  -s  (i.e.,
>> > mount_mfs assumes a default size), mdmfs should be fixed to do that
>> > too; although I don't know what the default size should be.
>>
>> Well, with the mfs the default is the size of the partition (da0b).
>
> I suppose it might be possible to  use, say, half the swap size of the
> system. I  think what mfs allegedly  did is quite silly;  who wants to
> use their *entire* swap space for a filesystem?

The filesystems used swap space dynamicly,  AFAIK. As in Solaris -- your
files in /tmp  and your swap use  grew to meet each other  in the middle
somewhere. That way, you did not have to guess, how much to allocate for
FS and how much to leave availalble.

>> For  those, who  modified the  /etc/fstab, it  should be  possible to
>> specify the type of md as a mount option explicitly.
>
> What do you mean  "type of md"? As in malloc-  or swap-backed? You can
> do that. This:
> 
> 	/dev/md			/tmp		mfs	-M,rw,async

Yes, that's what I meant. Cool.

	-mi



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe cvs-all" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200106200325.f5K3PF628641>