Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 3 Mar 2000 17:03:38 -0800 (PST)
From:      Matthew Dillon <dillon@apollo.backplane.com>
To:        Brian Beattie <beattie@aracnet.com>
Cc:        Michael Bacarella <mbac@bsd4.nyct.net>, Julian Elischer <julian@elischer.org>, David Scheidt <dscheidt@enteract.com>, freebsd-hackers@FreeBSD.ORG
Subject:   Re: Copy-on-write filesystem
Message-ID:  <200003040103.RAA12519@apollo.backplane.com>
References:   <Pine.LNX.4.21.0003031447050.22409-100000@shell1.aracnet.com>

next in thread | previous in thread | raw e-mail | index | archive | help
:> > I think. One problem with "Copy-on-write, when applied to 
:> > file copies is that you need to assign the blocks up front, even if you
:> > don't copy the data, as otherwise you could run out of space
:> > when the copy is actually needed.
:> 
:> That's the only real drawback I've considered.
:> 
:> People accept it (barely) when the OS commits to providing virtual memory
:> it does not have, killing processes when the system falls into debt.
:> 
:> No one will appreciate that happening to their "permanent" data,
:> especially if the OS decides that the best way to get out of debt is by
:> deleting a file :)
:>
:
:Actually, since this is copy-on-write, you do not need the block, until
:you write.  If you need to make a copy, it will be on a write system call
:(possibly an inode update), just fail the write ENOSPC or whatever.  Or am
:I missing something simple here.

    The issue here is to ensure that you have sufficient swap.  There are two
    ways to do this:  One, make sure you have enough swap to cover likely
    operations done on the overlay, or Two, pre-reserve the entire partition's
    worth of space in swap.

    Both these options already exist for fresh swap-backed VN filesystems 
    under 4.0 so you'd get them for free if we were to implement overlay
    functionality.

    Trying to do anything fancier then that will create more problems then
    it solves.

						-Matt



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




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200003040103.RAA12519>