xfs
[Top] [All Lists]

Re: pagebuf page cleaner and page aging

To: Marcelo Tosatti <marcelo@xxxxxxxxxxxxxxxx>
Subject: Re: pagebuf page cleaner and page aging
From: Christoph Hellwig <hch@xxxxxxxxxx>
Date: Fri, 19 Jan 2001 18:38:03 +0100
Cc: Steve Lord <lord@xxxxxxx>, Rajagopal Ananthanarayanan <ananth@xxxxxxx>, linux-xfs@xxxxxxxxxxx
In-reply-to: <Pine.LNX.4.21.0101191123270.5416-100000@freak.distro.conectiva>; from marcelo@conectiva.com.br on Fri, Jan 19, 2001 at 12:38:27PM -0200
References: <200101191510.f0JFAHs02250@jen.americas.sgi.com> <Pine.LNX.4.21.0101191123270.5416-100000@freak.distro.conectiva>
Sender: owner-linux-xfs@xxxxxxxxxxx
On Fri, Jan 19, 2001 at 12:38:27PM -0200, Marcelo Tosatti wrote:
> Obviously the VM does not have knowledge about the filesystem low-level
> information, which is also needed to get write clustering right.

IMHO we should try to give the VM knowlede how to cluster without
having to go into the low-level code at all.
The idea I had for such information is the notation of a virtual extent.
A virtual extend would simply be a list-head in the struct page that links
a pages together that are worth clustering because they are either real
extents (because the fs supports it) or just continguos aligned on disk
for block-based filesystems (yeah, we have to do magic on readtime to
get this information, best by a readpage op).

The disadvantage of this appropeach is that it bloats struct page even more...

        Christoph

-- 
Whip me.  Beat me.  Make me maintain AIX.

<Prev in Thread] Current Thread [Next in Thread>