From owner-freebsd-hackers Mon Aug 7 12:51:45 2000 Delivered-To: freebsd-hackers@freebsd.org Received: from pedigree.cs.ubc.ca (pedigree.cs.ubc.ca [142.103.6.50]) by hub.freebsd.org (Postfix) with ESMTP id 8B14937BBF9 for ; Mon, 7 Aug 2000 12:51:28 -0700 (PDT) (envelope-from ycoady@cs.ubc.ca) Received: from newcastle.cs.ubc.ca (ycoady@newcastle.cs.ubc.ca [142.103.6.21]) by pedigree.cs.ubc.ca (8.8.8/8.6.9) with ESMTP id MAA00039 for ; Mon, 7 Aug 2000 12:51:17 -0700 (PDT) Received: from localhost (ycoady@localhost) by newcastle.cs.ubc.ca (8.9.3/8.6.12) with ESMTP id MAA20608 for ; Mon, 7 Aug 2000 12:51:17 -0700 (PDT) X-Authentication-Warning: newcastle.cs.ubc.ca: ycoady owned process doing -bs Date: Mon, 7 Aug 2000 12:51:16 -0700 (PDT) From: Yvonne Coady To: freebsd-hackers@freebsd.org Subject: ffs_getpages() Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Just wondering if anyone can shed some light on the changes in the page fault handling code in 3.3 vs 4.0. I've noticed that the async read ahead that used to work for mmaped files via a VOP_READ in ffs_getpages() is no longer there... I've run some very simple (and by no means exhaustive) tests, and on the surface the tradeoff appears to be that in 3.3 the disk is accessed synchronously, 16 pages at a time, whereas in 4.0 access may be asynchronous, 32 pages at a time. I am aware of the fact that in 3.3, using the vnode vlastr field was maybe not a perfect way to detect seq access for "normal" objects in ffs_getpages, but I thought the asynch read ahead was a real bonus - I might be overlooking something obvious here, but can anyone tell me why this has changed? Thanks, Yvonne To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message