From mboxrd@z Thu Jan 1 00:00:00 1970 From: Suparna Bhattacharya Subject: Re: [PATCH, RFC] map multiple blocks at a time in mpage_readpage(s) Date: Fri, 21 Oct 2005 11:48:40 +0530 Message-ID: <20051021061840.GA3958@in.ibm.com> References: <20051004194246.GA18432@lst.de> <20051019115639.GA13589@lst.de> <20051020150631.1f860552.akpm@osdl.org> Reply-To: suparna@in.ibm.com Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Christoph Hellwig , linux-fsdevel@vger.kernel.org Return-path: Received: from e5.ny.us.ibm.com ([32.97.182.145]:64219 "EHLO e5.ny.us.ibm.com") by vger.kernel.org with ESMTP id S964881AbVJUGJh (ORCPT ); Fri, 21 Oct 2005 02:09:37 -0400 Received: from d01relay04.pok.ibm.com (d01relay04.pok.ibm.com [9.56.227.236]) by e5.ny.us.ibm.com (8.12.11/8.12.11) with ESMTP id j9L69V63008525 for ; Fri, 21 Oct 2005 02:09:31 -0400 Received: from d01av02.pok.ibm.com (d01av02.pok.ibm.com [9.56.224.216]) by d01relay04.pok.ibm.com (8.12.10/NCO/VERS6.7) with ESMTP id j9L69ViT093972 for ; Fri, 21 Oct 2005 02:09:31 -0400 Received: from d01av02.pok.ibm.com (loopback [127.0.0.1]) by d01av02.pok.ibm.com (8.12.11/8.13.3) with ESMTP id j9L69UwU031416 for ; Fri, 21 Oct 2005 02:09:30 -0400 To: Andrew Morton Content-Disposition: inline In-Reply-To: <20051020150631.1f860552.akpm@osdl.org> Sender: linux-fsdevel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Thu, Oct 20, 2005 at 03:06:31PM -0700, Andrew Morton wrote: > Christoph Hellwig wrote: > > > > On Tue, Oct 04, 2005 at 09:42:46PM +0200, Christoph Hellwig wrote: > > > This patch changes mpage_readpage/mpage_readpages to use a get_blocks > > > call that gets the disk mapping information for multiple blocks at the > > > same time, similar to the way direct I/O code works. For extent based > > > filesystems like jfs, xfs or reiser4 this allows to reduce the overhead > > > of the allocator calls, which is especially nice when that codepath > > > is rather heavyweight (as it is for example in xfs). For filesystems > > > that don't have an allocator that make use of it nothing changes. > > > > > > I've tested this heavily on XFS with block size = page size and 512 byte > > > blocks, and it passes the XFS QA regression test suite fine. I've done > > > some basic fsx testing on all the other affected filesystems. > > > > I think the patch should go into -mm after no one had serious objections. > > It adds a bunch of complexity and overhead which will speed up a filesystem > which few people use while slowing down the filesystems which most people > use. > > If/when we have an ext3 implementation of get_blocks() and some benchmarks > which show it's worthwhile then OK. We do already have this based on reservation based allocation (Mingming posted the patches a while back). I think we need to combine the work we did for mpage_writepages with the readpages work. > > Could we please see the xfs benchmark results? > > first_logical_block should have type sector_t. > > Did you consider aggregating those nine args to do_mpage_readpage() into a > struct, passing the address of that? Regards Suparna > > - > To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- Suparna Bhattacharya (suparna@in.ibm.com) Linux Technology Center IBM Software Lab, India