From: Badari Pulavarty <pbadari@us.ibm.com>
To: Nathan Scott <nathans@sgi.com>
Cc: christoph <hch@lst.de>,
mcao@us.ibm.com, akpm@osdl.org,
lkml <linux-kernel@vger.kernel.org>,
jeremy@sgi.com, linux-fsdevel <linux-fsdevel@vger.kernel.org>
Subject: Re: [PATCH 0/3] map multiple blocks in get_block() and mpage_readpages()
Date: Mon, 20 Feb 2006 15:06:11 -0800 [thread overview]
Message-ID: <1140476772.22756.28.camel@dyn9047017100.beaverton.ibm.com> (raw)
In-Reply-To: <20060221085953.H9484650@wobbly.melbourne.sgi.com>
On Tue, 2006-02-21 at 08:59 +1100, Nathan Scott wrote:
> On Mon, Feb 20, 2006 at 01:21:27PM -0800, Badari Pulavarty wrote:
> > Hi,
>
> Hi Badari,
>
> > Following patches add support to map multiple blocks in ->get_block().
> > This is will allow us to handle mapping of multiple disk blocks for
> > mpage_readpages() and mpage_writepages() etc. Instead of adding new
> > argument, I use "b_size" to indicate the amount of disk mapping needed
> > for get_block(). And also, on success get_block() actually indicates
> > the amount of disk mapping it did.
>
> Thanks for doing this work!
>
> > Now that get_block() can handle multiple blocks, there is no need
> > for ->get_blocks() which was added for DIO.
> >
> > [PATCH 1/3] pass b_size to ->get_block()
> >
> > [PATCH 2/3] map multiple blocks for mpage_readpages()
> >
> > [PATCH 3/3] remove ->get_blocks() support
> >
> > I noticed decent improvements (reduced sys time) on JFS, XFS and ext3.
> > (on simple "dd" read tests).
> >
> > (rc3.mm1) (rc3.mm1 + patches)
> > real 0m18.814s 0m18.482s
> > user 0m0.000s 0m0.004s
> > sys 0m3.240s 0m2.912s
> >
> > Andrew, Could you include it in -mm tree ?
> >
> > Comments ?
>
> I've been running these patches in my development tree for awhile
> and have not seen any problems. My one (possibly minor) concern
> is that we pass get_block a size in units of bytes, e.g....
>
> bh->b_size = 1 << inode->i_blkbits;
> err = get_block(inode, block, bh, 1);
>
> And b_size is a u32. We have had the situation in the past where
> people (I'm looking at you, Jeremy ;) have been issuing multiple-
> gigabyte direct reads/writes through XFS. The syscall interface
> takes an (s)size_t in bytes, which on 64 bit platforms is a 64 bit
> byte count.
> I wonder if this change will end up ruining things for the lunatic
> fringe issuing these kinds of IOs? Maybe the get_block call could
> take a block count rather than a byte count?
Yes. I thought about it too.. I wanted to pass "block count" instead
of "byte count". Right now it does ..
bh->b_size = 1 << inode->i_blkbits;
call get_block();
First thing get_block() does is
blocks = bh->b_size >> inode->i_blkbits;
All, the unnecessary shifting around for nothing :(
But, I ended up doing "byte count" just to avoid confusion of
asking in "blocks" getting back in "bytes".
I have no problem making b_size as "size_t" to handle 64-bit.
But again, if we are fiddling with buffer_head - may be its time
to look at alternative to "buffer_head" with the information exactly
we need for getblock() ?
Thanks,
Badari
next prev parent reply other threads:[~2006-02-20 23:04 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2006-02-20 21:21 [PATCH 0/3] map multiple blocks in get_block() and mpage_readpages() Badari Pulavarty
2006-02-20 21:23 ` [PATCH 1/3] pass b_size to ->get_block() Badari Pulavarty
2006-02-20 21:23 ` [PATCH 2/3] map multiple blocks for mpage_readpages() Badari Pulavarty
2006-02-23 3:29 ` Suparna Bhattacharya
2006-02-23 22:18 ` Badari Pulavarty
2006-02-20 21:24 ` [PATCH 3/3] remove ->get_blocks() support Badari Pulavarty
2006-02-20 21:59 ` [PATCH 0/3] map multiple blocks in get_block() and mpage_readpages() Nathan Scott
2006-02-20 23:06 ` Badari Pulavarty [this message]
2006-02-20 23:16 ` Nathan Scott
2006-02-21 2:41 ` Jeremy Higdon
2006-02-21 16:03 ` Badari Pulavarty
2006-02-21 21:39 ` Nathan Scott
2006-02-22 15:12 ` christoph
2006-02-22 16:58 ` Badari Pulavarty
2006-02-22 16:59 ` christoph
2006-02-23 1:40 ` Nathan Scott
2006-02-23 1:59 ` Andrew Morton
2006-02-23 16:28 ` [PATCH] change b_size to size_t Badari Pulavarty
2006-02-23 16:32 ` Benjamin LaHaise
2006-02-23 17:20 ` Badari Pulavarty
2006-02-23 17:28 ` Badari Pulavarty
2006-02-23 17:29 ` Benjamin LaHaise
2006-02-23 18:46 ` Badari Pulavarty
2006-02-23 17:40 ` Badari Pulavarty
2006-02-23 16:40 ` Dave Kleikamp
2006-02-22 17:23 ` [PATCH 0/3] map multiple blocks in get_block() and mpage_readpages() Peter Staubach
2006-02-22 18:37 ` Dave Kleikamp
2006-02-22 19:00 ` Badari Pulavarty
2006-02-24 17:19 ` Badari Pulavarty
2006-03-06 10:03 ` Suparna Bhattacharya
2006-03-06 22:39 ` Nathan Scott
2006-03-07 9:00 ` Badari Pulavarty
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1140476772.22756.28.camel@dyn9047017100.beaverton.ibm.com \
--to=pbadari@us.ibm.com \
--cc=akpm@osdl.org \
--cc=hch@lst.de \
--cc=jeremy@sgi.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mcao@us.ibm.com \
--cc=nathans@sgi.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).