[7921] in linux-scsi channel archive
Re: request merge changes?
daemon@ATHENA.MIT.EDU (Jens Axboe)
Tue Jan 25 05:55:38 2000
Date: Tue, 25 Jan 2000 03:32:55 +0100
From: Jens Axboe <axboe@suse.de>
To: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: linux-scsi@vger.rutgers.edu
Message-ID: <20000125033255.H3512@suse.de>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <E12CvbO-0007RE-00@the-village.bc.nu>; from alan@lxorguk.ukuu.org.uk on Tue, Jan 25, 2000 at 02:21:05AM +0000
On Tue, Jan 25 2000, Alan Cox wrote:
> > I was relieing on the ll_rw_block layer to correctly handle
> > request merging and this was fine in 2.3.26. I would receive
> > a single clustered request. In 2.3.39 I actually get two requests,
> > one with the first block and a second with the remaining merged.
> >
> > Surely this isn't the intended effect. Comments?
>
> Plugging appears to be broken for the scsi layer. I suspect its bust in 2.2
> as well but 2.3.39 is getting the I/O out faster.
This is actually IDE, I haven't inspected SCSI merging. The error
seems to be at the ll_rw_block level, though.
> I have lots of raid folks bitching because they cant get nice 64K I/O's
> from Linux which is what they all want (that or 128K I/O's). I think its
> all related somewhere, but Im not a scsi layer guru
Now I'm bitching too ;). What really annoys me right now that instead
of getting
rq sector 0 - 127, clustered
rq sector 128 - 255, clustered
for a continous 128KB write, I get
rq sector 0 - 3
rq sector 4 - 131
rq sector 132 - 255
As I want writes to be aligned on a sector boundary of my choice,
this is broken.
--
* Jens Axboe <axboe@suse.de>
* Linux CD-ROM Maintainer
* http://www.kernel.dk
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.rutgers.edu
