[1507] in linux-scsi channel archive
Re: elevator sorting for the scsi subsystem.
daemon@ATHENA.MIT.EDU (Andy Poling)
Tue Mar 4 06:20:59 1997
Date: Sun, 2 Mar 1997 23:41:15 -0500 (EST)
From: Andy Poling <andy@realbig.com>
To: "Leonard N. Zubkoff" <lnz@dandelion.com>
cc: Dario_Ballabio@milano.europe.dg.com, linux-scsi@vger.rutgers.edu
In-Reply-To: <199703022044.MAA04228@dandelion.com>
On Sun, 2 Mar 1997, Leonard N. Zubkoff wrote:
> If overlapping I/O's are generated and the driver sorts the requests, then you
> are correct that we could have a problem, but I don't think you can convince
> the kernel to rewrite sector 2 until the first I/O writing sectors 2 and 3 is
> completed. Once an I/O request is made in make_request, the buffer header is
> locked (preventing further requests) until the I/O completes in
> end_scsi_request.
Maybe I'm being naive, but what the heck...
It strikes me that the optimal thing for the kernel to do in this case is to
invalidate the earlier queued write of sector 2 since it is no longer needed
(it has been/will be over-written by the later request).
Am I over-simplifying again? :-)
-Andy
Global Auctions
