Subject: Re: The meaning of scsipi_minphys
To: Manuel Bouyer <email@example.com>
From: Lennart Augustsson <firstname.lastname@example.org>
Date: 03/27/2000 23:32:28
Manuel Bouyer wrote:
> On Mon, Mar 27, 2000 at 01:01:39PM +0200, Lennart Augustsson wrote:
> > What is the purpose of the scsipi_minphys() function pointer in the
> > scsipi_adapter struct?
> > I thought it could be used to give an upper bound on the amount of
> > data that is given to the scsipi_cmd() function.
> > This interpretation is obviously wrong, because I use the minphys
> > function to limit the bp->b_bcount to 16K, but scsipi_cmd() is happily
> > given larger chunks to transfer despite this.
> yes, because for a block I/O scsipi_minphys() is not used.
Which makes scsipi_minphys() pretty useless in my opinion...
> > If this is not the way to limit the size, how can I limit it? If I can't limit
> > it, what is the minimum that a driver has to be able to handle?
> Currently MAXPHYS.
> Some time ago I haked something so that physical drivers could pass this info
> to upper layer to use instead of MAXPHYS (I think this info was stored in
> 'struct disk'), and posted the patch in tech-kern.
> I can't reach mail-index rigth now but I think you should be able tp find
> it here (look for MAXPHYS). This was not especially hard to implement,
> only a few hours work.
Well, I'll make sure to handle MAXPHYS for now.