Subject: Re: MAXBSIZE Was Re: bread() returns EINVAL during vfs mount
To: Thor Lancelot Simon <tls@rek.tjls.com>
From: Reinoud Zandijk <reinoud@netbsd.org>
List: tech-kern
Date: 08/27/2005 22:17:02
--ZPt4rx8FFjLCG7dd
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline

On Sat, Aug 27, 2005 at 12:13:03PM -0400, Thor Lancelot Simon wrote:
> On Sat, Aug 27, 2005 at 05:56:44PM +0200, Reinoud Zandijk wrote:
> > 
> > I wonder how i can implement sector sizes bigger than MAXBSIZE (normally 
> > around 4k AFAIK) with normal bread and bwrite calls though.
> 
> I'm sorry, what?  I am unaware of any port that has a MAXBSIZE smaller
> than 31K.

really? now where did i read this then.... a scout and a printf() tells me 
you're right though i wonder where i read this was `just' 4kb... 
interesting. Is NetBSD different from other BSD's in this? or was it a 
linux thing i read... *sigh* i need to `grep' more and read less.

> You can hand device drivers I/O requests of up to MAXPHYS in size (in
> practice, you can hand them larger requests, too, but don't do that).
> See the FFS code for examples.

more than MAXPHYS prolly won't help much as the SCSI layer will prolly 
object :)

thanks,
Reinoud


--ZPt4rx8FFjLCG7dd
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.6 (NetBSD)

iQEVAwUBQxDKMIKcNwBDyKpoAQJ6Dwf6A1ojM2UCwRhfx0tAKjI+qYlz1IPDB0es
jjx60G+ufMLJiJ7noxMV6zN6xp+2aDCyDA9XWanSiGK9DmbNhCtqXEBR8BLU/xlh
5IRAwwjF3SuxH2keFjApExy40PsKlFiEtQUe7DP0fy8SdehZKaXCbQT+MZeXDg1n
gIndPssTFat+/zla7L++YGiyI19NkR1gFcSaIyW0Recq3nPPZLKVlO6OGEC3fiG4
jhsiENqvJHHkXtVW46jqILarcKPSBMOyaSHX2BcmRIGJroheP+D1zUgrLFmANYce
P89m/pAc2HOMdG4ENP7SWbrBELcgNRaWvJRrJaZUvU/ky78cJqLHqQ==
=sbju
-----END PGP SIGNATURE-----

--ZPt4rx8FFjLCG7dd--