Date: Sat, 10 Mar 2007 14:38:35 +0200 From: John Hay <jhay@meraka.org.za> To: Sam Leffler <sam@errno.com> Cc: freebsd-arm@freebsd.org, Olivier Houchard <cognet@freebsd.org> Subject: Re: redboot based boot loader for kernels? Message-ID: <20070310123835.GA83779@zibbi.meraka.csir.co.za> In-Reply-To: <45F19438.5030109@errno.com> References: <20070228102459.GB28669@zibbi.meraka.csir.co.za> <45E5A73E.20503@errno.com> <20070228.094155.660269855.imp@bsdimp.com> <20070305194018.GA73100@zibbi.meraka.csir.co.za> <45ECBA38.7030607@errno.com> <20070306064715.GA3932@zibbi.meraka.csir.co.za> <45F084C5.8020601@errno.com> <20070309064843.GA3384@zibbi.meraka.csir.co.za> <45F19438.5030109@errno.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> > > > Well in effect I only get half of the bytes, but not the first or second > > half. With every 16 bit read I get 2 bytes. But some are missing. So if > > the start of the sector on the disk looks like this: > > > > 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 > > 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 > > > > What I read is this: > > > > 01 02 > > 05 06 > > 09 10 > > 13 14 > > 17 18 > > 21 22 > > 25 26 > > 29 30 > > > > So after 128 reads you start to read zeros because all the data has been > > read. > > So you're reading 4 bytes at a time, swab'ing 'em, then taking only 2. Yip, and now after finishing most of the other stuff in my boot2 program, I decided to look at the 16 bit mode again... And it was working without me doing anything. Hmmm I know I have twiddled some of the compile options in my makefile... Maybe it was that. > > In another email you asked: > > > >> I'm probably wrong and it would explain some issues we had with reset. > >> Feel free to send us a patch. I know I used both the appnote and the > >> linux driver as a basis for avila_ata.c ... > > > > I'll have a look at it as soon as I am happy with my boot2 loader. I'll > > just have to figure out how to map CS2. At the moment only CS1 is mapped, > > but for ALT_STATUS and DRV_CNTRL you need CS2. > > Look in ixp425.c for the CS1 entry. Add CS2 and map it in the driver. > But I'm not sure how to deal with gluing this into the ata driver unless > you can just special case those registers and fetch from the alternate > map'd area. All this is kinda surprising as I did a bunch of tests on > my cf w/o seeing any issues--but I'm certainly no ata expert. Everywhere in the Intel appnote (302456-003) where they describe "True IDE Mode", they say that they implemented it with CS1 and CS2. The register set wrap around on an 8 byte boundary and I think we have just been lucky that a write to 0x1e actually went to the ATA_DRIVE register and didn't cause any problems. But the CF is then not really reset. This probably didn't matter too much because the CF is in a pretty sane state. Something else, appart from the Avila boards, I also have a Pronghorn Metro from Adi Engineering. It is very similar to the Avila software wise. The 2 main differences that I have found is, they use the other serial port for the console and they used CS3 and CS4 for the IDE/CF card. It seems silly to make a whole new kernel config for it. Would it be acceptable if one could detect the board type somehow on startup and use that to select the correct serial port and chip selects for the CF? One last thing, I see that npe0 and npe1 use the same mac address. The uClinux that came installed use 2 different addresses: ixp400_eth: Using MAC address 00:d0:12:02:87:10 for port 0 ixp400_eth: Using MAC address 00:d0:12:12:87:10 for port 1 Not sure where they get it though. John -- John Hay -- John.Hay@meraka.csir.co.za / jhay@FreeBSD.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20070310123835.GA83779>