Return to BSD News archive
Path: sserve!newshost.anu.edu.au!munnari.oz.au!constellation!osuunx.ucc.okstate.edu!newsfeed.ksu.ksu.edu!moe.ksu.ksu.edu!vixen.cso.uiuc.edu!howland.reston.ans.net!xlink.net!zib-berlin.de!unidui!rrz.uni-koeln.de!RRZ.Uni-Koeln.DE!RRZ.Uni-Koeln.DE!news From: se@fileserv1.MI.Uni-Koeln.DE (Stefan Esser) Newsgroups: comp.os.386bsd.questions Subject: Re: Any experience with ASUS PCI SCSI controller? Date: 9 Jun 1994 01:38:23 GMT Organization: Institute for Mathematics, University of Cologne, Germany Lines: 71 Distribution: world Message-ID: <2t5rqfINN171r@rs1.rrz.Uni-Koeln.DE> References: <1994Jun4.135711.12888@hellgate.utah.edu> <2t0n6r$23s@huey.cc.utexas.edu> NNTP-Posting-Host: fileserv1.mi.uni-koeln.de Keywords: PCI SCSI NCR-53c810 |> ASUS is using the NCR 53c810 SCSI chip on their PCI boards. It |> is very fast but as of yet there are no drivers for it. Persons |> on the FreeBSD core team are working on writing one. In addition |> someone is almost finished w/ one for 386bsd which could easily be |> ported over to FreeBSD. Until then BSD won't work w/ the ASUS PCI |> SCSI cont. Sorry, I've been wanting to get one too. Since I've been involved in the development of this driver (or at least such *a* driver :), I'll comment on its release state: The driver is for real, and it has been successfully installed on a number of systems. (See below, if you want to try it *now*). It has been posted to a group of beta testers, and there have been reports of quite impressive throughput, but there is still room for further improvement (less cpu load). Eg. completely reading a DEC RZ25L (535MB) with 'dd if=/dev/rsd0d of=/dev/null bs=64K' => 3MB/s sustained. We are waiting for the delivery of a new fast drive (>4MB/s) to be used for tuning and the test of advanced features (Tagged Command Queues). It compiles (at least) under 386BSD-0.1+patches and FreeBSD-1.1, and doesn't require any changes to Julian Elischer's generic SCSI driver (ie. its just a new chip level interface and thus can drive all devices supported by Elischer's code, not only disk drives). We didn't try to integrate it into NetBSD, but if there are no big differences in the VM system (functions to map the PCI bus and NCR registers into KVM and locking of buffer pages), it shouldn't be much of a problem (... famous last words :). There isn't support for synch. SCSI, yet (will be a matter of a few hours of hacking, but we didn't consider it important for the first release, it doesn't make much of a difference, given a system with a single fast 3.5" drive or two ...). If interoperatibility has been shown over a wide range of drives, than synch. SCSI will be provided and problems that show up thereafter may be attributed to synch. SCSI. But for the time being we'd like to exclude this possibility. Tagged command queing is planned (and has been considered an important design goal from the beginning), but will not be officially supported for some time, since we don't generally trust all drives to implement it correctly. Single interrupt per transfer (ie. no host interaction for disconnect/ reconnect required) is supported by use of the NCR 53c810 SCRIPTS language (well, there seems to be a chip bug (in selection time out), which is easy to work around by using a second interrupt, but we'll get rid of this in the release version). CPU load is less than 10% per 1MB/s on a 486DX2/66. This will allow such a system to fully saturate the 10MB/s FAST SCSI bus, given the right (number/type of) drives. If you want to become a beta tester (and know that this doesn't come for free, you'll have to install a few files in your source directory, build a new kernel, and hope it won't blow up your system on first boot :), then send a request, stating your system type (CPU + OS) and the disk drive model you intend to use, either to Wolfgang Stanglmeier <wolf@dentaro.GUN.de> or reply to my address. There isn't a boot floppy, yet. You need a working 386BSD or FreeBSD system to build the kernel and to install it on your SCSI disk, if you intend to make it your boot disk. (If there is any interest, I'll make a generic kernel binary with PCI support available via anon. FTP). -- Stefan Esser Internet: <se@MI.Uni-Koeln.DE> Mathematisches Institut Tel: +49 221 4706010 Universitaet zu Koeln FAX: +49 221 4705160 Weyertal 80 50931 Koeln