From owner-cvs-sys Wed Jan 17 08:03:07 1996 Return-Path: owner-cvs-sys Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id IAA18221 for cvs-sys-outgoing; Wed, 17 Jan 1996 08:03:07 -0800 (PST) Received: from Sysiphos (Sysiphos.MI.Uni-Koeln.DE [134.95.212.10]) by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id IAA18173 Wed, 17 Jan 1996 08:01:40 -0800 (PST) Received: by Sysiphos id AA08461 (5.67b/IDA-1.5); Wed, 17 Jan 1996 17:01:35 +0100 Message-Id: <199601171601.AA08461@Sysiphos> From: se@zpr.uni-koeln.de (Stefan Esser) Date: Wed, 17 Jan 1996 17:01:34 +0100 In-Reply-To: "Justin T. Gibbs" "Re: cvs commit: src/sys/pci ncr.c" (Jan 16, 21:30) X-Mailer: Mail User's Shell (7.2.6 alpha(2) 7/9/95) To: "Justin T. Gibbs" Subject: Re: cvs commit: src/sys/pci ncr.c Cc: CVS-committers@freefall.freebsd.org, cvs-sys@freefall.freebsd.org Sender: owner-cvs-sys@FreeBSD.ORG Precedence: bulk On Jan 16, 21:30, "Justin T. Gibbs" wrote: } Subject: Re: cvs commit: src/sys/pci ncr.c } >se 96/01/14 16:10:22 } > } > Modified: sys/pci ncr.c } > Log: } > Make the NCR driver on the boot floppy kernel not negotiate synch. SCSI } > transfers for CDROM drives. } > } > Revision Changes Path } > 1.57 +3 -3 src/sys/pci/ncr.c } } Does config still add a -DKERNEL_NAME to the build process? AFAIK it still does for 2.1, but doesn't do it anymore for -current. I want to suggest some "official" way to make a kernel abstain from trying generally useful performance obtimizations, which might fail in very rare situations, though. There should be a kernel variable (like bootverbose), whose default value should be set by a kernel CONFIG OPTION (e.g. using options FAILSAVEMODE or something like that) and should be made available as a BOOT OPTION (is -f already used ?). Install kernels must work on as much (even buggy) hardware as possible, while the kernel run everyday should be able to make best use of available performance features, IMHO. All INSTALL kernels should default to failsave, and I'd make this have the following effects in drivers I'm working on: 1) switch off NCR tagged commands (failed for 2 drives in the last 20 months) 2) disable synch. negotiation requests for CDROM drives (which might be a show stopper if the drive else does lock up) 3) PCI might use shorter burst length or other performance options that might make an install fail could be disabled I don't know which other drivers might be able to take advantage of such a feature. Does this sound reasonable ? Regards, STefan -- Stefan Esser, Zentrum fuer Paralleles Rechnen Tel: +49 221 4706021 Universitaet zu Koeln, Weyertal 80, 50931 Koeln FAX: +49 221 4705160 ============================================================================== http://www.zpr.uni-koeln.de/~se