Date: Wed, 10 Jul 1996 10:19:40 -0400 From: "Clement T. Cole" <clemc@echo.CCC.COM> To: jkh@cdrom.com Subject: FreeBSD 2.1 and Adaptec SCSI Strangeness Message-ID: <9607101019.aa00497@echo.ccc.com.> Resent-Message-ID: <7528.837072481@time.cdrom.com>
next in thread | raw e-mail | index | archive | help
Has anybody else seen this type of situtation? Is there helpf or me? At Locus, we have been running FreeBSD 2.1 on a PCI based Pentium, w/16 megs of memory and an Adaptec 2740 SCSI -- which all installed and runs fine (but failed on an older Gateway with an ESDI controller - Bruce and I worked through that. In fact, I switched to the SCSI to fix the ESDI problems). At home I have been running 2.0.5 on a 386 ``production'' system and have run it off and on on my 486 (which is my ``blow up system).'' Anyway -- the important data point is that on two >>PCI systems<< with an Adaptec, FreeBSD 2.1 has been shown to run fine. And the 2.0.5 runs fine on my two ISA (x86) base systems at home... I'm on the subscription service and have had the 2.1 CD, but until now no time to convert. When over the 4th, I finally had a litte free time so... so I wanted to upgrade the 2.0.5 system at my house. They are: 1.) 386/40 16 Megs (generic MB), Adaptec 1542B SCSI 2.) 486/33 16 Megs (Micronic ``Geminii'' MB), Adaptec 1542B SCSI Remember 386 is running FreeBSD 2.05 no sweat (it what I composed this message on). I try to use the 2.1 update scheme -- and I get install errors -- disk corruption (like SCSI driver problems).... (so I rolled back to 2.05 via backup tape and I'm fine for now but..). I said --- well Jordon told that the upate stuff was still being debugged, so I'll try a different path. I said -- let's try a fresh install, but I'll borrow another CPU (which runs a different OS normaly) and do the install on a new disk, and then do a disk swap So... I go over to the 486 which has run: 1.) SVR3 2.) Win/311 3.) Win95 4.) Win/NT 5.) FreeBSD 2.0.5 I disconnect my system disk, and take a new Hitachi 2 Gig disk and install DOS/WIN in the first 300 Megs. I run diagnostics. Hardware is very happy. I even install NT on the system to push the hardware configuration. Note: I had cut the SCSI chain down to: Adaptec @ 7, Disk @ 0, and Toshiba CD-ROM @ 6. So I reformat the disk again, and reinstall DOS/WIN at the bottom. Then Boot DOS, and run the FreeBSD 2.1 install in ``Novice'' mode. And.... I get failures consistantly, during the cpio of the floppies/root.cpio area. So... I borrow a different copy of the 2.1 CD from work and try again, but get the same type of failures at a different place. So... I make a boot floppy. This is worse. I hang during boot in the SCSI driver with the SCSI access LED stuck on (not I boot from CD fine mind you...) .... ahb0 not found aha0: aha_cmd, cmd/data port full aha0: at 0x330-0x333 irq 11 drq 5 on isa aha0 waiting for scsi devices to settle (aha0:0:0) timed out adapter not taking commands.. frozen?? Debugger("aha1542") called. (aha0:0:0) timed out adapter not taking commands.. frozen?? Debugger("aha1542") called. AGAIN aha0: MBO 02 and not 00 (free) <-- this repeats many times before I kill it --> So... I think -- wait a minute this used to work. So for grins, I'll try the FreeBSD 2.0.5 CD's. Well - It works fine on this hardware (just like the 386). I even set up a disk exsercisor that I had make sure. And yes it works all under 2.0.5. So... I conclude something changed in the 2.1 SCSI driver Other interesting tid bit... the 386 system has run Linux no sweat. One reason I became FreeBSD fan, was that this type of hang was typically of Linux on the 486. I later found out that the Linux had some over lapped I/O problems with drivers that performed true (so called ``first party'' DMA). I gave up and switched to FreeBSD which worked (just like AT&T UNIX did). But sadly - it looks like we picked up the a problem similar to Linux. So... my questions: What changed? Anybody else see this? Was this ``fixed'' in 2.1x? and... What do you think should be my next step? Thanks again, Clem PS One other minor bug in sysinstall. The auto size code for the FreeBSD partitions/slice stuff gets the -u parameter to newfs wrong consistantly for the last slice you have (/var or /usr which ever). The number seems calculated, not random, but wrong none the less. / and all other UFS's before the last one get a -u param of 4096. But the last UFS slice is a strange number. Note: I have seen this on all 2.1 installs at LCC and meant to tell you about it, and frankly I forgot. If you watch the logs of the newfs (on ttyv1) you'll see an error go past.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9607101019.aa00497>