Date: Wed, 20 Aug 1997 03:49:04 -0700 (PDT) From: Howard Lew <hlew@www2.shoppersnet.com> To: "Jordan K. Hubbard" <jkh@time.cdrom.com> Cc: dkelly@HiWAAY.net, michael@blueneptune.com, freebsd-questions@FreeBSD.ORG, freebsd-isp@FreeBSD.ORG Subject: Re: Is there still problems with Adaptec UW controllers (fwd) Message-ID: <Pine.BSF.3.91.970820023126.1242A-100000@www2.shoppersnet.com> In-Reply-To: <16256.872068444@time.cdrom.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 20 Aug 1997, Jordan K. Hubbard wrote: > > I have never ran make worlds so I don't know if my cpu works doing that. > > However, just because there are reported problems with make world doesn't > > mean the cpu doesn't work at all with FreeBSD because that simply isn't > > true. In short, the AMD K6-200 runs FreeBSD fine *for almost everything*. > > The sticky point is whether it can/can not do make worlds successfully. > > If you never need to do make worlds, then you don't have any problems. Jordan, If it pleases you, I'll take back that last line above because I can't guarantee (based on what you said) that there won't be other problems. But regardless of that point I was saying there is no problem using the Adaptec 2940UW with the K6-200 as far as I can tell on 2.1.7.1R on "my" system. I have not heard anyone else saying there are other problems. Granted, I don't have 300 users on it like an ISP might, but given that I have several simultaneous users running X on it, there are no problems that I have seen with it and the drives do rattle quite a bit. I think the original question (if I recall correctly) was about whether the Adaptec 2940UW works with 2.2.2R or a RELENG one? I think he said that it worked in 2.1.7 before, but not on the later release. Then someone else said it was likely his cpu and suggested him try another cpu first. Why would it work in 2.1.7 before then? > The load imposed on the CPU by "make world" is not particularly > noteworthy, nor are any special secret x86 instructions executed > during the process which the average user is otherwise protected from. > It is also a virtual certainty that the failures we experienced with > make world on the K6 could be reproduced under other types of > perfectly typical load on a serious web or FTP site - it might not > happen as quickly or be as obvious to the admins when it did (hardly a > feature), but it's a serious risk nonetheless. True it can happen, but running X on it for several users I haven't seen it give any random seg faults. > Your advice is tantamount to telling someone that just because a car > is unable to make left turns, it is hardly a serious issue for people > like yourself who just happen to be able to get everywhere they need > to go by making right turns only. That's all very nice for you, but > back here in the real world where most of the commercial folk live, a > piece of equipment which fails one of its significant acceptance tests > is still considered to be a broken piece of equipment, period. Having taken back that last line above, all I am saying is it works for me. Your mileage may vary. > > A make world failure in testing is considered all the more disturbing > (by those who understand what's going on, anyway) because it tests > "general" system stability rather than focusing conveniently on some > specific driver or system feature which one might conceivably be able > to do without. A failure in this area is indicative of a more general > problem, one which could strike at any time given varying load or > memory usage, and as such it greatly erodes the confidence one is able > to place in that system (to put it mildly). Yes, I agree -- that is a possibility. But until we are certain about the problem should we say that just because he has a K6, he has problems running FreeBSD with a SCSI card? I am sure there are others using a K6 and a 2940UW without problems too. Maybe this guy has a newer rev K6 and has SCSI trouble? Do we not help him because he is a K6 owner and just use the cpu as the scapegoat for any kind of problems with FreeBSD? I don't think it is fair for him to be treated that way. I know there are some FreeBSD users with K6s that feel bad about their chip, but for now there isn't much that can be done except to wait for the puzzle to unravel given some time. I would hate to see FreeBSD become just a genuine Intel only OS and not support any user who doesn't own an Intel chip. > > In any case, it also does not appear that you've tested your system to > any significant degree, chosing instead to run a single and somewhat > obsolete version of the OS on it (where are your 2.2.x tests, for > example?) and in situations which do not appear to exert much strain > on the system at all. If you understand anything at all about > testing, you'll know that it involves placing the maximum projected > strain on something things before pronouncing it fit for duty, not the > minimum strain. Such a testing methodology would only be a recipe for > building bridges which fall down during actual use and of interest > purely to fools and masochists. > Harsh language I must say... This box running 2.1.7.1R is on a production machine for X11R6 that I really can't have go down. It's the only FreeBSD machine with the 2940UW, but is running 2.1.7.1R. I do have a 2.2.2 RELENG and a 3.0 SNAP testing with a new Cyrix 6x86MX chip at the moment. I have another 2.2.2 RELENG machine running on a Pentium 133. So yes, I do try to keep up with newer releases, but I really don't like taking down a perfectly good important system and upgrading to find out that the patchwork upgrade breaks certain things where a complete reinstall would not.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.91.970820023126.1242A-100000>