Date: Wed, 3 Apr 1996 15:03:11 -0700 (MST) From: Terry Lambert <terry@lambert.org> To: mikko@dynas.se (Mikko Tyolajarvi) Cc: freebsd-questions@FreeBSD.ORG Subject: Re: Intel Endeavor install problems (639K != 640K -> SIG 11) Message-ID: <199604032203.PAA20092@phaeton.artisoft.com> In-Reply-To: <4judc4$cb1@spirit.dynas.se> from "Mikko Tyolajarvi" at Apr 3, 96 07:43:32 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> Some background, > > We've got some machines with with the Intel Endeavor (Advanced/EV) > motherboard, Pentium 90 or 100, 16 or 32 MB of memory and various ~1GB I= > DE > disks. > > When booting, all report: > > =09BIOS basemem (639K) !=3D RTC basemem (640K) > > This flashes by very quickly, and I wouldn't even have noticed, if > the machine I was installing on had not failed to install. It gets > as far as ftp-ing the bin distribution (from a local machine with > a CD drive), then dies on a SIG 11. [ ... ] > Looking at /sys/i386/i386/machdep.c, I can locate the source of the > error message: biosboot (or is that bootbios?) locates 639/xxx K > memory, but the kernes uses something it finds in RTC CMOS RAM, > and reports the discrepancy. > > But not knowing (not wanting to know :-) enough about PC BIOSes and > such, I don't really know what it implies, or where the failing 1K > has gone. > > > > Questions: > > What does the error message mean? Is it related to the install failure? It is unrelated. The message means one of three things: 1) The CMOS memory is stored using option base 0 instead of option base 1. This usually results in "639K" reported. This is typical of AT&T 6386/WGS. The original 386BSD 0.1 blew the memory calcualtion following that and lost its mind. This error has its genesis in the first patch I ever put in the patch kit, though the code isn't derivative any more. The warning is there to indicate that an assumption (which may not be correct) is being made. 2) Something is eating some amount of memory. Generally, this is 4K, and results in "636K" reported. The HP Vectra, specifically, *also* lost its mind on 386BSD 0.1. It stole the 4k, and created an "adjustment hole" to store the drive type 44 and drive type 45 C/H/S values for configurable drives, copied from CMOS. Some boot sector viruses also protect themselve the same way, but obvious could affect the second stage boot, but not FreeBSD itself. 3) There is a memory problem, like the 128k from 512k to 640k being "not there" (generally required a "back fill" via a memory extender board on very old Intel 386 hardware). This will be obvious, since a power of 2 of some kind, smaller than 640k, will be reported. In any case, BSD loads above 1M, so this should not cause a boot failure. > Is it harmful after installation? I.e. once the machine runs, will > it continue to run? (It has so far...) It is never harmful; it is only a diagnostic, and can generally be ignored. > I will probably want to put FreeBSD on more of these machines, > how do I install on them without resorting to moving the disk > to another machine? Probably the install problem is related to the compressed kernel image. I believe newer snaps (ie: after the 2.1R release) deal with the problem. Generally, it's indicative of a motherboard cache coherency problem (apparently, we've invented the only working diagnostic in the industry 8-)). Turning of the cache during the install should fix the problem, after which you can turn it back on; if there is really a cache problem (it may in fact be related to instruction cache writeback, which was not supported prior to the P5), then you will see additional problems, and you will need to get an ECO board upgrade, or disable writeback, or the type of caching that causes the problem for your particular chipset. Most likely, you won't see any problems. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199604032203.PAA20092>