Date: Thu, 15 Mar 2001 19:14:26 -0500 From: The Babbler <bts@babbleon.org> To: Nate Williams <nate@yogotech.com>, mobile@freebsd.org Subject: Babbleon the idiot / status (was: Bridging with 3C589D-COMBO on 4.2-RELEASE?) Message-ID: <3AB15AE1.5F37390C@babbleon.org> References: <3AAC4C03.13000DE@babbleon.org> <3AAC4E83.2C281B90@babbleon.org> <20010312174852.T20830@speedy.gsinet> <3AAF06E8.103042C6@babbleon.org> <15023.44543.137285.702518@nomad.yogotech.com> <3AB036DB.3CE7E3C3@babbleon.org> <15025.897.571649.16251@nomad.yogotech.com> <3AB11C3B.7578E134@babbleon.org> <15025.7561.937102.558775@nomad.yogotech.com> <3AB12177.6DA79EF1@babbleon.org> <15025.8842.184792.656585@nomad.yogotech.com>
next in thread | previous in thread | raw e-mail | index | archive | help
NON-TECHNICAL STUFF: Ok, first, let me say that most of y'all have the patience of saints. I was getting out of line with the tone of my posts, and WAY out of line with the volume. I posted a straightforward "is this normal" question; the answer was no. That should have been that. One poster took a tone that I found condesending and I started lashing out at everybody. I'm not sure if my interpretation of his posts as condesending was fair to him in the first place, though I find few things quite as annoying as telling somebody that something's wrong and hearing "no, it's fine." But it was CERTAINLY unfair to everybody else. I came across as an idiot who was unwilling to learn anything new, and a pro-Linux advocate. I like to think that I'm neither; indeed, I like almost everything else about FreeBSD better, but the area of PCMCIA/network support has given me trouble. I actually ran FreeBSD back in 1996ish but switched to Linux because of the (at the time unquestionably) superior (as in, extant) PCMCIA support. I expected to have some trouble with this area and that probably colored the way I've seen it. This also makes me an A-1 hypocrite since a major reason I wanted to switch was that I was finding modern Linux distributions (like Mandrake) too automatic. Indeed, with FreeBSD I was able to fairly easily get Windows machines to print via Samba to my printer, a task which had eluded me for months under Linux. The reason I could do that under FreeBSD is because I controlled all the scripts & parameters so I could figure out where the process was breaking down and fix it. Yet here I am complaining over much the same thing w/r/t PCMCIA. Sheesh. Finally, another confession: I didn't actually use *this* network card under Linux. Thus, my comments about relative stability are bogus anyway; maybe the card is problematic for any operating system, for all the more I know. I got this card (3com 3c589D) specifically for FreeBSD after giving up on getting my Linux card working under FreeBSD at all (see J. Rodrigo Refandez M.'s post of 22 January 2001 in this list if you are interested in the details of *that*). I'm using the same card in my other machine, for much the same reason--I was never able to get things properly configured with the exact same cards I used under Linux and this card "seemed" to work better under FreeBSD (and was recommended to me by Brian Dean, my local FreeBSD guru). My guess is that the double-file-loss on crash was coincidence and/or a korganizer bug (it perhaps re-writes the config file every time it starts up, though why it should do such a thing eludes me). I'm not really interested in investigating that one unless it happens again. Three times starts to stetch my ability to explain away things as coincidence. Anyway, I apologize and I hope I can sort of "start over" on the right foot. TECHNICAL STUFF: I'm in the process of upgrading my kernel to -STABLE from -RELEASE, building in debugging, diagnostic, and ddb code, and double-checking my IRQs. I'm having trouble getting the STABLE kernel built, which certainly argues in favor of the thesis that I no longer have any functioning brain cells. However, I'm bugging the poor folks on the freebsd-stable list about that so most of you can take a break in that regard. If I get any information from the kernel debugging I've enabled, I'll post the information back here. IRQ: Guess what! Sure enough, it looks like I have two IRQ conflicts after all. I really *thought* I'd checked this long ago, but I guess I checked these on my other computer and not on this one . . . and one of the conflicts is with the PCMCIA card, sho' nuff. I could swear that I'd checked this long ago, but I now suspect that I'd checked it only on my *other* computer. In practice, I hadn't had any trouble with the 3com card on this computer. The crashes I've had on this computer have involved changes to the bridges and firewall code. Changes which I never even thought to connect with the PCMCIA itself until this thread started. They are with devices I never use, but perhaps they generate interrupts sometimes anyway. IRQ 3: Used by sio1 and the ep0 PCMCIA card. This one might explain a lot. IRQ 5: Used by the USB controller and the sound card. I don't ever use USB devices, and I don't even have usb support built into the kernel. Is this a potential problem anyway, or is it harmless? As for IRQ 3, I've now set up pccard.conf to use IRQ 9 (which is unused) and we'll see what that does for the state of my system. BUT: Still, when all is said & done I find it odd that changing bridge/firewall settting might possibly crash the computer, even if the hardware ethernet settings are silly, when this never happens without bridging or firewalling set. I mean, I'll change it in the hopes that things will be better, but does it make sense as a possible explanation? Nate Williams wrote: > > > > > Do you mean that popping them out might cost reliability, or that the > > > > Linux code that tolerates them better itself makes Linux less reliable? > > > > > > Ahh, the light-bulb is starting to come on now. :) > > > > > > As an example of this kind of behavior, what happens when you pop out a > > > CF memory card or an ATA hard-disk in the middle of writing data to it? > > > > > > Something very bad has happened, because the data has not been > > > completely written to the disk. > > > > Well, yes, of course. And if an ethernet card is in the middle of > > writing packets back things could happen there, too. But if it's > > quiesent and there aren't NFS mounts going across it or anything I > > wouldn't expect to have to run explicit commands. > > You got it. That's why I said it's impossible to do 100% reliability. > However, you can get pretty darn close, and I do hot-swap with my cards > running FreeBSD. (I'm the author of the hot-swap code in FreeBSD, which > was written over 2 years ago, ). > > Minimizing the window of vulnerability is *hard*, but possible. With > all the changes that came about in recent FreeBSD kernels, the window is > bigger. Probably no bigger than in Linux, but certainly bigger than I > was comfortable with. > > In any case, even if the kernel crashes, it should rarely (if ever) > corrupt your hard-disk, unless there is a misconfiguration. > > Nate -- "Brian, the man from babble-on" bts@babbleon.org Brian T. Schellenberger http://www.babbleon.org Support http://www.eff.org. Support decss defendents. Support http://www.programming-freedom.org. Boycott amazon.com. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-mobile" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3AB15AE1.5F37390C>