RE: [SLE] [RAM] Test memory
At a theoretical level, Linux can map bad blocks of memory out of the way and use good blocks. I remember reading about a set of kernel patches that were written by someone who was exploring the idea of selling 'faulty' RAM (which is a lot cheaper) and mapping out the bad bits. Since RAM has come down in price so much over the last 2 years, I suspect that this idea has been dropped. Richard
-----Original Message----- From: John McNulty [SMTP:john@jmtl.com] Sent: 23 October 2001 01:36 To: David A. Riggs; suse-linux-e@suse.com Subject: Re: [SLE] [RAM] Test memory
On Tuesday 23 October 2001 1:17 am, David A. Riggs wrote:
Could you point me towards some information on excluding bad blocks of RAM? Is this something that can only be setup in BIOS, or is it possible to specify at boot? I was under the impression that a dud
stick
was destined for keychain duty.
you're quite correct ! keychain duty it is. Remapping bad blocks is a repair function that applies to disks, not memory.
-- To unsubscribe send e-mail to suse-linux-e-unsubscribe@suse.com For additional commands send e-mail to suse-linux-e-help@suse.com
Also check the FAQ at http://www.suse.com/support/faq and the archives at http://lists.suse.com
There is a patch, the BadRAM patch. It had nothing to do with trying to sell faulty RAM though. Read about it on: http://rick.vanrein.org/linux/badram Mazzel, Marcel On Tuesday 23 October 2001 14:57, you wrote:
At a theoretical level, Linux can map bad blocks of memory out of the way and use good blocks. I remember reading about a set of kernel patches that were written by someone who was exploring the idea of selling 'faulty' RAM (which is a lot cheaper) and mapping out the bad bits.
Since RAM has come down in price so much over the last 2 years, I suspect that this idea has been dropped.
Richard
-----Original Message----- From: John McNulty [SMTP:john@jmtl.com] Sent: 23 October 2001 01:36 To: David A. Riggs; suse-linux-e@suse.com Subject: Re: [SLE] [RAM] Test memory
On Tuesday 23 October 2001 1:17 am, David A. Riggs wrote:
Could you point me towards some information on excluding bad blocks of RAM? Is this something that can only be setup in BIOS, or is it possible to specify at boot? I was under the impression that a dud
stick
was destined for keychain duty.
you're quite correct ! keychain duty it is. Remapping bad blocks is a repair function that applies to disks, not memory.
-- To unsubscribe send e-mail to suse-linux-e-unsubscribe@suse.com For additional commands send e-mail to suse-linux-e-help@suse.com
Also check the FAQ at http://www.suse.com/support/faq and the archives at http://lists.suse.com
On Tuesday 23 October 2001 21:57, you wrote:
At a theoretical level, Linux can map bad blocks of memory out of the way and use good blocks. I remember reading about a set of kernel patches that were written by someone who was exploring the idea of selling 'faulty' RAM (which is a lot cheaper) and mapping out the bad bits.
Since RAM has come down in price so much over the last 2 years, I suspect that this idea has been dropped.
Richard
I'm thinking that, as Rick van Rein's site ( http://home.zonnet.nl/vanrein/badram/ ) states, memtest would let a person use RAM that became faulty due to manufacturer defects AND damaged due to static discharge. The average linux guy can use this to actually make the bad RAM effective rather than spend more money on functional RAM. For instance, about 2 years ago, I bought RAM over the internet and when I received it and put it in my machine it kept faulting. The retailer wouldn't accept a return so I was stuck with bad RAM that I hadn't even had the chance to use! Using memtest, I can now map out the bad parts of the RAM and get it to work with my machine. This is cool but I'm thinking it will work only with Linux unless the bad parts can be locked out physically (can this be done?) so it can work with Windows or other OSs. The thing I don't understand about the above site is that the author is trying to sell an idea: selling faulty' RAM (which is a lot cheaper) and mapping out the bad bits. That's not bad in itself, but how is he going to map out bad bits of RAM to work with a system other than Linux? By what I read, the mapping will be done with Linux-specific software. What if this idea of his is implemented and someone with a dualboot Linux/W98 system wants to use this "cheaper" RAM? They probably won't be able to. The RAM would have to be labeled "Linux compatible only" OR his memory mapping software would have to be ported to Windows. He stated that he's not going to attempt it because Windows isn't open-source and isn't an efficient OS to begin with. I'm no programmer. I've attempted very rudimentary coding and that $hit is hard. I'll give him props for his cool ideas and his memory mapping. He's entitled to his own opinions and can do whatever he likes with whatever OS he wants. I liked his info on his site but once I read about his Windows philosophy, I pretty much stopped reading. There's no need to adopt the same attitude as Microsoft's lackeys in regards to programming. The thing I hate most of the Linux movement is that most are so into this open-source kick that they crap on other OSs for the simple fact that closed-source programming is more difficult. The same thing happened with winmodems and the Linux community and only now are you seeing progress with winmodems...not because hardware developers are finally releasing the code bue because some programmers are putting their spiteful attitudes aside. I've lost count of the number of 'net conversations that state, "my modem won't work with Linux/you have a winmodem -- get a REAL modem." It's funny...hackers will hack almost anything that's hackable but a winmodem, for simple philosophical reasons. Almost everything about Linux was hacked by Unix/Linux programmers...if everyone one of them had the same attitude that many have about winmodems, there would be no Linux (or, Linux would be a real POS). Sure, Linux is open-source but adopting bad attitudes like the above amounts to closed source...it limits the Linux OS as a whole. I'm just putting out my thoughts on the memory test issue and my take on the Linux movement. Keep in mind, all of what you just read is meant to be related. I'm probably missing some things that others, with their vast knowledge of coding and experience with Linux, will want to add. The above isn't meant as a flame. I'm just putting out my opinion and willing to see other points-of-view. Regards. -- Ron Sinclair @ http://www.wigglit.com
--- Ron Sinclair
Since RAM has come down in price so much over the last 2 years, I suspect that this idea has been dropped.
Not entirely. These tools were written for and by people who spend WAAY more than $40 on a DIMM.
This is cool but I'm thinking it will work only with Linux unless
Yes. As is well-acknowledged in the web sites related to these tools. Note many references to the beauty of Free Software.
the bad parts can be locked out physically (can this be done?) so
No.
it can work with Windows or other OSs.
Outside of the intended scope of application. See how the author heavily chides Microsoft and others for not being friendly to these ideas...
The thing I don't understand about the above site is that the author is trying to sell an idea: selling faulty' RAM (which is a lot cheaper) and mapping out the bad bits.
No. You've missed the point entirely. Being the service guy at a large vendor of computational clusters, I once recieved a plain but heavy box, a little bigger than that which your personal banking checks are mailed to your house in (about half the size of your average software package, for those that don't use checks or cheques?). It was packed to the bursting point with 256MB, ECC DIMMs that somebody, at some point, had paid nearly a thousand dollars a stick for. Each one of those DIMMs had been tested as "bad" by crude means, and been shipped back to me for replacement. This box landed on my desk merely hours before the badmem patches were announced on /. or some place like that, and I had half a mind to find out exactly which ones of these DIMMs could be put to use in less important roles.
What if this idea of his is implemented and someone with a dualboot Linux/W98 system wants to use this "cheaper" RAM?
This was never the intent. American profiteering wins again: a perfectly good idea spoiled by somebody trying to make a buck, or at least accusing someone else of it...
I liked his info on his site but once I read about his Windows philosophy, I pretty much stopped reading.
This is a new addition to me...I must see what you're talking about.
There's no need to adopt the same attitude as Microsoft's lackeys in regards to programming.
Whoa, whoa, whoa! Back this truck up. Microsoft has some of the best programmers in the world; I know a few. The problem with Microsoft is not the programmers, but the marketers and the support staff. The marketers hear that the programmers have something neat, develop a branding/packaging scheme, and rip it from the hands of the programmers before it's ready for sale. The programmers get stuck in a never-ending cycle of fixing problems that never should have seen the light of day, all the while the original codebase, concepts, and intended application are entirely forgotten. The consumer support division, on the other hand, *does* hire general morons. I digress.
The same thing happened with winmodems and the Linux community and only now are you seeing progress with winmodems...not because hardware developers are finally releasing the code bue because some programmers are putting their spiteful attitudes aside.
No, I think this is largely due to the fact that Lucent, Connexant, PCTel, and the others finally realize that HSP is not a new idea, everybody else (the other chipmakers) already knows how to do it, and are releasing hardware specs. Lucent and Connexant wrote the first HSP-for-Linux drivers in-house, if you want proof. Without specifications, a winmodem truly is useless; you can't debug a circuit if you can't turn it on. Now that the code is available, other people are trying to make it better without prejudice...if the code were free from the start, it would have always been this way. For the record, I'm still staunchly against using the CPU to make beeps and warbles on a phone line for communication; this is a job best left to a dedicated circuit. Using an HSP device for grander things can be cool (like a PBX system), but gamers and road warriors would largely be better off without the technology. I'm all about saving money, but there are some corners you just shouldn't cut...
I'm just putting out my thoughts on the memory test issue and my take on the Linux movement.
Thanks for your input, but I've been in the user trenches of Linux for seven years now. Movement? What movement? You mean that marketing fad two years ago? ;) I'd still be using Linux, even if it weren't, or were never, hip. ===== -- -=|JP|=- Hit me! - http://www.xanga.com/cowboydren/ Jon Pennington | Debian 2.3 -o) cowboydren @ yahoo . com | Auto Enthusiast /\\ Kansas City, MO, USA | ICQ UIN 69 67 29 31 _\_V __________________________________________________ Do You Yahoo!? Make a great connection at Yahoo! Personals. http://personals.yahoo.com
On Wednesday 24 October 2001 13:00, Jon Pennington wrote:
--- Ron Sinclair
wrote: Since RAM has come down in price so much over the last 2 years, I suspect that this idea has been dropped.
Not entirely. These tools were written for and by people who spend WAAY more than $40 on a DIMM.
Previous poster before me wrote that, not I. I just included it with my own text, as reference. :o)
it can work with Windows or other OSs.
Outside of the intended scope of application. See how the author heavily chides Microsoft and others for not being friendly to these ideas...
Microsoft may be closed-source but there's a hell of a lot of shareware and freeware out there for that OS. The OS is unstable, as the author states, but that's a copout, IMO. The OS itself may be unstable, yes, but there's some awesome software that will run on this unstable OS (sim games like Falcon 4.0, Quake III, Flanker 2.0....). Just because the development of software for Microsoft may be difficult doesn't mean you can't create code that will get the job done.
The thing I don't understand about the above site is that the author is trying to sell an idea: selling faulty' RAM (which is a lot cheaper) and mapping out the bad bits.
No. You've missed the point entirely. Being the service guy at a large vendor of computational clusters, I once recieved a plain but heavy box, a little bigger than that which your personal banking checks are mailed to your house in (about half the size of your average software package, for those that don't use checks or cheques?). It was packed to the bursting point with 256MB, ECC DIMMs that somebody, at some point, had paid nearly a thousand dollars a stick for. Each one of those DIMMs had been tested as "bad" by crude means, and been shipped back to me for replacement. This box landed on my desk merely hours before the badmem patches were announced on /. or some place like that, and I had half a mind to find out exactly which ones of these DIMMs could be put to use in less important roles.
I got the point, but what I was talking about what the parts that you didn't include. His idea was to sell faulty RAM after mapping out the bad parts but how is one to sell it if it won't work on non-Linux systems? He'd have to label the product "for Linux use only." To people in the corporate environment running Linux, this may cut costs, but then again, costs could be cut by just buying cheaper RAM in the first place instead of $100 clips. I see this as benefiting those in the home and educational environments but not corporate. Does Linux-specific hardware sell well? Oh yeah....how would a company go about this? They'd sell the patch (which contains the remapping) AND the chip? You said the chip couldn't be hardcoded. The company would have to test the chip for faults, locate them, then use code to go around the bad part(s), but once you remove that chip and put it in another Linux machine, would that same remapping code go with the relocated chip?
What if this idea of his is implemented and someone with a dualboot Linux/W98 system wants to use this "cheaper" RAM?
This was never the intent. American profiteering wins again: a perfectly good idea spoiled by somebody trying to make a buck, or at least accusing someone else of it...
Dunno about that. He did offer to post any Windows code that anyone hacked. Paraphrasing here but what I saw on his pages seemed to say, "I'm not willing to port this procedure to Microsoft OSs but if someone else does, I'll be willing to post the info on my page." He offered, so there's at least SOME intent, just no intent on HIS behalf. :o)
I liked his info on his site but once I read about his Windows philosophy, I pretty much stopped reading.
This is a new addition to me...I must see what you're talking about.
I'm talking about the part where he chides MS about it's closed OS and unstability. Just seems to be a lame excuse. People develop for both OSs all the time (think VMWare). Software like CuteFTP was probably hard to design when considering that W95/98/2000 is closed-source but that particular program is very solid. I don't know what the CuteFTP creator's Windows philosophy is and I'm sure he may have run into problems when designing his product but because he prevailed, he's now got one hell of a product. Most of us on this list know of MS's faults and we have our negative opinions on the company and its OS. I just think Rick should have said that to port his code to Windows would have went beyond the scope of what he was doing. I was at his pages to read what was going on with his BadRAM project, not get a philosophy lesson. It's his site, yeah, but it helps to be a bit less subjective...everything else on his site was very much objective.
There's no need to adopt the same attitude as Microsoft's lackeys in regards to programming.
Whoa, whoa, whoa! Back this truck up. Microsoft has some of the best programmers in the world; I know a few. The problem with Microsoft is not the programmers, but the marketers and the support staff. The marketers hear that the programmers have something neat, develop a branding/packaging scheme, and rip it from the hands of the programmers before it's ready for sale. The programmers get stuck in a never-ending cycle of fixing problems that never should have seen the light of day, all the while the original codebase, concepts, and intended application are entirely forgotten. The consumer support division, on the other hand, *does* hire general morons. I digress.
The marketers and support staff for Microsoft have nothing directly to do with BSODs and CTDs. It's mostly software issues (and maybe hardware, depending on the case, I guess) that get you blue screens of death and crashes to desktop. Bottom line is that the product isn't ready to sell because the code isn't finished. Taking what you said above, I'm sure this happens with Linux products too. When the stable 2.4.0 kernel was released by Linus, many mentioned that it was too early yet, but Linus said that if it wasn't ready, it would never be, then he released it. The thing is, when things crash in Linux, the whole box doesn't go down. Some things still work. In the MS case, how many times have you seen a Windows box semi-functional...its either up or down. W98 has been around for some time now and even with the most current OS patches, its NEVER stable. How many years have MS's programmers had to make it stable, marketing department aside? Do marketers and support staff delve with patches? Probably not, or at least not at the same extent as when the OS is released when new.
For the record, I'm still staunchly against using the CPU to make beeps and warbles on a phone line for communication; this is a job best left to a dedicated circuit. Using an HSP device for grander things can be cool (like a PBX system), but gamers and road warriors would largely be better off without the technology. I'm all about saving money, but there are some corners you just shouldn't cut...
True. Then again, running a winmodem with a 1.5GHz CPU should be VERY trivial. The CPU cycles used to run an controllerless modem on, say, a 400MHz system would be needed elswhere but when you have a CPU that's 3 times as powerful, theres no REAL grounds for complaint.
I'm just putting out my thoughts on the memory test issue and my take on the Linux movement.
Thanks for your input, but I've been in the user trenches of Linux for seven years now. Movement? What movement? You mean that marketing fad two years ago? ;) I'd still be using Linux, even if it weren't, or were never, hip.
I've been using Linux since around 1997. Not entrenched but learning both Linux and Windows. I'm just now getting to the point to where I'm comfortable using a Linux-only system, but I'm only doing it now because I have no choice (hard disk with Windows crashed). When I say Linux movement, I'm not talking about it just being hip. I'm talking about all the Stallman wannabes out there that swear their mothers' souls to Linux. The OS is very good, but I have a life and like games and sims....not too many sims for Linux, and even then, I have a hard enough time installing smaller software. I can't imaging installing a flight sim in Linux...I can imagine lib errors and such. But anyways, IMO, there are things other than Linux. Liked how I played Devil's advocate in this post? :o) Laters! -- Ron Sinclair @ http://www.wigglit.com
--- Ron Sinclair
On Wednesday 24 October 2001 13:00, Jon Pennington wrote:
Not entirely. These tools were written for and by people who spend WAAY more than $40 on a DIMM.
Previous poster before me wrote that, not I. I just included it with my own text, as reference. :o)
And I referenced your reference! ;)
Microsoft may be closed-source but there's a hell of a lot of shareware and freeware out there for that OS.
Unlike tools such as LiteStep (http://newer.litestep.com/) and 98lite (http://www.98lite.net/), in order for an incarnation of the Linux badram patch to work, you'd have to have access to the kernel. I can hear it now: "Microsoft Tech Support, this is Jim, may I have your card number and expiration date, please... Thank you. May I have your product key please... Thank you. How can I help you, sir? No, we don't have any popcorn, uh, but there should be a patch soon..."
I got the point, but what I was talking about what the parts that you didn't include. His idea was to sell faulty RAM after mapping out the bad parts but how is one to sell it if it won't work on non- Linux systems?
No, dammit! :D *HE* isn't selling hardware! It's an *IDEA*!!! A possible implementation of a good thing gone bad turned good again! You make it sound as if he's profiteering on innocent consumers looking for a deal. (I'm keeping this conversation very friendly, btw; chuckling doesn't translate well into email, I'm afraid...)
You said the chip couldn't be hardcoded. The company would have to test the chip for faults, locate them, then use code to go around the bad part(s), but once you remove that chip and put it in another Linux machine, would that same remapping code go with the relocated chip?
Yes, because the only thing you're doing is telling which physical registers on the DIMM the kernel is not allowed to map. You start a kernel in very low memory space; some place hopefully safe. You run memtest86, which probes each range of registers, looking for sticky bits (a 0 or 1 that will not change). memtest86 gives you a report of which addresses are bad, which you write down with a pencil or something clever like that. Edit /etc/lilo.conf to hold this information, and boot with a badmem-patched kernel; lilo tells the kernel to exclude those addresses. It's simple, and it's all handled by the OS at the kernel level.
I'm talking about the part where he chides MS about it's closed OS and unstability. Just seems to be a lame excuse.
Agreed. I'm not a Microsoft-hater, but I know where the (whoops, some) flaws are, and sometimes I refuse to put up with them. I think some people get a little militant, though, which is just as dangerous in the long run.
People develop for both OSs all the time (think VMWare).
Better; think StarOffice. I'll get back to VMWare...
Software like CuteFTP was probably hard to design when considering that W95/98/2000 is closed-source but that particular program is very solid.
Not really. The APIs are defined, but if something is wrong with the API itself or even just the documentation...that's where it gets hard.
I just think Rick should have said that to port his code to Windows would have went beyond the scope of what he was doing. I was at his pages to read what was going on with his BadRAM project, not get a philosophy lesson.
Granted, in whole. Refusing to port to Windows was probably more of
a reflection on the idea that he doesn't have as much money to invest
in it as VMWare did to get the licensing. ;)
Bottom line is that the product isn't ready to sell because the code isn't finished.
...which is not the programmers fault, dammit! ;) You know how it goes; MS announces a release date. The programmers are saying, yet, *screaming* "NOOOO!!!!" Date get's pushed back. A new date is announced; programmers *might* make that deadline. Date gets pushed back. Date comes and goes. Next date comes and goes. Programmers' wives file for divorces in alarming numbers. Marketers get upset. Executives start threatening terminations...products appear in boxes on shelves. Programmers spend the next week standing at the coffee pot growling at each other and casting bets on *which* bug is going to be on the Top Ten list on Letterman tonight. It's marketing's fault the product isn't finished, and we all know it.
How many years have MS's programmers had to make it stable, marketing department aside?
MS-DOS 5.0. 4 was a joke, 6 was a bridge. 5 was good...
True. Then again, running a winmodem with a 1.5GHz CPU should be VERY trivial.
Except that you're cluttering my precious PCI bus with odd-sized packets at wierd intervals with your trivial data. CPU cycles are unimportant; I/O is a different story! ;)
Liked how I played Devil's advocate in this post? :o)
You've not seen the last of me... ;) ===== -- -=|JP|=- Hit me! - http://www.xanga.com/cowboydren/ Jon Pennington | Debian 2.3 -o) cowboydren @ yahoo . com | Auto Enthusiast /\\ Kansas City, MO, USA | ICQ UIN 69 67 29 31 _\_V __________________________________________________ Do You Yahoo!? Make a great connection at Yahoo! Personals. http://personals.yahoo.com
On Wed, Oct 24, 2001 at 12:18:35PM +0900, Ron Sinclair wrote:
On Tuesday 23 October 2001 21:57, you wrote: <snip>
I want to add a couple of comments to what you say. Firstly I have not read the article on bad-ram mapping, but I tip my hat to anyone clever enough to work out how to do it - but of course the real problem is bad-ram getting out there in the first place. I have a small collection of old EDO chips that were born sick and never got beyoned running DOS..but my local shop at the time decided it would not swap bad ram. With Windows it is not bad programmers btw. They probably have a collection of very good programmers, but the blue-screen phenomena is regarded as acceptable by the bean-counters. Blue-screens are often caused by faulty hardware, ctrl/alt/del always fixes it :). As regards win-modems, as you say you are not a programmer, you need to know what it is you are trying to program when you write device drivers. Winmodems, are not modems, even Windows recognises them as something it cutely calls a "communications device", and will do squat until the software is loaded that does all the modem business for it. If you do not have the specification of this "communication device" it is damn difficult to know how you communicate with it. It is like trying to learn a foreign language without a book on the rules of grammer and a dictionary. I don't believe this is just some bloody-mindedness from Open Source developers, look at the phenomenal efforts that are made to let Open Source systems co-exist with Windows, and even run Windows applications - all without a shred of help from Microsoft. I recently installed Windows ME on a system with Linux on it, I wanted to put it on the second hard drive, so I physically unplugged the first hard drive while I did it because I had no idea what the ME installation program would do to it, in order to boot it I have to use a trick in Lilo to swap the drive Bios designations, a trick put in *specifically* to allow booting of Microsoft OS'es..(and OS/2 as well). Long ago in the early days of DOS it was not exactly open source but you used to get a phenomenal amount of technical information on it - it was then an O/S waiting for applications to be written for it. The explosion of programs written for it in those days were probably written by the same kind of people who write open source programs today. That was my Eurocent... -- Regards Cliff
participants (5)
-
Cliff Sarginson
-
Jon Pennington
-
Marcel Broekman
-
Richard Clyne
-
Ron Sinclair