Memtest knocks Puppy out?

Booting, installing, newbie
Locked
Message
Author
JohnRoberts
Posts: 145
Joined: Thu 30 Nov 2006, 00:04
Location: Greece

Memtest knocks Puppy out?

#1 Post by JohnRoberts »

when running from a Live CD (using 2.13, 2.14 & 2.16), memtest freezes Puppy. Does this work at all? If it is buggy, instead of having memtest inside Puppy, why not embed a memtest86+ bootable image in the ISO (making it hence multi-bootable...). Actual image may be about 100k...
Help M$ become a Linux distro maintainer...
Force-feed them with Open-Source faster than they can produce patents

User avatar
steevieb
Posts: 289
Joined: Sun 31 Dec 2006, 00:11
Location: Poole, Dorset. UK

#2 Post by steevieb »

memtest v. 2.93.1
(C) 2000 Charles Cazabon <memtest@discworld.dyndns.org>
Original v.1 (C) 1999 Simon Kirby <sim@stormix.com> <sim@neato.org>

Current limits:
RLIMIT_RSS 0xffffffff
RLIMIT_VMEM 0x8000
Raising limits...
Unable to malloc 4293918720 bytes.
Unable to malloc 4289724416 bytes.
etc,
Unable to malloc 1253048320 bytes.
Unable to malloc 1248854016 bytes.
Unable to malloc 1244659712 bytes.
etc, until
Allocated 1240465408 bytes...trying mlock.../usr/sbin/memtestshell: line 12: 91 44 Killed memtest all 1
Test completed. Press ENTER to continue...
This is what I get if it helps anyone. I use SDRAM 168pin PC133 Sync CL3 Registered ECC memory (server memory)

Bruce B

#3 Post by Bruce B »

I had a problem with memtest, but I don't remember specifics. I just don't use it anymore.

For things like that I use SystemRescueCD

My brother (who's smarter than I am), says you should test the memory 3x, because some times the first pass doesn't catch an error, but subsequent passes will.

Also I like doing things like this at a low level rather than on top of a full working operating system.

Memtest takes a lot of time when you have a lot of RAM, so I do it while asleep and let it run over and over. Then read results in the morning or afternoon. Depends on when I get up :)

jonyo

#4 Post by jonyo »

Someone reported this baaaack in feb. :shock: Tried it out on~ 2-3 live cd setups (diff 2.1+series vers too) & pup locked up solid. Only way i knew at the time how to get out was to cut the power. Since that time, found out about control>alt>delete & control>alt>backspace options but never tried the mem test since. :?

JohnRoberts
Posts: 145
Joined: Thu 30 Nov 2006, 00:04
Location: Greece

#5 Post by JohnRoberts »

Nothing wrong with my RAM (for sure!!). When I build a machine I usally run a thoroughbred memtest for at least 8-10 hours to be absolutely sure the DIMMs are not defective. If I get even a single error, on any of the tests the DIMM's go back to the store for replacement.

...my argument still stands...If memtest cannot work properly or freezes puppy why not get rid of it alltogether and simply embed a memtest image in the ISO with a boot option to load memtest instead of running Puppy...


@ BarryK: Barry, if you read this, what do you think? Memtest86+ 1.70 is maybe 100k (minus the binaries that could be deleted from the FS...), say about 85k...All it takes to include the image in the ISO and make it muti-bootable...
Help M$ become a Linux distro maintainer...
Force-feed them with Open-Source faster than they can produce patents

User avatar
Gn2
Posts: 943
Joined: Mon 16 Oct 2006, 05:33
Location: virtual - Veni vidi, nihil est adpulerit

#6 Post by Gn2 »

Of course it will "kill" Puppy when running in RAM
Memtest sooner or later will try to access the same memory range as the O/System !
The error interupt process cannot recover- either a kernel dump or freeze will occur
The ABI scheduling queue cannot be user-space altered.
A memory register dump cannot be executed - Puppy was running in RAM !

But regardless of running in Ram or not > ,Memtest was meant to be booted directly NOT run from O/System
Any utility/ App or forked process running in background -
must be mem_alloc own (protected kernel memory) space

Next > It is absolutely pointless to run for hours -
Each pass is exactly the same access/read sequence of memory flagged sectors of RAM -
Repeated endlessly

Instead, use such as md5sum against LARGE database such as graphical storage.
Large compiles also stresses RAM far more than simple reads !
That accesses memory randomly - the correct way to test Ram banks

Heat is the killer

It may not even be the memory - just the position "pointer" on <>off: (actually a transistor, open<> closed gate flagging state).
When a faulty bank IS found - first try swapping slots
Often the unique sequence to access bad (one byte only) portion will NEVER be duplicated again

User avatar
Gn2
Posts: 943
Joined: Mon 16 Oct 2006, 05:33
Location: virtual - Veni vidi, nihil est adpulerit

#7 Post by Gn2 »

Yes Barry - Memtest should never have been scripted to run from within Puppy !

Knowing what it does TESTS MEMORY - the reason why
should have been very obvious.

BTW it functions just about the same as any P.O.S.T. pre-boot by BIOS > Reads ram -
Difference - has options to read/exclude/re-run etc.

A single pass should take no more than one > two minutes tops IF over ONE GIG of physical RAM
Parity or not ("registered - only means it has one extra "check sector)
MORE important is latency of RAM,. matching of types &
same Mfg batch.

Bruce B

#8 Post by Bruce B »

It is absolutely pointless to run for hours -
Each pass is exactly the same access/read sequence of memory flagged sectors of RAM -
Repeated endlessly
This above a reply to what I wrote:

The software (Memtest) and the hardware (RAM) would have to be absolutely static in order to make running memtest more than 1x absolutely pointless as you contend, rebutting my brother's recommendation of running 3x tests.

If the RAM was flat out dead, there wouldn't be a way or reason to run Memtest.

If RAM is slightly degrading but still working, it may show errors under stringent tests, depending on the level of degradation. One reason why is, degrading RAM is not static, it is degrading.

Good RAM will pass a statistic of 1x as well as a statistic of 3x.

But all you get are statistics.

I can come back and say, "I have a statistic of three tests and found no problem."

You can come back and say you have a statistic of one test and it would be absolutely pointless to test any further. The implication being the RAM is absolutely good.

User avatar
Gn2
Posts: 943
Joined: Mon 16 Oct 2006, 05:33
Location: virtual - Veni vidi, nihil est adpulerit

#9 Post by Gn2 »

Once again- you can't read - where was it stated to run only one pass !

Or write ??
WTH is
a statistic of 1x as well as a statistic of 3x.

User avatar
JohnMurga
Site Admin
Posts: 555
Joined: Wed 04 May 2005, 04:26
Location: Far to the east
Contact:

#10 Post by JohnMurga »

Gn2 wrote:Once again- you can't read - where was it stated to run only one pass !
OK, this was your final chance ...

I have just caught up with your posts for the last few days, and there is more of the behavior that you have been warned about over-and-over again.

I have just had enough, I gave you several chances.

You are the first person to be banned.

:-(

Cheers
JohnM

Locked