Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Alan McKinnon]

On Friday 12 February 2010 10:54:53 Alan Mackenzie wrote:
> Hi, Gentoo,
> 
> My new Gentoo box has become unusably unstable.
> 
> The first sign was when the compiler threw a segfault whilst emerging
> the xfce window manager.  I "solved" this by emerging Openbox instead.
> 
> Then I got another compiler segfault whilst emerging firefox (yes, I
> know there's a binary for this).



everything you mention below is indicative of failing hardware, especially RAM 
closely followed by PSU.

Swap them out with known good items and test thoroughly *before* doing 
anything else.





> 
> Then, on somebody's advice (not fully understood), I did
> 
>     # emerge -e gcc
> 
> , to try and get a consistent working gcc.  This crashed.  I repeated
> the invocation, and it crashed more quickly.  :-(
> 
> At this point, I thought, just reload "everything" from the stage3, with
> 
>     # cd / ; bunzip2 /stage3-amd64-20100121.tar.gz
> 
> , which didn't help either.  I emerged gentoolkit, to see if I could get
> some handle on the mess.  Then
> 
>     # revdep-rebuild -p
> 
> threw a segfault.
> 
> At this point, I'm feeling a bit sad.  My rough guess is that there's
> some conflict somewhere between 32-bit and 64-bit code, and some of my
> USE flags are inconsistent with some others, or the kernel, or something
> like that.
> 
> One other thing I remember vaguely is that early on, some emerge told me
> I had to "revdep-rebuild" something.  I wasn't able to do this through
> not yet knowing what "revdep-rebuild" meant, and not having any file of
> that name on my system.  Could this be the cause?
> 
> Finally, is there a way of reloading/rebuilding ALL the executables
> onto/on my system without discarding all my painfully wrought config
> files and without portage getting confused?
> 
> Thanks in advance for the help!

-- 
alan dot mckinnon at gmail dot com

[gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Alan Mackenzie]

Hi, Gentoo,

My new Gentoo box has become unusably unstable.

The first sign was when the compiler threw a segfault whilst emerging
the xfce window manager.  I "solved" this by emerging Openbox instead.

Then I got another compiler segfault whilst emerging firefox (yes, I
know there's a binary for this).

Then, on somebody's advice (not fully understood), I did

    # emerge -e gcc

, to try and get a consistent working gcc.  This crashed.  I repeated
the invocation, and it crashed more quickly.  :-(

At this point, I thought, just reload "everything" from the stage3, with

    # cd / ; bunzip2 /stage3-amd64-20100121.tar.gz

, which didn't help either.  I emerged gentoolkit, to see if I could get
some handle on the mess.  Then

    # revdep-rebuild -p

threw a segfault.

At this point, I'm feeling a bit sad.  My rough guess is that there's
some conflict somewhere between 32-bit and 64-bit code, and some of my
USE flags are inconsistent with some others, or the kernel, or something
like that.

One other thing I remember vaguely is that early on, some emerge told me
I had to "revdep-rebuild" something.  I wasn't able to do this through
not yet knowing what "revdep-rebuild" meant, and not having any file of
that name on my system.  Could this be the cause?

Finally, is there a way of reloading/rebuilding ALL the executables
onto/on my system without discarding all my painfully wrought config
files and without portage getting confused?

Thanks in advance for the help!

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[William Kenworthy]

I am (was) getting this as well on a new amd64 install (zotac ION n330)
- it appears there is a problem with certain glibc 32bit libraries -
there is a bug about it.  Upgraded glibc and most things are now happy -
but I still cant build gcc.

BillK



On Fri, 2010-02-12 at 10:50 +0200, Alan McKinnon wrote:
> On Friday 12 February 2010 10:54:53 Alan Mackenzie wrote:
> > Hi, Gentoo,
> > 
> > My new Gentoo box has become unusably unstable.
> > 
> > The first sign was when the compiler threw a segfault whilst emerging
> > the xfce window manager.  I "solved" this by emerging Openbox instead.
> > 
> > Then I got another compiler segfault whilst emerging firefox (yes, I
> > know there's a binary for this).
> 
> 
> 
> everything you mention below is indicative of failing hardware, especially RAM 
> closely followed by PSU.
> 
> Swap them out with known good items and test thoroughly *before* doing 
> anything else.
> 
> 
> 
> 
> 
> > 
> > Then, on somebody's advice (not fully understood), I did
> > 
> >     # emerge -e gcc
> > 
> > , to try and get a consistent working gcc.  This crashed.  I repeated
> > the invocation, and it crashed more quickly.  :-(
> > 
> > At this point, I thought, just reload "everything" from the stage3, with
> > 
> >     # cd / ; bunzip2 /stage3-amd64-20100121.tar.gz
> > 
> > , which didn't help either.  I emerged gentoolkit, to see if I could get
> > some handle on the mess.  Then
> > 
> >     # revdep-rebuild -p
> > 
> > threw a segfault.
> > 
> > At this point, I'm feeling a bit sad.  My rough guess is that there's
> > some conflict somewhere between 32-bit and 64-bit code, and some of my
> > USE flags are inconsistent with some others, or the kernel, or something
> > like that.
> > 
> > One other thing I remember vaguely is that early on, some emerge told me
> > I had to "revdep-rebuild" something.  I wasn't able to do this through
> > not yet knowing what "revdep-rebuild" meant, and not having any file of
> > that name on my system.  Could this be the cause?
> > 
> > Finally, is there a way of reloading/rebuilding ALL the executables
> > onto/on my system without discarding all my painfully wrought config
> > files and without portage getting confused?
> > 
> > Thanks in advance for the help!
> 
-- 
William Kenworthy <billk@iinet.net.au>
Home in Perth!

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Alan Mackenzie]

On Fri, Feb 12, 2010 at 10:50:46AM +0200, Alan McKinnon wrote:
> On Friday 12 February 2010 10:54:53 Alan Mackenzie wrote:
> > Hi, Gentoo,

> > My new Gentoo box has become unusably unstable.

> > The first sign was when the compiler threw a segfault whilst emerging
> > the xfce window manager.  I "solved" this by emerging Openbox
> > instead.

> > Then I got another compiler segfault whilst emerging firefox (yes, I
> > know there's a binary for this).



> everything you mention below is indicative of failing hardware,
> especially RAM closely followed by PSU.

> Swap them out with known good items and test thoroughly *before* doing
> anything else.

I hope you're not right here.  ;-)  The hardware is spanking brand new;
so new, in fact, that it's still gleaming.

Are there any handy utility programs around to test RAM exhaustively?

[ .... ]

> -- 
> alan dot mckinnon at gmail dot com

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Helmut Jarausch]

On 12 Feb, Alan Mackenzie wrote:
> On Fri, Feb 12, 2010 at 10:50:46AM +0200, Alan McKinnon wrote:
>> On Friday 12 February 2010 10:54:53 Alan Mackenzie wrote:
>> > Hi, Gentoo,
> 
>> > My new Gentoo box has become unusably unstable.
> 
>> > The first sign was when the compiler threw a segfault whilst emerging
>> > the xfce window manager.  I "solved" this by emerging Openbox
>> > instead.
> 
>> > Then I got another compiler segfault whilst emerging firefox (yes, I
>> > know there's a binary for this).
> 
> 
> 
>> everything you mention below is indicative of failing hardware,
>> especially RAM closely followed by PSU.
> 
>> Swap them out with known good items and test thoroughly *before* doing
>> anything else.
> 
> I hope you're not right here.  ;-)  The hardware is spanking brand new;
> so new, in fact, that it's still gleaming.
> 
> Are there any handy utility programs around to test RAM exhaustively?

I have made good experience with sys-apps/memtester. If you have not
more than 4 Gb RAM, you can use the SystemRescueCD.
Otherwise the GRML64 RescueCD has a true 64 bit version of memtester.

I have made the experience that memtester finds errors more quickly than
memtest86+ .

Helmut.


-- 
Helmut Jarausch

Lehrstuhl fuer Numerische Mathematik
RWTH - Aachen University
D 52056 Aachen, Germany

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Lewis Handy]

<snip>
>
> I hope you're not right here.  ;-)  The hardware is spanking brand new;
> so new, in fact, that it's still gleaming.
>
> Are there any handy utility programs around to test RAM exhaustively?
>

Most live CD's these days will come with memtest86 as a boot option,
that would give your RAM a good workout and tell you how it looks.

It could be that the RAM itself is perfectly fine, but just not
working too well with your motherboard hardware, I have seen this many
times before where in one system the RAM is completely unusable, but
in a different system is perfectly fine.

> [ .... ]
>
>> --
>> alan dot mckinnon at gmail dot com
>
> --
> Alan Mackenzie (Nuremberg, Germany).
>
>

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Daniel Wagener]

On Fri, 12 Feb 2010 09:46:33 +0000
Alan Mackenzie <acm@muc.de> wrote:

> On Fri, Feb 12, 2010 at 10:50:46AM +0200, Alan McKinnon wrote:
> > On Friday 12 February 2010 10:54:53 Alan Mackenzie wrote:
> > > Hi, Gentoo,
> 
> > > My new Gentoo box has become unusably unstable.
> 
> > > The first sign was when the compiler threw a segfault whilst
> > > emerging the xfce window manager.  I "solved" this by emerging
> > > Openbox instead.
> 
> > > Then I got another compiler segfault whilst emerging firefox
> > > (yes, I know there's a binary for this).
> 
> 
> 
> > everything you mention below is indicative of failing hardware,
> > especially RAM closely followed by PSU.
> 
> > Swap them out with known good items and test thoroughly *before*
> > doing anything else.
> 
> I hope you're not right here.  ;-)  The hardware is spanking brand
> new; so new, in fact, that it's still gleaming.
> 
> Are there any handy utility programs around to test RAM exhaustively?
> 
> [ .... ]
> 
> > -- 
> > alan dot mckinnon at gmail dot com
> 

you may wanna look at memtest86+, its in portage
http://www.memtest.org/

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Roy Wright]

[-- Attachment #1: Type: text/plain, Size: 1121 bytes --]


On Feb 12, 2010, at 3:46 AM, Alan Mackenzie wrote:

> On Fri, Feb 12, 2010 at 10:50:46AM +0200, Alan McKinnon wrote:
> 
> I hope you're not right here.  ;-)  The hardware is spanking brand new;
> so new, in fact, that it's still gleaming.
> 
> Are there any handy utility programs around to test RAM exhaustively?
> 

I use a kubuntu live disk, one of the boot options is to run memtest.  I typically let it run a several hours on a new box.

The second issue might be heat (or cooling depending on your point of view).  A couple of things, easiest is to install the lm_sensors and your favorite capture utility.  I also will use an IR thermometer and DMM with temperature probe to verify cooling.

With lm_sensors running, I'll run several stress tests to find the hottest the system will run.  If the system gets too hot, I'll increase cooling (more/faster fans).

Another thing to check is that the memory is seated properly.  I'd remove the modules, inspect the connectors, then very carefully reinstall them.  While you are at it, make sure all connectors are seated correctly.


HTH,
Roy



[-- Attachment #2: Type: text/html, Size: 1604 bytes --]

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Roy Wright]

Another thing to try, change compile flag to -J1 (MAKEOPTS="-J1" emerge ...)

Last time I built up a system there were a couple of packages that couldn't handle parallel compiles.

HTH,
Roy

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Alan McKinnon]

On Friday 12 February 2010 11:46:33 Alan Mackenzie wrote:
> On Fri, Feb 12, 2010 at 10:50:46AM +0200, Alan McKinnon wrote:
> > On Friday 12 February 2010 10:54:53 Alan Mackenzie wrote:
> > > Hi, Gentoo,
> > > 
> > > My new Gentoo box has become unusably unstable.
> > > 
> > > The first sign was when the compiler threw a segfault whilst emerging
> > > the xfce window manager.  I "solved" this by emerging Openbox
> > > instead.
> > > 
> > > Then I got another compiler segfault whilst emerging firefox (yes, I
> > > know there's a binary for this).
> > 
> > everything you mention below is indicative of failing hardware,
> > especially RAM closely followed by PSU.
> > 
> > Swap them out with known good items and test thoroughly *before* doing
> > anything else.
> 
> I hope you're not right here.  ;-)  The hardware is spanking brand new;
> so new, in fact, that it's still gleaming.

The engineering phrase "infant mortality" comes to mind :-)


> 
> Are there any handy utility programs around to test RAM exhaustively?

memtest is pretty good at this. Let it run for many hours, there's lots of 
tips out there in Googleland on how to get good results



-- 
alan dot mckinnon at gmail dot com

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Alan Mackenzie]

Hi, Alan,

On Fri, Feb 12, 2010 at 10:50:46AM +0200, Alan McKinnon wrote:
> On Friday 12 February 2010 10:54:53 Alan Mackenzie wrote:
> > Hi, Gentoo,

> > My new Gentoo box has become unusably unstable.

> > The first sign was when the compiler threw a segfault whilst emerging
> > the xfce window manager.  I "solved" this by emerging Openbox instead.

> > Then I got another compiler segfault whilst emerging firefox (yes, I
> > know there's a binary for this).



> everything you mention below is indicative of failing hardware, especially RAM 
> closely followed by PSU.

Yes, you're right.  :-(

When I run memtest86 from the gentoo boot disk, it signals millions of
failures in b11 of 32 bit words.

I'll try unplugging and replugging these.

Thanks for the tip.

> alan dot mckinnon at gmail dot com

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Mark Knecht]

On Fri, Feb 12, 2010 at 12:54 AM, Alan Mackenzie <acm@muc.de> wrote:
> Hi, Gentoo,
>
> My new Gentoo box has become unusably unstable.
>
> The first sign was when the compiler threw a segfault whilst emerging
> the xfce window manager.  I "solved" this by emerging Openbox instead.
>
> Then I got another compiler segfault whilst emerging firefox (yes, I
> know there's a binary for this).
>
> Then, on somebody's advice (not fully understood), I did
>
>    # emerge -e gcc
>
> , to try and get a consistent working gcc.  This crashed.  I repeated
> the invocation, and it crashed more quickly.  :-(
>
> At this point, I thought, just reload "everything" from the stage3, with
>
>    # cd / ; bunzip2 /stage3-amd64-20100121.tar.gz
>
> , which didn't help either.  I emerged gentoolkit, to see if I could get
> some handle on the mess.  Then
>
>    # revdep-rebuild -p
>
> threw a segfault.
>
> At this point, I'm feeling a bit sad.  My rough guess is that there's
> some conflict somewhere between 32-bit and 64-bit code, and some of my
> USE flags are inconsistent with some others, or the kernel, or something
> like that.
>
> One other thing I remember vaguely is that early on, some emerge told me
> I had to "revdep-rebuild" something.  I wasn't able to do this through
> not yet knowing what "revdep-rebuild" meant, and not having any file of
> that name on my system.  Could this be the cause?
>
> Finally, is there a way of reloading/rebuilding ALL the executables
> onto/on my system without discarding all my painfully wrought config
> files and without portage getting confused?
>
> Thanks in advance for the help!
>
> --
> Alan Mackenzie (Nuremberg, Germany).
>
>

+1 for memtest86.

Since it's new hardware possibly your CPU fan isn't well seated?
Failures while compiling sound like heat to me. Failures at random
times sound like heat, PSU and memory problems.

Hope tis helps,
Mark

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Helmut Jarausch]

On 12 Feb, Alan Mackenzie wrote:
> Hi, Gentoo,
> 
> My new Gentoo box has become unusably unstable.
> 

Just one more reason.

Have you got more than 4 Gb RAM installed?

If yes, you might have to reduce memory speed.
Some CPUs (among them AMD Phenom) have difficulties
at the specified RAM speed if more than 1 memory bank is
installed.
I have been burnt by that several times.
Helmut.

-- 
Helmut Jarausch

Lehrstuhl fuer Numerische Mathematik
RWTH - Aachen University
D 52056 Aachen, Germany

[gentoo-user] Ramifications of memtest86

[Alan Mackenzie]

On Fri, Feb 12, 2010 at 11:22:21AM +0000, Alan Mackenzie wrote:

> On Fri, Feb 12, 2010 at 10:50:46AM +0200, Alan McKinnon wrote:
> > On Friday 12 February 2010 10:54:53 Alan Mackenzie wrote:
> > > Hi, Gentoo,

> > > My new Gentoo box has become unusably unstable.

> > > The first sign was when the compiler threw a segfault whilst
> > > emerging the xfce window manager.  I "solved" this by emerging
> > > Openbox instead.

> > > Then I got another compiler segfault whilst emerging firefox (yes,
> > > I know there's a binary for this).



> > everything you mention below is indicative of failing hardware,
> > especially RAM closely followed by PSU.

> Yes, you're right.  :-(

> When I run memtest86 from the gentoo boot disk, it signals millions of
> failures in b11 of 32 bit words.

> I'll try unplugging and replugging these.

No help.  :-(

No matter how I plug in the RAM (4 combinations of 2 sticks into 2 pairs
of slots) it is always b11 which fails and always at an address ending
in (hex) 0 or 8.

How is this DDR3 Ram organised?  Is each stick 64 bits wide, or are they
32 bits wide, being accessed by the motherboard pairwisely?  If the
latter, I would have exected the failure to move to address ....4 and
....C when I swap the two sticks.  Am I being prematurely pessimistic in
thinking the motherboard might be the fault?

> > alan dot mckinnon at gmail dot com

> -- 
> Alan Mackenzie (Nuremberg, Germany).

Re: [gentoo-user] Ramifications of memtest86

[Dale]

chrome://messenger/locale/messengercompose/composeMsgs.properties:
> On Fri, Feb 12, 2010 at 11:22:21AM +0000, Alan Mackenzie wrote:
>
>    
>> On Fri, Feb 12, 2010 at 10:50:46AM +0200, Alan McKinnon wrote:
>>      
>>> On Friday 12 February 2010 10:54:53 Alan Mackenzie wrote:
>>>        
>>>> Hi, Gentoo,
>>>>          
>    
>>>> My new Gentoo box has become unusably unstable.
>>>>          
>    
>>>> The first sign was when the compiler threw a segfault whilst
>>>> emerging the xfce window manager.  I "solved" this by emerging
>>>> Openbox instead.
>>>>          
>    
>>>> Then I got another compiler segfault whilst emerging firefox (yes,
>>>> I know there's a binary for this).
>>>>          
>
>
>    
>>> everything you mention below is indicative of failing hardware,
>>> especially RAM closely followed by PSU.
>>>        
>    
>> Yes, you're right.  :-(
>>      
>    
>> When I run memtest86 from the gentoo boot disk, it signals millions of
>> failures in b11 of 32 bit words.
>>      
>    
>> I'll try unplugging and replugging these.
>>      
> No help.  :-(
>
> No matter how I plug in the RAM (4 combinations of 2 sticks into 2 pairs
> of slots) it is always b11 which fails and always at an address ending
> in (hex) 0 or 8.
>
> How is this DDR3 Ram organised?  Is each stick 64 bits wide, or are they
> 32 bits wide, being accessed by the motherboard pairwisely?  If the
> latter, I would have exected the failure to move to address ....4 and
> ....C when I swap the two sticks.  Am I being prematurely pessimistic in
> thinking the motherboard might be the fault?
>
>    
>>> alan dot mckinnon at gmail dot com
>>>        
>    
>> -- 
>> Alan Mackenzie (Nuremberg, Germany).
>>      

Sounds to me like the mobo may have issues.  If swapping the ram around 
doesn't move the error, then the chips on the mobo that connect to the 
ram may be bad.

I would still rule out power issues if you can.  A bad power supply, or 
a weak one, can cause some pretty weird problems.

Hmmm, I would also try putting in two sticks, then running the test and 
seeing what that says.  If it still errors, try the other two sticks.  
If it still fails in the same place, see if you can put the sticks in 
the other two slots.  Some mobos don't care what slot you have them in.  
May want to refer to the manual on that.

Let's hope for something simple and cheap.  ;-)

Dale

:-)  :-)

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Volker Armin Hemmann]

On Freitag 12 Februar 2010, Alan Mackenzie wrote:
> Hi, Alan,
> 
> On Fri, Feb 12, 2010 at 10:50:46AM +0200, Alan McKinnon wrote:
> > On Friday 12 February 2010 10:54:53 Alan Mackenzie wrote:
> > > Hi, Gentoo,
> > > 
> > > My new Gentoo box has become unusably unstable.
> > > 
> > > The first sign was when the compiler threw a segfault whilst emerging
> > > the xfce window manager.  I "solved" this by emerging Openbox instead.
> > > 
> > > Then I got another compiler segfault whilst emerging firefox (yes, I
> > > know there's a binary for this).
> > 
> > everything you mention below is indicative of failing hardware,
> > especially RAM closely followed by PSU.
> 
> Yes, you're right.  :-(
> 
> When I run memtest86 from the gentoo boot disk, it signals millions of
> failures in b11 of 32 bit words.
> 
> I'll try unplugging and replugging these.
> 
> Thanks for the tip.
> 
> > alan dot mckinnon at gmail dot com

you can try upping the voltage of the ram by 0.05V. I had a stick that threw 
errors unless I gave it a bit more. After that the system was stable and none 
of the memtesting apps found any errors.

Re: [gentoo-user] Ramifications of memtest86

[Alan McKinnon]

On Friday 12 February 2010 14:43:30 Alan Mackenzie wrote:
> > failures in b11 of 32 bit words.
> > 
> > I'll try unplugging and replugging these.
> 
> No help.  :-(
> 
> No matter how I plug in the RAM (4 combinations of 2 sticks into 2 pairs
> of slots) it is always b11 which fails and always at an address ending
> in (hex) 0 or 8.
> 
> How is this DDR3 Ram organised?  Is each stick 64 bits wide, or are they
> 32 bits wide, being accessed by the motherboard pairwisely?  If the
> latter, I would have exected the failure to move to address ....4 and
> ....C when I swap the two sticks.  Am I being prematurely pessimistic in
> thinking the motherboard might be the fault?

I strongly suspect a motherboard fault too. What you are describing sounds 
like an address decoding fault.


-- 
alan dot mckinnon at gmail dot com

Re: [gentoo-user] Ramifications of memtest86

[Alan Mackenzie]

Hi, Dale,

On Fri, Feb 12, 2010 at 06:43:46AM -0600, Dale wrote:
> chrome://messenger/locale/messengercompose/composeMsgs.properties:
> >On Fri, Feb 12, 2010 at 11:22:21AM +0000, Alan Mackenzie wrote:


> >>When I run memtest86 from the gentoo boot disk, it signals millions
> >>of failures in b11 of 32 bit words.


> >>I'll try unplugging and replugging these.

> >No help.  :-(

> >No matter how I plug in the RAM (4 combinations of 2 sticks into 2
> >pairs of slots) it is always b11 which fails and always at an address
> >ending in (hex) 0 or 8.

> >How is this DDR3 Ram organised?  Is each stick 64 bits wide, or are
> >they 32 bits wide, being accessed by the motherboard pairwisely?  If
> >the latter, I would have exected the failure to move to address ....4
> >and ....C when I swap the two sticks.  Am I being prematurely
> >pessimistic in thinking the motherboard might be the fault?


> Sounds to me like the mobo may have issues.  If swapping the ram around
> doesn't move the error, then the chips on the mobo that connect to the
> ram may be bad.

So the RAM sticks are each 32 bits wide, then?

> I would still rule out power issues if you can.  A bad power supply, or 
> a weak one, can cause some pretty weird problems.

The power supply is brand new, from a reputable manufacturer.  Surely if
the power supply were dicky, I'd get RAM errors in some other bit
position inside a 32-bit word.  Or maybe the RAM into which memtest86 is
loaded is also faulty.  ;-(

> Hmmm, I would also try putting in two sticks, then running the test and
> seeing what that says.  If it still errors, try the other two sticks.
> If it still fails in the same place, see if you can put the sticks in
> the other two slots.  Some mobos don't care what slot you have them in.
> May want to refer to the manual on that.

Sorry, I wasn't clear about that.  I have just two RAM sticks for a total
of 4Gb, but there are two pairs of slots on the motherboard.  The manual
says "put the sticks for preference in the orange slots for 'better
overclocking capability'".  But they (almost) work just as well (badly)
in the black slots.

> Let's hope for something simple and cheap.  ;-)

Oh, yes!

> Dale

> :-)  :-)

Re: [gentoo-user] Ramifications of memtest86

[Dale]

chrome://messenger/locale/messengercompose/composeMsgs.properties:
> Hi, Dale,
>
> On Fri, Feb 12, 2010 at 06:43:46AM -0600, Dale wrote:
>    
>> chrome://messenger/locale/messengercompose/composeMsgs.properties:
>>      
>>> On Fri, Feb 12, 2010 at 11:22:21AM +0000, Alan Mackenzie wrote:
>>>        
>
>    
>>>> When I run memtest86 from the gentoo boot disk, it signals millions
>>>> of failures in b11 of 32 bit words.
>>>>          
>
>    
>>>> I'll try unplugging and replugging these.
>>>>          
>    
>>> No help.  :-(
>>>        
>    
>>> No matter how I plug in the RAM (4 combinations of 2 sticks into 2
>>> pairs of slots) it is always b11 which fails and always at an address
>>> ending in (hex) 0 or 8.
>>>        
>    
>>> How is this DDR3 Ram organised?  Is each stick 64 bits wide, or are
>>> they 32 bits wide, being accessed by the motherboard pairwisely?  If
>>> the latter, I would have exected the failure to move to address ....4
>>> and ....C when I swap the two sticks.  Am I being prematurely
>>> pessimistic in thinking the motherboard might be the fault?
>>>        
>
>    
>> Sounds to me like the mobo may have issues.  If swapping the ram around
>> doesn't move the error, then the chips on the mobo that connect to the
>> ram may be bad.
>>      
> So the RAM sticks are each 32 bits wide, then?
>    

I run 32 bit here.  Mine doesn't care where they are but yours may be 
different.

>    
>> I would still rule out power issues if you can.  A bad power supply, or
>> a weak one, can cause some pretty weird problems.
>>      
> The power supply is brand new, from a reputable manufacturer.  Surely if
> the power supply were dicky, I'd get RAM errors in some other bit
> position inside a 32-bit word.  Or maybe the RAM into which memtest86 is
> loaded is also faulty.  ;-(
>
>    
>> Dale
>>      
>

Where the error is could depend on a single transistor that is maybe not 
as sensitive as the others.  It's sort of like a chain.  It's only as 
strong as its weakest link.  It could be that whatever is going wrong 
could be right on the edge of others not working either.  The one that 
is failing is just the first if it is a power problem.  That's where the 
power problem thought comes from.  Have you had a look here for well 
tested power supplies?

http://www.overclockers.com/forums/showthread.php?t=458204

That said, it could be a lot of things.  It could be a bad chip on the 
mobo, a piece of dust in the wrong place or any number of other things.  
It's finding it that is so much fun.

I would also check to see if it is a heat related issue.  If it runs 
fine cold, that could point to the "dust" theory since it is 
consistently broke.  If it only does it when hot, then that could be a 
chip problem.

Again, lots of things it could be.  So far, everybody has replied with 
good ideas to check.  There are lots of them.

Dale

:-)  :-)

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Paul Hartman]

On Fri, Feb 12, 2010 at 6:33 AM, Helmut Jarausch
<jarausch@igpm.rwth-aachen.de> wrote:
> On 12 Feb, Alan Mackenzie wrote:
>> Hi, Gentoo,
>>
>> My new Gentoo box has become unusably unstable.
>>
>
> Just one more reason.
>
> Have you got more than 4 Gb RAM installed?
>
> If yes, you might have to reduce memory speed.
> Some CPUs (among them AMD Phenom) have difficulties
> at the specified RAM speed if more than 1 memory bank is
> installed.
> I have been burnt by that several times.
> Helmut.

I had exactly the situation you described. 4 GB of RAM worked great at
800MHz but when I increased to 8 GB I had to lower speed to something
like 533 MHz otherwise system was very unstable. In my case Intel Core
2 Duo E6600. I contacted the RAM manufacturer and this was their
recommendation as well.

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Volker Armin Hemmann]

On Freitag 12 Februar 2010, Helmut Jarausch wrote:
> On 12 Feb, Alan Mackenzie wrote:
> > Hi, Gentoo,
> > 
> > My new Gentoo box has become unusably unstable.
> 
> Just one more reason.
> 
> Have you got more than 4 Gb RAM installed?
> 
> If yes, you might have to reduce memory speed.
> Some CPUs (among them AMD Phenom) have difficulties
> at the specified RAM speed if more than 1 memory bank is
> installed.
> I have been burnt by that several times.
> Helmut.

no, not entirely true. There is a problem with Phenoms and DDR2 1066 when for 
modules are installed. You can get that configuration stable with an high 
quality board and some voltage tricks.

If you have problems at lower speed and 4banks, you have a crappy mainboard. 
Maybe increasing voltages a bit (0.05 to 0.1V) will help you. It did for me 
with one certain Asrock board.

Re: [gentoo-user] New Gentoo system has become unstable and unusable - help, please!

[Alan Mackenzie]

Hi, Volker,

On Fri, Feb 12, 2010 at 01:56:48PM +0100, Volker Armin Hemmann wrote:

> you can try upping the voltage of the ram by 0.05V. I had a stick that
> threw errors unless I gave it a bit more. After that the system was
> stable and none of the memtesting apps found any errors.

Tried that with the voltage everywhere from 1.5v to 1.6v in 0.02v steps.
It didn't help.  It turns out, one of the RAM sticks was kaputt.

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: [gentoo-user] Ramifications of memtest86. Got it!

[Alan Mackenzie]

Hi, Dale,

On Fri, Feb 12, 2010 at 08:27:01AM -0600, Dale wrote:

> Where the error is could depend on a single transistor that is maybe
> not as sensitive as the others.  It's sort of like a chain.  It's only
> as strong as its weakest link.  It could be that whatever is going
> wrong could be right on the edge of others not working either.  The
> one that is failing is just the first if it is a power problem.
> That's where the power problem thought comes from.  Have you had a
> look here for well tested power supplies?

> That said, it could be a lot of things.  It could be a bad chip on the 
> mobo, a piece of dust in the wrong place or any number of other things.  
> It's finding it that is so much fun.

The shop who sold me the components suggested running memtest86 with
just one RAM stick at a time.  It turns out, one was duff, the other's
just fine.  (It went ~20 minutes on memtest86 without any errors.)  So
it looks like I'll be running on 2Gb only until I get a replacement for
the broken one.

> Again, lots of things it could be.  So far, everybody has replied with
> good ideas to check.  There are lots of them.

Many thanks to all who helped me track this one down!

> Dale

> :-)  :-)

-- 
Alan Mackenzie (Nuremberg, Germany).

[gentoo-user] Re: Ramifications of memtest86. Got it!

[Nikos Chantziaras]

On 02/12/2010 09:25 PM, Alan Mackenzie wrote:
> Hi, Dale,
>
> On Fri, Feb 12, 2010 at 08:27:01AM -0600, Dale wrote:
>
>> Where the error is could depend on a single transistor that is maybe
>> not as sensitive as the others.  It's sort of like a chain.  It's only
>> as strong as its weakest link.  It could be that whatever is going
>> wrong could be right on the edge of others not working either.  The
>> one that is failing is just the first if it is a power problem.
>> That's where the power problem thought comes from.  Have you had a
>> look here for well tested power supplies?
>
>> That said, it could be a lot of things.  It could be a bad chip on the
>> mobo, a piece of dust in the wrong place or any number of other things.
>> It's finding it that is so much fun.
>
> The shop who sold me the components suggested running memtest86 with
> just one RAM stick at a time.

The correct procedure is to test just one stick in at least two 
different slots.  This is to ensure that it's not the mainboard at fault 
(like a faulty RAM slot.)

Re: [gentoo-user] Ramifications of memtest86. Got it!

[Dale]

chrome://messenger/locale/messengercompose/composeMsgs.properties:
> Hi, Dale,
>
> On Fri, Feb 12, 2010 at 08:27:01AM -0600, Dale wrote:
>
>    
>> Where the error is could depend on a single transistor that is maybe
>> not as sensitive as the others.  It's sort of like a chain.  It's only
>> as strong as its weakest link.  It could be that whatever is going
>> wrong could be right on the edge of others not working either.  The
>> one that is failing is just the first if it is a power problem.
>> That's where the power problem thought comes from.  Have you had a
>> look here for well tested power supplies?
>>      
>    
>> That said, it could be a lot of things.  It could be a bad chip on the
>> mobo, a piece of dust in the wrong place or any number of other things.
>> It's finding it that is so much fun.
>>      
> The shop who sold me the components suggested running memtest86 with
> just one RAM stick at a time.  It turns out, one was duff, the other's
> just fine.  (It went ~20 minutes on memtest86 without any errors.)  So
> it looks like I'll be running on 2Gb only until I get a replacement for
> the broken one.
>
>    
>> Again, lots of things it could be.  So far, everybody has replied with
>> good ideas to check.  There are lots of them.
>>      
> Many thanks to all who helped me track this one down!
>
>    
>> Dale
>>      
>    
>> :-)  :-)
>>      

There you go.  Most likely one little transistor that went belly up.  
Considering there are millions of those little devils on there, no 
surprise at all.

Glad you got it sorted out and that is better than a lot of other 
options.  Since it is new, I hope you have a good warranty that will 
make it a cheap fix as well.

Dale

:-)  :-)

[gentoo-user] Re: New Gentoo system has become unstable and unusable - help, please!

[walt]

Volker Armin Hemmann wrote:

> Maybe increasing voltages a bit (0.05 to 0.1V) will help you. It did for me 
> with one certain Asrock board.

This is the first I've heard of adjustable voltages.  Where/how do you make
the adjustment?

Re: [gentoo-user] Re: New Gentoo system has become unstable and unusable - help, please!

[Dale]

chrome://messenger/locale/messengercompose/composeMsgs.properties:
> Volker Armin Hemmann wrote:
>
>    
>> Maybe increasing voltages a bit (0.05 to 0.1V) will help you. It did for me
>> with one certain Asrock board.
>>      
> This is the first I've heard of adjustable voltages.  Where/how do you make
> the adjustment?
>
>    

Mine is in the BIOS.

Dale

:-)  :-)

Re: [gentoo-user] Re: New Gentoo system has become unstable and unusable - help, please!

[Volker Armin Hemmann]

On Samstag 13 Februar 2010, walt wrote:
> Volker Armin Hemmann wrote:
> > Maybe increasing voltages a bit (0.05 to 0.1V) will help you. It did for
> > me with one certain Asrock board.
> 
> This is the first I've heard of adjustable voltages.  Where/how do you make
> the adjustment?

Bios

Re: [gentoo-user] Ramifications of memtest86. Got it!

[Alan Mackenzie]

Hi, Dale,

On Fri, Feb 12, 2010 at 05:27:55PM -0600, Dale wrote:
> chrome://messenger/locale/messengercompose/composeMsgs.properties:

> >The shop who sold me the components suggested running memtest86 with
> >just one RAM stick at a time.  It turns out, one was duff, the other's
> >just fine.  (It went ~20 minutes on memtest86 without any errors.)  So
> >it looks like I'll be running on 2Gb only until I get a replacement
> >for the broken one.


> >Many thanks to all who helped me track this one down!


> >>Dale


> >>:-)  :-)


> There you go.  Most likely one little transistor that went belly up.  
> Considering there are millions of those little devils on there, no 
> surprise at all.

Oh, don't be like that!  You're saying, like, another "little" transistor
will soon be going.  ;-)

> Glad you got it sorted out and that is better than a lot of other
> options.  Since it is new, I hope you have a good warranty that will
> make it a cheap fix as well.

I bought my PC components from a premium quality shop, the sort that
behaves like a gentleman and honours its guarantees.  Its email support
gets back to you within an hour or so (in business hours).  The
proprietor said I needed to send back _both_ RAM sticks (since they have
a joint serial number), but he's sending me a replacement pair first, so
my machine remains working.

For all that, the cost of this PC was less than half that of its
predecessor, a 1.2 GHz Athlon machine from ~2001.  With desktop PCs now
being so ridiculously cheap anyhow, it seems false economy to buy from a
lesser vendor.

> Dale

> :-)  :-)

-- 
Alan Mackenzie (Nuremberg, Germany).

Re: [gentoo-user] Ramifications of memtest86. Got it!

[Volker Armin Hemmann]

On Samstag 13 Februar 2010, Alan Mackenzie wrote:
> Hi, Dale,
> 
> On Fri, Feb 12, 2010 at 05:27:55PM -0600, Dale wrote:
> > chrome://messenger/locale/messengercompose/composeMsgs.properties:
> > >The shop who sold me the components suggested running memtest86 with
> > >just one RAM stick at a time.  It turns out, one was duff, the other's
> > >just fine.  (It went ~20 minutes on memtest86 without any errors.)  So
> > >it looks like I'll be running on 2Gb only until I get a replacement
> > >for the broken one.
> > >
> > >
> > >Many thanks to all who helped me track this one down!
> > >
> > >>Dale
> > >>
> > >>:-)  :-)
> > 
> > There you go.  Most likely one little transistor that went belly up.
> > Considering there are millions of those little devils on there, no
> > surprise at all.
> 
> Oh, don't be like that!  You're saying, like, another "little" transistor
> will soon be going.  ;-)
> 
> > Glad you got it sorted out and that is better than a lot of other
> > options.  Since it is new, I hope you have a good warranty that will
> > make it a cheap fix as well.
> 
> I bought my PC components from a premium quality shop, the sort that
> behaves like a gentleman and honours its guarantees.  Its email support
> gets back to you within an hour or so (in business hours).  The
> proprietor said I needed to send back _both_ RAM sticks (since they have
> a joint serial number), but he's sending me a replacement pair first, so
> my machine remains working.
> 
> For all that, the cost of this PC was less than half that of its
> predecessor, a 1.2 GHz Athlon machine from ~2001.  With desktop PCs now
> being so ridiculously cheap anyhow, it seems false economy to buy from a
> lesser vendor.
> 
> > Dale
> > 
> > :-)  :-)

nothing beats a local shop where the personal knows you and they just replace 
your stuff when you tell them that it is defective...

I gladly pay a little premium for that.

Re: [gentoo-user] Ramifications of memtest86. Got it!

[Dale]

chrome://messenger/locale/messengercompose/composeMsgs.properties:
> Hi, Dale,
>
> On Fri, Feb 12, 2010 at 05:27:55PM -0600, Dale wrote:
>    
>> chrome://messenger/locale/messengercompose/composeMsgs.properties:
>>      
>    
>>> The shop who sold me the components suggested running memtest86 with
>>> just one RAM stick at a time.  It turns out, one was duff, the other's
>>> just fine.  (It went ~20 minutes on memtest86 without any errors.)  So
>>> it looks like I'll be running on 2Gb only until I get a replacement
>>> for the broken one.
>>>        
>
>    
>>> Many thanks to all who helped me track this one down!
>>>        
>
>    
>>>> Dale
>>>>          
>
>    
>>>> :-)  :-)
>>>>          
>
>    
>> There you go.  Most likely one little transistor that went belly up.
>> Considering there are millions of those little devils on there, no
>> surprise at all.
>>      
> Oh, don't be like that!  You're saying, like, another "little" transistor
> will soon be going.  ;-)
>    

That's not what I meant at all.  Consider the odds.  There are millions 
of little circuits on those chips, even a 0.0001% failure rate can mean 
the chip is bad.  They either all work or the chip doesn't work.


>    
>> Glad you got it sorted out and that is better than a lot of other
>> options.  Since it is new, I hope you have a good warranty that will
>> make it a cheap fix as well.
>>      
> I bought my PC components from a premium quality shop, the sort that
> behaves like a gentleman and honours its guarantees.  Its email support
> gets back to you within an hour or so (in business hours).  The
> proprietor said I needed to send back _both_ RAM sticks (since they have
> a joint serial number), but he's sending me a replacement pair first, so
> my machine remains working.
>
> For all that, the cost of this PC was less than half that of its
> predecessor, a 1.2 GHz Athlon machine from ~2001.  With desktop PCs now
> being so ridiculously cheap anyhow, it seems false economy to buy from a
> lesser vendor.
>
>    
>> Dale
>>      
>    
>> :-)  :-)
>>      
>    

I bought mine from newegg and they stand behind theirs too.  Things is, 
the person you buy them from doesn't decide if it is a bad one or not.  
It's just a luck of the draw.  I doubt there are many companies that 
want to sell something that is broken.  It's not good for business.

If you bought the memory sticks as a set, they do usually want them 
returned as a set.  That's normal.  I have seen that many times.

Glad you got something working tho.

Dale

:-)  :-)

Re: [gentoo-user] Ramifications of memtest86. Got it!

[Volker Armin Hemmann]

On Sonntag 14 Februar 2010, Dale wrote:
> chrome://messenger/locale/messengercompose/composeMsgs.properties:
> > Hi, Dale,
> > 
> > On Fri, Feb 12, 2010 at 05:27:55PM -0600, Dale wrote:
> >> chrome://messenger/locale/messengercompose/composeMsgs.properties:
> >>> The shop who sold me the components suggested running memtest86 with
> >>> just one RAM stick at a time.  It turns out, one was duff, the other's
> >>> just fine.  (It went ~20 minutes on memtest86 without any errors.)  So
> >>> it looks like I'll be running on 2Gb only until I get a replacement
> >>> for the broken one.
> >>> 
> >>> 
> >>> 
> >>> Many thanks to all who helped me track this one down!
> >>> 
> >>>> Dale
> >>>> 
> >>>> :-)  :-)
> >> 
> >> There you go.  Most likely one little transistor that went belly up.
> >> Considering there are millions of those little devils on there, no
> >> surprise at all.
> > 
> > Oh, don't be like that!  You're saying, like, another "little" transistor
> > will soon be going.  ;-)
> 
> That's not what I meant at all.  Consider the odds.  There are millions
> of little circuits on those chips, even a 0.0001% failure rate can mean
> the chip is bad.  They either all work or the chip doesn't work.
> 
> >> Glad you got it sorted out and that is better than a lot of other
> >> options.  Since it is new, I hope you have a good warranty that will
> >> make it a cheap fix as well.
> > 
> > I bought my PC components from a premium quality shop, the sort that
> > behaves like a gentleman and honours its guarantees.  Its email support
> > gets back to you within an hour or so (in business hours).  The
> > proprietor said I needed to send back _both_ RAM sticks (since they have
> > a joint serial number), but he's sending me a replacement pair first, so
> > my machine remains working.
> > 
> > For all that, the cost of this PC was less than half that of its
> > predecessor, a 1.2 GHz Athlon machine from ~2001.  With desktop PCs now
> > being so ridiculously cheap anyhow, it seems false economy to buy from a
> > lesser vendor.
> > 
> >> Dale
> >> 
> >> :-)  :-)
> 
> I bought mine from newegg and they stand behind theirs too.  Things is,
> the person you buy them from doesn't decide if it is a bad one or not.
> It's just a luck of the draw.  I doubt there are many companies that
> want to sell something that is broken.  It's not good for business.

no, but some shops just say 'ok, that is bad luck, here is a new set, try 
these'. And some others let you wait or just don't believe you, demand the 
output of certain test apps etc pp.

Some years ago a friend had a stick that would produce errors only in certain 
scenarios (it was a muster dependent error). Most checkers back then did not 
find it. Our favorite shop just exchanged it....

And that kind of service is a good reason to buy there. 

Re: [gentoo-user] Ramifications of memtest86. Got it!

[Alan McKinnon]

On Saturday 13 February 2010 19:43:47 Alan Mackenzie wrote:
> > There you go.  Most likely one little transistor that went belly up.  
> > Considering there are millions of those little devils on there, no 
> > surprise at all.
> 
> Oh, don't be like that!  You're saying, like, another "little" transistor
> will soon be going.  ;-)

Considering the enormous complexity of modern PC's, I'm always amazed that nay 
of the bits and pieces inside them ever work at all

:-)

> > Glad you got it sorted out and that is better than a lot of other
> > options.  Since it is new, I hope you have a good warranty that will
> > make it a cheap fix as well.
> 
> I bought my PC components from a premium quality shop, the sort that
> behaves like a gentleman and honours its guarantees.  Its email support
> gets back to you within an hour or so (in business hours).  The
> proprietor said I needed to send back both RAM sticks (since they have
> a joint serial number), but he's sending me a replacement pair first, so
> my machine remains working.

Nice to see that decent responsible vendors still exist out their who do 
TheRightThing(tm) for their customers, even the ones with little orders.

I get that level of service from Dell, but I think that's more because we're 
ordering 1000+ servers a year all with extended 3-5 year warranty and Next Day 
Corporate Service Contract

-- 
alan dot mckinnon at gmail dot com