Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
[gentoo-user] Any way of tracing kernel freezes?
9 views
Skip to first unread message
Frank Steinmetzger
Jun 13, 2013, 9:50:02 AM6/13/13
to
Hello list,
recently my netbook got the habit of freezing sporadically. It happens
apparently out of the blue. Until now I thought it might have something
to do with audio playback, because I remember always having music
playing when it happens. But just now it happened without music:
One second I was surfing along, the other second -- dead. While I was
already 3 hours in of a 4-hour emerge. :-(
If music is playing at a freeze, the last bit of it (less than a
second's worth) is looped indefinitely. The screen output is still
there, but there is no reaction to any input, not even sysrequests.
I also tried SSHing in without success.
I have no experience with Kernel debugging. Is there any possibility at
all of finding out what the problem is? Perhaps enable some log output?
Being a netbook, there is no serial port (which I don't have the means
to connect to anyway). In the hopes that the problem started with a 3.9
kernel, I am building a 3.8.13 again (using kernel.org archives).
I ran a memtest a few months back and stopped it after 8 successful
passes. I might try that again as soon as I find out how¹, but I'd
think that a corrupt memory would cause something different than a full
freeze.
I'm running a 64 bit kernel and 32 bit userland with march=atom. I'd
really hate to find out the machine has a hardware problem, it's a very
neat piece of silicon with a great screen. Who cares that the 6.5 Watt
CPU is a tad slow. It's perfect for a console environment. :-)
¹ I have the in-kernel memtest enabled, so I tried booting it with the
"memtest" option right after the last freeze happened. But all I got
was a black screen with blinking cursor.
--
Gruß | Greetings | Qapla’
Please do not share anything from, with or about me with any Facebook service.
The circle is the parallel to the dot.
recently my netbook got the habit of freezing sporadically. It happens
apparently out of the blue. Until now I thought it might have something
to do with audio playback, because I remember always having music
playing when it happens. But just now it happened without music:
One second I was surfing along, the other second -- dead. While I was
already 3 hours in of a 4-hour emerge. :-(
If music is playing at a freeze, the last bit of it (less than a
second's worth) is looped indefinitely. The screen output is still
there, but there is no reaction to any input, not even sysrequests.
I also tried SSHing in without success.
I have no experience with Kernel debugging. Is there any possibility at
all of finding out what the problem is? Perhaps enable some log output?
Being a netbook, there is no serial port (which I don't have the means
to connect to anyway). In the hopes that the problem started with a 3.9
kernel, I am building a 3.8.13 again (using kernel.org archives).
I ran a memtest a few months back and stopped it after 8 successful
passes. I might try that again as soon as I find out how¹, but I'd
think that a corrupt memory would cause something different than a full
freeze.
I'm running a 64 bit kernel and 32 bit userland with march=atom. I'd
really hate to find out the machine has a hardware problem, it's a very
neat piece of silicon with a great screen. Who cares that the 6.5 Watt
CPU is a tad slow. It's perfect for a console environment. :-)
¹ I have the in-kernel memtest enabled, so I tried booting it with the
"memtest" option right after the last freeze happened. But all I got
was a black screen with blinking cursor.
--
Gruß | Greetings | Qapla’
Please do not share anything from, with or about me with any Facebook service.
The circle is the parallel to the dot.
Nikos Chantziaras
Jun 13, 2013, 10:10:01 AM6/13/13
to
On 13/06/13 16:47, Frank Steinmetzger wrote:
> recently my netbook got the habit of freezing sporadically. [...]
>[...]
still worth a shot. It's very easy. Emerge "sys-apps/memtest86+" and
add this grub entry:
title=Memtest86+
root (hd0,0)
kernel /boot/memtest86plus/memtest.bin
(Adapt the "hd0,0" of course to be the same disk as the one you're using
to boot your kernel.)
That's it. Now your boot menu will include a "Memtest86+" option. This
is for grub 1 though. If you migrated to grub 2 by now, then I don't
know how that boot entry would look like. I suspect it will be some
300-line monstrosity or something :-|
> recently my netbook got the habit of freezing sporadically. [...]
>[...]
> I ran a memtest a few months back and stopped it after 8 successful
> passes. I might try that again as soon as I find out how¹, but I'd
> think that a corrupt memory would cause something different than a full
> freeze.
It usually manifests in segfaults that seem to come at random. But it's
> passes. I might try that again as soon as I find out how¹, but I'd
> think that a corrupt memory would cause something different than a full
> freeze.
still worth a shot. It's very easy. Emerge "sys-apps/memtest86+" and
add this grub entry:
title=Memtest86+
root (hd0,0)
kernel /boot/memtest86plus/memtest.bin
(Adapt the "hd0,0" of course to be the same disk as the one you're using
to boot your kernel.)
That's it. Now your boot menu will include a "Memtest86+" option. This
is for grub 1 though. If you migrated to grub 2 by now, then I don't
know how that boot entry would look like. I suspect it will be some
300-line monstrosity or something :-|
Neil Bothwick
Jun 13, 2013, 11:20:03 AM6/13/13
to
On Thu, 13 Jun 2013 17:05:15 +0300, Nikos Chantziaras wrote:
> add this grub entry:
>
> title=Memtest86+
> root (hd0,0)
> kernel /boot/memtest86plus/memtest.bin
>
> (Adapt the "hd0,0" of course to be the same disk as the one you're
> using to boot your kernel.)
>
> That's it. Now your boot menu will include a "Memtest86+" option.
> This is for grub 1 though. If you migrated to grub 2 by now, then I
> don't know how that boot entry would look like. I suspect it will be
> some 300-line monstrosity or something :-|
With GRUB2 you don't have to add anything. /etc/grub.d/39_memtest86+ is
> add this grub entry:
>
> title=Memtest86+
> root (hd0,0)
> kernel /boot/memtest86plus/memtest.bin
>
> (Adapt the "hd0,0" of course to be the same disk as the one you're
> using to boot your kernel.)
>
> That's it. Now your boot menu will include a "Memtest86+" option.
> This is for grub 1 though. If you migrated to grub 2 by now, then I
> don't know how that boot entry would look like. I suspect it will be
> some 300-line monstrosity or something :-|
created by the ebuild, containing slightly less than 300 lines
#!/bin/sh
exec tail -n +3 $0
menuentry "memtest86+ 4.20" {
linux16 /boot/memtest86plus/memtest
}
menuentry "memtest86+ 4.20 (netbsd)" {
insmod bsd
knetbsd /boot/memtest86plus/memtest.netbsd
}
This is added to your menu automatically by grub2-mkconfig.
This is one of the great things about GRUB2, it allows install scripts,
whether they are package installers or complete distro installers, to add
to your GRUB menu without disturbing, or even needing to know about, your
existing configuration.
--
Neil Bothwick
Give a man a fish and you feed him for a day; teach him
to use the Net and he won't bother you for weeks.
William Kenworthy
Jun 13, 2013, 5:10:02 PM6/13/13
to
BillK
Paul Hartman
Jun 13, 2013, 5:20:01 PM6/13/13
to
menuentry "Memtest86+ 4.20" {
linux16 /memtest86plus/memtest.bin
}
That's from my grub.cfg... I don't use the grub auto-configuration
tools. I just made a manual grub.cfg like in the grub1 days. It is
quite similar to the old grub syntax, but more can do more stuff.
BTW there is a new version 5.00 of memtest86+ which is on rc1 release
right now, it supports multi-core processors (and tests/uses them),
shows system temperature while testing, is much faster than the old
one, and has a few new tests. There is a link to it on the memtest
forums.
Frank Steinmetzger
Jun 14, 2013, 10:50:02 AM6/14/13
to
On Fri, Jun 14, 2013 at 05:04:15AM +0800, William Kenworthy wrote:
> On 13/06/13 22:05, Nikos Chantziaras wrote:
> >> but I'd think that a corrupt memory would cause something different
> >> than a full freeze.
> >
> > It usually manifests in segfaults that seem to come at random. But it's
> > still worth a shot. It's very easy. Emerge "sys-apps/memtest86+" and
> > add this grub entry:
I found my old USB Gentoo which has memtest installed (and which I used
> On 13/06/13 22:05, Nikos Chantziaras wrote:
> >> but I'd think that a corrupt memory would cause something different
> >> than a full freeze.
> >
> > It usually manifests in segfaults that seem to come at random. But it's
> > still worth a shot. It's very easy. Emerge "sys-apps/memtest86+" and
> > add this grub entry:
for my first test also). I let it run for two passes, both successful.
So the RAM seems fine.
> Temperature? - check for dust puppies clogging the heatsinks, cooling etc.
again right now (3.5 hours in), and it never goes above 66°. Besides, the
freezes also happened when I didn’t do anything heavy. Just surfing
(though Firefox can also be heave for an Atom ^^).
Anyhoo, I’m running 3.8.13 again now. I didn’t have my config anymore,
so I oldconfig'ed it from 3.9. Let’s see whether it’s more stable. If
yes, hm... I can’t really report this to the kernel devs: “My netbook
freezes since 3.9, that’s all I know. Here, have my configs.” :-I
--
Gruß | Greetings | Qapla’
Please do not share anything from, with or about me with any Facebook service.
The problem with FORTRAN jokes is that they never fit into a single li
Gruß | Greetings | Qapla’
Please do not share anything from, with or about me with any Facebook service.
Thomas Mueller
Jun 15, 2013, 2:00:02 AM6/15/13
to
Excerpt from Frank Steinmetzger:
> I found my old USB Gentoo which has memtest installed (and which I used
> for my first test also). I let it run for two passes, both successful.
> So the RAM seems fine.
> > Temperature? - check for dust puppies clogging the heatsinks, cooling etc.
> It's a netbook with a 6.5 Watt CPU. I'm going through the big emerge
> again right now (3.5 hours in), and it never goes above 66°. Besides, the
> freezes also happened when I didn't do anything heavy. Just surfing
> (though Firefox can also be heave for an Atom ^^).
> Anyhoo, I'm running 3.8.13 again now. I didn't have my config anymore,
> so I oldconfig'ed it from 3.9. Let's see whether it's more stable. If
> yes, hm... I can't really report this to the kernel devs: "My netbook
> freezes since 3.9, that's all I know. Here, have my configs." :-I
Is your temperature of 66° F or C?
System temperature or surrounding room temperature?
I have an old computer whose fan has quit as happened once before.
CPUs generate considerable heat, I see system temperature and realize the fan is much more critical than whether the room temperature is a chilly 20 C or sweaty (for humans) 35 C.
I don't use that old 2001 computer much, am getting ready to put together a new computer from parts to run FreeBSD and Linux, likely Gentoo; otherwise I'd order a Socket A fan.
When I do use that old computer, I open the case and prop a hair dryer to run at low, ambient-temperature air pointed at the CPU. This keeps the CPU down to 47 C according to the BIOS/CMOS screen. This is cool for the CPU if not for us humans.
Tom
> I found my old USB Gentoo which has memtest installed (and which I used
> for my first test also). I let it run for two passes, both successful.
> So the RAM seems fine.
> > Temperature? - check for dust puppies clogging the heatsinks, cooling etc.
> It's a netbook with a 6.5 Watt CPU. I'm going through the big emerge
> again right now (3.5 hours in), and it never goes above 66°. Besides, the
> freezes also happened when I didn't do anything heavy. Just surfing
> (though Firefox can also be heave for an Atom ^^).
> Anyhoo, I'm running 3.8.13 again now. I didn't have my config anymore,
> so I oldconfig'ed it from 3.9. Let's see whether it's more stable. If
> yes, hm... I can't really report this to the kernel devs: "My netbook
> freezes since 3.9, that's all I know. Here, have my configs." :-I
System temperature or surrounding room temperature?
I have an old computer whose fan has quit as happened once before.
CPUs generate considerable heat, I see system temperature and realize the fan is much more critical than whether the room temperature is a chilly 20 C or sweaty (for humans) 35 C.
I don't use that old 2001 computer much, am getting ready to put together a new computer from parts to run FreeBSD and Linux, likely Gentoo; otherwise I'd order a Socket A fan.
When I do use that old computer, I open the case and prop a hair dryer to run at low, ambient-temperature air pointed at the CPU. This keeps the CPU down to 47 C according to the BIOS/CMOS screen. This is cool for the CPU if not for us humans.
Tom
Frank Steinmetzger
Jun 19, 2013, 9:30:01 PM6/19/13
to
On Fri, Jun 14, 2013 at 10:56:51PM -0700, Thomas Mueller wrote:
> Excerpt from Frank Steinmetzger:
> System temperature or surrounding room temperature?
CPU Temp. It’s the only sensor I’ve got.
> I have an old computer whose fan has quit as happened once before.
The fan is fine.
> CPUs generate considerable heat […]
On an on-topic sidenote: I had a freeze the other day with the 3.8
kernel. So it could likely be a hardware issue.
After I accidentially cut the WiFi cable where it goes through the
hinge, I installed another netbook's antenna and connected it to the
metal foil in the back of the lid. That can't possibly be the culprit,
can it? WiFi is working fine, as far as I can tell.
> Excerpt from Frank Steinmetzger:
> > > Temperature? - check for dust puppies clogging the heatsinks, cooling etc.
>
> > It's a netbook with a 6.5 Watt CPU. I'm going through the big emerge
> > again right now (3.5 hours in), and it never goes above 66°. Besides, the
> > freezes also happened when I didn't do anything heavy. Just surfing
> > (though Firefox can also be heave for an Atom ^^).
>
> > Anyhoo, I'm running 3.8.13 again now. I didn't have my config anymore,
> > so I oldconfig'ed it from 3.9. Let's see whether it's more stable. If
> > yes, hm... I can't really report this to the kernel devs: "My netbook
> > freezes since 3.9, that's all I know. Here, have my configs." :-I
>
> Is your temperature of 66° F or C?
The proper degrees of course ;-) i.e. Celsius.
>
> > It's a netbook with a 6.5 Watt CPU. I'm going through the big emerge
> > again right now (3.5 hours in), and it never goes above 66°. Besides, the
> > freezes also happened when I didn't do anything heavy. Just surfing
> > (though Firefox can also be heave for an Atom ^^).
>
> > Anyhoo, I'm running 3.8.13 again now. I didn't have my config anymore,
> > so I oldconfig'ed it from 3.9. Let's see whether it's more stable. If
> > yes, hm... I can't really report this to the kernel devs: "My netbook
> > freezes since 3.9, that's all I know. Here, have my configs." :-I
>
> Is your temperature of 66° F or C?
> System temperature or surrounding room temperature?
> I have an old computer whose fan has quit as happened once before.
> CPUs generate considerable heat […]
> When I do use that old computer, I open the case and prop a hair dryer
> to run at low, ambient-temperature air pointed at the CPU.
Please re-read the first paragraph you quoted from me in your reply.
> to run at low, ambient-temperature air pointed at the CPU.
On an on-topic sidenote: I had a freeze the other day with the 3.8
kernel. So it could likely be a hardware issue.
After I accidentially cut the WiFi cable where it goes through the
hinge, I installed another netbook's antenna and connected it to the
metal foil in the back of the lid. That can't possibly be the culprit,
can it? WiFi is working fine, as far as I can tell.
--
Gruß | Greetings | Qapla’
Please do not share anything from, with or about me with any Facebook service.
I hate being bi-polar. It’s fantastic!
Gruß | Greetings | Qapla’
Please do not share anything from, with or about me with any Facebook service.
0 new messages