On 9/7/07, <b class="gmail_sendername">Peter Watkins</b> <<a href="mailto:peterw@tux.org" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">peterw@tux.org</a>> wrote:<div><span class="gmail_quote">
</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div bgcolor="#ffffff" text="#000000">
A couple weeks ago I upgraded my 0.19 system (based on Fedora Core 4
i386 &
Jarod's guide) to 0.20.x on Ubuntu 7.04 amd64, in preparation for
Schedules Direct.<br>
<br>
The system has been spontaneously rebooting ever since -- uptime
between 7 minutes and 9 hours. Before this change, it was rock solid.
Sometimes the HD3000/DVB tuner
would fail (and record 0 bytes), but the hardware & OS were very
stable. It /seems/ less stable when recording SD shows via IVTV and my
PVR 500, but it /has/ successfully recorded some two hour movies, so
it's not failing predictably. <br>
<br>
I've tried to test a number of components and came across
the MythTV/IVTV wiki entry warning "Do <i>not</i> attempt to enable
pre-emptive kernel locks in your kernel
configuration--the ivtv modules do not get along with them and will
cause <i>spontaneous reboots</i>." Is this possibly what I'm seeing?
How can I tell: first, if my amd64 kernel supports pre-emptive kernel
locks, and, second, if that's enabled? And can I disable them through
/proc or a boot option?<br>
<br>
What has changed: <br>
I replaced two older ATA drives (replaced Seagate 200G and 250G with a
matched pair of Seagate 500G), added a 1GB memory module (now has 2 GB
RAM), and installed
an additional 120mm case fan to help cool the drives & system.
Instead of Fedora Code 4 i386 on /dev/hda I now have Ubuntu 7.04 amd64
running on md raid (raid1 for /boot and /) -- which also means many
upgraded software packages, including IVTV and NVidia drivers. I've
replaced the 18 month old Antec TruePower II 550w power supply with a
new Sea Sonic 550w supply.<br>
<br>
Troubleshooting:<br>
The RAM checked out fine with an ~18 hour memtest86+ run. No errors
have
been reported on the drives by partitioning tools, SMART tools, etc.
I've tried removing the new RAM, moving the RAM around and
disconnecting the new case fan. I have disconnected a seldom-used SCSI
tape drive and physically removed its host adapter. I switched SATA
power cables for the SATA drives that I am still using. I have run
StressCPU from UBCD for over three hours without
trouble. I have not tried disconnecting either of the new 500G hard
drives. The system temperature looks OK. There's no lm_sensors support,
but when rebooting after normal use, the system temp is usually around
40 C and the CPU temp around 45 C. As stated above, I have replaced the
power supply. I have run a 'cp' for a 4 GB file in a while [] loop to
test the disk/md/lvm.<br>
<br>
Considering the previous stability, the new power supply, the success
with memtest86+, and the results of my hardware
tests, it seems very plausible that I'm having software trouble.<br>
<br>
System details:<br>
Athlon64 with 2 ATA and 2 SATA hard drives, using LVM atop "md" RAID1
arrays for the MythTV recordings (MySQL data is on the simple raid1 md
array for /). PS is a 550w Sea Sonic S12, which
gets AC through an APC UPS. Video is an Nvidia 6200 LE (AGP), feeding
the TV via DVI, using the Nvidia binary driver (rev 9631, I believe --
the version suggested by Ubuntu's proprietary driver manager app, using
NvAGP "2" for better picture quality). Three PCI cards: a pcHDTV HD3000
tuner used for ATSC, a Hauppauge PVR-500 used for SD/cable, and an old
Ensoniq 1371 sound card. One DVD-RW on the second IDE channel, and,
yes, a
floppy drive. I'm running the Ubuntu kernel "2.6.20-16-generic #2 SMP
Thu Jun 7 19:00:28 UTC 2007 x86_64 GNU/Linux" and Ubuntu's IVTV 0.10.1.
The config for that kernel (in /boot) has the following "preempt"
kernel settings:<br>
<br>
# CONFIG_PREEMPT_NONE is not set<br>
CONFIG_PREEMPT_VOLUNTARY=y<br>
# CONFIG_PREEMPT is not set<br>
CONFIG_PREEMPT_BKL=y<br>
<br>
Also, sometimes Ubuntu fails to boot, dropping me to a busybox prompt
(looks like this problem, but the piix initramfs approach has not
sufficed for me:
<a href="https://bugs.launchpad.net/ubuntu/+source/linux-source-2.6.20/+bug/106864" target="_blank" onclick="return top.js.OpenExtLink(window,event,this)">https://bugs.launchpad.net/ubuntu/+source/linux-source-2.6.20/+bug/106864
</a>;
I haven't tried the quiet/splash/irqpoll boot settings yet). With the
UPS, I'm not too worried about this; I only mention it in case someone
thinks it relevant to the spontaneous reboots.<br>
<br>
Any help appreciated!<br>
<br>
-Peter</div></blockquote><div><br>I feel your pain.<br><br>I can't tell if you're compiling your own kernel. If you are, don't do it. Open the box, pull out all the cards and memory and put them back, making sure all the cards are sitting nicely in their slots. Then install 32 bit version of Ubuntu (64 bit is not really worth it, been there done that), and don't do any customizations until you system stabilizes. Out of the box Ubuntu
7.04 works with ivtv and mythtv great.<br><br>Keep the box opened and point a room fan at it to remove any possibility that your problem is cooling related.
<br><br>If it still crashes, start removing components, starting with HD3000. Give the system enough time to crash in between and stop when it stops crashing. Sometimes putting cards in different pci slots helps.<br><br>
My current BE looks a lot like yours but instead of HD3000 I have
HDHomeRun (external, no pci issues), 1Gb ram and only one Seagate 500G
drive. I stopped doing all the lvm, raid/md funny stuff. Just plain XFS
on a dedicated tv disk. Recording and watching both SD/HD, system is
rock solid. I'm very careful about changes I make, always doing _one_ thing at the time, giving enough time in between to see if the change destabilized it (usually couple of days is good enough).<br>
<br>HTH<br></div></div>