On 08/20/2014 09:09 PM, Jacco Ligthart wrote:
* I quite often see messages like "Unknown HZ value! (45) Assume 100."
which is probably due a kernel/system mismatch somewhere (the kernel in
this image is not Redsleeve !)
I have seen this on other distros, too. Not sure what the cause is but
it is harmless.
I also think it is harmless, but if you google it, you'll find
references to a rootkit.
f.i.
http://www.bigismore.com/web-server-security/unknown-hz-value-assume-100-youve-been-hacked/
It's easy enough to verify whether procps binaries have been replaced by
checking the package signatures. For this you need signed packages,
though, so if you are using an image that predates the signed packages,
you might first want to
yum reinstall `rpm -qa`
I think it is a bug in procps, but I did not find an 'agreed' patch. The
issue is ofcourse due to a non RS6 kernel.
I'm not really sure about the root cause. EL6 should have procps new
enough that it doesn't matter.
* In a previous attempt to use redsleeve I found, that I could not
compile new kernel modules due to missing kernel source.
The way around this would be to build el6 rpms of the Pidora kernel
source rpms. They would probably need a bit of tweaking but it
shouldn't be insurmountable.
If you provide a suitable source rpms I'm happy to include them in the
official packages. It would probably help work around package build
issues on some packages when using mock, too, since the only kernel
included is the Marvell Kirkwood one, and even that is ancient since
the official sources are broken for everything except x86, and even
then it sometimes only works due to luck with the .config file used,
e.g.:
https://bugzilla.redhat.com/show_bug.cgi?id=773107
Hmm, I'm not sure if what I just did to the raspi image is compatible
with this.
I updated the kernel to 3.12.26+ which is the "official" latest version from
https://github.com/raspberrypi/firmware
(for reference with the rasbian-esque tool
https://github.com/Hexxeh/rpi-update)
So there was no building at all.
Indeed, you'd need to get the source and .config and roll those into a RPM.
After this I tried something similar to this:
http://www.raspberrypi.org/forums/viewtopic.php?p=315376#p315376
which resulted in a kernel source tree with which I could build
additional modules.
Take a look at the archived Fedora src.rpm for kernels of similar
vintage. Look at the .spec file, and replace the source tar ball and
config file(s) with the Pi specific one, and rebuild it from that. The
end result should be a packaged kernel for the pi.
Ideally you should also include modules for all the non platform
specific things that are in the standard EL6 kernel.
Total result is thus that I have a couple of checkouts from git, which
we could ofcourse package. But I'm not sure if this kernel source tree
is good/smart to use for non-raspi projects.
It doesn't matter if it is not a kernel that will work on other boards.
It will work on Pis, and it will still bring the relevant headers to
ensure that dependencies for them in some packages that require them can
be satisfied.
Question: all files under /lib/firmware are not installed by an RPM. can
you remember where they came from? should that be replaced by firmware
from kernel-firmware.armv5tel ?
It doesn't really matter where they come from. I think the standard set
comes from the kernel-firmware package, but I may have dropped in a few
that are handy for other devices. If you have a Pi specific kernel for
use with a Pi, use all the firmware blobs from that as that will have
been much more thoroughly tested on that device.
* I think pidora has some scripts somewhere to "resize the image" and to
"create a swap file", should we include them into redsleeve somewhere,
or is that to big a deviation from PNAELV
It's not a deviation at all. A script that puts a swap partition at
the end of the media and runs "resize2fs /" is a minor change.
I found that they have those scripts in firstboot. which is quite nice.
Yup, that is probably a good way to do it.
I'll check if the modules are compatible with our firstboot
I think it would be good to install firstboot on such an image. on the
downside, firstboot "needs" an awful lot of packages, including all
kinds of X-libs.
Not ideal. :(
* I still miss the RTC on these machines. is there a script somewhere
that sets the time to the last known date, so that (in case of missing
network connectivity) the clock never goes backwards. maybe something
like this: http://zenit.senecac.on.ca/wiki/index.php/Nortc
Not in EL6, but I even Chromebooks have interesting issues syncing RTC
to hardware clock. Check the wiki page for those, a look at the script
in /etc/sysconfig/modules/. You should be able to replace that script
with something more appropriate for other machines.
Are you sure it doesn't have a RTC? It could be that it just doesn't
sync the RTC with the hardware clock, same as the chromebook.
Dead sure. Actually, I personally don't have that much of a problem. I
bought some RTC's cheap on ebay :)
my /etc/sysconfig/modules/rtc.modules looks like:
#!/bin/sh
/sbin/modprobe rtc-ds1307
echo ds1307 0x68 > /sys/class/i2c-adapter/i2c-1/new_device
sleep 1
/usr/sbin/hwclock -s
Indeed, the best solution to a lack of RTC is to get an RTC. :)
Gordan
_______________________________________________
users mailing list
[email protected]
http://lists.redsleeve.org/mailman/listinfo/users
