2.6.14, 2.6.15 md mystery partially solved (was: 2.6.15: mdrun, udev

2.6.14, 2.6.15 md mystery partially solved (was: 2.6.15: mdrun, udev

[Moshe Yudkowsky]

Last month, I posted about the a mystery: I could not boot into 2.6.15 
when using RAID. As of yesterday I found that I couldn't boot into 
2.6.14 any more, either -- right now I'm back on 2.6.12. At the time the 
theory was that udev and md were in some sort of race condition.

The lack of 2.6.14 boot motivated me to see if I could solve the 2.6.15 
problem.

Part of the mystery is solved: someone -- I have to assume it was a 
prank -- renamed the md module the "md-mod" module. Therefore it never 
showed up in the boot image. My mind boggles at the thought of renaming 
the module without warning and without catching it during depmod. Oh, 
and another gotcha: "mdadm --auto" doesn't create the /dev/md directory; 
you have to do it yourself. A bit of missing documentation there. (Sorry 
  if I sound a bit bitter, but renaming the module retrospectively? It's 
a major gotcha.)

Other problems, however, persist in 2.6.15: the standard script "mdrun" 
tries to create a /dev/md/0 even though I don't have an array by that 
name.  I created a custom hook and custom script which uses 
/etc/mdamd.conf to create the correct devices. As a side benefit of 
these custom scripts, which use "mdadm --auto", I don't get the the 
create dozens of /dev/md* entries created by the standard "mdrun."

Unfortunately I get a kernel panic when I try to boot the 2.6.15 image, 
for unknown reasons; I can't even tell where the problem occurs just 
yet. But as sad as that may be, at least this "md" problem turns out not 
to be a udev problem.

I'd send the partial solution off to the Debian developers, but for the 
life of me I can't figure out who should get the solution.

So thanks to everyone who commented during the previous thread!

-- 
  Moshe Yudkowsky
  work: http://www.Disaggregate.com
  book: http://www.PebbleAndAvalanche.com


-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid\x110944&bid$1720&dat\x121642
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

Re: 2.6.14, 2.6.15 md mystery partially solved (was: 2.6.15: mdrun,

[Moshe Yudkowsky]

If anyone is interested, the reason I couldn't boot 2.6.15 wasn't udev 
-- it was initramfs-tools. Several separate bugs!

* the md module wasn't being loaded because it'd been renamed to md-mod, 
and even when that was fixed md-mod wasn't being loaded because modprobe 
doesn't put out the insmod line we'd expect from the modprobe documentation.

* mdrun attempts to create an /dev/md/0. I gave up and replaced mdrun 
with "mkdir /dev/md; mdadm --scan --auto".

<http://bugs.debian.org/cgi-bin/bugreport.cgi?bug55580> for full details.

And I'm pleased to report that udev properly creates the symlinks from 
/dev/mdN to /dev/md/N -- mdrun's symlinking is not required.

-- 
  Moshe Yudkowsky
  work: http://www.Disaggregate.com
  book: http://www.PebbleAndAvalanche.com


-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid\x110944&bid$1720&dat\x121642
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel