Hi Christian.

While at SLES9, would you recommend hcp be used instead of vmcp at least
during boot?  The Basevol/Guestvol script uses hcp, but since it sounded
like vmcp was preferred over hcp, I changed the script to invoke vmcp
instead.

I see udev in the boot process before the vmcp is issued (Basevol/Guestvol
script). However, it is within 4 seconds. If it can take up to 10 seconds,
it sounds like the timing of this not consistent enough at SLES9.




09:24:45 VFS: Mounted root (ext2 filesystem).
Starting udev
Creating devices
09:24:46 Loading kernel/drivers/s390/char/vmcp.ko
z/VM CP interface loaded
.
.
.
09:24:49 ""
Basevol/Guestvol script begins... .""
Welcome to the Linux disk model userid!""
All directories will be in R/W mode.""
Basevol/Guestvol script ends.""
""



thanks again for your help.

susan

----------------------------------------------------------------------

Date:    Tue, 3 Jul 2007 09:30:58 +0200
From:    Christian Borntraeger <[EMAIL PROTECTED]>
Subject: Re: vmcp during boot results in  Error: Could not open device
/dev/vmcp: No such device

Am Montag, 2. Juli 2007 schrieb Susan Zimmerman:
> Hi Mark.
>=20
> I noticed /sys/class/misc/vmcp was 10 63
>=20
> OSA-LNX1:~ # cat /sys/class/misc/vmcp/dev
> 10:63
>=20
> So I deleted /dev/vmcp and manually created the node
>=20
> as 10 63
>=20
> and now it works just fine.
>=20
> Thanks for your help.

Yes, vmcp registers a misc device with dynamic minor number. Therefore it
i=
s=20
normally not a good idea to have a permanent device node for vmcp, unless
y=
ou=20
know the load sequence of all your misc devices. Instead udev should be
use=
d.=20
I know that this is getting trickier on an initrd.=20

There is a problem assigned with udev: device node creation takes time. So=
=20
after modprobe <module>, it can take some time to create a device node.=20
(usually milliseconds, but it can take longer)

Udev has improved a lot on SLES10 and RHEL5. On these systems you can=20
use "udevsettle" to be sure that udev has finished all its work that was=20
scheduled until now:

# modprobe vmcp
# udevsettle
# vmcp my fancy command

will always work as long as the kernel module was able to load.
Furthermore=
,=20
both systems use a tmpfs filesystem as /dev so your device folder is=20
re-initialized during boot without stale device nodes.=20


Things are much trickier on SLES9 and RHEL4. Both use an older udev
version=
=20
without udevsettle.
SLES9 does not use tmpfs. SLES9 also does not use udev for devices,
which=20
exist during boot, unless you enable the boot.udev init script.=20

RHEL4 uses udev for all devices, but has another issue. There is a race in
=
the=20
network hotplug scripts (I forgot the bugzilla number). In these script,=20
hotplug is temporarily disabled and udev can loose events. If udev misses
a=
n=20
event it will wait up to 10 seconds until it continues. Thats why on=20
RHEL4 "modprobe vmcp -> /dev/vmcp appears" will often take 10 seconds.

=2D-=20
IBM Deutschland Entwicklung GmbH
Vorsitzender des Aufsichtsrats: Martin Jetter
Gesch=E4ftsf=FChrung: Herbert Kircher=20
Sitz der Gesellschaft: B=F6blingen
Registergericht: Amtsgericht Stuttgart, HRB 243294

----------------------------------------------------------------------

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to