This request was evaluated by Red Hat Product Management for
inclusion, but this component is not scheduled to be updated in
the current Red Hat Enterprise Linux release. If you would like
this request to be reviewed for the next minor release, ask your
support representative to set the next rhel-x.y flag to "?".

I think we should be writing out a file to /etc/modprobe.d rather than placing:

    OPTIONS="layer=2"

in the ifcfg file.  For example, we should be writing /etc/modprobe.d/qeth.conf with this line:

    options qeth layer=2

Because it doesn't look like the OPTIONS variable from the ifcfg file in RHEL-5 is not used by the scripts in /etc/sysconfig/network-scripts.  The modules are just loaded using modprobe, so options need to be put in the modprobe configuration files.

Can someone at IBM confirm?

(In reply to comment #6)
> ------- Comment From dcantrell 2008-10-27 17:50:50 EDT-------
> I think we should be writing out a file to /etc/modprobe.d rather than placing:

No, William's configurations files all look sane to me. Device configuration should have been done before automatically by /lib/udev/ccw_init, which has been interpreting and applying SUBCHANNELS, PORTNAME,  CTCPROT, and OPTIONS since earlier releases. The qeth driver module does not have any parameters (modinfo qeth) since things such as OPTIONS (including layer2 or portno) are per device and not per driver.

> Because it doesn't look like the OPTIONS variable from the ifcfg file in RHEL-5
> is not used by the scripts in /etc/sysconfig/network-scripts.

Right, network-scripts (ifup*, network-functions) don't use OPTIONS since they only configure network layer stuff and not the device/hardware themselves.

We have to figure out, why the device wouldn't come up and qeth prints error messages to the kernel log.

Hi David,
to the theory I would recommend to check the Device Drivers, Features and Commands Book from October 2005 stream in developerWorks, chapter 9, pages 112-116,124 and  110:
http://www.ibm.com/developerworks/linux/linux390/october2005_documentation.html
That is, layer2 is not a module parameter, but an attribute in sysfs.

OPTIONS="attribute1=value1 attribute2=value2"
in ifcfg-ethXX files has been working since RHEL 4.4 and RHEL 5 thanks to the file/script ccw_init, which from RHEL 4 to RHEL 5 got moved to:
/lib/udev/ccw_init
and it is called by udev from /etc/udev/rules.d/55-ccw.rules 

Maybe you could check with Harald (if he still is the initscripts/udev mantainer for RHEL 5) if any changes where done since RHEL 5.2 in initscripts/udev. As mentioned, in RHEL 5.2 was working.

Hope it helps,
Gonzalo Muelas Serrano.

I see what's happening.  We're writing out this line to the ifcfg file:

    OPTIONS="layer=2"

The /lib/udev/ccw_init script is expecting (see lines 37-42 of ccw_init):

    OPTIONS="layer2=1"

Here's a patch to anaconda to fix up what we write out to the ifcfg file:

diff --git a/loader2/net.c b/loader2/net.c
index a40ee18..e595b59 100644
--- a/loader2/net.c
+++ b/loader2/net.c
@@ -1849,7 +1849,7 @@ int writeNetInfo(const char * fn, struct networkDeviceConfig * dev) {
         fprintf(f, "CTCPROT=%s\n", dev->ctcprot);
 
     if (dev->layer2 && !strcmp(dev->layer2, "1"))
-       strcat(osa_opts, "layer=2");
+       strcat(osa_opts, "layer2=1");
     else if (dev->subchannels)
        fprintf(f, "ARP=no\n");
     if (dev->portno && !strcmp(dev->portno, "1")) {

To init/activate the device after writing the configuration, s.th. like this is needed:

# SUBSYSTEM="ccw" DEVPATH="bus/ccwgroup/drivers/qeth/${SUBCHANNELS//,*/}" /lib/udev/ccw_init

(In reply to comment #9)
> ------- Comment From dcantrell 2008-10-28 18:04:45 EDT-------

Good spotting, David!

> diff --git a/loader2/net.c b/loader2/net.c
> @@ -1849,7 +1849,7 @@ int writeNetInfo(const char * fn, struct
> if (dev->layer2 && !strcmp(dev->layer2, "1"))
> -       strcat(osa_opts, "layer=2");
> +       strcat(osa_opts, "layer2=1");

This seems to be a regression which came in with support for portno:

diff -rup rhel5.2/anaconda-11.1.2.113-1.src.rpm_FILES/anaconda-11.1.2.113/loader
2/net.c rhel5.3-alpha-20081006/anaconda-11.1.2.136-2.src.rpm_FILES/anaconda-11.1
.2.136/loader2/net.c
@@ -1733,10 +1847,20 @@ int writeNetInfo(const char * fn, struct
fprintf(f, "NETTYPE=%s\n", dev->nettype);
if (dev->ctcprot)
fprintf(f, "CTCPROT=%s\n", dev->ctcprot);
+
if (dev->layer2 && !strcmp(dev->layer2, "1"))
-        fprintf(f, "OPTIONS=\"layer2=1\"\n");
+       strcat(osa_opts, "layer=2");
### this is where the regression came in ###
else if (dev->subchannels)
fprintf(f, "ARP=no\n");
+    if (dev->portno && !strcmp(dev->portno, "1")) {
+       if (strlen(osa_opts) != 0) {
+           strcat(osa_opts, " ");
+       }
+       strcat(osa_opts, "portno=1");
+    }
+    if ((strlen(osa_opts) > 0))
+        fprintf(f, "OPTIONS=\"%s\"\n", osa_opts);
+
if (dev->macaddr)
fprintf(f, "MACADDR=%s\n", dev->macaddr);
Since layer2 used to be the only options, it used to write out the whole string "OPTIONS=\"layer2=1\"\n". Now that we have another OPTION portno, the arguments to OPTION are accumulated in osa_opts and finally the OPTION string gets written out.

Thanks for your fix.

This bugzilla has Keywords: Regression.  

Since no regressions are allowed between releases, 
it is also being proposed as a blocker for this release.  

Please resolve ASAP.

Patch will be in anaconda-11.1.2.151-1

This is verified fixed in Snapshot1.

An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on therefore solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.

http://rhn.redhat.com/errata/RHBA-2009-0164.html