Steps to reproduce:
- Create parm file which includes CMSDASD parameter with uppercase hex, e.g. CMSDASD=2F1
- Punch and IPL the RHEL5.6 installation files
- Observe that the latter stages of /sbin/init:readcmsfile() fail to offline the device and rmmod dasd_eckd_mod and dasd_mod
- Specifically, the sysecho command uses the uppercase version whilst the /sys/bus/ccw/... path uses lowercase

Effect:
Returning from readcmsfile() without validating its final sysecho and rmmod commands leaves us in an undefined state. The correct operation of loader to insmod again, creating fresh /dev/dasd* devices defined by DASD= is broken. Ultimately loader fails to read the kickstart file specified by ks=hd:dasdb1:/path/to/kickstart.file and prompts the user.


Bug or feature?
I vote bug, given that the module itself and parse_dasd() functions in linuxrc.s390 take care to allow both uppercase and lowercase hex for dasd device addresses; because the documentation does not specify case-sensitivity; because uppercase DASD= addresses are acceptable within the CONF file.


Suggested fixes:
The comment in readcmsfile() would work ("more robust ... printf"), but I'd suggest lowercasing the address as soon as it's read in MAIN:#Parse configuration.

Also, test the return values of sysecho and rmmod; do not ignore errors. Test the return value of readcmsfile itself.

Also, make loader's rmmod more verbose on failure, like real rmmod. (This would have greatly helped in tracking down the problem.)