From Bugzilla Helper: User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.6) Gecko/20040113 Description of problem: If one requests grub to generate a new device.map file and the machine has a SmartArray Controller, which names its devices in /dev/cciss/c0d0, the resultant file only contains (fd0) /dev/fd0 Version-Release number of selected component (if applicable): grub-0.94-4 How reproducible: Always Steps to Reproduce: 1.grub --device-file=/tmp/foo (where foo does not exist) 2. 3. Actual Results: file only contains (fd0) /dev/fd0 Expected Results: (fd0) /dev/fd0 (hd0) /dev/cciss/c0d0 Additional info: whatever anaconda does, it DOES populate /boot/grub/device.map correctly as can be seen on any successfully installed system.
I found the source of the problem... In the module lib/device.c, the section of code that checks for SCSI disks, calls the function get_scsi_disk_name(), which in turn tries to find all the scsi devices with the name of /dev/sda-p and therefore never sees the cciss devices, which is why nothing ever gets written to the device.map file. To test my theory I changed get_scsi_disk_name to look for SCSI devices with the name format from: sprintf (name, "/dev/sd%c", unit + 'a'); to: sprintf (name, "/dev/cciss/c0d%c", unit + '0'); The first time I tried this my system crashed because it couldn't handle 16 device names, so I changed the look calling this to only look at devices 0,1 and it created the following device map: (fd0) /dev/fd0 (hd0) /dev/cciss/c0d0 (hd1) /dev/cciss/c0d1 It feels like the thing to do is build a new function called something like get_cciss_disk_name and call it in addition to the get_scsi_disk_name with the appropriate look counter. If there is anything more I can provide, let me know...
http://testing.fedora.redhat.com/tree/ Is this report still an issue in RHEL4 beta1 or beta2, or FC3? Try the FC3 RC install from the URL above. Is Bug #125808 related to this issue?
Fixed in grub-0.95-4 and later
I believe this bug correlates to the following in the changelog: * Mon Aug 01 2005 Peter Jones <pjones> - 0.93-4.8 - minor fix to the --recheck fix. So if this is the case, this bug should be closed. J