Red Hat Bugzilla – Bug 242262
Fedora 7 2.6.21-1.3 kernel has IDE disabled by default
Last modified: 2007-11-30 17:12:05 EST
Description of problem: When upgrading from FC6 to F7 via yum, the new kernel 2.6.21 omitted a bunch of IDE modules and includes that are needed on some hardware. Was this intended? A lot of equipment from online dedicated server centers still use IDE drives and users are going to have upgrade problems because of this. The IDE modules were included in 2.6.18 and this omission seems to be the largest omission when comparing the two kernel config files. It was not mentioned in the release notes that these were going to be disabled. Version-Release number of selected component (if applicable): kernel 2.6.18.fc6 kernel 2.6.21.fc7 How reproducible: On any working FC6 system using IDE controller, after upgrading to FC7, the machine will not reboot properly. Steps to Reproduce: 1. Install FC6 on IDE controlled machine 2. Upgrade FC6 to latest 3. Issue RPM command to download latest FC7 release file from repo 4. Issue yum update to go to FC7 5. Reboot will panic. Expected results: Existing IDE machines should be able to work with this release. Additional info: Will attach config-2.6.18, config-2.6.21 from Fedora. Will also attach .config of make oldconfig on 2.6.21.3 from config-2.6.18
Created attachment 155993 [details] Config files from FC6 and FC7 as well as working 2.6.21.3 config upgraded from FC6 config Config files from FC6 and FC7 as well as working 2.6.21.3 config upgraded from FC6 config
Found this in the release notes for FC7: http://docs.fedoraproject.org/release-notes/f7/en_US/sn-Kernel.html New IDE drivers that use the same libata code as the SATA drivers. [Important] IDE Device Names Changed The new IDE drivers now cause all IDE drives to have device names such as /dev/sdX instead of /dev/hdX. If the /etc/fstab or /etc/crypttab files reference these devices by name, they must be migrated before the system can access those partitions. Not sure if this would be the cause of it as rebooting remotely and updating through YUM, would not be able to diagnose if harddisks weren't accessible.
Need console output to get an idea why the system isn't booting, but indeed, the removal of the old IDE drivers was intentional.
I'm pretty sure this is a NOTABUG. We didn't update /etc/fstab or anything to change hardware locations to sdX as indicated in those release notes. Sorry for the confusion!
If this is NOT A BUG, then redhat/fedora appears to be subverted by microsift. I guess next step will be adopting ``advanced'' nameing scheme: C:\, D:\, etc. >:> This ``intentional removal of the old IDE drivers'' and transition to SCSI API leads to lots of problems. One critical point: scsi drives can only have 14 partitions, while IDE discs have 64 reserved minor numbers and can contain 62 partitions (zero is the whole disc and one partition is ``wasted' for EXT). If fedora maintainers cannot imagine installations using more than 16 partitions that doesn't mean they do not exist. And BTW, LVM is not an adequate replacement for such partitioning as partitions could be used as RAID members, while logical volumes do not. This is just an example (and this partitioning problem DOES really exist), there are other problems as well (performance, stability, reliability, etc). That's why I call this ``solution'' idiotic and now tend giving up using fedora, fortunatelly, there are other Linux distributions. :(