> I don't know what Jeronimo has against initrd's, but I can tell you > what I have against them. > > An initrd complicates things. > It's one more thing to remember to build. > It's one more thing that can potentially break. the good news is that make install makes this easy and automatic > > If you just build into the kernel whatever you need to get to the > point of mounting the root fs (or want to have available early) and > then anything else you need as modules, then an initrd is pretty > pointless, and it's a much more simple setup IMO. it's not if you need udev to mount your root filesystem for example, or mount-by-label. And before you say "you never need that", that is only the case for simple setups. More complex setups don't have fixed device nodes for disks... (like fiber channel or SAS or USB storage or ..) SELinux is another case.. "init" needs to start after selinux policy is loaded iirc... so initrd is needed for that. > > > It's not like using an initrd has drawbacks that I know of, nor is it > > hard; if you use "make install" it's automatic as I said, and that's a > > convenient thing to use anyway (because it does the bootloader stuff for > > you) > > Personally I consider "make install" dangerous. > > First of all it assumes that I want my kernel to be named > /boot/vmlinuz and happily overwrites any previous kernel image that > may exist by that name. So if my new kernel doesn't boot, and I only > have that one entry in my lilo.conf, then I'm in trouble and have to > go find a CD to boot from to recover. hmmm I've not seen that; in fact it uses the vmlinuz-2.6.12-extraversion layout on my systems, same for System.map etc. -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
