Quote:
Originally Posted by jefro
|
Well, those deal with how the kernel and initrd are built and how they operate once they are loaded. In order to get those files loaded, GRUB does have internal knowledge of filesystem structure and can parse that structure without, technically**, mounting it. It is much the same as the way
fsck,
debugfs, and the like parse the internal structure of an unmounted filesystem.
In the case of GRUB legacy, the stage 1.5 bootloader is built with knowledge of
one filesystem, that being the filesystem that contains /boot.
GRUB2 starts out in a similar manner, but once the bootloader has gained access to the /boot/grub2 directory there are many loadable modules available to handle various filesystems as well as video drivers, networking, encryption, USB serial devices, etc. All filesystems are handled by parsing their structure and not by mounting them. Mounting would imply that there is some mount point directory on which to mount them, and there is not (yet) a root filesystem on which to do that.
** Note that the word "mount" here is being used in the specific sense of attaching a filesystem to a point on the overall directory tree and not in the more general sense of "make available for processing," such as mounting a device that contains a tar archive.