GA.MCK%ISUMVS.BITNET@cunyvm.cuny.edu (Bret A. Mckee) (01/30/88)
I posted a note about parts of this some time ago. I have began a version of protected mode minix. Unfortunately it has been put on the back burner for a while due to certain unforseen events. Hopefully I will be able to get back to it soon. I began be eliminating the TDS segment model in favor of a true 32 bit pointer model. I have finished with this and now have a system in which all kernel <=> user data exchanges are done via 32 bit pointers. At present I am running in real mode and the function char far * umap(int process, char far * address) returns the pointer it was passed. When in protected mode it will create an LDT alias for the segment. This seemed to be the "clean" way of doing things. As a bonus it should be possible to painlessly implement the larger memory models under this scheme. The next step will be to convert the OS tables from their current format into the format of the corresponding hardware tables. This will allow me to test things in real mode first. When this conversion is done it will be time to actually begin executing in protected mode. I plan to implement the interrupt routines as tasks initially because of the "clean-ness" of this method. It provides complete seperation of the tasks. Comments are welcome. Bret Mckee Iowa State University Ames, Ia 50010 p.s. Please direct any mail responses to f2bam%isumvs.bitnet@cunyvm.cuny.edu which is not the account I am posting this from.