On Fri, Jul 12, 2013 at 3:57 PM, Hesham Moustafa <heshamelmat...@gmail.com> wrote: > Hi, > > The current test cases for libmm use hard-coded memory region addresses to > apply memory attributes on. We want to remove that BSP dependency and hide > any target specs to enable test cases to run for any BSP. I would suggest > some solutions for your review : > > - Use the current allocators like partition/region managers. This should be possible. I think at least one of them was initially designed to support MMU protection in either the ORKID or RTEID standards. I can hunt down the document that describes it. We converted the design documents during last GCI (code-in).
> - Define an area of memory in linkcmds that acts as a pool for libmm to > dynamically allocate and apply attributes for. It does not make sense to carve out part of the memory address space just for the libmm use. > - Use the free/busy list (stable enough ?) It's not ready yet, but it would be a good way to manage the memory by providing page-sized "buffers". What you would end up with is something very similar to a slab allocator. > > Would it make sense to enable the user to specify hard-coded addresses (that > are not within heap or workspace) in the application code ? > Yes. Other ideas that might make sense would include mapping a relationship between names and memory regions. Perhaps with a filesystem to specify the names, and nodes (files) that define the memory region and attributes. > Regards, > Hesham > > _______________________________________________ > rtems-devel mailing list > rtems-devel@rtems.org > http://www.rtems.org/mailman/listinfo/rtems-devel > _______________________________________________ rtems-devel mailing list rtems-devel@rtems.org http://www.rtems.org/mailman/listinfo/rtems-devel