Thanks Axel, Stupid of me to point in the direction of app_server. So in my previous message a search and replace should be done for app_server -> kernel :- ) > We don't yet support this flag. And we also don't have the modules (or whatever is needed) for it I take it then. The kernel itself relies on the system BIOS then about caching setup etc I take it as well. (as it should be according to some Intel spec now read) > There is no point in publishing these interfaces - they are for > internal kernel use only (and only the kernel loads modules). Indeed. If it was published though I could start from there (if I wanted to setup this myself). But it's not. OK. > Under BeOS it works like this: you specify the B_MTR_* flags (as > specified in KernelExport.h) as flag when creating the area. The > kernel > will load the correct CPU module for you at boot time, and mark the > area as write combined or whatever you ask for using that module. > There is no need to involve the app_server into this, only the > graphics > driver has to take care for this flag when creating the frame buffer > area. I was confused because the syslog entries start with 'app_server' (if I remember correctly) if I (kernel graphics driver) request a mapping with MTR flags. > If you need to duplicate that behaviour on existing BeOS, you would > indeed need to find out the private module interface used there, but > for Haiku, we're free to chose our own. Indeed. I will likely setup a driver for now just modifying/extending the BIOS setup done for the variable range MTRR registers. (found also an example doing this in the R3 days from Be in a newsletter now) Thanks for responding and setting me straight :-) Rudolf.