On 03/11/2012 02:10 PM, Martin Sucha wrote: > Hello, > > On Sunday 11 March 2012 13:38:44 Jakub Jermar wrote: >> MS: >> - amd64 SMP problem fixed >> - ZONE_RESERVED must be allowed for physmem_map() >> - MD: reserved zones can be dangerous to touch, perhaps a firmware >> bug? > - Are there any checks I can run to determine if it is a firmware bug?
Don't know. A bug in this case means a bug with respect to our understanding of the reserved attribute. Allegedly, the reserved attribute here constitutes a warning that touching such memory can have unpredictable consequences (e.g. hang the system). > - We may add a flag to physmem_map to only allow ZONE_RESERVED if requested > and > only drivers that are known to require this will use the flag (those may also > check some device ids to determine if such a quirk is required). Is this a > good idea? I'd rather avoid any exceptions and quirks. I think the current way of dealing with the IO APIC in a reserved zone can stay as it is and we will simply wait to see if/when something breaks in the future. Another clean solution would be to always include the IO APIC range in a dedicated parea and revert physmem_map() to only succeed on firmware zones. Jakub _______________________________________________ HelenOS-devel mailing list [email protected] http://lists.modry.cz/cgi-bin/listinfo/helenos-devel
