Re: [realtime] Re: [rtl] virtual memory in RTL

[Paolo Mantegazza <mantegazza@aero.polimi.it>]

yodaiken@fsmlabs.com wrote:
> 
> On Thu, Mar 23, 2000 at 06:10:55PM +0000, Paolo Mantegazza wrote:
> > Hi,
> >
> > yodaiken@fsmlabs.com wrote:
> >
> > > So you make RT calls via  "int something" ? How do you deal with
> > > the memory map? Does the RT system switch mmu context?
> 
> > LXRT does all what Linux does in switching its processes, i.e:
> >
> >         rthal.gdt_table[next->tss.tr >> 3].b &=
> 
> etc.
> 
> So what happens if Linux is in the middle of a fork or if the user didn't
> mlock all pages?

Page locking is a must for hard real time under LXRT. We made available
a function to allow it also to nonroot users. It should not be seen as a
limitation, I think that it is a must in all hard real time commercial
UNIX like systems.

AFAIK nothing special happens. Take into account that if you ask LXRT
for hard real time it guarantee Linux task switching to be atomic.
Pierre Cloutier wanted "forked bomb" tests, both thruogh fork and
pthread_create, and there seem to be no problems.

Ciao, Paolo.