Doing egrep NR_TASKS thru the kernel source shows me that
there are special dependencies on NR_TASKS against a few
architecture specific things:
- SPARC has something special with regard of sun4c
memory management, and perhaps more generic iommu
management too
Right, and for this reason it really should be defined
per-architecture.
- i386 APM BIOS has some odd dependency
Didn't notice this one before.
And few otherwise difficult details:
- kernel/sched.c: struct task_struct * task[NR_TASKS];
I tried a while back to eliminate all references to task[] array, only
remaining piece of code which cares is TSS allocation on Intel. For
this reason, when I said I wanted to zap task[] entirely, Linus said
"Yes, but not right now, some time later."
Later,
David "Sparc" Miller
davem@caip.rutgers.edu