read.me for tasktes1.zip Hardware multi-tasking example
Version 1.0, Mar 25, 1998
Sample code
by John S. Fine
[email protected]
I do not place any restrictions on your use of this source code
I do not provide any warranty of the correctness of this source code
____________________________________________________________________________
To run:
vload tasktest.bin
It runs 200 simultaneous tasks. Each task has a different combination
of the scheduling parameters "defer" and "quantum". Each task just
updates a counter on the screen, so you can see how fast they are
running.
Every task has a private TSS with a private CR3 value (a private page
directory) and one private page table and one private page (for its
stack). All tasks share the same code (as well as GDT, IDT etc.)
It uses only a single TSS descriptor in the GDT. It modifies the
TSS descriptor in the GDT to always point to the current task.
To rebuild (If you don't have gcc, you can skip that step):
You need JLOC version 0.6 or later.
gcc -c -O3 counter.c
gcc -c -O3 showcoun.c
NASM -f OBJ tasktest.asm
NASM -f OBJ tasks_h.asm
NASM -f OBJ idt2.asm
NASM -f OBJ set_vec.asm
NASM -f OBJ dump.asm
NASM -f OBJ 8259.asm
NASM -f OBJ mempool.asm
NASM -f OBJ meminit.asm
NASM -f OBJ pri_heap.asm
JLOC tasktest.lnk tasktest.bin tasktest.map
____________________________________________________________________________
This example includes a very flexible scheduler. It is based on the real-
time event scheduler in my 8254 examples. Some of the documentation there
might help to understand its inner working. This version is a normal
scheduler and not a real-time event scheduler.
The explanation of the scheduler parameters (as well as the code for
creating and time-slicing tasks) is in TASKS_H.ASM.
The code for maintaining the priority heap of active tasks is in
PRI_HEAP.ASM.
____________________________________________________________________________
This example include a manager for three pools of free 4K pages. The
kernel of a paged OS should manage pools of 4K pages. This is a good
example of how to do that using no significant memory overhead outside of
the free pages themselves. This code is in MEMPOOL.ASM
In this example, the initial set of free pages is identified and "freed"
by MEMINIT.ASM. That module also moves the entire kernel and all its
paging and other data to new pages allocated from the last pool. This
step might be required in an OS that needs low memory for DMA buffers
or similar purposes.
____________________________________________________________________________
Another interesting feature is the use of the "EMPTY" section. In
TASKTEST.LNK the EMPTY section is forced to be 4Kb aligned and beyond the
end of the stack. In MEMINIT.ASM, any memory allocated to that section is
returned to the free pool. In TASKS_H.ASM, memory is requested for
portions of that section as required by the kernel. This demonstrates a
common feature in mapped kernels. You can allocate a large amount of
address space for potentially large structures, and you can wait to
allocate RAM to that address space until you actually need it.
