VKernel progress update - 7 Jan 2006

[Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx>]

    I've made phenominal progress.  The virtual kernel can now actually
    run a /bin/csh on the console given a suitable replacement to init
    (see code at end).

    It can't run init yet, or much else without crashing (not the real 
    system, just the virtual kernel process), but the fact that it
    gives me a shell prompt at all is pretty significant :-).  Just for 
    yucks I did the old 'dd' test:

    [ on real kernel ]
    test28# dd if=/dev/zero of=/dev/null bs=32k count=102400
    102400+0 records in
    102400+0 records out
    3355443200 bytes transferred in 0.732878 secs (4578447352 bytes/sec)
    test28# 

    [ on virtual kernel ]
    ./kernel.debug -m 64m -r /var/vkernel/rootimg.01
    ...
    %dd if=/dev/zero of=/dev/null bs=32k count=102400
    102400+0 records in
    102400+0 records out
    3355443200 bytes transferred in 1.989408 secs (1686654142 bytes/sec)

    Don't read much into the times, it's doing some very, very nasty stuff
    to keep the page tables in order, more nasty stuff to implement its
    clock 'interrupt', console, context handling, etc.  It's not preloading
    any pages and the real kernel has to take a fault to mark a page as
    modified in the virtual kernel's page table.  And so on and so forth.

    I'm not going to try optimizing anything until I actually have the
    thing booting normally with a real init and running stable.

						-Matt

/*
 * /mnt/sbin/init replacement (vkernel root image) to get a shell prompt.
 *
 * cc blah.c -o /mnt/sbin/init -static
 */
#include <sys/types.h>
#include <stdio.h>
#include <stdlib.h>
#include <fcntl.h>
#include <unistd.h>

static int x = 1;

int
main(int ac, char **av)
{
        int fd;
        char buf[32];

        fd = open("/dev/console", O_RDWR);
        dup2(fd, 1);
        dup2(fd, 2);
        write(1, "TEST\n", 5);
        execl("/bin/csh", "csh", "-i", NULL);
        return(0);
}