Re: [Freemint-discuss] XaAES often crashes during boot after a reboot

[Jo E. S. <jo...@on...>]

Good tips! I changed the program to set the free memory to 0x55, and if I run this program before mint.prg I can reproduce the crash almost every time on both the Milan and Falcon/CT60. Different network hardware, different drivers, same snapshot. Crashes somewhere between route (in mint.cnf) and starting the desktop.

Also, it does not happen if I don't run route. ifconfig alone does not cause the crash, ifconfig+route does.

I've attached the test program (no I haven't, the mail was rejected because of the .zip attachment. Get it here: https://atari.joska.no/download/memset.zip). It would be very interesting if you could reproduce this on your Falcon.

--------------->8------------------

#include <stdio.h>
#include <string.h>
#include <tos.h>

#define VALUE 0x55

int main(int argc, char **argv)
{
   long s, total = 0;

    printf("Setting all available memory to 0x%02x:\n", VALUE);

    printf("\nST-RAM:\n");
    while ((s = (long) Mxalloc(-1L, 0)) > 0)
    {
        char *m = Mxalloc(s, 0);
        printf("0x%08p (%ld Kb)\n", m, s / 1024L);
        memset(m, VALUE, s);
        total += s;
    }

    printf("\nFast-RAM:\n");
    while ((s = (long) Mxalloc(-1L, 1)) > 0)
    {
        char *m = Mxalloc(s, 1);
        printf("0x%08p (%ld Kb)\n", m, s / 1024L);
        memset(m, VALUE, s);
        total += s;
    }

    printf("\n%ld Kb set.\n\n", total / 1024L);
    return 0;
}

--------------8<-----------------

It must be run before mint.prg, if I run it from mint.cnf before ifconfig/route it doesn't make a difference. Same with clearing the RAM instead - do it before mint.prg and the crash never happens. So it looks like the kernel assumes that atleast one byte of the RAM it allocates is zeroed.

I run ifconfig/route as the last thing before GEM= in mint.cnf. GlueSTiK is started immediately before. It doesn't matter if GlueSTiK is started or not.

Jo Even

On Tue, 21 Jan 2025 14:04:05 +0100, "Miro Kropáček" <mir...@gm...> wrote:

>> - since your CT60 scenario is far less reproducible: what about filling RAM
>> with 0xFF instead of 0x00? Does it lead to more reliable crashes? (just to
>> be on the safe side, check on Milan before, too - the 0xFF may be one of
>> those "safe" ones)
>>
>> - do not run both ifconfig and route, just one of them and alternatively,
>> do not configure the given network interface, just the loopback one