Miro Kropáček

> But definitely, FreeMiNT should hook vectors as less of possible, thus preferring solutions in the above order. Actually, this is up for debate - one could argue that FreeMiNT...

Hmmm... looking into asix.c has revealed some interesting information: - the assert comes from kernel's `kfree` function and `asixpoll` does indeed calls kfree() in `asix_poll_thread` (which is in turn awakened...

Unfortunately, I can't confirm your findings. In the meantime I have found quite good test case: replacing this kmalloc: https://github.com/freemint/freemint/blob/master/sys/sockets/xif/asix.c#L617-L621 with `static unsigned char recv_buf[AX_RX_URB_SIZE];`, this triggers the assert every...

I'll try to play with it tonight. Maybe you can try my "repro", i.e. the static buffer + `-m68030` to see what happens (although with that buffer it crashes for...

Scratch that, I really am an idiot. So another round of testing...

Ok, so one more time. This is my current set of changes: https://github.com/mikrosk/freemint/commit/6275226012e650482f25c945137d4dfebbdf43bc. However all my reports with the static buffer are worthless because only an hour ago I realised...

> there is one thing you could investigate: Whether the BSS of .xif modules is properly dimensioned and zeroed when they are loaded. I think this should be covered here:...

Even after two evenings spent on this I'm no closer to at least some clue. I've set up quite good debugging environment (serial port output on another computer) but due...

No worries Christian, that was just my personal grievance. :) However, just a few minutes after posting it, I got really lucky. I had inet4.xdd + usb.km + eth.udd in...

Indeed. I also took a look at this afterwards (I was exhausted that day) and yes, the only obvious thing is that if "something" happens before ifconfig/route command, then it's...