I replaced ALL rte instructions with "jbra check_rte", and in check_rte I make sure the stack pointer is correct (0x2ffa) if the saved SR has the supervisor bit clear (which means it is returning back to user space). I also test the return address against the value 0x0000002f. If either of those tests fails, it halts right there. Guess what? Nothing failed the test, but the address exception still happens. :'(

Okay, so I found the problem. It is in this line in set_state() in sched.c:
               if (time_before(p->p_deadline, current->p_deadline)) {
Somehow the value 1 is passed in for p, and the p_deadline member is at offset 0x2e in the proc structure, so this is attempting to access a long value at p->p_deadline at address 0x2f.

So the address exception doesn't contain the return address but rather the address that some code attempted to access non-word aligned.

Now I finally bother reading M68000PRM.pdf. Here's the stack frame for address error exceptions (in pretty ASCII art ):
15 14 13 12 11 10  9  8  7  6  5  4  3  2  1  0
+-----------------------------------------------+
|                                |RW|IN|  func  | <-- lower address (%sp)
+-----------------------------------------------+
|             access address (high)             |
+                                               +
|             access address (low)              |
+-----------------------------------------------+
|             instruction register              |
+-----------------------------------------------+
|               status register                 |
+-----------------------------------------------+
|             program counter (high)            |
+                                               +
|             program counter (low)             |
+-----------------------------------------------+
RW (Read/Write): Write = 0, Read = 1
IN (Instruction/Not): Instruction = 0, Not = 1

Therefore this address exception and the stack frame make more sense.

I still have to figure out what is calling set_state() with the value 1.

EDIT:   I just found it. I called set_run (which calls set_state) without any arguments, so whatever value happened to be on the stack was used as the value for p. FAIL  

EDIT 2: Lo and behold, that fixed the problem!! I guess I should use function prototypes more thoroughly, since TIGCC would give an error if I had a prototype for sched_run().

EDIT 3: Actually, calling sched_run() from that place (in tsleep() in process.c) was a bug itself. The process was already scheduled to run at that time so it was trying to schedule a process that was already running. I just had to remove the line completely. Now my scheduler and (due to my recent clock/timer fixes) timers are extremely stable! 

Good
So my wild guess about a read at offset 0x2E from address 1 actually proved to be correct...

The bug you've reported on the TIEmu SF tracker looks odd. TIEmu has no business reading data one byte off their valid locations...

Yes, TIEmu looks for 0xCC 0xCC 0xCC 0xCC. Not when importing a TIB (src/core/images.c), but in src/core/ti_hw/flash.c::find_ssp_and_pc. And if the OS's size happens to end with 0xCC (which you describe), bad things occur...

=> for the time being, just add two bytes of waste to your OS, and you should be fine. I'll fix TIEmu

This is the first time I see a quadruple post here I think, although it's fine due to the large period of time between all. That goes to tell how we need a few more 68K users.

I'm glad this is still progressing well.

Heh, it's actually five. I didn't realize I posted so many in a row when I did it. You're right, we do need more 68k folks around here.

Anyway, I just added a busy indicator to the bottom status line (next to the battery status). I think it's pretty cool watching it flicker as programs run and pause:


What does everyone think about the status icon itself? I don't particularly care for it, but it was the best I could come up with. Can anyone draw a better one? (Should I ask this in the Pixel Art and Drawing board instead?)

Ok, I changed it to a tiny hourglass. I already tried to draw an hourglass before I posted, but I didn't like the way it turned out. Speaking of the status icons, I don't really like how the hand icon looks either. I think it looks like a claw. I don't really have a use for the "claw" button anyway. Maybe I could use it to switch to different virtual terminals using claw+F1 through F8.

I just released Beta 3, now with vfork()! This means multi-tasking works now.

The current shell can't run any pipelines or background processes, so multi-tasking is only marginally useful at this stage, but it's an important piece of the puzzle for later on.

Download the TIB file now, or read the release notes.