R4000 end-of-page bug

[Frank Wille <frank%phoenix.owl.de@localhost>]

Hi,

the current sgimips kernels are still unstable on a R4000 Indy. I had some
difficulties to complete sysinst, with frequent drops into 'db' while untaring
the archives. Example:

---------------------------------------------------------------------
0%|                                                                             
      | 384 KiB  382.86 KiB/s  04:10 ETA
Stopped in pid 390.1 (tar) at   0x880c5e74:     lw      v0,12(v0)
db> bt
0+880c5e74 (b,3000,1,cb4a196e) ra 0 sz 0
User-level: pid 390.1
db> show reg
...
v0        0xc0175fe0

During normal use programs will spontaneously crash with segmentation fault
when launched (once running, they are usually stable). For example I can
reproduce such a crash running "/bin/sh" 20 or 30 times. Analyzing the core
dump with gdb doesn't help much, as the PC and/or the stack is completely
out of bounds.


So I'm wondering if we have workarounds for all known R4000-bugs? What about
the "end-of-page" bug, for example? Does our gcc take care of it? And I
didn't see an -mfix-r4000 option in the Makefiles either, for the other
workarounds.


-- 
    _  Frank Wille (frank%phoenix.owl.de@localhost)
 _ //  http://sun.hasenbraten.de/~frank/
 \X/   Phx @ #AmigaGer