RV670 AGP card with agpmode=-1, d-r-t kernel as it was 16/03/11

Since

5066f7cd0e010db46b42f8f08a33c10cb34f8726
r600g: add rv670 flushing workaround.

I am getting segfaults with openarena, nexuiz, etqw and a few mesa demos - 

gltestperf,clearspd and ipers.

The bt of the segfault seems to repeat for ever.

Starting program: /home/andy/Src/Mesa-git/demos/src/demos/ipers 
[Thread debugging using libthread_db enabled]
IperS V1.0
Written by David Bucciarelli (tech.hmw@plus.it)
[New Thread 0xb726f700 (LWP 14143)]

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0xb726f700 (LWP 14143)]
r600_context_flush (ctx=0x8074e10) at r600_hw_context.c:1282
1282    {
(gdb) bt
#0  r600_context_flush (ctx=0x8074e10) at r600_hw_context.c:1282
#1  0xb6eeed3b in r600_context_flush_all (ctx=0x8074e10, flush_flags=128) at r600_hw_context.c:786
#2  0xb6eeef18 in r600_context_bo_flush (ctx=0x8074e10, flush_flags=33554496, flush_mask=0, rbo=0x83c1488) at r600_hw_context.c:814
#3  0xb6eef030 in r600_context_flush_dest_caches (ctx=0x8074e10) at r600_hw_context.c:1197
#4  0xb6eee833 in r600_context_flush (ctx=0x8074e10) at r600_hw_context.c:1298
#5  0xb6eeed3b in r600_context_flush_all (ctx=0x8074e10, flush_flags=128) at r600_hw_context.c:786
#6  0xb6eeef18 in r600_context_bo_flush (ctx=0x8074e10, flush_flags=33554496, flush_mask=0, rbo=0x83c1488) at r600_hw_context.c:814
#7  0xb6eef030 in r600_context_flush_dest_caches (ctx=0x8074e10) at r600_hw_context.c:1197
#8  0xb6eee833 in r600_context_flush (ctx=0x8074e10) at r600_hw_context.c:1298
#9  0xb6eeed3b in r600_context_flush_all (ctx=0x8074e10, flush_flags=128) at r600_hw_context.c:786
#10 0xb6eeef18 in r600_context_bo_flush (ctx=0x8074e10, flush_flags=33554496, flush_mask=0, rbo=0x83c1488) at r600_hw_context.c:814
#11 0xb6eef030 in r600_context_flush_dest_caches (ctx=0x8074e10) at r600_hw_context.c:1197
#12 0xb6eee833 in r600_context_flush (ctx=0x8074e10) at r600_hw_context.c:1298