re: panic on dual processor UltraSPARC-IIIi

[matthew green <mrg%eterna.com.au@localhost>]

i've seen the fpusave panic, and i've seen what appears to be semi-random
data corruption.  my build.sh runs tend to fail either with a panic from
savefpu (though i have a patch to retry, and that always works so far) or
usually with a userland failure.  often ./configure can't create executables
that run...

   trap type 0x34: cpu 1, pc=100ae6c npc=100ae70 pstate=0x44820006<PRIV,IE>
   panic: mem address not aligned
   cpu1: Begin traceback...
   cpu1: End traceback...
   Frame pointer is at 0x12782a21
   Call traceback:
   12bf5ec(18a5108, 12740880, ff0f0000000001, 185a400, 75, 0, 12782af1) fp = 
12782af1
   1410d90(104, 0, ffff, 1773da1, 140ff80, 0, 12782bb1) fp = 12782bb1
   146269c(1764b60, 127836e0, 1278356e, 181c000, 104, 184e400, 12782c81) fp = 
12782c81
   1009070(127836e0, 34, 100ae6c, 44820006, 11fb5a0, 800, 12782e31) fp = 
12782e31
   1354850(1276cc00, 40942000, 464748494a4b5155, 508, a1, badcafe, 12783011) fp 
= 12783011
   1354f84(124ea440, 40942000, 1c82390, 18cfd30, 0, 1, 127830d1) fp = 127830d1
   14e4f4c(1276cc00, 40942000, 40a22000, 1770400, 13033650, 18a1c00, 12783191) 
fp = 12783191
   14e52ac(12771e40, 2000, 12fbde10, 12783b18, 0, 0, 12783261) fp = 12783261
   12639cc(12771e40, 13936e70, ffffffffffffffff, 1, 0, 12783b18, 12783331) fp = 
12783331
   12640bc(12740880, 0, 0, ffffffffffff8a88, 4022a600, badcafe, 12783451) fp = 
12783451
   1439f0c(12740880, 12783dc0, 12783e00, 1, 2228e0, badcafe, 12783511) fp = 
12783511
   10092b4(12783ed0, 12783f48, 40940f08, 0, 40940f08, 800, 12783621) fp = 
12783621
   116130(0, 40c12a5c, 0, badcafe, badcafe, badcafe, ffffffffffff8421) fp = 
ffffffffffff8421

you can use gdb to figure out what addresses these are.  of particular
interest is 0x100ae6c .. 
   
   (during boot, fsck already warned me the fs was not clean, but still)

this one is likely a corrupted disk or similar, ..
   
   panic: kernel diagnostic assertion "data & TLB_V" failed: file 
"/disk1/home/kiers/nbcurrent/src/sys/arch/sparc64/sparc64/pmap.c", line 3265
   Stopped in pid 19415.1 (sh) at  netbsd:cpu_Debugger+0x4:        nop
   __kernassert(16f8908, 1743748, cc1, 1743898, 43, badcafe) at 
netbsd:__kernassert+0x2c
   pmap_remove_pv(12922280, 0, 1c845c8, 18cfd30, 0, 1) at 
netbsd:pmap_remove_pv+0x24c

i haven't seen this one before.  it may be another symptom of all the
other issues seen...  i'm not sure what is wrong yet.

i am testing nakayama-san's IPI change for IIIi cpus.


.mrg.