arlad crashes/loops (memory corruption?)

[Nickolai Zeldovich]

> We have stopped supporting fbsd 3.x, xfs will probably break some day soon
> for you. Any special reson you hasn't upgraded to 4.x ?

I'm infact hoping to get the machine running 4.2-RELEASE before the
new year. If it looks like a compiler/kernel problem, maybe it will
go away with the new OS/userland version.

(There's a large amount of glue/scripts holding the system together
that I need to preserve when upgrading...)

> It might be interesting that some stuff contains invalid values that look
> like pointers (like call->conn->peer.{burstWait.usec,host}).

Looks like the host value is a pointer to a rx_peer struct -- at
least the values make sense:

(gdb) print {struct rx_peer}call->conn->peer.host
$21 = {next = 0x8119180, host = 808842701, port = 22555, packetSize = 1444, 
  idleWhen = 0, refCount = 3, burstSize = 0 '\000', burst = 0 '\000', 
  burstWait = {sec = 0, usec = 0}, congestionQueue = {prev = 0x811941c, 
    next = 0x811941c}, rtt = 0, rtt_dev = 0, timeout = {sec = 2, usec = 0}, 
  nSent = 34, reSends = 0, inPacketSkew = 0, outPacketSkew = 0, rateFlag = 2, 
  maxWindow = 15, spare = 0}

It looks suspiciously like what call->conn->peer.next, 4 bytes away,
should perhaps be, so maybe you're right about it being a subtle
compiler bug of some sort.

I'm not sure what burstWait.usec might be; could be an rx_queue(?),
since the pairs of pointer values following it look like empty queues.

(gdb) x/8x call->conn->peer.burstWait.usec
0x81e3d00 <_kafs_debug+1307068>:        0x00000000      0x00000000      0x081e3d08      0x081e3d08
0x81e3d10 <_kafs_debug+1307084>:        0x081e3d10      0x081e3d10      0x08119300      0x08119328

> +>zepa.net		# Zepa
> ?

>zepa.net                  #Kolya/zepa.net cell
205.245.53.21                   #perseus.zepa.net.
205.245.53.48                   #neptune.zepa.net.

(It should also have AFSDB records.)

Thanks,

-- kolya