Re: Freeze (kind of)

Peter Holm wrote:

>With a GENERIC current from 16 Sep 09:25 UTC I have
>seen quite a lot of freezes where I'm able to unfreeze by
>pinging from another machine. Here's a ps from a frozen box:
>
>http://www.holm.cc/stress/log/cons75.html
>
>I'm working on getting a dump. 
>



>1472 c1c89e00 d1340000 1001  1311   531 0000002 [RUNQ] sh
> 1471 c206d1c0 d180e000 1001  1464   531 0000002 [LOCK  Giant c17f4180] sh
> 1470 c206b1c0 d160b000 1001  1311   531 0004002 [LOCK vm object c1bc2440] sh
> 1469 c206d380 d180f000 1001  1308   531 0004002 [LOCK CMAPCADDR12 c1ae9600] random
> 1468 c1aae700 d125d000 1001  1457   531 0004002 [LOCK  Giant c17f4180] rw
> 1466 c186e700 d1119000 1001  1290   531 0004002 [LOCK  Giant c17f4180] dir
> 1464 c206b000 d160a000 1001  1311   531 0004002 [RUNQ] sh
> 1462 c1c898c0 d131f000 1001  1308   531 0004002 [SLPQ nanslp 0xc08be0ac][SLP] memory
> 1461 c206b380 d17e1000 1001  1455   531 0004002 [LOCK  Giant c17f4180] rw
> 1460 c206d540 d1810000 1001  1425   531 0004002 [SLPQ nanslp 0xc08be0ac][SLP] sleep
> 1459 c1a8a700 d11e3000 1001  1423   531 0000002 [SLPQ sbwait 0xc1ba7754][SLP] net
> 1457 c1c89a80 d1320000 1001  1311   531 0004002 [SLPQ wait 0xc1c89a80][SLP] sh
> 1455 c1aae380 d125b000 1001  1311   531 0004002 [SLPQ wait 0xc1aae380][SLP] sh
> 1454 c1f31380 d17df000 1001  1450   531 0004002 [LOCK  Giant c17f4180] rw
> 1453 c1a68c40 d1192000 1001  1413   531 0000002 [SLPQ sbwait 0xc18164cc][SLP] net
> 1452 c1a84000 d1198000 1001  1446   531 0004002 [RUNQ] rw
> 1450 c1a688c0 d1190000 1001  1311   531 0004002 [SLPQ wait 0xc1a688c0][SLP] sh
> 1449 c1aae8c0 d125e000 1001  1444   531 0004002 [LOCK  Giant c17f4180] rw
> 1446 c1f31e00 d17ea000 1001  1311   531 0004002 [SLPQ wait 0xc1f31e00][SLP] sh
> 1444 c1c8d700 d1345000 1001  1311   531 0004002 [SLPQ wait 0xc1c8d700][SLP] sh
> 1441 c1c8d380 d1343000 1001  1419   531 0004002 [SLPQ nanslp 0xc08be0ac][SLP] sleep
> 1439 c1c891c0 d12c1000 1001  1425   531 0004002 [RUNQ] crashme
> 1438 c1aaa8c0 d1255000 1001  1419   531 0004002 [RUNQ] crashme
> 1435 c186e1c0 d1116000 1001  1290   531 0004002 [LOCK  Giant c17f4180] dir
> 1433 c1aaa540 d1217000 1001  1308   531 0004002 [LOCK  Giant c17f4180] memory
> 1429 c1aaa000 d1214000 1001  1290   531 0004002 [SLPQ getblk 0xc667d5d8][SLP] dir
> 1425 c1f318c0 d17e7000 1001  1287   531 0004002 [SLPQ wait 0xc1f318c0][SLP] sh
> 1423 c186ee00 d1159000 1001  1309   531 0004002 [RUNQ] net
> 1421 c1c8d540 d1344000 1001  1290   531 0004002 [RUNQ] dir
> 1419 c1aae000 d1259000 1001  1287   531 0004002 [SLPQ wait 0xc1aae000][SLP] sh
> 1416 c1c8de00 d1349000 1001  1290   531 0004002 [RUNQ] dir
> 1413 c206ba80 d17f7000 1001  1309   531 0004002 [RUNQ] net
> 1411 c1aae1c0 d125a000 1001  1290   531 0004002 [RUNQ] dir
> 1407 c1aaac40 d1257000 1001  1290   531 0004002 [RUNQ] dir
> 1392 c1a68e00 d1193000 1001  1308   531 0004002 [RUNQ] memory
> 1389 c1c89700 d131e000 1001  1308   531 0004002 [RUNQ] memory
> 1386 c206b8c0 d17f6000 1001  1308   531 0004002 [RUNQ] memory
> 1383 c1f31540 d17e5000 1001  1308   531 0004002 [RUNQ] memory
> 1380 c186e8c0 d111a000 1001  1308   531 0004002 [RUNQ] memory
> 1377 c1c89540 d131d000 1001  1308   531 0004002 [RUNQ] memory
> 1374 c206b540 d17ec000 1001  1308   531 0004002 [RUNQ] memory
> 1371 c1a68700 d118f000 1001  1308   531 0004002 [CPU 0] memory
>
"show locks"

might be interesting..
looks a bit like 1371 might be spinning while holding Giant?
also

add:

options         KTR
options         KTR_ENTRIES=1024
options         
KTR_COMPILE=(KTR_INTR|KTR_PROC|KTR_RUNQ|KTR_SYSC|KTR_CLK|KTR_TRA
P)
options         KTR_MASK=(KTR_INTR|KTR_PROC|KTR_RUNQ|KTR_SYSC|K

to you rkernel and then you can do a
   show ktr
as well, which is probably going to be very instructive.

>1368 c1f31a80 d17e8000 1001  1308   531 0004002 [RUNQ] memory
> 1365 c1a8ac40 d11e6000 1001  1308   531 0004002 [RUNQ] memory
> 1345 c1aaee00 d1261000 1001  1310   531 0004002 [SLPQ nanslp 0xc08be0ac][SLP] sleep
> 1340 c1c89000 d12a6000 1001  1314   531 0004002 [SLPQ nanslp 0xc08be0ac][SLP] sleep
>