needbuf

[Patrick Welche <prlw1%cam.ac.uk@localhost>]

On a (today's) -current/amd64 box, when I try to (disk)scrub a
raid6a, the system grinds to a halt with processes (including
ioflush) stuck in needbuf, e.g.,

PID    LID S CPU     FLAGS       STRUCT LWP *               NAME WAIT
1027     1 3   2         0   fffffe81ca842680                 sh needbuf
863      1 3   2         0   fffffe81cdc34640                 sh wait
980      1 3   2        80   fffffe821397f960               cron pipe_rd
160      1 3   1        80   fffffe81cdc34a60               sshd netio
290      1 3   2         0   fffffe82129fba00               sshd needbuf
137    650 5   1         0   fffffe81ca842aa0              scrub
137      1 3   1   1000000   fffffe821c579740              scrub needbuf
225      1 3   2         0   fffffe8211443200                 sh needbuf
1142     1 3   0        80   fffffe82137d11e0                 sh wait
...
0       84 3   0       200   fffffe821bfc8b80            physiod physiod
0       83 3   2       200   fffffe821c842700           aiodoned aiodoned
0       82 3   2       200   fffffe821cacc2c0            ioflush needbuf
0       81 3   0       200   fffffe821c842b20           pgdaemon pgdaemon
0       78 3   2       200   fffffe821cacc6e0            raidio8 needbuf
0       77 3   1       200   fffffe821caccb00              raid8 rfnodeq
0       76 3   3       200   fffffe821d03e2a0            raidio7 needbuf
0       75 3   2       200   fffffe821d03e6c0              raid7 rfnodeq
0       74 3   2       200   fffffe821d03eae0            raidio6 needbuf
0       73 3   3       200   fffffe821d2e4280              raid6 rfnodeq
0       72 3   0       200   fffffe821d2e46a0            raidio5 needbuf
0       71 3   0       200   fffffe821d2e4ac0              raid5 rfnodeq
0       70 3   0       200   fffffe810e6c6260            raidio9 needbuf

Presumably this shouldn't happen? (This is repeatable)

Cheers,

Patrick