The following is a demonstration of the lockbyproc.d script,
dtrace: description 'lockstat:::adaptive-block ' matched 1 probe
^C
pageout 49438
mysql-test-run 96414
oracle 149086
sched 220601
The above output shows that threads belonging to sched, the kernel, spent
a total of 220 microseconds waiting for an adaptive mutex lock.
This example sampled for a longer interval,
dtrace: description 'lockstat:::adaptive-block ' matched 1 probe
^C
init 136228
java 371896
oracle 783402
sched 2315779
mysqltest 9428277
mysql-test-run 10093658
mysqld 17412999
fsflush 19676738
Here we can see threads belonging to fsflush have spent a total of 19.7 ms
waiting for an adaptive mutex. Note: it's not easy to say that it means a
19.7 ms delay in the completion of the fsflush program, as this value is
the sum of the block times across all the threads. So it is possible that
many threads were blocked at the same time, eg, it could have been 19 threads
blocked during the same 1 ms.