Scott Cheloha wrote: > I might have a winner already, though. Earlier today I saw iridium busy > sleeping: repeat nanosleeps with zero'd timespec structs. I have a ktrace > here, but I have yet to reproduce. Need to fuss with it a bit more. > > fwiw, here's an exerpt: > > 90240/241378 iridium 1545768212.972668 CALL nanosleep(0x84928d42a78,0) > 90240/123819 iridium 1545768212.972670 CALL nanosleep(0x84928d42a78,0) > 90240/241378 iridium 1545768212.972703 STRU struct timespec { 0 } > 90240/123819 iridium 1545768212.972706 STRU struct timespec { 0 } > 90240/169304 iridium 1545768212.983015 CALL nanosleep(0x84928d42a78,0) > 90240/169304 iridium 1545768212.983095 STRU struct timespec { 0 } > 90240/525129 iridium 1545768212.983100 CALL nanosleep(0x84928d42a78,0) > 90240/525129 iridium 1545768212.984049 STRU struct timespec { 0 } > 90240/241378 iridium 1545768212.993270 CALL nanosleep(0x84928d42a78,0) > 90240/241378 iridium 1545768212.993351 STRU struct timespec { 0 } > 90240/123819 iridium 1545768212.997352 CALL nanosleep(0x84928d42a78,0) > 90240/123819 iridium 1545768212.997431 STRU struct timespec { 0 } > 90240/558012 iridium 1545768212.998511 CALL nanosleep(0x84928d42a78,0) > 90240/558012 iridium 1545768212.998516 STRU struct timespec { 0 } > > ... in my 13 second trace there are ~3000 zero'd nanosleep calls across > several threads. It went like that, eating up 1.5 CPUs worth of time until > I killed it... ugh, chrome...
and you want to change things to use *more* CPU??? :)