You'll need to get used to comparing it to more "usual" output, I guess. Ioctl() is a syscall you'd very rarely expect to see many of in most code. If I'm seein

You'll need to get used to comparing it to more "usual" output, I guess. Ioctl() is a syscall you'd very rarely expect to see many of in most code. If I'm seeing more than half a dozen ioctl's, I'll be wondering what is going on (doesn't mean it has to be something wrong, just unusual). But also, you'll want to look at the wall time and the microseconds per call - the raw syscall count is much less interesting. Once you then spot something odd, dropping the "-c" and using "--trace=" to set which syscalls you're interested in tracing, and getting a live dump of the syscalls while you're testing often reveals very "interesting" patterns (favourite pet peeve: Looking for small reads/writes - context switch overhead almost always makes it pay to buffer in userspace, but a surprising number of applications fail to do so) Basically, my assumption w/networking code is that if I'm stress-testing it and the wall-time isn't dominated by read/write, something is very wrong.