e...@math.uni-bonn.de (Edgar =?iso-8859-1?B?RnXf?=) writes: >Even if tar (and pax, which I also tried) was to use stdio, this would not >explain being bound by /system/ cpu time.
With a smaller stdio buffer, you do more read/write syscalls and add more overhead. >It must be something in the kernel. Likely vnode/FFS, but maybe at a lower >level (RAIDframe, scsipi, mfii), that's why I asked up to what level time >spent in the kernel is attributed to the user process. The spend in a syscall is attributed to the process. But the actual I/O is often deferred to kernel threads and handled in interrupts. Interrupt time is not accounted to user processes.
