When running a SCP transfer, sometimes it zips along at 90KB/s. Other
times it crawls 7 KB/s. When this happens, a lok at top shows
ksoftirq using over 90% of the CPU. I did a google on it, and I found
other "non-Foxboard" cases, but I did not see, or did not understand
what was happening. is ksoftirqd emulating the interrupt for ethernet
and so spinning in a tight loop so it does not miss any data? and if
so, why does it affect performance sometimes and not others? -Chuck-
Mem: 11384K used, 2432K free, 0K shrd, 920K buff, 5200K cached
Load average: 2.18, 1.50, 0.68 (State: S=sleeping R=running,
W=waiting)
PID USER STATUS RSS PPID %CPU %MEM COMMAND
2 root RWN 0 1 91.7 0.0 ksoftirqd/0
762 root S 496 708 5.0 3.5 dropbear
766 root R 384 723 2.1 2.7 top
764 root R 312 763 0.3 2.2 scp
647 root S 232 1 0.3 1.6 respawnd
746 root S 504 708 0.0 3.6 dropbear
747 root S 392 746 0.0 2.8 sh
708 root S 384 1 0.0 2.7 dropbear
723 root S 376 1 0.0 2.7 sh
713 root S 344 1 0.0 2.4 boa
763 root S 312 762 0.0 2.2 sh
650 root S 280 1 0.0 2.0 syslogd
653 root S 256 1 0.0 1.8 klogd
743 root S 256 1 0.0 1.8 udhcpc
1 root S 232 0 0.0 1.6 init
696 root S 224 1 0.0 1.6 vftpd
691 root S 152 1 0.0 1.1 utelnetd
4 root SW< 0 1 0.0 0.0 khelper
578 root SW 0 1 0.0 0.0 mtdblockd
