cvs commit: src/sys/i386/i386 trap.c src/sys/kern kern_poll.c

Matthew Dillon dillon at
Wed May 25 01:11:38 PDT 2005

:On Tue, May 24, 2005 at 02:18:27PM -0700, Matthew Dillon wrote:
:>   Get rid of the bad hack that was doing network polling from the trap code.
:Why? This allowed to kernel to get an extra round for free, now we have
:to wait for the timer to trigger.

    Putting a hook in the trap code does not result in any sort of
    deterministic operation.  I don't know who put that hack in there
    (inherited from FreeBSDland) and I don't want to know.  That whole mess
    was put in because hardclock (the normal polling point) is a totally
    inadequate place to trigger the poll.

    A nice clean solution would simply be to give the polling subsystem
    its own systimer.   It just so happens that DragonFly has the 
    timer facilities necessary to do it right.

    How would you like to do that?  It couldn't be more then an hour's
    worth of programming, and the user could then choose the polling
    interval on the fly rather then screw around with the system's heartbeat.

					Matthew Dillon 
					<dillon at xxxxxxxxxxxxx>

More information about the Commits mailing list