Re: cvs commit: src/sys/netinet tcp_input.c

[Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx>]

:Can't you just explicitly release and reacquire giant in the sysctl? It
:should be MP safe already.
:
:Joerg

    It looks like that ought to work.  The sysctl already iterates the
    cpus by migrating to each one in turn.  But it won't solve the larger
    problem which is (I believe) one cpu holding the BGL for a long period
    time and causing the other cpu's callout timers to stall.  For the
    kernel to be robust we have to be able to hold the BGL for a long period
    of time without the system blowing up.

					-Matt
					Matthew Dillon 
					<dillon@backplane.com>