Re: new if_sk locking patch...

Bjoern A. Zeeb wrote this message on Tue, Aug 17, 2004 at 19:26 +0000:
> On Tue, 17 Aug 2004, John-Mark Gurney wrote:
> 
> > > Does that also fix the "hangs" a lot of people can see ?
> > > I guess not ?
> >
> > Probably not...  it only fixed the LOR's on attach...  I just
> > got my card, but if you could get me a back trace on the hang,
> > I can look at it more...  Though the hangs might be due to other
> > problems...
> 
> it's not a problem of kernel hang but sk<n> is up/active but no data
> passes the interface and established connections drop. In most cases
> connectivity comes back after some amount of time.  See kern/69879.

I took a quick look, and did some tests last night, and I couldn't
get the box to hang.  I will admit I don't have the fastest hardware,
I was using my Apple TiBook doing testing w/ ssh and nfs.

With ssh (I believe blowfish), I was able to tar up and transfer ~31gigs
of data, getting around 6.5meg/sec.  With nfs, I did a bunch of transfers
of multigigs of data, and then wrote a program (to lessen the cpu usage
on the laptop) to just read the file, and got about 13meg/sec for a 1gig
VOB.  Never did I see the sk interface hang.

I was running with debug.mpsafenet=1 on a Uniproc box though.  I'm going
to commit my locking patch shortly, and could you test again?

Thanks.

-- 
  John-Mark Gurney				Voice: +1 415 225 5579

     "All that I will do, has been done, All that I have, has not."