if_re locking patch...

John-Mark Gurney gurney_j at resnet.uoregon.edu
Sat Sep 4 16:57:01 PDT 2004


Daniel Eriksson wrote this message on Sat, Sep 04, 2004 at 12:33 +0200:
> John-Mark Gurney wrote:
> 
> > To expand upon the patch posted previously, I have cribbed bms's rl
> > locking, and ported it to re.  This makes the interrupt MPSAFE along
> > with the rest of the driver.  No more GIANT LOCKED messages for re. :)
> 
> I saw that this was committed to HEAD yesterday. I have a machine with a
> 8169-based NIC that is doing ~300GB on a daily basis that I'd love to try to
> run with debug.mpsafenet=1. Do you think the patch that went into HEAD has
> been tested enough that it is safe to try it?
> 
> It's a UP machine using POLLING mode (HZ=2000), if that makes any
> difference.

Well, as demonstrated by ru's commit (thanks ru!).. I had not tested
DEVICE_POLLING...  But this afternoon I did some testing, and did not
have any troubles...   The only problems w/ netperf was that the *RR*
tests were a bit slow, but this is to be expected due to DEVICE_POLLING..

But as I mentioned, the re driver was originally based on the rl, so I
doubt there'll be much trouble since rl's locking has been in the tree
for quite a while longer...

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

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


More information about the freebsd-current mailing list