HFSC issues in RELENG_6

[Bill Marquette]

On 2/26/06, Andrew Thompson <thompsa at freebsd.org> wrote:
> On Sat, Feb 25, 2006 at 10:12:55PM -0600, Bill Marquette wrote:
> > On 2/25/06, Andrew Thompson <thompsa at freebsd.org> wrote:
> > > You may want to test with another network card to verify that sis(4) is
> > > actually working correctly with ALTQ. I am having ALTQ problems which
> > > went away when using a xl(4) card. I am waiting for a sis card to arrive
> > > in the mail so I can diagnose it further.
> >
> > You may be onto something here.  I swapped my compact flash over to a
> > Nexcom I have laying around with fxp nic's in it and so far haven't
> > seen the same queue behavior with either the small test case or a full
> > ruleset.
>
> Thats interesting indeed. As I said I am waiting for a pci card to
> arrive so I can debug further (rather than on my soekris). I would
> really like to have this fixed for 6.1 but time is short so no
> guarantees.

OK, more info.  I just created an OpenBSD 3.9 snapshot image for my
Soekris 4801 so I could do an apples to apples comparison.  Same pf
config (queue name changed accidentally, but that's a nop), same
hardware - hfsc works like a champ.

queue root_sis1 bandwidth 100Mb priority 0 hfsc( upperlimit 768Kb ) {qWANRoot}
  [ pkts:          0  bytes:          0  dropped pkts:      0 bytes:      0 ]
  [ qlength:   0/ 50 ]
  [ measured:     0.0 packets/s, 0 b/s ]
queue  qWANRoot bandwidth 100Mb priority 7 hfsc( default realtime
128Kb upperlimit 128Kb )
  [ pkts:       1100  bytes:     230995  dropped pkts:      0 bytes:      0 ]
  [ qlength:   0/ 50 ]
  [ measured:    39.8 packets/s, 67.13Kb/s ]
queue root_sis0 bandwidth 100Mb priority 0 hfsc( upperlimit 6Mb ) {qLANRoot}
  [ pkts:          0  bytes:          0  dropped pkts:      0 bytes:      0 ]
  [ qlength:   0/ 50 ]
  [ measured:     0.0 packets/s, 0 b/s ]
queue  qLANRoot bandwidth 100Mb priority 7 hfsc( default realtime
128Kb upperlimit 128Kb )
  [ pkts:       1285  bytes:     246653  dropped pkts:      0 bytes:      0 ]
  [ qlength:   0/ 50 ]
  [ measured:    45.6 packets/s, 72.27Kb/s ]


Notice two differences here. 100Mbit instead of 10Mbit - this looks
like driver? and no backlogged queue - yay!  I'm going to start poking
more at the driver and see if I can come up with anything, I still
find it strange that fxp works (I'm going to poke at that too and see
if  anything pops out at me).

--Bill