Subject: Re: gem stalls
To: None <port-sparc64@netbsd.org>
From: Miles Nordin <carton@Ivy.NET>
List: port-sparc64
Date: 07/25/2005 13:26:03
--pgp-sign-Multipart_Mon_Jul_25_13:26:03_2005-1
Content-Type: text/plain; charset=US-ASCII

>>>>> "mh" == Martin Husemann <martin@duskware.de> writes:

    mh> provide more info.

what would you like to know?  

The box is still running NetBSD.  I can build kernels.  I can
reproduce the problem in an hour or two.  My colo has a terminal
server hooked up to the LOM port so I can execute commands at a prompt
while it's frozen, I can use ddb, but IIRC crash dumps don't work on
sparc64.

    mh> dig into it yourself

Certainly I'd like to buy some books and try reading code carefully
and experiment and become more skilled, but first I need my Internet
connection and that of my housemates to stop dropping out.  I hope you
won't take it personally, but I can't continue working under these
conditions, on NetBSD nor on anything else, and since I'm not the only
one affected it's gotten to the point where it's not up to me to offer
the box for experiment.

I've been planning, testing, staging this whole setup with two NetBSD
boxes running ALTQ on either side of a T1 since February.  The box
that's failing now has been running for two months under test, but now
as the traffic increases this nasty subtle problem creeps up.  I need
to have fewer of these problems, and when they arise at this late
stage I need to fix them quickly.

I'm also asking myself why on earth would I spec NetBSD on a
configuration that exists basically _just_ to do ALTQ, when NetBSD
stands out as the only BSD that stubbornly hasn't integrated the new
ALTQ in spite of repeatedly-contributed patches.  You want me to dig
into this problem myself, then what?  Dig into porting ALTQ to the 3.0
release?  So that I can find it's still broken just like it was in the
2.0 release?  And then fix it, so that I can join at least three other
people since 2003 in having my patches rejected by Darren and friends
for political reasons?  It doesn't make sense for me.

Rather I'd like to warn other people using sparc64 away from this
subtle and disastrous problem that arises when using it as a router,
or when using PF, or something---also a reasonable purpose for
posting, since it's been costly and unexpected for me.

--pgp-sign-Multipart_Mon_Jul_25_13:26:03_2005-1
Content-Type: application/pgp-signature
Content-Transfer-Encoding: 7bit

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.6 (NetBSD)

iQCVAwUAQuUgq4nCBbTaW/4dAQI1HgP+L7mLOT5vgBbT8hMFqVagjnDcgbpSEebY
Dt0wxAV6gMEEIGWMl9FEs44tZOQEBWbWKUAYmwgDNtwRk7uLexZK6sSpfqHqwNY1
jNbeNb4H6KF9guLe36FeEUKNIHhWRUdbDSBLLrkalLzWXwnwuuC6s/eiPoow1gZu
wN4lAV1m3eE=
=n3lt
-----END PGP SIGNATURE-----

--pgp-sign-Multipart_Mon_Jul_25_13:26:03_2005-1--