Azureus + 7-STABLE == Slow download + No Upload - some idea

[Marcin Cieslak]

Greg Lewis wrote:
> On Tue, Apr 01, 2008 at 09:04:46PM -0300, Marc G. Fournier wrot
>>> On FreeBSD 6.1/2/3 Azureus shows the same behaviour (at least on my PC)
>>> when _not_ using /etc/libmap.conf to map libthread to libc_r.
>>>
>>> Work-arounds like using other clients are one thing, but I would like
>>> to see a fix for this on FreeBSD 7. Is there anything that can be done
>>> to help identify / solve the problem?

> The reason that we've persisted with doing patchsets for Java is so that
> everybody can get at the source and investigate/fix things which are
> important to them.  I've persistently argued against switching solely to
> binary releases, when that has come up, for just this reason.

Greg and friends,

I have taken the opportunity to look at the source and compile jdk16 
with debug patches to investigate. Thanks for providing us possibility 
to do this for years already.

 > None of this will even require getting your hands dirty with the code, it
 > will just take some time commitment to do it and analyse the results.


There was another, simple problem report on the freebsd-java:

http://thread.gmane.org/gmane.os.freebsd.devel.java/10146

> If you wanted to go further you could start trying to mess around with
> gdb and/or trace statements in the code.

I have decided to go dirty anyway :)

I have traced this a bit and came up with a testcase code to open a UDP 
IPv4 connection via the IPv6 socket:

http://www.freebsd.org/cgi/query-pr.cgi?pr=kern/127057

This testcase is the way JDK16 does UDP connection translated to C code 
(okay, Java sets first bytes of the sockaddr * structures to zero, but 
this doesn't matter and doesn't change anything).

Either:

1) something is wrong in this code and Java should be fixed (something 
around:

j2se/src/solaris/native/java/net/PlainDatagramSocketImpl.c 
j2se/src/solaris/native/java/net/net_util_md.c

2) there is wrong in the UDP in the kernel. Even worse, EINVAL is not 
documented in the manpage, but is returned by udp6_sendto and friends.

I've had difficulty single-stepping the kernel but maybe there is 
something in here.

Can anyone replicate my results (I am running an amd64 machine, but this 
code gives me the same result on a i386 machine installed from the same 
codebase)?

--Marcin

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 273 bytes
Desc: OpenPGP digital signature
Url : http://lists.freebsd.org/pipermail/freebsd-java/attachments/20080906/f4f4fa9f/signature.pgp