Any progress for: amd64/91799: ggated core dumps on amd64 after client connection

[Gabriele Cecchetti]

Pawel Jakub Dawidek ha scritto:
> On Wed, Oct 11, 2006 at 06:33:35PM +0200, Gabriele Cecchetti wrote:
>> Any progress about problem amd64/91799 ?
>> (info at: http://www.freebsd.org/cgi/query-pr.cgi?pr=91799)
>>
>> A patch was proposed, but for some reason has not yet been submitted.
>> Any chance to see a working patch inside 6-STABLE ?
> 
> I'm sorry that it takes so long...
> 
> I just tried to reproduce it on:
> 
> 	beast:root:~# uname -a
> 	FreeBSD beast.wheel.pl 6.1-STABLE FreeBSD 6.1-STABLE #2: Thu May 11 02:05:14 CEST 2006     root at beast.wheel.pl:/usr/home/obj/usr/home/src/REL6/src/sys/BEAST  amd64
> 
> No luck, it works just fine.
> 
> I need more info to reproduce it. Could you start both ggated and ggatec
> with -v option and send me the output from right before the crash?
> 

Sorry for the late.

I've reproduced it on:
FreeBSD feanor2.cluster 6.2-PRERELEASE FreeBSD 6.2-PRERELEASE #1: Tue 
Oct 17 14:13:34 CEST 2006 
root at feanor2.retis:/usr/obj/usr/src/sys/SMP  amd64

on the secondary node I run:
# ggated -v
while on the primary node I run:
# ggatec create 10.0.0.2 /dev/ad4s1h

After a lot of output ggated core dump with the following last messages:

info: Connection from: 10.0.0.1.
debug: Receiving version packet.
debug: Version packet received.
debug: Receiving initial packet.
debug: Initial packet received.
debug: Connection created [10.0.0.1, /dev/ad4s1h].
debug: New connection created (token=3889378918).
debug: Sending initial packet.
info: Connection from: 10.0.0.1.
debug: Receiving version packet.
debug: Version packet received.
debug: Receiving initial packet.
debug: Initial packet received.
debug: Found existing connection (token=3889378918).
debug: Connection added [10.0.0.1, /dev/ad4s1h].
debug: Sending initial packet.
debug: Connection removed [10.0.0.1 /dev/ad4s1h].
debug: Process created [/dev/ad4s1h].
notice: disk_thread: started [/dev/ad4s1h]!
notice: send_thread: started [/dev/ad4s1h]!
notice: recv_thread: started [/dev/ad4s1h]!
debug: Received hdr packet.
debug: recv_thread: offset=26029891072 length=512
Assertion failed: ((req->r_offset % conn->c_sectorsize) == 0), function 
disk_thread, file /usr/src/sbin/ggate/ggated/ggated.c, line 713.
info: Connection from: 10.0.0.1.
debug: Receiving version packet.
debug: Version packet received.
debug: Receiving initial packet.
debug: Initial packet received.
debug: Connection created [10.0.0.1, /dev/ad4s1h].
debug: New connection created (token=3536653038).
debug: Sending initial packet.
error: accept(): Too many open files.
error: Exiting.

I think the main problem is pointed out by: Assertion failed: (...)
and I could not get /dev/ggate0

If I run the same test with the proposed patch I get:
debug: Receiving version packet.
debug: Version packet received.
debug: Receiving initial packet.
debug: Initial packet received.
debug: Connection created [10.0.0.1, /dev/ad4s1h].
debug: New connection created (token=1330107011).
debug: Sending initial packet.
info: Connection from: 10.0.0.1.
debug: Receiving version packet.
debug: Version packet received.
debug: Receiving initial packet.
debug: Initial packet received.
debug: Found existing connection (token=1330107011).
debug: Connection added [10.0.0.1, /dev/ad4s1h].
debug: Sending initial packet.
debug: Connection removed [10.0.0.1 /dev/ad4s1h].
debug: Process created [/dev/ad4s1h].
notice: disk_thread: started [/dev/ad4s1h]!
notice: send_thread: started [/dev/ad4s1h]!
notice: recv_thread: started [/dev/ad4s1h]!
error: Error while receiving hdr packet: Resource temporarily unavailable.
error: Exiting.

nothing of good right now

Thanks
Gabriele