FS#34832 - [glibc] Assertion in sysdeps/posix/getaddrinfo.c forces openJDK to exit with some networking apps
Attached to Project:
Arch Linux
Opened by Old User New ID (u2012) - Wednesday, 17 April 2013, 23:43 GMT
Last edited by Allan McRae (Allan) - Friday, 25 October 2013, 21:17 GMT
Opened by Old User New ID (u2012) - Wednesday, 17 April 2013, 23:43 GMT
Last edited by Allan McRae (Allan) - Friday, 25 October 2013, 21:17 GMT
|
Details
Description:
Recently (starting from April 13th), the JVM running I2P started to exit unexpectedly at seemingly random times due to this assertion: java: ../sysdeps/posix/getaddrinfo.c:1738: rfc3484_sort: Assertion `src->results[i].native == -1 || src->results[i].native == a2_native' failed. ---- Note that I run a relatively unusual setup were short-lived processes create their own namespace with a dedicated hostname and network stack, connected with the master namespace through a veth master/slave interface. ---- I initially suspected that this patch was to blame: https://projects.archlinux.org/svntogit/packages.git/plain/trunk/glibc-2.17-getaddrinfo-stack-overflow.patch?h=packages/glibc&id=27d80958180562e033e57483ac2a58ac49dab8e5 But I found another Archlinux user with the same issue but with another java app here: https://groups.google.com/forum/?fromgroups=#!topic/omnetpp/exI342W4_P0 As this was reported on April 6th, I'm starting to think the GCC 4.8 rebuild is to blame. Additional info: * package version(s) 2.17-5 (with reason to believe it existed in 2.17-4) * config and/or log files etc. java: ../sysdeps/posix/getaddrinfo.c:1738: rfc3484_sort: Assertion `src->results[i].native == -1 || src->results[i].native == a2_native' failed. JVM received a signal UNKNOWN (6). JVM process is gone. JVM exited unexpectedly. |
This task depends upon
Closed by Allan McRae (Allan)
Friday, 25 October 2013, 21:17 GMT
Reason for closing: Fixed
Additional comments about closing: glibc-2.18-9 in [testing]
Friday, 25 October 2013, 21:17 GMT
Reason for closing: Fixed
Additional comments about closing: glibc-2.18-9 in [testing]
Debian has a patch for this:
http://patch-tracker.debian.org/patch/series/view/eglibc/2.11.3-4/any/submitted-getaddrinfo-lo.diff
Upstream report:
http://sourceware.org/bugzilla/show_bug.cgi?id=9954
Maybe also related, though no idea if it is needed:
http://patch-tracker.debian.org/patch/series/view/eglibc/2.13-38/any/local-getaddrinfo-interface.diff
Yes. But I added that a long time ago.
127.0.0.1 localhost.localdomain localhost
127.0.0.2 localhost2.localdomain localhost2
::1 localhost.localdomain localhost local6host
::2 local6host2.localdomain local6host2
So, I built glibc with the patch and "epoch=1" so it wouldn't be upgraded.
I'll report back in a couple of weeks.
svn co svn+ssh://mikro_sk@svn.code.sf.net/p/mxplay/code/trunk mxplay
svn: ../sysdeps/posix/getaddrinfo.c:1732: rfc3484_sort: Assertion `src->results[i].native == -1 || src->results[i].native == a1_native' failed.
Aborted (core dumped)
Content of /etc/hosts is:
#
# /etc/hosts: static lookup table for host names
#
#<ip-address> <hostname.domain.org> <hostname>
127.0.0.1 localhost.localdomain localhost
::1 localhost.localdomain localhost
# End of file
It must be something new (April 2013 onwards) because it used to work perfectly, I haven't changed my system configuration at all.
Didn't see the assertion since June 30. I lost my earlier logs.
So, the patch seems to fix the issue although that's hard to proof.
"Removing "myhostname" from /etc/nsswitch.conf is a workaround that works for me."
FS#37191is caused by aur/linux-pf. I'll do further tests later.