[Pkg-ofed-devel] Bug#608873: opensm: segfaults at startup

[Lucas Nussbaum]

Package: opensm
Version: 3.2.6-20090317-2
Severity: grave

Hi,

opensm segfaults at startup:
graphene-49:~# /etc/init.d/opensm-boot start
Starting opensm on 0x0002c903000b93d1: 
Starting opensm on 0x0002c903000b93d2:

in dmesg:
[  540.340322] opensm[20476]: segfault at ffffffff ip 0000000000416e24 sp 00007f9bd8e72cc0 error 6 in opensm[400000+75000] in opensm[400000+75000]

In /var/log/opensm.*:
Jan 04 10:26:27 557299 [FF4A5700] 0x03 -> OpenSM 3.2.6_20090317
Jan 04 10:26:27 557335 [FF4A5700] 0x80 -> OpenSM 3.2.6_20090317
Jan 04 10:26:27 557566 [FF4A5700] 0x02 -> osm_vendor_init: 1000 pending umads specified
Jan 04 10:26:27 557680 [FF4A5700] 0x80 -> Entering DISCOVERING state
Jan 04 10:31:28 605970 [595FC700] 0x03 -> OpenSM 3.2.6_20090317
Jan 04 10:31:28 605996 [595FC700] 0x80 -> OpenSM 3.2.6_20090317
Jan 04 10:31:28 606226 [595FC700] 0x02 -> osm_vendor_init: 1000 pending umads specified
Jan 04 10:31:28 606315 [595FC700] 0x80 -> Entering DISCOVERING state
Jan 04 10:31:32 111937 [9FBA700] 0x03 -> OpenSM 3.2.6_20090317
Jan 04 10:31:32 111962 [9FBA700] 0x80 -> OpenSM 3.2.6_20090317
Jan 04 10:31:32 112185 [9FBA700] 0x02 -> osm_vendor_init: 1000 pending umads specified
Jan 04 10:31:32 112272 [9FBA700] 0x80 -> Entering DISCOVERING state
Jan 04 10:26:27 557298 [CB6A0700] 0x03 -> OpenSM 3.2.6_20090317
Jan 04 10:26:27 557335 [CB6A0700] 0x80 -> OpenSM 3.2.6_20090317
Jan 04 10:26:27 557567 [CB6A0700] 0x02 -> osm_vendor_init: 1000 pending umads specified
Jan 04 10:26:27 557620 [CB6A0700] 0x01 -> osm_db_init: ERR 6101: Failed to create the db directory:/var/cache/opensm/
Jan 04 10:31:28 606838 [8517E700] 0x03 -> OpenSM 3.2.6_20090317
Jan 04 10:31:28 606864 [8517E700] 0x80 -> OpenSM 3.2.6_20090317
Jan 04 10:31:28 607089 [8517E700] 0x02 -> osm_vendor_init: 1000 pending umads specified
Jan 04 10:31:28 607168 [8517E700] 0x80 -> Entering DISCOVERING state
Jan 04 10:31:32 112291 [DD43F700] 0x03 -> OpenSM 3.2.6_20090317
Jan 04 10:31:32 112316 [DD43F700] 0x80 -> OpenSM 3.2.6_20090317
Jan 04 10:31:32 112557 [DD43F700] 0x02 -> osm_vendor_init: 1000 pending umads specified
Jan 04 10:31:32 112641 [DD43F700] 0x80 -> Entering DISCOVERING state

Running it manually:
graphene-49:~# opensm -v
-------------------------------------------------
OpenSM 3.2.6_20090317
Command Line Arguments:
 Verbose option -v (log flags = 0x7)
 Log File: /var/log/opensm.log
-------------------------------------------------
OpenSM 3.2.6_20090317

Entering DISCOVERING state

Using default GUID 0x2c903000b93d1
Segmentation fault

IB hardware we are using:
02:00.0 InfiniBand: Mellanox Technologies MT26418 [ConnectX VPI PCIe 2.0 5GT/s - IB DDR / 10GigE] (rev a0)

- Lucas

-- System Information:
Debian Release: squeeze/sid
  APT prefers testing
  APT policy: (500, 'testing')
Architecture: amd64 (x86_64)

Kernel: Linux 2.6.32-5-amd64 (SMP w/4 CPU cores)
Locale: LANG=C, LC_CTYPE=C (charmap=ANSI_X3.4-1968)
Shell: /bin/sh linked to /bin/dash

Versions of packages opensm depends on:
ii  infiniband-diags      1.4.4-20090314-1   InfiniBand diagnostic programs
ii  libc6                 2.11.2-7           Embedded GNU C Library: Shared lib
ii  libibcommon1          1.1.2-20090314-1   InfiniBand management library
ii  libibumad1            1.2.3-20090314-1.1 InfiniBand Userspace Management Da
ii  libopensm2            3.2.6-20090317-2   Infiniband subnet manager librarie

opensm recommends no packages.