[ofa-general] Mellanox Gen3, Linux and ibpanic - "Resource Temporarily unavailable"

Robert Dunkley Robert at saq.co.uk
Tue Nov 25 07:21:10 PST 2008


Hi Hal,

Thanks again, I will try this in a minute. I think I have found the
moment it went bad on Machine A using Dmesg:
ib_mthca 0000:87:00.0: Catastrophic error detected: unknown error
ib_mthca 0000:87:00.0:   buf[00]: ffffffff
ib_mthca 0000:87:00.0:   buf[01]: ffffffff
ib_mthca 0000:87:00.0:   buf[02]: ffffffff
ib_mthca 0000:87:00.0:   buf[03]: ffffffff
ib_mthca 0000:87:00.0:   buf[04]: ffffffff
ib_mthca 0000:87:00.0:   buf[05]: ffffffff
ib_mthca 0000:87:00.0:   buf[06]: ffffffff
ib_mthca 0000:87:00.0:   buf[07]: ffffffff
ib_mthca 0000:87:00.0:   buf[08]: ffffffff
ib_mthca 0000:87:00.0:   buf[09]: ffffffff
ib_mthca 0000:87:00.0:   buf[0a]: ffffffff
ib_mthca 0000:87:00.0:   buf[0b]: ffffffff
ib_mthca 0000:87:00.0:   buf[0c]: ffffffff
ib_mthca 0000:87:00.0:   buf[0d]: ffffffff
ib_mthca 0000:87:00.0:   buf[0e]: ffffffff
ib_mthca 0000:87:00.0:   buf[0f]: ffffffff
ib_mthca 0000:87:00.0: HW2SW_MPT failed (-11)
ib0: ib_query_gid() failed
ib_mthca 0000:87:00.0: HW2SW_MPT failed (-11)
ib0: ib_query_port failed
ib0: Failed to modify QP to ERROR state
ib0: timing out; 1 sends 250 receives not completed
ib0: Failed to modify QP to RESET state
ib_mthca 0000:87:00.0: HW2SW_MPT failed (-11)
ib_mthca 0000:87:00.0: HW2SW_CQ failed (-11)
ib_mthca 0000:87:00.0: HW2SW_MPT failed (-11)
ib_mthca 0000:87:00.0: HW2SW_CQ failed (-11)
ib_mthca 0000:87:00.0: HW2SW_MPT failed (-11)
ib_mthca 0000:87:00.0: HW2SW_SRQ failed (-11)
ib_mthca 0000:87:00.0: HW2SW_MPT failed (-11)
ib_mthca 0000:87:00.0: HW2SW_MPT failed (-11)
ib_mthca 0000:87:00.0: HW2SW_MPT failed (-11)
ib_mthca 0000:87:00.0: HW2SW_MPT failed (-11)
ib_mthca 0000:87:00.0: HW2SW_MPT failed (-11)

Does this help to pinpoint what might have caused this?

Thanks,

Rob


-----Original Message-----
From: Hal Rosenstock [mailto:hal.rosenstock at gmail.com] 
Sent: 25 November 2008 15:19
To: Robert Dunkley
Subject: Re: [ofa-general] Mellanox Gen3, Linux and ibpanic - "Resource
Temporarily unavailable"

Hi Rob,

On Tue, Nov 25, 2008 at 10:01 AM, Robert Dunkley <Robert at saq.co.uk>
wrote:
> Hi Hal,
>
> Machine A is definitely on and I have had the cable connection
checked.
> I'm afraid I'm not much of a techy, how do I unload and reload the IB
> stack?

It depends on what you have running... Is it just OpenSM and IPoIB ?

Kill off opensm

Use modprobe -r to remove all the ib_ modules. You can find them via
lsmod | grep ib_. There is a dependency order.

If you can get them all unloaded, reload them in the reverse order and
hopefully things will be better...

-- Hal

> Thanks,
>
> Rob
>
>
> -----Original Message-----
> From: Hal Rosenstock [mailto:hal.rosenstock at gmail.com]
> Sent: 25 November 2008 15:00
> To: Robert Dunkley
> Cc: Baur, Eric; general at lists.openfabrics.org
> Subject: Re: [ofa-general] Mellanox Gen3, Linux and ibpanic -
"Resource
> Temporarily unavailable"
>
> Hi Rob,
>
> On Tue, Nov 25, 2008 at 9:54 AM, Robert Dunkley <Robert at saq.co.uk>
> wrote:
>> Hi Hal,
>>
>> Thank you for your help.
>>
>> Ibstat on MachineB:
>> CA 'mthca0'
>>        CA type: MT25204
>>        Number of ports: 1
>>        Firmware version: 1.2.0
>>        Hardware version: a0
>>        Node GUID: 0x0002c9020022d428
>>        System image GUID: 0x0002c9020022d42b
>>        Port 1:
>>                State: Down
>
> Is machine A on ? Is mthca loaded there ? If so, this should at least
> be init but the driver errors below may preclude this from occurring.
>
>>                Physical state: Polling
>>                Rate: 10
>>                Base lid: 0
>>                LMC: 0
>>                SM lid: 0
>>                Capability mask: 0x02510a6a
>>                Port GUID: 0x0002c9020022d429
>>
>> Machine A is operating normally with the exception of Infiniband
which
>> broke after powering down Machine B and did not recover once Machine
B
>> was powered on again. An extract from the log of Machine A:
>> Nov 25 14:30:21 mrhappy kernel: ib_mthca 0000:87:00.0: HW2SW_MPT
> failed
>> (-11)
>> Nov 25 14:30:31 mrhappy kernel: ib_mthca 0000:87:00.0: HW2SW_CQ
failed
>> (-11)
>> Nov 25 14:30:41 mrhappy kernel: ib_mthca 0000:87:00.0: HW2SW_MPT
> failed
>> (-11)
>> Nov 25 14:30:51 mrhappy kernel: ib_mthca 0000:87:00.0: HW2SW_CQ
failed
>> (-11)
>> Nov 25 14:31:01 mrhappy kernel: ib_mthca 0000:87:00.0: HW2SW_MPT
> failed
>> (-11)
>> Nov 25 14:31:11 mrhappy kernel: ib_mthca 0000:87:00.0: HW2SW_SRQ
> failed
>> (-11)
>> Nov 25 14:31:21 mrhappy kernel: ib_mthca 0000:87:00.0: HW2SW_MPT
> failed
>> (-11)
>> Nov 25 14:32:01 mrhappy last message repeated 3 times
>> Nov 25 14:32:11 mrhappy kernel: ib_mthca 0000:87:00.0: HW2SW_MPT
> failed
>> (-11)
>
> -11 is EAGAIN. Not sure what this is used for in the mthca driver.
>
> Can you unload and reload the IB stack especially mthca driver ?
>
> -- Hal
>
>> Thanks again,
>>
>> Rob
>>
>> -----Original Message-----
>> From: Hal Rosenstock [mailto:hal.rosenstock at gmail.com]
>> Sent: 25 November 2008 14:49
>> To: Robert Dunkley
>> Cc: Baur, Eric; general at lists.openfabrics.org
>> Subject: Re: [ofa-general] Mellanox Gen3, Linux and ibpanic -
> "Resource
>> Temporarily unavailable"
>>
>> On Tue, Nov 25, 2008 at 9:39 AM, Robert Dunkley <Robert at saq.co.uk>
>> wrote:
>>> Hi Eric,
>>>
>>> Thanks for the response. OpenSM is running and set to start on
bootup
>> on
>>> MachineB:
>>> ps aux | grep open
>>> root      5616  0.0  0.1 142004  1396 ?        Sl   13:39   0:00
>>> /usr/sbin/opensm -t 200 -f /var/log/opensm.log -g 0
>>>
>>> The log on Machine B just logs this every 10 seconds:
>>> Nov 25 14:34:21 148541 [477A7940] 0x01 ->
>>> __osm_sm_state_mgr_signal_error: ERR 3207: Invalid signal
>>> OSM_SM_SIGNAL_DISCOVER in state IB_SMINFO_STATE_DISCOVERING
>>> Nov 25 14:34:31 153173 [477A7940] 0x80 -> SM port is down
>>>
>>> Ibstat confirms port is in polling state on MachineB.
>>
>> Is the port in init or down ?
>>
>>> MachineA however is in a bad state,
>>
>> Any additional details on this ?
>>
>> Can you kill/unload all the ib stuff and reload it ? That would be
>> gentler than rebooting.
>>
>> -- Hal
>>
>>>I tried the openibd restart command, it accepted the
>>> command but after 5 minutes shows no progress of doing anything and
> is
>>> just at the cursor. Is some sort of forced restart of openibd
>> possible?
>>>
>>> Thanks,
>>>
>>> Rob
>>>
>>>
>>> -----Original Message-----
>>> From: Baur, Eric [mailto:Eric.Baur at gs.com]
>>> Sent: 25 November 2008 14:31
>>> To: Robert Dunkley
>>> Subject: RE: [ofa-general] Mellanox Gen3,Linux and ibpanic -
> "Resource
>>> Temporarily unavailable"
>>>
>>> Robert-
>>>
>>> Is OpenSM set to start on boot?
>>>                chkconfig --list | grep opensmd
>>>
>>> If not:         chkconfig opensmd on
>>> and:            /etc/init.d/opensmd start
>>>
>>> You can also restart openib without rebooting the machines.
>>>                /etc/init.d/openibd restart
>>>
>>> -Eric
>>>
>>> -----Original Message-----
>>> From: general-bounces at lists.openfabrics.org
>>> [mailto:general-bounces at lists.openfabrics.org] On Behalf Of Robert
>>> Dunkley
>>> Sent: Tuesday, November 25, 2008 9:21 AM
>>> To: general at lists.openfabrics.org
>>> Subject: [ofa-general] Mellanox Gen3,Linux and ibpanic - "Resource
>>> Temporarily unavailable"
>>>
>>> Hi everyone,
>>>
>>> I'm using a setup of two machines (Lets call them A and B) directly
>>> connected by 1 cable. Each machine has a Mellanox MT25204 (Gen3
>> Mellanox
>>> PCI-E Infiniband card) and uses IPOIB, they run Centos 5.2 with OFED
>> 1.3
>>> installed, Machine B runs OpenSM.
>>>
>>> All was working fine. I shutdown Machine A did some maintenance and
>> then
>>> powered it on again, everything is OK again. I then shutdown Machine
> B
>>> (The one running OpenSM), this seemed to really upset Machine A.
> After
>>> booting Machine B again, Machine B looks OK with the port down and
in
>>> polling state. Machine A however gives the following error if I run
>>> ibstat: ibpanic: [11406] main: stat of IB device 'mthca0' failed:
>>> (Resource temporarily unavailable)
>>>
>>> I don't want to reboot Machine A as it must synch data with Machine
B
>>> over the Infiniband link first. Does anyone have any idea how to fix
>>> machine A?
>>>
>>> Thanks,
>>>
>>> Rob
>>>
>>> The SAQ Group
>>>
>>> Registered Office: 18 Chapel Street, Petersfield, Hampshire GU32 3DZ
>>> SEMTEC Limited Trading as SAQ is Registered in England & Wales
>>> Company Number: 06481952
>>>
>>>
>>>
>>> http://www.saqnet.co.uk AS29219
>>>
>>> SAQ Group Delivers high quality, honestly priced communication and
>> I.T.
>>> services to UK Business.
>>>
>>> DSL : Domains : Email : Hosting : CoLo : Servers : Racks : Transit :
>>> Backups : Managed Networks : Remote Support.
>>>
>>> Find us in http://www.thebestof.co.uk/petersfield
>>>
>>> _______________________________________________
>>> general mailing list
>>> general at lists.openfabrics.org
>>> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
>>>
>>> To unsubscribe, please visit
>>> http://openib.org/mailman/listinfo/openib-general
>>> _______________________________________________
>>> general mailing list
>>> general at lists.openfabrics.org
>>> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
>>>
>>> To unsubscribe, please visit
>> http://openib.org/mailman/listinfo/openib-general
>>>
>>
>



More information about the general mailing list