Re: [Freeipmi-devel] Inverted IPMI responses

[Albert Chu]

Hey Marc,

I think I see the problem(s).  I've added some tweaks, can you give the
branch "ipmb_network_reorder_race3" from github a shot.

For the ipmi-raw case, you should see a trace message just like last
time that says "reversed obj_cmd responses".  And for the new case,
where there is an error in the first response, you should see the trace
"accept out-of-order with bad completion code".

The ipmi debug packets will still be incorrect though.

Thanks,

Al

On Thu, 2019-01-31 at 15:30 +, GIRARD, MARC wrote:
> Hi Albert,
> 
> I am afraid your fix doesn't solve all cases.
> Maybe it is necessary to reconsider the existence of a race
> condition. I thing the problem is on the BMC side.
> 
> I found two 'unusual' cases : one with ipmi-raw and another one with
> IntelNM and an invalid domain_id.
> See attachments for details
> 
> 
> I apologize for my English: I wish I could be more nuanced and more
> eloquent. I hope you understood me.
> Kind regards / Cordialement
> 
> Marc Girard
> Power Efficiency team
> Atos
-- 
Albert Chu
ch...@llnl.gov
Computer Scientist
High Performance Systems Division
Lawrence Livermore National Laboratory



___
Freeipmi-devel mailing list
Freeipmi-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/freeipmi-devel

Re: [Freeipmi-devel] Inverted IPMI responses

[GIRARD, MARC]

Hi Albert

Session sequence numbers looks correct.

Kind regards / Cordialement

Marc Girard
Power Efficiency team
marc.gir...@atos.net
https://eu.yourcircuit.com/#/email/marc.gir...@atos.net

-Original Message-
From: achudr...@gmail.com  On Behalf Of Albert Chu
Sent: Wednesday, January 16, 2019 4:59 PM
To: GIRARD, MARC 
Cc: freeipmi-devel@gnu.org
Subject: Re: [Freeipmi-devel] Inverted IPMI responses

Hi Marc,

It's certainly possible it's a implementation bug.  It's also possibly a 
network routing race where the packets race.  Could you show the full 
headers/trailers of the packets, I would like to see the session sequence 
numbers.

Al

On Wed, Jan 16, 2019 at 12:58 AM GIRARD, MARC  wrote:
>
> Hi FreeIPMI Team,
>
> We experiment a unusual problem with Intel PCSD S2600BPB host.
> With Intel Node Manager OEM command, sometime response messages come in 
> reverse order : OEM Intelnm Get Node Manager Statistics Response first then 
> Send Message Response.
> See side by side trace analysis in .pdf joined
>
> With my knowledge of the IPMI protocol, this is not allowed and I suspect a 
> BMC implementation bug.
> Can you give me your opinion?
>
> Kind regards / Cordialement
> Marc Girard
> Power Efficiency team - Atos
> marc.gir...@atos.net
> https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Feu.
> yourcircuit.com%2F%23%2Femail%2Fmarc.girard%40atos.netdata=02%7C0
> 1%7Cmarc.girard%40atos.net%7Cbf0d012bdd5841e5cb9108d67bcb9087%7C33440f
> c6b7c7412cbb730e70b0198d5a%7C0%7C0%7C636832511564980186sdata=aEZc
> ItBQyxHnFSssND0gPa2ERqiEU7Upf%2Bdoiqnlras%3Dreserved=0
> ___
> Freeipmi-devel mailing list
> Freeipmi-devel@gnu.org
> https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Flis
> ts.gnu.org%2Fmailman%2Flistinfo%2Ffreeipmi-develdata=02%7C01%7Cma
> rc.girard%40atos.net%7Cbf0d012bdd5841e5cb9108d67bcb9087%7C33440fc6b7c7
> 412cbb730e70b0198d5a%7C0%7C0%7C636832511564980186sdata=DmcRkkGnJV
> Yx5HRYyxNavwyUbtV6rvEVrXJ3WH23%2Bz4%3Dreserved=0


stderr_ko
Description: stderr_ko
___
Freeipmi-devel mailing list
Freeipmi-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/freeipmi-devel