Netgate SG-1000 microFirewall

Author Topic: Intel Atom C2xxx LPC failures  (Read 20198 times)

0 Members and 1 Guest are viewing this topic.

Offline virgiliomi

  • Sr. Member
  • ****
  • Posts: 563
  • Karma: +74/-4
    • View Profile
Intel Atom C2xxx LPC failures
« on: February 05, 2017, 07:29:31 am »
Get ready folks... this is going to be a fun ride soon. :) Cisco has started having routers and switches fail due to an LPC clock failure. Coincidentally, Intel has updated the errata of their Atom C2000-series chips, indicating that an LPC clock failure can prevent the system from booting. Cisco didn't name-and-shame the company producing the failed part in their gear, but it's pretty coincidental that Intel happens to update their errata at the same time Cisco announces issues with their hardware indicating the same failure.

Cisco claims that the failure can start after as little as 18 months of use.

Intel claims to have a platform-level workaround that can be used. Of course, there are no details in the errata about the workaround.

This could make for some fun times soon, given all of the Rangeley chips being used in systems running pfSense.

Article on The Register, the update at the bottom indicates the Atom may be at fault in Cisco's gear.

Offline w0w

  • Sr. Member
  • ****
  • Posts: 560
  • Karma: +35/-6
  • kernel panic attack
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #1 on: February 05, 2017, 09:27:21 am »
http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/atom-c2000-family-spec-update.pdf
Not fun at all and not only for pfSense users. But what "A platform level change has been identified and may be implemented as a workaround for this erratum." means?

Offline gcu_greyarea

  • Jr. Member
  • **
  • Posts: 77
  • Karma: +9/-2
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #2 on: February 05, 2017, 04:05:57 pm »
Oh Dear.
I've got a ASA5506, SG-2220, 2 x SG-2440 and a SG-4860.

Can't wait to hear about that "workaround".

Would love to learn how pfSense/ADI are planning to handle this issue?

Would prefer a software/firmware patch, and hope there won't be a need for Hardware replacement.

I think cisco are doing HW replacements.

Offline VAMike

  • Sr. Member
  • ****
  • Posts: 407
  • Karma: +65/-11
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #3 on: February 05, 2017, 04:14:46 pm »
if it turns out to be true it's gonna really chap a lot of people who were convinced they really needed to buy "server grade" components. :D

Offline gcu_greyarea

  • Jr. Member
  • **
  • Posts: 77
  • Karma: +9/-2
    • View Profile

Offline gcu_greyarea

  • Jr. Member
  • **
  • Posts: 77
  • Karma: +9/-2
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #5 on: February 05, 2017, 11:06:28 pm »

Offline athurdent

  • Hero Member
  • *****
  • Posts: 654
  • Karma: +36/-7
  • N00b.
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #6 on: February 05, 2017, 11:45:27 pm »
Has anyone contacted Supermicro about this?

Offline gcu_greyarea

  • Jr. Member
  • **
  • Posts: 77
  • Karma: +9/-2
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #7 on: February 06, 2017, 03:59:44 am »
All the information available online leads me to believe that the issue is not limited to Cisco. I.e. Intel is very clear about AVR54 and the affected processors. It's safe to assume that the ADI/NetGate/pfSense boxes are affected, too (as well as some Synology NAS')

Unless there's some magic "firmware bullet" I'm convinced that most vendors will just ride this issue out - or at best manage the issue on a Case by Case basis. If Cisco could've pushed out a firmware fix they would've done it in a heartbeat.

I'd be more than happy to receive replacements boards from Netgate/pfSense (for my 2240, 2440 and 4860's) but my hopes are indeed very slim.
The pfSense store shows One year manufacturer's warranty and my boxes are already 1+ year old. That absolves pfSense of the responsibility to replace my SG Series Appliances - which by the way have not yet even failed.

Do I sleep comfortable that my network sits on a time bomb ? No.
Do I sleep comfortable that my employer's network sits on a time bomb ? No.

Perhaps ADI/NetGate/PfSense do not have the same level of clout with intel as Cisco does, but I'd surely hope ADI/NetGate/PfSense will work out some sort of arrangement with intel to reduce the impact on existing customers.

I would suggest that affected Rangely/Avoton customers should receive a heavy discount when buying from the pfSense Store again.

Offline Wolf666

  • Full Member
  • ***
  • Posts: 277
  • Karma: +31/-1
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #8 on: February 06, 2017, 01:07:47 pm »
This issue has nothing to do with warranty since its a design flaw...replacement or upgrade program is needed.
Personally I have my core network based on C2000, pfSense and Synology units.
There is an open thread on Synology. I also wrote to Supermicro in order to get a solution before failure occurs.
« Last Edit: February 06, 2017, 01:23:10 pm by Wolf666 »
Modem Draytek Vigor 130
pfSense 2.4 Supermicro A1SRi-2558 - 8GB ECC RAM - Intel S3500 SSD 80GB - M350 Case
Switch Cisco SG350-10
AP Netgear R7000 (Stock FW)
HTPC Intel NUC5i3RYH
NAS Synology DS1515+
NAS Synology DS213+

Offline gcu_greyarea

  • Jr. Member
  • **
  • Posts: 77
  • Karma: +9/-2
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #9 on: February 06, 2017, 07:13:35 pm »
It's eerily quiet here ....  :-\

Offline awontroba

  • Newbie
  • *
  • Posts: 3
  • Karma: +0/-0
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #10 on: February 06, 2017, 08:05:28 pm »
iXsystems FreeNAS Mini at risk too.

Offline virgiliomi

  • Sr. Member
  • ****
  • Posts: 563
  • Karma: +74/-4
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #11 on: February 06, 2017, 08:52:53 pm »
I also wrote to Supermicro in order to get a solution before failure occurs.
Please let us know what they say... there are plenty here that have SuperMicro boards with the Atom C2xxx processors on them.

Offline athurdent

  • Hero Member
  • *****
  • Posts: 654
  • Karma: +36/-7
  • N00b.
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #12 on: February 07, 2017, 12:02:30 am »
I also contaced Supermicro yesterday, seems at least the European support did not know about the issue yet. Sent them an explanation and some links. Now they are checking with the PM of the motherboard.
« Last Edit: February 07, 2017, 12:14:54 am by athurdent »

Offline athurdent

  • Hero Member
  • *****
  • Posts: 654
  • Karma: +36/-7
  • N00b.
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #13 on: February 07, 2017, 03:14:39 am »
Hmm, the Intel doc says stepping B0 is affected. My Supermicro board says:

Code: [Select]
CPU: Intel(R) Atom(TM) CPU  C2758  @ 2.40GHz (2400.07-MHz K8-class CPU)
  Origin="GenuineIntel"  Id=0x406d8  Family=0x6  Model=0x4d  Stepping=8

From what I have gathered about steppings, the version normally consist of a letter followed by a number. So what could "8" mean?
I'd love to think that Cisco got the whole B0 stepping, but then again my (and all the googled dmesg) results are missing the letter...

Any experts on this?

Offline Creep89

  • Newbie
  • *
  • Posts: 5
  • Karma: +0/-0
    • View Profile
Re: Intel Atom C2xxx LPC failures
« Reply #14 on: February 07, 2017, 03:37:30 am »
Code: [Select]
CPU: Intel(R) Atom(TM) CPU  C2558  @ 2.40GHz (2400.06-MHz K8-class CPU)
  Origin="GenuineIntel"  Id=0x406d8  Family=0x6  Model=0x4d  Stepping=8

Well, yeah, that is not very helpful at all.