To be fair, I haven't ruled out the possibility that we did something stupid, 
and left a security door wide open somewhere.  But I'm having a hard time 
figuring out what that would have been.

I also went through and checked a handfulof 8000s in the field, and haven't 
found any others that look like they either have an FTP server running, or that 
look like they have served as a dumping ground for these sorts of worms.  We 
program each CPE identically when they get installed, and the configuration is 
stamped out of a template (we configured one CPE how we wanted it, then backed 
up the config and restore that config file to a new CPE during installation).

This is a weird one, to be sure.

-- Nathan

-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of 
Robert
Sent: Thursday, February 09, 2017 3:48 AM
To: Adam Moffett; [email protected]
Subject: Re: [Telrad] UE upgrade failure rate

Sometimes, you feel like reaching out and finding a particular software 
engineer and slapping them upside the back of the head... Over and over 
and over and over....  WTF were they thinking...

On 2/9/17 2:56 AM, Adam Moffett wrote:
> WOW.
>
> ------ Original Message ------
> From: "Nathan Anderson" <[email protected] <mailto:[email protected]>>
> To: "Telrad List" <[email protected] <mailto:[email protected]>>
> Sent: 2/9/2017 5:54:27 AM
> Subject: Re: [Telrad] UE upgrade failure rate
>
>> After much trial and error, I managed to learn that the white port is
>> a TTL-level serial interface.  And there was much rejoicing.
>>
>>
>>
>> ALSO, I FIGURED OUT WHAT HAS BEEN KILLING (at least our) CPE8000s.
>>
>>
>>
>> Remember that problem that the EPC firmware had back when it was first
>> released?  Back when root access was still available on the EPC
>> firmware, there was an FTP server running on it that accepted
>> connections via the PDN IP address, and if you didn't change the root
>> password from the default insecure one (which was ironically named),
>> then infected machines trying to spread that stupid Photo.scr worm
>> would successfully log into the EPC via FTP and, thinking that it had
>> managed to log into a public web server somewhere, upload a bajillion
>> copies of the virus to it in various directories, filling up the disk.
>>
>>
>>
>> The exact same thing is happening here, believe it or not.  It hadn't
>> ever occurred to me to test for this, but it turns out that under
>> certain circumstances that I haven't yet managed to nail down, the
>> CPE8000 firmware actually starts running an FTP server.  Even worse,
>> this FTP server, once enabled, does not ask for any credentials.  You
>> can literally type in any username when prompted, and you are in.  I
>> see no config option on the web interface for the CPE that allows you
>> to turn this on and off...but whatever is triggering it ends up
>> creating a ready and completely unsecured backdoor to the CPE.
>>
>>
>>
>> *headdesk*
>>
>>
>>
>> If you guys give out routable IPs to your LTE users, or if you have
>> somebody on your network that has a PC infected with this particular
>> virus, then it might be that this could also explain your CPE8000
>> firmware upgrade problems.
>>
>>
>>
>> After figuring out the serial port bit and examining the "dead" CPEs
>> more in-depth, I found the filesystems littered with files named
>> things like Photo.scr, IMG001.scr, Info.zip, etc.  Once the writable
>> partition with the CPE configuration is completely full, if at that
>> point you issue either a reset-to-defaults, or upload a configuration
>> backup, or initiate a firmware upgrade (which has to migrate your
>> configuration from the old firmware version during the process), your
>> CPE gets bricked because there isn't enough disk space left for it to
>> properly finish writing the config changes to disk.  So it gets only
>> half-done, and the configuration is left in an inconsistent state.
>>
>>
>>
>> I've managed to fix my dead units, and also found the mechanism for
>> disabling the FTP server.  Still not sure how it is getting toggled on
>> in the first place (perhaps there is some other vulnerability that is
>> getting exploited first?), but I'll keep looking.
>>
>>
>>
>> I'll write up some instructions for y'all and post them here soon.
>>
>>
>>
>> -- Nathan
>>
>>
>>
>> *From:*[email protected] <mailto:[email protected]>
>> [mailto:[email protected] <mailto:[email protected]>]
>> *On Behalf Of *Nathan Anderson
>> *Sent:* Wednesday, February 08, 2017 1:49 PM
>> *To:* Telrad List; Adam Moffett
>> *Subject:* Re: [Telrad] UE upgrade failure rate
>>
>>
>>
>> Does anybody happen to know if the 6-pin white connector on the 8000's
>> board is either a serial port or a JTAG interface?
>>
>>
>>
>> -- Nathan
>>
>>
>>
>> *From:*[email protected] <mailto:[email protected]>
>> [mailto:[email protected]] *On Behalf Of *Nathan Anderson
>> *Sent:* Wednesday, February 08, 2017 1:40 PM
>> *To:* Telrad List; Adam Moffett
>> *Subject:* Re: [Telrad] UE upgrade failure rate
>>
>>
>>
>> This thread is interesting because I was just complaining last night
>> to our vendor about how fragile the software on the CPE8000s seems to be.
>>
>>
>>
>> We have not had specific issues with flashing CPEs "over the air" from
>> the web interface, but sometimes ACS-initiated updates don't complete
>> correctly.  On 7000s it usually takes the form of the upgrade not
>> completing and the UE falling off of the ACS, but the radio stays up
>> and attached to the network.  We go in via the web interface OTA and
>> reboot it and it comes back with the same version of firmware it was
>> already running.  Second time is usually the charm, and I'm thinking
>> that perhaps if the UE had been freshly-rebooted before attempting the
>> update, we might have a higher success rate.  (We have also seen 7000s
>> just stop talking to the ACS without us touching the firmware, and
>> even though they are otherwise working fine.  Again, rebooting the CPE
>> fixes this.  Although it is rare, we have seen this even on the latest
>> .116)
>>
>>
>>
>> We once had a 7000 that did drop off the network after pushing the
>> update via ACS.  We never checked what state it was in from the
>> ethernet side, but we had the customer powercycle it themselves and it
>> came back...again running the same firmware.  So the upgrade did not
>> take, but it didn't brick it either and resetting config to defaults
>> on the UE was not (and at least for us never has been) necessary.
>>
>>
>>
>> So we have never had to truck-roll to a 7000 as a result of a failed
>> firmware upgrade.  The 8000s, however, seem to be another story.  I am
>> so scared to touch the ones we have in the field anymore.  We have had
>> a couple that seem to get their configs corrupted after a firmware
>> change, and get into very funky states.
>>
>>
>>
>> One of them had these symptoms: defaulted to a 192.168.0.1 IP on the
>> ethernet (!), no web server running, no DHCP server running, had
>> telnet access that didn't prompt for a password (!!).  Fixed it by
>> resetting to defaults (found a shell script that performs this
>> function on the CPE's filesystem).  I got lucky with this one.
>>
>>
>>
>> One that I have sitting on my desk now is one that we tried to
>> rollback the firmware on (customer was experiencing random network
>> detaches, and the latest 8000 firmware doesn't reattach for 15 minutes
>> on-the-dot, so customer was -- I think justifiably -- getting a bit
>> pissy).  Current symptoms are: NO IPv4 on the ethernet, IPv6
>> link-local responds, no web server running, no DHCP server running,
>> telnet responds (calls itself "KZTECH") but default root/root123
>> doesn't work, so I have NO way to get in and reset the damn thing, and
>> the 8000s don't seem to have a reset button.  Thus it seems that it is
>> possible for a scrambled config to completely brick an 8000.
>>
>>
>>
>> If anybody has reliable information on how to get the 8000 to wipe its
>> config during bootup even though it seemingly lacks a reset button, I
>> would be eternally grateful...
>>
>>
>>
>> -- Nathan
>>
>>
>>
>> *From:*[email protected] <mailto:[email protected]>
>> [mailto:[email protected]] *On Behalf Of *Matthew Carpenter
>> *Sent:* Wednesday, February 08, 2017 6:38 AM
>> *To:* Adam Moffett; Telrad List
>> *Subject:* Re: [Telrad] UE upgrade failure rate
>>
>>
>>
>> Hi,
>>
>>
>>
>> So far only 1 CPE8000 UE that did not come back after a firmware
>> update.  Normally a hard reboot would fix it, but in this case we had
>> to replace it.
>>
>> I have that CPE8000 on my desk and need to see what the status is from
>> the LAN side.  Thanks for the info on defaulting it, will try it.
>>
>>
>>
>> Matt C.
>>
>>
>>
>>
>>
>>
>>
>> On Wed, Feb 8, 2017 at 8:23 AM, Adam Moffett <[email protected]
>> <mailto:[email protected]>> wrote:
>>
>> We've had a helluva time upgrading UE firmware over the air.  It was
>> worse with Wimax.  On Wimax it was more like 75% of the time we would
>> lose the channel scan table and have to go on site to add it back in.
>> It became SOP to leave the operator password at default so we had the
>> option of having the customer log in and fix the scan table for us.
>>
>>
>>
>> I think we've had more success since going to LTE.  However, failed
>> firmware updates was one of the incentives to set up a dedicated
>> management bearer.  I was hoping it would help with these things.  We
>> haven't pushed out an update recently enough to say whether it helped.
>>
>>
>>
>> -Adam
>>
>>
>>
>>
>>
>>
>>
>> ------ Original Message ------
>>
>> From: "Shayne Lebrun" <[email protected]
>> <mailto:[email protected]>>
>>
>> To: [email protected] <mailto:[email protected]>
>>
>> Sent: 2/8/2017 9:14:36 AM
>>
>> Subject: [Telrad] UE upgrade failure rate
>>
>>
>>
>>     Does anybody else experience a ten to fifteen percent failure rate
>>     when upgrading UEs?  The behavior is, you upgrade the firmware,
>>     reboot, and the device doesn't come back.  Logging into the UE's
>>     management from LAN, you'll see it's stuck in 'device init.'
>>     Defaulting the unit and rebooting allows it to boot and attach.
>>
>>
>>
>>     We're not using the residential gateway device or anything, and
>>     the only config we put in is device name, SNMP and ACS settings.
>>     Sometimes we hardcode the client's device in the DHCP server, to
>>     turn on DMZ to allow port forwarding, but that doesn't seem to be
>>     a causal factor.
>>
>>
>> _______________________________________________
>> Telrad mailing list
>> [email protected] <mailto:[email protected]>
>> http://lists.wispa.org/mailman/listinfo/telrad
>>
>>
>>
>>
>>
>> --
>>
>> *Matthew Carpenter*
>>
>> *806-316-5071 office*
>>
>> *806-236-9558 cell*
>>
>>
>>
>
>
> _______________________________________________
> Telrad mailing list
> [email protected]
> http://lists.wispa.org/mailman/listinfo/telrad
>
_______________________________________________
Telrad mailing list
[email protected]
http://lists.wispa.org/mailman/listinfo/telrad

_______________________________________________
Telrad mailing list
[email protected]
http://lists.wispa.org/mailman/listinfo/telrad

Reply via email to