Faulty ac68?

[Chrysalis]

New router is laggy to use webui, ssh etc.

e.g. even ls has a noticeable pause.

This process is constantly hogging resources.

PID PPID USER STAT VSZ %VSZ CPU %CPU COMMAND
258 2 admin SW 0 0.0 1 30.4 [mtdblock3]

Is faulty unit?

 

[hggomes]

Press RESET button 5sec without releasing it or FACTORY DEFAULT RESTORE from the WEBUI, and try again.

 

[Chrysalis]

This is brand new setup after a reset.

 

[hggomes]

What FW and VERSION? That process is always running at 30% usage?

 

[Chrysalis]

found this in log on nand scan

Jan 1 00:00:13 kernel: Scanning device for bad blocks
Jan 1 00:00:13 kernel: Bad eraseblock 802 at 0x000006440000

 

[Chrysalis]

is john's fork latest version, same version as I run on my ac66.

the mtdblock is at 30+ % 95% of the time, if it goes away its only for one top refresh.

overall cpu usage is stuck at 40-55% sys usage.

Right now the device feels about 4x as slow as my ac66.

 

[john9527]

found this in log on nand scan

Jan 1 00:00:13 kernel: Scanning device for bad blocks
Jan 1 00:00:13 kernel: Bad eraseblock 802 at 0x000006440000

A bad block (or two) is normal...I've had two of them ever since I got my router. Only a concern if they start growing.

The mtd allocations are different between code levels....on my fork on AC68 mtd3 should be 'rootfs'

admin@AC68R-xxxxx:/tmp/home/root# cat /proc/mtd
dev: size erasesize name
mtd0: 00080000 00020000 "boot"
mtd1: 00180000 00020000 "nvram"
mtd2: 01e00000 00020000 "linux"
mtd3: 01c598f0 00020000 "rootfs"
mtd4: 06000000 00020000 "brcmnand"
mtd5: 02000000 00020000 "jffs2"

if you run 'cat /proc/mtd' does it match?

 

[hggomes]

It would help if you could answer all the questions first.

About that message you shouldnt worry too much about it, more info:

http://wiki.openmoko.org/wiki/NAND_bad_blocks

 

[Chrysalis]

yep it matches john.

so an issue somewhere with rootfs.

I used your tool to migrate from the ac66, I guess I should try to completely manually input my settings to see if the problem goes away?

 

[hggomes]

You cant migrate different models like that, yes you should configure it manually.

 

[Chrysalis]

hggomes I suggest you read john's tool thread his tool has a migrate function for changing between different models.

anyway I be back as soon as possible, I will do a full manual setup which is a massive job with my setup.

 

[john9527]

If you don't mind, before you do I'd like to collect a little data in case of a problem with the tool (but yes, a reset would be the next step).

Try and open the 'Tools' tab....does it display correctly?
Try and remove any USB devices and reboot.
Try and disable/re-enable JFFS and see if it helps.

 

[hggomes]

Just configure it manually and test it again please or simply RESET the device and test it without any configuration (just enable SSH), i bet you will not see that happening.

About the tool i dropped it at the time because of several issues in the past, not sure if its working better by now. I didnt knew it was able of doing it on different models, does it? We will see

Thanks for the info.

TBH i always prefeer doing it the hardway, my setup can be done in a couple of minutes anyway.

 

[Chrysalis]

I am back already, seems it is jffs/(flash) related.

Due to an ipv6 issue the syslog was active.

If I disable ipv6 the issue goes away.
or
if I enable ipv6 but disable syslog logging to jffs the issue also goes away.

and no worries hggomes, I think I will be doing that anyway as I have to be sure its not down to what I did.

The tool certianly works wonders when upgrading firmware, I used it many times now without issues, however this is the first time I used it for a migration.

 

[john9527]

One other thought.....did you have Entware/Optware installed and then restore the /jffs partition?

 

[Chrysalis]

entware was messy I had to fix it.

So basically I did my normal process which was this.

flash firmware, reboot
reset to defaults and reboot
enable telnet, enable jffs, enabled format jffs
reboot
login to telnet and restore backup (in this case restore migration backup)
enable #a wireless
reboot
enable usb to mount over jffs (i did this to stop syslog writing to the nand but as I discovered today it still writes to the nand doing this anyway) (this is effectively copying configs and scripts to jffs)
reboot
normally at this point everything works including existing entware install

however I discovered the entware binaries for the ac66 dont work on ac68, so basically I had to rerun entware-setup.sh and it makes a new install using different binaries, so to answer your question entware has been refreshed for the new device. Also currently I disabled the jffs remount via usb.

 

[Chrysalis]

by the way remember that weird radvd issue we discussed? I seem to have accidently resolved it.

 

[john9527]

Thanks for the info.....I need to add a warning about restoring jffs between MIPS and ARM if using Entware/Optware.

Regarding the IPv6 syslog 'flood', we've been working on your ISPs bad servers with IPv6, but that must be some error flood if it slows down the router writing to jffs. Can you send me some sample syslog output?

 

[john9527]

by the way remember that weird radvd issue we discussed? I seem to have accidently resolved it.

It's a different kernel.....but interesting. Any idea of something specific you may have done?

 

[Chrysalis]

john I reset to defaults, enabled ipv6, telnet, pppoe, jffs and left everything else alone.

Problem is still there so in my opinion it rules out your tool been the problem, looks like a router issue to me.

and ignore what I said about radvd I didnt fix anything sorry

here is my really powerful new router idle

Mem: 45956K used, 209696K free, 0K shrd, 312K buff, 9716K cached
CPU: 1.4% usr 52.6% sys 0.0% nic 32.0% idle 13.1% io 0.0% irq 0.5% sirq
Load average: 2.47 1.20 0.50 2/76 2108
PID PPID USER STAT VSZ %VSZ CPU %CPU COMMAND
258 2 admin RW 0 0.0 0 34.3 [mtdblock3]
1 0 admin S 2204 0.8 1 1.5 /sbin/preinit
997 1 admin S 2192 0.8 0 0.2 usbled