OpenWrt Forum Archive

Topic: DG834G V3 loop rebooting, help needed

The content of this topic has been archived on 28 Apr 2018. There are no obvious gaps in this topic, but there may still be some posts missing at the end.

Hi,

I managed to brick my DG834G V3. I was doing a routine upgrade, and flashed an image through the web panel. Something went wrong but seemingly something like that can happen when the image is close to the 4 MB mark. Thought nothing of it as I entered recovery mode, and went on with:

# ./nftp -u eth0 ../Dropbox/DG834/DG834GV3_V4.01.40_DGTeam_1022_eng_adsldrv703.img
Sending frame on eth0 (a:b:c:d:e:f)
Found a router at x:y:z:t:u:w
Router is a DG834V3
Current version is 4128
If everything above looks good, press Enter now.

Erasing flash, this will take around ten seconds...
Upgrading:       1024/4063232 bytes

However, erasing took a lot more than 10 seconds, and Upgrading never went further than 1024. After about half an hour of this I gave up and pulled the plug.

Now whenever power is applied, the Power light goes solid green, but Test keeps blinking. Then the LAN light comes up, but shortly after that the router resets itself: Power and Test turn off, Internet and 1-2-3-4 light up briefly and then it's back to square 1.

I've tried the 30/30/30 reset thing but whenever the reset button is pressed, the router just keeps rebooting furiously. I've also attempted setting up a static route but so far with no luck at all. Is there anything I can do to get it unstuck, or JTAG is my only option?

BTW the router was working well before the bad flash, still I wonder if it's possible that it failed twice because of some previously undetected hardware problem?

Try 4.01.30 or lower from recovery mode.

Thanks for the reply.

It's no longer possible to enter recovery mode as it can only be activated by holding the reset button pressed while plugging it in, and this just produce continuous rebooting.

I need to stop the rebooting but all my attempts failed: direct connection, using another DG834, ping, ftp, tftp, telnet... nothing seems to work. The activity LED shows it's receiving, but I get no reply.

Suppose JTAG is the only way out, unless someone has an ace up his sleeve and can save my sorry ass... smile

PS: I've been googling like a madman looking for solutions, after posting I dug out some info I'm going to share in case someone with the same problem finds this thread. Apparently most people (and possibly everybody) who have tried that specific image (DGTeam  1122) have also bricked their routers.

So avoid it at all costs, and for those who have already flashed it, try recovery mode and load either an official or the last good DGTeam image (0849 for V1/V2 or 1018 for the others). The latter can be found on this site: http://jake-tm.co.uk/?page_id=259

Don't do the same mistake I did...

Hey,

Yeh your going to need to JTAG in adam2 again, the images are prefixed with the adam2 bin when uploading in recovery mode. Not entirely sure but I think it overwrites it... however when I tested this with another bootloader inserted I still got some adam2 sign on so not too sure, but it is definitely hosed. The only way I upgrade my dg834gv3 is via the recovery mode using nftp like that, in the ar7 image makefile uncomment the sercomm stuff and point it at the adam2 bin for that router, and you have a working recovery mode image.
Hope this helps.

Also while we are on the topic of DG834G's for anyone which has/had this problem if your router enters 'ASMode' in a normal boot. Like this,

ADAM2 Revision 0.18.01
(C) Copyright 1996-2003 Texas Instruments Inc. All Rights Reserved.
(C) Copyright 2003 Telogy Networks, Inc.
Usage: setmfreq [-d] [-s sys_freq, in MHz] [cpu_freq, in MHz]
Memory optimization Complete!
ASMode.
Copying download from b0017000 to b4020000

Your ENV partition need to be re JTAGd in as its got a little funky, luckily I had a donor router I could pull it from and flash in smile

(Last edited by Wipster on 2 Feb 2011, 20:05)

Wipster wrote:

the images are prefixed with the adam2 bin when uploading in recovery mode. Not entirely sure but I think it overwrites it...

I took a peek at the borked imaged (1122) with Okteta and there are strings that look like they belong to Adam2 at the beginning. I started comparing manually it to the firmware I was using before (849) and I couldn't spot any difference so I fired up KDiff3: they're byte for byte the same up to roughly the first 640 kiB or 15%. Then thrown the official 4.01.40 in the mix, and that's equal, too.

So there's no reason I can see why nftp would fail flashing 1122 and not 849 or the official image... Still, it stalled at 1024B. If my router's faulty it should have started acting up well before the flash, I guess. That leaves me looking at nftp in a different light...

@ fyi, any particular reason to prefer 4.01.30 over 4.01.40 ?

Frex wrote:

So there's no reason I can see why nftp would fail flashing 1122 and not 849 or the official image... Still, it stalled at 1024B. If my router's faulty it should have started acting up well before the flash, I guess. That leaves me looking at nftp in a different light...

I don't think that the router is faulty, I think the stall is pure coincidence. It has happened to me once when my adapter was trying to get an IP address from the router DHCP style I told it not too while NFTP was starting up, opps and it broke the stream.
Easy fix though and its not happened since with many many re-flashes.

(Last edited by Wipster on 3 Feb 2011, 02:46)

You might very well be right again smile Do you think that could also explain the long time it took to erase the flash?

Now I'm looking into my options for adding the necessary headers and building the Xilinx cable discussed in this thread. I'm not any good at soldering but this looks a task I can tackle and a chance to get a bit better.

Frex wrote:

So there's no reason I can see why nftp would fail flashing 1122 and not 849 or the official image...

Does the image size matter?

Frex wrote:

@ fyi, any particular reason to prefer 4.01.30 over 4.01.40 ?

Because it has been mentioned in the wiki.

Access serial console to find out what's going on.

fyi wrote:

Does the image size matter?

Not at all, each and every one is the same size, I guess the flash is already maxed out in the official FW.

fyi wrote:

Because it has been mentioned in the wiki.

Strangely enough 'site:openwrt.org dg834 "4.01."' only brings up this thread :-?

fyi wrote:

Access serial console to find out what's going on.

AFAICT Adam2 is locked down in V3 so JTAG is really the only option. The unbuffered Xilinx is very simple to build but the serial cable is not because of the IC, so I'll think about building one further down the road.

I've been trying to decide whether to buy some 15W iron or a soldering station. The good stuff costs way too much so I'm only looking at the cheap ones, the dilemma is I want to keep the losses as low as reasonably possible in case I'm a total disaster, but if things turn out for the better, I'd like to try more projects.

(Last edited by Frex on 7 Feb 2011, 22:11)

Frex wrote:
fyi wrote:

Access serial console to find out what's going on.

AFAICT Adam2 is locked down in V3 so JTAG is really the only option.

Don't you want to know what's wrong?

Netgear DG834G v3 - OpenWrt Wiki

None of the datacables of my old cellphones seem to have the necessary IC, and the pre-built stuff is more expensive than I like, thus the serial cable will have to wait till I feel comfortable enough with a soldering iron smile

(Last edited by Frex on 7 Feb 2011, 22:11)

Thanks for the link. I was still hoping to get it built for much less than that, though.

And, on second thought, I happen to have an Ericsson data cable which [url=http://rtfm.vtt.net/pinouts/data/ericsson_cable_pinout.shtml.htm]might be a serial-TTL cable after all[/url, however only the end with the proprietary connector can be opened and there's nothing inside.

Is there a way to check if there's a MAX3232 inside it without opening it or connecting it to the equipment?

If there's 9 pin D-SUB FEMALE connector at one end, there must be a MAX3232/MAX232 inside. Youu need a multi-meter to check the signals.

Ok fyi, I've managed to get the headers and connectors today, so I've started tinkering with the wires.

Cut off the proprietary connector and stripped the wires, going to check the connections using this guide as a reference: http://www.neolics.com/pdfs/dku-5.pdf

Update: wasted a lot of time on the Ericsson cable, it has to be broken or something, I'm only reading xx mV. Plugged in another cable (Nokia 3310 this time) and I'm seeing ~2.25 on the TX/RX pins. Most 3310 circuit schematics out there call for a MAX232, elsewhere I'm reading the F-Bus runs at 3.3V and this seems consistent with voltage I'm measuring. I suspect people back in the day the 3232 was hard to find for whatever reason, and so they used the 232 and adjusted the rest circuit accordingly.

(Last edited by Frex on 9 Feb 2011, 03:41)

Ok, after recovering from the shock of some ugly, ugly pass through soldering on my poor V3 board, I managed to go back and make the serial console work. This is what I'm getting:

DG834 > addr=90020000
File for wrong Endian!
gocommand even2
Download.
Copying download from b0017000 to b4020000

CAUSE    = 0x30808028  Reserved Instruction
STATUS   = 0x00400002      EPC      = 0xb4020000
BADVADDR = 0xeffbdbb7      ERROREPC = 0x90000000

$ 0(zr):0x00000000  $ 8(t0):0x94010b0c  $16(s0):0x00000001  $24(t8):0x00000000
$ 1(at):0x00000001  $ 9(t1):0x00000020  $17(s1):0x00000001  $25(t9):0xbb6eef5f
$ 2(v0):0xffffffff  $10(t2):0x94010ad4  $18(s2):0x94010cb4  $26(k0):0x00000000
$ 3(v1):0xffffffff  $11(t3):0x00000010  $19(s3):0xb03f070b  $27(k1):0x00000000
$ 4(a0):0x00000001  $12(t4):0x00000078  $20(s4):0x9000889c  $28(gp):0x94008d70
$ 5(a1):0x90005bd0  $13(t5):0x94010b0c  $21(s5):0x3ffffe63  $29(sp):0x94010c4c
$ 6(a2):0x00006000  $14(t6):0x00000000  $22(s6):0xdbf9fffb  $30(s8):0x01000000
$ 7(a3):0xfffffff9  $15(t7):0x00000061  $23(s7):0xed7feb4f  $31(ra):0x9000235c

ADAM2 Revision 0.22.02
(C) Copyright 1996-2003 Texas Instruments Inc. All Rights Reserved.
(C) Copyright 2003 Telogy Networks, Inc.
memsize == 0x01000000Usage: setmfreq [-d] [-s sys_freq, in MHz] [cpu_freq, in MHz]
maca                  x:y:z:u:w:t
macb                  x:y:z:u:w:t+1
memsize               0x01000000
flashsize             0x00400000
modetty0              115200,n,8,1,hw
modetty1              115200,n,8,1,hw
bootserport           tty0
cpufrequency          211968000
sysfrequency          105984000
bootloaderVersion     0.22.02
ProductID             DG834
HWRevision            Unknown
SerialNumber          none
my_ipaddress          169.254.87.1
prompt                DG834
firstfreeaddress      0x9401bd20
req_fullrate_freq     125000000
mtd0                  0x900d0000,0x903e0000
mtd1                  0x90020000,0x900d0000
mtd2                  0x90000000,0x90020000
mtd3                  0x903e0000,0x903f0000
mtd4                  0x903f0000,0x90400000
oam_lb_timeout        100
modulation            MMODE

(Last edited by Frex on 11 Feb 2011, 04:50)

Frex wrote:
my_ipaddress          169.254.87.1

Is it useful?

I doubt it, Mad Professor had reported a similar situation in this thread (although mine should be a standard V3) and unbricked it via JTAG.

Anyway, I've started working on the Xilinx cable. Took me a fair amount of time, the colors had no discernible relationship with the schemes on wikipedia (the particular cable I cannibalized was over 20 yo IIRC, maybe that's why). I'm probably going to ask a silly question but I would hate to make a mistake here: the circuit layout on the Ciclamab site has been drawn looking at the DB25 connector from the back?

http://i54.tinypic.com/2vl8adv.jpg

The discussion might have continued from here.