Firmware v1.9.2.7 CR6, updated to CR6a ... now CR6b [Oleg]

[alien433]

i sent a mail to you oleg

[nils]

Dear Oleg,

first of all let me thank you for your great work

I run into some trouble, when I tried to connect a netgear wg500v2 wlan card with the wg-500g with your 1.9.2.7 CR6b Firmware.

I watched the behaviour that the card connects well and a first transmission of data from a website was succesfull. But after a short time (a minute or two) the transmission went down and no more packets could be received.

After some troubleshooting I found that the problem comes from the packet size / mtu settings of the card. The default was set to 3000. When I change it to 1490 everything went well.

I am no IP-expert, but I wonder if there is something wrong with the fragmentation handling in the firmware. I would have expect, that the router would fragment the respond packets, even if the default mtu settings form the netgear card are nonsense.

[auduny]

I still have problems with random resets of my wl500g, I've redirected the syslog to a nearby linuxbox without me getting any closer to the problem

Firmware: 1.9.2.7-6b

I've turned off most of the features, only extra features is wshaper, dropbear ssh and a connected raw-printer. I also usually have a 250GB USB disk connected, but I've removed this as smb usually leakes memory and starts killing of processes.

Relevant loglines:

Sep 5 23:45:53 jens.thehive.org udhcpc[125]: Lease of xxx.xxx.xxx.xxx obtained, lease time 60
Sep 5 23:45:54 jens.thehive.org kernel: HTB init, kernel part version 3.16
Sep 5 23:45:59 jens.thehive.org kernel: NET: 26 messages suppressed.
Sep 5 23:45:59 jens.thehive.org kernel: Neighbour table overflow.
Sep 5 23:46:03 jens.thehive.org kernel: NET: 27 messages suppressed.
Sep 5 23:46:03 jens.thehive.org kernel: Neighbour table overflow.
Sep 5 23:46:08 jens.thehive.org kernel: NET: 37 messages suppressed.
Sep 5 23:46:08 jens.thehive.org kernel: Neighbour table overflow.
Sep 5 23:46:30 jens.thehive.org syslogd started: BusyBox v1.00 (2005.05.11-18:29+0000)

First crash, then a crash just after.

Sep 5 23:46:42 jens.thehive.org udhcpc[78]: Received DHCP NAK
Sep 5 23:46:54 jens.thehive.org last message repeated 2 times
Sep 5 23:46:57 jens.thehive.org udhcpc[78]: Lease of xxx.xxx.xxx.xxx obtained, lease time 60
Sep 5 23:46:59 jens.thehive.org kernel: HTB init, kernel part version 3.16
Sep 5 23:47:28 jens.thehive.org udhcpc[129]: Lease of xxx.xxx.xxx.xxx obtained, lease time 60
Sep 5 23:47:29 jens.thehive.org kernel: HTB init, kernel part version 3.16
Sep 5 23:47:58 jens.thehive.org udhcpc[129]: Lease of xxx.xxx.xxx.xxx obtained, lease time 60
Sep 5 23:47:59 jens.thehive.org kernel: HTB init, kernel part version 3.16
Sep 5 23:48:24 192.168.1.1 System log daemon exiting.
Sep 5 23:48:45 jens.thehive.org syslogd started: BusyBox v1.00 (2005.05.11-18:29+0000)

One strange thing I've discovered using the broadband down here in Australia (I'm actually norwegian), I get the same IP as the ADSL router.

Sep 6 00:16:23 jens.thehive.org dhcp client: bound IP : xxx.xxx.xxx.xxx from yyy.yyy.yyy.yyy

where x and y are identical. So i get a dhcp lease from the same ip as i am given.This lookes strange in my routing table as well. Default gateway is the same as the interface on eth1, but it works and i can stilll access the router by going to 10.1.1.1

I also get this if I connect a laptop directly to the router.

I am not sure if this is relevant. Anyway the router resets every day. Some days it can be stable for several hours with heavy load. And some days it crashes every 10 minutes.

Any ideas? And is there any way to make the dhcpd-client have a longer lease period?

What is the watchdog process checking?

[Oleg]

Sep 6 00:16:23 jens.thehive.org dhcp client: bound IP : xxx.xxx.xxx.xxx from yyy.yyy.yyy.yyy

where x and y are identical.

Any ideas? And is there any way to make the dhcpd-client have a longer lease period?

You've to adjust DHCP server settings - most likely your modem is running DHCP and gives really strange lease time of 1 minute, as well as forms incorrect answers... Most likely, this is the reason for reboots, as wl500g is reconfiguring over and over, triggering wl500g wan port hw bug...

[auduny]

Unfortunatly. The adsl-modem isn't configurable to go into any other modus than this strange switch-modus and the dhcp-server is either non configurable, or a passthorugh from my isp. Could the dhcp-client be configured to have a longer leasetime? Or maybe it would be better to set a static ip and see how that works out... I'll try the last idea.

Tried, and failed. Still reboots even with a using static ip. It still dies when i peak the my adsl line. I guess the ADSL modems sends icmp packages back telling it to lower it's troughput and the server does not like that and dies..
But why does it die? Running hot? Bad eth1 driver? What is the whachdog doing? Rebooting if not able to write to a file?

Sep 6 11:29:37 jens.thehive.org kernel: Neighbour table overflow.
Sep 6 11:29:42 jens.thehive.org kernel: NET: 34 messages suppressed.

Are these anything to worry about?

[dunatos]

is it suitable for WL500g?

[auduny]

Is there any way to check the 10/100 half/full duplex modus of eth1 from the commandline?

like ethtool on linux.

ay@ray:~$ sudo ethtool eth0
Password:
Settings for eth0:
Supported ports: [ TP MII ]
Supported link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
Supports auto-negotiation: Yes
Advertised link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
Advertised auto-negotiation: Yes
Speed: 100Mb/s
Duplex: Full
Port: MII
PHYAD: 1
Transceiver: internal
Auto-negotiation: on
Supports Wake-on: g
Wake-on: g
Current message level: 0x00000007 (7)
Link detected: yes

And maybe even modify it on the run?

[oleo]

Just a quick tought. Are there any llimitations to upgrade kernel from 2.4.20 to somewhat higher (eg 2.4.31)

I have thist annoying messages:

KERNEL: assertion (newsk->state != TCP_SYN_RECV) failed at tcp.c(2225)
KERNEL: assertion ((1<<sk2->state)&(TCPF_ESTABLISHED|TCPF_CLOSE_WAIT|TCPF_CLOSE)) failed at af_inet.c(689)
KERNEL: assertion (newsk->state != TCP_SYN_RECV) failed at tcp.c(2225)
KERNEL: assertion ((1<<sk2->state)&(TCPF_ESTABLISHED|TCPF_CLOSE_WAIT|TCPF_CLOSE)) failed at af_inet.c(689)
KERNEL: assertion (newsk->state != TCP_SYN_RECV) failed at tcp.c(2225)
KERNEL: assertion ((1<<sk2->state)&(TCPF_ESTABLISHED|TCPF_CLOSE_WAIT|TCPF_CLOSE)) failed at af_inet.c(689)
KERNEL: assertion (newsk->state != TCP_SYN_RECV) failed at tcp.c(2225)
KERNEL: assertion ((1<<sk2->state)&(TCPF_ESTABLISHED|TCPF_CLOSE_WAIT|TCPF_CLOSE)) failed at af_inet.c(689)
KERNEL: assertion (newsk->state != TCP_SYN_RECV) failed at tcp.c(2225)
KERNEL: assertion ((1<<sk2->state)&(TCPF_ESTABLISHED|TCPF_CLOSE_WAIT|TCPF_CLOSE)) failed at af_inet.c(689)

Which is fixed in 2.4.21-pre5 and later.

Another annoying thing which is hopefuly fixed in latest kernels is USB timeout issue on busy server with gift and torrent with load rarely under 1. But for this I've found solution (Revoltec case and WD3200 disk).


Now if you skipped bla-bla from above. The question is? What are the issues when upgrading to other kernel? Are kernel pathes needed for WL500GX? Does stock kernel suffice? Only .config needed, lzma patch, usb patch ...?

I've compiled Oleg source and unslug tree w/o problem. But bricking router with naive replacement with other kernel bothers me.

[Oleg]

I guess the ADSL modems sends icmp packages back telling it to lower it's troughput and the server does not like that and dies..
But why does it die? Running hot? Bad eth1 driver? What is the whachdog doing? Rebooting if not able to write to a file?

Sep 6 11:29:37 jens.thehive.org kernel: Neighbour table overflow.
Sep 6 11:29:42 jens.thehive.org kernel: NET: 34 messages suppressed.

Are these anything to worry about?

Try replacing your ADSL modem with other model. Looks like it's really crazy... Neighbour table overflow means, that too many hosts are around.
Try dumping it using
cat /proc/net/arp

Do you've any multicast in your network?

Also, replacing with wl500g deluxe could help, as it does not have ethernet controller hw bugs...

Watchdog does not reboot wl500g, it just respawns some processes...

[Oleg]

Is there any way to check the 10/100 half/full duplex modus of eth1 from the commandline?

mii-tool

And maybe even modify it on the run?

Use web interface for this.

[Oleg]

Just a quick tought. Are there any llimitations to upgrade kernel from 2.4.20 to somewhat higher (eg 2.4.31)

yes, binary wl.o.

Which is fixed in 2.4.21-pre5 and later.

backport a fix to 2.4.20.

Another annoying thing which is hopefuly fixed in latest kernels is USB timeout issue on busy server with gift and torrent with load rarely under 1. But for this I've found solution (Revoltec case and WD3200 disk).

usb stuff is from the 2.4.25 kernel.

Now if you skipped bla-bla from above. The question is? What are the issues when upgrading to other kernel? Are kernel pathes needed for WL500GX? Does stock kernel suffice? Only .config needed, lzma patch, usb patch ...?

Binary wl.o, as well as other broadcom customizations to the kernel.

Well, openwrt uses now the latest from 2.4.21, as they've ported broadcom stuff and adjusted kernel to be wl.o compatible, but this requires a lot of testing to be absilutely sure, that everything still works...

[oleo]

backport a fix to 2.4.20.

Found this patch which should help with assertion (newsk->state != TCP_SYN_RECV) failed at tcp.c(2225)

diff -urN linux-2.4.20/net/ipv4/tcp_minisocks.c linux-2.4.21/net/ipv4/tcp_minisocks.c
--- linux-2.4.20/net/ipv4/tcp_minisocks.c       2002-11-28 15:53:15.000000000 -0 800
+++ linux-2.4.21/net/ipv4/tcp_minisocks.c       2003-06-13 07:51:39.000000000 -0 700
@@ -935,6 +938,12 @@
        if (flg & (TCP_FLAG_RST|TCP_FLAG_SYN))
                goto embryonic_reset;

+       /* ACK sequence verified above, just make sure ACK is
+        * set.  If ACK not set, just silently drop the packet.
+        */
+       if (!(flg & TCP_FLAG_ACK))
+               return NULL;
+
        /* If TCP_DEFER_ACCEPT is set, drop bare ACK. */
        if (tp->defer_accept && TCP_SKB_CB(skb)->end_seq == req->rcv_isn+1) {
                req->acked = 1;

But I've tought of something like quantum leap in kernel (2.4.20->.31) as there are so much bugfixes which I would like to have.

Anyway TNX. Will test this patch and confirm if it helped.

[spank_santa]

Hi,

The wl-hdd forum section is scarely quiet that why post this question here.

I have a wl-hdd with Oleg's latest firmware 1.9.2.7-6b installed. When I try to upload a file with FTP (which is bigger than 2GB) to the HD it stops at 2 GB. Do I do something wrong ? or is this a known "feature" Do I have to change something in the stupid-ftp conf ?

BJ

[Oleg]

stupid-ftpd is not designed to handle files > 2GB.

[spank_santa]

Hi Oleg,

Thanks for your answer. I will try to upload VSFTP, I read on another site how to implement this, so I will give it a try. Thanks for your support and keep up the good work.

BJ