updated squashfs to 3.1
cyberstorm
1.0:
$ wc -c upgrade_code.bin
3539000 wag354_neptune_upgrade_code.bin
3.1:
$ wc -c upgrade_code.bin
3293240 upgrade_code.bin
you can checkout the new version
cyberstorm
-dnsmasq update, I hope that dns problems using dhcp have been fixed
now
-upgrade from web interface should now not erase the rw partition.
-you can specify to execute a script at startup.
What we still need to fix:
-Firewall rules.
-Log view does not work. I think it's because log layout has changed
with the new busybox.
cheers
marco
Rail Aliev
2007/4/5, cyberstorm <marc...@gmail.com>:
toneworks
correctly to new). Anyway, some things I'd like to ask:
1) I saw you changed monitor_ps.c a little bit. Since monitor_ps and
cron work, why don't we simply add dnsmasq, httpd and (maybe) udhcpd
processes to its check ps tasks? I mean dnsmasq crashed from time to
time (specially with p2p apps), httpd more rarely. It could be a good
thing.
2) I wanna make a restore settings script at boot (in case router has
lost settings). But this cannot happen in case rwpart_enable and
cush_enable are set to 0 on defaults.c Could you set these at 1 on
default or is it a problem?
3) I checked ipconntrack values (eg. /proc/sys/net/ipv4/netfilter/
ip_conntrack_tcp_timeout_close_wait). Some of them are really bad set.
Do you think a tweaking would be a good idea? Wag's settings and
values now are these:
/proc/sys/net/ipv4/netfilter/ip_conntrack_max 8192
/proc/sys/net/ipv4/netfilter/ip_conntrack_generic_timeout 600
/proc/sys/net/ipv4/netfilter/ip_conntrack_tcp_timeout_close 10
/proc/sys/net/ipv4/netfilter/ip_conntrack_tcp_timeout_close_wait
259200
/proc/sys/net/ipv4/netfilter/ip_conntrack_tcp_timeout_established
432000
/proc/sys/net/ipv4/netfilter/ip_conntrack_tcp_timeout_fin_wait 120
/proc/sys/net/ipv4/netfilter/ip_conntrack_tcp_timeout_time_wait 120
/proc/sys/net/ipv4/netfilter/ip_conntrack_udp_timeout 46
Some other settings that might need change are these:
proc/sys/net/ipv4/tcp_max_syn_backlog
proc/sys/net/ipv4/tcp_rfc1337
net.core.rmem_default
net.core.rmem_max
net.core.wmem_default
net.core.wmem_max
net.ipv4.tcp_wmem
net.ipv4.tcp_rmem
net.ipv4.tcp_mem
4) Is there any possibility for some tools to be added in busybox?
like gunzip, or awk, in case we need to make a better script. I made a
copy of nvram and it took me 65k! half of mnt space! gziped is 6k. I
tried compression with tar but got no luck.
toneworks
Ο/Η cyberstorm έγραψε:
> -Log view does not work. I think it's because log layout has changed
> with the new busybox.
I can't say it's fixed, because page has lost some of its
functionality. But at least now someone can browse/erase logs from
httpd. What he can't do is select filter to view specific logs (this
means probably we will need to change page layout too). Maybe a box
[ ] to enter how many rows of log he wants per page is a good
substitute.
Review change: http://users.forthnet.gr/her/gojira/log.tar.bz2
cyberstorm
> Ok, I'm on 129 but with the old dnsmasq (can't get make install
> correctly to new).
fixed, I forgot to commit the Makefile
Anyway, some things I'd like to ask:
>
> 1) I saw you changed monitor_ps.c a little bit. Since monitor_ps and
> cron work, why don't we simply add dnsmasq, httpd and (maybe) udhcpd
> processes to its check ps tasks? I mean dnsmasq crashed from time to
> time (specially with p2p apps), httpd more rarely. It could be a good
> thing.
yes, this could be done. I noticed that the bad thing of monitor_ps
is that it checks existence of process, despite of being effectively
selected in your custom configuration. E.g. I dont use upnpd, but
monitor_ps continue to check its existence, and try to start it if it
is not running.
Maybe we could change it into a shell script. When a process starts,
it adds a line to the list of process to be monitored. When the
process stop the line is removed.
What do u think?
>
> 2) I wanna make a restore settings script at boot (in case router has
> lost settings). But this cannot happen in case rwpart_enable and
> cush_enable are set to 0 on defaults.c Could you set these at 1 on
> default or is it a problem?
>
Well, I dont know. A common user probably will never use these 2
functionalities, so including them as default into every startup could
slow down starting.
> 3) I checked ipconntrack values (eg. /proc/sys/net/ipv4/netfilter/
> ip_conntrack_tcp_timeout_close_wait). Some of them are really bad set.
> Do you think a tweaking would be a good idea? Wag's settings and
> values now are these:
>
> /proc/sys/net/ipv4/netfilter/ip_conntrack_max 8192
> /proc/sys/net/ipv4/netfilter/ip_conntrack_generic_timeout 600
> /proc/sys/net/ipv4/netfilter/ip_conntrack_tcp_timeout_close 10
> /proc/sys/net/ipv4/netfilter/ip_conntrack_tcp_timeout_close_wait
> 259200
> /proc/sys/net/ipv4/netfilter/ip_conntrack_tcp_timeout_established
> 432000
> /proc/sys/net/ipv4/netfilter/ip_conntrack_tcp_timeout_fin_wait 120
> /proc/sys/net/ipv4/netfilter/ip_conntrack_tcp_timeout_time_wait 120
> /proc/sys/net/ipv4/netfilter/ip_conntrack_udp_timeout 46
>
> Some other settings that might need change are these:
> proc/sys/net/ipv4/tcp_max_syn_backlog
> proc/sys/net/ipv4/tcp_rfc1337
> net.core.rmem_default
> net.core.rmem_max
> net.core.wmem_default
> net.core.wmem_max
> net.ipv4.tcp_wmem
> net.ipv4.tcp_rmem
> net.ipv4.tcp_mem
How would u change these settings?
I really dont know which are the optimal values.
>
> 4) Is there any possibility for some tools to be added in busybox?
> like gunzip, or awk, in case we need to make a better script. I made a
> copy of nvram and it took me 65k! half of mnt space! gziped is 6k. I
> tried compression with tar but got no luck.
Of course, I would like to add:
-find
-less
-which
-syslogd circular buffer support
-gzip
-awk
maybe others?
toneworks
http://img84.imageshack.us/my.php?image=logsoh6.jpg
Menu is aligned on left side. The ex-filter scroll down menu now
chooses lines per page.I also changed dimension of table (notice that
logs are not wrapped, there are scroll bars on the sides). (hmm we
need to fix that "5bUpnp" crap).
I send you a newer dhcp_reserved.asp too. In firefox everything was
ok, but in IE it had a minor glitch. If you find something similar,
tell me so.
Review and change what you like:
http://users.forthnet.gr/her/gojira/log_view.tar.bz2
About monitor_ps thing:
Some important and standard processes should be monitored, that's for
sure.
I was thinking about a nvram approach. (check if it exists in ps and
configured in nvram).
But as I say, you know better than me.
On those defaults.c matter, it's ok.I usually lose settings every
couple of days (maybe I push too hard my wag). So I need to do
something for router to restore nvram on its own. I found gzip for
mipsel, it's about 30k compressed. I must say I'm closer to my goal :D
In conntrack settings, I've posted everything that imho needs
checkout. For instance, router keeps for 5 days
ip_conntrack_tcp_timeout_close_wait ips. If you use p2p, this is
really dumb. We could use settings from openwrt, or other optimised
fw. As far as I can remember, on other touched firmwares this setting
was for 20 mins instead of 5 days.
That's for now, C ya
Jb boin
produce this error when using it "traceroute: unknown protocol icmp:
No such file or directory".
Jb boin
toneworks
It's clearly a busybox issue.
I tried to mess around with traceroute.c, that didn't work out.
So the best solution with minimal changes to busybox is simply to add
protocols file in /etc/.
I moved protocols from my linuxbox to /etc/, editing Makefile in /misc
(adding a line like the one for neptune banner). I tested it and
traceroute works like a charm.
I also added two parameters in Status_DSL.asp: the ones concerning max
ds/us attainable bitrate. As I figured out, it's pretty easy to add
any stat from avsar_modem_stats to .asp using existing functions.
Changes are here:
http://users.forthnet.gr/her/gojira/max_trace.tar.bz2
Jb boin
just modify files without having to make a new firm?
toneworks
"Makefile" and "protocols" in trunk/router/misc/ and do "make
install", make rom", "make upgrade" for a working traceroute.
I think the best option is wait for cyberstorm to update svn and get
changes from there.
It won't take long.
Marco Vedovati
Hash: SHA1
yes, sorry, I am busy with other stuff in this period. I will commit
your patch as soon as possible.
It would be great to build the module "mini_fo" for our router, it
allows to coexist both squashfs and a editable file system, using a
second partition to store changes....
cheers
Il giorno 09/apr/07, alle ore 18:40, toneworks ha scritto:
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)
iD8DBQFGGno88PPSbh9Zs4YRAgJwAJwJIWl/mcMKQ2rSGRPbNl85Mha3UgCcDUqb
/KezG9DIwjfah5sGU4o2z38=
=RSId
-----END PGP SIGNATURE-----
toneworks
had some segmentation faults. It also occupied 150kb more space, that
sucks. I returned to original version.
I have a problem with internet led. It randomly turns off (even if I'm
having internet).
Since noone has reported such an error so far, it must have been one
of my latest changes (those aren't on svn yet). I believe it's the one
using avsar_get_stats function.
So don't patch status.c and Status_DSL.asp (those are on
max_trace.tar.bz2).
When I'm sure about it, I'll post again.
Marco Vedovati
Hash: SHA1
Il giorno 10/apr/07, alle ore 17:27, toneworks ha scritto:
>
> I tried to do some things with iptables, i did compile 1.3.6 but .bin
> had some segmentation faults. It also occupied 150kb more space, that
> sucks. I returned to original version.
>
yes I've tried too, but I get a problem compiling one extension, so I
give up...
> I have a problem with internet led. It randomly turns off (even if I'm
> having internet).
> Since noone has reported such an error so far, it must have been one
> of my latest changes (those aren't on svn yet). I believe it's the one
> using avsar_get_stats function.
> So don't patch status.c and Status_DSL.asp (those are on
> max_trace.tar.bz2).
>
> When I'm sure about it, I'll post again.
ok, I have never seen the problem of led...
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)
iD8DBQFGG9Eb8PPSbh9Zs4YRAs53AJ9s1rJVJMNctxB8qC35uvweXNnAJgCgmN64
Uo8SCbvKuqlowx6PkBi7NQU=
=7EeT
-----END PGP SIGNATURE-----
toneworks
It was iptables update causing the problem (even if i replaced
folder).
I updated all /router from svn. I patched newer updates and all run
smoothly.
This time I'm gonna give .bin for any testers to make sure it's ok.
http://users.forthnet.gr/her/gojira/neptune.135.unstable.v1.annexa.bin
Changes (from svn #135)
1) Icmp error fixed (traceroute command)
2) Log view change (Log.asp)
3) Max ds/us attainable bitrate on Status_DSL.asp
4) httpd, dnsmasq, ntpd now monitored by monitor_ps. (this means that
ntpd runs even if qos is enabled at boot. Many users want ntpd and qos
and this is the easiest way to fix things).
If you notice internet led turn off while you're online, or when you
enter "ps" in telnet you get "segmentation fault" this means problems.
These two happen an the same time (ps error is permanent, led turns on
later).
If you test it, please report success/failure.