Is a PC the destination behind the port forward rule?  Have you disabled the firewall on the PC and checked?

@starcms
As you guessed changing to eth1 did not make any difference in SQM QoS. Still need to define download and upload both. Thought if I played some with leaving the download at 0 and playing with the upload I could get things to a happy place. Not true. I don't play games so some buffer bloat is not a big thing to me. It is more of just playing to see what I can do with the router and features.

Latest of David's builds has turned my wrt3200 into to the  router I thought it should be. Fast and reliable. Runs well with Brainslayer's latest DD-WRT builds as well, but for some reason 5GHz wireless on his builds is slower then David's and latest Linksys OEM on my simple iperf test. He is using the latest drivers and 4.9 kernel.

I'm on Cox 300/30 plan. I get about 400 down with a SB6183 and can use SQM under FQ_CODEL (simplest.qos) full speed. CAKE choked about about 160 ish. Late at night I will actually see momentary bursts into the mid 500s.

I'm only using it on upstream at the moment, its very rare we saturate the WAN link here. Took almost all of my upstream bufferbloat away. This is the most asymmetric link I've ever used, by far.

A SB6141 will be tough on a node, but if you are getting 300 with a SB6141 it must not be as busy of a node.

(Last edited by PolarisX on 4 Sep 2017, 02:46)

Wow, can't believe you are getting 400-500+ on Cox's 300/30 plan.  Is that sustained downloads or only for a 10-20 second burst at the beginning of a download?  If its sustained, that makes me think that since 300/30 is their fastest plan, they uncap the modem for those who have it.  I would upgrade my modem, but with Cox rolling out either DOCSIS 3.1 or GigaBlast (FTTN) to basically all their service areas in the next year, I don't want to waste my money buying a DOCSIS 3.1 modem now, and then wind up getting GigaBlast in my neighborhood instead.  I get the full 300/30 on my SB6141, but no more than that.  From your experience, I'm sure I could get more with a 16 channel modem, but since the only reason I even have their 300/30 plan is they are giving to me for $5/month less than the regular price of their 50/5 plan for 12 months for being a long-time customer, I'm plenty happy and hardly ever come close to using 300mbps down anyway (hence why I had been on their 50/5 plan for years, back since it was 25/5, and before that 10/2, and going way back to 1999, 5/0.5).  I had upgraded to my SB6141 just about 2-3 years ago when they bumped the speed to 50/5 because my old DOCSIS 2.0 modem from 1999 maxed out around 33mbps.  I had been getting 60/6 on their 50/5 plan with my SB6141.

Back on topic, so it is CAKE that chokes at 150-160mbps, not SQM or the router, thanks for the info! I wonder why that is -- if it's a bug or if the way it works requires much more processing power.  Since Cox has minimal to no buffer-bloat on the downstream, I'll stick with CAKE for now, since I have the downstream set at 0 and it works great for me like that, practically no buffer-bloat on the downstream to begin with and CAKE takes away practically all buffer-bloat from the upstream, always score an A or A+ on DSLReports' test.

(Last edited by starcms on 4 Sep 2017, 07:27)

@david, good thing you hadn't released a new build yet, because the kernel just got another bump, now at 4.9.47

(Last edited by starcms on 4 Sep 2017, 07:17)

Actually it's buffalo nas with bittorent. I want to access it's web ui outside my home. I dont think it has a firewall built in

@all,

A suggestion to make your network "invisible" from the internet.  Very simple way to make any wanna-be hackers pass-by your IP address because it will appear to them that there is nothing there to hack into.

First, in LuCi, under Network --> Firewall, I recommend changing the drop-down setting for the WAN Zone Input from reject to drop.  Click Save.  This will make all of your ports stealth instead of simply closed.

Then, click on the Traffic Rules tab, and uncheck the enable checkbox for Allow-Ping.  Pretty obvious what this does --if anyone tries to ping your WAN IP address, they won't get a response. Click Save & Apply.

Together, both settings make your router/network appear to not exist to the outside world.

(Last edited by starcms on 4 Sep 2017, 23:34)

That is controversial advice, not "highly recommended".
https://wiki.openwrt.org/doc/uci/firewa … _vs_reject

Greets,

The question of how to install davidc502's community build on a 1900AC(v1, Mamba) (currently installed Cybrnook's lede community build on both partitions) has been on my mind. The obvious and easy procedure is sys-upgrade via Luci (downld/check sums first). Is it that easy then?

I have read and searched this thread.. and other resources.

Its okay to lose configs. I assume 4.9 kernel is still twitchy on Mamba..

Thanks,

Duster

Since LEDE is on both partitions, you should be able to just do a sysupgrade from LuCi using the .bin file. Also, configurations should not be lost unless you want them to be.

For Mamba owners, I still recommend the 4.4.x build over Kernel 4.9.

It's "highly recommended" by me   I have edited my post to clarify. 

For the majority of users, it is much safer for their ports to be steath instead of closed and there are no side effects.  Yes, I know alot of the "experts" each have their own opinion, and I don't want to start a debate here as it's not the place for it, so again, it's just highly recommended by me.  I'm not forcing my advice on anyone.  It's just a suggestion that users might not have known they had.  Thank you for posting that link so everyone can read up on it and make their own decision.

Edit: Congrats, that was your 3,000th post!

(Last edited by starcms on 4 Sep 2017, 23:35)

Ok, I finally have a question.  Are both odhcpd and odhcp6c required?  From my understanding, it seems odhcp6c is required for WAN DHCPv6 and odhcpd is for LAN DHCPv6 and RA.  Is that correct?  Or does odhcpd handle everything now and odhcp6c is no longer needed?

@Duster, are you coming from Cybernook's "stable" build or his "beta" build?  His stable build was built from the stable LEDE branch, while his beta build is built from the snapshot/master branch just as @david's builds are. 

Because alot has changed between the two branches, especially since the last time Cybernook released a build, I would highly recommend not saving/restoring your settings/config, but starting fresh, especially/definitely if you are coming from his "stable" build (which you most likely are because if I remember right, his beta builds only had kernel 4.9 and his stable builds only had kernel 4.4).

(Last edited by starcms on 4 Sep 2017, 23:51)

Looking at the two I would think you are right.... Would love to investigate odhcp6c, but my ISP doesn't believe in IPV6

@starcms -- I should have time to start working on a new build this weekend. I'm hoping besides a kernel bump that we also have a new wifi commit to try.....

You have FTTH but no IPv6?  WTF?! lol

You don't get any IPv6 connectivity?  Your ISP doesn't assign you an IPv6 WAN address?  I thought pretty much all ISP's were IPv6 ready by now...

(Last edited by starcms on 5 Sep 2017, 02:36)

Yeah, the mwlwifi github has been quiet as of late.  Are you just hoping we get a new commit or is there an issue being worked on that @yuhhaurlin is about to release a commit for?

When the mwlwifi driver is finally updated to give the 3200ACM MU-MIMO support, I'll probably pick one up.  Is 160MHz and/or 80+80MHz functional yet?  Because without the "802.11ac wave 2" features working yet on the 3200ACM, I'm perfectly happy with my 1200AC.

@david,

In your next build, could you include this patch from @hnyman which enables CPU frequency scaling?

All info about it is here: https://forum.lede-project.org/t/cpu-fr … m-etc/2808

This is a link to the commit: https://github.com/hnyman/source/commit … e0ce41c1ac and this is the link to the downloadable patch: https://github.com/hnyman/source/commit … c1ac.patch

Instructions from @hnyman on how to use it seem simple enough:

Go to your buildroot root
Download the patch with ever from github. Direct link above.
Apply the patch with "patch -p 1 -i filename
Recompile firmware (after make clean)

The patch allows the SoC/CPU to scale down to 50% speed when not needed to save power and reduce temperatures, and scale back up to 100% when needed.  As it is right now, it always runs at 100% speed.   From the responses in the lede-forum thread linked above, it seems to work great with reducing temps, and no problems were reported.  It appears it is only for Kernel 4.9.

(Last edited by starcms on 5 Sep 2017, 03:10)

Fairly ridiculous right? I've called a couple of times, and each time they talk about a project for IPv6, but with a unknown completion date.

All I can really do is to stand by until it is rolled out.

From what I can gather, reading the github posts, yuhhaurlin is currently busy on other projects. At some point he will circle back around to finish MU-MIMO, but "later"

I haven't tested 160MHz in a couple of months, but when I did, 5Ghz crashed after radar was detected, and hence had nowhere to go... lol  It could be the problem is fixed by now.

Scan Ebay, as there are some pretty good deals for refurbished units as well as pre-owned.  I'm confident the driver is good enough to recommend... I have a work at home wife who is on wifi all the time, and we don't have any issue. However, besides streaming video from the FireTV, Wifi is used pretty lightly, so heavy users would probably have a different experience.

I'll circle back to this towards the end of the week along with another item we talked about last week.  Will communicate with you as the time draws nearer. Will probably want me and you to test for a day or so before releasing to help ensure there are no issue.

By the way... from command line running "htop", processors on the acs and acm seem to sit idle most of the time unless under heavy load. Is there something htop is not showing when it comes to cpu "speed"?

Honestly though, without MU-MIMO and possibly without 160/80+80MHz, I have nothing to gain from upgrading from my 1200AC. 

By the time the driver finally does support MU-MIMO, then they will be even cheaper

Sounds good, look forward to hearing from you and testing out some builds.

No, since the CPUs do sit idle for the majority of the time, the patch would allow the CPUs to scale down to 50% of their maximum speed to reduce temperatures and power consumption.  That's the biggest advantage of adding the patch.  The CPUs will only scale up to 100% when under heavy load.  Plus it will actually let us view the CPU speeds
As it is now without the patch, the CPUs are always at 100% speed, even when idle, burning excess power and raising temps.

(Last edited by starcms on 5 Sep 2017, 03:42)

Could you add a guest network on your next release?

If you search the wiki, you'll find all the instructions.
https://wiki.openwrt.org/doc/recipes/guest-wlan and https://wiki.openwrt.org/doc/recipes/gu … binterface.