OpenWrt Forum Archive

Topic: davidc502 1900ac 3200acm builds

The content of this topic has been archived between 26 Feb 2018 and 7 May 2018. Unfortunately there are posts – most likely complete pages – missing.

jefftk99 wrote:

Hi,

Any idea when the next build will be released with the "amsdu" patch incorporated into the build ?

The plan is to incorporate it into the build, but you don't have to wait.  Just download the appropriate driver for the Kernel build you are on, and install.

As for the next build, there is not a specific time frame. Generally, it is subjective as to when, and is based on the number of changes, and importance of those changes that kind of dictate when the next build will take place. On July 28th we had a Kernel bump, and expect another one to be coming soon.

jefftk99 wrote:

Hi,

Any idea when the next build will be released with the "amsdu" patch incorporated into the build ?

@david usually releases a new build when something somewhat major changes: a new kernel version, new driver version, etc.  Just follow the directions and install it manually.  It works great.  Been running @david's r4651 with the patched mwlwifi driver with amsdu disabled for about a week now.  Never better.

jvquintero1021 wrote:

@starcms & @bill1228

Ok so I've restarted from scratch loading LEDE fresh from factory. First thing I did was I set the DNS settings as per @starcms instructions. Put custom DNS in DNS forwardings of DHCP and DNS settings, and checked the box for ignore resolve file and everything worked as advertised. So appears you are correct sir's.

Now If one of you could maybe help me further troubleshoot? I continued my setup for the rest of the router and what do you know I get the same problem as before after installing and setting up openvpn-policy-routing package. again unchecking the checkbox for ignore resolve file fixes it. I have done a check for any DNS leaks after unchecking ignore resolve file and it appears that DNS is not going through the ISP DNS server. So do either of you have any insight that could maybe help me out?

NOTE 1: I noticed that before checking the box for ignore resolve file that the line below it "Resolve file" was populated with one of the /tmp/resolv.conf files. After unchecking ignore resolve file again the line "Resolve file" no longer was populated with the /tmp/resolv.conf file it is just blank now. So maybe just deleting this line has the same effect as checking the box?

NOTE 2: I did reboot the router after changing any of the settings just to be sure.

Thanks for all your help fellas, I really appreciate it!

I can't really help with openvpn, no experience.  As for unchecking the box for ignore resolve file, even if the field for resolve file is left blank, it will still pull DNS servers from /etc/resolv.conf if the box is unchecked (/etc/resolv.conf and /tmp/resolv.conf are both auto-generated based on your ISP's DNS servers).  At least for my setup, the contents of /tmp/resolv.conf and /etc/resolv.conf are always the same, filled with my ISP's DNS servers.

(Last edited by starcms on 9 Aug 2017, 23:08)

Pasxalisk wrote:

I am facing issues with the latest David's built.
It is like it doesn't boot completely.
Some commands i have inside rc.local are not executed, and i cannot perform a reboot neither from command line, nor from Luci (it just refreshes the page when i hit reboot).
I checked the logs, but i didn't see something relevant there.
Tried a couple of times to re-flash it, but the problem remained.
Any idea what can cause it?

Sounds like when I grab the wrong file to flash. Make sure you are grabbing the file for your router, you did not say what model. Then the next question is are you upgrading from within LEDE  or Linksys UI?

You have squashfs-factory  --Use for flashing from Linksysy UI
squashfs-sysupgrade.  -- Use to upgrade from within LEDE.

If this is not it I have no idea,

Just a quick follow up to the page loading problem, I went back to LEDE 17.01.0 to make sure I wasn't remembering wrong and that the page loading problem wasn't present in it, spent the past few days running it with dnscrypt and without dnscrypt, no page loading problems either way, and like starcms mentioned nonwildcard is set to 0 by default, so if you're still having problems with page loading just go back to 17.01.0 and you should be good.

starcms wrote:

@david usually releases a new build when something somewhat major changes: a new kernel version, new driver version, etc.  Just follow the directions and install it manually.  It works great.  Been running @david's r4651 with the patched mwlwifi driver with amsdu disabled for about a week now.  Never better.

On r4576 with my WRT3200acm and no issues here ;-P

And loving it...

(Last edited by sean.p.hopkins on 10 Aug 2017, 01:27)

beachbum wrote:

Just a quick follow up to the page loading problem, I went back to LEDE 17.01.0 to make sure I wasn't remembering wrong and that the page loading problem wasn't present in it, spent the past few days running it with dnscrypt and without dnscrypt, no page loading problems either way, and like starcms mentioned nonwildcard is set to 0 by default, so if you're still having problems with page loading just go back to 17.01.0 and you should be good.

Or just use @david's latest build and set nonwildcard equal to 0 in /etc/config/dhcp.  And/or if you are using custom DNS servers, enter them in DHCP and DNS settings and check the box for Ignore Resolv File as I had mentioned.  Either or both of those things should fix the DNS resolving issue.  It's been confirmed that this solves any resolving issues a few people were having. 

LEDE 17.01.0 is built from the "stable" branch/not the development branch like @david's builds, is a few months out of date (which puts it several months behind the development branch, especially regarding mwlwifi driver), and most importantly isn't compiled by @david with all the added features/packages and the amsdu fix in the patched driver.  Doesn't belong in this thread.  Pretty much anyone would see decreased performance with 17.01.0, especially those with a 3200ACM.

(Last edited by starcms on 10 Aug 2017, 02:21)

bill1228 wrote:

Sounds like when I grab the wrong file to flash. Make sure you are grabbing the file for your router, you did not say what model. Then the next question is are you upgrading from within LEDE  or Linksys UI?

You have squashfs-factory  --Use for flashing from Linksysy UI
squashfs-sysupgrade.  -- Use to upgrade from within LEDE.

If this is not it I have no idea,


I have WRT1900ACS v1.
I am flashing the correct file and i am using the factory.img from Linksys UI.
It is really weird, and annoying.
It is like something doesn't run like it is supposed to.
Is there any command to see which processes are forbiding the reboot, for example?
Because, i am pretty sure that the one that doesn't allow me to reboot, also doesn't start correctly and it messes with the boot sequence.

I see there is a new mwlwifi driver, 10.3.4.0-20170810. Again this doesn't seem to have anything exciting. A one line change to, I think pcie, at least it is to the pcie/dev.h file. The devs will know better.

bill1228 wrote:

I see there is a new mwlwifi driver, 10.3.4.0-20170810. Again this doesn't seem to have anything exciting. A one line change to, I think pcie, at least it is to the pcie/dev.h file. The devs will know better.

There are a few changes:

  • Fixing RCU stall (still needs to be tested per issue 198 on the driver github)

  • Fixing a build issue with Kernel 4.4.x and OpenWRT (including a needed header file)

  • Fixing a comparison error

  • Updating the driver version to reflect these changes.

I agree though, I am not 100% sure on the impact this will have.

AjkayAlan wrote:
bill1228 wrote:

I see there is a new mwlwifi driver, 10.3.4.0-20170810. Again this doesn't seem to have anything exciting. A one line change to, I think pcie, at least it is to the pcie/dev.h file. The devs will know better.

There are a few changes:

  • Fixing RCU stall (still needs to be tested per issue 198 on the driver github)

  • Fixing a build issue with Kernel 4.4.x and OpenWRT (including a needed header file)

  • Fixing a comparison error

  • Updating the driver version to reflect these changes.

I agree though, I am not 100% sure on the impact this will have.

Tonight, I should be able to make a wifi driver, based on these changes, for Kernel 4.4 and 4.9.  Will also include other changes we've done for ourselves.

Great. Will give it a go when you make it available. Know it is easier for you to just do the driver rather then dealing with a whole new build for no reason other then new driver.

Hey Guys. Couple days ago I asked about decreasing txpower on wrt1900acv1 because the settings in /etc/config/wireless apparently have no effect on actual txpower, despite what the driver reports.

I found the reason why:

https://github.com/kaloz/mwlwifi/issues/70

"Power table is loaded according to setting in dts. Mwlwifi driver does not support specific tx power setting now."

So to decrease txpower, you would have to change the powertable in dts file for wrt1900ac found here:
https://github.com/lede-project/source/ … bles.patch

Regarding the amsdu issue.  Has this been highlighted tto yuhhaurlin on the mwlwifi github page?  Would be nice to know the developer thoughts on this

Will the patch be made public and posted here for those that compile their own?

Thanks

Phil-bkk wrote:

Regarding the amsdu issue.  Has this been highlighted tto yuhhaurlin on the mwlwifi github page?  Would be nice to know the developer thoughts on this

Will the patch be made public and posted here for those that compile their own?

Thanks

When the new driver is complete it will be posted and available for download.  All of the sources are already public and available for anyone to download and compile.

By the way... there have been issues getting the new driver compiled... I've run out of time, so I'll have to work on it some more tomorrow.

Phil-bkk wrote:

Regarding the amsdu issue.  Has this been highlighted tto yuhhaurlin on the mwlwifi github page?  Would be nice to know the developer thoughts on this

Will the patch be made public and posted here for those that compile their own?

Thanks

@yuhhaurlin said he will get back to fixing it after adding features to the 3200acm (88W8964).  The amsdu issue only affects the 88W8864 (1200/1900ac/acs models).

davidc502 wrote:

By the way... there have been issues getting the new driver compiled... I've run out of time, so I'll have to work on it some more tomorrow.

No rush!  Just to make sure, you will continue to include the tweak to keep amsdu disabled?

Thanks much as always!!

kyzhk wrote:

Hey Guys. Couple days ago I asked about decreasing txpower on wrt1900acv1 because the settings in /etc/config/wireless apparently have no effect on actual txpower, despite what the driver reports.

I found the reason why:

https://github.com/kaloz/mwlwifi/issues/70

"Power table is loaded according to setting in dts. Mwlwifi driver does not support specific tx power setting now."

So to decrease txpower, you would have to change the powertable in dts file for wrt1900ac found here:
https://github.com/lede-project/source/ … bles.patch

First, it's important to understand that on the 1200AC V2, 1900ACS V2, and 3200ACM, the power table is built into EEPROM so it is impossible to modify it.  On the other models (1200AC V1, 1900AC V1 and V2, 1900ACS V1) the power table is in firmware, exactly the file that you posted.  So it can be modified to basically anything.

Next, the power table you linked there allows power to be changed from 4dBm to a maximum of either 23dBm or 30dBm depending on the channel. 

On the 1200AC V1, 1900AC V1 and V2, 1900ACS V1, you can change the transmit power.  And you can change the country (if you look at that link for the power table, you will see a few different countries listed). I have tested it on my 1200AC V1. 

Addition/correction: However, after looking at that link for the power table, for the 1900AC V1 (mamba), the power table is restricted further and also only has 2 countries to choose from (FCC and ETSI).  This might be why you are having issues.  All the other models I just listed and that are included in that power table link let you choose from AU, CA, CN, ETSI, and FCC for the country.  So it seems the mamba is a special case (of the models that read the power table from firmware) I never realized before and doesn't allow changing transmit power.

On the 1200AC V2, 1900ACS V2, and the 3200ACM, you cannot change transmit power, country code, or anything relating to the regulatory laws and regulations.

Addition2:  However, since you have the 1900AC V1 (mamba) and can edit the power table since it is stored in firmware, search this thread for reghack2.  If you apply that (it needs to be applied everytime you update the firmware), you can change the power level as low as you want and as high as 30dBm on all channels, it disables DFS, and it enables wifi channels 12 and 13.

(Last edited by starcms on 11 Aug 2017, 04:41)

ok thanks for the info David.  I havent compiled for a few months as time does not permit

I see @starcms comments on github regarding the issue.  Work on the driver for the older devices is low priority. 

Its a bit wait and see...

@Phil-bkk, It is only a one liner, but a patch it is not too difficult. Just take the change from a commit someone has made, add .patch to the URL,

mkdir -f package/kernel/mwlwifi/patches
vi package/kernel/mwlwifi/patches/100-amsdu.patch

copy and paste patch into the file and rebuild mwlwifi package.

Just be sure to keep an eye on further mwlwifi commits as a change may break things.

Villeneuve wrote:

@Phil-bkk, It is only a one liner, but a patch it is not too difficult. Just take the change from a commit someone has made, add .patch to the URL,

mkdir -f package/kernel/mwlwifi/patches
vi package/kernel/mwlwifi/patches/100-amsdu.patch

copy and paste patch into the file and rebuild mwlwifi package.

Just be sure to keep an eye on further mwlwifi commits as a change may break things.

Good job making that patch out of the info to disable amsdu!  Should make things easier for @david.

I'm assuming // is the comment symbol for kernel patches as opposed to #? 

However, I have no idea how to build mwlwifi; I'll wait for @david to release the updated 0810 driver with amsdu disabled.

(Last edited by starcms on 11 Aug 2017, 04:25)

It is just a comment in C, allowed at the beginning of a line. I did open a PR for the latest mwlwifi , but things seems to have slowed as of late, EU on August break I assume.

Thanks starcms & Villenueve for the effort and the patch

This thread has become my No. 1 resource for all things Linksys WRT, so thanks to davidc502 also

Thanks starcms.

I looked at the source code for reghack (http://luci.subsignal.org/~jow/reghack/) and looks like I only need to change these two lines replacing 30 to 4:

    struct ieee80211_reg_rule r2 = REG_RULE(2400, 2483, 40, 0, 30, 0);
    struct ieee80211_reg_rule r5 = REG_RULE(5140, 5860, 160, 0, 30, 0);

and recompile an arm version to patch the powertable without recompiling cfg80211.ko