OpenWrt Forum Archive

Topic: Update on Linksys WRT1900AC support

The content of this topic has been archived between 16 Sep 2014 and 7 May 2018. Unfortunately there are posts – most likely complete pages – missing.

JohnnySL wrote:
gufus wrote:
davidc502 wrote:

That's too funny.... doesn't surprise me.

I don't think it's funny. it's BS

Both openwrt and dd-wrt have started developing firmware also.

I'm curious, what exactly do you think is BS. That the vendor does this, or that Giraffe states this? Not really following you on the firmware development part either. Could you explain more what you've meant here?

I'm talking about the vendor.

This has been going on for a long time sad
https://forum.openwrt.org/viewtopic.php?id=50173

The fact that they are touting full OpenWRT CC support is imho BS, as it maybe "work", but is far from completely stable yet.

Hopefully the new commits that have been done today will change that

JohnnySL wrote:

Hopefully the new commits that have been done today will change that

Yep.

JohnnySL wrote:

BTW: what is the upgradepath to trunk from CC? i've read that it would be best to flash back to stock and then forward to trunk again. Has this changed already, can we directly upgrade now?

It's not a requirement, however it does help prevent potential issues.  With that being said, I've only had to flash back to stock 3 or four times prior to flashing trunk.

JW0914 wrote:

It's not a requirement, however it does help prevent potential issues.

Agreed.

northbound wrote:
lifehacksback wrote:

Anyone willing to test the new firmware? I'm building an image with the new wifi driver. I will test it (very limited it's 3am pst and i have a test tomorrow) with gufus's deamon fan ctrl (I will manually add this it wont come with the img). Let's see how well it works.


I'm willing to test it.

Do I need to build it for testing?

i.trankolov wrote:
northbound wrote:
lifehacksback wrote:

Anyone willing to test the new firmware? I'm building an image with the new wifi driver. I will test it (very limited it's 3am pst and i have a test tomorrow) with gufus's deamon fan ctrl (I will manually add this it wont come with the img). Let's see how well it works.


I'm willing to test it.

Do I need to build it for testing?

It has not been pushed to trunk yet so if you build it I will try it I am curious if just replacing the mwlwifi.ko will fly?
Of course it will still have to be built.

northbound wrote:
i.trankolov wrote:
northbound wrote:

I'm willing to test it.

Do I need to build it for testing?

It has not been pushed to trunk yet so if you build it I will try it I am curious if just replacing the mwlwifi.ko will fly?
Of course it will still have to be built.

Yeah .. the building is my issue. I was kind of asking if someone has already compiled it. smile

i.trankolov wrote:
northbound wrote:
i.trankolov wrote:

Do I need to build it for testing?

It has not been pushed to trunk yet so if you build it I will try it I am curious if just replacing the mwlwifi.ko will fly?
Of course it will still have to be built.

Yeah .. the building is my issue. I was kind of asking if someone has already compiled it. smile

smile Understood I don't have a nix box but am tempted to fire up one of my old computers load unbunto and learn how to build my own images.

northbound wrote:

smile Understood I don't have a nix box but am tempted to fire up one of my old computers load unbunto and learn how to build my own images.

You could try vmware's vmware player which is free and surprisingly fast (graphics too). Or oracle's virtual box. But nothing (that isnt server grade or enthusiast) beats bare metal install on a quad-core intel w/ hyper-threading. I'd recommend ubuntu lts or 15.04 (tho .10 is coming next week) server since building doesn't require graphical user interface just cli.

(Last edited by lifehacksback on 16 Oct 2015, 16:06)

i.trankolov wrote:

Yeah .. the building is my issue. I was kind of asking if someone has already compiled it. smile

I took the plunge.  I was reluctant to do so because my running build has been so stable (both operational and thermal).  But I have some free time today and figured I'd at least take a look.

strings /lib/modules/3.18.21/mwlwifi.ko | grep 10.3 now returns:

mwlwifi-10.3.0.10-20151016/mac80211.c
mwlwifi-10.3.0.10-20151016/fwcmd.c
mwlwifi-10.3.0.10-20151016/tx.c

I'm a bit frustrated that I have not seen a router reboot in the last few weeks but have gone through 3 power outages.

Sample of 1 but just last night I did a large file transfer over wireless and was getting ~22MB/s.  Just did another and was getting ~28MB/s.

InkblotAdmirer wrote:
i.trankolov wrote:

Yeah .. the building is my issue. I was kind of asking if someone has already compiled it. smile

I took the plunge.  I was reluctant to do so because my running build has been so stable (both operational and thermal).  But I have some free time today and figured I'd at least take a look.

strings /lib/modules/3.18.21/mwlwifi.ko | grep 10.3 now returns:

mwlwifi-10.3.0.10-20151016/mac80211.c
mwlwifi-10.3.0.10-20151016/fwcmd.c
mwlwifi-10.3.0.10-20151016/tx.c

I'm a bit frustrated that I have not seen a router reboot in the last few weeks but have gone through 3 power outages.

Sample of 1 but just last night I did a large file transfer over wireless and was getting ~22MB/s.  Just did another and was getting ~28MB/s.


Any chance of you posting the mwlwifi.ko?

@lifehacksback

Thanks for the info much appreciated.

northbound wrote:

Any chance of you posting the mwlwifi.ko?

+1, my wifi is far from stable at the moment, so would love to test this on a WRT1900AC-V2

lifehacksback wrote:
plamka00 wrote:
plamka00 wrote:

push

hey guys whats up? why nobody gives me an answer?

Not an atheros device


OK, yes its an Marvell device, but anyway is there a possibility to set the router up in repeater mode? And if, than how?

Thanks

InkblotAdmirer wrote:
i.trankolov wrote:

Yeah .. the building is my issue. I was kind of asking if someone has already compiled it. smile

I took the plunge.  I was reluctant to do so because my running build has been so stable (both operational and thermal).  But I have some free time today and figured I'd at least take a look.

strings /lib/modules/3.18.21/mwlwifi.ko | grep 10.3 now returns:

mwlwifi-10.3.0.10-20151016/mac80211.c
mwlwifi-10.3.0.10-20151016/fwcmd.c
mwlwifi-10.3.0.10-20151016/tx.c

I'm a bit frustrated that I have not seen a router reboot in the last few weeks but have gone through 3 power outages.

Sample of 1 but just last night I did a large file transfer over wireless and was getting ~22MB/s.  Just did another and was getting ~28MB/s.

After I figured out that I MUST not change the 5Ghz settings, I can transfer huge files (30GB) over wireless with speeds of up to 50MB/s. That of course after I managed to keep the CPU and WIFI below 58C.

@InkblotAdmirer

Would you mind posting the bin file from /lib/firmware/mwlwifi

Does anyone know if he spinlock was related to Apple devices. Should this fix Apple device problem also?

InkblotAdmirer wrote:

strings /lib/modules/3.18.21/mwlwifi.ko | grep 10.3 now returns:

mwlwifi-10.3.0.10-20151016/mac80211.c
mwlwifi-10.3.0.10-20151016/fwcmd.c
mwlwifi-10.3.0.10-20151016/tx.c

Nice smile

Now we wait and see eh.

gufus wrote:
InkblotAdmirer wrote:

strings /lib/modules/3.18.21/mwlwifi.ko | grep 10.3 now returns:

mwlwifi-10.3.0.10-20151016/mac80211.c
mwlwifi-10.3.0.10-20151016/fwcmd.c
mwlwifi-10.3.0.10-20151016/tx.c

Nice smile

Now we wait and see eh.

Can someone guide me how to compile the new driver?

I've been debugging a bit why the V2 keeps a maximum speed and doesn't throttle down based on the governer.

found this in the start of the kernel logs (had to hack around a bit as the kernel is currently compiled with to little memory to store every line straigh from the start):

<snip>
[    0.149944] cpuidle: using governor ladder
[    0.151872] mvebu-pmsu: CPU hotplug support is currently broken on Armada 38x                                                                                              : disabling
[    0.151880] mvebu-pmsu: CPU idle is currently broken on Armada 38x: disabling
<6>[    0.180322] reg-fixed-voltage xhci0-vbus: could not find pctldev for node /soc/internal-regs/pinctrl@18000/xhci0-vbus-pins, deferring probe
</snip>

so i guess that bold part means that it doesn't thottle and keeps running full speed? fixing that would make it run a lot cooler and more power efficient I'd think?

(Last edited by JohnnySL on 16 Oct 2015, 23:07)

i.trankolov wrote:
gufus wrote:
InkblotAdmirer wrote:

strings /lib/modules/3.18.21/mwlwifi.ko | grep 10.3 now returns:

mwlwifi-10.3.0.10-20151016/mac80211.c
mwlwifi-10.3.0.10-20151016/fwcmd.c
mwlwifi-10.3.0.10-20151016/tx.c

Nice smile

Now we wait and see eh.

Can someone guide me how to compile the new driver?

I'm trying to compile it now but am running into compile errors.

make[4]: Entering directory '/home/ctalbot/openwrt/build_dir/target-arm_cortex-a9+vfpv3_musl-1.1.11_eabi/linux-mvebu/linux-3.18.21'
  CC [M]  /home/ctalbot/openwrt/build_dir/target-arm_cortex-a9+vfpv3_musl-1.1.11_eabi/linux-mvebu/mwlwifi-10.3.0.10-20151016/main.o
/home/ctalbot/openwrt/build_dir/target-arm_cortex-a9+vfpv3_musl-1.1.11_eabi/linux-mvebu/mwlwifi-10.3.0.10-20151016/main.c: In function 'mwl_set_ht_caps':
/home/ctalbot/openwrt/build_dir/target-arm_cortex-a9+vfpv3_musl-1.1.11_eabi/linux-mvebu/mwlwifi-10.3.0.10-20151016/main.c:398:12: error: invalid operands to binary | (have 'long unsigned int[1]' and 'int')
  hw->flags |= IEEE80211_HW_AMPDU_AGGREGATION;
            ^
/home/ctalbot/openwrt/build_dir/target-arm_cortex-a9+vfpv3_musl-1.1.11_eabi/linux-mvebu/mwlwifi-10.3.0.10-20151016/main.c: In function 'mwl_wl_init':
/home/ctalbot/openwrt/build_dir/target-arm_cortex-a9+vfpv3_musl-1.1.11_eabi/linux-mvebu/mwlwifi-10.3.0.10-20151016/main.c:504:12: error: invalid operands to binary | (have 'long unsigned int[1]' and 'int')
  hw->flags |= IEEE80211_HW_SIGNAL_DBM | IEEE80211_HW_HAS_RATE_CONTROL;
            ^
/home/ctalbot/openwrt/build_dir/target-arm_cortex-a9+vfpv3_musl-1.1.11_eabi/linux-mvebu/mwlwifi-10.3.0.10-20151016/main.c:514:12: error: invalid operands to binary | (have 'long unsigned int[1]' and 'int')
  hw->flags |= IEEE80211_HW_AP_LINK_PS;
            ^

It appears gcc does not like the different types when using a bitwise operator. The first error is with the line below with hw->flags

#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 2, 0)
        hw->flags |= IEEE80211_HW_AMPDU_AGGREGATION;
#else
        ieee80211_hw_set(hw, AMPDU_AGGREGATION);
#endif
i.trankolov wrote:

Can someone guide me how to compile the new driver?

Here's what I did.  In package/kernel/mwlwifi change the appropriate lines in Makefile:

PKG_VERSION:=10.3.0.10-20151016
PKG_SOURCE_VERSION:=7a6aff718d2de7dee6f9d82eae8af91667e8ee86

Then

make package/mwlwifi/{clean,compile}
make target/linux/clean
make V=s

If you're sure it's going to work I think you can probably just clean Linux and make, however I wanted to be sure the driver was going to compile first.

I'd rather not be posting binaries around just yet - if I'm doing something wrong or incomplete I don't want garbage floating around.

I will say that CPU0 is cheating toward 60% full-time now, all soft-IRQ related.

@alirz, you can get the bin files from here:  https://github.com/kaloz/mwlwifi

Just download the zip and extract.

EDIT:  don't capitalize linux...  (the editor here is auto-capping...)

(Last edited by InkblotAdmirer on 17 Oct 2015, 00:09)

@InkblotAdmirer

Still it would be nice to play with. smile Changing affinity might help.  The Trunk  buildbot has not had a successful build since the 8th.
And still no activity from openwrt-commits@lists.openwrt.org mailing list.

Thank you Kaloz!

Best Regards,