151 (edited by subterfugium 2015-11-29 17:24:18)

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

mastabog wrote:

Thanks for the test. It doesn't appear that anything was improved. No crypto wouldn't help. It feels like the ath10k driver in openwrt doesn't do MPDU aggregation, or that it aggregates fewer frames ... but it could be something else.

Are there similar differences on Broadcom chips? Did anyone test native firmwares that give 800+ Mbps on Broadcom wifi chips versus OpenWRT?

@mastabog, have you been able to test latest openwrt with iperf LAN<-> 5GHz WLAN?

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

I have had to revert back to 15.05 as I updated to yesterdays trunk, installed luci, ath10k, etc with opkg and when I rebooted it lost most of my settings and refused to save any new ones.  I had to wipe the jffs partition and restore a backup to get my settings back.

I also noticed that while the Note 4 dropped the connection less on trunk, it seemed to spend longer trying to actually get any data flowing.  eg It took way longer for YouTube to start playing or web pages to load.  So I think I will wait for the next stable release.

ISP: Zen Internet (74Mbit/16Mbit) Steam/XBOX: Alex Atkin UK PSN/WiiU: AlexAtkinUK Website: http://csdprojects.co.uk
Router (pfSense): Intel DN2800MT WiFi (OpenWRT): TP-Link Archer C7 v2, TP-Link WDR3600 x2, Buffalo WZR-HP-G300NH x2.

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

Alex Atkin UK wrote:

I also noticed that while the Note 4 dropped the connection less on trunk, it seemed to spend longer trying to actually get any data flowing.  eg It took way longer for YouTube to start playing or web pages to load.  So I think I will wait for the next stable release.

I have noticed this on my Nexus 4 as well... it usually eventually works, but the wifi seems sluggish only on this device.  Everything else works absolutely (all Apple household) perfectly these days (trunk from 3 days ago or so), even without having to manually update the ath10k driver.

154 (edited by Alex Atkin UK 2015-11-29 22:39:11)

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

emmeff wrote:
Alex Atkin UK wrote:

I also noticed that while the Note 4 dropped the connection less on trunk, it seemed to spend longer trying to actually get any data flowing.  eg It took way longer for YouTube to start playing or web pages to load.  So I think I will wait for the next stable release.

I have noticed this on my Nexus 4 as well... it usually eventually works, but the wifi seems sluggish only on this device.  Everything else works absolutely (all Apple household) perfectly these days (trunk from 3 days ago or so), even without having to manually update the ath10k driver.

That's the thing though, on 15.05 my Note 4 is the only device to have stability problems.  My laptop does up to 45MBytes/s from my NAS and that reduced to 35MBytes/s on trunk.

Also while the Note 4 drops (it pops up the connected to SSID message), it doesn't usually last long enough to disrupt what I'm doing.

I'm tending to blame the phone as it was dropping much less on 15.05 after one update and then became less stable again after another, with no changes to the Archer at all.

ISP: Zen Internet (74Mbit/16Mbit) Steam/XBOX: Alex Atkin UK PSN/WiiU: AlexAtkinUK Website: http://csdprojects.co.uk
Router (pfSense): Intel DN2800MT WiFi (OpenWRT): TP-Link Archer C7 v2, TP-Link WDR3600 x2, Buffalo WZR-HP-G300NH x2.

155 (edited by mastabog 2016-01-07 04:26:18)

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

subterfugium wrote:
mastabog wrote:

Thanks for the test. It doesn't appear that anything was improved. No crypto wouldn't help. It feels like the ath10k driver in openwrt doesn't do MPDU aggregation, or that it aggregates fewer frames ... but it could be something else.

Are there similar differences on Broadcom chips? Did anyone test native firmwares that give 800+ Mbps on Broadcom wifi chips versus OpenWRT?

@mastabog, have you been able to test latest openwrt with iperf LAN<-> 5GHz WLAN?

No, I've been swamped and was hoping someone else could try - I documented everything in this thread's 1st post.

Also, can we please keep this thread about WLAN to LAN only? This is not a thread about general Archer C7 things. This thread has nothing to do with NAT, WAN, flashing issues/commands etc. Please create new threads for those!

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

Can i ask something? I use arokhs build on a archer c5/7 -> https://forum.openwrt.org/viewtopic.php?id=50914

My 5Ghz wont start after a restart if channel is setup to auto. Im not sure what you need. Pls give me an advice. Im not ne CLI Noob but im not so familar with all that deamon/configstuff in Linux.

Its setup to AC/Chanel Auto/80Mhz and Germany - If its setup to chanel 36 or so boot is fine.

root@OpenWrt ~# iw wlan0 info
Interface wlan0
        ifindex 20
        wdev 0x2
        addr e8:de:27:6f:f4:64
        type AP
        wiphy 0
        txpower 0.00 dBm
root@OpenWrt ~#
root@OpenWrt ~# logread | grep wlan0
Fri Jan  8 16:24:22 2016 kern.info kernel: [   35.900374] IPv6: ADDRCONF(NETDEV_UP): wlan0: link is not ready
Fri Jan  8 16:24:22 2016 kern.info kernel: [   36.168130] device wlan0 entered promiscuous mode
Fri Jan  8 16:24:22 2016 daemon.notice netifd: radio0 (1696): wlan0: interface state UNINITIALIZED->COUNTRY_UPDATE
Fri Jan  8 16:24:22 2016 daemon.notice netifd: radio0 (1696): wlan0: interface state COUNTRY_UPDATE->ACS
Fri Jan  8 16:24:22 2016 daemon.notice netifd: radio0 (1696): wlan0: ACS-STARTED
Fri Jan  8 16:25:55 2016 daemon.notice netifd: radio0 (1696): wlan0: ACS-COMPLETED freq=5580 channel=116
Fri Jan  8 16:25:55 2016 daemon.notice netifd: radio0 (1696): wlan0: interface state ACS->HT_SCAN
Fri Jan  8 16:25:55 2016 daemon.notice netifd: radio0 (1696): wlan0: interface state HT_SCAN->DFS
Fri Jan  8 16:25:55 2016 daemon.notice netifd: radio0 (1696): wlan0: DFS-CAC-START freq=5580 chan=116 sec_chan=1, width=1, seg0=122, seg1=0, cac_time=60s
Fri Jan  8 16:25:55 2016 daemon.notice netifd: radio0 (1696): wlan0: interface state DFS->DISABLED
Fri Jan  8 16:25:55 2016 daemon.notice netifd: radio0 (1696): wlan0: AP-DISABLED
Fri Jan  8 16:25:55 2016 daemon.notice netifd: radio0 (1696): wlan0: interface state DISABLED->DISABLED
Fri Jan  8 16:25:55 2016 daemon.notice netifd: radio0 (1696): wlan0: AP-DISABLED
Fri Jan  8 16:25:55 2016 daemon.notice netifd: radio0 (1696): hostapd_free_hapd_data: Interface wlan0 wasn't started
Fri Jan  8 16:25:55 2016 daemon.notice netifd: radio0 (1696): nl80211: deinit ifname=wlan0 disabled_11b_rates=0
Fri Jan  8 16:25:55 2016 kern.info kernel: [   51.535462] device wlan0 left promiscuous mode
Fri Jan  8 16:25:55 2016 kern.info kernel: [   51.540063] br-lan: port 2(wlan0) entered disabled state
Fri Jan  8 16:25:56 2016 kern.info kernel: [   52.896067] IPv6: ADDRCONF(NETDEV_UP): wlan0: link is not ready
Fri Jan  8 16:25:56 2016 kern.info kernel: [   52.951126] device wlan0 entered promiscuous mode

157 (edited by mastabog 2016-01-08 21:49:46)

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

@Hexxer, please create a new thread. This thread is about a different issue, as it says in its title.

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

I installed 15.05 just one day after I bought the router and I didn't feel it slower but I could see it was less stable than the stock firmware (sometimes the WiFi connection gets lost and it reconnects automatically after some seconds).

Did anyone else experience the same? Do you guys recommend previous openwrt builds or 15.05 for this router?

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

juanmartinbravo wrote:

I installed 15.05 just one day after I bought the router and I didn't feel it slower but I could see it was less stable than the stock firmware (sometimes the WiFi connection gets lost and it reconnects automatically after some seconds).


I have tested file transfer over LAN and notice that the client reconnects from the AP automatically. This only happens when the client is using lots of bandwidth. When the client is idle it stays connected.

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

heavymetal wrote:
juanmartinbravo wrote:

I installed 15.05 just one day after I bought the router and I didn't feel it slower but I could see it was less stable than the stock firmware (sometimes the WiFi connection gets lost and it reconnects automatically after some seconds).


I have tested file transfer over LAN and notice that the client reconnects from the AP automatically. This only happens when the client is using lots of bandwidth. When the client is idle it stays connected.

You are right but that doesn't happens with stock firmware.

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

yep, speed up come from HW accelleration which is visible at the stock tp-link linux kernel 2.6.31. / available for free download at tp-link site as GPL free code /


In the stock linux kernel ( 2.6.31),besides the accelleration for Hardware NAT there is also HW acceleration with DMA support for:
- HW fragmentation and defragmentation
- HW csum
and
- ATH_HOTSPOT and ATH_PERF - which are optimizing the ATH bridge support



I hope tha someone could build an openwrt firmware of Archer C7 v2 with "stock" GPL tp-link kernel, so that the HW accelleration engines could be used on Archer C7 / max speed / and keep the openwrt community user level packages to bring more flexability / for instance use of openvpn which is not available in the stock firmware /.

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

Alex Atkin UK wrote:

...
I have just updated to trunk with firmware-5.bin_10.2.4.70.12-2 and am really regretting it as now I'm getting half the speed I was on 15.05 with firmware-4.bin_10.2.4.48. The only reason I updated is that after the last firmware update on my Note 4, its become unstable again. sad

Interestingly, if I remove firmware-5 and put firmware-4.bin_10.2.4.48 in there instead (named firmware-4.bin), the performance is the same.
So it seems its probably not a firmware bug but something specific about OpenWRT trunk itself, perhaps some extra debugging enabled in the ath10k driver?
...

I'm facing similar problems and tried several times to update the ath10k firmware. I'm still having problems to understand which versions are compatible.

Chaos Calmer 15.05 (stable):
1) Can I use firmware-5 or only the latest firmware-4?
2) If yes, do I have to rename the firmware-5 to firmware-4?
3) Do I also need to replace the board.bin?
4) Is there an easy way to update the firmware directly. I always use wget to download to my personal folder and then copy to the firmware folder.

163

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

todorcolov wrote:

yep, speed up come from HW accelleration which is visible at the stock tp-link linux kernel 2.6.31. / available for free download at tp-link site as GPL free code /

In the stock linux kernel ( 2.6.31),besides the accelleration for Hardware NAT there is also HW acceleration with DMA support for:
- HW fragmentation and defragmentation
- HW csum
and
- ATH_HOTSPOT and ATH_PERF - which are optimizing the ATH bridge support

I hope tha someone could build an openwrt firmware of Archer C7 v2 with "stock" GPL tp-link kernel, so that the HW accelleration engines could be used on Archer C7 / max speed / and keep the openwrt community user level packages to bring more flexability / for instance use of openvpn which is not available in the stock firmware /.

Near as I can tell, OpenWRT hasn't used the 2.6.X kernel for ~5 years. I doubt anyone will be be getting openwrt going on the tp-link kernel.

164 (edited by Alex Atkin UK 2016-02-13 03:50:06)

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

tim2000 wrote:
Alex Atkin UK wrote:

...
I have just updated to trunk with firmware-5.bin_10.2.4.70.12-2 and am really regretting it as now I'm getting half the speed I was on 15.05 with firmware-4.bin_10.2.4.48. The only reason I updated is that after the last firmware update on my Note 4, its become unstable again. sad

Interestingly, if I remove firmware-5 and put firmware-4.bin_10.2.4.48 in there instead (named firmware-4.bin), the performance is the same.
So it seems its probably not a firmware bug but something specific about OpenWRT trunk itself, perhaps some extra debugging enabled in the ath10k driver?
...

I'm facing similar problems and tried several times to update the ath10k firmware. I'm still having problems to understand which versions are compatible.

Chaos Calmer 15.05 (stable):
1) Can I use firmware-5 or only the latest firmware-4?
2) If yes, do I have to rename the firmware-5 to firmware-4?
3) Do I also need to replace the board.bin?
4) Is there an easy way to update the firmware directly. I always use wget to download to my personal folder and then copy to the firmware folder.

I believe if you put firmware-5 in the correct directory then it will be used if the ath10k package is new enough to support it and ignored if it isn't.

Renaming it to firmware-4 will not work as the numbers specifically refer to changes in the ath10k firmware API.

Are people trying to fix any particular issues at this point?  The only device I have any issues with is my Note 4 which has unstable periods I am convinced is device related.

I also now have the FireTV 4k and it streams 4k without any problems, my laptop can do 40MB/s from my NAS and since switching ISPs I get full speed from my broadband too.  Performance seems good enough for now.

ISP: Zen Internet (74Mbit/16Mbit) Steam/XBOX: Alex Atkin UK PSN/WiiU: AlexAtkinUK Website: http://csdprojects.co.uk
Router (pfSense): Intel DN2800MT WiFi (OpenWRT): TP-Link Archer C7 v2, TP-Link WDR3600 x2, Buffalo WZR-HP-G300NH x2.

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

Alex Atkin UK wrote:

Are people trying to fix any particular issues at this point?  The only device I have any issues with is my Note 4 which has unstable periods I am convinced is device related.

I also now have the FireTV 4k and it streams 4k without any problems, my laptop can do 40MB/s from my NAS and since switching ISPs I get full speed from my broadband too.  Performance seems good enough for now.

I got the impression that since this router isn't the latest and greatest anymore, it wasn't getting much attention. With that said, my throughput lately is very comparable to that of the stock firmware.

Now if only I could get IPSec passthrough working... but that's a topic for another thread.

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

emmeff wrote:
Alex Atkin UK wrote:

Are people trying to fix any particular issues at this point?  The only device I have any issues with is my Note 4 which has unstable periods I am convinced is device related.

I also now have the FireTV 4k and it streams 4k without any problems, my laptop can do 40MB/s from my NAS and since switching ISPs I get full speed from my broadband too.  Performance seems good enough for now.

I got the impression that since this router isn't the latest and greatest anymore, it wasn't getting much attention. With that said, my throughput lately is very comparable to that of the stock firmware.

Now if only I could get IPSec passthrough working... but that's a topic for another thread.

I don't think that is really relevant to OpenWRT as no single router gets focus and they just slowly bring in code from upstream package updates.

Atheros are still updating the firmware and the developers of Ath10k are still working, so ultimately the WiFi performance can still improve.  This is naturally a lot slower on OpenWRT than stock Linux as they are focused on stability over bleeding-edge software/kernels.  But I don't think there is any reason to think things cannot improve further.

ISP: Zen Internet (74Mbit/16Mbit) Steam/XBOX: Alex Atkin UK PSN/WiiU: AlexAtkinUK Website: http://csdprojects.co.uk
Router (pfSense): Intel DN2800MT WiFi (OpenWRT): TP-Link Archer C7 v2, TP-Link WDR3600 x2, Buffalo WZR-HP-G300NH x2.

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

Alex Atkin UK wrote:

...

I believe if you put firmware-5 in the correct directory then it will be used if the ath10k package is new enough to support it and ignored if it isn't.

Renaming it to firmware-4 will not work as the numbers specifically refer to changes in the ath10k firmware API.

Are people trying to fix any particular issues at this point?  The only device I have any issues with is my Note 4 which has unstable periods I am convinced is device related.

I also now have the FireTV 4k and it streams 4k without any problems, my laptop can do 40MB/s from my NAS and since switching ISPs I get full speed from my broadband too.  Performance seems good enough for now.

Just in case someone else has the same question. I just tried several things. On CC (stable release) firmware-5.bin cannot be configured via LuCI. If renamed to firmware-4.bin, it can be configured, but I was not able to connect with any device. The latest firmware-4.bin_10.2.4.48 (named to firmware-4.bin) however, works fine so far.

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

I was under impression that firmware-5 files require the newer kernel, so they are incompatible with CC.

169 (edited by Alex Atkin UK 2016-02-16 06:10:40)

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

stangri wrote:

I was under impression that firmware-5 files require the newer kernel, so they are incompatible with CC.

Correct, although I believe OpenWRT do backport some kernel stuff.

If I recall correctly, firmware-4 was supported in an earlier kernel than the ath10k website claimed it was used on.

ISP: Zen Internet (74Mbit/16Mbit) Steam/XBOX: Alex Atkin UK PSN/WiiU: AlexAtkinUK Website: http://csdprojects.co.uk
Router (pfSense): Intel DN2800MT WiFi (OpenWRT): TP-Link Archer C7 v2, TP-Link WDR3600 x2, Buffalo WZR-HP-G300NH x2.

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

stangri wrote:

I was under impression that firmware-5 files require the newer kernel, so they are incompatible with CC.

A few days ago, new ath10k code was added to the 15.05 branch which now defaults to firmware-5.bin, or – to be precise – firmware-5.bin_10.2.4.97-1. The kernel for CC was updated to 3.18.27.

On the weekend, I compiled a new snapshot and can confirm, it actually uses the newer firmware. So far all my clients work well with the new firmware, but I didn't have any issues before either, so it's hard to tell a difference.

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

Great news! I hope this is a step towards the 15.05.1 release.

So the firmware is now a separate package on CC, just like it is on trunk?

172 (edited by silentcreek 2016-03-07 21:08:36)

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

stangri wrote:

So the firmware is now a separate package on CC, just like it is on trunk?

Yes. In my snapshot from early February the firmware files were included in the package kmod-ath10k, whereas now with the new build from last Saturday, I have a package called ath10k-firmware-qca988x which contains the actual firmware files.

173 (edited by Alex Atkin UK 2016-03-07 22:50:54)

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

I'm still having major compatibility issues with my Galaxy Note 4 on OpenWRT, but it seems to apply to ALL OpenWRT devices. sad

The weird thing is it seems to work fine on the LAN but has issues with web related activity.  I can't fathom why it would make any difference but I'm hoping its something that will be fixed in the next Android update as it seems to start with the previous update, so probably not OpenWRT related at all.  Its just weird that it only affects OpenWRT APs.

Example, I load the LIFX app and it detects my bulbs instantly but takes about 15 seconds to display my Scenes as those require accessing the website.  Connect my phone to a none OpenWRT AP and the scenes appear instantly.  Same with web pages, the initial load takes many times longer when connected to OpenWRT and I have no idea why.  Actual throughput is fine, once it gets going, but some apps can timeout.

WiFi does still sometimes drop as well, I think THAT is the only bit where ath10k is to blame.

ISP: Zen Internet (74Mbit/16Mbit) Steam/XBOX: Alex Atkin UK PSN/WiiU: AlexAtkinUK Website: http://csdprojects.co.uk
Router (pfSense): Intel DN2800MT WiFi (OpenWRT): TP-Link Archer C7 v2, TP-Link WDR3600 x2, Buffalo WZR-HP-G300NH x2.

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

silentcreek wrote:

On the weekend, I compiled a new snapshot and can confirm, it actually uses the newer firmware. So far all my clients work well with the new firmware, but I didn't have any issues before either, so it's hard to tell a difference.

Well, it seems the changes have introduced a new issue. And it's kinda odd:
When I do a cold boot (power is off before booting), everything works. When I do a simple reboot, the ath10k firmware or module won't load anymore. I tested this repeadtedly and it all leads to the same result: cold boot - ok, reboot - not ok.
I noticed it because I have a cronjob to reboot the router once a week and after the reboot the 5GHz wifi was gone.

I created a ticket here and hope it get's resolved eventually.

Does anybody know what might be the difference between a soft reboot and a cold boot that might explain this issue?

Re: TP-Link Archer C7: 802.11ac slower in OpenWRT than native firmware

Timo, I can't replicate that with DD r48973. Are you running a newer version? I might very well be quite a few revisions behind.