BitTorrent uses both TCP and UDP. Here is an example of a UDP congestion control that bit torrent uses and is probably the source of your thousands of connections: http://www.bittorrent.org/beps/bep_0029.html
BitTorrent uses tons of connections and can eat up a good chunk of bandwidth too.
Not too worried about that number of connections with BitTorrent. Here is my morning’s stats with just some light browsing:
16k connections is not normal, i have limit to max 50 seeds for one torrent and 500 for all of them, no problem for rtorrent, but when i downloading by synology download station it generates couple of more time load of cpu in ,y pfsense router with little more tweaked firewall list
All looks correct to me, I followed that wiki page to set up the AP. Ignore DHCP is ticked and firewall, dnsmasq and odhcpd are all disabled at startup and not running.
Also, I looked at the system logs at the time when the connection wasn't working and here's the output:
Fri Jan 1 16:50:50 2021 daemon.err hostapd: nl80211: kernel reports: key addition failed
Fri Jan 1 16:50:50 2021 daemon.info hostapd: wlan0: STA 9a:ef:44:2e:13:28 IEEE 802.11: associated (aid 1)
Fri Jan 1 16:50:52 2021 daemon.err dawn[2855]: Neigbor-Report is NULL!
Fri Jan 1 16:50:52 2021 daemon.err dawn[2855]: Neigbor-Report is NULL!
Fri Jan 1 16:51:32 2021 daemon.err dawn[2855]: Neigbor-Report is NULL!
Fri Jan 1 16:52:14 2021 daemon.err hostapd: nl80211: kernel reports: key addition failed
Fri Jan 1 16:52:14 2021 daemon.info hostapd: wlan0: STA 9a:ef:44:2e:13:28 IEEE 802.11: associated (aid 1)
Fri Jan 1 16:52:16 2021 daemon.err dawn[2855]: Neigbor-Report is NULL!
Fri Jan 1 16:52:31 2021 daemon.err dawn[2855]: Neigbor-Report is NULL!
Fri Jan 1 16:52:31 2021 daemon.err dawn[2855]: Neigbor-Report is NULL!
Related at all do you think?
I’d remove the dawn and dawn luci package. If it works - the dawn network controller may be causing issues for your clients.
got it
will test it asap and let you know
thanks
I'm familiar with that, but it's explicity set to TCP only: 
, and as I mentioned earlier on familiar setup (openwrt 12.09 + qcom sdk), I didn't had conntrack table full warning in kernel log (but it had 32k max. net.nf_conntrack_max -- just checked now). Torrents causing conntrack table to overflow
Anyway it's pretty stable on C2600, haven't run into a single issue (which related to NSS / build). Just flashed C2600-20201231 @ yesterday, no issues.
btw:
/sys/kernel/debug/qca-nss-drv/stats/pppoe
RX_BYTES RX_PACKETS + RX_BYTES always zero, is that normal?
Is there a way to get the NSS CPU cores usage?
p.s.: did u revert openwrt's branch?
C2600-20201228 was r15430+1-774b4f318f
C2600-20201231 is r15366+83-dbb542f194
Dunno why the stats are zero - that is interesting. Haven’t seen a way to measure NSS cores usage. I’d need some help to figure that out.
The version gets a little wonky when you have add on commits. Versions are managed by the get version script - getver.sh. There are 67 commits (the rebase commit is never counted) so here is the reasonings for the version changes:
C2600-20201228 was r15430+1 = “15431” (I made 1 local commit “+1”, no commits from master in the timeframe from the build the prior day so I didn’t rebase with master- the get version script gets tricked and thinks that 65 of my branch’s commits are a part of master)
C2600-20201231 is r15366+83 = “15449” (this is a more common nomenclature, it has +66 of my branch’s commits and +17 commits from the last time I rebased with master on Dec 28th - “+83”)
So “15449” is the newer build. 
Thanks, gave that a try as well but seems to have failed as well.
Seems my iDevices will never be happy! Appreciate all your help 
Just flash your firmware of EA7500 V1. Booting is ok !
I'm not sure how to verify nss driver is working or not.need research previous posts
thank you very much for your great work!
I also will test r7500 v2 later and let you know
With NSS you should see full line speed (940mbps) and minimal CPU usage (if you have gig speed from your ISP). Wifi should have a small improvement too.
Enjoy!
I'm trying to build by myself from your repo
I have built it ,now I want to update to the latest openwrt master
So I do
1 git stash
2 git fetch upstream && git rebase upstream/maste
but I got following,I'm a new openwrt, please help me
thanks
Applying: Fix NSS for NBG6817 (Zyxel Armor Z2)
Applying: Fix NSS for C2600
Applying: Fix NSS for EA7500 v1
Applying: rebase
Applying: Add qca package for qsdk10.0
Using index info to reconstruct a base tree...
.git/rebase-apply/patch:109: trailing whitespace.
.git/rebase-apply/patch:116: trailing whitespace.
.git/rebase-apply/patch:118: space before tab in indent.
crypto_bench_debug("#");
.git/rebase-apply/patch:119: trailing whitespace.
.git/rebase-apply/patch:120: space before tab in indent.
/* get start time */
warning: squelched 985 whitespace errors
warning: 990 lines add whitespace errors.
Falling back to patching base and 3-way merge...
warning: Cannot merge binary files: package/qca/qca-nss-drv/files/nss-firmware/qca-nss1-retail.bin (HEAD vs. Add qca package for qsdk10.0)
warning: Cannot merge binary files: package/qca/qca-nss-drv/files/nss-firmware/qca-nss0-retail.bin (HEAD vs. Add qca package for qsdk10.0)
CONFLICT (add/add): Merge conflict in package/qca/shortcut-fe/shortcut-fe/files/etc/init.d/shortcut-fe
Auto-merging package/qca/shortcut-fe/shortcut-fe/files/etc/init.d/shortcut-fe
CONFLICT (add/add): Merge conflict in package/qca/qca-nss-ecm/Makefile
Auto-merging package/qca/qca-nss-ecm/Makefile
CONFLICT (add/add): Merge conflict in package/qca/qca-nss-drv/files/nss-firmware/qca-nss1-retail.bin
Auto-merging package/qca/qca-nss-drv/files/nss-firmware/qca-nss1-retail.bin
CONFLICT (add/add): Merge conflict in package/qca/qca-nss-drv/files/nss-firmware/qca-nss0-retail.bin
Auto-merging package/qca/qca-nss-drv/files/nss-firmware/qca-nss0-retail.bin
CONFLICT (add/add): Merge conflict in package/qca/qca-nss-crypto/Makefile
Auto-merging package/qca/qca-nss-crypto/Makefile
error: Failed to merge in the changes.
Patch failed at 0067 Add qca package for qsdk10.0
hint: Use 'git am --show-current-patch' to see the failed patch
Resolve all conflicts manually, mark them as resolved with
"git add/rm <conflicted_files>", then run "git rebase --continue".
You can instead skip this commit: run "git rebase --skip".
To abort and get back to the state before "git rebase", run "git rebase --abort".
git clone -b kernel5.4-nss-qsdk10.0 https://github.com/ACwifidude/openwrt.git
cd openwrt
git remote add upstream https://git.openwrt.org/openwrt/openwrt.git
git fetch upstream && git rebase upstream/master && ./scripts/feeds update -a && ./scripts/feeds install -a && cp diffconfig .config && make defconfig && ./scripts/getver.sh
make -j5
You can run just this line to update with master and get ready for make (I run something similar to this everytime I rebuild). Git stash is not required.
If git stash threw things off you can start off fresh with a new clone. A good technique is to keep a backup of your work (I have a github and a backup folder!).
Git rebase like it is written here grabs the head of master and puts all the NSS commits on top of it. Every now and then you’ll run in to an issue if there is a issue with a master commit or a conflict with one of the NSS commits and master.
Got it !
will try soon
Thanks a lot
I just tested your firmware of r7500 v2 ,and will test ea7500 v1 too
I added something your recommend at startup
here are some screensnap
I didn't see 940mbps on ethernet (772 MB is ethernet and 302 MB is Wi-FI)
is there someting wrong of my configuration?
thanks
Here are ea7500 v1 ,same configuration on startup
it looks ethernet speed is good ,better than r7500 v2
it's amazing !
You should be able to max out the ethernet iperf 3 tests, which is about 930-940 Mbps. Try using two parallel stream using the '-P2' switch in the client side. I found that my recent work Windows 10 Dell laptap doesn't seem to be able to saturate the ethernet link when sending, while my 10 year old iMac has no problem saturating the link sending packets.
I just try use "iperf3 -c 10.1.1.2 -P 2 -t 30" instead of "iperf3 -c 10.1.1.2 -i 1 -t 30" on client side
but no 930-940 Mbps,it's almost same like before
maybe my client pc is too old,it's samsuang R429,should be 10 years old too
thank you @quarky
update news:
you are right,after I change to thinkapd x250 as client to test.it's 930-940 Mbps on ea 7500 v1 now !
it's awesome !
r7500 v2 of your firmware gets speedtest 612 Mbits/sec on my lenovo pc x250
I don't konw why,same device and firmware.samsuang r429 gets 772 Mbits/sec
Appreciate the testing. EA7500 v1 looks good. Interesting that the r7500 v2 is getting lower results. What does speedtest.net look like?
