Build for Netgear R7800

m4niacjp · 2019-08-24T13:25:32.240Z

Hi, i've just installed your 19.07 build but when i'm trying to install a package i have this error.

Installing iptables-mod-tproxy (1.8.3-1) to root...
Downloading http://downloads.openwrt.org/releases/19.07-SNAPSHOT/targets/ipq806x/generic/packages/iptables-mod-tproxy_1.8.3-1_arm_cortex-a15_neon-vfpv4.ipk
Collected errors:
 * satisfy_dependencies_for: Cannot satisfy the following dependencies for iptables-mod-tproxy:
 *      kernel (= 4.14.138-1-59af6c05c48bd57aceabd02fb462389d)
 * opkg_install_cmd: Cannot install package iptables-mod-tproxy

hnyman · 2019-08-24T16:09:32.914Z

You can't install kernel depending packages to private builds.

wired · 2019-08-25T15:51:31.746Z

I tried r10825-2ceee0e023-20190818 ("old") and it's very unstable. It locks up between minutes and hours, I have to do a hard power cycle. Pings to the AP are all over the place and when I ssh in it's very laggy as if there's high packet loss.

I also tried firmware-5.bin_10.4-3.10-00047 (from https://github.com/kvalo/ath10k-firmware/tree/master/QCA9984/hw1.0) with it but it's not any better, same issues.

Also, a pile of these in dmesg:

[  255.816870] ath10k_warn: 14 callbacks suppressed
[  255.816877] ath10k_pci 0001:01:00.0: SWBA overrun on vdev 1, skipped old beacon
[  255.820679] ath10k_pci 0001:01:00.0: SWBA overrun on vdev 0, skipped old beacon

Last usable build for me remains r10443-ea4e1dac71 ("old"), but even that is not without issues for me, it still reboots once in a while, but at least it reboots instead of having hard lockups.

wired · 2019-08-25T16:15:58.066Z

Update on the above, managed to capture a firmware crash from dmesg, if useful to anyone:

[  152.122093] ath10k_pci 0000:01:00.0: firmware crashed! (guid n/a)
[  152.123934] ath10k_pci 0001:01:00.0: firmware register dump:
[  152.131193] ath10k_pci 0000:01:00.0: qca9984/qca9994 hw1.0 target 0x01000000 chip_id 0x00000000 sub 168c:cafe
[  152.131198] ath10k_pci 0000:01:00.0: kconfig debug 0 debugfs 1 tracing 0 dfs 1 testmode 1
[  152.133278] ath10k_pci 0000:01:00.0: firmware ver 10.4-3.9.0.2-00046 api 5 features no-p2p,mfp,peer-flow-ctrl,btcoex-param,allows-mesh-bcast,no-ps crc32 a09d5bfe
[  152.145788] ath10k_pci 0001:01:00.0: [00]: 0x0000000A 0x000015B3 0x0096353E 0x00975B31
[  152.145807] ath10k_pci 0001:01:00.0: [04]: 0x0096353E 0x00060934 0x00000018 0x00000000
[  152.153300] ath10k_pci 0000:01:00.0: board_file api 2 bmi_id 0:1 crc32 85498734
[  152.160186] ath10k_pci 0001:01:00.0: [08]: 0x00400000 0x00963508 0x00000034 0x00418840
[  152.167453] ath10k_pci 0000:01:00.0: htt-ver 2.2 wmi-op 6 htt-op 4 cal pre-cal-file max-sta 512 raw 0 hwcrypto 1
[  152.167584] ath10k_pci 0000:01:00.0: firmware register dump:
[  152.177899] ath10k_pci 0001:01:00.0: [12]: 0x00000009 0x00001557 0x009650E0 0x009650E9
[  152.183806] ath10k_pci 0000:01:00.0: [00]: 0x0000000A 0x000015B3 0x0096353E 0x00975B31
[  152.189614] ath10k_pci 0001:01:00.0: [16]: 0x0096353E 0x00972009 0x009606CA 0x0096F92A
[  152.199418] ath10k_pci 0000:01:00.0: [04]: 0x0096353E 0x00060934 0x00000018 0x00000000
[  152.207609] ath10k_pci 0001:01:00.0: [20]: 0x4096353E 0x00401BF0 0x00975A00 0x00415620
[  152.221964] ath10k_pci 0000:01:00.0: [08]: 0x00400000 0x00963508 0x00000034 0x00418840
[  152.221969] ath10k_pci 0000:01:00.0: [12]: 0x00000009 0x00001557 0x009650E0 0x009650E9
[  152.221975] ath10k_pci 0000:01:00.0: [16]: 0x0096353E 0x00972009 0x009606CA 0x0096F7FB
[  152.221980] ath10k_pci 0000:01:00.0: [20]: 0x4096353E 0x00401BF0 0x00975A00 0x00415620
[  152.221985] ath10k_pci 0000:01:00.0: [24]: 0x80963F50 0x00401C50 0x004293E0 0xC096353E
[  152.221989] ath10k_pci 0000:01:00.0: [28]: 0x809C5BC8 0x00401C70 0x0000001E 0x00402300
[  152.222003] ath10k_pci 0000:01:00.0: [32]: 0x809608C1 0x00401C90 0x0000001E 0x00400000
[  152.229820] ath10k_pci 0001:01:00.0: [24]: 0x80963F50 0x00401C50 0x004293AC 0xC096353E
[  152.237677] ath10k_pci 0000:01:00.0: [36]: 0x409605F5 0x00401CB0 0x0000001E 0x0000001E
[  152.237683] ath10k_pci 0000:01:00.0: [40]: 0x00000000 0x00401CD0 0x00050024 0x00000009
[  152.237688] ath10k_pci 0000:01:00.0: [44]: 0x00422198 0x0000001F 0x0000001E 0x00000008
[  152.237693] ath10k_pci 0000:01:00.0: [48]: 0x00000000 0x00401CD0 0x00050024 0x00000009
[  152.237698] ath10k_pci 0000:01:00.0: [52]: 0x00000009 0x00406724 0x00406728 0x0040672C
[  152.237703] ath10k_pci 0000:01:00.0: [56]: 0x00000000 0x00000000 0x00000000 0x00000000
[  152.237715] ath10k_pci 0000:01:00.0: Copy Engine register dump:
[  152.244839] ath10k_pci 0001:01:00.0: [28]: 0x809C5BC8 0x00401C70 0x0000001E 0x00402300
[  152.252899] ath10k_pci 0000:01:00.0: [00]: 0x0004a000   1   1   3   3
[  152.263215] ath10k_pci 0001:01:00.0: [32]: 0x809608C1 0x00401C90 0x0000001E 0x00400000
[  152.268862] ath10k_pci 0000:01:00.0: [01]: 0x0004a400   6   6  14  15
[  152.276585] ath10k_pci 0001:01:00.0: [36]: 0x409605F5 0x00401CB0 0x0000001E 0x00000000
[  152.284412] ath10k_pci 0000:01:00.0: [02]: 0x0004a800  33  33  32  33
[  152.292394] ath10k_pci 0001:01:00.0: [40]: 0x00000000 0x00401CD0 0x00050024 0x0000001F
[  152.300254] ath10k_pci 0000:01:00.0: [03]: 0x0004ac00   4   4   4   4
[  152.300263] ath10k_pci 0000:01:00.0: [04]: 0x0004b000 6828 6828 154 134
[  152.300270] ath10k_pci 0000:01:00.0: [05]: 0x0004b400   8   8 103 104
[  152.300278] ath10k_pci 0000:01:00.0: [06]: 0x0004b800  18  18  18  18
[  152.300291] ath10k_pci 0000:01:00.0: [07]: 0x0004bc00   1   1   1   1
[  152.308200] ath10k_pci 0001:01:00.0: [44]: 0x00422198 0x0000001F 0x0000001E 0x00000000
[  152.316062] ath10k_pci 0000:01:00.0: [08]: 0x0004c000   0   0 127   0
[  152.316070] ath10k_pci 0000:01:00.0: [09]: 0x0004c400   0   0   0   0
[  152.316078] ath10k_pci 0000:01:00.0: [10]: 0x0004c800   0   0   0   0
[  152.316086] ath10k_pci 0000:01:00.0: [11]: 0x0004cc00   0   0   0   0
[  152.535145] ath10k_pci 0001:01:00.0: [48]: 0x00000000 0x00401CD0 0x00050024 0x0000001F
[  152.541457] ath10k_pci 0001:01:00.0: [52]: 0x00406720 0x00406724 0x00406728 0x0040672C
[  152.549382] ath10k_pci 0001:01:00.0: [56]: 0x00000000 0x00000000 0x00000000 0x00000000
[  152.557282] ath10k_pci 0001:01:00.0: Copy Engine register dump:
[  152.565170] ath10k_pci 0001:01:00.0: [00]: 0x0004a000   1   1   3   3
[  152.570890] ath10k_pci 0001:01:00.0: [01]: 0x0004a400  26  26  34  35
[  152.577598] ath10k_pci 0001:01:00.0: [02]: 0x0004a800  21  21  20  21
[  152.583909] ath10k_pci 0001:01:00.0: [03]: 0x0004ac00  11  11  11  11
[  152.590437] ath10k_pci 0001:01:00.0: [04]: 0x0004b000 4019 4019  81  49
[  152.596840] ath10k_pci 0001:01:00.0: [05]: 0x0004b400  30  30  61  62
[  152.603179] ath10k_pci 0001:01:00.0: [06]: 0x0004b800  10  10  10  10
[  152.609867] ath10k_pci 0001:01:00.0: [07]: 0x0004bc00   1   1   1   1
[  152.616303] ath10k_pci 0001:01:00.0: [08]: 0x0004c000   0   0 127   0
[  152.622623] ath10k_pci 0001:01:00.0: [09]: 0x0004c400   1   1   1   1
[  152.629144] ath10k_pci 0001:01:00.0: [10]: 0x0004c800   0   0   0   0
[  152.635578] ath10k_pci 0001:01:00.0: [11]: 0x0004cc00   0   0   0   0
[  152.728008] ieee80211 phy0: Hardware restart was requested
[  153.681035] ieee80211 phy1: Hardware restart was requested

johnnysl · 2019-08-25T19:23:36.438Z

how does /etc/config/wireless look like? could you post it here? (just delete the password lines)
i'm curious if you are running something special or not.

wired · 2019-08-25T21:07:02.154Z

See below and thanks for taking a look.

I'd be so happy if it were as simple as some setting and I could have a stable AP.

I run it in AP mode only (almost all services disabled), but I need 3 VLANs hence the different network values. 2 VLANs use eth0 and 1 uses eth1 if that makes a difference in terms of stability. Let me know if the switch/VLAN config is important and I can post that too.

config wifi-device 'radio0'
	option type 'mac80211'
	option channel '149'
	option hwmode '11a'
	option path 'soc/1b500000.pci/pci0000:00/0000:00:00.0/0000:01:00.0'
	option htmode 'VHT80'
	option country 'US'
	option legacy_rates '0'
	option noscan '1'

config wifi-iface 'default_radio0'
	option device 'radio0'
	option network 'lan'
	option mode 'ap'
	option ssid 'REDACTED'
	option encryption 'psk2+ccmp'
	option key 'REDACTED'
	option wps_pushbutton '0'
	option short_preamble '1'

config wifi-device 'radio1'
	option type 'mac80211'
	option hwmode '11g'
	option path 'soc/1b700000.pci/pci0001:00/0001:00:00.0/0001:01:00.0'
	option htmode 'HT20'
	option channel '6'
	option country 'US'
	option legacy_rates '0'
	option noscan '1'

config wifi-iface 'default_radio1'
	option device 'radio1'
	option network 'lan'
	option mode 'ap'
	option ssid 'REDACTED'
	option encryption 'psk2+ccmp'
	option key 'REDACTED'
	option wps_pushbutton '0'
	option short_preamble '1'

config wifi-iface
	option device 'radio1'
	option mode 'ap'
	option ssid 'REDACTED'
	option encryption 'psk2+ccmp'
	option key 'REDACTED'
	option wps_pushbutton '0'
	option isolate '1'
	option network 'guest'
	option disabled '0'
	option short_preamble '1'

config wifi-iface
	option device 'radio0'
	option mode 'ap'
	option ssid 'REDACTED'
	option encryption 'psk2+ccmp'
	option key 'REDACTED'
	option wps_pushbutton '0'
	option isolate '1'
	option short_preamble '1'
	option network 'work'

RainGater · 2019-09-03T02:04:19.030Z

Installed your 08/29 build (R7800-owrt1907-r10352) on my R7800 and looks pretty good. Thank you for a wonderful build.

I have a 400/20 Mbps line and was wondering what is the max throughput I can expect on this build? I get about 120 Mbps and was wondering whether I need to make any special tweaks to get it to at least in the 300 Mbps range.

DSL Reports – 2 Sep 19

Speed test result

I tested my connection at DSLReports.com/speedtest ..

FYI - I enabled SQM QOS and set the download link to 360000 kbps and use cake/piece_of_cake QD. Remaining settings are pretty much basic.

hnyman · 2019-09-03T06:43:44.035Z

RainGater:

I have a 400/20 Mbps line

With those speeds, you might drop cake and just use simple.sqm with fq_codel. Or at least stop modifying the download speed. I feel that cake is too computational intensinve for gigagit class speeds. With fq_codel I get some 200/18 by myself, maxing out my connection.

If you look at the R7800 discussion, you can see that there is some discussion on the processor speedup tweaks, mostly regarding the CPU clock and memory clock speeds vs. frequency scaling. Right now I have no special modifications in my own build.

RainGater · 2019-09-03T12:34:19.203Z

hnyman:

With those speeds, you might drop cake and just use simple.sqm with fq_codel.

Thank you. After using fq_codel with simple.qos, I see it is little bit better. I still have the download speed to 360K and do you recommend not to set any limits or have a limit for simple.qos?

DSL Reports – 3 Sep 19

Speed test result

I tested my connection at DSLReports.com/speedtest ..

hnyman · 2019-09-03T14:26:52.376Z

Are you talking about wired speed from PC, or about wireless? Wireless could be slow for other reasons in any case.

Depending on your user profile I see no actual need for setting the download speed limit to 360. Why spend the CPU cycles for QoS calculations if you are not going to saturate the local link in any case. You might just try unlimited download speed.

RainGater · 2019-09-03T16:43:53.153Z

I am talking about wired. I have it setup at 360,000 for download and 18,000 for upload - that is 10% below my subscribed limit. I read online that usually you setup 10% below your available link speed.

So, you suggest that I set it as 0 for both upload and download?

moeller0 · 2019-09-03T17:38:26.887Z

RainGater:

So, you suggest that I set it as 0 for both upload and download?

This will effectively disable sqm...

hnyman · 2019-09-03T18:10:12.431Z

moeller0:

RainGater:

So, you suggest that I set it as 0 for both upload and download?

This will effectively disable sqm...

Yep.

I simply suggest that with a 400 Mbit downlink it makes usually no sense to spend CPU cycles for throttling the download speed (as likely in real life there will not be a continuous download bandwidth from external internet exceeding that speed). You might still need to monitor the upload bandwidth as you "only" have 20 Mbit upload speed.

At least test it.

Additionally the "leave 10% of bandwidth free" advice relates to times of a few megabits traffic to both direction. Then it made sense to to leave at least some bandwidth for protocol traffic. With 400 Mbit, it sounds like an overkill to leave 40 Mbit free for that protocol traffic.

The qdiscs used for QoS tools like SQM have usually been designed for 5-100 Mbit/s speeds where you would normally have a CPU power surplus compared to traffic volume. With gigabit speeds you do not have that with normal home routers (except with some x86 and maybe mvebu based ones). Thus the qdisc's intended usage of spending the CPU power surplus for calculating the optimal traffic in narrow upstream internet bandwidth gets invalid. Now with gigabits speeds you have surplus traffic volume that threatens to choke the CPU calculation power, so you really need to keep things simple in order to sustain the traffic.

RainGater · 2019-09-03T18:13:39.088Z

Thanks. I mistook hnyman's comment that he suggested to not set anything for download and upload. I see hynman responded and it explains very well.

I setup 0 for download and 19,000 for upload and will see how it works...

RainGater · 2019-09-03T18:15:41.502Z

hnyman:

You might still need to monitor the upload bandwidth as you "only" have 20 Mbit upload speed.

Thank you very much for explaining it very well. I setup 0 for download and 19,000 for upload and will monitor the performance.

moeller0 · 2019-09-03T18:27:11.949Z

hnyman:

I simply suggest that with a 400 Mbit downlink it makes usually no sense to spend CPU cycles for throttling the download speed (as likely in real life there will not be a continuous download bandwidth from external internet exceeding that speed). You might still need to monitor the upload bandwidth as you "only" have 20 Mbit upload speed.

hnyman:

At least test it.

+1

hnyman:

Additionally the "leave 10% of bandwidth free" advice relates to times of a few megabits traffic to both direction. Then it made sense to to leave at least some bandwidth for protocol traffic. With 400 Mbit, it sounds like an overkill to leave 40 Mbit free for that protocol traffic.

Nah, for ingress we need our artificial bottleneck to be lower than the real bottleneck, the lower in relation to the true biottleneck the easier it is to keep packets out of the often under-managed & over-sized upstream buffers. Around 10% works reasonably enough for a number of people and is easy enough to configure, but it is by no means a magic number. And for egress one often can reach close to 100% of the actual link rate (but one needs to know the applicable per-packet-overhead and the link's technology), again 90% of link rate will work just as well (at a higher bandwidth cost) so we also recommend a reduction by 10% points. For users that just want it simple 90%/90% often works and people more into the details quickly learn how to iteratively find their personal acceptable trade-off between latency-increase-under-load and bandwidth-sacrifice

Just as an aside, and really not applicable for a 400Mbps link, ATM cell header add almost 10% of overhead, so for ADSL lnks the 90%/90% recommendation is actually to permissive, but I digress...

hnyman:

The qdiscs used for QoS tools like SQM have usually been designed for 5-100 Mbit/s speeds where you would normally have a CPU power surplus compared to traffic volume.

Not sure that this was the design goal. My interpretation is more they were developed mainly on relatively beefy desktop-class CPUs that can tolerate the shaping load even up to 1Gbps, and the rather anemic router CPU's simply can not keep up (I am simplifying, it might not be the CPU core but the whole package including memory and DMA bandwidth).

hnyman:

Now with gigabits speeds you have surplus traffic volume that threatens to choke the CPU calculation power, so you really need to keep things simple in order to sustain the traffic.

+1; this is really always a trade-off, and not over-taxing the router CPU needs to be factored in.

Stevie · 2019-09-05T20:24:16.249Z

Hi,

last kernel compile was some years ago :-/ ...and perhaps, meanwhile I missed some things regarding compiling images for devices like the r7800.
A lot of the discussions in this thread spins around config-file parameters and so on. Wouldn't it be of advantage, to include the basic configurations (beside of the user space packages) into the github repo (.config)?

TIA Stevie

RainGater · 2019-09-06T00:45:06.416Z

I see in Kong's build, a user posted the following to run in startup. Would this help?

echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
echo performance > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor

echo 400000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq
echo 400000 > /sys/devices/system/cpu/cpu1/cpufreq/scaling_min_freq

echo 20 > /sys/devices/system/cpu/cpufreq/ondemand/up_threshold

echo 2 > /proc/irq/31/smp_affinity
echo 2 > /proc/irq/32/smp_affinity

exit 0

ratking · 2019-09-06T03:13:18.162Z

~# cat /sys/devices/system/cpu/cpu1/cpufreq/scaling_available_frequ
encies
384000 600000 800000 1000000 1400000 1725000

400000 isnt a frequency though

jerrytouille · 2019-09-06T15:05:42.244Z

for those with 200+mbps simply disable the entire sqm or set the associated line to 0. so for example 200/20 line sqm would be 0/18, 200/200 line would be to disable the entire sqm (you don't really need it)