Custom firmware build for R7800 v. 1.0.2.32SF

[mr.fabe]

Hello,

For those of you who use the VPN option on this firmware, are you able to connect to your router? While using a cellular connection, I am able to ping devices all devices on my home network, however, I cannot connect and login to the router. I was wondering if this is due to the IPv6 issue with some cellular connections but I was able to connect and login to the router on the older firmware before.

Thank you,
mr.fabe

 

[Charles]

Hi Voxel,
I hope you're having a good break. Some comments for when you return.

I started with LEDE on my R7800 routers installed at various locations, which is my preferred distribution for its flexibility, configurability, and openness. However, the ath10k wireless stack is not stable enough for real-world use so I have stopped using it.

Your firmware is the next best alternative, offering the stability of manufacturer firmware with additional flexibility and features. I have settled on your firmware for all my R7800 installations. I'm making little tweaks to get the functionality I need.

Some questions:

1) DNSCrypt: Could you compile DNSCrypt with plugins support? This would allow us to get increased performance as described here, and other useful features such as filtering.

2) DNSCrypt: NTP server passthrough. In LEDE, it's recommended to have DNS queries to the NTP server bypass DNSCrypt and go through standard DNS, in case the time is set wrong, so DNSCrypt would never work. Could this issue occur in your firmware?

3) dnsmasq: In LEDE, all entries in "etc / hosts" applies to all devices in the LAN. However in your firmware, it only applies to the router itself. Do you know how to make it apply to all devices in the LAN?

Thanks

 

[Voxel]

Guys, sorry for a silence from my side: I was absent (practically no Internet access) and arrived only today night. The only good is that all this time my R7800 worked OK (uptime 25 days), some kind of test

I'll comment/answer your questions/requests but step-by-step.

Voxel.

 

[Voxel]

Hi Voxel,
I hope you're having a good break. Some comments for when you return.

I started with LEDE on my R7800 routers installed at various locations, which is my preferred distribution for its flexibility, configurability, and openness. However, the ath10k wireless stack is not stable enough for real-world use so I have stopped using it.

Your firmware is the next best alternative, offering the stability of manufacturer firmware with additional flexibility and features. I have settled on your firmware for all my R7800 installations. I'm making little tweaks to get the functionality I need.

Some questions:

1) DNSCrypt: Could you compile DNSCrypt with plugins support? This would allow us to get increased performance as described here, and other useful features such as filtering.

2) DNSCrypt: NTP server passthrough. In LEDE, it's recommended to have DNS queries to the NTP server bypass DNSCrypt and go through standard DNS, in case the time is set wrong, so DNSCrypt would never work. Could this issue occur in your firmware?

3) dnsmasq: In LEDE, all entries in "etc / hosts" applies to all devices in the LAN. However in your firmware, it only applies to the router itself. Do you know how to make it apply to all devices in the LAN?

Thanks

In general your questions are what I need, i.e. feedback from the user of my firmware with concrete user's needs.

I'll check 1) 02) and answer you later. If I am not mistaken, I did something like this with 3). Will check too.

Voxel.

 

[Voxel]

Hello,

For those of you who use the VPN option on this firmware, are you able to connect to your router? While using a cellular connection, I am able to ping devices all devices on my home network, however, I cannot connect and login to the router. I was wondering if this is due to the IPv6 issue with some cellular connections but I was able to connect and login to the router on the older firmware before.

Thank you,
mr.fabe

Probably you tried to connect to older version of firmware using TAP. Now you try to use smartphone config with TUN (because of default config for smartphones is TUN). See this comment to similar report:

https://www.snbforums.com/threads/c...r-r7800-v-1-0-2-31sf.39142/page-8#post-330276

Voxel.

 

[Voxel]

I have an issue, not related to any changes you made, but hoping you may have some tips.

Voxel I am getting low USB 3.0 Speeds ~40/20 MB/s R/W. I'm using an OEM USB 3.0 case with a Hitachi 2.5" 5400RPM drive, I tried both NTFS/EXT4 filesystems.

I made the following changes and still no difference, do you have any tips? Do you think it would be possible to include "irqbalance" package for load balancing, not sure if that would help.

cat /sys/block/sda/device/max_sectors (240 Default)
echo 1024 > /sys/block/sda/device/max_sectors
cat /sys/block/sda/queue/read_ahead_kb (128 Default)
echo 1024 > /sys/block/sda/queue/read_ahead_kb

Difficult to say. Maybe specific of concrete case/drive. I have about 80MB/s R/W with my old WD (2.5", USB3). Maybe it has sense to try hdparm package from Entware. And check system logs when you attach your drive to router. Maybe it is recognized as USB2?

irqbalance: I am not sure that it will help. OK, I'll think over this.

Voxel.

 

[avtella]

Yeah it was odd because the same drive/cases hit 100+ MB/s when directly connected to a computer.

 

[Voxel]

Hi Voxel,

I would like to report that DDNS at times wont work. Manually need to go to the status button for sometimes or disable and enable it for it to work again

Seems you are right. There is a bug in the stock FW, when wget is placed in /usr/sbin directory, but some programs try to run it from /usr/bin. One prebuilt module /usr/sbin/noip2 contains such run. OK, I'll correct this bug in the next release adding symlink to /usr/bin

You can try to create such symlink from console to check that problem disappears:

ln -sf /usr/sbin/wget /usr/bin/wget

Also the date is wiered.

I could not reproduce. Maybe it is because of the bug above. My report:

***.mynetgear.com updated successfully at 08:33 AM, 07/05/2017

Voxel.

 

[RogerSC]

Seems you are right. There is a bug in the stock FW, when wget is placed in /usr/sbin directory, but some programs try to run it from /usr/bin. One prebuilt module /usr/sbin/noip2 contains such run. OK, I'll correct this bug in the next release adding symlink to /usr/bin

You can try to create such symlink from console to check that problem disappears:

ln -sf /usr/sbin/wget /usr/bin/wget

Voxel.

Wouldn't /usr/bin and /usr/sbin be "read-only" directories? In that case you can't create the symlink on a live router *smile*.

 

[Voxel]

Wouldn't /usr/bin and /usr/sbin be "r buildead-only" directories? In that case you can't create the symlink on a live router *smile*.

No problems for root to create/modify all files and directories. Changes are in /overlay

Filesystem           1K-blocks      Used Available Use% Mounted on
rootfs                   51032       376     48016   1% /
/dev/root                23040     23040         0 100% /rom
tmpfs                      512         0       512   0% /dev
ubi0:overlay_volume      51032       376     48016   1% /overlay
overlayfs:/overlay       51032       376     48016   1% /

Voxel.

 

[RogerSC]

No problems for root to create/modify all files and directories. Changes are in /overlay

Filesystem           1K-blocks      Used Available Use% Mounted on
rootfs                   51032       376     48016   1% /
/dev/root                23040     23040         0 100% /rom
tmpfs                      512         0       512   0% /dev
ubi0:overlay_volume      51032       376     48016   1% /overlay
overlayfs:/overlay       51032       376     48016   1% /

Voxel.

Okay, all the other firmware that I've played with had everything except /tmp (and jffs, if one were using that) as "read-only". Haven't delved too deeply into this one, obviously *smile*.

 

[Voxel]

Okay, all the other firmware that I've played with had everything except /tmp (and jffs, if one were using that) as "read-only". Haven't delved too deeply into this one, obviously *smile*.

This scheme is used in OpenWRT . Stock firmware (and thus my build) are based on OpenWRT.

Voxel.

 

[Supersaiyan_IV]

Dual-band ac devices can crash a R7800's WAN when HT160 is enabled.
To reproduce: dualband-ac device + make device generate ~0.6Gbit of internet traffic (eg. speedtest). Expect the R7800 crash crash and burn during the ~640Mbit/s upload phase.

@Voxel: Think this might be a Netgear issue?

 

[avtella]

It's not related to Voxel's work. There is an issue with the Intel7260AC and HT160 not playing nice, but Intel was supposedly working on that according to Netgear. Then again there really aren't many HT160 devices available. Intel 9260 will probably be the first laptop/desktop HT160 device and it's not even out yet.

 

[ulaganath]

a client can crash is so bad. It needs to drop the client when such issues . Not hold it till it die

 

[Supersaiyan_IV]

@avtella: Sounds about right. At least until my Oneplus 5 arrived.
Its dual-band chip also crashed the WAN when HT160 is enabled. And worked fine when disabled: http://www.speedtest.net/my-result/a/3015566544

I'm suspecting this isn't Netgear's fault either - but Qualcomm's itself.
Because this only happens at high speeds it's probably a data-race, or a bad attempt at synchronization via 'sleep'. Don't know if it's possible to make a ThreadSanitizer debug build of the firmware, but that would easily catch a trace of the fault during runtime. But I suppose that would be doing Qualcomm/Netgear's work for them?

 

[ulaganath]

Hi Voxel,

I do want to report. That with this current firmware after downgrading to v2.0 QOS DB files the streamboost is stable and never crash on idle timed out. But even with this firmware with QOS v4.0 crash .

At that time the clients connected wirelessly wont work meaning it will be still associated but no internet . I have to disconect the client and re-connect.

Now even though the boost is working i am facing similar issues. Why? i check for the Devices session it is active. But for wireless devices nope there is nil activity. I check other devices as well.

Sometimes my phone will go on sleep and wont actively scan wifi though still connected. once i re-connect it works fine.

 

[Killhippie]

Hi Voxel,

I do want to report. That with this current firmware after downgrading to v2.0 QOS DB files the streamboost is stable and never crash on idle timed out. But even with this firmware with QOS v4.0 crash .

At that time the clients connected wirelessly wont work meaning it will be still associated but no internet . I have to disconect the client and re-connect.

Now even though the boost is working i am facing similar issues. Why? i check for the Devices session it is active. But for wireless devices nope there is nil activity. I check other devices as well.

Sometimes my phone will go on sleep and wont actively scan wifi though still connected. once i re-connect it works fine.

Do you suffer power issues in your country? Brownouts power surges etc? I ask because I noticed 'sometimes' If i power my router down for some reason and remove the power supply from the unit the QoS GUI shows 'low' for all devices upon boot up. But the QoS can easily be reset to how it was if you have any special QoS rules rather than letting it autonomously decide on the rules for each device. QoS does not crash I have never had any wireless device not connect or not scan for Wi-Fi upon wake up in the time i have had my R7800 which was pretty much from the release date. Mine however is protected by a Smart 1000VA sine wave UPS along with my modem, 27" 5K iMac, printer, external hard drive for Time machine, NAS and a very old but still gorgeous sounding Logitech Z2300 2.1 speaker set up.

I'm wondering if power issues in your area could be what is causing you to have this issue with v4 of the database because when the router is up and running (last time before I powered it down it had been for well over 31 days) and the database worked fine, the GUI showed traffic like YouTube for my iMac or Netflix on my smart TV there was no crashing at all and there never has been. it acts as it did in the previous version. Also you mentioned some devices didn't have a + sign on them, if they are dormant on the network but still seen by the router often they don't but as soon as you use them the + symbol appears and you can see what traffic is being used eg 'general internet' or maybe YouTube, and also the amount of traffic is indicated Numerically in Mbps (4K HDR Netflix is about 16-20Mbps and the blue bar for visual effect as well.

i am just wondering if your equipment is at risk due to power issues which might cause the issues you are having. No one on this forum (unless I have missed them) or the Netgear forums has mentioned this, apart from you so the logical deduction would be (I can always be wrong here so sorry if I am) is that the issue is at your end and not in the data base apart from its new behaviour with loss of power (I repeat it does not always do this oddly enough when power is lost) because apart from that odd issue it works just fine. I have noticed if you note what the QoS settings were for each device at and re apply those setting on boot of the router they eventually do stick when the power goes off, so maybe there is a bug here, but the GUI for QoS has always been somewhat buggy on this router, and i have seen devices loose QoS rules before after a power cycle, just not all of them. This leads me as I said to think the issue is local to you so maybe devices cant fine the network because the router is just rebooting after a brief power outage possibly?

 

[ulaganath]

No even i have APC ups 1000 IN and never have router powercycle unless it my UPS fails it will be completely off the backup.

I do see many reported the similar issue not only on R7800 but other x series as well.

Again i knew that when devices dont show + symbol it may not be doing active traffic but will do show once its showing some traffic.

Coming to GUI QOS. yes it is very slow and only show after few second by the time the isp speed will be reached on capable site or application use. But it is for sure broken.

Here is a video i had already shared. The only way to see if it is completely broken or crashed and Streamboost is not working is

sort by application in the first part of filter. You will see a blank page.

https://www.dropbox.com/s/9huonvvoe...database corrupting the streamboost.webm?dl=0

Again after many attempts from the engineer's remote session it was reported my desktop IP is causing the problem. Which is wrong. As i am connected via lan port. 2nd its not only way i crash it was previously crashed via wifi and even while remote access it crashed after all types of clients IDLE timedout which revokes re-authentication when i give the credentials instantly it crash itself but not in the v2.0 db.

something is doing the instability.

I am really frustated as it also affects most wifi devices to loose internet.

 

[ulaganath]

Today i had remote access to my router no local devices. It was connected for a while and QOS working fine. After 45 min of client was disconnect from net my office laptop. I reconnected and browser in Attached devices page. I clicked on that again after connecting to Internet. It did re-authentication. Boom even now v2.0 remote client can crash the QOS

Which did happened on v4.0 as well. But v2.0 holded and never before. Something is so wiered.

As this crash will be fixed on router reboot. But it not the point. If it is say months running and a single reboot fix is not a problem But within few days is not at all good for such a high end router costing twice as R7000 which is so goood which i now felt i made a mistake on give away r7000 and getting this hardware failure router.