Issues with multiple wifi clients

[jeff288]

When I have certain guests with wifi devices, I lose internet access and can't access my cable modem's setup page which is bridged to my N66U. I rarely ever have issues but when I have a few guests over using wifi, that's when I lose internet access. I can still access the N66U setup page, I just have to reboot my modem. Anyone have an answer to this oddity?

Edit: 380.65.4 firmware

 

[jph1589]

Are they iPhones or iPads? Do they possibly plug in to charge? If so, are they set up to backup to iCloud when plugged in?

I won't bother going into details on what might be happening unless you answer yes to those.

 

[jeff288]

Are they iPhones or iPads? Do they possibly plug in to charge? If so, are they set up to backup to iCloud when plugged in?

I won't bother going into details on what might be happening unless you answer yes to those.

1 iphone and a couple android tablets and phones. Yes on the second and I think yes to the third question. But this happens seemingly randomly but it does seem to happen almost every weekend when they're over including some weekdays when visiting and when not, I rarely ever have issues. But a reboot tends to fix it until it happens again which may be 1-3 times on the weekend.

 

[jph1589]

1 iphone and a couple android tablets and phones. Yes on the second and I think yes to the third question. But this happens seemingly randomly but it does seem to happen almost every weekend when they're over including some weekdays when visiting and when not, I rarely ever have issues. But a reboot tends to fix it until it happens again which may be 1-3 times on the weekend.

I had this issue on my network and it made me crazy getting a solution. What I found was that Apple products in particular when they plug in to charge and are setup to backup to icloud will completely bog down the upstream bandwidth. That makes it so that other network processes cant get through up or down. Within 2 or 3 minutes of them plugging in and starting their backup, the network link would go now. I had to reboot the router or modem or both. This sometimes happens with dropbox and other apps like that also. It happens more often with people who do not have high upstream bandwidth. The best way to test it is to have them turn off their backups before they plug in to charge. That will tell you if that is the problem.

It became a real problem for my network because two of my users take a lot of pictures on their iphones. When those backups started it just killed my network. The ethernet connected devices lost their link and all the wifi came to a complete stop. So now I am running Merlin firmware with my RT-AC3100 and I used the QOS bandwidth limiter on specific devices so that they can only use a part of my upstream bandwidth. I can set that up for all the Apple devices that us iCloud backups and photo sharing. I implemented this fix about 2 months ago and haven't had an issue since.

I hope this helps.

 

[sfx2000]

I had this issue on my network and it made me crazy getting a solution. What I found was that Apple products in particular when they plug in to charge and are setup to backup to icloud will completely bog down the upstream bandwidth. That makes it so that other network processes cant get through up or down. Within 2 or 3 minutes of them plugging in and starting their backup, the network link would go now. I had to reboot the router or modem or both. This sometimes happens with dropbox and other apps like that also. It happens more often with people who do not have high upstream bandwidth. The best way to test it is to have them turn off their backups before they plug in to charge. That will tell you if that is the problem.

FWIW - we don't see a lot of comments with other vendors, mostly it's Asus devices that seem to have some issues like this - and it's worthwhile to spend some time digging into this, and figure out why... not saying it's Asus specific, but a majority of the concerns are being voiced on the Asus sub-forums here on SmallNetBuilder...

The one unique thing that I've found with iOS compared to Android (or Windows/Linux/MacOS) is that IOS supports Multipath TCP - see RFC6824

TCP option 30 used to be a reserved field - and the RFC was only recently released - I'm wondering if this is invoking some kind of bug either inside Broadcom's code, or perhaps one of the third party libraries (Trend, or even Asus written code), and causing things to get weird here...

@RMerlin - feel free to reach out, and maybe we can sort a debug effort here.

(merged with a post below for context)

Additional background on MPTCP...

From IP networking lab - http://multipath-tcp.org/pmwiki.php/Users/ConfigureMPTCP

OpenWRT - https://wiki.openwrt.org/doc/uci/mptcp

Some third party using OpenWRT - http://mytechpg.blogspot.it/2016/01/multipath-tcp-mptcp-for-openwrt.html

Here's a test case - http://mytechpg.blogspot.com/2016/01/multipath-tcp-how-to-test-mptcp-with.html

I don't have an AsusWRT device to tinker with, so... I think perhaps @RMerlin has the right stuff to do this, or perhaps @kvic @hggomes or John perhaps...

sfx

 

[jph1589]

FWIW - we don't see a lot of comments with other vendors, mostly it's Asus devices that seem to have some issues like this - and it's worthwhile to spend some time digging into this, and figure out why... not saying it's Asus specific, but a majority of the concerns are being voiced on the Asus sub-forums here on SmallNetBuilder...

The one unique thing that I've found with iOS compared to Android (or Windows/Linux/MacOS) is that IOS supports Multipath TCP - see RFC6824

TCP option 30 used to be a reserved field - and the RFC was only recently released - I'm wondering if this is invoking some kind of bug either inside Broadcom's code, or perhaps one of the third party libraries (Trend, or even Asus written code), and causing things to get weird here...

@RMerlin - feel free to reach out, and maybe we can sort a debug effort here.

sfx

I would be eager to add any additional data for this or be a tester if you need one.

 

[jeff288]

I had this issue on my network and it made me crazy getting a solution. What I found was that Apple products in particular when they plug in to charge and are setup to backup to icloud will completely bog down the upstream bandwidth. That makes it so that other network processes cant get through up or down. Within 2 or 3 minutes of them plugging in and starting their backup, the network link would go now. I had to reboot the router or modem or both. This sometimes happens with dropbox and other apps like that also. It happens more often with people who do not have high upstream bandwidth. The best way to test it is to have them turn off their backups before they plug in to charge. That will tell you if that is the problem.

It became a real problem for my network because two of my users take a lot of pictures on their iphones. When those backups started it just killed my network. The ethernet connected devices lost their link and all the wifi came to a complete stop. So now I am running Merlin firmware with my RT-AC3100 and I used the QOS bandwidth limiter on specific devices so that they can only use a part of my upstream bandwidth. I can set that up for all the Apple devices that us iCloud backups and photo sharing. I implemented this fix about 2 months ago and haven't had an issue since.

I hope this helps.

Thanks. Come to think of it, don't think I had connection issues when the iphone user was away. I turned on traditional QOS, N66U doesn't have the fancier QOS option, so maybe it'll help.

 

[jph1589]

Thanks. Come to think of it, don't think I had connection issues when the iphone user was away. I turned on traditional QOS, N66U doesn't have the fancier QOS option, so maybe it'll help.

I tried the traditional QOS on mine and it didn't help. What are your down and up speeds from your ISP?

 

[jeff288]

I tried the traditional QOS on mine and it didn't help. What are your down and up speeds from your ISP?

15 down, 1 up. Doesn't take much to saturate my upload. Just did your option, I misunderstood at first, to limit upload bandwidth and hopefully will help.

 

[jph1589]

15 down, 1 up. Doesn't take much to saturate my upload. Just did your option, I misunderstood at first, to limit upload bandwidth and hopefully will help.

I have 1 up also. I have all my Apple devices limited to .9 Mbps upstream and that works perfectly.

 

[RMerlin]

I don't have an AsusWRT device to tinker with, so... I think perhaps @RMerlin has the right stuff to do this, or perhaps @kvic @hggomes or John perhaps...

There are no Apple product in this house, and there won't be any for the foreseeable future.

 

[sfx2000]

There are no Apple product in this house, and there won't be any for the foreseeable future.

So in other words - no interest...

 

[jph1589]

So in other words - no interest...

I have plenty of Apple products and can recreate the issue at will. If someone else can do the coding, I can do the testing.

 

[jeff288]

I have plenty of Apple products and can recreate the issue at will. If someone else can do the coding, I can do the testing.

I'll try also. For an issue caused by such ubiquitous hardware and seemingly simple problem, the apathy in fixing this is confusing. Maybe it should be stated somewhere when being sold that it's incompatible with Apple products if indeed it is because it's a hell of an issue to have to figure out later.

 

[sfx2000]

I think there's a way to block the mptcp connection negotiation with iptables...

This would probably be put in the forwarding change, something like below - I'll leave it to others on how to advise putting this within the chains on AsusWRT... the first would block inbound, the next blocks outbound - clients then fall back to normal TCP connections

iptables -A INPUT -p tcp --tcp-option 30 -m state --state NEW -j DROP
iptables -A OUTPUT -p tcp --tcp-option 30 -m state --state NEW -j DROP

​

 

[sfx2000]

Maybe it should be stated somewhere when being sold that it's incompatible with Apple products if indeed it is because it's a hell of an issue to have to figure out later.

Apple is not entirely without fault -- they have brought their own issues with Bonjour problems in some versions of iOS/MacOS, and they're sensitive to regions and 802.11h headers...

Good chance that the iptables item above might help...

 

[sfx2000]

To test the iptables change - one can enable mptcp on ubuntu, and then grab a pcap with a filter similar to below;

tcpdump –i eth0 ‘tcp[13] & 0x12 = 0x12’ and tcp[40] = 0x1e
​

To enable mptcp on ubuntu, check sysctl

sysctl –w net.mptcp.mptcp_enabled=1​

(set the bit to 0 to disable afterwords)

 

[jph1589]

To test the iptables change - one can enable mptcp on ubuntu, and then grab a pcap with a filter similar to below;

tcpdump –i eth0 ‘tcp[13] & 0x12 = 0x12’ and tcp[40] = 0x1e
​

To enable mptcp on ubuntu, check sysctl

sysctl –w net.mptcp.mptcp_enabled=1​

(set the bit to 0 to disable afterwords)

If I could send an image of my eyes glazing over, this would be the time.

 

[jph1589]

To test the iptables change - one can enable mptcp on ubuntu, and then grab a pcap with a filter similar to below;

tcpdump –i eth0 ‘tcp[13] & 0x12 = 0x12’ and tcp[40] = 0x1e
​

To enable mptcp on ubuntu, check sysctl

sysctl –w net.mptcp.mptcp_enabled=1​

(set the bit to 0 to disable afterwords)

Before you typed this one I was about to ask if I could just telnet in and

iptables -save
(enter your two commands)
Test
iptables -restore

Is that anywhere near being on the right page?

 

[sfx2000]

If I could send an image of my eyes glazing over, this would be the time.

Hehe... but perhaps someone that is running RMerlin's build and has an iDevice (or ubuntu, even inside a VM) can give this a try...

If it works, then no code changes needed, and it would solve a long standing issue...