Yes, if I remember correctly I had similar behaviour, but I cannot be 100% sure - at that moment i've not had a USB-TTY cable and there may have been problems with one of the wireless clients.
I can check if reboots persist on the latest stock from Linksys. Can I assume that if stock firmware is running normal, than its not a hardware issue and opposite if problem persists?
If I remember correctly - there is no temperature/sensor monitors on stock firmware. Is there any way to monitor temperatures from, for example, USB-TTY connection?
I think u need to remove dnsmasq before you can install dnsmasq-full.
I did it in the past but i'm not sure how i did it but i'm pretty sure that i had vpn policy-based routing installed with one of your builds and it was working great!
I'm very confused atm...
edit: okay so i was able to install dnsmasq-full but it was quite a hassle...
I needed to install a few packages offline before i was finally able to install dnsmasq-full.
I guess there is something messed up with my config. Normally opkg install should automatically install all the needed packages if i remember right.
Very weird...
root@lede:~# opkg list |grep kmod-tcp-bbr
kmod-tcp-bbr - 4.14.66-1 - Kernel module for BBR (Bottleneck Bandwidth and RTT) TCP congestion control. It requires the fq ("Fair Queue") pacing packet scheduler. For kernel 4.13+, TCP internal pacing is implemented as fallback.
I would check the stock fw because it if it isn't stable, then I'd say there is a hardware issue going on. If it doesn't have issues then it would likely be an issue in the LEDE software or configuration within LEDE. If that were the case, I'd request a re-install/flash of LEDE without saving any configurations, and wait for a failure.
There are a couple of options (probably more) for monitoring the CPU temperature. First would be to write a simple script to check cpu temperature and write the result to a text file, and have cron run it as often as you like. Or you could use SNMP to monitor the temperature using a client. There's also a way you could probably add it to Collectd, and have it graph the CPU temperature.
EDIT
Looking at collectd, it's a simple check box to add "Sensors" to get CPU temperature.
Yes, this process is simple on LEDE. But I've no idea how to do it on stock firmware - it even has no ssh/telnet/smth to connect and USB-TTY connection does not give you any way to issue any commands.
Yes, that is the the normal cpu temperature of the 3200acm.
It's been probably 4 years since I've even looked at the stock firmware, but seem to remember there was a method to get ssh working OR there was an administrative page that gave many different types of statistics though I don't know if temp was one of them.
Speaking as part of the admin team for a large university network I can tell you that bcp38 is incredibly important and should be implemented across the Internet, but I'm not entirely convinced of it's relevance on individual home routers. Allow me to hold forth...
Apologies for trying to teach anyone to suck eggs, but here goes: BCP38 is part of an RFC to help prevent DDoS attacks. It should usually be implemented on a router that aggregates user access to an ISPs core and prevents ingress of packets with a source address that does not match the sending host. Packets with a spoofed source address are most commonly seen as part of a DDoS attack on a target host. As an attacker, if you have control of 10,000 owned hosts in a zombie cluster and they all send a DNS request at the same time, but with a spoofed source address of your target host, that target then receives 10,000 replies, saturating it's network connection. This gets even worse if the attacker makes use of an amplification attack where the response consists of many more packets than the original request.
These kinds of attacks could be easily stopped if ISPs dropped packets originating from users that don't have a source address matching their known network ranges. The most efficient place to implement it would be where you aggregate your broadband gear (be it cable modem UBRs, ADSL DSLAMs or whatever) before it gets to the core of your network, and certainly before it reaches your peering points with other ISPs or carriers.
I can see how implementing BCP38 on a home router would be a nice thing to do, every little helps after all, but if you're concerned about traffic such as this egressing your home/corporate network I'd say you should be more worried about securing your equipment in the first place to ensure it's not compromised and part of a botnet.
edit fwiw, that "package" is just a handful of firewall rules that you could implement yourself pretty easily if you wanted your router to be BCP38 compliant
Well, I've got a reboot on stock - https://pastebin.com/1Gxyqccr
But reboot happened not under the circumstances I've managed to reproduce situation on LEDE (different builds).
Right now, after reboot, it keeps working.
Thus, it's a high probability of hardware issue. What part of the hardware can provoke such behaviour?
I've tested memory with mtest on bootloader, I've put a high load on CPU on LEDE with WiFi down, I've checked nand for bad sectors - this tests went ok.
I may have used not the most painful test for CPU - heat issue still may be a problem. How can I test test WiFi chip?
I first purchased a wrt32x certified junk unit and it would not allow anything to work so I returned it and I then bought a brand new one , guess being cheap gets you what you pay for,lol, anyway I installed the openwrt firmware and got stuck , had to do a forced reinstall to factory which was a pain, then reinstalled everything the right way back to the current openwrt so I am guessing that there could be a HW issue with some of these units...
After reading about your temperature issue I have kept a pretty close eye on it, I did what I use to do with my old E3000 I put a small usb fan with a small filter under the unit(I use an old phone charger plug for the fan).I have been running about 55C very steady and under load about 60C ....
might want to try the fan idea and see if you can get the temps down, not that they are terrible but my units last longer staying cooler:)
I have 76-94 for CPU, 51-63 for Wifi. AC3200, working stable, outside temp are 32-36.
AC1900v1 i have before was a bit more hot by CPU, but default cooler in the box solve the problem.
I think temperature here is not a source of the problem, but symptom - seems like you got high CPU load.
I have a problem with my AC1900v1. In the actual firm i have installed (r6868) my internet speed is 450/500 over a 600/600 line. When i install the last firm all seems to be ok, but when i enable the wifi nets speed downs 250/250. If i disable wifi nets its go back to 450/500. Also i have randoms reboots....
It sounds like you're hitting the CPU limit with SIRQs, there. My WRT32X runs around 60-70% SIRQ (ie leaving ~30-40% of CPU for the OS after system IRQs) when saturating my wifi, copying files from a local server on the wired network at ~1Gb/s. The 1900AC has a dual core 1.6GHz CPU, whereas the WRT32X has a dual core 1.8GHz CPU, so you've got a little bit less processing power.
System IRQs are when some hardware (typically, in a router, the ethernet interfaces) interrupts the OS so it can instruct the CPU to do something else (reading or writing from memory, typically). The faster you go, the more packets you forward, the more the network interface interrupts the CPU. Forwarding packets from the wireless adapters to the wired (your internet connection) is using more hardware resources than simply forwarding from wired to wired, so that could be the problem.
SSH into your 1900AC, run 'top', set a download going from a fast site you know can saturate your network connection and have a look at the 'sirq' figure in top. If that figure is getting very high you'll know that your router basically isn't powerful enough for your broadband speed.
Personally, I would have liked a little fan like we used to have on the V1, but seems these cpu's are designed to run much hotter. Just doing the hand test above the case, it doesn't seem hot. However, I remember the V1 used to be like a little volcano as it always felt hot. I remember the wife putting a book on top of the V1, and remember not being happy
Another thing that helps increase the CPU Temperatures is CPU_Idle being disabled.
I have had two 80mm AC Infinity USB Component Fans running on my WRT1900AC v1 and now my WRT32X.
They both have been running almost non-stop for 3 years without a problem. I turn them off for a few minutes every 3-4 months to clean.
I was concerned about dust build-up inside the router, but after over 2 years of the fans constantly running I was surprised to see barely anything when I took the WRT1900AC case apart.
My WRT32X's CPU temperatures are consistently around 57°C, 37°C, & 36°C.
I'd buy a fan, but is there anything that tells me it is of any benefit? In other words, my WRT1900ACS v2 has temperatures of 75°C, 47°C, and 49°C. I can't find any information which tells me if those temperatures are inconsistent with the design of the hardware. Granted, a fan isn't exactly a large investment, however my engineering background tells me if these operating temperatures are within the design specification there's really little benefit to having a fan.
