I changed to the performance governor last week and wanted to share this surprising (to me) CPU graph:
Both Core 0 and Core 1 showed a similar drop in processor usage. I did nothing but change the CPU governor from ondemand to performance at that time.
Now the thermals:
The temperature increased about 2-3℃.
No complaints or problems to report; just wanted to share this data. I'll be leaving my governor on performance; the Luci web interface feels snappier, could be placebo.
Hey folks, I've been running a Linksys E8450 with the latest OpenWRT for about 45 days now. Yesterday, I accidently unplugged it, and now it will not boot up. Getting no LED lights or activity.
I remember reading here something about the governor and dead reboots, but was wondering if no LEDs on is consistent with that. Would appreciate being pointed in the right direction.
I did use a volt meter to confirm my power supply is still good.
Thanks!
Another one! Sorry to hear this @jamesm113.
The OKOD - OpenWrt Kiss of Death - strikes again.
We don’t know why, we don’t know when; but it’s only a matter of time, before OKOD strikes another poor RT3200 again. All was once well, but that was only then; once it was five, but five became ten. So we ain’t got no zen, knowing OKOD will strike for sure anytime again. WiFi 6, tricks and nice cake-mix, but too many bricks; we therefore place all our hope in @daniel and his promised fix.
I had similar issue with WED enabled on latest release. Disabling it results in perfect stability for extended periods:
OpenWrt 23.05.2, r23630-842932a63d
-----------------------------------------------------
root@OpenWrt:~# uptime
12:10:28 up 78 days, 8:17, load average: 0.25, 0.18, 0.08
It is surprising but .... logical
%CPU usage represents the ratio between time processor is working vs time processor is doing nothing (idle)
In your case:
ondemand / processor is running at half its max speed: it needs to run 6 seconds every 100 seconds to perform its basic job. %CPU = 6/100 = 6%
Performance / processor is running at full speed: it needs only to run 3 seconds every 100 seconds to perform the same job. %CPU = 3/100 = 3%
"ondemand = half speed" is only to illustrate what happens (not the right value).
ondemand is suppose to raise CPU freq when needed, but it is not immediate.
It starts from low frequency and ramp up to max speed.
CPU is often used when not running at full speed producing the effect you have observed (higher %CPU using ondemand)
Just read through all that and wow that is concerning. So should I avoid rebooting my E8450 while waiting for the promised fix?
Please correct me if you know more, but from my current knowledge about the issue: It is not known what the root cause for the reported boot or reboot problems is. It is not known if your device is affected by the issue. How should someone answer specific questions about handling the issue while we are still trying to solve a puzzle?
If we want to solve this puzzle then please don’t just ask how to handle the issue, let’s work on analyze it. Waiting for a solution does not help. I also do not see that a fix was promised.
I suggest to raise min_freq CPU clock speed or switch to performance governor since we have one report in the past that a minimum CPU clock speed of 600 MHz increased reboot stability. This should not affect cold boot problems so we still don’t know what the issue is about but it’s one idea what could contribute to the issue.
I suggest to note the model and production date of the router before the issue might occur to narrow the issue down if there are patterns of models or production dates that are affected or unaffected.
How does one find the model and production date? I'm not seeing anything obvious on the body or the board. Is there anything else I can do to contribute and/or revive my bricked E8450?
To add some data points:
I have 6 RT3200 (and 1 E8450 if that matters). 1 of the RT3200 started to suffer from min_feq soft reboot issue only after a sysupgrade, 23.05-rcX -> 23.05.2 IIRC. (Versions may not be accurate. The device is at a remote site so I don't play with it so often). Setting min_feq in rc.local mitigates the issue. All other devices work without setting min_feq.
I have a friend also purchased a few RT3200's, and 1 of his bricked in a recent morning. According to his description: "Internet was down in the morning but wifi was connected and router was accessible from luci (-- The internet issue maybe totally unrelated). Then after a reboot, router didn't come up at all. Power supply is good (swapped) but no LED." I promised him to jtag and try recover it next month or 2. Please remind me what info I need to collect (uboot console log and mtd dump?) before writing anything to it? I'm very curious too but I haven't received that brick.
I reboot mine daily at 3:30am. I've set the min frequency to 600MHz and don't have any issues.
Before setting the 600MHz min frequency, sometimes I had problems rebooting. I had to unplug and replug the device to turn on.
Cheers!
Why? Can’t this be avoided?
I do it because I want it to.
I know people here have their routers with incredible uptime, but I prefer to do a reboot daily to have if always fresh. That's just me.
Cheers!
Another user wrote that putting his device into the freezer got it back to life and cold boot. I would not recommend this since powering the device up while being the most cold area in the room should get air humidity condensation on and in the device starting instantly and this water together with electricity could lead to hardware damage but you could read here, and decide if you would like to take the risk.
How to read out the production date:
The latest openwrt commit fixed the offloading issue with 6.1 kernel. Thanks, Felix! 
I'm still on kernel 5.15 OpenWrt SNAPSHOT r24660-c22aa0be3e. How to switch to 6.1 using Attended Sysupgrade?
You need to build your own image, as the main branch for the mediatek target is still on 5.15 as of this morning:
SNAPSHOT r25154-f3cbdaec29 (kernel 5.15.148)
Keep an eye on this issue and you'll probably see when 6.1 hits main for these devices:
This is my RT3200 manufacturer date when I got it.
sw_version=1.0.0.1
modelNumber=RT3200
cert_region=US
hw_revision=1
hw_version=48SAR601.0GB
manufacturer_date=2021/09/05
I had to set a min frequency of 600MHz to avoid not powering up when rebooting on 23.05.x
This seems like pretty big news!
@daniel just pushed out a bunch of updates on master/snapshot, including the very significant mediatek package update to Linux kernel 6.1 and the reliability updates for E8450/RT3200 he talked about earlier!
Be sure to read his warning about updating first. I don't know where to find the updated installer he mentions, or maybe it isn't released yet. Basically, he ends by saying: DON'T USE YET.
mediatek: mt7622: modernize Linksys E8450 / Belkin RT3200 UBI build
Move fip and factory into UBI static volumes.
Use fitblk instead of partition parser.
!! RUN INSTALLER FIRST !!
Existing users of previous OpenWrt releases or snapshot builds will
have to **re-run the updated installer** before upgrading to firmware
after this commit.
DO NOT flash or run even just the initramfs image unless you have
run the updated installer which moves the content of the 'factory'
partition into a UBI volume.
tl;dr: DON'T USE YET!
Somebody please correct me if I'm wrong, but I believe this means if you're on SNAPSHOT right now and attempt a straight update, your device will cease to function (temporarily).
Daniel was merging that stuff as I was posting about 6.1 just above 
@daniel's github installer site currently still contains the old version 1.0.3 of the ubi installer.
But the site has also the "fitblk" branch that seems to enable building the all-ubi layout installer. That is looks like the new installer.
But there is no built installer release yet, as the installer build script build_installer.sh requires buildbot builds as components, so a bit waiting still needed until buildbot has built these new commits, after which the installer can be built (and released)...
