Integrating the generation of the installer image into the current OpenWrt git tree would require quite a lot of structural changes. Currently we don't have a way to generate a specific per-device initramfs, and including yet another initramfs image as well as other build artifacts in an initramfs image (the installer) is completely out of scope.
What I was planning to do is simply to generate an installer based on 22.03.0 once it gets released.
I agree that forcing boot into recovery image based on pstore/ramoops may be confusing -- on the other hand it helps a lot to find and document hidden bugs which we do want to find and fix, so keeping it enabled has already brought as to a state where now (with nftables/firewall4 and no longer using the iptables flow-offloading hack) there just haven't been any reports of kernel crashes for months. In my opinion OpenWrt doesn't try to make a consumer-grade product but actually targets developers and enthusiasts, so I'm not feeling too bad about burdening users a bit if that results in overall better software quality (ie. meaningful bug reports of things which would otherwise go unnoticed).
This is kinda similar to the debate about the "reboot every 24h" cron-job you will find in some community mesh networks -- I understand the convenience and the need for it, but ultimately workarounds like this have the effect that things will just not get fixed and in the long-term you will then need to reboot (or live with kernel oops related reboots) more and more often.
Hence I'm tempted to just leave it on and maybe document it better (incl. how users can disable that feature) in the wiki.
How about a package e.g. "toggle-bootcmd-pstore-e8450" that installs a script file as /etc/toggle-bootcmd that by default displays the current bootcmd and offers cmdline options to toggle the initramfs fallback on/off with fw_setenv ? So the users would more easily be able to toggle the feature. The package might be included by default to E8450/RT3200.
Something like that might help casual users (although they naturally would need to find the command first even if the package would be installed by default).
Ps. Personal anecdote: I toggled that pstore initramfs feature off when I installed two new RT3200s for my sister. Her family would never be debugging anything, so having the routers possibly stuck on initramfs was not feasible. I think that this will true for many users.
Something like this:
(I made just minimal change and even left the if clause)
root@router4:~# cat /etc/toggle_bootcmd
#!/bin/sh
#
# This script can change the bootcmd in the u-boot bootloader
# of E8450 / RT3200 UBI variant to boot either into
# * initramfs recovery if pstore crash logs are detected (default)
# * main OpenWrt despite the crash log files in pstore
cur_bootcmd=$(fw_printenv -n bootcmd)
echo "Current setting: bootcmd=$cur_bootcmd"
echo -e "\nOptions:\n" \
"-i Boot into initramfs if crash files are present\n" \
"-o Boot always into normal OpenWrt\n"
if [ "$1" = "-i" ]
then
echo "Set bootcmd to use initramfs if pstore files present"
fw_setenv bootcmd "if pstore check ; then run boot_recovery ; else run boot_ubi ; fi"
elif [ "$1" = "-o" ]
then
echo "Set bootcmd to always boot into normal OpenWrt"
fw_setenv bootcmd "if pstore check ; then run boot_ubi ; else run boot_ubi ; fi"
else
echo -e "\nNo operation selected"
fi
sync
fw_printenv bootcmd
I generally like the idea, but it should be in /sbin rather than /etc (which contains configuration rather than executable stuff).
Maybe it could even be integrated with luci-app-advanced-reboot so we won't need an additional package for that.
Hey @hnyman thanks so much. Last night I left the build running following your simple build instructions, however it was too late here and I fell asleep. It worked fine. Yes, my build also shows the problem has been fixed. Thanks for sharing your diffconfig. I didn't know that the build comes straightforward without any configuration file.
good evening ! on March 20 will take place the freezing of the next release but paul says that it takes more time before the officialization 22.03
my question may be stupid but how long do you think thank you in advance
Maybe you can try iptables-nft with those commands?
Or disable the built in dns server and put the adblocking one directly on port 53.
Or config the built in dns server to use the adblock as the upstream dns server. I setup DHCP on my router to hand out pihole as DNS1 and the routers DNS as DNS2 if pihole/server would go down for some reason.
Realized my setup was actually adding an additional hop by forwarding all the traffic from 53 to 5353. Ended up finding the official documentation (https://openwrt.org/docs/guide-user/services/dns/adguard-home) which basically just says to move DNSMasq to 5353 and run AdGuard native on 53. Been working amazingly so far. 4ms average processing time. This router is a beast.
To resolve it temporarily for me, I've interposed a dumb gbps switch between the E8450 and the slow devices, so the router only sees gbps ethernet ports now. That resolved the issue for me and allows me to track snapshot builds until a real bug fix is found for the original commit's error.
I am on a Snapshot from earlier this year (beginning of january). It has been very solid for me and normally i have no reason to update. Unfortunately there is a major CVE in the Linux kernel. I am not certain if this CVE compromises my system, but better safe than sorry...
I tried to sysupgrade, but it is giving me the following error: opkg_install_cmd: Cannot install package iptables.
What should i do? Build from latest snapshot and reflash?
Ok. Just to clarify. I build from last snapshot on my linux system and flash via Flash new firmware image. Do i have to enter recovery mode or flash the "recovery firmware" before flashing the firmware like when i did the first install?
What happens to my settings and installed packages when i flash the self-build firmware?
I just wanted to share a recent experience I had with the RT3200 that might be of use. I have been using openwrt on this Belkin since Jan2022. I used the guide from dangowrt and all seemed well until yesterday. Comcast service went down for a few hours in the morning. When it came back up my 5ghz radio was down. Everything in luci appeared to be normal. So I did a reboot through the GUI. When the router came back up it was in recovery mode. I tried to upload a new snapshot with no luck it would always boot into recovery. I went back to the install and tried the suggestion to ssh into the router to remove /sys/fs/pstore*. This also removed the luci GUI so after reinstalling it did boot into production.
I'm not sure why this happened I've had other outages and reboots over time with no issues. Maybe this was known on older snapshots ot a fluke but I thought it someone else comes across this it might help.
my first step was to install openwrt-mediatek-mt7622-linksys_e8450-ubi-initramfs-recovery-installer.itb from the stock firmware. it appears that this has installed an openwrt initramfs flat image.
i saved the mtd0,mtd1 mtd2, mtd3 images from this flash.
reading through the guide, it is later suggested to do a full backup by flashing openwrt-mediatek-mt7622-linksys_e8450-ubi-initramfs-recovery.itb from stock. however, it appears that the 4 mtd files that i saved from after my first openwrt flash are not really the original mtd - it appears that i should have first flashed the second file (recovery.itb, not recovery-installer.itb)
is there any hope for restoring original firmware now?
if not - i'll just proceed
suggest - to @daniel (dangowrt) - to reword the instructions jsut a little?? thanks - im looking forward to proceeding
