Hi Samm, so I started working on something here https://github.com/f00b4r0/my-lede-source/tree/rbcfg

It's working as expected on my boards. I'll try to collect hexdumps from other devices' soft_config see if the tag is present there: if it isn't (and cpu frequency is not supported) then I think it's "problem solved": if I read the code correctly, rbcfg will simply ignore tags it can't find. If it is but shouldn't be used then we'll have to add a board detection routine. I have several ideas on how to do this, not sure which is best yet.

You'll notice that for now I have just marked the various options as "normal", "low1/2" and "high1/2" since the change will be dependent on the base frequency (clock ref). I think it's okay that way for now.

Let me know what you think!

Hi, i still think that its good idea to collect more data from other devices before making it GA. E.g. on one of my boards (493G) router allows 8 selections. I cant get dump of this partition because it is in prod.

(Last edited by samm on 26 Feb 2017, 20:11)

Yeah collecting more info is definitely the plan right now.
If you get a chance, can you maybe just look whether your 493G uses the same ID (0xc) for the flag? A simple hexdump -C /dev/mtdblockN where mtdblockN is the soft_config partition should be able to tell. It will also tell what is the "normal" value (hopefully it'll be 0x10).

Thanks!

Unfortunately it is in prod and running routeros, so i am not able to get mtd* content May be we should ask in lede list to get more data?

Definitely, I might help here with some dumps too.

Here's the dump of a rb493G:

root@hostname:/# hexdump -C /dev/mtd3
00000000  74 66 6f 53 1e 43 bb f6  00 04 00 0c 00 00 03 60  |tfoS.C.........`|
00000010  00 04 00 0a 00 00 00 02  00 04 00 01 00 00 00 00  |................|
00000020  00 04 00 02 00 00 00 02  00 04 00 03 00 00 00 01  |................|
00000030  00 04 00 04 00 00 00 00  00 04 00 05 00 00 00 01  |................|
00000040  00 04 00 07 00 00 00 00  00 04 00 09 00 00 00 00  |................|
00000050  00 04 00 0f 00 00 00 00  00 04 00 0d 00 00 00 00  |................|
00000060  00 08 00 06 32 2e 32 38  00 00 00 00 00 00 00 00  |....2.28........|
00000070  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
*
00001000

The tag is there and the format is apparently different. Looks like it's not going to be as easy after all

Email posted: http://lists.infradead.org/pipermail/le … 06355.html
Thanks!

I'm starting to see a pattern with some boards (the RB493G aside for now). I'm thinking I'll try emailing mikrotik see if they would willingly share the algorithm for this value computation (you never know ;-)
Otherwise I'll keep reverse-engineering it

(Last edited by f00b4r0 on 27 Feb 2017, 22:30)

So in next version they will encode boot loader config, haha ;-) I personally never had good conversation with most of the vendors, but you can try, of course.

(Last edited by samm on 28 Feb 2017, 13:42)

So they replied and they turned me down
I'll have tried.

Anyway I think I have enough data to support QCA953x and AR9344 devices for now. I'm going to update my github branch soon.

See PR #930

Looks good, thank you! I was thinking to read board type from the hw config partition, but this way is probably even better! What kind of reply it was?

Well the thing is this is not a per-board thing, it's a per-cpu one. And CPU type isn't in hard_config. Also adding support for hard_config accessors would have required a lot more code

As for the reply, it was "we can't provide you with that information" ;P

(Last edited by f00b4r0 on 15 Mar 2017, 21:58)

As expected. I am more and more thinking to experiment with some no-name boards on the same chipsets because of this stupid policy.

Can't blame you. I wouldn't recommend MikroTik hw at this point, and will stay away from it now.

(Last edited by f00b4r0 on 15 Mar 2017, 22:08)

BTW, a bit offtopic, but are you aware of some cheap alternative which can be used as CPE? Something like 32-64 RAM, 16 Flash, etc.

No specific recommendation at this point. I'll be looking for equivalents to the mAP series myself for a project I have.

Hi.
Recently tried to upgrade hAP lite:

/tmp# sysupgrade lede-ar71xx-mikrotik-rb-nor-flash-16M-squashfs-sysupgrade.bin 
Invalid image. Use *-sysupgrade.bin files on this board
Image check 'platform_check_image' failed.

Please, explain. Is it already possible to upgrade to next snapshot directly from LEDE already installed in device or I still should upgrade first booting from

lede-ar71xx-mikrotik-rb-nor-flash-16M-initramfs-kernel.bin

and only then to run sysupgrade ?
Thanks.

Image format was recently changed in the Lede because of many bugfixes and re-work on this platform. So i would really recommend to boot from the initramfs-kernel.bin. Next upgrade should run smoothly

I understood. Logically there is no another way. Very thanks for the answer.

Just wonted to pop by and say that i have a RB750 r2 (hEX lite) and is working lovely with LEDE snapshot build at the moment, I have tryed the stable LEDE 17.01.1 but no joy on this as it gived me non-working LAN ports!

At the moment the LEDE snapshot builds shows the USR LED is flashing away, Anyone eles have the same issue on other models? or is it becouse i still have it running in the RAM? (initramfs)

Edit:
I have decided to go for the sysupgrade bit, However i get this error but i waited for about 1 min and then i see the LEDs at the top come back on so i have no idea what went on? So would anyone recommend me to do a "mtd erase firmware" and then sysupgrade lede-ar71xx-mikrotik-rb-nor-flash-16M-squashfs-sysupgrade.bin etc?

I have unplugged my RB750 r2 heX lte from the power and back on again so is definitely stored, Installed LuCI fine no issues.

Reboot (SNAPSHOT, r4049-9412fc2) the snapshot build dated 27/04/2017

root@LEDE:/tmp# sysupgrade lede-ar71xx-mikrotik-rb-nor-flash-16M-squashfs-sysupgrade.bin
Cannot save config while running from ramdisk.
killall: watchdog: no process killed
Unlocking firmware ...
Erasing firmware ...
Sending TERM to remaining processes ... udhcpc odhcp6c ntpd dnsmasq ubusd askfirst logd netifd odhcpd
Sending KILL to remaining processes ... askfirst
Performing system upgrade...
Unlocking firmware ..
Writing from <stdin> to firmware ...
Appending jffs2 data from /tmp/sysupgrade.tgz to firmware..
.File /tmp/sysupgrade.tgz does not exist
Upgrade completed
Rebooting system...
umount: can't unmount /dev/pts: Resource busy
umount: can't unmount /dev: Resource busy
umount: can't unmount /tmp: Resource busy
umount: can't unmount /: Invalid argument

Edit
Everything seems to be fine according to LEDE bug system report
https://bugs.lede-project.org/index.php … ask_id=741

(Last edited by francisuk on 2 May 2017, 14:06)

I have tried to install LEDE and OpenWRT on a Mikrotik RB962 hAP ac.
It has (I think) made an initial boot from tftp on the latest
openwrt-ar71xx-mikrotik-vmlinux-initramfs.elf and on
lede-ar71xx-mikrotik-rb-nor-flash-16M-ac-initramfs-kernel.bin
but it has not answered ping on 192.168.1.1 and I cannot access it with telnet to 192.168.1.1.

Is it possible to run LEDE og OpenWRT on RB962 and what is the way to get it started ?

hello.
i have  RB941-2nD (hAP lite) with QCA9531 and try to add USB-mod support.
Can anyone give the backup on Hard and Soft partitions of other mikrotik which one has support of USB from the stock?

To narrow down your search a bit: https://lede-project.org/toh/views/toh_ … _*~%5D=2.0

Filtered for Mikrotik + QCA + USB.

Hello

Im trying to install openwrt on my hAP lite, i was able to load via dhcp/tftp  lede-17.01.4-ar71xx-mikrotik-vmlinux-initramfs.elf and able to access LUCI via 192.168.1.1   i was trying to flash the lede-17.01.4-ar71xx-mikrotik-nand-large-squashfs-sysupgrade.bin  but it is not accepting the .bin file and from what ive read its necessary to have a .trx file is there a way to convert the image file from windows?
Does anyone have a working .trx file that can be shared to help me complete the installation

or am i doing somehting wrong?

let me know!

I appreciate your time, have a good day.