How can I go from Asuswrt-Merlin to OpenWrt?

[jbaugh]

I have an Asus RT-AC68U router currently running Asuswrt-Merlin FW 378.56_2. I'm interested in converting to the OpenWrt firmware since it apparently has features that allow some mitigation of the buffer bloat problem I'm experiencing. I'm pretty new to this issue of flashing router firmware and I'm not sure how to proceed. Can I simply perform a firmware upgrade through my router's "Firmware Upgrade" tab or do I need to do some intermediary step(s)?
Thanks.
John

 

[Nullity]

It would probably be much easier to use Tomato. It supports fq_CoDel/CoDel.

 

[cybrnook]

Or just enable QoS in Asus Merlin and set your upload and download caps to be 1-5Mbps, or so, slower than your MAX sustained available to fix buffer bloat.

For example if you have a 50/10 Internet line, and for argument sake thats what a speed test is giving you as well. Then set your QoS limits to 48/9

 

[jbaugh]

Or just enable QoS in Asus Merlin and set your upload and download caps to be 1-5Mbps, or so, slower than your MAX sustained available to fix buffer bloat.

For example if you have a 50/10 Internet line, and for argument sake thats what a speed test is giving you as well. Then set your QoS limits to 48/9

Thanks,
I have tried to set the QoS as you have suggested. But something is not working. I've toggled on the "Enable QoS button and selected "Adaptive QoS". My bandwidth is about 120 down/12 up. I've set the "upload Bandwidth" to 9 and the "Download Bandwidth" to 90 and hit "Apply". When I click the "Customize:" button to set priorities I just get a screen with "Highest" at the top and "Lowest" at the bottom. There are no categories to drag and drop. And when I test my bandwidth speeds at dslreports.com I still get download speeds of 120+ and uploads of 11+ with "F" rating for buffer bloat.
Any ideas?
John

 

[Zirescu]

Thanks,
I have tried to set the QoS as you have suggested. But something is not working. I've toggled on the "Enable QoS button and selected "Adaptive QoS". My bandwidth is about 120 down/12 up. I've set the "upload Bandwidth" to 9 and the "Download Bandwidth" to 90 and hit "Apply". When I click the "Customize:" button to set priorities I just get a screen with "Highest" at the top and "Lowest" at the bottom. There are no categories to drag and drop. And when I test my bandwidth speeds at dslreports.com I still get download speeds of 120+ and uploads of 11+ with "F" rating for buffer bloat.
Any ideas?
John

Make sure to reboot afterwards.

 

[Nullity]

AsusWRT does not include any AQM except RED (because of Linux kernel versions being too old?), so bufferbloat is still a concern that is best solved by another firmware that supports CoDel.

 

[sfx2000]

Respect OP's ask...

Either answer the question - which is how to migrate to OpenWRT?

Or don't...

 

[Nullity]

Respect OP's ask...

Either answer the question - which is how to migrate to OpenWRT?

Or don't...

Chill out bro. His primary focus was bufferbloat, not OpenWRT.

What, did you just read the topic and post a knee-jerk reaction without reading the body of his post? (I've done that... embarrassingly often.)

 

[sfx2000]

Chill out bro. His primary focus was bufferbloat, not OpenWRT.

What, did you just read the topic and post a knee-jerk reaction without reading the body of his post? (I've done that... embarrassingly often.)

Not upset, not at all...

He asked a basic question - and got everything but an answer - it's his choice after all...

 

[RMerlin]

I have an Asus RT-AC68U router currently running Asuswrt-Merlin FW 378.56_2. I'm interested in converting to the OpenWrt firmware since it apparently has features that allow some mitigation of the buffer bloat problem I'm experiencing. I'm pretty new to this issue of flashing router firmware and I'm not sure how to proceed. Can I simply perform a firmware upgrade through my router's "Firmware Upgrade" tab or do I need to do some intermediary step(s)?
Thanks.
John

You will have to check the OpenWRT documentation for this. Treat Asuswrt-Merlin exactly as if you were still running the stock firmware when it comes to flashing another third party firmware.

 

[Nullity]

You will have to check the OpenWRT documentation for this. Treat Asuswrt-Merlin exactly as if you were still running the stock firmware when it comes to flashing another third party firmware.

If there's a simple answer, what's the latest official version of the Linux kernel for MIPS & ARM AsusWRT firmwares?

The most applicable info I have is an old post by the most active fq_CoDel contributor, which defines Linux kernel version 3.10.36 to be the oldest to support fq_CoDel. (Though, I think I saw info that d-taht had backported to 2.6.x, but I could be wrong.)

 

[RMerlin]

If there's a simple answer, what's the latest official version of the Linux kernel for MIPS & ARM AsusWRT firmwares?

2.6.22.19 (BCM MIPS), 2.6.36 (BCM ARM). No idea what Atheros and Qualcomm-based models use.

The most applicable info I have is an old post by the most active fq_CoDel contributor, which defines Linux kernel version 3.10.36 to be the oldest to support fq_CoDel.

For *full* support you need one of the 3.xx kernels AND you need specific support in the Ethernet driver itself. You can get partial support in 2.6 however.

(Though, I think I saw info that d-taht had backported to 2.6.x, but I could be wrong.)

merlin@mint-dev ~/asuswrt $ git log qos
commit df007e360fc38de125bda0f70e6598edcde3a420
Author: Eric Sauvageau <rmerl@lostrealm.ca>
Date:   Tue Apr 12 00:03:36 2016 -0400

    iproute2-3.x: Updated to 3.19.0 (includes fq_codel support)

commit 4436e88c3e8cfeb38d5a7bb271755cde192c4fb6
Author: Eric Sauvageau <rmerl@lostrealm.ca>
Date:   Fri Apr 8 17:30:29 2016 -0400

    qos: Add support for codel and fq_codel to Traditionnal QoS and Bandwidth Limiter (nvram-controlable for now - experimental)

commit 8db76e615ee26aa61dfa38eaa62be23adb2ed9a7
Author: Eric Sauvageau <rmerl@lostrealm.ca>
Date:   Fri Apr 8 17:26:19 2016 -0400

    kernel-arm: Implement codel and fq_codel support
  
    2.6.36 patches developped by Kyle Sanderson, for Tomato.

commit 77191edbec0311a29c8cef496ff7e6ae35f6ead2
Author: Eric Sauvageau <rmerl@lostrealm.ca>
Date:   Fri Apr 8 17:22:34 2016 -0400

    kernel-arm: Implement vzalloc macro (backport from kernel.org upstream)

 

[Nullity]

2.6.22.19 (BCM MIPS), 2.6.36 (BCM ARM). No idea what Atheros and Qualcomm-based models use.



For *full* support you need one of the 3.xx kernels AND you need specific support in the Ethernet driver itself. You can get partial support in 2.6 however.

merlin@mint-dev ~/asuswrt $ git log qos
commit df007e360fc38de125bda0f70e6598edcde3a420
Author: Eric Sauvageau <rmerl@lostrealm.ca>
Date:   Tue Apr 12 00:03:36 2016 -0400

    iproute2-3.x: Updated to 3.19.0 (includes fq_codel support)

commit 4436e88c3e8cfeb38d5a7bb271755cde192c4fb6
Author: Eric Sauvageau <rmerl@lostrealm.ca>
Date:   Fri Apr 8 17:30:29 2016 -0400

    qos: Add support for codel and fq_codel to Traditionnal QoS and Bandwidth Limiter (nvram-controlable for now - experimental)

commit 8db76e615ee26aa61dfa38eaa62be23adb2ed9a7
Author: Eric Sauvageau <rmerl@lostrealm.ca>
Date:   Fri Apr 8 17:26:19 2016 -0400

    kernel-arm: Implement codel and fq_codel support

    2.6.36 patches developped by Kyle Sanderson, for Tomato.

commit 77191edbec0311a29c8cef496ff7e6ae35f6ead2
Author: Eric Sauvageau <rmerl@lostrealm.ca>
Date:   Fri Apr 8 17:22:34 2016 -0400

    kernel-arm: Implement vzalloc macro (backport from kernel.org upstream)

Wait... you recently added fq_codel support? Great!

Is this an AsusWRT-merlin exclusive feature, or does AsusWRT proper support it, or have any plan to (or any obvious reason to not add the feature?)

 

[RMerlin]

Wait... you recently added fq_codel support? Great!

Is this an AsusWRT-merlin exclusive feature, or does AsusWRT proper support it, or have any plan to (or any obvious reason to not add the feature?)

It's purely experimental at this point, and the experiment is pretty much stalled right now since Asus/Broadcom broke the kernel traffic classifier on all SDK7 devices. I don't know yet if the kernel change would conflict with the closed source components or not.

IF (and that's big one) the kernel modules are compatible with the closed source components and IF they actually are more than snake oil (something which I'll have to be convinced - all of those bufferbloat complains can already be resolved simply by properly configuring Adaptive QoS), then the plan would be to implement a third QoS method based on OpenWRT's SQM scripts (my initial tests were done using their simplest.qos and simple.qos SQMs).

But at this point it's still mostly theory, and until Asus/Broadcom fixes what they broke, it's impossible for me to fully test things out. That bug caps my 30 Mbps downstream to about 10 Mbps, which means my link hardly gets congested.

 

[Nullity]

It's purely experimental at this point, and the experiment is pretty much stalled right now since Asus/Broadcom broke the kernel traffic classifier on all SDK7 devices. I don't know yet if the kernel change would conflict with the closed source components or not.

IF (and that's big one) the kernel modules are compatible with the closed source components and IF they actually are more than snake oil (something which I'll have to be convinced - all of those bufferbloat complains can already be resolved simply by properly configuring Adaptive QoS), then the plan would be to implement a third QoS method based on OpenWRT's SQM scripts (my initial tests were done using their simplest.qos and simple.qos SQMs).

But at this point it's still mostly theory, and until Asus/Broadcom fixes what they broke, it's impossible for me to fully test things out. That bug caps my 30 Mbps downstream to about 10 Mbps, which means my link hardly gets congested.

Are these (proprietary?) problems mitigated by disabling CTF?

 

[john9527]

Are these (proprietary?) problems mitigated by disabling CTF?

Last time this came up in discussion, it was pretty sure it would only work with CTF disabled. As Merlin alluded, Tomato pulled it in a while back, but they don't have to worry about CTF. Not sure if that would have changed.

 

[RMerlin]

CTF would be the least of my worries, as it can easily be disabled (it already has to be disabled with Traditional QoS anyway, as packet marking occurs in netfilter). The potential issue lies in the kernel symbol changes that can occur when you change kernel settings - this is what is preventing me from enabling IPSEC support in the kernel for instance. The kernels symbol changes prevent the DPI modules from loading due to symbol version mistmatches.

Tomato only has to worry about the Paragon FS modules and the wireless driver - everything else is open sourced. I have to deal with Tuxera and Trend Micro on top of these.

 

[kvic]

It's a big loss to disable CTF. It helps network traffic at reduced CPU load in every way, including LAN <> WLAN and WLAN <> WLAN traffic..

 

[Nullity]

It's a big loss to disable CTF. It helps network traffic at reduced CPU load in every way, including LAN <> WLAN and WLAN <> WLAN traffic..

I thought it was primarily a routing (WAN<->LAN) enhancement, not a switching enhancement.

Regardless, disabling CTF might be a big loss from a maximum throughput perspective, but there are many who are more concerned with functionality rather than maximum throughput; Many traffic monitoring solutions, especially tcpdump, fail when CTF is enabled.

CTF has it's pros & cons. It's not a cure-all.

 

[sfx2000]

I thought it was primarily a routing (WAN<->LAN) enhancement, not a switching enhancement.

Regardless, disabling CTF might be a big loss from a maximum throughput perspective, but there are many who are more concerned with functionality rather than maximum throughput; Many traffic monitoring solutions, especially tcpdump, fail when CTF is enabled.

CTF has it's pros & cons. It's not a cure-all.

The biggest challenge with CTF is that it's not very well documented without signing an NDA with Broadcom - and as a result, it makes things very difficult for 3rd party independent developers to code around it.