CakeQOS CakeQOS-Merlin

[Wade Coxon]

I just tried a manual Trend signature update here, killed cake.
Manual restart of cake required.

[Edit] just repeated, yup, kills Cake.

You know you have to disable the built in QoS entirely before you enable cake?

Updating the signatures likely restarted QoS which would have overwritten the cake configuration.

 

[Adamm]

My signature update details are identical to yours.

Result:
[$] /opt/bin/cake-qos status general

#########################################################

CakeQOS-Merlin: Not running...

#########################################################


[*] Press Enter To Continue...

Please post a syslog snippet from before and after you check for updates and cake gets disabled.

 

[Treadler]

You know you have to disable the built in QoS entirely before you enable cake?

Updating the signatures likely restarted QoS which would have overwritten the cake configuration.

The Asus QoS is ‘Off’.

 

[Treadler]

Please post a syslog snippet from before and after you check for updates and cake gets disabled.

This is after signature check, & before manual restart:

Jul 16 20:01:50 rc_service: httpds 1301:notify_rc start_sig_check
Jul 16 20:01:50 custom_script: Running /jffs/scripts/service-event (args: start sig_check)
Jul 16 20:01:54 kernel: The For ALL DEVICES flag of Prof 1 has been set to ENABLE
Jul 16 20:01:54 BWDPI: fun bitmap = 47f

After manual restart of Cake:

Jul 16 20:07:49 CakeQOS-Merlin: Starting - ( 98Mbit | 40Mbit | besteffort | ack-filter-aggressive | ack-filter-aggressive )
Jul 16 20:07:49 kernel: Cpuidle Host Clock divider is disabled

 

[Treadler]

Please post a syslog snippet from before and after you check for updates and cake gets disabled.

There is no mention of Cake shutting down, but it happens.

 

[brummygit]

You know you have to disable the built in QoS entirely before you enable cake?

Updating the signatures likely restarted QoS which would have overwritten the cake configuration.

No built-in QoS has ever been enabled since the last nuclear reset, however I am running AiProtect.

This is after signature check, & before manual restart:

Jul 16 20:01:50 rc_service: httpds 1301:notify_rc start_sig_check
Jul 16 20:01:50 custom_script: Running /jffs/scripts/service-event (args: start sig_check)
Jul 16 20:01:54 kernel: The For ALL DEVICES flag of Prof 1 has been set to ENABLE
Jul 16 20:01:54 BWDPI: fun bitmap = 47f

After manual restart of Cake:

Jul 16 20:07:49 CakeQOS-Merlin: Starting - ( 98Mbit | 40Mbit | besteffort | ack-filter-aggressive | ack-filter-aggressive )
Jul 16 20:07:49 kernel: Cpuidle Host Clock divider is disabled

This is exactly the same as I see

 

[ttgapers]

No built-in QoS has ever been enabled since the last nuclear reset, however I am running AiProtect.


This is exactly the same as I see

Interesting. I have never and will never use any TM bloatware, so I definitely will not see this. @Adamm or any other users, if you do use AIProtection, can we confirm please? I suspect this might be another "not supported with AIP etc." until we can chase it down. I currently use Suricata for IPS/IDS.

Cheers. Good on you to open the issue, let's track and see if/how we can work pass this once it's confirmed as a reproducible issue (which it appears to be).

 

[Chris0815]

Interesting. I have never and will never use any TM bloatware, so I definitely will not see this. @Adamm or any other users, if you do use AIProtection, can we confirm please? I suspect this might be another "not supported with AIP etc." until we can chase it down. I currently use Suricata for IPS/IDS.

I have setup up a second 86U this week running Merlin 384.18, Diversion, Unbound, Cake and the native AI-Protection (no Suricata on this Router). And it is the same here:
as soon as I start a manual update of the Trend-Micro Signature, cake stops to work...

 

[brummygit]

Interesting. I have never and will never use any TM bloatware, so I definitely will not see this. @Adamm or any other users, if you do use AIProtection, can we confirm please? I suspect this might be another "not supported with AIP etc." until we can chase it down. I currently use Suricata for IPS/IDS.

Cheers. Good on you to open the issue, let's track and see if/how we can work pass this once it's confirmed as a reproducible issue (which it appears to be).

@dave14305 seems to have had the same challenge in FlexQoS. His change log lists

Trigger persistence check immediately after TrendMicro signature check/update (restart_wrs)

I think this could be the commit in question (I'm not a programmer) https://github.com/dave14305/FlexQoS/commit/cae6694cc11dabcf570c3883ff4013e22ba8ab96 but maybe @dave14305 can advise.

 

[Treadler]

FWIW, I use AiProtect.
Looks like that might be the issue?

 

[dave14305]

@dave14305 seems to have had the same challenge in FlexQoS. His change log lists


I think this could be the commit in question (I'm not a programmer) https://github.com/dave14305/FlexQoS/commit/cae6694cc11dabcf570c3883ff4013e22ba8ab96 but maybe @dave14305 can advise.

I only noticed it while helping someone with a sig update issue. But it wipes QoS and mangle rules.

 

[brummygit]

I only noticed it while helping someone with a sig update issue. But it wipes QoS and mangle rules.

Thanks

 

[rgnldo]

Interesting. I have never and will never use any TM bloatware, so I definitely will not see this. @Adamm or any other users, if you do use AIProtection, can we confirm please? I suspect this might be another "not supported with AIP etc." until we can chase it down. I currently use Suricata for IPS/IDS.

Cheers. Good on you to open the issue, let's track and see if/how we can work pass this once it's confirmed as a reproducible issue (which it appears to be).

Is it another incompatibility?

Work fine!

Select Status Option:
[1]  --> Download Status
[2]  --> Upload Status
[3]  --> General Status

[1-3]: 3

[$] /opt/bin/cake-qos status general

#########################################################

CakeQOS-Merlin: > Download Status:
qdisc cake 8004: dev ifb9eth0 root refcnt 2 bandwidth unlimited besteffort dual-dsthost nat wash ingress no-ack-filter split-gso rtt 100.0ms raw overhead 0 no-sce

CakeQOS-Merlin: > Upload Status:
qdisc cake 8001: dev eth0 root refcnt 2 bandwidth unlimited besteffort dual-srchost nat nowash no-ack-filter split-gso rtt 100.0ms raw overhead 0 no-sce

#########################################################


[*] Press Enter To Continue...

 

[Bensam123]

Can this be installed while FreshJr is still installed, but QoS is disabled? Also what would be the best settings for prioritizing gaming traffic?

 

[brummygit]

Can this be installed while FreshJr is still installed, but QoS is disabled? Also what would be the best settings for prioritizing gaming traffic?

Yes it can, however note the conversation above that Trend Micro signature updates stop cake-qos and it needs restarting afterwards. Test this at your own risk as it's generally not advised.

There is no traffic prioritisation in cake-qos, all devices and traffic streams receive a fair share of the bandwidth but with much improved latency etc. If you are in the position that your bandwidth is over subscribed requiring prioritisation to back off low priority demands leaving bandwidth for your gaming you are better off sticking with FreshJR, FlexQoS or standard Adaptive QoS. It might be worth a try though as many games have low bandwidth requirements, but need latency and jitter to be as low as possible.

 

[brummygit]

Is it another incompatibility?

I don't think so - it's just that the service restart caused by the Trend Micro update wipes out the changes made to apply cake. All that's needed is code to restart cake after this event.

Unfortunately my coding is not good enough to help, although it has been catered for in FlexQoS so should be possible.

 

[Kingp1n]

I have a dumb question...I noticed the new firmwares have a Gear Accelerator (Game Device priority) under the Games tab. This basically enables Traditional QoS to give priority to your gaming devices if you wish however, at this point Cake-QoS wouldn't work anymore correct?

 

[Sinner]

very nice... i assume the 250mb recomendation is due to cpu limitations? also is cake more or less cpu intensive than fq_codel? i think its higher isnt it?

 

[sfx2000]

very nice... i assume the 250mb recomendation is due to cpu limitations? also is cake more or less cpu intensive than fq_codel? i think its higher isnt it?

Not really - it's about the same... while I'm responding to a particular post, it's more the thread in general.

With SQM in general - flow offloading generally tends to break it - whether it's Broadcom CTF or Qualcomm NSS, or even NAT flow offloading in SW for supported kernels.

Basically anything that interferes with flow management is going to impact SQM - SQM need to manage all the flows - for AsusWRT on HND, one has to basically disable all the accelerators to play with Cake SQM

Did a fair amount of work/study on SQM approaches for both BSD and Linux...

 

[sfx2000]

Interesting read - https://www.kernel.org/doc/Documentation/networking/nf_flowtable.txt

enabling SW/HW flows breaks SQM at the moment with some distros...

Basically any fast path approaches will break SQM for now...