pfSense

Perhaps it is obvious, but it looks like the calls to "codel_alloc(100, 5, 0);" in one/all of the "altq_codel.diff" need to change to "codel_alloc(5, 100, 0);".

I think there is already a (newer) patch merged, but according to a few sources, the patch is not working.

https://github.com/pfsense/pfsense-tools/commit/3108a902bd816036a3abffd3ec669767140891a7

Revert it until someone can do a proper fix?

My assumption that this bug was so simple to fix that I could forgoe building pfSense to confirm the patch's functionality was plain stupid. No more code submissions from me until I can build pfSense reliably.

Ben Cook wrote:

I think there is already a (newer) patch merged, but according to a few sources, the patch is not working.

https://github.com/pfsense/pfsense-tools/commit/3108a902bd816036a3abffd3ec669767140891a7

Revert it until someone can do a proper fix?

My assumption that this bug was so simple to fix that I could forgoe building pfSense to confirm the patch's functionality was plain stupid. No more code submissions from me until I can build pfSense reliably.

Which patch you say is not working? commit:3108a902bd or the one you submitted?

Did you try a recent 2.2.4 snapshot? Did you get same problem?

Renato Botelho wrote:

Ben Cook wrote:

I think there is already a (newer) patch merged, but according to a few sources, the patch is not working.

https://github.com/pfsense/pfsense-tools/commit/3108a902bd816036a3abffd3ec669767140891a7

Revert it until someone can do a proper fix?

My assumption that this bug was so simple to fix that I could forgoe building pfSense to confirm the patch's functionality was plain stupid. No more code submissions from me until I can build pfSense reliably.

Which patch you say is not working? commit:3108a902bd or the one you submitted?

Did you try a recent 2.2.4 snapshot? Did you get same problem?

commit:3108a902bd is mine from github, I think.

I will try out a snapshot.

As far as I can see, the interval is correctly set to 100 on both the 20150721 and 20150719 nightly builds, the target can be set using 'queue limit'.

Kieran Cawthray wrote:

As far as I can see, the interval is correctly set to 100 on both the 20150721 and 20150719 nightly builds, the target can be set using 'queue limit'.

I just checked 20150708 nightly and the interval is 100. I also tested 2.2.3 release and got an interval of 5, so it was fixed somewhere between 20150625 and 20150708, probably commit 3108a902bd816036a3abffd3ec669767140891a7 on 20150629?

Good to hear.

There are two different methods of employing codel.

1. Where codel is the one and only scheduling discipline for a single parent queue with no other sub-queues.

2. Where codel is a sub-discipline ("Codel Active Queue" check-box) under one of the primary schedulers (HFSC, CBQ, PRIQ, FAIRQ).

The pfctl command in my original post will show the values of codel in a method-1 situation, but does anyone know how to display codel's configuration values when codel is a sub-discipline of another scheduler like in method-2?

The codel code (reviewed and) committed in FreeBSD is now backported to 2.3.