[solved] 25.03.b.20250610.1659 re-enabling limiters leads to syslog kernel messages "update_fs ..."
-
Do you have Nexus/MIM enabled?
-
@stephenw10 Nope
-
@stephenw10 additionally, but also relating to Limiters, I'd like to make you aware of this possible regression as noted for 2.8.0: https://forum.netgate.com/topic/197859/2-8-0-limiter-rule-not-honored-on-lan-download-with-multiple-limiters-queues
I saw the same issue in 25.03 the other day, but today, after recreating the limiters yesterday, I get a different result. It is still failing though, just slightly differently, and I need to do some more testing before I can comfortably raise a bug on the issue.
-
@pst
I'm having odd issues with limiters too and I have not really chased it down yet due to other (recently resolved) issues.My thread was here and I'm going to look back into it:
https://forum.netgate.com/topic/197395/25-03-beta-bufferbloat-fq-codel-issues?_=1750539387871
️ -
@RobbieTT good, the more evidence we can gather the better. I don't think it's related to the new if_pppoe though as I don't use that (and neither would the 2.8.0 user in the thread I referenced above).
I plan to do some comparison tests between near default setups of 2.7.2, 2.8.0 and the current 25.03 in the next few days.
-
To summarize the results from the testing I've done so far on 25.03 (june10 beta):
Without the floating rule for buffer bloat prevention, limiters on LAN are working, both with policy routing and default.
With a floating rule for buffer bloat prevention configured:
- for LANs with policy routing, both UL and DL LAN limiters are disregarded
- for LANs without policy routing, LAN DL limit is disregarded, the LAN UL limit is adhered to
-
How is your floating rule defined?
Are you using if_pppoe?
-
@stephenw10 said in [solved] 25.03.b.20250610.1659 re-enabling limiters leads to syslog kernel messages "update_fs ...":
How is your floating rule defined?
as per the latest Netgate recipe
(ignore the fact the rule's got no states/bytes as it's been deactivated for a few days)
Are you using if_pppoe?
No.
IMHO, the problem resembles, on the surface at least, one from long ago: redmines 13026 and 14039
-
So pass - quick - outbound on WAN only?
Are you using pppoe at all? Or dhcp WAN?
-
@stephenw10 said in [solved] 25.03.b.20250610.1659 re-enabling limiters leads to syslog kernel messages "update_fs ...":
So pass - quick - outbound on WAN only?
yes, here's the rule:
[25.03-BETA][admin@felicity.local.lan]/root: pfctl -sr | grep -i buffer pass out quick on igb0 route-to (igb0 xxx.xxx.xxx.xx1) inet from xxx.xxx.xxx.xx3 to any flags S/SA keep state (if-bound) label "USER_RULE: From bufferbloat recipe" label "id:1750159398" label "gw:WAN_DHCP" ridentifier 1750159398 dnqueue(2, 1)
Are you using pppoe at all? Or dhcp WAN?
no pppoe, only dhcp v4/v6 on WAN
I should mention that I disabled IPv6 during the testing as to not interfere. Redmine #16201 mentions that the IPv6 rule needs to look slightly different as there is no NAT involved, so the source will not be WAN but rather the client's LAN address (so I skipped IPv6 for now)
