Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - pfry

#1
Quote from: dmcgough on November 20, 2026, 06:40:45 PM[...]I don't want to suddenly start routing all of my actual traffic to 8.8.8.8 via WAN2[...]

You could include a policy routing rule in your enhancement request. In the mean time you could configure (or more) one manually, or ping an IP you're not using for other purposes (e.g. ping 8.8.4.4 if you use 8.8.8.8 as a resolver). I don't think a policy route for ping from the firewall specifically would be detrimental in any meaningful way. Then again, I'm just assuming configuration of such is a simple matter in pf, as I don't use policy routing.

Aside: It's a bit disappointing that Cloudflare uses 1.0.0.1, but does not name it "SOS". Aside aside: I keep seeing "gateway" as "getaway".
#2
As I understand it, "prio" is a queue setting (assuming that's the parameter you're matching); if you're matching a packet element, "tos" ("Match TOS / DSCP" in OPNsense) may be more appropriate. I haven't used it, though.
#3
26.7, 26.10 Series / Re: Speed drops?
November 19, 2026, 03:51:01 AM
Quote from: RouterGuy2019 on November 18, 2026, 09:50:20 PM[...]If you mean when it gets slow I will check the next time it happens.

For top, yes, best to look during the issue. The netstat counters will accrue and persist until cleared or the firewall is restarted.

QuoteCan diddling the offloads have any serious negative side effects?

Yes, but it depends on the driver, and in the case of Intel hardware, the NVM. (Many other devices use firmware loaded from the driver.) There is no actual fail-safe set of parameters (too many variables), but for the most part, disabling offloads is the safer bet. You might try digging through the giant i226 thread, for instance, for experiences/recommendations (although it's mostly firmware-related).

Re-reading...
Quote[...]But sometimes I have to reboot the modem to get the speed back to normal[...]

...could suggest a link- or load-related issue, but solving the problem by poking the modem doesn't tend to implicate the firewall. Still, it's best to rule it out as well as possible. Unfortunately the process for (successfully) troubleshooting ISP issues is usually  "customer determines the issue and presents it to the ISP".
#4
26.7, 26.10 Series / Re: Speed drops?
November 18, 2026, 09:36:42 PM
Diddling the offloads may be helpful. I'd do one at a time, and test. EEE is garbage, but it usually results in link failures, which should be pretty obvious. Can't hurt to disable it, and see if it makes any difference. I would not fix speed and duplex unless autonegotiation fails to deliver proper speed and duplex.

Any other symptoms? What does the CPU widget (or better, top from a shell) indicate? How about various netstat displays ("Interfaces: Diagnostics: Netstat" or shell) - any bad counters incrementing?
#5
Quote from: sam1970 on November 16, 2026, 11:09:48 PM[...]The Protectli VP3230 has only two Intel i226-V ports, but apparently supports PCIe expansion:[...]

Looks like it trades storage for expansion. I'm too paranoid to rely on a non-replaceable EMMC (for anything besides booting a tiny ARM); I suppose you could rig the x1 M.2 slot for SATA and the x4 for network, but that would not leverage your existing SSDs. It's an inherent disadvantage of the Atom: not much in the way of expansion. I'd be tempted to look for a surplus office-type PC with similar specs, but you may not be willing to pay the power and size penalty - it's tough to match the N300s watt-for-watt.

Then again, you could use the x4 M.2 for storage and the x1 for network if you just need a single 2.5Gb or multiple 1Gb interfaces. I didn't look at the accessory cost to accomplish that.
#6
Quote from: nero355 on November 10, 2026, 08:13:51 PM[...]none of those things changes anything to a Android/Apple/Linux/Windows Client ofcourse ;)

Ha! I forgot about those entirely. Trapped in my bubble (even though I'm using Win10).

Quote from: tonys on November 10, 2026, 10:56:45 PM[...]At this point, I'm wondering why the state table grows so large[...]

What protocol/service? TCP sessions tend to hang around; I usually have a bunch of TCP DNS sessions in TIME_WAIT, as DNS can be a spew. My old Fortigate had granular control of session timeouts, one of its few advantages. Not a functional issue, really - I mostly cut UDP DNS timeouts to reduce clutter when browsing the state table. I don't bother with OPNsense.

I'd expect sessions to be more of a memory issue than processor, and I wouldn't expect a problem with 16GB. Those session spikes may be more of a symptom than a cause - the protocol may be revealing. (I have static IPs and servers; I average a few hundred sessions, but lots of bots/scrapers/whatever can put that up to 5000 or so for hours or days. That doesn't even account for the constant stream of blocked packets.)
#7
Quote from: nero355 on November 10, 2026, 03:47:03 PM[...]so that's something we will all just have to learn to live with :)[...]

rc.conf:
ip6addrctl_enable = "YES"
ip6addrctl_policy = "ipv4_prefer"

OPNsense: "System: Settings: General" -> "Networking" -> "Prefer IPv4 over IPv6"

I also only use IPv4 only, with no IPv6 address assignments or discovery, and "net.inet6.ip6.auto_linklocal = 0", so I can't comment on effect.
#8
From your subject I figured you were talking about "Bind states to interface". Personally I'd dig into why normal "State Type: keep state" does not work for you. It suggests a client/server TCP issue, packet loss, and/or partial bypass of the firewall. Packet loss at least should be obvious - it kills performance. I can't think of a practical security exploit for sloppy state offhand, but that doesn't say much.
#9
That's a strange issue. I'd've thought the MAC would always pad short packets, as 64B has been the standard for... 50 years? Looking at it initially, I thought it might be a carrier extension issue, but apparently not.

I haven't noticed ARP issues (other than proxy issues, which do not result from lost requests). I do not currently have any Broadcom devices in use - just Intel and Realtek at the moment. I imagine the particular switch might be an issue as well.
#10
Quote from: abulafia on November 05, 2026, 08:23:44 PM[...]In my case, they are UDPv4 packets incoming on WAN and have "dstport 0" (which might explain why they are blocked).

In that case the "block all targeting port 0" automatic rules should be listed (unless disabled). Most unlabeled "block" logs are from the scrub (normalization, which may also be disabled) - in other words, the packet had some other issue, and the scrub precedes other rules. I don't know about unlabeled "pass" logs as above - that's a new one to me.
#11
Quote from: Brilliant Technologies Lt on November 02, 2026, 05:46:23 PM[...]Live View showed TCP/443 from the affected client to 95.100.110.24 blocked under 'Default deny / state violation rule'. TCP flags and exact correlation with a fresh connection attempt were not captured. Kernel logs also show 'pf: BAD state' and 'pf: loose state match' involving other clients; these are not proven related to this client's Lark failure.[...]

Looks like a failed session setup, i.e. no state = timeout. I'd start with the info (specifically TCP flags) from that state violation - it's easiest to view from the "i" in the Live View; it's also in the session log. A loose match may be normal (it's the default); state mismatches can stem from path inconsistencies (bypassing the firewall). Given the loose state default, bypassing the expected interface is generally less of an issue.
#12
Huh! I'd assumed they were looking at the TCP header Data Offset field (possible using netfilter) - I didn't look closely at the CVE. Makes it easy to filter (as I've always done by default).

For folks like me who sit in the GUI and are too lazy to grep, "Firewall: Diagnostics: Statistics": -> "rules", expand the rulesets, CTRL-f "opts". Should be one: "let out anything from firewall host itself".
#13
Interesting. From [configctl commands list?], "configctl configd actions", in particular "configctl service ...". I don't know if the GUI would appreciate being stopped and started... I'll let others find out the hard way.
#14
Hardware and Performance / Re: Semi parts bin new build
October 23, 2026, 01:19:34 AM
Quote from: Greg_E on October 22, 2026, 10:33:13 PM[...]New 10Gtek quad port is about $125usd.[...]

10GTek is a standard OEM - they usually modify their PC boards and heat sinks (from the Intel references), and their firmware is updateable using Intel's utilities. (YMMV, of course. Intel-based cards only; my experience is with one X550.)

Heh. I haven't seen a quad X710 for $125, though (unless $150 counts).

Edit: It looks like 8257x cards are still available. I'd avoid those for anything other than Linux.
#15
Hardware and Performance / Re: Semi parts bin new build
October 22, 2026, 07:21:41 PM
Quote from: opnsenseguy on October 22, 2026, 06:25:41 PM[...]B.  I do have SFP switch so can use SFP so any particular cards recommended?[...]

I'd say X710 for low power, 82599/X520/X540 for lower purchase price. 82598 is probably not worth buying these days; E810 is fairly expensive and consumes more power. DAC cables are cheap and your lowest power option. You need at least a v3x4 slot for the X710 for full bandwidth from two ports; v2x8 for the 82599s. As usual, be aware of clone vs. OEM vs. actual Intel.

You didn't mention bandwidth; I assume you're looking for a few (2-4) ports. For a gigabit-class setup, a couple of i210s or an i350 are fairly inexpensive and low power, and may be more suitable from a connectivity (internal and external) standpoint. Then again, if your SFP-equipped switch is managed and you plan to use it in any case, an SFP card might be a good option.

How about form factor and other parts?