----------------------------------------------------------------------------------
@MSGID: <10m4gbu$2gva$1@dont-email.me> ec26f1e1
@REPLY: <10m2ru8$3gdmm$1@dont-email.me> 554ff2d8
@PID: PyGate 1.5.11
@TID: PyGate/Linux 1.5.11
@CHRS: CP1252 2
@TZUTC: 0000
@REPLYADDR tnp@invalid.invalid
@REPLYTO 3:633/10 UUCP
On 05/02/2026 19:44, Tauno Voipio wrote:
> On 5.2.2026 16.27, The Natural Philosopher wrote:
>> On 05/02/2026 14:04, Lars Poulsen wrote:
>>> On 2026-02-04, The Natural Philosopher <tnp@invalid.invalid> wrote:
>>>> If you remember I had constricted a bridge from wifi to ethernet to act
>>>> as a bridged access point. On a PI 4B as a test platform
>>>>
>>>> The problem was that whilst the bridge was reasonably OK accessing my
>>>> LAN, up to 90% packet loss was experienced when accessing the internet
>>>> via my edge router.
>>>>
>>>> Two further points have been established but the exact reason for the
>>>> behaviour still remains a mystery
>>>>
>>>> 1/. A friend with a Pi 5 attempted to duplicate the setup, could not
>>>> get
>>>> it to work and instead used the Network Manager GUI to set up a
>>>> (routed?) access point which worked ok. It turns out that you cannot
>>>> use
>>>> the GUI tool to set up a bridge at all. Only nmcli.
>>>>
>>>> 2/. After a long time with traceroutes and pings I realised that this
>>>> particular machine was the *only one wired directly to the router via a
>>>> single gigabit Ethernet cable*. Everything else went via an ancient
>>>> 100Mbps switch that I inherited from an office clearout. In a rash of
>>>> `well I tried everything else` I unplugged the Pi from the Gigabit
>>>> router socket and put it into the 100Mbps switch and bingo!... Pretty
>>>> decent internet performance. Yes extremely long transfers sometimes
>>>> fail, but its very useable
>>>>
>>>> What I cannot for the life of me understand is *why* this worked. The
>>>> same [Gigabit] link was involved in both local and Internet access.
>>>> The
>>>> only difference being that local access ALSO went through a 100Mbps
>>>> switch.
>>>>
>>>> If anyone can shed light on this I would appreciate it.
>>>>
>>>> If it matters, the router is a Draytek Vigor2762Vac running PPPoE
>>>> via an
>>>> Openrach ONT to an optical fibre for Internet and thence to the ISP.
>>>
>>> The link-level connection involves a negotiation handshake to find
>>> compatible parameters. You may read up on MII (Media Independent
>>> Interface). When the state machines in the MII part of the MAC block in
>>> the Ethernet part of your SoC chip encounters an MII state machine it
>>> has not seen before, there may be timing dependent glitches.
>>> One of our customers has an installation on a remote island where
>>> the link between a microprocessor in his seismic gear connected via an
>>> ethernet switch to our radio locks up every 6 to 12 months and needs a
>>> remote-triggered power cycle to reset. We suggested he try another
>>> switch next time he can get a service tech to the island.
>>>
>>> It is also possible that a port data rate of a Gigabit may occasionally
>>> cause bus contention on some internal data bus in the PI triggering
>>> a bus error, while 100Mbps avoids that contention. I have seen such
>>> bus contention cause glitches in memory controllers in a few systems
>>> over my career.
>>>
>>
>> Thank you for that.
>>
>> Since the pi was always connected via the gigabit and the router
>> hardware, and performed well when then routed by an external switch,
>> it seems unlikely that it was the PI<=>router link that was at fault.
>>
>> I am leaning more towards the router buffering Internet data into
>> large Gigabit Ethernet bursts that overwhelmed the Pi when it was
>> forwarding to wifi. Limiting the data rate to 100Mbps allowed the Pis
>> Ethernet to function well enough not to overload the wifi.
>>
>> It was only the Internet<=>Router<=gigabit=>Pi4<=wifi=>client that broke
>>
>> Without the wifi the ethernet was OK., Without the gigabit the wifi
>> was OK.
>>
>> Anyway I think we both agree that it is not something that can be
>> programmed around . I will test again when I get a Pi 5 and if it
>> still sucks, a wifi access point is not that expensive. The Pi is also
>> pretty crippled in wifi speed.
>>
>> I suspect the PI wifi hardware was never really designed for AP
>> usage: More for client access to a Wifi station.
>>
>> I may try adding a wifi dongle at some point
>>
>
> It may be as simple as the cable from your router to the Pi, if it
> is a different one than with the switch. Verify that you`re using
> a CAT6 patch cable.
>
Well the cable worked fine for all other traffic except wifi traffic *to
the internet* so I doubt that it is the problem.
> I`m running a Zyxel GS-1200 switch with gigabit ports to a Pi3B+
> and WLAN bridging without extra problems. The WLAN in Pi is not
> completely as good a radio than a dedicated WLAN base station,
> due to different antenna systems.
>
I am going to back burner this project until I can get a pi 5 (at a
reasonable price. RAM is just silly right now) and if it doesn`t work
properly use a dedicated wifi unit. I may try with a dedicated wifi-USB
dongle
My conclusion is that I had arrived at the correct software setup, but
the hardware or firmware was an issue. Since I can`t change either in
practical terms apart from trying a wifi dongle the matter tests where
it is.
--
Civilization exists by geological consent, subject to change without notice.
? Will Durant
--- PyGate Linux v1.5.11
* Origin: Dragon`s Lair, PyGate NNTP<>Fido Gate (3:633/10)
SEEN-BY: 19/10 50/109 153/757 218/840 840 220/70
221/1 6 360 226/17 100
SEEN-BY: 229/426 240/1120 267/800 301/1 113 812
310/31 335/364 341/66 463/68
SEEN-BY: 633/10 280 414 418 420 422 509 2744
712/848 770/1 3 100 340 350
SEEN-BY: 772/210 220 230 5019/40 5020/715 848 1042
4441 12000 5030/49 722
SEEN-BY: 5030/1081 1474 5053/55 5061/133 5075/128
@PATH: 633/10 280 770/1 218/840 221/6 301/1
5020/1042 4441