Network extremely slow after upgrading from Bullseye to Bookworm (Intel I219-LM)

[DanielKr]

Dear all,

I am running a headless home server which I use as file server (Samba) and webserver (Docker running Nextcloud, Wordpress, ...). On Nov. 4, I upgraded from Debian Bullseye to Bookworm. Since then, the network is incredibly slow. Initially, I thought the root-cause would be my Samba configuration, since some Samba config files were changed during upgrade but it turned out, that all network services are impacted.

System info:
- Fujitsu D3644-B mainboard with onboard 1 GBit LAN
- Ethernet Intel I219-LM
- Kernel version: 6.1.0-13-amd64
- Debian version: Debian 12.2.0-14

Networking details:

Code: Select all

root@server:~# lspci 
(...)
00:1f.6 Ethernet controller: Intel Corporation Ethernet Connection (7) I219-LM (rev 10)

Code: Select all

root@server:~# lspci -nnk
(...)
00:1f.6 Ethernet controller [0200]: Intel Corporation Ethernet Connection (7) I219-LM [8086:15bb] (rev 10)
	DeviceName: Onboard - Ethernet
	Subsystem: Fujitsu Technology Solutions Ethernet Connection (7) I219-LM [1734:1248]
	Kernel driver in use: e1000e
	Kernel modules: e1000e

Code: Select all

root@server:~# ifconfig eno1
eno1: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>  mtu 1500
        inet 192.168.17.30  netmask 255.255.255.0  broadcast 192.168.17.255
        inet6 2003:fa:71b:e00:4e52:62ff:fe1b:1b0c  prefixlen 64  scopeid 0x0<global>
        inet6 fe80::4e52:62ff:fe1b:1b0c  prefixlen 64  scopeid 0x20<link>
        ether 4c:52:62:1b:1b:0c  txqueuelen 1000  (Ethernet)
        RX packets 49839558  bytes 22273690633 (20.7 GiB)
        RX errors 0  dropped 2869910  overruns 0  frame 0
        TX packets 41279564  bytes 44224604051 (41.1 GiB)
        TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0
        device interrupt 16  memory 0xa1200000-a1220000 

Observed impact:


- Samba: Copying files to and from the server from a Samba client is extremely slow (approx. 1 MB/s). This is independent from using a Linux or a Mac client. Also switching to SMB 2.1 protocol does not change anything.

- SCP: Copying files from the server via scp is equally slow as Samba

- iperf: Also extremely slow (client and server connected via LAN cable)

Code: Select all

daniel@daniel-ThinkPad-X1:~$ iperf -c 192.168.17.30
------------------------------------------------------------
Client connecting to 192.168.17.30, TCP port 5001
TCP window size: 16.0 KByte (default)
------------------------------------------------------------
[  1] local 192.168.17.20 port 33844 connected with 192.168.17.30 port 5001 (icwnd/mss/irtt=14/1448/729)
[ ID] Interval       Transfer     Bandwidth
[  1] 0.0000-11.2565 sec  12.6 MBytes  9.41 Mbits/sec

- FritzBox: I am using a FRITZ!Box 7530 AX for my home network. The ethernet connector (LAN2) is configured for "Power Mode" (1 Gbit/s). Interestingly, the FritzBox only recognizes the server (which is connected with a short approx. 40 cm cable) as 10 Mbit/s device!!! --> With Debian 11 (Bullseye) everything worked well so I suppose there must be some error with the LAN (Kernel) driver.

Does anyone have a suggestion what might have been broken in the upgrading process or is anyone aware of such issues with the Intel I219-LM driver?

Thanks for your help!

With best regards,
Daniel

[Aki]

Hello,

[..]
I am using a FRITZ!Box 7530 AX for my home network. The ethernet connector (LAN2) is configured for "Power Mode" (1 Gbit/s). Interestingly, the FritzBox only recognizes the server (which is connected with a short approx. 40 cm cable) as 10 Mbit/s device!!! --> With Debian 11 (Bullseye) everything worked well so I suppose there must be some error with the LAN (Kernel) driver.

Does anyone have a suggestion what might have been broken in the upgrading process or is anyone aware of such issues with the Intel I219-LM driver?

Welcome to the forum.

We had a similar topic in the past and it was probably related to a hardware issue (it was consistent on different operating systems):

• Intel I219-LM only 100 Mbit out of its 1 Gigabit

Take a look at it for some hints.

You could
1) check your system logs (systemd journal) for messages from the Linux kernel about your NIC.
2) use the ethtool command to change the configuration of your NIC.
3) set the debug level of the kernel module (e1000e) that drive you NIC to gather more debug informations; you can check the available parameter with the command:

Code: Select all

/sbin/modinfo e1000e

Hope this helps.

[DanielKr]

Hello,

@Aki : Thanks for your advice and link to the other forum topic.

In the meantime, I booted the old kernel from the previous Bullseye installation (5.10.0-26-amd64) to exclude the possibility that the kernel driver could be the root cause but the speed results were exactly the same like using the new kernel (6.1.0-13-amd64).

Using ethtools to get info on the ethernet connection confirmed that my server is only connected with 10Mb/s (as can be seen on the FritzBox).

Code: Select all

root@server:~# ethtool eno1
Settings for eno1:
	Supported ports: [ TP ]
	Supported link modes:   10baseT/Half 10baseT/Full
	                        100baseT/Half 100baseT/Full
	                        1000baseT/Full
	Supported pause frame use: No
	Supports auto-negotiation: Yes
	Supported FEC modes: Not reported
	Advertised link modes:  10baseT/Half 10baseT/Full
	                        100baseT/Half 100baseT/Full
	                        1000baseT/Full
	Advertised pause frame use: No
	Advertised auto-negotiation: Yes
	Advertised FEC modes: Not reported
	Speed: 10Mb/s
	Duplex: Full
	Auto-negotiation: on
	Port: Twisted Pair
	PHYAD: 1
	Transceiver: internal
	MDI-X: on (auto)
	Supports Wake-on: pumbg
	Wake-on: g
        Current message level: 0x00000007 (7)
                               drv probe link
	Link detected: yes

I then used the command from the mentioned post Intel I219-LM only 100 Mbit out of its 1 Gigabit and set the connection speed manually:

Code: Select all

 root@server:~# ethtool -s eno1 speed 1000 duplex full autoneg off

After that, connection speed is manually set to 1 Gb/s:

Code: Select all

root@server:~# ethtool eno1
Settings for eno1:
	Supported ports: [ TP ]
	Supported link modes:   10baseT/Half 10baseT/Full
	                        100baseT/Half 100baseT/Full
	                        1000baseT/Full
	Supported pause frame use: No
	Supports auto-negotiation: Yes
	Supported FEC modes: Not reported
	Advertised link modes:  1000baseT/Full
	Advertised pause frame use: No
	Advertised auto-negotiation: Yes
	Advertised FEC modes: Not reported
	Speed: 1000Mb/s
	Duplex: Full
	Auto-negotiation: on
	Port: Twisted Pair
	PHYAD: 1
	Transceiver: internal
	MDI-X: on (auto)
	Supports Wake-on: pumbg
	Wake-on: g
        Current message level: 0x00000007 (7)
                               drv probe link
	Link detected: yes

Both sides (Server and FritzBox) now indicate a speed of 1 Gb/s.

Running iperf again shows significant improvement. Also Samba is much faster now:

Code: Select all

daniel@daniel-ThinkPad-X1:~$ iperf -c 192.168.17.30
------------------------------------------------------------
Client connecting to 192.168.17.30, TCP port 5001
TCP window size: 16.0 KByte (default)
------------------------------------------------------------
[  1] local 192.168.17.20 port 45584 connected with 192.168.17.30 port 5001 (icwnd/mss/irtt=14/1448/603)
[ ID] Interval       Transfer     Bandwidth
[  1] 0.0000-10.0281 sec  1.10 GBytes   941 Mbits/sec

I will keep the topic open until I can confirm that the connection is stable. As of now it appears as if only the speed negotiation part is broken but if setting speed manually, everything works fine.

Best regards,
Daniel

[fabien]

We had a similar topic in the past and it was probably related to a hardware issue

Code: Select all

        RX packets 49839558  bytes 22273690633 (20.7 GiB)
        RX errors 0  dropped 2869910  overruns 0  frame 0
        TX packets 41279564  bytes 44224604051 (41.1 GiB)
        TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0

There is a lot of dropped received packets, should be 0.

Interestingly, the FritzBox only recognizes the server (which is connected with a short approx. 40 cm cable) as 10 Mbit/s device!!!

Short cables can cause problems, could you test with a longer cable? (and cross your fingers)

[Aki]

Hello,

[..]
I will keep the topic open until I can confirm that the connection is stable. As of now it appears as if only the speed negotiation part is broken but if setting speed manually, everything works fine.

Have you made any progress ?