Hi all, I noticed that my ISP requires me to reduce my MTU to 1472. As my ethernet devices are used via bridge devices, I wonder if: A) it is enough to lower the MTU on the physical ethernet device on my linux box.. B) also lower the MTU of the related bridge device C) also required to lower the MTU of the other (virtual) devices connected to this bridge? Kind regards, Hans. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 01/04/2018 10:54 AM, suse@a-domani.nl wrote:
Hi all,
I noticed that my ISP requires me to reduce my MTU to 1472. As my ethernet devices are used via bridge devices, I wonder if:
A) it is enough to lower the MTU on the physical ethernet device on my linux box.. B) also lower the MTU of the related bridge device C) also required to lower the MTU of the other (virtual) devices connected to this bridge?
I assume you have a router/firewall between the ISP and your network. If so, you just change the WAN interface MTU to 1472 (that's an unusual size). Anything behind the firewall can continue to use 1500, if you wish, though there's no harm in reducing the MTU there as well. You can often do that with DHCP. Regardless, 1500 on the local network is not a problem, as IPv4 can use Path MTU Discovery or fragmentation to deal with the smaller MTU. IPv6 uses PMTUD only. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
suse@a-domani.nl wrote:
Hi all,
I noticed that my ISP requires me to reduce my MTU to 1472. As my ethernet devices are used via bridge devices, I wonder if:
A) it is enough to lower the MTU on the physical ethernet device on my linux box.. B) also lower the MTU of the related bridge device C) also required to lower the MTU of the other (virtual) devices connected to this bridge?
It's enough to lower the MTU on the uplink device. -- Per Jessen, Zürich (14.9°C) http://www.hostsuisse.com/ - dedicated server rental in Switzerland. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Thursday, 2018-01-04 at 17:05 +0100, Per Jessen wrote:
suse@a-domani.nl wrote:
Hi all,
I noticed that my ISP requires me to reduce my MTU to 1472. As my ethernet devices are used via bridge devices, I wonder if:
A) it is enough to lower the MTU on the physical ethernet device on my linux box.. B) also lower the MTU of the related bridge device C) also required to lower the MTU of the other (virtual) devices connected to this bridge?
It's enough to lower the MTU on the uplink device.
Yes, but that will cause a 1500 byte packet inside to be divided in two packets when going outside. More work for the router, and slowdown. The 1472 figure rings a fuzzy bell with old modems. :-? - -- Cheers, Carlos E. R. (from openSUSE 42.2 x86_64 "Malachite" at Telcontar) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iEYEARECAAYFAlpOhcoACgkQtTMYHG2NR9Vj/ACeP5IN2Yk6c6AJok0MCWpMYs5T 4RIAn3hoxtPuCtMcEGoe8BPx41baHsFt =EA+r -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On Thu, 4 Jan 2018 20:51:30 +0100 (CET)
"Carlos E. R."
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
On Thursday, 2018-01-04 at 17:05 +0100, Per Jessen wrote:
suse@a-domani.nl wrote:
Hi all,
I noticed that my ISP requires me to reduce my MTU to 1472. As my ethernet devices are used via bridge devices, I wonder if:
A) it is enough to lower the MTU on the physical ethernet device on my linux box.. B) also lower the MTU of the related bridge device C) also required to lower the MTU of the other (virtual) devices connected to this bridge?
It's enough to lower the MTU on the uplink device.
Yes, but that will cause a 1500 byte packet inside to be divided in two packets when going outside. More work for the router, and slowdown.
The 1472 figure rings a fuzzy bell with old modems. :-?
1472 is the maximum size ping packet that will fit inside a 1500 byte network packet, given the 20 byte IP header and 8 byte ICMP header (according to stackoverflow or https://support.aa.net.uk/MTU). 1492 for PPPoE, BTW -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 01/04/2018 02:51 PM, Carlos E. R. wrote:
It's enough to lower the MTU on the uplink device. Yes, but that will cause a 1500 byte packet inside to be divided in two packets when going outside. More work for the router, and slowdown.
With path MTU discovery, the packet size is automatically adjusted to accommodate smaller MTUs. It does this by a router sending a "too big" ICMP message back to the source, with the recommended MTU. Linux appears to use it for all traffic, but Window for TCP only. Of course, with IPv6, PMTUD is mandatory.
The 1472 figure rings a fuzzy bell with old modems. :-?
I remember 576 from the dial up modem days and 1492 for ADSL. I don't ever recall hearing 1472 until now. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Thursday, 2018-01-04 at 15:18 -0500, James Knott wrote:
On 01/04/2018 02:51 PM, Carlos E. R. wrote:
It's enough to lower the MTU on the uplink device. Yes, but that will cause a 1500 byte packet inside to be divided in two packets when going outside. More work for the router, and slowdown.
With path MTU discovery, the packet size is automatically adjusted to accommodate smaller MTUs. It does this by a router sending a "too big" ICMP message back to the source, with the recommended MTU. Linux appears to use it for all traffic, but Window for TCP only. Of course, with IPv6, PMTUD is mandatory.
Interesting.
The 1472 figure rings a fuzzy bell with old modems. :-?
I remember 576 from the dial up modem days and 1492 for ADSL. I don't ever recall hearing 1472 until now.
Maybe I got the figure wrong. - -- Cheers, Carlos E. R. (from openSUSE 42.2 x86_64 "Malachite" at Telcontar) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iEYEARECAAYFAlpOrAYACgkQtTMYHG2NR9VFwACeI9gg5sp2wlR6twnG2vqDbCBd iJoAoIYYsJy/vWQw8ueBQK7ZdFIU6jAy =s5h0 -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 01/04/2018 05:34 PM, Carlos E. R. wrote:
With path MTU discovery, the packet size is automatically adjusted to accommodate smaller MTUs. It does this by a router sending a "too big" ICMP message back to the source, with the recommended MTU. Linux appears to use it for all traffic, but Window for TCP only. Of course, with IPv6, PMTUD is mandatory. Interesting.
https://en.wikipedia.org/wiki/Path_MTU_Discovery
The 1472 figure rings a fuzzy bell with old modems. :-?
I remember 576 from the dial up modem days and 1492 for ADSL. I don't ever recall hearing 1472 until now. Maybe I got the figure wrong.
Perhaps it was 1492, as that is the usual for PPPoE, which is used for ADSL. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 01/04/2018 07:54 AM, suse@a-domani.nl wrote:
A) it is enough to lower the MTU on the physical ethernet device on my linux box.. B) also lower the MTU of the related bridge device C) also required to lower the MTU of the other (virtual) devices connected to this bridge?
I think you got enough info already through those all replies. So I'll just add some that was missed. 1) You want to adjust mtu size at the device interface where the size is not the default size, like 1500 for copper eth (BTW you can increase it to more than 9000 for fiber GigE) for below 2) reason. PMTUD is supposed to find out the smallest MTU on the path between end to end and set the packet size below it before sending packets down to the network. However, as described at Wikipedia [https://en.wikipedia.org/wiki/Path_MTU_Discovery] IPv4 based method often doesn't work due to the fact ICMP messages are blocked by some hops/network devices between end to end. Therefore effective PMTUD needs to rely on TCP based discovery. It most unlikely works for any UDP traffic. 2) If you set MTU lower at the link interface where it's not necessary, like at your PC, all packets use that link need to be fragmented if larger than the MTU size, such as packets to your printer, network storage device, etc. You should avoid that as much as possible not to slow your LAN down. Router/FW are designed to perform it most effectively (by hardware/ASIC) than anything else. Let them do their job. Toshi -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 2018-01-05 06:59, Toshi Esumi wrote:
On 01/04/2018 07:54 AM, suse@a-domani.nl wrote:
A) it is enough to lower the MTU on the physical ethernet device on my linux box.. B) also lower the MTU of the related bridge device C) also required to lower the MTU of the other (virtual) devices connected to this bridge?
I think you got enough info already through those all replies. So I'll just add some that was missed.
1) You want to adjust mtu size at the device interface where the size is not the default size, like 1500 for copper eth (BTW you can increase it to more than 9000 for fiber GigE) for below 2) reason. PMTUD is supposed to find out the smallest MTU on the path between end to end and set the packet size below it before sending packets down to the network. However, as described at Wikipedia [https://en.wikipedia.org/wiki/Path_MTU_Discovery] IPv4 based method often doesn't work due to the fact ICMP messages are blocked by some hops/network devices between end to end. Therefore effective PMTUD needs to rely on TCP based discovery. It most unlikely works for any UDP traffic.
2) If you set MTU lower at the link interface where it's not necessary, like at your PC, all packets use that link need to be fragmented if larger than the MTU size, such as packets to your printer, network storage device, etc. You should avoid that as much as possible not to slow your LAN down. Router/FW are designed to perform it most effectively (by hardware/ASIC) than anything else. Let them do their job.
Toshi
Yes I certainly got quite some responses. And I do remember that old modems required a value of 1492 (just 8 bytes), but my max payload is 1472. Reason for asking, is thatI used a SuSE box as a firewall. One leg to cable-modem (the one with the 1472 MTU) others to internal servers, clients or wifi-modem Besides that, on the machine itself I have some virtualized machines doing common tasks like dns So all other (internal) interfaces (and bridges) I kept at the default MTU 1500 bytes, But one physical interface should be changed for certain, but I wonder if the corresponding virtual-bridge device should also lower its mtu -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
suse@a-domani.nl wrote:
Reason for asking, is thatI used a SuSE box as a firewall. One leg to cable-modem (the one with the 1472 MTU) others to internal servers, clients or wifi-modem Besides that, on the machine itself I have some virtualized machines doing common tasks like dns
So all other (internal) interfaces (and bridges) I kept at the default MTU 1500 bytes, But one physical interface should be changed for certain, but I wonder if the corresponding virtual-bridge device should also lower its mtu
Your WAN interface is part of bridge? I am not sure, but I expect it should be sufficient to change the bridge MTU. -- Per Jessen, Zürich (10.3°C) http://www.hostsuisse.com/ - virtual servers, made in Switzerland. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 2018-01-05 11:10, Per Jessen wrote:
suse@a-domani.nl wrote:
Reason for asking, is thatI used a SuSE box as a firewall. One leg to cable-modem (the one with the 1472 MTU) others to internal servers, clients or wifi-modem Besides that, on the machine itself I have some virtualized machines doing common tasks like dns
So all other (internal) interfaces (and bridges) I kept at the default MTU 1500 bytes, But one physical interface should be changed for certain, but I wonder if the corresponding virtual-bridge device should also lower its mtu
Your WAN interface is part of bridge? I am not sure, but I expect it should be sufficient to change the bridge MTU.
eth1 is connected to the cable modem; fw6:~ # ifconfig eth1 eth1 Link encap:Ethernet HWaddr 00:08:C7:E6:9A:3D inet6 addr: fe80::208:c7ff:fee6:9a3d/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:8531668002 errors:0 dropped:0 overruns:1 frame:1 TX packets:4249592124 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:1000 RX bytes:11815554011517 (11268190.3 Mb) TX bytes:416408316189 (397117.9 Mb) fw6:~ # fw6:~ # ifconfig br1 br1 Link encap:Ethernet HWaddr 00:08:C7:E6:9A:3D inet addr:172.16.11.1 Bcast:172.16.11.255 Mask:255.255.255.0 inet6 addr: 2001:xyz:1f01:3785:1::1/80 Scope:Global inet6 addr: fe80::208:c7ff:fee6:9a3d/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:8498737383 errors:0 dropped:0 overruns:0 frame:0 TX packets:4249592140 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:0 RX bytes:11695013314138 (11153233.8 Mb) TX bytes:416408323489 (397117.9 Mb) fw6:~ # brctl show br1 bridge name bridge id STP enabled interfaces br1 8000.0008c7e69a3d no eth1 Currently both eth1 AND br1 have a MTU of 1500 -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
suse@a-domani.nl wrote:
On 2018-01-05 11:10, Per Jessen wrote:
suse@a-domani.nl wrote:
Reason for asking, is thatI used a SuSE box as a firewall. One leg to cable-modem (the one with the 1472 MTU) others to internal servers, clients or wifi-modem Besides that, on the machine itself I have some virtualized machines doing common tasks like dns
So all other (internal) interfaces (and bridges) I kept at the default MTU 1500 bytes, But one physical interface should be changed for certain, but I wonder if the corresponding virtual-bridge device should also lower its mtu
Your WAN interface is part of bridge? I am not sure, but I expect it should be sufficient to change the bridge MTU.
eth1 is connected to the cable modem; fw6:~ # ifconfig eth1 eth1 Link encap:Ethernet HWaddr 00:08:C7:E6:9A:3D inet6 addr: fe80::208:c7ff:fee6:9a3d/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:8531668002 errors:0 dropped:0 overruns:1 frame:1 TX packets:4249592124 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:1000 RX bytes:11815554011517 (11268190.3 Mb) TX bytes:416408316189 (397117.9 Mb)
fw6:~ # fw6:~ # ifconfig br1 br1 Link encap:Ethernet HWaddr 00:08:C7:E6:9A:3D inet addr:172.16.11.1 Bcast:172.16.11.255 Mask:255.255.255.0 inet6 addr: 2001:xyz:1f01:3785:1::1/80 Scope:Global inet6 addr: fe80::208:c7ff:fee6:9a3d/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:8498737383 errors:0 dropped:0 overruns:0 frame:0 TX packets:4249592140 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:0 RX bytes:11695013314138 (11153233.8 Mb) TX bytes:416408323489 (397117.9 Mb)
fw6:~ # brctl show br1 bridge name bridge id STP enabled interfaces br1 8000.0008c7e69a3d no eth1
Currently both eth1 AND br1 have a MTU of 1500
Hi Hans Having thought about it, I think you need to change the MTU of the physical device, eth1. Assuming you have physical access, I would just try it out ip link set mtu xxxx dev eth1 then check the device see if the mtu changes propagate to br1. The MTU is just as easily reset. /Per -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 01/05/2018 05:32 AM, suse@a-domani.nl wrote:
On 2018-01-05 11:10, Per Jessen wrote:
suse@a-domani.nl wrote:
Reason for asking, is thatI used a SuSE box as a firewall. One leg to cable-modem (the one with the 1472 MTU) others to internal servers, clients or wifi-modem Besides that, on the machine itself I have some virtualized machines doing common tasks like dns
So all other (internal) interfaces (and bridges) I kept at the default MTU 1500 bytes, But one physical interface should be changed for certain, but I wonder if the corresponding virtual-bridge device should also lower its mtu
Your WAN interface is part of bridge? I am not sure, but I expect it should be sufficient to change the bridge MTU.
eth1 is connected to the cable modem; fw6:~ # ifconfig eth1 eth1 Link encap:Ethernet HWaddr 00:08:C7:E6:9A:3D inet6 addr: fe80::208:c7ff:fee6:9a3d/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:8531668002 errors:0 dropped:0 overruns:1 frame:1 TX packets:4249592124 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:1000 RX bytes:11815554011517 (11268190.3 Mb) TX bytes:416408316189 (397117.9 Mb)
fw6:~ # fw6:~ # ifconfig br1 br1 Link encap:Ethernet HWaddr 00:08:C7:E6:9A:3D inet addr:172.16.11.1 Bcast:172.16.11.255 Mask:255.255.255.0 inet6 addr: 2001:xyz:1f01:3785:1::1/80 Scope:Global inet6 addr: fe80::208:c7ff:fee6:9a3d/64 Scope:Link UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:8498737383 errors:0 dropped:0 overruns:0 frame:0 TX packets:4249592140 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:0 RX bytes:11695013314138 (11153233.8 Mb) TX bytes:416408323489 (397117.9 Mb)
fw6:~ # brctl show br1 bridge name bridge id STP enabled interfaces br1 8000.0008c7e69a3d no eth1
Currently both eth1 AND br1 have a MTU of 1500
That's a strange situation. Does that bridge device actually talk to the ISP??? If not, get it off of there. Also, you shouldn't be using RFC1918 addresses on the WAN side, unless your ISP provides them. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 01/05/2018 04:49 AM, suse@a-domani.nl wrote:
So all other (internal) interfaces (and bridges) I kept at the default MTU 1500 bytes, But one physical interface should be changed for certain, but I wonder if the corresponding virtual-bridge device should also lower its mtu
Only if it's on the same physical interface as the WAN. Anything behind the firewall/router, no. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 01/05/2018 12:59 AM, Toshi Esumi wrote:
On 01/04/2018 07:54 AM, suse@a-domani.nl wrote:
A) it is enough to lower the MTU on the physical ethernet device on my linux box.. B) also lower the MTU of the related bridge device C) also required to lower the MTU of the other (virtual) devices connected to this bridge?
I think you got enough info already through those all replies. So I'll just add some that was missed.
1) You want to adjust mtu size at the device interface where the size is not the default size, like 1500 for copper eth (BTW you can increase it to more than 9000 for fiber GigE) for below 2) reason. PMTUD is supposed to find out the smallest MTU on the path between end to end and set the packet size below it before sending packets down to the network. However, as described at Wikipedia [https://en.wikipedia.org/wiki/Path_MTU_Discovery] IPv4 based method often doesn't work due to the fact ICMP messages are blocked by some hops/network devices between end to end. Therefore effective PMTUD needs to rely on TCP based discovery. It most unlikely works for any UDP traffic.
You can set MTU much bigger than 1500 on a lot of hardware. Most gigabit gear supports 9K or more on copper. Also, when PMTUD fails, another mechanism kicks in where it's assumed the failure is caused by a small MTU somewhere and so the MTU is set smaller so that it works. There is also provision for testing larger MTUs from that point.
2) If you set MTU lower at the link interface where it's not necessary, like at your PC, all packets use that link need to be fragmented if larger than the MTU size, such as packets to your printer, network storage device, etc. You should avoid that as much as possible not to slow your LAN down. Router/FW are designed to perform it most effectively (by hardware/ASIC) than anything else. Let them do their job.
If you set a smaller MTU on the local link, you won't get fragmentation, which you shouldn't see anyway. You'll just have a less efficient network.
Toshi
-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
On 01/05/2018 07:16 AM, James Knott wrote:
You can set MTU much bigger than 1500 on a lot of hardware. Most gigabit gear supports 9K or more on copper. Also, when PMTUD fails, another mechanism kicks in where it's assumed the failure is caused by a small MTU somewhere and so the MTU is set smaller so that it works. There is also provision for testing larger MTUs from that point.
You may be wondering about that 1500 byte limit on Ethernet. That only applies to 802.3, not Ethernet II, which IP uses. There is no hard limit to Ethernet II, so the MTU can be as big as the hardware supports. https://en.wikipedia.org/wiki/Ethernet_frame -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse+owner@opensuse.org
participants (6)
-
Carlos E. R. -
Dave Howorth -
James Knott -
Per Jessen -
suse＠a-domani.nl -
Toshi Esumi