Page 1 of 1

Poss crash raspbian bug with jumbo ping / long ping -s of 4000 or more

Posted: Mon Sep 11, 2017 1:11 pm
by CecilWard
I was logged in using ssh. I did an

ethconfig eth0 mtu 9000

that being the usual built-in wired ethernet, was all good,
and then

ping -s 4000 some_global_ipv4_address

and that either stuffed my ssh session, or killed the o/s, hung the process, I can't tell which, I couldn't recover from it anyway. I don't have a screen or keyboard. I power-cycled the pi to recover.

Could someone try reproducing this perhaps? - but on a non-headless system too. Which I physically can't do (due to disability - long story skipped - see my earlier posts).

And would someone be kind enough to report this as a bug in <x> as you think appropriate as I am not able to do so myself.

Pings of size 1800 (and ~2000 too iirc) were ok. It's not that there is a limit around 1500 anyway.

For what it's worth:
The ping was directed to a box in my own Lan. The eth0 interface was using a global routable IPv4 address assigned by dhcp and also various global routable ipv6 address obtained from prefix ra from my router as I have an ipv6 isp and a global /48. The ethernet i/f also had additional static IPv4 addresses as well assigned to eth0:0 eth0:1 these being in the and 169.254.*.* ranges plus a link-local ipv6 address.

Keywords long ping jumbo frames raspbian jessie ping -s

Re: Poss crash raspbian bug with jumbo ping / long ping -s of 4000 or more

Posted: Mon Dec 24, 2018 11:52 pm
by ka9q
I just encountered the same problem on one of my Pi3s. I set the MTU to 9200 (which seems to be something of an unofficial standard for jumbograms; my switches accept up to 9216 bytes with the MAC header attached) and hung the network interface by sending a single large ping.

My x86 systems running Debian Linux work fine with 9200 byte MTUs so I suspect a limitation of the specific Ethernet controller used by the Pi.