kernel: in6_delayed_cksum: delayed m_pullup

[Dries Michiels]

Tom,

I’m using the igb (intel I210) driver for my LAN interface and the em (intel I219-V) driver for the WAN interface.
I use this FreeBSD box as my home server so I wouldn’t say that it has a heavy ipv6 workload.

Regards,
Dries

From: Tom Jones
Sent: zondag 29 april 2018 21:05
To: Dries Michiels
Cc: freebsd-net at freebsd.org
Subject: Re: kernel: in6_delayed_cksum: delayed m_pullup

On Sun, Apr 29, 2018 at 08:30:28PM +0200, Dries Michiels wrote:
> Dear mailing list,
> 
> After upgrading my FreeBSD server from source from:
> FreeBSD 11.1-STABLE (VADOS) #9 r331859: Sun Apr  1 12:09:18 CEST 2018 
> to
> FreeBSD 11.2-PRERELEASE (VADOS) #10 r333091: Sun Apr 29 16:48:44 CEST 2018
> 
> My /var/log/messages is getting spammed by the following notice/error:
> Apr 29 19:51:42 vados kernel: in6_delayed_cksum: delayed m_pullup, m->len: 48 plen 68 off 56 csum_flags=400<CSUM_IP6_TCP>
> Apr 29 19:55:34 vados last message repeated 11 times
> Apr 29 20:11:56 vados last message repeated 10 times
> Apr 29 20:12:42 vados last message repeated 4 times
> 
> Does anyone have a clue what this indicates?

An error it seems:

     if (offset + sizeof(u_short) > m->m_len) {
         printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
             "csum_flags=%b\n", __func__, m->m_len, plen, offset,
             (int)m->m_pkthdr.csum_flags, CSUM_BITS);
         /*
          * XXX this should not happen, but if it does, the correct
          * behavior may be to insert the checksum in the appropriate
          * next mbuf in the chain.
          */
         return;
     }

> I did not have this message on my older system (r331859).

What network hardware are you using? 

Do you have a heavy ipv6 workload that might be a source?

- [tj]