Bug#1059390: linux-image-6.5.0-5-amd64: Large number of missed packets reported by "e1000e" driver
Control: tags -1 + moreinfo
Hi Matthias,
On Sun, Dec 24, 2023 at 10:39:22AM +0000, Matthias Scheler wrote:
> Package: src:linux
> Version: 6.5.13-1
> Severity: normal
> Tags: upstream
>
> Dear Maintainer,
>
> I'm using a Shuttle DS20UV2 as my broadband router and firewall. After
> updating it from Debian 12.4 to Trixie yesterday one of its network
> interfaces started reporting a large and ever increasing number of
> missed packets:
>
> > ip -s link show eno1
> 3: eno1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1508 qdisc fq_codel state UP mode DEFAULT group default qlen 1000
> link/ether 80:ee:73:XX:XX:XX brd ff:ff:ff:ff:ff:ff
> RX: bytes packets errors dropped missed mcast
> 991043595 846955 0 0 3691 0
> TX: bytes packets errors dropped carrier collsns
> 890811461 793387 0 0 0 0
> altname enp0s31f6
>
> The number was much larger before I rebooted the machine to
> test out some kernel parameter changes.
>
> I originally had similar problems with Debian 12.x. But increasing
> the RX ring size from 256 to 1024 fixed the problem. Under Trixie
> neither this nor increasing the RX ring size to the maximum of 4096
> helps. I also tried to restrict the maximum C-state to 3 (down
> from 9) which again didn't help.
>
> The interface in question is using an MTU of 1508 bytes to allow
> an MTU of 1500 bytes on PPPoE session running over it. The broadband
> link's maximum speed is 1 Gbit/sec. As such the incoming packet rate
> can be high at times.
>
> There is another ehternet interface in this machine which is driven
> by the "igc" driver. It reports zero missed packets but it also not
> using mini jumbo frames.
And I assume you have the same as well with 6.6.8-1 from the newest
stable series as present in unstable?
If so I guess you are in a good position to report that upstream, but
please keep this downstream bug as well in the loop (or ping it about
updates).
Regards,
Salvatore
Reply to: