netdev - 3.12-git Intel e1000e hardware unit hang / tx queue timeouts

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <52594DBD.3070108@tomt.net>
Date:	Sat, 12 Oct 2013 15:25:17 +0200
From:	Andre Tomt <andre@...t.net>
To:	netdev@...r.kernel.org
Subject: 3.12-git Intel e1000e hardware unit hang / tx queue timeouts

After some interesting commits in 3.12 I figured I'd try to run this 
system with the normal packet coalescing offloads enabled. Normally I 
disable them to get smoother packet scheduling. But some hours after 
turning GSO/GRO/TSO it is getting a bunch of tx timeouts and hardware 
unit hangs.

The port hanging is a 82579LM running with hfsc + fq_codel for egress, 
and redirecting ingress to a similarly configured ifb device. No 
ethernet flow control. This port is internet-facing. em2 in the dumps.

The internal port are only using fq_codel as root qdisc, no hfsc 
involved. Also an e1000e, 82574L. This port uses VLAN's, with one rarely 
used VLAN device on a bridge for KVM. The majority of traffic moves over 
a "clean" VLAN device. No ethernet flow control. em1 in the dumps.

Turning the offloads back off seems to cure it.

Kernel is linus git as of friday 11. oct (yesterday). I'm not sure when 
this surfaced as I've not been running with the offloads for a good 
while. The ports are on-board, motherboard is an Intel DQ77KB mini-itx. 
BIOS identifies as KBQ7710H.86A.0052.2013.0708.1336 (the latest one with 
the biosdevname fixes). More info is attached.

I'm going to boot 3.10.16 on it now, and see how it fares.

View attachment "qdisc.txt" of type "text/plain" (2036 bytes)

View attachment "lspci.txt" of type "text/plain" (29915 bytes)

View attachment "dmesg.txt" of type "text/plain" (72612 bytes)