netdev - TX on r8169 causes latency and messes up real-time apps, on some boards

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 for Android: free password hash cracker in your pocket

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

Message-ID: <cb0375e10908160339i4526b957nf73b0576f19da2bd@mail.gmail.com>
Date:	Sun, 16 Aug 2009 06:39:37 -0400
From:	Andrew Lutomirski <luto@....edu>
To:	romieu@...zoreil.com
Cc:	netdev@...r.kernel.org
Subject: TX on r8169 causes latency and messes up real-time apps, on some 
	boards

Hi-

I'm having a bit of a mystery problem.  I have a bunch of identical
machines (Intel D201GLY2A + Trendnet TEG-PCITXR) which all work
wonderfully.

I have four brand-new machines that should be basically identical,
except that they are Intel D201GLY2 boards to which I have attached
fans myself (that's supposed to be the only difference), and for which
I bought the Trendnet TEG-PCITXR cards much later.  On those four new
machines, transmitting with the Trendnet card breaks my real-time app.

These are 2.6.30.1, although I've reproduced the behavior on
2.6.29.something.  All machines in question are running bitwise
identical kernels and userspace.

I've diffed the dmesg results.  The good machines differ only in
BogoMIPS (tiny difference), MAC addresses (obviously), and RTC time.
The difference from good to bad is attached and also shows which r8169
device I have.  (The USB stuff at the bottom is the real-time device
in question.)

The working machines do have a newer BIOS, and the only difference I
see that might make a difference is that the PWM fan control changed,
which might affect SMI latency, but I don't see why this should only
matter when sending data over the network.

Any ideas?  I'm happy to test things.

Thanks,
Andy

View attachment "dmesg_diff.txt" of type "text/plain" (12991 bytes)