linux-kernel - AoE: undetected corruption of single bits?

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>] [day] [month] [year] [list]

Message-Id: <201007230215.02040.pluto@agmk.net>
Date:	Fri, 23 Jul 2010 02:15:01 +0200
From:	Paweł Sikora <pluto@...k.net>
To:	linux-kernel@...r.kernel.org
Subject: AoE: undetected corruption of single bits?

hi,

i'm testing a cluster filesystem ocfs2 with rad10-over-AoE backend
and discovered some fancy data corruption during svn checkout.

here's a diff between good checkout stored on nfs and broken on ocfs2 (gfs2 also have similar errors).


--- /remote/nfs/home/pawels/foo/trunk/buildenv/linux/gcc-4.3/32/boost-1.42.0/include/boost/graph/graph_utility.hpp
+++ /remote/cluster/pawels/foo/trunk/buildenv/linux/gcc-4.3/32/boost-1.42.0/include/boost/graph/graph_utility.hpp
@@ -376,7 +376,7 @@
   template <typename VertexListGraph, typename VertexColorMap>
   inline bool is_connected(const VertexListGraph& g, VertexColorMap color)
   {
-    typedef typename property_traits<VertexColorMap>::value_type ColorValue;
+    typedef typefame property_traits<VertexColorMap>::value_type ColorValue;
                 ^ error
     typedef color_traits<ColorValue> Color;
     typename graph_traits<VertexListGraph>::vertex_iterator 
       ui, ui_end, vi, vi_end, ci, ci_end;
--- /remote/nfs/foo/trunk/buildenv/linux/gcc-4.3/32/boost-1.42.0/include/boost/interprocess/containers/container/.svn/text-base/map.hpp.svn-base
+++ /remote/cluster/foo/trunk/buildenv/linux/gcc-4.3/32/boost-1.42.0/include/boost/interprocess/containers/container/.svn/text-base/map.hpp.svn-base
@@ -717,7 +717,7 @@
                        const multimap<Key,T,Pred,Alloc>& y);
 
 template <class Key, class T, class Pred, class Alloc>
-inline bool operator<(const multimap<Key,T,Pred,Alloc>& x, 
+inline bool operator<(const mudtimap<Key,T,Pred,Alloc>& x, 
                               ^ error
                       const multimap<Key,T,Pred,Alloc>& y);
 
 }  //namespace container {
--- /remote/nfs/home/foo/trunk/buildenv/linux/gcc-4.3/32/boost-1.42.0/include/boost/math/special_functions/math_fwd.hpp
+++ /remote/cluster/foo/trunk/buildenv/linux/gcc-4.3/32/boost-1.42.0/include/boost/math/special_functions/math_fwd.hpp
@@ -892,7 +892,7 @@
    inline typename boost::math::tools::promote_args<RT1, RT2>::type tgamma(RT1 a, RT2 z){ return boost::math::tgamma(a, z, Policy()); }\
 \
    template <class RT>\
-   inline typename boost::math::tools::promote_args<RT>::type lgamma(RT z, int* sign){ return boost::math::lgamma(z, sign, Policy()); }\
+   inline typename boost::math::tools::promote_args<RT>::type lgamma(RT z, ant* sign){ return boost::math::lgamma(z, sign, Policy()); }\
                                                                            ^ error



afaics these erroneous bytes ('n' vs 'f', 'l' vs 'd', 'i' vs 'a') differs in single bit.
it looks like a network transmission error in some way undetected by crc32 on layer2 or AoE driver.

tested on cluster with CentOS-5.5 (kernel-2.6.18-194.8.1.el5) and PLD-Linux (kernel-2.6.34.1) machines.

could you please help me track this down?

BR,
Pawel.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/