<div dir="ltr"><div class="gmail_extra"><div class="gmail_quote">2015-02-07 23:58 GMT+09:00 Jouni Malinen <span dir="ltr"><<a href="mailto:j@w1.fi" target="_blank">j@w1.fi</a>></span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">On Sat, Feb 07, 2015 at 01:41:34PM +0200, Jouni Malinen wrote:<br>
> It looks like adding just two bytes to the sockaddr_ll buffer is enough<br>
> to make valgrind not complain. I'm not sure how padding could really<br>
> cause this, though, taken into account how valgrind code for this is<br>
> implemented in pre_mem_read_sockaddr(). That implementation is not aware<br>
> of sockaddr_ll (maybe it would be a good idea to make it aware of<br>
> that..), but it seems to be checking the area following sa_family (i.e.,<br>
> sll_family in sockaddr_ll) in a way that should not go beyond the<br>
> 20-byte buffer regardless of how the fields are padded.<br>
<br>
</span>Well.. I was obviously looking at the trunk version of Valgrind rather<br>
than 3.10.0 that is used in Ubuntu 14.10. This was indeed broken and<br>
fixed after that release:<br>
<a href="https://bugs.kde.org/show_bug.cgi?id=342221" target="_blank">https://bugs.kde.org/show_bug.cgi?id=342221</a><br>
<br>
I confirmed that the issue does not show up with the current Valgrind<br>
trunk snapshot. As such, I'm dropping this l2_packet patch since this<br>
was caused by a Valgrind bug that has already been fixed (but not yet<br>
included in a formal release).<br></blockquote></div><br></div><div class="gmail_extra">Thank you and sorry for false report.<br><br></div></div>