<html>
<head>
<base href="http://llvm.org/bugs/" />
</head>
<body><table border="1" cellspacing="0" cellpadding="8">
<tr>
<th>Bug ID</th>
<td><a class="bz_bug_link
bz_status_NEW "
title="NEW --- - getgrnam_r interceptor can crash"
href="http://llvm.org/bugs/show_bug.cgi?id=19734">19734</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>getgrnam_r interceptor can crash
</td>
</tr>
<tr>
<th>Product</th>
<td>compiler-rt
</td>
</tr>
<tr>
<th>Version</th>
<td>unspecified
</td>
</tr>
<tr>
<th>Hardware</th>
<td>PC
</td>
</tr>
<tr>
<th>OS</th>
<td>Linux
</td>
</tr>
<tr>
<th>Status</th>
<td>NEW
</td>
</tr>
<tr>
<th>Severity</th>
<td>normal
</td>
</tr>
<tr>
<th>Priority</th>
<td>P
</td>
</tr>
<tr>
<th>Component</th>
<td>compiler-rt
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>davem@iabyn.com
</td>
</tr>
<tr>
<th>CC</th>
<td>llvmbugs@cs.uiuc.edu
</td>
</tr>
<tr>
<th>Classification</th>
<td>Unclassified
</td>
</tr></table>
<p>
<div>
<pre>This relates to sanitizer_common_interceptors.inc @r208530.
The interceptor for getgrnam_r() (and possibly others, including getgrid_r())
can crash, because it unpoisons the fields of the struct group even when no
matches have been found.
getgrnam_r(...., &result) is documented to return zero on lack of error (even
including no matches), and to set result to a non-NULL value if a match occurs.
The interceptor only checks for !res and not for result before unpoisoning.
On my system, the fields of grp on no-match happen to point to the last entry
in /etc/group, so this is harmless; on another system (which I don't have
access to)
and which has 'group: compat' in nsswitch.conf (which may or may not be
relevant), the fields of grp on no-match contain garbage. Here's what gdb shows
just after the return from the libc getgrnam_r() call on that system:
$3 = {gr_name = 0xbebebebebebebebe <Address 0xbebebebebebebebe out of bounds>,
gr_passwd = 0xbebebebebebebebe <Address 0xbebebebebebebebe out of bounds>,
gr_gid = 3200171710, gr_mem = 0xbebebebebebebebe}
So clearly the getgrnam_r interceptor shouldn't call unpoison_group() (and
possibly some other actions too?) unless result is non-NULL.
Also I think the places where it checks for result being non-null should be
checking *result????
This was originally discovered in the test suite for the perl interpreter;
under ASan on one person's test server, it was consistently crashing when doing
a getgrnam on an unknown name on a perl built using getgrnam_r() (i.e. a
threaded interpreter).
The original thread can be seen here:
<a href="http://www.nntp.perl.org/group/perl.perl5.porters/2014/05/msg215176.html">http://www.nntp.perl.org/group/perl.perl5.porters/2014/05/msg215176.html</a></pre>
</div>
</p>
<hr>
<span>You are receiving this mail because:</span>
<ul>
<li>You are on the CC list for the bug.</li>
</ul>
</body>
</html>