article.xml revision 109400
1<!-- 
2	FreeBSD errata document.  Unlike some of the other RELNOTESng
3	files, this file should remain as a single SGML file, so that
4	the dollar FreeBSD dollar header has a meaningful modification
5	time.  This file is all but useless without a datestamp on it,
6	so we'll take some extra care to make sure it has one.
7
8	(If we didn't do this, then the file with the datestamp might
9	not be the one that received the last change in the document.)
10
11-->
12
13<!DOCTYPE article PUBLIC "-//FreeBSD//DTD DocBook V4.1-Based Extension//EN" [
14<!ENTITY % man PUBLIC "-//FreeBSD//ENTITIES DocBook Manual Page Entities//EN">
15%man;
16<!ENTITY % authors PUBLIC  "-//FreeBSD//ENTITIES DocBook Author Entities//EN">
17%authors;
18<!ENTITY % mlists PUBLIC "-//FreeBSD//ENTITIES DocBook Mailing List Entities//EN">
19%mlists;
20<!ENTITY % release PUBLIC "-//FreeBSD//ENTITIES Release Specification//EN">
21%release;
22]>
23
24<article>
25  <articleinfo>
26    <title>&os;
27    &release.next;
28    Errata</title>
29
30    <corpauthor>
31    The &os; Project
32    </corpauthor>
33
34    <pubdate>$FreeBSD: head/release/doc/en_US.ISO8859-1/errata/article.sgml 109400 2003-01-16 23:58:28Z bmah $</pubdate>
35
36    <copyright>
37      <year>2000</year>
38      <year>2001</year>
39      <year>2002</year>
40      <year>2003</year>
41      <holder role="mailto:doc@FreeBSD.org">The FreeBSD Documentation Project</holder>
42    </copyright>
43  </articleinfo>
44
45  <abstract>
46    <para>This document lists errata items for &os; 
47      &release.next;,
48      containing significant information discovered after the release.
49      This information includes security advisories, as well as news
50      relating to the software or documentation that could affect its
51      operation or usability.  An up-to-date version of this document
52      should always be consulted before installing this version of
53      &os;.</para>
54
55    <para>This errata document for &os; 
56      &release.next;
57      will be maintained until the release of &os; 5.1-RELEASE.</para>
58  </abstract>
59
60  <sect1 id="intro">
61    <title>Introduction</title>
62
63    <para>This errata document contains <quote>late-breaking news</quote>
64      about &os;
65      &release.next;.
66      Before installing this version, it is important to consult this
67      document to learn about any post-release discoveries or problems
68      that may already have been found and fixed.</para>
69
70    <para>Any version of this errata document actually distributed
71      with the release (for example, on a CDROM distribution) will be
72      out of date by definition, but other copies are kept updated on
73      the Internet and should be consulted as the <quote>current
74      errata</quote> for this release.  These other copies of the
75      errata are located at <ulink
76      url="http://www.FreeBSD.org/releases/"></ulink>, plus any sites
77      which keep up-to-date mirrors of this location.</para>
78
79    <para>Source and binary snapshots of &os; &release.branch; also
80      contain up-to-date copies of this document (as of the time of
81      the snapshot).</para>
82
83    <para>For a list of all &os; CERT security advisories, see <ulink
84      url="http://www.FreeBSD.org/security/"></ulink> or <ulink
85      url="ftp://ftp.FreeBSD.org/pub/FreeBSD/CERT/"></ulink>.</para>
86
87  </sect1>
88
89  <sect1 id="security">
90    <title>Security Advisories</title>
91
92    <para>No active security advisories.</para>
93
94  </sect1>
95
96  <sect1 id="late-news">
97    <title>Late-Breaking News</title>
98
99    <para>The &man.geom.4;-based disk partitioning code in the kernel
100      will not allow an open partition to be overwritten.  This
101      usually prevents the use of <command>disklabel -B</command> to
102      update the boot blocks on a disk because the
103      <literal>a</literal> partition overlaps the space where the boot
104      blocks are stored.  A suggested workaround is to boot from an
105      alternate disk, a CDROM, or a fixit floppy.</para>
106
107    <para>When using disk media with sector sizes larger than 512
108      bytes (for instance, &man.gbde.4; encrypted disks), the
109      &man.dump.8; program fails to respect the larger sector size and
110      cannot dump the partition.  One possible workaround is to copy
111      the entire file system in raw format and dump the copy.  It is,
112      for instance, possible to dump a file system stored in a regular
113      file:</para>
114
115      <screen>&prompt.root; <userinput>dd if=/dev/ad0s1d.bde of=/junk/ad0.dd bs=1m</userinput>
116&prompt.root; <userinput>dump 0f - /junk/ad0.dd | ...</userinput></screen>
117
118    <para>A simpler workaround is to use &man.tar.1; or &man.cpio.1;
119      to make backup copies.</para>
120
121    <para>Hangs have been reported during &os; 5.0 snapshot
122      installations when installing to &man.mly.4;-supported RAID
123      arrays, in hardware configurations that appear to work fine
124      under &os; 4.7-RELEASE.</para>
125
126    <para>NETNCP and nwfs appear to be as-yet unadapted for KSE, and
127      hence not working.</para>
128
129    <para>During installation, the &man.iir.4; controller appears to
130      probe correctly, but finds no disk devices.</para>
131
132    <para>&man.truss.1; appears to contain a race condition during the
133      start-up of debugging, which can result in &man.truss.1; failing
134      to attach to the process before it exists.  The symptom is that
135      &man.truss.1; reports that it cannot open the &man.procfs.5;
136      node supporting the process being debugged.  A bug also appears
137      to exist wherein &man.truss.1; will hang if &man.execve.2;
138      returns <literal>ENOENT</literal> A further race appears to
139      exist in which &man.truss.1; will return <errorname>PIOCWAIT:
140      Input/output error</errorname> occasionally on startup.  The fix
141      for this sufficiently changes process execution handling that it
142      has been deferred until after 5.0.</para>
143
144    <para>Some bugs have been reported in &man.sysinstall.8; disk
145      partitioning.  One observed problem on the i386 is that
146      &man.sysinstall.8; cannot recalculate the free space left on a
147      disk after changing the type of an FDISK-type partition.</para>
148
149    <para>In some case, documentation (such as the FAQ or Handbook)
150      has not been updated to take into account &os; &release.current;
151      features.  Examples of areas where documentation is still
152      needed include &man.gbde.8; and the new <quote>fast
153      IPsec</quote> implementation.</para>
154
155    <para>Attempting to unmount smbfs shares may fail with
156      <errorname>Device busy</errorname> errors even when the
157      mount-point is not really busy.  A workaround is to keep trying
158      to unmount the share until it eventually succeeds.  This bug has
159      been fixed in 5.0-CURRENT.</para>
160
161    <para>Forcefully unmounting (<command>umount -f</command>) smbfs
162      shares may cause a kernel panic.  This bug has been fixed in
163      5.0-CURRENT.</para>
164
165    <para>When called on a connected socket file descriptor,
166      &man.fstat.2; is supposed to return the number of bytes
167      available to read in the <varname>st_size</varname> member of
168      <varname>struct stat</varname>. However,
169      <varname>st_size</varname> is always erroneously reported as
170      <literal>0</literal> on TCP sockets.  This bug has been fixed in
171      5.0-CURRENT.</para>
172
173    <para>The &man.kqueue.2; <literal>EVFILT_READ</literal> filter
174      erroneously indicates that <literal>0</literal> bytes are
175      available to be read on TCP sockets, regardless of the number of
176      bytes that are actually available. The
177      <literal>NOTE_LOWAT</literal> flag for
178      <literal>EVFILT_READ</literal> is also broken on TCP sockets.
179      This bug has been fixed in 5.0-CURRENT.</para>
180
181    <para>&os; 5.0-RELEASE introduced support for POSIX named semaphores
182      but the implementation contains a critical bug that causes
183      &man.sem.open.3; to incorrectly handle the opening of the same
184      semaphore multiple times by the same process, and that causes
185      &man.sem.close.3; to crash calling programs.  This bug has been
186      fixed in 5.0-CURRENT.</para>
187
188    <para>&os; 5.0-RELEASE has a minor bug in how the permissions of
189      <filename>/dev/tty</filename> are handled.  This can be
190      triggered by logging in as a non-<username>root</username>,
191      non-<groupname>tty</groupname> group user, and using &man.su.1;
192      to switch to a second non-<username>root</username>,
193      non-<groupname>tty</groupname> group user.  &man.ssh.1; will
194      fail because it cannot open <filename>/dev/tty</filename>.  This
195      bug has been fixed in 5.0-CURRENT.</para>
196
197    <para>&man.growfs.8; no longer works on &man.vinum.4; volumes (and
198      presumably, on &man.geom.4; entities) since these subsystems no
199      longer fake disklabels, but &man.growfs.8; insists on examining
200      a label.</para>
201
202  </sect1>
203</article>
204