Red Hat Enterprise Linux 3 Update 1 *Beta* Availability Announcement

[Note: this message contains an attached text file.]

Red Hat is pleased to announce availability of our *beta* release of 
Update 1 for Red Hat Enterprise Linux 3 family of products, including:
 - Red Hat Enterprise Linux AS for x86, Itanium, AMD64, 
	IBM iSeries and pSeries, IBM zSeries, and IBM S/390
 - Red Hat Enterprise Linux ES for x86
 - Red Hat Enterprise Linux WS for x86, Itanium, and AMD64

The beta update packages and CD images are intended for testing purposes 
only, and will be followed up with official, supported update packages 
delivered as official product errata when testing is complete.  Red Hat 
may, at any time, change the contents of any or all of these beta updates.
The beta testing period is scheduled to continue through January 9, 2003.

Please remember that this beta software is not supported and 
is not intended for production environments!  Please do not publish
any benchmark or performance results based on this beta release.  
Upgrades will *not* be supported from the beta update to the 
subsequent official update; a fresh base install will be 
required.

This message includes details on obtaining the early access software, 
reporting bugs, and communicating with Red Hat and other testers via 
mailing lists during the early access period.

Of special interest to application developers, a note on application
compatibility with Red Hat Enterprise Linux 3 Update 1 is attached
to this message.

This update includes updates to the following packages:
anaconda, ant, compat-gcc, compat-libstdc++, cpp, devlabel, 
ethereal, freeradius, gcc, gdb, glibc, glibc-utils, gnupg, 
hotplug, httpd, hwdata, initscripts, iproute, jaf, javamail, 
junit, kdegames, kernel, kernel-utils, kinput2-canna-wnn6, 
kudzu, libf2c, libgcc, libgcj, libgnat, libobjc, libstdc++, 
mdadm, modutils, mod_ssl, net-snmp, netdump, netdump-server, 
nptl-devel, nscd, nss_ldap, openssl, oprofile, pstack, 
quagga, redhat-config-kickstart, redhat-config-network, 
redhat-config-network, redhat-config-packages, 
redhat-config-xfree86, rh-postgresql, rsync, shadow-utils, 
strace, up2date, XFree86, ypbind.

For specific package changelogs and version levels, consult the
architecture-specific channel package lists on Red Hat Network.

Accessing the Software
----------------------

The beta update packages are available via Red Hat Network
(https://rhn.redhat.com) to all current Red Hat Enterprise Linux
customers.  The updates are available via two methods:
 
(a) Install from new ISO images

Once you have logged in to RHN, click on the "Software" tab and choose
the appropriate child channel entitled "<product name> Beta Updates".  
Click on the 'Downloads' subtab, and follow the directions 
provided to download installable ISO images for your platform and
release.  

For Red Hat Enterprise Linux 3 Update 1 Beta, an updated
installation CD1 is provided with installer updates and updated
packages, which can be used with CD2-9 of the original Red Hat 
Enterprise Linux 3 ISO set.

(b) Update existing test installation using RHN beta channel

Once you have logged in to RHN, click on the "Software" tab and choose
the appropriate child channel entitled "<product name> Beta
Updates".  Click on the "Target Systems" subtab, and select all of the 
non-production, test systems you wish to run in the beta errata 
program.  Click on the "subscribe" button to apply the 
selection(s).

Once the beta channel has been subscribed, run the following command 
from a shell prompt on each target system to update any packages for
which new versions exist:

  up2date

You will see a list of packages that are available to be installed.
Select all packages and step through the instructions. Note: you will
optionally be able to skip installation of the kernel. Do not skip the
kernel!  Be sure to select it as well for installation.

For more information, Red Hat Network user guides and FAQs are 
available on the RHN website at https://rhn.redhat.com.

* Note: Systems that are updated with the beta update will not be 
upgradeable to the official update.  A fresh install will be 
required to ensure support! *

Bug Reporting
-------------

To ensure accurate tracking and follow-up, all bugs should be reported
using Bugzilla at:

  http://bugzilla.redhat.com

If you do not have a bugzilla account, please visit this URL and 
create one.
 
If you already have a bugzilla account:
 
1.  On bugzilla's home page at http://bugzilla.redhat.com, choose the
"New" tab at the top of the web page.  Proceed to Step 3, Choosing 
Your Product by clicking that button at the bottom of the web page.
 
2.  Choose "Red Hat Enterprise Linux Public Beta".  Proceed to Step 4, 
Choosing Component by clicking that button at the bottom of the 
web page.
 
3.  Choose the component against which you want to report a problem,
such as kernel, glibc, etc.

4.  Scroll the page to Step 5, Choosing Version and Architecture.  In
the Version field, choose 'rhel3-update'.  Choose the appropriate
architecture.  Proceed to Step 6, Additional Info, by clicking that 
button at the bottom of the web page.
 
5.  Provide the information about the problem you're reporting by
entering information in the appropriate fields. Then click the "Open
Bugzilla Entry Form" button. In the summary of the bug, and in the
text describing your problem, please state clearly that you are using 
a *BETA* package and provide the full package versions of any components 
you are experiencing problems with (as beta packages may be updated 
during the beta period).  Check to make sure that all information is 
accurate and click the Commit button to submit your problem report.

Mailing List
------------

Red Hat has created a public mailing list for communication during 
this and all future Red Hat Enterprise Linux 3 beta programs. 

* [email protected]

  A public general discussion mailing list for beta testers of 
  Red Hat Enterprise Linux 3 updates.  Subscribe at:
    https://www.redhat.com/mailman/listinfo/taroon-beta-list

Thank you for participating in the Red Hat Enterprise Linux 3
Update 1 beta program. 

Sincerely,

The Red Hat Enterprise Linux Team

Notes on Application Compatibility in Red Hat Enterprise Linux 3 Update 1 Beta
------------------------------------------------------------------------------

The POSIX standard defines interfaces to lock files or parts of files to 
ensure that they are not concurrently changed by other programs. The fcntl()
interface with the F_GETLK, F_SETLK, and F_SETLKW interfaces can be used
for this purpose.  The standard defines that attempts to lock an already 
locked file from another process must fail.  Also part of POSIX is the 
lockf() interface which serves the same purpose.  Not in POSIX, but 
falling into the same category, is the flock() interface.

The original kernel in the initial Red Hat Enterprise Linux 3 release not 
only prevents locking of the file from other processes, it also prevents 
locking from other threads in the same process.  This is not correct and 
does not correspond to the behavior in the 2.6 Linux kernel and correct 
POSIX semantics.  The kernel update in Red Hat Enterprise Linux 3 Update 1 
introduces the correct behavior for programs using NPTL.  There is no 
impact to programs that use LinuxThreads nor to programs that are not 
multi-threaded.

A side effect of the original Red Hat Enterprise Linux 3 kernel is that 
the l_pid field, which is part of the flock structure which is returned 
by a call to fcntl() with the F_GETLK command, contains a kernel-internal 
number of the thread.  With the Red Hat Enterprise Linux 3 Update 1 kernel, 
this field contains the process ID for all threads of the process.

If a non-POSIX-compliant multi-threaded application uses the NPTL 
library (which is the default), and that application performs file 
locking with the fcntl() system call, it is possible that the application 
might not work as originally intended when the Red Hat Enterprise Linux 3 
Update 1 kernel is used.  Red Hat recommends that any such application be 
requalified on the updated kernel, and if a problem is encountered, 
one of the following courses of action should be taken:

1) Set the LD_ASSUME_KERNEL environment variable to 2.4.19 for the 
   application process to prevent it from using the NPTL library.  Note 
   that if the application doesn't reset the LD_ASSUME_KERNEL environment 
   variable, this setting is inherited by all child processes.

2) Modify the application so that it no longer expects behavior
   that is inconsistent with POSIX compliance.  This will improve the 
   program's overall portability.  

Red Hat believes that few (if any) multi-threaded applications will 
be negatively impacted by the file locking fixes contained in the 
updated kernel.

Similar problems exist with the kernel functionality which provides
credentials for transport for Unix sockets and the System V message
queues, semaphores, and shared memory.  In the first case, calls like

getsockopt (fd, SOL_SOCKET, SO_PEERCRED, &caller, &optlen)

return information in the structure 'caller' which is of type struct 
ucred.  This structure contains a field 'pid'.  In the original Red Hat 
Enterprise Linux 3 kernel, this field was filled with an kernel-internal 
value representing the thread which sent the data.  With Red Hat Enterprise 
Linux 3 Update 1, the field correctly contains the process ID, i.e., the 
same value for all threads in a process.

The struct msqid_s type, defined in <sys/msq.h>, which is used to get
information about a System V message queue, contains fields msg_lspid and
msg_lrpid.  Both fields previously contained a kernel-internal value, 
whereas now they correctly contain the PID of the process calling 
msgsnd() and msgrcv() respectively.

When the GETPID command, defined in <sys/sem.h>, is used with the 
semctl() function, the original Red Hat Enterprise Linux 3 kernel 
returned a kernel-internal value. As of Red Hat Enterprise Linux 3 
Update 1, the returned value is the process ID.

The struct shmid_s type, defined in <sys/shm.h>, which is used to get
information about a System V shared memory segment, contains fields
shm_lpid and shm_cpid.  Both fields previously contained kernel-internal
values, now they correctly contain the PID of the process performing the
last operation, and creating the segment, respectively.