general question about lazy loading of shared libraries
by John Ellson
This is general linux question, rather than specifically FC4 devel, but
I'm hoping there are experts here that will know.
Are shared libraries on linux loaded lazily? i.e. not until the
application invokes a function in the library?
If so, how can I convince myself that it is working?
If I run my application in a way that should not use a particular .so,
how can I see that it is not loaded?
Thanks in advance for any advice.
John Ellson
19 years, 5 months
Self-Introduction: Orion Poplawski
by Orion Poplawski
Full Legal Name: Orion Edward Poplawski
Country, City: USA, Boulder, CO
Profession: System Administrator
Company: Northwest Research Associates, CoRA division
Goals:
Integration of various tools for scientific computing and clustering. I
currently maintain a modified Fedora distribution for in-house use.
Packages I've worked on include:
- octave-forge, updated octave
- python-gmt
- python-pyx
- python-opendap
- qhull
- ScientificPython
- hdf5
- ginac
I also maintain updated versions of amanda, pam, ypbind, logwatch solely
for internal use, but some changes I'd like to see upstream or more
widely tested.
Historical qualifications:
My name is easily googled :-) and you'll turn up plenty of boring posts
to various lists and newsgroups. I've submitted minor patches to
various open source projects including PHP, and filed plenty of bug
reports. I've been a Unix SysAdmin for just over 10 years, been
programming for over 23 years. I'm proficient in C, Perl, PHP,
(ba|k)sh, awk, sed. Familiar with C++, Fortran, IDL.
pub 1024D/D142DBFB 2004-11-11 Orion Poplawski <orion(a)cora.nwra.com>
Key fingerprint = 5C5A A7EF B441 C533 C656 A9AC 3919 F3AE D142 DBFB
sub 1024g/B7B050FE 2004-11-11
--
Orion Poplawski
System Administrator 303-415-9701 x222
Colorado Research Associates/NWRA FAX: 303-415-9702
3380 Mitchell Lane, Boulder CO 80301 http://www.co-ra.com
19 years, 5 months
exec-shield broken in newer kernels?
by utz lehmann
Hi
I just noticed that exec-shield looks broken in newer kernels.
All test are on i686 with disabled prelinking and
kernel.exec-shield = 1
kernel.exec-shield-randomize = 1
vm.legacy_va_layout = 0
A cat /proc/self/maps looks like this:
2.6.8-1.521smp (FC 2):
00cbf000-00cd4000 r-xp 00000000 03:02 327682 /lib/ld-2.3.3.so
00cd4000-00cd5000 r--p 00014000 03:02 327682 /lib/ld-2.3.3.so
00cd5000-00cd6000 rw-p 00015000 03:02 327682 /lib/ld-2.3.3.so
00dad000-00ec3000 r-xp 00000000 03:02 229387 /lib/tls/libc-2.3.3.so
00ec3000-00ec4000 r--p 00116000 03:02 229387 /lib/tls/libc-2.3.3.so
00ec4000-00ec7000 rw-p 00117000 03:02 229387 /lib/tls/libc-2.3.3.so
00ec7000-00ec9000 rw-p 00ec7000 00:00 0
08048000-0804c000 r-xp 00000000 03:02 459964 /bin/cat
0804c000-0804d000 rw-p 00003000 03:02 459964 /bin/cat
09dc8000-09de9000 rw-p 09dc8000 00:00 0
f6dc4000-f6dc5000 r--p 00b1e000 03:02 475909 /usr/lib/locale/locale-archive
f6dc5000-f6df8000 r--p 00aea000 03:02 475909 /usr/lib/locale/locale-archive
f6df8000-f6ff8000 r--p 00000000 03:02 475909 /usr/lib/locale/locale-archive
f6ff8000-f6ff9000 rw-p f6ff8000 00:00 0
feee0000-ff000000 rw-p feee0000 00:00 0
ffffd000-ffffe000 ---p 00000000 00:00 0
This is the familiar behavior. Only the exec mapping have the x bit and they
are randomized. The shared libs are placed under the binary.
2.6.9-1.667smp (FC 3):
00111000-00112000 r-xp 00a46000 03:01 640087 /usr/lib/locale/locale-archive
00112000-00113000 r-xp 00a99000 03:01 640087 /usr/lib/locale/locale-archive
002ef000-00304000 r-xp 00000000 03:01 704213 /lib/ld-2.3.3.so
00304000-00305000 r-xp 00014000 03:01 704213 /lib/ld-2.3.3.so
00305000-00306000 rwxp 00015000 03:01 704213 /lib/ld-2.3.3.so
00485000-00685000 r-xp 00000000 03:01 640087 /usr/lib/locale/locale-archive
0079c000-008bd000 r-xp 00000000 03:01 704943 /lib/tls/libc-2.3.3.so
008bd000-008bf000 r-xp 00120000 03:01 704943 /lib/tls/libc-2.3.3.so
008bf000-008c1000 rwxp 00122000 03:01 704943 /lib/tls/libc-2.3.3.so
008c1000-008c3000 rwxp 008c1000 00:00 0
009e8000-009e9000 rwxp 009e8000 00:00 0
00d8b000-00dbd000 r-xp 00a12000 03:01 640087 /usr/lib/locale/locale-archive
00f85000-00fb9000 r-xp 00a64000 03:01 640087 /usr/lib/locale/locale-archive
08048000-0804c000 r-xp 00000000 03:01 441615 /bin/cat
0804c000-0804d000 rwxp 00003000 03:01 441615 /bin/cat
0804d000-0806e000 rwxp 0804d000 00:00 0
feffe000-ff000000 rw-p feffe000 00:00 0
ffffe000-fffff000 ---p 00000000 00:00 0
Here all mappings except the stack have the x bit and they are randomized. I
think this will cause that the local-archive mappings are placed under the
binary too.
2.6.9-1.678_FC3 (FC 3):
08048000-0804c000 r-xp 00000000 03:01 6296799 /bin/cat
0804c000-0804d000 rwxp 00003000 03:01 6296799 /bin/cat
0804d000-0806e000 rwxp 0804d000 00:00 0
f6c37000-f6c38000 r-xp 00b71000 03:01 27280385 /usr/lib/locale/locale-archive
f6c38000-f6c6c000 r-xp 00b3c000 03:01 27280385 /usr/lib/locale/locale-archive
f6c6c000-f6c6d000 r-xp 00b1e000 03:01 27280385 /usr/lib/locale/locale-archive
f6c6d000-f6c9f000 r-xp 00aea000 03:01 27280385 /usr/lib/locale/locale-archive
f6c9f000-f6e9f000 r-xp 00000000 03:01 27280385 /usr/lib/locale/locale-archive
f6e9f000-f6ea0000 rwxp f6e9f000 00:00 0
f6ea0000-f6fc1000 r-xp 00000000 03:01 31481922 /lib/tls/libc-2.3.3.so
f6fc1000-f6fc3000 r-xp 00120000 03:01 31481922 /lib/tls/libc-2.3.3.so
f6fc3000-f6fc5000 rwxp 00122000 03:01 31481922 /lib/tls/libc-2.3.3.so
f6fc5000-f6fc7000 rwxp f6fc5000 00:00 0
f6fe9000-f6ffe000 r-xp 00000000 03:01 16798526 /lib/ld-2.3.3.so
f6ffe000-f6fff000 r-xp 00014000 03:01 16798526 /lib/ld-2.3.3.so
f6fff000-f7000000 rwxp 00015000 03:01 16798526 /lib/ld-2.3.3.so
feffe000-ff000000 rw-p feffe000 00:00 0
ffffe000-fffff000 ---p 00000000 00:00 0
Here all mappings including the executable mapping are placed top down. Only
the stack doesn't have the x bit and there is no randomization at all.
I think here is exec-shield mostly not working (except the non-exec stack).
With a segment based executable protection (no NX bit) the addressspace is
up to 0xf7000000 executable.
utz
19 years, 5 months
yum update problems on fc3
by Chapman, Matt
Any ideas on how to fix this? It happens everytime I run yum update
now.
Traceback (most recent call last):
File "/usr/bin/yum", line 8, in ?
yummain.main(sys.argv[1:])
File "/usr/share/yum-cli/yummain.py", line 51, in main
base.getOptionsConfig(args)
File "/usr/share/yum-cli/cli.py", line 133, in getOptionsConfig
self.conf = yumconf(configfile = yumconffile, root=root)
File "/usr/lib/python2.4/site-packages/yum/config.py", line 227, in
__init__
self._doFileRepo(fn)
File "/usr/lib/python2.4/site-packages/yum/config.py", line 299, in
_doFileRep
o
doRepoSection(self, repoconf, section)
File "/usr/lib/python2.4/site-packages/yum/config.py", line 313, in
doRepoSect
ion
mirrorurls = getMirrorList(mirrorlist)
File "/usr/lib/python2.4/site-packages/yum/config.py", line 390, in
getMirrorL
ist
fo = urlresolver.urlopen(url)
File "/usr/lib/python2.4/site-packages/urlgrabber/grabber.py", line
427, in ur
lopen
return default_grabber.urlopen(url, **kwargs)
File "/usr/lib/python2.4/site-packages/urlgrabber/grabber.py", line
555, in ur
lopen
return self._retry(opts, retryfunc, url)
File "/usr/lib/python2.4/site-packages/urlgrabber/grabber.py", line
527, in _r
etry
return apply(func, (opts,) + args, {})
File "/usr/lib/python2.4/site-packages/urlgrabber/grabber.py", line
554, in re
tryfunc
return URLGrabberFileObject(url, filename=None, opts=opts)
File "/usr/lib/python2.4/site-packages/urlgrabber/grabber.py", line
703, in __
init__
self._do_open()
File "/usr/lib/python2.4/site-packages/urlgrabber/grabber.py", line
747, in _d
o_open
fo, hdr = self._make_request(req, opener)
File "/usr/lib/python2.4/site-packages/urlgrabber/grabber.py", line
823, in _m
ake_request
fo = opener.open(req)
File "/usr/lib/python2.4/urllib2.py", line 364, in open
response = meth(req, response)
File "/usr/lib/python2.4/urllib2.py", line 468, in http_response
code, msg, hdrs = response.code, response.msg, response.info()
AttributeError: HTTPResponse instance has no attribute 'code'
Thanks,
Matt Chapman
--
Matthew Chapman
Network Engineer
Orange County Public Schools
"Hard work, sacrifice and focus will never show up in tests."
- Lance Armstrong
19 years, 5 months
rawhide report: 20041119 changes
by Build System
Updated Packages:
bash-3.0-24
-----------
* Thu Nov 18 2004 Tim Waugh <twaugh(a)redhat.com> 3.0-24
- Use upstream patch to fix bug #139575 and bug #139306.
* Thu Nov 18 2004 Tim Waugh <twaugh(a)redhat.com> 3.0-23
- Fixed last patch to avoid regressions (bug #139575).
file-4.10-3
-----------
* Thu Nov 18 2004 Radek Vokal <rvokal(a)redhat.com> 4.10-3
- set of patches from debian.org
- new magic types (#128763)
- zlib added to BuildReq (#125294)
* Tue Oct 12 2004 Tim Waugh <twaugh(a)redhat.com> 4.10-2
- Fixed occasional segfault (bug #131892).
* Wed Aug 11 2004 Radek Vokal <rvokal(a)redhat.com>
- zlib patch deleted, note patch deleted, rh patch updated, debian patch updated
- upgrade to file-4.10
grep-2.5.1-40
-------------
* Thu Nov 18 2004 Tim Waugh <twaugh(a)redhat.com> 2.5.1-40
- Fixed a bug in the fgrep patch, exposed by the dfa-optional patch
(bug #138558).
httpd-2.0.52-6
--------------
* Tue Nov 16 2004 Joe Orton <jorton(a)redhat.com> 2.0.52-6
- add security fix for CVE CAN-2004-0942 (memory consumption DoS)
- SELinux: run httpd -t under runcon in configtest (Steven Smalley)
- fix SSLSessionCache comment for distcache in ssl.conf
- restart using SIGHUP not SIGUSR1 after logrotate
- add ap_save_brigade fix (upstream #31247)
- mod_ssl: fix possible segfault in auth hook (upstream #31848)
- add htsslpass(1) and configure as default SSLPassPhraseDialog (#128677)
- apachectl: restore use of $OPTIONS
- apachectl, httpd.init: refuse to restart if $HTTPD -t fails
- apachectl: run $HTTPD -t in user SELinux context for configtest
- update for pcre-5.0 header locations
isdn4k-utils-3.2-19
-------------------
* Thu Nov 18 2004 Than Ngo <than(a)redhat.com> 3.2-19
- update cvs snapshot
- workaround, add capi devices
kdebindings-3.3.1-3
-------------------
* Thu Nov 18 2004 Than Ngo <than(a)redhat.com> 3.3.1-3
- rebuilt against python 2.4
* Mon Oct 18 2004 Than Ngo <than(a)redhat.com> 3.3.1-2
- rebuilt
libselinux-1.19.1-6
-------------------
* Thu Nov 18 2004 Dan Walsh <dwalsh(a)redhat.com> 1.19.1-6
- Add avcstat program
man-pages-ja-20041115-1
-----------------------
* Fri Nov 19 2004 Akira TAGOH <tagoh(a)redhat.com> - 20041115-1
- updates to 20041115.
mod_python-3.1.3-7
------------------
* Thu Nov 18 2004 Joe Orton <jorton(a)redhat.com> 3.1.3-7
- require python-abi
* Thu Nov 18 2004 Joe Orton <jorton(a)redhat.com> 3.1.3-6
- rebuild for Python 2.4
pcre-5.0-2
----------
* Thu Nov 18 2004 Joe Orton <jorton(a)redhat.com> 5.0-2
- include LICENCE, AUTHORS in docdir
- run make check
- move %configure to %build
* Thu Nov 18 2004 Than Ngo <than(a)redhat.com> 5.0-1
- update to 5.0
- change License: BSD
- fix header location #64248
perl-DBD-MySQL-2.9004-1
-----------------------
policycoreutils-1.18.1-3
------------------------
* Thu Nov 18 2004 Dan Walsh <dwalsh(a)redhat.com> 1.18.1-3
- Fix run_init.8 to refer to correct location of initrc_context
rdesktop-1.3.1-6
----------------
* Thu Nov 18 2004 Than Ngo <than(a)redhat.com> 1.3.1-6
- add cvs patch to make krdc working again
rpmdb-fedora-1:4-0.20041119
---------------------------
selinux-doc-1.15.1-1
--------------------
* Wed Jun 30 2004 Dan Walsh <dwalsh(a)redhat.com> 1.15-1
- Upgrade to match NSA
- Change installation directory to follow standards
selinux-policy-strict-1.19.2-1
------------------------------
* Thu Nov 18 2004 Dan Walsh <dwalsh(a)redhat.com> 1.19-2-1
- Upgrade to upstream
selinux-policy-targeted-1.19.2-1
--------------------------------
* Thu Nov 18 2004 Dan Walsh <dwalsh(a)redhat.com> 1.19-2-1
- Upgrade to upstream
setup-2.5.38-1
--------------
* Thu Nov 18 2004 Bill Nottingham <notting(a)redhat.com> 2.5.38-1
- fix bash/tcsh coredump size inconsistency (#139821)
system-config-users-1.2.28-1
----------------------------
* Wed Nov 10 2004 Nils Philippsen <nphilipp(a)redhat.com> - 1.2.28-1
- check for running processes of a user about to be deleted (#132902)
vnc-4.0-9
---------
* Wed Nov 17 2004 Tim Waugh <twaugh(a)redhat.com> 4.0-9
- Don't build hw/xfree86/etc.
- Prevent xorg-x11 build failure in rman.
- Build requires autoconf, flex, bison.
- Use correct PATH setting when started from initscript (bug #84167).
webalizer-2.01_10-26
--------------------
* Fri Nov 19 2004 Joe Orton <jorton(a)redhat.com> 2.0_10-26
- rebuild
19 years, 5 months
ls colours question
by Stuart MacDonald
I just installed FC3. I can't find a Fedora specific user list; if
this is not the appropriate place for my question, please point me on
my way.
ls colours in FC3 are much much darker than in previous RH releases.
It's almost as if my old RH 7.2 system was using light blue for
directories, and FC3 is using dark blue. I've compared the LD_COLORS
env variable, they are identical. I've compared the actual ASNI codes
that ls outputs (--color=always > file) and they are the same!
So that implies that the terminal is doing the colors differently.
I tried slogging through the termcap/terminfo, I really did. I was
unable to find anything that looked like a color table/specification/
layout/map.
Why is "blue" now "dark blue"? How can I change that so the new blue
is the same as the old blue (aka "light blue")?
..Stu
19 years, 5 months
anaconda ftp install
by Michael A. Peters
Since I am going to be installing fc3 in lots of different
configurations etc. for testing purposes, I set up an old iMac I had
lying around with YDL and set it up as a ftp server for install and
updates, figuring ftp install is the easiest way to do thing (no disk
swapping)
Two attempts at install failed. First attempt failed at installing
ImageMagick - so I tried again, wondering if maybe the ImageMagick rpm
on the server was bad. Second install failed earlier, with the kernel
(which had installed first go around)
It seems anaconda just gives up when it has downloaded a bad package.
Now maybe the nic or something on that iMac isn't perfect, but there's
a simple solution for remote installs - check the GPG sig of the
package. If it fails grab it again. That way a clean package is there
for install, even if it takes more than one fetch to get it.
Can this be added to Anaconda for FC4?
--
Cheap Linux CD's
http://mpeters.us/linux/
19 years, 5 months
ntpd and selinux error in fc3
by Farkas Levente
hi,
while the i start ntpd we'got (even after restorecon -v /usr/sbin/ntpd,
fixfiles relabel):
---------------------------------------------
audit(1100821314.712:0): avc: denied { read } for pid=6616
exe=/usr/sbin/ntpd name=mtab dev=hda2 ino=211823
scontext=root:system_r:ntpd_t tcontext=system_u:object_r:etc_runtime_t
tclass=file
audit(1100821314.713:0): avc: denied { read } for pid=6616
exe=/usr/sbin/ntpd name=meminfo dev=proc ino=-268435454
scontext=root:system_r:ntpd_t tcontext=system_u:object_r:proc_t tclass=file
---------------------------------------------
yours
--
Levente "Si vis pacem para bellum!"
19 years, 5 months
Remote bin client/server project (mainly for stateless, but could be used for any bin server)
by Alexandre Nault
Here is an idea I had recently about a new (?) project. Please post about any
comment, idea, suggestion, participation, feasability, usability, need for
it, ...
=========
The idea
=========
A client/server architecture for up to date bin distribution.
Server part: This is where the big stuff is. The latest up to date code for a
particular program/package (or any version, but the goal is to always have
the latest up to date version) would be on the server and the actual binary
would be produced on the fly depending on the options required by the client.
This would require either a collaboration from developpers or some other mean
(like rsync or ?). Of course there is a caching mechanism for not recompiling
the same binary/options combination all the time.
Client part: The client mount a directory (like binfs, libfs or sbinfs) with
either general options or program/package specific options. There is a
preconfigured default on the server side so the client only need to provide
the cpu arch. The client could also have a caching mechanism depending on the
configuration.
============
implementation
============
This could be implemented in a diskless client mounting a NFS root either on
the client side, or on the NFS server side.
It could also be implemented on any kind of workstation with a syncronisation
mechanism when the computer is linked to the network.
The actual software implementation is not yet "on the table", it's still in
the thought process.
This is just the beginning of the idea. One can easily think about the many
possibilities. Please feel free to discuss about it.
-
Alexandre
19 years, 5 months
fc3 on VirtualPC2004?
by Rex Dieter
Any ideas/pointers on getting FC3 running on (MS) VirtualPC (2004)? It
installs fine, but doesn't completely get into multiuser (or even
single-user) mode after that. Can't get in far enough to much snooping
around or even look at the logs.
-- Rex
19 years, 5 months