On Tue, 2006-07-18 at 16:15 +0100, Paul Howarth wrote:
Marc Schwartz (via MN) wrote:
> On Tue, 2006-07-18 at 14:11 +0100, Paul Howarth wrote:
> # Run restorecon to restore the SELinux context after re-building
> 10 01 * * * root /sbin/restorecon -rv /var/dcc
> 11 01 * * * root /sbin/restorecon -v /usr/local/bin/dccproc
>
>
>> I could add this to the dcc policy (or a separate local module for
>> restorecon):
>> cron_system_entry(restorecon_t)
>>
>> However, I think that the "right" way to fix this would be to use
>> restorecond and add an entry to /etc/selinux/restorecond.conf (the
>> manpage for restorecond suggests
>> /etc/selinux/POLICYTYPE/restorconfiles.conf but that file doesn't seem
>> to exist, and the restorecond binary includes the string
>> /etc/selinux/restorecond.conf).
>>
>> Maybe Dan could comment on that?
Try this instead of your cron jobs and see what happens:
# chkconfig --add restorecond
# chkconfig restorecond on
Edit /etc/selinux/restorecond.conf and add a line:
/usr/local/bin/dccproc
All the above done.
(is the restorecon of /var/dcc actually needed?)
Hopefully not. We did that as a temp fix b/c the context would change
after each nightly build.
If you have network-mounted home directories, you may want to remove
the
lines starting with "~"
Then:
# service restorecond start
Done.
>>> type=AVC msg=audit(1153053408.030:4599): avc: denied {
execmod } for pid=6019 comm="ld-linux.so.2"
name="libGLcore.so.1.0.8762" d ev=hdc7 ino=3116816
scontext=user_u:system_r:prelink_t:s0 tcontext=root:object_r:lib_t:s0 tclass=file
>>> type=SYSCALL msg=audit(1153053408.030:4599): arch=40000003 syscall=125
success=no exit=-13 a0=5c8000 a1=78e000 a2=5 a3=bf84c100 item s=0 pid=6019 auid=0 uid=0
gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=(none)
comm="ld-linux.so.2" exe="/lib/ld-2.4.so" sub
j=user_u:system_r:prelink_t:s0
>>> type=AVC_PATH msg=audit(1153053408.030:4599):
path="/usr/lib/libGLcore.so.1.0.8762"
>>> type=AVC msg=audit(1153053408.034:4600): avc: denied { execmod } for
pid=6022 comm="ld-linux.so.2" name="libnvidia-tls.so.1.0.876 2"
dev=hdc7 ino=3117829 scontext=user_u:system_r:prelink_t:s0 tcontext=root:object_r:lib_t:s0
tclass=file
>>> type=SYSCALL msg=audit(1153053408.034:4600): arch=40000003 syscall=125
success=no exit=-13 a0=a3e000 a1=1000 a2=5 a3=bfc98d40 items= 0 pid=6022 auid=0 uid=0
gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=(none)
comm="ld-linux.so.2" exe="/lib/ld-2.4.so" subj=
user_u:system_r:prelink_t:s0
>>> type=AVC_PATH msg=audit(1153053408.034:4600):
path="/usr/lib/tls/libnvidia-tls.so.1.0.8762"
>> Do you have nvidia video drivers installed using the nvidia installer
>> rather than an RPM package? If so, you should probably see:
>>
http://www.city-fan.org/tips/ProprietaryVideoDriverWarning
>
> Yep. I have never had a problem with them (dating back to RH 8.0, all
> on Dell laptops) and this is the first time that I had noted any avc's
> related to them.
>
> I have a script that I ran when I first moved to FC5 to set the
> following:
>
> /usr/sbin/setsebool -P allow_execstack=1
> /usr/sbin/setsebool -P allow_execmod=1
>
> based upon documents that I had found elsewhere.
That's somewhat overkill and I wouldn't want to do that.
Curiously, that approach is still noted in a variety of places,
including
FedoraFaq.org:
http://www.fedorafaq.org/#nvidia
and others:
http://www.mjmwired.net/resources/mjm-fedora-fc5.html#nvidia
http://stanton-finley.net/fedora_core_5_installation_notes.html#nVidia
Though I noted that it has been updated similar to your recommendation
in other places now, including the nVidia forums:
http://www.nvnews.net/vbulletin/showthread.php?t=68681
Undo it with:
# setsebool -P allow_execstack 0
# setsebool -P allow_execmod 0
Then fix the file contexts instead:
# semanage fcontext -a -f -- -t textrel_shlib_t
'/usr/lib/libGL(core)?\.so(\.[^/]*)*'
# semanage fcontext -a -f -- -t textrel_shlib_t
'/usr/lib/libnvidia.*\.so(\.[^/]*)*'
# restorecon -v /usr/lib/libGL* /usr/lib/libnvidia*
Please check that these files have context type textrel_shlib_t after
doing this.
# ls -lZ /usr/lib/libGL*
lrwxrwxrwx root root root:object_r:lib_t /usr/lib/libGLcore.so.1 ->
libGLcore.so.1.0.8762
-rwxr-xr-x root root system_u:object_r:textrel_shlib_t /usr/lib/libGLcore.so.1.0.8762
-rw-r--r-- root root root:object_r:lib_t /usr/lib/libGL.la
lrwxrwxrwx root root root:object_r:lib_t /usr/lib/libGL.so -> libGL.so.1
lrwxrwxrwx root root root:object_r:lib_t /usr/lib/libGL.so.1 ->
libGL.so.1.0.8762
-rwxr-xr-x root root system_u:object_r:textrel_shlib_t /usr/lib/libGL.so.1.0.8762
lrwxrwxrwx root root system_u:object_r:lib_t /usr/lib/libGLU.so ->
libGLU.so.1
lrwxrwxrwx root root system_u:object_r:lib_t /usr/lib/libGLU.so.1 ->
libGLU.so.1.3.060402
-rwxr-xr-x root root system_u:object_r:textrel_shlib_t /usr/lib/libGLU.so.1.3.060402
lrwxrwxrwx root root system_u:object_r:lib_t /usr/lib/libGLw.so ->
libGLw.so.1
lrwxrwxrwx root root system_u:object_r:lib_t /usr/lib/libGLw.so.1 ->
libGLw.so.1.0.0
-rwxr-xr-x root root system_u:object_r:lib_t /usr/lib/libGLw.so.1.0.0
# ls -lZ /usr/lib/libnvidia*
lrwxrwxrwx root root root:object_r:lib_t /usr/lib/libnvidia-cfg.so ->
libnvidia-cfg.so.1
lrwxrwxrwx root root root:object_r:lib_t /usr/lib/libnvidia-cfg.so.1 ->
libnvidia-cfg.so.1.0.8762
-rwxr-xr-x root root system_u:object_r:textrel_shlib_t
/usr/lib/libnvidia-cfg.so.1.0.8762
lrwxrwxrwx root root root:object_r:lib_t /usr/lib/libnvidia-tls.so.1 ->
libnvidia-tls.so.1.0.8762
-rwxr-xr-x root root system_u:object_r:textrel_shlib_t
/usr/lib/libnvidia-tls.so.1.0.8762
The upstream policy has file contexts set for this but I fear
they're
falling foul of the file contexts ordering rules.
>>> type=PATH msg=audit(1153138559.711:8): item=0 name="/var/run/utmp"
inode=87750 dev=fd:01 mode=0100664 ouid=0 ogid=22 rdev=00:00 obj=
system_u:object_r:init_var_run_t:s0
>>> type=AVC msg=audit(1153138559.715:9): avc: denied { read } for pid=2374
comm="mingetty" name="utmp" dev=dm-1 ino=87750 scontext=s
ystem_u:system_r:getty_t:s0 tcontext=system_u:object_r:init_var_run_t:s0 tclass=file
>>> type=SYSCALL msg=audit(1153138559.715:9): arch=40000003 syscall=5 success=no
exit=-13 a0=48909fd4 a1=0 a2=804a000 a3=48909fda items= 1 pid=2374 auid=4294967295 uid=0
gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=(none) comm="mingetty"
exe="/sbin/mingetty" s ubj=system_u:system_r:getty_t:s0
>>> type=CWD msg=audit(1153138559.715:9): cwd="/"
>>> type=PATH msg=audit(1153138559.715:9): item=0 name="/var/run/utmp"
inode=87750 dev=fd:01 mode=0100664 ouid=0 ogid=22 rdev=00:00 obj=
system_u:object_r:init_var_run_t:s0
>>> type=AVC msg=audit(1153138559.715:10): avc: denied { read write } for
pid=2374 comm="mingetty" name="utmp" dev=dm-1 ino=87750 sco
ntext=system_u:system_r:getty_t:s0 tcontext=system_u:object_r:init_var_run_t:s0
tclass=file
>>> type=SYSCALL msg=audit(1153138559.715:10): arch=40000003 syscall=5
success=no exit=-13 a0=48909fd4 a1=2 a2=0 a3=48909fda items=1 pid =2374 auid=4294967295
uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=(none)
comm="mingetty" exe="/sbin/mingetty" subj=s
ystem_u:system_r:getty_t:s0
>>> type=CWD msg=audit(1153138559.715:10): cwd="/"
>> Your /var/run/utmp appears to be labelled init_var_run_t rather than
>> initrc_var_run_t.
>
> Yep:
>
> # ls -lZ /var/run/utmp
> -rw-rw-r-- root utmp system_u:object_r:init_var_run_t /var/run/utmp
>
> # restorecon /var/run/utmp
>
> # ls -lZ /var/run/utmp
> -rw-rw-r-- root utmp system_u:object_r:initrc_var_run_t /var/run/utmp
Running restorecond should stop this happening again.
Paul.
OK. All done.
So far, no more avc's, but I'll keep track overnight and through a
couple of re-boots tomorrow.
Thanks,
Marc