Bug 151145 - segfaults when changing context
segfaults when changing context
Status: CLOSED NOTABUG
Product: Fedora
Classification: Fedora
Component: vixie-cron (Show other bugs)
rawhide
All Linux
medium Severity high
: ---
: ---
Assigned To: Marcela Mašláňová
Brock Organ
:
Depends On:
Blocks: 163883
  Show dependency treegraph
 
Reported: 2005-03-15 07:49 EST by Enrico Scholz
Modified: 2007-11-30 17:11 EST (History)
0 users

See Also:
Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Environment:
Last Closed: 2007-01-08 10:42:00 EST
Type: ---
Regression: ---
Mount Type: ---
Documentation: ---
CRM:
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---


Attachments (Terms of Use)

  None (edit)
Description Enrico Scholz 2005-03-15 07:49:45 EST
Description of problem:

crond dies when executing jobs:

| Program received signal SIGSEGV, Segmentation fault.
| 0x97fddfba in *setexeccon_internal (context=0x11d3cc <Address 0x11d3cc out of bounds>)
|     at setexeccon.c:15
| 15                      ret = write(fd, context, strlen(context)+1);
| (gdb) bt
| #0  0x97fddfba in *setexeccon_internal (context=0x11d3cc <Address 0x11d3cc out of bounds>)
|     at setexeccon.c:15
| #1  0x6aab1988 in do_command (e=0x6aab86f0, u=0x6aab8698) at do_command.c:334
| #2  0x6aab11fc in job_runqueue () at job.c:66
| #3  0x6aaaed3e in main (argc=1, argv=0x4236e629) at cron.c:256


Previous breakpoints were

| Breakpoint 2, do_command (e=0x6aab86f0, u=0x6aab8698) at do_command.c:71
| 71              Debug(DPROC, ("[%ld] do_command(%s, (%s,%ld,%ld))\n",
| (gdb) p u
| $1 = (user *) 0x6aab8698
| (gdb) p *u
| $2 = {next = 0x0, prev = 0x6aab8190, name = 0x6aab81c0 "*system*", mtime = 1109646298, 
|   crontab = 0x6aab9700, scontext = 0x11d3cc <Address 0x11d3cc out of bounds>}
| (gdb) c
| Continuing.
| Attaching after fork to child process 28298.
| Attaching after fork to child process 28299.
| [Switching to process 28299]
| 
| Breakpoint 1, 0x97fddf7e in *setexeccon_internal (context=0x11d3cc <Address 0x11d3cc out of bounds>)
|     at setexeccon.c:7
| 7       {
| (gdb) c


As you see, 'u->scontext' is uninitialized which is caused in
get_security_context() (user.c) by

|       if (get_default_context(name, NULL, &scontext)) {
|               if (security_getenforce() > 0) {
|                       log_it(name, getpid(), "No SELinux security context",tabname);
|                       return -1;
|               } else {
|                       log_it(name, getpid(), "No security context but SELinux in permissive mode, co
|               }
|       }

On my system (a chroot environment), get_default_context() fails and
'scontext' gets never be initialized.



Version-Release number of selected component (if applicable):

vixie-cron-4.1-24_FC4
libselinux-1.22-1
Comment 1 Jason Vas Dias 2005-03-15 09:41:19 EST
This is now fixed with vixie-cron-4.1-26_FC4 .

A workaround for this bug would be to copy the 
/etc/security/default_contexts or the /usr/home/$user/.default_contexts
into the chroot environment .

Please note that any increase in security obtained through use of 
a chroot environment is more than offset by disabling SELinux, and 
that by using SELinux properly, a far more secure environment could
be achieved than that obtained by use of a chroot .

Also note that if you wish to disable certain users from using cron,
a much better way of doing so would be to use PAM access.conf or 
limits.conf .

However, crond should not segfault if the context for a user cannot
be found. This is now fixed, and will be in Rawhide / FC4t2 shortly.
Comment 2 Enrico Scholz 2005-03-15 09:48:54 EST
yes, I found this workaround also.

My chroot is secure as I am using Linux-Vserver
(http://linux-vserver.org) technology for it.

PAM access.conf is a little bit problematic as it requires a login-TTY
which is not set by the most servers. So using is generally in
system-auth is not possible. Adding it manually to pam.d/crond would
be possible but may cause problems on program updates.
Comment 3 Jason Vas Dias 2005-03-15 10:03:14 EST
Linux-Vserver is still way less secure than SELinux when properly
configured.
Also adding the line 
  'account  required  /lib/security/$ISA/pam_access.so'
to /etc/pam.d/crond will remain across updates because it is
installed as a %config file - crond uses the 'crond' PAM_TTY
"login-TTY", which has no more significance to PAM than the name - 
it is not a real "terminal" .
Comment 4 Enrico Scholz 2005-03-15 10:27:47 EST
> Linux-Vserver is still way less secure than SELinux 

I am not aware of any attack against Linux-Vserver which would not
work against SELinux also. But SELinux has the following, additional
problems:

* it has a strange kernel-API; relying on the /proc + /selinux
  filesystem makes it unsuitable for any chroot tasks

* the missing chroot functionality does not allow virtualization

* it is very difficultly to configure. The default setup does not fit
  for more advanced tasks so you have to configure it yourself. When
  doing this, the upgrade path with 'fixfiles -C ...' does not work
  and you are just busy with rebooting and relabeling the entire
  filesystem.


> when properly configured.

yes, this is a problem with SELinux. It is very difficultly to configure
while VServer uses the well-known Unix configuration style. So, most
SELinux machines will not be "properly configured" and less secure than
corresponding vservers.
Comment 5 Marcela Mašláňová 2006-08-28 07:16:52 EDT
Well, are you content with vixie-cron-4.1-26_FC4? 
I know there are still some problems with SElinux. If you have some problems 
with configuration and I have to fix something in cron, let me know.

Note You need to log in before you can comment on or make changes to this bug.