On Thursday, September 28, 2017 at 5:59:09 PM UTC+2, steve.coleman wrote:
> On 09/26/2017 05:35 AM, pels wrote:
> > On Sunday, September 24, 2017 at 6:19:15 PM UTC+2, cooloutac wrote:
> >> On Sunday, September 24, 2017 at 12:17:33 PM UTC-4, cooloutac wrote:
> >>> On Sunday, September 24, 2017 at 12:16:34 PM UTC-4, cooloutac wrote:
> >>>> On Thursday, September 21, 2017 at 4:40:42 AM UTC-4, pels wrote:
> >>>>> On Wednesday, September 20, 2017 at 2:54:31 PM UTC+2, cooloutac wrote:
> >>>>>> On Wednesday, September 20, 2017 at 4:41:58 AM UTC-4, pels wrote:
> >>>>>>> I'd like to activate SELINUX(enforcing) in VMs (f25 and f25-minimal), 
> >>>>>>> but fails:
> >>>>>>>
> >>>>>>> [    1.510532] audit: type=1404 audit(1505894636.317:2): enforcing=1 
> >>>>>>> old_enforcing=0 auid=4294967295 ses=4294967295
> >>>>>>> [    1.601491] audit: type=1403 audit(1505894636.408:3): policy 
> >>>>>>> loaded auid=4294967295 ses=4294967295
> >>>>>>> [    1.605815] systemd[1]: Successfully loaded SELinux policy in 
> >>>>>>> 95.611ms.
> >>>>>>> [    1.617897] systemd[1]: Failed to mount tmpfs at /run: Permission 
> >>>>>>> denied
> >>>>>>> [.[0;1;31m!!!!!!.[0m] Failed to mount API filesystems, freezing.
> >>>>>>> [    1.621206] systemd[1]: Freezing execution.
> >>>>>>>
> >>>>>>> I had it enabled  in fedora 24 but after upgrading failed
> >>>>>>> I create a new template (f25 and f25-minimal) with same effect.
> >>>>>>>
> >>>>>>> I have tried to reset SELinux to its initial state:
> >>>>>>> yum remove selinux-policy
> >>>>>>> rm -rf /etc/selinux
> >>>>>>> yum install selinux-policy-targeted
> >>>>>>> fixfiles -f -F relabel
> >>>>>>> reboot
> >>>>>>>
> >>>>>>> Any ideas?
> >>>>>>>
> >>>>>>> Thank you very much
> >>>>>>>
> >>>>>>> Best Regards
> >>>>>>    
> >>>>>>    Is this a vm, if so do we really care if systemd is running in it?  
> >>>>>>  You sure thats selinux?  what does sestatus say?
> >>>>>>
> >>>>>> When googling this error seems people have same issue when running 
> >>>>>> docker.  And you have to set seccomp to unconfined.
> >>>>>
> >>>>> Thank you cooloutac
> >>>>>
> >>>>> -Is this a vm
> >>>>> It happens in Templates and VMs.
> >>>>>
> >>>>> -Is this a vm, if so do we really care if systemd is running in it?
> >>>>> The problem is when i enable SELINUX VMs/templates doesn't "boot" or 
> >>>>> fail to start.
> >>>>> If I disable SELINUX, the templates/VMs start whithout problems and 
> >>>>> systemd is activated.
> >>>>>
> >>>>> -You sure thats selinux?
> >>>>> Yes i'm pretty sure, it's exactly the same config that i had in 
> >>>>> fedora24.
> >>>>> In dom0
> >>>>> qvm-prefs -s fedora-25 kernelopts "nopat security=selinux selinux=1"
> >>>>> and in VMs/Templats
> >>>>> /etc/selinux/config
> >>>>>
> >>>>> SELINUX=enforcing
> >>>>> SELINUXTYPE=targeted
> >>>>>
> >>>>> Default selinux config
> >>>>>
> >>>>> -what does sestatus say?
> >>>>> I can't execute anything in template/VMs
> >>>>> in dom0:
> >>>>> qvm-run fedora-25 --nogui -pass-io -u root "sestatus"
> >>>>> Error(fedora-25): Domain 'fedora-25':qreexec not connected
> >>>>>
> >>>>> -When googling this error seems people have same issue when running 
> >>>>> docker.  And you have to set seccomp to unconfined
> >>>>>
> >>>>> Yes, i've read it, but i don't know how disable seccomp and the 
> >>>>> consequences...
> >>>>>
> >>>>>
> >>>>> Could you make me a big favour and try to activate SELINUX?
> >>>>>
> >>>>> Thank you very much
> >>>>>
> >>>>> Best regards
> >>>>
> >>>> Probably only useful in the template vm.  But still not sure how 
> >>>> beneficial it would be was my point though.  Its probably not compatible 
> >>>> with qubes, sounds like it breaks qrexec, maybe not worth the headache 
> >>>> man.
> >>>
> 
> Try running SELinux in permissive mode then use the SELinux audit2allow 
> to turn the permission violations into new permit rules. Those permit 
> rules will allow the system to boot normally once applied to the system 
> policy. This process may need to be repeated.
> 
> I used to use this script back when I was using tcsh on a stock Fedora 
> system, but it would be trivial to change to bash or just use the 
> command line:
> 
> #!/bin/tcsh
> echo
> # find the avc entries for the application in the log file
> grep "$argv[1]" /var/log/audit/audit.log >  "my$argv[1].out"
> 
> # pass those entries to audit to allow
> cat  "my$argv[1].out" | audit2allow -M "my$argv[1]"
> 
> # display the generated policy so you can learn what its doing
> cat "my$argv[1].te"
> 
> echo
> 
> # Just output to stdout the command needed to commit the new policy,
> # this is for cut and paste command execution
> 
> echo semodule -i my$argv[1].pp
> echo
> 
> Basically you give the script a search string and it grep locates all 
> avc permission violations containing that application-name/error-message 
> and creates a policy file to fix those specific problems, and then it 
> echos the command needed to actually fix it to the console. If you agree 
> with those permissions displayed you simply copy and paste that command 
> back into the command line and press enter. Keep the *.te,*.pp files for 
> any future upgrades should you need to repeat this on another system.
> 
> If you were running in permissive mode nothing will fail but the system 
> will still log those avc violations. Once you have run the *.te file to 
> update the policy, on the next boot those avc messages should be 
> resolved, if so, turn on the se enforcement (targeted mode) and try 
> booting again. Rinse and repeat as necessary until your system boots 
> properly, then you can focus on the user application side of things. 
> Targeted mode is best for the user, as it denies any system services 
> from running that you didn't give permission to run but does not block 
> the user from getting their work done.
> 
> As for Xen itself, there is also the Flask architecture (SELinux equiv) 
> built into Xen, so that might be interesting to see what it can do with 
> regards to a determined hacker. I have not seen how to enable that in a 
> stock qubes system without rebuilding everything.
> 
> I personally feel that simply enabling the permissive audit logging 
> could be quite protective simply by detecting any tampering, but only if 
> there were a realtime process that could monitor the log to alert the 
> user to the presence of a hacker/malware tampering with the system. The 
> adversary has to try things that would normally get logged, and thus 
> leave fingerprints on your system. Hardware enforced separation is 
> great, but if you have an intruder then you really do want to know when 
> to put the shields up and begin to analyze the damage to plan your next 
> actions. The sooner you know you are not alone the better. How you 
> recover can be highly dependent on how long you have been subverted by 
> your adversary, and what the did to get a foothold on your system. AVC 
> audits may show you exactly how they did it, thus once you recover you 
> may be able to plug those holes. Without this important information they 
> will be back, you can count on it.
> 
> If you do pursue this challenge please keep me informed of your progress 
> off line preferably. While I agree with the qubes developers that don't 
> see it as a big preventive measure, given that users all have sudo, I do 
> however think that getting a first-warning upon their entry can be key 
> to keeping the system safe to use. What you do after they have a 
> foothold is dependent on what you are protecting.
> 
> good luck!
> 
> Steve Coleman

Thank you very much Steve Coleman. 

-- 
You received this message because you are subscribed to the Google Groups 
"qubes-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/qubes-users/f675acae-9f00-48df-99b3-59fda4a60e71%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to