Craig White wrote:
On Mon, 2008-06-16 at 11:39 +0100, Paul Howarth wrote:Craig White wrote:The conventional unix ownership and permissions make very little difference as far as SELinux is concerned, so although you need to get them right, they're not going to affect the file contexts needed.On Sat, 2008-06-14 at 16:51 +0100, Paul Howarth wrote:On Sat, 14 Jun 2008 08:05:56 -0700 Craig White <craigwhite azapple com> wrote:I'm a bit confused myself because in essence, httpd is just a proxy to the ruby/rails 'mongrel' which is a http server in ruby running the ruby processes and is providing dhtml on higher ports as the user. FWIW...httpd runs as user 'apache' (as ususal) mongrels run as regular 'user' (me) all files and folders inside the subdirectory we are discussing... (/home/craig/svn-new) are owned by me (not root, not apache)What context is mongrels running in (try the -Z option of ps)? How does that process get started (via an initscript?)?---- yes, a SysV initscript...(running 2 mongrels at present... port & pid #'s 3000 & 3001) # ps auxZ|grep mongrel unconfined_u:unconfined_r:unconfined_t:SystemLow-SystemHigh root 7079 0.0 0.0 4120 732 pts/6 S+ 05:02 0:00 grep mongrel root:unconfined_r:unconfined_t:-s0:c0.c255 craig 27313 0.0 3.0 45068 30164 ? Sl Jun15 0:10 /usr/bin/ruby /usr/bin/mongrel_rails start -d -e development -a 127.0.0.1 -c /home/craig/svn-new/th-db/branches/phase5 --user craig --group craig -p 3000 -P log/mongrel.3000.pid -l log/mongrel.3000.log root:unconfined_r:unconfined_t:-s0:c0.c255 craig 27316 0.0 2.9 45052 29468 ? Sl Jun15 0:10 /usr/bin/ruby /usr/bin/mongrel_rails start -d -e development -a 127.0.0.1 -c /home/craig/svn-new/th-db/branches/phase5 --user craig --group craig -p 3001 -P log/mongrel.3001.pid -l log/mongrel.3001.log ----
OK, so they're running as unconfined_t at the moment.
You probably need to run the ruby process confined so that it generates files that are readable by httpd. It might actually work ok running as httpd_t given how closely related the processes are.I ran the commands that you suggested (ignoring the alerts raised by the first command) and then restarted httpd service and got a new alert... SELinux is preventing the httpd from using potentially mislabeled files (2F7661722F746D702F6B646563616368652D63726169672F6B70632F6B64652D69636F6E2D63616368652E64617461202864656C6574656429). Detailed Description[SELinux is in permissive mode, the operation would have been denied but was permitted due to permissive mode.]SELinux has denied httpd access to potentially mislabeled file(s) (2F7661722F746D702F6B646563616368652D63726169672F6B70632F6B64652D69636F6E2D63616368652E64617461202864656C6574656429). This means that SELinux will not allow httpd to use these files. It is common for users to edit files in their home directory or tmp directories and then move (mv) them to system directories. The problem is that the files end up with the wrong file context which confined applications are not allowed to access. Allowing AccessIf you want httpd to access this files, you need to relabel them using restorecon -v '2F7661722F746D702F6B646563616368652D63726169672F6B70632F6B64652D69636F6E2D63616368652E64617461202864656C6574656429'. You might want to relabel the entire directory using restorecon -R -v ''. Additional InformationSource Context: unconfined_u:system_r:httpd_tTarget Context: unconfined_u:object_r:user_tmp_tTarget Objects: 2F7661722F746D702F6B646563616368652D63726169672F6B70632F6B64652D69636F6E2D63616368652E64617461202864656C6574656429 This is my new development system and I obviously will be doing svn commit/update operations in this directory and it was created by acheckout.There is a 'tmp' directory in the RAILS_ROOT directory (/home/craig/svn-new/th-db/branches/phase5) which holds... - temporary pdf files put there by ruby before 'merging' database data with pdftk - subdirectories but the only 'non-empty' subdirectory is one called 'pids' which holds the pid for the backgroundrd (a separate ruby process that runs long running processes in the background). I'm wondering if this directory shouldn't have some different contexts... My desire is to have a plan to manage selinux contexts when it comes time to merge this on my production server.---- I'm sort of unclear on what you are telling me here. What did happen after I made the change you suggested on Saturday is that the 4:02 rotation log restart of httpd stopped triggering selinux alerts but a full restart of httpd service does generate the latest alert. I could conceivably run the mongrels as user 'apache' except that the permissions on some of the folders would have to be changed because there are some directories that files are written into by the ruby web server...so I try to just run as user.
Don't change anything about the regular Unix permissions at the moment; I guess that for a production server you'd create a separate account for the Ruby stuff to run as.
What would be an interesting experiment would be to run the Ruby stuff in the same SELinux context as httpd. Try changing the context type of /usr/bin/mongrel_rails to httpd_exec_t and restart the services.
# chcon -t httpd_exec_t /usr/bin/mongrel_railsI'm not sure whether this will make things better or worse but it should eliminate some problems for the two httpd-like bits talking to each other.
Paul.