On Fri, 2006-05-26 at 14:18 -0400, Daniel J Walsh wrote: > Paul Howarth wrote: > > I made something similar for my moin wiki running under mod_fcgid: > > > > te file: > > > > policy_module(apache, 0.2.1) > > > > require { > > type devpts_t; > > type httpd_t; > > type httpd_log_t; > > type httpd_sys_script_exec_t; > > type var_run_t; > > }; > > > > # ========================================================== > > # Create and use httpd_fastcgi_script_t for mod_fcgid apps > > # ========================================================== > > > > apache_content_template(fastcgi) > > kernel_read_kernel_sysctls(httpd_fastcgi_script_t) > > > > # Allow FastCGI applications to live alongside regular CGI apps > > allow httpd_fastcgi_script_t httpd_sys_script_exec_t:dir { > > search_dir_perms }; > > > > # Allow FastCGI applications to listen for FastCGI requests on their > > # sockets and respond to them > > allow httpd_fastcgi_script_t httpd_t:unix_stream_socket { > > rw_stream_socket_perms }; > > > > # FastCGI application doing something to the httpd error log > > dontaudit httpd_fastcgi_script_t httpd_log_t:file ioctl; > > > > # Not sure what this is doing (happens when fastcgi scripts start) > > dontaudit httpd_t devpts_t:chr_file ioctl; > > > > # mod_fcgid setting attr of its socket dir > > allow httpd_t var_run_t:dir setattr; > Why not create a context for its socket dir so you don't need this for > var_run? The obvious type to use would really be httpd_var_run_t rather than creating a new type (comparing with other users of /var/run). In fact I think I tried that but it seemed worse than leaving it the default var_run_t and adding the one allow rule. What would you suggest? > > fc file: > > > > /srv/www/tips/cgi-bin/moin.fcgi -- > > gen_context(system_u:object_r:httpd_fastcgi_script_exec_t,s0) > > /var/www/tips/cgi-bin/moin.fcgi -- > > gen_context(system_u:object_r:httpd_fastcgi_script_exec_t,s0) > > > > Paul. > > I think it might be a good idea to add this (fastcgi that is) policy to > base. Have you tried to submit it upstream? Not yet; it probably needs more work to add further capabilities, as I've only use one application with FastCGI myself, and I can see that httpd_sys_script_t has far more capabilities that I've so far allowed to httpd_fastcgi_script_t. Perhaps there should be a interface that goes further than apache_content_template and adds capabilities needed by most server-side scripts (e.g. the kernel_read_kernel_sysctls from above), for use in developing custom types like httpd_fastcgi_script_t? Paul. -- fedora-selinux-list mailing list fedora-selinux-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-selinux-list