| 1921 |
not working. |
not working. |
| 1922 |
|
|
| 1923 |
This problem happens on little endian platforms (e.g. x86). |
This problem happens on little endian platforms (e.g. x86). |
| 1924 |
|
|
| 1925 |
|
Fix 2009/04/20 |
| 1926 |
|
|
| 1927 |
|
@ Update recvmsg() hooks. |
| 1928 |
|
|
| 1929 |
|
Since 1.5.0, I was doing network access control for incoming UDP and RAW |
| 1930 |
|
packets inside skb_recv_datagram(). But to synchronize with LSM version, |
| 1931 |
|
I moved ccs_recv_datagram_permission() hook from skb_recv_datagram() to |
| 1932 |
|
udp_recvmsg()/udpv6_recvmsg()/raw_recvmsg()/rawv6_recvmsg() with name |
| 1933 |
|
change to ccs_recvmsg_permission(). |
| 1934 |
|
|
| 1935 |
|
Fix 2009/05/08 |
| 1936 |
|
|
| 1937 |
|
@ Add condition for symlink's target pathname. |
| 1938 |
|
|
| 1939 |
|
Until now, "allow_symlink" keyword allows creation of a symlink but does |
| 1940 |
|
not check the symlink's target. Usually it is no problem because |
| 1941 |
|
permission checks are done using dereferenced pathname. But in some |
| 1942 |
|
cases, we should restrict the symlink's target. For example, |
| 1943 |
|
"ln -s .htpasswd /var/www/html/readme.html" by CGI program should be |
| 1944 |
|
blocked because we will allow Apache to read both |
| 1945 |
|
/var/www/html/readme.html and /var/www/html/.htpasswd . |
| 1946 |
|
|
| 1947 |
|
Thus, I added new condition, "symlink.target". |
| 1948 |
|
|
| 1949 |
|
allow_symlink /var/www/html/\*.html if symlink.target="\*.html" |
| 1950 |
|
|
| 1951 |
|
allow_symlink /var/www/html/\*\-.\* if symlink.target="\*\-.\*" |
| 1952 |
|
|
| 1953 |
|
@ Don't return -EAGAIN at ccs_socket_recvmsg_permission(). |
| 1954 |
|
|
| 1955 |
|
It turned out that it is not permitted for accept() and recvmsg() to |
| 1956 |
|
return -EAGAIN if poll() said connections/datagrams are ready. However, |
| 1957 |
|
recvmsg() may return -EAGAIN and potentially confuse some applications |
| 1958 |
|
because ccs_socket_recvmsg_permission() is returning -EAGAIN. |
| 1959 |
|
|
| 1960 |
|
Thus, I modified ccs_socket_recvmsg_permission() to return -ENOMEM |
| 1961 |
|
rather than -EAGAIN. |
| 1962 |
|
|
| 1963 |
|
Fix 2009/05/19 |
| 1964 |
|
|
| 1965 |
|
@ Don't call get_fs_type() with a mutex held. |
| 1966 |
|
|
| 1967 |
|
Until now, when ccs_update_mount_acl() is called with unsupported |
| 1968 |
|
filesystem, /sbin/modprobe is executed from get_fs_type() to load |
| 1969 |
|
filesystem module. And get_fs_type() does not return until /sbin/modprobe |
| 1970 |
|
finishes. |
| 1971 |
|
|
| 1972 |
|
This means that it will cause deadlock if /sbin/modprobe (which is |
| 1973 |
|
executed via get_fs_type() in ccs_update_mount_acl()) calls |
| 1974 |
|
ccs_update_mount_acl(); although it won't happen unless an administrator |
| 1975 |
|
inserts execute_handler to call mount() requests in learning mode or to |
| 1976 |
|
add "allow_mount" entries to /proc/ccs/system_policy . |
| 1977 |
|
|
| 1978 |
|
I modified to unlock the mutex before calling get_fs_type(). |
| 1979 |
|
|
| 1980 |
|
Version 1.6.8 2009/05/?? Feature enhancement release. |
TOMOYO
Parent Directory
Revision Log
Patch