| 1569 |
a process which violated policy, administrator can update policy while |
a process which violated policy, administrator can update policy while |
| 1570 |
the process is sleeping. This "try again" response allows administrator |
the process is sleeping. This "try again" response allows administrator |
| 1571 |
to restart policy checks from the beginning after updating policy. |
to restart policy checks from the beginning after updating policy. |
| 1572 |
|
|
| 1573 |
|
Fix 2008/09/11 |
| 1574 |
|
|
| 1575 |
|
@ Remember whether the process is allowed to write to /proc/ccs/ interface. |
| 1576 |
|
|
| 1577 |
|
Since programs for manipulating policy (e.g. ccs-queryd ) are installed |
| 1578 |
|
in the form of RPM/DEB packages, these programs lose the original |
| 1579 |
|
pathnames when they are updated by the package manager. The package |
| 1580 |
|
manager renames these programs before deleting these programs so that |
| 1581 |
|
the package manager can rollback the operation. |
| 1582 |
|
This causes a problem when the programs are listed into /proc/ccs/manager |
| 1583 |
|
using pathnames, as the programs will no longer be allowed to write to |
| 1584 |
|
/proc/ccs/ interface while the process of old version of the program is |
| 1585 |
|
alive. |
| 1586 |
|
|
| 1587 |
|
To solve this problem, I modified to remember the fact that the process |
| 1588 |
|
is once allowed to write to /proc/ccs/ interface until the process |
| 1589 |
|
attempts to execute a different program. |
| 1590 |
|
This change makes it impossible to revoke permission to write to |
| 1591 |
|
/proc/ccs/ interface without killing the process, but it will be better |
| 1592 |
|
than nonfunctioning ccs-queryd program. |
| 1593 |
|
|
| 1594 |
|
Fix 2008/09/19 |
| 1595 |
|
|
| 1596 |
|
@ Allow selecting a domain by PID. |
| 1597 |
|
|
| 1598 |
|
Sometimes we want to know what ACLs are given to specific PID, but |
| 1599 |
|
finding a domainname for that PID from /proc/ccs/.process_status and |
| 1600 |
|
reading ACLs from /proc/ccs/domain_policy by the domainname is very slow. |
| 1601 |
|
Thus, I modified /proc/ccs/domain_policy to allow selecting a domain by |
| 1602 |
|
PID. For example, to read domain ACL of current process from bash, |
| 1603 |
|
run as follows. |
| 1604 |
|
|
| 1605 |
|
# exec 100<>/proc/ccs/domain_policy |
| 1606 |
|
# echo select pid=$$ >&100 |
| 1607 |
|
# while read -u 100; do echo $REPLY; done |
| 1608 |
|
|
| 1609 |
|
If a domain is once selected by PID, reading /proc/ccs/domain_policy will |
| 1610 |
|
print only that domain if that PID exists or print nothing otherwise. |
| 1611 |
|
|
| 1612 |
|
@ Disallow concurrent /proc/ccs/ access using the same file descriptor. |
| 1613 |
|
|
| 1614 |
|
Until now, one process can read() from /proc/ccs/ while other process |
| 1615 |
|
that shares the file descriptor can write() to /proc/ccs/ . |
| 1616 |
|
But to implement "Allow selecting a domain by PID" feature, I disabled |
| 1617 |
|
concurrent read()/write() because the feature need to modify read buffer |
| 1618 |
|
while writing. |
| 1619 |
|
|
| 1620 |
|
Fix 2008/10/01 |
| 1621 |
|
|
| 1622 |
|
@ Add retry counter into /proc/ccs/query . |
| 1623 |
|
|
| 1624 |
|
To be able to handle some of queries from /proc/ccs/query without user's |
| 1625 |
|
interaction, I added retry counter for avoiding infinite loop caused by |
| 1626 |
|
"try again" response. |
| 1627 |
|
|
| 1628 |
|
Fix 2008/10/07 |
| 1629 |
|
|
| 1630 |
|
@ Don't transit to new domain until do_execve() succeeds. |
| 1631 |
|
|
| 1632 |
|
Until now, a process's domain was updated to new domain which the process |
| 1633 |
|
will belongs to before do_execve() succeeds so that the kernel can do |
| 1634 |
|
permission checks for interpreters and environment variables based on |
| 1635 |
|
new domain. But this caused a subtle problem when other process sends |
| 1636 |
|
signals to the process, for the process returns to old domain if |
| 1637 |
|
do_execve() failed. |
| 1638 |
|
|
| 1639 |
|
So, I modified to pass new domain to functions so that I can avoid |
| 1640 |
|
modifying a process's domain before do_execve() succeeds. |
| 1641 |
|
|
| 1642 |
|
@ Use old task state for audit logs. |
| 1643 |
|
|
| 1644 |
|
Until now, audit logs were generated using the task state after |
| 1645 |
|
processing "; set task.state" part. But to generate accurate logs, |
| 1646 |
|
I modified to save the task state before processing "; set task.state" |
| 1647 |
|
part and use the saved state for audit logs. |
| 1648 |
|
|
| 1649 |
|
@ Use a structure for passing parameters. |
| 1650 |
|
|
| 1651 |
|
As the number of parameters is increasing, I modified to use a structure |
| 1652 |
|
for passing parameters. |
| 1653 |
|
|
| 1654 |
|
Fix 2008/10/11 |
| 1655 |
|
|
| 1656 |
|
@ Remove domain_acl_lock mutex. |
| 1657 |
|
|
| 1658 |
|
I noticed that I don't need to keep all functions that modify an ACL of |
| 1659 |
|
a domain mutually exclusive. Since each functions handles different type |
| 1660 |
|
of ACL, locking is needed only when they append an ACL to a domain. |
| 1661 |
|
So, I modified to use local locks. |
| 1662 |
|
|
| 1663 |
|
Fix 2008/10/14 |
| 1664 |
|
|
| 1665 |
|
@ Fix ccs_check_condition() bug. |
| 1666 |
|
|
| 1667 |
|
Due to a bug in ccs_check_condition(), it was impossible to use |
| 1668 |
|
task.state[0] task.state[1] task.state[2] inside condition part |
| 1669 |
|
if the ACL does not treat a pathname. For example, an ACL like |
| 1670 |
|
|
| 1671 |
|
allow_network TCP connect @HTTP_SERVERS 80 if task.state[0]=100 |
| 1672 |
|
|
| 1673 |
|
didn't work. |
| 1674 |
|
|
| 1675 |
|
Fix 2008/10/15 |
| 1676 |
|
|
| 1677 |
|
@ Show process information in /proc/ccs/.process_status . |
| 1678 |
|
|
| 1679 |
|
To be able to determine a process's type, I added a command "info PID" |
| 1680 |
|
which returns process information of the specified PID in |
| 1681 |
|
"PID manager=\* execute_handler=\* state[0]=\$ state[1]=\$ state[2]=\$" |
| 1682 |
|
format. |
| 1683 |
|
|
| 1684 |
|
Fix 2008/10/20 |
| 1685 |
|
|
| 1686 |
|
@ Use rcu_dereference() when walking the list. |
| 1687 |
|
|
| 1688 |
|
I was using "dependency ordering" for appending an element to a list |
| 1689 |
|
without asking the reader to take a lock. But "dependency ordering" |
| 1690 |
|
is not respected by DEC Alpha or by some aggressive value-speculation |
| 1691 |
|
compiler optimizations. |
| 1692 |
|
|
| 1693 |
|
On such environment, use of "dependency ordering" can lead to system |
| 1694 |
|
crash because the reader might read uninitialized value of newly |
| 1695 |
|
appended element. |
| 1696 |
|
|
| 1697 |
|
To prevent the reader from reading uninitialized value of newly appended |
| 1698 |
|
element, I inserted rcu_dereference() when walking the list. |
| 1699 |
|
|
| 1700 |
|
Fix 2008/11/04 |
| 1701 |
|
|
| 1702 |
|
@ Use sys_getpid() instead for current->pid. |
| 1703 |
|
|
| 1704 |
|
Kernel 2.6.24 introduced PID namespace. |
| 1705 |
|
|
| 1706 |
|
To compare PID given from userland, I can't use current->pid. |
| 1707 |
|
So, I modified to use sys_getpid() instead for current->pid. |
| 1708 |
|
|
| 1709 |
|
I modified to use task_tgid_nr_ns() for 2.6.25 and later instead for |
| 1710 |
|
current->tgid when checking /proc/self/ in get_absolute_path(). |
TOMOYO
Parent Directory
Revision Log
Patch