2088 |
@ Transit to new domain before do_execve() succeeds. |
@ Transit to new domain before do_execve() succeeds. |
2089 |
|
|
2090 |
Permission checks for interpreters and environment variables are |
Permission checks for interpreters and environment variables are |
2091 |
done using new domain. In order to be allow ccs-queryd to reach the new |
done using new domain. In order to allow ccs-queryd to reach the new |
2092 |
domain via global PID, I reverted "Don't transit to new domain until |
domain via global PID, I reverted "Don't transit to new domain until |
2093 |
do_execve() succeeds." made on 2008/10/07. |
do_execve() succeeds." made on 2008/10/07. |
2094 |
|
|
2305 |
operations (e.g. mkdir()) even if mode=disabled . It is a waste of CPU |
operations (e.g. mkdir()) even if mode=disabled . It is a waste of CPU |
2306 |
resource to check DAC permissions when MAC permissions are not checked. |
resource to check DAC permissions when MAC permissions are not checked. |
2307 |
Thus, I modified to skip DAC permission checks if mode=disabled . |
Thus, I modified to skip DAC permission checks if mode=disabled . |
2308 |
|
|
2309 |
|
Fix 2009/12/19 |
2310 |
|
|
2311 |
|
@ Fix memory leak in ccs_environ(). |
2312 |
|
|
2313 |
|
When I fixed a bug that a permission like |
2314 |
|
|
2315 |
|
allow_env PATH if exec.envp["PATH"]="/" |
2316 |
|
|
2317 |
|
was not working (2009/11/02), I allocated two buffers but only one buffer |
2318 |
|
was released. |
2319 |
|
|
2320 |
|
This bug will trigger OOM killer if environment variable checking is |
2321 |
|
enabled. |
2322 |
|
|
2323 |
|
Fix 2010/01/17 |
2324 |
|
|
2325 |
|
@ Use current domain's name for execute_handler audit log. |
2326 |
|
|
2327 |
|
Since 1.6.7 , /proc/ccs/grant_log was by error using next domain's name |
2328 |
|
when auditing current domain's "execute_handler" line. |
2329 |
|
|
2330 |
|
Fix 2010/03/02 |
2331 |
|
|
2332 |
|
@ Allow domain transition without execve(). |
2333 |
|
|
2334 |
|
To be able to split permissions for Apache's CGI programs which are |
2335 |
|
executed without execve(), I added special domain transition which is |
2336 |
|
performed by atomically writing '\0'-terminated binary string to |
2337 |
|
/proc/ccs/.transition interface. For example, a process which belongs to |
2338 |
|
"<kernel> /usr/sbin/httpd" domain will transit to |
2339 |
|
"<kernel> /usr/sbin/httpd //app=cgi1\040id=10000" domain by atomically |
2340 |
|
writing "app=cgi1 id=10000" + '\0' to /proc/ccs/.transition using |
2341 |
|
Apache's ap_hook_handler() functionality. |
2342 |
|
|
2343 |
|
Note that '\0'-terminated binary string is converted to TOMOYO's string |
2344 |
|
inside kernel and prefix "//" is automatically added to the string so |
2345 |
|
that domainname does not conflict with domainnames created by execve(). |
2346 |
|
Without this prefix, if "<kernel> /usr/sbin/sshd /bin/bash" domain is |
2347 |
|
allowed to open /proc/ccs/.transition for writing and |
2348 |
|
"<kernel> /usr/sbin/sshd /bin/bash /usr/bin/passwd" domain is allowed to |
2349 |
|
access /etc/shadow , /bin/bash will be able to access /etc/shadow by |
2350 |
|
atomically writing "/usr/bin/passwd" + '\0' to /proc/ccs/.transition . |
2351 |
|
Allowing /bin/bash to access /etc/shadow is not what people want. |
2352 |
|
|
2353 |
|
Permission for this operation is checked by "allow_transit" keyword. |
2354 |
|
Unlike "allow_execute" keyword, the string parameter for "allow_transit" |
2355 |
|
keyword does not refer a real file on filesystem's namespace. Therefore, |
2356 |
|
you can store any combination of parameters like LDAP's DN entry in the |
2357 |
|
string parameter for "allow_transit" keyword. |
2358 |
|
|
2359 |
|
Fix 2010/03/08 |
2360 |
|
|
2361 |
|
@ Allow building as loadable kernel module. |
2362 |
|
|
2363 |
|
To be able to minimize filesize increment of vmlinux, I made it |
2364 |
|
possible to compile TOMOYO Linux as loadable kernel module. |
2365 |
|
Although patching the kernel source and recompiling the kernel are |
2366 |
|
inevitable, this change will make it easier to enable TOMOYO Linux |
2367 |
|
when there is a filesize limitation on vmlinux (e.g. embedded systems). |
2368 |
|
|
2369 |
|
Fix 2010/03/25 |
2370 |
|
|
2371 |
|
@ Fix ccs_get_ipv6_address() bug. |
2372 |
|
|
2373 |
|
Since 1.7.0 , ccs_get_ipv6_address() was by error returning address of |
2374 |
|
"struct list_head ccs_address_list" if memory allocation failed. |
2375 |
|
As a result, ccs_put_ipv6_address() will modify memory near |
2376 |
|
"struct list_head ccs_address_list" if memory allocation failed. |
2377 |
|
|
2378 |
|
Fix 2010/03/26 |
2379 |
|
|
2380 |
|
@ Fix ccs_lport_reserved() bug. |
2381 |
|
|
2382 |
|
Since 1.7.0 , ccs_lport_reserved() was by error checking wrong port |
2383 |
|
number. As a result, "deny_autobind" keyword was not working as expected. |
2384 |
|
|
2385 |
|
Version 1.7.2 2010/04/01 Feature enhancement release. |
2386 |
|
|
2387 |
|
Fix 2010/04/10 |
2388 |
|
|
2389 |
|
@ Fix invalid "struct nameidata" to "struct path" conversion macro. |
2390 |
|
|
2391 |
|
Regarding kernels 2.6.24 and earlier, I was converting "struct nameidata" |
2392 |
|
to "struct path" in caller side so that I can unify the callee function's |
2393 |
|
parameter type. But it turned out that the macro I used did not follow C |
2394 |
|
standards and did not work with gcc 4.x . As a result, "allow_pivot_root" |
2395 |
|
keyword was not working as expected. |
2396 |
|
|
2397 |
|
Fix 2010/05/05 |
2398 |
|
|
2399 |
|
@ Fix incorrect audit on/off control. |
2400 |
|
|
2401 |
|
The grant_log= and reject_log= parameters of CONFIG::misc::env were not |
2402 |
|
used because I forgot to update request type. As a result, those of |
2403 |
|
CONFIG::file::execute were used for CONFIG::misc::env . |
2404 |
|
|
2405 |
|
Those of CONFIG::file::rewrite were not used because I forgot to update |
2406 |
|
request type. As a result, those of CONFIG::file::truncate were used for |
2407 |
|
CONFIG::file::rewrite . |
2408 |
|
|
2409 |
|
Fix 2010/05/10 |
2410 |
|
|
2411 |
|
@ Fix incorrect out of memory warning. |
2412 |
|
|
2413 |
|
Out of memory warnings were not printed in some cases by error. |
2414 |
|
|
2415 |
|
Fix 2010/05/27 |
2416 |
|
|
2417 |
|
@ Add missing rcu_dereference() for ccs_find_execute_handler(). |
2418 |
|
|
2419 |
|
Since 1.7.0 , ccs_find_execute_handler() was by error using |
2420 |
|
list_for_each_entry() rather than list_for_each_entry_rcu(). |
2421 |
|
This bug affects only Alpha architecture. |
2422 |
|
|
2423 |
|
Fix 2010/06/03 |
2424 |
|
|
2425 |
|
@ Fix missing sanity check for "file_pattern". |
2426 |
|
|
2427 |
|
Since 1.7.0 , ccs_write_pattern_policy() was by error accepting |
2428 |
|
invalid pathname. |
2429 |
|
|
2430 |
|
Fix 2010/06/09 |
2431 |
|
|
2432 |
|
@ Add missing ccs_put_name() in ccs_parse_envp(). |
2433 |
|
|
2434 |
|
Since 1.7.0 , ccs_parse_envp() was not calling ccs_put_name() if |
2435 |
|
environment variable's value ('if exec.envp["name"]="value"' condition) |
2436 |
|
was invalid. |
2437 |
|
|
2438 |
|
@ Add missing NULL check in ccs_condition(). |
2439 |
|
|
2440 |
|
Since 1.7.0 , if 'if symlink.target=' part was given against non-file |
2441 |
|
permissions (e.g. allow_env PATH if symlink.target="/"), it triggered |
2442 |
|
NULL pointer dereference. |
2443 |
|
|
2444 |
|
Fix 2010/10/28 |
2445 |
|
|
2446 |
|
@ Fix umount() pathname calculation. |
2447 |
|
|
2448 |
|
"mount --bind /path/to/file1 /path/to/file2" is legal. |
2449 |
|
Therefore, "umount /path/to/file2" is also legal. |
2450 |
|
Do not automatically append trailing '/' if pathname to be unmounted |
2451 |
|
does not end with '/'. |
2452 |
|
|
2453 |
|
@ Add preserve KABI compatibility option. (2.6 kernels only) |
2454 |
|
|
2455 |
|
TOMOYO needs "struct ccs_domain_info *" and "u32" for each |
2456 |
|
"struct task_struct". But embedding these variables into |
2457 |
|
"struct task_struct" breaks KABI for prebuilt kernel modules (which |
2458 |
|
means that you will need to rebuild prebuilt kernel modules). |
2459 |
|
|
2460 |
|
Since KABI is commonly used (compared to 5 years ago), asking users to |
2461 |
|
rebuild kernel modules which are not included in kernel package is no |
2462 |
|
longer preferable. Therefore, I added a new option that keeps |
2463 |
|
"struct task_struct" unmodified in order to keep KABI. |
2464 |
|
|
2465 |
|
Note that you have to use ccs-patch-2.6.\*.diff which patches |
2466 |
|
kernel/fork.c in order to use this option. Otherwise, TOMOYO will leak |
2467 |
|
memory whenever "struct task_struct" is released. |
2468 |
|
|
2469 |
|
@ Change directives. |
2470 |
|
|
2471 |
|
I removed "allow_" prefix from directives. New directives for files are |
2472 |
|
prefixed with "file ". For example, "allow_read" changed to "file read", |
2473 |
|
"allow_ioctl" changed to "file ioctl". New directive for "allow_network |
2474 |
|
TCP" is "network inet stream", "allow_network UDP" is "network inet |
2475 |
|
dgram", "allow_network RAW" is "network inet raw". New directive for |
2476 |
|
"allow_env" is "misc env". New directive for "allow_signal" is "ipc |
2477 |
|
signal". New directive for "allow_capability" is "capability". These new |
2478 |
|
directives correspond with keywords used by profile's CONFIG lines. |
2479 |
|
|
2480 |
|
I removed "deny_rewrite" and "allow_rewrite" directives and introduced |
2481 |
|
"file append" directive. Thus, permission for open(O_WRONLY | O_APPEND) |
2482 |
|
changed from "allow_write" + "allow_rewrite" to "file append". |
2483 |
|
|
2484 |
|
I removed "SYS_MOUNT", "SYS_UMOUNT", "SYS_CHROOT", "SYS_KILL", |
2485 |
|
"SYS_LINK", "SYS_SYMLINK", "SYS_RENAME", "SYS_UNLINK", "SYS_CHMOD", |
2486 |
|
"SYS_CHOWN", "SYS_IOCTL", "SYS_PIVOT_ROOT" keywords from capabilities |
2487 |
|
because these permissions can be checked by other directives (e.g. |
2488 |
|
"file mount", "ipc signal"). |
2489 |
|
|
2490 |
|
I also removed "conceal_mount" keyword from capabilities because this |
2491 |
|
check requires hooks in filesystem part while almost all hooks for |
2492 |
|
filesystem part have moved to LSM by Linux 2.6.34. |
2493 |
|
|
2494 |
|
New directive for "execute_handler" is "task auto_execute_handler", |
2495 |
|
"denied_execute_handler" is "task denied_execute_handler". |
2496 |
|
|
2497 |
|
@ Distinguish send() and recv() operations. |
2498 |
|
|
2499 |
|
Until now, it was impossible for UDP and IP sockets to allow either |
2500 |
|
only sending or only receiving because permissions were aggregated with |
2501 |
|
"connect" keyword. I broke "connect" keyword into "send" and "recv" |
2502 |
|
keywords so that you can keep access control for send() operation enabled |
2503 |
|
when you have to disable access control for recv() operation due to |
2504 |
|
application breakage by discarding incoming datagram. |
2505 |
|
|
2506 |
|
@ Add Unix domain socket restriction support. |
2507 |
|
|
2508 |
|
Until now, it was possible to restrict only inet domain sockets (i.e. |
2509 |
|
TCP/UDP/RAW). I added restriction for Unix domain sockets (i.e. stream/ |
2510 |
|
dgram/seqpacket). New directive "network unix" is added as well as |
2511 |
|
"network inet" directive. |
2512 |
|
|
2513 |
|
@ Allow specifying multiple permissions in a line. |
2514 |
|
|
2515 |
|
Until now, only "allow_read/write" can be specified for combination of |
2516 |
|
"allow_read" + "allow_write". Now, you can combine other permissions as |
2517 |
|
long as type of parameters for these permissions is same. For example, |
2518 |
|
"file read/write/append/execute/unlink/truncate /tmp/file" is correct |
2519 |
|
but "file read/write/create /tmp/file" is wrong because "file create" |
2520 |
|
requires create mode whereas "file read" and "file write" do not. |
2521 |
|
|
2522 |
|
@ Allow wildcard for execute permission and domainname. |
2523 |
|
|
2524 |
|
Until now, to execute programs with temporary names, "aggregator" is |
2525 |
|
needed. To simplify code, I modified to accept wildcards for execute |
2526 |
|
permission and domainname. Now, you can directly specify |
2527 |
|
"file execute /tmp/logrotate.\?\?\?\?\?\?" and use |
2528 |
|
"/tmp/logrotate.\?\?\?\?\?\?" within domainnames. |
2529 |
|
|
2530 |
|
@ Change pathname for non-rename()able filesystems. |
2531 |
|
|
2532 |
|
LSM version of TOMOYO wants to use /proc/self/ rather than /proc/$PID/ if |
2533 |
|
$PID matches current thread's process ID in order to prevent current |
2534 |
|
thread from accessing other process's information unless needed. |
2535 |
|
But since procfs can be mounted on various locations (e.g. /proc/ /proc2/ |
2536 |
|
/p/ /tmp/foo/100/p/ ), LSM version of TOMOYO cannot tell that whether the |
2537 |
|
numeric part in the string returned by __d_path() represents process ID |
2538 |
|
or not. |
2539 |
|
|
2540 |
|
Therefore, to be able to convert from $PID to self no matter where procfs |
2541 |
|
is mounted, I changed pathname representations for filesystems which do |
2542 |
|
not support rename() operation (e.g. proc, sysfs, securityfs). |
2543 |
|
|
2544 |
|
Now, "/proc/self/mounts" changed to "proc:/self/mounts" and |
2545 |
|
"/sys/kernel/security/" changed to "sys:/kernel/security/" and |
2546 |
|
"/dev/pts/0" changed to "devpts:/0". |
2547 |
|
|
2548 |
|
@ Add a new keyword "any" for domain transition control. |
2549 |
|
|
2550 |
|
To be able to make it easier to apply auto_execute_handler on each |
2551 |
|
domain, I added "any" keyword to domain transition control keywords. Now, |
2552 |
|
"initialize_domain /usr/sbin/sshd" changed to |
2553 |
|
"initialize_domain /usr/sbin/sshd from any" and |
2554 |
|
"keep_domain <kernel> /usr/sbin/sshd /bin/bash" changed to |
2555 |
|
"keep_domain any from <kernel> /usr/sbin/sshd /bin/bash". |
2556 |
|
|
2557 |
|
"keep_domain /path/to/auto_execute_handler from any" will allow you to |
2558 |
|
apply auto_execute_handler for any domains without creating domains for |
2559 |
|
auto_execute_handler. |
2560 |
|
|
2561 |
|
@ Change buffering mode for reading policy. |
2562 |
|
|
2563 |
|
To be able to read() very very long lines correctly, I changed the way |
2564 |
|
TOMOYO buffers policy for reading. |
2565 |
|
|
2566 |
|
@ Introduce "acl_group" keyword. |
2567 |
|
|
2568 |
|
Until now, it was possible to specify only "allow_read" and "allow_env" |
2569 |
|
keywords in the exception policy. |
2570 |
|
|
2571 |
|
Since some operations like "file read/write/append /dev/null" and |
2572 |
|
"network UDP send/recv @DNS_SERVER 53" are very common and should be |
2573 |
|
permitted to all domains, I introduced "acl_group" keyword for giving |
2574 |
|
such permissions. |
2575 |
|
|
2576 |
|
For example, specify "acl_group 0 file read/write/append /dev/null" in |
2577 |
|
the exception policy and specify "use_group 0" from the domains in the |
2578 |
|
domain policy. |
2579 |
|
|
2580 |
|
"ignore_global_allow_read" and "ignore_global_allow_env" directives were |
2581 |
|
removed from domain policy and "use_group" keyword was added. |
2582 |
|
|
2583 |
|
@ Remove "if" and "; set" keyword. |
2584 |
|
|
2585 |
|
I removed need for specifying these keyword. |
2586 |
|
You can simply specify like below. |
2587 |
|
|
2588 |
|
file read /etc/shadow task.uid=0 |
2589 |
|
|
2590 |
|
@ Remove "file_pattern" keyword. |
2591 |
|
|
2592 |
|
I removed "file_pattern" keyword because it is impossible to predefine |
2593 |
|
all possible pathname patterns. Also, learning pathnames using incomplete |
2594 |
|
patterns makes it difficult to later replace using "path_group" keyword. |
2595 |
|
|
2596 |
|
@ Replace verbose= parameter with statistic interface. |
2597 |
|
|
2598 |
|
Since it is noisy if a lot of policy violation messages are printed, |
2599 |
|
I removed printk(). To be able to check whether policy violation occurred |
2600 |
|
or not, I introduced /proc/ccs/stat interface which counts number of |
2601 |
|
policy violations occurred. You can firstly check /proc/ccs/stat and then |
2602 |
|
check /proc/ccs/reject_log . |
2603 |
|
|
2604 |
|
@ Remove global preference. |
2605 |
|
|
2606 |
|
I removed global preference in order to make code simpler. |
2607 |
|
|
2608 |
|
@ Allow controlling generation of access granted logs for per an entry |
2609 |
|
basis. |
2610 |
|
|
2611 |
|
I added per-entry flag which controls generation of grant logs because |
2612 |
|
Xen and KVM issues ioctl requests so frequently. For example, |
2613 |
|
|
2614 |
|
file ioctl /dev/null 0x5401 grant_log=no |
2615 |
|
|
2616 |
|
will suppress /proc/ccs/grant_log even if preference says grant_log=yes . |
2617 |
|
|
2618 |
|
file ioctl /dev/null 0x5401 grant_log=yes |
2619 |
|
|
2620 |
|
will generate /proc/ccs/grant_log even if preference says grant_log=no . |
2621 |
|
|
2622 |
|
file ioctl /dev/null 0x5401 |
2623 |
|
|
2624 |
|
will generate /proc/ccs/grant_log only if preference says grant_log=yes . |
2625 |
|
|
2626 |
|
This flag is intended for frequently accessed resources like |
2627 |
|
|
2628 |
|
file read /var/www/html/\{\*\}/\*.html grant_log=no |
2629 |
|
|
2630 |
|
. |
2631 |
|
|
2632 |
|
@ Automatically create domain by execve() even if enforcing mode. |
2633 |
|
|
2634 |
|
Until now, new domains are not created if the domain was not defined and |
2635 |
|
current domain is enforcing mode ("CONFIG::file::execute=enforcing"). |
2636 |
|
|
2637 |
|
To be able to restrict shell session without using "keep_domain", |
2638 |
|
I changed to create new domains automatically even if current domain is |
2639 |
|
enforcing mode. |
2640 |
|
|
2641 |
|
@ Replace "task.state" with "auto_domain_transition". |
2642 |
|
|
2643 |
|
task.state is difficult to use. Thus, I replaced task.state with |
2644 |
|
auto_domain_transition which performs domain transition instead of |
2645 |
|
changing current process's state variables. |
2646 |
|
|
2647 |
|
If domain transition failed, current process will be killed by SIGKILL |
2648 |
|
signal. This should not happen in normal circumstances, for you know the |
2649 |
|
domain to transit to and thereby you will define the domain beforehand |
2650 |
|
when you use "auto_domain_transition" keyword. |
2651 |
|
|
2652 |
|
@ Replace "allow_transit" with "task manual_domain_transition". |
2653 |
|
|
2654 |
|
I changed this directive to specify absolute domainname (e.g. |
2655 |
|
"<kernel> /usr/sbin/httpd //app=cgi1\040id=10000") rather than virtual |
2656 |
|
pathname (e.g. "//app=cgi1\040id=10000") because you know the domain to |
2657 |
|
transit to and thereby you will define the domain beforehand when you use |
2658 |
|
"task manual_domain_transition" directive. |
2659 |
|
|
2660 |
|
This change allows you to jump to arbitrary domain. |
2661 |
|
|
2662 |
|
@ Add "task auto_domain_transition". |
2663 |
|
|
2664 |
|
This is similar to "task manual_domain_transition", but is automatically |
2665 |
|
applied whenever conditions are met. For example, |
2666 |
|
|
2667 |
|
task auto_domain_transition <kernel> //./non-root task.uid!=0 |
2668 |
|
|
2669 |
|
will automatically jump to "<kernel> //./non-root" domain if current |
2670 |
|
process's UID is not 0 whereas |
2671 |
|
|
2672 |
|
task manual_domain_transition <kernel> //./non-root task.uid!=0 |
2673 |
|
|
2674 |
|
will jump to "<kernel> //./non-root" domain if current process's UID is |
2675 |
|
not 0 and current process wrote "<kernel> //./non-root" + '\n' to |
2676 |
|
/proc/ccs/.transition interface. |
2677 |
|
|
2678 |
|
If domain transition failed, current process will be killed by SIGKILL |
2679 |
|
signal. |
2680 |
|
|
2681 |
|
@ Optimize for object's size. |
2682 |
|
|
2683 |
|
I merged similar code in order to reduce object's filesize. |
2684 |
|
|
2685 |
|
Version 1.8.0 2010/XX/XX Feature enhancement release. |