オープンソース・ソフトウェアの開発とダウンロード

Subversion リポジトリの参照

Diff of /trunk/1.8.x/ccs-patch/README.ccs

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

trunk/1.7.x/ccs-patch/README.ccs revision 3502 by kumaneko, Mon Mar 8 08:44:55 2010 UTC trunk/1.8.x/ccs-patch/README.ccs revision 4170 by kumaneko, Wed Dec 1 06:53:26 2010 UTC
# Line 1723  Fix 2008/11/07 Line 1723  Fix 2008/11/07
1723        to see process information. The content of /proc/ccs/.execute_handler is        to see process information. The content of /proc/ccs/.execute_handler is
1724        identical to /proc/ccs/.process_status .        identical to /proc/ccs/.process_status .
1725    
1726  Version 1.6.5   2008/11/11   Third anniversary release.  Version 1.6.5 2008/11/11   Third anniversary release.
1727    
1728  Fix 2008/12/01  Fix 2008/12/01
1729    
# Line 2243  Fix 2009/11/09 Line 2243  Fix 2009/11/09
2243        /proc/ccs/profile (which results in not protecting the system at all),        /proc/ccs/profile (which results in not protecting the system at all),
2244        I added a check for PROFILE_VERSION= .        I added a check for PROFILE_VERSION= .
2245    
2246  Version 1.7.1   2009/11/11   Fourth anniversary release.  Version 1.7.1 2009/11/11   Fourth anniversary release.
2247    
2248  Fix 2009/11/13  Fix 2009/11/13
2249    
# Line 2342  Fix 2010/03/02 Line 2342  Fix 2010/03/02
2342    
2343        Note that '\0'-terminated binary string is converted to TOMOYO's string        Note that '\0'-terminated binary string is converted to TOMOYO's string
2344        inside kernel and prefix "//" is automatically added to the string so        inside kernel and prefix "//" is automatically added to the string so
2345        that domainname does not confclict with domainnames created by execve().        that domainname does not conflict with domainnames created by execve().
2346        Without this prefix, if "<kernel> /usr/sbin/sshd /bin/bash" domain is        Without this prefix, if "<kernel> /usr/sbin/sshd /bin/bash" domain is
2347        allowed to open /proc/ccs/.transition for writing and        allowed to open /proc/ccs/.transition for writing and
2348        "<kernel> /usr/sbin/sshd /bin/bash /usr/bin/passwd" domain is allowed to        "<kernel> /usr/sbin/sshd /bin/bash /usr/bin/passwd" domain is allowed to
# Line 2360  Fix 2010/03/08 Line 2360  Fix 2010/03/08
2360    
2361      @ Allow building as loadable kernel module.      @ Allow building as loadable kernel module.
2362    
2363        To be able to minimize filesize increment on vmlinux, I made it        To be able to minimize filesize increment of vmlinux, I made it
2364        possible to compile TOMOYO Linux as loadable kernel module.        possible to compile TOMOYO Linux as loadable kernel module.
2365        Although patching the kernel source and recompiling the kernel are        Although patching the kernel source and recompiling the kernel are
2366        inevitable, this change will make it easier to enable TOMOYO Linux        inevitable, this change will make it easier to enable TOMOYO Linux
2367        when there is a filesize limitation on vmlinux (e.g. embedded systems).        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" to
2676          /proc/ccs/self_domain 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 2010/11/11   Fifth anniversary release.
2686    
2687    Fix 2010/12/01
2688    
2689        @ Use same interface for audit logs.
2690    
2691          To be able to perform fine grained filtering by /usr/sbin/ccs-auditd ,
2692          I merged /proc/ccs/grant_log and /proc/ccs/reject_log as
2693          /proc/ccs/audit and added granted=yes or granted=no to audit logs.

Legend:
Removed from v.3502  
changed lines
  Added in v.4170

Back to OSDN">Back to OSDN
ViewVC Help
Powered by ViewVC 1.1.26