)]}'
{
  "log": [
    {
      "commit": "d84f4f992cbd76e8f39c488cf0c5d123843923b1",
      "tree": "fc4a0349c42995715b93d0f7a3c78e9ea9b3f36e",
      "parents": [
        "745ca2475a6ac596e3d8d37c2759c0fbe2586227"
      ],
      "author": {
        "name": "David Howells",
        "email": "dhowells@redhat.com",
        "time": "Fri Nov 14 10:39:23 2008 +1100"
      },
      "committer": {
        "name": "James Morris",
        "email": "jmorris@namei.org",
        "time": "Fri Nov 14 10:39:23 2008 +1100"
      },
      "message": "CRED: Inaugurate COW credentials\n\nInaugurate copy-on-write credentials management.  This uses RCU to manage the\ncredentials pointer in the task_struct with respect to accesses by other tasks.\nA process may only modify its own credentials, and so does not need locking to\naccess or modify its own credentials.\n\nA mutex (cred_replace_mutex) is added to the task_struct to control the effect\nof PTRACE_ATTACHED on credential calculations, particularly with respect to\nexecve().\n\nWith this patch, the contents of an active credentials struct may not be\nchanged directly; rather a new set of credentials must be prepared, modified\nand committed using something like the following sequence of events:\n\n\tstruct cred *new \u003d prepare_creds();\n\tint ret \u003d blah(new);\n\tif (ret \u003c 0) {\n\t\tabort_creds(new);\n\t\treturn ret;\n\t}\n\treturn commit_creds(new);\n\nThere are some exceptions to this rule: the keyrings pointed to by the active\ncredentials may be instantiated - keyrings violate the COW rule as managing\nCOW keyrings is tricky, given that it is possible for a task to directly alter\nthe keys in a keyring in use by another task.\n\nTo help enforce this, various pointers to sets of credentials, such as those in\nthe task_struct, are declared const.  The purpose of this is compile-time\ndiscouragement of altering credentials through those pointers.  Once a set of\ncredentials has been made public through one of these pointers, it may not be\nmodified, except under special circumstances:\n\n  (1) Its reference count may incremented and decremented.\n\n  (2) The keyrings to which it points may be modified, but not replaced.\n\nThe only safe way to modify anything else is to create a replacement and commit\nusing the functions described in Documentation/credentials.txt (which will be\nadded by a later patch).\n\nThis patch and the preceding patches have been tested with the LTP SELinux\ntestsuite.\n\nThis patch makes several logical sets of alteration:\n\n (1) execve().\n\n     This now prepares and commits credentials in various places in the\n     security code rather than altering the current creds directly.\n\n (2) Temporary credential overrides.\n\n     do_coredump() and sys_faccessat() now prepare their own credentials and\n     temporarily override the ones currently on the acting thread, whilst\n     preventing interference from other threads by holding cred_replace_mutex\n     on the thread being dumped.\n\n     This will be replaced in a future patch by something that hands down the\n     credentials directly to the functions being called, rather than altering\n     the task\u0027s objective credentials.\n\n (3) LSM interface.\n\n     A number of functions have been changed, added or removed:\n\n     (*) security_capset_check(), -\u003ecapset_check()\n     (*) security_capset_set(), -\u003ecapset_set()\n\n     \t Removed in favour of security_capset().\n\n     (*) security_capset(), -\u003ecapset()\n\n     \t New.  This is passed a pointer to the new creds, a pointer to the old\n     \t creds and the proposed capability sets.  It should fill in the new\n     \t creds or return an error.  All pointers, barring the pointer to the\n     \t new creds, are now const.\n\n     (*) security_bprm_apply_creds(), -\u003ebprm_apply_creds()\n\n     \t Changed; now returns a value, which will cause the process to be\n     \t killed if it\u0027s an error.\n\n     (*) security_task_alloc(), -\u003etask_alloc_security()\n\n     \t Removed in favour of security_prepare_creds().\n\n     (*) security_cred_free(), -\u003ecred_free()\n\n     \t New.  Free security data attached to cred-\u003esecurity.\n\n     (*) security_prepare_creds(), -\u003ecred_prepare()\n\n     \t New. Duplicate any security data attached to cred-\u003esecurity.\n\n     (*) security_commit_creds(), -\u003ecred_commit()\n\n     \t New. Apply any security effects for the upcoming installation of new\n     \t security by commit_creds().\n\n     (*) security_task_post_setuid(), -\u003etask_post_setuid()\n\n     \t Removed in favour of security_task_fix_setuid().\n\n     (*) security_task_fix_setuid(), -\u003etask_fix_setuid()\n\n     \t Fix up the proposed new credentials for setuid().  This is used by\n     \t cap_set_fix_setuid() to implicitly adjust capabilities in line with\n     \t setuid() changes.  Changes are made to the new credentials, rather\n     \t than the task itself as in security_task_post_setuid().\n\n     (*) security_task_reparent_to_init(), -\u003etask_reparent_to_init()\n\n     \t Removed.  Instead the task being reparented to init is referred\n     \t directly to init\u0027s credentials.\n\n\t NOTE!  This results in the loss of some state: SELinux\u0027s osid no\n\t longer records the sid of the thread that forked it.\n\n     (*) security_key_alloc(), -\u003ekey_alloc()\n     (*) security_key_permission(), -\u003ekey_permission()\n\n     \t Changed.  These now take cred pointers rather than task pointers to\n     \t refer to the security context.\n\n (4) sys_capset().\n\n     This has been simplified and uses less locking.  The LSM functions it\n     calls have been merged.\n\n (5) reparent_to_kthreadd().\n\n     This gives the current thread the same credentials as init by simply using\n     commit_thread() to point that way.\n\n (6) __sigqueue_alloc() and switch_uid()\n\n     __sigqueue_alloc() can\u0027t stop the target task from changing its creds\n     beneath it, so this function gets a reference to the currently applicable\n     user_struct which it then passes into the sigqueue struct it returns if\n     successful.\n\n     switch_uid() is now called from commit_creds(), and possibly should be\n     folded into that.  commit_creds() should take care of protecting\n     __sigqueue_alloc().\n\n (7) [sg]et[ug]id() and co and [sg]et_current_groups.\n\n     The set functions now all use prepare_creds(), commit_creds() and\n     abort_creds() to build and check a new set of credentials before applying\n     it.\n\n     security_task_set[ug]id() is called inside the prepared section.  This\n     guarantees that nothing else will affect the creds until we\u0027ve finished.\n\n     The calling of set_dumpable() has been moved into commit_creds().\n\n     Much of the functionality of set_user() has been moved into\n     commit_creds().\n\n     The get functions all simply access the data directly.\n\n (8) security_task_prctl() and cap_task_prctl().\n\n     security_task_prctl() has been modified to return -ENOSYS if it doesn\u0027t\n     want to handle a function, or otherwise return the return value directly\n     rather than through an argument.\n\n     Additionally, cap_task_prctl() now prepares a new set of credentials, even\n     if it doesn\u0027t end up using it.\n\n (9) Keyrings.\n\n     A number of changes have been made to the keyrings code:\n\n     (a) switch_uid_keyring(), copy_keys(), exit_keys() and suid_keys() have\n     \t all been dropped and built in to the credentials functions directly.\n     \t They may want separating out again later.\n\n     (b) key_alloc() and search_process_keyrings() now take a cred pointer\n     \t rather than a task pointer to specify the security context.\n\n     (c) copy_creds() gives a new thread within the same thread group a new\n     \t thread keyring if its parent had one, otherwise it discards the thread\n     \t keyring.\n\n     (d) The authorisation key now points directly to the credentials to extend\n     \t the search into rather pointing to the task that carries them.\n\n     (e) Installing thread, process or session keyrings causes a new set of\n     \t credentials to be created, even though it\u0027s not strictly necessary for\n     \t process or session keyrings (they\u0027re shared).\n\n(10) Usermode helper.\n\n     The usermode helper code now carries a cred struct pointer in its\n     subprocess_info struct instead of a new session keyring pointer.  This set\n     of credentials is derived from init_cred and installed on the new process\n     after it has been cloned.\n\n     call_usermodehelper_setup() allocates the new credentials and\n     call_usermodehelper_freeinfo() discards them if they haven\u0027t been used.  A\n     special cred function (prepare_usermodeinfo_creds()) is provided\n     specifically for call_usermodehelper_setup() to call.\n\n     call_usermodehelper_setkeys() adjusts the credentials to sport the\n     supplied keyring as the new session keyring.\n\n(11) SELinux.\n\n     SELinux has a number of changes, in addition to those to support the LSM\n     interface changes mentioned above:\n\n     (a) selinux_setprocattr() no longer does its check for whether the\n     \t current ptracer can access processes with the new SID inside the lock\n     \t that covers getting the ptracer\u0027s SID.  Whilst this lock ensures that\n     \t the check is done with the ptracer pinned, the result is only valid\n     \t until the lock is released, so there\u0027s no point doing it inside the\n     \t lock.\n\n(12) is_single_threaded().\n\n     This function has been extracted from selinux_setprocattr() and put into\n     a file of its own in the lib/ directory as join_session_keyring() now\n     wants to use it too.\n\n     The code in SELinux just checked to see whether a task shared mm_structs\n     with other tasks (CLONE_VM), but that isn\u0027t good enough.  We really want\n     to know if they\u0027re part of the same thread group (CLONE_THREAD).\n\n(13) nfsd.\n\n     The NFS server daemon now has to use the COW credentials to set the\n     credentials it is going to use.  It really needs to pass the credentials\n     down to the functions it calls, but it can\u0027t do that until other patches\n     in this series have been applied.\n\nSigned-off-by: David Howells \u003cdhowells@redhat.com\u003e\nAcked-by: James Morris \u003cjmorris@namei.org\u003e\nSigned-off-by: James Morris \u003cjmorris@namei.org\u003e\n"
    }
  ]
}