kernel/linux.git/security, branch v5.15.3

apparmor: fix error check

2021-11-18T18:16:58+00:00

[ Upstream commit d108370c644b153382632b3e5511ade575c91c86 ] clang static analysis reports this representative problem: label.c:1463:16: warning: Assigned value is garbage or undefined label->hname = name; ^ ~~~~ In aa_update_label_name(), this the problem block of code if (aa_label_acntsxprint(&name, ...) == -1) return res; On failure, aa_label_acntsxprint() has a more complicated return that just -1. So check for a negative return. It was also noted that the aa_label_acntsxprint() main comment refers to a nonexistent parameter, so clean up the comment. Fixes: f1bd904175e8 ("apparmor: add the base fns() for domain labels") Signed-off-by: Tom Rix Reviewed-by: Nick Desaulniers Signed-off-by: John Johansen Signed-off-by: Sasha Levin

smackfs: use netlbl_cfg_cipsov4_del() for deleting cipso_v4_doi

2021-11-18T18:16:39+00:00

[ Upstream commit 0934ad42bb2c5df90a1b9de690f93de735b622fe ] syzbot is reporting UAF at cipso_v4_doi_search() [1], for smk_cipso_doi() is calling kfree() without removing from the cipso_v4_doi_list list after netlbl_cfg_cipsov4_map_add() returned an error. We need to use netlbl_cfg_cipsov4_del() in order to remove from the list and wait for RCU grace period before kfree(). Link: https://syzkaller.appspot.com/bug?extid=93dba5b91f0fed312cbd [1] Reported-by: syzbot Signed-off-by: Tetsuo Handa Fixes: 6c2e8ac0953fccdd ("netlabel: Update kernel configuration API") Signed-off-by: Casey Schaufler Signed-off-by: Sasha Levin

ima: fix deadlock when traversing "ima_default_rules".

2021-11-18T18:16:31+00:00

[ Upstream commit eb0782bbdfd0d7c4786216659277c3fd585afc0e ] The current IMA ruleset is identified by the variable "ima_rules" that default to "&ima_default_rules". When loading a custom policy for the first time, the variable is updated to "&ima_policy_rules" instead. That update isn't RCU-safe, and deadlocks are possible. Indeed, some functions like ima_match_policy() may loop indefinitely when traversing "ima_default_rules" with list_for_each_entry_rcu(). When iterating over the default ruleset back to head, if the list head is "ima_default_rules", and "ima_rules" have been updated to "&ima_policy_rules", the loop condition (&entry->list != ima_rules) stays always true, traversing won't terminate, causing a soft lockup and RCU stalls. Introduce a temporary value for "ima_rules" when iterating over the ruleset to avoid the deadlocks. Signed-off-by: liqiong Reviewed-by: THOBY Simon Fixes: 38d859f991f3 ("IMA: policy can now be updated multiple times") Reported-by: kernel test robot (Fix sparse: incompatible types in comparison expression.) Signed-off-by: Mimi Zohar Signed-off-by: Sasha Levin

smackfs: use __GFP_NOFAIL for smk_cipso_doi()

2021-11-18T18:16:19+00:00

[ Upstream commit f91488ee15bd3cac467e2d6a361fc2d34d1052ae ] syzbot is reporting kernel panic at smk_cipso_doi() due to memory allocation fault injection [1]. The reason for need to use panic() was not explained. But since no fix was proposed for 18 months, for now let's use __GFP_NOFAIL for utilizing syzbot resource on other bugs. Link: https://syzkaller.appspot.com/bug?extid=89731ccb6fec15ce1c22 [1] Reported-by: syzbot Signed-off-by: Tetsuo Handa Signed-off-by: Casey Schaufler Signed-off-by: Sasha Levin

smackfs: Fix use-after-free in netlbl_catmap_walk()

2021-11-18T18:16:10+00:00

[ Upstream commit 0817534ff9ea809fac1322c5c8c574be8483ea57 ] Syzkaller reported use-after-free bug as described in [1]. The bug is triggered when smk_set_cipso() tries to free stale category bitmaps while there are concurrent reader(s) using the same bitmaps. Wait for RCU grace period to finish before freeing the category bitmaps in smk_set_cipso(). This makes sure that there are no more readers using the stale bitmaps and freeing them should be safe. [1] https://lore.kernel.org/netdev/000000000000a814c505ca657a4e@google.com/ Reported-by: syzbot+3f91de0b813cc3d19a80@syzkaller.appspotmail.com Signed-off-by: Pawan Gupta Signed-off-by: Casey Schaufler Signed-off-by: Sasha Levin

evm: mark evm_fixmode as __ro_after_init

2021-11-18T18:15:58+00:00

commit 32ba540f3c2a7ef61ed5a577ce25069a3d714fc9 upstream. The evm_fixmode is only configurable by command-line option and it is never modified outside initcalls, so declaring it with __ro_after_init is better. Signed-off-by: Austin Kim Cc: stable@vger.kernel.org Signed-off-by: Mimi Zohar Signed-off-by: Greg Kroah-Hartman

selinux: fix race condition when computing ocontext SIDs

2021-11-18T18:15:57+00:00

commit cbfcd13be5cb2a07868afe67520ed181956579a7 upstream. Current code contains a lot of racy patterns when converting an ocontext's context structure to an SID. This is being done in a "lazy" fashion, such that the SID is looked up in the SID table only when it's first needed and then cached in the "sid" field of the ocontext structure. However, this is done without any locking or memory barriers and is thus unsafe. Between commits 24ed7fdae669 ("selinux: use separate table for initial SID lookup") and 66f8e2f03c02 ("selinux: sidtab reverse lookup hash table"), this race condition lead to an actual observable bug, because a pointer to the shared sid field was passed directly to sidtab_context_to_sid(), which was using this location to also store an intermediate value, which could have been read by other threads and interpreted as an SID. In practice this caused e.g. new mounts to get a wrong (seemingly random) filesystem context, leading to strange denials. This bug has been spotted in the wild at least twice, see [1] and [2]. Fix the race condition by making all the racy functions use a common helper that ensures the ocontext::sid accesses are made safely using the appropriate SMP constructs. Note that security_netif_sid() was populating the sid field of both contexts stored in the ocontext, but only the first one was actually used. The SELinux wiki's documentation on the "netifcon" policy statement [3] suggests that using only the first context is intentional. I kept only the handling of the first context here, as there is really no point in doing the SID lookup for the unused one. I wasn't able to reproduce the bug mentioned above on any kernel that includes commit 66f8e2f03c02, even though it has been reported that the issue occurs with that commit, too, just less frequently. Thus, I wasn't able to verify that this patch fixes the issue, but it makes sense to avoid the race condition regardless. [1] https://github.com/containers/container-selinux/issues/89 [2] https://lists.fedoraproject.org/archives/list/selinux@lists.fedoraproject.org/thread/6DMTAMHIOAOEMUAVTULJD45JZU7IBAFM/ [3] https://selinuxproject.org/page/NetworkStatements#netifcon Cc: stable@vger.kernel.org Cc: Xinjie Zheng Reported-by: Sujithra Periasamy Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2") Signed-off-by: Ondrej Mosnacek Signed-off-by: Paul Moore Signed-off-by: Greg Kroah-Hartman

binder: use cred instead of task for selinux checks

2021-11-12T14:05:48+00:00

commit 52f88693378a58094c538662ba652aff0253c4fe upstream. Since binder was integrated with selinux, it has passed 'struct task_struct' associated with the binder_proc to represent the source and target of transactions. The conversion of task to SID was then done in the hook implementations. It turns out that there are race conditions which can result in an incorrect security context being used. Fix by using the 'struct cred' saved during binder_open and pass it to the selinux subsystem. Cc: stable@vger.kernel.org # 5.14 (need backport for earlier stables) Fixes: 79af73079d75 ("Add security hooks to binder and implement the hooks for SELinux.") Suggested-by: Jann Horn Signed-off-by: Todd Kjos Acked-by: Casey Schaufler Signed-off-by: Paul Moore Signed-off-by: Greg Kroah-Hartman

Merge branch 'ucount-fixes-for-v5.15' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace

2021-10-22T03:27:17+00:00

Pull ucounts fixes from Eric Biederman: "There has been one very hard to track down bug in the ucount code that we have been tracking since roughly v5.14 was released. Alex managed to find a reliable reproducer a few days ago and then I was able to instrument the code and figure out what the issue was. It turns out the sigqueue_alloc single atomic operation optimization did not play nicely with ucounts multiple level rlimits. It turned out that either sigqueue_alloc or sigqueue_free could be operating on multiple levels and trigger the conditions for the optimization on more than one level at the same time. To deal with that situation I have introduced inc_rlimit_get_ucounts and dec_rlimit_put_ucounts that just focuses on the optimization and the rlimit and ucount changes. While looking into the big bug I found I couple of other little issues so I am including those fixes here as well. When I have time I would very much like to dig into process ownership of the shared signal queue and see if we could pick a single owner for the entire queue so that all of the rlimits can count to that owner. That should entirely remove the need to call get_ucounts and put_ucounts in sigqueue_alloc and sigqueue_free. It is difficult because Linux unlike POSIX supports setuid that works on a single thread" * 'ucount-fixes-for-v5.15' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace: ucounts: Move get_ucounts from cred_alloc_blank to key_change_session_keyring ucounts: Proper error handling in set_cred_ucounts ucounts: Pair inc_rlimit_ucounts with dec_rlimit_ucoutns in commit_creds ucounts: Fix signal ucount refcounting

ucounts: Move get_ucounts from cred_alloc_blank to key_change_session_keyring

2021-10-20T15:34:20+00:00

Setting cred->ucounts in cred_alloc_blank does not make sense. The uid and user_ns are deliberately not set in cred_alloc_blank but instead the setting is delayed until key_change_session_keyring. So move dealing with ucounts into key_change_session_keyring as well. Unfortunately that movement of get_ucounts adds a new failure mode to key_change_session_keyring. I do not see anything stopping the parent process from calling setuid and changing the relevant part of it's cred while keyctl_session_to_parent is running making it fundamentally necessary to call get_ucounts in key_change_session_keyring. Which means that the new failure mode cannot be avoided. A failure of key_change_session_keyring results in a single threaded parent keeping it's existing credentials. Which results in the parent process not being able to access the session keyring and whichever keys are in the new keyring. Further get_ucounts is only expected to fail if the number of bits in the refernece count for the structure is too few. Since the code has no other way to report the failure of get_ucounts and because such failures are not expected to be common add a WARN_ONCE to report this problem to userspace. Between the WARN_ONCE and the parent process not having access to the keys in the new session keyring I expect any failure of get_ucounts will be noticed and reported and we can find another way to handle this condition. (Possibly by just making ucounts->count an atomic_long_t). Cc: stable@vger.kernel.org Fixes: 905ae01c4ae2 ("Add a reference to ucounts for each cred") Link: https://lkml.kernel.org/r/7k0ias0uf.fsf_-_@disp2133 Tested-by: Yu Zhao Reviewed-by: Alexey Gladkov Signed-off-by: "Eric W. Biederman"