Merge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security

Pull security subsystem updates from James Morris:
 "Highlights:

   - TPM core and driver updates/fixes
   - IPv6 security labeling (CALIPSO)
   - Lots of Apparmor fixes
   - Seccomp: remove 2-phase API, close hole where ptrace can change
     syscall #"

* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security: (156 commits)
  apparmor: fix SECURITY_APPARMOR_HASH_DEFAULT parameter handling
  tpm: Add TPM 2.0 support to the Nuvoton i2c driver (NPCT6xx family)
  tpm: Factor out common startup code
  tpm: use devm_add_action_or_reset
  tpm2_i2c_nuvoton: add irq validity check
  tpm: read burstcount from TPM_STS in one 32-bit transaction
  tpm: fix byte-order for the value read by tpm2_get_tpm_pt
  tpm_tis_core: convert max timeouts from msec to jiffies
  apparmor: fix arg_size computation for when setprocattr is null terminated
  apparmor: fix oops, validate buffer size in apparmor_setprocattr()
  apparmor: do not expose kernel stack
  apparmor: fix module parameters can be changed after policy is locked
  apparmor: fix oops in profile_unpack() when policy_db is not present
  apparmor: don't check for vmalloc_addr if kvzalloc() failed
  apparmor: add missing id bounds check on dfa verification
  apparmor: allow SYS_CAP_RESOURCE to be sufficient to prlimit another task
  apparmor: use list_next_entry instead of list_entry_next
  apparmor: fix refcount race when finding a child profile
  apparmor: fix ref count leak when profile sha1 hash is read
  apparmor: check that xindex is in trans_table bounds
  ...

1 files changed, 4 insertions, 2 deletions

diff --git a/security/apparmor/resource.c b/security/apparmor/resource.c
index 748bf0ca6c9f..67a6072ead4b 100644
--- a/security/apparmor/resource.c
+++ b/security/apparmor/resource.c
@@ -101,9 +101,11 @@ int aa_task_setrlimit(struct aa_profile *profile, struct task_struct *task,
 	/* TODO: extend resource control to handle other (non current)
 	 * profiles.  AppArmor rules currently have the implicit assumption
 	 * that the task is setting the resource of a task confined with
-	 * the same profile.
+	 * the same profile or that the task setting the resource of another
+	 * task has CAP_SYS_RESOURCE.
 	 */
-	if (profile != task_profile ||
+	if ((profile != task_profile &&
+	     aa_capable(profile, CAP_SYS_RESOURCE, 1)) ||
 	    (profile->rlimits.mask & (1 << resource) &&
 	     new_rlim->rlim_max > profile->rlimits.limits[resource].rlim_max))
 		error = -EACCES;