ima: Avoid blocking in RCU read-side critical section

A panic happens in ima_match_policy:

BUG: unable to handle kernel NULL pointer dereference at 0000000000000010
PGD 42f873067 P4D 0
Oops: 0000 [#1] SMP NOPTI
CPU: 5 PID: 1286325 Comm: kubeletmonit.sh
Kdump: loaded Tainted: P
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
BIOS 0.0.0 02/06/2015
RIP: 0010:ima_match_policy+0x84/0x450
Code: 49 89 fc 41 89 cf 31 ed 89 44 24 14 eb 1c 44 39
7b 18 74 26 41 83 ff 05 74 20 48 8b 1b 48 3b 1d
f2 b9 f4 00 0f 84 9c 01 00 00 <44> 85 73 10 74 ea
44 8b 6b 14 41 f6 c5 01 75 d4 41 f6 c5 02 74 0f
RSP: 0018:ff71570009e07a80 EFLAGS: 00010207
RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000200
RDX: ffffffffad8dc7c0 RSI: 0000000024924925 RDI: ff3e27850dea2000
RBP: 0000000000000000 R08: 0000000000000000 R09: ffffffffabfce739
R10: ff3e27810cc42400 R11: 0000000000000000 R12: ff3e2781825ef970
R13: 00000000ff3e2785 R14: 000000000000000c R15: 0000000000000001
FS: 00007f5195b51740(0000)
GS:ff3e278b12d40000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000010 CR3: 0000000626d24002 CR4: 0000000000361ee0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
ima_get_action+0x22/0x30
process_measurement+0xb0/0x830
? page_add_file_rmap+0x15/0x170
? alloc_set_pte+0x269/0x4c0
? prep_new_page+0x81/0x140
? simple_xattr_get+0x75/0xa0
? selinux_file_open+0x9d/0xf0
ima_file_check+0x64/0x90
path_openat+0x571/0x1720
do_filp_open+0x9b/0x110
? page_counter_try_charge+0x57/0xc0
? files_cgroup_alloc_fd+0x38/0x60
? __alloc_fd+0xd4/0x250
? do_sys_open+0x1bd/0x250
do_sys_open+0x1bd/0x250
do_syscall_64+0x5d/0x1d0
entry_SYSCALL_64_after_hwframe+0x65/0xca

Commit c7423dbdbc9e ("ima: Handle -ESTALE returned by
ima_filter_rule_match()") introduced call to ima_lsm_copy_rule within a
RCU read-side critical section which contains kmalloc with GFP_KERNEL.
This implies a possible sleep and violates limitations of RCU read-side
critical sections on non-PREEMPT systems.

Sleeping within RCU read-side critical section might cause
synchronize_rcu() returning early and break RCU protection, allowing a
UAF to happen.

The root cause of this issue could be described as follows:
| Thread A | Thread B |
| |ima_match_policy |
| | rcu_read_lock |
|ima_lsm_update_rule | |
| synchronize_rcu | |
| | kmalloc(GFP_KERNEL)|
| | sleep |
==> synchronize_rcu returns early
| kfree(entry) | |
| | entry = entry->next|
==> UAF happens and entry now becomes NULL (or could be anything).
| | entry->action |
==> Accessing entry might cause panic.

To fix this issue, we are converting all kmalloc that is called within
RCU read-side critical section to use GFP_ATOMIC.

Fixes: c7423dbdbc9e ("ima: Handle -ESTALE returned by ima_filter_rule_match()")
Cc: [e-mail]
Signed-off-by: GUO Zihua <[e-mail]>
Acked-by: John Johansen <[e-mail]>
Reviewed-by: Mimi Zohar <[e-mail]>
Reviewed-by: Casey Schaufler <[e-mail]>
[PM: fixed missing comment, long lines, !CONFIG_IMA_LSM_RULES case]
Signed-off-by: Paul Moore <[e-mail]>

include/linux/lsm_hook_defs.h

···
 
 #ifdef CONFIG_AUDIT
 LSM_HOOK(int, 0, audit_rule_init, u32 field, u32 op, char *rulestr,
-	 void **lsmrule)
+	 void **lsmrule, gfp_t gfp)
 LSM_HOOK(int, 0, audit_rule_known, struct audit_krule *krule)
 LSM_HOOK(int, 0, audit_rule_match, u32 secid, u32 field, u32 op, void *lsmrule)
 LSM_HOOK(void, LSM_RET_VOID, audit_rule_free, void *lsmrule)

include/linux/security.h

···
 
 #ifdef CONFIG_AUDIT
 #ifdef CONFIG_SECURITY
-int security_audit_rule_init(u32 field, u32 op, char *rulestr, void **lsmrule);
+int security_audit_rule_init(u32 field, u32 op, char *rulestr, void **lsmrule,
+			     gfp_t gfp);
 int security_audit_rule_known(struct audit_krule *krule);
 int security_audit_rule_match(u32 secid, u32 field, u32 op, void *lsmrule);
 void security_audit_rule_free(void *lsmrule);
···
 #else
 
 static inline int security_audit_rule_init(u32 field, u32 op, char *rulestr,
-					   void **lsmrule)
+					   void **lsmrule, gfp_t gfp)
 {
 	return 0;
 }

kernel/auditfilter.c

···
 			entry->rule.buflen += f_val;
 			f->lsm_str = str;
 			err = security_audit_rule_init(f->type, f->op, str,
-						       (void **)&f->lsm_rule);
+						       (void **)&f->lsm_rule,
+						       GFP_KERNEL);
 			/* Keep currently invalid fields around in case they
 			 * become valid after a policy reload. */
 			if (err == -EINVAL) {
···
 
 	/* our own (refreshed) copy of lsm_rule */
 	ret = security_audit_rule_init(df->type, df->op, df->lsm_str,
-				       (void **)&df->lsm_rule);
+				       (void **)&df->lsm_rule, GFP_KERNEL);
 	/* Keep currently invalid fields around in case they
 	 * become valid after a policy reload. */
 	if (ret == -EINVAL) {

security/apparmor/audit.c

···
 	}
 }
 
-int aa_audit_rule_init(u32 field, u32 op, char *rulestr, void **vrule)
+int aa_audit_rule_init(u32 field, u32 op, char *rulestr, void **vrule, gfp_t gfp)
 {
 	struct aa_audit_rule *rule;
 
···
 		return -EINVAL;
 	}
 
-	rule = kzalloc(sizeof(struct aa_audit_rule), GFP_KERNEL);
+	rule = kzalloc(sizeof(struct aa_audit_rule), gfp);
 
 	if (!rule)
 		return -ENOMEM;
 
 	/* Currently rules are treated as coming from the root ns */
 	rule->label = aa_label_parse(&root_ns->unconfined->label, rulestr,
-				     GFP_KERNEL, true, false);
+				     gfp, true, false);
 	if (IS_ERR(rule->label)) {
 		int err = PTR_ERR(rule->label);
 		aa_audit_rule_free(rule);

security/apparmor/include/audit.h

···
 }
 
 void aa_audit_rule_free(void *vrule);
-int aa_audit_rule_init(u32 field, u32 op, char *rulestr, void **vrule);
+int aa_audit_rule_init(u32 field, u32 op, char *rulestr, void **vrule, gfp_t gfp);
 int aa_audit_rule_known(struct audit_krule *rule);
 int aa_audit_rule_match(u32 sid, u32 field, u32 op, void *vrule);
 

security/integrity/ima/ima.h

···
 #else
 
 static inline int ima_filter_rule_init(u32 field, u32 op, char *rulestr,
-				       void **lsmrule)
+				       void **lsmrule, gfp_t gfp)
 {
 	return -EINVAL;
 }

security/integrity/ima/ima_policy.c

···
 	kfree(entry);
 }
 
-static struct ima_rule_entry *ima_lsm_copy_rule(struct ima_rule_entry *entry)
+static struct ima_rule_entry *ima_lsm_copy_rule(struct ima_rule_entry *entry,
+						gfp_t gfp)
 {
 	struct ima_rule_entry *nentry;
 	int i;
···
 	 * Immutable elements are copied over as pointers and data; only
 	 * lsm rules can change
 	 */
-	nentry = kmemdup(entry, sizeof(*nentry), GFP_KERNEL);
+	nentry = kmemdup(entry, sizeof(*nentry), gfp);
 	if (!nentry)
 		return NULL;
 
···
 
 		ima_filter_rule_init(nentry->lsm[i].type, Audit_equal,
 				     nentry->lsm[i].args_p,
-				     &nentry->lsm[i].rule);
+				     &nentry->lsm[i].rule,
+				     gfp);
 		if (!nentry->lsm[i].rule)
 			pr_warn("rule for LSM \'%s\' is undefined\n",
 				nentry->lsm[i].args_p);
···
 	int i;
 	struct ima_rule_entry *nentry;
 
-	nentry = ima_lsm_copy_rule(entry);
+	nentry = ima_lsm_copy_rule(entry, GFP_KERNEL);
 	if (!nentry)
 		return -ENOMEM;
 
···
 		}
 
 		if (rc == -ESTALE && !rule_reinitialized) {
-			lsm_rule = ima_lsm_copy_rule(rule);
+			lsm_rule = ima_lsm_copy_rule(rule, GFP_ATOMIC);
 			if (lsm_rule) {
 				rule_reinitialized = true;
 				goto retry;
···
 	entry->lsm[lsm_rule].type = audit_type;
 	result = ima_filter_rule_init(entry->lsm[lsm_rule].type, Audit_equal,
 				      entry->lsm[lsm_rule].args_p,
-				      &entry->lsm[lsm_rule].rule);
+				      &entry->lsm[lsm_rule].rule,
+				      GFP_KERNEL);
 	if (!entry->lsm[lsm_rule].rule) {
 		pr_warn("rule for LSM \'%s\' is undefined\n",
 			entry->lsm[lsm_rule].args_p);

security/security.c

···
  * @op: rule operator
  * @rulestr: rule context
  * @lsmrule: receive buffer for audit rule struct
+ * @gfp: GFP flag used for kmalloc
  *
  * Allocate and initialize an LSM audit rule structure.
  *
  * Return: Return 0 if @lsmrule has been successfully set, -EINVAL in case of
  *         an invalid rule.
  */
-int security_audit_rule_init(u32 field, u32 op, char *rulestr, void **lsmrule)
+int security_audit_rule_init(u32 field, u32 op, char *rulestr, void **lsmrule,
+			     gfp_t gfp)
 {
-	return call_int_hook(audit_rule_init, field, op, rulestr, lsmrule);
+	return call_int_hook(audit_rule_init, field, op, rulestr, lsmrule, gfp);
 }
 
 /**

security/selinux/include/audit.h

···
  *	@op: the operator the rule uses
  *	@rulestr: the text "target" of the rule
  *	@rule: pointer to the new rule structure returned via this
+ *	@gfp: GFP flag used for kmalloc
  *
  *	Returns 0 if successful, -errno if not.  On success, the rule structure
  *	will be allocated internally.  The caller must free this structure with
  *	selinux_audit_rule_free() after use.
  */
-int selinux_audit_rule_init(u32 field, u32 op, char *rulestr, void **rule);
+int selinux_audit_rule_init(u32 field, u32 op, char *rulestr, void **rule,
+			    gfp_t gfp);
 
 /**
  *	selinux_audit_rule_free - free an selinux audit rule structure.

security/selinux/ss/services.c

···
 	}
 }
 
-int selinux_audit_rule_init(u32 field, u32 op, char *rulestr, void **vrule)
+int selinux_audit_rule_init(u32 field, u32 op, char *rulestr, void **vrule,
+			    gfp_t gfp)
 {
 	struct selinux_state *state = &selinux_state;
 	struct selinux_policy *policy;
···
 		return -EINVAL;
 	}
 
-	tmprule = kzalloc(sizeof(struct selinux_audit_rule), GFP_KERNEL);
+	tmprule = kzalloc(sizeof(struct selinux_audit_rule), gfp);
 	if (!tmprule)
 		return -ENOMEM;
 	context_init(&tmprule->au_ctxt);

security/smack/smack_lsm.c

···
  * @op: required testing operator (=, !=, >, <, ...)
  * @rulestr: smack label to be audited
  * @vrule: pointer to save our own audit rule representation
+ * @gfp: type of the memory for the allocation
  *
  * Prepare to audit cases where (@field @op @rulestr) is true.
  * The label to be audited is created if necessay.
  */
-static int smack_audit_rule_init(u32 field, u32 op, char *rulestr, void **vrule)
+static int smack_audit_rule_init(u32 field, u32 op, char *rulestr, void **vrule,
+				 gfp_t gfp)
 {
 	struct smack_known *skp;
 	char **rule = (char **)vrule;