Hook improvements (take 2) (#563)

Hi @tiann.

Thanks for the great project, I had great fun playing around with it.

This PR mainly tries to further minimize the possible delays caused by
KernelSU hooking.

There are 3 major changes:
- Processes with 0 < UID < 2000 are blocked straight-up before going
through the allow_list.
I don't see any need for such processes to be interested in root, and
this allows returning early before going through a more expensive
lookup.
If there's an expected breakage due to this change, I'll remove it. Let
me know.
- A page-sized (4K) bitmap is added.
This allows O(1) lookup for UID <= 32767.
This speeds up `ksu_is_allow_uid()` by about 4.8x by sacrificing a 4K
memory. IMHO, a good trade-off.
Most notably, this reduces the 99.999% result previously from worrying
milliseconds scale to microseconds scale.
For UID > 32767, another page-sized (4K) sequential array is used to
cache allow_list.

Compared to the previous PR #557, this new approach gives another nice
25% performance boost in average, 63-96% boost in worst cases.

Benchmark results are available at
https://docs.google.com/spreadsheets/d/1w_tO1zRLPNMFRer49pL1TQfL6ndEhilRrDU1XFIcWXY/edit?usp=sharing

Thanks!

---------

Signed-off-by: Juhyung Park <qkrwngud825@gmail.com>

kernel/allowlist.c

@@ -29,6 +29,38 @@ static DEFINE_MUTEX(allowlist_mutex);
 static struct root_profile default_root_profile;
 static struct non_root_profile default_non_root_profile;
 
+static int allow_list_arr[PAGE_SIZE / sizeof(int)] __read_mostly __aligned(PAGE_SIZE);
+static int allow_list_pointer __read_mostly = 0;
+
+static void remove_uid_from_arr(uid_t uid)
+{
+	int *temp_arr;
+	int i, j;
+
+	if (allow_list_pointer == 0)
+		return;
+
+	temp_arr = kmalloc(sizeof(allow_list_arr), GFP_KERNEL);
+	if (temp_arr == NULL) {
+		pr_err("%s: unable to allocate memory\n", __func__);
+		return;
+	}
+
+	for (i = j = 0; i < allow_list_pointer; i++) {
+		if (allow_list_arr[i] == uid)
+			continue;
+		temp_arr[j++] = allow_list_arr[i];
+	}
+
+	allow_list_pointer = j;
+
+	for (; j < ARRAY_SIZE(allow_list_arr); j++)
+		temp_arr[j] = -1;
+
+	memcpy(&allow_list_arr, temp_arr, PAGE_SIZE);
+	kfree(temp_arr);
+}
+
 static void init_default_profiles()
 {
 	default_root_profile.uid = 0;
@@ -51,6 +83,9 @@ struct perm_data {
 
 static struct list_head allow_list;
 
+static uint8_t allow_list_bitmap[PAGE_SIZE] __read_mostly __aligned(PAGE_SIZE);
+#define BITMAP_UID_MAX ((sizeof(allow_list_bitmap) * BITS_PER_BYTE) - 1)
+
 #define KERNEL_SU_ALLOWLIST "/data/adb/ksu/.allowlist"
 
 static struct work_struct ksu_save_work;
@@ -110,6 +145,11 @@ static bool profile_valid(struct app_profile *profile)
 		return false;
 	}
 
+	if (profile->current_uid < 2000) {
+		pr_err("uid lower than 2000 is unsupported: %d\n", profile->current_uid);
+		return false;
+	}
+
 	if (profile->version < KSU_APP_PROFILE_VER) {
 		pr_info("Unsupported profile version: %d\n", profile->version);
 		return false;
@@ -147,7 +187,7 @@ bool ksu_set_app_profile(struct app_profile *profile, bool persist)
 			// found it, just override it all!
 			memcpy(&p->profile, profile, sizeof(*profile));
 			result = true;
-			goto exit;
+			goto out;
 		}
 	}
 
@@ -170,9 +210,31 @@ bool ksu_set_app_profile(struct app_profile *profile, bool persist)
 			profile->nrp_config.profile.umount_modules);
 	}
 	list_add_tail(&p->list, &allow_list);
+
+out:
+	if (profile->current_uid <= BITMAP_UID_MAX) {
+		if (profile->allow_su)
+			allow_list_bitmap[profile->current_uid / BITS_PER_BYTE] |= 1 << (profile->current_uid % BITS_PER_BYTE);
+		else
+			allow_list_bitmap[profile->current_uid / BITS_PER_BYTE] &= ~(1 << (profile->current_uid % BITS_PER_BYTE));
+	} else {
+		if (profile->allow_su) {
+			/*
+			 * 1024 apps with uid higher than BITMAP_UID_MAX
+			 * registered to request superuser?
+			 */
+			if (allow_list_pointer >= ARRAY_SIZE(allow_list_arr)) {
+				pr_err("too many apps registered\n");
+				WARN_ON(1);
+				return false;
+			}
+			allow_list_arr[allow_list_pointer++] = profile->current_uid;
+		} else {
+			remove_uid_from_arr(profile->current_uid);
+		}
+	}
 	result = true;
 
-exit:
 	// check if the default profiles is changed, cache it to a single struct to accelerate access.
 	if (unlikely(!strcmp(profile->key, "$"))) {
 		// set default non root profile
@@ -192,21 +254,26 @@ exit:
 	return result;
 }
 
-bool ksu_is_allow_uid(uid_t uid)
+bool __ksu_is_allow_uid(uid_t uid)
 {
-	struct perm_data *p = NULL;
-	struct list_head *pos = NULL;
+	int i;
 
-	if (uid == 0) {
+	if (unlikely(uid == 0)) {
 		// already root, but only allow our domain.
 		return is_ksu_domain();
 	}
 
-	list_for_each (pos, &allow_list) {
-		p = list_entry(pos, struct perm_data, list);
-		// pr_info("is_allow_uid uid :%d, allow: %d\n", p->uid, p->allow);
-		if (uid == p->profile.current_uid) {
-			return p->profile.allow_su;
+	if (uid < 2000) {
+		// do not bother going through the list if it's system
+		return false;
+	}
+
+	if (likely(uid <= BITMAP_UID_MAX)) {
+		return !!(allow_list_bitmap[uid / BITS_PER_BYTE] & (1 << (uid % BITS_PER_BYTE)));
+	} else {
+		for (i = 0; i < allow_list_pointer; i++) {
+			if (allow_list_arr[i] == uid)
+				return true;
 		}
 	}
 
@@ -386,6 +453,9 @@ void ksu_prune_allowlist(bool (*is_uid_exist)(uid_t, void *), void *data)
 			modified = true;
 			pr_info("prune uid: %d\n", uid);
 			list_del(&np->list);
+			allow_list_bitmap[uid / BITS_PER_BYTE] &= ~(1 << (uid % BITS_PER_BYTE));
+			remove_uid_from_arr(uid);
+			smp_mb();
 			kfree(np);
 		}
 	}
@@ -409,6 +479,14 @@ bool ksu_load_allow_list(void)
 
 void ksu_allowlist_init(void)
 {
+	int i;
+
+	BUILD_BUG_ON(sizeof(allow_list_bitmap) != PAGE_SIZE);
+	BUILD_BUG_ON(sizeof(allow_list_arr) != PAGE_SIZE);
+
+	for (i = 0; i < ARRAY_SIZE(allow_list_arr); i++)
+		allow_list_arr[i] = -1;
+
 	INIT_LIST_HEAD(&allow_list);
 
 	INIT_WORK(&ksu_save_work, do_save_allow_list);