kernel: Syncronize upstream changes (#198)

* Cherry-picked range: (kernel)
ebea31daa8..6915b62b9a

* Also merged unmerged pr:
https://github.com/tiann/KernelSU/pull/ 2909

Co-authored-by: Ylarod <me@ylarod.cn>
Co-authored-by: 5ec1cff <56485584+5ec1cff@users.noreply.github.com>
Co-authored-by: backslashxx <118538522+backslashxx@users.noreply.github.com>
Co-authored-by: u9521 <63995396+u9521@users.noreply.github.com>
Co-authored-by: Wang Han <416810799@qq.com>

kernel/Kconfig

@@ -8,6 +8,11 @@ config KSU
 	  To compile as a module, choose M here: the
 	  module will be called kernelsu.
 
+# For easier extern ifdef handling
+config RKSU
+	bool "RKSU compat, do not modify"
+	default y
+
 config KSU_DEBUG
 	bool "KernelSU debug mode"
 	depends on KSU
@@ -16,12 +21,11 @@ config KSU_DEBUG
 	  Enable KernelSU debug mode.
 
 config KSU_ALLOWLIST_WORKAROUND
-	bool "KernelSU Session init keyring workaround"
+	bool "KernelSU allowlist workaround"
 	depends on KSU
 	default n
 	help
-	  Enable session keyring init workaround for problematic devices.
+	  Enable workaround for broken allowlist save
-	  Useful for situations where the SU allowlist is not kept after a reboot.
 
 config KPM
 	bool "Enable SukiSU KPM"
@@ -35,9 +39,21 @@ config KPM
 	select KALLSYMS_ALL
 
 config KSU_MANUAL_HOOK
-    bool "Hook KernelSU manually"
+	bool "KernelSU manual hook mode."
-    depends on KSU != m
+	depends on KSU && KSU != m
+	default y if !KPROBES
+	default n
 	help
-      If enabled, Hook required KernelSU syscalls with manually-patched function.
+	  Enable manual hook support.
+
+config KSU_SHOULD_USE_NEW_TP
+	bool "KernelSU tracepoint+kretprobe hook"
+	depends on KSU && !KSU_MANUAL_HOOK
+	depends on KRETPROBES && KPROBES && HAVE_SYSCALL_TRACEPOINTS
+	default y if KPROBES && KRETPROBES && HAVE_SYSCALL_TRACEPOINTS
+	default n
+	help
+	  Enable KPROBES, KRETPROBES and TRACEPOINT hook for KernelSU core.
+	  This should not be used on kernel below 5.10.
 
 endmenu

kernel/Makefile

@@ -1,17 +1,22 @@
 kernelsu-objs := ksu.o
 kernelsu-objs += allowlist.o
 kernelsu-objs += dynamic_manager.o
+kernelsu-objs += app_profile.o
 kernelsu-objs += apk_sign.o
 kernelsu-objs += sucompat.o
+kernelsu-objs += syscall_hook_manager.o
 kernelsu-objs += throne_tracker.o
 kernelsu-objs += pkg_observer.o
-kernelsu-objs += core_hook.o
+kernelsu-objs += setuid_hook.o
+kernelsu-objs += lsm_hooks.o
+kernelsu-objs += kernel_compat.o
+kernelsu-objs += kernel_umount.o
 kernelsu-objs += supercalls.o
 kernelsu-objs += feature.o
 kernelsu-objs += throne_tracker.o
 kernelsu-objs += ksud.o
 kernelsu-objs += embed_ksud.o
-kernelsu-objs += kernel_compat.o
+kernelsu-objs += seccomp_cache.o
 kernelsu-objs += file_wrapper.o
 kernelsu-objs += throne_comm.o
 
@@ -84,14 +89,33 @@ endif
 ccflags-y += -DKSU_VERSION=$(KSU_VERSION)
 ccflags-y += -DKSU_VERSION_FULL=\"$(KSU_VERSION_FULL)\"
 
+# RKSU: checks for available hook
+## Logic flipped for HAVE_KSU_HOOK: 0 is success, 1 is failure, but not with KSU_DRY_RUN
+HAVE_KSU_HOOK ?= 1
+KSU_DRY_RUN ?= 0
+
 # Checks hooks state
+ifeq ($(CONFIG_KSU_SHOULD_USE_NEW_TP), y)
+$(info -- KernelSU: SHOULD_USE_NEW_TP)
+ccflags-y += -DKSU_SHOULD_USE_NEW_TP
+# Let's make it 0, so it would pass.
+HAVE_KSU_HOOK := 0
+endif
+
 ifeq ($(CONFIG_KSU_MANUAL_HOOK), y)
-ccflags-y += -DKSU_MANUAL_HOOK
+HAVE_KSU_HOOK := $(shell grep -q "ksu_handle_faccessat" $(srctree)/fs/open.c; echo $$?)
-$(info -- SukiSU: KSU_MANUAL_HOOK)
+ifeq ($(HAVE_KSU_HOOK),0)
+$(info -- KernelSU: CONFIG_KSU_MANUAL_HOOK)
+endif
+endif
+
+ifeq ($(KSU_DRY_RUN),0)
+ifneq ($(HAVE_KSU_HOOK),0)
+$(error -- KernelSU: No hooks were defined, bail!)
+endif
 else
-ccflags-y += -DKSU_HAVE_SYSCALL_TRACEPOINTS_HOOK
+$(info -- KernelSU in dry run mode, skip hook checks)
-ccflags-y += -DKSU_KPROBES_HOOK
+
-$(info -- SukiSU: KSU_TRACEPOINT_HOOK)
 endif
 
 # SELinux drivers check
@@ -122,16 +146,12 @@ ifeq ($(shell grep -q "task_security_struct\s\+\*selinux_cred" $(srctree)/securi
 ccflags-y += -DKSU_OPTIONAL_SELINUX_CRED
 endif
 
-# Check if __poll_t exists in linux or uapi types headers (handles backports)
-ifneq ($(shell grep -q "__poll_t" \
-    $(srctree)/include/linux/types.h \
-    $(srctree)/include/uapi/linux/types.h; echo $$?),0)
-ccflags-y += -DKSU_NO___POLL_T
-endif
-
 ifeq ($(shell grep -q "anon_inode_getfd_secure" $(srctree)/fs/anon_inodes.c; echo $$?),0)
 ccflags-y += -DKSU_HAS_GETFD_SECURE
 endif
+ifeq ($(shell grep -A1 "^int vfs_getattr" $(srctree)/fs/stat.c | grep -q "query_flags"; echo $$?),0)
+ccflags-y += -DKSU_HAS_NEW_VFS_GETATTR
+endif
 
 # Checks Samsung
 ifeq ($(shell grep -q "CONFIG_KDP_CRED" $(srctree)/kernel/cred.c; echo $$?),0)

kernel/allowlist.c

@@ -19,12 +19,13 @@
 #include <linux/compiler_types.h>
 #endif
 
-#include "ksu.h"
 #include "klog.h" // IWYU pragma: keep
+#include "ksud.h"
 #include "selinux/selinux.h"
-#include "kernel_compat.h"
 #include "allowlist.h"
 #include "manager.h"
+#include "kernel_compat.h"
+#include "syscall_hook_manager.h"
 
 #define FILE_MAGIC 0x7f4b5355 // ' KSU', u32
 #define FILE_FORMAT_VERSION 3 // u32
@@ -50,7 +51,7 @@ static void remove_uid_from_arr(uid_t uid)
 	if (allow_list_pointer == 0)
 		return;
 
-	temp_arr = kmalloc(sizeof(allow_list_arr), GFP_KERNEL);
+	temp_arr = kzalloc(sizeof(allow_list_arr), GFP_KERNEL);
 	if (temp_arr == NULL) {
 		pr_err("%s: unable to allocate memory\n", __func__);
 		return;
@@ -205,7 +206,7 @@ bool ksu_set_app_profile(struct app_profile *profile, bool persist)
 	}
 
 	// not found, alloc a new node!
-	p = (struct perm_data *)kmalloc(sizeof(struct perm_data), GFP_KERNEL);
+	p = (struct perm_data *)kzalloc(sizeof(struct perm_data), GFP_KERNEL);
 	if (!p) {
 		pr_err("ksu_set_app_profile alloc failed\n");
 		return false;
@@ -264,8 +265,11 @@ out:
 		       sizeof(default_root_profile));
 	}
 
-	if (persist)
+	if (persist) {
 		persistent_allow_list();
+		// FIXME: use a new flag
+		ksu_mark_running_process();
+	}
 
 	return result;
 }
@@ -432,12 +436,7 @@ void persistent_allow_list(void)
 		goto put_task;
 	}
 	cb->func = do_persistent_allow_list;
-
+	ksu_task_work_add(tsk, cb, TWA_RESUME);
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 7, 8)
-	task_work_add(tsk, cb, TWA_RESUME);
-#else
-	task_work_add(tsk, cb, true);
-#endif
 
 put_task:
 	put_task_struct(tsk);
@@ -506,6 +505,11 @@ void ksu_prune_allowlist(bool (*is_uid_valid)(uid_t, char *, void *),
 	struct perm_data *np = NULL;
 	struct perm_data *n = NULL;
 
+	if (!ksu_boot_completed) {
+		pr_info("boot not completed, skip prune\n");
+		return;
+	}
+
 	bool modified = false;
 	// TODO: use RCU!
 	mutex_lock(&allowlist_mutex);
@@ -554,8 +558,6 @@ void ksu_allowlist_exit(void)
 	struct perm_data *np = NULL;
 	struct perm_data *n = NULL;
 
-	persistent_allow_list();
-
 	// free allowlist
 	mutex_lock(&allowlist_mutex);
 	list_for_each_entry_safe (np, n, &allow_list, list) {

kernel/allowlist.h

@@ -2,7 +2,12 @@
 #define __KSU_H_ALLOWLIST
 
 #include <linux/types.h>
-#include "ksu.h"
+#include <linux/uidgid.h>
+#include "app_profile.h"
+
+#define PER_USER_RANGE 100000
+#define FIRST_APPLICATION_UID 10000
+#define LAST_APPLICATION_UID 19999
 
 void ksu_allowlist_init(void);
 
@@ -31,4 +36,10 @@ bool ksu_set_app_profile(struct app_profile *, bool persist);
 
 bool ksu_uid_should_umount(uid_t uid);
 struct root_profile *ksu_get_root_profile(uid_t uid);
+
+static inline bool is_appuid(uid_t uid)
+{
+	uid_t appid = uid % PER_USER_RANGE;
+	return appid >= FIRST_APPLICATION_UID && appid <= LAST_APPLICATION_UID;
+}
 #endif

kernel/apk_sign.c

@@ -38,7 +38,7 @@ static struct sdesc *init_sdesc(struct crypto_shash *alg)
 	int size;
 
 	size = sizeof(struct shash_desc) + crypto_shash_descsize(alg);
-	sdesc = kmalloc(size, GFP_KERNEL);
+	sdesc = kzalloc(size, GFP_KERNEL);
 	if (!sdesc)
 		return ERR_PTR(-ENOMEM);
 	sdesc->shash.tfm = alg;

kernel/app_profile.c

@@ -0,0 +1,309 @@
+#include <linux/version.h>
+#include <linux/capability.h>
+#include <linux/cred.h>
+#include <linux/err.h>
+#include <linux/fdtable.h>
+#include <linux/file.h>
+#include <linux/fs.h>
+#include <linux/proc_ns.h>
+#include <linux/pid.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 10, 0)
+#include <linux/sched/signal.h> // signal_struct
+#include <linux/sched/task.h>
+#endif
+#include <linux/sched.h>
+#include <linux/seccomp.h>
+#include <linux/thread_info.h>
+#include <linux/uidgid.h>
+#include <linux/syscalls.h>
+
+#include "allowlist.h"
+#include "app_profile.h"
+#include "arch.h"
+#include "kernel_compat.h"
+#include "klog.h" // IWYU pragma: keep
+#include "selinux/selinux.h"
+#include "syscall_hook_manager.h"
+
+static struct group_info root_groups = { .usage = ATOMIC_INIT(2) };
+
+void setup_groups(struct root_profile *profile, struct cred *cred)
+{
+	if (profile->groups_count > KSU_MAX_GROUPS) {
+		pr_warn("Failed to setgroups, too large group: %d!\n",
+			profile->uid);
+		return;
+	}
+
+	if (profile->groups_count == 1 && profile->groups[0] == 0) {
+		// setgroup to root and return early.
+		if (cred->group_info)
+			put_group_info(cred->group_info);
+		cred->group_info = get_group_info(&root_groups);
+		return;
+	}
+
+	u32 ngroups = profile->groups_count;
+	struct group_info *group_info = groups_alloc(ngroups);
+	if (!group_info) {
+		pr_warn("Failed to setgroups, ENOMEM for: %d\n", profile->uid);
+		return;
+	}
+
+	int i;
+	for (i = 0; i < ngroups; i++) {
+		gid_t gid = profile->groups[i];
+		kgid_t kgid = make_kgid(current_user_ns(), gid);
+		if (!gid_valid(kgid)) {
+			pr_warn("Failed to setgroups, invalid gid: %d\n", gid);
+			put_group_info(group_info);
+			return;
+		}
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 9, 0)
+		group_info->gid[i] = kgid;
+#else
+		GROUP_AT(group_info, i) = kgid;
+#endif
+	}
+
+	groups_sort(group_info);
+	set_groups(cred, group_info);
+	put_group_info(group_info);
+}
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 17, 0)
+extern long SYS_SETNS_SYMBOL(const struct pt_regs *regs);
+static long ksu_sys_setns(int fd, int flags)
+{
+	struct pt_regs regs;
+	memset(&regs, 0, sizeof(regs));
+
+	PT_REGS_PARM1(&regs) = fd;
+	PT_REGS_PARM2(&regs) = flags;
+
+#if (defined(__aarch64__) || defined(__x86_64__))
+	return SYS_SETNS_SYMBOL(&regs);
+#else
+	return -ENOSYS;
+#endif
+}
+#else
+static long ksu_sys_setns(int fd, int flags)
+{
+	return sys_setns(fd, flags);
+}
+
+__weak int ksys_unshare(unsigned long unshare_flags)
+{
+	return sys_unshare(unshare_flags);
+}
+#endif
+
+static void setup_mount_namespace(int32_t ns_mode)
+{
+	pr_info("setup mount namespace for pid: %d\n", current->pid);
+
+	if (ns_mode == 0) {
+		pr_info("mount namespace mode: inherit\n");
+		// do nothing
+		return;
+	}
+
+	if (ns_mode > 2) {
+		pr_warn("unknown mount namespace mode: %d\n", ns_mode);
+		return;
+	}
+
+	const struct cred *old_cred = NULL;
+	struct cred *new_cred = NULL;
+	if (!(capable(CAP_SYS_ADMIN) && capable(CAP_SYS_CHROOT))) {
+		pr_info("process dont have CAP_SYS_ADMIN or CAP_SYS_CHROOT, adding it temporarily.\n");
+		new_cred = prepare_creds();
+		if (!new_cred) {
+			pr_warn("failed to prepare new credentials\n");
+			return;
+		}
+		cap_raise(new_cred->cap_effective, CAP_SYS_ADMIN);
+		cap_raise(new_cred->cap_effective, CAP_SYS_CHROOT);
+		old_cred = override_creds(new_cred);
+	}
+
+	if (ns_mode == 1) {
+		pr_info("mount namespace mode: global\n");
+		struct file *ns_file;
+		struct path ns_path;
+		struct task_struct *pid1_task = NULL;
+		struct pid *pid_struct = NULL;
+		rcu_read_lock();
+		// find init
+		pid_struct = find_pid_ns(1, &init_pid_ns);
+		if (unlikely(!pid_struct)) {
+			rcu_read_unlock();
+			pr_warn("failed to find pid_struct for PID 1\n");
+			goto try_drop_caps;
+		}
+
+		pid1_task = get_pid_task(pid_struct, PIDTYPE_PID);
+		rcu_read_unlock();
+		if (unlikely(!pid1_task)) {
+			pr_warn("failed to get task_struct for PID 1\n");
+			goto try_drop_caps;
+		}
+		// mabe you can use &init_task for first stage init?
+		long ret = ns_get_path(&ns_path, pid1_task, &mntns_operations);
+		put_task_struct(pid1_task);
+		if (ret) {
+			pr_warn("failed to get path for init's mount namespace: %ld\n",
+				ret);
+			goto try_drop_caps;
+		}
+		ns_file = dentry_open(&ns_path, O_RDONLY, current_cred());
+
+		path_put(&ns_path);
+		if (IS_ERR(ns_file)) {
+			pr_warn("failed to open file for init's mount namespace: %ld\n",
+				PTR_ERR(ns_file));
+			goto try_drop_caps;
+		}
+
+		int fd = get_unused_fd_flags(O_CLOEXEC);
+		if (fd < 0) {
+			pr_warn("failed to get an unused fd: %d\n", fd);
+			fput(ns_file);
+			goto try_drop_caps;
+		}
+
+		fd_install(fd, ns_file);
+		pr_info("calling sys_setns with fd : %d\n", fd);
+
+		ret = ksu_sys_setns(fd, CLONE_NEWNS);
+		if (ret) {
+			pr_warn("sys_setns failed: %ld\n", ret);
+		}
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 11, 0)
+		close_fd(fd);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(4, 17, 0)
+		ksys_close(fd);
+#else
+		sys_close(fd);
+#endif
+	}
+
+	if (ns_mode == 2) {
+		long ret;
+		pr_info("mount namespace mode: independent\n");
+
+		ret = ksys_unshare(CLONE_NEWNS);
+		if (ret) {
+			pr_warn("call ksys_unshare failed: %ld\n", ret);
+		}
+	}
+
+try_drop_caps:
+	if (old_cred) {
+		pr_info("dropping temporarily capability.\n");
+		revert_creds(old_cred);
+		put_cred(new_cred);
+	}
+	return;
+}
+
+// RKSU: Use it wisely, not static.
+void disable_seccomp(struct task_struct *tsk)
+{
+	if (unlikely(!tsk))
+		return;
+
+	assert_spin_locked(&tsk->sighand->siglock);
+
+	// disable seccomp
+#if defined(CONFIG_GENERIC_ENTRY) &&                                           \
+	LINUX_VERSION_CODE >= KERNEL_VERSION(5, 11, 0)
+	clear_syscall_work(SECCOMP);
+#else
+	clear_thread_flag(TIF_SECCOMP);
+#endif
+
+#ifdef CONFIG_SECCOMP
+	tsk->seccomp.mode = 0;
+	if (tsk->seccomp.filter) {
+		// 5.9+ have filter_count and use seccomp_filter_release
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 9, 0)
+		seccomp_filter_release(tsk);
+		atomic_set(&tsk->seccomp.filter_count, 0);
+#else
+		// for 6.11+ kernel support?
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 9, 0)
+		put_seccomp_filter(tsk);
+#endif
+		tsk->seccomp.filter = NULL;
+#endif
+	}
+#endif
+}
+
+void escape_with_root_profile(void)
+{
+	struct cred *cred;
+	// a bit useless, but we just want less ifdefs
+	struct task_struct *p = current;
+
+	if (current_euid().val == 0) {
+		pr_warn("Already root, don't escape!\n");
+		return;
+	}
+
+	cred = prepare_creds();
+	if (!cred) {
+		pr_warn("prepare_creds failed!\n");
+		return;
+	}
+
+	struct root_profile *profile = ksu_get_root_profile(cred->uid.val);
+
+	cred->uid.val = profile->uid;
+	cred->suid.val = profile->uid;
+	cred->euid.val = profile->uid;
+	cred->fsuid.val = profile->uid;
+
+	cred->gid.val = profile->gid;
+	cred->fsgid.val = profile->gid;
+	cred->sgid.val = profile->gid;
+	cred->egid.val = profile->gid;
+	cred->securebits = 0;
+
+	BUILD_BUG_ON(sizeof(profile->capabilities.effective) !=
+		     sizeof(kernel_cap_t));
+
+	// setup capabilities
+	// we need CAP_DAC_READ_SEARCH becuase `/data/adb/ksud` is not accessible for non root process
+	// we add it here but don't add it to cap_inhertiable, it would be dropped automaticly after exec!
+	u64 cap_for_ksud =
+		profile->capabilities.effective | CAP_DAC_READ_SEARCH;
+	memcpy(&cred->cap_effective, &cap_for_ksud,
+	       sizeof(cred->cap_effective));
+	memcpy(&cred->cap_permitted, &profile->capabilities.effective,
+	       sizeof(cred->cap_permitted));
+	memcpy(&cred->cap_bset, &profile->capabilities.effective,
+	       sizeof(cred->cap_bset));
+
+	setup_groups(profile, cred);
+
+	commit_creds(cred);
+
+	// Refer to kernel/seccomp.c: seccomp_set_mode_strict
+	// When disabling Seccomp, ensure that current->sighand->siglock is held during the operation.
+	spin_lock_irq(&p->sighand->siglock);
+	disable_seccomp(p);
+	spin_unlock_irq(&p->sighand->siglock);
+
+	setup_selinux(profile->selinux_domain);
+	setup_mount_namespace(profile->namespaces);
+
+#ifdef KSU_SHOULD_USE_NEW_TP
+	struct task_struct *t;
+	for_each_thread (p, t) {
+		ksu_set_task_tracepoint_flag(t);
+	}
+#endif
+}

kernel/app_profile.h

@@ -0,0 +1,66 @@
+#ifndef __KSU_H_APP_PROFILE
+#define __KSU_H_APP_PROFILE
+
+#include <linux/types.h>
+
+// Forward declarations
+struct cred;
+
+#define KSU_APP_PROFILE_VER 2
+#define KSU_MAX_PACKAGE_NAME 256
+// NGROUPS_MAX for Linux is 65535 generally, but we only supports 32 groups.
+#define KSU_MAX_GROUPS 32
+#define KSU_SELINUX_DOMAIN 64
+
+struct root_profile {
+	int32_t uid;
+	int32_t gid;
+
+	int32_t groups_count;
+	int32_t groups[KSU_MAX_GROUPS];
+
+	// kernel_cap_t is u32[2] for capabilities v3
+	struct {
+		u64 effective;
+		u64 permitted;
+		u64 inheritable;
+	} capabilities;
+
+	char selinux_domain[KSU_SELINUX_DOMAIN];
+
+	int32_t namespaces;
+};
+
+struct non_root_profile {
+	bool umount_modules;
+};
+
+struct app_profile {
+	// It may be utilized for backward compatibility, although we have never explicitly made any promises regarding this.
+	u32 version;
+
+	// this is usually the package of the app, but can be other value for special apps
+	char key[KSU_MAX_PACKAGE_NAME];
+	int32_t current_uid;
+	bool allow_su;
+
+	union {
+		struct {
+			bool use_default;
+			char template_name[KSU_MAX_PACKAGE_NAME];
+
+			struct root_profile profile;
+		} rp_config;
+
+		struct {
+			bool use_default;
+
+			struct non_root_profile profile;
+		} nrp_config;
+	};
+};
+
+// Escalate current process to root with the appropriate profile
+void escape_with_root_profile(void);
+
+#endif

kernel/arch.h

@@ -21,19 +21,13 @@
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 16, 0)
 #define REBOOT_SYMBOL "__arm64_sys_reboot"
 #define SYS_READ_SYMBOL "__arm64_sys_read"
-#define SYS_NEWFSTATAT_SYMBOL "__arm64_sys_newfstatat"
-#define SYS_FSTATAT64_SYMBOL "__arm64_sys_fstatat64"
-#define SYS_FACCESSAT_SYMBOL "__arm64_sys_faccessat"
 #define SYS_EXECVE_SYMBOL "__arm64_sys_execve"
-#define SYS_EXECVE_COMPAT_SYMBOL "__arm64_compat_sys_execve"
+#define SYS_SETNS_SYMBOL __arm64_sys_setns
 #else
 #define REBOOT_SYMBOL "sys_reboot"
 #define SYS_READ_SYMBOL "sys_read"
-#define SYS_NEWFSTATAT_SYMBOL "sys_newfstatat"
-#define SYS_FSTATAT64_SYMBOL "sys_fstatat64"
-#define SYS_FACCESSAT_SYMBOL "sys_faccessat"
 #define SYS_EXECVE_SYMBOL "sys_execve"
-#define SYS_EXECVE_COMPAT_SYMBOL "compat_sys_execve"
+#define SYS_SETNS_SYMBOL sys_setns
 #endif
 
 #elif defined(__x86_64__)
@@ -51,26 +45,21 @@
 #define __PT_RC_REG ax
 #define __PT_SP_REG sp
 #define __PT_IP_REG ip
+
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 16, 0)
 #define REBOOT_SYMBOL "__x64_sys_reboot"
 #define SYS_READ_SYMBOL "__x64_sys_read"
-#define SYS_NEWFSTATAT_SYMBOL "__x64_sys_newfstatat"
-#define SYS_FSTATAT64_SYMBOL "__x64_sys_fstatat64"
-#define SYS_FACCESSAT_SYMBOL "__x64_sys_faccessat"
 #define SYS_EXECVE_SYMBOL "__x64_sys_execve"
-#define SYS_EXECVE_COMPAT_SYMBOL "__x64_compat_sys_execve"
+#define SYS_SETNS_SYMBOL __x64_sys_setns
 #else
-#define PRCTL_SYMBOL "sys_reboot"
+#define REBOOT_SYMBOL "sys_reboot"
 #define SYS_READ_SYMBOL "sys_read"
-#define SYS_NEWFSTATAT_SYMBOL "sys_newfstatat"
-#define SYS_FSTATAT64_SYMBOL "sys_fstatat64"
-#define SYS_FACCESSAT_SYMBOL "sys_faccessat"
 #define SYS_EXECVE_SYMBOL "sys_execve"
-#define SYS_EXECVE_COMPAT_SYMBOL "compat_sys_execve"
+#define SYS_SETNS_SYMBOL sys_setns
 #endif
 
 #else
-#ifdef KSU_KPROBES_HOOK
+#ifdef KSU_SHOULD_USE_NEW_TP
 #error "Unsupported arch"
 #endif
 #endif

kernel/core_hook.c

@@ -1,835 +0,0 @@
-#include <linux/compiler.h>
-#include <linux/version.h>
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
-#include <linux/sched/task_stack.h>
-#else
-#include <linux/sched.h>
-#endif
-#include <linux/slab.h>
-#include <linux/task_work.h>
-#include <linux/thread_info.h>
-#include <linux/seccomp.h>
-#include <linux/capability.h>
-#include <linux/cred.h>
-#include <linux/dcache.h>
-#include <linux/fs.h>
-#include <linux/init.h>
-#include <linux/init_task.h>
-#include <linux/kernel.h>
-#include <linux/kprobes.h>
-#include <linux/mm.h>
-#include <linux/mount.h>
-#include <linux/namei.h>
-#include <linux/nsproxy.h>
-#include <linux/path.h>
-#include <linux/printk.h>
-#include <linux/sched.h>
-#include <linux/stddef.h>
-#include <linux/string.h>
-#include <linux/types.h>
-#include <linux/uaccess.h>
-#include <linux/uidgid.h>
-#ifndef KSU_HAS_PATH_UMOUNT
-#include <linux/syscalls.h> // sys_umount (<4.17) & ksys_umount (4.17+)
-#endif
-#include <linux/binfmts.h>
-#include <linux/err.h>
-
-#include "allowlist.h"
-#include "arch.h"
-#include "core_hook.h"
-#include "feature.h"
-#include "klog.h" // IWYU pragma: keep
-#include "ksu.h"
-#include "ksud.h"
-#include "manager.h"
-#include "selinux/selinux.h"
-#include "throne_tracker.h"
-#include "throne_comm.h"
-#include "kernel_compat.h"
-#include "supercalls.h"
-#include "sucompat.h"
-
-bool ksu_module_mounted __read_mostly = false;
-
-#ifndef DEVPTS_SUPER_MAGIC
-#define DEVPTS_SUPER_MAGIC	0x1cd1
-#endif
-
-extern int __ksu_handle_devpts(struct inode *inode); // sucompat.c
-
-#ifdef CONFIG_COMPAT
-bool ksu_is_compat __read_mostly = false;
-#endif
-
-static bool ksu_kernel_umount_enabled = true;
-static bool ksu_enhanced_security_enabled = false;
-
-static int kernel_umount_feature_get(u64 *value)
-{
-	*value = ksu_kernel_umount_enabled ? 1 : 0;
-	return 0;
-}
-
-static int kernel_umount_feature_set(u64 value)
-{
-	bool enable = value != 0;
-	ksu_kernel_umount_enabled = enable;
-	pr_info("kernel_umount: set to %d\n", enable);
-	return 0;
-}
-
-static const struct ksu_feature_handler kernel_umount_handler = {
-	.feature_id = KSU_FEATURE_KERNEL_UMOUNT,
-	.name = "kernel_umount",
-	.get_handler = kernel_umount_feature_get,
-	.set_handler = kernel_umount_feature_set,
-};
-
-static int enhanced_security_feature_get(u64 *value)
-{
-	*value = ksu_enhanced_security_enabled ? 1 : 0;
-	return 0;
-}
-
-static int enhanced_security_feature_set(u64 value)
-{
-	bool enable = value != 0;
-	ksu_enhanced_security_enabled = enable;
-	pr_info("enhanced_security: set to %d\n", enable);
-	return 0;
-}
-
-static const struct ksu_feature_handler enhanced_security_handler = {
-	.feature_id = KSU_FEATURE_ENHANCED_SECURITY,
-	.name = "enhanced_security",
-	.get_handler = enhanced_security_feature_get,
-	.set_handler = enhanced_security_feature_set,
-};
-
-static inline bool is_allow_su(void)
-{
-	if (is_manager()) {
-		// we are manager, allow!
-		return true;
-	}
-	return ksu_is_allow_uid_for_current(current_uid().val);
-}
-
-static inline bool is_unsupported_uid(uid_t uid)
-{
-#define LAST_APPLICATION_UID 19999
-	uid_t appid = uid % 100000;
-	return appid > LAST_APPLICATION_UID;
-}
-
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 7, 0)
-static struct group_info root_groups = {
-	.usage = REFCOUNT_INIT(2),
-};
-#else
-static struct group_info root_groups = { .usage = ATOMIC_INIT(2) };
-#endif
-
-static void setup_groups(struct root_profile *profile, struct cred *cred)
-{
-	if (profile->groups_count > KSU_MAX_GROUPS) {
-		pr_warn("Failed to setgroups, too large group: %d!\n",
-			profile->uid);
-		return;
-	}
-
-	if (profile->groups_count == 1 && profile->groups[0] == 0) {
-		// setgroup to root and return early.
-		if (cred->group_info)
-			put_group_info(cred->group_info);
-		cred->group_info = get_group_info(&root_groups);
-		return;
-	}
-
-	u32 ngroups = profile->groups_count;
-	struct group_info *group_info = groups_alloc(ngroups);
-	if (!group_info) {
-		pr_warn("Failed to setgroups, ENOMEM for: %d\n", profile->uid);
-		return;
-	}
-
-	int i;
-	for (i = 0; i < ngroups; i++) {
-		gid_t gid = profile->groups[i];
-		kgid_t kgid = make_kgid(current_user_ns(), gid);
-		if (!gid_valid(kgid)) {
-			pr_warn("Failed to setgroups, invalid gid: %d\n", gid);
-			put_group_info(group_info);
-			return;
-		}
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 9, 0)
-		group_info->gid[i] = kgid;
-#else
-		GROUP_AT(group_info, i) = kgid;
-#endif
-	}
-
-	groups_sort(group_info);
-	set_groups(cred, group_info);
-	put_group_info(group_info);
-}
-
-static void disable_seccomp(struct task_struct *tsk)
-{
-	assert_spin_locked(&tsk->sighand->siglock);
-
-	// disable seccomp
-#if defined(CONFIG_GENERIC_ENTRY) &&                                           \
-	LINUX_VERSION_CODE >= KERNEL_VERSION(5, 11, 0)
-	clear_syscall_work(SECCOMP);
-#else
-	clear_thread_flag(TIF_SECCOMP);
-#endif
-
-#ifdef CONFIG_SECCOMP
-	tsk->seccomp.mode = 0;
-	if (tsk->seccomp.filter) {
-		// 5.9+ have filter_count and use seccomp_filter_release
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 9, 0)
-		seccomp_filter_release(tsk);
-		atomic_set(&tsk->seccomp.filter_count, 0);
-#else
-		// for 6.11+ kernel support?
-#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 9, 0)
-		put_seccomp_filter(tsk);
-#endif
-		tsk->seccomp.filter = NULL;
-#endif
-	}
-#endif
-}
-
-void escape_to_root(void)
-{
-	struct cred *cred;
-#ifdef KSU_SHOULD_USE_NEW_TP
-	struct task_struct *p = current;
-	struct task_struct *t;
-#endif
-
-	cred = prepare_creds();
-	if (!cred) {
-		pr_warn("prepare_creds failed!\n");
-		return;
-	}
-
-	if (cred->euid.val == 0) {
-		pr_warn("Already root, don't escape!\n");
-		abort_creds(cred);
-		return;
-	}
-
-	struct root_profile *profile = ksu_get_root_profile(cred->uid.val);
-
-	cred->uid.val = profile->uid;
-	cred->suid.val = profile->uid;
-	cred->euid.val = profile->uid;
-	cred->fsuid.val = profile->uid;
-
-	cred->gid.val = profile->gid;
-	cred->fsgid.val = profile->gid;
-	cred->sgid.val = profile->gid;
-	cred->egid.val = profile->gid;
-	cred->securebits = 0;
-
-	BUILD_BUG_ON(sizeof(profile->capabilities.effective) !=
-		     sizeof(kernel_cap_t));
-
-	// setup capabilities
-	// we need CAP_DAC_READ_SEARCH becuase `/data/adb/ksud` is not accessible for non root process
-	// we add it here but don't add it to cap_inhertiable, it would be dropped automaticly after exec!
-	u64 cap_for_ksud =
-		profile->capabilities.effective | CAP_DAC_READ_SEARCH;
-	memcpy(&cred->cap_effective, &cap_for_ksud,
-	       sizeof(cred->cap_effective));
-	memcpy(&cred->cap_permitted, &profile->capabilities.effective,
-	       sizeof(cred->cap_permitted));
-	memcpy(&cred->cap_bset, &profile->capabilities.effective,
-	       sizeof(cred->cap_bset));
-
-	setup_groups(profile, cred);
-
-	commit_creds(cred);
-
-	spin_lock_irq(&current->sighand->siglock);
-	disable_seccomp(current);
-	spin_unlock_irq(&current->sighand->siglock);
-
-	setup_selinux(profile->selinux_domain);
-
-#ifdef KSU_SHOULD_USE_NEW_TP
-	for_each_thread (p, t) {
-		ksu_set_task_tracepoint_flag(t);
-	}
-#endif
-}
-
-extern void ext4_unregister_sysfs(struct super_block *sb);
-void nuke_ext4_sysfs(void)
-{
-#ifdef CONFIG_EXT4_FS
-	struct path path;
-	int err = kern_path("/data/adb/modules", 0, &path);
-	if (err) {
-		pr_err("%s: failed to get path, err %d\n", __func__, err);
-		return;
-	}
-
-	struct super_block *sb = path.dentry->d_inode->i_sb;
-	const char *name = sb->s_type->name;
-	if (strcmp(name, "ext4") != 0) {
-		pr_info("nuke_module: skipping s_type: %s\n", name);
-		path_put(&path);
-		return;
-	}
-
-	ext4_unregister_sysfs(sb);
-	pr_info("nuke_module: ext4 sysfs unregistered.\n");
-	path_put(&path);
-#endif
-}
-
-static bool is_appuid(kuid_t uid)
-{
-#define PER_USER_RANGE 100000
-#define FIRST_APPLICATION_UID 10000
-#define LAST_APPLICATION_UID 19999
-
-	uid_t appid = uid.val % PER_USER_RANGE;
-	return appid >= FIRST_APPLICATION_UID && appid <= LAST_APPLICATION_UID;
-}
-
-static bool should_umount(struct path *path)
-{
-	if (!path) {
-		return false;
-	}
-
-	if (current->nsproxy->mnt_ns == init_nsproxy.mnt_ns) {
-		pr_info("ignore global mnt namespace process: %d\n",
-			current_uid().val);
-		return false;
-	}
-
-	if (path->mnt && path->mnt->mnt_sb && path->mnt->mnt_sb->s_type) {
-		const char *fstype = path->mnt->mnt_sb->s_type->name;
-		return strcmp(fstype, "overlay") == 0;
-	}
-	return false;
-}
-
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 9, 0) ||                           \
-	defined(KSU_HAS_PATH_UMOUNT)
-extern int path_umount(struct path *path, int flags);
-#define ksu_umount_mnt(__unused, path, flags) (path_umount(path, flags))
-#else
-static int ksu_sys_umount(const char *mnt, int flags)
-{
-	char __user *usermnt = (char __user *)mnt;
-	mm_segment_t old_fs;
-	int ret; // although asmlinkage long
-
-	old_fs = get_fs();
-	set_fs(KERNEL_DS);
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 17, 0)
-	ret = ksys_umount(usermnt, flags);
-#else
-	ret = sys_umount(usermnt, flags); // cuz asmlinkage long sys##name
-#endif
-	set_fs(old_fs);
-	return ret;
-}
-
-#define ksu_umount_mnt(mnt, __unused, flags)                                   \
-	({                                                                     \
-		int ret;                                                       \
-		path_put(__unused);                                            \
-		ret = ksu_sys_umount(mnt, flags);                              \
-		ret;                                                           \
-	})
-
-#endif
-
-static void try_umount(const char *mnt, bool check_mnt, int flags)
-{
-	struct path path;
-	int ret;
-	int err = kern_path(mnt, 0, &path);
-	if (err) {
-		return;
-	}
-
-	if (path.dentry != path.mnt->mnt_root) {
-		// it is not root mountpoint, maybe umounted by others already.
-		path_put(&path);
-		return;
-	}
-
-	// we are only interest in some specific mounts
-	if (check_mnt && !should_umount(&path)) {
-		path_put(&path);
-		return;
-	}
-
-	ret = ksu_umount_mnt(mnt, &path, flags);
-	if (ret) {
-#ifdef CONFIG_KSU_DEBUG
-		pr_info("%s: path: %s, ret: %d\n", __func__, mnt, ret);
-#endif
-	}
-}
-
-static void ksu_do_umount_lists(void)
-{
-	// fixme: use `collect_mounts` and `iterate_mount` to iterate all mountpoint and
-	// filter the mountpoint whose target is `/data/adb`
-	try_umount("/odm", true, 0);
-	try_umount("/system", true, 0);
-	try_umount("/vendor", true, 0);
-	try_umount("/product", true, 0);
-	try_umount("/system_ext", true, 0);
-	try_umount("/data/adb/modules", false, MNT_DETACH);
-
-	try_umount("/debug_ramdisk", false, MNT_DETACH);
-	try_umount("/sbin", false, MNT_DETACH);
-}
-
-#if defined(MODULE) || defined(KSU_KPROBES_HOOK)
-struct umount_tw {
-	struct callback_head cb;
-	const struct cred *old_cred;
-};
-
-static void umount_tw_func(struct callback_head *cb)
-{
-	struct umount_tw *tw = container_of(cb, struct umount_tw, cb);
-	const struct cred *saved = NULL;
-	if (tw->old_cred) {
-		saved = override_creds(tw->old_cred);
-	}
-
-	ksu_do_umount_lists();
-
-	if (saved)
-		revert_creds(saved);
-
-	if (tw->old_cred)
-		put_cred(tw->old_cred);
-
-	kfree(tw);
-}
-#endif
-
-// force_sig kcompat, TODO: move it out of core_hook.c
-// https://elixir.bootlin.com/linux/v5.3-rc1/source/kernel/signal.c#L1613
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 3, 0)
-#define __force_sig(sig) force_sig(sig)
-#else
-#define __force_sig(sig) force_sig(sig, current)
-#endif
-
-int ksu_handle_setuid(struct cred *new, const struct cred *old)
-{
-	if (!new || !old) {
-		return 0;
-	}
-
-	kuid_t new_uid = new->uid;
-	kuid_t old_uid = old->uid;
-	// pr_info("handle_setuid from %d to %d\n", old_uid.val, new_uid.val);
-
-	if (0 != old_uid.val) {
-		// old process is not root, ignore it.
-		if (ksu_enhanced_security_enabled) {
-			// disallow any non-ksu domain escalation from non-root to root!
-			if (unlikely(new_uid.val) == 0) {
-				if (!is_ksu_domain()) {
-					pr_warn("find suspicious EoP: %d %s, from %d to %d\n",
-						current->pid, current->comm,
-						old_uid.val, new_uid.val);
-					__force_sig(SIGKILL);
-					return 0;
-				}
-			}
-			// disallow appuid decrease to any other uid if it is allowed to su
-			if (is_appuid(old_uid)) {
-				if (new_uid.val < old_uid.val &&
-				    !ksu_is_allow_uid_for_current(
-					    old_uid.val)) {
-					pr_warn("find suspicious EoP: %d %s, from %d to %d\n",
-						current->pid, current->comm,
-						old_uid.val, new_uid.val);
-					__force_sig(SIGKILL);
-					return 0;
-				}
-			}
-		}
-		return 0;
-	}
-
-#ifdef KSU_SHOULD_USE_NEW_TP
-	if (new_uid.val == 2000 && ksu_su_compat_enabled) {
-		ksu_set_task_tracepoint_flag(current);
-	}
-#endif
-
-	if (!is_appuid(new_uid) || is_unsupported_uid(new_uid.val)) {
-		// pr_info("handle setuid ignore non application or isolated uid: %d\n", new_uid.val);
-		return 0;
-	}
-
-	// if on private space, see if its possibly the manager
-	if (new_uid.val > 100000 &&
-	    new_uid.val % 100000 == ksu_get_manager_uid()) {
-		ksu_set_manager_uid(new_uid.val);
-	}
-
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 10, 0)
-	if (ksu_get_manager_uid() == new_uid.val) {
-		pr_info("install fd for ksu manager(uid=%d)\n", new_uid.val);
-		ksu_install_fd();
-		spin_lock_irq(&current->sighand->siglock);
-		ksu_seccomp_allow_cache(current->seccomp.filter, __NR_reboot);
-		spin_unlock_irq(&current->sighand->siglock);
-		ksu_set_task_tracepoint_flag(current);
-		return 0;
-	}
-
-	if (ksu_is_allow_uid_for_current(new_uid.val)) {
-		if (current->seccomp.mode == SECCOMP_MODE_FILTER &&
-		    current->seccomp.filter) {
-			spin_lock_irq(&current->sighand->siglock);
-			ksu_seccomp_allow_cache(current->seccomp.filter,
-						__NR_reboot);
-			spin_unlock_irq(&current->sighand->siglock);
-		}
-		if (ksu_su_compat_enabled) {
-			ksu_set_task_tracepoint_flag(current);
-		}
-	} else {
-		if (ksu_su_compat_enabled) {
-			// Disable syscall tracepoint sucompat for non-allowed processes
-			ksu_clear_task_tracepoint_flag(current);
-		}
-	}
-#else
-	if (ksu_is_allow_uid_for_current(new_uid.val)) {
-		spin_lock_irq(&current->sighand->siglock);
-		disable_seccomp(current);
-		spin_unlock_irq(&current->sighand->siglock);
-
-		if (ksu_get_manager_uid() == new_uid.val) {
-			pr_info("install fd for ksu manager(uid=%d)\n",
-				new_uid.val);
-			ksu_install_fd();
-		}
-
-		return 0;
-	}
-#endif
-
-	// this hook is used for umounting overlayfs for some uid, if there isn't any module mounted, just ignore it!
-	if (!ksu_module_mounted) {
-		return 0;
-	}
-
-	if (!ksu_kernel_umount_enabled) {
-		return 0;
-	}
-
-	if (!ksu_uid_should_umount(new_uid.val)) {
-		return 0;
-	} else {
-#ifdef CONFIG_KSU_DEBUG
-		pr_info("uid: %d should not umount!\n", current_uid().val);
-#endif
-	}
-
-	// check old process's selinux context, if it is not zygote, ignore it!
-	// because some su apps may setuid to untrusted_app but they are in global mount namespace
-	// when we umount for such process, that is a disaster!
-	if (!is_zygote(old)) {
-		pr_info("handle umount ignore non zygote child: %d\n",
-			current->pid);
-		return 0;
-	}
-#ifdef CONFIG_KSU_DEBUG
-	// umount the target mnt
-	pr_info("handle umount for uid: %d, pid: %d\n", new_uid.val,
-		current->pid);
-#endif
-
-#if defined(MODULE) || defined(KSU_KPROBES_HOOK)
-	struct umount_tw *tw;
-	tw = kmalloc(sizeof(*tw), GFP_ATOMIC);
-	if (!tw)
-		return 0;
-
-	tw->old_cred = get_current_cred();
-	tw->cb.func = umount_tw_func;
-
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 7, 8)
-	int err = task_work_add(current, &tw->cb, TWA_RESUME);
-#else
-	int err = task_work_add(current, &tw->cb, true);
-#endif
-
-	if (err) {
-		if (tw->old_cred) {
-			put_cred(tw->old_cred);
-		}
-		kfree(tw);
-		pr_warn("unmount add task_work failed\n");
-	}
-#else
-	ksu_do_umount_lists();
-#endif
-
-	return 0;
-}
-
-int ksu_inode_permission(struct inode *inode, int mask)
-{
-	if (inode && inode->i_sb 
-		&& unlikely(inode->i_sb->s_magic == DEVPTS_SUPER_MAGIC)) {
-		//pr_info("%s: handling devpts for: %s \n", __func__, current->comm);
-		__ksu_handle_devpts(inode);
-	}
-	return 0;
-}
-
-int ksu_bprm_check(struct linux_binprm *bprm)
-{
-	char *filename = (char *)bprm->filename;
-	
-	if (likely(!ksu_execveat_hook))
-		return 0;
-
-#ifdef CONFIG_COMPAT
-	static bool compat_check_done __read_mostly = false;
-	if ( unlikely(!compat_check_done) && unlikely(!strcmp(filename, "/data/adb/ksud"))
-		&& !memcmp(bprm->buf, "\x7f\x45\x4c\x46", 4) ) {
-		if (bprm->buf[4] == 0x01 )
-			ksu_is_compat = true;
-
-		pr_info("%s: %s ELF magic found! ksu_is_compat: %d \n", __func__, filename, ksu_is_compat);
-		compat_check_done = true;
-	}
-#endif
-
-	ksu_handle_pre_ksud(filename);
-
-	return 0;
-
-}
-
-int ksu_handle_sys_reboot(int magic1, int magic2, unsigned int cmd,
-			  void __user **arg)
-{
-	if (magic1 != KSU_INSTALL_MAGIC1)
-		return 0;
-
-#ifdef CONFIG_KSU_DEBUG
-	pr_info("sys_reboot: intercepted call! magic: 0x%x id: %d\n", magic1,
-		magic2);
-#endif
-
-	// Check if this is a request to install KSU fd
-	if (magic2 == KSU_INSTALL_MAGIC2) {
-		int fd = ksu_install_fd();
-		// downstream: dereference all arg usage!
-		if (copy_to_user((void __user *)*arg, &fd, sizeof(fd))) {
-			pr_err("install ksu fd reply err\n");
-		}
-		return 0;
-	}
-
-	// extensions
-
-	return 0;
-}
-
-// -- For old kernel compat?
-#if !defined(MODULE) && !defined(KSU_KPROBE_HOOK)
-static int ksu_task_fix_setuid(struct cred *new, const struct cred *old,
-			       int flags)
-{
-	return ksu_handle_setuid(new, old);
-}
-
-// kernel 4.4 and 4.9
-#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 10, 0) ||                           \
-	defined(CONFIG_IS_HW_HISI) || defined(CONFIG_KSU_ALLOWLIST_WORKAROUND)
-static int ksu_key_permission(key_ref_t key_ref, const struct cred *cred,
-			      unsigned perm)
-{
-	if (init_session_keyring != NULL) {
-		return 0;
-	}
-	if (strcmp(current->comm, "init")) {
-		// we are only interested in `init` process
-		return 0;
-	}
-	init_session_keyring = cred->session_keyring;
-	pr_info("kernel_compat: got init_session_keyring\n");
-	return 0;
-}
-#endif
-
-#include <linux/lsm_hooks.h>
-static struct security_hook_list ksu_hooks[] = {
-	LSM_HOOK_INIT(task_fix_setuid, ksu_task_fix_setuid),
-	LSM_HOOK_INIT(inode_permission, ksu_inode_permission),
-	#ifndef KSU_KPROBES_HOOK
-	LSM_HOOK_INIT(bprm_check_security, ksu_bprm_check),
-#endif
-#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 10, 0) ||                           \
-	defined(CONFIG_IS_HW_HISI) || defined(CONFIG_KSU_ALLOWLIST_WORKAROUND)
-	LSM_HOOK_INIT(key_permission, ksu_key_permission)
-#endif
-};
-
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 8, 0)
-static const struct lsm_id ksu_lsmid = {
-	.name = "ksu",
-	.id = 912,
-};
-#endif
-
-static void ksu_lsm_hook_init(void)
-{
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 8, 0)
-	security_add_hooks(ksu_hooks, ARRAY_SIZE(ksu_hooks), &ksu_lsmid);
-#elif LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
-	security_add_hooks(ksu_hooks, ARRAY_SIZE(ksu_hooks), "ksu");
-#else
-	// https://elixir.bootlin.com/linux/v4.10.17/source/include/linux/lsm_hooks.h#L1892
-	security_add_hooks(ksu_hooks, ARRAY_SIZE(ksu_hooks));
-#endif
-	pr_info("LSM hooks initialized.\n");
-}
-#else
-static void ksu_lsm_hook_init(void)
-{
-}
-#endif
-
-// -- For KPROBE and LKM handler
-#if defined(MODULE) || defined(KSU_KPROBES_HOOK)
-static int reboot_handler_pre(struct kprobe *p, struct pt_regs *regs)
-{
-	struct pt_regs *real_regs = PT_REAL_REGS(regs);
-	int magic1 = (int)PT_REGS_PARM1(real_regs);
-	int magic2 = (int)PT_REGS_PARM2(real_regs);
-	int cmd = (int)PT_REGS_PARM3(real_regs);
-	void __user **arg = (void __user **)&PT_REGS_SYSCALL_PARM4(real_regs);
-
-	return ksu_handle_sys_reboot(magic1, magic2, cmd, arg);
-}
-
-static struct kprobe reboot_kp = {
-	.symbol_name = REBOOT_SYMBOL,
-	.pre_handler = reboot_handler_pre,
-};
-
-// 2. cap_task_fix_setuid hook for handling setuid
-static int cap_task_fix_setuid_handler_pre(struct kprobe *p,
-					   struct pt_regs *regs)
-{
-	struct cred *new = (struct cred *)PT_REGS_PARM1(regs);
-	const struct cred *old = (const struct cred *)PT_REGS_PARM2(regs);
-
-	ksu_handle_setuid(new, old);
-
-	return 0;
-}
-
-static struct kprobe cap_task_fix_setuid_kp = {
-	.symbol_name = "cap_task_fix_setuid",
-	.pre_handler = cap_task_fix_setuid_handler_pre,
-};
-
-static int ksu_kprobe_init(void)
-{
-	int rc = 0;
-
-	// Register reboot kprobe
-	rc = register_kprobe(&reboot_kp);
-	if (rc) {
-		pr_err("reboot kprobe failed: %d\n", rc);
-		return rc;
-	}
-	pr_info("reboot kprobe registered successfully\n");
-
-	// Register cap_task_fix_setuid kprobe
-	rc = register_kprobe(&cap_task_fix_setuid_kp);
-	if (rc) {
-		pr_err("cap_task_fix_setuid kprobe failed: %d\n", rc);
-		unregister_kprobe(&reboot_kp);
-		return rc;
-	}
-	pr_info("cap_task_fix_setuid kprobe registered successfully\n");
-
-	return 0;
-}
-
-static void ksu_kprobe_exit(void)
-{
-	unregister_kprobe(&cap_task_fix_setuid_kp);
-	unregister_kprobe(&reboot_kp);
-}
-
-void __init ksu_core_init(void)
-{
-	int rc = ksu_kprobe_init();
-	if (rc) {
-		pr_err("ksu_kprobe_init failed: %d\n", rc);
-	}
-
-	if (ksu_register_feature_handler(&kernel_umount_handler)) {
-		pr_err("Failed to register umount feature handler\n");
-	}
-
-	if (ksu_register_feature_handler(&enhanced_security_handler)) {
-		pr_err("Failed to register enhanced security feature handler\n");
-	}
-}
-
-void ksu_core_exit(void)
-{
-	pr_info("ksu_core_exit\n");
-	ksu_kprobe_exit();
-	ksu_unregister_feature_handler(KSU_FEATURE_KERNEL_UMOUNT);
-}
-#else
-
-void __init ksu_core_init(void)
-{
-	ksu_lsm_hook_init();
-
-	if (ksu_register_feature_handler(&kernel_umount_handler)) {
-		pr_err("Failed to register umount feature handler\n");
-	}
-
-	if (ksu_register_feature_handler(&enhanced_security_handler)) {
-		pr_err("Failed to register enhanced security feature handler\n");
-	}
-}
-
-void ksu_core_exit(void)
-{
-	ksu_unregister_feature_handler(KSU_FEATURE_KERNEL_UMOUNT);
-	
-	ksu_uid_exit();
-	ksu_throne_comm_exit();
-
-	ksu_unregister_feature_handler(KSU_FEATURE_ENHANCED_SECURITY);
-}
-
-#endif

kernel/core_hook.h

@@ -1,16 +0,0 @@
-#ifndef __KSU_H_KSU_CORE
-#define __KSU_H_KSU_CORE
-
-#include <linux/init.h>
-#include "apk_sign.h"
-
-void __init ksu_core_init(void);
-void ksu_core_exit(void);
-
-void escape_to_root(void);
-
-void nuke_ext4_sysfs(void);
-
-extern bool ksu_module_mounted;
-
-#endif

kernel/feature.c

@@ -157,7 +157,7 @@ void ksu_feature_init(void)
 		feature_handlers[i] = NULL;
 	}
 
-	pr_info("%s: feature management initialized\n", __func__);
+	pr_info("feature: feature management initialized\n");
 }
 
 void ksu_feature_exit(void)
@@ -172,5 +172,5 @@ void ksu_feature_exit(void)
 
 	mutex_unlock(&feature_mutex);
 
-	pr_info("%s: feature management cleaned up\n", __func__);
+	pr_info("feature: feature management cleaned up\n");
 }

kernel/file_wrapper.c

@@ -1,68 +1,48 @@
-#include "linux/export.h"
+#include <linux/export.h>
 #include <linux/anon_inodes.h>
+#include <linux/aio.h> // kernel 3.18
 #include <linux/capability.h>
 #include <linux/cred.h>
 #include <linux/err.h>
 #include <linux/file.h>
 #include <linux/fs.h>
-#include <linux/poll.h>
+#include <linux/seq_file.h>
 #include <linux/slab.h>
-#include <linux/types.h>
 #include <linux/uaccess.h>
 #include <linux/version.h>
 
-#include "allowlist.h"
 #include "klog.h" // IWYU pragma: keep
-#include "ksu.h"
-#include "ksud.h"
-#include "manager.h"
 #include "selinux/selinux.h"
-#include "core_hook.h"
-#include "objsec.h"
 
 #include "file_wrapper.h"
 
-#ifdef KSU_NO___POLL_T
+static loff_t ksu_wrapper_llseek(struct file *fp, loff_t off, int flags) {
-typedef unsigned int ksu_poll_t;
-#else
-typedef __poll_t ksu_poll_t;
-#endif
-
-static loff_t ksu_wrapper_llseek(struct file *fp, loff_t off, int flags)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->llseek(data->orig, off, flags);
 }
 
-static ssize_t ksu_wrapper_read(struct file *fp, char __user *ptr, size_t sz,
+static ssize_t ksu_wrapper_read(struct file *fp, char __user *ptr, size_t sz, loff_t *off) {
-				 loff_t *off)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->read(orig, ptr, sz, off);
 }
 
-static ssize_t ksu_wrapper_write(struct file *fp, const char __user *ptr,
+static ssize_t ksu_wrapper_write(struct file *fp, const char __user *ptr, size_t sz, loff_t *off) {
-				  size_t sz, loff_t *off)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->write(orig, ptr, sz, off);
 }
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 16, 0)
-static ssize_t ksu_wrapper_read_iter(struct kiocb *iocb, struct iov_iter *iovi)
+static ssize_t ksu_wrapper_read_iter(struct kiocb *iocb, struct iov_iter *iovi) {
-{
 	struct ksu_file_wrapper* data = iocb->ki_filp->private_data;
 	struct file* orig = data->orig;
 	iocb->ki_filp = orig;
 	return orig->f_op->read_iter(iocb, iovi);
 }
 
-static ssize_t ksu_wrapper_write_iter(struct kiocb *iocb,
+static ssize_t ksu_wrapper_write_iter(struct kiocb *iocb, struct iov_iter *iovi) {
-				       struct iov_iter *iovi)
-{
 	struct ksu_file_wrapper* data = iocb->ki_filp->private_data;
 	struct file* orig = data->orig;
 	iocb->ki_filp = orig;
@@ -71,17 +51,14 @@ static ssize_t ksu_wrapper_write_iter(struct kiocb *iocb,
 #endif
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 1, 0)
-static int ksu_wrapper_iopoll(struct kiocb *kiocb, struct io_comp_batch *icb,
+static int ksu_wrapper_iopoll(struct kiocb *kiocb, struct io_comp_batch* icb, unsigned int v) {
-			       unsigned int v)
-{
 	struct ksu_file_wrapper* data = kiocb->ki_filp->private_data;
 	struct file* orig = data->orig;
 	kiocb->ki_filp = orig;
 	return orig->f_op->iopoll(kiocb, icb, v);
 }
 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(5, 1, 0)
-static int ksu_wrapper_iopoll(struct kiocb *kiocb, bool spin)
+static int ksu_wrapper_iopoll(struct kiocb *kiocb, bool spin) {
-{
 	struct ksu_file_wrapper* data = kiocb->ki_filp->private_data;
 	struct file* orig = data->orig;
 	kiocb->ki_filp = orig;
@@ -89,50 +66,51 @@ static int ksu_wrapper_iopoll(struct kiocb *kiocb, bool spin)
 }
 #endif
 
-#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 6, 0)
+#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 6, 0) && (LINUX_VERSION_CODE > KERNEL_VERSION(3, 11, 0) || defined(KSU_HAS_ITERATE_DIR))
-static int ksu_wrapper_iterate(struct file *fp, struct dir_context *dc)
+static int ksu_wrapper_iterate (struct file *fp, struct dir_context *dc) {
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->iterate(orig, dc);
 }
 #endif 
 
+// int (*readdir) (struct file *, void *, filldir_t);
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 11, 0) && !defined(KSU_HAS_ITERATE_DIR)
+static int ksu_wrapper_readdir(struct file *fp, void *ptr, filldir_t filler) {
+	struct ksu_file_wrapper* data = fp->private_data;
+	struct file* orig = data->orig;
+	return orig->f_op->readdir(orig, ptr, filler);
+}
+#endif
+
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 7, 0)
-static int ksu_wrapper_iterate_shared(struct file *fp, struct dir_context *dc)
+static int ksu_wrapper_iterate_shared(struct file *fp, struct dir_context *dc) {
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->iterate_shared(orig, dc);
 }
 #endif
 
-static ksu_poll_t ksu_wrapper_poll(struct file *fp,
+// typedef unsigned __bitwise __poll_t;
-				  struct poll_table_struct *pts)
+static unsigned __bitwise ksu_wrapper_poll(struct file *fp, struct poll_table_struct *pts) {
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->poll(orig, pts);
 }
 
-static long ksu_wrapper_unlocked_ioctl(struct file *fp, unsigned int cmd,
+static long ksu_wrapper_unlocked_ioctl(struct file *fp, unsigned int cmd, unsigned long arg) {
-					unsigned long arg)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->unlocked_ioctl(orig, cmd, arg);
 }
 
-static long ksu_wrapper_compat_ioctl(struct file *fp, unsigned int cmd,
+static long ksu_wrapper_compat_ioctl(struct file *fp, unsigned int cmd, unsigned long arg) {
-				      unsigned long arg)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->compat_ioctl(orig, cmd, arg);
 }
 
-static int ksu_wrapper_mmap(struct file *fp, struct vm_area_struct *vma)
+static int ksu_wrapper_mmap(struct file *fp, struct vm_area_struct * vma) {
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->mmap(orig, vma);
@@ -140,47 +118,41 @@ static int ksu_wrapper_mmap(struct file *fp, struct vm_area_struct *vma)
 
 // static unsigned long mmap_supported_flags {}
 
-static int ksu_wrapper_open(struct inode *ino, struct file *fp)
+static int ksu_wrapper_open(struct inode *ino, struct file *fp) {
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	struct inode *orig_ino = file_inode(orig);
 	return orig->f_op->open(orig_ino, orig);
 }
 
-static int ksu_wrapper_flush(struct file *fp, fl_owner_t id)
+static int ksu_wrapper_flush(struct file *fp, fl_owner_t id) {
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->flush(orig, id);
 }
 
-static int ksu_wrapper_fsync(struct file *fp, loff_t off1, loff_t off2,
+
-			      int datasync)
+static int ksu_wrapper_fsync(struct file *fp, loff_t off1, loff_t off2, int datasync) {
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->fsync(orig, off1, off2, datasync);
 }
 
-static int ksu_wrapper_fasync(int arg, struct file *fp, int arg2)
+static int ksu_wrapper_fasync(int arg, struct file *fp, int arg2) {
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->fasync(arg, orig, arg2);
 }
 
-static int ksu_wrapper_lock(struct file *fp, int arg1, struct file_lock *fl)
+static int ksu_wrapper_lock(struct file *fp, int arg1, struct file_lock *fl) {
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	return orig->f_op->lock(orig, arg1, fl);
 }
 
+
 #if LINUX_VERSION_CODE < KERNEL_VERSION(6, 6, 0)
-static ssize_t ksu_wrapper_sendpage(struct file *fp, struct page *pg, int arg1,
+static ssize_t ksu_wrapper_sendpage(struct file *fp, struct page *pg, int arg1, size_t sz, loff_t *off, int arg2) {
-				     size_t sz, loff_t *off, int arg2)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->sendpage) {
@@ -190,25 +162,18 @@ static ssize_t ksu_wrapper_sendpage(struct file *fp, struct page *pg, int arg1,
 }
 #endif
 
-static unsigned long ksu_wrapper_get_unmapped_area(struct file *fp,
+static unsigned long ksu_wrapper_get_unmapped_area(struct file *fp, unsigned long arg1, unsigned long arg2, unsigned long arg3, unsigned long arg4) {
-						    unsigned long arg1,
-						    unsigned long arg2,
-						    unsigned long arg3,
-						    unsigned long arg4)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->get_unmapped_area) {
-		return orig->f_op->get_unmapped_area(orig, arg1, arg2, arg3,
+		return orig->f_op->get_unmapped_area(orig, arg1, arg2, arg3, arg4);
-						     arg4);
 	}
 	return -EINVAL;
 }
 
 // static int ksu_wrapper_check_flags(int arg) {}
 
-static int ksu_wrapper_flock(struct file *fp, int arg1, struct file_lock *fl)
+static int ksu_wrapper_flock(struct file *fp, int arg1, struct file_lock *fl) {
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->flock) {
@@ -217,10 +182,7 @@ static int ksu_wrapper_flock(struct file *fp, int arg1, struct file_lock *fl)
 	return -EINVAL;
 }
 
-static ssize_t ksu_wrapper_splice_write(struct pipe_inode_info *pii,
+static ssize_t ksu_wrapper_splice_write(struct pipe_inode_info * pii, struct file *fp, loff_t *off, size_t sz, unsigned int arg1) {
-					 struct file *fp, loff_t *off,
-					 size_t sz, unsigned int arg1)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->splice_write) {
@@ -229,10 +191,7 @@ static ssize_t ksu_wrapper_splice_write(struct pipe_inode_info *pii,
 	return -EINVAL;
 }
 
-static ssize_t ksu_wrapper_splice_read(struct file *fp, loff_t *off,
+static ssize_t ksu_wrapper_splice_read(struct file *fp, loff_t *off, struct pipe_inode_info *pii, size_t sz, unsigned int arg1) {
-					struct pipe_inode_info *pii, size_t sz,
-					unsigned int arg1)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->splice_read) {
@@ -242,8 +201,7 @@ static ssize_t ksu_wrapper_splice_read(struct file *fp, loff_t *off,
 }
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 6, 0)
-void ksu_wrapper_splice_eof(struct file *fp)
+void ksu_wrapper_splice_eof(struct file *fp) {
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->splice_eof) {
@@ -253,9 +211,7 @@ void ksu_wrapper_splice_eof(struct file *fp)
 #endif
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 12, 0)
-static int ksu_wrapper_setlease(struct file *fp, int arg1,
+static int ksu_wrapper_setlease(struct file *fp, int arg1, struct file_lease **fl, void **p) {
-				 struct file_lease **fl, void **p)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->setlease) {
@@ -264,9 +220,7 @@ static int ksu_wrapper_setlease(struct file *fp, int arg1,
 	return -EINVAL;
 }
 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(6, 6, 0)
-static int ksu_wrapper_setlease(struct file *fp, int arg1,
+static int ksu_wrapper_setlease(struct file *fp, int arg1, struct file_lock **fl, void **p) {
-				 struct file_lock **fl, void **p)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->setlease) {
@@ -274,11 +228,8 @@ static int ksu_wrapper_setlease(struct file *fp, int arg1,
 	}
 	return -EINVAL;
 }
-// int (*setlease)(struct file *, long, struct file_lock **, void **);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3, 18, 0) // int (*setlease)(struct file *, long, struct file_lock **, void **);
-#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3, 18, 0)
+static int ksu_wrapper_setlease(struct file *fp, long arg1, struct file_lock **fl, void **p) {
-static int ksu_wrapper_setlease(struct file *fp, long arg1,
-				 struct file_lock **fl, void **p)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->setlease) {
@@ -287,9 +238,7 @@ static int ksu_wrapper_setlease(struct file *fp, long arg1,
 	return -EINVAL;
 }
 #else // int (*setlease)(struct file *, long, struct file_lock **);
-static int ksu_wrapper_setlease(struct file *fp, long arg1,
+static int ksu_wrapper_setlease(struct file *fp, long arg1, struct file_lock **fl) {
-				 struct file_lock **fl)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->setlease) {
@@ -299,9 +248,7 @@ static int ksu_wrapper_setlease(struct file *fp, long arg1,
 }
 #endif
 
-static long ksu_wrapper_fallocate(struct file *fp, int mode, loff_t offset,
+static long ksu_wrapper_fallocate(struct file *fp, int mode, loff_t offset, loff_t len) {
-				   loff_t len)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->fallocate) {
@@ -311,19 +258,16 @@ static long ksu_wrapper_fallocate(struct file *fp, int mode, loff_t offset,
 }
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 19, 0)
-static void ksu_wrapper_show_fdinfo(struct seq_file *m, struct file *f)
+static void ksu_wrapper_show_fdinfo(struct seq_file *m, struct file *f) {
-{
+	struct ksu_file_wrapper* data = f->private_data;
-	struct ksu_file_wrapper *data = m->file->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->show_fdinfo) {
 		orig->f_op->show_fdinfo(m, orig);
 	}
 }
-
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3, 8, 0)
-#else
+static int ksu_wrapper_show_fdinfo(struct seq_file *m, struct file *f) {
-static int ksu_wrapper_show_fdinfo(struct seq_file *m, struct file *f)
+	struct ksu_file_wrapper* data = f->private_data;
-{
-	struct ksu_file_wrapper *data = m->file->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->show_fdinfo) {
 		orig->f_op->show_fdinfo(m, orig);
@@ -333,42 +277,38 @@ static int ksu_wrapper_show_fdinfo(struct seq_file *m, struct file *f)
 #endif
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 7, 0)
-static ssize_t ksu_wrapper_copy_file_range(struct file *f1, loff_t off1,
+// https://cs.android.com/android/kernel/superproject/+/common-android-mainline:common/fs/read_write.c;l=1593-1606;drc=398da7defe218d3e51b0f3bdff75147e28125b60
-					    struct file *f2, loff_t off2,
+static ssize_t ksu_wrapper_copy_file_range(struct file *file_in, loff_t pos_in, struct file *file_out,
-					    size_t sz, unsigned int flags)
+		loff_t pos_out, size_t len, unsigned int flags) {
-{
+	struct ksu_file_wrapper* data = file_out->private_data;
-	// TODO: determine which file to use
-	struct ksu_file_wrapper *data = f1->private_data;
 	struct file* orig = data->orig;
-	if (orig->f_op->copy_file_range) {
+	return orig->f_op->copy_file_range(file_in, pos_in, orig, pos_out, len, flags);
-		return orig->f_op->copy_file_range(orig, off1, f2, off2, sz,
-						   flags);
-	}
-	return -EINVAL;
 }
 #endif
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 20, 0)
+// no REMAP_FILE_DEDUP: use file_in
+// https://cs.android.com/android/kernel/superproject/+/common-android-mainline:common/fs/read_write.c;l=1598-1599;drc=398da7defe218d3e51b0f3bdff75147e28125b60
+// https://cs.android.com/android/kernel/superproject/+/common-android-mainline:common/fs/remap_range.c;l=403-404;drc=398da7defe218d3e51b0f3bdff75147e28125b60
+// REMAP_FILE_DEDUP: use file_out
+// https://cs.android.com/android/kernel/superproject/+/common-android-mainline:common/fs/remap_range.c;l=483-484;drc=398da7defe218d3e51b0f3bdff75147e28125b60
 static loff_t ksu_wrapper_remap_file_range(struct file *file_in, loff_t pos_in,
-					    struct file *file_out,
+				struct file *file_out, loff_t pos_out,
-					    loff_t pos_out, loff_t len,
+				loff_t len, unsigned int remap_flags) {
-					    unsigned int remap_flags)
+	if (remap_flags & REMAP_FILE_DEDUP) {
-{
+		struct ksu_file_wrapper* data = file_out->private_data;
-	// TODO: determine which file to use
+		struct file* orig = data->orig;
+		return orig->f_op->remap_file_range(file_in, pos_in, orig, pos_out, len, remap_flags);
+	} else {
 		struct ksu_file_wrapper* data = file_in->private_data;
 		struct file* orig = data->orig;
-	if (orig->f_op->remap_file_range) {
+		return orig->f_op->remap_file_range(orig, pos_in, file_out, pos_out, len, remap_flags);
-		return orig->f_op->remap_file_range(orig, pos_in, file_out,
-						    pos_out, len, remap_flags);
 	}
-	return -EINVAL;
 }
 #endif
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 19, 0)
-static int ksu_wrapper_fadvise(struct file *fp, loff_t off1, loff_t off2,
+static int ksu_wrapper_fadvise(struct file *fp, loff_t off1, loff_t off2, int flags) {
-				int flags)
-{
 	struct ksu_file_wrapper* data = fp->private_data;
 	struct file* orig = data->orig;
 	if (orig->f_op->fadvise) {
@@ -378,16 +318,13 @@ static int ksu_wrapper_fadvise(struct file *fp, loff_t off1, loff_t off2,
 }
 #endif
 
-static int ksu_wrapper_release(struct inode *inode, struct file *filp)
+static int ksu_wrapper_release(struct inode *inode, struct file *filp) {
-{
 	ksu_delete_file_wrapper(filp->private_data);
     return 0;
 }
 
-struct ksu_file_wrapper *ksu_create_file_wrapper(struct file *fp)
+struct ksu_file_wrapper* ksu_create_file_wrapper(struct file* fp) {
-{
+	struct ksu_file_wrapper* p = kcalloc(sizeof(struct ksu_file_wrapper), 1, GFP_KERNEL);
-	struct ksu_file_wrapper *p =
-		kcalloc(sizeof(struct ksu_file_wrapper), 1, GFP_KERNEL);
 	if (!p) {
 		return NULL;
 	}
@@ -401,24 +338,23 @@ struct ksu_file_wrapper *ksu_create_file_wrapper(struct file *fp)
 	p->ops.write = fp->f_op->write ? ksu_wrapper_write : NULL;
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 16, 0)
 	p->ops.read_iter = fp->f_op->read_iter ? ksu_wrapper_read_iter : NULL;
-	p->ops.write_iter =
+	p->ops.write_iter = fp->f_op->write_iter ? ksu_wrapper_write_iter : NULL;
-		fp->f_op->write_iter ? ksu_wrapper_write_iter : NULL;
 #endif
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 1, 0)
 	p->ops.iopoll = fp->f_op->iopoll ? ksu_wrapper_iopoll : NULL;
 #endif
-#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 6, 0)
+#if LINUX_VERSION_CODE < KERNEL_VERSION(6, 6, 0) && (LINUX_VERSION_CODE > KERNEL_VERSION(3, 11, 0) || defined(KSU_HAS_ITERATE_DIR))
 	p->ops.iterate = fp->f_op->iterate ? ksu_wrapper_iterate : NULL;
 #endif
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 11, 0) && !defined(KSU_HAS_ITERATE_DIR)
+	p->ops.readdir = fp->f_op->readdir ? ksu_wrapper_readdir : NULL;
+#endif
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 7, 0)
-	p->ops.iterate_shared =
+	p->ops.iterate_shared = fp->f_op->iterate_shared ? ksu_wrapper_iterate_shared : NULL;
-		fp->f_op->iterate_shared ? ksu_wrapper_iterate_shared : NULL;
 #endif
 	p->ops.poll = fp->f_op->poll ? ksu_wrapper_poll : NULL;
-	p->ops.unlocked_ioctl =
+	p->ops.unlocked_ioctl = fp->f_op->unlocked_ioctl ? ksu_wrapper_unlocked_ioctl : NULL;
-		fp->f_op->unlocked_ioctl ? ksu_wrapper_unlocked_ioctl : NULL;
+	p->ops.compat_ioctl = fp->f_op->compat_ioctl ? ksu_wrapper_compat_ioctl : NULL;
-	p->ops.compat_ioctl =
-		fp->f_op->compat_ioctl ? ksu_wrapper_compat_ioctl : NULL;
 	p->ops.mmap = fp->f_op->mmap ? ksu_wrapper_mmap : NULL;
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 12, 0)
 	p->ops.fop_flags = fp->f_op->fop_flags;
@@ -434,42 +370,33 @@ struct ksu_file_wrapper *ksu_create_file_wrapper(struct file *fp)
 #if LINUX_VERSION_CODE < KERNEL_VERSION(6, 6, 0)
 	p->ops.sendpage = fp->f_op->sendpage ? ksu_wrapper_sendpage : NULL;
 #endif
-	p->ops.get_unmapped_area = fp->f_op->get_unmapped_area ?
+	p->ops.get_unmapped_area = fp->f_op->get_unmapped_area ? ksu_wrapper_get_unmapped_area : NULL;
-					   ksu_wrapper_get_unmapped_area :
-					   NULL;
 	p->ops.check_flags = fp->f_op->check_flags;
 	p->ops.flock = fp->f_op->flock ? ksu_wrapper_flock : NULL;
-	p->ops.splice_write =
+	p->ops.splice_write = fp->f_op->splice_write ? ksu_wrapper_splice_write : NULL;
-		fp->f_op->splice_write ? ksu_wrapper_splice_write : NULL;
+	p->ops.splice_read = fp->f_op->splice_read ? ksu_wrapper_splice_read : NULL;
-	p->ops.splice_read =
-		fp->f_op->splice_read ? ksu_wrapper_splice_read : NULL;
 	p->ops.setlease = fp->f_op->setlease ? ksu_wrapper_setlease : NULL;
 	p->ops.fallocate = fp->f_op->fallocate ? ksu_wrapper_fallocate : NULL;
-	p->ops.show_fdinfo =
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 8, 0)
-		fp->f_op->show_fdinfo ? ksu_wrapper_show_fdinfo : NULL;
+	p->ops.show_fdinfo = fp->f_op->show_fdinfo ? ksu_wrapper_show_fdinfo : NULL;
+#endif
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 7, 0)
-	p->ops.copy_file_range =
+	p->ops.copy_file_range = fp->f_op->copy_file_range ? ksu_wrapper_copy_file_range : NULL;
-		fp->f_op->copy_file_range ? ksu_wrapper_copy_file_range : NULL;
 #endif
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 20, 0)
-	p->ops.remap_file_range = fp->f_op->remap_file_range ?
+	p->ops.remap_file_range = fp->f_op->remap_file_range ? ksu_wrapper_remap_file_range : NULL;
-					  ksu_wrapper_remap_file_range :
-					  NULL;
 #endif
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 19, 0)
 	p->ops.fadvise = fp->f_op->fadvise ? ksu_wrapper_fadvise : NULL;
 #endif
-
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 6, 0)
-	p->ops.splice_eof =
+	p->ops.splice_eof = fp->f_op->splice_eof ? ksu_wrapper_splice_eof : NULL;
-		fp->f_op->splice_eof ? ksu_wrapper_splice_eof : NULL;
 #endif
 
 	return p;
 }
 
-void ksu_delete_file_wrapper(struct ksu_file_wrapper *data)
+void ksu_delete_file_wrapper(struct ksu_file_wrapper* data) {
-{
 	fput((struct file*) data->orig);
 	kfree(data);
 }

kernel/file_wrapper.h

@@ -1,5 +1,5 @@
-#ifndef __KSU_H_FILE_WRAPPER
+#ifndef KSU_FILE_WRAPPER_H
-#define __KSU_H_FILE_WRAPPER
+#define KSU_FILE_WRAPPER_H
 
 #include <linux/file.h>
 #include <linux/fs.h>
@@ -12,4 +12,4 @@ struct ksu_file_wrapper {
 struct ksu_file_wrapper *ksu_create_file_wrapper(struct file *fp);
 void ksu_delete_file_wrapper(struct ksu_file_wrapper *data);
 
-#endif
+#endif // KSU_FILE_WRAPPER_H

kernel/include/ksu_hook.h

@@ -1,28 +0,0 @@
-#ifndef __KSU_H_KSHOOK
-#define __KSU_H_KSHOOK
-
-#include <linux/fs.h>
-#include <linux/types.h>
-
-// For sucompat
-
-int ksu_handle_faccessat(int *dfd, const char __user **filename_user, int *mode,
-			 int *flags);
-
-int ksu_handle_stat(int *dfd, const char __user **filename_user, int *flags);
-
-// For ksud
-
-int ksu_handle_vfs_read(struct file **file_ptr, char __user **buf_ptr,
-			size_t *count_ptr, loff_t **pos);
-
-// For ksud and sucompat
-
-int ksu_handle_execveat(int *fd, struct filename **filename_ptr, void *argv,
-			void *envp, int *flags);
-
-// For volume button
-int ksu_handle_input_handle_event(unsigned int *type, unsigned int *code,
-				  int *value);
-
-#endif

kernel/kernel_compat.c

@@ -1,14 +1,11 @@
 #include <linux/version.h>
 #include <linux/fs.h>
-#include <linux/nsproxy.h>
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 10, 0)
 #include <linux/sched/task.h>
 #else
 #include <linux/sched.h>
 #endif
 #include <linux/uaccess.h>
-#include <linux/filter.h>
-#include <linux/seccomp.h>
 #include "klog.h" // IWYU pragma: keep
 #include "kernel_compat.h"
 
@@ -17,10 +14,12 @@
 #include <linux/key.h>
 #include <linux/errno.h>
 #include <linux/cred.h>
+#include <linux/lsm_hooks.h>
 
 extern int install_session_keyring_to_cred(struct cred *, struct key *);
 struct key *init_session_keyring = NULL;
-static inline int install_session_keyring(struct key *keyring)
+
+static int install_session_keyring(struct key *keyring)
 {
 	struct cred *new;
 	int ret;
@@ -124,117 +123,3 @@ long ksu_strncpy_from_user_nofault(char *dst, const void __user *unsafe_addr,
 	return ret;
 }
 #endif
-
-long ksu_strncpy_from_user_retry(char *dst, const void __user *unsafe_addr,
-				 long count)
-{
-	long ret;
-
-	ret = ksu_strncpy_from_user_nofault(dst, unsafe_addr, count);
-	if (likely(ret >= 0))
-		return ret;
-
-	// we faulted! fallback to slow path
-	if (unlikely(!ksu_access_ok(unsafe_addr, count))) {
-#ifdef CONFIG_KSU_DEBUG
-		pr_err("%s: faulted!\n", __func__);
-#endif
-		return -EFAULT;
-	}
-
-	// why we don't do like how strncpy_from_user_nofault?
-	ret = strncpy_from_user(dst, unsafe_addr, count);
-
-	if (ret >= count) {
-		ret = count;
-		dst[ret - 1] = '\0';
-	} else if (likely(ret >= 0)) {
-		ret++;
-	}
-
-	return ret;
-}
-
-long ksu_copy_from_user_nofault(void *dst, const void __user *src, size_t size)
-{
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 8, 0)
-	return copy_from_user_nofault(dst, src, size);
-#else
-	// https://elixir.bootlin.com/linux/v5.8/source/mm/maccess.c#L205
-	long ret = -EFAULT;
-	mm_segment_t old_fs = get_fs();
-
-	set_fs(USER_DS);
-	// tweaked to use ksu_access_ok
-	if (ksu_access_ok(src, size)) {
-		pagefault_disable();
-		ret = __copy_from_user_inatomic(dst, src, size);
-		pagefault_enable();
-	}
-	set_fs(old_fs);
-
-	if (ret)
-		return -EFAULT;
-	return 0;
-#endif
-}
-
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 10, 0)
-
-struct action_cache {
-	DECLARE_BITMAP(allow_native, SECCOMP_ARCH_NATIVE_NR);
-#ifdef SECCOMP_ARCH_COMPAT
-	DECLARE_BITMAP(allow_compat, SECCOMP_ARCH_COMPAT_NR);
-#endif
-};
-
-struct seccomp_filter {
-	refcount_t refs;
-	refcount_t users;
-	bool log;
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 1, 0)
-	bool wait_killable_recv;
-#endif
-	struct action_cache cache;
-	struct seccomp_filter *prev;
-	struct bpf_prog *prog;
-	struct notification *notif;
-	struct mutex notify_lock;
-	wait_queue_head_t wqh;
-};
-
-void ksu_seccomp_clear_cache(struct seccomp_filter *filter, int nr)
-{
-	if (!filter) {
-		return;
-	}
-
-	if (nr >= 0 && nr < SECCOMP_ARCH_NATIVE_NR) {
-		clear_bit(nr, filter->cache.allow_native);
-	}
-
-#ifdef SECCOMP_ARCH_COMPAT
-	if (nr >= 0 && nr < SECCOMP_ARCH_COMPAT_NR) {
-		clear_bit(nr, filter->cache.allow_compat);
-	}
-#endif
-}
-
-void ksu_seccomp_allow_cache(struct seccomp_filter *filter, int nr)
-{
-	if (!filter) {
-		return;
-	}
-
-	if (nr >= 0 && nr < SECCOMP_ARCH_NATIVE_NR) {
-		set_bit(nr, filter->cache.allow_native);
-	}
-
-#ifdef SECCOMP_ARCH_COMPAT
-	if (nr >= 0 && nr < SECCOMP_ARCH_COMPAT_NR) {
-		set_bit(nr, filter->cache.allow_compat);
-	}
-#endif
-}
-
-#endif

kernel/kernel_compat.h

@@ -3,6 +3,7 @@
 
 #include <linux/fs.h>
 #include <linux/version.h>
+#include <linux/task_work.h>
 #include "ss/policydb.h"
 #include "linux/key.h"
 
@@ -27,21 +28,9 @@
 #endif
 #endif
 
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 10, 0) && defined(KSU_KPROBE_HOOK)
-#define KSU_SHOULD_USE_NEW_TP
-#endif
-
 extern long ksu_strncpy_from_user_nofault(char *dst,
 					  const void __user *unsafe_addr,
 					  long count);
-extern long ksu_strncpy_from_user_retry(char *dst,
-					const void __user *unsafe_addr,
-					long count);
-
-#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 10, 0) ||                           \
-	defined(CONFIG_IS_HW_HISI) || defined(CONFIG_KSU_ALLOWLIST_WORKAROUND)
-extern struct key *init_session_keyring;
-#endif
 
 extern struct file *ksu_filp_open_compat(const char *filename, int flags,
 					 umode_t mode);
@@ -67,9 +56,9 @@ static long ksu_copy_from_user_retry(void *to,
 	return copy_from_user(to, from, count);
 }
 
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 10, 0)
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 10, 0) ||                           \
-extern void ksu_seccomp_clear_cache(struct seccomp_filter *filter, int nr);
+	defined(CONFIG_IS_HW_HISI) || defined(CONFIG_KSU_ALLOWLIST_WORKAROUND)
-extern void ksu_seccomp_allow_cache(struct seccomp_filter *filter, int nr);
+extern struct key *init_session_keyring;
 #endif
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 0, 0)
@@ -78,4 +67,18 @@ extern void ksu_seccomp_allow_cache(struct seccomp_filter *filter, int nr);
 #define ksu_access_ok(addr, size) access_ok(VERIFY_READ, addr, size)
 #endif
 
+// Linux >= 5.7
+// task_work_add (struct, struct, enum)
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 7, 0)
+#define ksu_task_work_add(tsk, cb, notify) task_work_add(tsk, cb, notify)
+#else
+// Linux pre-5.7
+// task_work_add (struct, struct, bool)
+#define ksu_task_work_add(tsk, cb, notify) task_work_add(tsk, cb, notify)
+// Decoy, so it wouldn't complain.
+#ifndef TWA_RESUME
+#define TWA_RESUME	true
+#endif
+#endif
+
 #endif

kernel/kernel_umount.c

@@ -0,0 +1,227 @@
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/task_work.h>
+#include <linux/cred.h>
+#include <linux/fs.h>
+#include <linux/mount.h>
+#include <linux/namei.h>
+#include <linux/nsproxy.h>
+#include <linux/path.h>
+#include <linux/printk.h>
+#include <linux/types.h>
+#ifndef KSU_HAS_PATH_UMOUNT
+#include <linux/syscalls.h>
+#endif
+
+#include "kernel_umount.h"
+#include "klog.h" // IWYU pragma: keep
+#include "allowlist.h"
+#include "kernel_compat.h"
+#include "selinux/selinux.h"
+#include "feature.h"
+#include "ksud.h"
+
+static bool ksu_kernel_umount_enabled = true;
+
+static int kernel_umount_feature_get(u64 *value)
+{
+	*value = ksu_kernel_umount_enabled ? 1 : 0;
+	return 0;
+}
+
+static int kernel_umount_feature_set(u64 value)
+{
+	bool enable = value != 0;
+	ksu_kernel_umount_enabled = enable;
+	pr_info("kernel_umount: set to %d\n", enable);
+	return 0;
+}
+
+static const struct ksu_feature_handler kernel_umount_handler = {
+	.feature_id = KSU_FEATURE_KERNEL_UMOUNT,
+	.name = "kernel_umount",
+	.get_handler = kernel_umount_feature_get,
+	.set_handler = kernel_umount_feature_set,
+};
+
+static bool should_umount(struct path *path)
+{
+	if (!path) {
+		return false;
+	}
+
+	if (current->nsproxy->mnt_ns == init_nsproxy.mnt_ns) {
+		pr_info("ignore global mnt namespace process: %d\n",
+			current_uid().val);
+		return false;
+	}
+
+	if (path->mnt && path->mnt->mnt_sb && path->mnt->mnt_sb->s_type) {
+		const char *fstype = path->mnt->mnt_sb->s_type->name;
+		return strcmp(fstype, "overlay") == 0;
+	}
+	return false;
+}
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 9, 0) || defined(KSU_HAS_PATH_UMOUNT)
+extern int path_umount(struct path *path, int flags);
+static void ksu_umount_mnt(const char *__never_use_mnt, struct path *path, int flags)
+{
+	int err = path_umount(path, flags);
+	if (err) {
+		pr_info("umount %s failed: %d\n", path->dentry->d_iname, err);
+	}
+}
+#else
+static void ksu_sys_umount(const char *mnt, int flags)
+{
+	char __user *usermnt = (char __user *)mnt;
+	mm_segment_t old_fs;
+
+	old_fs = get_fs();
+	set_fs(KERNEL_DS);
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 17, 0)
+	ksys_umount(usermnt, flags);
+#else
+	sys_umount(usermnt, flags); // cuz asmlinkage long sys##name
+#endif
+	set_fs(old_fs);
+}
+
+#define ksu_umount_mnt(mnt, __unused, flags)                                   \
+	({                                                                     \
+		path_put(__unused);                                            \
+		ksu_sys_umount(mnt, flags);                              \
+	})
+
+#endif
+
+static void try_umount(const char *mnt, bool check_mnt, int flags)
+{
+	struct path path;
+	int err = kern_path(mnt, 0, &path);
+	if (err) {
+		return;
+	}
+
+	if (path.dentry != path.mnt->mnt_root) {
+		// it is not root mountpoint, maybe umounted by others already.
+		path_put(&path);
+		return;
+	}
+
+	// we are only interest in some specific mounts
+	if (check_mnt && !should_umount(&path)) {
+		path_put(&path);
+		return;
+	}
+
+	ksu_umount_mnt(mnt, &path, flags);
+}
+
+static inline void do_ksu_umount_lists(void)
+{
+	// fixme: use `collect_mounts` and `iterate_mount` to iterate all mountpoint and
+	// filter the mountpoint whose target is `/data/adb`
+	try_umount("/odm", true, 0);
+	try_umount("/system", true, 0);
+	try_umount("/vendor", true, 0);
+	try_umount("/product", true, 0);
+	try_umount("/system_ext", true, 0);
+	try_umount("/data/adb/modules", false, MNT_DETACH);
+}
+
+#ifdef KSU_SHOULD_USE_NEW_TP
+struct umount_tw {
+	struct callback_head cb;
+	const struct cred *old_cred;
+};
+
+static void umount_tw_func(struct callback_head *cb)
+{
+	struct umount_tw *tw = container_of(cb, struct umount_tw, cb);
+	const struct cred *saved = NULL;
+	if (tw->old_cred) {
+		saved = override_creds(tw->old_cred);
+	}
+
+	do_ksu_umount_lists();
+
+	if (saved)
+		revert_creds(saved);
+
+	if (tw->old_cred)
+		put_cred(tw->old_cred);
+
+	kfree(tw);
+}
+#endif
+
+int ksu_handle_umount(uid_t old_uid, uid_t new_uid)
+{
+	// this hook is used for umounting overlayfs for some uid, if there isn't any module mounted, just ignore it!
+	if (!ksu_module_mounted) {
+		return 0;
+	}
+
+	if (!ksu_kernel_umount_enabled) {
+		return 0;
+	}
+
+	// FIXME: isolated process which directly forks from zygote is not handled
+	if (!is_appuid(new_uid)) {
+		return 0;
+	}
+
+	if (!ksu_uid_should_umount(new_uid)) {
+		return 0;
+	}
+
+	// check old process's selinux context, if it is not zygote, ignore it!
+	// because some su apps may setuid to untrusted_app but they are in global mount namespace
+	// when we umount for such process, that is a disaster!
+	bool is_zygote_child = is_zygote(get_current_cred());
+	if (!is_zygote_child) {
+		pr_info("handle umount ignore non zygote child: %d\n",
+			current->pid);
+		return 0;
+	}
+	// umount the target mnt
+	pr_info("handle umount for uid: %d, pid: %d\n", new_uid, current->pid);
+
+#ifdef KSU_SHOULD_USE_NEW_TP
+	struct umount_tw *tw;
+	tw = kzalloc(sizeof(*tw), GFP_ATOMIC);
+	if (!tw)
+		return 0;
+
+	tw->old_cred = get_current_cred();
+	tw->cb.func = umount_tw_func;
+
+	int err = ksu_task_work_add(current, &tw->cb, TWA_RESUME);
+	if (err) {
+		if (tw->old_cred) {
+			put_cred(tw->old_cred);
+		}
+		kfree(tw);
+		pr_warn("unmount add task_work failed\n");
+	}
+#else
+	// Using task work for non-kp context is expansive?
+	do_ksu_umount_lists();
+#endif
+
+	return 0;
+}
+
+void ksu_kernel_umount_init(void)
+{
+	if (ksu_register_feature_handler(&kernel_umount_handler)) {
+		pr_err("Failed to register kernel_umount feature handler\n");
+	}
+}
+
+void ksu_kernel_umount_exit(void)
+{
+	ksu_unregister_feature_handler(KSU_FEATURE_KERNEL_UMOUNT);
+}

kernel/kernel_umount.h

@@ -0,0 +1,12 @@
+#ifndef __KSU_H_KERNEL_UMOUNT
+#define __KSU_H_KERNEL_UMOUNT
+
+#include <linux/types.h>
+
+void ksu_kernel_umount_init(void);
+void ksu_kernel_umount_exit(void);
+
+// Handler function to be called from setresuid hook
+int ksu_handle_umount(uid_t old_uid, uid_t new_uid);
+
+#endif

kernel/ksu.c

@@ -1,5 +1,6 @@
 #include <linux/export.h>
 #include <linux/fs.h>
+#include <linux/printk.h>
 #include <linux/kobject.h>
 #include <linux/module.h>
 #include <linux/workqueue.h>
@@ -9,15 +10,17 @@
 
 #include "allowlist.h"
 #include "arch.h"
-#include "core_hook.h"
 #include "feature.h"
 #include "klog.h" // IWYU pragma: keep
 #include "ksu.h"
 #include "throne_tracker.h"
-#include "sucompat.h"
+#include "syscall_hook_manager.h"
 #include "ksud.h"
 #include "supercalls.h"
 
+#include "throne_comm.h"
+#include "dynamic_manager.h"
+
 static struct workqueue_struct *ksu_workqueue;
 
 bool ksu_queue_work(struct work_struct *work)
@@ -25,6 +28,17 @@ bool ksu_queue_work(struct work_struct *work)
     return queue_work(ksu_workqueue, work);
 }
 
+void sukisu_custom_config_init(void)
+{
+}
+
+void sukisu_custom_config_exit(void)
+{
+    ksu_uid_exit();
+    ksu_throne_comm_exit();
+    ksu_dynamic_manager_exit();
+}
+
 extern int ksu_handle_execveat_sucompat(int *fd, struct filename **filename_ptr,
 					void *argv, void *envp, int *flags);
 extern int ksu_handle_execveat_ksud(int *fd, struct filename **filename_ptr,
@@ -39,8 +53,10 @@ int ksu_handle_execveat(int *fd, struct filename **filename_ptr, void *argv,
 
 int __init kernelsu_init(void)
 {
+#ifndef DDK_ENV
 	pr_info("Initialized on: %s (%s) with driver version: %u\n",
 		UTS_RELEASE, UTS_MACHINE, KSU_VERSION);
+#endif
 
 #ifdef CONFIG_KSU_DEBUG
 	pr_alert("*************************************************************");
@@ -56,7 +72,9 @@ int __init kernelsu_init(void)
 
 	ksu_supercalls_init();
 
-	ksu_core_init();
+	sukisu_custom_config_init();
+
+	ksu_syscall_hook_manager_init();
 
 	ksu_workqueue = alloc_ordered_workqueue("kernelsu_work_queue", 0);
 
@@ -64,8 +82,6 @@ int __init kernelsu_init(void)
 
 	ksu_throne_tracker_init();
 
-	ksu_sucompat_init();
-
 	ksu_ksud_init();
 
 #ifdef MODULE
@@ -89,9 +105,11 @@ void kernelsu_exit(void)
 
 	ksu_ksud_exit();
 
-	ksu_sucompat_exit();
+	ksu_syscall_hook_manager_exit();
 
-	ksu_core_exit();
+	sukisu_custom_config_exit();
+
+	ksu_supercalls_exit();
 
 	ksu_feature_exit();
 }

kernel/ksu.h

@@ -13,12 +13,6 @@ extern bool ksu_uid_scanner_enabled;
 #define EVENT_BOOT_COMPLETED 2
 #define EVENT_MODULE_MOUNTED 3
 
-#define KSU_APP_PROFILE_VER 2
-#define KSU_MAX_PACKAGE_NAME 256
-// NGROUPS_MAX for Linux is 65535 generally, but we only supports 32 groups.
-#define KSU_MAX_GROUPS 32
-#define KSU_SELINUX_DOMAIN 64
-
 // SukiSU Ultra kernel su version full strings
 #ifndef KSU_VERSION_FULL 
 #define KSU_VERSION_FULL "v3.x-00000000@unknown"
@@ -47,53 +41,7 @@ struct manager_list_info {
     } managers[2];
 };
 
-struct root_profile {
+bool ksu_queue_work(struct work_struct *work);
-	int32_t uid;
-	int32_t gid;
-
-	int32_t groups_count;
-	int32_t groups[KSU_MAX_GROUPS];
-
-	// kernel_cap_t is u32[2] for capabilities v3
-	struct {
-		u64 effective;
-		u64 permitted;
-		u64 inheritable;
-	} capabilities;
-
-	char selinux_domain[KSU_SELINUX_DOMAIN];
-
-	int32_t namespaces;
-};
-
-struct non_root_profile {
-	bool umount_modules;
-};
-
-struct app_profile {
-	// It may be utilized for backward compatibility, although we have never explicitly made any promises regarding this.
-	u32 version;
-
-	// this is usually the package of the app, but can be other value for special apps
-	char key[KSU_MAX_PACKAGE_NAME];
-	int32_t current_uid;
-	bool allow_su;
-
-	union {
-		struct {
-			bool use_default;
-			char template_name[KSU_MAX_PACKAGE_NAME];
-
-			struct root_profile profile;
-		} rp_config;
-
-		struct {
-			bool use_default;
-
-			struct non_root_profile profile;
-		} nrp_config;
-	};
-};
 
 #if 0
 static inline int startswith(char *s, char *prefix)
@@ -111,6 +59,4 @@ static inline int endswith(const char *s, const char *t)
 }
 #endif
 
-bool ksu_queue_work(struct work_struct *work);
-
 #endif

kernel/ksud.c

@@ -21,16 +21,25 @@
 #include <linux/printk.h>
 #include <linux/types.h>
 #include <linux/uaccess.h>
+#include <linux/namei.h>
 #include <linux/workqueue.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
+#include <linux/sched/signal.h>
+#else
+#include <linux/sched.h>
+#endif
 
+#include "manager.h"
 #include "allowlist.h"
 #include "arch.h"
+#include "kernel_compat.h"
 #include "klog.h" // IWYU pragma: keep
 #include "ksud.h"
-#include "kernel_compat.h"
 #include "selinux/selinux.h"
-#include "manager.h"
+#include "throne_tracker.h"
-#include "sucompat.h"
+
+bool ksu_module_mounted __read_mostly = false;
+bool ksu_boot_completed __read_mostly = false;
 
 static const char KERNEL_SU_RC[] =
 	"\n"
@@ -59,7 +68,7 @@ static void stop_vfs_read_hook(void);
 static void stop_execve_hook(void);
 static void stop_input_hook(void);
 
-#ifdef KSU_KPROBES_HOOK
+#ifdef KSU_SHOULD_USE_NEW_TP
 static struct work_struct stop_vfs_read_work;
 static struct work_struct stop_execve_hook_work;
 static struct work_struct stop_input_hook_work;
@@ -78,13 +87,12 @@ void on_post_fs_data(void)
 {
 	static bool done = false;
 	if (done) {
-		pr_info("%s already done\n", __func__);
+		pr_info("on_post_fs_data already done\n");
 		return;
 	}
 	done = true;
-	pr_info("%s!\n", __func__);
+	pr_info("on_post_fs_data!\n");
 	ksu_load_allow_list();
-	ksu_mark_running_process();
 	ksu_observer_init();
 	// sanity check, this may influence the performance
 	stop_input_hook();
@@ -96,6 +104,43 @@ void on_post_fs_data(void)
 	pr_info("devpts sid: %d\n", ksu_file_sid);
 }
 
+extern void ext4_unregister_sysfs(struct super_block *sb);
+static void nuke_ext4_sysfs(void)
+{
+	struct path path;
+	int err = kern_path("/data/adb/modules", 0, &path);
+	if (err) {
+		pr_err("nuke path err: %d\n", err);
+		return;
+	}
+
+	struct super_block *sb = path.dentry->d_inode->i_sb;
+	const char *name = sb->s_type->name;
+	if (strcmp(name, "ext4") != 0) {
+		pr_info("nuke but module aren't mounted\n");
+		path_put(&path);
+		return;
+	}
+
+	ext4_unregister_sysfs(sb);
+	path_put(&path);
+}
+
+void on_module_mounted(void)
+{
+	pr_info("on_module_mounted!\n");
+	ksu_module_mounted = true;
+	nuke_ext4_sysfs();
+}
+
+void on_boot_completed(void)
+{
+	ksu_boot_completed = true;
+	pr_info("on_boot_completed!\n");
+	track_throne(true);
+}
+
+#define MAX_ARG_STRINGS 0x7FFFFFFF
 struct user_arg_ptr {
 #ifdef CONFIG_COMPAT
 	bool is_compat;
@@ -113,7 +158,9 @@ static void on_post_fs_data_cbfun(struct callback_head *cb)
 	on_post_fs_data();
 }
 
-static struct callback_head on_post_fs_data_cb = { .func = on_post_fs_data_cbfun };
+static struct callback_head on_post_fs_data_cb = {
+	.func = on_post_fs_data_cbfun
+};
 
 // since _ksud handler only uses argv and envp for comparisons
 // this can probably work
@@ -207,13 +254,11 @@ first_app_process:
 		rcu_read_lock();
 		init_task = rcu_dereference(current->real_parent);
 		if (init_task) {
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 7, 8)
+			ksu_task_work_add(init_task, &on_post_fs_data_cb,
-			task_work_add(init_task, &on_post_fs_data_cb, TWA_RESUME);
+				      TWA_RESUME);
-#else
-			task_work_add(init_task, &on_post_fs_data_cb, true);
-#endif
 		}
 		rcu_read_unlock();
+
 		stop_execve_hook();
 	}
 
@@ -315,7 +360,7 @@ static ssize_t read_iter_proxy(struct kiocb *iocb, struct iov_iter *to)
 int ksu_handle_vfs_read(struct file **file_ptr, char __user **buf_ptr,
 			       size_t *count_ptr, loff_t **pos)
 {
-#ifndef KSU_KPROBES_HOOK
+#ifndef KSU_SHOULD_USE_NEW_TP
 	if (!ksu_vfs_read_hook) {
 		return 0;
 	}
@@ -428,7 +473,7 @@ static bool is_volumedown_enough(unsigned int count)
 int ksu_handle_input_handle_event(unsigned int *type, unsigned int *code,
 				  int *value)
 {
-#ifndef KSU_KPROBES_HOOK
+#ifndef KSU_SHOULD_USE_NEW_TP
 	if (!ksu_input_hook) {
 		return 0;
 	}
@@ -470,7 +515,8 @@ bool ksu_is_safe_mode(void)
 	return false;
 }
 
-#ifdef KSU_KPROBES_HOOK
+#ifdef KSU_SHOULD_USE_NEW_TP
+
 static int sys_execve_handler_pre(struct kprobe *p, struct pt_regs *regs)
 {
 	/*
@@ -689,7 +735,7 @@ int __maybe_unused ksu_handle_compat_execve_ksud(
 
 static void stop_vfs_read_hook(void)
 {
-#ifdef KSU_KPROBES_HOOK
+#ifdef KSU_SHOULD_USE_NEW_TP
 	bool ret = schedule_work(&stop_vfs_read_work);
 	pr_info("unregister vfs_read kprobe: %d!\n", ret);
 #else
@@ -700,7 +746,7 @@ static void stop_vfs_read_hook(void)
 
 static void stop_execve_hook(void)
 {
-#ifdef KSU_KPROBES_HOOK
+#ifdef KSU_SHOULD_USE_NEW_TP
 	bool ret = schedule_work(&stop_execve_hook_work);
 	pr_info("unregister execve kprobe: %d!\n", ret);
 #else
@@ -711,7 +757,7 @@ static void stop_execve_hook(void)
 
 static void stop_input_hook(void)
 {
-#ifdef KSU_KPROBES_HOOK
+#ifdef KSU_SHOULD_USE_NEW_TP
 	static bool input_hook_stopped = false;
 	if (input_hook_stopped) {
 		return;
@@ -731,7 +777,7 @@ static void stop_input_hook(void)
 // ksud: module support
 void ksu_ksud_init(void)
 {
-#ifdef KSU_KPROBES_HOOK
+#ifdef KSU_SHOULD_USE_NEW_TP
 	int ret;
 
 	ret = register_kprobe(&execve_kp);
@@ -751,7 +797,7 @@ void ksu_ksud_init(void)
 
 void ksu_ksud_exit(void)
 {
-#ifdef KSU_KPROBES_HOOK
+#ifdef KSU_SHOULD_USE_NEW_TP
 	unregister_kprobe(&execve_kp);
 	// this should be done before unregister vfs_read_kp
 	// unregister_kprobe(&vfs_read_kp);

kernel/ksud.h

@@ -5,14 +5,18 @@
 
 #define KSUD_PATH "/data/adb/ksud"
 
-void ksu_ksud_init();
+void ksu_ksud_init(void);
-void ksu_ksud_exit();
+void ksu_ksud_exit(void);
 
 void on_post_fs_data(void);
+void on_module_mounted(void);
+void on_boot_completed(void);
 
 bool ksu_is_safe_mode(void);
 
 extern u32 ksu_file_sid;
+extern bool ksu_module_mounted;
+extern bool ksu_boot_completed;
 
 extern bool ksu_execveat_hook __read_mostly;
 extern int ksu_handle_pre_ksud(const char *filename);

kernel/lsm_hooks.c

@@ -0,0 +1,122 @@
+#include <linux/lsm_hooks.h>
+#include <linux/uidgid.h>
+#include <linux/version.h>
+#include <linux/binfmts.h>
+#include <linux/err.h>
+
+#include "klog.h" // IWYU pragma: keep
+#include "kernel_compat.h"
+#include "setuid_hook.h"
+
+#ifndef KSU_SHOULD_USE_NEW_TP
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 10, 0) ||                           \
+	defined(CONFIG_IS_HW_HISI) || defined(CONFIG_KSU_ALLOWLIST_WORKAROUND)
+static int ksu_key_permission(key_ref_t key_ref, const struct cred *cred,
+			      unsigned perm)
+{
+	if (init_session_keyring != NULL) {
+		return 0;
+	}
+	if (strcmp(current->comm, "init")) {
+		// we are only interested in `init` process
+		return 0;
+	}
+	init_session_keyring = cred->session_keyring;
+	pr_info("kernel_compat: got init_session_keyring\n");
+	return 0;
+}
+#endif
+
+static int ksu_task_fix_setuid(struct cred *new, const struct cred *old,
+			       int flags)
+{
+	kuid_t new_uid = new->uid;
+	kuid_t new_euid = new->euid;
+
+	return ksu_handle_setresuid((uid_t)new_uid.val, (uid_t)new_euid.val,
+				    (uid_t)new_uid.val);
+}
+
+#ifndef DEVPTS_SUPER_MAGIC
+#define DEVPTS_SUPER_MAGIC	0x1cd1
+#endif
+
+extern int __ksu_handle_devpts(struct inode *inode); // sucompat.c
+
+#ifdef CONFIG_COMPAT
+bool ksu_is_compat __read_mostly = false;
+#endif
+
+int ksu_inode_permission(struct inode *inode, int mask)
+{
+	if (inode && inode->i_sb 
+		&& unlikely(inode->i_sb->s_magic == DEVPTS_SUPER_MAGIC)) {
+		//pr_info("%s: handling devpts for: %s \n", __func__, current->comm);
+		__ksu_handle_devpts(inode);
+	}
+	return 0;
+}
+
+int ksu_bprm_check(struct linux_binprm *bprm)
+{
+	char *filename = (char *)bprm->filename;
+	
+	if (likely(!ksu_execveat_hook))
+		return 0;
+
+#ifdef CONFIG_COMPAT
+	static bool compat_check_done __read_mostly = false;
+	if ( unlikely(!compat_check_done) && unlikely(!strcmp(filename, "/data/adb/ksud"))
+		&& !memcmp(bprm->buf, "\x7f\x45\x4c\x46", 4) ) {
+		if (bprm->buf[4] == 0x01 )
+			ksu_is_compat = true;
+
+		pr_info("%s: %s ELF magic found! ksu_is_compat: %d \n", __func__, filename, ksu_is_compat);
+		compat_check_done = true;
+	}
+#endif
+
+	ksu_handle_pre_ksud(filename);
+
+	return 0;
+
+}
+
+static struct security_hook_list ksu_hooks[] = {
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 10, 0) ||                           \
+	defined(CONFIG_IS_HW_HISI) || defined(CONFIG_KSU_ALLOWLIST_WORKAROUND)
+	LSM_HOOK_INIT(key_permission, ksu_key_permission),
+#endif
+	LSM_HOOK_INIT(inode_permission, ksu_inode_permission),
+#ifndef KSU_SHOULD_USE_NEW_TP
+	LSM_HOOK_INIT(bprm_check_security, ksu_bprm_check),
+#endif
+	LSM_HOOK_INIT(task_fix_setuid, ksu_task_fix_setuid)
+};
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 8, 0)
+static const struct lsm_id ksu_lsmid = {
+	.name = "ksu",
+	.id = 912,
+};
+#endif
+
+void ksu_lsm_hook_init(void)
+{
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 8, 0)
+	security_add_hooks(ksu_hooks, ARRAY_SIZE(ksu_hooks), &ksu_lsmid);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
+	security_add_hooks(ksu_hooks, ARRAY_SIZE(ksu_hooks), "ksu");
+#else
+	// https://elixir.bootlin.com/linux/v4.10.17/source/include/linux/lsm_hooks.h#L1892
+	security_add_hooks(ksu_hooks, ARRAY_SIZE(ksu_hooks));
+#endif
+	pr_info("LSM hooks initialized.\n");
+}
+#else
+void ksu_lsm_hook_init(void)
+{
+	return;
+}
+#endif

kernel/pkg_observer.c

@@ -31,7 +31,7 @@ static KSU_DECL_FSNOTIFY_OPS(ksu_handle_generic_event)
 	if (ksu_fname_len(file_name) == 13 &&
 	    !memcmp(ksu_fname_arg(file_name), "packages.list", 13)) {
 		pr_info("packages.list detected (mask=%d)\n", mask);
-		track_throne();
+		track_throne(false);
 	}
 	return 0;
 }

kernel/seccomp_cache.c

@@ -0,0 +1,67 @@
+#include <linux/version.h>
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 10, 0)
+#include <linux/fs.h>
+#include <linux/sched/task.h>
+#include <linux/uaccess.h>
+#include <linux/filter.h>
+#include <linux/seccomp.h>
+#include "klog.h" // IWYU pragma: keep
+#include "seccomp_cache.h"
+
+struct action_cache {
+	DECLARE_BITMAP(allow_native, SECCOMP_ARCH_NATIVE_NR);
+#ifdef SECCOMP_ARCH_COMPAT
+	DECLARE_BITMAP(allow_compat, SECCOMP_ARCH_COMPAT_NR);
+#endif
+};
+
+struct seccomp_filter {
+	refcount_t refs;
+	refcount_t users;
+	bool log;
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 1, 0)
+	bool wait_killable_recv;
+#endif
+	struct action_cache cache;
+	struct seccomp_filter *prev;
+	struct bpf_prog *prog;
+	struct notification *notif;
+	struct mutex notify_lock;
+	wait_queue_head_t wqh;
+};
+
+void ksu_seccomp_clear_cache(struct seccomp_filter *filter, int nr)
+{
+	if (!filter) {
+		return;
+	}
+
+	if (nr >= 0 && nr < SECCOMP_ARCH_NATIVE_NR) {
+		clear_bit(nr, filter->cache.allow_native);
+	}
+
+#ifdef SECCOMP_ARCH_COMPAT
+	if (nr >= 0 && nr < SECCOMP_ARCH_COMPAT_NR) {
+		clear_bit(nr, filter->cache.allow_compat);
+	}
+#endif
+}
+
+void ksu_seccomp_allow_cache(struct seccomp_filter *filter, int nr)
+{
+	if (!filter) {
+		return;
+	}
+
+	if (nr >= 0 && nr < SECCOMP_ARCH_NATIVE_NR) {
+		set_bit(nr, filter->cache.allow_native);
+	}
+
+#ifdef SECCOMP_ARCH_COMPAT
+	if (nr >= 0 && nr < SECCOMP_ARCH_COMPAT_NR) {
+		set_bit(nr, filter->cache.allow_compat);
+	}
+#endif
+}
+#endif

kernel/seccomp_cache.h

@@ -0,0 +1,12 @@
+#ifndef __KSU_H_SECCOMP_CACHE
+#define __KSU_H_SECCOMP_CACHE
+
+#include <linux/fs.h>
+#include <linux/version.h>
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 10, 0)
+extern void ksu_seccomp_clear_cache(struct seccomp_filter *filter, int nr);
+extern void ksu_seccomp_allow_cache(struct seccomp_filter *filter, int nr);
+#endif
+
+#endif

kernel/selinux/rules.c

@@ -6,7 +6,7 @@
 #include "selinux.h"
 #include "sepolicy.h"
 #include "ss/services.h"
-#include "linux/lsm_audit.h"
+#include "linux/lsm_audit.h" // IWYU pragma: keep
 #include "xfrm.h"
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 10, 0)
@@ -37,6 +37,7 @@ static struct policydb *get_policydb(void)
 }
 
 static DEFINE_MUTEX(ksu_rules);
+
 void apply_kernelsu_rules(void)
 {
 	struct policydb *db;
@@ -95,7 +96,6 @@ void apply_kernelsu_rules(void)
 	ksu_allow(db, "init", "adb_data_file", "file", ALL);
 	ksu_allow(db, "init", "adb_data_file", "dir", ALL); // #1289
 	ksu_allow(db, "init", KERNEL_SU_DOMAIN, ALL, ALL);
-
 	// we need to umount modules in zygote
 	ksu_allow(db, "zygote", "adb_data_file", "dir", "search");
 
@@ -139,9 +139,6 @@ void apply_kernelsu_rules(void)
 	ksu_allow(db, "system_server", KERNEL_SU_DOMAIN, "process", "getpgid");
 	ksu_allow(db, "system_server", KERNEL_SU_DOMAIN, "process", "sigkill");
 
-	// https://android-review.googlesource.com/c/platform/system/logging/+/3725346
-	ksu_dontaudit(db, "untrusted_app", KERNEL_SU_DOMAIN, "dir", "getattr");
-
 	mutex_unlock(&ksu_rules);
 }
 
@@ -158,15 +155,15 @@ void apply_kernelsu_rules(void)
 #define CMD_GENFSCON 9
 
 struct sepol_data {
-	uint32_t cmd;
+	u32 cmd;
-	uint32_t subcmd;
+	u32 subcmd;
-	uint64_t sepol1;
+	u64 sepol1;
-	uint64_t sepol2;
+	u64 sepol2;
-	uint64_t sepol3;
+	u64 sepol3;
-	uint64_t sepol4;
+	u64 sepol4;
-	uint64_t sepol5;
+	u64 sepol5;
-	uint64_t sepol6;
+	u64 sepol6;
-	uint64_t sepol7;
+	u64 sepol7;
 };
 
 static int get_object(char *buf, char __user *user_object, size_t buf_sz,
@@ -185,14 +182,12 @@ static int get_object(char *buf, char __user *user_object, size_t buf_sz,
 
 	return 0;
 }
-
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 4, 0) ||                           \
 	!defined(KSU_COMPAT_USE_SELINUX_STATE)
 extern int avc_ss_reset(u32 seqno);
 #else
 extern int avc_ss_reset(struct selinux_avc *avc, u32 seqno);
 #endif
-
 // reset avc cache table, otherwise the new rules will not take effect if already denied
 static void reset_avc_cache(void)
 {
@@ -222,7 +217,7 @@ int handle_sepolicy(unsigned long arg3, void __user *arg4)
 		pr_info("SELinux permissive or disabled when handle policy!\n");
 	}
 
-	struct sepol_data data = { 0 };
+	struct sepol_data data;
 	if (copy_from_user(&data, arg4, sizeof(struct sepol_data))) {
 		pr_err("sepol: copy sepol_data failed.\n");
 		return -EINVAL;
@@ -236,7 +231,6 @@ int handle_sepolicy(unsigned long arg3, void __user *arg4)
 	db = get_policydb();
 
 	int ret = -EINVAL;
-
 	switch (cmd) {
 	case CMD_NORMAL_PERM: {
 		char src_buf[MAX_SEPOL_LEN];

kernel/selinux/selinux.c

@@ -59,6 +59,7 @@ is_ksu_transition(const struct task_security_struct *old_tsec,
 }
 #endif
 
+
 void setup_selinux(const char *domain)
 {
 	if (transive_to_domain(domain)) {
@@ -106,7 +107,7 @@ static int __security_secid_to_secctx(u32 secid, struct lsm_context *cp)
 }
 static void __security_release_secctx(struct lsm_context *cp)
 {
-	return security_release_secctx(cp->context, cp->len);
+	security_release_secctx(cp->context, cp->len);
 }
 #else
 #define __security_secid_to_secctx security_secid_to_secctx
@@ -139,7 +140,7 @@ bool is_ksu_domain(void)
 	return is_task_ksu_domain(current_cred());
 }
 
-bool is_zygote(const struct cred *cred)
+bool is_context(const struct cred *cred, const char *context)
 {
 	if (!cred) {
 		return false;
@@ -154,11 +155,21 @@ bool is_zygote(const struct cred *cred)
 	if (err) {
 		return false;
 	}
-	result = strncmp("u:r:zygote:s0", ctx.context, ctx.len) == 0;
+	result = strncmp(context, ctx.context, ctx.len) == 0;
 	__security_release_secctx(&ctx);
 	return result;
 }
 
+bool is_zygote(const struct cred *cred)
+{
+	return is_context(cred, "u:r:zygote:s0");
+}
+
+bool is_init(const struct cred *cred)
+{
+	return is_context(cred, "u:r:init:s0");
+}
+
 #define KSU_FILE_DOMAIN "u:object_r:ksu_file:s0"
 
 u32 ksu_get_ksu_file_sid(void)

kernel/selinux/selinux.h

@@ -3,7 +3,7 @@
 
 #include "linux/types.h"
 #include "linux/version.h"
-#include "linux/sched.h"
+#include "linux/cred.h"
 
 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(5, 10, 0)) ||                        \
 	defined(KSU_COMPAT_HAS_SELINUX_STATE)
@@ -14,14 +14,16 @@ void setup_selinux(const char *);
 
 void setenforce(bool);
 
-bool is_task_ksu_domain(const struct cred *cred);
-
 bool getenforce(void);
 
+bool is_task_ksu_domain(const struct cred *cred);
+
 bool is_ksu_domain(void);
 
 bool is_zygote(const struct cred *cred);
 
+bool is_init(const struct cred *cred);
+
 void apply_kernelsu_rules(void);
 
 u32 ksu_get_ksu_file_sid(void);

kernel/selinux/sepolicy.c

@@ -355,7 +355,7 @@ static void add_xperm_rule_raw(struct policydb *db, struct type_datum *src,
 
 		if (datum->u.xperms == NULL) {
 			datum->u.xperms =
-				(struct avtab_extended_perms *)(kmalloc(
+				(struct avtab_extended_perms *)(kzalloc(
 					sizeof(xperms), GFP_KERNEL));
 			if (!datum->u.xperms) {
 				pr_err("alloc xperms failed\n");
@@ -555,7 +555,7 @@ static bool add_filename_trans(struct policydb *db, const char *s,
 		trans = (struct filename_trans_datum *)kcalloc(sizeof(*trans),
 							       1, GFP_ATOMIC);
 		struct filename_trans_key *new_key =
-			(struct filename_trans_key *)kmalloc(sizeof(*new_key),
+			(struct filename_trans_key *)kzalloc(sizeof(*new_key),
 							     GFP_ATOMIC);
 		*new_key = key;
 		new_key->name = kstrdup(key.name, GFP_ATOMIC);
@@ -585,7 +585,7 @@ static bool add_filename_trans(struct policydb *db, const char *s,
 			return false;
 		}
 		struct filename_trans *new_key =
-			(struct filename_trans *)kmalloc(sizeof(*new_key),
+			(struct filename_trans *)kzalloc(sizeof(*new_key),
 							 GFP_ATOMIC);
 		if (!new_key) {
 			pr_err("add_filename_trans: Failed to alloc new_key\n");

kernel/setuid_hook.c

@@ -0,0 +1,186 @@
+#include <linux/compiler.h>
+#include <linux/version.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 10, 0)
+#include <linux/sched/signal.h>
+#endif
+#include <linux/slab.h>
+#include <linux/task_work.h>
+#include <linux/thread_info.h>
+#include <linux/seccomp.h>
+#include <linux/capability.h>
+#include <linux/cred.h>
+#include <linux/dcache.h>
+#include <linux/err.h>
+#include <linux/fs.h>
+#include <linux/init.h>
+#include <linux/init_task.h>
+#include <linux/kernel.h>
+#include <linux/kprobes.h>
+#include <linux/mm.h>
+#include <linux/mount.h>
+#include <linux/namei.h>
+#include <linux/nsproxy.h>
+#include <linux/path.h>
+#include <linux/printk.h>
+#include <linux/sched.h>
+#include <linux/stddef.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/uaccess.h>
+#include <linux/uidgid.h>
+
+#include "allowlist.h"
+#include "setuid_hook.h"
+#include "feature.h"
+#include "klog.h" // IWYU pragma: keep
+#include "manager.h"
+#include "selinux/selinux.h"
+#include "seccomp_cache.h"
+#include "supercalls.h"
+#include "syscall_hook_manager.h"
+#include "kernel_umount.h"
+
+static bool ksu_enhanced_security_enabled = false;
+
+static int enhanced_security_feature_get(u64 *value)
+{
+	*value = ksu_enhanced_security_enabled ? 1 : 0;
+	return 0;
+}
+
+static int enhanced_security_feature_set(u64 value)
+{
+	bool enable = value != 0;
+	ksu_enhanced_security_enabled = enable;
+	pr_info("enhanced_security: set to %d\n", enable);
+	return 0;
+}
+
+static const struct ksu_feature_handler enhanced_security_handler = {
+	.feature_id = KSU_FEATURE_ENHANCED_SECURITY,
+	.name = "enhanced_security",
+	.get_handler = enhanced_security_feature_get,
+	.set_handler = enhanced_security_feature_set,
+};
+
+static inline bool is_allow_su(void)
+{
+	if (is_manager()) {
+		// we are manager, allow!
+		return true;
+	}
+	return ksu_is_allow_uid_for_current(current_uid().val);
+}
+
+// force_sig kcompat, TODO: move it out of core_hook.c
+// https://elixir.bootlin.com/linux/v5.3-rc1/source/kernel/signal.c#L1613
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 3, 0)
+#define __force_sig(sig) force_sig(sig)
+#else
+#define __force_sig(sig) force_sig(sig, current)
+#endif
+
+extern void disable_seccomp(struct task_struct *tsk);
+int ksu_handle_setresuid(uid_t ruid, uid_t euid, uid_t suid)
+{
+	// we rely on the fact that zygote always call setresuid(3) with same uids
+	uid_t new_uid = ruid;
+	uid_t old_uid = current_uid().val;
+
+	if (old_uid != new_uid)
+		pr_info("handle_setresuid from %d to %d\n", old_uid, new_uid);
+
+	// if old process is root, ignore it.
+	if (old_uid != 0 && ksu_enhanced_security_enabled) {
+		// disallow any non-ksu domain escalation from non-root to root!
+		// euid is what we care about here as it controls permission
+		if (unlikely(euid == 0)) {
+			if (!is_ksu_domain()) {
+				pr_warn("find suspicious EoP: %d %s, from %d to %d\n",
+					current->pid, current->comm, old_uid,
+					new_uid);
+				__force_sig(SIGKILL);
+				return 0;
+			}
+		}
+		// disallow appuid decrease to any other uid if it is not allowed to su
+		if (is_appuid(old_uid)) {
+			if (euid < current_euid().val &&
+			    !ksu_is_allow_uid_for_current(old_uid)) {
+				pr_warn("find suspicious EoP: %d %s, from %d to %d\n",
+					current->pid, current->comm, old_uid,
+					new_uid);
+				__force_sig(SIGKILL);
+				return 0;
+			}
+		}
+		return 0;
+	}
+
+	// if on private space, see if its possibly the manager
+	if (new_uid > PER_USER_RANGE &&
+	    new_uid % PER_USER_RANGE == ksu_get_manager_uid()) {
+		ksu_set_manager_uid(new_uid);
+	}
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 10, 0)
+	if (ksu_get_manager_uid() == new_uid) {
+		pr_info("install fd for ksu manager(uid=%d)\n", new_uid);
+		ksu_install_fd();
+		spin_lock_irq(&current->sighand->siglock);
+		ksu_seccomp_allow_cache(current->seccomp.filter, __NR_reboot);
+		ksu_set_task_tracepoint_flag(current);
+		spin_unlock_irq(&current->sighand->siglock);
+		return 0;
+	}
+
+	if (ksu_is_allow_uid_for_current(new_uid)) {
+		if (current->seccomp.mode == SECCOMP_MODE_FILTER &&
+		    current->seccomp.filter) {
+			spin_lock_irq(&current->sighand->siglock);
+			ksu_seccomp_allow_cache(current->seccomp.filter,
+						__NR_reboot);
+			spin_unlock_irq(&current->sighand->siglock);
+		}
+		ksu_set_task_tracepoint_flag(current);
+	} else {
+		ksu_clear_task_tracepoint_flag_if_needed(current);
+	}
+#else
+	if (ksu_is_allow_uid_for_current(new_uid)) {
+		spin_lock_irq(&current->sighand->siglock);
+		disable_seccomp(current);
+		spin_unlock_irq(&current->sighand->siglock);
+
+		if (ksu_get_manager_uid() == new_uid) {
+			pr_info("install fd for ksu manager(uid=%d)\n",
+				new_uid);
+			ksu_install_fd();
+		}
+
+		return 0;
+	}
+#endif
+
+	// Handle kernel umount
+	ksu_handle_umount(old_uid, new_uid);
+
+	return 0;
+}
+
+extern void ksu_lsm_hook_init(void);
+void ksu_setuid_hook_init(void)
+{
+	ksu_kernel_umount_init();
+	ksu_lsm_hook_init(); // <4.11
+	if (ksu_register_feature_handler(&enhanced_security_handler)) {
+		pr_err("Failed to register enhanced security feature handler\n");
+	}
+}
+
+void ksu_setuid_hook_exit(void)
+{
+	pr_info("ksu_core_exit\n");
+	ksu_kernel_umount_exit();
+	ksu_unregister_feature_handler(KSU_FEATURE_ENHANCED_SECURITY);
+}

kernel/setuid_hook.h

@@ -0,0 +1,13 @@
+#ifndef __KSU_H_KSU_SETUID_HOOK
+#define __KSU_H_KSU_SETUID_HOOK
+
+#include <linux/init.h>
+#include <linux/types.h>
+
+void ksu_setuid_hook_init(void);
+void ksu_setuid_hook_exit(void);
+
+// Handler functions for hook_manager
+int ksu_handle_setresuid(uid_t ruid, uid_t euid, uid_t suid);
+
+#endif

kernel/sucompat.c

@@ -1,109 +1,35 @@
-#include <linux/dcache.h>
-#include <linux/security.h>
 #include <asm/current.h>
 #include <linux/cred.h>
 #include <linux/err.h>
 #include <linux/fs.h>
-#include <linux/kprobes.h>
 #include <linux/types.h>
 #include <linux/uaccess.h>
 #include <linux/version.h>
 #include <linux/ptrace.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 14, 0)
+#include <linux/compiler.h>
+#endif
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
 #include <linux/sched/task_stack.h>
 #else
 #include <linux/sched.h>
 #endif
 
-#include "objsec.h"
 #include "allowlist.h"
-#include "arch.h"
 #include "feature.h"
 #include "klog.h" // IWYU pragma: keep
 #include "ksud.h"
-#include "kernel_compat.h"
 #include "sucompat.h"
-#include "core_hook.h"
+#include "app_profile.h"
+#include "syscall_hook_manager.h"
 
 #define SU_PATH "/system/bin/su"
 #define SH_PATH "/system/bin/sh"
 
-static const char su[] = SU_PATH;
-static const char ksud_path[] = KSUD_PATH;
 bool ksu_su_compat_enabled __read_mostly = true;
 
-#ifdef KSU_SHOULD_USE_NEW_TP
+static const char su[] = SU_PATH;
-
+static const char ksud_path[] = KSUD_PATH;
-#include "linux/compiler.h"
-#include "linux/printk.h"
-#include "selinux/selinux.h"
-
-#include <linux/tracepoint.h>
-#include <linux/spinlock.h>
-#include <asm/syscall.h>
-#include <trace/events/syscalls.h>
-
-void ksu_mark_running_process(void)
-{
-	struct task_struct *p, *t;
-	read_lock(&tasklist_lock);
-	for_each_process_thread (p, t) {
-		if (!t->mm) { // only user processes
-			continue;
-		}
-		int uid = task_uid(t).val;
-		bool ksu_root_process =
-			uid == 0 && is_task_ksu_domain(get_task_cred(t));
-		if (ksu_root_process || ksu_is_allow_uid(uid)) {
-			ksu_set_task_tracepoint_flag(t);
-			pr_info("sucompat: mark process: pid:%d, uid: %d, comm:%s\n",
-				t->pid, uid, t->comm);
-		}
-	}
-	read_unlock(&tasklist_lock);
-}
-
-static void handle_process_mark(bool mark)
-{
-	struct task_struct *p, *t;
-	read_lock(&tasklist_lock);
-	for_each_process_thread (p, t) {
-		if (mark)
-			ksu_set_task_tracepoint_flag(t);
-		else
-			ksu_clear_task_tracepoint_flag(t);
-	}
-	read_unlock(&tasklist_lock);
-}
-
-static void mark_all_process(void)
-{
-	handle_process_mark(true);
-	pr_info("sucompat: mark all user process done!\n");
-}
-
-static void unmark_all_process(void)
-{
-	handle_process_mark(false);
-	pr_info("sucompat: unmark all user process done!\n");
-}
-#else
-void ksu_mark_running_process(void)
-{
-}
-
-static void handle_process_mark(bool mark)
-{
-}
-
-static void mark_all_process(void)
-{
-}
-
-static void unmark_all_process(void)
-{
-}
-#endif
 
 static int su_compat_feature_get(u64 *value)
 {
@@ -114,21 +40,8 @@ static int su_compat_feature_get(u64 *value)
 static int su_compat_feature_set(u64 value)
 {
 	bool enable = value != 0;
-
-	if (enable == ksu_su_compat_enabled) {
-		pr_info("su_compat: no need to change\n");
-		return 0;
-	}
-
-	if (enable) {
-		ksu_sucompat_enable();
-	} else {
-		ksu_sucompat_disable();
-	}
-
 	ksu_su_compat_enabled = enable;
 	pr_info("su_compat: set to %d\n", enable);
-
 	return 0;
 }
 
@@ -139,29 +52,30 @@ static const struct ksu_feature_handler su_compat_handler = {
 	.set_handler = su_compat_feature_set,
 };
 
-static inline void __user *userspace_stack_buffer(const void *d, size_t len)
+static void __user *userspace_stack_buffer(const void *d, size_t len)
 {
-	/* To avoid having to mmap a page in userspace, just write below the stack
+	// To avoid having to mmap a page in userspace, just write below the stack
-   * pointer. */
+	// pointer.
 	char __user *p = (void __user *)current_user_stack_pointer() - len;
 
 	return copy_to_user(p, d, len) ? NULL : p;
 }
 
-static inline char __user *sh_user_path(void)
+static char __user *sh_user_path(void)
 {
-	const char sh_path[] = SH_PATH;
+	static const char sh_path[] = "/system/bin/sh";
+
 	return userspace_stack_buffer(sh_path, sizeof(sh_path));
 }
 
-static inline char __user *ksud_user_path(void)
+static char __user *ksud_user_path(void)
 {
 	return userspace_stack_buffer(ksud_path, sizeof(ksud_path));
 }
 
 static inline bool __is_su_allowed(const void *ptr_to_check)
 {
-#ifdef KSU_KPROBES_HOOK
+#ifndef KSU_SHOULD_USE_NEW_TP
 	if (!ksu_su_compat_enabled)
 		return false;
 #endif
@@ -182,7 +96,7 @@ static int ksu_sucompat_user_common(const char __user **filename_user,
 	char path[sizeof(su)]; // sizeof includes nullterm already!
 	memset(path, 0, sizeof(path));
 
-	ksu_strncpy_from_user_retry(path, *filename_user, sizeof(path));
+	ksu_strncpy_from_user_nofault(path, *filename_user, sizeof(path));
 
 	if (memcmp(path, su, sizeof(su)))
 		return 0;
@@ -190,7 +104,7 @@ static int ksu_sucompat_user_common(const char __user **filename_user,
 	if (escalate) {
 		pr_info("%s su found\n", syscall_name);
 		*filename_user = ksud_user_path();
-		escape_to_root(); // escalate !!
+		escape_with_root_profile(); // escalate !!
 	} else {
 		pr_info("%s su->sh!\n", syscall_name);
 		*filename_user = sh_user_path();
@@ -245,14 +159,14 @@ int ksu_handle_execveat_sucompat(int *fd, struct filename **filename_ptr,
 	pr_info("do_execveat_common su found\n");
 	memcpy((void *)filename->name, ksud_path, sizeof(ksud_path));
 
-	escape_to_root();
+	escape_with_root_profile();
 
 	return 0;
 }
 
 int __ksu_handle_devpts(struct inode *inode)
 {
-#ifndef KSU_KPROBES_HOOK
+#ifndef KSU_SHOULD_USE_NEW_TP
 	if (!ksu_su_compat_enabled)
 		return 0;
 #endif
@@ -281,193 +195,15 @@ int __ksu_handle_devpts(struct inode *inode)
 	return 0;
 }
 
-#ifdef KSU_SHOULD_USE_NEW_TP
+// sucompat: permitted process can execute 'su' to gain root access.
-#ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS
-
-// Tracepoint probe for sys_enter
-static void sucompat_sys_enter_handler(void *data, struct pt_regs *regs,
-				       long id)
-{
-	// Handle newfstatat
-	if (unlikely(id == __NR_newfstatat)) {
-		int *dfd = (int *)&PT_REGS_PARM1(regs);
-		const char __user **filename_user =
-			(const char __user **)&PT_REGS_PARM2(regs);
-		int *flags = (int *)&PT_REGS_SYSCALL_PARM4(regs);
-		ksu_handle_stat(dfd, filename_user, flags);
-		return;
-	}
-
-	// Handle faccessat
-	if (unlikely(id == __NR_faccessat)) {
-		int *dfd = (int *)&PT_REGS_PARM1(regs);
-		const char __user **filename_user =
-			(const char __user **)&PT_REGS_PARM2(regs);
-		int *mode = (int *)&PT_REGS_PARM3(regs);
-		ksu_handle_faccessat(dfd, filename_user, mode, NULL);
-		return;
-	}
-
-	// Handle execve
-	if (unlikely(id == __NR_execve)) {
-		const char __user **filename_user =
-			(const char __user **)&PT_REGS_PARM1(regs);
-		ksu_handle_execve_sucompat(AT_FDCWD, filename_user, NULL, NULL,
-					   NULL);
-		return;
-	}
-}
-
-#endif // CONFIG_HAVE_SYSCALL_TRACEPOINTS
-
-#ifdef CONFIG_KRETPROBES
-
-static struct kretprobe *init_kretprobe(const char *name,
-					kretprobe_handler_t handler)
-{
-	struct kretprobe *rp = kzalloc(sizeof(struct kretprobe), GFP_KERNEL);
-	if (!rp)
-		return NULL;
-	rp->kp.symbol_name = name;
-	rp->handler = handler;
-	rp->data_size = 0;
-	rp->maxactive = 0;
-
-	int ret = register_kretprobe(rp);
-	pr_info("sucompat: register_%s kretprobe: %d\n", name, ret);
-	if (ret) {
-		kfree(rp);
-		return NULL;
-	}
-
-	return rp;
-}
-
-static void destroy_kretprobe(struct kretprobe **rp_ptr)
-{
-	struct kretprobe *rp = *rp_ptr;
-	if (!rp)
-		return;
-	unregister_kretprobe(rp);
-	synchronize_rcu();
-	kfree(rp);
-	*rp_ptr = NULL;
-}
-
-static int tracepoint_reg_count = 0;
-static DEFINE_SPINLOCK(tracepoint_reg_lock);
-
-static int syscall_regfunc_handler(struct kretprobe_instance *ri,
-				   struct pt_regs *regs)
-{
-	unsigned long flags;
-	spin_lock_irqsave(&tracepoint_reg_lock, flags);
-	if (tracepoint_reg_count < 1) {
-		// while install our tracepoint, mark our processes
-		unmark_all_process();
-		ksu_mark_running_process();
-	} else {
-		// while installing other tracepoint, mark all processes
-		mark_all_process();
-	}
-	tracepoint_reg_count++;
-	spin_unlock_irqrestore(&tracepoint_reg_lock, flags);
-	return 0;
-}
-
-static int syscall_unregfunc_handler(struct kretprobe_instance *ri,
-				     struct pt_regs *regs)
-{
-	unsigned long flags;
-	spin_lock_irqsave(&tracepoint_reg_lock, flags);
-	if (tracepoint_reg_count <= 1) {
-		// while uninstall our tracepoint, unmark all processes
-		unmark_all_process();
-	} else {
-		// while uninstalling other tracepoint, mark our processes
-		unmark_all_process();
-		ksu_mark_running_process();
-	}
-	tracepoint_reg_count--;
-	spin_unlock_irqrestore(&tracepoint_reg_lock, flags);
-	return 0;
-}
-
-static struct kretprobe *syscall_regfunc_rp = NULL;
-static struct kretprobe *syscall_unregfunc_rp = NULL;
-#endif
-#endif
-
-void ksu_sucompat_enable(void)
-{
-#ifdef KSU_SHOULD_USE_NEW_TP
-	int ret;
-	pr_info("sucompat: ksu_sucompat_enable called\n");
-
-#ifdef CONFIG_KRETPROBES
-	// Register kretprobe for syscall_regfunc
-	syscall_regfunc_rp =
-		init_kretprobe("syscall_regfunc", syscall_regfunc_handler);
-	// Register kretprobe for syscall_unregfunc
-	syscall_unregfunc_rp =
-		init_kretprobe("syscall_unregfunc", syscall_unregfunc_handler);
-#endif
-
-#ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS
-	ret = register_trace_sys_enter(sucompat_sys_enter_handler, NULL);
-#ifndef CONFIG_KRETPROBES
-	unmark_all_process();
-	ksu_mark_running_process();
-#endif
-	if (ret) {
-		pr_err("sucompat: failed to register sys_enter tracepoint: %d\n",
-		       ret);
-	} else {
-		pr_info("sucompat: sys_enter tracepoint registered\n");
-	}
-#endif
-#else
-	ksu_su_compat_enabled = true;
-	pr_info("init sucompat\n");
-#endif
-}
-
-void ksu_sucompat_disable(void)
-{
-#ifdef KSU_SHOULD_USE_NEW_TP
-	pr_info("sucompat: ksu_sucompat_disable called\n");
-#ifdef CONFIG_HAVE_SYSCALL_TRACEPOINTS
-	unregister_trace_sys_enter(sucompat_sys_enter_handler, NULL);
-	tracepoint_synchronize_unregister();
-	pr_info("sucompat: sys_enter tracepoint unregistered\n");
-#endif
-
-#ifdef CONFIG_KRETPROBES
-	destroy_kretprobe(&syscall_regfunc_rp);
-	destroy_kretprobe(&syscall_unregfunc_rp);
-#endif
-#else
-	ksu_su_compat_enabled = false;
-	pr_info("deinit sucompat\n");
-#endif
-}
-
-// sucompat: permited process can execute 'su' to gain root access.
 void ksu_sucompat_init(void)
 {
 	if (ksu_register_feature_handler(&su_compat_handler)) {
 		pr_err("Failed to register su_compat feature handler\n");
 	}
-
-	if (ksu_su_compat_enabled) {
-		ksu_sucompat_enable();
-	}
 }
 
 void ksu_sucompat_exit(void)
 {
-	if (ksu_su_compat_enabled) {
-		ksu_sucompat_disable();
-	}
 	ksu_unregister_feature_handler(KSU_FEATURE_SU_COMPAT);
 }

kernel/sucompat.h

@@ -1,49 +1,18 @@
 #ifndef __KSU_H_SUCOMPAT
 #define __KSU_H_SUCOMPAT
-#include <linux/sched.h>
+#include <linux/types.h>
-#include <linux/version.h>
 
 extern bool ksu_su_compat_enabled;
 
 void ksu_sucompat_init(void);
 void ksu_sucompat_exit(void);
 
-void ksu_sucompat_enable(void);
+// Handler functions exported for hook_manager
-void ksu_sucompat_disable(void);
+int ksu_handle_faccessat(int *dfd, const char __user **filename_user, int *mode,
-
+			 int *__unused_flags);
-void ksu_mark_running_process(void);
+int ksu_handle_stat(int *dfd, const char __user **filename_user, int *flags);
-
+int ksu_handle_execve_sucompat(int *fd, const char __user **filename_user,
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 10, 0) && defined(KSU_KPROBE_HOOK)
+			       void *__never_use_argv, void *__never_use_envp,
-
+			       int *__never_use_flags);
-#include <linux/thread_info.h>
-
-static inline void ksu_set_task_tracepoint_flag(struct task_struct *t)
-{
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 11, 0)
-	set_task_syscall_work(t, SYSCALL_TRACEPOINT);
-#else
-	set_tsk_thread_flag(t, TIF_SYSCALL_TRACEPOINT);
-#endif
-}
-
-static inline void ksu_clear_task_tracepoint_flag(struct task_struct *t)
-{
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 11, 0)
-	clear_task_syscall_work(t, SYSCALL_TRACEPOINT);
-#else
-	clear_tsk_thread_flag(t, TIF_SYSCALL_TRACEPOINT);
-#endif
-}
-#else
-static inline void ksu_set_task_tracepoint_flag(struct task_struct *t)
-{
-	return;
-}
-
-static inline void ksu_clear_task_tracepoint_flag(struct task_struct *t)
-{
-	return;
-}
-#endif
 
 #endif

kernel/supercalls.c

@@ -4,23 +4,26 @@
 #include <linux/capability.h>
 #include <linux/cred.h>
 #include <linux/err.h>
+#include <linux/fdtable.h>
 #include <linux/file.h>
 #include <linux/fs.h>
 #include <linux/slab.h>
+#include <linux/kprobes.h>
+#include <linux/syscalls.h>
+#include <linux/task_work.h>
 #include <linux/uaccess.h>
 #include <linux/version.h>
 
+#include "arch.h"
 #include "allowlist.h"
 #include "feature.h"
 #include "klog.h" // IWYU pragma: keep
-#include "ksu.h"
 #include "ksud.h"
 #include "manager.h"
 #include "selinux/selinux.h"
-#include "core_hook.h"
 #include "objsec.h"
 #include "file_wrapper.h"
-#include "kernel_compat.h"
+#include "syscall_hook_manager.h"
 #include "throne_comm.h"
 #include "dynamic_manager.h"
 
@@ -50,8 +53,7 @@ bool always_allow(void)
 
 bool allowed_for_su(void)
 {
-    bool is_allowed =
+	bool is_allowed = is_manager() || ksu_is_allow_uid_for_current(current_uid().val);
-		is_manager() || ksu_is_allow_uid_for_current(current_uid().val);
 	return is_allowed;
 }
 
@@ -73,7 +75,7 @@ static int do_grant_root(void __user *arg)
 	// we already check uid above on allowed_for_su()
 
 	pr_info("allow root for: %d\n", current_uid().val);
-    escape_to_root();
+	escape_with_root_profile();
 
 	return 0;
 }
@@ -123,13 +125,13 @@ static int do_report_event(void __user *arg)
 		if (!boot_complete_lock) {
 			boot_complete_lock = true;
 			pr_info("boot_complete triggered\n");
+			on_boot_completed();
 		}
 		break;
 	}
 	case EVENT_MODULE_MOUNTED: {
-        ksu_module_mounted = true;
 		pr_info("module mounted!\n");
-        nuke_ext4_sysfs();
+		on_module_mounted();
 		break;
 	}
 	default:
@@ -345,13 +347,11 @@ static int do_set_feature(void __user *arg)
 	return 0;
 }
 
-static int do_get_wrapper_fd(void __user *arg)
+static int do_get_wrapper_fd(void __user *arg) {
-{
 	if (!ksu_file_sid) {
 		return -EINVAL;
 	}
 
-	const char *anon_name = "[ksu_fdwrapper]";
 	struct ksu_get_wrapper_fd_cmd cmd;
 	int ret;
 
@@ -373,37 +373,26 @@ static int do_get_wrapper_fd(void __user *arg)
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 12, 0)
 #define getfd_secure anon_inode_create_getfd
-#elif defined(KSU_HAS_GETFD_SECURE)
+#else
 #define getfd_secure anon_inode_getfd_secure
 #endif
-
+	ret = getfd_secure("[ksu_fdwrapper]", &data->ops, data, f->f_flags, NULL);
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(6, 12, 0) || defined(KSU_HAS_GETFD_SECURE)
-	ret = getfd_secure(anon_name, &data->ops, data, f->f_flags, NULL);
-#else
-	ret = anon_inode_getfd(anon_name, &data->ops, data, f->f_flags);
-#endif
-
 	if (ret < 0) {
 		pr_err("ksu_fdwrapper: getfd failed: %d\n", ret);
 		goto put_wrapper_data;
 	}
-
 	struct file* pf = fget(ret);
+
 	struct inode* wrapper_inode = file_inode(pf);
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 1, 0) ||                           \
+	// copy original inode mode
-	defined(KSU_OPTIONAL_SELINUX_INODE)
+	wrapper_inode->i_mode = file_inode(f)->i_mode;
 	struct inode_security_struct *sec = selinux_inode(wrapper_inode);
-#else
-	struct inode_security_struct *sec =
-		(struct inode_security_struct *)wrapper_inode->i_security;
-#endif
 	if (sec) {
 		sec->sid = ksu_file_sid;
 	}
 
 	fput(pf);
 	goto put_orig_file;
-
 put_wrapper_data:
 	ksu_delete_file_wrapper(data);
 put_orig_file:
@@ -412,6 +401,71 @@ put_orig_file:
 	return ret;
 }
 
+static int do_manage_mark(void __user *arg)
+{
+	struct ksu_manage_mark_cmd cmd;
+	int ret = 0;
+
+	if (copy_from_user(&cmd, arg, sizeof(cmd))) {
+		pr_err("manage_mark: copy_from_user failed\n");
+		return -EFAULT;
+	}
+
+	switch (cmd.operation) {
+	case KSU_MARK_GET: {
+		// Get task mark status
+		ret = ksu_get_task_mark(cmd.pid);
+		if (ret < 0) {
+			pr_err("manage_mark: get failed for pid %d: %d\n", cmd.pid, ret);
+			return ret;
+		}
+		cmd.result = (u32)ret;
+		break;
+	}
+	case KSU_MARK_MARK: {
+		if (cmd.pid == 0) {
+			ksu_mark_all_process();
+		} else {
+			ret = ksu_set_task_mark(cmd.pid, true);
+			if (ret < 0) {
+				pr_err("manage_mark: set_mark failed for pid %d: %d\n", cmd.pid,
+					   ret);
+				return ret;
+			}
+		}
+		break;
+	}
+	case KSU_MARK_UNMARK: {
+		if (cmd.pid == 0) {
+			ksu_unmark_all_process();
+		} else {
+			ret = ksu_set_task_mark(cmd.pid, false);
+			if (ret < 0) {
+				pr_err("manage_mark: set_unmark failed for pid %d: %d\n",
+					   cmd.pid, ret);
+				return ret;
+			}
+		}
+		break;
+	}
+	case KSU_MARK_REFRESH: {
+		ksu_mark_running_process();
+		pr_info("manage_mark: refreshed running processes\n");
+		break;
+	}
+	default: {
+		pr_err("manage_mark: invalid operation %u\n", cmd.operation);
+		return -EINVAL;
+	}
+	}
+	if (copy_to_user(arg, &cmd, sizeof(cmd))) {
+		pr_err("manage_mark: copy_to_user failed\n");
+		return -EFAULT;
+	}
+
+	return 0;
+}
+
 // 100. GET_FULL_VERSION - Get full version string
 static int do_get_full_version(void __user *arg)
 {
@@ -435,13 +489,12 @@ static int do_get_full_version(void __user *arg)
 static int do_get_hook_type(void __user *arg)
 {
 	struct ksu_hook_type_cmd cmd = {0};
-    const char *type = "Unknown";
+	const char *type = "Tracepoint";
 
-#if defined(KSU_HAVE_SYSCALL_TRACEPOINTS_HOOK)
+#if defined(CONFIG_KSU_MANUAL_HOOK)
-    type = "Tracepoint";
-#elif defined(KSU_MANUAL_HOOK)
 	type = "Manual";
 #endif
+
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 13, 0)
 	strscpy(cmd.hook_type, type, sizeof(cmd.hook_type));
 #else
@@ -587,7 +640,8 @@ static const struct ksu_ioctl_cmd_map ksu_ioctl_handlers[] = {
 	{ .cmd = KSU_IOCTL_SET_APP_PROFILE, .name = "SET_APP_PROFILE", .handler = do_set_app_profile, .perm_check = only_manager },
 	{ .cmd = KSU_IOCTL_GET_FEATURE, .name = "GET_FEATURE", .handler = do_get_feature, .perm_check = manager_or_root },
 	{ .cmd = KSU_IOCTL_SET_FEATURE, .name = "SET_FEATURE", .handler = do_set_feature, .perm_check = manager_or_root },
-    { .cmd = KSU_IOCTL_PROXY_FILE, .name = "PROXY_FILE", .handler = do_get_wrapper_fd, .perm_check = manager_or_root },
+	{ .cmd = KSU_IOCTL_GET_WRAPPER_FD, .name = "GET_WRAPPER_FD", .handler = do_get_wrapper_fd, .perm_check = manager_or_root },
+	{ .cmd = KSU_IOCTL_MANAGE_MARK, .name = "MANAGE_MARK", .handler = do_manage_mark, .perm_check = manager_or_root },
 	{ .cmd = KSU_IOCTL_GET_FULL_VERSION,.name = "GET_FULL_VERSION", .handler = do_get_full_version, .perm_check = always_allow},
 	{ .cmd = KSU_IOCTL_HOOK_TYPE,.name = "GET_HOOK_TYPE", .handler = do_get_hook_type, .perm_check = manager_or_root},
 	{ .cmd = KSU_IOCTL_ENABLE_KPM, .name = "GET_ENABLE_KPM", .handler = do_enable_kpm, .perm_check = manager_or_root},
@@ -600,18 +654,126 @@ static const struct ksu_ioctl_cmd_map ksu_ioctl_handlers[] = {
 	{ .cmd = 0, .name = NULL, .handler = NULL, .perm_check = NULL} // Sentine
 };
 
+#ifdef KSU_SHOULD_USE_NEW_TP
+struct ksu_install_fd_tw {
+	struct callback_head cb;
+	int __user *outp;
+};
+
+static void ksu_install_fd_tw_func(struct callback_head *cb)
+{
+	struct ksu_install_fd_tw *tw =
+		container_of(cb, struct ksu_install_fd_tw, cb);
+	int fd = ksu_install_fd();
+	pr_info("[%d] install ksu fd: %d\n", current->pid, fd);
+
+	if (copy_to_user(tw->outp, &fd, sizeof(fd))) {
+		pr_err("install ksu fd reply err\n");
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 11, 0)
+		close_fd(fd);
+#else
+		ksys_close(fd);
+#endif
+	}
+
+	kfree(tw);
+}
+
+static int reboot_handler_pre(struct kprobe *p, struct pt_regs *regs)
+{
+	struct pt_regs *real_regs = PT_REAL_REGS(regs);
+	int magic1 = (int)PT_REGS_PARM1(real_regs);
+	int magic2 = (int)PT_REGS_PARM2(real_regs);
+	unsigned long arg4;
+
+	// Check if this is a request to install KSU fd
+	if (magic1 == KSU_INSTALL_MAGIC1 && magic2 == KSU_INSTALL_MAGIC2) {
+		struct ksu_install_fd_tw *tw;
+
+		arg4 = (unsigned long)PT_REGS_SYSCALL_PARM4(real_regs);
+
+		tw = kzalloc(sizeof(*tw), GFP_ATOMIC);
+		if (!tw)
+			return 0;
+
+		tw->outp = (int __user *)arg4;
+		tw->cb.func = ksu_install_fd_tw_func;
+
+		if (ksu_task_work_add(current, &tw->cb, TWA_RESUME)) {
+			kfree(tw);
+			pr_warn("install fd add task_work failed\n");
+		}
+	}
+
+	return 0;
+}
+
+static struct kprobe reboot_kp = {
+	.symbol_name = REBOOT_SYMBOL,
+	.pre_handler = reboot_handler_pre,
+};
+#else
+int ksu_handle_sys_reboot(int magic1, int magic2, unsigned int cmd,
+			  void __user **arg)
+{
+	if (magic1 != KSU_INSTALL_MAGIC1)
+		return 0;
+
+#ifdef CONFIG_KSU_DEBUG
+	pr_info("sys_reboot: intercepted call! magic: 0x%x id: %d\n", magic1,
+		magic2);
+#endif
+
+	// Check if this is a request to install KSU fd
+	if (magic2 == KSU_INSTALL_MAGIC2) {
+		int fd = ksu_install_fd();
+		// downstream: dereference all arg usage!
+		if (copy_to_user((void __user *)*arg, &fd, sizeof(fd))) {
+			pr_err("install ksu fd reply err\n");
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 11, 0)
+			close_fd(fd);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(4, 17, 0)
+			ksys_close(fd);
+#else
+			sys_close(fd);
+#endif
+		}
+		return 0;
+	}
+
+	return 0;
+}
+#endif
+
 void ksu_supercalls_init(void)
 {
 	int i;
 
 	pr_info("KernelSU IOCTL Commands:\n");
 	for (i = 0; ksu_ioctl_handlers[i].handler; i++) {
-        pr_info("  %-18s = 0x%08x\n", ksu_ioctl_handlers[i].name, ksu_ioctl_handlers[i].cmd);
+		pr_info("  %-18s = 0x%08x\n", ksu_ioctl_handlers[i].name,
+			ksu_ioctl_handlers[i].cmd);
 	}
+#ifdef KSU_SHOULD_USE_NEW_TP
+	int rc = register_kprobe(&reboot_kp);
+	if (rc) {
+		pr_err("reboot kprobe failed: %d\n", rc);
+	} else {
+		pr_info("reboot kprobe registered successfully\n");
+	}
+#endif
+}
+
+void ksu_supercalls_exit(void)
+{
+#ifdef KSU_SHOULD_USE_NEW_TP
+	unregister_kprobe(&reboot_kp);
+#endif
 }
 
 // IOCTL dispatcher
-static long anon_ksu_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
+static long anon_ksu_ioctl(struct file *filp, unsigned int cmd,
+			   unsigned long arg)
 {
 	void __user *argp = (void __user *)arg;
 	int i;
@@ -630,8 +792,7 @@ static long anon_ksu_ioctl(struct file *filp, unsigned int cmd, unsigned long ar
 				return -EPERM;
 			}
 			// Execute handler
-            int ret = ksu_ioctl_handlers[i].handler(argp);
+			return ksu_ioctl_handlers[i].handler(argp);
-            return ret;
 		}
 	}
 
@@ -642,7 +803,9 @@ static long anon_ksu_ioctl(struct file *filp, unsigned int cmd, unsigned long ar
 // File release handler
 static int anon_ksu_release(struct inode *inode, struct file *filp)
 {
+#ifdef CONFIG_KSU_DEBUG
 	pr_info("ksu fd released\n");
+#endif
 	return 0;
 }
 
@@ -668,7 +831,8 @@ int ksu_install_fd(void)
 	}
 
 	// Create anonymous inode file
-    filp = anon_inode_getfile("[ksu_driver]", &anon_ksu_fops, NULL, O_RDWR | O_CLOEXEC);
+	filp = anon_inode_getfile("[ksu_driver]", &anon_ksu_fops, NULL,
+				  O_RDWR | O_CLOEXEC);
 	if (IS_ERR(filp)) {
 		pr_err("ksu_install_fd: failed to create anon inode file\n");
 		put_unused_fd(fd);
@@ -678,7 +842,8 @@ int ksu_install_fd(void)
 	// Install fd
 	fd_install(fd, filp);
 
-    pr_info("ksu fd[%d] installed for %s/%d\n", fd, current->comm, current->pid);
+	pr_info("ksu fd[%d] installed for %s/%d\n", fd, current->comm,
+		current->pid);
 
 	return fd;
 }

kernel/supercalls.h

@@ -3,7 +3,7 @@
 
 #include <linux/types.h>
 #include <linux/ioctl.h>
-#include "ksu.h"
+#include "app_profile.h"
 
 #ifdef CONFIG_KPM
 #include "kpm/kpm.h"
@@ -105,10 +105,21 @@ struct ksu_enable_uid_scanner_cmd {
 };
 
 struct ksu_get_wrapper_fd_cmd {
-	__u32 fd;
+	__u32 fd; // Input: userspace fd
-	__u32 flags; // CLOEXEC
+	__u32 flags; // Input: flags of userspace fd
 };
 
+struct ksu_manage_mark_cmd {
+	__u32 operation; // Input: KSU_MARK_*
+	__s32 pid; // Input: target pid (0 for all processes)
+	__u32 result; // Output: for get operation - mark status or reg_count
+};
+
+#define KSU_MARK_GET 1
+#define KSU_MARK_MARK 2
+#define KSU_MARK_UNMARK 3
+#define KSU_MARK_REFRESH 4
+
 // IOCTL command definitions
 #define KSU_IOCTL_GRANT_ROOT _IOC(_IOC_NONE, 'K', 1, 0)
 #define KSU_IOCTL_GET_INFO _IOC(_IOC_READ, 'K', 2, 0)
@@ -124,7 +135,8 @@ struct ksu_get_wrapper_fd_cmd {
 #define KSU_IOCTL_SET_APP_PROFILE _IOC(_IOC_WRITE, 'K', 12, 0)
 #define KSU_IOCTL_GET_FEATURE _IOC(_IOC_READ|_IOC_WRITE, 'K', 13, 0)
 #define KSU_IOCTL_SET_FEATURE _IOC(_IOC_WRITE, 'K', 14, 0)
-#define KSU_IOCTL_PROXY_FILE _IOC(_IOC_NONE, 'K', 15, 0)
+#define KSU_IOCTL_GET_WRAPPER_FD _IOC(_IOC_WRITE, 'K', 15, 0)
+#define KSU_IOCTL_MANAGE_MARK _IOC(_IOC_READ | _IOC_WRITE, 'K', 16, 0)
 // Other IOCTL command definitions
 #define KSU_IOCTL_GET_FULL_VERSION _IOC(_IOC_READ, 'K', 100, 0)
 #define KSU_IOCTL_HOOK_TYPE _IOC(_IOC_READ, 'K', 101, 0)
@@ -148,6 +160,6 @@ struct ksu_ioctl_cmd_map {
 // Install KSU fd to current process
 int ksu_install_fd(void);
 
-void ksu_supercalls_init();
+void ksu_supercalls_init(void);
-
+void ksu_supercalls_exit(void);
 #endif // __KSU_H_SUPERCALLS

kernel/syscall_hook_manager.c

@@ -0,0 +1,381 @@
+#ifdef KSU_SHOULD_USE_NEW_TP
+#include "linux/compiler.h"
+#include "linux/cred.h"
+#include "linux/printk.h"
+#include "selinux/selinux.h"
+#include <linux/spinlock.h>
+#include <linux/kprobes.h>
+#include <linux/tracepoint.h>
+#include <asm/syscall.h>
+#include <linux/ptrace.h>
+#include <linux/slab.h>
+#include <trace/events/syscalls.h>
+
+#include "allowlist.h"
+#include "arch.h"
+#include "klog.h" // IWYU pragma: keep
+#include "syscall_hook_manager.h"
+#include "sucompat.h"
+#include "setuid_hook.h"
+#include "selinux/selinux.h"
+
+// Tracepoint registration count management
+// == 1: just us
+// >  1: someone else is also using syscall tracepoint e.g. ftrace
+static int tracepoint_reg_count = 0;
+static DEFINE_SPINLOCK(tracepoint_reg_lock);
+
+void ksu_clear_task_tracepoint_flag_if_needed(struct task_struct *t)
+{
+	unsigned long flags;
+	spin_lock_irqsave(&tracepoint_reg_lock, flags);
+	if (tracepoint_reg_count <= 1) {
+		ksu_clear_task_tracepoint_flag(t);
+	}
+	spin_unlock_irqrestore(&tracepoint_reg_lock, flags);
+}
+
+// Process marking management
+static void handle_process_mark(bool mark)
+{
+	struct task_struct *p, *t;
+	read_lock(&tasklist_lock);
+	for_each_process_thread (p, t) {
+		if (mark)
+			ksu_set_task_tracepoint_flag(t);
+		else
+			ksu_clear_task_tracepoint_flag(t);
+	}
+	read_unlock(&tasklist_lock);
+}
+
+void ksu_mark_all_process(void)
+{
+	handle_process_mark(true);
+	pr_info("hook_manager: mark all user process done!\n");
+}
+
+void ksu_unmark_all_process(void)
+{
+	handle_process_mark(false);
+	pr_info("hook_manager: unmark all user process done!\n");
+}
+
+static void ksu_mark_running_process_locked()
+{
+	struct task_struct *p, *t;
+	read_lock(&tasklist_lock);
+	for_each_process_thread (p, t) {
+		if (!t->mm) { // only user processes
+			continue;
+		}
+		int uid = task_uid(t).val;
+		const struct cred *cred = get_task_cred(t);
+		bool ksu_root_process = uid == 0 && is_task_ksu_domain(cred);
+		bool is_zygote_process = is_zygote(cred);
+		bool is_shell = uid == 2000;
+		// before boot completed, we shall mark init for marking zygote
+		bool is_init = t->pid == 1;
+		if (ksu_root_process || is_zygote_process || is_shell ||
+		    is_init || ksu_is_allow_uid(uid)) {
+			ksu_set_task_tracepoint_flag(t);
+			pr_info("hook_manager: mark process: pid:%d, uid: %d, comm:%s\n",
+				t->pid, uid, t->comm);
+		} else {
+			ksu_clear_task_tracepoint_flag(t);
+			pr_info("hook_manager: unmark process: pid:%d, uid: %d, comm:%s\n",
+				t->pid, uid, t->comm);
+		}
+		put_cred(cred);
+	}
+	read_unlock(&tasklist_lock);
+}
+
+void ksu_mark_running_process()
+{
+	unsigned long flags;
+	spin_lock_irqsave(&tracepoint_reg_lock, flags);
+	if (tracepoint_reg_count <= 1) {
+		ksu_mark_running_process_locked();
+	} else {
+		pr_info("hook_manager: not mark running process since syscall tracepoint is in use\n");
+	}
+	spin_unlock_irqrestore(&tracepoint_reg_lock, flags);
+}
+
+// Get task mark status
+// Returns: 1 if marked, 0 if not marked, -ESRCH if task not found
+int ksu_get_task_mark(pid_t pid)
+{
+	struct task_struct *task;
+	int marked = -ESRCH;
+
+	rcu_read_lock();
+	task = find_task_by_vpid(pid);
+	if (task) {
+		get_task_struct(task);
+		rcu_read_unlock();
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 11, 0)
+		marked = test_task_syscall_work(task, SYSCALL_TRACEPOINT) ? 1 :
+									    0;
+#else
+		marked = test_tsk_thread_flag(task, TIF_SYSCALL_TRACEPOINT) ?
+				 1 :
+				 0;
+#endif
+		put_task_struct(task);
+	} else {
+		rcu_read_unlock();
+	}
+
+	return marked;
+}
+
+// Set task mark status
+// Returns: 0 on success, -ESRCH if task not found
+int ksu_set_task_mark(pid_t pid, bool mark)
+{
+	struct task_struct *task;
+	int ret = -ESRCH;
+
+	rcu_read_lock();
+	task = find_task_by_vpid(pid);
+	if (task) {
+		get_task_struct(task);
+		rcu_read_unlock();
+		if (mark) {
+			ksu_set_task_tracepoint_flag(task);
+			pr_info("hook_manager: marked task pid=%d comm=%s\n",
+				pid, task->comm);
+		} else {
+			ksu_clear_task_tracepoint_flag(task);
+			pr_info("hook_manager: unmarked task pid=%d comm=%s\n",
+				pid, task->comm);
+		}
+		put_task_struct(task);
+		ret = 0;
+	} else {
+		rcu_read_unlock();
+	}
+
+	return ret;
+}
+
+static struct kretprobe *init_kretprobe(const char *name,
+					kretprobe_handler_t handler)
+{
+	struct kretprobe *rp = kzalloc(sizeof(struct kretprobe), GFP_KERNEL);
+	if (!rp)
+		return NULL;
+	rp->kp.symbol_name = name;
+	rp->handler = handler;
+	rp->data_size = 0;
+	rp->maxactive = 0;
+
+	int ret = register_kretprobe(rp);
+	pr_info("hook_manager: register_%s kretprobe: %d\n", name, ret);
+	if (ret) {
+		kfree(rp);
+		return NULL;
+	}
+
+	return rp;
+}
+
+static void destroy_kretprobe(struct kretprobe **rp_ptr)
+{
+	struct kretprobe *rp = *rp_ptr;
+	if (!rp)
+		return;
+	unregister_kretprobe(rp);
+	synchronize_rcu();
+	kfree(rp);
+	*rp_ptr = NULL;
+}
+
+static int syscall_regfunc_handler(struct kretprobe_instance *ri,
+				   struct pt_regs *regs)
+{
+	unsigned long flags;
+	spin_lock_irqsave(&tracepoint_reg_lock, flags);
+	if (tracepoint_reg_count < 1) {
+		// while install our tracepoint, mark our processes
+		ksu_mark_running_process_locked();
+	} else if (tracepoint_reg_count == 1) {
+		// while other tracepoint first added, mark all processes
+		ksu_mark_all_process();
+	}
+	tracepoint_reg_count++;
+	spin_unlock_irqrestore(&tracepoint_reg_lock, flags);
+	return 0;
+}
+
+static int syscall_unregfunc_handler(struct kretprobe_instance *ri,
+				     struct pt_regs *regs)
+{
+	unsigned long flags;
+	spin_lock_irqsave(&tracepoint_reg_lock, flags);
+	tracepoint_reg_count--;
+	if (tracepoint_reg_count <= 0) {
+		// while no tracepoint left, unmark all processes
+		ksu_unmark_all_process();
+	} else if (tracepoint_reg_count == 1) {
+		// while just our tracepoint left, unmark disallowed processes
+		ksu_mark_running_process_locked();
+	}
+	spin_unlock_irqrestore(&tracepoint_reg_lock, flags);
+	return 0;
+}
+
+static struct kretprobe *syscall_regfunc_rp = NULL;
+static struct kretprobe *syscall_unregfunc_rp = NULL;
+
+static inline bool check_syscall_fastpath(int nr)
+{
+	switch (nr) {
+	case __NR_newfstatat:
+	case __NR_faccessat:
+	case __NR_execve:
+	case __NR_setresuid:
+		return true;
+	default:
+		return false;
+	}
+}
+
+// Unmark init's child that are not zygote, adbd or ksud
+int ksu_handle_init_mark_tracker(const char __user **filename_user)
+{
+	char path[64];
+
+	if (unlikely(!filename_user))
+		return 0;
+
+	memset(path, 0, sizeof(path));
+	ksu_strncpy_from_user_nofault(path, *filename_user, sizeof(path));
+
+	if (likely(strstr(path, "/app_process") == NULL &&
+		   strstr(path, "/adbd") == NULL &&
+		   strstr(path, "/ksud") == NULL)) {
+		pr_info("hook_manager: unmark %d exec %s", current->pid, path);
+		ksu_clear_task_tracepoint_flag_if_needed(current);
+	}
+
+	return 0;
+}
+
+// Generic sys_enter handler that dispatches to specific handlers
+static void ksu_sys_enter_handler(void *data, struct pt_regs *regs, long id)
+{
+	if (unlikely(check_syscall_fastpath(id))) {
+		if (ksu_su_compat_enabled) {
+			// Handle newfstatat
+			if (id == __NR_newfstatat) {
+				int *dfd = (int *)&PT_REGS_PARM1(regs);
+				const char __user **filename_user =
+					(const char __user **)&PT_REGS_PARM2(
+						regs);
+				int *flags =
+					(int *)&PT_REGS_SYSCALL_PARM4(regs);
+				ksu_handle_stat(dfd, filename_user, flags);
+				return;
+			}
+
+			// Handle faccessat
+			if (id == __NR_faccessat) {
+				int *dfd = (int *)&PT_REGS_PARM1(regs);
+				const char __user **filename_user =
+					(const char __user **)&PT_REGS_PARM2(
+						regs);
+				int *mode = (int *)&PT_REGS_PARM3(regs);
+				ksu_handle_faccessat(dfd, filename_user, mode,
+						     NULL);
+				return;
+			}
+
+			// Handle execve
+			if (id == __NR_execve) {
+				const char __user **filename_user =
+					(const char __user **)&PT_REGS_PARM1(
+						regs);
+				if (current->pid != 1 &&
+				    is_init(get_current_cred())) {
+					ksu_handle_init_mark_tracker(
+						filename_user);
+				} else {
+					ksu_handle_execve_sucompat(
+						NULL, filename_user, NULL,
+						NULL, NULL);
+				}
+				return;
+			}
+		}
+
+		// Handle setresuid
+		if (id == __NR_setresuid) {
+			uid_t ruid = (uid_t)PT_REGS_PARM1(regs);
+			uid_t euid = (uid_t)PT_REGS_PARM2(regs);
+			uid_t suid = (uid_t)PT_REGS_PARM3(regs);
+			ksu_handle_setresuid(ruid, euid, suid);
+			return;
+		}
+	}
+}
+
+void ksu_syscall_hook_manager_init(void)
+{
+	int ret;
+	pr_info("hook_manager: ksu_hook_manager_init called\n");
+
+	// Register kretprobe for syscall_regfunc
+	syscall_regfunc_rp =
+		init_kretprobe("syscall_regfunc", syscall_regfunc_handler);
+	// Register kretprobe for syscall_unregfunc
+	syscall_unregfunc_rp =
+		init_kretprobe("syscall_unregfunc", syscall_unregfunc_handler);
+
+	ret = register_trace_sys_enter(ksu_sys_enter_handler, NULL);
+	if (ret) {
+		pr_err("hook_manager: failed to register sys_enter tracepoint: %d\n",
+		       ret);
+	} else {
+		pr_info("hook_manager: sys_enter tracepoint registered\n");
+	}
+
+	ksu_setuid_hook_init();
+	ksu_sucompat_init();
+}
+
+void ksu_syscall_hook_manager_exit(void)
+{
+	pr_info("hook_manager: ksu_hook_manager_exit called\n");
+	unregister_trace_sys_enter(ksu_sys_enter_handler, NULL);
+	tracepoint_synchronize_unregister();
+	pr_info("hook_manager: sys_enter tracepoint unregistered\n");
+
+	destroy_kretprobe(&syscall_regfunc_rp);
+	destroy_kretprobe(&syscall_unregfunc_rp);
+
+	ksu_sucompat_exit();
+	ksu_setuid_hook_exit();
+}
+#else
+#include "klog.h" // IWYU pragma: keep
+#include "syscall_hook_manager.h"
+#include "sucompat.h"
+#include "setuid_hook.h"
+
+void ksu_syscall_hook_manager_init(void)
+{
+	pr_info("hook_manager: initializing..\n");
+	ksu_setuid_hook_init();
+	ksu_sucompat_init();
+}
+
+void ksu_syscall_hook_manager_exit(void)
+{
+	pr_info("hook_manager: exiting..\n");
+	ksu_sucompat_exit();
+	ksu_setuid_hook_exit();
+}
+#endif

kernel/syscall_hook_manager.h

@@ -0,0 +1,85 @@
+#ifndef __KSU_H_HOOK_MANAGER
+#define __KSU_H_HOOK_MANAGER
+
+#include <linux/version.h>
+#include <linux/sched.h>
+#include <linux/thread_info.h>
+#include "kernel_compat.h"
+
+// Hook manager initialization and cleanup
+void ksu_syscall_hook_manager_init(void);
+void ksu_syscall_hook_manager_exit(void);
+
+#ifdef KSU_SHOULD_USE_NEW_TP
+// Process marking for tracepoint
+void ksu_mark_all_process(void);
+void ksu_unmark_all_process(void);
+void ksu_mark_running_process(void);
+
+// Per-task mark operations
+int ksu_get_task_mark(pid_t pid);
+int ksu_set_task_mark(pid_t pid, bool mark);
+
+static inline void ksu_set_task_tracepoint_flag(struct task_struct *t)
+{
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 11, 0)
+	set_task_syscall_work(t, SYSCALL_TRACEPOINT);
+#else
+	set_tsk_thread_flag(t, TIF_SYSCALL_TRACEPOINT);
+#endif
+}
+
+static inline void ksu_clear_task_tracepoint_flag(struct task_struct *t)
+{
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 11, 0)
+	clear_task_syscall_work(t, SYSCALL_TRACEPOINT);
+#else
+	clear_tsk_thread_flag(t, TIF_SYSCALL_TRACEPOINT);
+#endif
+}
+
+void ksu_clear_task_tracepoint_flag_if_needed(struct task_struct *t);
+#else
+// Process marking for tracepoint
+static inline void ksu_mark_all_process(void)
+{
+	return;
+}
+
+static inline void ksu_unmark_all_process(void)
+{
+	return;
+}
+
+static inline void ksu_mark_running_process(void)
+{
+	return;
+}
+
+// Per-task mark operations
+static inline int ksu_get_task_mark(pid_t pid)
+{
+	return 0;
+}
+
+static inline int ksu_set_task_mark(pid_t pid, bool mark)
+{
+	return 0;
+}
+
+static inline void ksu_set_task_tracepoint_flag(struct task_struct *t)
+{
+	return;
+}
+
+static inline void ksu_clear_task_tracepoint_flag(struct task_struct *t)
+{
+	return;
+}
+static inline void ksu_clear_task_tracepoint_flag_if_needed(struct task_struct *t)
+{
+	return;
+}
+#endif
+
+#endif

kernel/throne_tracker.c

@@ -10,11 +10,9 @@
 
 #include "allowlist.h"
 #include "klog.h" // IWYU pragma: keep
-#include "ksu.h"
-#include "apk_sign.h"
 #include "manager.h"
 #include "throne_tracker.h"
-#include "kernel_compat.h"
+#include "apk_sign.h"
 #include "dynamic_manager.h"
 #include "throne_comm.h"
 
@@ -40,7 +38,7 @@ static int uid_from_um_list(struct list_head *uid_list)
 	ssize_t nr;
 	int cnt = 0;
 
-	fp = ksu_filp_open_compat(KSU_UID_LIST_PATH, O_RDONLY, 0);
+	fp = filp_open(KSU_UID_LIST_PATH, O_RDONLY, 0);
 	if (IS_ERR(fp))
 		return -ENOENT;
 
@@ -57,7 +55,7 @@ static int uid_from_um_list(struct list_head *uid_list)
 		return -ENOMEM;
 	}
 
-	nr = ksu_kernel_read_compat(fp, buf, size, &pos);
+	nr = kernel_read(fp, buf, size, &pos);
 	filp_close(fp, NULL);
 	if (nr != size) {
 		pr_err("uid_list: short read %zd/%lld\n", nr, size);
@@ -235,7 +233,6 @@ struct my_dir_context {
 #define FILLDIR_ACTOR_CONTINUE 0
 #define FILLDIR_ACTOR_STOP -EINVAL
 #endif
-
 FILLDIR_RETURN_TYPE my_actor(struct dir_context *ctx, const char *name,
 							 int namelen, loff_t off, u64 ino,
 							 unsigned int d_type)
@@ -271,7 +268,7 @@ FILLDIR_RETURN_TYPE my_actor(struct dir_context *ctx, const char *name,
 
 	if (d_type == DT_DIR && my_ctx->depth > 0 &&
 		(my_ctx->stop && !*my_ctx->stop)) {
-		struct data_path *data = kmalloc(sizeof(struct data_path), GFP_ATOMIC);
+		struct data_path *data = kzalloc(sizeof(struct data_path), GFP_ATOMIC);
 
 		if (!data) {
 			pr_err("Failed to allocate memory for %s\n", dirpath);
@@ -306,29 +303,24 @@ FILLDIR_RETURN_TYPE my_actor(struct dir_context *ctx, const char *name,
 			// Check for dynamic sign or multi-manager signatures
 			if (is_multi_manager && (signature_index == DYNAMIC_SIGN_INDEX || signature_index >= 2)) {
 				crown_manager(dirpath, my_ctx->private_data, signature_index);
+			} else if (is_manager_apk(dirpath)) {
+				crown_manager(dirpath, my_ctx->private_data, 0);
+				*my_ctx->stop = 1;
+			}
 
-				struct apk_path_hash *apk_data = kmalloc(sizeof(struct apk_path_hash), GFP_ATOMIC);
+			struct apk_path_hash *apk_data = kzalloc(sizeof(*apk_data), GFP_ATOMIC);
 			if (apk_data) {
 				apk_data->hash  = hash;
 				apk_data->exists = true;
 				list_add_tail(&apk_data->list, &apk_path_hash_list);
 			}
-			} else if (is_manager_apk(dirpath)) {
-				crown_manager(dirpath, my_ctx->private_data, 0);
-				*my_ctx->stop = 1;
 
+			if (is_manager_apk(dirpath)) {
 				// Manager found, clear APK cache list
 				list_for_each_entry_safe(pos, n, &apk_path_hash_list, list) {
 					list_del(&pos->list);
 					kfree(pos);
 				}
-			} else {
-				struct apk_path_hash *apk_data = kmalloc(sizeof(struct apk_path_hash), GFP_ATOMIC);
-				if (apk_data) {
-					apk_data->hash = hash;
-					apk_data->exists = true;
-					list_add_tail(&apk_data->list, &apk_path_hash_list);
-				}
 			}
 		}
 	}
@@ -368,9 +360,10 @@ void search_manager(const char *path, int depth, struct list_head *uid_data)
 			struct file *file;
 
 			if (!stop) {
-				file = ksu_filp_open_compat(pos->dirpath, O_RDONLY | O_NOFOLLOW, 0);
+				file = filp_open(pos->dirpath, O_RDONLY | O_NOFOLLOW, 0);
 				if (IS_ERR(file)) {
-					pr_err("Failed to open directory: %s, err: %ld\n", pos->dirpath, PTR_ERR(file));
+					pr_err("Failed to open directory: %s, err: %ld\n",
+						   pos->dirpath, PTR_ERR(file));
 					goto skip_iterate;
 				}
 
@@ -378,7 +371,8 @@ void search_manager(const char *path, int depth, struct list_head *uid_data)
 				if (!data_app_magic) {
 					if (file->f_inode->i_sb->s_magic) {
 						data_app_magic = file->f_inode->i_sb->s_magic;
-						pr_info("%s: dir: %s got magic! 0x%lx\n", __func__, pos->dirpath, data_app_magic);
+						pr_info("%s: dir: %s got magic! 0x%lx\n", __func__,
+								pos->dirpath, data_app_magic);
 					} else {
 						filp_close(file, NULL);
 						goto skip_iterate;
@@ -386,7 +380,8 @@ void search_manager(const char *path, int depth, struct list_head *uid_data)
 				}
 
 				if (file->f_inode->i_sb->s_magic != data_app_magic) {
-					pr_info("%s: skip: %s magic: 0x%lx expected: 0x%lx\n", __func__, pos->dirpath, 
+					pr_info("%s: skip: %s magic: 0x%lx expected: 0x%lx\n",
+							__func__, pos->dirpath,
 							file->f_inode->i_sb->s_magic, data_app_magic);
 					filp_close(file, NULL);
 					goto skip_iterate;
@@ -427,7 +422,7 @@ static bool is_uid_exist(uid_t uid, char *package, void *data)
 	return exist;
 }
 
-void track_throne(void)
+void track_throne(bool prune_only)
 {
 	struct list_head uid_list;
 	struct uid_data *np, *n;
@@ -456,7 +451,7 @@ void track_throne(void)
 	}
 
 	{
-		fp = ksu_filp_open_compat(SYSTEM_PACKAGES_LIST_PATH, O_RDONLY, 0);
+		fp = filp_open(SYSTEM_PACKAGES_LIST_PATH, O_RDONLY, 0);
 		if (IS_ERR(fp)) {
 			pr_err("%s: open " SYSTEM_PACKAGES_LIST_PATH " failed: %ld\n", __func__, PTR_ERR(fp));
 			return;
@@ -464,13 +459,13 @@ void track_throne(void)
 
 		for (;;) {
 			ssize_t count =
-				ksu_kernel_read_compat(fp, &chr, sizeof(chr), &pos);
+				kernel_read(fp, &chr, sizeof(chr), &pos);
 			if (count != sizeof(chr))
 				break;
 			if (chr != '\n')
 				continue;
 
-			count = ksu_kernel_read_compat(fp, buf, sizeof(buf),
+			count = kernel_read(fp, buf, sizeof(buf),
 							&line_start);
 			struct uid_data *data =
 				kzalloc(sizeof(struct uid_data), GFP_ATOMIC);
@@ -504,6 +499,9 @@ void track_throne(void)
 	}
 
 uid_ready:
+	if (prune_only)
+		goto prune;
+
 	// first, check if manager_uid exist!
 	list_for_each_entry(np, &uid_list, list) {
 		if (np->uid == current_manager_uid) {
@@ -547,6 +545,7 @@ uid_ready:
 		pr_info("Manager search finished\n");
 	}
 	
+prune:
 	// then prune the allowlist
 	ksu_prune_allowlist(is_uid_exist, &uid_list);
 out:

kernel/throne_tracker.h

@@ -5,6 +5,6 @@ void ksu_throne_tracker_init(void);
 
 void ksu_throne_tracker_exit(void);
 
-void track_throne(void);
+void track_throne(bool prune_only);
 
 #endif