Magisk/native/jni/su/su_daemon.cpp
topjohnwu ec3705f2ed Redesign of MagiskSU's sepolicy model
Introduce new domain `magisk_client` and new file type `magisk_exec`.

Connection to magiskd's always-on socket is restricted to magisk_client
only. Whitelisted process domains can transit to magisk_client through
executing files labelled magisk_exec. The main magisk binary shall be
the only file labelled as magisk_exec throughout the whole system.
All processes thus are no longer allowed to connect to magiskd directly
without going through the proper magisk binary.

Connection failures are silenced from audit logs with dontaudit rules,
so crazy processes which traverse through all unix domain sockets to try
connection can no longer check logcat to know the actual reason behind
EACCES, leaking the denied process policy (which is u:r:magisk:s0).

This also allows us to remove many rules that open up holes in
untrusted_app domains that were used to make remote shell work properly.
Since all processes establishing the remote shell are now restricted to
the magisk_client domain, all these rules are moved to magisk_client.
This makes Magisk require fewer compromises in Android's security model.

Note: as of this commit, requesting new root access via Magisk Manager
will stop working as Magisk Manager can no longer communicate with
magiskd directly. This will be addressed in a future commit that
involves changes in both native and application side.
2020-06-03 23:29:42 -07:00

354 lines
8.3 KiB
C++

#include <unistd.h>
#include <pthread.h>
#include <stdlib.h>
#include <fcntl.h>
#include <string.h>
#include <signal.h>
#include <pwd.h>
#include <time.h>
#include <sys/socket.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/wait.h>
#include <sys/mount.h>
#include <logging.hpp>
#include <daemon.hpp>
#include <utils.hpp>
#include <selinux.hpp>
#include "su.hpp"
#include "pts.hpp"
using namespace std;
static pthread_mutex_t cache_lock = PTHREAD_MUTEX_INITIALIZER;
static shared_ptr<su_info> cached;
su_info::su_info(unsigned uid) :
uid(uid), access(DEFAULT_SU_ACCESS), mgr_st({}),
timestamp(0), _lock(PTHREAD_MUTEX_INITIALIZER) {}
su_info::~su_info() {
pthread_mutex_destroy(&_lock);
}
mutex_guard su_info::lock() {
return mutex_guard(_lock);
}
bool su_info::is_fresh() {
timespec ts;
clock_gettime(CLOCK_MONOTONIC, &ts);
long current = ts.tv_sec * 1000L + ts.tv_nsec / 1000000L;
return current - timestamp < 3000; /* 3 seconds */
}
void su_info::refresh() {
timespec ts;
clock_gettime(CLOCK_MONOTONIC, &ts);
timestamp = ts.tv_sec * 1000L + ts.tv_nsec / 1000000L;
}
static void database_check(const shared_ptr<su_info> &info) {
int uid = info->uid;
get_db_settings(info->cfg);
get_db_strings(info->str);
// Check multiuser settings
switch (info->cfg[SU_MULTIUSER_MODE]) {
case MULTIUSER_MODE_OWNER_ONLY:
if (info->uid / 100000) {
uid = -1;
info->access = NO_SU_ACCESS;
}
break;
case MULTIUSER_MODE_OWNER_MANAGED:
uid = info->uid % 100000;
break;
case MULTIUSER_MODE_USER:
default:
break;
}
if (uid > 0)
get_uid_policy(info->access, uid);
// We need to check our manager
if (info->access.log || info->access.notify)
validate_manager(info->str[SU_MANAGER], uid / 100000, &info->mgr_st);
}
static shared_ptr<su_info> get_su_info(unsigned uid) {
LOGD("su: request from uid=[%d]\n", uid);
shared_ptr<su_info> info;
{
mutex_guard lock(cache_lock);
if (!cached || cached->uid != uid || !cached->is_fresh())
cached = make_shared<su_info>(uid);
cached->refresh();
info = cached;
}
auto g = info->lock();
if (info->access.policy == QUERY) {
// Not cached, get data from database
database_check(info);
// If it's root or the manager, allow it silently
if (info->uid == UID_ROOT || (info->uid % 100000) == (info->mgr_st.st_uid % 100000)) {
info->access = SILENT_SU_ACCESS;
return info;
}
// Check su access settings
switch (info->cfg[ROOT_ACCESS]) {
case ROOT_ACCESS_DISABLED:
LOGW("Root access is disabled!\n");
info->access = NO_SU_ACCESS;
break;
case ROOT_ACCESS_ADB_ONLY:
if (info->uid != UID_SHELL) {
LOGW("Root access limited to ADB only!\n");
info->access = NO_SU_ACCESS;
}
break;
case ROOT_ACCESS_APPS_ONLY:
if (info->uid == UID_SHELL) {
LOGW("Root access is disabled for ADB!\n");
info->access = NO_SU_ACCESS;
}
break;
case ROOT_ACCESS_APPS_AND_ADB:
default:
break;
}
if (info->access.policy != QUERY)
return info;
// If still not determined, check if manager exists
if (info->str[SU_MANAGER][0] == '\0') {
info->access = NO_SU_ACCESS;
return info;
}
} else {
return info;
}
// If still not determined, ask manager
struct sockaddr_un addr;
int sockfd = create_rand_socket(&addr);
// Connect manager
app_socket(addr.sun_path + 1, info);
int fd = socket_accept(sockfd, 60);
if (fd < 0) {
info->access.policy = DENY;
} else {
socket_send_request(fd, info);
int ret = read_int_be(fd);
info->access.policy = ret < 0 ? DENY : static_cast<policy_t>(ret);
close(fd);
}
close(sockfd);
return info;
}
static void set_identity(unsigned uid) {
/*
* Set effective uid back to root, otherwise setres[ug]id will fail
* if uid isn't root.
*/
if (seteuid(0)) {
PLOGE("seteuid (root)");
}
if (setresgid(uid, uid, uid)) {
PLOGE("setresgid (%u)", uid);
}
if (setresuid(uid, uid, uid)) {
PLOGE("setresuid (%u)", uid);
}
}
void su_daemon_handler(int client, struct ucred *credential) {
LOGD("su: request from pid=[%d], client=[%d]\n", credential->pid, client);
su_context ctx = {
.info = get_su_info(credential->uid),
.req = su_request(true),
.pid = credential->pid
};
// Read su_request
xxread(client, &ctx.req, sizeof(su_req_base));
ctx.req.shell = read_string(client);
ctx.req.command = read_string(client);
if (ctx.info->access.log)
app_log(ctx);
else if (ctx.info->access.notify)
app_notify(ctx);
// Fail fast
if (ctx.info->access.policy == DENY) {
LOGW("su: request rejected (%u)", ctx.info->uid);
ctx.info.reset();
write_int(client, DENY);
close(client);
return;
} else if (int child = xfork(); child) {
ctx.info.reset();
// Wait result
LOGD("su: waiting child pid=[%d]\n", child);
int status, code;
if (waitpid(child, &status, 0) > 0)
code = WEXITSTATUS(status);
else
code = -1;
LOGD("su: return code=[%d]\n", code);
write(client, &code, sizeof(code));
close(client);
return;
}
/* The child process will need to setsid, open a pseudo-terminal
* if needed, and will eventually run exec.
* The parent process will wait for the result and
* send the return code back to our client
*/
LOGD("su: fork handler\n");
// Abort upon any error occurred
log_cb.ex = exit;
// ack
write_int(client, 0);
// Become session leader
xsetsid();
// Get pts_slave
char *pts_slave = read_string(client);
// The FDs for each of the streams
int infd = recv_fd(client);
int outfd = recv_fd(client);
int errfd = recv_fd(client);
if (pts_slave[0]) {
LOGD("su: pts_slave=[%s]\n", pts_slave);
// Check pts_slave file is owned by daemon_from_uid
struct stat st;
xstat(pts_slave, &st);
// If caller is not root, ensure the owner of pts_slave is the caller
if(st.st_uid != ctx.info->uid && ctx.info->uid != 0)
LOGE("su: Wrong permission of pts_slave");
// Opening the TTY has to occur after the
// fork() and setsid() so that it becomes
// our controlling TTY and not the daemon's
int ptsfd = xopen(pts_slave, O_RDWR);
if (infd < 0)
infd = ptsfd;
if (outfd < 0)
outfd = ptsfd;
if (errfd < 0)
errfd = ptsfd;
}
free(pts_slave);
// Swap out stdin, stdout, stderr
xdup2(infd, STDIN_FILENO);
xdup2(outfd, STDOUT_FILENO);
xdup2(errfd, STDERR_FILENO);
// Unleash all streams from SELinux hell
setfilecon("/proc/self/fd/0", "u:object_r:" SEPOL_FILE_TYPE ":s0");
setfilecon("/proc/self/fd/1", "u:object_r:" SEPOL_FILE_TYPE ":s0");
setfilecon("/proc/self/fd/2", "u:object_r:" SEPOL_FILE_TYPE ":s0");
close(infd);
close(outfd);
close(errfd);
close(client);
// Handle namespaces
if (ctx.req.mount_master)
ctx.info->cfg[SU_MNT_NS] = NAMESPACE_MODE_GLOBAL;
switch (ctx.info->cfg[SU_MNT_NS]) {
case NAMESPACE_MODE_GLOBAL:
LOGD("su: use global namespace\n");
break;
case NAMESPACE_MODE_REQUESTER:
LOGD("su: use namespace of pid=[%d]\n", ctx.pid);
if (switch_mnt_ns(ctx.pid))
LOGD("su: setns failed, fallback to global\n");
break;
case NAMESPACE_MODE_ISOLATE:
LOGD("su: use new isolated namespace\n");
xunshare(CLONE_NEWNS);
xmount(nullptr, "/", nullptr, MS_PRIVATE | MS_REC, nullptr);
break;
}
const char *argv[] = { nullptr, nullptr, nullptr, nullptr };
argv[0] = ctx.req.login ? "-" : ctx.req.shell;
if (ctx.req.command[0]) {
argv[1] = "-c";
argv[2] = ctx.req.command;
}
// Setup environment
umask(022);
char path[32];
snprintf(path, sizeof(path), "/proc/%d/cwd", ctx.pid);
chdir(path);
snprintf(path, sizeof(path), "/proc/%d/environ", ctx.pid);
char buf[4096] = { 0 };
int fd = xopen(path, O_RDONLY);
read(fd, buf, sizeof(buf));
close(fd);
clearenv();
for (size_t pos = 0; buf[pos];) {
putenv(buf + pos);
pos += strlen(buf + pos) + 1;
}
if (!ctx.req.keepenv) {
struct passwd *pw;
pw = getpwuid(ctx.req.uid);
if (pw) {
setenv("HOME", pw->pw_dir, 1);
setenv("USER", pw->pw_name, 1);
setenv("LOGNAME", pw->pw_name, 1);
setenv("SHELL", ctx.req.shell, 1);
}
}
const char *ld_path = getenv("LD_LIBRARY_PATH");
if (ld_path && strncmp(ld_path, ":/apex/com.android.runtime/lib", 30) == 0)
unsetenv("LD_LIBRARY_PATH");
// Unblock all signals
sigset_t block_set;
sigemptyset(&block_set);
sigprocmask(SIG_SETMASK, &block_set, nullptr);
set_identity(ctx.req.uid);
execvp(ctx.req.shell, (char **) argv);
fprintf(stderr, "Cannot execute %s: %s\n", ctx.req.shell, strerror(errno));
PLOGE("exec");
exit(EXIT_FAILURE);
}