Merge "tombstoned: log where we're writing the tombstone."
diff --git a/adb/adb_utils.h b/adb/adb_utils.h
index 20c63b3..11c0ec9 100644
--- a/adb/adb_utils.h
+++ b/adb/adb_utils.h
@@ -74,13 +74,18 @@
template <typename Fn>
void PopAll(Fn fn) {
- std::unique_lock<std::mutex> lock(mutex);
- cv.wait(lock, [this]() { return !queue.empty(); });
+ std::vector<T> popped;
- for (const T& t : queue) {
+ {
+ std::unique_lock<std::mutex> lock(mutex);
+ cv.wait(lock, [this]() { return !queue.empty(); });
+ popped = std::move(queue);
+ queue.clear();
+ }
+
+ for (const T& t : popped) {
fn(t);
}
- queue.clear();
}
};
diff --git a/adb/client/usb_libusb.cpp b/adb/client/usb_libusb.cpp
index 7e77b5e..e7f44c6 100644
--- a/adb/client/usb_libusb.cpp
+++ b/adb/client/usb_libusb.cpp
@@ -180,10 +180,6 @@
if (port_count < 0) return "";
return StringPrintf("/dev/bus/usb/%03u/%03u", libusb_get_bus_number(device), ports[0]);
}
-
-static bool is_device_accessible(libusb_device* device) {
- return access(get_device_dev_path(device).c_str(), R_OK | W_OK) == 0;
-}
#endif
static bool endpoint_is_output(uint8_t endpoint) {
@@ -377,9 +373,10 @@
{
std::unique_lock<std::mutex> lock(usb_handles_mutex);
usb_handles[device_address] = std::move(result);
- }
- register_usb_transport(usb_handle_raw, device_serial.c_str(), device_address.c_str(), writable);
+ register_usb_transport(usb_handle_raw, device_serial.c_str(), device_address.c_str(),
+ writable);
+ }
LOG(INFO) << "registered new usb device '" << device_serial << "'";
}
@@ -388,18 +385,12 @@
static void device_connected(libusb_device* device) {
#if defined(__linux__)
// Android's host linux libusb uses netlink instead of udev for device hotplug notification,
- // which means we can get hotplug notifications before udev has updated ownership/perms on
- // the device. Since we're not going to be able to link against the system's libudev any
- // time soon, hack around this by checking for accessibility in a loop.
+ // which means we can get hotplug notifications before udev has updated ownership/perms on the
+ // device. Since we're not going to be able to link against the system's libudev any time soon,
+ // hack around this by inserting a sleep.
auto thread = std::thread([device]() {
std::string device_path = get_device_dev_path(device);
- auto start = std::chrono::steady_clock::now();
- while (std::chrono::steady_clock::now() - start < 500ms) {
- if (is_device_accessible(device)) {
- break;
- }
- std::this_thread::sleep_for(10ms);
- }
+ std::this_thread::sleep_for(1s);
process_device(device);
if (--connecting_devices == 0) {
@@ -422,8 +413,10 @@
if (!it->second->device_handle) {
// If the handle is null, we were never able to open the device.
unregister_usb_transport(it->second.get());
+ usb_handles.erase(it);
+ } else {
+ // Closure of the transport will erase the usb_handle.
}
- usb_handles.erase(it);
}
}
diff --git a/adb/transport.cpp b/adb/transport.cpp
index 308ee8d..2bbbefd 100644
--- a/adb/transport.cpp
+++ b/adb/transport.cpp
@@ -603,15 +603,15 @@
static void transport_unref(atransport* t) {
CHECK(t != nullptr);
- std::lock_guard<std::mutex> lock(transport_lock);
- CHECK_GT(t->ref_count, 0u);
- t->ref_count--;
- if (t->ref_count == 0) {
+ size_t old_refcount = t->ref_count--;
+ CHECK_GT(old_refcount, 0u);
+
+ if (old_refcount == 1u) {
D("transport: %s unref (kicking and closing)", t->serial);
t->close(t);
remove_transport(t);
} else {
- D("transport: %s unref (count=%zu)", t->serial, t->ref_count);
+ D("transport: %s unref (count=%zu)", t->serial, old_refcount - 1);
}
}
diff --git a/adb/transport.h b/adb/transport.h
index 57fc988..4a89ed9 100644
--- a/adb/transport.h
+++ b/adb/transport.h
@@ -61,7 +61,7 @@
// class in one go is a very large change. Given how bad our testing is,
// it's better to do this piece by piece.
- atransport(ConnectionState state = kCsOffline) : connection_state_(state) {
+ atransport(ConnectionState state = kCsOffline) : ref_count(0), connection_state_(state) {
transport_fde = {};
protocol_version = A_VERSION;
max_payload = MAX_PAYLOAD;
@@ -88,7 +88,7 @@
int fd = -1;
int transport_socket = -1;
fdevent transport_fde;
- size_t ref_count = 0;
+ std::atomic<size_t> ref_count;
uint32_t sync_token = 0;
bool online = false;
TransportType type = kTransportAny;
diff --git a/adb/transport_local.cpp b/adb/transport_local.cpp
index a9e583f..6cedd92 100644
--- a/adb/transport_local.cpp
+++ b/adb/transport_local.cpp
@@ -388,6 +388,25 @@
D("transport: qemu_socket_thread() exiting");
return;
}
+
+// If adbd is running inside the emulator, it will normally use QEMUD pipe (aka
+// goldfish) as the transport. This can either be explicitly set by the
+// service.adb.transport property, or be inferred from ro.kernel.qemu that is
+// set to "1" for ranchu/goldfish.
+static bool use_qemu_goldfish() {
+ // Legacy way to detect if adbd should use the goldfish pipe is to check for
+ // ro.kernel.qemu, keep that behaviour for backward compatibility.
+ if (android::base::GetBoolProperty("ro.kernel.qemu", false)) {
+ return true;
+ }
+ // If service.adb.transport is present and is set to "goldfish", use the
+ // QEMUD pipe.
+ if (android::base::GetProperty("service.adb.transport", "") == "goldfish") {
+ return true;
+ }
+ return false;
+}
+
#endif // !ADB_HOST
void local_init(int port)
@@ -401,13 +420,7 @@
#else
// For the adbd daemon in the system image we need to distinguish
// between the device, and the emulator.
- if (android::base::GetBoolProperty("ro.kernel.qemu", false)) {
- // Running inside the emulator: use QEMUD pipe as the transport.
- func = qemu_socket_thread;
- } else {
- // Running inside the device: use TCP socket as the transport.
- func = server_socket_thread;
- }
+ func = use_qemu_goldfish() ? qemu_socket_thread : server_socket_thread;
debug_name = "server";
#endif // !ADB_HOST
diff --git a/debuggerd/Android.bp b/debuggerd/Android.bp
index 5565cfd..42c3023 100644
--- a/debuggerd/Android.bp
+++ b/debuggerd/Android.bp
@@ -7,6 +7,7 @@
"-Wno-nullability-completeness",
"-Os",
],
+ cpp_std: "experimental",
local_include_dirs: ["include"],
}
diff --git a/debuggerd/crash_dump.cpp b/debuggerd/crash_dump.cpp
index 558bc72..3ca9c92 100644
--- a/debuggerd/crash_dump.cpp
+++ b/debuggerd/crash_dump.cpp
@@ -44,6 +44,9 @@
#include <private/android_filesystem_config.h>
#include <procinfo/process.h>
+#define ATRACE_TAG ATRACE_TAG_BIONIC
+#include <utils/Trace.h>
+
#include "backtrace.h"
#include "tombstone.h"
#include "utility.h"
@@ -92,15 +95,11 @@
return false;
}
- // Put the task into ptrace-stop state.
- if (ptrace(PTRACE_INTERRUPT, tid, 0, 0) != 0) {
- PLOG(FATAL) << "failed to interrupt thread " << tid;
- }
-
return true;
}
static bool activity_manager_notify(pid_t pid, int signal, const std::string& amfd_data) {
+ ATRACE_CALL();
android::base::unique_fd amfd(socket_local_client(
"/data/system/ndebugsocket", ANDROID_SOCKET_NAMESPACE_FILESYSTEM, SOCK_STREAM));
if (amfd.get() == -1) {
@@ -176,6 +175,7 @@
}
static void drop_capabilities() {
+ ATRACE_CALL();
__user_cap_header_struct capheader;
memset(&capheader, 0, sizeof(capheader));
capheader.version = _LINUX_CAPABILITY_VERSION_3;
@@ -194,6 +194,8 @@
}
int main(int argc, char** argv) {
+ atrace_begin(ATRACE_TAG, "before reparent");
+
pid_t target = getppid();
bool tombstoned_connected = false;
unique_fd tombstoned_socket;
@@ -261,6 +263,8 @@
PLOG(FATAL) << "parent died";
}
+ atrace_end(ATRACE_TAG);
+
// Reparent ourselves to init, so that the signal handler can waitpid on the
// original process to avoid leaving a zombie for non-fatal dumps.
pid_t forkpid = fork();
@@ -270,19 +274,27 @@
exit(0);
}
+ ATRACE_NAME("after reparent");
+
// Die if we take too long.
alarm(2);
+ std::string process_name = get_process_name(main_tid);
std::string attach_error;
- // Seize the main thread.
- if (!ptrace_seize_thread(target_proc_fd, main_tid, &attach_error)) {
- LOG(FATAL) << attach_error;
- }
-
- // Seize the siblings.
std::map<pid_t, std::string> threads;
+
{
+ ATRACE_NAME("ptrace_interrupt");
+
+ // Seize the main thread.
+ if (!ptrace_seize_thread(target_proc_fd, main_tid, &attach_error)) {
+ LOG(FATAL) << attach_error;
+ }
+
+ threads.emplace(main_tid, get_thread_name(main_tid));
+
+ // Seize its siblings.
std::set<pid_t> siblings;
if (!android::procinfo::GetProcessTids(target, &siblings)) {
PLOG(FATAL) << "failed to get process siblings";
@@ -296,31 +308,48 @@
for (pid_t sibling_tid : siblings) {
if (!ptrace_seize_thread(target_proc_fd, sibling_tid, &attach_error)) {
LOG(WARNING) << attach_error;
- } else {
- threads.emplace(sibling_tid, get_thread_name(sibling_tid));
+ continue;
}
+ threads.emplace(sibling_tid, get_thread_name(sibling_tid));
}
}
// Collect the backtrace map, open files, and process/thread names, while we still have caps.
- std::unique_ptr<BacktraceMap> backtrace_map(BacktraceMap::Create(main_tid));
- if (!backtrace_map) {
- LOG(FATAL) << "failed to create backtrace map";
+ std::unique_ptr<BacktraceMap> backtrace_map;
+ {
+ ATRACE_NAME("backtrace map");
+ backtrace_map.reset(BacktraceMap::Create(main_tid));
+ if (!backtrace_map) {
+ LOG(FATAL) << "failed to create backtrace map";
+ }
}
// Collect the list of open files.
OpenFilesList open_files;
- populate_open_files_list(target, &open_files);
-
- std::string process_name = get_process_name(main_tid);
- threads.emplace(main_tid, get_thread_name(main_tid));
+ {
+ ATRACE_NAME("open files");
+ populate_open_files_list(target, &open_files);
+ }
// Drop our capabilities now that we've attached to the threads we care about.
drop_capabilities();
- const DebuggerdDumpType dump_type_enum = static_cast<DebuggerdDumpType>(dump_type);
- LOG(INFO) << "obtaining output fd from tombstoned, type: " << dump_type_enum;
- tombstoned_connected = tombstoned_connect(target, &tombstoned_socket, &output_fd, dump_type_enum);
+ {
+ ATRACE_NAME("tombstoned_connect");
+ const DebuggerdDumpType dump_type_enum = static_cast<DebuggerdDumpType>(dump_type);
+ LOG(INFO) << "obtaining output fd from tombstoned, type: " << dump_type_enum;
+ tombstoned_connected = tombstoned_connect(target, &tombstoned_socket, &output_fd, dump_type_enum);
+ }
+
+ // Pause the threads.
+ {
+ ATRACE_NAME("ptrace_interrupt");
+ for (const auto& [sibling_tid, _] : threads) {
+ if (ptrace(PTRACE_INTERRUPT, sibling_tid, 0, 0) != 0) {
+ PLOG(FATAL) << "failed to interrupt thread " << sibling_tid;
+ }
+ }
+ }
// Write a '\1' to stdout to tell the crashing process to resume.
// It also restores the value of PR_SET_DUMPABLE at this point.
@@ -349,9 +378,12 @@
}
siginfo_t siginfo = {};
- if (!wait_for_signal(main_tid, &siginfo)) {
- printf("failed to wait for signal in tid %d: %s\n", main_tid, strerror(errno));
- exit(1);
+ {
+ ATRACE_NAME("wait_for_signal");
+ if (!wait_for_signal(main_tid, &siginfo)) {
+ printf("failed to wait for signal in tid %d: %s\n", main_tid, strerror(errno));
+ exit(1);
+ }
}
int signo = siginfo.si_signo;
@@ -373,8 +405,10 @@
std::string amfd_data;
if (backtrace) {
+ ATRACE_NAME("dump_backtrace");
dump_backtrace(output_fd.get(), backtrace_map.get(), target, main_tid, process_name, threads, 0);
} else {
+ ATRACE_NAME("engrave_tombstone");
engrave_tombstone(output_fd.get(), backtrace_map.get(), &open_files, target, main_tid,
process_name, threads, abort_address, fatal_signal ? &amfd_data : nullptr);
}
diff --git a/debuggerd/debuggerd_test.cpp b/debuggerd/debuggerd_test.cpp
index 9008b95..da8ad37 100644
--- a/debuggerd/debuggerd_test.cpp
+++ b/debuggerd/debuggerd_test.cpp
@@ -89,7 +89,7 @@
} while (0)
static void tombstoned_intercept(pid_t target_pid, unique_fd* intercept_fd, unique_fd* output_fd,
- DebuggerdDumpType intercept_type) {
+ InterceptStatus* status, DebuggerdDumpType intercept_type) {
intercept_fd->reset(socket_local_client(kTombstonedInterceptSocketName,
ANDROID_SOCKET_NAMESPACE_RESERVED, SOCK_SEQPACKET));
if (intercept_fd->get() == -1) {
@@ -136,7 +136,7 @@
<< ", received " << rc;
}
- ASSERT_EQ(InterceptStatus::kRegistered, response.status);
+ *status = response.status;
}
class CrasherTest : public ::testing::Test {
@@ -180,7 +180,9 @@
FAIL() << "crasher hasn't been started";
}
- tombstoned_intercept(crasher_pid, &this->intercept_fd, output_fd, intercept_type);
+ InterceptStatus status;
+ tombstoned_intercept(crasher_pid, &this->intercept_fd, output_fd, &status, intercept_type);
+ ASSERT_EQ(InterceptStatus::kRegistered, status);
}
void CrasherTest::FinishIntercept(int* result) {
@@ -598,7 +600,9 @@
pid_t pid = 123'456'789 + i;
unique_fd intercept_fd, output_fd;
- tombstoned_intercept(pid, &intercept_fd, &output_fd, kDebuggerdTombstone);
+ InterceptStatus status;
+ tombstoned_intercept(pid, &intercept_fd, &output_fd, &status, kDebuggerdTombstone);
+ ASSERT_EQ(InterceptStatus::kRegistered, status);
{
unique_fd tombstoned_socket, input_fd;
@@ -630,7 +634,9 @@
pid_t pid = pid_base + dump;
unique_fd intercept_fd, output_fd;
- tombstoned_intercept(pid, &intercept_fd, &output_fd, kDebuggerdTombstone);
+ InterceptStatus status;
+ tombstoned_intercept(pid, &intercept_fd, &output_fd, &status, kDebuggerdTombstone);
+ ASSERT_EQ(InterceptStatus::kRegistered, status);
// Pretend to crash, and then immediately close the socket.
unique_fd sockfd(socket_local_client(kTombstonedCrashSocketName,
@@ -661,7 +667,9 @@
pid_t pid = pid_base + dump;
unique_fd intercept_fd, output_fd;
- tombstoned_intercept(pid, &intercept_fd, &output_fd, kDebuggerdTombstone);
+ InterceptStatus status;
+ tombstoned_intercept(pid, &intercept_fd, &output_fd, &status, kDebuggerdTombstone);
+ ASSERT_EQ(InterceptStatus::kRegistered, status);
{
unique_fd tombstoned_socket, input_fd;
@@ -685,3 +693,65 @@
thread.join();
}
}
+
+TEST(tombstoned, java_trace_intercept_smoke) {
+ // Using a "real" PID is a little dangerous here - if the test fails
+ // or crashes, we might end up getting a bogus / unreliable stack
+ // trace.
+ const pid_t self = getpid();
+
+ unique_fd intercept_fd, output_fd;
+ InterceptStatus status;
+ tombstoned_intercept(self, &intercept_fd, &output_fd, &status, kDebuggerdJavaBacktrace);
+ ASSERT_EQ(InterceptStatus::kRegistered, status);
+
+ // First connect to tombstoned requesting a native backtrace. This
+ // should result in a "regular" FD and not the installed intercept.
+ const char native[] = "native";
+ unique_fd tombstoned_socket, input_fd;
+ ASSERT_TRUE(tombstoned_connect(self, &tombstoned_socket, &input_fd, kDebuggerdNativeBacktrace));
+ ASSERT_TRUE(android::base::WriteFully(input_fd.get(), native, sizeof(native)));
+ tombstoned_notify_completion(tombstoned_socket.get());
+
+ // Then, connect to tombstoned asking for a java backtrace. This *should*
+ // trigger the intercept.
+ const char java[] = "java";
+ ASSERT_TRUE(tombstoned_connect(self, &tombstoned_socket, &input_fd, kDebuggerdJavaBacktrace));
+ ASSERT_TRUE(android::base::WriteFully(input_fd.get(), java, sizeof(java)));
+ tombstoned_notify_completion(tombstoned_socket.get());
+
+ char outbuf[sizeof(java)];
+ ASSERT_TRUE(android::base::ReadFully(output_fd.get(), outbuf, sizeof(outbuf)));
+ ASSERT_STREQ("java", outbuf);
+}
+
+TEST(tombstoned, multiple_intercepts) {
+ const pid_t fake_pid = 1'234'567;
+ unique_fd intercept_fd, output_fd;
+ InterceptStatus status;
+ tombstoned_intercept(fake_pid, &intercept_fd, &output_fd, &status, kDebuggerdJavaBacktrace);
+ ASSERT_EQ(InterceptStatus::kRegistered, status);
+
+ unique_fd intercept_fd_2, output_fd_2;
+ tombstoned_intercept(fake_pid, &intercept_fd_2, &output_fd_2, &status, kDebuggerdNativeBacktrace);
+ ASSERT_EQ(InterceptStatus::kFailedAlreadyRegistered, status);
+}
+
+TEST(tombstoned, intercept_any) {
+ const pid_t fake_pid = 1'234'567;
+
+ unique_fd intercept_fd, output_fd;
+ InterceptStatus status;
+ tombstoned_intercept(fake_pid, &intercept_fd, &output_fd, &status, kDebuggerdNativeBacktrace);
+ ASSERT_EQ(InterceptStatus::kRegistered, status);
+
+ const char any[] = "any";
+ unique_fd tombstoned_socket, input_fd;
+ ASSERT_TRUE(tombstoned_connect(fake_pid, &tombstoned_socket, &input_fd, kDebuggerdAnyIntercept));
+ ASSERT_TRUE(android::base::WriteFully(input_fd.get(), any, sizeof(any)));
+ tombstoned_notify_completion(tombstoned_socket.get());
+
+ char outbuf[sizeof(any)];
+ ASSERT_TRUE(android::base::ReadFully(output_fd.get(), outbuf, sizeof(outbuf)));
+ ASSERT_STREQ("any", outbuf);
+}
diff --git a/debuggerd/tombstoned/tombstoned.cpp b/debuggerd/tombstoned/tombstoned.cpp
index b4a2bde..b920ebc 100644
--- a/debuggerd/tombstoned/tombstoned.cpp
+++ b/debuggerd/tombstoned/tombstoned.cpp
@@ -165,7 +165,7 @@
};
// Whether java trace dumps are produced via tombstoned.
-static constexpr bool kJavaTraceDumpsEnabled = false;
+static constexpr bool kJavaTraceDumpsEnabled = true;
/* static */ CrashQueue* const CrashQueue::tombstone =
new CrashQueue("/data/tombstones", "tombstone_" /* file_name_prefix */, 10 /* max_artifacts */,
diff --git a/fs_mgr/fs_mgr_fstab.cpp b/fs_mgr/fs_mgr_fstab.cpp
index 70cd608..7a41b14 100644
--- a/fs_mgr/fs_mgr_fstab.cpp
+++ b/fs_mgr/fs_mgr_fstab.cpp
@@ -337,16 +337,8 @@
dirent* dp;
while ((dp = readdir(fstabdir.get())) != NULL) {
- // skip over name and compatible
- if (dp->d_type != DT_DIR) {
- continue;
- }
-
- // skip if its not 'vendor', 'odm' or 'system'
- if (strcmp(dp->d_name, "odm") && strcmp(dp->d_name, "system") &&
- strcmp(dp->d_name, "vendor")) {
- continue;
- }
+ // skip over name, compatible and .
+ if (dp->d_type != DT_DIR || dp->d_name[0] == '.') continue;
// create <dev> <mnt_point> <type> <mnt_flags> <fsmgr_flags>\n
std::vector<std::string> fstab_entry;
diff --git a/fs_mgr/fs_mgr_verity.cpp b/fs_mgr/fs_mgr_verity.cpp
index 5fa10bc..0177fdf 100644
--- a/fs_mgr/fs_mgr_verity.cpp
+++ b/fs_mgr/fs_mgr_verity.cpp
@@ -689,27 +689,55 @@
return read_verity_state(fstab->verity_loc, offset, mode);
}
-static void update_verity_table_blk_device(char *blk_device, char **table)
-{
- std::string result, word;
+// Update the verity table using the actual block device path.
+// Two cases:
+// Case-1: verity table is shared for devices with different by-name prefix.
+// Example:
+// verity table token: /dev/block/bootdevice/by-name/vendor
+// blk_device-1 (non-A/B): /dev/block/platform/soc.0/7824900.sdhci/by-name/vendor
+// blk_device-2 (A/B): /dev/block/platform/soc.0/f9824900.sdhci/by-name/vendor_a
+//
+// Case-2: append A/B suffix in the verity table.
+// Example:
+// verity table token: /dev/block/platform/soc.0/7824900.sdhci/by-name/vendor
+// blk_device: /dev/block/platform/soc.0/7824900.sdhci/by-name/vendor_a
+static void update_verity_table_blk_device(const std::string& blk_device, char** table,
+ bool slot_select) {
+ bool updated = false;
+ std::string result, ab_suffix;
auto tokens = android::base::Split(*table, " ");
+ // If slot_select is set, it means blk_device is already updated with ab_suffix.
+ if (slot_select) ab_suffix = fs_mgr_get_slot_suffix();
+
for (const auto& token : tokens) {
- if (android::base::StartsWith(token, "/dev/block/") &&
- android::base::StartsWith(blk_device, token.c_str())) {
- word = blk_device;
+ std::string new_token;
+ if (android::base::StartsWith(token, "/dev/block/")) {
+ if (token == blk_device) return; // no need to update if they're already the same.
+ std::size_t found1 = blk_device.find("by-name");
+ std::size_t found2 = token.find("by-name");
+ if (found1 != std::string::npos && found2 != std::string::npos &&
+ blk_device.substr(found1) == token.substr(found2) + ab_suffix) {
+ new_token = blk_device;
+ }
+ }
+
+ if (!new_token.empty()) {
+ updated = true;
+ LINFO << "Verity table: updated block device from '" << token << "' to '" << new_token
+ << "'";
} else {
- word = token;
+ new_token = token;
}
if (result.empty()) {
- result = word;
+ result = new_token;
} else {
- result += " " + word;
+ result += " " + new_token;
}
}
- if (result.empty()) {
+ if (!updated) {
return;
}
@@ -825,10 +853,9 @@
LINFO << "Enabling dm-verity for " << mount_point.c_str()
<< " (mode " << params.mode << ")";
- if (fstab->fs_mgr_flags & MF_SLOTSELECT) {
- // Update the verity params using the actual block device path
- update_verity_table_blk_device(fstab->blk_device, ¶ms.table);
- }
+ // Update the verity params using the actual block device path
+ update_verity_table_blk_device(fstab->blk_device, ¶ms.table,
+ fstab->fs_mgr_flags & MF_SLOTSELECT);
// load the verity mapping table
if (load_verity_table(io, mount_point, verity.data_size, fd, ¶ms,
diff --git a/init/devices.cpp b/init/devices.cpp
index 40f9740..c52d8f8 100644
--- a/init/devices.cpp
+++ b/init/devices.cpp
@@ -175,7 +175,7 @@
if (s.MatchWithSubsystem(path, subsystem)) s.SetPermissions(path);
}
- if (access(path.c_str(), F_OK) == 0) {
+ if (!skip_restorecon_ && access(path.c_str(), F_OK) == 0) {
LOG(VERBOSE) << "restorecon_recursive: " << path;
if (selinux_android_restorecon(path.c_str(), SELINUX_ANDROID_RESTORECON_RECURSE) != 0) {
PLOG(ERROR) << "selinux_android_restorecon(" << path << ") failed";
@@ -467,12 +467,13 @@
DeviceHandler::DeviceHandler(std::vector<Permissions> dev_permissions,
std::vector<SysfsPermissions> sysfs_permissions,
- std::vector<Subsystem> subsystems)
+ std::vector<Subsystem> subsystems, bool skip_restorecon)
: dev_permissions_(std::move(dev_permissions)),
sysfs_permissions_(std::move(sysfs_permissions)),
subsystems_(std::move(subsystems)),
- sehandle_(selinux_android_file_context_handle()) {}
+ sehandle_(selinux_android_file_context_handle()),
+ skip_restorecon_(skip_restorecon) {}
DeviceHandler::DeviceHandler()
: DeviceHandler(std::vector<Permissions>{}, std::vector<SysfsPermissions>{},
- std::vector<Subsystem>{}) {}
+ std::vector<Subsystem>{}, false) {}
diff --git a/init/devices.h b/init/devices.h
index 50f49fc..09a0ce3 100644
--- a/init/devices.h
+++ b/init/devices.h
@@ -114,14 +114,21 @@
DeviceHandler();
DeviceHandler(std::vector<Permissions> dev_permissions,
std::vector<SysfsPermissions> sysfs_permissions,
- std::vector<Subsystem> subsystems);
+ std::vector<Subsystem> subsystems, bool skip_restorecon);
~DeviceHandler(){};
void HandleDeviceEvent(const Uevent& uevent);
+
+ void FixupSysPermissions(const std::string& upath, const std::string& subsystem) const;
+
+ void HandlePlatformDeviceEvent(const Uevent& uevent);
+ void HandleBlockDeviceEvent(const Uevent& uevent) const;
+ void HandleGenericDeviceEvent(const Uevent& uevent) const;
+
std::vector<std::string> GetBlockDeviceSymlinks(const Uevent& uevent) const;
+ void set_skip_restorecon(bool value) { skip_restorecon_ = value; }
private:
- void FixupSysPermissions(const std::string& upath, const std::string& subsystem) const;
std::tuple<mode_t, uid_t, gid_t> GetDevicePermissions(
const std::string& path, const std::vector<std::string>& links) const;
void MakeDevice(const std::string& path, int block, int major, int minor,
@@ -129,15 +136,13 @@
std::vector<std::string> GetCharacterDeviceSymlinks(const Uevent& uevent) const;
void HandleDevice(const std::string& action, const std::string& devpath, int block, int major,
int minor, const std::vector<std::string>& links) const;
- void HandlePlatformDeviceEvent(const Uevent& uevent);
- void HandleBlockDeviceEvent(const Uevent& uevent) const;
- void HandleGenericDeviceEvent(const Uevent& uevent) const;
std::vector<Permissions> dev_permissions_;
std::vector<SysfsPermissions> sysfs_permissions_;
std::vector<Subsystem> subsystems_;
PlatformDeviceList platform_devices_;
selabel_handle* sehandle_;
+ bool skip_restorecon_;
};
// Exposed for testing
diff --git a/init/firmware_handler.cpp b/init/firmware_handler.cpp
index 1471aeb..844c605 100644
--- a/init/firmware_handler.cpp
+++ b/init/firmware_handler.cpp
@@ -18,6 +18,7 @@
#include <fcntl.h>
#include <sys/sendfile.h>
+#include <sys/wait.h>
#include <unistd.h>
#include <string>
@@ -103,14 +104,29 @@
if (uevent.subsystem != "firmware" || uevent.action != "add") return;
// Loading the firmware in a child means we can do that in parallel...
- // (We ignore SIGCHLD rather than wait for our children.)
+ // We double fork instead of waiting for these processes.
pid_t pid = fork();
- if (pid == 0) {
- Timer t;
- ProcessFirmwareEvent(uevent);
- LOG(INFO) << "loading " << uevent.path << " took " << t;
- _exit(EXIT_SUCCESS);
- } else if (pid == -1) {
+ if (pid == -1) {
PLOG(ERROR) << "could not fork to process firmware event for " << uevent.firmware;
+ return;
}
+
+ if (pid == 0) {
+ pid = fork();
+ if (pid == -1) {
+ PLOG(ERROR) << "could not fork a sceond time to process firmware event for "
+ << uevent.firmware;
+ _exit(EXIT_FAILURE);
+ }
+ if (pid == 0) {
+ Timer t;
+ ProcessFirmwareEvent(uevent);
+ LOG(INFO) << "loading " << uevent.path << " took " << t;
+ _exit(EXIT_SUCCESS);
+ }
+
+ _exit(EXIT_SUCCESS);
+ }
+
+ waitpid(pid, nullptr, 0);
}
diff --git a/init/init_first_stage.cpp b/init/init_first_stage.cpp
index 8a7d9a2..0f2b1f3 100644
--- a/init/init_first_stage.cpp
+++ b/init/init_first_stage.cpp
@@ -60,9 +60,8 @@
virtual bool SetUpDmVerity(fstab_rec* fstab_rec) = 0;
bool need_dm_verity_;
- // Device tree fstab entries.
+
std::unique_ptr<fstab, decltype(&fs_mgr_free_fstab)> device_tree_fstab_;
- // Eligible first stage mount candidates, only allow /system, /vendor and/or /odm.
std::vector<fstab_rec*> mount_fstab_recs_;
std::set<std::string> required_devices_partition_names_;
DeviceHandler device_handler_;
@@ -115,12 +114,10 @@
LOG(ERROR) << "Failed to read fstab from device tree";
return;
}
- for (auto mount_point : {"/system", "/vendor", "/odm"}) {
- fstab_rec* fstab_rec =
- fs_mgr_get_entry_for_mount_point(device_tree_fstab_.get(), mount_point);
- if (fstab_rec != nullptr) {
- mount_fstab_recs_.push_back(fstab_rec);
- }
+ // Stores device_tree_fstab_->recs[] into mount_fstab_recs_ (vector<fstab_rec*>)
+ // for easier manipulation later, e.g., range-base for loop.
+ for (int i = 0; i < device_tree_fstab_->num_entries; i++) {
+ mount_fstab_recs_.push_back(&device_tree_fstab_->recs[i]);
}
}
@@ -420,7 +417,7 @@
// Public functions
// ----------------
-// Mounts /system, /vendor, and/or /odm if they are present in the fstab provided by device tree.
+// Mounts partitions specified by fstab in device tree.
bool DoFirstStageMount() {
// Skips first stage mount if we're in recovery mode.
if (IsRecoveryMode()) {
diff --git a/init/uevent_listener.cpp b/init/uevent_listener.cpp
index 27c5d23..01b8250 100644
--- a/init/uevent_listener.cpp
+++ b/init/uevent_listener.cpp
@@ -165,7 +165,7 @@
return RegenerateUeventsForDir(d.get(), callback);
}
-static const char* kRegenerationPaths[] = {"/sys/class", "/sys/block", "/sys/devices"};
+const char* kRegenerationPaths[] = {"/sys/class", "/sys/block", "/sys/devices"};
void UeventListener::RegenerateUevents(RegenerateCallback callback) const {
for (const auto path : kRegenerationPaths) {
diff --git a/init/uevent_listener.h b/init/uevent_listener.h
index ba31aaa..8e6f3b4 100644
--- a/init/uevent_listener.h
+++ b/init/uevent_listener.h
@@ -35,6 +35,8 @@
using RegenerateCallback = std::function<RegenerationAction(const Uevent&)>;
using PollCallback = std::function<void(const Uevent&)>;
+extern const char* kRegenerationPaths[3];
+
class UeventListener {
public:
UeventListener();
diff --git a/init/ueventd.cpp b/init/ueventd.cpp
index bd21a3e..31e4106 100644
--- a/init/ueventd.cpp
+++ b/init/ueventd.cpp
@@ -22,10 +22,15 @@
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
+#include <sys/wait.h>
+
+#include <set>
+#include <thread>
#include <android-base/logging.h>
#include <android-base/properties.h>
#include <android-base/stringprintf.h>
+#include <selinux/android.h>
#include <selinux/selinux.h>
#include "devices.h"
@@ -35,6 +40,194 @@
#include "ueventd_parser.h"
#include "util.h"
+// At a high level, ueventd listens for uevent messages generated by the kernel through a netlink
+// socket. When ueventd receives such a message it handles it by taking appropriate actions,
+// which can typically be creating a device node in /dev, setting file permissions, setting selinux
+// labels, etc.
+// Ueventd also handles loading of firmware that the kernel requests, and creates symlinks for block
+// and character devices.
+
+// When ueventd starts, it regenerates uevents for all currently registered devices by traversing
+// /sys and writing 'add' to each 'uevent' file that it finds. This causes the kernel to generate
+// and resend uevent messages for all of the currently registered devices. This is done, because
+// ueventd would not have been running when these devices were registered and therefore was unable
+// to receive their uevent messages and handle them appropriately. This process is known as
+// 'cold boot'.
+
+// 'init' currently waits synchronously on the cold boot process of ueventd before it continues
+// its boot process. For this reason, cold boot should be as quick as possible. One way to achieve
+// a speed up here is to parallelize the handling of ueventd messages, which consume the bulk of the
+// time during cold boot.
+
+// Handling of uevent messages has two unique properties:
+// 1) It can be done in isolation; it doesn't need to read or write any status once it is started.
+// 2) It uses setegid() and setfscreatecon() so either care (aka locking) must be taken to ensure
+// that no file system operations are done while the uevent process has an abnormal egid or
+// fscreatecon or this handling must happen in a separate process.
+// Given the above two properties, it is best to fork() subprocesses to handle the uevents. This
+// reduces the overhead and complexity that would be required in a solution with threads and locks.
+// In testing, a racy multithreaded solution has the same performance as the fork() solution, so
+// there is no reason to deal with the complexity of the former.
+
+// One other important caveat during the boot process is the handling of SELinux restorecon.
+// Since many devices have child devices, calling selinux_android_restorecon() recursively for each
+// device when its uevent is handled, results in multiple restorecon operations being done on a
+// given file. It is more efficient to simply do restorecon recursively on /sys during cold boot,
+// than to do restorecon on each device as its uevent is handled. This only applies to cold boot;
+// once that has completed, restorecon is done for each device as its uevent is handled.
+
+// With all of the above considered, the cold boot process has the below steps:
+// 1) ueventd regenerates uevents by doing the /sys traversal and listens to the netlink socket for
+// the generated uevents. It writes these uevents into a queue represented by a vector.
+//
+// 2) ueventd forks 'n' separate uevent handler subprocesses and has each of them to handle the
+// uevents in the queue based on a starting offset (their process number) and a stride (the total
+// number of processes). Note that no IPC happens at this point and only const functions from
+// DeviceHandler should be called from this context.
+//
+// 3) In parallel to the subprocesses handling the uevents, the main thread of ueventd calls
+// selinux_android_restorecon() recursively on /sys/class, /sys/block, and /sys/devices.
+//
+// 4) Once the restorecon operation finishes, the main thread calls waitpid() to wait for all
+// subprocess handlers to complete and exit. Once this happens, it marks coldboot as having
+// completed.
+//
+// At this point, ueventd is single threaded, poll()'s and then handles any future uevents.
+
+// Lastly, it should be noted that uevents that occur during the coldboot process are handled
+// without issue after the coldboot process completes. This is because the uevent listener is
+// paused while the uevent handler and restorecon actions take place. Once coldboot completes,
+// the uevent listener resumes in polling mode and will handle the uevents that occurred during
+// coldboot.
+
+class ColdBoot {
+ public:
+ ColdBoot(UeventListener& uevent_listener, DeviceHandler& device_handler)
+ : uevent_listener_(uevent_listener),
+ device_handler_(device_handler),
+ num_handler_subprocesses_(std::thread::hardware_concurrency() ?: 4) {}
+
+ void Run();
+
+ private:
+ void UeventHandlerMain(unsigned int process_num, unsigned int total_processes);
+ void RegenerateUevents();
+ void ForkSubProcesses();
+ void DoRestoreCon();
+ void WaitForSubProcesses();
+
+ UeventListener& uevent_listener_;
+ DeviceHandler& device_handler_;
+
+ unsigned int num_handler_subprocesses_;
+ std::vector<Uevent> uevent_queue_;
+
+ std::set<pid_t> subprocess_pids_;
+};
+
+void ColdBoot::UeventHandlerMain(unsigned int process_num, unsigned int total_processes) {
+ for (unsigned int i = process_num; i < uevent_queue_.size(); i += total_processes) {
+ auto& uevent = uevent_queue_[i];
+ if (uevent.action == "add" || uevent.action == "change" || uevent.action == "online") {
+ device_handler_.FixupSysPermissions(uevent.path, uevent.subsystem);
+ }
+
+ if (uevent.subsystem == "block") {
+ device_handler_.HandleBlockDeviceEvent(uevent);
+ } else {
+ device_handler_.HandleGenericDeviceEvent(uevent);
+ }
+ }
+ _exit(EXIT_SUCCESS);
+}
+
+void ColdBoot::RegenerateUevents() {
+ uevent_listener_.RegenerateUevents([this](const Uevent& uevent) {
+ HandleFirmwareEvent(uevent);
+
+ // This is the one mutable part of DeviceHandler, in which platform devices are
+ // added to a vector for later reference. Since there is no communication after
+ // fork()'ing subprocess handlers, all platform devices must be in the vector before
+ // we fork, and therefore they must be handled in this loop.
+ if (uevent.subsystem == "platform") {
+ device_handler_.HandlePlatformDeviceEvent(uevent);
+ }
+
+ uevent_queue_.emplace_back(std::move(uevent));
+ return RegenerationAction::kContinue;
+ });
+}
+
+void ColdBoot::ForkSubProcesses() {
+ for (unsigned int i = 0; i < num_handler_subprocesses_; ++i) {
+ auto pid = fork();
+ if (pid < 0) {
+ PLOG(FATAL) << "fork() failed!";
+ }
+
+ if (pid == 0) {
+ UeventHandlerMain(i, num_handler_subprocesses_);
+ }
+
+ subprocess_pids_.emplace(pid);
+ }
+}
+
+void ColdBoot::DoRestoreCon() {
+ for (const char* path : kRegenerationPaths) {
+ selinux_android_restorecon(path, SELINUX_ANDROID_RESTORECON_RECURSE);
+ }
+ device_handler_.set_skip_restorecon(false);
+}
+
+void ColdBoot::WaitForSubProcesses() {
+ // Treat subprocesses that crash or get stuck the same as if ueventd itself has crashed or gets
+ // stuck.
+ //
+ // When a subprocess crashes, we fatally abort from ueventd. init will restart ueventd when
+ // init reaps it, and the cold boot process will start again. If this continues to fail, then
+ // since ueventd is marked as a critical service, init will reboot to recovery.
+ //
+ // When a subprocess gets stuck, keep ueventd spinning waiting for it. init has a timeout for
+ // cold boot and will reboot to the bootloader if ueventd does not complete in time.
+ while (!subprocess_pids_.empty()) {
+ int status;
+ pid_t pid = TEMP_FAILURE_RETRY(waitpid(-1, &status, 0));
+ if (pid == -1) {
+ PLOG(ERROR) << "waitpid() failed";
+ continue;
+ }
+
+ auto it = std::find(subprocess_pids_.begin(), subprocess_pids_.end(), pid);
+ if (it == subprocess_pids_.end()) continue;
+
+ if (WIFEXITED(status)) {
+ if (WEXITSTATUS(status) == EXIT_SUCCESS) {
+ subprocess_pids_.erase(it);
+ } else {
+ LOG(FATAL) << "subprocess exited with status " << WEXITSTATUS(status);
+ }
+ } else if (WIFSIGNALED(status)) {
+ LOG(FATAL) << "subprocess killed by signal " << WTERMSIG(status);
+ }
+ }
+}
+
+void ColdBoot::Run() {
+ Timer cold_boot_timer;
+
+ RegenerateUevents();
+
+ ForkSubProcesses();
+
+ DoRestoreCon();
+
+ WaitForSubProcesses();
+
+ close(open(COLDBOOT_DONE, O_WRONLY | O_CREAT | O_CLOEXEC, 0000));
+ LOG(INFO) << "Coldboot took " << cold_boot_timer;
+}
+
DeviceHandler CreateDeviceHandler() {
Parser parser;
@@ -64,11 +257,10 @@
parser.ParseConfig("/ueventd." + hardware + ".rc");
return DeviceHandler(std::move(dev_permissions), std::move(sysfs_permissions),
- std::move(subsystems));
+ std::move(subsystems), true);
}
-int ueventd_main(int argc, char **argv)
-{
+int ueventd_main(int argc, char** argv) {
/*
* init sets the umask to 077 for forked processes. We need to
* create files with exact permissions, without modification by
@@ -76,13 +268,6 @@
*/
umask(000);
- /* Prevent fire-and-forget children from becoming zombies.
- * If we should need to wait() for some children in the future
- * (as opposed to none right now), double-forking here instead
- * of ignoring SIGCHLD may be the better solution.
- */
- signal(SIGCHLD, SIG_IGN);
-
InitKernelLogging(argv);
LOG(INFO) << "ueventd started!";
@@ -95,16 +280,8 @@
UeventListener uevent_listener;
if (access(COLDBOOT_DONE, F_OK) != 0) {
- Timer t;
-
- uevent_listener.RegenerateUevents([&device_handler](const Uevent& uevent) {
- HandleFirmwareEvent(uevent);
- device_handler.HandleDeviceEvent(uevent);
- return RegenerationAction::kContinue;
- });
-
- close(open(COLDBOOT_DONE, O_WRONLY | O_CREAT | O_CLOEXEC, 0000));
- LOG(INFO) << "Coldboot took " << t;
+ ColdBoot cold_boot(uevent_listener, device_handler);
+ cold_boot.Run();
}
uevent_listener.DoPolling([&device_handler](const Uevent& uevent) {
diff --git a/libbacktrace/Android.bp b/libbacktrace/Android.bp
index 7dd9227..a643a29 100644
--- a/libbacktrace/Android.bp
+++ b/libbacktrace/Android.bp
@@ -113,6 +113,7 @@
static_libs: ["libasync_safe", "libcutils"],
},
},
+ whole_static_libs: ["libdemangle"],
}
cc_library_shared {
diff --git a/libbacktrace/Backtrace.cpp b/libbacktrace/Backtrace.cpp
index 3545661..e46d353 100644
--- a/libbacktrace/Backtrace.cpp
+++ b/libbacktrace/Backtrace.cpp
@@ -27,6 +27,8 @@
#include <backtrace/Backtrace.h>
#include <backtrace/BacktraceMap.h>
+#include <demangle.h>
+
#include "BacktraceLog.h"
#include "thread_utils.h"
#include "UnwindCurrent.h"
@@ -62,8 +64,7 @@
if (map->start == 0 || (map->flags & PROT_DEVICE_MAP)) {
return "";
}
- std::string func_name = GetFunctionNameRaw(pc, offset);
- return func_name;
+ return demangle(GetFunctionNameRaw(pc, offset).c_str());
}
bool Backtrace::VerifyReadWordArgs(uintptr_t ptr, word_t* out_value) {
diff --git a/libcutils/sched_policy.cpp b/libcutils/sched_policy.cpp
index e29a844..7e38535 100644
--- a/libcutils/sched_policy.cpp
+++ b/libcutils/sched_policy.cpp
@@ -340,7 +340,7 @@
return 0;
}
-static void set_timerslack_ns(int tid, unsigned long long slack) {
+static void set_timerslack_ns(int tid, unsigned long slack) {
// v4.6+ kernels support the /proc/<tid>/timerslack_ns interface.
// TODO: once we've backported this, log if the open(2) fails.
if (__sys_supports_timerslack) {
@@ -348,7 +348,7 @@
snprintf(buf, sizeof(buf), "/proc/%d/timerslack_ns", tid);
int fd = open(buf, O_WRONLY | O_CLOEXEC);
if (fd != -1) {
- int len = snprintf(buf, sizeof(buf), "%llu", slack);
+ int len = snprintf(buf, sizeof(buf), "%lu", slack);
if (write(fd, buf, len) != len) {
SLOGE("set_timerslack_ns write failed: %s\n", strerror(errno));
}
diff --git a/liblog/include/log/log_event_list.h b/liblog/include/log/log_event_list.h
index 057be5d..bb1ce34 100644
--- a/liblog/include/log/log_event_list.h
+++ b/liblog/include/log/log_event_list.h
@@ -177,6 +177,12 @@
return *this;
}
+ android_log_event_list& operator<<(bool value) {
+ int retval = android_log_write_int32(ctx, value ? 1 : 0);
+ if (retval < 0) ret = retval;
+ return *this;
+ }
+
android_log_event_list& operator<<(int64_t value) {
int retval = android_log_write_int64(ctx, value);
if (retval < 0) ret = retval;