liblog: use a rwlock for writer initialization
The current system of using atomics isn't thread safe and may result
in doubly closing FDs or closing actively used FDs. The safest way to
do this is to use a rwlock, which should not have a much higher
overhead than the atomics do, as a vast majority of the time, there
will not be writers.
This moves us further away from using the transport interface, which
will be removed. Each writer should be self contained, without a
separate open or available function.
Also, keep the pmsg fd open if it is opened by
__android_log_pmsg_file_write(). This fd was closed due to issues
with zygote, but it looks like it is only called by recovery now, so
there is no reason to close this fd at the end of that function.
Test: logging works, liblog-unit-tests
Change-Id: I345c9a5d18c55b11a280c8362df854784abf46fd
diff --git a/liblog/logd_writer.cpp b/liblog/logd_writer.cpp
index a22c3be..283a979 100644
--- a/liblog/logd_writer.cpp
+++ b/liblog/logd_writer.cpp
@@ -30,97 +30,80 @@
#include <time.h>
#include <unistd.h>
+#include <shared_mutex>
+
#include <cutils/sockets.h>
#include <private/android_filesystem_config.h>
#include <private/android_logger.h>
#include "log_portability.h"
#include "logger.h"
+#include "rwlock.h"
#include "uio.h"
-static int logdAvailable(log_id_t LogId);
-static int logdOpen();
-static void logdClose();
-static int logdWrite(log_id_t logId, struct timespec* ts, struct iovec* vec, size_t nr);
+static int LogdWrite(log_id_t logId, struct timespec* ts, struct iovec* vec, size_t nr);
+static void LogdClose();
struct android_log_transport_write logdLoggerWrite = {
.name = "logd",
.logMask = 0,
- .context.sock = -EBADF,
- .available = logdAvailable,
- .open = logdOpen,
- .close = logdClose,
- .write = logdWrite,
+ .available = [](log_id_t) { return 0; },
+ .open = [] { return 0; },
+ .close = LogdClose,
+ .write = LogdWrite,
};
-/* log_init_lock assumed */
-static int logdOpen() {
- int i, ret = 0;
+static int logd_socket;
+static RwLock logd_socket_lock;
- i = atomic_load(&logdLoggerWrite.context.sock);
- if (i < 0) {
- int sock = TEMP_FAILURE_RETRY(socket(PF_UNIX, SOCK_DGRAM | SOCK_CLOEXEC | SOCK_NONBLOCK, 0));
- if (sock < 0) {
- ret = -errno;
- } else {
- struct sockaddr_un un;
- memset(&un, 0, sizeof(struct sockaddr_un));
- un.sun_family = AF_UNIX;
- strcpy(un.sun_path, "/dev/socket/logdw");
-
- if (TEMP_FAILURE_RETRY(connect(sock, (struct sockaddr*)&un, sizeof(struct sockaddr_un))) <
- 0) {
- ret = -errno;
- switch (ret) {
- case -ENOTCONN:
- case -ECONNREFUSED:
- case -ENOENT:
- i = atomic_exchange(&logdLoggerWrite.context.sock, ret);
- [[fallthrough]];
- default:
- break;
- }
- close(sock);
- } else {
- ret = atomic_exchange(&logdLoggerWrite.context.sock, sock);
- if ((ret >= 0) && (ret != sock)) {
- close(ret);
- }
- ret = 0;
- }
- }
+static void OpenSocketLocked() {
+ logd_socket = TEMP_FAILURE_RETRY(socket(PF_UNIX, SOCK_DGRAM | SOCK_CLOEXEC | SOCK_NONBLOCK, 0));
+ if (logd_socket <= 0) {
+ return;
}
- return ret;
-}
+ sockaddr_un un = {};
+ un.sun_family = AF_UNIX;
+ strcpy(un.sun_path, "/dev/socket/logdw");
-static void __logdClose(int negative_errno) {
- int sock = atomic_exchange(&logdLoggerWrite.context.sock, negative_errno);
- if (sock >= 0) {
- close(sock);
+ if (TEMP_FAILURE_RETRY(
+ connect(logd_socket, reinterpret_cast<sockaddr*>(&un), sizeof(sockaddr_un))) < 0) {
+ close(logd_socket);
+ logd_socket = 0;
}
}
-static void logdClose() {
- __logdClose(-EBADF);
+static void OpenSocket() {
+ auto lock = std::unique_lock{logd_socket_lock};
+ if (logd_socket > 0) {
+ // Someone raced us and opened the socket already.
+ return;
+ }
+
+ OpenSocketLocked();
}
-static int logdAvailable(log_id_t logId) {
- if (logId >= LOG_ID_MAX || logId == LOG_ID_KERNEL) {
- return -EINVAL;
+static void ResetSocket(int old_socket) {
+ auto lock = std::unique_lock{logd_socket_lock};
+ if (old_socket != logd_socket) {
+ // Someone raced us and reset the socket already.
+ return;
}
- if (atomic_load(&logdLoggerWrite.context.sock) < 0) {
- if (access("/dev/socket/logdw", W_OK) == 0) {
- return 0;
- }
- return -EBADF;
- }
- return 1;
+ close(logd_socket);
+ logd_socket = 0;
+ OpenSocketLocked();
}
-static int logdWrite(log_id_t logId, struct timespec* ts, struct iovec* vec, size_t nr) {
+static void LogdClose() {
+ auto lock = std::unique_lock{logd_socket_lock};
+ if (logd_socket > 0) {
+ close(logd_socket);
+ }
+ logd_socket = 0;
+}
+
+static int LogdWrite(log_id_t logId, struct timespec* ts, struct iovec* vec, size_t nr) {
ssize_t ret;
- int sock;
static const unsigned headerLength = 1;
struct iovec newVec[nr + headerLength];
android_log_header_t header;
@@ -128,15 +111,16 @@
static atomic_int dropped;
static atomic_int droppedSecurity;
- sock = atomic_load(&logdLoggerWrite.context.sock);
- if (sock < 0) switch (sock) {
- case -ENOTCONN:
- case -ECONNREFUSED:
- case -ENOENT:
- break;
- default:
- return -EBADF;
- }
+ auto lock = std::shared_lock{logd_socket_lock};
+ if (logd_socket <= 0) {
+ lock.unlock();
+ OpenSocket();
+ lock.lock();
+ }
+
+ if (logd_socket <= 0) {
+ return -EBADF;
+ }
/* logd, after initialization and priv drop */
if (__android_log_uid() == AID_LOGD) {
@@ -155,41 +139,39 @@
newVec[0].iov_base = (unsigned char*)&header;
newVec[0].iov_len = sizeof(header);
- if (sock >= 0) {
- int32_t snapshot = atomic_exchange_explicit(&droppedSecurity, 0, memory_order_relaxed);
- if (snapshot) {
- android_log_event_int_t buffer;
+ int32_t snapshot = atomic_exchange_explicit(&droppedSecurity, 0, memory_order_relaxed);
+ if (snapshot) {
+ android_log_event_int_t buffer;
- header.id = LOG_ID_SECURITY;
- buffer.header.tag = LIBLOG_LOG_TAG;
- buffer.payload.type = EVENT_TYPE_INT;
- buffer.payload.data = snapshot;
+ header.id = LOG_ID_SECURITY;
+ buffer.header.tag = LIBLOG_LOG_TAG;
+ buffer.payload.type = EVENT_TYPE_INT;
+ buffer.payload.data = snapshot;
- newVec[headerLength].iov_base = &buffer;
- newVec[headerLength].iov_len = sizeof(buffer);
+ newVec[headerLength].iov_base = &buffer;
+ newVec[headerLength].iov_len = sizeof(buffer);
- ret = TEMP_FAILURE_RETRY(writev(sock, newVec, 2));
- if (ret != (ssize_t)(sizeof(header) + sizeof(buffer))) {
- atomic_fetch_add_explicit(&droppedSecurity, snapshot, memory_order_relaxed);
- }
+ ret = TEMP_FAILURE_RETRY(writev(logd_socket, newVec, 2));
+ if (ret != (ssize_t)(sizeof(header) + sizeof(buffer))) {
+ atomic_fetch_add_explicit(&droppedSecurity, snapshot, memory_order_relaxed);
}
- snapshot = atomic_exchange_explicit(&dropped, 0, memory_order_relaxed);
- if (snapshot && __android_log_is_loggable_len(ANDROID_LOG_INFO, "liblog", strlen("liblog"),
- ANDROID_LOG_VERBOSE)) {
- android_log_event_int_t buffer;
+ }
+ snapshot = atomic_exchange_explicit(&dropped, 0, memory_order_relaxed);
+ if (snapshot && __android_log_is_loggable_len(ANDROID_LOG_INFO, "liblog", strlen("liblog"),
+ ANDROID_LOG_VERBOSE)) {
+ android_log_event_int_t buffer;
- header.id = LOG_ID_EVENTS;
- buffer.header.tag = LIBLOG_LOG_TAG;
- buffer.payload.type = EVENT_TYPE_INT;
- buffer.payload.data = snapshot;
+ header.id = LOG_ID_EVENTS;
+ buffer.header.tag = LIBLOG_LOG_TAG;
+ buffer.payload.type = EVENT_TYPE_INT;
+ buffer.payload.data = snapshot;
- newVec[headerLength].iov_base = &buffer;
- newVec[headerLength].iov_len = sizeof(buffer);
+ newVec[headerLength].iov_base = &buffer;
+ newVec[headerLength].iov_len = sizeof(buffer);
- ret = TEMP_FAILURE_RETRY(writev(sock, newVec, 2));
- if (ret != (ssize_t)(sizeof(header) + sizeof(buffer))) {
- atomic_fetch_add_explicit(&dropped, snapshot, memory_order_relaxed);
- }
+ ret = TEMP_FAILURE_RETRY(writev(logd_socket, newVec, 2));
+ if (ret != (ssize_t)(sizeof(header) + sizeof(buffer))) {
+ atomic_fetch_add_explicit(&dropped, snapshot, memory_order_relaxed);
}
}
@@ -208,49 +190,26 @@
}
}
- /*
- * The write below could be lost, but will never block.
- *
- * ENOTCONN occurs if logd has died.
- * ENOENT occurs if logd is not running and socket is missing.
- * ECONNREFUSED occurs if we can not reconnect to logd.
- * EAGAIN occurs if logd is overloaded.
- */
- if (sock < 0) {
- ret = sock;
- } else {
- ret = TEMP_FAILURE_RETRY(writev(sock, newVec, i));
- if (ret < 0) {
- ret = -errno;
- }
+ // The write below could be lost, but will never block.
+ // EAGAIN occurs if logd is overloaded, other errors indicate that something went wrong with
+ // the connection, so we reset it and try again.
+ ret = TEMP_FAILURE_RETRY(writev(logd_socket, newVec, i));
+ if (ret < 0 && errno != EAGAIN) {
+ int old_socket = logd_socket;
+ lock.unlock();
+ ResetSocket(old_socket);
+ lock.lock();
+
+ ret = TEMP_FAILURE_RETRY(writev(logd_socket, newVec, i));
}
- switch (ret) {
- case -ENOTCONN:
- case -ECONNREFUSED:
- case -ENOENT:
- if (__android_log_trylock()) {
- return ret; /* in a signal handler? try again when less stressed */
- }
- __logdClose(ret);
- ret = logdOpen();
- __android_log_unlock();
- if (ret < 0) {
- return ret;
- }
-
- ret = TEMP_FAILURE_RETRY(writev(atomic_load(&logdLoggerWrite.context.sock), newVec, i));
- if (ret < 0) {
- ret = -errno;
- }
- [[fallthrough]];
- default:
- break;
+ if (ret < 0) {
+ ret = -errno;
}
if (ret > (ssize_t)sizeof(header)) {
ret -= sizeof(header);
- } else if (ret == -EAGAIN) {
+ } else if (ret < 0) {
atomic_fetch_add_explicit(&dropped, 1, memory_order_relaxed);
if (logId == LOG_ID_SECURITY) {
atomic_fetch_add_explicit(&droppedSecurity, 1, memory_order_relaxed);