Merge changes I12d6aa4e,I9e802113,I2b30e0fc,Iaf387b9e
* changes:
logcat: remove dead label code
logcat: do not stop on unexpected log ID
Revert "logd: Add minimum time bucket statistics"
liblog: Instrument logging of logd write drops
diff --git a/liblog/Android.mk b/liblog/Android.mk
index 5756c54..9353546 100644
--- a/liblog/Android.mk
+++ b/liblog/Android.mk
@@ -16,6 +16,8 @@
LOCAL_PATH := $(my-dir)
include $(CLEAR_VARS)
+liblog_cflags := $(shell sed -n 's/^\([0-9]*\)[ \t]*liblog[ \t].*/-DLIBLOG_LOG_TAG=\1/p' $(LOCAL_PATH)/event.logtags)
+
ifneq ($(TARGET_USES_LOGD),false)
liblog_sources := logd_write.c
else
@@ -45,7 +47,7 @@
uio.c
endif
-liblog_host_sources := $(liblog_sources) fake_log_device.c
+liblog_host_sources := $(liblog_sources) fake_log_device.c event.logtags
liblog_target_sources := $(liblog_sources) log_time.cpp log_is_loggable.c
ifneq ($(TARGET_USES_LOGD),false)
liblog_target_sources += log_read.c
@@ -57,7 +59,7 @@
# ========================================================
LOCAL_MODULE := liblog
LOCAL_SRC_FILES := $(liblog_host_sources)
-LOCAL_CFLAGS := -DFAKE_LOG_DEVICE=1 -Werror
+LOCAL_CFLAGS := -DFAKE_LOG_DEVICE=1 -Werror $(liblog_cflags)
LOCAL_MULTILIB := both
include $(BUILD_HOST_STATIC_LIBRARY)
@@ -76,13 +78,13 @@
include $(CLEAR_VARS)
LOCAL_MODULE := liblog
LOCAL_SRC_FILES := $(liblog_target_sources)
-LOCAL_CFLAGS := -Werror
+LOCAL_CFLAGS := -Werror $(liblog_cflags)
include $(BUILD_STATIC_LIBRARY)
include $(CLEAR_VARS)
LOCAL_MODULE := liblog
LOCAL_WHOLE_STATIC_LIBRARIES := liblog
-LOCAL_CFLAGS := -Werror
+LOCAL_CFLAGS := -Werror $(liblog_cflags)
include $(BUILD_SHARED_LIBRARY)
include $(call first-makefiles-under,$(LOCAL_PATH))
diff --git a/liblog/README b/liblog/README
index 461dfbe..f29ac04 100644
--- a/liblog/README
+++ b/liblog/README
@@ -131,6 +131,33 @@
when opening the sub-log. It is recommended to open the log
ANDROID_LOG_RDONLY in these cases.
+ERRORS
+ If messages fail, a negative error code will be returned to the caller.
+
+ The -ENOTCONN return code indicates that the logger daemon is stopped.
+
+ The -EBADF return code indicates that the log access point can not be
+ opened, or the log buffer id is out of range.
+
+ For the -EAGAIN return code, this means that the logging message was
+ temporarily backed-up either because of Denial Of Service (DOS) logging
+ pressure from some chatty application or service in the Android system,
+ or if too small of a value is set in /proc/sys/net/unix/max_dgram_qlen.
+ To aid in diagnosing the occurence of this, a binary event from liblog
+ will be sent to the log daemon once a new message can get through
+ indicating how many messages were dropped as a result. Please take
+ action to resolve the structural problems at the source.
+
+ It is generally not advised for the caller to retry the -EAGAIN return
+ code as this will only make the problem(s) worse and cause your
+ application to temporarily drop to the logger daemon priority, BATCH
+ scheduling policy and background task cgroup. If you require a group of
+ messages to be passed atomically, merge them into one message with
+ embedded newlines to the maximum length LOGGER_ENTRY_MAX_PAYLOAD.
+
+ Other return codes from writing operation can be returned. Since the
+ library retries on EINTR, -EINTR should never be returned.
+
SEE ALSO
syslogd(8)
diff --git a/liblog/event.logtags b/liblog/event.logtags
new file mode 100644
index 0000000..72ecab1
--- /dev/null
+++ b/liblog/event.logtags
@@ -0,0 +1,36 @@
+# The entries in this file map a sparse set of log tag numbers to tag names.
+# This is installed on the device, in /system/etc, and parsed by logcat.
+#
+# Tag numbers are decimal integers, from 0 to 2^31. (Let's leave the
+# negative values alone for now.)
+#
+# Tag names are one or more ASCII letters and numbers or underscores, i.e.
+# "[A-Z][a-z][0-9]_". Do not include spaces or punctuation (the former
+# impacts log readability, the latter makes regex searches more annoying).
+#
+# Tag numbers and names are separated by whitespace. Blank lines and lines
+# starting with '#' are ignored.
+#
+# Optionally, after the tag names can be put a description for the value(s)
+# of the tag. Description are in the format
+# (<name>|data type[|data unit])
+# Multiple values are separated by commas.
+#
+# The data type is a number from the following values:
+# 1: int
+# 2: long
+# 3: string
+# 4: list
+#
+# The data unit is a number taken from the following list:
+# 1: Number of objects
+# 2: Number of bytes
+# 3: Number of milliseconds
+# 4: Number of allocations
+# 5: Id
+# 6: Percent
+# Default value for data of type int/long is 2 (bytes).
+#
+# TODO: generate ".java" and ".h" files with integer constants from this file.
+
+1005 liblog (dropped|1)
diff --git a/liblog/logd_write.c b/liblog/logd_write.c
index 0208c73..c8a6162 100644
--- a/liblog/logd_write.c
+++ b/liblog/logd_write.c
@@ -13,12 +13,14 @@
* See the License for the specific language governing permissions and
* limitations under the License.
*/
+#include <endian.h>
#include <errno.h>
#include <fcntl.h>
#if !defined(_WIN32)
#include <pthread.h>
#endif
#include <stdarg.h>
+#include <stdatomic.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
@@ -172,6 +174,7 @@
size_t i, payload_size;
static uid_t last_uid = AID_ROOT; /* logd *always* starts up as AID_ROOT */
static pid_t last_pid = (pid_t) -1;
+ static atomic_int_fast32_t dropped;
if (last_uid == AID_ROOT) { /* have we called to get the UID yet? */
last_uid = getuid();
@@ -206,7 +209,6 @@
pmsg_header.uid = last_uid;
pmsg_header.pid = last_pid;
- header.id = log_id;
header.tid = gettid();
header.realtime.tv_sec = ts.tv_sec;
header.realtime.tv_nsec = ts.tv_nsec;
@@ -216,6 +218,28 @@
newVec[1].iov_base = (unsigned char *) &header;
newVec[1].iov_len = sizeof(header);
+ if (logd_fd > 0) {
+ int32_t snapshot = atomic_exchange_explicit(&dropped, 0, memory_order_relaxed);
+ if (snapshot) {
+ android_log_event_int_t buffer;
+
+ header.id = LOG_ID_EVENTS;
+ buffer.header.tag = htole32(LIBLOG_LOG_TAG);
+ buffer.payload.type = EVENT_TYPE_INT;
+ buffer.payload.data = htole32(snapshot);
+
+ newVec[2].iov_base = &buffer;
+ newVec[2].iov_len = sizeof(buffer);
+
+ ret = TEMP_FAILURE_RETRY(writev(logd_fd, newVec + 1, 2));
+ if (ret != (ssize_t)(sizeof(header) + sizeof(buffer))) {
+ atomic_fetch_add_explicit(&dropped, snapshot, memory_order_relaxed);
+ }
+ }
+ }
+
+ header.id = log_id;
+
for (payload_size = 0, i = header_length; i < nr + header_length; i++) {
newVec[i].iov_base = vec[i - header_length].iov_base;
payload_size += newVec[i].iov_len = vec[i - header_length].iov_len;
@@ -281,6 +305,8 @@
if (ret > (ssize_t)sizeof(header)) {
ret -= sizeof(header);
+ } else if (ret == -EAGAIN) {
+ atomic_fetch_add_explicit(&dropped, 1, memory_order_relaxed);
}
#endif
diff --git a/logcat/logcat.cpp b/logcat/logcat.cpp
index cd5003e..7e4eadd 100644
--- a/logcat/logcat.cpp
+++ b/logcat/logcat.cpp
@@ -38,14 +38,12 @@
struct logger *logger;
struct logger_list *logger_list;
bool printed;
- char label;
log_device_t* next;
- log_device_t(const char* d, bool b, char l) {
+ log_device_t(const char* d, bool b) {
device = d;
binary = b;
- label = l;
next = NULL;
printed = false;
}
@@ -61,9 +59,7 @@
static int g_outFD = -1;
static off_t g_outByteCount = 0;
static int g_printBinary = 0;
-static int g_devCount = 0;
-
-static EventTagMap* g_eventTagMap = NULL;
+static int g_devCount = 0; // >1 means multiple
static int openLogFile (const char *pathname)
{
@@ -133,8 +129,15 @@
char binaryMsgBuf[1024];
if (dev->binary) {
+ static bool hasOpenedEventTagMap = false;
+ static EventTagMap *eventTagMap = NULL;
+
+ if (!eventTagMap && !hasOpenedEventTagMap) {
+ eventTagMap = android_openEventTagMap(EVENT_TAG_MAP_FILE);
+ hasOpenedEventTagMap = true;
+ }
err = android_log_processBinaryLogBuffer(&buf->entry_v1, &entry,
- g_eventTagMap,
+ eventTagMap,
binaryMsgBuf,
sizeof(binaryMsgBuf));
//printf(">>> pri=%d len=%d msg='%s'\n",
@@ -147,16 +150,6 @@
}
if (android_log_shouldPrintLine(g_logformat, entry.tag, entry.priority)) {
- if (false && g_devCount > 1) {
- binaryMsgBuf[0] = dev->label;
- binaryMsgBuf[1] = ' ';
- bytesWritten = write(g_outFD, binaryMsgBuf, 2);
- if (bytesWritten < 0) {
- perror("output error");
- exit(-1);
- }
- }
-
bytesWritten = android_log_printLogLine(g_logformat, g_outFD, &entry);
if (bytesWritten < 0) {
@@ -331,7 +324,6 @@
const char *forceFilters = NULL;
log_device_t* devices = NULL;
log_device_t* dev;
- bool needBinary = false;
bool printDividers = false;
struct logger_list *logger_list;
unsigned int tail_lines = 0;
@@ -469,7 +461,6 @@
devices = dev = NULL;
android::g_devCount = 0;
- needBinary = false;
for(int i = LOG_ID_MIN; i < LOG_ID_MAX; ++i) {
const char *name = android_log_id_to_name((log_id_t)i);
log_id_t log_id = android_name_to_log_id(name);
@@ -479,7 +470,7 @@
}
bool binary = strcmp(name, "events") == 0;
- log_device_t* d = new log_device_t(name, binary, *name);
+ log_device_t* d = new log_device_t(name, binary);
if (dev) {
dev->next = d;
@@ -488,26 +479,20 @@
devices = dev = d;
}
android::g_devCount++;
- if (binary) {
- needBinary = true;
- }
}
break;
}
bool binary = strcmp(optarg, "events") == 0;
- if (binary) {
- needBinary = true;
- }
if (devices) {
dev = devices;
while (dev->next) {
dev = dev->next;
}
- dev->next = new log_device_t(optarg, binary, optarg[0]);
+ dev->next = new log_device_t(optarg, binary);
} else {
- devices = new log_device_t(optarg, binary, optarg[0]);
+ devices = new log_device_t(optarg, binary);
}
android::g_devCount++;
}
@@ -641,14 +626,14 @@
}
if (!devices) {
- dev = devices = new log_device_t("main", false, 'm');
+ dev = devices = new log_device_t("main", false);
android::g_devCount = 1;
if (android_name_to_log_id("system") == LOG_ID_SYSTEM) {
- dev = dev->next = new log_device_t("system", false, 's');
+ dev = dev->next = new log_device_t("system", false);
android::g_devCount++;
}
if (android_name_to_log_id("crash") == LOG_ID_CRASH) {
- dev = dev->next = new log_device_t("crash", false, 'c');
+ dev = dev->next = new log_device_t("crash", false);
android::g_devCount++;
}
}
@@ -848,17 +833,15 @@
//LOG_EVENT_LONG(11, 0x1122334455667788LL);
//LOG_EVENT_STRING(0, "whassup, doc?");
- if (needBinary)
- android::g_eventTagMap = android_openEventTagMap(EVENT_TAG_MAP_FILE);
-
dev = NULL;
+ log_device_t unexpected("unexpected", false);
while (1) {
struct log_msg log_msg;
log_device_t* d;
int ret = android_logger_list_read(logger_list, &log_msg);
if (ret == 0) {
- fprintf(stderr, "read: Unexpected EOF!\n");
+ fprintf(stderr, "read: unexpected EOF!\n");
exit(EXIT_FAILURE);
}
@@ -868,7 +851,7 @@
}
if (ret == -EIO) {
- fprintf(stderr, "read: Unexpected EOF!\n");
+ fprintf(stderr, "read: unexpected EOF!\n");
exit(EXIT_FAILURE);
}
if (ret == -EINVAL) {
@@ -885,8 +868,9 @@
}
}
if (!d) {
- fprintf(stderr, "read: Unexpected log ID!\n");
- exit(EXIT_FAILURE);
+ android::g_devCount = 2; // set to Multiple
+ d = &unexpected;
+ d->binary = log_msg.id() == LOG_ID_EVENTS;
}
if (dev != d) {
diff --git a/logd/LogBuffer.cpp b/logd/LogBuffer.cpp
index 3d0b38f..edda6c4 100644
--- a/logd/LogBuffer.cpp
+++ b/logd/LogBuffer.cpp
@@ -93,8 +93,7 @@
}
LogBuffer::LogBuffer(LastLogTimes *times)
- : dgramQlenStatistics(false)
- , mTimes(*times) {
+ : mTimes(*times) {
pthread_mutex_init(&mLogElementsLock, NULL);
static const char global_tuneable[] = "persist.logd.size"; // Settings App
@@ -150,23 +149,6 @@
while (last != mLogElements.begin()) {
--it;
if ((*it)->getRealTime() <= realtime) {
- // halves the peak performance, use with caution
- if (dgramQlenStatistics) {
- LogBufferElementCollection::iterator ib = it;
- unsigned short buckets, num = 1;
- for (unsigned short i = 0; (buckets = stats.dgramQlen(i)); ++i) {
- buckets -= num;
- num += buckets;
- while (buckets && (--ib != mLogElements.begin())) {
- --buckets;
- }
- if (buckets) {
- break;
- }
- stats.recordDiff(
- elem->getRealTime() - (*ib)->getRealTime(), i);
- }
- }
break;
}
last = it;
diff --git a/logd/LogBuffer.h b/logd/LogBuffer.h
index 879baea..b0003de 100644
--- a/logd/LogBuffer.h
+++ b/logd/LogBuffer.h
@@ -37,7 +37,6 @@
pthread_mutex_t mLogElementsLock;
LogStatistics stats;
- bool dgramQlenStatistics;
PruneList mPrune;
@@ -63,11 +62,6 @@
// *strp uses malloc, use free to release.
void formatStatistics(char **strp, uid_t uid, unsigned int logMask);
- void enableDgramQlenStatistics() {
- stats.enableDgramQlenStatistics();
- dgramQlenStatistics = true;
- }
-
void enableStatistics() {
stats.enableStatistics();
}
diff --git a/logd/LogStatistics.cpp b/logd/LogStatistics.cpp
index 53036e6..5a70689 100644
--- a/logd/LogStatistics.cpp
+++ b/logd/LogStatistics.cpp
@@ -395,71 +395,11 @@
LogStatistics::LogStatistics()
: mStatistics(false)
- , dgramQlenStatistics(false)
, start(CLOCK_MONOTONIC) {
log_id_for_each(i) {
mSizes[i] = 0;
mElements[i] = 0;
}
-
- for(unsigned short bucket = 0; dgramQlen(bucket); ++bucket) {
- mMinimum[bucket].tv_sec = mMinimum[bucket].tv_sec_max;
- mMinimum[bucket].tv_nsec = mMinimum[bucket].tv_nsec_max;
- }
-}
-
-// Each bucket below represents a dgramQlen of log messages. By
-// finding the minimum period of time from start to finish
-// of each dgramQlen, we can get a performance expectation for
-// the user space logger. The net result is that the period
-// of time divided by the dgramQlen will give us the average time
-// between log messages; at the point where the average time
-// is greater than the throughput capability of the logger
-// we will not longer require the benefits of the FIFO formed
-// by max_dgram_qlen. We will also expect to see a very visible
-// knee in the average time between log messages at this point,
-// so we do not necessarily have to compare the rate against the
-// measured performance (BM_log_maximum_retry) of the logger.
-//
-// for example (reformatted):
-//
-// Minimum time between log events per dgramQlen:
-// 1 2 3 5 10 20 30 50 100 200 300 400 500 600
-// 5u2 12u 13u 15u 16u 27u 30u 36u 407u 3m1 3m3 3m9 3m9 5m5
-//
-// demonstrates a clear knee rising at 100, so this means that for this
-// case max_dgram_qlen = 100 would be more than sufficient to handle the
-// worst that the system could stuff into the logger. The
-// BM_log_maximum_retry performance (derated by the log collection) on the
-// same system was 33.2us so we would almost be fine with max_dgram_qlen = 50.
-// BM_log_maxumum_retry with statistics off is roughly 20us, so
-// max_dgram_qlen = 20 would work. We will be more than willing to have
-// a large engineering margin so the rule of thumb that lead us to 100 is
-// fine.
-//
-// bucket dgramQlen are tuned for /proc/sys/net/unix/max_dgram_qlen = 300
-const unsigned short LogStatistics::mBuckets[] = {
- 1, 2, 3, 5, 10, 20, 30, 50, 100, 200, 300, 400, 500, 600
-};
-
-unsigned short LogStatistics::dgramQlen(unsigned short bucket) {
- if (bucket >= sizeof(mBuckets) / sizeof(mBuckets[0])) {
- return 0;
- }
- return mBuckets[bucket];
-}
-
-unsigned long long LogStatistics::minimum(unsigned short bucket) {
- if (mMinimum[bucket].tv_sec == mMinimum[bucket].tv_sec_max) {
- return 0;
- }
- return mMinimum[bucket].nsec();
-}
-
-void LogStatistics::recordDiff(log_time diff, unsigned short bucket) {
- if ((diff.tv_sec || diff.tv_nsec) && (mMinimum[bucket] > diff)) {
- mMinimum[bucket] = diff;
- }
}
void LogStatistics::add(unsigned short size,
@@ -709,55 +649,6 @@
pids.clear();
}
- if (dgramQlenStatistics) {
- const unsigned short spaces_time = 6;
- const unsigned long long max_seconds = 100000;
- spaces = 0;
- string.append("\n\nMinimum time between log events per max_dgram_qlen:\n");
- for(unsigned short i = 0; dgramQlen(i); ++i) {
- oldLength = string.length();
- if (spaces < 0) {
- spaces = 0;
- }
- string.appendFormat("%*s%u", spaces, "", dgramQlen(i));
- spaces += spaces_time + oldLength - string.length();
- }
- string.append("\n");
- spaces = 0;
- unsigned short n;
- for(unsigned short i = 0; (n = dgramQlen(i)); ++i) {
- unsigned long long duration = minimum(i);
- if (duration) {
- duration /= n;
- if (duration >= (NS_PER_SEC * max_seconds)) {
- duration = NS_PER_SEC * (max_seconds - 1);
- }
- oldLength = string.length();
- if (spaces < 0) {
- spaces = 0;
- }
- string.appendFormat("%*s", spaces, "");
- if (duration >= (NS_PER_SEC * 10)) {
- string.appendFormat("%llu",
- (duration + (NS_PER_SEC / 2))
- / NS_PER_SEC);
- } else if (duration >= (NS_PER_SEC / (1000 / 10))) {
- string.appendFormat("%llum",
- (duration + (NS_PER_SEC / 2 / 1000))
- / (NS_PER_SEC / 1000));
- } else if (duration >= (NS_PER_SEC / (1000000 / 10))) {
- string.appendFormat("%lluu",
- (duration + (NS_PER_SEC / 2 / 1000000))
- / (NS_PER_SEC / 1000000));
- } else {
- string.appendFormat("%llun", duration);
- }
- spaces -= string.length() - oldLength;
- }
- spaces += spaces_time;
- }
- }
-
log_id_for_each(i) {
if (!(logMask & (1 << i))) {
continue;
diff --git a/logd/LogStatistics.h b/logd/LogStatistics.h
index f6c4329..f892cd0 100644
--- a/logd/LogStatistics.h
+++ b/logd/LogStatistics.h
@@ -145,7 +145,6 @@
size_t mElements[LOG_ID_MAX];
bool mStatistics;
- bool dgramQlenStatistics;
static const unsigned short mBuckets[14];
log_time mMinimum[sizeof(mBuckets) / sizeof(mBuckets[0])];
@@ -157,11 +156,7 @@
LidStatistics &id(log_id_t log_id) { return LogIds[log_id]; }
- void enableDgramQlenStatistics() { dgramQlenStatistics = true; }
void enableStatistics() { mStatistics = true; }
- static unsigned short dgramQlen(unsigned short bucket);
- unsigned long long minimum(unsigned short bucket);
- void recordDiff(log_time diff, unsigned short bucket);
void add(unsigned short size, log_id_t log_id, uid_t uid, pid_t pid);
void subtract(unsigned short size, log_id_t log_id, uid_t uid, pid_t pid);
diff --git a/logd/README.property b/logd/README.property
index b7fcece..60542b2 100644
--- a/logd/README.property
+++ b/logd/README.property
@@ -7,12 +7,6 @@
logd.statistics bool depends Enable logcat -S statistics.
ro.config.low_ram bool false if true, logd.statistics default false
ro.build.type string if user, logd.statistics default false
-logd.statistics.dgram_qlen bool false Record dgram_qlen statistics. This
- represents a performance impact and
- is used to determine the platform's
- minimum domain socket network FIFO
- size (see source for details) based
- on typical load (logcat -S to view)
persist.logd.size number 256K default size of the buffer for all
log ids at initial startup, at runtime
use: logcat -b all -G <value>
diff --git a/logd/main.cpp b/logd/main.cpp
index 7a1ae54..243bee4 100644
--- a/logd/main.cpp
+++ b/logd/main.cpp
@@ -155,9 +155,6 @@
LogBuffer *logBuf = new LogBuffer(times);
- if (property_get_bool("logd.statistics.dgram_qlen", false)) {
- logBuf->enableDgramQlenStatistics();
- }
{
char property[PROPERTY_VALUE_MAX];
property_get("ro.build.type", property, "");
diff --git a/logd/tests/logd_test.cpp b/logd/tests/logd_test.cpp
index 96877a9..46bd9c0 100644
--- a/logd/tests/logd_test.cpp
+++ b/logd/tests/logd_test.cpp
@@ -192,164 +192,6 @@
EXPECT_TRUE(NULL != events_logs);
#endif
- // Parse timing stats
-
- cp = strstr(cp, "Minimum time between log events per dgram_qlen:");
-
- if (cp) {
- while (*cp && (*cp != '\n')) {
- ++cp;
- }
- if (*cp == '\n') {
- ++cp;
- }
-
- char *list_of_spans = cp;
- EXPECT_NE('\0', *list_of_spans);
-
- unsigned short number_of_buckets = 0;
- unsigned short *dgram_qlen = NULL;
- unsigned short bucket = 0;
- while (*cp && (*cp != '\n')) {
- bucket = 0;
- while (isdigit(*cp)) {
- bucket = bucket * 10 + *cp - '0';
- ++cp;
- }
- while (*cp == ' ') {
- ++cp;
- }
- if (!bucket) {
- break;
- }
- unsigned short *new_dgram_qlen = new unsigned short[number_of_buckets + 1];
- EXPECT_TRUE(new_dgram_qlen != NULL);
- if (dgram_qlen) {
- memcpy(new_dgram_qlen, dgram_qlen, sizeof(*dgram_qlen) * number_of_buckets);
- delete [] dgram_qlen;
- }
-
- dgram_qlen = new_dgram_qlen;
- dgram_qlen[number_of_buckets++] = bucket;
- }
-
- char *end_of_spans = cp;
- EXPECT_NE('\0', *end_of_spans);
-
- EXPECT_LT(5, number_of_buckets);
-
- unsigned long long *times = new unsigned long long [number_of_buckets];
- ASSERT_TRUE(times != NULL);
-
- memset(times, 0, sizeof(*times) * number_of_buckets);
-
- while (*cp == '\n') {
- ++cp;
- }
-
- unsigned short number_of_values = 0;
- unsigned long long value;
- while (*cp && (*cp != '\n')) {
- EXPECT_GE(number_of_buckets, number_of_values);
-
- value = 0;
- while (isdigit(*cp)) {
- value = value * 10ULL + *cp - '0';
- ++cp;
- }
-
- switch(*cp) {
- case ' ':
- case '\n':
- value *= 1000ULL;
- /* FALLTHRU */
- case 'm':
- value *= 1000ULL;
- /* FALLTHRU */
- case 'u':
- value *= 1000ULL;
- /* FALLTHRU */
- case 'n':
- default:
- break;
- }
- while (*++cp == ' ');
-
- if (!value) {
- break;
- }
-
- times[number_of_values] = value;
- ++number_of_values;
- }
-
-#ifdef TARGET_USES_LOGD
- EXPECT_EQ(number_of_values, number_of_buckets);
-#endif
-
- FILE *fp;
- ASSERT_TRUE(NULL != (fp = fopen("/proc/sys/net/unix/max_dgram_qlen", "r")));
-
- unsigned max_dgram_qlen = 0;
- fscanf(fp, "%u", &max_dgram_qlen);
-
- fclose(fp);
-
- // Find launch point
- unsigned short launch = 0;
- unsigned long long total = 0;
- do {
- total += times[launch];
- } while (((++launch < number_of_buckets)
- && ((total / launch) >= (times[launch] / 8ULL)))
- || (launch == 1)); // too soon
-
- bool failure = number_of_buckets <= launch;
- if (!failure) {
- unsigned short l = launch;
- if (l >= number_of_buckets) {
- l = number_of_buckets - 1;
- }
- failure = max_dgram_qlen < dgram_qlen[l];
- }
-
- // We can get failure if at any time liblog_benchmarks has been run
- // because designed to overload /proc/sys/net/unix/max_dgram_qlen even
- // at excessive values like 20000. It does so to measure the raw processing
- // performance of logd.
- if (failure) {
- cp = find_benchmark_spam(cp);
- }
-
- if (cp) {
- // Fake a failure, but without the failure code
- if (number_of_buckets <= launch) {
- printf ("Expected: number_of_buckets > launch, actual: %u vs %u\n",
- number_of_buckets, launch);
- }
- if (launch >= number_of_buckets) {
- launch = number_of_buckets - 1;
- }
- if (max_dgram_qlen < dgram_qlen[launch]) {
- printf ("Expected: max_dgram_qlen >= dgram_qlen[%d],"
- " actual: %u vs %u\n",
- launch, max_dgram_qlen, dgram_qlen[launch]);
- }
- } else
-#ifndef TARGET_USES_LOGD
- if (total)
-#endif
- {
- EXPECT_GT(number_of_buckets, launch);
- if (launch >= number_of_buckets) {
- launch = number_of_buckets - 1;
- }
- EXPECT_GE(max_dgram_qlen, dgram_qlen[launch]);
- }
-
- delete [] dgram_qlen;
- delete [] times;
- }
delete [] buf;
}