Merge "Refine incidentd error handling and logging" (bc6a68cd) · Commits · e / os / android_frameworks_base

cmds/incidentd/Android.bp

+1 −0

Original line number	Diff line number	Diff line
		@@ -53,6 +53,7 @@ cc_binary {
		"libprotoutil",
		"libservices",
		"libutils",
		"libprotobuf-cpp-lite",
		],

		init_rc: ["incidentd.rc"],

cmds/incidentd/src/FdBuffer.cpp

+16 −5

Original line number	Diff line number	Diff line
		@@ -47,9 +47,13 @@ status_t FdBuffer::read(int fd, int64_t timeout) {
		while (true) {
		if (mBuffer.size() >= MAX_BUFFER_COUNT * BUFFER_SIZE) {
		mTruncated = true;
		VLOG("Truncating data");
		break;
		}
		if (mBuffer.writeBuffer() == NULL) return NO_MEMORY;
		if (mBuffer.writeBuffer() == NULL) {
		VLOG("No memory");
		return NO_MEMORY;
		}

		int64_t remainingTime = (mStartTime + timeout) - uptimeMillis();
		if (remainingTime <= 0) {
		@@ -58,7 +62,7 @@ status_t FdBuffer::read(int fd, int64_t timeout) {
		break;
		}

		int count = poll(&pfds, 1, remainingTime);
		int count = TEMP_FAILURE_RETRY(poll(&pfds, 1, remainingTime));
		if (count == 0) {
		VLOG("timed out due to block calling poll");
		mTimedOut = true;
		@@ -102,7 +106,10 @@ status_t FdBuffer::readFully(int fd) {
		VLOG("Truncating data");
		break;
		}
		if (mBuffer.writeBuffer() == NULL) return NO_MEMORY;
		if (mBuffer.writeBuffer() == NULL) {
		VLOG("No memory");
		return NO_MEMORY;
		}

		ssize_t amt =
		TEMP_FAILURE_RETRY(::read(fd, mBuffer.writeBuffer(), mBuffer.currentToWrite()));
		@@ -144,10 +151,14 @@ status_t FdBuffer::readProcessedDataInStream(int fd, unique_fd toFd, unique_fd f
		// This is the buffer used to store processed data
		while (true) {
		if (mBuffer.size() >= MAX_BUFFER_COUNT * BUFFER_SIZE) {
		VLOG("Truncating data");
		mTruncated = true;
		break;
		}
		if (mBuffer.writeBuffer() == NULL) return NO_MEMORY;
		if (mBuffer.writeBuffer() == NULL) {
		VLOG("No memory");
		return NO_MEMORY;
		}

		int64_t remainingTime = (mStartTime + timeoutMs) - uptimeMillis();
		if (remainingTime <= 0) {
		@@ -157,7 +168,7 @@ status_t FdBuffer::readProcessedDataInStream(int fd, unique_fd toFd, unique_fd f
		}

		// wait for any pfds to be ready to perform IO
		int count = poll(pfds, 3, remainingTime);
		int count = TEMP_FAILURE_RETRY(poll(pfds, 3, remainingTime));
		if (count == 0) {
		VLOG("timed out due to block calling poll");
		mTimedOut = true;

cmds/incidentd/src/IncidentService.cpp

+3 −5

Original line number	Diff line number	Diff line
		@@ -42,13 +42,11 @@ enum { WHAT_RUN_REPORT = 1, WHAT_SEND_BACKLOG_TO_DROPBOX = 2 };
		#define DEFAULT_BYTES_SIZE_LIMIT (20 * 1024 * 1024) // 20MB
		#define DEFAULT_REFACTORY_PERIOD_MS (24 * 60 * 60 * 1000) // 1 Day

		// Skip logs (1100 - 1108) because they are already in the bug report
		// Skip 1200, 1201, 1202, 3018 because they take too long
		// TODO(120079956): Skip 3008, 3015 because of error
		// Skip these sections for dumpstate only. Dumpstate allows 10s max for each service to dump.
		// Skip logs (1100 - 1108) and traces (1200 - 1202) because they are already in the bug report.
		// Skip 3018 because it takes too long.
		#define SKIPPED_SECTIONS { 1100, 1101, 1102, 1103, 1104, 1105, 1106, 1107, 1108, /* Logs */ \
		1200, 1201, 1202, /* Native, hal, java traces */ \
		3008, /* "package --proto" */ \
		3015, /* "activity --proto processes" */ \
		3018 /* "meminfo -a --proto" */ }

		namespace android {

cmds/incidentd/src/PrivacyBuffer.cpp

+12 −3

Original line number	Diff line number	Diff line
		@@ -96,7 +96,12 @@ status_t PrivacyBuffer::stripField(const Privacy* parentPolicy, const PrivacySpe
		uint64_t token = mProto.start(encode_field_id(policy));
		while (mData.rp()->pos() - start != msgSize) {
		status_t err = stripField(policy, spec, depth + 1);
		if (err != NO_ERROR) return err;
		if (err != NO_ERROR) {
		VLOG("Bad value when stripping id %d, wiretype %d, tag %#x, depth %d, size %d, "
		"relative pos %zu, ", fieldId, read_wire_type(fieldTag), fieldTag, depth,
		msgSize, mData.rp()->pos() - start);
		return err;
		}
		}
		mProto.end(token);
		return NO_ERROR;
		@@ -117,9 +122,13 @@ status_t PrivacyBuffer::strip(const PrivacySpec& spec) {
		}
		while (mData.hasNext()) {
		status_t err = stripField(mPolicy, spec, 0);
		if (err != NO_ERROR) return err;
		if (err != NO_ERROR) return err; // Error logged in stripField.
		}
		if (mData.bytesRead() != mData.size()) {
		VLOG("Buffer corrupted: expect %zu bytes, read %zu bytes",
		mData.size(), mData.bytesRead());
		return BAD_VALUE;
		}
		if (mData.bytesRead() != mData.size()) return BAD_VALUE;
		mSize = mProto.size();
		mData.rp()->rewind(); // rewind the read pointer back to beginning after the strip.
		return NO_ERROR;

cmds/incidentd/src/Reporter.cpp

+8 −4

Original line number	Diff line number	Diff line
		@@ -129,6 +129,7 @@ Reporter::run_report_status_t Reporter::runReport(size_t* reportByteSize) {
		bool needMainFd = false;
		int mainFd = -1;
		int mainDest = -1;
		int sectionCount = 0;
		HeaderSection headers;
		MetadataSection metadataSection;
		std::string buildType = android::base::GetProperty("ro.build.type", "");
		@@ -180,12 +181,12 @@ Reporter::run_report_status_t Reporter::runReport(size_t* reportByteSize) {
		for (const Section** section = SECTION_LIST; *section; section++) {
		const int id = (*section)->id;
		if ((*section)->userdebugAndEngOnly && !isUserdebugOrEng) {
		ALOGD("Skipping incident report section %d '%s' because it's limited to userdebug/eng",
		VLOG("Skipping incident report section %d '%s' because it's limited to userdebug/eng",
		id, (*section)->name.string());
		continue;
		}
		if (this->batch.containsSection(id)) {
		ALOGD("Taking incident report section %d '%s'", id, (*section)->name.string());
		VLOG("Taking incident report section %d '%s'", id, (*section)->name.string());
		for (ReportRequestSet::iterator it = batch.begin(); it != batch.end(); it++) {
		if ((it)->listener != NULL && (it)->args.containsSection(id)) {
		(*it)->listener->onReportSectionStatus(
		@@ -198,11 +199,12 @@ Reporter::run_report_status_t Reporter::runReport(size_t* reportByteSize) {
		int64_t startTime = uptimeMillis();
		err = (*section)->Execute(&batch);
		int64_t endTime = uptimeMillis();
		stats->set_success(err == NO_ERROR);
		stats->set_exec_duration_ms(endTime - startTime);
		if (err != NO_ERROR) {
		ALOGW("Incident section %s (%d) failed: %s. Stopping report.",
		(*section)->name.string(), id, strerror(-err));
		// Execute() has already recorded this status. Only update if there's new failure.
		stats->set_success(false);
		goto DONE;
		}
		(*reportByteSize) += stats->report_size_bytes();
		@@ -214,11 +216,13 @@ Reporter::run_report_status_t Reporter::runReport(size_t* reportByteSize) {
		id, IIncidentReportStatusListener::STATUS_FINISHED);
		}
		}
		ALOGD("Finish incident report section %d '%s'", id, (*section)->name.string());
		VLOG("Finish incident report section %d '%s'", id, (*section)->name.string());
		sectionCount++;
		}
		}

		DONE:
		ALOGD("Incident reporting took %d sections.", sectionCount);
		// Reports the metdadata when taking the incident report.
		if (!isTest) metadataSection.Execute(&batch);