As mentioned in the comment inside the code, the Intel documentation states that the internal CPU buffer is flushed out to RAM only when tracing is disabled. Otherwise, the buffer on RAM might be stale. This diff disables tracing when the trace buffer is going to be read. This is a quite safe operation, as the reading is done when the inferior is paused at a breakpoint, so we are not losing any packets because there's no code being executed. After the reading is finished, tracing is enabled back. It's a bit hard to write a test for this now, but Greg Clayton and I will refactor the PT support and writing tests for it will be easier. However I tested it manually by doing a script that automates the following flow ``` (lldb) b main Breakpoint 1: where = a.out`main + 15 at main.cpp:4:7, address = 0x000000000040050f (lldb) r Process 3078226 stopped * thread #1, name = 'a.out', stop reason = breakpoint 1.1 frame #0: 0x000000000040050f a.out`main at main.cpp:4:7 (lldb) processor-trace start (lldb) b 5 Breakpoint 2: where = a.out`main + 22 at main.cpp:5:12, address = 0x0000000000400516 (lldb) c Process 3078226 resuming Process 3078226 stopped * thread #1, name = 'a.out', stop reason = breakpoint 2.1 frame #0: 0x0000000000400516 a.out`main at main.cpp:5:12 (lldb) processor-trace show-instr-log thread #1: tid=3078226 0x40050f <+15>: movl $0x0, -0x8(%rbp) >>> Before, some runs of the script up to this point lead to empty traces (lldb) b 6 Breakpoint 3: where = a.out`main + 42 at main.cpp:6:14, address = 0x000000000040052a (lldb) c Process 3092991 resuming Process 3092991 stopped * thread #1, name = 'a.out', stop reason = breakpoint 3.1 frame #0: 0x000000000040052a a.out`main at main.cpp:6:14 (lldb) processor-trace show-instr-log thread #1: tid=3092991 0x40050f <+15>: movl $0x0, -0x8(%rbp) 0x400516 <+22>: movl $0x0, -0xc(%rbp) 0x40051d <+29>: cmpl $0x2710, -0xc(%rbp) ; imm = 0x2710 0x400524 <+36>: jge 0x400546 ; <+70> at main.cpp 0x400524 <+36>: jge 0x400546 ; <+70> at main.cpp >>> The trace was re-enabled correctly and includes the instruction of the first reading. ``` Those instructions correspond to these lines ``` 3 int main() { 4 int z = 0; 5 for (int i = 0; i < 10000; i++) { 6 z += fun(z) ... ``` Differential Revision: https://reviews.llvm.org/D85241
416 lines
13 KiB
C++
416 lines
13 KiB
C++
//===-- ProcessorTrace.cpp ------------------------------------------------===//
|
|
//
|
|
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
|
|
// See https://llvm.org/LICENSE.txt for license information.
|
|
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#include <algorithm>
|
|
#include <fstream>
|
|
|
|
#include "llvm/ADT/StringRef.h"
|
|
#include "llvm/Support/Error.h"
|
|
#include "llvm/Support/MathExtras.h"
|
|
|
|
#include "Plugins/Process/POSIX/ProcessPOSIXLog.h"
|
|
#include "ProcessorTrace.h"
|
|
#include "lldb/Host/linux/Support.h"
|
|
|
|
#include <sys/ioctl.h>
|
|
#include <sys/syscall.h>
|
|
|
|
using namespace lldb;
|
|
using namespace lldb_private;
|
|
using namespace process_linux;
|
|
using namespace llvm;
|
|
|
|
lldb::user_id_t ProcessorTraceMonitor::m_trace_num = 1;
|
|
|
|
Status ProcessorTraceMonitor::GetTraceConfig(TraceOptions &config) const {
|
|
#ifndef PERF_ATTR_SIZE_VER5
|
|
llvm_unreachable("perf event not supported");
|
|
#else
|
|
Status error;
|
|
|
|
config.setType(lldb::TraceType::eTraceTypeProcessorTrace);
|
|
config.setMetaDataBufferSize(m_mmap_meta->data_size);
|
|
|
|
config.setTraceBufferSize(m_mmap_meta->aux_size);
|
|
|
|
error = GetCPUType(config);
|
|
|
|
return error;
|
|
#endif
|
|
}
|
|
|
|
Status ProcessorTraceMonitor::StartTrace(lldb::pid_t pid, lldb::tid_t tid,
|
|
const TraceOptions &config) {
|
|
#ifndef PERF_ATTR_SIZE_VER5
|
|
llvm_unreachable("perf event not supported");
|
|
#else
|
|
Status error;
|
|
Log *log(ProcessPOSIXLog::GetLogIfAllCategoriesSet(POSIX_LOG_PTRACE));
|
|
|
|
LLDB_LOG(log, "called thread id {0}", tid);
|
|
uint64_t page_size = getpagesize();
|
|
uint64_t bufsize = config.getTraceBufferSize();
|
|
uint64_t metabufsize = config.getMetaDataBufferSize();
|
|
|
|
uint64_t numpages = static_cast<uint64_t>(
|
|
llvm::PowerOf2Floor((bufsize + page_size - 1) / page_size));
|
|
numpages = std::max<uint64_t>(1, numpages);
|
|
bufsize = page_size * numpages;
|
|
|
|
numpages = static_cast<uint64_t>(
|
|
llvm::PowerOf2Floor((metabufsize + page_size - 1) / page_size));
|
|
metabufsize = page_size * numpages;
|
|
|
|
perf_event_attr attr;
|
|
memset(&attr, 0, sizeof(attr));
|
|
attr.size = sizeof(attr);
|
|
attr.exclude_kernel = 1;
|
|
attr.sample_type = PERF_SAMPLE_TIME;
|
|
attr.sample_id_all = 1;
|
|
attr.exclude_hv = 1;
|
|
attr.exclude_idle = 1;
|
|
attr.mmap = 1;
|
|
|
|
int intel_pt_type = 0;
|
|
|
|
auto ret = llvm::MemoryBuffer::getFileAsStream(
|
|
"/sys/bus/event_source/devices/intel_pt/type");
|
|
if (!ret) {
|
|
LLDB_LOG(log, "failed to open Config file");
|
|
return ret.getError();
|
|
}
|
|
|
|
StringRef rest = ret.get()->getBuffer();
|
|
if (rest.empty() || rest.trim().getAsInteger(10, intel_pt_type)) {
|
|
LLDB_LOG(log, "failed to read Config file");
|
|
error.SetErrorString("invalid file");
|
|
return error;
|
|
}
|
|
|
|
rest.trim().getAsInteger(10, intel_pt_type);
|
|
LLDB_LOG(log, "intel pt type {0}", intel_pt_type);
|
|
attr.type = intel_pt_type;
|
|
|
|
LLDB_LOG(log, "meta buffer size {0}", metabufsize);
|
|
LLDB_LOG(log, "buffer size {0} ", bufsize);
|
|
|
|
if (error.Fail()) {
|
|
LLDB_LOG(log, "Status in custom config");
|
|
|
|
return error;
|
|
}
|
|
|
|
errno = 0;
|
|
auto fd =
|
|
syscall(SYS_perf_event_open, &attr, static_cast<::tid_t>(tid), -1, -1, 0);
|
|
if (fd == -1) {
|
|
LLDB_LOG(log, "syscall error {0}", errno);
|
|
error.SetErrorString("perf event syscall Failed");
|
|
return error;
|
|
}
|
|
|
|
m_fd = std::unique_ptr<int, file_close>(new int(fd), file_close());
|
|
|
|
errno = 0;
|
|
auto base =
|
|
mmap(nullptr, (metabufsize + page_size), PROT_WRITE, MAP_SHARED, fd, 0);
|
|
|
|
if (base == MAP_FAILED) {
|
|
LLDB_LOG(log, "mmap base error {0}", errno);
|
|
error.SetErrorString("Meta buffer allocation failed");
|
|
return error;
|
|
}
|
|
|
|
m_mmap_meta = std::unique_ptr<perf_event_mmap_page, munmap_delete>(
|
|
reinterpret_cast<perf_event_mmap_page *>(base),
|
|
munmap_delete(metabufsize + page_size));
|
|
|
|
m_mmap_meta->aux_offset = m_mmap_meta->data_offset + m_mmap_meta->data_size;
|
|
m_mmap_meta->aux_size = bufsize;
|
|
|
|
errno = 0;
|
|
auto mmap_aux = mmap(nullptr, bufsize, PROT_READ, MAP_SHARED, fd,
|
|
static_cast<long int>(m_mmap_meta->aux_offset));
|
|
|
|
if (mmap_aux == MAP_FAILED) {
|
|
LLDB_LOG(log, "second mmap done {0}", errno);
|
|
error.SetErrorString("Trace buffer allocation failed");
|
|
return error;
|
|
}
|
|
m_mmap_aux = std::unique_ptr<uint8_t, munmap_delete>(
|
|
reinterpret_cast<uint8_t *>(mmap_aux), munmap_delete(bufsize));
|
|
return error;
|
|
#endif
|
|
}
|
|
|
|
llvm::MutableArrayRef<uint8_t> ProcessorTraceMonitor::GetDataBuffer() {
|
|
#ifndef PERF_ATTR_SIZE_VER5
|
|
llvm_unreachable("perf event not supported");
|
|
#else
|
|
return MutableArrayRef<uint8_t>(
|
|
(reinterpret_cast<uint8_t *>(m_mmap_meta.get()) +
|
|
m_mmap_meta->data_offset),
|
|
m_mmap_meta->data_size);
|
|
#endif
|
|
}
|
|
|
|
llvm::MutableArrayRef<uint8_t> ProcessorTraceMonitor::GetAuxBuffer() {
|
|
#ifndef PERF_ATTR_SIZE_VER5
|
|
llvm_unreachable("perf event not supported");
|
|
#else
|
|
return MutableArrayRef<uint8_t>(m_mmap_aux.get(), m_mmap_meta->aux_size);
|
|
#endif
|
|
}
|
|
|
|
Status ProcessorTraceMonitor::GetCPUType(TraceOptions &config) {
|
|
|
|
Status error;
|
|
uint64_t cpu_family = -1;
|
|
uint64_t model = -1;
|
|
uint64_t stepping = -1;
|
|
std::string vendor_id;
|
|
|
|
Log *log(ProcessPOSIXLog::GetLogIfAllCategoriesSet(POSIX_LOG_PTRACE));
|
|
|
|
auto BufferOrError = getProcFile("cpuinfo");
|
|
if (!BufferOrError)
|
|
return BufferOrError.getError();
|
|
|
|
LLDB_LOG(log, "GetCPUType Function");
|
|
|
|
StringRef Rest = BufferOrError.get()->getBuffer();
|
|
while (!Rest.empty()) {
|
|
StringRef Line;
|
|
std::tie(Line, Rest) = Rest.split('\n');
|
|
|
|
SmallVector<StringRef, 2> columns;
|
|
Line.split(columns, StringRef(":"), -1, false);
|
|
|
|
if (columns.size() < 2)
|
|
continue; // continue searching
|
|
|
|
columns[1] = columns[1].trim(" ");
|
|
if (columns[0].contains("cpu family") &&
|
|
columns[1].getAsInteger(10, cpu_family))
|
|
continue;
|
|
|
|
else if (columns[0].contains("model") && columns[1].getAsInteger(10, model))
|
|
continue;
|
|
|
|
else if (columns[0].contains("stepping") &&
|
|
columns[1].getAsInteger(10, stepping))
|
|
continue;
|
|
|
|
else if (columns[0].contains("vendor_id")) {
|
|
vendor_id = columns[1].str();
|
|
if (!vendor_id.empty())
|
|
continue;
|
|
}
|
|
LLDB_LOG(log, "{0}:{1}:{2}:{3}", cpu_family, model, stepping, vendor_id);
|
|
|
|
if ((cpu_family != static_cast<uint64_t>(-1)) &&
|
|
(model != static_cast<uint64_t>(-1)) &&
|
|
(stepping != static_cast<uint64_t>(-1)) && (!vendor_id.empty())) {
|
|
auto params_dict = std::make_shared<StructuredData::Dictionary>();
|
|
params_dict->AddIntegerItem("cpu_family", cpu_family);
|
|
params_dict->AddIntegerItem("cpu_model", model);
|
|
params_dict->AddIntegerItem("cpu_stepping", stepping);
|
|
params_dict->AddStringItem("cpu_vendor", vendor_id);
|
|
|
|
llvm::StringRef intel_custom_params_key("intel-pt");
|
|
|
|
auto intel_custom_params = std::make_shared<StructuredData::Dictionary>();
|
|
intel_custom_params->AddItem(
|
|
intel_custom_params_key,
|
|
StructuredData::ObjectSP(std::move(params_dict)));
|
|
|
|
config.setTraceParams(intel_custom_params);
|
|
return error; // we are done
|
|
}
|
|
}
|
|
|
|
error.SetErrorString("cpu info not found");
|
|
return error;
|
|
}
|
|
|
|
llvm::Expected<ProcessorTraceMonitorUP>
|
|
ProcessorTraceMonitor::Create(lldb::pid_t pid, lldb::tid_t tid,
|
|
const TraceOptions &config,
|
|
bool useProcessSettings) {
|
|
|
|
Log *log(ProcessPOSIXLog::GetLogIfAllCategoriesSet(POSIX_LOG_PTRACE));
|
|
|
|
Status error;
|
|
if (tid == LLDB_INVALID_THREAD_ID) {
|
|
error.SetErrorString("thread not specified");
|
|
return error.ToError();
|
|
}
|
|
|
|
ProcessorTraceMonitorUP pt_monitor_up(new ProcessorTraceMonitor);
|
|
|
|
error = pt_monitor_up->StartTrace(pid, tid, config);
|
|
if (error.Fail())
|
|
return error.ToError();
|
|
|
|
pt_monitor_up->SetThreadID(tid);
|
|
|
|
if (useProcessSettings) {
|
|
pt_monitor_up->SetTraceID(0);
|
|
} else {
|
|
pt_monitor_up->SetTraceID(m_trace_num++);
|
|
LLDB_LOG(log, "Trace ID {0}", m_trace_num);
|
|
}
|
|
return std::move(pt_monitor_up);
|
|
}
|
|
|
|
Status
|
|
ProcessorTraceMonitor::ReadPerfTraceAux(llvm::MutableArrayRef<uint8_t> &buffer,
|
|
size_t offset) {
|
|
#ifndef PERF_ATTR_SIZE_VER5
|
|
llvm_unreachable("perf event not supported");
|
|
#else
|
|
// Disable the perf event to force a flush out of the CPU's internal buffer.
|
|
// Besides, we can guarantee that the CPU won't override any data as we are
|
|
// reading the buffer.
|
|
//
|
|
// The Intel documentation says:
|
|
//
|
|
// Packets are first buffered internally and then written out asynchronously.
|
|
// To collect packet output for postprocessing, a collector needs first to
|
|
// ensure that all packet data has been flushed from internal buffers.
|
|
// Software can ensure this by stopping packet generation by clearing
|
|
// IA32_RTIT_CTL.TraceEn (see “Disabling Packet Generation” in
|
|
// Section 35.2.7.2).
|
|
//
|
|
// This is achieved by the PERF_EVENT_IOC_DISABLE ioctl request, as mentioned
|
|
// in the man page of perf_event_open.
|
|
ioctl(*m_fd, PERF_EVENT_IOC_DISABLE);
|
|
|
|
Log *log(ProcessPOSIXLog::GetLogIfAllCategoriesSet(POSIX_LOG_PTRACE));
|
|
Status error;
|
|
uint64_t head = m_mmap_meta->aux_head;
|
|
|
|
LLDB_LOG(log, "Aux size -{0} , Head - {1}", m_mmap_meta->aux_size, head);
|
|
|
|
/**
|
|
* When configured as ring buffer, the aux buffer keeps wrapping around
|
|
* the buffer and its not possible to detect how many times the buffer
|
|
* wrapped. Initially the buffer is filled with zeros,as shown below
|
|
* so in order to get complete buffer we first copy firstpartsize, followed
|
|
* by any left over part from beginning to aux_head
|
|
*
|
|
* aux_offset [d,d,d,d,d,d,d,d,0,0,0,0,0,0,0,0,0,0,0] aux_size
|
|
* aux_head->||<- firstpartsize ->|
|
|
*
|
|
* */
|
|
|
|
ReadCyclicBuffer(buffer, GetAuxBuffer(), static_cast<size_t>(head), offset);
|
|
LLDB_LOG(log, "ReadCyclic BUffer Done");
|
|
|
|
// Reenable tracing now we have read the buffer
|
|
ioctl(*m_fd, PERF_EVENT_IOC_ENABLE);
|
|
return error;
|
|
#endif
|
|
}
|
|
|
|
Status
|
|
ProcessorTraceMonitor::ReadPerfTraceData(llvm::MutableArrayRef<uint8_t> &buffer,
|
|
size_t offset) {
|
|
#ifndef PERF_ATTR_SIZE_VER5
|
|
llvm_unreachable("perf event not supported");
|
|
#else
|
|
Log *log(ProcessPOSIXLog::GetLogIfAllCategoriesSet(POSIX_LOG_PTRACE));
|
|
uint64_t bytes_remaining = buffer.size();
|
|
Status error;
|
|
|
|
uint64_t head = m_mmap_meta->data_head;
|
|
|
|
/*
|
|
* The data buffer and aux buffer have different implementations
|
|
* with respect to their definition of head pointer. In the case
|
|
* of Aux data buffer the head always wraps around the aux buffer
|
|
* and we don't need to care about it, whereas the data_head keeps
|
|
* increasing and needs to be wrapped by modulus operator
|
|
*/
|
|
|
|
LLDB_LOG(log, "bytes_remaining - {0}", bytes_remaining);
|
|
|
|
auto data_buffer = GetDataBuffer();
|
|
|
|
if (head > data_buffer.size()) {
|
|
head = head % data_buffer.size();
|
|
LLDB_LOG(log, "Data size -{0} Head - {1}", m_mmap_meta->data_size, head);
|
|
|
|
ReadCyclicBuffer(buffer, data_buffer, static_cast<size_t>(head), offset);
|
|
bytes_remaining -= buffer.size();
|
|
} else {
|
|
LLDB_LOG(log, "Head - {0}", head);
|
|
if (offset >= head) {
|
|
LLDB_LOG(log, "Invalid Offset ");
|
|
error.SetErrorString("invalid offset");
|
|
buffer = buffer.slice(buffer.size());
|
|
return error;
|
|
}
|
|
|
|
auto data = data_buffer.slice(offset, (head - offset));
|
|
auto remaining = std::copy(data.begin(), data.end(), buffer.begin());
|
|
bytes_remaining -= (remaining - buffer.begin());
|
|
}
|
|
buffer = buffer.drop_back(bytes_remaining);
|
|
return error;
|
|
#endif
|
|
}
|
|
|
|
void ProcessorTraceMonitor::ReadCyclicBuffer(
|
|
llvm::MutableArrayRef<uint8_t> &dst, llvm::MutableArrayRef<uint8_t> src,
|
|
size_t src_cyc_index, size_t offset) {
|
|
|
|
Log *log(ProcessPOSIXLog::GetLogIfAllCategoriesSet(POSIX_LOG_PTRACE));
|
|
|
|
if (dst.empty() || src.empty()) {
|
|
dst = dst.drop_back(dst.size());
|
|
return;
|
|
}
|
|
|
|
if (dst.data() == nullptr || src.data() == nullptr) {
|
|
dst = dst.drop_back(dst.size());
|
|
return;
|
|
}
|
|
|
|
if (src_cyc_index > src.size()) {
|
|
dst = dst.drop_back(dst.size());
|
|
return;
|
|
}
|
|
|
|
if (offset >= src.size()) {
|
|
LLDB_LOG(log, "Too Big offset ");
|
|
dst = dst.drop_back(dst.size());
|
|
return;
|
|
}
|
|
|
|
llvm::SmallVector<MutableArrayRef<uint8_t>, 2> parts = {
|
|
src.slice(src_cyc_index), src.take_front(src_cyc_index)};
|
|
|
|
if (offset > parts[0].size()) {
|
|
parts[1] = parts[1].slice(offset - parts[0].size());
|
|
parts[0] = parts[0].drop_back(parts[0].size());
|
|
} else if (offset == parts[0].size()) {
|
|
parts[0] = parts[0].drop_back(parts[0].size());
|
|
} else {
|
|
parts[0] = parts[0].slice(offset);
|
|
}
|
|
auto next = dst.begin();
|
|
auto bytes_left = dst.size();
|
|
for (auto part : parts) {
|
|
size_t chunk_size = std::min(part.size(), bytes_left);
|
|
next = std::copy_n(part.begin(), chunk_size, next);
|
|
bytes_left -= chunk_size;
|
|
}
|
|
dst = dst.drop_back(bytes_left);
|
|
}
|