Kernel+Profiler: Improve profiling subsystem

This turns the perfcore format into more a log than it was before,
which lets us properly log process, thread and region
creation/destruction. This also makes it unnecessary to dump the
process' regions every time it is scheduled like we did before.

Incidentally this also fixes 'profile -c' because we previously ended
up incorrectly dumping the parent's region map into the profile data.

Log-based mmap support enables profiling shared libraries which
are loaded at runtime, e.g. via dlopen().

This enables profiling both the parent and child process for
programs which use execve(). Previously we'd discard the profiling
data for the old process.

The Profiler tool has been updated to not treat thread IDs as
process IDs anymore. This enables support for processes with more
than one thread. Also, there's a new widget to filter which
process should be displayed.

Kernel/PerformanceEventBuffer.cpp

@@ -20,20 +20,20 @@ PerformanceEventBuffer::PerformanceEventBuffer(NonnullOwnPtr<KBuffer> buffer)
 {
 }
 
-KResult PerformanceEventBuffer::append(int type, FlatPtr arg1, FlatPtr arg2)
+NEVER_INLINE KResult PerformanceEventBuffer::append(int type, FlatPtr arg1, FlatPtr arg2, const StringView& arg3)
 {
     FlatPtr ebp;
     asm volatile("movl %%ebp, %%eax"
                  : "=a"(ebp));
     auto current_thread = Thread::current();
-    auto eip = current_thread->get_register_dump_from_stack().eip;
-    return append_with_eip_and_ebp(eip, ebp, type, arg1, arg2);
+    return append_with_eip_and_ebp(current_thread->pid(), current_thread->tid(), 0, ebp, type, arg1, arg2, arg3);
 }
 
 static Vector<FlatPtr, PerformanceEvent::max_stack_frame_count> raw_backtrace(FlatPtr ebp, FlatPtr eip)
 {
     Vector<FlatPtr, PerformanceEvent::max_stack_frame_count> backtrace;
-    backtrace.append(eip);
+    if (eip != 0)
+        backtrace.append(eip);
     FlatPtr stack_ptr_copy;
     FlatPtr stack_ptr = (FlatPtr)ebp;
     // FIXME: Figure out how to remove this SmapDisabler without breaking profile stacks.
@@ -55,7 +55,8 @@ static Vector<FlatPtr, PerformanceEvent::max_stack_frame_count> raw_backtrace(Fl
     return backtrace;
 }
 
-KResult PerformanceEventBuffer::append_with_eip_and_ebp(u32 eip, u32 ebp, int type, FlatPtr arg1, FlatPtr arg2)
+KResult PerformanceEventBuffer::append_with_eip_and_ebp(ProcessID pid, ThreadID tid,
+    u32 eip, u32 ebp, int type, FlatPtr arg1, FlatPtr arg2, const StringView& arg3)
 {
     if (count() >= capacity())
         return ENOBUFS;
@@ -73,6 +74,39 @@ KResult PerformanceEventBuffer::append_with_eip_and_ebp(u32 eip, u32 ebp, int ty
     case PERF_EVENT_FREE:
         event.data.free.ptr = arg1;
         break;
+    case PERF_EVENT_MMAP:
+        event.data.mmap.ptr = arg1;
+        event.data.mmap.size = arg2;
+        memset(event.data.mmap.name, 0, sizeof(event.data.mmap.name));
+        if (!arg3.is_empty())
+            memcpy(event.data.mmap.name, arg3.characters_without_null_termination(), min(arg3.length(), sizeof(event.data.mmap.name) - 1));
+        break;
+    case PERF_EVENT_MUNMAP:
+        event.data.mmap.ptr = arg1;
+        event.data.mmap.size = arg2;
+        break;
+    case PERF_EVENT_PROCESS_CREATE:
+        event.data.process_create.parent_pid = arg1;
+        memset(event.data.process_create.executable, 0, sizeof(event.data.process_create.executable));
+        if (!arg3.is_empty()) {
+            memcpy(event.data.process_create.executable, arg3.characters_without_null_termination(),
+                min(arg3.length(), sizeof(event.data.process_create.executable) - 1));
+        }
+        break;
+    case PERF_EVENT_PROCESS_EXEC:
+        memset(event.data.process_exec.executable, 0, sizeof(event.data.process_exec.executable));
+        if (!arg3.is_empty()) {
+            memcpy(event.data.process_exec.executable, arg3.characters_without_null_termination(),
+                min(arg3.length(), sizeof(event.data.process_exec.executable) - 1));
+        }
+        break;
+    case PERF_EVENT_PROCESS_EXIT:
+        break;
+    case PERF_EVENT_THREAD_CREATE:
+        event.data.thread_create.parent_tid = arg1;
+        break;
+    case PERF_EVENT_THREAD_EXIT:
+        break;
     default:
         return EINVAL;
     }
@@ -81,7 +115,8 @@ KResult PerformanceEventBuffer::append_with_eip_and_ebp(u32 eip, u32 ebp, int ty
     event.stack_size = min(sizeof(event.stack) / sizeof(FlatPtr), static_cast<size_t>(backtrace.size()));
     memcpy(event.stack, backtrace.data(), event.stack_size * sizeof(FlatPtr));
 
-    event.tid = Thread::current()->tid().value();
+    event.pid = pid.value();
+    event.tid = tid.value();
     event.timestamp = TimeManagement::the().uptime_ms();
     at(m_count++) = event;
     return KSuccess;
@@ -114,7 +149,38 @@ bool PerformanceEventBuffer::to_json_impl(Serializer& object) const
             event_object.add("type", "free");
             event_object.add("ptr", static_cast<u64>(event.data.free.ptr));
             break;
+        case PERF_EVENT_MMAP:
+            event_object.add("type", "mmap");
+            event_object.add("ptr", static_cast<u64>(event.data.mmap.ptr));
+            event_object.add("size", static_cast<u64>(event.data.mmap.size));
+            event_object.add("name", event.data.mmap.name);
+            break;
+        case PERF_EVENT_MUNMAP:
+            event_object.add("type", "munmap");
+            event_object.add("ptr", static_cast<u64>(event.data.munmap.ptr));
+            event_object.add("size", static_cast<u64>(event.data.munmap.size));
+            break;
+        case PERF_EVENT_PROCESS_CREATE:
+            event_object.add("type", "process_create");
+            event_object.add("parent_pid", static_cast<u64>(event.data.process_create.parent_pid));
+            event_object.add("executable", event.data.process_create.executable);
+            break;
+        case PERF_EVENT_PROCESS_EXEC:
+            event_object.add("type", "process_exec");
+            event_object.add("executable", event.data.process_exec.executable);
+            break;
+        case PERF_EVENT_PROCESS_EXIT:
+            event_object.add("type", "process_exit");
+            break;
+        case PERF_EVENT_THREAD_CREATE:
+            event_object.add("type", "thread_create");
+            event_object.add("parent_tid", static_cast<u64>(event.data.thread_create.parent_tid));
+            break;
+        case PERF_EVENT_THREAD_EXIT:
+            event_object.add("type", "thread_exit");
+            break;
         }
+        event_object.add("pid", event.pid);
         event_object.add("tid", event.tid);
         event_object.add("timestamp", event.timestamp);
         auto stack_array = event_object.add_array("stack");
@@ -132,25 +198,6 @@ bool PerformanceEventBuffer::to_json_impl(Serializer& object) const
 bool PerformanceEventBuffer::to_json(KBufferBuilder& builder) const
 {
     JsonObjectSerializer object(builder);
-
-    auto processes_array = object.add_array("processes");
-    for (auto& it : m_processes) {
-        auto& process = *it.value;
-        auto process_object = processes_array.add_object();
-        process_object.add("pid", process.pid.value());
-        process_object.add("executable", process.executable);
-
-        auto regions_array = process_object.add_array("regions");
-        for (auto& region : process.regions) {
-            auto region_object = regions_array.add_object();
-            region_object.add("name", region.name);
-            region_object.add("base", region.range.base().get());
-            region_object.add("size", region.range.size());
-        }
-    }
-
-    processes_array.finish();
-
     return to_json_impl(object);
 }
 
@@ -162,35 +209,30 @@ OwnPtr<PerformanceEventBuffer> PerformanceEventBuffer::try_create_with_size(size
     return adopt_own(*new PerformanceEventBuffer(buffer.release_nonnull()));
 }
 
-void PerformanceEventBuffer::add_process(const Process& process)
+void PerformanceEventBuffer::add_process(const Process& process, ProcessEventType event_type)
 {
-    // FIXME: What about threads that have died?
-
     ScopedSpinLock locker(process.space().get_lock());
 
     String executable;
     if (process.executable())
         executable = process.executable()->absolute_path();
+    else
+        executable = String::formatted("<{}>", process.name());
+
+    [[maybe_unused]] auto rc = append_with_eip_and_ebp(process.pid(), 0, 0, 0,
+        event_type == ProcessEventType::Create ? PERF_EVENT_PROCESS_CREATE : PERF_EVENT_PROCESS_EXEC,
+        process.pid().value(), 0, executable.characters());
 
-    auto sampled_process = adopt_own(*new SampledProcess {
-        .pid = process.pid().value(),
-        .executable = executable,
-        .threads = {},
-        .regions = {},
-    });
     process.for_each_thread([&](auto& thread) {
-        sampled_process->threads.set(thread.tid());
+        [[maybe_unused]] auto rc = append_with_eip_and_ebp(process.pid(), thread.tid().value(),
+            0, 0, PERF_EVENT_THREAD_CREATE, 0, 0, nullptr);
         return IterationDecision::Continue;
     });
 
     for (auto& region : process.space().regions()) {
-        sampled_process->regions.append(SampledProcess::Region {
-            .name = region->name(),
-            .range = region->range(),
-        });
+        [[maybe_unused]] auto rc = append_with_eip_and_ebp(process.pid(), 0,
+            0, 0, PERF_EVENT_MMAP, region->range().base().get(), region->range().size(), region->name().characters());
     }
-
-    m_processes.set(process.pid(), move(sampled_process));
 }
 
 }

Kernel/PerformanceEventBuffer.h

@@ -23,25 +23,61 @@ struct [[gnu::packed]] FreePerformanceEvent {
     FlatPtr ptr;
 };
 
+struct [[gnu::packed]] MmapPerformanceEvent {
+    size_t size;
+    FlatPtr ptr;
+    char name[64];
+};
+
+struct [[gnu::packed]] MunmapPerformanceEvent {
+    size_t size;
+    FlatPtr ptr;
+};
+
+struct [[gnu::packed]] ProcessCreatePerformanceEvent {
+    pid_t parent_pid;
+    char executable[64];
+};
+
+struct [[gnu::packed]] ProcessExecPerformanceEvent {
+    char executable[64];
+};
+
+struct [[gnu::packed]] ThreadCreatePerformanceEvent {
+    pid_t parent_tid;
+};
+
 struct [[gnu::packed]] PerformanceEvent {
     u8 type { 0 };
     u8 stack_size { 0 };
+    u32 pid { 0 };
     u32 tid { 0 };
     u64 timestamp;
     union {
         MallocPerformanceEvent malloc;
         FreePerformanceEvent free;
+        MmapPerformanceEvent mmap;
+        MunmapPerformanceEvent munmap;
+        ProcessCreatePerformanceEvent process_create;
+        ProcessExecPerformanceEvent process_exec;
+        ThreadCreatePerformanceEvent thread_create;
     } data;
     static constexpr size_t max_stack_frame_count = 64;
     FlatPtr stack[max_stack_frame_count];
 };
 
+enum class ProcessEventType {
+    Create,
+    Exec
+};
+
 class PerformanceEventBuffer {
 public:
     static OwnPtr<PerformanceEventBuffer> try_create_with_size(size_t buffer_size);
 
-    KResult append(int type, FlatPtr arg1, FlatPtr arg2);
-    KResult append_with_eip_and_ebp(u32 eip, u32 ebp, int type, FlatPtr arg1, FlatPtr arg2);
+    KResult append(int type, FlatPtr arg1, FlatPtr arg2, const StringView& arg3);
+    KResult append_with_eip_and_ebp(ProcessID pid, ThreadID tid, u32 eip, u32 ebp,
+        int type, FlatPtr arg1, FlatPtr arg2, const StringView& arg3);
 
     void clear()
     {
@@ -57,23 +93,11 @@ public:
 
     bool to_json(KBufferBuilder&) const;
 
-    void add_process(const Process&);
+    void add_process(const Process&, ProcessEventType event_type);
 
 private:
     explicit PerformanceEventBuffer(NonnullOwnPtr<KBuffer>);
 
-    struct SampledProcess {
-        ProcessID pid;
-        String executable;
-        HashTable<ThreadID> threads;
-
-        struct Region {
-            String name;
-            Range range;
-        };
-        Vector<Region> regions;
-    };
-
     template<typename Serializer>
     bool to_json_impl(Serializer&) const;
 
@@ -81,8 +105,9 @@ private:
 
     size_t m_count { 0 };
     NonnullOwnPtr<KBuffer> m_buffer;
-
-    HashMap<ProcessID, NonnullOwnPtr<SampledProcess>> m_processes;
 };
 
+extern bool g_profiling_all_threads;
+extern PerformanceEventBuffer* g_global_perf_events;
+
 }

Kernel/Process.cpp

@@ -242,8 +242,14 @@ Process::~Process()
     VERIFY(thread_count() == 0); // all threads should have been finalized
     VERIFY(!m_alarm_timer);
 
+    if (g_profiling_all_threads) {
+        VERIFY(g_global_perf_events);
+        [[maybe_unused]] auto rc = g_global_perf_events->append_with_eip_and_ebp(
+            pid(), 0, 0, 0, PERF_EVENT_PROCESS_EXIT, 0, 0, nullptr);
+    }
+
     {
-        ScopedSpinLock processses_lock(g_processes_lock);
+        ScopedSpinLock processes_lock(g_processes_lock);
         if (prev() || next())
             g_processes->remove(this);
     }
@@ -675,7 +681,7 @@ bool Process::create_perf_events_buffer_if_needed()
 {
     if (!m_perf_event_buffer) {
         m_perf_event_buffer = PerformanceEventBuffer::try_create_with_size(4 * MiB);
-        m_perf_event_buffer->add_process(*this);
+        m_perf_event_buffer->add_process(*this, ProcessEventType::Create);
     }
     return !!m_perf_event_buffer;
 }

Kernel/Process.h

@@ -20,6 +20,7 @@
 #include <Kernel/Forward.h>
 #include <Kernel/FutexQueue.h>
 #include <Kernel/Lock.h>
+#include <Kernel/PerformanceEventBuffer.h>
 #include <Kernel/ProcessGroup.h>
 #include <Kernel/StdLib.h>
 #include <Kernel/Thread.h>
@@ -528,6 +529,11 @@ private:
 
     void clear_futex_queues_on_exec();
 
+    inline PerformanceEventBuffer* current_perf_events_buffer()
+    {
+        return g_profiling_all_threads ? g_global_perf_events : m_perf_event_buffer.ptr();
+    }
+
     Process* m_prev { nullptr };
     Process* m_next { nullptr };
 

Kernel/Scheduler.cpp

@@ -22,9 +22,6 @@
 
 namespace Kernel {
 
-extern bool g_profiling_all_threads;
-extern PerformanceEventBuffer* g_global_perf_events;
-
 class SchedulerPerProcessorData {
     AK_MAKE_NONCOPYABLE(SchedulerPerProcessorData);
     AK_MAKE_NONMOVABLE(SchedulerPerProcessorData);
@@ -513,12 +510,6 @@ void Scheduler::timer_tick(const RegisterState& regs)
         //        That will be an interesting mode to add in the future. :^)
         if (current_thread != Processor::current().idle_thread()) {
             perf_events = g_global_perf_events;
-            if (current_thread->process().space().enforces_syscall_regions()) {
-                // FIXME: This is very nasty! We dump the current process's address
-                //        space layout *every time* it's sampled. We should figure out
-                //        a way to do this less often.
-                perf_events->add_process(current_thread->process());
-            }
         }
     } else if (current_thread->process().is_profiling()) {
         VERIFY(current_thread->process().perf_events());
@@ -526,7 +517,9 @@ void Scheduler::timer_tick(const RegisterState& regs)
     }
 
     if (perf_events) {
-        [[maybe_unused]] auto rc = perf_events->append_with_eip_and_ebp(regs.eip, regs.ebp, PERF_EVENT_SAMPLE, 0, 0);
+        [[maybe_unused]] auto rc = perf_events->append_with_eip_and_ebp(
+            current_thread->pid(), current_thread->tid(),
+            regs.eip, regs.ebp, PERF_EVENT_SAMPLE, 0, 0, nullptr);
     }
 
     if (current_thread->tick())

Kernel/Syscalls/execve.cpp

@@ -628,9 +628,9 @@ KResult Process::do_exec(NonnullRefPtr<FileDescription> main_program_description
     tss.cr3 = space().page_directory().cr3();
     tss.ss2 = pid().value();
 
-    // Throw away any recorded performance events in this process.
-    if (m_perf_event_buffer)
-        m_perf_event_buffer->clear();
+    if (auto* event_buffer = current_perf_events_buffer()) {
+        event_buffer->add_process(*this, ProcessEventType::Exec);
+    }
 
     {
         ScopedSpinLock lock(g_scheduler_lock);

Kernel/Syscalls/exit.cpp

@@ -5,6 +5,7 @@
  */
 
 #include <Kernel/KSyms.h>
+#include <Kernel/PerformanceEventBuffer.h>
 #include <Kernel/Process.h>
 
 namespace Kernel {
@@ -16,6 +17,11 @@ void Process::sys$exit(int status)
         m_termination_status = status;
         m_termination_signal = 0;
     }
+
+    if (auto* event_buffer = current_perf_events_buffer()) {
+        [[maybe_unused]] auto rc = event_buffer->append(PERF_EVENT_THREAD_EXIT, Thread::current()->tid().value(), 0, nullptr);
+    }
+
     die();
     Thread::current()->die_if_needed();
     VERIFY_NOT_REACHED();

Kernel/Syscalls/fork.cpp

@@ -7,6 +7,7 @@
 #include <Kernel/Debug.h>
 #include <Kernel/FileSystem/Custody.h>
 #include <Kernel/FileSystem/FileDescription.h>
+#include <Kernel/PerformanceEventBuffer.h>
 #include <Kernel/Process.h>
 #include <Kernel/VM/Region.h>
 
@@ -84,6 +85,11 @@ KResultOr<pid_t> Process::sys$fork(RegisterState& regs)
         g_processes->prepend(child);
     }
 
+    if (g_profiling_all_threads) {
+        VERIFY(g_global_perf_events);
+        g_global_perf_events->add_process(*child, ProcessEventType::Create);
+    }
+
     ScopedSpinLock lock(g_scheduler_lock);
     child_first_thread->set_affinity(Thread::current()->affinity());
     child_first_thread->set_state(Thread::State::Runnable);

Kernel/Syscalls/mmap.cpp

@@ -246,6 +246,12 @@ KResultOr<FlatPtr> Process::sys$mmap(Userspace<const Syscall::SC_mmap_params*> u
 
     if (!region)
         return ENOMEM;
+
+    if (auto* event_buffer = current_perf_events_buffer()) {
+        [[maybe_unused]] auto res = event_buffer->append(PERF_EVENT_MMAP, region->vaddr().get(),
+            region->size(), name.is_null() ? region->name().characters() : name.characters());
+    }
+
     region->set_mmap(true);
     if (map_shared)
         region->set_shared(true);
@@ -430,6 +436,9 @@ KResultOr<int> Process::sys$set_mmap_name(Userspace<const Syscall::SC_set_mmap_n
         return EINVAL;
     if (!region->is_mmap())
         return EPERM;
+    if (auto* event_buffer = current_perf_events_buffer()) {
+        [[maybe_unused]] auto res = event_buffer->append(PERF_EVENT_MMAP, region->vaddr().get(), region->size(), name.characters());
+    }
     region->set_name(move(name));
     return 0;
 }
@@ -453,8 +462,13 @@ KResultOr<int> Process::sys$munmap(Userspace<void*> addr, size_t size)
     if (auto* whole_region = space().find_region_from_range(range_to_unmap)) {
         if (!whole_region->is_mmap())
             return EPERM;
+        auto base = whole_region->vaddr();
+        auto size = whole_region->size();
         bool success = space().deallocate_region(*whole_region);
         VERIFY(success);
+        if (auto* event_buffer = current_perf_events_buffer()) {
+            [[maybe_unused]] auto res = event_buffer->append(PERF_EVENT_MUNMAP, base.get(), size, nullptr);
+        }
         return 0;
     }
 
@@ -479,6 +493,11 @@ KResultOr<int> Process::sys$munmap(Userspace<void*> addr, size_t size)
         for (auto* new_region : new_regions) {
             new_region->map(space().page_directory());
         }
+
+        if (auto* event_buffer = current_perf_events_buffer()) {
+            [[maybe_unused]] auto res = event_buffer->append(PERF_EVENT_MUNMAP, range_to_unmap.base().get(), range_to_unmap.size(), nullptr);
+        }
+
         return 0;
     }
 
@@ -521,6 +540,10 @@ KResultOr<int> Process::sys$munmap(Userspace<void*> addr, size_t size)
         new_region->map(space().page_directory());
     }
 
+    if (auto* event_buffer = current_perf_events_buffer()) {
+        [[maybe_unused]] auto res = event_buffer->append(PERF_EVENT_MUNMAP, range_to_unmap.base().get(), range_to_unmap.size(), nullptr);
+    }
+
     return 0;
 }
 

Kernel/Syscalls/perf_event.cpp

@@ -13,7 +13,7 @@ KResultOr<int> Process::sys$perf_event(int type, FlatPtr arg1, FlatPtr arg2)
 {
     if (!create_perf_events_buffer_if_needed())
         return ENOMEM;
-    return perf_events()->append(type, arg1, arg2);
+    return perf_events()->append(type, arg1, arg2, nullptr);
 }
 
 }

Kernel/Syscalls/profiling.cpp

@@ -12,8 +12,8 @@
 
 namespace Kernel {
 
-PerformanceEventBuffer* g_global_perf_events;
 bool g_profiling_all_threads;
+PerformanceEventBuffer* g_global_perf_events;
 
 KResultOr<int> Process::sys$profiling_enable(pid_t pid)
 {
@@ -27,6 +27,11 @@ KResultOr<int> Process::sys$profiling_enable(pid_t pid)
             g_global_perf_events->clear();
         else
             g_global_perf_events = PerformanceEventBuffer::try_create_with_size(32 * MiB).leak_ptr();
+        ScopedSpinLock lock(g_processes_lock);
+        Process::for_each([](auto& process) {
+            g_global_perf_events->add_process(process, ProcessEventType::Create);
+            return IterationDecision::Continue;
+        });
         g_profiling_all_threads = true;
         return 0;
     }

Kernel/Syscalls/thread.cpp

@@ -8,6 +8,7 @@
 #include <AK/String.h>
 #include <AK/StringBuilder.h>
 #include <AK/StringView.h>
+#include <Kernel/PerformanceEventBuffer.h>
 #include <Kernel/Process.h>
 #include <Kernel/VM/MemoryManager.h>
 #include <Kernel/VM/PageDirectory.h>
@@ -69,6 +70,10 @@ KResultOr<int> Process::sys$create_thread(void* (*entry)(void*), Userspace<const
     if (tsr_result.is_error())
         return tsr_result.error();
 
+    if (m_perf_event_buffer) {
+        [[maybe_unused]] auto rc = m_perf_event_buffer->append(PERF_EVENT_THREAD_CREATE, thread->tid().value(), 0, nullptr);
+    }
+
     ScopedSpinLock lock(g_scheduler_lock);
     thread->set_priority(requested_thread_priority);
     thread->set_state(Thread::State::Runnable);
@@ -84,6 +89,10 @@ void Process::sys$exit_thread(Userspace<void*> exit_value)
         this->sys$exit(0);
     }
 
+    if (m_perf_event_buffer) {
+        [[maybe_unused]] auto rc = m_perf_event_buffer->append(PERF_EVENT_THREAD_EXIT, Thread::current()->tid().value(), 0, nullptr);
+    }
+
     Thread::current()->exit(reinterpret_cast<void*>(exit_value.ptr()));
     VERIFY_NOT_REACHED();
 }

Kernel/UnixTypes.h

@@ -46,9 +46,18 @@ enum {
     _SC_CLK_TCK,
 };
 
-#define PERF_EVENT_SAMPLE 0
-#define PERF_EVENT_MALLOC 1
-#define PERF_EVENT_FREE 2
+enum {
+    PERF_EVENT_SAMPLE,
+    PERF_EVENT_MALLOC,
+    PERF_EVENT_FREE,
+    PERF_EVENT_MMAP,
+    PERF_EVENT_MUNMAP,
+    PERF_EVENT_PROCESS_CREATE,
+    PERF_EVENT_PROCESS_EXEC,
+    PERF_EVENT_PROCESS_EXIT,
+    PERF_EVENT_THREAD_CREATE,
+    PERF_EVENT_THREAD_EXIT
+};
 
 #define WNOHANG 1
 #define WUNTRACED 2

Userland/DevTools/Profiler/CMakeLists.txt

@@ -1,11 +1,13 @@
 set(SOURCES
     DisassemblyModel.cpp
     main.cpp
-        IndividualSampleModel.cpp
-        Profile.cpp
+    IndividualSampleModel.cpp
+    Process.cpp
+    ProcessPickerWidget.cpp
+    Profile.cpp
     ProfileModel.cpp
-        ProfileTimelineWidget.cpp
-        SamplesModel.cpp
+    ProfileTimelineWidget.cpp
+    SamplesModel.cpp
 )
 
 serenity_app(Profiler ICON app-profiler)

Userland/DevTools/Profiler/DisassemblyModel.cpp

@@ -48,10 +48,14 @@ DisassemblyModel::DisassemblyModel(Profile& profile, ProfileNode& node)
         kernel_elf = make<ELF::Image>((const u8*)m_kernel_file->data(), m_kernel_file->size());
         elf = kernel_elf.ptr();
     } else {
-        // FIXME: This is kinda rickety looking with all the -> -> ->
-        auto library_data = node.process(profile)->library_metadata->library_containing(node.address());
+        auto process = node.process(profile, node.timestamp());
+        if (!process) {
+            dbgln("no process for address {}", node.address());
+            return;
+        }
+        auto library_data = process->library_metadata.library_containing(node.address());
         if (!library_data) {
-            dbgln("no library data");
+            dbgln("no library data for address {}", node.address());
             return;
         }
         elf = &library_data->object->elf;

Userland/DevTools/Profiler/Process.cpp

@@ -0,0 +1,119 @@
+/*
+ * Copyright (c) 2018-2021, Andreas Kling <kling@serenityos.org>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#include "Process.h"
+
+Thread* Process::find_thread(pid_t tid, u64 timestamp)
+{
+    auto it = threads.find(tid);
+    if (it == threads.end())
+        return nullptr;
+    for (auto& thread : it->value) {
+        if (thread.start_valid < timestamp && (thread.end_valid == 0 || thread.end_valid > timestamp))
+            return &thread;
+    }
+    return nullptr;
+}
+
+void Process::handle_thread_create(pid_t tid, u64 timestamp)
+{
+    auto it = threads.find(tid);
+    if (it == threads.end()) {
+        threads.set(tid, {});
+        it = threads.find(tid);
+    }
+
+    auto thread = Thread { tid, timestamp, 0 };
+    it->value.append(move(thread));
+}
+
+void Process::handle_thread_exit(pid_t tid, u64 timestamp)
+{
+    auto* thread = find_thread(tid, timestamp);
+    if (!thread)
+        return;
+    thread->end_valid = timestamp;
+}
+
+HashMap<String, OwnPtr<MappedObject>> g_mapped_object_cache;
+
+static MappedObject* get_or_create_mapped_object(const String& path)
+{
+    if (auto it = g_mapped_object_cache.find(path); it != g_mapped_object_cache.end())
+        return it->value.ptr();
+
+    auto file_or_error = MappedFile::map(path);
+    if (file_or_error.is_error()) {
+        g_mapped_object_cache.set(path, {});
+        return nullptr;
+    }
+    auto elf = ELF::Image(file_or_error.value()->bytes());
+    if (!elf.is_valid()) {
+        g_mapped_object_cache.set(path, {});
+        return nullptr;
+    }
+    auto new_mapped_object = adopt_own(*new MappedObject {
+        .file = file_or_error.release_value(),
+        .elf = elf,
+    });
+    auto* ptr = new_mapped_object.ptr();
+    g_mapped_object_cache.set(path, move(new_mapped_object));
+    return ptr;
+}
+
+void LibraryMetadata::handle_mmap(FlatPtr base, size_t size, const String& name)
+{
+    String path;
+    if (name.contains("Loader.so"))
+        path = "Loader.so";
+    else if (!name.contains(":"))
+        return;
+    else
+        path = name.substring(0, name.view().find_first_of(":").value());
+
+    String full_path;
+    if (name.contains(".so"))
+        full_path = String::formatted("/usr/lib/{}", path);
+    else
+        full_path = path;
+
+    auto* mapped_object = get_or_create_mapped_object(full_path);
+    if (!mapped_object) {
+        full_path = String::formatted("/usr/local/lib/{}", path);
+        mapped_object = get_or_create_mapped_object(full_path);
+        if (!mapped_object)
+            return;
+    }
+
+    FlatPtr text_base {};
+    mapped_object->elf.for_each_program_header([&](const ELF::Image::ProgramHeader& ph) {
+        if (ph.is_executable())
+            text_base = ph.vaddr().get();
+        return IterationDecision::Continue;
+    });
+
+    m_libraries.set(name, adopt_own(*new Library { base, size, name, text_base, mapped_object }));
+}
+
+String LibraryMetadata::Library::symbolicate(FlatPtr ptr, u32* offset) const
+{
+    if (!object)
+        return String::formatted("?? <{:p}>", ptr);
+
+    return object->elf.symbolicate(ptr - base + text_base, offset);
+}
+
+const LibraryMetadata::Library* LibraryMetadata::library_containing(FlatPtr ptr) const
+{
+    for (auto& it : m_libraries) {
+        if (!it.value)
+            continue;
+        auto& library = *it.value;
+        if (ptr >= library.base && ptr < (library.base + library.size))
+            return &library;
+    }
+    return nullptr;
+}

Userland/DevTools/Profiler/Process.h

@@ -0,0 +1,68 @@
+/*
+ * Copyright (c) 2018-2021, Andreas Kling <kling@serenityos.org>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#pragma once
+
+#include <AK/HashMap.h>
+#include <AK/MappedFile.h>
+#include <AK/OwnPtr.h>
+#include <AK/Vector.h>
+#include <LibELF/Image.h>
+
+struct MappedObject {
+    NonnullRefPtr<MappedFile> file;
+    ELF::Image elf;
+};
+
+extern HashMap<String, OwnPtr<MappedObject>> g_mapped_object_cache;
+
+class LibraryMetadata {
+public:
+    struct Library {
+        FlatPtr base;
+        size_t size;
+        String name;
+        FlatPtr text_base;
+        MappedObject* object { nullptr };
+
+        String symbolicate(FlatPtr, u32* offset) const;
+    };
+
+    void handle_mmap(FlatPtr base, size_t size, const String& name);
+    const Library* library_containing(FlatPtr) const;
+
+private:
+    mutable HashMap<String, OwnPtr<Library>> m_libraries;
+};
+
+struct Thread {
+    pid_t tid;
+    u64 start_valid;
+    u64 end_valid { 0 };
+
+    bool valid_at(u64 timestamp) const
+    {
+        return timestamp >= start_valid && (end_valid == 0 || timestamp <= end_valid);
+    }
+};
+
+struct Process {
+    pid_t pid {};
+    String executable;
+    HashMap<int, Vector<Thread>> threads {};
+    LibraryMetadata library_metadata {};
+    u64 start_valid;
+    u64 end_valid { 0 };
+
+    Thread* find_thread(pid_t tid, u64 timestamp);
+    void handle_thread_create(pid_t tid, u64 timestamp);
+    void handle_thread_exit(pid_t tid, u64 timestamp);
+
+    bool valid_at(u64 timestamp) const
+    {
+        return timestamp >= start_valid && (end_valid == 0 || timestamp <= end_valid);
+    }
+};

Userland/DevTools/Profiler/ProcessPickerWidget.cpp

@@ -0,0 +1,48 @@
+/*
+ * Copyright (c) 2021, Gunnar Beutner <gunnar@beutner.name>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#include "ProcessPickerWidget.h"
+#include "Profile.h"
+#include <LibGUI/BoxLayout.h>
+#include <LibGUI/ItemListModel.h>
+#include <LibGUI/Label.h>
+
+ProcessPickerWidget::ProcessPickerWidget(Profile& profile)
+    : m_profile(profile)
+{
+    set_layout<GUI::HorizontalBoxLayout>();
+    set_fixed_height(30);
+
+    set_frame_shape(Gfx::FrameShape::NoFrame);
+
+    auto& label = add<GUI::Label>("Process:");
+    label.set_fixed_width(50);
+    label.set_text_alignment(Gfx::TextAlignment::CenterRight);
+
+    m_process_combo = add<GUI::ComboBox>();
+    m_process_combo->set_only_allow_values_from_model(true);
+
+    m_processes.append("All processes");
+
+    for (auto& process : m_profile.processes())
+        m_processes.append(String::formatted("{}: {}", process.pid, process.executable));
+
+    m_process_combo->set_model(*GUI::ItemListModel<String>::create(m_processes));
+    m_process_combo->set_selected_index(0);
+    m_process_combo->on_change = [this](auto&, const GUI::ModelIndex& index) {
+        if (index.row() == 0) {
+            m_profile.clear_process_filter();
+        } else {
+            auto& process = m_profile.processes()[index.row() - 1];
+            auto end_valid = process.end_valid == 0 ? m_profile.last_timestamp() : process.end_valid;
+            m_profile.set_process_filter(process.pid, process.start_valid, end_valid);
+        }
+    };
+}
+
+ProcessPickerWidget::~ProcessPickerWidget()
+{
+}

Userland/DevTools/Profiler/ProcessPickerWidget.h

@@ -0,0 +1,31 @@
+/*
+ * Copyright (c) 2021, Gunnar Beutner <gunnar@beutner.name>
+ *
+ * SPDX-License-Identifier: BSD-2-Clause
+ */
+
+#pragma once
+
+#include <LibGUI/ComboBox.h>
+#include <LibGUI/Frame.h>
+
+class Profile;
+
+class ProcessPickerWidget final : public GUI::Frame {
+    C_OBJECT(ProcessPickerWidget)
+public:
+    virtual ~ProcessPickerWidget() override;
+
+    struct Process {
+        pid_t pid;
+        String name;
+    };
+
+private:
+    explicit ProcessPickerWidget(Profile&);
+
+    Profile& m_profile;
+    Vector<String> m_processes;
+
+    RefPtr<GUI::ComboBox> m_process_combo;
+};

Userland/DevTools/Profiler/Profile.cpp

@@ -11,6 +11,7 @@
 #include <AK/HashTable.h>
 #include <AK/LexicalPath.h>
 #include <AK/MappedFile.h>
+#include <AK/NonnullOwnPtrVector.h>
 #include <AK/QuickSort.h>
 #include <AK/RefPtr.h>
 #include <LibCore/File.h>
@@ -60,12 +61,10 @@ GUI::Model& Profile::samples_model()
 
 void Profile::rebuild_tree()
 {
-    u32 filtered_event_count = 0;
     Vector<NonnullRefPtr<ProfileNode>> roots;
 
     auto find_or_create_root = [&roots](FlyString object_name, String symbol, u32 address, u32 offset, u64 timestamp, pid_t pid) -> ProfileNode& {
-        for (size_t i = 0; i < roots.size(); ++i) {
-            auto& root = roots[i];
+        for (auto root : roots) {
             if (root->symbol() == symbol) {
                 return root;
             }
@@ -84,18 +83,23 @@ void Profile::rebuild_tree()
             live_allocations.remove(event.ptr);
     });
 
-    Optional<size_t> first_filtered_event_index;
+    m_filtered_event_indices.clear();
 
     for (size_t event_index = 0; event_index < m_events.size(); ++event_index) {
         auto& event = m_events.at(event_index);
+
         if (has_timestamp_filter_range()) {
             auto timestamp = event.timestamp;
             if (timestamp < m_timestamp_filter_range_start || timestamp > m_timestamp_filter_range_end)
                 continue;
         }
 
-        if (!first_filtered_event_index.has_value())
-            first_filtered_event_index = event_index;
+        if (has_process_filter()) {
+            if (event.pid != m_process_filter_pid || event.timestamp < m_process_filter_start_valid || event.timestamp > m_process_filter_end_valid)
+                continue;
+        }
+
+        m_filtered_event_indices.append(event_index);
 
         if (event.type == "malloc" && !live_allocations.contains(event.ptr))
             continue;
@@ -130,9 +134,9 @@ void Profile::rebuild_tree()
 
                 // FIXME: More cheating with intentional mixing of TID/PID here:
                 if (!node)
-                    node = &find_or_create_root(object_name, symbol, address, offset, event.timestamp, event.tid);
+                    node = &find_or_create_root(object_name, symbol, address, offset, event.timestamp, event.pid);
                 else
-                    node = &node->find_or_create_child(object_name, symbol, address, offset, event.timestamp, event.tid);
+                    node = &node->find_or_create_child(object_name, symbol, address, offset, event.timestamp, event.pid);
 
                 node->increment_event_count();
                 if (is_innermost_frame) {
@@ -156,11 +160,11 @@ void Profile::rebuild_tree()
 
                     // FIXME: More PID/TID mixing cheats here:
                     if (!node) {
-                        node = &find_or_create_root(object_name, symbol, address, offset, event.timestamp, event.tid);
+                        node = &find_or_create_root(object_name, symbol, address, offset, event.timestamp, event.pid);
                         root = node;
                         root->will_track_seen_events(m_events.size());
                     } else {
-                        node = &node->find_or_create_child(object_name, symbol, address, offset, event.timestamp, event.tid);
+                        node = &node->find_or_create_child(object_name, symbol, address, offset, event.timestamp, event.pid);
                     }
 
                     if (!root->has_seen_event(event_index)) {
@@ -177,14 +181,10 @@ void Profile::rebuild_tree()
                 }
             }
         }
-
-        ++filtered_event_count;
     }
 
     sort_profile_nodes(roots);
 
-    m_filtered_event_count = filtered_event_count;
-    m_first_filtered_event_index = first_filtered_event_index.value_or(0);
     m_roots = move(roots);
     m_model->update();
 }
@@ -201,45 +201,6 @@ Result<NonnullOwnPtr<Profile>, String> Profile::load_from_perfcore_file(const St
 
     auto& object = json.value().as_object();
 
-    auto processes_value = object.get("processes");
-    if (processes_value.is_null())
-        return String { "Invalid perfcore format (no processes)" };
-
-    if (!processes_value.is_array())
-        return String { "Invalid perfcore format (processes is not an array)" };
-
-    Vector<Process> sampled_processes;
-
-    for (auto& process_value : processes_value.as_array().values()) {
-        if (!process_value.is_object())
-            return String { "Invalid perfcore format (process value is not an object)" };
-        auto& process = process_value.as_object();
-        auto regions_value = process.get("regions");
-        if (!regions_value.is_array())
-            return String { "Invalid perfcore format (regions is not an array)" };
-
-        Process sampled_process {
-            .pid = (pid_t)process.get("pid").to_i32(),
-            .executable = process.get("executable").to_string(),
-            .threads = {},
-            .regions = {},
-            .library_metadata = make<LibraryMetadata>(regions_value.as_array()),
-        };
-
-        for (auto& region_value : regions_value.as_array().values()) {
-            if (!region_value.is_object())
-                return String { "Invalid perfcore format (region is not an object)" };
-            auto& region = region_value.as_object();
-            sampled_process.regions.append(Process::Region {
-                .name = region.get("name").to_string(),
-                .base = region.get("base").to_u32(),
-                .size = region.get("size").to_u32(),
-            });
-        }
-
-        sampled_processes.append(move(sampled_process));
-    }
-
     auto file_or_error = MappedFile::map("/boot/Kernel");
     OwnPtr<ELF::Image> kernel_elf;
     if (!file_or_error.is_error())
@@ -250,9 +211,9 @@ Result<NonnullOwnPtr<Profile>, String> Profile::load_from_perfcore_file(const St
         return String { "Malformed profile (events is not an array)" };
 
     auto& perf_events = events_value.as_array();
-    if (perf_events.is_empty())
-        return String { "No events captured (targeted process was never on CPU)" };
 
+    NonnullOwnPtrVector<Process> all_processes;
+    HashMap<pid_t, Process*> current_processes;
     Vector<Event> events;
 
     for (auto& perf_event_value : perf_events.values()) {
@@ -262,6 +223,7 @@ Result<NonnullOwnPtr<Profile>, String> Profile::load_from_perfcore_file(const St
 
         event.timestamp = perf_event.get("timestamp").to_number<u64>();
         event.type = perf_event.get("type").to_string();
+        event.pid = perf_event.get("pid").to_i32();
         event.tid = perf_event.get("tid").to_i32();
 
         if (event.type == "malloc") {
@@ -269,6 +231,65 @@ Result<NonnullOwnPtr<Profile>, String> Profile::load_from_perfcore_file(const St
             event.size = perf_event.get("size").to_number<size_t>();
         } else if (event.type == "free") {
             event.ptr = perf_event.get("ptr").to_number<FlatPtr>();
+        } else if (event.type == "mmap") {
+            event.ptr = perf_event.get("ptr").to_number<FlatPtr>();
+            event.size = perf_event.get("size").to_number<size_t>();
+            event.name = perf_event.get("name").to_string();
+
+            auto it = current_processes.find(event.pid);
+            if (it != current_processes.end())
+                it->value->library_metadata.handle_mmap(event.ptr, event.size, event.name);
+            continue;
+        } else if (event.type == "munmap") {
+            event.ptr = perf_event.get("ptr").to_number<FlatPtr>();
+            event.size = perf_event.get("size").to_number<size_t>();
+            continue;
+        } else if (event.type == "process_create") {
+            event.parent_pid = perf_event.get("parent_pid").to_number<FlatPtr>();
+            event.executable = perf_event.get("executable").to_string();
+
+            auto sampled_process = adopt_own(*new Process {
+                .pid = event.pid,
+                .executable = event.executable,
+                .start_valid = event.timestamp,
+            });
+
+            current_processes.set(sampled_process->pid, sampled_process);
+            all_processes.append(move(sampled_process));
+            continue;
+        } else if (event.type == "process_exec") {
+            event.executable = perf_event.get("executable").to_string();
+
+            auto old_process = current_processes.get(event.pid).value();
+            old_process->end_valid = event.timestamp - 1;
+
+            current_processes.remove(event.pid);
+
+            auto sampled_process = adopt_own(*new Process {
+                .pid = event.pid,
+                .executable = event.executable,
+                .start_valid = event.timestamp });
+
+            current_processes.set(sampled_process->pid, sampled_process);
+            all_processes.append(move(sampled_process));
+            continue;
+        } else if (event.type == "process_exit") {
+            auto old_process = current_processes.get(event.pid).value();
+            old_process->end_valid = event.timestamp - 1;
+
+            current_processes.remove(event.pid);
+            continue;
+        } else if (event.type == "thread_create") {
+            event.parent_tid = perf_event.get("parent_tid").to_i32();
+            auto it = current_processes.find(event.pid);
+            if (it != current_processes.end())
+                it->value->handle_thread_create(event.tid, event.timestamp);
+            continue;
+        } else if (event.type == "thread_exit") {
+            auto it = current_processes.find(event.pid);
+            if (it != current_processes.end())
+                it->value->handle_thread_exit(event.tid, event.timestamp);
+            continue;
         }
 
         auto stack_array = perf_event.get("stack").as_array();
@@ -283,22 +304,19 @@ Result<NonnullOwnPtr<Profile>, String> Profile::load_from_perfcore_file(const St
                 if (kernel_elf) {
                     symbol = kernel_elf->symbolicate(ptr, &offset);
                 } else {
-                    symbol = "??";
+                    symbol = String::formatted("?? <{:p}>", ptr);
                 }
             } else {
-                auto it = sampled_processes.find_if([&](auto& entry) {
-                    // FIXME: This doesn't support multi-threaded programs!
-                    return entry.pid == event.tid;
-                });
+                auto it = current_processes.find(event.pid);
                 // FIXME: This logic is kinda gnarly, find a way to clean it up.
                 LibraryMetadata* library_metadata {};
-                if (!it.is_end())
-                    library_metadata = it->library_metadata.ptr();
+                if (it != current_processes.end())
+                    library_metadata = &it->value->library_metadata;
                 if (auto* library = library_metadata ? library_metadata->library_containing(ptr) : nullptr) {
                     object_name = library->name;
                     symbol = library->symbolicate(ptr, &offset);
                 } else {
-                    symbol = "??";
+                    symbol = String::formatted("?? <{:p}>", ptr);
                 }
             }
 
@@ -314,7 +332,21 @@ Result<NonnullOwnPtr<Profile>, String> Profile::load_from_perfcore_file(const St
         events.append(move(event));
     }
 
-    return adopt_own(*new Profile(move(sampled_processes), move(events)));
+    if (events.is_empty())
+        return String { "No events captured (targeted process was never on CPU)" };
+
+    quick_sort(all_processes, [](auto& a, auto& b) {
+        if (a.pid == b.pid)
+            return a.start_valid < b.start_valid;
+        else
+            return a.pid < b.pid;
+    });
+
+    Vector<Process> processes;
+    for (auto& it : all_processes)
+        processes.append(move(it));
+
+    return adopt_own(*new Profile(move(processes), move(events)));
 }
 
 void ProfileNode::sort_children()
@@ -344,6 +376,32 @@ void Profile::clear_timestamp_filter_range()
     m_samples_model->update();
 }
 
+void Profile::set_process_filter(pid_t pid, u64 start_valid, u64 end_valid)
+{
+    if (m_has_process_filter && m_process_filter_pid == pid && m_process_filter_start_valid == start_valid && m_process_filter_end_valid == end_valid)
+        return;
+    m_has_process_filter = true;
+
+    m_process_filter_pid = pid;
+    m_process_filter_start_valid = start_valid;
+    m_process_filter_end_valid = end_valid;
+
+    rebuild_tree();
+    if (m_disassembly_model)
+        m_disassembly_model->update();
+    m_samples_model->update();
+}
+void Profile::clear_process_filter()
+{
+    if (!m_has_process_filter)
+        return;
+    m_has_process_filter = false;
+    rebuild_tree();
+    if (m_disassembly_model)
+        m_disassembly_model->update();
+    m_samples_model->update();
+}
+
 void Profile::set_inverted(bool inverted)
 {
     if (m_inverted == inverted)
@@ -381,87 +439,6 @@ GUI::Model* Profile::disassembly_model()
     return m_disassembly_model;
 }
 
-HashMap<String, OwnPtr<MappedObject>> g_mapped_object_cache;
-
-static MappedObject* get_or_create_mapped_object(const String& path)
-{
-    if (auto it = g_mapped_object_cache.find(path); it != g_mapped_object_cache.end())
-        return it->value.ptr();
-
-    auto file_or_error = MappedFile::map(path);
-    if (file_or_error.is_error()) {
-        g_mapped_object_cache.set(path, {});
-        return nullptr;
-    }
-    auto elf = ELF::Image(file_or_error.value()->bytes());
-    if (!elf.is_valid()) {
-        g_mapped_object_cache.set(path, {});
-        return nullptr;
-    }
-    auto new_mapped_object = adopt_own(*new MappedObject {
-        .file = file_or_error.release_value(),
-        .elf = move(elf),
-    });
-    auto* ptr = new_mapped_object.ptr();
-    g_mapped_object_cache.set(path, move(new_mapped_object));
-    return ptr;
-}
-
-LibraryMetadata::LibraryMetadata(JsonArray regions)
-    : m_regions(move(regions))
-{
-    for (auto& region_value : m_regions.values()) {
-        auto& region = region_value.as_object();
-        auto base = region.get("base").as_u32();
-        auto size = region.get("size").as_u32();
-        auto name = region.get("name").as_string();
-
-        String path;
-        if (name.contains("Loader.so"))
-            path = "Loader.so";
-        else if (!name.contains(":"))
-            continue;
-        else
-            path = name.substring(0, name.view().find_first_of(":").value());
-
-        if (name.contains(".so"))
-            path = String::formatted("/usr/lib/{}", path);
-
-        auto* mapped_object = get_or_create_mapped_object(path);
-        if (!mapped_object)
-            continue;
-
-        FlatPtr text_base {};
-        mapped_object->elf.for_each_program_header([&](const ELF::Image::ProgramHeader& ph) {
-            if (ph.is_executable())
-                text_base = ph.vaddr().get();
-            return IterationDecision::Continue;
-        });
-
-        m_libraries.set(name, adopt_own(*new Library { base, size, name, text_base, mapped_object }));
-    }
-}
-
-String LibraryMetadata::Library::symbolicate(FlatPtr ptr, u32* offset) const
-{
-    if (!object)
-        return "??"sv;
-
-    return object->elf.symbolicate(ptr - base + text_base, offset);
-}
-
-const LibraryMetadata::Library* LibraryMetadata::library_containing(FlatPtr ptr) const
-{
-    for (auto& it : m_libraries) {
-        if (!it.value)
-            continue;
-        auto& library = *it.value;
-        if (ptr >= library.base && ptr < (library.base + library.size))
-            return &library;
-    }
-    return nullptr;
-}
-
 ProfileNode::ProfileNode(const String& object_name, String symbol, u32 address, u32 offset, u64 timestamp, pid_t pid)
     : m_symbol(move(symbol))
     , m_pid(pid)
@@ -478,7 +455,7 @@ ProfileNode::ProfileNode(const String& object_name, String symbol, u32 address,
     m_object_name = LexicalPath(object).basename();
 }
 
-const Process* ProfileNode::process(Profile& profile) const
+const Process* ProfileNode::process(Profile& profile, u64 timestamp) const
 {
-    return profile.find_process(m_pid);
+    return profile.find_process(m_pid, timestamp);
 }

Userland/DevTools/Profiler/Profile.h

@@ -6,6 +6,7 @@
 
 #pragma once
 
+#include "Process.h"
 #include <AK/Bitmap.h>
 #include <AK/FlyString.h>
 #include <AK/JsonArray.h>
@@ -24,49 +25,6 @@ class Profile;
 class ProfileModel;
 class SamplesModel;
 
-struct MappedObject {
-    NonnullRefPtr<MappedFile> file;
-    ELF::Image elf;
-};
-
-extern HashMap<String, OwnPtr<MappedObject>> g_mapped_object_cache;
-
-class LibraryMetadata {
-public:
-    explicit LibraryMetadata(JsonArray regions);
-
-    struct Library {
-        FlatPtr base;
-        size_t size;
-        String name;
-        FlatPtr text_base;
-        MappedObject* object { nullptr };
-
-        String symbolicate(FlatPtr, u32* offset) const;
-    };
-
-    const Library* library_containing(FlatPtr) const;
-
-private:
-    mutable HashMap<String, OwnPtr<Library>> m_libraries;
-    JsonArray m_regions;
-};
-
-struct Process {
-    pid_t pid {};
-    String executable;
-    HashTable<int> threads;
-
-    struct Region {
-        String name;
-        FlatPtr base {};
-        size_t size {};
-    };
-    Vector<Region> regions;
-
-    NonnullOwnPtr<LibraryMetadata> library_metadata;
-};
-
 class ProfileNode : public RefCounted<ProfileNode> {
 public:
     static NonnullRefPtr<ProfileNode> create(FlyString object_name, String symbol, u32 address, u32 offset, u64 timestamp, pid_t pid)
@@ -137,7 +95,7 @@ public:
 
     pid_t pid() const { return m_pid; }
 
-    const Process* process(Profile&) const;
+    const Process* process(Profile&, u64 timestamp) const;
 
 private:
     explicit ProfileNode(const String& object_name, String symbol, u32 address, u32 offset, u64 timestamp, pid_t);
@@ -165,10 +123,10 @@ public:
     GUI::Model& samples_model();
     GUI::Model* disassembly_model();
 
-    const Process* find_process(pid_t pid) const
+    const Process* find_process(pid_t pid, u64 timestamp) const
     {
         auto it = m_processes.find_if([&](auto& entry) {
-            return entry.pid == pid;
+            return entry.pid == pid && entry.valid_at(timestamp);
         });
         return it.is_end() ? nullptr : &(*it);
     }
@@ -189,15 +147,18 @@ public:
         String type;
         FlatPtr ptr { 0 };
         size_t size { 0 };
+        String name;
+        int parent_pid { 0 };
+        int parent_tid { 0 };
+        String executable;
+        int pid { 0 };
         int tid { 0 };
         bool in_kernel { false };
         Vector<Frame> frames;
     };
 
-    u32 first_filtered_event_index() const { return m_first_filtered_event_index; }
-    u32 filtered_event_count() const { return m_filtered_event_count; }
-
     const Vector<Event>& events() const { return m_events; }
+    const Vector<size_t>& filtered_event_indices() const { return m_filtered_event_indices; }
 
     u64 length_in_ms() const { return m_last_timestamp - m_first_timestamp; }
     u64 first_timestamp() const { return m_first_timestamp; }
@@ -208,6 +169,10 @@ public:
     void clear_timestamp_filter_range();
     bool has_timestamp_filter_range() const { return m_has_timestamp_filter_range; }
 
+    void set_process_filter(pid_t pid, u64 start_valid, u64 end_valid);
+    void clear_process_filter();
+    bool has_process_filter() const { return m_has_process_filter; }
+
     bool is_inverted() const { return m_inverted; }
     void set_inverted(bool);
 
@@ -216,6 +181,8 @@ public:
     bool show_percentages() const { return m_show_percentages; }
     void set_show_percentages(bool);
 
+    const Vector<Process>& processes() const { return m_processes; }
+
     template<typename Callback>
     void for_each_event_in_filter_range(Callback callback)
     {
@@ -241,8 +208,7 @@ private:
     GUI::ModelIndex m_disassembly_index;
 
     Vector<NonnullRefPtr<ProfileNode>> m_roots;
-    u32 m_filtered_event_count { 0 };
-    size_t m_first_filtered_event_index { 0 };
+    Vector<size_t> m_filtered_event_indices;
     u64 m_first_timestamp { 0 };
     u64 m_last_timestamp { 0 };
 
@@ -253,6 +219,11 @@ private:
     u64 m_timestamp_filter_range_start { 0 };
     u64 m_timestamp_filter_range_end { 0 };
 
+    bool m_has_process_filter { false };
+    pid_t m_process_filter_pid { 0 };
+    u64 m_process_filter_start_valid { 0 };
+    u64 m_process_filter_end_valid { 0 };
+
     u32 m_deepest_stack_depth { 0 };
     bool m_inverted { false };
     bool m_show_top_functions { false };

Userland/DevTools/Profiler/ProfileModel.cpp

@@ -108,12 +108,12 @@ GUI::Variant ProfileModel::data(const GUI::ModelIndex& index, GUI::ModelRole rol
     if (role == GUI::ModelRole::Display) {
         if (index.column() == Column::SampleCount) {
             if (m_profile.show_percentages())
-                return ((float)node->event_count() / (float)m_profile.filtered_event_count()) * 100.0f;
+                return ((float)node->event_count() / (float)m_profile.filtered_event_indices().size()) * 100.0f;
             return node->event_count();
         }
         if (index.column() == Column::SelfCount) {
             if (m_profile.show_percentages())
-                return ((float)node->self_count() / (float)m_profile.filtered_event_count()) * 100.0f;
+                return ((float)node->self_count() / (float)m_profile.filtered_event_indices().size()) * 100.0f;
             return node->self_count();
         }
         if (index.column() == Column::ObjectName)

Userland/DevTools/Profiler/SamplesModel.cpp

@@ -22,7 +22,7 @@ SamplesModel::~SamplesModel()
 
 int SamplesModel::row_count(const GUI::ModelIndex&) const
 {
-    return m_profile.filtered_event_count();
+    return m_profile.filtered_event_indices().size();
 }
 
 int SamplesModel::column_count(const GUI::ModelIndex&) const
@@ -37,6 +37,8 @@ String SamplesModel::column_name(int column) const
         return "#";
     case Column::Timestamp:
         return "Timestamp";
+    case Column::ProcessID:
+        return "PID";
     case Column::ThreadID:
         return "TID";
     case Column::ExecutableName:
@@ -50,7 +52,7 @@ String SamplesModel::column_name(int column) const
 
 GUI::Variant SamplesModel::data(const GUI::ModelIndex& index, GUI::ModelRole role) const
 {
-    u32 event_index = m_profile.first_filtered_event_index() + index.row();
+    u32 event_index = m_profile.filtered_event_indices()[index.row()];
     auto& event = m_profile.events().at(event_index);
 
     if (role == GUI::ModelRole::Custom) {
@@ -61,12 +63,14 @@ GUI::Variant SamplesModel::data(const GUI::ModelIndex& index, GUI::ModelRole rol
         if (index.column() == Column::SampleIndex)
             return event_index;
 
+        if (index.column() == Column::ProcessID)
+            return event.pid;
+
         if (index.column() == Column::ThreadID)
             return event.tid;
 
         if (index.column() == Column::ExecutableName) {
-            // FIXME: More abuse of the PID/TID relationship:
-            if (auto* process = m_profile.find_process(event.tid))
+            if (auto* process = m_profile.find_process(event.pid, event.timestamp))
                 return process->executable;
             return "";
         }

Userland/DevTools/Profiler/SamplesModel.h

@@ -20,6 +20,7 @@ public:
     enum Column {
         SampleIndex,
         Timestamp,
+        ProcessID,
         ThreadID,
         ExecutableName,
         InnermostStackFrame,

Userland/DevTools/Profiler/main.cpp

@@ -5,6 +5,7 @@
  */
 
 #include "IndividualSampleModel.h"
+#include "ProcessPickerWidget.h"
 #include "Profile.h"
 #include "ProfileTimelineWidget.h"
 #include <LibCore/ArgsParser.h>
@@ -86,6 +87,7 @@ int main(int argc, char** argv)
     main_widget.set_layout<GUI::VerticalBoxLayout>();
 
     main_widget.add<ProfileTimelineWidget>(*profile);
+    main_widget.add<ProcessPickerWidget>(*profile);
 
     auto& tab_widget = main_widget.add<GUI::TabWidget>();
 

Userland/Libraries/LibC/serenity.h

@@ -75,9 +75,18 @@ int futex(uint32_t* userspace_address, int futex_op, uint32_t value, const struc
 
 int purge(int mode);
 
-#define PERF_EVENT_SAMPLE 0
-#define PERF_EVENT_MALLOC 1
-#define PERF_EVENT_FREE 2
+enum {
+    PERF_EVENT_SAMPLE,
+    PERF_EVENT_MALLOC,
+    PERF_EVENT_FREE,
+    PERF_EVENT_MMAP,
+    PERF_EVENT_MUNMAP,
+    PERF_EVENT_PROCESS_CREATE,
+    PERF_EVENT_PROCESS_EXEC,
+    PERF_EVENT_PROCESS_EXIT,
+    PERF_EVENT_THREAD_CREATE,
+    PERF_EVENT_THREAD_EXIT
+};
 
 int perf_event(int type, uintptr_t arg1, uintptr_t arg2);