/*
 * Copyright 2019 Google, Inc.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are
 * met: redistributions of source code must retain the above copyright
 * notice, this list of conditions and the following disclaimer;
 * redistributions in binary form must reproduce the above copyright
 * notice, this list of conditions and the following disclaimer in the
 * documentation and/or other materials provided with the distribution;
 * neither the name of the copyright holders nor the names of its
 * contributors may be used to endorse or promote products derived from
 * this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * Authors: Gabe Black
 */

#include "arch/arm/fastmodel/iris/thread_context.hh"

#include <utility>

#include "iris/detail/IrisCppAdapter.h"
#include "iris/detail/IrisObjects.h"
#include "mem/fs_translating_port_proxy.hh"
#include "mem/se_translating_port_proxy.hh"

namespace Iris
{

void
ThreadContext::initFromIrisInstance(const ResourceMap &resources)
{
    bool enabled = false;
    call().perInstanceExecution_getState(_instId, enabled);
    _status = enabled ? Active : Suspended;

    suspend();

    call().memory_getMemorySpaces(_instId, memorySpaces);
    call().memory_getUsefulAddressTranslations(_instId, translations);

    typedef ThreadContext Self;
    iris::EventSourceInfo evSrcInfo;

    client.registerEventCallback<Self, &Self::breakpointHit>(
            this, "ec_IRIS_BREAKPOINT_HIT",
            "Handle hitting a breakpoint", "Iris::ThreadContext");
    call().event_getEventSource(_instId, evSrcInfo, "IRIS_BREAKPOINT_HIT");
    call().eventStream_create(_instId, breakpointEventStreamId,
            evSrcInfo.evSrcId, client.getInstId());

    for (auto it = bps.begin(); it != bps.end(); it++)
        installBp(it);
}

iris::ResourceId
ThreadContext::extractResourceId(
        const ResourceMap &resources, const std::string &name)
{
    return resources.at(name).rscId;
}

void
ThreadContext::extractResourceMap(
        ResourceIds &ids, const ResourceMap &resources,
        const IdxNameMap &idx_names)
{
    for (const auto &idx_name: idx_names) {
        int idx = idx_name.first;
        const std::string &name = idx_name.second;

        if (idx >= ids.size())
            ids.resize(idx + 1, iris::IRIS_UINT64_MAX);

        ids[idx] = extractResourceId(resources, name);
    }
}

void
ThreadContext::maintainStepping()
{
    Tick now = 0;

    while (true) {
        if (comInstEventQueue.empty()) {
            // Set to 0 to deactivate stepping.
            call().step_setup(_instId, 0, "instruction");
            break;
        }

        Tick next = comInstEventQueue.nextTick();
        if (!now)
            now = getCurrentInstCount();

        if (next <= now) {
            comInstEventQueue.serviceEvents(now);
            // Start over now that comInstEventQueue has likely changed.
            continue;
        }

        // Set to the number of instructions still to step through.
        Tick remaining = next - now;
        call().step_setup(_instId, remaining, "instruction");
        break;
    }
}

ThreadContext::BpInfoIt
ThreadContext::getOrAllocBp(Addr pc)
{
    auto pc_it = bps.find(pc);

    if (pc_it != bps.end())
        return pc_it;

    auto res = bps.emplace(std::make_pair(pc, new BpInfo(pc)));
    panic_if(!res.second, "Inserting breakpoint failed.");
    return res.first;
}

void
ThreadContext::installBp(BpInfoIt it)
{
    BpId id;
    Addr pc = it->second->pc;
    auto space_id = getBpSpaceId(pc);
    call().breakpoint_set_code(_instId, id, pc, space_id, 0, true);
    it->second->id = id;
}

void
ThreadContext::uninstallBp(BpInfoIt it)
{
    call().breakpoint_delete(_instId, it->second->id);
    it->second->clearId();
}

void
ThreadContext::delBp(BpInfoIt it)
{
    panic_if(!it->second->empty(),
             "BP info still had events associated with it.");

    if (it->second->validId())
        uninstallBp(it);

    bps.erase(it);
}

iris::IrisErrorCode
ThreadContext::instanceRegistryChanged(
        uint64_t esId, const iris::IrisValueMap &fields, uint64_t time,
        uint64_t sInstId, bool syncEc, std::string &error_message_out)
{
    const std::string &event = fields.at("EVENT").getString();
    const iris::InstanceId id = fields.at("INST_ID").getU64();
    const std::string &name = fields.at("INST_NAME").getString();

    if (name != "component." + _irisPath)
        return iris::E_ok;

    if (event == "added")
        _instId = id;
    else if (event == "removed")
        _instId = iris::IRIS_UINT64_MAX;
    else
        panic("Unrecognized event type %s", event);

    return iris::E_ok;
}

iris::IrisErrorCode
ThreadContext::phaseInitLeave(
        uint64_t esId, const iris::IrisValueMap &fields, uint64_t time,
        uint64_t sInstId, bool syncEc, std::string &error_message_out)
{
    std::vector<iris::ResourceInfo> resources;
    call().resource_getList(_instId, resources);

    ResourceMap resourceMap;
    for (auto &resource: resources)
        resourceMap[resource.name] = resource;

    initFromIrisInstance(resourceMap);

    return iris::E_ok;
}

iris::IrisErrorCode
ThreadContext::simulationTimeEvent(
        uint64_t esId, const iris::IrisValueMap &fields, uint64_t time,
        uint64_t sInstId, bool syncEc, std::string &error_message_out)
{
    if (fields.at("RUNNING").getAsBool()) {
        // If this is just simulation time starting up, don't do anything.
        return iris::E_ok;
    }

    // If simulation time has stopped for any reason, IRIS helpfully clears
    // all stepping counters and we need to set them back. We might also need
    // to service events based on the current number of executed instructions.
    maintainStepping();

    // Restart simulation time to make sure things progress once we give
    // control back.
    call().simulationTime_run(iris::IrisInstIdSimulationEngine);

    return iris::E_ok;
}

iris::IrisErrorCode
ThreadContext::breakpointHit(
        uint64_t esId, const iris::IrisValueMap &fields, uint64_t time,
        uint64_t sInstId, bool syncEc, std::string &error_message_out)
{
    Addr pc = fields.at("PC").getU64();

    auto it = getOrAllocBp(pc);

    auto e_it = it->second->events.begin();
    while (e_it != it->second->events.end()) {
        PCEvent *e = *e_it;
        // Advance e_it here since e might remove itself from the list.
        e_it++;
        e->process(this);
    }

    return iris::E_ok;
}

ThreadContext::ThreadContext(
        BaseCPU *cpu, int id, System *system, ::BaseTLB *dtb, ::BaseTLB *itb,
        iris::IrisConnectionInterface *iris_if, const std::string &iris_path) :
    _cpu(cpu), _threadId(id), _system(system), _dtb(dtb), _itb(itb),
    _irisPath(iris_path), vecRegs(ArmISA::NumVecRegs),
    vecPredRegs(ArmISA::NumVecPredRegs),
    comInstEventQueue("instruction-based event queue"),
    client(iris_if, "client." + iris_path)
{
    iris::InstanceInfo info;
    auto ret_code = noThrow().instanceRegistry_getInstanceInfoByName(
                info, "component." + iris_path);
    if (ret_code == iris::E_ok) {
        // The iris instance registry already new about this path.
        _instId = info.instId;
    } else {
        // This path doesn't (yet) exist. Set the ID to something invalid.
        _instId = iris::IRIS_UINT64_MAX;
    }

    typedef ThreadContext Self;
    iris::EventSourceInfo evSrcInfo;

    client.registerEventCallback<Self, &Self::instanceRegistryChanged>(
            this, "ec_IRIS_INSTANCE_REGISTRY_CHANGED",
            "Install the iris instance ID", "Iris::ThreadContext");
    call().event_getEventSource(iris::IrisInstIdGlobalInstance, evSrcInfo,
            "IRIS_INSTANCE_REGISTRY_CHANGED");
    regEventStreamId = iris::IRIS_UINT64_MAX;
    static const std::vector<std::string> fields =
        { "EVENT", "INST_ID", "INST_NAME" };
    call().eventStream_create(iris::IrisInstIdGlobalInstance, regEventStreamId,
            evSrcInfo.evSrcId, client.getInstId(), &fields);

    client.registerEventCallback<Self, &Self::phaseInitLeave>(
            this, "ec_IRIS_SIM_PHASE_INIT_LEAVE",
            "Initialize register contexts", "Iris::ThreadContext");
    call().event_getEventSource(iris::IrisInstIdSimulationEngine, evSrcInfo,
            "IRIS_SIM_PHASE_INIT_LEAVE");
    initEventStreamId = iris::IRIS_UINT64_MAX;
    call().eventStream_create(
            iris::IrisInstIdSimulationEngine, initEventStreamId,
            evSrcInfo.evSrcId, client.getInstId());

    client.registerEventCallback<Self, &Self::simulationTimeEvent>(
            this, "ec_IRIS_SIMULATION_TIME_EVENT",
            "Handle simulation time stopping for breakpoints or stepping",
            "Iris::ThreadContext");
    call().event_getEventSource(iris::IrisInstIdSimulationEngine, evSrcInfo,
            "IRIS_SIMULATION_TIME_EVENT");
    timeEventStreamId = iris::IRIS_UINT64_MAX;
    call().eventStream_create(
            iris::IrisInstIdSimulationEngine, timeEventStreamId,
            evSrcInfo.evSrcId, client.getInstId());

    breakpointEventStreamId = iris::IRIS_UINT64_MAX;
}

ThreadContext::~ThreadContext()
{
    call().eventStream_destroy(
            iris::IrisInstIdSimulationEngine, initEventStreamId);
    initEventStreamId = iris::IRIS_UINT64_MAX;
    client.unregisterEventCallback("ec_IRIS_SIM_PHASE_INIT_LEAVE");

    call().eventStream_destroy(
            iris::IrisInstIdGlobalInstance, regEventStreamId);
    regEventStreamId = iris::IRIS_UINT64_MAX;
    client.unregisterEventCallback("ec_IRIS_INSTANCE_REGISTRY_CHANGED");

    call().eventStream_destroy(
            iris::IrisInstIdGlobalInstance, timeEventStreamId);
    timeEventStreamId = iris::IRIS_UINT64_MAX;
    client.unregisterEventCallback("ec_IRIS_SIMULATION_TIME_EVENT");
}

bool
ThreadContext::schedule(PCEvent *e)
{
    auto it = getOrAllocBp(e->pc());
    it->second->events.push_back(e);

    if (_instId != iris::IRIS_UINT64_MAX && !it->second->validId())
        installBp(it);

    return true;
}

bool
ThreadContext::remove(PCEvent *e)
{
    auto it = getOrAllocBp(e->pc());
    it->second->events.remove(e);

    if (it->second->empty())
        delBp(it);

    return true;
}

bool
ThreadContext::translateAddress(Addr &paddr, iris::MemorySpaceId p_space,
                                Addr vaddr, iris::MemorySpaceId v_space)
{
    iris::MemoryAddressTranslationResult result;
    auto ret = noThrow().memory_translateAddress(
            _instId, result, v_space, vaddr, p_space);

    if (ret != iris::E_ok) {
        // Check if there was  a legal translation between these two spaces.
        // If so, something else went wrong.
        for (auto &trans: translations)
            if (trans.inSpaceId == v_space && trans.outSpaceId == p_space)
                return false;

        panic("No legal translation IRIS address translation found.");
    }

    if (result.address.empty())
        return false;

    if (result.address.size() > 1) {
        warn("Multiple mappings for address %#x.", vaddr);
        return false;
    }

    paddr = result.address[0];
    return true;
}

void
ThreadContext::scheduleInstCountEvent(Event *event, Tick count)
{
    Tick now = getCurrentInstCount();
    comInstEventQueue.schedule(event, count);
    if (count <= now)
        call().simulationTime_stop(iris::IrisInstIdSimulationEngine);
    else
        maintainStepping();
}

void
ThreadContext::descheduleInstCountEvent(Event *event)
{
    comInstEventQueue.deschedule(event);
    maintainStepping();
}

Tick
ThreadContext::getCurrentInstCount()
{
    uint64_t count;
    auto ret = call().step_getStepCounterValue(_instId, count, "instruction");
    panic_if(ret != iris::E_ok, "Failed to get instruction count.");
    return count;
}

void
ThreadContext::initMemProxies(::ThreadContext *tc)
{
    if (FullSystem) {
        assert(!physProxy && !virtProxy);
        physProxy.reset(new PortProxy(_cpu->getSendFunctional(),
                                      _cpu->cacheLineSize()));
        virtProxy.reset(new FSTranslatingPortProxy(tc));
    } else {
        assert(!virtProxy);
        virtProxy.reset(new SETranslatingPortProxy(
                        _cpu->getSendFunctional(), getProcessPtr(),
                        SETranslatingPortProxy::NextPage));
    }
}

ThreadContext::Status
ThreadContext::status() const
{
    return _status;
}

void
ThreadContext::setStatus(Status new_status)
{
    if (new_status == Active) {
        if (_status != Active)
            call().perInstanceExecution_setState(_instId, true);
    } else {
        if (_status == Active)
            call().perInstanceExecution_setState(_instId, false);
    }
    _status = new_status;
}

ArmISA::PCState
ThreadContext::pcState() const
{
    ArmISA::CPSR cpsr = readMiscRegNoEffect(ArmISA::MISCREG_CPSR);
    ArmISA::PCState pc;

    pc.thumb(cpsr.t);
    pc.nextThumb(pc.thumb());
    pc.jazelle(cpsr.j);
    pc.nextJazelle(cpsr.j);
    pc.aarch64(!cpsr.width);
    pc.nextAArch64(!cpsr.width);
    pc.illegalExec(false);

    iris::ResourceReadResult result;
    call().resource_read(_instId, result, pcRscId);
    Addr addr = result.data.at(0);
    if (cpsr.width && cpsr.t)
        addr = addr & ~0x1;
    pc.set(addr);

    return pc;
}
void
ThreadContext::pcState(const ArmISA::PCState &val)
{
    Addr pc = val.pc();

    ArmISA::CPSR cpsr = readMiscRegNoEffect(ArmISA::MISCREG_CPSR);
    if (cpsr.width && cpsr.t)
        pc = pc | 0x1;

    iris::ResourceWriteResult result;
    call().resource_write(_instId, result, pcRscId, pc);
}

Addr
ThreadContext::instAddr() const
{
    return pcState().instAddr();
}

Addr
ThreadContext::nextInstAddr() const
{
    return pcState().nextInstAddr();
}

RegVal
ThreadContext::readMiscRegNoEffect(RegIndex misc_reg) const
{
    iris::ResourceReadResult result;
    call().resource_read(_instId, result, miscRegIds.at(misc_reg));
    return result.data.at(0);
}

void
ThreadContext::setMiscRegNoEffect(RegIndex misc_reg, const RegVal val)
{
    iris::ResourceWriteResult result;
    call().resource_write(_instId, result, miscRegIds.at(misc_reg), val);
}

RegVal
ThreadContext::readIntReg(RegIndex reg_idx) const
{
    ArmISA::CPSR cpsr = readMiscRegNoEffect(ArmISA::MISCREG_CPSR);

    iris::ResourceReadResult result;
    if (cpsr.width)
        call().resource_read(_instId, result, intReg32Ids.at(reg_idx));
    else
        call().resource_read(_instId, result, intReg64Ids.at(reg_idx));
    return result.data.at(0);
}

void
ThreadContext::setIntReg(RegIndex reg_idx, RegVal val)
{
    ArmISA::CPSR cpsr = readMiscRegNoEffect(ArmISA::MISCREG_CPSR);

    iris::ResourceWriteResult result;
    if (cpsr.width)
        call().resource_write(_instId, result, intReg32Ids.at(reg_idx), val);
    else
        call().resource_write(_instId, result, intReg64Ids.at(reg_idx), val);
}

RegVal
ThreadContext::readCCRegFlat(RegIndex idx) const
{
    if (idx >= ccRegIds.size())
        return 0;
    iris::ResourceReadResult result;
    call().resource_read(_instId, result, ccRegIds.at(idx));
    return result.data.at(0);
}

void
ThreadContext::setCCRegFlat(RegIndex idx, RegVal val)
{
    panic_if(idx >= ccRegIds.size(),
            "CC reg %d is not supported by fast model.", idx);
    iris::ResourceWriteResult result;
    call().resource_write(_instId, result, ccRegIds.at(idx), val);
}

const ArmISA::VecRegContainer &
ThreadContext::readVecReg(const RegId &reg_id) const
{
    const RegIndex idx = reg_id.index();
    // Ignore accesses to registers which aren't architected. gem5 defines a
    // few extra registers which it uses internally in the implementation of
    // some instructions.
    if (idx >= vecRegIds.size())
        return vecRegs.at(idx);
    ArmISA::VecRegContainer &reg = vecRegs.at(idx);

    iris::ResourceReadResult result;
    call().resource_read(_instId, result, vecRegIds.at(idx));
    size_t data_size = result.data.size() * (sizeof(*result.data.data()));
    size_t size = std::min(data_size, reg.SIZE);
    memcpy(reg.raw_ptr<void>(), (void *)result.data.data(), size);

    return reg;
}

const ArmISA::VecRegContainer &
ThreadContext::readVecRegFlat(RegIndex idx) const
{
    return readVecReg(RegId(VecRegClass, idx));
}

const ArmISA::VecPredRegContainer &
ThreadContext::readVecPredReg(const RegId &reg_id) const
{
    RegIndex idx = reg_id.index();
    if (idx >= vecPredRegIds.size())
        return vecPredRegs.at(idx);

    ArmISA::VecPredRegContainer &reg = vecPredRegs.at(idx);

    iris::ResourceReadResult result;
    call().resource_read(_instId, result, vecPredRegIds.at(idx));

    size_t offset = 0;
    size_t num_bits = reg.NUM_BITS;
    uint8_t *bytes = (uint8_t *)result.data.data();
    while (num_bits > 8) {
        reg.set_bits(offset, 8, *bytes);
        offset += 8;
        num_bits -= 8;
        bytes++;
    }
    if (num_bits)
        reg.set_bits(offset, num_bits, *bytes);

    return reg;
}

const ArmISA::VecPredRegContainer &
ThreadContext::readVecPredRegFlat(RegIndex idx) const
{
    return readVecPredReg(RegId(VecPredRegClass, idx));
}

} // namespace Iris