blob: a22637f9835b9af274f3f98d02497b56215f4618 [file] [log] [blame]
/*
* Copyright (c) 2012 ARM Limited
* All rights reserved
*
* The license below extends only to copyright in the software and shall
* not be construed as granting a license to any other intellectual
* property including but not limited to intellectual property relating
* to a hardware implementation of the functionality of the software
* licensed hereunder. You may use the software subject to the license
* terms below provided that you ensure that this notice is replicated
* unmodified and in its entirety in all distributions of the software,
* modified or unmodified, in source code or in binary form.
*
* 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: Andreas Sandberg
*/
#ifndef __CPU_KVM_BASE_HH__
#define __CPU_KVM_BASE_HH__
#include <pthread.h>
#include <csignal>
#include <memory>
#include <queue>
#include "base/statistics.hh"
#include "cpu/kvm/perfevent.hh"
#include "cpu/kvm/timer.hh"
#include "cpu/kvm/vm.hh"
#include "cpu/base.hh"
#include "cpu/simple_thread.hh"
/** Signal to use to trigger exits from KVM */
#define KVM_KICK_SIGNAL SIGRTMIN
// forward declarations
class ThreadContext;
struct BaseKvmCPUParams;
/**
* Base class for KVM based CPU models
*
* All architecture specific KVM implementation should inherit from
* this class. The most basic CPU models only need to override the
* updateKvmState() and updateThreadContext() methods to implement
* state synchronization between gem5 and KVM.
*
* The architecture specific implementation is also responsible for
* delivering interrupts into the VM. This is typically done by
* overriding tick() and checking the thread context before entering
* into the VM. In order to deliver an interrupt, the implementation
* then calls KvmVM::setIRQLine() or BaseKvmCPU::kvmInterrupt()
* depending on the specifics of the underlying hardware/drivers.
*/
class BaseKvmCPU : public BaseCPU
{
public:
BaseKvmCPU(BaseKvmCPUParams *params);
virtual ~BaseKvmCPU();
void init() override;
void startup() override;
void regStats() override;
void serializeThread(CheckpointOut &cp, ThreadID tid) const override;
void unserializeThread(CheckpointIn &cp, ThreadID tid) override;
DrainState drain() override;
void drainResume() override;
void notifyFork() override;
void switchOut() override;
void takeOverFrom(BaseCPU *cpu) override;
void verifyMemoryMode() const override;
MasterPort &getDataPort() override { return dataPort; }
MasterPort &getInstPort() override { return instPort; }
void wakeup(ThreadID tid = 0) override;
void activateContext(ThreadID thread_num) override;
void suspendContext(ThreadID thread_num) override;
void deallocateContext(ThreadID thread_num);
void haltContext(ThreadID thread_num) override;
long getVCpuID() const { return vcpuID; }
ThreadContext *getContext(int tn) override;
Counter totalInsts() const override;
Counter totalOps() const override;
/**
* Callback from KvmCPUPort to transition the CPU out of RunningMMIOPending
* when all timing requests have completed.
*/
void finishMMIOPending();
/** Dump the internal state to the terminal. */
virtual void dump() const;
/**
* Force an exit from KVM.
*
* Send a signal to the thread owning this vCPU to get it to exit
* from KVM. Ignored if the vCPU is not executing.
*/
void kick() const { pthread_kill(vcpuThread, KVM_KICK_SIGNAL); }
/**
* A cached copy of a thread's state in the form of a SimpleThread
* object.
*
* Normally the actual thread state is stored in the KVM vCPU. If KVM has
* been running this copy is will be out of date. If we recently handled
* some events within gem5 that required state to be updated this could be
* the most up-to-date copy. When getContext() or updateThreadContext() is
* called this copy gets updated. The method syncThreadContext can
* be used within a KVM CPU to update the thread context if the
* KVM state is dirty (i.e., the vCPU has been run since the last
* update).
*/
SimpleThread *thread;
/** ThreadContext object, provides an interface for external
* objects to modify this thread's state.
*/
ThreadContext *tc;
KvmVM &vm;
protected:
/**
*
* @dot
* digraph {
* Idle;
* Running;
* RunningService;
* RunningServiceCompletion;
* RunningMMIOPending;
*
* Idle -> Idle;
* Idle -> Running [label="activateContext()", URL="\ref activateContext"];
* Running -> Running [label="tick()", URL="\ref tick"];
* Running -> RunningService [label="tick()", URL="\ref tick"];
* Running -> Idle [label="suspendContext()", URL="\ref suspendContext"];
* Running -> Idle [label="drain()", URL="\ref drain"];
* Idle -> Running [label="drainResume()", URL="\ref drainResume"];
* RunningService -> RunningServiceCompletion [label="handleKvmExit()", URL="\ref handleKvmExit"];
* RunningService -> RunningMMIOPending [label="handleKvmExit()", URL="\ref handleKvmExit"];
* RunningMMIOPending -> RunningServiceCompletion [label="finishMMIOPending()", URL="\ref finishMMIOPending"];
* RunningServiceCompletion -> Running [label="tick()", URL="\ref tick"];
* RunningServiceCompletion -> RunningService [label="tick()", URL="\ref tick"];
* }
* @enddot
*/
enum Status {
/** Context not scheduled in KVM.
*
* The CPU generally enters this state when the guest execute
* an instruction that halts the CPU (e.g., WFI on ARM or HLT
* on X86) if KVM traps this instruction. Ticks are not
* scheduled in this state.
*
* @see suspendContext()
*/
Idle,
/** Running normally.
*
* This is the normal run state of the CPU. KVM will be
* entered next time tick() is called.
*/
Running,
/** Requiring service at the beginning of the next cycle.
*
* The virtual machine has exited and requires service, tick()
* will call handleKvmExit() on the next cycle. The next state
* after running service is determined in handleKvmExit() and
* depends on what kind of service the guest requested:
* <ul>
* <li>IO/MMIO (Atomic): RunningServiceCompletion
* <li>IO/MMIO (Timing): RunningMMIOPending
* <li>Halt: Idle
* <li>Others: Running
* </ul>
*/
RunningService,
/** Timing MMIO request in flight or stalled.
*
* The VM has requested IO/MMIO and we are in timing mode. A timing
* request is either stalled (and will be retried with recvReqRetry())
* or it is in flight. After the timing request is complete, the CPU
* will transition to the RunningServiceCompletion state.
*/
RunningMMIOPending,
/** Service completion in progress.
*
* The VM has requested service that requires KVM to be
* entered once in order to get to a consistent state. This
* happens in handleKvmExit() or one of its friends after IO
* exits. After executing tick(), the CPU will transition into
* the Running or RunningService state.
*/
RunningServiceCompletion,
};
/** CPU run state */
Status _status;
/**
* Execute the CPU until the next event in the main event queue or
* until the guest needs service from gem5.
*/
void tick();
/**
* Get the value of the hardware cycle counter in the guest.
*
* This method is supposed to return the total number of cycles
* executed in hardware mode relative to some arbitrary point in
* the past. It's mainly used when estimating the number of cycles
* actually executed by the CPU in kvmRun(). The default behavior
* of this method is to use the cycles performance counter, but
* some architectures may want to use internal registers instead.
*
* @return Number of host cycles executed relative to an undefined
* point in the past.
*/
virtual uint64_t getHostCycles() const;
/**
* Request KVM to run the guest for a given number of ticks. The
* method returns the approximate number of ticks executed.
*
* @note The returned number of ticks can be both larger or
* smaller than the requested number of ticks. A smaller number
* can, for example, occur when the guest executes MMIO. A larger
* number is typically due to performance counter inaccuracies.
*
* @note This method is virtual in order to allow implementations
* to check for architecture specific events (e.g., interrupts)
* before entering the VM.
*
* @note It is the response of the caller (normally tick()) to
* make sure that the KVM state is synchronized and that the TC is
* invalidated after entering KVM.
*
* @note This method does not normally cause any state
* transitions. However, if it may suspend the CPU by suspending
* the thread, which leads to a transition to the Idle state. In
* such a case, kvm <i>must not</i> be entered.
*
* @param ticks Number of ticks to execute, set to 0 to exit
* immediately after finishing pending operations.
* @return Number of ticks executed (see note)
*/
virtual Tick kvmRun(Tick ticks);
/**
* Request the CPU to run until draining completes.
*
* This function normally calls kvmRun(0) to make KVM finish
* pending MMIO operations. Architecures implementing
* archIsDrained() must override this method.
*
* @see BaseKvmCPU::archIsDrained()
*
* @return Number of ticks executed
*/
virtual Tick kvmRunDrain();
/**
* Get a pointer to the kvm_run structure containing all the input
* and output parameters from kvmRun().
*/
struct kvm_run *getKvmRunState() { return _kvmRun; };
/**
* Retrieve a pointer to guest data stored at the end of the
* kvm_run structure. This is mainly used for PIO operations
* (KVM_EXIT_IO).
*
* @param offset Offset as specified by the kvm_run structure
* @return Pointer to guest data
*/
uint8_t *getGuestData(uint64_t offset) const {
return (uint8_t *)_kvmRun + offset;
};
/**
* @addtogroup KvmInterrupts
* @{
*/
/**
* Send a non-maskable interrupt to the guest
*
* @note The presence of this call depends on Kvm::capUserNMI().
*/
void kvmNonMaskableInterrupt();
/**
* Send a normal interrupt to the guest
*
* @note Make sure that ready_for_interrupt_injection in kvm_run
* is set prior to calling this function. If not, an interrupt
* window must be requested by setting request_interrupt_window in
* kvm_run to 1 and restarting the guest.
*
* @param interrupt Structure describing the interrupt to send
*/
void kvmInterrupt(const struct kvm_interrupt &interrupt);
/** @} */
/** @{ */
/**
* Get/Set the register state of the guest vCPU
*
* KVM has two different interfaces for accessing the state of the
* guest CPU. One interface updates 'normal' registers and one
* updates 'special' registers. The distinction between special
* and normal registers isn't very clear and is architecture
* dependent.
*/
void getRegisters(struct kvm_regs &regs) const;
void setRegisters(const struct kvm_regs &regs);
void getSpecialRegisters(struct kvm_sregs &regs) const;
void setSpecialRegisters(const struct kvm_sregs &regs);
/** @} */
/** @{ */
/**
* Get/Set the guest FPU/vector state
*/
void getFPUState(struct kvm_fpu &state) const;
void setFPUState(const struct kvm_fpu &state);
/** @} */
/** @{ */
/**
* Get/Set single register using the KVM_(SET|GET)_ONE_REG API.
*
* @note The presence of this call depends on Kvm::capOneReg().
*/
void setOneReg(uint64_t id, const void *addr);
void setOneReg(uint64_t id, uint64_t value) { setOneReg(id, &value); }
void setOneReg(uint64_t id, uint32_t value) { setOneReg(id, &value); }
void getOneReg(uint64_t id, void *addr) const;
uint64_t getOneRegU64(uint64_t id) const {
uint64_t value;
getOneReg(id, &value);
return value;
}
uint32_t getOneRegU32(uint64_t id) const {
uint32_t value;
getOneReg(id, &value);
return value;
}
/** @} */
/**
* Get and format one register for printout.
*
* This function call getOneReg() to retrieve the contents of one
* register and automatically formats it for printing.
*
* @note The presence of this call depends on Kvm::capOneReg().
*/
std::string getAndFormatOneReg(uint64_t id) const;
/** @{ */
/**
* Update the KVM state from the current thread context
*
* The base CPU calls this method before starting the guest CPU
* when the contextDirty flag is set. The architecture dependent
* CPU implementation is expected to update all guest state
* (registers, special registers, and FPU state).
*/
virtual void updateKvmState() = 0;
/**
* Update the current thread context with the KVM state
*
* The base CPU after the guest updates any of the KVM state. In
* practice, this happens after kvmRun is called. The architecture
* dependent code is expected to read the state of the guest CPU
* and update gem5's thread state.
*/
virtual void updateThreadContext() = 0;
/**
* Update a thread context if the KVM state is dirty with respect
* to the cached thread context.
*/
void syncThreadContext();
/**
* Get a pointer to the event queue owning devices.
*
* Devices always live in a separate device event queue when
* running in multi-core mode. We need to temporarily migrate to
* this queue when accessing devices. By convention, devices and
* the VM use the same event queue.
*/
EventQueue *deviceEventQueue() { return vm.eventQueue(); }
/**
* Update the KVM if the thread context is dirty.
*/
void syncKvmState();
/** @} */
/** @{ */
/**
* Main kvmRun exit handler, calls the relevant handleKvmExit*
* depending on exit type.
*
* @return Number of ticks spent servicing the exit request
*/
virtual Tick handleKvmExit();
/**
* The guest performed a legacy IO request (out/inp on x86)
*
* @return Number of ticks spent servicing the IO request
*/
virtual Tick handleKvmExitIO();
/**
* The guest requested a monitor service using a hypercall
*
* @return Number of ticks spent servicing the hypercall
*/
virtual Tick handleKvmExitHypercall();
/**
* The guest exited because an interrupt window was requested
*
* The guest exited because an interrupt window was requested
* (request_interrupt_window in the kvm_run structure was set to 1
* before calling kvmRun) and it is now ready to receive
*
* @return Number of ticks spent servicing the IRQ
*/
virtual Tick handleKvmExitIRQWindowOpen();
/**
* An unknown architecture dependent error occurred when starting
* the vCPU
*
* The kvm_run data structure contains the hardware error
* code. The defaults behavior of this method just prints the HW
* error code and panics. Architecture dependent implementations
* may want to override this method to provide better,
* hardware-aware, error messages.
*
* @return Number of ticks delay the next CPU tick
*/
virtual Tick handleKvmExitUnknown();
/**
* An unhandled virtualization exception occured
*
* Some KVM virtualization drivers return unhandled exceptions to
* the user-space monitor. This interface is currently only used
* by the Intel VMX KVM driver.
*
* @return Number of ticks delay the next CPU tick
*/
virtual Tick handleKvmExitException();
/**
* KVM failed to start the virtualized CPU
*
* The kvm_run data structure contains the hardware-specific error
* code.
*
* @return Number of ticks delay the next CPU tick
*/
virtual Tick handleKvmExitFailEntry();
/** @} */
/**
* Is the architecture specific code in a state that prevents
* draining?
*
* This method should return false if there are any pending events
* in the guest vCPU that won't be carried over to the gem5 state
* and thus will prevent correct checkpointing or CPU handover. It
* might, for example, check for pending interrupts that have been
* passed to the vCPU but not acknowledged by the OS. Architecures
* implementing this method <i>must</i> override
* kvmRunDrain().
*
* @see BaseKvmCPU::kvmRunDrain()
*
* @return true if the vCPU is drained, false otherwise.
*/
virtual bool archIsDrained() const { return true; }
/**
* Inject a memory mapped IO request into gem5
*
* @param paddr Physical address
* @param data Pointer to the source/destination buffer
* @param size Memory access size
* @param write True if write, False if read
* @return Number of ticks spent servicing the memory access
*/
Tick doMMIOAccess(Addr paddr, void *data, int size, bool write);
/** @{ */
/**
* Set the signal mask used in kvmRun()
*
* This method allows the signal mask of the thread executing
* kvmRun() to be overridden inside the actual system call. This
* allows us to mask timer signals used to force KVM exits while
* in gem5.
*
* The signal mask can be disabled by setting it to NULL.
*
* @param mask Signals to mask
*/
void setSignalMask(const sigset_t *mask);
/** @} */
/**
* @addtogroup KvmIoctl
* @{
*/
/**
* vCPU ioctl interface.
*
* @param request KVM vCPU request
* @param p1 Optional request parameter
*
* @return -1 on error (error number in errno), ioctl dependent
* value otherwise.
*/
int ioctl(int request, long p1) const;
int ioctl(int request, void *p1) const {
return ioctl(request, (long)p1);
}
int ioctl(int request) const {
return ioctl(request, 0L);
}
/** @} */
/**
* KVM memory port. Uses default MasterPort behavior and provides an
* interface for KVM to transparently submit atomic or timing requests.
*/
class KVMCpuPort : public MasterPort
{
public:
KVMCpuPort(const std::string &_name, BaseKvmCPU *_cpu)
: MasterPort(_name, _cpu), cpu(_cpu), activeMMIOReqs(0)
{ }
/**
* Interface to send Atomic or Timing IO request. Assumes that the pkt
* and corresponding req have been dynamically allocated and deletes
* them both if the system is in atomic mode.
*/
Tick submitIO(PacketPtr pkt);
/** Returns next valid state after one or more IO accesses */
Status nextIOState() const;
protected:
/** KVM cpu pointer for finishMMIOPending() callback */
BaseKvmCPU *cpu;
/** Pending MMIO packets */
std::queue<PacketPtr> pendingMMIOPkts;
/** Number of MMIO requests in flight */
unsigned int activeMMIOReqs;
bool recvTimingResp(PacketPtr pkt) override;
void recvReqRetry() override;
};
/** Port for data requests */
KVMCpuPort dataPort;
/** Unused dummy port for the instruction interface */
KVMCpuPort instPort;
/**
* Be conservative and always synchronize the thread context on
* KVM entry/exit.
*/
const bool alwaysSyncTC;
/**
* Is the gem5 context dirty? Set to true to force an update of
* the KVM vCPU state upon the next call to kvmRun().
*/
bool threadContextDirty;
/**
* Is the KVM state dirty? Set to true to force an update of
* the KVM vCPU state upon the next call to kvmRun().
*/
bool kvmStateDirty;
/** KVM internal ID of the vCPU */
const long vcpuID;
/** ID of the vCPU thread */
pthread_t vcpuThread;
private:
/**
* Service MMIO requests in the mmioRing.
*
*
* @return Number of ticks spent servicing the MMIO requests in
* the MMIO ring buffer
*/
Tick flushCoalescedMMIO();
/**
* Setup a signal handler to catch the timer signal used to
* switch back to the monitor.
*/
void setupSignalHandler();
/**
* Discard a (potentially) pending signal.
*
* @param signum Signal to discard
* @return true if the signal was pending, false otherwise.
*/
bool discardPendingSignal(int signum) const;
/**
* Thread-specific initialization.
*
* Some KVM-related initialization requires us to know the TID of
* the thread that is going to execute our event queue. For
* example, when setting up timers, we need to know the TID of the
* thread executing in KVM in order to deliver the timer signal to
* that thread. This method is called as the first event in this
* SimObject's event queue.
*
* @see startup
*/
void startupThread();
/** Try to drain the CPU if a drain is pending */
bool tryDrain();
/** Execute the KVM_RUN ioctl */
void ioctlRun();
/** KVM vCPU file descriptor */
int vcpuFD;
/** Size of MMAPed kvm_run area */
int vcpuMMapSize;
/**
* Pointer to the kvm_run structure used to communicate parameters
* with KVM.
*
* @note This is the base pointer of the MMAPed KVM region. The
* first page contains the kvm_run structure. Subsequent pages may
* contain other data such as the MMIO ring buffer.
*/
struct kvm_run *_kvmRun;
/**
* Coalesced MMIO ring buffer. NULL if coalesced MMIO is not
* supported.
*/
struct kvm_coalesced_mmio_ring *mmioRing;
/** Cached page size of the host */
const long pageSize;
EventFunctionWrapper tickEvent;
/**
* Setup an instruction break if there is one pending.
*
* Check if there are pending instruction breaks in the CPU's
* instruction event queue and schedule an instruction break using
* PerfEvent.
*
* @note This method doesn't currently handle the main system
* instruction event queue.
*/
void setupInstStop();
/** @{ */
/** Setup hardware performance counters */
void setupCounters();
/**
* Setup the guest instruction counter.
*
* Setup the guest instruction counter and optionally request a
* signal every N instructions executed by the guest. This method
* will re-attach the counter if the counter has already been
* attached and its sampling settings have changed.
*
* @param period Signal period, set to 0 to disable signaling.
*/
void setupInstCounter(uint64_t period = 0);
/** Currently active instruction count breakpoint */
uint64_t activeInstPeriod;
/**
* Guest cycle counter.
*
* This is the group leader of all performance counters measuring
* the guest system. It can be used in conjunction with the
* PerfKvmTimer (see perfControlledByTimer) to trigger exits from
* KVM.
*/
PerfKvmCounter hwCycles;
/**
* Guest instruction counter.
*
* This counter is typically only used to measure the number of
* instructions executed by the guest. However, it can also be
* used to trigger exits from KVM if the configuration script
* requests an exit after a certain number of instructions.
*
* @see setupInstBreak
* @see scheduleInstStop
*/
PerfKvmCounter hwInstructions;
/**
* Does the runTimer control the performance counters?
*
* The run timer will automatically enable and disable performance
* counters if a PerfEvent-based timer is used to control KVM
* exits.
*/
bool perfControlledByTimer;
/** @} */
/**
* Timer used to force execution into the monitor after a
* specified number of simulation tick equivalents have executed
* in the guest. This counter generates the signal specified by
* KVM_TIMER_SIGNAL.
*/
std::unique_ptr<BaseKvmTimer> runTimer;
/** Host factor as specified in the configuration */
float hostFactor;
public:
/* @{ */
Stats::Scalar numInsts;
Stats::Scalar numVMExits;
Stats::Scalar numVMHalfEntries;
Stats::Scalar numExitSignal;
Stats::Scalar numMMIO;
Stats::Scalar numCoalescedMMIO;
Stats::Scalar numIO;
Stats::Scalar numHalt;
Stats::Scalar numInterrupts;
Stats::Scalar numHypercalls;
/* @} */
/** Number of instructions executed by the CPU */
Counter ctrInsts;
};
#endif