kvm: Add experimental support for a perf-based execution timer

Add support for using the CPU cycle counter instead of a normal POSIX
timer to generate timed exits to gem5. This should, in theory, provide
better resolution when requesting timer signals.

The perf-based timer requires a fairly recent kernel since it requires
a working PERF_EVENT_IOC_PERIOD ioctl. This ioctl has existed in the
kernel for a long time, but it used to be completely broken due to an
inverted match when the kernel copied things from user
space. Additionally, the ioctl does not change the sample period
correctly on all kernel versions which implement it. It is currently
only known to work reliably on kernel version 3.7 and above on ARM.

1 files changed, 9 insertions, 6 deletions

diff --git a/src/cpu/kvm/base.hh b/src/cpu/kvm/base.hh
index 90ebae644..0554f913e 100644
--- a/src/cpu/kvm/base.hh
+++ b/src/cpu/kvm/base.hh
@@ -460,12 +460,6 @@ class BaseKvmCPU : public BaseCPU
     /** Setup hardware performance counters */
     void setupCounters();
 
-    /** @{ */
-    /** Start/stop counting HW performance events */
-    void startCounters();
-    void stopCounters();
-    /** @} */
-
     /** KVM vCPU file descriptor */
     int vcpuFD;
     /** Size of MMAPed kvm_run area */
@@ -496,6 +490,15 @@ class BaseKvmCPU : public BaseCPU
     /** @} */
 
     /**
+     * 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