svn commit: r298926 - head/sys/dev/xen/timer

[Roger Pau Monné]

Author: royger
Date: Mon May  2 16:13:11 2016
New Revision: 298926
URL: https://svnweb.freebsd.org/changeset/base/298926

Log:
  xen/x86: don't lose event interrupts
  
  On slow platforms with unreliable TSC, such as QEMU emulated machines,
  it is possible for the FreeBSD kernel to request the next event in the
  past. In that case, in the current implementation of
  xentimer_vcpu_start_timer, we simply return -ETIME. To be precise Xen
  returns -ETIME and we pass it on. As a consequence we need to loop
  around to function to make sure that the timer is properly set.
  
  Instead it is better to always ask the hypervisor for a timer event,
  even if the timeout is past. To do that, remove the VCPU_SSHOTTMR_future
  flag.
  
  Submitted by:	Stefano Stabellini <sstabellini at kernel.org>
  Reviewed by:	royger
  MFC after:	2 weeks

Modified:
  head/sys/dev/xen/timer/timer.c

Modified: head/sys/dev/xen/timer/timer.c
==============================================================================
--- head/sys/dev/xen/timer/timer.c	Mon May  2 15:14:16 2016	(r298925)
+++ head/sys/dev/xen/timer/timer.c	Mon May  2 16:13:11 2016	(r298926)
@@ -267,7 +267,8 @@ xentimer_vcpu_start_timer(int vcpu, uint
 	struct vcpu_set_singleshot_timer single;
 
 	single.timeout_abs_ns = next_time;
-	single.flags          = VCPU_SSHOTTMR_future;
+	/* Get an event anyway, even if the timeout is already expired */
+	single.flags          = 0;
 	return (HYPERVISOR_vcpu_op(VCPUOP_set_singleshot_timer, vcpu, &single));
 }