x86: vmm: Finalize to owning_proc, not cur_proc.

Similar to a previous bug, x86_finalize_vmtf() assumed the TF belonged to
cur_proc, but it actually belongs to owning_proc.  If we finalize a TF on a
core that runs proc's/kthreads concurrently, then we could have a situation
where cur_proc != owning_proc.  Then we'd try finding a GPC for the other
process, instead of the VMM.  Yikes!

This was relatively easy to make happen regularly: run vmrunkernel as an
SCP under strace from ssh.  I think I triggered it with perf at some point
too.

Here's the main debugging info that pointed me the right way:

	couldn't find a gpc, p 284, guest_pcoreid 0

	kernel panic at kern/arch/x86/vmm/vmm.c:206, from core 0: assertion failed:
	gpc
	Entering Nanwan's Dungeon on Core 0 (Ints off):
	Type 'help' for a list of commands.
	ROS(Core 0)> ps
	     PID Name                 State      Parent
	-------------------------------------------------
	      15 /bin/cs              WAITING         0
	      12 /bin/ipconfig        WAITING         0
	       1 bash                 WAITING         0
	     269 /bin/dropbear        WAITING         0
	     284 strace               RUNNABLE_S    275
	     274 /bin/dropbear        WAITING       269
	     270 /bin/bash            WAITING         1
	     285 vmrunkernel          RUNNING_S     284
	     275 -sh                  WAITING       274
	ROS(Core 0)> bt
	Stack Backtrace on Core 0:
	#01 [<0xffffffffc201ed74>] in mon_backtrace
	#02 [<0xffffffffc201fd77>] in monitor
	#03 [<0xffffffffc200ca1a>] in _panic
	#04 [<0xffffffffc2134e9c>] in unload_guest_pcore
	#05 [<0xffffffffc21320d8>] in arch_finalize_ctx
	#06 [<0xffffffffc205d1bb>] in copy_current_ctx_to
	#07 [<0xffffffffc204d70c>] in __notify
	#08 [<0xffffffffc205d71f>] in process_routine_kmsg
	#09 [<0xffffffffc2051665>] in proc_restartcore

Note that the lookup was using PID 284 (strace), but the VM was 285.

Signed-off-by: Barret Rhoden <brho@cs.berkeley.edu>

kern/arch/x86/trap.c

diff --git a/kern/arch/x86/trap.c b/kern/arch/x86/trap.c
index 412df60..d21f0b2 100644
--- a/kern/arch/x86/trap.c
+++ b/kern/arch/x86/trap.c
@@ -1147,7 +1147,7 @@
 	struct per_cpu_info *pcpui = &per_cpu_info[core_id()];
 
 	x86_vmtf_clear_partial(tf);
-	unload_guest_pcore(pcpui->cur_proc, pcpui->guest_pcoreid);
+	unload_guest_pcore(pcpui->owning_proc, pcpui->guest_pcoreid);
 }
 
 /* Makes sure that the user context is fully saved into ctx and not split across

kern/src/syscall.c

diff --git a/kern/src/syscall.c b/kern/src/syscall.c
index beb6e1d..5802642 100644
--- a/kern/src/syscall.c
+++ b/kern/src/syscall.c
@@ -717,6 +717,7 @@
 {
 	uintptr_t temp;
 	int ret;
+	struct per_cpu_info *pcpui = &per_cpu_info[core_id()];
 
 	// TODO: right now we only support fork for single-core processes
 	if (e->state != PROC_RUNNING_S) {
@@ -736,6 +737,7 @@
 		set_errno(EINVAL);
 		return -1;
 	}
+	assert(pcpui->cur_proc == pcpui->owning_proc);
 	copy_current_ctx_to(&env->scp_ctx);
 
 	/* Make the new process have the same VMRs as the older.  This will copy the
@@ -870,6 +872,7 @@
 	}
 	/* Preemptively copy out the cur_ctx, in case we fail later (easier on
 	 * cur_ctx if we do this now) */
+	assert(pcpui->cur_proc == pcpui->owning_proc);
 	copy_current_ctx_to(&p->scp_ctx);
 	/* Check the size of the argenv array, error out if too large. */
 	if ((argenv_l < sizeof(struct argenv)) || (argenv_l > ARG_MAX)) {

kern/src/trap.c

diff --git a/kern/src/trap.c b/kern/src/trap.c
index a9983e4..5e3824d 100644
--- a/kern/src/trap.c
+++ b/kern/src/trap.c
@@ -59,6 +59,7 @@
 	uint32_t vcoreid = pcpui->owning_vcoreid;
 	struct preempt_data *vcpd = &p->procdata->vcore_preempt_data[vcoreid];
 
+	assert(pcpui->cur_proc == pcpui->owning_proc);
 	if (!proc_is_vcctx_ready(p))
 		return -1;
 	if (vcpd->notif_disabled)