Don't zero td_runtime when billing thread CPU usage to the process;

maintain a separate td_incruntime to hold unbilled CPU usage for
the thread that has the previous properties of td_runtime.

When thread information is requested using the thread monitoring
sysctls, export thread td_runtime instead of process rusage runtime
in kinfo_proc.

This restores the display of individual ithread and other kernel
thread CPU usage since inception in ps -H and top -SH, as well for
libthr user threads, valuable debugging information lost with the
move to try kthreads since they are no longer independent processes.

There is universal agreement that we should rewrite the process and
thread export sysctls, but this commit gets things going a bit
better in the mean time.  Likewise, there are resevations about the
continued validity of statclock given the speed of modern processors.

Reviewed by:		attilio, emaste, jhb, julian

1 files changed, 13 insertions, 7 deletions

diff --git a/sys/kern/kern_proc.c b/sys/kern/kern_proc.c
index be70eedbb039..c398a8d97670 100644
--- a/sys/kern/kern_proc.c
+++ b/sys/kern/kern_proc.c
@@ -84,7 +84,8 @@ MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
 static void doenterpgrp(struct proc *, struct pgrp *);
 static void orphanpg(struct pgrp *pg);
 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
-static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp);
+static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
+    int preferthread);
 static void pgadjustjobc(struct pgrp *pgrp, int entering);
 static void pgdelete(struct pgrp *);
 static int proc_ctor(void *mem, int size, void *arg, int flags);
@@ -765,11 +766,12 @@ fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
 }
 
 /*
- * Fill in information that is thread specific.
- * Must be called with p_slock locked.
+ * Fill in information that is thread specific.  Must be called with p_slock
+ * locked.  If 'preferthread' is set, overwrite certain process-related
+ * fields that are maintained for both threads and processes.
  */
 static void
-fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp)
+fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
 {
 	struct proc *p;
 
@@ -829,6 +831,9 @@ fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp)
 	kp->ki_pri.pri_class = td->td_pri_class;
 	kp->ki_pri.pri_user = td->td_user_pri;
 
+	if (preferthread)
+		kp->ki_runtime = cputick2usec(td->td_runtime);
+
 	/* We can't get this anymore but ps etc never used it anyway. */
 	kp->ki_rqindex = 0;
 
@@ -848,7 +853,7 @@ fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
 	fill_kinfo_proc_only(p, kp);
 	PROC_SLOCK(p);
 	if (FIRST_THREAD_IN_PROC(p) != NULL)
-		fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp);
+		fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
 	PROC_SUNLOCK(p);
 }
 
@@ -918,7 +923,8 @@ sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
 	if (flags & KERN_PROC_NOTHREADS) {
 		PROC_SLOCK(p);
 		if (FIRST_THREAD_IN_PROC(p) != NULL)
-			fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), &kinfo_proc);
+			fill_kinfo_thread(FIRST_THREAD_IN_PROC(p),
+			    &kinfo_proc, 0);
 		PROC_SUNLOCK(p);
 		error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
 				   sizeof(kinfo_proc));
@@ -926,7 +932,7 @@ sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
 		PROC_SLOCK(p);
 		if (FIRST_THREAD_IN_PROC(p) != NULL)
 			FOREACH_THREAD_IN_PROC(p, td) {
-				fill_kinfo_thread(td, &kinfo_proc);
+				fill_kinfo_thread(td, &kinfo_proc, 1);
 				error = SYSCTL_OUT(req, (caddr_t)&kinfo_proc,
 						   sizeof(kinfo_proc));
 				if (error)