Re: [PATCH v4 -rcu 1/4] rcu/segcblist: Do not depend on rcl->len to store the segcb len during merge

[Joel Fernandes <joel@joelfernandes.org>]

On Wed, Aug 26, 2020 at 07:20:28AM -0700, Paul E. McKenney wrote:
[...]
> > > Or better yet, please see below, which should allow getting rid of both
> > > of them.
> > > 
> > > >  	rcu_segcblist_extract_done_cbs(src_rsclp, &donecbs);
> > > >  	rcu_segcblist_extract_pend_cbs(src_rsclp, &pendcbs);
> > > > -	rcu_segcblist_insert_count(dst_rsclp, &donecbs);
> > > > +
> > > > +	rcu_segcblist_add_len(dst_rsclp, src_len);
> > > >  	rcu_segcblist_insert_done_cbs(dst_rsclp, &donecbs);
> > > >  	rcu_segcblist_insert_pend_cbs(dst_rsclp, &pendcbs);
> > > 
> > > Rather than adding the blank lines, why not have the rcu_cblist structures
> > > carry the lengths?  You are already adjusting one of the two call sites
> > > that care (rcu_do_batch()), and the other is srcu_invoke_callbacks().
> > > That should shorten this function a bit more.  And make callback handling
> > > much more approachable, I suspect.
> > 
> > Sorry, I did not understand. The rcu_cblist structure already has a length
> > field. I do modify rcu_segcblist_extract_done_cbs() and
> > rcu_segcblist_extract_pend_cbs() to carry the length already, in a later
> > patch.
> > 
> > Just to emphasize, this patch is just a small refactor to avoid an issue in
> > later patches. It aims to keep current functionality unchanged.
> 
> True enough.  I am just suggesting that an equally small refactor in
> a slightly different direction should get to a better place.  The key
> point enabling this slightly different direction is that this code is
> an exception to the "preserve ->cblist.len" rule because it is invoked
> only from the CPU hotplug code.
> 
> So you could use the rcu_cblist .len field to update the ->cblist.len
> field, thus combining the _cbs and _count updates.  One thing that helps
> is that setting th e rcu_cblist .len field doesn't hurt the other use
> cases that require careful handling of ->cblist.len.

Thank you for the ideas. I am trying something like this on top of this
series based on the ideas. One thing I concerned a bit is if getting rid of
the rcu_segcblist_xchg_len() function (which has memory barriers in them)
causes issues in the hotplug path. I am now directly updating the length
without additional memory barriers. I will test it more and try to reason
more about it as well.

---8<-----------------------

From: Joel Fernandes <joelaf@google.com>
Date: Thu, 27 Aug 2020 18:30:25 -0400
Subject: [PATCH] fixup! rcu/segcblist: Do not depend on donecbs ->len to store
 the segcb len during merge

Signed-off-by: Joel Fernandes <joelaf@google.com>
---
 kernel/rcu/rcu_segcblist.c | 38 ++++----------------------------------
 1 file changed, 4 insertions(+), 34 deletions(-)

diff --git a/kernel/rcu/rcu_segcblist.c b/kernel/rcu/rcu_segcblist.c
index 79c2cbe388c5..c33abbc97a07 100644
--- a/kernel/rcu/rcu_segcblist.c
+++ b/kernel/rcu/rcu_segcblist.c
@@ -175,26 +175,6 @@ void rcu_segcblist_inc_len(struct rcu_segcblist *rsclp)
 	rcu_segcblist_add_len(rsclp, 1);
 }
 
-/*
- * Exchange the numeric length of the specified rcu_segcblist structure
- * with the specified value.  This can cause the ->len field to disagree
- * with the actual number of callbacks on the structure.  This exchange is
- * fully ordered with respect to the callers accesses both before and after.
- */
-static long rcu_segcblist_xchg_len(struct rcu_segcblist *rsclp, long v)
-{
-#ifdef CONFIG_RCU_NOCB_CPU
-	return atomic_long_xchg(&rsclp->len, v);
-#else
-	long ret = rsclp->len;
-
-	smp_mb(); /* Up to the caller! */
-	WRITE_ONCE(rsclp->len, v);
-	smp_mb(); /* Up to the caller! */
-	return ret;
-#endif
-}
-
 /*
  * Initialize an rcu_segcblist structure.
  */
@@ -361,6 +341,7 @@ void rcu_segcblist_extract_done_cbs(struct rcu_segcblist *rsclp,
 		if (rsclp->tails[i] == rsclp->tails[RCU_DONE_TAIL])
 			WRITE_ONCE(rsclp->tails[i], &rsclp->head);
 	rcu_segcblist_set_seglen(rsclp, RCU_DONE_TAIL, 0);
+	rcu_segcblist_add_len(rsclp, -(rclp->len));
 }
 
 /*
@@ -414,17 +395,7 @@ void rcu_segcblist_extract_pend_cbs(struct rcu_segcblist *rsclp,
 		WRITE_ONCE(rsclp->tails[i], rsclp->tails[RCU_DONE_TAIL]);
 		rcu_segcblist_set_seglen(rsclp, i, 0);
 	}
-}
-
-/*
- * Insert counts from the specified rcu_cblist structure in the
- * specified rcu_segcblist structure.
- */
-void rcu_segcblist_insert_count(struct rcu_segcblist *rsclp,
-				struct rcu_cblist *rclp)
-{
-	rcu_segcblist_add_len(rsclp, rclp->len);
-	rclp->len = 0;
+	rcu_segcblist_add_len(rsclp, -(rclp->len));
 }
 
 /*
@@ -448,6 +419,7 @@ void rcu_segcblist_insert_done_cbs(struct rcu_segcblist *rsclp,
 			break;
 	rclp->head = NULL;
 	rclp->tail = &rclp->head;
+	rcu_segcblist_add_len(rsclp, rclp->len);
 }
 
 /*
@@ -463,6 +435,7 @@ void rcu_segcblist_insert_pend_cbs(struct rcu_segcblist *rsclp,
 	rcu_segcblist_add_seglen(rsclp, RCU_NEXT_TAIL, rclp->len);
 	WRITE_ONCE(*rsclp->tails[RCU_NEXT_TAIL], rclp->head);
 	WRITE_ONCE(rsclp->tails[RCU_NEXT_TAIL], rclp->tail);
+	rcu_segcblist_add_len(rsclp, rclp->len);
 }
 
 /*
@@ -601,16 +574,13 @@ void rcu_segcblist_merge(struct rcu_segcblist *dst_rsclp,
 {
 	struct rcu_cblist donecbs;
 	struct rcu_cblist pendcbs;
-	long src_len;
 
 	rcu_cblist_init(&donecbs);
 	rcu_cblist_init(&pendcbs);
 
-	src_len = rcu_segcblist_xchg_len(src_rsclp, 0);
 	rcu_segcblist_extract_done_cbs(src_rsclp, &donecbs);
 	rcu_segcblist_extract_pend_cbs(src_rsclp, &pendcbs);
 
-	rcu_segcblist_add_len(dst_rsclp, src_len);
 	rcu_segcblist_insert_done_cbs(dst_rsclp, &donecbs);
 	rcu_segcblist_insert_pend_cbs(dst_rsclp, &pendcbs);
 
-- 
2.28.0.402.g5ffc5be6b7-goog