sunrpc/cache: make sure deferred requests eventually get revisited.

While deferred requests normally get revisited quite quickly,
it is possible for a request to remain in the deferral queue
when the cache item is discarded.  We can easily make sure that
doesn't happen by calling cache_revisit_request just before
the final 'put'.

Also there is a small chance that a race would cause one thread to
defer a request against a cache item while another thread is failing
to queue an upcall for that item.  So when the upcall fails, make
sure to revisit all deferred requests.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

net/sunrpc/cache.c

@@ -221,6 +221,7 @@ int cache_check(struct cache_detail *detail,
 			switch (cache_make_upcall(detail, h)) {
 			case -EINVAL:
 				clear_bit(CACHE_PENDING, &h->flags);
+				cache_revisit_request(h);
 				if (rv == -EAGAIN) {
 					set_bit(CACHE_NEGATIVE, &h->flags);
 					cache_fresh_unlocked(h, detail,
@@ -473,8 +474,10 @@ static int cache_clean(void)
 		if (!ch)
 			current_index ++;
 		spin_unlock(&cache_list_lock);
-		if (ch)
+		if (ch) {
+			cache_revisit_request(ch);
 			cache_put(ch, d);
+		}
 	} else
 		spin_unlock(&cache_list_lock);