credit2: respect credit2_runqueue=all when arranging runqueues

Documentation for credit2_runqueue=all says it should create one queue
for all pCPUs on the host. But since introduction
sched_credit2_max_cpus_runqueue, it actually created separate runqueue
per socket, even if the CPUs count is below
sched_credit2_max_cpus_runqueue.

Adjust the condition to skip syblink check in case of
credit2_runqueue=all.

Fixes: 8e2aa76dc167 ("xen: credit2: limit the max number of CPUs in a runqueue")
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Reviewed-by: Juergen Gross <jgross@suse.com>

diff --git a/docs/misc/xen-command-line.pandoc b/docs/misc/xen-command-line.pandoc
index 63dfda4a7a9e5412c13f5abe4e9a1e504a735850..f68deaa6a97a8166722ac0ee24ed9f509fe99c1d 100644 (file)
--- a/docs/misc/xen-command-line.pandoc
+++ b/docs/misc/xen-command-line.pandoc
@@ -726,6 +726,11 @@ Available alternatives, with their meaning, are:
 * `all`: just one runqueue shared by all the logical pCPUs of
          the host
 
+Regardless of the above choice, Xen attempts to respect
+`sched_credit2_max_cpus_runqueue` limit, which may mean more than one runqueue
+for the `all` value. If that isn't intended, raise
+the `sched_credit2_max_cpus_runqueue` value.
+
 ### dbgp
 > `= ehci[ <integer> | @pci<bus>:<slot>.<func> ]`
 > `= xhci[ <integer> | @pci<bus>:<slot>.<func> ][,share=<bool>|hwdom]`

diff --git a/xen/common/sched/credit2.c b/xen/common/sched/credit2.c
index 0e3f89e5378e6b98923ecd750d7ea4f3c328c701..ae55feea345993a4aeb4f65bc612090386d7418b 100644 (file)
--- a/xen/common/sched/credit2.c
+++ b/xen/common/sched/credit2.c
@@ -996,9 +996,14 @@ cpu_add_to_runqueue(const struct scheduler *ops, unsigned int cpu)
              *
              * Otherwise, let's try to make sure that siblings stay in the
              * same runqueue, pretty much under any cinrcumnstances.
+             *
+             * Furthermore, try to respect credit2_runqueue=all, as long as
+             * max_cpus_runq isn't violated.
              */
-            if ( rqd->refcnt < max_cpus_runq && (ops->cpupool->gran != SCHED_GRAN_cpu ||
-                  cpu_runqueue_siblings_match(rqd, cpu, max_cpus_runq)) )
+            if ( rqd->refcnt < max_cpus_runq &&
+                 (ops->cpupool->gran != SCHED_GRAN_cpu ||
+                  cpu_runqueue_siblings_match(rqd, cpu, max_cpus_runq) ||
+                  opt_runqueue == OPT_RUNQUEUE_ALL) )
             {
                 /*
                  * This runqueue is ok, but as we said, we also want an even