File 60be0e24-credit2-pick-runnable-unit.patch of Package xen.32200

From eb8e66b6041068ffaca12f18700071dc3b6a8fe9 Mon Sep 17 00:00:00 2001
From: Dario Faggioli <dfaggioli@suse.com>
Date: Fri, 28 May 2021 17:12:48 +0200
Subject: [PATCH 1/3] credit2: make sure we pick a runnable unit from the runq
 if there is one
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

A !runnable unit (temporarily) present in the runq may cause us to
stop scanning the runq itself too early. Of course, we don't run any
non-runnable vCPUs, but we end the scan and we fallback to picking
the idle unit. In other word, this prevent us to find there and pick
the actual unit that we're meant to start running (which might be
further ahead in the runq).

Depending on the vCPU pinning configuration, this may lead to such
unit to be stuck in the runq for long time, causing malfunctioning
inside the guest.

Fix this by checking runnable/non-runnable status up-front, in the runq
scanning function.

Reported-by: Michał Leszczyński <michal.leszczynski@cert.pl>
Reported-by: Dion Kant <g.w.kant@hunenet.nl>
Signed-off-by: Dario Faggioli <dfaggioli@suse.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
---
 xen/common/sched_credit2.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/xen/common/sched_credit2.c b/xen/common/sched_credit2.c
index d6ebd126de..d89e340905 100644
--- a/xen/common/sched_credit2.c
+++ b/xen/common/sched_credit2.c
@@ -3361,6 +3361,10 @@ runq_candidate(struct csched2_runqueue_data *rqd,
                         (unsigned char *)&d);
         }
 
+        /* Skip non runnable vcpus that we (temporarily) have in the runq */
+        if ( unlikely(!vcpu_runnable(svc->vcpu)) )
+            continue;
+
         /* Only consider vcpus that are allowed to run on this processor. */
         if ( !cpumask_test_cpu(cpu, svc->vcpu->cpu_hard_affinity) )
             continue;
-- 
2.32.0