memory: bail from page scrubbing when CPU is no longer online

Scrubbing can significantly delay the offlining (parking) of a CPU (e.g.
because of booting into in smt=0 mode), to a degree that the "CPU <n>
still not dead..." messages logged on x86 in 1s intervals can be seen
multiple times. There are no softirqs involved in this process, so
extend the existing preemption check in the scrubbing logic to also exit
when the CPU is no longer observed online.

Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: 3c9fd69416f8ffc611705fb24dfb383203ddc84f
master date: 2021-01-29 11:34:37 +0100

diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
index d9578abb7c6230fafe76c3406290c2a94ff42f3e..8859fa4fd1ce2585fea187b61a9563ba36724734 100644 (file)
--- a/xen/common/page_alloc.c
+++ b/xen/common/page_alloc.c
@@ -1328,9 +1328,11 @@ bool scrub_free_pages(void)
                      * Scrub a few (8) pages before becoming eligible for
                      * preemption. But also count non-scrubbing loop iterations
                      * so that we don't get stuck here with an almost clean
-                     * heap.
+                     * heap. Consider the CPU no longer being seen as online as
+                     * a request to preempt immediately, to not unduly delay
+                     * its offlining.
                      */
-                    if ( cnt > 800 && softirq_pending(cpu) )
+                    if ( !cpu_online(cpu) || (cnt > 800 && softirq_pending(cpu)) )
                     {
                         preempt = true;
                         break;