summaryrefslogtreecommitdiff
path: root/test
diff options
context:
space:
mode:
authorRichard Braun <rbraun@sceen.net>2018-01-08 00:01:24 +0100
committerRichard Braun <rbraun@sceen.net>2018-01-08 00:01:24 +0100
commitb56d4a2bf6de3aa96f0b37a4eb04d41d8b1ca2c4 (patch)
tree17de1a79e260c3f21f182ed0e318f369a1aba742 /test
parentd63c9bbc67a60567b384e066583d5e5428a6ea1f (diff)
test/test_xcall: update test description
Diffstat (limited to 'test')
-rw-r--r--test/test_xcall.c8
1 files changed, 6 insertions, 2 deletions
diff --git a/test/test_xcall.c b/test/test_xcall.c
index 49706cde..cd335947 100644
--- a/test/test_xcall.c
+++ b/test/test_xcall.c
@@ -92,8 +92,12 @@ test_run(void *arg)
for (unsigned int i = 0; i < cpu_count(); i++) {
/*
- * FIXME There seems to be an initialization race of the local APIC when
- * sending IPIs early from CPU 1 to CPU 2 or more.
+ * Send IPIs from CPU 1 first, in order to better trigger any
+ * initialization race that may prevent correct IPI transmission.
+ * This assumes CPUs are initialized sequentially, and that CPU 1
+ * may have finished initialization much earlier than the last CPU.
+ * CPU 0 isn't used since it's the one normally initializing remote
+ * CPUs.
*/
cpu = (1 + i) % cpu_count();