base: wait a bit at the end of the smp test

Fix #3306
This commit is contained in:
Stefan Kalkowski 2019-04-25 15:00:45 +02:00 committed by Christian Helmuth
parent 918b9a9fa4
commit b85071174a

View File

@ -304,7 +304,17 @@ namespace Tlb_shootdown_test {
log("TLB: all threads are up and running...");
destroy(heap, ram_ds);
log("TLB: ram dataspace destroyed, all will fault...");
log("TLB: ram dataspace destroyed, all have to fail...");
/**
* The more cores are existing the more threads have to fail.
* The bottleneck is core's page-fault messages all printed
* over a lazy serial line from core 0.
* We have to wait here, for some time so that all fault
* messages are received before the test finishes.
*/
for (volatile unsigned i = 0; i < (0x2000000 * cpus.total()); i++) ;
for (unsigned i = 1; i < cpus.total(); i++) destroy(heap, threads[i]);
destroy(heap, threads);