Context Navigation

Changes between Initial Version and Version 1 of Ticket #12117, comment 1

-              initial
+              v1
 This information may weaken the multi-processor synchronization conjecture. The effect is definitely worse with multi-cpu's in the guest, but transfers still hang with single-cpu guest configuration:
+More information relevant to the multi-processor synchronization conjecture. The effect is definitely worse with multi-cpu's in the guest, but transfers still hang with single-cpu guest configuration:
 -CPU guest VM: ~ 6 minutes to fail
 …
 chuck@rosman:~/ohci-dbg$
 }}}
+Can we expect the DevOHCI code to be executing in the host environment, and the ohci-hcd code to be executing in the guest environment? IIRC, the VM gets a dedicated CPU in a multiprocessor host environment, thus they will always be executing on different CPUs. Having 4 instead of 1 guest CPUs apparently increases the stress level.