if someone has time, could take a look at code.illumos.org/c/illumos-gate/+/3161 , I have done a couple of tests but I don't have systems with a lot of memory to see the difference.

→ CODE REVIEW 3161: 15859 vmm reservoir could defer zeroing to taskq (NEW) | illumos.org/issues/15859

neirac: I believe it's going to be a bit more complicated than that

the memory cannot go back into the free pool while it's being zeroed

and there are some remaining questions about visibility into the zeroing if a simultaneous reservoir allocation request comes in which has to wait for the to-be-zeroed memory in order to be successfully fulfilled

pmooney thank you for taking a look, oh yes this code just sends the zeroing to a taskq. I'll close this , as this patch seems to be doing harm, I just tested in a vm using and releasign memory from the reservoir, provisioning several vms from the reservoir should trigger a bug as the pages are being zero while at the same time being transferred to the free pool.

is there a reason pciadm is not in the PATH for root ?

it's such a useful utility

because it's not a Committed interface yet