nvme/nda: Fail all nvme I/Os after controller fails

Once the controller has failed, fail all I/O w/o sending it to the
device. The reset of the nvme driver won't schedule any I/O to the
failed device, and the controller is in an indeterminate state and can't
accept I/O. Fail both at the top end of the sim and the bottom
end. Don't bother queueing up the I/O for failure in a different task.

Reviewed by:		chuck
Sponsored by:		Netflix
Differential Revision:	https://reviews.freebsd.org/D31341
This commit is contained in:
Warner Losh 2021-09-17 14:56:58 -06:00
parent e05d9788b7
commit 4b977e6dda
2 changed files with 7 additions and 5 deletions

View File

@ -1078,12 +1078,10 @@ _nvme_qpair_submit_request(struct nvme_qpair *qpair, struct nvme_request *req)
if (qpair->ctrlr->is_failed) {
/*
* The controller has failed. Post the request to a
* task where it will be aborted, so that we do not
* invoke the request's callback in the context
* of the submission.
* The controller has failed, so fail the request.
*/
nvme_ctrlr_post_failed_request(qpair->ctrlr, req);
nvme_qpair_manual_complete_request(qpair, req,
NVME_SCT_GENERIC, NVME_SC_ABORTED_BY_REQUEST);
} else {
/*
* Put the request on the qpair's request queue to be

View File

@ -258,6 +258,10 @@ nvme_sim_action(struct cam_sim *sim, union ccb *ccb)
break;
case XPT_NVME_IO: /* Execute the requested I/O operation */
case XPT_NVME_ADMIN: /* or Admin operation */
if (ctrlr->is_failed) {
ccb->ccb_h.status = CAM_DEV_NOT_THERE;
break;
}
nvme_sim_nvmeio(sim, ccb);
return; /* no done */
default: