Commit 8531a41f authored by Jacek Kalwas's avatar Jacek Kalwas Committed by Tomasz Zawadzki
Browse files

nvme/tcp: add util to cond schedule qpair poll 



This is a prework for further changes.

There is behavior change for accel operations callbacks i.e. extra qpair
states are checked but there's no harm in setting needs_poll more than
needed moreover those states are unlikely (updated condition to reflect
that).

Change-Id: Ibaa23819b4f69f40e57fe818f848bef7a1428aaf
Signed-off-by: default avatarJacek Kalwas <jacek.kalwas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/23359


Reviewed-by: default avatarKonrad Sztyber <konrad.sztyber@intel.com>
Tested-by: default avatarSPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: default avatarJim Harris <jim.harris@samsung.com>
parent b10f50b0

lib/nvme/nvme_tcp.c

+33 −36
Original line number Diff line number Diff line
@@ -425,32 +425,43 @@ nvme_tcp_ctrlr_destruct(struct spdk_nvme_ctrlr *ctrlr) 
	return 0;

}



static void

pdu_write_done(void *cb_arg, int err)

{

	struct nvme_tcp_pdu *pdu = cb_arg;

	struct nvme_tcp_qpair *tqpair = pdu->qpair;

	struct nvme_tcp_poll_group *pgroup;



/* If there are queued requests, we assume they are queued because they are waiting

 * for resources to be released. Those resources are almost certainly released in

	 * response to a PDU completing here. However, to attempt to make forward progress

 * response to a PDU completing. However, to attempt to make forward progress

 * the qpair needs to be polled and we can't rely on another network event to make

	 * that happen. Add it to a list of qpairs to poll regardless of network activity

	 * here.

 * that happen. Add it to a list of qpairs to poll regardless of network activity.

 *

 * Besides, when tqpair state is NVME_TCP_QPAIR_STATE_FABRIC_CONNECT_POLL or

 * NVME_TCP_QPAIR_STATE_INITIALIZING, need to add it to needs_poll list too to make

 * forward progress in case that the resources are released after icreq's or CONNECT's

 * resp is processed. */

	if (tqpair->qpair.poll_group && !tqpair->needs_poll && (!STAILQ_EMPTY(&tqpair->qpair.queued_req) ||

			tqpair->state == NVME_TCP_QPAIR_STATE_FABRIC_CONNECT_POLL ||

			tqpair->state == NVME_TCP_QPAIR_STATE_INITIALIZING)) {

		pgroup = nvme_tcp_poll_group(tqpair->qpair.poll_group);

static void

nvme_tcp_cond_schedule_qpair_polling(struct nvme_tcp_qpair *tqpair)

{

	struct nvme_tcp_poll_group *pgroup;



	if (tqpair->needs_poll || !tqpair->qpair.poll_group) {

		return;

	}



	if (STAILQ_EMPTY(&tqpair->qpair.queued_req) &&

	    spdk_likely(tqpair->state != NVME_TCP_QPAIR_STATE_FABRIC_CONNECT_POLL &&

			tqpair->state != NVME_TCP_QPAIR_STATE_INITIALIZING)) {

		return;

	}



	pgroup = nvme_tcp_poll_group(tqpair->qpair.poll_group);

	TAILQ_INSERT_TAIL(&pgroup->needs_poll, tqpair, link);

	tqpair->needs_poll = true;

}



static void

pdu_write_done(void *cb_arg, int err)

{

	struct nvme_tcp_pdu *pdu = cb_arg;

	struct nvme_tcp_qpair *tqpair = pdu->qpair;



	nvme_tcp_cond_schedule_qpair_polling(tqpair);

	TAILQ_REMOVE(&tqpair->send_queue, pdu, tailq);



	if (err != 0) {
@@ -1263,18 +1274,11 @@ nvme_tcp_recv_payload_seq_cb(void *cb_arg, int status) 
	struct nvme_tcp_req *treq = cb_arg;

	struct nvme_request *req = treq->req;

	struct nvme_tcp_qpair *tqpair = treq->tqpair;

	struct nvme_tcp_poll_group *group;



	assert(treq->ordering.bits.in_progress_accel);

	treq->ordering.bits.in_progress_accel = 0;



	/* We need to force poll the qpair to make sure any queued requests will be resubmitted, see

	 * comment in pdu_write_done(). */

	if (tqpair->qpair.poll_group && !tqpair->needs_poll && !STAILQ_EMPTY(&tqpair->qpair.queued_req)) {

		group = nvme_tcp_poll_group(tqpair->qpair.poll_group);

		TAILQ_INSERT_TAIL(&group->needs_poll, tqpair, link);

		tqpair->needs_poll = true;

	}

	nvme_tcp_cond_schedule_qpair_polling(tqpair);



	req->accel_sequence = NULL;

	if (spdk_unlikely(status != 0)) {
@@ -1392,7 +1396,6 @@ nvme_tcp_accel_recv_compute_crc32_done(void *cb_arg, int status) 
	struct nvme_tcp_pdu *pdu;

	struct nvme_tcp_qpair *tqpair;

	int rc;

	struct nvme_tcp_poll_group *pgroup;

	int dummy_reaped = 0;



	pdu = tcp_req->pdu;
@@ -1404,13 +1407,7 @@ nvme_tcp_accel_recv_compute_crc32_done(void *cb_arg, int status) 
	assert(tcp_req->ordering.bits.in_progress_accel);

	tcp_req->ordering.bits.in_progress_accel = 0;



	/* We need to force poll the qpair to make sure any queued requests will be resubmitted, see

	 * comment in pdu_write_done(). */

	if (tqpair->qpair.poll_group && !tqpair->needs_poll && !STAILQ_EMPTY(&tqpair->qpair.queued_req)) {

		pgroup = nvme_tcp_poll_group(tqpair->qpair.poll_group);

		TAILQ_INSERT_TAIL(&pgroup->needs_poll, tqpair, link);

		tqpair->needs_poll = true;

	}

	nvme_tcp_cond_schedule_qpair_polling(tqpair);



	if (spdk_unlikely(status)) {

		SPDK_ERRLOG("Failed to compute the data digest for pdu =%p\n", pdu);