Commit 51c8574f authored by Changpeng Liu's avatar Changpeng Liu Committed by Jim Harris
Browse files

nvmf/vfio-user: introduce request state 



Unlike tcp/rdma transport, the vfio-user transport doesn't need to
wait for the data buffers, so here we add two request states for
now.

The request state will help us for coming request abort API.

Change-Id: Ibbb193fbbd358333f81aa29341493c19ab7bd108
Signed-off-by: default avatarChangpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7435


Community-CI: Broadcom CI
Tested-by: default avatarSPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: default avatarJim Harris <james.r.harris@intel.com>
Reviewed-by: default avatarBen Walker <benjamin.walker@intel.com>
parent 7abbad3c

lib/nvmf/vfio_user.c

+13 −5
Original line number Diff line number Diff line
@@ -76,11 +76,17 @@ typedef int (*nvmf_vfio_user_req_cb_fn)(struct nvmf_vfio_user_req *req, void *cb 
/* 1 more for PRP2 list itself */

#define NVMF_VFIO_USER_MAX_IOVECS	(NVMF_REQ_MAX_BUFFERS + 1)



enum nvmf_vfio_user_req_state {

	VFIO_USER_REQUEST_STATE_FREE = 0,

	VFIO_USER_REQUEST_STATE_EXECUTING,

};



struct nvmf_vfio_user_req  {

	struct spdk_nvmf_request		req;

	struct spdk_nvme_cpl			rsp;

	struct spdk_nvme_cmd			cmd;



	enum nvmf_vfio_user_req_state		state;

	nvmf_vfio_user_req_cb_fn		cb_fn;

	void					*cb_arg;
@@ -2068,6 +2074,7 @@ _nvmf_vfio_user_req_free(struct nvmf_vfio_user_qpair *vu_qpair, struct nvmf_vfio 
	memset(&vu_req->cmd, 0, sizeof(vu_req->cmd));

	memset(&vu_req->rsp, 0, sizeof(vu_req->rsp));

	vu_req->iovcnt = 0;

	vu_req->state = VFIO_USER_REQUEST_STATE_FREE;



	TAILQ_INSERT_TAIL(&vu_qpair->reqs, vu_req, link);

}
@@ -2265,7 +2272,7 @@ handle_cmd_req(struct nvmf_vfio_user_ctrlr *ctrlr, struct spdk_nvme_cmd *cmd, 
	       struct spdk_nvmf_request *req)

{

	int err;

	struct nvmf_vfio_user_req *vfio_user_req;

	struct nvmf_vfio_user_req *vu_req;



	assert(ctrlr != NULL);

	assert(cmd != NULL);
@@ -2280,9 +2287,9 @@ handle_cmd_req(struct nvmf_vfio_user_ctrlr *ctrlr, struct spdk_nvme_cmd *cmd, 
		return -1;

	}



	vfio_user_req = SPDK_CONTAINEROF(req, struct nvmf_vfio_user_req, req);

	vfio_user_req->cb_fn = handle_cmd_rsp;

	vfio_user_req->cb_arg = SPDK_CONTAINEROF(req->qpair, struct nvmf_vfio_user_qpair, qpair);

	vu_req = SPDK_CONTAINEROF(req, struct nvmf_vfio_user_req, req);

	vu_req->cb_fn = handle_cmd_rsp;

	vu_req->cb_arg = SPDK_CONTAINEROF(req->qpair, struct nvmf_vfio_user_qpair, qpair);

	req->cmd->nvme_cmd = *cmd;

	if (nvmf_qpair_is_admin_queue(req->qpair)) {

		err = map_admin_cmd_req(ctrlr, req);
@@ -2295,9 +2302,10 @@ handle_cmd_req(struct nvmf_vfio_user_ctrlr *ctrlr, struct spdk_nvme_cmd *cmd, 
			    ctrlr_id(ctrlr), cmd->opc);

		req->rsp->nvme_cpl.status.sc = SPDK_NVME_SC_INTERNAL_DEVICE_ERROR;

		req->rsp->nvme_cpl.status.sct = SPDK_NVME_SCT_GENERIC;

		return handle_cmd_rsp(vfio_user_req, vfio_user_req->cb_arg);

		return handle_cmd_rsp(vu_req, vu_req->cb_arg);

	}



	vu_req->state = VFIO_USER_REQUEST_STATE_EXECUTING;

	spdk_nvmf_request_exec(req);



	return 0;