Commit 72e4a4d4 authored by Shuhei Matsumoto's avatar Shuhei Matsumoto Committed by Tomasz Zawadzki
Browse files

bdev/nvme: Each nvme_bdev_channel caches its current io_path 



Signed-off-by: default avatarShuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I3ec3a588ff741cf04383e89f5a701e33bf1987a6
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9894


Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: default avatarSPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: default avatarBen Walker <benjamin.walker@intel.com>
Reviewed-by: default avatarKonrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: default avatarAleksey Marchuk <alexeymar@mellanox.com>
parent 32697257

module/bdev/nvme/bdev_nvme.c

+52 −6
Original line number Diff line number Diff line
@@ -581,6 +581,8 @@ _bdev_nvme_add_io_path(struct nvme_bdev_channel *nbdev_ch, struct nvme_ns *nvme_ 
	io_path->nbdev_ch = nbdev_ch;

	STAILQ_INSERT_TAIL(&nbdev_ch->io_path_list, io_path, stailq);



	nbdev_ch->current_io_path = NULL;



	return 0;

}
@@ -589,6 +591,8 @@ _bdev_nvme_delete_io_path(struct nvme_bdev_channel *nbdev_ch, struct nvme_io_pat 
{

	struct spdk_io_channel *ch;



	nbdev_ch->current_io_path = NULL;



	STAILQ_REMOVE(&nbdev_ch->io_path_list, io_path, nvme_io_path, stailq);



	TAILQ_REMOVE(&io_path->ctrlr_ch->io_path_list, io_path, tailq);
@@ -724,6 +728,10 @@ bdev_nvme_find_io_path(struct nvme_bdev_channel *nbdev_ch) 
{

	struct nvme_io_path *io_path, *non_optimized = NULL;



	if (spdk_likely(nbdev_ch->current_io_path != NULL)) {

		return nbdev_ch->current_io_path;

	}



	STAILQ_FOREACH(io_path, &nbdev_ch->io_path_list, stailq) {

		if (spdk_unlikely(!nvme_io_path_is_connected(io_path))) {

			/* The device is currently resetting. */
@@ -736,6 +744,7 @@ bdev_nvme_find_io_path(struct nvme_bdev_channel *nbdev_ch) 


		switch (io_path->nvme_ns->ana_state) {

		case SPDK_NVME_ANA_OPTIMIZED_STATE:

			nbdev_ch->current_io_path = io_path;

			return io_path;

		case SPDK_NVME_ANA_NON_OPTIMIZED_STATE:

			if (non_optimized == NULL) {
@@ -870,6 +879,7 @@ bdev_nvme_io_complete_nvme_status(struct nvme_bdev_io *bio, 
	    spdk_nvme_cpl_is_aborted_sq_deletion(cpl) ||

	    !nvme_io_path_is_available(bio->io_path) ||

	    nvme_io_path_is_failed(bio->io_path)) {

		nbdev_ch->current_io_path = NULL;

		if (spdk_nvme_cpl_is_ana_error(cpl)) {

			bio->io_path->nvme_ns->ana_state_updating = true;

			nvme_ctrlr_read_ana_log_page(nvme_ctrlr);
@@ -916,6 +926,8 @@ bdev_nvme_io_complete(struct nvme_bdev_io *bio, int rc) 
	case -ENXIO:

		nbdev_ch = spdk_io_channel_get_ctx(spdk_bdev_io_get_io_channel(bdev_io));



		nbdev_ch->current_io_path = NULL;



		if (!bdev_nvme_io_type_is_admin(bdev_io->type) &&

		    any_io_path_may_become_available(nbdev_ch)) {

			bdev_nvme_queue_retry_io(nbdev_ch, bio, 1000ULL);
@@ -932,6 +944,16 @@ bdev_nvme_io_complete(struct nvme_bdev_io *bio, int rc) 
	spdk_bdev_io_complete(bdev_io, io_status);

}



static void

_bdev_nvme_clear_io_path_cache(struct nvme_ctrlr_channel *ctrlr_ch)

{

	struct nvme_io_path *io_path;



	TAILQ_FOREACH(io_path, &ctrlr_ch->io_path_list, tailq) {

		io_path->nbdev_ch->current_io_path = NULL;

	}

}



static struct nvme_ctrlr_channel *

nvme_poll_group_get_ctrlr_channel(struct nvme_poll_group *group,

				  struct spdk_nvme_qpair *qpair)
@@ -954,6 +976,8 @@ bdev_nvme_destroy_qpair(struct nvme_ctrlr_channel *ctrlr_ch) 
		spdk_nvme_ctrlr_free_io_qpair(ctrlr_ch->qpair);

		ctrlr_ch->qpair = NULL;

	}



	_bdev_nvme_clear_io_path_cache(ctrlr_ch);

}



static void
@@ -2766,15 +2790,21 @@ nvme_ctrlr_set_ana_states(const struct spdk_nvme_ana_group_descriptor *desc, 
}



static void

nvme_ctrlr_read_ana_log_page_done(void *ctx, const struct spdk_nvme_cpl *cpl)

bdev_nvme_clear_io_path_cache(struct spdk_io_channel_iter *i)

{

	struct nvme_ctrlr *nvme_ctrlr = ctx;

	struct spdk_io_channel *_ch = spdk_io_channel_iter_get_channel(i);

	struct nvme_ctrlr_channel *ctrlr_ch = spdk_io_channel_get_ctx(_ch);



	if (cpl != NULL && spdk_nvme_cpl_is_success(cpl)) {

		bdev_nvme_parse_ana_log_page(nvme_ctrlr, nvme_ctrlr_set_ana_states,

					     nvme_ctrlr);

	_bdev_nvme_clear_io_path_cache(ctrlr_ch);



	spdk_for_each_channel_continue(i, 0);

}



static void

bdev_nvme_clear_io_path_cache_done(struct spdk_io_channel_iter *i, int status)

{

	struct nvme_ctrlr *nvme_ctrlr = spdk_io_channel_iter_get_io_device(i);



	pthread_mutex_lock(&nvme_ctrlr->mutex);



	assert(nvme_ctrlr->ana_log_page_updating == true);
@@ -2791,6 +2821,22 @@ nvme_ctrlr_read_ana_log_page_done(void *ctx, const struct spdk_nvme_cpl *cpl) 
	nvme_ctrlr_unregister(nvme_ctrlr);

}



static void

nvme_ctrlr_read_ana_log_page_done(void *ctx, const struct spdk_nvme_cpl *cpl)

{

	struct nvme_ctrlr *nvme_ctrlr = ctx;



	if (cpl != NULL && spdk_nvme_cpl_is_success(cpl)) {

		bdev_nvme_parse_ana_log_page(nvme_ctrlr, nvme_ctrlr_set_ana_states,

					     nvme_ctrlr);

	}



	spdk_for_each_channel(nvme_ctrlr,

			      bdev_nvme_clear_io_path_cache,

			      NULL,

			      bdev_nvme_clear_io_path_cache_done);

}



static void

nvme_ctrlr_read_ana_log_page(struct nvme_ctrlr *nvme_ctrlr)

{

module/bdev/nvme/bdev_nvme.h

+1 −0
Original line number Diff line number Diff line
@@ -185,6 +185,7 @@ struct nvme_io_path { 
};



struct nvme_bdev_channel {

	struct nvme_io_path			*current_io_path;

	STAILQ_HEAD(, nvme_io_path)		io_path_list;

	TAILQ_HEAD(retry_io_head, spdk_bdev_io)	retry_io_list;

	struct spdk_poller			*retry_io_poller;

test/unit/lib/bdev/nvme/bdev_nvme.c/bdev_nvme_ut.c

+9 −0
Original line number Diff line number Diff line
@@ -3778,9 +3778,13 @@ test_find_io_path(void) 
	nvme_ns1.ana_state = SPDK_NVME_ANA_OPTIMIZED_STATE;

	CU_ASSERT(bdev_nvme_find_io_path(&nbdev_ch) == &io_path1);



	nbdev_ch.current_io_path = NULL;



	nvme_ns1.ana_state = SPDK_NVME_ANA_NON_OPTIMIZED_STATE;

	CU_ASSERT(bdev_nvme_find_io_path(&nbdev_ch) == &io_path1);



	nbdev_ch.current_io_path = NULL;



	/* Test if io_path whose qpair is resetting is excluced. */



	ctrlr_ch1.qpair = NULL;
@@ -3798,8 +3802,12 @@ test_find_io_path(void) 
	nvme_ns2.ana_state = SPDK_NVME_ANA_OPTIMIZED_STATE;

	CU_ASSERT(bdev_nvme_find_io_path(&nbdev_ch) == &io_path2);



	nbdev_ch.current_io_path = NULL;



	nvme_ns2.ana_state = SPDK_NVME_ANA_NON_OPTIMIZED_STATE;

	CU_ASSERT(bdev_nvme_find_io_path(&nbdev_ch) == &io_path1);



	nbdev_ch.current_io_path = NULL;

}



static void
@@ -3947,6 +3955,7 @@ test_retry_io_if_ctrlr_is_resetting(void) 


	/* If ANA state of namespace is inaccessible, I/O should be queued. */

	nvme_ns->ana_state = SPDK_NVME_ANA_INACCESSIBLE_STATE;

	nbdev_ch->current_io_path = NULL;



	bdev_io1->internal.in_submit_request = true;