Skip to contentSkip to conversation

nvme: remove support or stream based temperature hint

PATCH In review Christoph Hellwig 4 participants 5 messages Mar 3, 2022 NVMe

lore
Share & tools
lore

Christoph Hellwig

author maintainer
#0 lore raw

PATCH 1/1 nvme: remove support or stream based temperature hint

This support was added for RocksDB, but RocksDB ended up not using it. At the same time drives on the open marked (vs those build for OEMs for non-Linux support) that actually support streams are extremly rare. Don't bloat the nvme driver for it.

---
 drivers/nvme/host/core.c | 143 ---------------------------------------
 1 file changed, 143 deletions(-)
drivers/nvme/host/core.c +0 -143
index f8084ded69e50..61f29c59b14cf 100644
@@ -77,10 +77,6 @@ module_param(apst_secondary_latency_tol_us, ulong, 0644);
 MODULE_PARM_DESC(apst_secondary_latency_tol_us,
 	"secondary APST latency tolerance in us");
 
-static bool streams;
-module_param(streams, bool, 0644);
-MODULE_PARM_DESC(streams, "turn on support for Streams write directives");
-
 /*
  * nvme_wq - hosts nvme related works that are not reset or delete
  * nvme_reset_wq - hosts nvme reset works
@@ -747,108 +743,6 @@ bool __nvme_check_ready(struct nvme_ctrl *ctrl, struct request *rq,
 }
 EXPORT_SYMBOL_GPL(__nvme_check_ready);
 
-static int nvme_toggle_streams(struct nvme_ctrl *ctrl, bool enable)
-{
-	struct nvme_command c = { };
-
-	c.directive.opcode = nvme_admin_directive_send;
-	c.directive.nsid = cpu_to_le32(NVME_NSID_ALL);
-	c.directive.doper = NVME_DIR_SND_ID_OP_ENABLE;
-	c.directive.dtype = NVME_DIR_IDENTIFY;
-	c.directive.tdtype = NVME_DIR_STREAMS;
-	c.directive.endir = enable ? NVME_DIR_ENDIR : 0;
-
-	return nvme_submit_sync_cmd(ctrl->admin_q, &c, NULL, 0);
-}
-
-static int nvme_disable_streams(struct nvme_ctrl *ctrl)
-{
-	return nvme_toggle_streams(ctrl, false);
-}
-
-static int nvme_enable_streams(struct nvme_ctrl *ctrl)
-{
-	return nvme_toggle_streams(ctrl, true);
-}
-
-static int nvme_get_stream_params(struct nvme_ctrl *ctrl,
-				  struct streams_directive_params *s, u32 nsid)
-{
-	struct nvme_command c = { };
-
-	memset(s, 0, sizeof(*s));
-
-	c.directive.opcode = nvme_admin_directive_recv;
-	c.directive.nsid = cpu_to_le32(nsid);
-	c.directive.numd = cpu_to_le32(nvme_bytes_to_numd(sizeof(*s)));
-	c.directive.doper = NVME_DIR_RCV_ST_OP_PARAM;
-	c.directive.dtype = NVME_DIR_STREAMS;
-
-	return nvme_submit_sync_cmd(ctrl->admin_q, &c, s, sizeof(*s));
-}
-
-static int nvme_configure_directives(struct nvme_ctrl *ctrl)
-{
-	struct streams_directive_params s;
-	u16 nssa;
-	int ret;
-
-	if (!(ctrl->oacs & NVME_CTRL_OACS_DIRECTIVES))
-		return 0;
-	if (!streams)
-		return 0;
-
-	ret = nvme_enable_streams(ctrl);
-	if (ret)
-		return ret;
-
-	ret = nvme_get_stream_params(ctrl, &s, NVME_NSID_ALL);
-	if (ret)
-		goto out_disable_stream;
-
-	nssa = le16_to_cpu(s.nssa);
-	if (nssa < BLK_MAX_WRITE_HINTS - 1) {
-		dev_info(ctrl->device, "too few streams (%u) available\n",
-					nssa);
-		/* this condition is not an error: streams are optional */
-		ret = 0;
-		goto out_disable_stream;
-	}
-
-	ctrl->nr_streams = min_t(u16, nssa, BLK_MAX_WRITE_HINTS - 1);
-	dev_info(ctrl->device, "Using %u streams\n", ctrl->nr_streams);
-	return 0;
-
-out_disable_stream:
-	nvme_disable_streams(ctrl);
-	return ret;
-}
-
-/*
- * Check if 'req' has a write hint associated with it. If it does, assign
- * a valid namespace stream to the write.
- */
-static void nvme_assign_write_stream(struct nvme_ctrl *ctrl,
-				     struct request *req, u16 *control,
-				     u32 *dsmgmt)
-{
-	enum rw_hint streamid = req->write_hint;
-
-	if (streamid == WRITE_LIFE_NOT_SET || streamid == WRITE_LIFE_NONE)
-		streamid = 0;
-	else {
-		streamid--;
-		if (WARN_ON_ONCE(streamid > ctrl->nr_streams))
-			return;
-
-		*control |= NVME_RW_DTYPE_STREAMS;
-		*dsmgmt |= streamid << 16;
-	}
-
-	if (streamid < ARRAY_SIZE(req->q->write_hints))
-		req->q->write_hints[streamid] += blk_rq_bytes(req) >> 9;
-}
-
 static inline void nvme_setup_flush(struct nvme_ns *ns,
 		struct nvme_command *cmnd)
 {
@@ -952,7 +846,6 @@ static inline blk_status_t nvme_setup_rw(struct nvme_ns *ns,
 		struct request *req, struct nvme_command *cmnd,
 		enum nvme_opcode op)
 {
-	struct nvme_ctrl *ctrl = ns->ctrl;
 	u16 control = 0;
 	u32 dsmgmt = 0;
 
@@ -975,9 +868,6 @@ static inline blk_status_t nvme_setup_rw(struct nvme_ns *ns,
 	cmnd->rw.apptag = 0;
 	cmnd->rw.appmask = 0;
 
-	if (req_op(req) == REQ_OP_WRITE && ctrl->nr_streams)
-		nvme_assign_write_stream(ctrl, req, &control, &dsmgmt);
-
 	if (ns->ms) {
 		/*
 		 * If formated with metadata, the block layer always provides a
@@ -1696,9 +1586,6 @@ static void nvme_config_discard(struct gendisk *disk, struct nvme_ns *ns)
 		return;
 	}
 
-	if (ctrl->nr_streams && ns->sws && ns->sgs)
-		size *= ns->sws * ns->sgs;
-
 	BUILD_BUG_ON(PAGE_SIZE / sizeof(struct nvme_dsm_range) <
 			NVME_DSM_MAX_RANGES);
 
@@ -1724,31 +1611,6 @@ static bool nvme_ns_ids_equal(struct nvme_ns_ids *a, struct nvme_ns_ids *b)
 		a->csi == b->csi;
 }
 
-static int nvme_setup_streams_ns(struct nvme_ctrl *ctrl, struct nvme_ns *ns,
-				 u32 *phys_bs, u32 *io_opt)
-{
-	struct streams_directive_params s;
-	int ret;
-
-	if (!ctrl->nr_streams)
-		return 0;
-
-	ret = nvme_get_stream_params(ctrl, &s, ns->head->ns_id);
-	if (ret)
-		return ret;
-
-	ns->sws = le32_to_cpu(s.sws);
-	ns->sgs = le16_to_cpu(s.sgs);
-
-	if (ns->sws) {
-		*phys_bs = ns->sws * (1 << ns->lba_shift);
-		if (ns->sgs)
-			*io_opt = *phys_bs * ns->sgs;
-	}
-
-	return 0;
-}
-
 static int nvme_configure_metadata(struct nvme_ns *ns, struct nvme_id_ns *id)
 {
 	struct nvme_ctrl *ctrl = ns->ctrl;
@@ -1841,7 +1703,6 @@ static void nvme_update_disk_info(struct gendisk *disk,
 	blk_integrity_unregister(disk);
 
 	atomic_bs = phys_bs = bs;
-	nvme_setup_streams_ns(ns->ctrl, ns, &phys_bs, &io_opt);
 	if (id->nabo == 0) {
 		/*
 		 * Bit 1 indicates whether NAWUPF is defined for this namespace
@@ -3133,10 +2994,6 @@ int nvme_init_ctrl_finish(struct nvme_ctrl *ctrl)
 	if (ret < 0)
 		return ret;
 
-	ret = nvme_configure_directives(ctrl);
-	if (ret < 0)
-		return ret;
-
 	ret = nvme_configure_acre(ctrl);
 	if (ret < 0)
 		return ret;
Signature · 1 line
2.30.2
Acked-by on v1 1/1
#1 lore raw
Quoted 4 lines · Christoph Hellwig

This support was added for RocksDB, but RocksDB ended up not using it. At the same time drives on the open marked (vs those build for OEMs for non-Linux support) that actually support streams are extremly rare. Don't bloat the nvme driver for it.

Signature · 1 line
Jens Axboe
maintainer Reviewed-by on v1 1/1
#2 lore raw
In reply to Christoph Hellwig
Quoted 4 lines · Christoph Hellwig

This support was added for RocksDB, but RocksDB ended up not using it. At the same time drives on the open marked (vs those build for OEMs for non-Linux support) that actually support streams are extremly rare. Don't bloat the nvme driver for it.

Fine with me. It looks like the early interest in streams never really materialized, so it's unlikely we'll hear any complaints.

You can remove nr_streams from 'struct nvme_ctrl' while you're at it.

And without nvme, there doesn't appear to be a real user for the write_hint anymore. All you can really do with it now is prevent merging, which doesn't seem useful.

maintainer
#3 lore raw
Quoted 9 lines · Keith Busch
Quoted 4 lines · Christoph Hellwig

This support was added for RocksDB, but RocksDB ended up not using it. At the same time drives on the open marked (vs those build for OEMs for non-Linux support) that actually support streams are extremly rare. Don't bloat the nvme driver for it.

Fine with me. It looks like the early interest in streams never really materialized, so it's unlikely we'll hear any complaints.

Quoted 7 lines

You can remove nr_streams from 'struct nvme_ctrl' while you're at it.

And without nvme, there doesn't appear to be a real user for the write_hint anymore. All you can really do with it now is prevent merging, which doesn't seem useful.

Yes, I would suggest that we just prune the write hint and just leave the fcntl() ops as stubs as a followup patch.

Signature · 1 line
Jens Axboe
author maintainer
#4 lore raw
Quoted 7 lines · Jens Axboe
Quoted 4 lines

And without nvme, there doesn't appear to be a real user for the write_hint anymore. All you can really do with it now is prevent merging, which doesn't seem useful.

Yes, I would suggest that we just prune the write hint and just leave the fcntl() ops as stubs as a followup patch.

So we need to keep i_write_hint anyway for f2fs, but the block layer removal is easy. It has some conflicts with other work, what tree would you like it against?