[01/13] block: move queues types to the block layer
diff mbox series

Message ID 20181129191310.9795-2-hch@lst.de
State New
Headers show
Series
  • [01/13] block: move queues types to the block layer
Related show

Commit Message

Christoph Hellwig Nov. 29, 2018, 7:12 p.m. UTC
Having another indirect all in the fast path doesn't really help
in our post-spectre world.  Also having too many queue type is just
going to create confusion, so I'd rather manage them centrally.

Note that the queue type naming and ordering changes a bit - the
first index now is the defauly queue for everything not explicitly
marked, the optional ones are read and poll queues.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-mq.h          | 21 +++++++------
 drivers/nvme/host/pci.c | 68 +++++++++++++++--------------------------
 include/linux/blk-mq.h  | 15 ++++-----
 3 files changed, 43 insertions(+), 61 deletions(-)

Comments

Jens Axboe Nov. 29, 2018, 7:50 p.m. UTC | #1
On 11/29/18 12:12 PM, Christoph Hellwig wrote:
> Having another indirect all in the fast path doesn't really help
> in our post-spectre world.  Also having too many queue type is just
> going to create confusion, so I'd rather manage them centrally.
> 
> Note that the queue type naming and ordering changes a bit - the
> first index now is the defauly queue for everything not explicitly
                         ^^^^^^^

default

> marked, the optional ones are read and poll queues.

Looks fine to me, was hoping NOT to bring this into the core, but
I guess it might be more manageable there in the long run. And it's
hard to argue with getting rid of the flags_to_type indirect.

Side not, should probably make the sysfs 'type' attribute a string
at this point.
Keith Busch Nov. 29, 2018, 8:19 p.m. UTC | #2
On Thu, Nov 29, 2018 at 08:12:58PM +0100, Christoph Hellwig wrote:
> +enum hctx_type {
> +	HCTX_TYPE_DEFAULT,	/* all I/O not otherwise accounted for */
> +	HCTX_TYPE_READ,		/* just for READ I/O */
> +	HCTX_TYPE_POLL,		/* polled I/O of any kind */
> +
> +	HCTX_MAX_TYPES,
>  };

Well, there goes my plan to use this with Weighted-Round-Robin NVMe IO
queues!

I'm not that sad about it, though.

Reviewed-by: Keith Busch <keith.busch@intel.com>
Jens Axboe Nov. 29, 2018, 8:25 p.m. UTC | #3
On 11/29/18 1:19 PM, Keith Busch wrote:
> On Thu, Nov 29, 2018 at 08:12:58PM +0100, Christoph Hellwig wrote:
>> +enum hctx_type {
>> +	HCTX_TYPE_DEFAULT,	/* all I/O not otherwise accounted for */
>> +	HCTX_TYPE_READ,		/* just for READ I/O */
>> +	HCTX_TYPE_POLL,		/* polled I/O of any kind */
>> +
>> +	HCTX_MAX_TYPES,
>>  };
> 
> Well, there goes my plan to use this with Weighted-Round-Robin NVMe IO
> queues!
> 
> I'm not that sad about it, though.

That's why I wanted these to be driver private, so you could just expand
at will. But it's not like we can't do that now, we'll just add some
extra types in here. The downside is that if we have a bunch of drivers
having dis-separate queues, then we end up with a bigger MAX number than
we would have with driver private queues.

IOW, I don't think this is ruining anything for you.
Christoph Hellwig Nov. 30, 2018, 7:56 a.m. UTC | #4
On Thu, Nov 29, 2018 at 07:50:09PM +0000, Jens Axboe wrote:
> > in our post-spectre world.  Also having too many queue type is just
> > going to create confusion, so I'd rather manage them centrally.
> > 
> > Note that the queue type naming and ordering changes a bit - the
> > first index now is the defauly queue for everything not explicitly
>                          ^^^^^^^
> 
> default

Fixed.

> Side not, should probably make the sysfs 'type' attribute a string
> at this point.

Fixed as well.
Christoph Hellwig Nov. 30, 2018, 8 a.m. UTC | #5
On Thu, Nov 29, 2018 at 01:19:14PM -0700, Keith Busch wrote:
> On Thu, Nov 29, 2018 at 08:12:58PM +0100, Christoph Hellwig wrote:
> > +enum hctx_type {
> > +	HCTX_TYPE_DEFAULT,	/* all I/O not otherwise accounted for */
> > +	HCTX_TYPE_READ,		/* just for READ I/O */
> > +	HCTX_TYPE_POLL,		/* polled I/O of any kind */
> > +
> > +	HCTX_MAX_TYPES,
> >  };
> 
> Well, there goes my plan to use this with Weighted-Round-Robin NVMe IO
> queues!

Wo between what do you even want to round robin?  If it is between
reads and writes that's easy.  If we want priority reads or writes
(separate from polling) that's also still fairly easily.

Btw, one thing I wanted to try once I get hold of the right hardware
is to mark the poll queues as priority queues and see if that makes
any differents in poll IOPS/latency.
Keith Busch Nov. 30, 2018, 2:40 p.m. UTC | #6
On Fri, Nov 30, 2018 at 12:00:13AM -0800, Christoph Hellwig wrote:
> On Thu, Nov 29, 2018 at 01:19:14PM -0700, Keith Busch wrote:
> > On Thu, Nov 29, 2018 at 08:12:58PM +0100, Christoph Hellwig wrote:
> > > +enum hctx_type {
> > > +	HCTX_TYPE_DEFAULT,	/* all I/O not otherwise accounted for */
> > > +	HCTX_TYPE_READ,		/* just for READ I/O */
> > > +	HCTX_TYPE_POLL,		/* polled I/O of any kind */
> > > +
> > > +	HCTX_MAX_TYPES,
> > >  };
> > 
> > Well, there goes my plan to use this with Weighted-Round-Robin NVMe IO
> > queues!
> 
> Wo between what do you even want to round robin?  If it is between
> reads and writes that's easy.  If we want priority reads or writes
> (separate from polling) that's also still fairly easily.

I was considering the IOPRIO_PRIO_CLASS. There are four of them, which
may roughly correspond to the four NVMe IO queues weights. Maybe even
through HIPRI flagged IOs with the RT class.

 
> Btw, one thing I wanted to try once I get hold of the right hardware
> is to mark the poll queues as priority queues and see if that makes
> any differents in poll IOPS/latency.

I doubt it will make much difference in IOPS, but should improve latency
on hipri IOs at the expense of normal IO since hipri will be fetched
ahead during command arbitrarion.
Jens Axboe Nov. 30, 2018, 3:20 p.m. UTC | #7
On 11/30/18 1:00 AM, Christoph Hellwig wrote:
> On Thu, Nov 29, 2018 at 01:19:14PM -0700, Keith Busch wrote:
>> On Thu, Nov 29, 2018 at 08:12:58PM +0100, Christoph Hellwig wrote:
>>> +enum hctx_type {
>>> +	HCTX_TYPE_DEFAULT,	/* all I/O not otherwise accounted for */
>>> +	HCTX_TYPE_READ,		/* just for READ I/O */
>>> +	HCTX_TYPE_POLL,		/* polled I/O of any kind */
>>> +
>>> +	HCTX_MAX_TYPES,
>>>  };
>>
>> Well, there goes my plan to use this with Weighted-Round-Robin NVMe IO
>> queues!
> 
> Wo between what do you even want to round robin?  If it is between
> reads and writes that's easy.  If we want priority reads or writes
> (separate from polling) that's also still fairly easily.
> 
> Btw, one thing I wanted to try once I get hold of the right hardware
> is to mark the poll queues as priority queues and see if that makes
> any differents in poll IOPS/latency.

Probably not a lot, if anything. Only for heavily mixed cases I'd
suspect it to make a difference. I can run some tests with it.

And beware that I've seen weird queue priority issues, ala the one
fixed by:

commit 9abd68ef454c824bfd18629033367b4382b5f390 (tag: for-linus-20180511)
Author: Jens Axboe <axboe@kernel.dk>
Date:   Tue May 8 10:25:15 2018 -0600

    nvme: add quirk to force medium priority for SQ creation

So we need to be careful with enabling priorities, I suspect. Hopefully
that's a standalone case.
Jens Axboe Nov. 30, 2018, 3:20 p.m. UTC | #8
On 11/30/18 12:56 AM, Christoph Hellwig wrote:
> On Thu, Nov 29, 2018 at 07:50:09PM +0000, Jens Axboe wrote:
>>> in our post-spectre world.  Also having too many queue type is just
>>> going to create confusion, so I'd rather manage them centrally.
>>>
>>> Note that the queue type naming and ordering changes a bit - the
>>> first index now is the defauly queue for everything not explicitly
>>                          ^^^^^^^
>>
>> default
> 
> Fixed.
> 
>> Side not, should probably make the sysfs 'type' attribute a string
>> at this point.
> 
> Fixed as well.

Thanks - are you going to post a v3? Would like to get this staged.
Christoph Hellwig Nov. 30, 2018, 3:21 p.m. UTC | #9
On Fri, Nov 30, 2018 at 03:20:51PM +0000, Jens Axboe wrote:
> Thanks - are you going to post a v3? Would like to get this staged.

Yes, will do.  Either late tonight or over the weekend.

Patch
diff mbox series

diff --git a/block/blk-mq.h b/block/blk-mq.h
index 7291e5379358..a664ea44ffd4 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -81,16 +81,14 @@  extern int blk_mq_hw_queue_to_node(struct blk_mq_queue_map *qmap, unsigned int);
 /*
  * blk_mq_map_queue_type() - map (hctx_type,cpu) to hardware queue
  * @q: request queue
- * @hctx_type: the hctx type index
+ * @type: the hctx type index
  * @cpu: CPU
  */
 static inline struct blk_mq_hw_ctx *blk_mq_map_queue_type(struct request_queue *q,
-							  unsigned int hctx_type,
+							  enum hctx_type type,
 							  unsigned int cpu)
 {
-	struct blk_mq_tag_set *set = q->tag_set;
-
-	return q->queue_hw_ctx[set->map[hctx_type].mq_map[cpu]];
+	return q->queue_hw_ctx[q->tag_set->map[type].mq_map[cpu]];
 }
 
 /*
@@ -103,12 +101,17 @@  static inline struct blk_mq_hw_ctx *blk_mq_map_queue(struct request_queue *q,
 						     unsigned int flags,
 						     unsigned int cpu)
 {
-	int hctx_type = 0;
+	enum hctx_type type = HCTX_TYPE_DEFAULT;
+
+	if (q->tag_set->nr_maps > HCTX_TYPE_POLL &&
+	    ((flags & REQ_HIPRI) && test_bit(QUEUE_FLAG_POLL, &q->queue_flags)))
+		type = HCTX_TYPE_POLL;
 
-	if (q->mq_ops->rq_flags_to_type)
-		hctx_type = q->mq_ops->rq_flags_to_type(q, flags);
+	else if (q->tag_set->nr_maps > HCTX_TYPE_READ &&
+		 ((flags & REQ_OP_MASK) == REQ_OP_READ))
+		type = HCTX_TYPE_READ;
 
-	return blk_mq_map_queue_type(q, hctx_type, cpu);
+	return blk_mq_map_queue_type(q, type, cpu);
 }
 
 /*
diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index 527907aa6903..a1bb4bb92e7f 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -95,13 +95,6 @@  struct nvme_queue;
 
 static void nvme_dev_disable(struct nvme_dev *dev, bool shutdown);
 
-enum {
-	NVMEQ_TYPE_READ,
-	NVMEQ_TYPE_WRITE,
-	NVMEQ_TYPE_POLL,
-	NVMEQ_TYPE_NR,
-};
-
 /*
  * Represents an NVM Express device.  Each nvme_dev is a PCI function.
  */
@@ -115,7 +108,7 @@  struct nvme_dev {
 	struct dma_pool *prp_small_pool;
 	unsigned online_queues;
 	unsigned max_qid;
-	unsigned io_queues[NVMEQ_TYPE_NR];
+	unsigned io_queues[HCTX_MAX_TYPES];
 	unsigned int num_vecs;
 	int q_depth;
 	u32 db_stride;
@@ -499,10 +492,10 @@  static int nvme_pci_map_queues(struct blk_mq_tag_set *set)
 
 		map->nr_queues = dev->io_queues[i];
 		if (!map->nr_queues) {
-			BUG_ON(i == NVMEQ_TYPE_READ);
+			BUG_ON(i == HCTX_TYPE_DEFAULT);
 
 			/* shared set, resuse read set parameters */
-			map->nr_queues = dev->io_queues[NVMEQ_TYPE_READ];
+			map->nr_queues = dev->io_queues[HCTX_TYPE_DEFAULT];
 			qoff = 0;
 			offset = queue_irq_offset(dev);
 		}
@@ -512,7 +505,7 @@  static int nvme_pci_map_queues(struct blk_mq_tag_set *set)
 		 * affinity), so use the regular blk-mq cpu mapping
 		 */
 		map->queue_offset = qoff;
-		if (i != NVMEQ_TYPE_POLL)
+		if (i != HCTX_TYPE_POLL)
 			blk_mq_pci_map_queues(map, to_pci_dev(dev->dev), offset);
 		else
 			blk_mq_map_queues(map);
@@ -961,16 +954,6 @@  static blk_status_t nvme_queue_rq(struct blk_mq_hw_ctx *hctx,
 	return ret;
 }
 
-static int nvme_rq_flags_to_type(struct request_queue *q, unsigned int flags)
-{
-	if ((flags & REQ_HIPRI) && test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
-		return NVMEQ_TYPE_POLL;
-	if ((flags & REQ_OP_MASK) == REQ_OP_READ)
-		return NVMEQ_TYPE_READ;
-
-	return NVMEQ_TYPE_WRITE;
-}
-
 static void nvme_pci_complete_rq(struct request *req)
 {
 	struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
@@ -1634,7 +1617,6 @@  static const struct blk_mq_ops nvme_mq_admin_ops = {
 #define NVME_SHARED_MQ_OPS					\
 	.queue_rq		= nvme_queue_rq,		\
 	.commit_rqs		= nvme_commit_rqs,		\
-	.rq_flags_to_type	= nvme_rq_flags_to_type,	\
 	.complete		= nvme_pci_complete_rq,		\
 	.init_hctx		= nvme_init_hctx,		\
 	.init_request		= nvme_init_request,		\
@@ -1785,9 +1767,9 @@  static int nvme_create_io_queues(struct nvme_dev *dev)
 	}
 
 	max = min(dev->max_qid, dev->ctrl.queue_count - 1);
-	if (max != 1 && dev->io_queues[NVMEQ_TYPE_POLL]) {
-		rw_queues = dev->io_queues[NVMEQ_TYPE_READ] +
-				dev->io_queues[NVMEQ_TYPE_WRITE];
+	if (max != 1 && dev->io_queues[HCTX_TYPE_POLL]) {
+		rw_queues = dev->io_queues[HCTX_TYPE_DEFAULT] +
+				dev->io_queues[HCTX_TYPE_READ];
 	} else {
 		rw_queues = max;
 	}
@@ -2076,9 +2058,9 @@  static void nvme_calc_io_queues(struct nvme_dev *dev, unsigned int nr_io_queues)
 	 * Setup read/write queue split
 	 */
 	if (nr_io_queues == 1) {
-		dev->io_queues[NVMEQ_TYPE_READ] = 1;
-		dev->io_queues[NVMEQ_TYPE_WRITE] = 0;
-		dev->io_queues[NVMEQ_TYPE_POLL] = 0;
+		dev->io_queues[HCTX_TYPE_DEFAULT] = 1;
+		dev->io_queues[HCTX_TYPE_READ] = 0;
+		dev->io_queues[HCTX_TYPE_POLL] = 0;
 		return;
 	}
 
@@ -2095,10 +2077,10 @@  static void nvme_calc_io_queues(struct nvme_dev *dev, unsigned int nr_io_queues)
 			this_p_queues = nr_io_queues - 1;
 		}
 
-		dev->io_queues[NVMEQ_TYPE_POLL] = this_p_queues;
+		dev->io_queues[HCTX_TYPE_POLL] = this_p_queues;
 		nr_io_queues -= this_p_queues;
 	} else
-		dev->io_queues[NVMEQ_TYPE_POLL] = 0;
+		dev->io_queues[HCTX_TYPE_POLL] = 0;
 
 	/*
 	 * If 'write_queues' is set, ensure it leaves room for at least
@@ -2112,11 +2094,11 @@  static void nvme_calc_io_queues(struct nvme_dev *dev, unsigned int nr_io_queues)
 	 * a queue set.
 	 */
 	if (!this_w_queues) {
-		dev->io_queues[NVMEQ_TYPE_WRITE] = 0;
-		dev->io_queues[NVMEQ_TYPE_READ] = nr_io_queues;
+		dev->io_queues[HCTX_TYPE_DEFAULT] = nr_io_queues;
+		dev->io_queues[HCTX_TYPE_READ] = 0;
 	} else {
-		dev->io_queues[NVMEQ_TYPE_WRITE] = this_w_queues;
-		dev->io_queues[NVMEQ_TYPE_READ] = nr_io_queues - this_w_queues;
+		dev->io_queues[HCTX_TYPE_DEFAULT] = this_w_queues;
+		dev->io_queues[HCTX_TYPE_READ] = nr_io_queues - this_w_queues;
 	}
 }
 
@@ -2138,8 +2120,8 @@  static int nvme_setup_irqs(struct nvme_dev *dev, int nr_io_queues)
 	 */
 	do {
 		nvme_calc_io_queues(dev, nr_io_queues);
-		irq_sets[0] = dev->io_queues[NVMEQ_TYPE_READ];
-		irq_sets[1] = dev->io_queues[NVMEQ_TYPE_WRITE];
+		irq_sets[0] = dev->io_queues[HCTX_TYPE_DEFAULT];
+		irq_sets[1] = dev->io_queues[HCTX_TYPE_READ];
 		if (!irq_sets[1])
 			affd.nr_sets = 1;
 
@@ -2226,12 +2208,12 @@  static int nvme_setup_io_queues(struct nvme_dev *dev)
 
 	dev->num_vecs = result;
 	result = max(result - 1, 1);
-	dev->max_qid = result + dev->io_queues[NVMEQ_TYPE_POLL];
+	dev->max_qid = result + dev->io_queues[HCTX_TYPE_POLL];
 
-	dev_info(dev->ctrl.device, "%d/%d/%d read/write/poll queues\n",
-					dev->io_queues[NVMEQ_TYPE_READ],
-					dev->io_queues[NVMEQ_TYPE_WRITE],
-					dev->io_queues[NVMEQ_TYPE_POLL]);
+	dev_info(dev->ctrl.device, "%d/%d/%d default/read/poll queues\n",
+					dev->io_queues[HCTX_TYPE_DEFAULT],
+					dev->io_queues[HCTX_TYPE_READ],
+					dev->io_queues[HCTX_TYPE_POLL]);
 
 	/*
 	 * Should investigate if there's a performance win from allocating
@@ -2332,13 +2314,13 @@  static int nvme_dev_add(struct nvme_dev *dev)
 	int ret;
 
 	if (!dev->ctrl.tagset) {
-		if (!dev->io_queues[NVMEQ_TYPE_POLL])
+		if (!dev->io_queues[HCTX_TYPE_POLL])
 			dev->tagset.ops = &nvme_mq_ops;
 		else
 			dev->tagset.ops = &nvme_mq_poll_noirq_ops;
 
 		dev->tagset.nr_hw_queues = dev->online_queues - 1;
-		dev->tagset.nr_maps = NVMEQ_TYPE_NR;
+		dev->tagset.nr_maps = HCTX_MAX_TYPES;
 		dev->tagset.timeout = NVME_IO_TIMEOUT;
 		dev->tagset.numa_node = dev_to_node(dev->dev);
 		dev->tagset.queue_depth =
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index 467f1dd21ccf..57eda7b20243 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -81,8 +81,12 @@  struct blk_mq_queue_map {
 	unsigned int queue_offset;
 };
 
-enum {
-	HCTX_MAX_TYPES = 3,
+enum hctx_type {
+	HCTX_TYPE_DEFAULT,	/* all I/O not otherwise accounted for */
+	HCTX_TYPE_READ,		/* just for READ I/O */
+	HCTX_TYPE_POLL,		/* polled I/O of any kind */
+
+	HCTX_MAX_TYPES,
 };
 
 struct blk_mq_tag_set {
@@ -118,8 +122,6 @@  struct blk_mq_queue_data {
 typedef blk_status_t (queue_rq_fn)(struct blk_mq_hw_ctx *,
 		const struct blk_mq_queue_data *);
 typedef void (commit_rqs_fn)(struct blk_mq_hw_ctx *);
-/* takes rq->cmd_flags as input, returns a hardware type index */
-typedef int (rq_flags_to_type_fn)(struct request_queue *, unsigned int);
 typedef bool (get_budget_fn)(struct blk_mq_hw_ctx *);
 typedef void (put_budget_fn)(struct blk_mq_hw_ctx *);
 typedef enum blk_eh_timer_return (timeout_fn)(struct request *, bool);
@@ -154,11 +156,6 @@  struct blk_mq_ops {
 	 */
 	commit_rqs_fn		*commit_rqs;
 
-	/*
-	 * Return a queue map type for the given request/bio flags
-	 */
-	rq_flags_to_type_fn	*rq_flags_to_type;
-
 	/*
 	 * Reserve budget before queue request, once .queue_rq is
 	 * run, it is driver's responsibility to release the