blk-mq: new multi-queue block IO queueing mechanism

Linux currently has two models for block devices:

- The classic request_fn based approach, where drivers use struct
  request units for IO. The block layer provides various helper
  functionalities to let drivers share code, things like tag
  management, timeout handling, queueing, etc.

- The "stacked" approach, where a driver squeezes in between the
  block layer and IO submitter. Since this bypasses the IO stack,
  driver generally have to manage everything themselves.

With drivers being written for new high IOPS devices, the classic
request_fn based driver doesn't work well enough. The design dates
back to when both SMP and high IOPS was rare. It has problems with
scaling to bigger machines, and runs into scaling issues even on
smaller machines when you have IOPS in the hundreds of thousands
per device.

The stacked approach is then most often selected as the model
for the driver. But this means that everybody has to re-invent
everything, and along with that we get all the problems again
that the shared approach solved.

This commit introduces blk-mq, block multi queue support. The
design is centered around per-cpu queues for queueing IO, which
then funnel down into x number of hardware submission queues.
We might have a 1:1 mapping between the two, or it might be
an N:M mapping. That all depends on what the hardware supports.

blk-mq provides various helper functions, which include:

- Scalable support for request tagging. Most devices need to
  be able to uniquely identify a request both in the driver and
  to the hardware. The tagging uses per-cpu caches for freed
  tags, to enable cache hot reuse.

- Timeout handling without tracking request on a per-device
  basis. Basically the driver should be able to get a notification,
  if a request happens to fail.

- Optional support for non 1:1 mappings between issue and
  submission queues. blk-mq can redirect IO completions to the
  desired location.

- Support for per-request payloads. Drivers almost always need
  to associate a request structure with some driver private
  command structure. Drivers can tell blk-mq this at init time,
  and then any request handed to the driver will have the
  required size of memory associated with it.

- Support for merging of IO, and plugging. The stacked model
  gets neither of these. Even for high IOPS devices, merging
  sequential IO reduces per-command overhead and thus
  increases bandwidth.

For now, this is provided as a potential 3rd queueing model, with
the hope being that, as it matures, it can replace both the classic
and stacked model. That would get us back to having just 1 real
model for block devices, leaving the stacked approach to dm/md
devices (as it was originally intended).

Contributions in this patch from the following people:

Shaohua Li <shli@fusionio.com>
Alexander Gordeev <agordeev@redhat.com>
Christoph Hellwig <hch@infradead.org>
Mike Christie <michaelc@cs.wisc.edu>
Matias Bjorling <m@bjorling.me>
Jeff Moyer <jmoyer@redhat.com>

Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

1 files changed, 204 insertions, 0 deletions

diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
new file mode 100644
index 000000000000..d64a02fb1f73
--- /dev/null
+++ b/block/blk-mq-tag.c
@@ -0,0 +1,204 @@
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/percpu_ida.h>
+
+#include <linux/blk-mq.h>
+#include "blk.h"
+#include "blk-mq.h"
+#include "blk-mq-tag.h"
+
+/*
+ * Per tagged queue (tag address space) map
+ */
+struct blk_mq_tags {
+        unsigned int nr_tags;
+        unsigned int nr_reserved_tags;
+        unsigned int nr_batch_move;
+        unsigned int nr_max_cache;
+
+        struct percpu_ida free_tags;
+        struct percpu_ida reserved_tags;
+};
+
+void blk_mq_wait_for_tags(struct blk_mq_tags *tags)
+{
+        int tag = blk_mq_get_tag(tags, __GFP_WAIT, false);
+        blk_mq_put_tag(tags, tag);
+}
+
+bool blk_mq_has_free_tags(struct blk_mq_tags *tags)
+{
+        return !tags ||
+                percpu_ida_free_tags(&tags->free_tags, nr_cpu_ids) != 0;
+}
+
+static unsigned int __blk_mq_get_tag(struct blk_mq_tags *tags, gfp_t gfp)
+{
+        int tag;
+
+        tag = percpu_ida_alloc(&tags->free_tags, gfp);
+        if (tag < 0)
+                return BLK_MQ_TAG_FAIL;
+        return tag + tags->nr_reserved_tags;
+}
+
+static unsigned int __blk_mq_get_reserved_tag(struct blk_mq_tags *tags,
+                                              gfp_t gfp)
+{
+        int tag;
+
+        if (unlikely(!tags->nr_reserved_tags)) {
+                WARN_ON_ONCE(1);
+                return BLK_MQ_TAG_FAIL;
+        }
+
+        tag = percpu_ida_alloc(&tags->reserved_tags, gfp);
+        if (tag < 0)
+                return BLK_MQ_TAG_FAIL;
+        return tag;
+}
+
+unsigned int blk_mq_get_tag(struct blk_mq_tags *tags, gfp_t gfp, bool reserved)
+{
+        if (!reserved)
+                return __blk_mq_get_tag(tags, gfp);
+
+        return __blk_mq_get_reserved_tag(tags, gfp);
+}
+
+static void __blk_mq_put_tag(struct blk_mq_tags *tags, unsigned int tag)
+{
+        BUG_ON(tag >= tags->nr_tags);
+
+        percpu_ida_free(&tags->free_tags, tag - tags->nr_reserved_tags);
+}
+
+static void __blk_mq_put_reserved_tag(struct blk_mq_tags *tags,
+                                      unsigned int tag)
+{
+        BUG_ON(tag >= tags->nr_reserved_tags);
+
+        percpu_ida_free(&tags->reserved_tags, tag);
+}
+
+void blk_mq_put_tag(struct blk_mq_tags *tags, unsigned int tag)
+{
+        if (tag >= tags->nr_reserved_tags)
+                __blk_mq_put_tag(tags, tag);
+        else
+                __blk_mq_put_reserved_tag(tags, tag);
+}
+
+static int __blk_mq_tag_iter(unsigned id, void *data)
+{
+        unsigned long *tag_map = data;
+        __set_bit(id, tag_map);
+        return 0;
+}
+
+void blk_mq_tag_busy_iter(struct blk_mq_tags *tags,
+                          void (*fn)(void *, unsigned long *), void *data)
+{
+        unsigned long *tag_map;
+        size_t map_size;
+
+        map_size = ALIGN(tags->nr_tags, BITS_PER_LONG) / BITS_PER_LONG;
+        tag_map = kzalloc(map_size * sizeof(unsigned long), GFP_ATOMIC);
+        if (!tag_map)
+                return;
+
+        percpu_ida_for_each_free(&tags->free_tags, __blk_mq_tag_iter, tag_map);
+        if (tags->nr_reserved_tags)
+                percpu_ida_for_each_free(&tags->reserved_tags, __blk_mq_tag_iter,
+                        tag_map);
+
+        fn(data, tag_map);
+        kfree(tag_map);
+}
+
+struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
+                                     unsigned int reserved_tags, int node)
+{
+        unsigned int nr_tags, nr_cache;
+        struct blk_mq_tags *tags;
+        int ret;
+
+        if (total_tags > BLK_MQ_TAG_MAX) {
+                pr_err("blk-mq: tag depth too large\n");
+                return NULL;
+        }
+
+        tags = kzalloc_node(sizeof(*tags), GFP_KERNEL, node);
+        if (!tags)
+                return NULL;
+
+        nr_tags = total_tags - reserved_tags;
+        nr_cache = nr_tags / num_possible_cpus();
+
+        if (nr_cache < BLK_MQ_TAG_CACHE_MIN)
+                nr_cache = BLK_MQ_TAG_CACHE_MIN;
+        else if (nr_cache > BLK_MQ_TAG_CACHE_MAX)
+                nr_cache = BLK_MQ_TAG_CACHE_MAX;
+
+        tags->nr_tags = total_tags;
+        tags->nr_reserved_tags = reserved_tags;
+        tags->nr_max_cache = nr_cache;
+        tags->nr_batch_move = max(1u, nr_cache / 2);
+
+        ret = __percpu_ida_init(&tags->free_tags, tags->nr_tags -
+                                tags->nr_reserved_tags,
+                                tags->nr_max_cache,
+                                tags->nr_batch_move);
+        if (ret)
+                goto err_free_tags;
+
+        if (reserved_tags) {
+                /*
+                 * With max_cahe and batch set to 1, the allocator fallbacks to
+                 * no cached. It's fine reserved tags allocation is slow.
+                 */
+                ret = __percpu_ida_init(&tags->reserved_tags, reserved_tags,
+                                1, 1);
+                if (ret)
+                        goto err_reserved_tags;
+        }
+
+        return tags;
+
+err_reserved_tags:
+        percpu_ida_destroy(&tags->free_tags);
+err_free_tags:
+        kfree(tags);
+        return NULL;
+}
+
+void blk_mq_free_tags(struct blk_mq_tags *tags)
+{
+        percpu_ida_destroy(&tags->free_tags);
+        percpu_ida_destroy(&tags->reserved_tags);
+        kfree(tags);
+}
+
+ssize_t blk_mq_tag_sysfs_show(struct blk_mq_tags *tags, char *page)
+{
+        char *orig_page = page;
+        int cpu;
+
+        if (!tags)
+                return 0;
+
+        page += sprintf(page, "nr_tags=%u, reserved_tags=%u, batch_move=%u,"
+                        " max_cache=%u\n", tags->nr_tags, tags->nr_reserved_tags,
+                        tags->nr_batch_move, tags->nr_max_cache);
+
+        page += sprintf(page, "nr_free=%u, nr_reserved=%u\n",
+                        percpu_ida_free_tags(&tags->free_tags, nr_cpu_ids),
+                        percpu_ida_free_tags(&tags->reserved_tags, nr_cpu_ids));
+
+        for_each_possible_cpu(cpu) {
+                page += sprintf(page, "  cpu%02u: nr_free=%u\n", cpu,
+                                percpu_ida_free_tags(&tags->free_tags, cpu));
+        }
+
+        return page - orig_page;
+}