Merge GEOM direct dispatch changes from the projects/camlock branch.

When safety requirements are met, it allows to avoid passing I/O requests
to GEOM g_up/g_down thread, executing them directly in the caller context.
That allows to avoid CPU bottlenecks in g_up/g_down threads, plus avoid
several context switches per I/O.

The defined now safety requirements are:
 - caller should not hold any locks and should be reenterable;
 - callee should not depend on GEOM dual-threaded concurency semantics;
 - on the way down, if request is unmapped while callee doesn't support it,
   the context should be sleepable;
 - kernel thread stack usage should be below 50%.

To keep compatibility with GEOM classes not meeting above requirements
new provider and consumer flags added:
 - G_CF_DIRECT_SEND -- consumer code meets caller requirements (request);
 - G_CF_DIRECT_RECEIVE -- consumer code meets callee requirements (done);
 - G_PF_DIRECT_SEND -- provider code meets caller requirements (done);
 - G_PF_DIRECT_RECEIVE -- provider code meets callee requirements (request).
Capable GEOM class can set them, allowing direct dispatch in cases where
it is safe.  If any of requirements are not met, request is queued to
g_up or g_down thread same as before.

Such GEOM classes were reviewed and updated to support direct dispatch:
CONCAT, DEV, DISK, GATE, MD, MIRROR, MULTIPATH, NOP, PART, RAID, STRIPE,
VFS, ZERO, ZFS::VDEV, ZFS::ZVOL, all classes based on g_slice KPI (LABEL,
MAP, FLASHMAP, etc).

To declare direct completion capability disk(9) KPI got new flag equivalent
to G_PF_DIRECT_SEND -- DISKFLAG_DIRECT_COMPLETION.  da(4) and ada(4) disk
drivers got it set now thanks to earlier CAM locking work.

This change more then twice increases peak block storage performance on
systems with manu CPUs, together with earlier CAM locking changes reaching
more then 1 million IOPS (512 byte raw reads from 16 SATA SSDs on 4 HBAs to
256 user-level threads).

Sponsored by:	iXsystems, Inc.
MFC after:	2 months

7 files changed, 9 insertions, 0 deletions

diff --git a/sys/geom/raid/g_raid.c b/sys/geom/raid/g_raid.c
index 9933cf1..a161f8a 100644
--- a/sys/geom/raid/g_raid.c
+++ b/sys/geom/raid/g_raid.c
@@ -792,6 +792,7 @@ g_raid_open_consumer(struct g_raid_softc *sc, const char *name)
 	if (pp == NULL)
 		return (NULL);
 	cp = g_new_consumer(sc->sc_geom);
+	cp->flags |= G_CF_DIRECT_RECEIVE;
 	if (g_attach(cp, pp) != 0) {
 		g_destroy_consumer(cp);
 		return (NULL);
@@ -1670,6 +1671,7 @@ g_raid_launch_provider(struct g_raid_volume *vol)
         }
 
 	pp = g_new_providerf(sc->sc_geom, "%s", name);
+	pp->flags |= G_PF_DIRECT_RECEIVE;
 	if (vol->v_tr->tro_class->trc_accept_unmapped) {
 		pp->flags |= G_PF_ACCEPT_UNMAPPED;
 		for (i = 0; i < vol->v_disks_count; i++) {
@@ -2255,6 +2257,7 @@ g_raid_taste(struct g_class *mp, struct g_provider *pp, int flags __unused)
 	 */
 	gp->orphan = g_raid_taste_orphan;
 	cp = g_new_consumer(gp);
+	cp->flags |= G_CF_DIRECT_RECEIVE;
 	g_attach(cp, pp);
 
 	geom = NULL;
diff --git a/sys/geom/raid/md_ddf.c b/sys/geom/raid/md_ddf.c
index 5a17301..4e1545b 100644
--- a/sys/geom/raid/md_ddf.c
+++ b/sys/geom/raid/md_ddf.c
@@ -2143,6 +2143,7 @@ g_raid_md_taste_ddf(struct g_raid_md_object *md, struct g_class *mp,
 	}
 
 	rcp = g_new_consumer(geom);
+	rcp->flags |= G_CF_DIRECT_RECEIVE;
 	g_attach(rcp, pp);
 	if (g_access(rcp, 1, 1, 1) != 0)
 		; //goto fail1;
diff --git a/sys/geom/raid/md_intel.c b/sys/geom/raid/md_intel.c
index eeb42d5..11917f5 100644
--- a/sys/geom/raid/md_intel.c
+++ b/sys/geom/raid/md_intel.c
@@ -1477,6 +1477,7 @@ search:
 	}
 
 	rcp = g_new_consumer(geom);
+	rcp->flags |= G_CF_DIRECT_RECEIVE;
 	g_attach(rcp, pp);
 	if (g_access(rcp, 1, 1, 1) != 0)
 		; //goto fail1;
diff --git a/sys/geom/raid/md_jmicron.c b/sys/geom/raid/md_jmicron.c
index a062215..2da4a33 100644
--- a/sys/geom/raid/md_jmicron.c
+++ b/sys/geom/raid/md_jmicron.c
@@ -923,6 +923,7 @@ search:
 	}
 
 	rcp = g_new_consumer(geom);
+	rcp->flags |= G_CF_DIRECT_RECEIVE;
 	g_attach(rcp, pp);
 	if (g_access(rcp, 1, 1, 1) != 0)
 		; //goto fail1;
diff --git a/sys/geom/raid/md_nvidia.c b/sys/geom/raid/md_nvidia.c
index 92d9f71..25cc2cc 100644
--- a/sys/geom/raid/md_nvidia.c
+++ b/sys/geom/raid/md_nvidia.c
@@ -919,6 +919,7 @@ search:
 	}
 
 	rcp = g_new_consumer(geom);
+	rcp->flags |= G_CF_DIRECT_RECEIVE;
 	g_attach(rcp, pp);
 	if (g_access(rcp, 1, 1, 1) != 0)
 		; //goto fail1;
diff --git a/sys/geom/raid/md_promise.c b/sys/geom/raid/md_promise.c
index 0007b20..b1e4427 100644
--- a/sys/geom/raid/md_promise.c
+++ b/sys/geom/raid/md_promise.c
@@ -1176,6 +1176,7 @@ search:
 	}
 
 	rcp = g_new_consumer(geom);
+	rcp->flags |= G_CF_DIRECT_RECEIVE;
 	g_attach(rcp, pp);
 	if (g_access(rcp, 1, 1, 1) != 0)
 		; //goto fail1;
diff --git a/sys/geom/raid/md_sii.c b/sys/geom/raid/md_sii.c
index 03bb03b..149b336 100644
--- a/sys/geom/raid/md_sii.c
+++ b/sys/geom/raid/md_sii.c
@@ -1012,6 +1012,7 @@ search:
 	}
 
 	rcp = g_new_consumer(geom);
+	rcp->flags |= G_CF_DIRECT_RECEIVE;
 	g_attach(rcp, pp);
 	if (g_access(rcp, 1, 1, 1) != 0)
 		; //goto fail1;