Allocate pager bufs from UMA instead of 80-ish mutex protected linked list.

o In vm_pager_bufferinit() create pbuf_zone and start accounting on how many
  pbufs are we going to have set.
  In various subsystems that are going to utilize pbufs create private zones
  via call to pbuf_zsecond_create(). The latter calls uma_zsecond_create(),
  and sets a limit on created zone. After startup preallocate pbufs according
  to requirements of all pbuf zones.

  Subsystems that used to have a private limit with old allocator now have
  private pbuf zones: md(4), fusefs, NFS client, smbfs, VFS cluster, FFS,
  swap, vnode pager.

  The following subsystems use shared pbuf zone: cam(4), nvme(4), physio(9),
  aio(4). They should have their private limits, but changing that is out of
  scope of this commit.

o Fetch tunable value of kern.nswbuf from init_param2() and while here move
  NSWBUF_MIN to opt_param.h and eliminate opt_swap.h, that was holding only
  this option.
  Default values aren't touched by this commit, but they probably should be
  reviewed wrt to modern hardware.

This change removes a tight bottleneck from sendfile(2) operation, that
uses pbufs in vnode pager. Other pagers also would benefit from faster
allocation.

Together with:	gallatin
Tested by:	pho

1 files changed, 10 insertions, 8 deletions

diff --git a/sys/sys/buf.h b/sys/sys/buf.h
index a099a972efc0..fe49224417d9 100644
--- a/sys/sys/buf.h
+++ b/sys/sys/buf.h
@@ -44,6 +44,7 @@
 #include <sys/queue.h>
 #include <sys/lock.h>
 #include <sys/lockmgr.h>
+#include <vm/uma.h>
 
 struct bio;
 struct buf;
@@ -275,6 +276,11 @@ struct buf {
 #define	PRINT_BUF_VFLAGS "\20\4bkgrderr\3bkgrdwait\2bkgrdinprog\1scanned"
 
 #ifdef _KERNEL
+
+#ifndef NSWBUF_MIN
+#define	NSWBUF_MIN	16
+#endif
+
 /*
  * Buffer locking
  */
@@ -287,7 +293,7 @@ extern const char *buf_wmesg;		/* Default buffer lock message */
  * Initialize a lock.
  */
 #define BUF_LOCKINIT(bp)						\
-	lockinit(&(bp)->b_lock, PRIBIO + 4, buf_wmesg, 0, 0)
+	lockinit(&(bp)->b_lock, PRIBIO + 4, buf_wmesg, 0, LK_NEW)
 /*
  *
  * Get a lock sleeping non-interruptably until it becomes available.
@@ -493,10 +499,6 @@ extern int	bdwriteskip;
 extern int	dirtybufferflushes;
 extern int	altbufferflushes;
 extern int	nswbuf;			/* Number of swap I/O buffer headers. */
-extern int	cluster_pbuf_freecnt;	/* Number of pbufs for clusters */
-extern int	vnode_pbuf_freecnt;	/* Number of pbufs for vnode pager */
-extern int	vnode_async_pbuf_freecnt; /* Number of pbufs for vnode pager,
-					     asynchronous reads */
 extern caddr_t	unmapped_buf;	/* Data address for unmapped buffers. */
 
 static inline int
@@ -537,7 +539,6 @@ void	brelse(struct buf *);
 void	bqrelse(struct buf *);
 int	vfs_bio_awrite(struct buf *);
 void	vfs_drain_busy_pages(struct buf *bp);
-struct buf *     getpbuf(int *);
 struct buf *incore(struct bufobj *, daddr_t);
 struct buf *gbincore(struct bufobj *, daddr_t);
 struct buf *getblk(struct vnode *, daddr_t, int, int, int, int);
@@ -549,6 +550,9 @@ int	bufwrite(struct buf *);
 void	bufdone(struct buf *);
 void	bd_speedup(void);
 
+extern uma_zone_t pbuf_zone;
+uma_zone_t pbuf_zsecond_create(char *name, int max);
+
 int	cluster_read(struct vnode *, u_quad_t, daddr_t, long,
 	    struct ucred *, long, int, int, struct buf **);
 int	cluster_wbuild(struct vnode *, long, daddr_t, int, int);
@@ -562,7 +566,6 @@ void	vfs_busy_pages(struct buf *, int clear_modify);
 void	vfs_unbusy_pages(struct buf *);
 int	vmapbuf(struct buf *, int);
 void	vunmapbuf(struct buf *);
-void	relpbuf(struct buf *, int *);
 void	brelvp(struct buf *);
 void	bgetvp(struct vnode *, struct buf *);
 void	pbgetbo(struct bufobj *bo, struct buf *bp);
@@ -571,7 +574,6 @@ void	pbrelbo(struct buf *);
 void	pbrelvp(struct buf *);
 int	allocbuf(struct buf *bp, int size);
 void	reassignbuf(struct buf *);
-struct	buf *trypbuf(int *);
 void	bwait(struct buf *, u_char, const char *);
 void	bdone(struct buf *);