[src/trunk]: src/sys/dev/raidframe Implement DIOCGSTRATEGY and DIOCSSTRATEGY ...

[buhrow <buhrow%NetBSD.org@localhost>]

details:   https://anonhg.NetBSD.org/src/rev/789750a6f82b
branches:  trunk
changeset: 780855:789750a6f82b
user:      buhrow <buhrow%NetBSD.org@localhost>
date:      Thu Aug 09 23:53:25 2012 +0000

description:
Implement DIOCGSTRATEGY and DIOCSSTRATEGY to allow raidframe to use
different buffer queue strategies.
Initialize raid sets to use the default buffer queue strategy for the given
architecture, rather than forcing raidframe to use  fcfs in all cases.
This should cause raidframe to use the same buffer queue strategy as the
underlying disks.
On I386, I see performance enhancements of between 14 and 16% with raid5
sets with no other change.
See http://mail-index.NetBSD.org/tech-kern/2012/08/08/msg013758.html
for a discussion of this issue.

diffstat:

 sys/dev/raidframe/rf_netbsdkintf.c |  50 ++++++++++++++++++++++++++++++++++---
 1 files changed, 46 insertions(+), 4 deletions(-)

diffs (99 lines):

diff -r 75c6950e0fec -r 789750a6f82b sys/dev/raidframe/rf_netbsdkintf.c
--- a/sys/dev/raidframe/rf_netbsdkintf.c        Thu Aug 09 20:25:05 2012 +0000
+++ b/sys/dev/raidframe/rf_netbsdkintf.c        Thu Aug 09 23:53:25 2012 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: rf_netbsdkintf.c,v 1.297 2012/04/07 01:39:38 christos Exp $    */
+/*     $NetBSD: rf_netbsdkintf.c,v 1.298 2012/08/09 23:53:25 buhrow Exp $      */
 
 /*-
  * Copyright (c) 1996, 1997, 1998, 2008-2011 The NetBSD Foundation, Inc.
@@ -101,7 +101,7 @@
  ***********************************************************/
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: rf_netbsdkintf.c,v 1.297 2012/04/07 01:39:38 christos Exp $");
+__KERNEL_RCSID(0, "$NetBSD: rf_netbsdkintf.c,v 1.298 2012/08/09 23:53:25 buhrow Exp $");
 
 #ifdef _KERNEL_OPT
 #include "opt_compat_netbsd.h"
@@ -376,7 +376,7 @@
        memset(raid_softc, 0, num * sizeof(struct raid_softc));
 
        for (raidID = 0; raidID < num; raidID++) {
-               bufq_alloc(&raid_softc[raidID].buf_queue, "fcfs", 0);
+               bufq_alloc(&raid_softc[raidID].buf_queue, BUFQ_DISK_DEFAULT_STRAT, BUFQ_SORT_RAWBLOCK);
 
                RF_Malloc(raidPtrs[raidID], sizeof(RF_Raid_t),
                          (RF_Raid_t *));
@@ -983,7 +983,7 @@
 {
        int     unit = raidunit(dev);
        int     error = 0;
-       int     part, pmask;
+       int     part, pmask, s;
        cfdata_t cf;
        struct raid_softc *rs;
        RF_Config_t *k_cfg, *u_cfg;
@@ -1036,6 +1036,7 @@
        case DIOCWLABEL:
        case DIOCAWEDGE:
        case DIOCDWEDGE:
+       case DIOCSSTRATEGY:
                if ((flag & FWRITE) == 0)
                        return (EBADF);
        }
@@ -1088,6 +1089,8 @@
        case RAIDFRAME_PARITYMAP_GET_DISABLE:
        case RAIDFRAME_PARITYMAP_SET_DISABLE:
        case RAIDFRAME_PARITYMAP_SET_PARAMS:
+       case DIOCGSTRATEGY:
+       case DIOCSSTRATEGY:
                if ((rs->sc_flags & RAIDF_INITED) == 0)
                        return (ENXIO);
        }
@@ -1861,6 +1864,45 @@
                    (struct dkwedge_list *)data, l);
        case DIOCCACHESYNC:
                return rf_sync_component_caches(raidPtr);
+
+       case DIOCGSTRATEGY:
+           {
+               struct disk_strategy *dks = (void *)data;
+
+               s = splbio();
+               strlcpy(dks->dks_name, bufq_getstrategyname(rs->buf_queue),
+                   sizeof(dks->dks_name));
+               splx(s);
+               dks->dks_paramlen = 0;
+
+               return 0;
+           }
+       
+       case DIOCSSTRATEGY:
+           {
+               struct disk_strategy *dks = (void *)data;
+               struct bufq_state *new;
+               struct bufq_state *old;
+
+               if (dks->dks_param != NULL) {
+                       return EINVAL;
+               }
+               dks->dks_name[sizeof(dks->dks_name) - 1] = 0; /* ensure term */
+               error = bufq_alloc(&new, dks->dks_name,
+                   BUFQ_EXACT|BUFQ_SORT_RAWBLOCK);
+               if (error) {
+                       return error;
+               }
+               s = splbio();
+               old = rs->buf_queue;
+               bufq_move(new, old);
+               rs->buf_queue = new;
+               splx(s);
+               bufq_free(old);
+
+               return 0;
+           }
+
        default:
                retcode = ENOTTY;
        }