On Sat, 4 Oct 2008 15:01:20 +0900
FUJITA Tomonori <fujita.tomonori@xxxxxxxxxxxxx> wrote:
> On Fri, 03 Oct 2008 13:20:50 +0300
> Alexander Nezhinsky <nezhinsky@xxxxxxxxx> wrote:
>
> > Implements a NULL I/O backing store, which shortcuts all scsi commands
> > and returns a success status without actually performing them.
> > Useful for performance measurements, by eliminating or significantly
> > reducing I/O side bottleneck.
> >
> > Not based on bs_threads, commands are handled immediately from
> > the context they are submitted from.
> >
> > Comparison with the thread-based implementation shows the impact of
> > threading associated overhead on performance:
> > Read, thread: 69kIOPS, non-thread:72kIOPS
> > Write, thread: 47kIOPS, non-thread:59kIOPS
> > In both Read and Write tests, CPU util. dropped from ~30% to ~10%
> > when switched to the non-thread based model.
> >
> > To add a device backed by this, "-E null" should be used with tgtadm,
> > also a dummy device name (any string) should be supplied,
> > because bs_null does not use any real devices but the current
> > code won't admit a null-string device name.
> >
> >
> > Signed-off-by: Alexander Nezhinsky <nezhinsky@xxxxxxxxx>
> > ---
> >
> > diff --git a/usr/Makefile b/usr/Makefile
> > index 82ddf07..3fb4bf4 100644
> > --- a/usr/Makefile
> > +++ b/usr/Makefile
> > @@ -1,5 +1,7 @@
> > mandir = /usr/share/man
> >
> > +TGTD_OBJS += bs_null.o
> > +
> > ifneq ($(IBMVIO),)
> > CFLAGS += -DIBMVIO -DUSE_KERNEL
> > TGTD_OBJS += $(addprefix ibmvio/, ibmvio.o)
> > diff --git a/usr/bs_null.c b/usr/bs_null.c
> > new file mode 100644
> > index 0000000..6de30d8
> > --- /dev/null
> > +++ b/usr/bs_null.c
> > @@ -0,0 +1,91 @@
> > +/*
> > + * NULL I/O backing store routine
> > + *
> > + * Copyright (C) 2008 Alexander Nezhinsky <nezhinskyf@xxxxxxxxx>
> > + *
> > + * This program is free software; you can redistribute it and/or
> > + * modify it under the terms of the GNU General Public License as
> > + * published by the Free Software Foundation, version 2 of the
> > + * License.
> > + *
> > + * This program is distributed in the hope that it will be useful, but
> > + * WITHOUT ANY WARRANTY; without even the implied warranty of
> > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> > + * General Public License for more details.
> > + *
> > + * You should have received a copy of the GNU General Public License
> > + * along with this program; if not, write to the Free Software
> > + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA
> > + * 02110-1301 USA
> > + */
> > +
> > +#include <errno.h>
> > +#include <inttypes.h>
> > +#include <stdio.h>
> > +#include <stdlib.h>
> > +
> > +#include "list.h"
> > +#include "tgtd.h"
> > +#include "scsi.h"
> > +
> > +#define NULL_BS_DEV_SIZE (1ULL << 40)
> > +
> > +int bs_null_cmd_submit(struct scsi_cmd *cmd)
> > +{
> > + int result = SAM_STAT_GOOD;
> > + uint16_t asc = 0;
> > + uint8_t key = 0;
> > + uint8_t scb_op = cmd->scb[0];
> > +
> > + set_cmd_async(cmd);
> > +
> > + if (scb_op == SYNCHRONIZE_CACHE || scb_op == SYNCHRONIZE_CACHE_16) {
> > + if (cmd->scb[1] & 0x2) {
> > + result = SAM_STAT_CHECK_CONDITION;
> > + key = ILLEGAL_REQUEST;
> > + asc = ASC_INVALID_FIELD_IN_CDB;
> > + }
> > + }
>
> By default, the sbc code enables WCE. So if this code does the above,
> it should disable WCE by default (an initiator sends SYNCHRONIZE_CACHE
> and sees an error).
I removed the above code and applied it. Enjoy performance
measurements!
Thanks,
--
To unsubscribe from this list: send the line "unsubscribe stgt" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
