public inbox for linux-block@vger.kernel.org
 help / color / mirror / Atom feed
From: Douglas Gilbert <dgilbert@interlog.com>
To: linux-scsi@vger.kernel.org, axboe@kernel.dk, linux-block@vger.kernel.org
Cc: martin.petersen@oracle.com, jejb@linux.vnet.ibm.com,
	hare@suse.de, kashyap.desai@broadcom.com
Subject: [PATCH v2] fio: add hipri option to sg engine
Date: Mon, 25 Jan 2021 15:26:51 -0500	[thread overview]
Message-ID: <20210125202651.346783-1-dgilbert@interlog.com> (raw)

Adds hipri option to the Linux sg driver engine. This turns on the
SGV4_FLAG_HIPRI flag in recent sg drivers (January 2021) on READ
and WRITE commands (and not on UNMAP (trim), VERIFY, etc). Uses
blk_poll() and the mq_poll() callback in SCSI LLDs. The mechanism
is also called "iopoll".

The Linux sg engine in fio uses the struct sg_io_hdr based interface
known as the sg driver "v3" interface.
Linux sg drivers in the kernel prior to January 2021 (sg version
4.0.12) will just ignore the SGV4_FLAG_HIPRI flag and do normal
completions where LLDs indicate command completion with a (software)
interrupt or similar mechanism.

Update fio.1 (manpage) with new hipri sg engine option.

Signed-off-by: Douglas Gilbert <dgilbert@interlog.com>
---
 engines/sg.c | 22 +++++++++++++++++++++-
 fio.1        | 10 ++++++++++
 2 files changed, 31 insertions(+), 1 deletion(-)

diff --git a/engines/sg.c b/engines/sg.c
index a1a6de4c..0c2d2c8b 100644
--- a/engines/sg.c
+++ b/engines/sg.c
@@ -60,6 +60,10 @@
 
 #ifdef FIO_HAVE_SGIO
 
+#ifndef SGV4_FLAG_HIPRI
+#define SGV4_FLAG_HIPRI 0x800
+#endif
+
 enum {
 	FIO_SG_WRITE		= 1,
 	FIO_SG_WRITE_VERIFY	= 2,
@@ -68,12 +72,22 @@ enum {
 
 struct sg_options {
 	void *pad;
+	unsigned int hipri;
 	unsigned int readfua;
 	unsigned int writefua;
 	unsigned int write_mode;
 };
 
 static struct fio_option options[] = {
+        {
+                .name   = "hipri",
+                .lname  = "High Priority",
+                .type   = FIO_OPT_STR_SET,
+                .off1   = offsetof(struct sg_options, hipri),
+                .help   = "Use polled IO completions",
+                .category = FIO_OPT_C_ENGINE,
+                .group  = FIO_OPT_G_SG,
+        },
 	{
 		.name	= "readfua",
 		.lname	= "sg engine read fua flag support",
@@ -527,6 +541,8 @@ static int fio_sgio_prep(struct thread_data *td, struct io_u *io_u)
 		else
 			hdr->cmdp[0] = 0x88; // read(16)
 
+		if (o->hipri)
+			hdr->flags |= SGV4_FLAG_HIPRI;
 		if (o->readfua)
 			hdr->cmdp[1] |= 0x08;
 
@@ -542,6 +558,8 @@ static int fio_sgio_prep(struct thread_data *td, struct io_u *io_u)
 				hdr->cmdp[0] = 0x2a; // write(10)
 			else
 				hdr->cmdp[0] = 0x8a; // write(16)
+			if (o->hipri)
+				hdr->flags |= SGV4_FLAG_HIPRI;
 			if (o->writefua)
 				hdr->cmdp[1] |= 0x08;
 			break;
@@ -865,6 +883,7 @@ static int fio_sgio_init(struct thread_data *td)
 {
 	struct sgio_data *sd;
 	struct sgio_trim *st;
+	struct sg_io_hdr *h3p;
 	int i;
 
 	sd = calloc(1, sizeof(*sd));
@@ -880,12 +899,13 @@ static int fio_sgio_init(struct thread_data *td)
 #ifdef FIO_SGIO_DEBUG
 	sd->trim_queue_map = calloc(td->o.iodepth, sizeof(int));
 #endif
-	for (i = 0; i < td->o.iodepth; i++) {
+	for (i = 0, h3p = sd->sgbuf; i < td->o.iodepth; i++, ++h3p) {
 		sd->trim_queues[i] = calloc(1, sizeof(struct sgio_trim));
 		st = sd->trim_queues[i];
 		st->unmap_param = calloc(td->o.iodepth + 1, sizeof(char[16]));
 		st->unmap_range_count = 0;
 		st->trim_io_us = calloc(td->o.iodepth, sizeof(struct io_u *));
+		h3p->interface_id = 'S';
 	}
 
 	td->io_ops_data = sd;
diff --git a/fio.1 b/fio.1
index d477b508..9636a85f 100644
--- a/fio.1
+++ b/fio.1
@@ -2114,6 +2114,16 @@ client and the server or in certain loopback configurations.
 Specify stat system call type to measure lookup/getattr performance.
 Default is \fBstat\fR for \fBstat\fR\|(2).
 .TP
+.BI (sg)hipri
+If this option is set, fio will attempt to use polled IO completions. This
+will have a similar effect as (io_uring)hipri. Only SCSI READ and WRITE
+commands will have the SGV4_FLAG_HIPRI set (not UNMAP (trim) nor VERIFY).
+Older versions of the Linux sg driver that do not support hipri will simply
+ignore this flag and do normal IO. The Linux SCSI Low Level Driver (LLD)
+that "owns" the device also needs to support hipri (also known as iopoll
+and mq_poll). The MegaRAID driver is an example of a SCSI LLD.
+Default: clear (0) which does normal (interrupted based) IO.
+.TP
 .BI (sg)readfua \fR=\fPbool
 With readfua option set to 1, read operations include the force
 unit access (fua) flag. Default: 0.
-- 
2.25.1


             reply	other threads:[~2021-01-25 20:28 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-01-25 20:26 Douglas Gilbert [this message]
2021-01-25 21:07 ` [PATCH v2] fio: add hipri option to sg engine Jens Axboe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210125202651.346783-1-dgilbert@interlog.com \
    --to=dgilbert@interlog.com \
    --cc=axboe@kernel.dk \
    --cc=hare@suse.de \
    --cc=jejb@linux.vnet.ibm.com \
    --cc=kashyap.desai@broadcom.com \
    --cc=linux-block@vger.kernel.org \
    --cc=linux-scsi@vger.kernel.org \
    --cc=martin.petersen@oracle.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox