All of lore.kernel.org
 help / color / mirror / Atom feed
From: Saul Tamari <stamari@gmail.com>
To: Anthony Liguori <anthony@codemonkey.ws>
Cc: qemu-devel@nongnu.org, kvm@vger.kernel.org
Subject: Re: [PATCH] qemu-kvm: clear only essential parts of VirtIOBlockReq on object allocation - RESUBMIT
Date: Wed, 18 Nov 2009 16:41:22 +0200	[thread overview]
Message-ID: <20091118144122.GC19099@localhost.localdomain> (raw)
In-Reply-To: <4B02FBE4.2060409@codemonkey.ws>

[-- Attachment #1: patch0.patch --]
[-- Type: text/plain, Size: 1797 bytes --]

ok,
Below is the patch. This time resent with mutt and hopefully without whitespace changes.
--------------------------------------------------------------------

This patch reduces the amount of memory being cleared on every virtio-blk IO operation.

Improve number of IOPS when using avirtio-blk device.

On every virtio-blk IO command passed to QEMU, virtio_blk_alloc_request()
allocates and clears (with qemu_mallocz()) a VirtIOBlockReq object.
The sizeof(VirtIOBlockReq) equals 41040 bytes on my x86-64 machine.
By moving the 'elem' variable to the end of VirtIOBlockReq and
clearing only upto the address of the 'elem.in_addr' field, the
memset() call now clears only 80 bytes.


Signed-off-by: Saul Tamari <stamari@gmail.com>
-------------------------------------------------
diff --git a/hw/virtio-blk.c b/hw/virtio-blk.c
index 2630b99..de74b00 100644
--- a/hw/virtio-blk.c
+++ b/hw/virtio-blk.c
@@ -79,12 +79,13 @@ static inline void virtio_identify_template(struct virtio_blk_config *bc)
 typedef struct VirtIOBlockReq
 {
     VirtIOBlock *dev;
-    VirtQueueElement elem;
     struct virtio_blk_inhdr *in;
     struct virtio_blk_outhdr *out;
     struct virtio_scsi_inhdr *scsi;
     QEMUIOVector qiov;
     struct VirtIOBlockReq *next;
+    /* Members that need clearing, must be added prior to elem */
+    VirtQueueElement elem;
 } VirtIOBlockReq;
 
 static void virtio_blk_req_complete(VirtIOBlockReq *req, int status)
@@ -139,7 +140,8 @@ static void virtio_blk_flush_complete(void *opaque, int ret)
 
 static VirtIOBlockReq *virtio_blk_alloc_request(VirtIOBlock *s)
 {
-    VirtIOBlockReq *req = qemu_mallocz(sizeof(*req));
+    VirtIOBlockReq *req = qemu_malloc(sizeof(*req));
+    memset(req, 0, offsetof(VirtIOBlockReq, elem.in_addr[0]));
     req->dev = s;
     return req;
 }

WARNING: multiple messages have this Message-ID (diff)
From: Saul Tamari <stamari@gmail.com>
To: Anthony Liguori <anthony@codemonkey.ws>
Cc: qemu-devel@nongnu.org, kvm@vger.kernel.org
Subject: [Qemu-devel] Re: [PATCH] qemu-kvm: clear only essential parts of VirtIOBlockReq on object allocation - RESUBMIT
Date: Wed, 18 Nov 2009 16:41:22 +0200	[thread overview]
Message-ID: <20091118144122.GC19099@localhost.localdomain> (raw)
In-Reply-To: <4B02FBE4.2060409@codemonkey.ws>

[-- Attachment #1: patch0.patch --]
[-- Type: text/plain, Size: 1797 bytes --]

ok,
Below is the patch. This time resent with mutt and hopefully without whitespace changes.
--------------------------------------------------------------------

This patch reduces the amount of memory being cleared on every virtio-blk IO operation.

Improve number of IOPS when using avirtio-blk device.

On every virtio-blk IO command passed to QEMU, virtio_blk_alloc_request()
allocates and clears (with qemu_mallocz()) a VirtIOBlockReq object.
The sizeof(VirtIOBlockReq) equals 41040 bytes on my x86-64 machine.
By moving the 'elem' variable to the end of VirtIOBlockReq and
clearing only upto the address of the 'elem.in_addr' field, the
memset() call now clears only 80 bytes.


Signed-off-by: Saul Tamari <stamari@gmail.com>
-------------------------------------------------
diff --git a/hw/virtio-blk.c b/hw/virtio-blk.c
index 2630b99..de74b00 100644
--- a/hw/virtio-blk.c
+++ b/hw/virtio-blk.c
@@ -79,12 +79,13 @@ static inline void virtio_identify_template(struct virtio_blk_config *bc)
 typedef struct VirtIOBlockReq
 {
     VirtIOBlock *dev;
-    VirtQueueElement elem;
     struct virtio_blk_inhdr *in;
     struct virtio_blk_outhdr *out;
     struct virtio_scsi_inhdr *scsi;
     QEMUIOVector qiov;
     struct VirtIOBlockReq *next;
+    /* Members that need clearing, must be added prior to elem */
+    VirtQueueElement elem;
 } VirtIOBlockReq;
 
 static void virtio_blk_req_complete(VirtIOBlockReq *req, int status)
@@ -139,7 +140,8 @@ static void virtio_blk_flush_complete(void *opaque, int ret)
 
 static VirtIOBlockReq *virtio_blk_alloc_request(VirtIOBlock *s)
 {
-    VirtIOBlockReq *req = qemu_mallocz(sizeof(*req));
+    VirtIOBlockReq *req = qemu_malloc(sizeof(*req));
+    memset(req, 0, offsetof(VirtIOBlockReq, elem.in_addr[0]));
     req->dev = s;
     return req;
 }

  reply	other threads:[~2009-11-18 14:40 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-11-09 18:04 [PATCH] qemu-kvm: clear only essential parts of VirtIOBlockReq on object allocation - RESUBMIT Saul Tamari
2009-11-09 18:04 ` [Qemu-devel] " Saul Tamari
2009-11-17 19:39 ` Anthony Liguori
2009-11-17 19:39   ` [Qemu-devel] " Anthony Liguori
2009-11-18 14:41   ` Saul Tamari [this message]
2009-11-18 14:41     ` Saul Tamari

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=20091118144122.GC19099@localhost.localdomain \
    --to=stamari@gmail.com \
    --cc=anthony@codemonkey.ws \
    --cc=kvm@vger.kernel.org \
    --cc=qemu-devel@nongnu.org \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.