From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from mailman by lists.gnu.org with tmda-scanned (Exim 4.43)
	id 1G2BCk-0007qi-TF
	for qemu-devel@nongnu.org; Sun, 16 Jul 2006 14:18:58 -0400
Received: from exim by lists.gnu.org with spam-scanned (Exim 4.43)
	id 1G2BCk-0007qF-4D
	for qemu-devel@nongnu.org; Sun, 16 Jul 2006 14:18:58 -0400
Received: from [199.232.76.173] (helo=monty-python.gnu.org)
	by lists.gnu.org with esmtp (Exim 4.43) id 1G2BCj-0007q9-PI
	for qemu-devel@nongnu.org; Sun, 16 Jul 2006 14:18:57 -0400
Received: from [70.116.9.243] (helo=localhost.localdomain)
	by monty-python.gnu.org with esmtps (TLS-1.0:RSA_AES_256_CBC_SHA:32)
	(Exim 4.52) id 1G2BFD-0007K7-Ce
	for qemu-devel@nongnu.org; Sun, 16 Jul 2006 14:21:31 -0400
Received: from localhost ([127.0.0.1])
	by localhost.localdomain with esmtp (Exim 4.60)
	(envelope-from <anthony@codemonkey.ws>) id 1G2B8U-0001Ns-VW
	for qemu-devel@nongnu.org; Sun, 16 Jul 2006 13:14:35 -0500
Message-ID: <44BA8205.7020304@codemonkey.ws>
Date: Sun, 16 Jul 2006 13:14:29 -0500
From: Anthony Liguori <anthony@codemonkey.ws>
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="------------030100010102030008050907"
Subject: [Qemu-devel] [PATCH 3/4] Improve character device read performance
Reply-To: qemu-devel@nongnu.org
List-Id: qemu-devel.nongnu.org
List-Unsubscribe: <http://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.gnu.org/pipermail/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <http://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: qemu-devel@nongnu.org

This is a multi-part message in MIME format.
--------------030100010102030008050907
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

Regards,

Anthony Liguori

--------------030100010102030008050907
Content-Type: text/x-patch;
 name="chrdev_partial_read.diff"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
 filename="chrdev_partial_read.diff"

# HG changeset patch
# User anthony@localhost.localdomain
# Node ID 32d5f77cf5e8c5a15b38b1fa86f588f374ebd3ad
# Parent  0e52f303222037247747558d43d72b5b89817828
Sometimes partial reads occur even though there's still data to be read.  This
patch makes sure that we read all the data that's available to read.

This has a huge performance impact for the shmem character device since.

diff -r 0e52f3032220 -r 32d5f77cf5e8 vl.c
--- a/vl.c	Sun Jul 16 17:04:01 2006
+++ b/vl.c	Sun Jul 16 17:23:21 2006
@@ -1205,6 +1205,29 @@
     return s->max_size;
 }
 
+static int fd_read_blocks(int fd)
+{
+    fd_set rfds;
+    struct timeval tv;
+    int ret;
+
+ again:
+    FD_ZERO(&rfds);
+    FD_SET(fd, &rfds);
+
+    tv.tv_sec = 0;
+    tv.tv_usec = 0;
+
+    ret = select(fd + 1, &rfds, NULL, NULL, &tv);
+    if (ret == -1) {
+	if (errno == EINTR)
+	    goto again;
+	return -1;
+    }
+
+    return !FD_ISSET(fd, &rfds);
+}
+
 static void fd_chr_read(void *opaque)
 {
     CharDriverState *chr = opaque;
@@ -1217,7 +1240,25 @@
         len = s->max_size;
     if (len == 0)
         return;
-    size = read(s->fd_in, buf, len);
+
+    for (size = 0; size < len;) {
+	int l;
+	l = read(s->fd_in, buf + size, len - size);
+	if (l == -1) {
+	    if (errno == EINTR)
+		continue;
+	    break;
+	} else if (l == 0) {
+	    break;
+	}
+
+	size += l;
+
+	if (size < len && fd_read_blocks(s->fd_in))
+	    break;
+    }
+
+
     if (size > 0) {
         s->fd_read(s->fd_opaque, buf, size);
     }

--------------030100010102030008050907--