Blame SOURCES/kvm-ui-track-how-much-decoded-data-we-consumed-when-doin.patch

5d360b
From 3be810287878138e5f72568d1ba1160b5bad22f8 Mon Sep 17 00:00:00 2001
5d360b
From: "Daniel P. Berrange" <berrange@redhat.com>
5d360b
Date: Thu, 8 Feb 2018 17:50:32 +0100
5d360b
Subject: [PATCH 18/27] ui: track how much decoded data we consumed when doing
5d360b
 SASL encoding
5d360b
MIME-Version: 1.0
5d360b
Content-Type: text/plain; charset=UTF-8
5d360b
Content-Transfer-Encoding: 8bit
5d360b
5d360b
RH-Author: Daniel P. Berrange <berrange@redhat.com>
5d360b
Message-id: <20180208175041.5634-19-berrange@redhat.com>
5d360b
Patchwork-id: 78951
5d360b
O-Subject: [RHEL-7.5 qemu-kvm PATCH v1 18/27] ui: track how much decoded data we consumed when doing SASL encoding
5d360b
Bugzilla: 1527405
5d360b
RH-Acked-by: Laszlo Ersek <lersek@redhat.com>
5d360b
RH-Acked-by: Gerd Hoffmann <kraxel@redhat.com>
5d360b
RH-Acked-by: Miroslav Rezanina <mrezanin@redhat.com>
5d360b
5d360b
From: "Daniel P. Berrange" <berrange@redhat.com>
5d360b
5d360b
When we encode data for writing with SASL, we encode the entire pending output
5d360b
buffer. The subsequent write, however, may not be able to send the full encoded
5d360b
data in one go though, particularly with a slow network. So we delay setting the
5d360b
output buffer offset back to zero until all the SASL encoded data is sent.
5d360b
5d360b
Between encoding the data and completing sending of the SASL encoded data,
5d360b
however, more data might have been placed on the pending output buffer. So it
5d360b
is not valid to set offset back to zero. Instead we must keep track of how much
5d360b
data we consumed during encoding and subtract only that amount.
5d360b
5d360b
With the current bug we would be throwing away some pending data without having
5d360b
sent it at all. By sheer luck this did not previously cause any serious problem
5d360b
because appending data to the send buffer is always an atomic action, so we
5d360b
only ever throw away complete RFB protocol messages. In the case of frame buffer
5d360b
updates we'd catch up fairly quickly, so no obvious problem was visible.
5d360b
5d360b
RHEL-7 note: context difference in the last argument to
5d360b
vnc_client_io_error() due to downstream lacking commit 04d2529da27d
5d360b
("ui: convert VNC server to use QIOChannelSocket", 2015-12-18).
5d360b
5d360b
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
5d360b
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
5d360b
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
5d360b
Message-id: 20171218191228.31018-6-berrange@redhat.com
5d360b
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
5d360b
(cherry picked from commit 8f61f1c5a6bc06438a1172efa80bc7606594fa07)
5d360b
Signed-off-by: Miroslav Rezanina <mrezanin@redhat.com>
5d360b
---
5d360b
 ui/vnc-auth-sasl.c | 3 ++-
5d360b
 ui/vnc-auth-sasl.h | 1 +
5d360b
 2 files changed, 3 insertions(+), 1 deletion(-)
5d360b
5d360b
diff --git a/ui/vnc-auth-sasl.c b/ui/vnc-auth-sasl.c
5d360b
index f3ad75d..804b8e7 100644
5d360b
--- a/ui/vnc-auth-sasl.c
5d360b
+++ b/ui/vnc-auth-sasl.c
5d360b
@@ -64,6 +64,7 @@ long vnc_client_write_sasl(VncState *vs)
5d360b
         if (err != SASL_OK)
5d360b
             return vnc_client_io_error(vs, -1, EIO);
5d360b
 
5d360b
+        vs->sasl.encodedRawLength = vs->output.offset;
5d360b
         vs->sasl.encodedOffset = 0;
5d360b
     }
5d360b
 
5d360b
@@ -75,7 +76,7 @@ long vnc_client_write_sasl(VncState *vs)
5d360b
 
5d360b
     vs->sasl.encodedOffset += ret;
5d360b
     if (vs->sasl.encodedOffset == vs->sasl.encodedLength) {
5d360b
-        vs->output.offset = 0;
5d360b
+        vs->output.offset -= vs->sasl.encodedRawLength;
5d360b
         vs->sasl.encoded = NULL;
5d360b
         vs->sasl.encodedOffset = vs->sasl.encodedLength = 0;
5d360b
     }
5d360b
diff --git a/ui/vnc-auth-sasl.h b/ui/vnc-auth-sasl.h
5d360b
index 8091d68..4ec6fb5 100644
5d360b
--- a/ui/vnc-auth-sasl.h
5d360b
+++ b/ui/vnc-auth-sasl.h
5d360b
@@ -54,6 +54,7 @@ struct VncStateSASL {
5d360b
      */
5d360b
     const uint8_t *encoded;
5d360b
     unsigned int encodedLength;
5d360b
+    unsigned int encodedRawLength;
5d360b
     unsigned int encodedOffset;
5d360b
     char *username;
5d360b
     char *mechlist;
5d360b
-- 
5d360b
1.8.3.1
5d360b