From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([140.186.70.92]:44275) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1QZ21h-00078L-A1 for qemu-devel@nongnu.org; Tue, 21 Jun 2011 10:34:02 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1QZ21f-0004fe-Ho for qemu-devel@nongnu.org; Tue, 21 Jun 2011 10:34:00 -0400 Received: from mail-gy0-f173.google.com ([209.85.160.173]:59642) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1QZ21f-0004fX-4D for qemu-devel@nongnu.org; Tue, 21 Jun 2011 10:33:59 -0400 Received: by gyg10 with SMTP id 10so1301375gyg.4 for ; Tue, 21 Jun 2011 07:33:58 -0700 (PDT) MIME-Version: 1.0 Date: Tue, 21 Jun 2011 15:33:57 +0100 Message-ID: From: Stefan Hajnoczi Content-Type: text/plain; charset=ISO-8859-1 Subject: [Qemu-devel] cirrus logic pathological case slow screen update List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel Hi, I've just seen a system running Windows 2008 64-bit under KVM doing a Windows Update. It was applying 1000s of updates in a text mode that triggers a pathological case with the cirrus logic emulation (I connected over VNC). The symptom is that the message being displayed "Applying update 1 out of 11666" is redrawn at an extremely slow speed - maybe one second to render the line of text. You can see portions of letters being blitted to the screen. The virtual machine appears blocked by this screen redraw so the updates apply *very* slowly. When -vga is changed from cirrus to std the issue goes away and the update proceeds at normal speed. I was given temporary VNC access to the problem VM so I don't know all the configuration details, but I did see the issue and that it was fixed by -vga cirrus -> -vga std. Any ideas? If it happens again perhaps the KVM stats will tell us that it is emulating framebuffer memory accesses or something very expensive. I didn't get a chance to try but would this be possible with cirrus and not std? Stefan