From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mailman by lists.gnu.org with tmda-scanned (Exim 4.43) id 1HWbyo-0004Jk-Ax for qemu-devel@nongnu.org; Wed, 28 Mar 2007 13:30:38 -0400 Received: from exim by lists.gnu.org with spam-scanned (Exim 4.43) id 1HWbym-0004JY-TP for qemu-devel@nongnu.org; Wed, 28 Mar 2007 13:30:38 -0400 Received: from [199.232.76.173] (helo=monty-python.gnu.org) by lists.gnu.org with esmtp (Exim 4.43) id 1HWbym-0004JV-O9 for qemu-devel@nongnu.org; Wed, 28 Mar 2007 12:30:36 -0500 Received: from wx-out-0506.google.com ([66.249.82.236]) by monty-python.gnu.org with esmtp (Exim 4.60) (envelope-from ) id 1HWbwH-0005Qe-Nk for qemu-devel@nongnu.org; Wed, 28 Mar 2007 13:28:01 -0400 Received: by wx-out-0506.google.com with SMTP id i30so4285730wxd for ; Wed, 28 Mar 2007 10:28:00 -0700 (PDT) Message-ID: <9a158e2e0703281028s246e7bf4m64961d5495864271@mail.gmail.com> Date: Wed, 28 Mar 2007 10:28:00 -0700 From: "Kyle Hubert" Subject: Re: [Qemu-devel] virtual machine halting and loosing connections In-Reply-To: <9b0d5f320703280721ga4ae040ydfed5a4a2e62bf74@mail.gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Content-Disposition: inline References: <9a158e2e0703271634x3b806b7ahe070aae2ea8a9631@mail.gmail.com> <9b0d5f320703280721ga4ae040ydfed5a4a2e62bf74@mail.gmail.com> Reply-To: qemu-devel@nongnu.org List-Id: qemu-devel.nongnu.org List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org On 3/28/07, WaxDragon wrote: > On 3/27/07, Kyle Hubert wrote: > > > Also, I have noticed that when using the ondemand governor for the > > host OS (AMD PowerNow) that QEMU doesn't register for some reason. My > > host CPU stays running at 1GHz, when it is capable of 2.6GHz. I can > > run anything on the host OS, and I see the cpu freq jump up, so it > > works on the host OS. I believe this may have something to do with > > kqemu's acceleration. > > I have seen the same behavior with cpufreq, but I run the > 'conservative' governor on my AMD. Haven't decided if it's a bug or a > feature yet. ;0) I was running ondemand, and I came to the conclusion it's a bug. If you see this paper: http://linux.inet.hr/ols2006_the_ondemand_governor.html You'll see that the ondemand governor policy is based on the idle/busy statistics from scheduler ticks. I believe that kqemu runs off of ioctl's from /dev/kqemu, and perhaps the system call doesn't accurately change this? How do the processes run in kqemu? Is it a kernel thread? Section 7.3 in that PDF talks of real time threads affecting the ondemand scheduler. NOTE: conservative is a fork of ondemand, so it applies here. I'll try running without kqemu, and see if I still have the problem. -Kyle