From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jan Kiszka Subject: Re: error while loading state for instance 0x0 of device 'kvmclock' Date: Mon, 02 Nov 2009 16:37:15 +0100 Message-ID: <4AEEFCAB.2070308@siemens.com> References: <4AEE99F5.5000509@web.de> <4AEEB456.1000006@siemens.com> <4AEEE04C.8080104@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: 7bit Cc: Marcelo Tosatti , Glauber Costa , kvm To: Avi Kivity Return-path: Received: from goliath.siemens.de ([192.35.17.28]:18121 "EHLO goliath.siemens.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755378AbZKBPhY (ORCPT ); Mon, 2 Nov 2009 10:37:24 -0500 In-Reply-To: <4AEEE04C.8080104@redhat.com> Sender: kvm-owner@vger.kernel.org List-ID: Avi Kivity wrote: > On 11/02/2009 12:28 PM, Jan Kiszka wrote: >> Jan Kiszka wrote: >> >>> Hi, >>> >>> current qemu-kvm.git gives me the message "qemu: warning: error while >>> loading state for instance 0x0 of device 'kvmclock'" when I run a simple >>> "savevm" followed by a "loadvm 1". What's broken here? >>> >> OK, this is due to "KVM: add flags to kvm_clock_data" (958b0c5497): the >> flags field is not cleared on KVM_SET_CLOCK. Will post a fix. >> >> But the above kernel commit is also broken: KVM_GET_CLOCK checks >> uninitialized user_ns.flags (probably instead of the user's value). This >> raises the question if the caller of KVM_GET_CLOCK is also supposed to >> pass kvm_clock_data with flags cleared down to the kernel. Could someone >> clarify this so I could fix it accordingly? >> > > I'd make KVM_GET_CLOCK set the flags, not get them. So if we add new > fields, we just set a new bit and userspace can read it. This makes sense and actually fixes the issue completely. Patch on its way... Jan -- Siemens AG, Corporate Technology, CT SE 2 Corporate Competence Center Embedded Linux