From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([140.186.70.92]:50616) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Rkvr0-0004E8-5P for qemu-devel@nongnu.org; Wed, 11 Jan 2012 05:56:38 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1Rkvqy-0001s3-Rr for qemu-devel@nongnu.org; Wed, 11 Jan 2012 05:56:25 -0500 Received: from mail-qy0-f173.google.com ([209.85.216.173]:51034) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Rkvqy-0001ro-Ph for qemu-devel@nongnu.org; Wed, 11 Jan 2012 05:56:24 -0500 Received: by qcsd15 with SMTP id d15so309754qcs.4 for ; Wed, 11 Jan 2012 02:56:23 -0800 (PST) MIME-Version: 1.0 In-Reply-To: <4F0D0C3E.8080706@suse.de> References: <1326213943-878-1-git-send-email-mark.langsdorf@calxeda.com> <1326227599-5257-1-git-send-email-mark.langsdorf@calxeda.com> <1326227599-5257-3-git-send-email-mark.langsdorf@calxeda.com> <4F0D0C3E.8080706@suse.de> Date: Wed, 11 Jan 2012 10:56:23 +0000 Message-ID: From: Peter Maydell Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Subject: Re: [Qemu-devel] [PATCH v7 2/6] arm: make the number of GIC interrupts configurable List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: =?UTF-8?Q?Andreas_F=C3=A4rber?= Cc: i.mitsyanko@gmail.com, edgar.iglesias@gmail.com, qemu-devel@nongnu.org, Mark Langsdorf On 11 January 2012 04:12, Andreas F=C3=A4rber wrote: > Am 10.01.2012 21:33, schrieb Mark Langsdorf: >> =C2=A0 =C2=A0 =C2=A0vmstate_register(&dev->qdev, -1, &vmstate_nvic, s); > > This should probably be removed now that you register it through > sysbus_register_withprop() below. Oops, yes, missed that bit. > I'm still wondering whether the num-irq property calls for a version > bump somewhere. My thinking is no, since the SysBus device will not be > user-created from the command line and therefore effectively has the > same values as hardcoded before; but in that case we should assert this > by setting .qdev.no_user =3D 1. I agree that we should probably set no_user=3D1 on more of the ARM internal devices than we currently do, but I think that's probably a separate issue rather than one we should deal with in this patch. Re version-bump: my thinking was that one isn't required because the onus is on the user not to migrate between systems where the devices have had different qdev property values set. So as long as "no properties set to non default values" doesn't change then we're ok whether the user put the device on the command line or not. (In practice (a) any config with a user-created NVIC device is always broken because there's no way for the user to wire it up properly, and (b) another patch in this series is bumping the CPU state version number so migration is going to fail anyhow.) -- PMM