From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andrew Cooper Subject: Re: RFC: EFER in crash notes Date: Tue, 25 Sep 2012 15:53:10 +0100 Message-ID: <5061C556.60406@citrix.com> References: <5061BD4E.6000905@citrix.com> <1348583626.11229.26.camel@zakaz.uk.xensource.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <1348583626.11229.26.camel@zakaz.uk.xensource.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Ian Campbell Cc: "Keir (Xen.org)" , Jan Beulich , "xen-devel@lists.xen.org" List-Id: xen-devel@lists.xenproject.org On 25/09/12 15:33, Ian Campbell wrote: > On Tue, 2012-09-25 at 15:18 +0100, Andrew Cooper wrote: >> While this patch is very simple, and I hope without any objection, it is >> RFC for the reason that our crash ABI is private. >> >> The comments for XEN_ELFNOTE_CRASH_REGS does state that it is >> architecture specific, and makes no indication about the size or >> contents of the crash note. However, any code trying to use one of >> these types of notes has to make an assumption that it if the note desc >> length is 4*8 bytes long, it is representing CR{0,2-4}. >> >> I guess my question boils down to whether it is acceptable to change a >> private ABI which is not really so private, or whether we should make a >> formal public ABI for all of the inards of the crash notes and use that. > Is it really private? (or even "not so private"), don't external tools > like crash support it? This is my point. It is not in xen/public but is used by crash/kdump and similar utilities, effectively making it a public ABI. include/public/elfnote.h even explicitly refers to include/xen/elfcore.h, which is not public by our definition. > > I suppose the size field in the notes is a sort of rudimentary version > field. Remember you can always add a new note type though. Yes, although looking through my code, I do raise an error if sizeof(note->desc) != sizeof(my structure representing this note), which was put in with the best of intentions, but will break with the this RFC change. On the other hand, adding a new crash note for every new register will not scale well, as it is per PCU. I guess the only sensible way to continue is to present a formal public ABI. > > Ian. > > -- Andrew Cooper - Dom0 Kernel Engineer, Citrix XenServer T: +44 (0)1223 225 900, http://www.citrix.com