Re: [Xen-devel] [PATCH RFC] stubdom: Change vTPM shared page ABI

["Fioravante, Matthew E." <Matthew.Fioravante@xxxxxxxxxx>]

________________________________________
From: Daniel De Graaf [dgdegra@xxxxxxxxxxxxx]
Sent: Friday, November 23, 2012 11:15 AM
To: Ian Campbell
Cc: Fioravante, Matthew E.; xen-devel@xxxxxxxxxxxxx
Subject: Re: [PATCH RFC] stubdom: Change vTPM shared page ABI

On 11/23/2012 06:04 AM, Ian Campbell wrote:
> On Tue, 2012-11-20 at 16:16 +0000, Fioravante, Matthew E. wrote:
>> This was something I've been wanting to do for a while so I'm very
>> happy you went ahead and fixed it yourself. Once you submit the linux
>> version I'll test it on my system.
>>
>> Acked by: Matthew Fioravante <matthew.fioravante@xxxxxxxxxx>
>
> Do we need to synchronise with anything else (Linux drivers?) or I can I
> just throw this into the tree once it is done?

I think the only Linux drivers are the ones that were just posted, and the
ones for XenLinux 2.6.18. Assuming the modified Linux module that I posted
works as it should, the v2 version will be the only one upstream.

> Are we hanging any existing users of the tpm stuff out to dry? Could we
> get a feature-protocol-v2 flag in xenstore so users at least get an
> error message from either the front or backend to give a hint why it
> isn't working?

I suppose we could do that, although I'm not sure the "feature-" prefix is
appropriate, since the code doesn't have the ability to fall back to v1.

I'd rather just let existing users hang out to dry. First off there aren't 
really 
any existing users. Anyone who is or might be a user is probably following 
these threads closely. Second the xen tpm driver hasn't even been 
accepted into linux yet so there really is not any "official" version 1.

I can update the documentation to state the requirements on the new
tpm driver version so that nobody tries to use the old version in the
xen 2.6.18 tree.

>>  void send_response(tpmcmd_t* cmd, tpmif_t* tpmif)  {
>> -   tpmif_tx_request_t* tx;
>> -   int offset;
>> -   int i;
>> -   uint32_t domid;
>> -   int tocopy;
>> +   vtpm_shared_page_t* shr;
>> +   unsigned int offset;
>>     int flags;
>> +#ifdef TPMBACK_PRINT_DEBUG
>> +int i;
>> +#endif
>
> Dodgy indentation? (I saw a bunch of these)

Whoops, I'll have to go fix that. The entire mini-os build is sprinkled with
tabs and spaces mostly at random, so perhaps a generic fix-indent patch is
needed after all the real changes are applied.

>> - * Copyright (c) 2005, IBM Corporation
>> - *
>> - * Author: Stefan Berger, stefanb@xxxxxxxxxx
>> - * Grant table support: Mahadevan Gomathisankaran
>
> Did you really remove everything these guys wrote?

>From this file, yes - unless the #ifndef and header/footer comments count, but
I was under the impression that that is too small to count for copyright, and
is duplicated across the other files in that directory anyway.

>> - * This code has been derived from tools/libxc/xen/io/netif.h
>> - *
>> - * Copyright (c) 2003-2004, Keir Fraser
>
> I guess I can buy this bit no longer being true given that you reworked
> the whole protocol.
>
> Ian

Basically, the file is a rewrite, although not detected by git/hg as such. I can
add an authorship line if that would be helpful (I see that's normal in other
header files).

--
Daniel De Graaf
National Security Agency

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel