Re: [PATCH v3 1/3] x86/ioreq server(patch for 4.7): Rename p2m_mmio_write_dm to p2m_ioreq_server.

From: "Yu, Zhang" <yu.c.zhang@linux.intel.com>
To: Paul Durrant <Paul.Durrant@citrix.com>,
	George Dunlap <George.Dunlap@citrix.com>
Cc: Kevin Tian <kevin.tian@intel.com>, Wei Liu <wei.liu2@citrix.com>,
	Jan Beulich <JBeulich@suse.com>,
	Andrew Cooper <Andrew.Cooper3@citrix.com>,
	"Tim (Xen.org)" <tim@xen.org>,
	"xen-devel@lists.xen.org" <xen-devel@lists.xen.org>,
	Zhiyuan Lv <zhiyuan.lv@intel.com>,
	Jun Nakajima <jun.nakajima@intel.com>,
	"Keir (Xen.org)" <keir@xen.org>
Subject: Re: [PATCH v3 1/3] x86/ioreq server(patch for 4.7): Rename p2m_mmio_write_dm to p2m_ioreq_server.
Date: Thu, 28 Apr 2016 10:47:21 +0800	[thread overview]
Message-ID: <027063a4-d263-40a6-1ac8-5bbb81b6b0fe@linux.intel.com> (raw)
In-Reply-To: <09057dd1763f4bf7975f307365e8be24@AMSPEX02CL03.citrite.net>

On 4/27/2016 10:42 PM, Paul Durrant wrote:
>> -----Original Message-----
>> From: George Dunlap
>> Sent: 27 April 2016 15:13
>> To: Paul Durrant
>> Cc: Yu, Zhang; Jan Beulich; Kevin Tian; Wei Liu; Andrew Cooper; Tim
>> (Xen.org); xen-devel@lists.xen.org; Zhiyuan Lv; Jun Nakajima; Keir (Xen.org)
>> Subject: Re: [Xen-devel] [PATCH v3 1/3] x86/ioreq server(patch for 4.7):
>> Rename p2m_mmio_write_dm to p2m_ioreq_server.
>>
>> On Mon, Apr 25, 2016 at 6:01 PM, Paul Durrant <Paul.Durrant@citrix.com>
>> wrote:
>>> For clarity, do you expect any existing use of HVMMEM_mmio_write_dm
>> to continue to *function*? I agree that things should continue to build, but if
>> they don't need to function then the now redundant p2m type should be
>> removed IMO and any attempt to set a page to HVMMEM_mmio_write_dm
>> (or the new HVMMEM_unused) name should result in -EINVAL. What is your
>> position on this?
>>
>> I sort of feel like we're playing some strange guessing game with the
>> color of this bike shed, where all 4 of us give a random combination
>> of constrants and then we have to figure out what the solution is. :-)
>>
>> There are two issues: the interface (HVMMEM_*) and the
>> internals.(p2m_*).
>>
>> Jan says that code that calls HVMOP_get_mem_type has to continue to
>> compile and function.  "Functioning" is easy, as you just don't return
>> that value and you're done.  Compiling just means having the #ifdef.
>>
>> It sounds like we all agree that HVMOP_set_mem_type with the current
>> HVMMEM_mmio_write_dm value should return -EINVAL.
>>
>> Regarding the p2m type which now should be impossible to set -- I
>> don't think it's critical to remove from the release, since it's just
>> internal.  I'd normally say just leave it for now to reduce code
>> churn.
>>
>> But mostly I think we just want to get this bike shed painted, so if
>> anyone thinks we should really remove the p2m type from this release,
>> then that's fine with me too (assuming it's OK with Wei).
>>
>> Does this cover everything?
>>
>
> I think so. Thanks for the clarification.
>
> Yu, are you happy to submit a patch with the #ifdef in the header, and that removes any ability to set the old type?
>

I'm fine with this, and thanks. :)
So my understanding is that the only difference between the new
patch and this current one is we do not replace p2m_mmio_write_dm
with p2m_ioreq_server, hence no need to introduce HVMMEM_ioreq_server.
Is this understanding correct?

Besides, do you think it acceptable we just replace p2m_mmio_write_dm
with p2m_ioreq_server in next version, or should we remove this p2m
type and define p2m_ioreq_server with a different value? :)

> I guess leaving the p2m type in place to avoid code churn is reasonable at this stage, but anyone looking at the p2m code is probably going to question why it's there in 4.7.
>
>   Paul
>
>>  -George

B.R.
Yu

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