[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <56E6BE7E.8010807@amd.com>
Date: Mon, 14 Mar 2016 20:37:02 +0700
From: Suravee Suthikulpanit <Suravee.Suthikulpanit@....com>
To: Peter Zijlstra <peterz@...radead.org>
CC: <joro@...tes.org>, <bp@...en8.de>, <mingo@...hat.com>,
<acme@...nel.org>, <andihartmann@...enet.de>, <vw@...mu.org>,
<labbott@...hat.com>, <linux-kernel@...r.kernel.org>,
<iommu@...ts.linux-foundation.org>
Subject: Re: [PATCH V5 02/10] perf/amd/iommu: Consolidate and move
perf_event_amd_iommu header
Hi,
On 3/14/16 16:58, Peter Zijlstra wrote:
> On Mon, Mar 14, 2016 at 12:26:00PM +0700, Suravee Suthikulpanit wrote:
>> Hi,
>>
>> On 03/12/2016 08:22 PM, Peter Zijlstra wrote:
>>> On Tue, Feb 23, 2016 at 08:12:36AM -0600, Suravee Suthikulpanit wrote:
>>>> From: Suravee Suthikulpanit <suravee.suthikulpanit@....com>
>>>>
>>>> First, this patch move arch/x86/events/amd/iommu.h to
>>>> arch/x86/include/asm/perf/amd/iommu.h so that we easily include
>>>> it in both perf-amd-iommu and amd-iommu drivers.
>>>>
>>>> Then, we consolidate declaration of AMD IOMMU performance counter
>>>> APIs into one file.
>>>
>>> These seem two independent thingies; should this therefore not be 2
>>> patches?
>>>
>>>> Reviewed-by: Joerg Roedel <jroedel@...e.de>
>>>> Signed-off-by: Suravee Suthikulpanit <Suravee.Suthikulpanit@....com>
>>>> ---
>>>> arch/x86/events/amd/iommu.c | 2 +-
>>>> arch/x86/events/amd/iommu.h | 40 ---------------------------------
>>>> arch/x86/include/asm/perf/amd/iommu.h | 42 +++++++++++++++++++++++++++++++++++
>>>
>>> That seems somewhat excessive. Not only do you create
>>> arch/x86/include/asm/perf/ you then put another directory on top of
>>> that.
>>>
>>
>> The original header files (arch/x86/events/amd/iommu.h and
>> drivers/iommu/amd_iommu_proto.h) has duplicate function declarations. So,
>> with the new header file being in the arch/x86/include/asm/perf/amd/iommu.h,
>> we can just have one function declaration.
>>
>> So, you just want to separate the file moving part and the part that removes
>> of the duplication?
>
> I'm fine with a new header, it just seems putting it in a two deep
> direcotry hierarchy of its own that seems excessive.
>
Basically, we are trying to match the current Perf hierarchy for AMD
IOMMU (arch/x86/events/amd/iommu.c). I can put it into
arch/x86/include/asm/perf_amd_iommu.h. What would you prefer?
Thanks,
Suravee
Powered by blists - more mailing lists