Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <b414a7af-7333-498c-55ba-d60ae0a0507c@linux.intel.com>
Date:   Thu, 22 Jun 2023 10:40:21 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.11.0
Subject: Re: [PATCH v2] drm/i915: Replace kmap() with kmap_local_page()
Content-Language: en-US
To:     =?UTF-8?Q?Thomas_Hellstr=c3=b6m_=28Intel=29?= 
        <thomas_os@shipmail.org>, Ira Weiny <ira.weiny@intel.com>,
        Sumitra Sharma <sumitraartsy@gmail.com>
Cc:     Jani Nikula <jani.nikula@linux.intel.com>,
        Joonas Lahtinen <joonas.lahtinen@linux.intel.com>,
        Rodrigo Vivi <rodrigo.vivi@intel.com>,
        David Airlie <airlied@gmail.com>,
        Daniel Vetter <daniel@ffwll.ch>,
        intel-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org,
        linux-kernel@vger.kernel.org, Fabio <fmdefrancesco@gmail.com>,
        Deepak R Varma <drv@mailo.com>
References: <20230620180759.GA419158@sumitra.com>
 <79e1f37f-3ffa-0195-860b-08cc890d810e@shipmail.org>
 <649326df1b895_1c0dd29486@iweiny-mobl.notmuch>
 <d58c3dfd-e1d5-a8c5-7f0f-295dbf99e91a@shipmail.org>
From:   Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Organization: Intel Corporation UK Plc
In-Reply-To: <d58c3dfd-e1d5-a8c5-7f0f-295dbf99e91a@shipmail.org>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk


On 21/06/2023 19:51, Thomas Hellström (Intel) wrote:
> 
> On 6/21/23 18:35, Ira Weiny wrote:
>> Thomas Hellström (Intel) wrote:
>>> I think one thing worth mentioning in the context of this patch is that
>>> IIRC kmap_local_page() will block offlining of the mapping CPU until
>>> kunmap_local(), so while I haven't seen any guidelines around the usage
>>> of this api for long-held mappings, I figure it's wise to keep the
>>> mapping duration short, or at least avoid sleeping with a
>>> kmap_local_page() map active.
>>>
>>> I figured, while page compression is probably to be considered "slow"
>>> it's probably not slow enough to motivate kmap() instead of
>>> kmap_local_page(), but if anyone feels differently, perhaps it should be
>>> considered.
>> What you say is all true.  But remember the mappings are only actually
>> created on a HIGHMEM system.  HIGHMEM systems are increasingly rare.  
>> Also
>> they must suffer such performance issues because there is just no other
>> way around supporting them.
>>
>> Also Sumitra, and our kmap conversion project in general, is focusing on
>> not using kmap* if at all possible.  Thus the reason V1 tried to use
>> page_address().
>>
>> Could we guarantee the i915 driver is excluded from all HIGHMEM systems?
> 
> The i915 maintainers might want to chime in here, but I would say no, we 
> can't, although we don't care much about optimizing for them. Same for 
> the new xe driver.

AFAIK i915 works on such systems so I don't think we can drop support 
just like that. Not sure what the process would be. Perhaps as part of a 
wider kernel deprecation would make most sense.

Regards,

Tvrtko