Re: [PATCH v2] drm/i915: Replace kmap() with kmap_local_page()

[Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>]

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