Re: [PATCH/RFC] Simplified Readahead

[Steven Pratt <slpratt@austin.ibm.com>]

Nick Piggin wrote:

> Steven Pratt wrote:
>
>> Andrew Morton wrote:
>>
>>> Steven Pratt <slpratt@austin.ibm.com> wrote:
>>>  
>>>
>>>> would like to offer up an alternative simplified design which will 
>>>> not only make the code easier to maintain,
>>>>   
>>>
>>>
>>> We won't know that until all functionality is in place.
>>>  
>>
>> Ok, but both you and Nick indicated that the queue congestion isn't 
>> needed,
>
> I would have thought that always doing the readahead would provide a
> more graceful degradation, assuming the readahead algorithm is fairly
> accurate, and copes with things like readahead thrashing (which we
> hope is the case).

Yes, that is exactly my thought.  I think this is what the new code does.

>
>>> I do think we should skip the I/O for POSIX_FADV_WILLNEED against a
>>> congested queue.  I can't immediately think of a good reason for 
>>> skipping
>>> the I/O for normal readahead.
>>>  
>>
>
> I don't see why you should skip the readahead for FADVISE_WILLNEED
> either. Presumably if someone needs this, they really need it. We
> should aim for optimal behaviour when the apis are being used 
> correctly... 

Ok, great, since this is what it does.

Thanks, Steve