Re: [RFC/PATCH 2/5] mm/fs: execute in place (V2)

[Carsten Otte <cotte@freenet.de>]

Christoph Hellwig wrote:

>On Wed, May 18, 2005 at 05:31:13PM +0200, Carsten Otte wrote:
>  
>
>>- generic_file_read           => xip_file_read
>>    
>>
>
>no need to have that one if you implement aio_read -> use do_sync_read
>
>  
>
>>- generic_file_aio_read    => xip_file_aio_read
>>- __generic_file_aio_read => __xip_file_aio_read
>>    
>>
>
>readv and aio_read are just wrappers around this one.
>
>  
>
>>- generic_file_sendfile     => xip_file_sendfile
>>    
>>
>
>pretty trivial
>
>  
>
>>- generic file_readv          => xip_file_readv
>>- generic_file_write          => xip_file_write
>>    
>>
>
>just use do_sync_write
>
>  
>
>>- generic_file_aio_write_nolock => xip_file_write_nolock
>>- __generic_file_write_nolock => __xip_file_write_nolock
>>- generic_file_write_nolock => xip_file_write_nolock
>>- generic_file_aio_write => xip_file_aio_write
>>    
>>
>
>you don't need all these.  Just writev and aio_write as wrappers around a common one
>
>  
>
>>- generic_file_mmap => xip_file_mmap
>>    
>>
>
>this one doesn't share code anyway
>
>  
>
>>- generic_file_readonly_mmap => xip_file_readonly_mmap
>>    
>>
>
>unless you want to implement a readonly filesystem with xip support you
>don't need this one.
>
>  
>
I agree that sync/async is not too much of a difference when you do a memcpy
behind, so you can just have wrappers. I am still not convinced that it
will stay
reasonably small with all that duplicated stuff, but since it's easy to
do I just
gonna give it a try to see how it'll look alike. Bet the patch size will
double.