[Devel] [RFC PATCH vz9 v6 15/62] dm-ploop: fast path execution for reads

[Alexander Atanasov]

On 13.12.24 14:48, Andrey Zhadchenko wrote:
> Honestly I do not see any benefit in changes within 
> ploop_process_one_deferred_bio().
> For now the fast path is supposed to work only with reads. Reads can be 
> either zero, top or some bottom delta.
> I think implementing small function to handle these cases is better than 
> adding many returns which no one uses

The idea of the fast path was to reduce latency by skipping threads but
it did not work due to rcu issues.
In later patches fast path and delayed writeback are disabled since they 
are not working properly - we probaly should decide what to do with 
them. i do not see a huge benefit from fast path reads /write we can not/.


Not on the particular case - I think it is better to have returns that 
are a bit more specific rather than void functions in a critical code.

Anyway if we decide to keep fast path then the code may undergoo an 
reorganisation.



>

-- 
Regards,
Alexander Atanasov