SMB2/3 Performance

[Stefan (metze) Metzmacher]

Hi David,

> On Fri, 22 Nov 2013 14:51:05 +0100
> "Stefan (metze) Metzmacher" <metze at samba.org> wrote:
> 
>>> Is there any advantage, aside from smb1 uniformity?  
>>
>> Yes, one less malloc() per request, which costs a lot.
> 
> Ok, but having the request processing code-path explicitly use the
> talloc hierarchy attached to the request instead of the stackframe would
> provide the same benefit (albeit with a huge change).

If we would get rid of talloc_tos() again, it would be possible,
but I guess this won't happen...

>> I'll add a comment to the commit message...
>>
>> The aim is to do at most one malloc() per request.
>> I have plans to reclaim unused talloc_stackframe_pool() memory,
>> instead of doing one malloc() event loop, but that might come later.
> 
> Can't say I like where this is heading. I'd prefer to keep this kind of
> garbage collection complexity and overhead out of talloc, but I'll leave
> it for others to decide.

It won't be in talloc it would be in the talloc_stackframe code.

metze