[ewg] Re: RFC: SRC API

[Michael S. Tsirkin]

> Quoting Roland Dreier <rdreier at cisco.com>:
> Subject: Re: RFC: SRC API
> 
>  > - manage SRC domains
> 
> OK, I guess, but why do we need so many functions to create, share,
> get shared, put SRC domains?
> 
> How about just two functions:
> 
> 	struct ibv_src_domain *ibv_open_src_domain(struct ibv_context *context,
> 						   int fd, int oflag);
> 
> where oflag is 0 or some combination of O_CREAT and O_EXCL, and:
> 
> 	int ibv_close_src_domain(struct ibv_src_domain);
> 
> 
> Is there some case where explicitly unsharing an SRC domain is useful
> that can't be dealt with just by unlinking whatever file was used to
> create it?  Other processes might be left with open file descriptors
> pointing to that file, but they could have already gotten the src
> domain anyway.

OK.

>  > - I envision implementing this sharing mechanism in kernel by means
>  >   of a per-device tree, with inode as a key and domain object
>  >   as a value.
> 
> Is the inode or the file descriptor the key?  Earlier you wrote
>
>  > 1:1 association between an SRC domain and a file.
> 
> so there's some ambiguity.

The inode.

> Also, what's the lifetime of SRC domains?  Are they implicitly
> destroyed when all processes with a reference exit?

Yes, that's the idea.

-- 
MST