[PATCH 03/14] tdb: Introduce tdb_chainlock_read_nonblock()

[Rusty Russell]

On Sat, 2 Oct 2010 11:13:42 pm Kirill Smelkov wrote:
> This (read+nonblock) case was missing...
> 
> But for consistency with other variants of chain locking, and also for
> cases when it's ok for readers to delay actual reads to minimize latency
> and use slightly outdated in-memory cache until next
> chainlock_read_nonblock() succeeds.

Yes, but AFAICT the only non-block user is ctdb.  And we really want to
discourage it until we have a proper solution (of course, TDB2 will
provide it :)

Cheers,
Rusty.