A discussion between database and kernel developers

> 6. Only writeback pages if explicitly synced.

> No, I'm sorry, that's never going to be possible. No user space
> application has all the facts. If we give you an interface to
> force unconditional holding of dirty pages in core you'll livelock
> the system eventually because you made a wrong decision to hold
> too many dirty pages.

Wouldn't a "deferred dirty page writeback mmap area" be like marking the mmap'ed area 'MAP_PRIVATE' and when writeback is allowed to happen turn it back into a 'MAP_SHARED'?

Because in that case I don't see how this wouldn't be exactly the same problem as Postgresql holding it into some large other private memory area. The kernel wouldn't be allowed to do anything with it in that case either (short of swapping it).