On Thu, Jan 11, 2018 at 05:47:50PM +0000, Al Viro wrote: > What I mean is that it would be nice to have do_select() and friends aware of that. > You are hiding the whole thing behind vfs_poll(); sure, we can't really exploit > that while we have the mix of converted and unconverted instances, but it would > be a nice payoff. I have tried to move it out, but I don't think it makes sense yet, given that we rely on passing the poll_table_struct everywhere. Btw, another issue: the procs and debugfs forwarders are in fact already broken right now for epoll and in-kernel callers. ->poll returning does not mean it is safe to release the module - it may still have an active wait_queue_entry on the waitqueue. For now I think I'll drop the support for the proc/debugfs forwarders, because they will need a real fix independent of this series.
