On Mon, Dec 07, 2020 at 05:07:55PM -0600, Eric W. Biederman wrote: > My mistake. I missed that the actual code was highly optimized and only > safe in the presence of an unshared files struct. That's a polite way to spell "overoptimized for no good reason" ;-) > What I saw and what I thought the comment was talking about was that the > result of the test isn't guaranteed to be stable without having an > exclusive access to the files. I figured and if something changes later > and it becomes safe to pass the file name to a script later we don't > really care. > > In any event thank you for the review. I will queue up a follow on > patch that makes this use get_close_on_exec. I would rather put that switch to get_close_on_exec() first in the queue (or fold it into this patch)...
