Re: [PATCH 0/3] *at syscalls: Intro

[Linus Torvalds]

On Thu, 15 Dec 2005, Nicholas Miell wrote:
> 
> Don't take this as an objection to implementation of the *at() syscalls
> in Linux, though; rather, look at is as a request for the addition of
> int pthread_attr_setfssharing_np(pthread_attr_t *attr, int share) and
> int pthread_attr_getfssharing_np(pthread_attr_t *attr) to glibc.

I don't think separate fs/files makes sense with pthreads.

Why? Signals.

Signals are shared in a pthread environment, so signal handlers can happen 
in any thread. If you don't share the filesystem thing, your signal 
handlers had better not do any file operations. You'd better not try to 
use the old "send a byte down a pipe" trick to wake somebody else up, 
because the thread you may take the signal in may not _have_ that pipe fd 
open.

Now, opening files in a signal handler may be unusual enough that 
having separate cwd/root (!CLONE_FS as opposed to !CLONE_FILES) might be 
more commonly usable, but it's still potentially asking for some really 
funky behaviour.

			Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/