Re: [PATCH] [RFC] Optimize select/poll by putting small data sets on the stack

[Andi Kleen]

On Wednesday 04 January 2006 01:43, Eric Dumazet wrote:

> And I would like to pinpoint that set_fd_set() implementation seems *buggy*
> :
>
> It should not use __copy_to_user() but the real one (copy_to_user())
> because the calling thread could have slept in do_select() and another
> thread played mm games during this sleep.

__ only skips the access_ok which checks the kernel/user boundary, and the 
kernel/user boundary doesn't change even while sleeping.

On very early 386s it did something more to work around a CPU bug, but that is 
racy on multithreaded processes in any case. Not really worth caring about.

Also I doubt any such machines are left in working condition. That old 
workaround code could be probably safely removed by now.

-Andi
-
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/