Em Wed, May 28, 2014 at 03:33:51PM -0600, Chris Friesen escreveu:

On 05/28/2014 01:50 PM, 'Arnaldo Carvalho de Melo' wrote:

quoted

What is being discussed here is how to return the EFAULT that may happen
_after_ datagram processing, be it interrupted by an EFAULT, signal, or
plain returning all that was requested, with no errors.

quoted

This EFAULT _after_ datagram processing may happen when updating the
remaining timeout, because then how can userspace both receive the
number of successfully copied datagrams (in any of the cases mentioned
in the previous paragraph) and know that that timeout can't be used
because there was a problem while trying to copy it to userspace
(EFAULT)?

How does select() handle this problem?  It updates the timeout and also
modifies other data.

Could we just check whether the timeout pointer is valid before doing
anything else?  Of course we could still fault the page out while waiting
for messages and then fail to fault it back in later, but that seems like a
not-very-likely scenario.

I'll check how select behaves, and yes, I think it is not-very-likely
and what we're doing now is reasonable for datagram protocols, i.e. to
return -EFAULT when updating the timeout fails, not reporting if packets
were successfully received, i.e. they end up being "dropped", as
userspace can't easily figure out if some was received short of painting
it with some pattern and then checking the ones that aren't with that
pattern.

- Arnaldo