Re: [PATCH/RFC] POLLHUP tinkering ...

[Davide Libenzi]

On Sun, 8 Jan 2006, David S. Miller wrote:

> From: Davide Libenzi <davidel@xxxxxxxxxxxxxxx>
> Date: Sun, 8 Jan 2006 16:11:10 -0800 (PST)
>
> > On Sun, 8 Jan 2006, David S. Miller wrote:
> >
> > > The extra last read is always necessary, it's an error synchronization
> > > barrier.  Did you know that?
> > >
> > > If a partial read or write hits an error, the successful amount of
> > > bytes read or written before the error occurred is returned.  Then any
> > > subsequent read or write will report the error immediately.
> >
> > Sorry for the missing info, but I was clearly talking about O_NONBLOCK
> > here.
>
> What I said still applies to O_NONBLOCK.

I thought you said in _not_ necessary, sorry. The extra read() for error 
discovery is just bogus, w/out proper Linux poll reporting. The epoll 
interface will have wait queue heads dropped inside the monitored devices 
wait queue, so I assume that an error condition would trigger a wakeup -> 
epoll event. If this is not true (but I'm pretty much sure it is), look at 
the extra read() for error reporting:

1) Good

read_loop();
--> Error happen on device
if (read() == ERROR)
	gotcha();

2)

read_loop();
if (read() == ERROR)
	whoops();
--> Error happen on device




- Davide


-
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/