From owner-freebsd-hackers  Sat Dec  2 20:01:53 1995
Return-Path: owner-hackers
Received: (from root@localhost)
          by freefall.freebsd.org (8.6.12/8.6.6) id UAA15661
          for hackers-outgoing; Sat, 2 Dec 1995 20:01:53 -0800
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id UAA15645
          for <freebsd-hackers@freebsd.org>; Sat, 2 Dec 1995 20:01:39 -0800
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id OAA07217; Sun, 3 Dec 1995 14:58:40 +1100
Date: Sun, 3 Dec 1995 14:58:40 +1100
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199512030358.OAA07217@godzilla.zeta.org.au>
To: freebsd-hackers@freebsd.org, witr@rwwa.com
Subject: Re: Proper way to determine non-blocking status...
Sender: owner-hackers@freebsd.org
Precedence: bulk

>Even though it seems like it is almost never done by
>existing driver code, would I be correct in assuming that
>the *right* way to determine if a device read is allowed to
>sleep or not is the following?

>int xxread(dev_t dev, struct uio *uio, int ioflag)
>{
>  if (ioflag & IO_NDELAY) {
>    /* Can't block here */
>    return EWOULDBLOCK;
>  } else {				/* Otherwise try again */
>    /* Can sleep here */
>    ...
>  }
>}

Yes.  See tty.c for examples.  In open/ioctl/close you have to check
O_NONBLOCK (aka FNONBLOCK) instead of IO_NDELAY.

The ugl^H^H^Hrightly formatted way is:

int
xxread(dev, uio, ioflag)
	dev_t dev;
	struct uio *uio;
	int ioflag;
{

	/* Return early for nonblocking reads. */
	if (ioflag & IO_NDELAY)
		return (EWOULDBLOCK);

	/* Can sleep here. */
	...
}

Bruce