[PATCH] n_tty: wake up poll(POLLRDNORM) on receiving data

[PATCH] n_tty: wake up poll(POLLRDNORM) on receiving data

[TATSUKAWA KOSUKE(立川 江介)]

Event POLLRDNORM should be equivalent to POLLIN when used as event in
poll().  However, in n_tty driver, POLLRDNORM does not return until
timeout even if there is terminal input, whereas POLLIN returns.

The following test program works until kernel-3.17, but the test stops
in poll() after commit 57087d515441 ("tty: Fix spurious poll() wakeups").

[Steps to run test program]
  $ cc -o test-pollrdnorm test-pollrdnorm.c
  $ ./test-pollrdnorm
  foo          <-- Type in something from the terminal followed by [RET].
                   The string should be echoed back.

  ------------------------< test-pollrdnorm.c >------------------------
  #include <stdio.h>
  #include <errno.h>
  #include <poll.h>
  #include <unistd.h>

  void main(void)
  {
	int		n;
	unsigned char	buf[8];
	struct pollfd	fds[1] = {{ 0, POLLRDNORM, 0 }};

	n = poll(fds, 1, -1);
	if (n < 0)
		perror("poll");
	n = read(0, buf, 8);
	if (n < 0)
		perror("read");
	if (n > 0)
		write(1, buf, n);
  }
  ------------------------------------------------------------------------

The attached patch fixes this problem.

Signed-off-by: Kosuke Tatsukawa <tatsu-ab1@nec.com>
Fixes: commit 57087d515441 ("tty: Fix spurious poll() wakeups")
Cc: stable@vger.kernel.org
---
 drivers/tty/n_tty.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/tty/n_tty.c b/drivers/tty/n_tty.c
index 5be6d02..b2b98fe 100644
--- a/drivers/tty/n_tty.c
+++ b/drivers/tty/n_tty.c
@@ -1369,7 +1369,7 @@ static void n_tty_receive_char_special(struct tty_struct *tty, unsigned char c)
 			put_tty_queue(c, ldata);
 			smp_store_release(&ldata->canon_head, ldata->read_head);
 			kill_fasync(&tty->fasync, SIGIO, POLL_IN);
-			wake_up_interruptible_poll(&tty->read_wait, EPOLLIN);
+			wake_up_interruptible_poll(&tty->read_wait, EPOLLIN | EPOLLRDNORM);
 			return;
 		}
 	}
@@ -1589,7 +1589,7 @@ static void __receive_buf(struct tty_struct *tty, const unsigned char *cp,
 
 	if (read_cnt(ldata)) {
 		kill_fasync(&tty->fasync, SIGIO, POLL_IN);
-		wake_up_interruptible_poll(&tty->read_wait, EPOLLIN);
+		wake_up_interruptible_poll(&tty->read_wait, EPOLLIN | EPOLLRDNORM);
 	}
 }
 

Re: [PATCH] n_tty: wake up poll(POLLRDNORM) on receiving data

[Greg Kroah-Hartman]

On Mon, Jan 24, 2022 at 08:17:22AM +0000, TATSUKAWA KOSUKE(立川 江介) wrote:
> Event POLLRDNORM should be equivalent to POLLIN when used as event in
> poll().

Where is that documented?

> However, in n_tty driver, POLLRDNORM does not return until
> timeout even if there is terminal input, whereas POLLIN returns.
> 
> The following test program works until kernel-3.17, but the test stops
> in poll() after commit 57087d515441 ("tty: Fix spurious poll() wakeups").
> 
> [Steps to run test program]
>   $ cc -o test-pollrdnorm test-pollrdnorm.c
>   $ ./test-pollrdnorm
>   foo          <-- Type in something from the terminal followed by [RET].
>                    The string should be echoed back.
> 
>   ------------------------< test-pollrdnorm.c >------------------------
>   #include <stdio.h>
>   #include <errno.h>
>   #include <poll.h>
>   #include <unistd.h>
> 
>   void main(void)
>   {
> 	int		n;
> 	unsigned char	buf[8];
> 	struct pollfd	fds[1] = {{ 0, POLLRDNORM, 0 }};
> 
> 	n = poll(fds, 1, -1);
> 	if (n < 0)
> 		perror("poll");
> 	n = read(0, buf, 8);
> 	if (n < 0)
> 		perror("read");
> 	if (n > 0)
> 		write(1, buf, n);
>   }
>   ------------------------------------------------------------------------
> 
> The attached patch fixes this problem.
> 
> Signed-off-by: Kosuke Tatsukawa <tatsu-ab1@nec.com>
> Fixes: commit 57087d515441 ("tty: Fix spurious poll() wakeups")

No need for "commit" here, please remove as the documentation asks you
to.

Can you resend this with this fixed up, and a pointer to where the
documentation is for this functionality.

thanks,

greg k-h

Re: [PATCH] n_tty: wake up poll(POLLRDNORM) on receiving data

[TATSUKAWA KOSUKE(立川 江介)]

Hi, thank you for the reply.

> On Mon, Jan 24, 2022 at 08:17:22AM +0000, TATSUKAWA KOSUKE(立川 江介) wrote:
> > Event POLLRDNORM should be equivalent to POLLIN when used as event in
> > poll().
> 
> Where is that documented?

It's in "man poll".  Below is an excerpt from the man page.
              POLLRDNORM
                     Equivalent to POLLIN.

I'll send a v2 patch in a separate message.

Many other calls to wake_up_interruptible_poll() in the kernel source
code already specify "POLLIN | POLLRDNORM", but there are places under
drivers/tty which only passes POLLIN (i.e. pty.c, tty_io.c,
tty_ldisc.c).  I haven't yet made test programs to hit the other
locations and I wasn't sure if they should also be fixed just for
consistency.

Best regards.


> > However, in n_tty driver, POLLRDNORM does not return until
> > timeout even if there is terminal input, whereas POLLIN returns.
> > 
> > The following test program works until kernel-3.17, but the test stops
> > in poll() after commit 57087d515441 ("tty: Fix spurious poll() wakeups").
> > 
> > [Steps to run test program]
> >   $ cc -o test-pollrdnorm test-pollrdnorm.c
> >   $ ./test-pollrdnorm
> >   foo          <-- Type in something from the terminal followed by [RET].
> >                    The string should be echoed back.
> > 
> >   ------------------------< test-pollrdnorm.c >------------------------
> >   #include <stdio.h>
> >   #include <errno.h>
> >   #include <poll.h>
> >   #include <unistd.h>
> > 
> >   void main(void)
> >   {
> > 	int		n;
> > 	unsigned char	buf[8];
> > 	struct pollfd	fds[1] = {{ 0, POLLRDNORM, 0 }};
> > 
> > 	n = poll(fds, 1, -1);
> > 	if (n < 0)
> > 		perror("poll");
> > 	n = read(0, buf, 8);
> > 	if (n < 0)
> > 		perror("read");
> > 	if (n > 0)
> > 		write(1, buf, n);
> >   }
> >   ------------------------------------------------------------------------
> > 
> > The attached patch fixes this problem.
> > 
> > Signed-off-by: Kosuke Tatsukawa <tatsu-ab1@nec.com>
> > Fixes: commit 57087d515441 ("tty: Fix spurious poll() wakeups")
> 
> No need for "commit" here, please remove as the documentation asks you
> to.
> 
> Can you resend this with this fixed up, and a pointer to where the
> documentation is for this functionality.
> 
> thanks,
> 
> greg k-h
---
Kosuke TATSUKAWA  | 1st Platform Software Division
                  | NEC Solution Innovators
                  | tatsu-ab1@nec.com