[3/3] tty: n_gsm: add TIOCMIWAIT support

From: Daniel Starke <daniel.starke@siemens.com>

Add support for the TIOCMIWAIT ioctl on the virtual ttys. This enables the
user to wait for virtual modem signals like RING.

Signed-off-by: Daniel Starke <daniel.starke@siemens.com>
---
 drivers/tty/n_gsm.c | 32 ++++++++++++++++++++++++++++++++
 1 file changed, 32 insertions(+)

> > Add support for the TIOCMIWAIT ioctl on the virtual ttys. This enables the
> > user to wait for virtual modem signals like RING.
> > 
> > Signed-off-by: Daniel Starke <daniel.starke@siemens.com>
> > ---
> >  drivers/tty/n_gsm.c | 32 ++++++++++++++++++++++++++++++++
> >  1 file changed, 32 insertions(+)
> > 
> > diff --git a/drivers/tty/n_gsm.c b/drivers/tty/n_gsm.c
> > index 118511c1fa37..48fb7dad44cd 100644
> > --- a/drivers/tty/n_gsm.c
> > +++ b/drivers/tty/n_gsm.c
> > @@ -1542,6 +1542,7 @@ static void gsm_process_modem(struct tty_struct *tty, struct gsm_dlci *dlci,
> >  	if (brk & 0x01)
> >  		tty_insert_flip_char(&dlci->port, 0, TTY_BREAK);
> >  	dlci->modem_rx = mlines;
> > +	wake_up_interruptible(&dlci->gsm->event);
> >  }
> >  
> >  /**
> > @@ -3848,6 +3849,35 @@ static int gsm_modem_update(struct gsm_dlci *dlci, u8 brk)
> >  	return -EPROTONOSUPPORT;
> >  }
> >  
> > +/**
> > + * gsm_wait_modem_change	-	wait for modem status line change
> 
> No need for tabs.
> 
> Or for kernel doc for static functions, but that's not a big deal.

All other functions follow the same style here. Nevertheless, I will
replace the tabs with spaces. Originally, I had planned a code clean-up
after committing all the changes from my original RFC.

> > + * @dlci: channel
> > + * @mask: modem status line bits
> > + */
> > +
> 
> No blank line please, didn't checkpatch warn about that?

No, checkpatch did not warn about this and all other functions follow the
same style. I will remove the blank line.

> > +static int gsm_wait_modem_change(struct gsm_dlci *dlci, u32 mask)
> > +{
> > +	struct gsm_mux *gsm = dlci->gsm;
> > +	u32 old = dlci->modem_rx & mask;
> > +	int ret;
> > +
> > +	if (gsm->dead)
> > +		return -ENODEV;
> > +
> > +	do {
> > +		ret = wait_event_interruptible(gsm->event, gsm->dead
> > +					      || old ^ (dlci->modem_rx & mask));
> > +		if (ret)
> > +			return ret;
> > +		if (dlci->state != DLCI_OPEN)
> > +			return -EL2NSYNC;
> > +		if (gsm->dead)
> > +			return -ENODEV;
> > +	} while ((old ^ (dlci->modem_rx & mask)) == 0);
> 
> No way to break out of the loop if it goes for forever?

I assume that this is the expected behavior for TIOCMIWAIT. The functions
returns if:
- the requested modem signal changed
- the wait function got interrupted (e.g. by a signal)
- the underlying DLCI was closed
- the underlying ldisc device was removed

I can add that the function returns immediate if a mask has been passed
which matches no modem signal if this is preferred?

Best regards,
Daniel Starke

On Wed, Feb 01, 2023 at 09:30:28AM +0000, Starke, Daniel wrote:
> > > Add support for the TIOCMIWAIT ioctl on the virtual ttys. This enables the
> > > user to wait for virtual modem signals like RING.
> > > 
> > > Signed-off-by: Daniel Starke <daniel.starke@siemens.com>
> > > ---
> > >  drivers/tty/n_gsm.c | 32 ++++++++++++++++++++++++++++++++
> > >  1 file changed, 32 insertions(+)
> > > 
> > > diff --git a/drivers/tty/n_gsm.c b/drivers/tty/n_gsm.c
> > > index 118511c1fa37..48fb7dad44cd 100644
> > > --- a/drivers/tty/n_gsm.c
> > > +++ b/drivers/tty/n_gsm.c
> > > @@ -1542,6 +1542,7 @@ static void gsm_process_modem(struct tty_struct *tty, struct gsm_dlci *dlci,
> > >  	if (brk & 0x01)
> > >  		tty_insert_flip_char(&dlci->port, 0, TTY_BREAK);
> > >  	dlci->modem_rx = mlines;
> > > +	wake_up_interruptible(&dlci->gsm->event);
> > >  }
> > >  
> > >  /**
> > > @@ -3848,6 +3849,35 @@ static int gsm_modem_update(struct gsm_dlci *dlci, u8 brk)
> > >  	return -EPROTONOSUPPORT;
> > >  }
> > >  
> > > +/**
> > > + * gsm_wait_modem_change	-	wait for modem status line change
> > 
> > No need for tabs.
> > 
> > Or for kernel doc for static functions, but that's not a big deal.
> 
> All other functions follow the same style here. Nevertheless, I will
> replace the tabs with spaces. Originally, I had planned a code clean-up
> after committing all the changes from my original RFC.

Ah, ok, nevermind then, I didn't see that as it's not in the diffs.

Cleaning up later is fine.

> > > +static int gsm_wait_modem_change(struct gsm_dlci *dlci, u32 mask)
> > > +{
> > > +	struct gsm_mux *gsm = dlci->gsm;
> > > +	u32 old = dlci->modem_rx & mask;
> > > +	int ret;
> > > +
> > > +	if (gsm->dead)
> > > +		return -ENODEV;
> > > +
> > > +	do {
> > > +		ret = wait_event_interruptible(gsm->event, gsm->dead
> > > +					      || old ^ (dlci->modem_rx & mask));
> > > +		if (ret)
> > > +			return ret;
> > > +		if (dlci->state != DLCI_OPEN)
> > > +			return -EL2NSYNC;
> > > +		if (gsm->dead)
> > > +			return -ENODEV;
> > > +	} while ((old ^ (dlci->modem_rx & mask)) == 0);
> > 
> > No way to break out of the loop if it goes for forever?
> 
> I assume that this is the expected behavior for TIOCMIWAIT. The functions
> returns if:
> - the requested modem signal changed
> - the wait function got interrupted (e.g. by a signal)
> - the underlying DLCI was closed
> - the underlying ldisc device was removed

Hm, I guess you are right.  But wow, reading that
wait_event_interruptible() condition is crazy, please document that
really well to explain this properly so you will be able to understand
it in a year when you next have to fix it up :)

> I can add that the function returns immediate if a mask has been passed
> which matches no modem signal if this is preferred?

I don't think that would work, try it on some existing serial ports to
see what they do.

thanks,

greg k-h