[PATCH] gatio: do not read if no read_handler

[PATCH] gatio: do not read if no read_handler

[Kristen Carlson Accardi]

[-- Attachment #1: Type: text/plain, Size: 1713 bytes --]

If no read_handler specified, leave data alone in case someone
else wants to read it.
---
 gatchat/gatio.c   |    4 ++++
 gatchat/gsmdial.c |   12 ++++--------
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/gatchat/gatio.c b/gatchat/gatio.c
index 61b0260..b4a3806 100644
--- a/gatchat/gatio.c
+++ b/gatchat/gatio.c
@@ -91,6 +91,10 @@ static gboolean received_data(GIOChannel *channel, GIOCondition cond,
 	if (cond & G_IO_NVAL)
 		return FALSE;
 
+	/* if nobody wants this data, leave it alone */
+	if (io->read_handler == NULL)
+		return TRUE;
+
 	/* Regardless of condition, try to read all the data available */
 	do {
 		toread = ring_buffer_avail_no_wrap(io->buf);
diff --git a/gatchat/gsmdial.c b/gatchat/gsmdial.c
index a531aa3..5fb406d 100644
--- a/gatchat/gsmdial.c
+++ b/gatchat/gsmdial.c
@@ -235,6 +235,8 @@ static void ppp_connect(const char *iface, const char *ip,
 static void ppp_disconnect(GAtPPPDisconnectReason reason, gpointer user_data)
 {
 	g_print("PPP Link down: %d\n", reason);
+	g_at_chat_resume(control);
+	g_at_chat_resume(modem);
 }
 
 static void connect_cb(gboolean ok, GAtResult *result, gpointer user_data)
@@ -249,14 +251,8 @@ static void connect_cb(gboolean ok, GAtResult *result, gpointer user_data)
 	/* get the data IO channel */
 	channel = g_at_chat_get_channel(modem);
 
-	/*
-	 * shutdown gatchat or else it tries to take all the input
-	 * from the modem and does not let PPP get it.
-	 */
-	g_at_chat_unref(control);
-	control = NULL;
-	g_at_chat_unref(modem);
-	modem = NULL;
+	g_at_chat_suspend(control);
+	g_at_chat_suspend(modem);
 
 	/* open ppp */
 	ppp = g_at_ppp_new(channel);
-- 
1.6.6.1

Re: [PATCH] gatio: do not read if no read_handler

[Denis Kenzior]

[-- Attachment #1: Type: text/plain, Size: 2175 bytes --]

Hi Kristen,

> If no read_handler specified, leave data alone in case someone
> else wants to read it.
> ---
>  gatchat/gatio.c   |    4 ++++
>  gatchat/gsmdial.c |   12 ++++--------
>  2 files changed, 8 insertions(+), 8 deletions(-)
> 
> diff --git a/gatchat/gatio.c b/gatchat/gatio.c
> index 61b0260..b4a3806 100644
> --- a/gatchat/gatio.c
> +++ b/gatchat/gatio.c
> @@ -91,6 +91,10 @@ static gboolean received_data(GIOChannel *channel,
>  GIOCondition cond, if (cond & G_IO_NVAL)
>  		return FALSE;
> 
> +	/* if nobody wants this data, leave it alone */
> +	if (io->read_handler == NULL)
> +		return TRUE;
> +

This is a really bad idea, we should never have cases where the read handler 
is unset, and if we do we should end up closing the socket when the buffer is 
full.

>  	/* Regardless of condition, try to read all the data available */
>  	do {
>  		toread = ring_buffer_avail_no_wrap(io->buf);
> diff --git a/gatchat/gsmdial.c b/gatchat/gsmdial.c
> index a531aa3..5fb406d 100644
> --- a/gatchat/gsmdial.c
> +++ b/gatchat/gsmdial.c
> @@ -235,6 +235,8 @@ static void ppp_connect(const char *iface, const char
>  *ip, static void ppp_disconnect(GAtPPPDisconnectReason reason, gpointer
>  user_data) {
>  	g_print("PPP Link down: %d\n", reason);
> +	g_at_chat_resume(control);
> +	g_at_chat_resume(modem);
>  }
> 
>  static void connect_cb(gboolean ok, GAtResult *result, gpointer user_data)
> @@ -249,14 +251,8 @@ static void connect_cb(gboolean ok, GAtResult *result,
>  gpointer user_data) /* get the data IO channel */
>  	channel = g_at_chat_get_channel(modem);
> 
> -	/*
> -	 * shutdown gatchat or else it tries to take all the input
> -	 * from the modem and does not let PPP get it.
> -	 */
> -	g_at_chat_unref(control);
> -	control = NULL;
> -	g_at_chat_unref(modem);
> -	modem = NULL;
> +	g_at_chat_suspend(control);
> +	g_at_chat_suspend(modem);
> 

Please send gsmdial patches separately.  Also, can you make g_at_ppp use the 
new_from_io constructor?  In that case you don't need the gatio modifications 
above.

>  	/* open ppp */
>  	ppp = g_at_ppp_new(channel);
> 

Regards,
-Denis

Re: [PATCH] gatio: do not read if no read_handler

[Kristen Carlson Accardi]

[-- Attachment #1: Type: text/plain, Size: 1018 bytes --]

On Mon, 10 May 2010 09:11:52 -0500
Denis Kenzior <denkenz@gmail.com> wrote:

> Hi Kristen,
> 
> > If no read_handler specified, leave data alone in case someone
> > else wants to read it.
> > ---
> >  gatchat/gatio.c   |    4 ++++
> >  gatchat/gsmdial.c |   12 ++++--------
> >  2 files changed, 8 insertions(+), 8 deletions(-)
> > 
> > diff --git a/gatchat/gatio.c b/gatchat/gatio.c
> > index 61b0260..b4a3806 100644
> > --- a/gatchat/gatio.c
> > +++ b/gatchat/gatio.c
> > @@ -91,6 +91,10 @@ static gboolean received_data(GIOChannel *channel,
> >  GIOCondition cond, if (cond & G_IO_NVAL)
> >  		return FALSE;
> > 
> > +	/* if nobody wants this data, leave it alone */
> > +	if (io->read_handler == NULL)
> > +		return TRUE;
> > +  
> 
> This is a really bad idea, we should never have cases where the read handler 
> is unset, and if we do we should end up closing the socket when the buffer is 
> full.

In your implementation of g_at_chat_suspend, you set the read handler to NULL.

Re: [PATCH] gatio: do not read if no read_handler

[Denis Kenzior]

[-- Attachment #1: Type: text/plain, Size: 1197 bytes --]

Hi Kristen,

> On Mon, 10 May 2010 09:11:52 -0500
> 
> Denis Kenzior <denkenz@gmail.com> wrote:
> > Hi Kristen,
> >
> > > If no read_handler specified, leave data alone in case someone
> > > else wants to read it.
> > > ---
> > >  gatchat/gatio.c   |    4 ++++
> > >  gatchat/gsmdial.c |   12 ++++--------
> > >  2 files changed, 8 insertions(+), 8 deletions(-)
> > >
> > > diff --git a/gatchat/gatio.c b/gatchat/gatio.c
> > > index 61b0260..b4a3806 100644
> > > --- a/gatchat/gatio.c
> > > +++ b/gatchat/gatio.c
> > > @@ -91,6 +91,10 @@ static gboolean received_data(GIOChannel *channel,
> > >  GIOCondition cond, if (cond & G_IO_NVAL)
> > >  		return FALSE;
> > >
> > > +	/* if nobody wants this data, leave it alone */
> > > +	if (io->read_handler == NULL)
> > > +		return TRUE;
> > > +
> >
> > This is a really bad idea, we should never have cases where the read
> > handler is unset, and if we do we should end up closing the socket when
> > the buffer is full.
> 
> In your implementation of g_at_chat_suspend, you set the read handler to
>  NULL.
> 

Correct, with the intent that someone else sets the read handler (e.g. 
g_at_ppp)

Regards,
-Denis

Re: [PATCH] gatio: do not read if no read_handler

[Kristen Carlson Accardi]

[-- Attachment #1: Type: text/plain, Size: 1485 bytes --]

On Mon, 10 May 2010 12:23:04 -0500
Denis Kenzior <denkenz@gmail.com> wrote:

> Hi Kristen,
> 
> > On Mon, 10 May 2010 09:11:52 -0500
> > 
> > Denis Kenzior <denkenz@gmail.com> wrote:
> > > Hi Kristen,
> > >
> > > > If no read_handler specified, leave data alone in case someone
> > > > else wants to read it.
> > > > ---
> > > >  gatchat/gatio.c   |    4 ++++
> > > >  gatchat/gsmdial.c |   12 ++++--------
> > > >  2 files changed, 8 insertions(+), 8 deletions(-)
> > > >
> > > > diff --git a/gatchat/gatio.c b/gatchat/gatio.c
> > > > index 61b0260..b4a3806 100644
> > > > --- a/gatchat/gatio.c
> > > > +++ b/gatchat/gatio.c
> > > > @@ -91,6 +91,10 @@ static gboolean received_data(GIOChannel *channel,
> > > >  GIOCondition cond, if (cond & G_IO_NVAL)
> > > >  		return FALSE;
> > > >
> > > > +	/* if nobody wants this data, leave it alone */
> > > > +	if (io->read_handler == NULL)
> > > > +		return TRUE;
> > > > +
> > >
> > > This is a really bad idea, we should never have cases where the read
> > > handler is unset, and if we do we should end up closing the socket when
> > > the buffer is full.
> > 
> > In your implementation of g_at_chat_suspend, you set the read handler to
> >  NULL.
> > 
> 
> Correct, with the intent that someone else sets the read handler (e.g. 
> g_at_ppp)

What if you receive data after the read handler is set to NULL, but before
someone else sets the read handler?  shouldn't we safe guard against that?

Re: [PATCH] gatio: do not read if no read_handler

[Denis Kenzior]

[-- Attachment #1: Type: text/plain, Size: 1863 bytes --]

Hi Kristen,

> On Mon, 10 May 2010 12:23:04 -0500
> 
> Denis Kenzior <denkenz@gmail.com> wrote:
> > Hi Kristen,
> >
> > > On Mon, 10 May 2010 09:11:52 -0500
> > >
> > > Denis Kenzior <denkenz@gmail.com> wrote:
> > > > Hi Kristen,
> > > >
> > > > > If no read_handler specified, leave data alone in case someone
> > > > > else wants to read it.
> > > > > ---
> > > > >  gatchat/gatio.c   |    4 ++++
> > > > >  gatchat/gsmdial.c |   12 ++++--------
> > > > >  2 files changed, 8 insertions(+), 8 deletions(-)
> > > > >
> > > > > diff --git a/gatchat/gatio.c b/gatchat/gatio.c
> > > > > index 61b0260..b4a3806 100644
> > > > > --- a/gatchat/gatio.c
> > > > > +++ b/gatchat/gatio.c
> > > > > @@ -91,6 +91,10 @@ static gboolean received_data(GIOChannel
> > > > > *channel, GIOCondition cond, if (cond & G_IO_NVAL)
> > > > >  		return FALSE;
> > > > >
> > > > > +	/* if nobody wants this data, leave it alone */
> > > > > +	if (io->read_handler == NULL)
> > > > > +		return TRUE;
> > > > > +
> > > >
> > > > This is a really bad idea, we should never have cases where the read
> > > > handler is unset, and if we do we should end up closing the socket
> > > > when the buffer is full.
> > >
> > > In your implementation of g_at_chat_suspend, you set the read handler
> > > to NULL.
> >
> > Correct, with the intent that someone else sets the read handler (e.g.
> > g_at_ppp)
> 
> What if you receive data after the read handler is set to NULL, but before
> someone else sets the read handler?  shouldn't we safe guard against that?
> 

We do safe guard, once the buffer runs out of space we close the GAtIO.  
However, letting GAtIO enter the event-loop (since only then the io watch will 
fire) without a read handler is a user error.  Having an active GAtIO without a 
read handler is useless...

Regards,
-Denis