[PATCH 1/2] Introduce tty_unregister_ldisc()

[PATCH 1/2] Introduce tty_unregister_ldisc()

[Alexey Dobriyan]

It's a bit strange to see tty_register_ldisc() call in modules' exit functions.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
---

 Documentation/tty.txt |    2 +-
 drivers/char/tty_io.c |   35 ++++++++++++++++++++++++-----------
 include/linux/tty.h   |    1 +
 3 files changed, 26 insertions(+), 12 deletions(-)

diff -uprN linux-tty_register_ldisc_000/Documentation/tty.txt linux-tty_register_ldisc_001/Documentation/tty.txt
--- linux-tty_register_ldisc_000/Documentation/tty.txt	2005-05-31 20:12:47.000000000 +0400
+++ linux-tty_register_ldisc_001/Documentation/tty.txt	2005-05-31 23:19:32.000000000 +0400
@@ -22,7 +22,7 @@ copy of the structure. You must not re-r
 discipline even with the same data or your computer again will be eaten by
 demons.
 
-In order to remove a line discipline call tty_register_ldisc passing NULL.
+In order to remove a line discipline call tty_unregister_ldisc().
 In ancient times this always worked. In modern times the function will
 return -EBUSY if the ldisc is currently in use. Since the ldisc referencing
 code manages the module counts this should not usually be a concern.
diff -uprN linux-tty_register_ldisc_000/drivers/char/tty_io.c linux-tty_register_ldisc_001/drivers/char/tty_io.c
--- linux-tty_register_ldisc_000/drivers/char/tty_io.c	2005-05-31 20:13:34.000000000 +0400
+++ linux-tty_register_ldisc_001/drivers/char/tty_io.c	2005-05-31 23:17:40.000000000 +0400
@@ -262,17 +262,10 @@ int tty_register_ldisc(int disc, struct 
 		return -EINVAL;
 	
 	spin_lock_irqsave(&tty_ldisc_lock, flags);
-	if (new_ldisc) {
-		tty_ldiscs[disc] = *new_ldisc;
-		tty_ldiscs[disc].num = disc;
-		tty_ldiscs[disc].flags |= LDISC_FLAG_DEFINED;
-		tty_ldiscs[disc].refcount = 0;
-	} else {
-		if(tty_ldiscs[disc].refcount)
-			ret = -EBUSY;
-		else
-			tty_ldiscs[disc].flags &= ~LDISC_FLAG_DEFINED;
-	}
+	tty_ldiscs[disc] = *new_ldisc;
+	tty_ldiscs[disc].num = disc;
+	tty_ldiscs[disc].flags |= LDISC_FLAG_DEFINED;
+	tty_ldiscs[disc].refcount = 0;
 	spin_unlock_irqrestore(&tty_ldisc_lock, flags);
 	
 	return ret;
@@ -280,6 +273,26 @@ int tty_register_ldisc(int disc, struct 
 
 EXPORT_SYMBOL(tty_register_ldisc);
 
+int tty_unregister_ldisc(int disc)
+{
+	unsigned long flags;
+	int ret = 0;
+
+	if (disc < N_TTY || disc >= NR_LDISCS)
+		return -EINVAL;
+
+	spin_lock_irqsave(&tty_ldisc_lock, flags);
+	if (tty_ldiscs[disc].refcount)
+		ret = -EBUSY;
+	else
+		tty_ldiscs[disc].flags &= ~LDISC_FLAG_DEFINED;
+	spin_unlock_irqrestore(&tty_ldisc_lock, flags);
+
+	return ret;
+}
+
+EXPORT_SYMBOL(tty_unregister_ldisc);
+
 struct tty_ldisc *tty_ldisc_get(int disc)
 {
 	unsigned long flags;
diff -uprN linux-tty_register_ldisc_000/include/linux/tty.h linux-tty_register_ldisc_001/include/linux/tty.h
--- linux-tty_register_ldisc_000/include/linux/tty.h	2005-05-31 20:15:51.000000000 +0400
+++ linux-tty_register_ldisc_001/include/linux/tty.h	2005-05-31 23:18:23.000000000 +0400
@@ -345,6 +345,7 @@ extern int tty_check_change(struct tty_s
 extern void stop_tty(struct tty_struct * tty);
 extern void start_tty(struct tty_struct * tty);
 extern int tty_register_ldisc(int disc, struct tty_ldisc *new_ldisc);
+extern int tty_unregister_ldisc(int disc);
 extern int tty_register_driver(struct tty_driver *driver);
 extern int tty_unregister_driver(struct tty_driver *driver);
 extern void tty_register_device(struct tty_driver *driver, unsigned index, struct device *dev);

Re: [PATCH 1/2] Introduce tty_unregister_ldisc()

[Paul Fulghum]

This patch is wrong. Please do not apply.

An unmodified ldisc driver (externally maintained) will continue to call
tty_register_ldisc with NULL, but the new behavior will be to set the
ldisc pointer to NULL but have LDISC_FLAG_DEFINED set.

If you feel it is absolutely necessary to add this new function
for cosmetic reasons, then leave the old behavior in place
and make tty_unregister_ldisc a wrapper to tty_register_ldisc
that uses a NULL pointer.

This preserves sanity and compatibility.

The existing function has a bug:
If new_ldisc != NULL and LDISC_FLAG_DEFINED is set,
the ldisc is switched to new_ldisc without checking
for non-zero refcount of current ldisc.

--
Paul Fulghum
paulkf@microgate.com


On Tue, 2005-05-31 at 23:56 +0400, Alexey Dobriyan wrote:
> It's a bit strange to see tty_register_ldisc() call in modules' exit functions.
> ...
> --- linux-tty_register_ldisc_000/Documentation/tty.txt	2005-05-31 20:12:47.000000000 +0400
> +++ linux-tty_register_ldisc_001/Documentation/tty.txt	2005-05-31 23:19:32.000000000 +0400
> @@ -22,7 +22,7 @@ copy of the structure. You must not re-r
>  discipline even with the same data or your computer again will be eaten by
>  demons.
>  
> -In order to remove a line discipline call tty_register_ldisc passing NULL.
> +In order to remove a line discipline call tty_unregister_ldisc().
>  In ancient times this always worked. In modern times the function will
>  return -EBUSY if the ldisc is currently in use. Since the ldisc referencing
>  code manages the module counts this should not usually be a concern.
> diff -uprN linux-tty_register_ldisc_000/drivers/char/tty_io.c linux-tty_register_ldisc_001/drivers/char/tty_io.c
> --- linux-tty_register_ldisc_000/drivers/char/tty_io.c	2005-05-31 20:13:34.000000000 +0400
> +++ linux-tty_register_ldisc_001/drivers/char/tty_io.c	2005-05-31 23:17:40.000000000 +0400
> @@ -262,17 +262,10 @@ int tty_register_ldisc(int disc, struct 
>  		return -EINVAL;
>  	
>  	spin_lock_irqsave(&tty_ldisc_lock, flags);
> -	if (new_ldisc) {
> -		tty_ldiscs[disc] = *new_ldisc;
> -		tty_ldiscs[disc].num = disc;
> -		tty_ldiscs[disc].flags |= LDISC_FLAG_DEFINED;
> -		tty_ldiscs[disc].refcount = 0;
> -	} else {
> -		if(tty_ldiscs[disc].refcount)
> -			ret = -EBUSY;
> -		else
> -			tty_ldiscs[disc].flags &= ~LDISC_FLAG_DEFINED;
> -	}
> +	tty_ldiscs[disc] = *new_ldisc;
> +	tty_ldiscs[disc].num = disc;
> +	tty_ldiscs[disc].flags |= LDISC_FLAG_DEFINED;
> +	tty_ldiscs[disc].refcount = 0;
>  	spin_unlock_irqrestore(&tty_ldisc_lock, flags);
>  	
>  	return ret;
> @@ -280,6 +273,26 @@ int tty_register_ldisc(int disc, struct 
>  
>  EXPORT_SYMBOL(tty_register_ldisc);
>  
> +int tty_unregister_ldisc(int disc)
> +{
> +	unsigned long flags;
> +	int ret = 0;
> +
> +	if (disc < N_TTY || disc >= NR_LDISCS)
> +		return -EINVAL;
> +
> +	spin_lock_irqsave(&tty_ldisc_lock, flags);
> +	if (tty_ldiscs[disc].refcount)
> +		ret = -EBUSY;
> +	else
> +		tty_ldiscs[disc].flags &= ~LDISC_FLAG_DEFINED;
> +	spin_unlock_irqrestore(&tty_ldisc_lock, flags);
> +
> +	return ret;
> +}
> +
> +EXPORT_SYMBOL(tty_unregister_ldisc);
> +
>  struct tty_ldisc *tty_ldisc_get(int disc)
>  {
>  	unsigned long flags;
> diff -uprN linux-tty_register_ldisc_000/include/linux/tty.h linux-tty_register_ldisc_001/include/linux/tty.h
> --- linux-tty_register_ldisc_000/include/linux/tty.h	2005-05-31 20:15:51.000000000 +0400
> +++ linux-tty_register_ldisc_001/include/linux/tty.h	2005-05-31 23:18:23.000000000 +0400
> @@ -345,6 +345,7 @@ extern int tty_check_change(struct tty_s
>  extern void stop_tty(struct tty_struct * tty);
>  extern void start_tty(struct tty_struct * tty);
>  extern int tty_register_ldisc(int disc, struct tty_ldisc *new_ldisc);
> +extern int tty_unregister_ldisc(int disc);
>  extern int tty_register_driver(struct tty_driver *driver);
>  extern int tty_unregister_driver(struct tty_driver *driver);
>  extern void tty_register_device(struct tty_driver *driver, unsigned index, struct device *dev);
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
-- 
--
Paul Fulghum
paulkf@microgate.com

Re: [PATCH 1/2] Introduce tty_unregister_ldisc()

[Paul Fulghum]

On Tue, 2005-05-31 at 17:28 -0500, Paul Fulghum wrote:
> An unmodified ldisc driver (externally maintained) will continue to call
> tty_register_ldisc with NULL, but the new behavior will be to set the
> ldisc pointer to NULL but have LDISC_FLAG_DEFINED set.

I was partially wrong.
After Alexey's patch, a NULL new_ldisc is accessed 
(not assigned) causing an oops. This is still not
good behavior.

As for the refcount bug, this patch should fix it.
Comments?

--
Paul Fulghum
paulkf@microgate.com

--- linux-2.6.11/drivers/char/tty_io.c	2005-03-02 01:38:10.000000000 -0600
+++ b/drivers/char/tty_io.c	2005-05-31 22:10:41.000000000 -0500
@@ -262,16 +262,16 @@ int tty_register_ldisc(int disc, struct 
 		return -EINVAL;
 	
 	spin_lock_irqsave(&tty_ldisc_lock, flags);
-	if (new_ldisc) {
+	if ((tty_ldiscs[disc].flags & LDISC_FLAG_DEFINED) &&
+	    tty_ldiscs[disc].refcount)
+		ret = -EBUSY;
+	else if (new_ldisc) {
 		tty_ldiscs[disc] = *new_ldisc;
 		tty_ldiscs[disc].num = disc;
 		tty_ldiscs[disc].flags |= LDISC_FLAG_DEFINED;
 		tty_ldiscs[disc].refcount = 0;
 	} else {
-		if(tty_ldiscs[disc].refcount)
-			ret = -EBUSY;
-		else
-			tty_ldiscs[disc].flags &= ~LDISC_FLAG_DEFINED;
+		tty_ldiscs[disc].flags &= ~LDISC_FLAG_DEFINED;
 	}
 	spin_unlock_irqrestore(&tty_ldisc_lock, flags);
 	

Re: [PATCH 1/2] Introduce tty_unregister_ldisc()

[Paul Fulghum]

The following patch would be more appropriate than
my last suggestion for the case of trying to register
a ldisc driver to an occupied slot.

It does not make sense to allow an existing
registered driver to be overwritten, even
if the refcount is zero.

This *should* not happen with unique ldisc numbers,
but it seems like a reasonable check. Even if
Alexey's patch is applied, this would be a
reasonable check to integrate.

-- 
Paul Fulghum
paulkf@microgate.com

--- linux-2.6.11/drivers/char/tty_io.c	2005-03-02 01:38:10.000000000 -0600
+++ b/drivers/char/tty_io.c	2005-06-01 08:34:05.000000000 -0500
@@ -263,10 +263,14 @@ int tty_register_ldisc(int disc, struct 
 	
 	spin_lock_irqsave(&tty_ldisc_lock, flags);
 	if (new_ldisc) {
-		tty_ldiscs[disc] = *new_ldisc;
-		tty_ldiscs[disc].num = disc;
-		tty_ldiscs[disc].flags |= LDISC_FLAG_DEFINED;
-		tty_ldiscs[disc].refcount = 0;
+		if (tty_ldiscs[disc].flags & LDISC_FLAG_DEFINED)
+			ret = -EBUSY;
+		else {
+			tty_ldiscs[disc] = *new_ldisc;
+			tty_ldiscs[disc].num = disc;
+			tty_ldiscs[disc].flags |= LDISC_FLAG_DEFINED;
+			tty_ldiscs[disc].refcount = 0;
+		}
 	} else {
 		if(tty_ldiscs[disc].refcount)
 			ret = -EBUSY;

Re: [PATCH 1/2] Introduce tty_unregister_ldisc()

[Greg KH]

On Tue, May 31, 2005 at 05:28:11PM -0500, Paul Fulghum wrote:
> This patch is wrong. Please do not apply.
> 
> An unmodified ldisc driver (externally maintained) will continue to call
> tty_register_ldisc with NULL, but the new behavior will be to set the
> ldisc pointer to NULL but have LDISC_FLAG_DEFINED set.

Then fix that code.  We can not do anything about externally maintained
code, sorry.  

> If you feel it is absolutely necessary to add this new function
> for cosmetic reasons, then leave the old behavior in place
> and make tty_unregister_ldisc a wrapper to tty_register_ldisc
> that uses a NULL pointer.
> 
> This preserves sanity and compatibility.

We don't care about compatibility in order to keep our sanity.  See
Documentation/stable_api_nonsense.txt for details :)

thanks,

greg k-h

Re: [PATCH 1/2] Introduce tty_unregister_ldisc()

[Christoph Hellwig]

On Tue, May 31, 2005 at 05:28:11PM -0500, Paul Fulghum wrote:
> This patch is wrong. Please do not apply.
> 
> An unmodified ldisc driver (externally maintained) will continue to call
> tty_register_ldisc with NULL, but the new behavior will be to set the
> ldisc pointer to NULL but have LDISC_FLAG_DEFINED set.
> 
> If you feel it is absolutely necessary to add this new function
> for cosmetic reasons, then leave the old behavior in place
> and make tty_unregister_ldisc a wrapper to tty_register_ldisc
> that uses a NULL pointer.

Nah, we don't want to allow staying out of tree folks out of sync.  Adding
a BUG_ON for ldisk beeing NULL is much more reasoable.

Re: [PATCH 1/2] Introduce tty_unregister_ldisc()

[Alexey Dobriyan]

On Wednesday 01 June 2005 09:42, Christoph Hellwig wrote:
> On Tue, May 31, 2005 at 05:28:11PM -0500, Paul Fulghum wrote:
> > An unmodified ldisc driver (externally maintained) will continue to call
> > tty_register_ldisc with NULL, but the new behavior will be to set the
> > ldisc pointer to NULL but have LDISC_FLAG_DEFINED set.
> > 
> > If you feel it is absolutely necessary to add this new function
> > for cosmetic reasons, then leave the old behavior in place
> > and make tty_unregister_ldisc a wrapper to tty_register_ldisc
> > that uses a NULL pointer.
> 
> Nah, we don't want to allow staying out of tree folks out of sync.  Adding
> a BUG_ON for ldisk beeing NULL is much more reasoable.

Why add BUG_ON?

	tty_register_ldisc(N_FOO, NULL);
			...
	tty_ldiscs[disc] = *new_ldisc;	<= NULL

[PATCH 1/2] Introduce tty_unregister_ldisc()

[Alexey Dobriyan]

It's a bit strange to see tty_register_ldisc call in modules' exit functions.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
---

 Documentation/tty.txt |    2 +-
 drivers/char/tty_io.c |   35 ++++++++++++++++++++++++-----------
 include/linux/tty.h   |    1 +
 3 files changed, 26 insertions(+), 12 deletions(-)

diff -uprN linux-tty_register_ldisc_000/Documentation/tty.txt linux-tty_register_ldisc_001/Documentation/tty.txt
--- linux-tty_register_ldisc_000/Documentation/tty.txt	2005-05-31 20:12:47.000000000 +0400
+++ linux-tty_register_ldisc_001/Documentation/tty.txt	2005-05-31 23:19:32.000000000 +0400
@@ -22,7 +22,7 @@ copy of the structure. You must not re-r
 discipline even with the same data or your computer again will be eaten by
 demons.
 
-In order to remove a line discipline call tty_register_ldisc passing NULL.
+In order to remove a line discipline call tty_unregister_ldisc().
 In ancient times this always worked. In modern times the function will
 return -EBUSY if the ldisc is currently in use. Since the ldisc referencing
 code manages the module counts this should not usually be a concern.
diff -uprN linux-tty_register_ldisc_000/drivers/char/tty_io.c linux-tty_register_ldisc_001/drivers/char/tty_io.c
--- linux-tty_register_ldisc_000/drivers/char/tty_io.c	2005-05-31 20:13:34.000000000 +0400
+++ linux-tty_register_ldisc_001/drivers/char/tty_io.c	2005-05-31 23:17:40.000000000 +0400
@@ -262,17 +262,10 @@ int tty_register_ldisc(int disc, struct 
 		return -EINVAL;
 	
 	spin_lock_irqsave(&tty_ldisc_lock, flags);
-	if (new_ldisc) {
-		tty_ldiscs[disc] = *new_ldisc;
-		tty_ldiscs[disc].num = disc;
-		tty_ldiscs[disc].flags |= LDISC_FLAG_DEFINED;
-		tty_ldiscs[disc].refcount = 0;
-	} else {
-		if(tty_ldiscs[disc].refcount)
-			ret = -EBUSY;
-		else
-			tty_ldiscs[disc].flags &= ~LDISC_FLAG_DEFINED;
-	}
+	tty_ldiscs[disc] = *new_ldisc;
+	tty_ldiscs[disc].num = disc;
+	tty_ldiscs[disc].flags |= LDISC_FLAG_DEFINED;
+	tty_ldiscs[disc].refcount = 0;
 	spin_unlock_irqrestore(&tty_ldisc_lock, flags);
 	
 	return ret;
@@ -280,6 +273,26 @@ int tty_register_ldisc(int disc, struct 
 
 EXPORT_SYMBOL(tty_register_ldisc);
 
+int tty_unregister_ldisc(int disc)
+{
+	unsigned long flags;
+	int ret = 0;
+
+	if (disc < N_TTY || disc >= NR_LDISCS)
+		return -EINVAL;
+
+	spin_lock_irqsave(&tty_ldisc_lock, flags);
+	if (tty_ldiscs[disc].refcount)
+		ret = -EBUSY;
+	else
+		tty_ldiscs[disc].flags &= ~LDISC_FLAG_DEFINED;
+	spin_unlock_irqrestore(&tty_ldisc_lock, flags);
+
+	return ret;
+}
+
+EXPORT_SYMBOL(tty_unregister_ldisc);
+
 struct tty_ldisc *tty_ldisc_get(int disc)
 {
 	unsigned long flags;
diff -uprN linux-tty_register_ldisc_000/include/linux/tty.h linux-tty_register_ldisc_001/include/linux/tty.h
--- linux-tty_register_ldisc_000/include/linux/tty.h	2005-05-31 20:15:51.000000000 +0400
+++ linux-tty_register_ldisc_001/include/linux/tty.h	2005-05-31 23:18:23.000000000 +0400
@@ -345,6 +345,7 @@ extern int tty_check_change(struct tty_s
 extern void stop_tty(struct tty_struct * tty);
 extern void start_tty(struct tty_struct * tty);
 extern int tty_register_ldisc(int disc, struct tty_ldisc *new_ldisc);
+extern int tty_unregister_ldisc(int disc);
 extern int tty_register_driver(struct tty_driver *driver);
 extern int tty_unregister_driver(struct tty_driver *driver);
 extern void tty_register_device(struct tty_driver *driver, unsigned index, struct device *dev);