Re: [PATCH] sx tty_driver add .owner field remove MOD_INC_DEC_USE_COUNT

Re: [PATCH] sx tty_driver add .owner field remove MOD_INC_DEC_USE_COUNT

[Greg KH]

On Tue, May 27, 2003 at 12:01:08AM +0000, Linux Kernel Mailing List wrote:
> ChangeSet 1.1424, 2003/05/26 17:01:08-07:00, rusty@rustcorp.com.au
> 
> 	[PATCH] sx tty_driver add .owner field remove MOD_INC_DEC_USE_COUNT
> 	
> 	From:  Hanna Linder <hannal@us.ibm.com>
> 
> 
> # This patch includes the following deltas:
> #	           ChangeSet	1.1423  -> 1.1424 
> #	   drivers/char/sx.c	1.29    -> 1.30   
> #
> 
>  sx.c |   42 +-----------------------------------------
>  1 files changed, 1 insertion(+), 41 deletions(-)
> 
> 
> diff -Nru a/drivers/char/sx.c b/drivers/char/sx.c
> --- a/drivers/char/sx.c	Mon May 26 18:26:48 2003
> +++ b/drivers/char/sx.c	Mon May 26 18:26:48 2003
> @@ -305,7 +305,6 @@
>  static int  sx_get_CD (void * ptr); 
>  static void sx_shutdown_port (void * ptr);
>  static int  sx_set_real_termios (void  *ptr);
> -static void sx_hungup (void  *ptr);

Ick, this patch should be reverted, it should not be removing
sx_hungup() for no reason.  I think Hanna agrees with this.

thanks,

greg k-h

Re: [PATCH] sx tty_driver add .owner field remove MOD_INC_DEC_USE_COUNT

[Hanna Linder]

--On Thursday, May 29, 2003 12:02:50 PM -0700 Greg KH <greg@kroah.com> wrote:

> 
> Ick, this patch should be reverted, it should not be removing
> sx_hungup() for no reason.  I think Hanna agrees with this.

Yup. Sorry. Not sure what happened there. Here is the patch
to replace the sx_hangup function. This is based off 2.5.70-bk3
and I have compiled it but dont have the hardware to test it.

Hanna

---
diff -Nrup -Xdontdiff linux-2.5.70-bk3/drivers/char/sx.c linux-sx/drivers/char/sx.c
--- linux-2.5.70-bk3/drivers/char/sx.c	Thu May 29 14:03:43 2003
+++ linux-sx/drivers/char/sx.c	Thu May 29 14:11:39 2003
@@ -299,6 +299,7 @@ static void sx_enable_rx_interrupts (voi
 static int  sx_get_CD (void * ptr); 
 static void sx_shutdown_port (void * ptr);
 static int  sx_set_real_termios (void  *ptr);
+static void sx_hungup (void  *ptr);
 static void sx_close (void  *ptr);
 static int sx_chars_in_buffer (void * ptr);
 static int sx_init_board (struct sx_board *board);
@@ -379,6 +380,7 @@ static struct real_driver sx_real_driver
 	sx_set_real_termios, 
 	sx_chars_in_buffer,
 	sx_close,
+	sx_hungup,
 };
 
 
@@ -1499,6 +1501,39 @@ static int sx_open  (struct tty_struct *
 }
 
 
+/* I haven't the foggiest why the decrement use count has to happen
+   here. The whole linux serial drivers stuff needs to be redesigned.
+   My guess is that this is a hack to minimize the impact of a bug
+   elsewhere. Thinking about it some more. (try it sometime) Try
+   running minicom on a serial port that is driven by a modularized
+   driver. Have the modem hangup. Then remove the driver module. Then
+   exit minicom.  I expect an "oops".  -- REW */
+static void sx_hungup (void *ptr)
+{
+  /*
+	struct sx_port *port = ptr; 
+  */
+	func_enter ();
+
+	/* Don't force the SX card to close. mgetty doesn't like it !!!!!! -- pvdl */
+	/* For some reson we added this code. Don't know why anymore ;-( -- pvdl */
+	/*
+	sx_setsignals (port, 0, 0);
+	sx_reconfigure_port(port);	
+	sx_send_command (port, HS_CLOSE, 0, 0);
+
+	if (sx_read_channel_byte (port, hi_hstat) != HS_IDLE_CLOSED) {
+		if (sx_send_command (port, HS_FORCE_CLOSED, -1, HS_IDLE_CLOSED) != 1) {
+			printk (KERN_ERR 
+			        "sx: sent the force_close command, but card didn't react\n");
+		} else
+			sx_dprintk (SX_DEBUG_CLOSE, "sent the force_close command.\n");
+	}
+	*/
+	func_exit ();
+}
+
+
 static void sx_close (void *ptr)
 {
 	struct sx_port *port = ptr; 

Re: [PATCH] sx tty_driver add .owner field remove MOD_INC_DEC_USE_COUNT

[Rusty Russell]

In message <12430000.1054244110@w-hlinder> you write:
> --On Thursday, May 29, 2003 12:02:50 PM -0700 Greg KH <greg@kroah.com> wrote:
> 
> > 
> > Ick, this patch should be reverted, it should not be removing
> > sx_hungup() for no reason.  I think Hanna agrees with this.
> 
> Yup. Sorry. Not sure what happened there. Here is the patch
> to replace the sx_hangup function. This is based off 2.5.70-bk3
> and I have compiled it but dont have the hardware to test it.

Yes, but I don't think you need to put back the comment about
decrementing use counts 8)

> +/* I haven't the foggiest why the decrement use count has to happen
> +   here. The whole linux serial drivers stuff needs to be redesigned.
> +   My guess is that this is a hack to minimize the impact of a bug
> +   elsewhere. Thinking about it some more. (try it sometime) Try
> +   running minicom on a serial port that is driven by a modularized
> +   driver. Have the modem hangup. Then remove the driver module. Then
> +   exit minicom.  I expect an "oops".  -- REW */

Cheers,
Rusty.
--
  Anyone who quotes me in their sig is an idiot. -- Rusty Russell.

Re: [PATCH] sx tty_driver add .owner field remove MOD_INC_DEC_USE_COUNT

[Hanna Linder]

Heh. I figured out why I deleted it, it was not a mistake. The whole
function is commented out! Except the func_enter and exit calls which
are printk wrappers. Besides, no one ever calls this function anyway...

So leave it out. No need to put it back in...

Hanna


--On Friday, May 30, 2003 06:05:24 PM +1000 Rusty Russell <rusty@au1.ibm.com> wrote:
> In message <12430000.1054244110@w-hlinder> you write:
>> --On Thursday, May 29, 2003 12:02:50 PM -0700 Greg KH <greg@kroah.com> wrote:
>> 
>> > 
>> > Ick, this patch should be reverted, it should not be removing
>> > sx_hungup() for no reason.  I think Hanna agrees with this.
>> 
>> Yup. Sorry. Not sure what happened there. Here is the patch
>> to replace the sx_hangup function. This is based off 2.5.70-bk3
>> and I have compiled it but dont have the hardware to test it.
> 
> Yes, but I don't think you need to put back the comment about
> decrementing use counts 8)
> 
>> +/* I haven't the foggiest why the decrement use count has to happen
>> +   here. The whole linux serial drivers stuff needs to be redesigned.
>> +   My guess is that this is a hack to minimize the impact of a bug
>> +   elsewhere. Thinking about it some more. (try it sometime) Try
>> +   running minicom on a serial port that is driven by a modularized
>> +   driver. Have the modem hangup. Then remove the driver module. Then
>> +   exit minicom.  I expect an "oops".  -- REW */
> 
> Cheers,
> Rusty.
> --
>   Anyone who quotes me in their sig is an idiot. -- Rusty Russell.
>