[PATCH] CONSOLE_LP_STRICT Kconfig option

[PATCH] CONSOLE_LP_STRICT Kconfig option

[Luca Falavigna]

This patch, built against kernel version 2.6.16-rc3, provides a Kconfig 
option in order to easily enable or disable CONSOLE_LP_STRICT variable 
in drivers/char/lp.c without modifying it directly.



Signed-off-by: Luca Falavigna <dktrkranz@gmail.com>

--- linux-2.6.16-rc3/drivers/char/lp.c.orig    2006-01-08 
16:48:14.000000000 +0100
+++ linux-2.6.16-rc3/drivers/char/lp.c    2006-02-14 13:43:41.000000000 
+0100
@@ -686,9 +686,13 @@ static struct file_operations lp_fops =
  #define CONSOLE_LP 0

  /* If the printer is out of paper, we can either lose the messages or
- * stall until the printer is happy again.  Define CONSOLE_LP_STRICT
- * non-zero to get the latter behaviour. */
-#define CONSOLE_LP_STRICT 1
+ * stall until the printer is happy again. If CONSOLE_LP_STRICT is
+ * non-zero to, we get the latter behaviour. */
+#ifdef CONFIG_LP_CONSOLE_STRICT
+# define CONSOLE_LP_STRICT 1
+#else
+# define CONSOLE_LP_STRICT 0
+#endif

  /* The console must be locked when we get here. */

@@ -697,7 +701,7 @@ static void lp_console_write (struct con
  {
      struct pardevice *dev = lp_table[CONSOLE_LP].dev;
      struct parport *port = dev->port;
-    ssize_t written;
+    ssize_t written = 0;

      if (parport_claim (dev))
          /* Nothing we can do. */
--- linux-2.6.16-rc3/drivers/char/Kconfig.orig    2006-02-14 
00:14:08.000000000 +0100
+++ linux-2.6.16-rc3/drivers/char/Kconfig    2006-02-14 
13:47:33.000000000 +0100
@@ -512,14 +512,21 @@ config LP_CONSOLE
        doing that; to actually get it to happen you need to pass the
        option "console=lp0" to the kernel at boot time.

-      If the printer is out of paper (or off, or unplugged, or too
-      busy..) the kernel will stall until the printer is ready again.
-      By defining CONSOLE_LP_STRICT to 0 (at your own risk) you
-      can make the kernel continue when this happens,
-      but it'll lose the kernel messages.
-
        If unsure, say N.

+config LP_CONSOLE_STRICT
+    bool "Wait for a ready printer"
+    depends on LP_CONSOLE
+    default y
+    ---help---
+      With this option enabled, if the printer is out of paper (or off,
+      or unplugged, or too busy..) the kernel will stall until the printer
+      is ready again. By turning this option off (at your own risk), you
+      can make the kernel continue when this happens, but it will lose
+      some kernel messages.
+
+      If unsure, say Y
+
  config PPDEV
      tristate "Support for user-space parallel port device drivers"
      depends on PARPORT



Regards,

-- 
Luca

Re: [PATCH] CONSOLE_LP_STRICT Kconfig option

[Andi Kleen]

Luca Falavigna <dktrkranz@gmail.com> writes:

> This patch, built against kernel version 2.6.16-rc3, provides a
> Kconfig option in order to easily enable or disable CONSOLE_LP_STRICT
> variable in drivers/char/lp.c without modifying it directly.

This shouldn't be a CONFIG. This should be a runtime option.
It's ridiculous to have to recompile your kernel just to fix some
problem with your printer.

e.g. sysctl, ioctl, sysfs entry, module parameter. Whatever is en
vogue these days. Easiest would be probably a module_param().

-Andi

Re: [PATCH] CONSOLE_LP_STRICT Kconfig option

[Luca Falavigna]

Oops, I noticed I sent twice my email. Sorry.

14 Feb 2006 15:59:56 +0100, Andi Kleen <ak@suse.de>:
> This shouldn't be a CONFIG. This should be a runtime option.
> It's ridiculous to have to recompile your kernel just to fix some
> problem with your printer.
>
> e.g. sysctl, ioctl, sysfs entry, module parameter. Whatever is en
> vogue these days. Easiest would be probably a module_param().

This feature only gets involved when passing console=lp0 parameter to
the bootloader. I never tried to load a new system console while
system was running so I'm not sure if it behaves correctly. If it
does, I will modify this patch following your advices.

Regards,

--
Luca

Re: [PATCH] CONSOLE_LP_STRICT Kconfig option

[Randy.Dunlap]

On Wed, 15 Feb 2006 10:03:30 +0100 Luca Falavigna wrote:

> Oops, I noticed I sent twice my email. Sorry.
> 
> 14 Feb 2006 15:59:56 +0100, Andi Kleen <ak@suse.de>:
> > This shouldn't be a CONFIG. This should be a runtime option.
> > It's ridiculous to have to recompile your kernel just to fix some
> > problem with your printer.
> >
> > e.g. sysctl, ioctl, sysfs entry, module parameter. Whatever is en
> > vogue these days. Easiest would be probably a module_param().
> 
> This feature only gets involved when passing console=lp0 parameter to
> the bootloader. I never tried to load a new system console while
> system was running so I'm not sure if it behaves correctly. If it
> does, I will modify this patch following your advices.

Andi's suggestion seems fine to me:  use a module_param() for
CONSOLE_LP_STRICT instead of a hidden build-time (non-Kconfig)
option.  Are you interested in making this change?

---
~Randy

Re: [PATCH] CONSOLE_LP_STRICT Kconfig option

[Randy.Dunlap]

On Wed, 12 Apr 2006 20:53:28 +0200 Luca Falavigna wrote:

> 2006/4/12, Randy.Dunlap <rdunlap@xenotime.net>:
> > On Wed, 15 Feb 2006 10:03:30 +0100 Luca Falavigna wrote:
> >
> > > Oops, I noticed I sent twice my email. Sorry.
> > >
> > > 14 Feb 2006 15:59:56 +0100, Andi Kleen <ak@suse.de>:
> > > > This shouldn't be a CONFIG. This should be a runtime option.
> > > > It's ridiculous to have to recompile your kernel just to fix some
> > > > problem with your printer.
> > > >
> > > > e.g. sysctl, ioctl, sysfs entry, module parameter. Whatever is en
> > > > vogue these days. Easiest would be probably a module_param().
> > >
> > > This feature only gets involved when passing console=lp0 parameter to
> > > the bootloader. I never tried to load a new system console while
> > > system was running so I'm not sure if it behaves correctly. If it
> > > does, I will modify this patch following your advices.
> >
> > Andi's suggestion seems fine to me:  use a module_param() for
> > CONSOLE_LP_STRICT instead of a hidden build-time (non-Kconfig)
> > option.  Are you interested in making this change?
> >
> > ---
> 
> I can give it a try. I'm not sure if this can be done when system is loaded.

I think that if you make it a module_param() and use (root-) writeable
permissions on it, it's just a variable that can be changed after the
driver is loaded and running.

Andi, did you want just a boot-time option or a run-time (changeable) option?

---
~Randy

Re: [PATCH] CONSOLE_LP_STRICT Kconfig option

[Luca Falavigna]

Randy.Dunlap ha scritto:
> I think that if you make it a module_param() and use (root-) writeable
> permissions on it, it's just a variable that can be changed after the
> driver is loaded and running.
> 
> Andi, did you want just a boot-time option or a run-time (changeable) option?

Here is a patch (built against 2.6.17-rc1) which does use of __setup()
call to dynamically adjust console_lp_strict value.
Actually the only way to get console line printer working is to enable
it at system startup so I adjusted Kconfig parameters accordingly.
Feel free to submit any feedback.



Signed-off-by: Luca Falavigna <dktrkranz@gmail.com>

--- ./drivers/char/lp.c.orig	2006-01-08 16:48:14.000000000 +0100
+++ ./drivers/char/lp.c	2006-04-14 00:31:57.000000000 +0200
@@ -686,9 +686,17 @@ static struct file_operations lp_fops =
 #define CONSOLE_LP 0

 /* If the printer is out of paper, we can either lose the messages or
- * stall until the printer is happy again.  Define CONSOLE_LP_STRICT
+ * stall until the printer is happy again.  Setting console_lp_strict
  * non-zero to get the latter behaviour. */
-#define CONSOLE_LP_STRICT 1
+static unsigned int console_lp_strict = 1;
+
+static int __init console_lp_strict_setup (char *str)
+{
+	console_lp_strict = simple_strtol(str, NULL, 0);
+	return 1;
+}
+
+__setup("console_lp_strict=", console_lp_strict_setup);

 /* The console must be locked when we get here. */

@@ -697,7 +705,7 @@ static void lp_console_write (struct con
 {
 	struct pardevice *dev = lp_table[CONSOLE_LP].dev;
 	struct parport *port = dev->port;
-	ssize_t written;
+	ssize_t written = 0;

 	if (parport_claim (dev))
 		/* Nothing we can do. */
@@ -737,9 +745,9 @@ static void lp_console_write (struct con
 				written = parport_write (port, crlf, i);
 				if (written > 0)
 					i -= written, crlf += written;
-			} while (i > 0 && (CONSOLE_LP_STRICT || written > 0));
+			} while (i > 0 && (console_lp_strict || written > 0));
 		}
-	} while (count > 0 && (CONSOLE_LP_STRICT || written > 0));
+	} while (count > 0 && (console_lp_strict || written > 0));

 	parport_release (dev);
 }
--- ./drivers/char/Kconfig.orig	2006-04-13 20:52:52.000000000 +0200
+++ ./drivers/char/Kconfig	2006-04-14 00:27:34.000000000 +0200
@@ -506,7 +506,7 @@ config PRINTER

 config LP_CONSOLE
 	bool "Support for console on line printer"
-	depends on PRINTER
+	depends on PRINTER = y
 	---help---
 	  If you want kernel messages to be printed out as they occur, you
 	  can have a console on the printer. This option adds support for
@@ -515,9 +515,9 @@ config LP_CONSOLE

 	  If the printer is out of paper (or off, or unplugged, or too
 	  busy..) the kernel will stall until the printer is ready again.
-	  By defining CONSOLE_LP_STRICT to 0 (at your own risk) you
-	  can make the kernel continue when this happens,
-	  but it'll lose the kernel messages.
+	  By passing the option "console_lp_strict=0" to the kernel at
+	  boot time (at your own risk) you can make the kernel continue
+	  when this happens, but it will lose the kernel messages.

 	  If unsure, say N.


Regards,

-- 
Luca

[PATCH] CONSOLE_LP_STRICT Kconfig option

[Luca Falavigna]

This patch, built against kernel version 2.6.16-rc3, provides a Kconfig
option in order to easily enable or disable CONSOLE_LP_STRICT variable
in drivers/char/lp.c without modifying it directly.



Signed-off-by: Luca Falavigna <dktrkranz@gmail.com>

--- linux-2.6.16-rc3/drivers/char/lp.c.orig    2006-01-08
16:48:14.000000000 +0100
+++ linux-2.6.16-rc3/drivers/char/lp.c    2006-02-14 13:43:41.000000000
+0100
@@ -686,9 +686,13 @@ static struct file_operations lp_fops =
 #define CONSOLE_LP 0

 /* If the printer is out of paper, we can either lose the messages or
- * stall until the printer is happy again.  Define CONSOLE_LP_STRICT
- * non-zero to get the latter behaviour. */
-#define CONSOLE_LP_STRICT 1
+ * stall until the printer is happy again. If CONSOLE_LP_STRICT is
+ * non-zero to, we get the latter behaviour. */
+#ifdef CONFIG_LP_CONSOLE_STRICT
+# define CONSOLE_LP_STRICT 1
+#else
+# define CONSOLE_LP_STRICT 0
+#endif

 /* The console must be locked when we get here. */

@@ -697,7 +701,7 @@ static void lp_console_write (struct con
 {
     struct pardevice *dev = lp_table[CONSOLE_LP].dev;
     struct parport *port = dev->port;
-    ssize_t written;
+    ssize_t written = 0;

     if (parport_claim (dev))
         /* Nothing we can do. */
--- linux-2.6.16-rc3/drivers/char/Kconfig.orig    2006-02-14
00:14:08.000000000 +0100
+++ linux-2.6.16-rc3/drivers/char/Kconfig    2006-02-14
13:47:33.000000000 +0100
@@ -512,14 +512,21 @@ config LP_CONSOLE
       doing that; to actually get it to happen you need to pass the
       option "console=lp0" to the kernel at boot time.

-      If the printer is out of paper (or off, or unplugged, or too
-      busy..) the kernel will stall until the printer is ready again.
-      By defining CONSOLE_LP_STRICT to 0 (at your own risk) you
-      can make the kernel continue when this happens,
-      but it'll lose the kernel messages.
-
       If unsure, say N.

+config LP_CONSOLE_STRICT
+    bool "Wait for a ready printer"
+    depends on LP_CONSOLE
+    default y
+    ---help---
+      With this option enabled, if the printer is out of paper (or off,
+      or unplugged, or too busy..) the kernel will stall until the printer
+      is ready again. By turning this option off (at your own risk), you
+      can make the kernel continue when this happens, but it will lose
+      some kernel messages.
+
+      If unsure, say Y
+
 config PPDEV
     tristate "Support for user-space parallel port device drivers"
     depends on PARPORT



Regards,

--
Luca

Re: [PATCH] CONSOLE_LP_STRICT Kconfig option

[Randy.Dunlap]

On Tue, 14 Feb 2006 22:05:37 +0100 Luca Falavigna wrote:

> This patch, built against kernel version 2.6.16-rc3, provides a Kconfig
> option in order to easily enable or disable CONSOLE_LP_STRICT variable
> in drivers/char/lp.c without modifying it directly.


The patch does not apply cleanly to 2.6.16-rc3 -- because tabs
have been converted to spaces, either by your email client or
by using copy-n-paste to create the email.

Couple of typo corrections below.


> Signed-off-by: Luca Falavigna <dktrkranz@gmail.com>
> 
> --- linux-2.6.16-rc3/drivers/char/lp.c.orig    2006-01-08
> 16:48:14.000000000 +0100
> +++ linux-2.6.16-rc3/drivers/char/lp.c    2006-02-14 13:43:41.000000000
> +0100
> @@ -686,9 +686,13 @@ static struct file_operations lp_fops =
>  #define CONSOLE_LP 0
> 
>  /* If the printer is out of paper, we can either lose the messages or
> - * stall until the printer is happy again.  Define CONSOLE_LP_STRICT
> - * non-zero to get the latter behaviour. */
> -#define CONSOLE_LP_STRICT 1
> + * stall until the printer is happy again. If CONSOLE_LP_STRICT is
> + * non-zero to, we get the latter behaviour. */
*              too,
> +#ifdef CONFIG_LP_CONSOLE_STRICT
> +# define CONSOLE_LP_STRICT 1
> +#else
> +# define CONSOLE_LP_STRICT 0
> +#endif
> 
>  /* The console must be locked when we get here. */
> 
> @@ -697,7 +701,7 @@ static void lp_console_write (struct con
>  {
>      struct pardevice *dev = lp_table[CONSOLE_LP].dev;
>      struct parport *port = dev->port;
> -    ssize_t written;
> +    ssize_t written = 0;
> 
>      if (parport_claim (dev))
>          /* Nothing we can do. */
> --- linux-2.6.16-rc3/drivers/char/Kconfig.orig    2006-02-14
> 00:14:08.000000000 +0100
> +++ linux-2.6.16-rc3/drivers/char/Kconfig    2006-02-14
> 13:47:33.000000000 +0100
> @@ -512,14 +512,21 @@ config LP_CONSOLE
> 
> +config LP_CONSOLE_STRICT
> +    bool "Wait for a ready printer"
> +    depends on LP_CONSOLE
> +    default y
> +    ---help---
> +      With this option enabled, if the printer is out of paper (or off,
> +      or unplugged, or too busy..) the kernel will stall until the printer
* add comma:                         ,
> +      is ready again. By turning this option off (at your own risk), you
> +      can make the kernel continue when this happens, but it will lose
> +      some kernel messages.
> +
> +      If unsure, say Y
> +
>  config PPDEV
>      tristate "Support for user-space parallel port device drivers"
>      depends on PARPORT


---
~Randy

Re: [PATCH] CONSOLE_LP_STRICT Kconfig option

[Luca Falavigna]

[-- Attachment #1: Type: text/plain, Size: 2598 bytes --]

Randy.Dunlap ha scritto:
> The patch does not apply cleanly to 2.6.16-rc3 -- because tabs
> have been converted to spaces, either by your email client or
> by using copy-n-paste to create the email.

Here is what happens when you use Outlook ;)



Signed-off-by: Luca Falavigna <dktrkranz@gmail.com>

--- linux-2.6.16-rc3/drivers/char/lp.c.orig	2006-01-08 
16:48:14.000000000 +0100
+++ linux-2.6.16-rc3/drivers/char/lp.c	2006-02-14 13:43:41.000000000 +0100
@@ -686,9 +686,13 @@ static struct file_operations lp_fops =
  #define CONSOLE_LP 0

  /* If the printer is out of paper, we can either lose the messages or
- * stall until the printer is happy again.  Define CONSOLE_LP_STRICT
- * non-zero to get the latter behaviour. */
-#define CONSOLE_LP_STRICT 1
+ * stall until the printer is happy again. If CONSOLE_LP_STRICT is
+ * non-zero, we get the latter behaviour. */
+#ifdef CONFIG_LP_CONSOLE_STRICT
+# define CONSOLE_LP_STRICT 1
+#else
+# define CONSOLE_LP_STRICT 0
+#endif

  /* The console must be locked when we get here. */

@@ -697,7 +701,7 @@ static void lp_console_write (struct con
  {
  	struct pardevice *dev = lp_table[CONSOLE_LP].dev;
  	struct parport *port = dev->port;
-	ssize_t written;
+	ssize_t written = 0;

  	if (parport_claim (dev))
  		/* Nothing we can do. */
--- linux-2.6.16-rc3/drivers/char/Kconfig.orig	2006-02-14 
00:14:08.000000000 +0100
+++ linux-2.6.16-rc3/drivers/char/Kconfig	2006-02-14 13:47:33.000000000 
+0100
@@ -512,14 +512,21 @@ config LP_CONSOLE
  	  doing that; to actually get it to happen you need to pass the
  	  option "console=lp0" to the kernel at boot time.

-	  If the printer is out of paper (or off, or unplugged, or too
-	  busy..), the kernel will stall until the printer is ready again.
-	  By defining CONSOLE_LP_STRICT to 0 (at your own risk) you
-	  can make the kernel continue when this happens,
-	  but it'll lose the kernel messages.
-
  	  If unsure, say N.

+config LP_CONSOLE_STRICT
+	bool "Wait for a ready printer"
+	depends on LP_CONSOLE
+	default y
+	---help---
+	  With this option enabled, if the printer is out of paper (or off,
+	  or unplugged, or too busy..) the kernel will stall until the printer
+	  is ready again. By turning this option off (at your own risk), you
+	  can make the kernel continue when this happens, but it will lose
+	  some kernel messages.
+	
+	  If unsure, say Y
+
  config PPDEV
  	tristate "Support for user-space parallel port device drivers"
  	depends on PARPORT



I attach a copy of this patch to this email in order to avoid these 
problems (at least I hope so)

Regards,

-- 
Luca

[-- Attachment #2: lp_console_strict.patch.txt --]
[-- Type: text/plain, Size: 2167 bytes --]

--- linux-2.6.16-rc3/drivers/char/lp.c.orig	2006-01-08 16:48:14.000000000 +0100
+++ linux-2.6.16-rc3/drivers/char/lp.c	2006-02-14 13:43:41.000000000 +0100
@@ -686,9 +686,13 @@ static struct file_operations lp_fops = 
 #define CONSOLE_LP 0
 
 /* If the printer is out of paper, we can either lose the messages or
- * stall until the printer is happy again.  Define CONSOLE_LP_STRICT
- * non-zero to get the latter behaviour. */
-#define CONSOLE_LP_STRICT 1
+ * stall until the printer is happy again. If CONSOLE_LP_STRICT is
+ * non-zero, we get the latter behaviour. */
+#ifdef CONFIG_LP_CONSOLE_STRICT
+# define CONSOLE_LP_STRICT 1
+#else
+# define CONSOLE_LP_STRICT 0
+#endif
 
 /* The console must be locked when we get here. */
 
@@ -697,7 +701,7 @@ static void lp_console_write (struct con
 {
 	struct pardevice *dev = lp_table[CONSOLE_LP].dev;
 	struct parport *port = dev->port;
-	ssize_t written;
+	ssize_t written = 0;
 
 	if (parport_claim (dev))
 		/* Nothing we can do. */
--- linux-2.6.16-rc3/drivers/char/Kconfig.orig	2006-02-14 00:14:08.000000000 +0100
+++ linux-2.6.16-rc3/drivers/char/Kconfig	2006-02-14 13:47:33.000000000 +0100
@@ -512,14 +512,21 @@ config LP_CONSOLE
 	  doing that; to actually get it to happen you need to pass the
 	  option "console=lp0" to the kernel at boot time.
 
-	  If the printer is out of paper (or off, or unplugged, or too
-	  busy..), the kernel will stall until the printer is ready again.
-	  By defining CONSOLE_LP_STRICT to 0 (at your own risk) you
-	  can make the kernel continue when this happens,
-	  but it'll lose the kernel messages.
-
 	  If unsure, say N.
 
+config LP_CONSOLE_STRICT
+	bool "Wait for a ready printer"
+	depends on LP_CONSOLE
+	default y
+	---help---
+	  With this option enabled, if the printer is out of paper (or off,
+	  or unplugged, or too busy..) the kernel will stall until the printer
+	  is ready again. By turning this option off (at your own risk), you
+	  can make the kernel continue when this happens, but it will lose
+	  some kernel messages.
+	  
+	  If unsure, say Y
+
 config PPDEV
 	tristate "Support for user-space parallel port device drivers"
 	depends on PARPORT