[U-Boot] [PATCH] [x86] Fix some bugs in the i8402 driver when no controller is present

[U-Boot] [PATCH] [x86] Fix some bugs in the i8402 driver when no controller is present

[Gabe Black]

If no controller is present, the i8402 driver should return immediately and
not attempt to operate on the missing hardware.

In kbd_input_empty, the status register is checked every millisecond to see
whether the input buffer is empty, up to a timeout which is tracked by
decrimenting a counter each time the check is performed. The decrement is
performed with a postfix -- operator, and the value of the counter is
checked in place. That means that when the counter reaches zero and the
loop terminates, it will actually be decrimented one more time and become
-1. That value is returned as the return value of the function. That would
give the right answer if it wasn't for that extra decrement because a
timeout would indicate that the buffer never became empty.

This change fixes both of those bugs.

Signed-off-by: Gabe Black <gabeblack@chromium.org>
---
 drivers/input/i8042.c |   12 +++++++++++-
 1 files changed, 11 insertions(+), 1 deletions(-)

diff --git a/drivers/input/i8042.c b/drivers/input/i8042.c
index 58094c9..3bdfa7d 100644
--- a/drivers/input/i8042.c
+++ b/drivers/input/i8042.c
@@ -315,6 +315,13 @@ static unsigned char ext_key_map [] =
     0x00    /* map end */
     };
 
+/******************************************************************************/
+
+static int kbd_controller_present (void)
+{
+    return in8(I8042_STATUS_REG) != 0xff;
+}
+
 /*******************************************************************************
  *
  * i8042_kbd_init - reset keyboard and init state flags
@@ -324,6 +331,9 @@ int i8042_kbd_init (void)
     int keymap, try;
     char *penv;
 
+    if (!kbd_controller_present())
+        return -1;
+
 #ifdef CONFIG_USE_CPCIDVI
     if ((penv = getenv ("console")) != NULL) {
 	    if (strncmp (penv, "serial", 7) == 0) {
@@ -629,7 +639,7 @@ static int kbd_input_empty (void)
     while ((in8 (I8042_STATUS_REG) & 0x02) && kbdTimeout--)
 	udelay(1000);
 
-    return kbdTimeout;
+    return kbdTimeout != -1;
 }
 
 /******************************************************************************/
-- 
1.7.3.1

[U-Boot] [PATCH] [x86] Fix some bugs in the i8402 driver when no controller is present

[Graeme Russ]

Hi Gabe,

On 08/11/11 20:48, Gabe Black wrote:
> If no controller is present, the i8402 driver should return immediately and
> not attempt to operate on the missing hardware.
> 
> In kbd_input_empty, the status register is checked every millisecond to see
> whether the input buffer is empty, up to a timeout which is tracked by
> decrimenting a counter each time the check is performed. The decrement is
> performed with a postfix -- operator, and the value of the counter is
> checked in place. That means that when the counter reaches zero and the
> loop terminates, it will actually be decrimented one more time and become
> -1. That value is returned as the return value of the function. That would
> give the right answer if it wasn't for that extra decrement because a
> timeout would indicate that the buffer never became empty.
> 
> This change fixes both of those bugs.
> 
> Signed-off-by: Gabe Black <gabeblack@chromium.org>
> ---
>  drivers/input/i8042.c |   12 +++++++++++-
>  1 files changed, 11 insertions(+), 1 deletions(-)
> 

total: 1 errors, 5 warnings, 30 lines checked

Patch is not checkpatch clean - Can you please fix checkpatch issues and
change tag to 'x86:' style and resubmit

Thanks,

Graeme

[U-Boot] [PATCH] [x86] Fix some bugs in the i8402 driver when no controller is present

[Gabe Black]

On Sat, Nov 12, 2011 at 2:26 AM, Graeme Russ <graeme.russ@gmail.com> wrote:

> Hi Gabe,
>
> On 08/11/11 20:48, Gabe Black wrote:
> > If no controller is present, the i8402 driver should return immediately
> and
> > not attempt to operate on the missing hardware.
> >
> > In kbd_input_empty, the status register is checked every millisecond to
> see
> > whether the input buffer is empty, up to a timeout which is tracked by
> > decrimenting a counter each time the check is performed. The decrement is
> > performed with a postfix -- operator, and the value of the counter is
> > checked in place. That means that when the counter reaches zero and the
> > loop terminates, it will actually be decrimented one more time and become
> > -1. That value is returned as the return value of the function. That
> would
> > give the right answer if it wasn't for that extra decrement because a
> > timeout would indicate that the buffer never became empty.
> >
> > This change fixes both of those bugs.
> >
> > Signed-off-by: Gabe Black <gabeblack@chromium.org>
> > ---
> >  drivers/input/i8042.c |   12 +++++++++++-
> >  1 files changed, 11 insertions(+), 1 deletions(-)
> >
>
> total: 1 errors, 5 warnings, 30 lines checked
>
> Patch is not checkpatch clean - Can you please fix checkpatch issues and
> change tag to 'x86:' style and resubmit
>
> Thanks,
>
> Graeme
>
>
checkpatch doesn't like it because I was consistent with the existing
incorrect style in the file. There are three or four options for how to
handle this.

1. Match the current style and ignore checkpatch (what I've already done).
2. Fix the style on just the lines I've modified.
3. Fix the style within a "local" area around my changes, for some
definition of local, along with my changes.
4. Submit a separate cleanup patch which fixes the style first.

What is the standard way of approaching this sort of situation? I'm ok with
option 1 if you are since I already have that ready to go, and I'd be fine
with 4 if you'd like to clean up this file first like you've been cleaning
up some others. 2 and 3 would make the style in that file inconsistent and
harder to read, but if that's standard practice I'll do that.

Gabe

[U-Boot] [PATCH] [x86] Fix some bugs in the i8402 driver when no controller is present

[Graeme Russ]

Hi Gabe,

On 13/11/11 13:52, Gabe Black wrote:
> 
> 
> On Sat, Nov 12, 2011 at 2:26 AM, Graeme Russ <graeme.russ@gmail.com
> <mailto:graeme.russ@gmail.com>> wrote:
> 
>     Hi Gabe,
> 
>     On 08/11/11 20:48, Gabe Black wrote:
>     > If no controller is present, the i8402 driver should return
>     immediately and
>     > not attempt to operate on the missing hardware.
>     >
>     > In kbd_input_empty, the status register is checked every millisecond
>     to see
>     > whether the input buffer is empty, up to a timeout which is tracked by
>     > decrimenting a counter each time the check is performed. The decrement is
>     > performed with a postfix -- operator, and the value of the counter is
>     > checked in place. That means that when the counter reaches zero and the
>     > loop terminates, it will actually be decrimented one more time and become
>     > -1. That value is returned as the return value of the function. That
>     would
>     > give the right answer if it wasn't for that extra decrement because a
>     > timeout would indicate that the buffer never became empty.
>     >
>     > This change fixes both of those bugs.
>     >
>     > Signed-off-by: Gabe Black <gabeblack@chromium.org
>     <mailto:gabeblack@chromium.org>>
>     > ---
>     >  drivers/input/i8042.c |   12 +++++++++++-
>     >  1 files changed, 11 insertions(+), 1 deletions(-)
>     >
> 
>     total: 1 errors, 5 warnings, 30 lines checked
> 
>     Patch is not checkpatch clean - Can you please fix checkpatch issues and
>     change tag to 'x86:' style and resubmit
> 
>     Thanks,
> 
>     Graeme
> 
> 
> checkpatch doesn't like it because I was consistent with the existing
> incorrect style in the file. There are three or four options for how to
> handle this.
> 
> 1. Match the current style and ignore checkpatch (what I've already done).
> 2. Fix the style on just the lines I've modified.
> 3. Fix the style within a "local" area around my changes, for some
> definition of local, along with my changes.
> 4. Submit a separate cleanup patch which fixes the style first.
> 
> What is the standard way of approaching this sort of situation? I'm ok with
> option 1 if you are since I already have that ready to go, and I'd be fine
> with 4 if you'd like to clean up this file first like you've been cleaning
> up some others. 2 and 3 would make the style in that file inconsistent and
> harder to read, but if that's standard practice I'll do that.

I think the standard we are now adopting (enforcing) is #4 - A two patch
series consisting of a checkpatch cleanup of the file and checkpatch clean
patch for the changes. Of course there are exceptions:

 - The checkpatch cleanup would be too onerous (touches too many files,
   the patched file is too large, too close to release) and the bugfix is
   critical
 - The bugfix covers too many files (again,
 - The file is from another repository (typically Linux) and the patch has
   been applied to that repository (in which case, a commit reference is
   required in the patch summary)

I'm happy for Wolfgang to correct me

Regards,

Graeme

[U-Boot] [PATCH] [x86] Fix some bugs in the i8402 driver when no controller is present

[Graeme Russ]

On 13/11/11 20:17, Graeme Russ wrote:
> Hi Gabe,

[snip]

>  - The checkpatch cleanup would be too onerous (touches too many files,
>    the patched file is too large, too close to release) and the bugfix is
>    critical
>  - The bugfix covers too many files (again,

Oops - forgot to proof-read ;)

Regards,

Graeme

[U-Boot] [PATCH v2] x86: Fix some bugs in the i8402 driver when no controller is present

[Gabe Black]

If no controller is present, the i8402 driver should return immediately and
not attempt to operate on the missing hardware.

In kbd_input_empty, the status register is checked every millisecond to see
whether the input buffer is empty, up to a timeout which is tracked by
decrimenting a counter each time the check is performed. The decrement is
performed with a postfix -- operator, and the value of the counter is
checked in place. That means that when the counter reaches zero and the
loop terminates, it will actually be decrimented one more time and become
-1. That value is returned as the return value of the function. That would
give the right answer if it wasn't for that extra decrement because a
timeout would indicate that the buffer never became empty.

This change fixes both of those bugs.

Signed-off-by: Gabe Black <gabeblack@chromium.org>
---
Changes in v2:
- Change summary tag style.
- Rebase onto checkpatch cleaned file.

 drivers/input/i8042.c |   12 +++++++++++-
 1 files changed, 11 insertions(+), 1 deletions(-)

diff --git a/drivers/input/i8042.c b/drivers/input/i8042.c
index 83b1bf4..c3bc536 100644
--- a/drivers/input/i8042.c
+++ b/drivers/input/i8042.c
@@ -313,6 +313,13 @@ static unsigned char ext_key_map[] = {
 	0x00  /* map end */
 	};
 
+/******************************************************************************/
+
+static int kbd_controller_present(void)
+{
+	return in8(I8042_STATUS_REG) != 0xff;
+}
+
 /*******************************************************************************
  *
  * i8042_kbd_init - reset keyboard and init state flags
@@ -322,6 +329,9 @@ int i8042_kbd_init(void)
 	int keymap, try;
 	char *penv;
 
+	if (!kbd_controller_present())
+		return -1;
+
 #ifdef CONFIG_USE_CPCIDVI
 	penv = getenv("console");
 	if (penv != NULL) {
@@ -603,7 +613,7 @@ static int kbd_input_empty(void)
 	while ((in8(I8042_STATUS_REG) & 0x02) && kbdTimeout--)
 		udelay(1000);
 
-	return kbdTimeout;
+	return kbdTimeout != -1;
 }
 
 /******************************************************************************/
-- 
1.7.3.1

[U-Boot] [PATCH v2] x86: Fix some bugs in the i8402 driver when no controller is present

[Graeme Russ]

Hi Gabe,

On 15/11/11 17:18, Gabe Black wrote:
> If no controller is present, the i8402 driver should return immediately and
> not attempt to operate on the missing hardware.
> 
> In kbd_input_empty, the status register is checked every millisecond to see
> whether the input buffer is empty, up to a timeout which is tracked by
> decrimenting a counter each time the check is performed. The decrement is
> performed with a postfix -- operator, and the value of the counter is
> checked in place. That means that when the counter reaches zero and the
> loop terminates, it will actually be decrimented one more time and become
> -1. That value is returned as the return value of the function. That would
> give the right answer if it wasn't for that extra decrement because a
> timeout would indicate that the buffer never became empty.
> 
> This change fixes both of those bugs.
> 
> Signed-off-by: Gabe Black <gabeblack@chromium.org>
> ---
> Changes in v2:
> - Change summary tag style.
> - Rebase onto checkpatch cleaned file.
> 
>  drivers/input/i8042.c |   12 +++++++++++-
>  1 files changed, 11 insertions(+), 1 deletions(-)

Applied to u-boot-x86/master

Thanks,

Graeme