serial: jsm: Use min_t instead of min

[PATCH] serial: jsm: Use min_t instead of min

Posted by Mohammad Mahfooz 2 years, 6 months ago

Use min_t instead of min to cut down n further if needed.

Signed-off-by: Mohammad Mahfooz <mohammadmahfoozpersonal@gmail.com>
---
 drivers/tty/serial/jsm/jsm_neo.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/tty/serial/jsm/jsm_neo.c b/drivers/tty/serial/jsm/jsm_neo.c
index 0c78f66276cd..4cce1e423b06 100644
--- a/drivers/tty/serial/jsm/jsm_neo.c
+++ b/drivers/tty/serial/jsm/jsm_neo.c
@@ -350,7 +350,7 @@ static void neo_copy_data_from_uart_to_queue(struct jsm_channel *ch)
 		 * IBM pSeries platform.
 		 * 15 bytes max appears to be the magic number.
 		 */
-		n = min((u32) n, (u32) 12);
+		n = min_t(u32, n, 12);
 
 		/*
 		 * Since we are grabbing the linestatus register, which
-- 
2.34.1

Re: [PATCH] serial: jsm: Use min_t instead of min

Posted by Jiri Slaby 2 years, 6 months ago

On 19. 02. 23, 21:53, Mohammad Mahfooz wrote:
> Use min_t instead of min to cut down n further if needed.
> 
> Signed-off-by: Mohammad Mahfooz <mohammadmahfoozpersonal@gmail.com>
> ---
>   drivers/tty/serial/jsm/jsm_neo.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/tty/serial/jsm/jsm_neo.c b/drivers/tty/serial/jsm/jsm_neo.c
> index 0c78f66276cd..4cce1e423b06 100644
> --- a/drivers/tty/serial/jsm/jsm_neo.c
> +++ b/drivers/tty/serial/jsm/jsm_neo.c
> @@ -350,7 +350,7 @@ static void neo_copy_data_from_uart_to_queue(struct jsm_channel *ch)
>   		 * IBM pSeries platform.
>   		 * 15 bytes max appears to be the magic number.
>   		 */
> -		n = min((u32) n, (u32) 12);
> +		n = min_t(u32, n, 12);

Nah, why is "n" signed in the first place? Could you fix that and all 
all those casts in min()s around instead?

-- 
js
suse labs

RE: [PATCH] serial: jsm: Use min_t instead of min

Posted by David Laight 2 years, 6 months ago

From: Jiri Slaby
> Sent: 20 February 2023 11:52
> 
> On 19. 02. 23, 21:53, Mohammad Mahfooz wrote:
> > Use min_t instead of min to cut down n further if needed.
> >
> > Signed-off-by: Mohammad Mahfooz <mohammadmahfoozpersonal@gmail.com>
> > ---
> >   drivers/tty/serial/jsm/jsm_neo.c | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/tty/serial/jsm/jsm_neo.c b/drivers/tty/serial/jsm/jsm_neo.c
> > index 0c78f66276cd..4cce1e423b06 100644
> > --- a/drivers/tty/serial/jsm/jsm_neo.c
> > +++ b/drivers/tty/serial/jsm/jsm_neo.c
> > @@ -350,7 +350,7 @@ static void neo_copy_data_from_uart_to_queue(struct jsm_channel *ch)
> >   		 * IBM pSeries platform.
> >   		 * 15 bytes max appears to be the magic number.
> >   		 */
> > -		n = min((u32) n, (u32) 12);
> > +		n = min_t(u32, n, 12);
> 
> Nah, why is "n" signed in the first place? Could you fix that and all
> all those casts in min()s around instead?

There is also a (IIRC) 'fifo_space -= 3;' in there that could
also generate negatives.

I took one look at that function and ran away.

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)

RE: [PATCH] serial: jsm: Use min_t instead of min

Posted by Ilpo Järvinen 2 years, 6 months ago

On Tue, 21 Feb 2023, David Laight wrote:

> From: Jiri Slaby
> > Sent: 20 February 2023 11:52
> > 
> > On 19. 02. 23, 21:53, Mohammad Mahfooz wrote:
> > > Use min_t instead of min to cut down n further if needed.
> > >
> > > Signed-off-by: Mohammad Mahfooz <mohammadmahfoozpersonal@gmail.com>
> > > ---
> > >   drivers/tty/serial/jsm/jsm_neo.c | 2 +-
> > >   1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/tty/serial/jsm/jsm_neo.c b/drivers/tty/serial/jsm/jsm_neo.c
> > > index 0c78f66276cd..4cce1e423b06 100644
> > > --- a/drivers/tty/serial/jsm/jsm_neo.c
> > > +++ b/drivers/tty/serial/jsm/jsm_neo.c
> > > @@ -350,7 +350,7 @@ static void neo_copy_data_from_uart_to_queue(struct jsm_channel *ch)
> > >   		 * IBM pSeries platform.
> > >   		 * 15 bytes max appears to be the magic number.
> > >   		 */
> > > -		n = min((u32) n, (u32) 12);
> > > +		n = min_t(u32, n, 12);
> > 
> > Nah, why is "n" signed in the first place? Could you fix that and all
> > all those casts in min()s around instead?
> 
> There is also a (IIRC) 'fifo_space -= 3;' in there that could
> also generate negatives.
> 
> I took one look at that function and ran away.

min_t() casts before compare so it would never work correctly if a 
negative number is fed into min_t(u32, ...).

Not that negative FIFO "space" makes any sense (and should be prevent 
at its source to begin with).

-- 
 i.

[PATCH] serial: jsm: Change n to unsigned int

Posted by Mohammad Mahfooz 2 years, 6 months ago

Change n to an unsigned int and remove casts from min()s.

Signed-off-by: Mohammad Mahfooz <mohammadmahfoozpersonal@gmail.com>
---
 drivers/tty/serial/jsm/jsm_neo.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/tty/serial/jsm/jsm_neo.c b/drivers/tty/serial/jsm/jsm_neo.c
index 4cce1e423b06..394121870436 100644
--- a/drivers/tty/serial/jsm/jsm_neo.c
+++ b/drivers/tty/serial/jsm/jsm_neo.c
@@ -277,7 +277,7 @@ static void neo_copy_data_from_uart_to_queue(struct jsm_channel *ch)
 	int qleft = 0;
 	u8 linestatus = 0;
 	u8 error_mask = 0;
-	int n = 0;
+	u32 n = 0;
 	int total = 0;
 	u16 head;
 	u16 tail;
@@ -342,15 +342,15 @@ static void neo_copy_data_from_uart_to_queue(struct jsm_channel *ch)
 			break;
 
 		/* Make sure we don't go over the end of our queue */
-		n = min(((u32) total), (RQUEUESIZE - (u32) head));
+		n = min(total, RQUEUESIZE - head);
 
 		/*
 		 * Cut down n even further if needed, this is to fix
 		 * a problem with memcpy_fromio() with the Neo on the
 		 * IBM pSeries platform.
-		 * 15 bytes max appears to be the magic number.
+		 * 12 bytes max appears to be the magic number.
 		 */
-		n = min_t(u32, n, 12);
+		n = min(n, 12U);
 
 		/*
 		 * Since we are grabbing the linestatus register, which
-- 
2.34.1

Re: [PATCH] serial: jsm: Change n to unsigned int

Posted by Jiri Slaby 2 years, 6 months ago

On 21. 02. 23, 2:03, Mohammad Mahfooz wrote:
> Change n to an unsigned int and remove casts from min()s.
> 
> Signed-off-by: Mohammad Mahfooz <mohammadmahfoozpersonal@gmail.com>
> ---
>   drivers/tty/serial/jsm/jsm_neo.c | 8 ++++----
>   1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/tty/serial/jsm/jsm_neo.c b/drivers/tty/serial/jsm/jsm_neo.c
> index 4cce1e423b06..394121870436 100644
> --- a/drivers/tty/serial/jsm/jsm_neo.c
> +++ b/drivers/tty/serial/jsm/jsm_neo.c
> @@ -277,7 +277,7 @@ static void neo_copy_data_from_uart_to_queue(struct jsm_channel *ch)
>   	int qleft = 0;
>   	u8 linestatus = 0;
>   	u8 error_mask = 0;
> -	int n = 0;
> +	u32 n = 0;
>   	int total = 0;
>   	u16 head;
>   	u16 tail;
> @@ -342,15 +342,15 @@ static void neo_copy_data_from_uart_to_queue(struct jsm_channel *ch)
>   			break;
>   
>   		/* Make sure we don't go over the end of our queue */
> -		n = min(((u32) total), (RQUEUESIZE - (u32) head));
> +		n = min(total, RQUEUESIZE - head);

This should warn now, right as total is signed too, right? And qleft. 
All should likely be unsigned. And the computation of qleft would need 
to be changed.

It's not as easy as this.

>   
>   		/*
>   		 * Cut down n even further if needed, this is to fix
>   		 * a problem with memcpy_fromio() with the Neo on the
>   		 * IBM pSeries platform.
> -		 * 15 bytes max appears to be the magic number.
> +		 * 12 bytes max appears to be the magic number.
>   		 */
> -		n = min_t(u32, n, 12);
> +		n = min(n, 12U);
>   
>   		/*
>   		 * Since we are grabbing the linestatus register, which

-- 
js
suse labs