[PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty

[PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty

[Jun Chen]

This patch check whether the fifo lenth is empty before writing new data to fifo.If condition
is true,ifx_spi_write need to trigger one mrdy_assert. If condition is false,the mrdy_assert
will be trigger by the next ifx_spi_io.
Cc: Bi Chao <chao.bi@intel.com>
Signed-off-by: Chen Jun <jun.d.chen@intel.com>
---
 drivers/tty/serial/ifx6x60.c |   14 +++++++++++---
 1 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/drivers/tty/serial/ifx6x60.c b/drivers/tty/serial/ifx6x60.c
index 5b9bc19..aa01989 100644
--- a/drivers/tty/serial/ifx6x60.c
+++ b/drivers/tty/serial/ifx6x60.c
@@ -469,9 +469,17 @@ static int ifx_spi_write(struct tty_struct *tty, const unsigned char *buf,
 {
 	struct ifx_spi_device *ifx_dev = tty->driver_data;
 	unsigned char *tmp_buf = (unsigned char *)buf;
-	int tx_count = kfifo_in_locked(&ifx_dev->tx_fifo, tmp_buf, count,
-				   &ifx_dev->fifo_lock);
-	mrdy_assert(ifx_dev);
+	int tx_count;
+	unsigned long flags;
+	bool is_fifo_empty;
+
+	spin_lock_irqsave(&ifx_dev->fifo_lock, flags);
+	is_fifo_empty = kfifo_is_empty(&ifx_dev->tx_fifo);
+	tx_count = kfifo_in(&ifx_dev->tx_fifo, tmp_buf, count);
+	spin_unlock_irqrestore(&ifx_dev->fifo_lock, flags);
+	if (is_fifo_empty)
+		mrdy_assert(ifx_dev);
+
 	return tx_count;
 }
 
-- 
1.7.4.1

[PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty.

[Jun Chen]

This patch check whether the fifo lenth is empty before writing new data to fifo.If condition
is true,ifx_spi_write need to trigger one mrdy_assert. If condition is false,the mrdy_assert
will be trigger by the next ifx_spi_io.
Cc: Bi Chao <chao.bi@intel.com>
Signed-off-by: Chen Jun <jun.d.chen@intel.com>
---
 drivers/tty/serial/ifx6x60.c |   14 +++++++++++---
 1 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/drivers/tty/serial/ifx6x60.c b/drivers/tty/serial/ifx6x60.c
index 5b9bc19..aa01989 100644
--- a/drivers/tty/serial/ifx6x60.c
+++ b/drivers/tty/serial/ifx6x60.c
@@ -469,9 +469,17 @@ static int ifx_spi_write(struct tty_struct *tty, const unsigned char *buf,
 {
 	struct ifx_spi_device *ifx_dev = tty->driver_data;
 	unsigned char *tmp_buf = (unsigned char *)buf;
-	int tx_count = kfifo_in_locked(&ifx_dev->tx_fifo, tmp_buf, count,
-				   &ifx_dev->fifo_lock);
-	mrdy_assert(ifx_dev);
+	int tx_count;
+	unsigned long flags;
+	bool is_fifo_empty;
+
+	spin_lock_irqsave(&ifx_dev->fifo_lock, flags);
+	is_fifo_empty = kfifo_is_empty(&ifx_dev->tx_fifo);
+	tx_count = kfifo_in(&ifx_dev->tx_fifo, tmp_buf, count);
+	spin_unlock_irqrestore(&ifx_dev->fifo_lock, flags);
+	if (is_fifo_empty)
+		mrdy_assert(ifx_dev);
+
 	return tx_count;
 }
 
-- 
1.7.4.1

Re: [PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty.

[Greg KH]

On Fri, Nov 16, 2012 at 06:08:37AM -0500, Jun Chen wrote:
> 
> This patch check whether the fifo lenth is empty before writing new data to fifo.If condition
> is true,ifx_spi_write need to trigger one mrdy_assert. If condition is false,the mrdy_assert
> will be trigger by the next ifx_spi_io.
> 
> Cc: Bi Chao <chao.bi@intel.com>
> Signed-off-by: Chen Jun <jun.d.chen@intel.com>
> ---
>  drivers/tty/serial/ifx6x60.c |   14 +++++++++++---
>  1 files changed, 11 insertions(+), 3 deletions(-)

What tree did you make this against?  It doesn't apply to my tty-next
tree at all :(

greg k-h

RE: [PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty.

[Chen, Jun D]

Hi,
   I make this patch on your tty-next tree,  Do I need to make it on the branch 3.7-rc6 against?

-----Original Message-----
From: Greg KH [mailto:gregkh@linuxfoundation.org] 
Sent: Friday, November 16, 2012 8:39 PM
To: Chen, Jun D
Cc: Alan Cox; Bi, Chao; serial; Gorby, Russ; linux-kernel@vger.kernel.org; Wu, Fengguang
Subject: Re: [PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty.

On Fri, Nov 16, 2012 at 06:08:37AM -0500, Jun Chen wrote:
> 
> This patch check whether the fifo lenth is empty before writing new 
> data to fifo.If condition is true,ifx_spi_write need to trigger one 
> mrdy_assert. If condition is false,the mrdy_assert will be trigger by the next ifx_spi_io.
> 
> Cc: Bi Chao <chao.bi@intel.com>
> Signed-off-by: Chen Jun <jun.d.chen@intel.com>
> ---
>  drivers/tty/serial/ifx6x60.c |   14 +++++++++++---
>  1 files changed, 11 insertions(+), 3 deletions(-)

What tree did you make this against?  It doesn't apply to my tty-next tree at all :(

greg k-h

Re: [PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty.

[Greg KH]

http://daringfireball.net/2007/07/on_top

On Sun, Nov 18, 2012 at 03:14:28AM +0000, Chen, Jun D wrote:
> Hi,
>    I make this patch on your tty-next tree,  Do I need to make it on the branch 3.7-rc6 against?

No, my tty-next branch.  But something is odd, it didn't apply at all.
Please try again.

thanks,

greg k-h

[PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty

[Jun Chen]

This patch check whether the fifo lenth is empty before writing new data to fifo.If condition
is true,ifx_spi_write need to trigger one mrdy_assert. If condition is false,the mrdy_assert
will be trigger by the next ifx_spi_io.
Cc: Bi Chao <chao.bi@intel.com>
Signed-off-by: Chen Jun <jun.d.chen@intel.com>
---
 drivers/tty/serial/ifx6x60.c |   13 ++++++++++---
 1 files changed, 10 insertions(+), 3 deletions(-)

diff --git a/drivers/tty/serial/ifx6x60.c b/drivers/tty/serial/ifx6x60.c
index a688f8f..4a21f5e 100644
--- a/drivers/tty/serial/ifx6x60.c
+++ b/drivers/tty/serial/ifx6x60.c
@@ -520,9 +520,16 @@ static int ifx_spi_write(struct tty_struct *tty, const unsigned char *buf,
 {
 	struct ifx_spi_device *ifx_dev = tty->driver_data;
 	unsigned char *tmp_buf = (unsigned char *)buf;
-	int tx_count = kfifo_in_locked(&ifx_dev->tx_fifo, tmp_buf, count,
-				   &ifx_dev->fifo_lock);
-	mrdy_assert(ifx_dev);
+	unsigned long flags;
+	bool is_fifo_empty;
+
+	spin_lock_irqsave(&ifx_dev->fifo_lock, flags);
+	is_fifo_empty = kfifo_is_empty(&ifx_dev->tx_fifo);
+	int tx_count = kfifo_in(&ifx_dev->tx_fifo, tmp_buf, count);
+	spin_unlock_irqrestore(&ifx_dev->fifo_lock, flags);
+	if (is_fifo_empty)
+		mrdy_assert(ifx_dev);
+
 	return tx_count;
 }
 
-- 
1.7.4.1

Re: [PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty

[Alan Cox]

On Wed, 07 Nov 2012 12:04:04 -0500
Jun Chen <jun.d.chen@intel.com> wrote:

> 
> This patch check whether the fifo lenth is empty before writing new data to fifo.If condition
> is true,ifx_spi_write need to trigger one mrdy_assert. If condition is false,the mrdy_assert
> will be trigger by the next ifx_spi_io.
> Cc: Bi Chao <chao.bi@intel.com>
> Signed-off-by: Chen Jun <jun.d.chen@intel.com>

Acked-by: Alan Cox <alan@linux.intel.com>

[PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty.

[Jun Chen]

This patch check whether the kfifo lenth equal to the count of write data.If condition
is true,ifx_spi_write need to trigger one mrdy_assert. If condition is false,the mrdy_assert
will be triggered by the next ifx_spi_io.
Cc: Bi Chao <chao.bi@intel.com>
Signed-off-by: Chen Jun <jun.d.chen@intel.com>
---
 drivers/tty/serial/ifx6x60.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/drivers/tty/serial/ifx6x60.c b/drivers/tty/serial/ifx6x60.c
index ac718e1..b52ef8d 100644
--- a/drivers/tty/serial/ifx6x60.c
+++ b/drivers/tty/serial/ifx6x60.c
@@ -522,7 +522,8 @@ static int ifx_spi_write(struct tty_struct *tty, const unsigned char *buf,
 	unsigned char *tmp_buf = (unsigned char *)buf;
 	int tx_count = kfifo_in_locked(&ifx_dev->tx_fifo, tmp_buf, count,
 				   &ifx_dev->fifo_lock);
-	mrdy_assert(ifx_dev);
+	if (kfifo_len(&ifx_dev->tx_fifo) == count)
+		mrdy_assert(ifx_dev);
 	return tx_count;
 }
 
-- 
1.7.4.1

Re: [PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty.

[Alan Cox]

On Tue, 06 Nov 2012 04:28:54 -0500
Jun Chen <jun.d.chen@intel.com> wrote:

> 
> This patch check whether the kfifo lenth equal to the count of write data.If condition
> is true,ifx_spi_write need to trigger one mrdy_assert. If condition is false,the mrdy_assert
> will be triggered by the next ifx_spi_io.
> Cc: Bi Chao <chao.bi@intel.com>
> Signed-off-by: Chen Jun <jun.d.chen@intel.com>
> ---
>  drivers/tty/serial/ifx6x60.c |    3 ++-
>  1 files changed, 2 insertions(+), 1 deletions(-)
> 
> diff --git a/drivers/tty/serial/ifx6x60.c b/drivers/tty/serial/ifx6x60.c
> index ac718e1..b52ef8d 100644
> --- a/drivers/tty/serial/ifx6x60.c
> +++ b/drivers/tty/serial/ifx6x60.c
> @@ -522,7 +522,8 @@ static int ifx_spi_write(struct tty_struct *tty, const unsigned char *buf,
>  	unsigned char *tmp_buf = (unsigned char *)buf;
>  	int tx_count = kfifo_in_locked(&ifx_dev->tx_fifo, tmp_buf, count,
>  				   &ifx_dev->fifo_lock);
> -	mrdy_assert(ifx_dev);
> +	if (kfifo_len(&ifx_dev->tx_fifo) == count)
> +		mrdy_assert(ifx_dev);
>  	return tx_count;

I'm not convinced this is safe. What stops the sending task from emptying
data out of the fifo between the kfifo_in_locked and the kfifo_len check ?

Can't you just check tx_count ?

Alan

RE: [PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty.

[Chen, Jun D]

Hi,Alan
	Thank for your question, I will update this patch in other email. 
 
-----Original Message-----
From: Alan Cox [mailto:alan@lxorguk.ukuu.org.uk] 
Sent: Tuesday, November 06, 2012 5:47 PM
To: Chen, Jun D
Cc: Alan Cox; serial; Gorby, Russ; Bi, Chao
Subject: Re: [PATCH] serial: ifx6x60: ifx_spi_write don't need to do mrdy_assert when fifo is not empty.

On Tue, 06 Nov 2012 04:28:54 -0500
Jun Chen <jun.d.chen@intel.com> wrote:

> 
> This patch check whether the kfifo lenth equal to the count of write 
> data.If condition is true,ifx_spi_write need to trigger one 
> mrdy_assert. If condition is false,the mrdy_assert will be triggered by the next ifx_spi_io.
> Cc: Bi Chao <chao.bi@intel.com>
> Signed-off-by: Chen Jun <jun.d.chen@intel.com>
> ---
>  drivers/tty/serial/ifx6x60.c |    3 ++-
>  1 files changed, 2 insertions(+), 1 deletions(-)
> 
> diff --git a/drivers/tty/serial/ifx6x60.c 
> b/drivers/tty/serial/ifx6x60.c index ac718e1..b52ef8d 100644
> --- a/drivers/tty/serial/ifx6x60.c
> +++ b/drivers/tty/serial/ifx6x60.c
> @@ -522,7 +522,8 @@ static int ifx_spi_write(struct tty_struct *tty, const unsigned char *buf,
>  	unsigned char *tmp_buf = (unsigned char *)buf;
>  	int tx_count = kfifo_in_locked(&ifx_dev->tx_fifo, tmp_buf, count,
>  				   &ifx_dev->fifo_lock);
> -	mrdy_assert(ifx_dev);
> +	if (kfifo_len(&ifx_dev->tx_fifo) == count)
> +		mrdy_assert(ifx_dev);
>  	return tx_count;

I'm not convinced this is safe. What stops the sending task from emptying data out of the fifo between the kfifo_in_locked and the kfifo_len check ?

Can't you just check tx_count ?

Alan