[PATCH 1/3] mtd: sh_flctl: fixup wait_for_completion_timeout return handling

[PATCH 1/3] mtd: sh_flctl: fixup wait_for_completion_timeout return handling

[Nicholas Mc Guire]

wait_for_completion_timeout() returns unsigned long not int so the check 
for <= should be == and the type unsigned long. This fixes up the return 
value handling and returns -ETIMEDOUT on timeout rather than 0 and 1 on 
on success rather than a more or less random remaining number of jiffies.

Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
---

call sites:
read_fiforeg,write_ec_fiforeg assume > 0 == success
and the comment in flctl_dma_fifo0_transfe states
        /* ret > 0 is success */
        return ret;
since it is only checking for > 0 in the call-sites
returning -ETIMEDOUT should be fine.

Patch was compile tested with ap325rxa_defconfig (implies
CONFIG_MTD_NAND_SH_FLCTL=y)

Patch is against 4.1-rc1 (localversion-next is -next-20150501)

 drivers/mtd/nand/sh_flctl.c |    9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/drivers/mtd/nand/sh_flctl.c b/drivers/mtd/nand/sh_flctl.c
index c3ce81c..4450864 100644
--- a/drivers/mtd/nand/sh_flctl.c
+++ b/drivers/mtd/nand/sh_flctl.c
@@ -354,6 +354,7 @@ static int flctl_dma_fifo0_transfer(struct sh_flctl *flctl, unsigned long *buf,
 	dma_cookie_t cookie = -EINVAL;
 	uint32_t reg;
 	int ret;
+	unsigned long time_left;
 
 	if (dir == DMA_FROM_DEVICE) {
 		chan = flctl->chan_fifo0_rx;
@@ -388,14 +389,16 @@ static int flctl_dma_fifo0_transfer(struct sh_flctl *flctl, unsigned long *buf,
 		goto out;
 	}
 
-	ret =
+	time_left =
 	wait_for_completion_timeout(&flctl->dma_complete,
 				msecs_to_jiffies(3000));
 
-	if (ret <= 0) {
+	if (time_left == 0) {
 		dmaengine_terminate_all(chan);
 		dev_err(&flctl->pdev->dev, "wait_for_completion_timeout\n");
-	}
+		ret = -ETIMEDOUT;
+	} else
+		ret = 1;	/* completion succeeded */
 
 out:
 	reg = readl(FLINTDMACR(flctl));
-- 
1.7.10.4

[PATCH 2/3] mtd: sh_flctl: fix alignment of function argument

[Nicholas Mc Guire]

This just is a minor coding style cleanup - align the function arguments
with the opening (.

Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
---

Patch was compile tested with ap325rxa_defconfig (implies
CONFIG_MTD_NAND_SH_FLCTL=y)

Patch is against 4.1-rc1 (localversion-next is -next-20150501)

 drivers/mtd/nand/sh_flctl.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/mtd/nand/sh_flctl.c b/drivers/mtd/nand/sh_flctl.c
index 4450864..b9f265a 100644
--- a/drivers/mtd/nand/sh_flctl.c
+++ b/drivers/mtd/nand/sh_flctl.c
@@ -391,7 +391,7 @@ static int flctl_dma_fifo0_transfer(struct sh_flctl *flctl, unsigned long *buf,
 
 	time_left =
 	wait_for_completion_timeout(&flctl->dma_complete,
-				msecs_to_jiffies(3000));
+				    msecs_to_jiffies(3000));
 
 	if (time_left == 0) {
 		dmaengine_terminate_all(chan);
-- 
1.7.10.4

Re: [PATCH 2/3] mtd: sh_flctl: fix alignment of function argument

[Laurent Pinchart]

Hi Nicholas,

Thank you for the patch.

On Friday 01 May 2015 16:16:02 Nicholas Mc Guire wrote:
> This just is a minor coding style cleanup - align the function arguments
> with the opening (.

I would just squash this into patch 1/3.

> Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
> ---
> 
> Patch was compile tested with ap325rxa_defconfig (implies
> CONFIG_MTD_NAND_SH_FLCTL=y)
> 
> Patch is against 4.1-rc1 (localversion-next is -next-20150501)
> 
>  drivers/mtd/nand/sh_flctl.c |    2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/mtd/nand/sh_flctl.c b/drivers/mtd/nand/sh_flctl.c
> index 4450864..b9f265a 100644
> --- a/drivers/mtd/nand/sh_flctl.c
> +++ b/drivers/mtd/nand/sh_flctl.c
> @@ -391,7 +391,7 @@ static int flctl_dma_fifo0_transfer(struct sh_flctl
> *flctl, unsigned long *buf,
> 
>  	time_left =
>  	wait_for_completion_timeout(&flctl->dma_complete,
> -				msecs_to_jiffies(3000));
> +				    msecs_to_jiffies(3000));
> 
>  	if (time_left == 0) {
>  		dmaengine_terminate_all(chan);

-- 
Regards,

Laurent Pinchart

[PATCH 3/3] mtd: sh_flctl: remove unused variable and assignment

[Nicholas Mc Guire]

Fixes a compile warning [-Wunused-but-set-variable] only.

Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
---

This fixes the compile time warning:
drivers/mtd/nand/sh_flctl.c: In function 'flctl_dma_fifo0_transfer':
drivers/mtd/nand/sh_flctl.c:354:15: warning: variable 'cookie' set but not
used [-Wunused-but-set-variable]

as dmaengine_submit only reads and returns desc->tx_submit(desc) which 
is then unused in flctl_dma_fifo0_transfer, this should be side-effect 
free to remove.

Patch was compile tested with ap325rxa_defconfig (implies
CONFIG_MTD_NAND_SH_FLCTL=y)

Patch is against 4.1-rc1 (localversion-next is -next-20150501)

 drivers/mtd/nand/sh_flctl.c |    2 --
 1 file changed, 2 deletions(-)

diff --git a/drivers/mtd/nand/sh_flctl.c b/drivers/mtd/nand/sh_flctl.c
index b9f265a..6a5c4f7 100644
--- a/drivers/mtd/nand/sh_flctl.c
+++ b/drivers/mtd/nand/sh_flctl.c
@@ -351,7 +351,6 @@ static int flctl_dma_fifo0_transfer(struct sh_flctl *flctl, unsigned long *buf,
 	struct dma_chan *chan;
 	enum dma_transfer_direction tr_dir;
 	dma_addr_t dma_addr;
-	dma_cookie_t cookie = -EINVAL;
 	uint32_t reg;
 	int ret;
 	unsigned long time_left;
@@ -377,7 +376,6 @@ static int flctl_dma_fifo0_transfer(struct sh_flctl *flctl, unsigned long *buf,
 
 		desc->callback = flctl_dma_complete;
 		desc->callback_param = flctl;
-		cookie = dmaengine_submit(desc);
 
 		dma_async_issue_pending(chan);
 	} else {
-- 
1.7.10.4

Re: [PATCH 3/3] mtd: sh_flctl: remove unused variable and assignment

[Laurent Pinchart]

Hi Nicholas,

Thank you for the patch.

On Friday 01 May 2015 16:16:03 Nicholas Mc Guire wrote:
> Fixes a compile warning [-Wunused-but-set-variable] only.
> 
> Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
> ---
> 
> This fixes the compile time warning:
> drivers/mtd/nand/sh_flctl.c: In function 'flctl_dma_fifo0_transfer':
> drivers/mtd/nand/sh_flctl.c:354:15: warning: variable 'cookie' set but not
> used [-Wunused-but-set-variable]
> 
> as dmaengine_submit only reads and returns desc->tx_submit(desc) which
> is then unused in flctl_dma_fifo0_transfer, this should be side-effect
> free to remove.

You can ignore its return value, but you can't remove the call to 
dmaengine_submit, as it's part of the DMA engine API and desc->tx_submit() 
does important work.

> Patch was compile tested with ap325rxa_defconfig (implies
> CONFIG_MTD_NAND_SH_FLCTL=y)
> 
> Patch is against 4.1-rc1 (localversion-next is -next-20150501)
> 
>  drivers/mtd/nand/sh_flctl.c |    2 --
>  1 file changed, 2 deletions(-)
> 
> diff --git a/drivers/mtd/nand/sh_flctl.c b/drivers/mtd/nand/sh_flctl.c
> index b9f265a..6a5c4f7 100644
> --- a/drivers/mtd/nand/sh_flctl.c
> +++ b/drivers/mtd/nand/sh_flctl.c
> @@ -351,7 +351,6 @@ static int flctl_dma_fifo0_transfer(struct sh_flctl
> *flctl, unsigned long *buf, struct dma_chan *chan;
>  	enum dma_transfer_direction tr_dir;
>  	dma_addr_t dma_addr;
> -	dma_cookie_t cookie = -EINVAL;
>  	uint32_t reg;
>  	int ret;
>  	unsigned long time_left;
> @@ -377,7 +376,6 @@ static int flctl_dma_fifo0_transfer(struct sh_flctl
> *flctl, unsigned long *buf,
> 
>  		desc->callback = flctl_dma_complete;
>  		desc->callback_param = flctl;
> -		cookie = dmaengine_submit(desc);
> 
>  		dma_async_issue_pending(chan);
>  	} else {

-- 
Regards,

Laurent Pinchart

Re: [PATCH 1/3] mtd: sh_flctl: fixup wait_for_completion_timeout return handling

[Laurent Pinchart]

Hi Nicholas,

Thank you for the patch.

On Friday 01 May 2015 16:16:01 Nicholas Mc Guire wrote:
> wait_for_completion_timeout() returns unsigned long not int so the check
> for <= should be == and the type unsigned long. This fixes up the return
> value handling and returns -ETIMEDOUT on timeout rather than 0 and 1 on
> on success rather than a more or less random remaining number of jiffies.
> 
> Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
> ---
> 
> call sites:
> read_fiforeg,write_ec_fiforeg assume > 0 == success
> and the comment in flctl_dma_fifo0_transfe states
>         /* ret > 0 is success */
>         return ret;
> since it is only checking for > 0 in the call-sites
> returning -ETIMEDOUT should be fine.
> 
> Patch was compile tested with ap325rxa_defconfig (implies
> CONFIG_MTD_NAND_SH_FLCTL=y)
> 
> Patch is against 4.1-rc1 (localversion-next is -next-20150501)
> 
>  drivers/mtd/nand/sh_flctl.c |    9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/mtd/nand/sh_flctl.c b/drivers/mtd/nand/sh_flctl.c
> index c3ce81c..4450864 100644
> --- a/drivers/mtd/nand/sh_flctl.c
> +++ b/drivers/mtd/nand/sh_flctl.c
> @@ -354,6 +354,7 @@ static int flctl_dma_fifo0_transfer(struct sh_flctl
> *flctl, unsigned long *buf, dma_cookie_t cookie = -EINVAL;
>  	uint32_t reg;
>  	int ret;
> +	unsigned long time_left;
> 
>  	if (dir == DMA_FROM_DEVICE) {
>  		chan = flctl->chan_fifo0_rx;
> @@ -388,14 +389,16 @@ static int flctl_dma_fifo0_transfer(struct sh_flctl
> *flctl, unsigned long *buf, goto out;
>  	}
> 
> -	ret =
> +	time_left =
>  	wait_for_completion_timeout(&flctl->dma_complete,
>  				msecs_to_jiffies(3000));
> 
> -	if (ret <= 0) {
> +	if (time_left == 0) {
>  		dmaengine_terminate_all(chan);
>  		dev_err(&flctl->pdev->dev, "wait_for_completion_timeout\n");
> -	}
> +		ret = -ETIMEDOUT;
> +	} else
> +		ret = 1;	/* completion succeeded */

I'd go one step further and make this function return < 0 on error and 0 on 
success, like usually done through the kernel API. You could then simplify the 
code using something like

        if (!wait_for_completion_timeout(&flctl->dma_complete,
                                         msecs_to_jiffies(3000))) {
                dmaengine_terminate_all(chan);
                dev_err(&flctl->pdev->dev, "wait_for_completion_timeout\n");
                ret = -ETIMEDOUT;
        }

(pre-initializing ret to 0)

> 
>  out:
>  	reg = readl(FLINTDMACR(flctl));

-- 
Regards,

Laurent Pinchart

Re: [PATCH 1/3] mtd: sh_flctl: fixup wait_for_completion_timeout return handling

[Nicholas Mc Guire]

On Sat, 02 May 2015, Laurent Pinchart wrote:

> Hi Nicholas,
> 
> Thank you for the patch.
> 
> On Friday 01 May 2015 16:16:01 Nicholas Mc Guire wrote:
> > wait_for_completion_timeout() returns unsigned long not int so the check
> > for <= should be == and the type unsigned long. This fixes up the return
> > value handling and returns -ETIMEDOUT on timeout rather than 0 and 1 on
> > on success rather than a more or less random remaining number of jiffies.
> > 
> > Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
> > ---
> > 
> > call sites:
> > read_fiforeg,write_ec_fiforeg assume > 0 == success
> > and the comment in flctl_dma_fifo0_transfe states
> >         /* ret > 0 is success */
> >         return ret;
> > since it is only checking for > 0 in the call-sites
> > returning -ETIMEDOUT should be fine.
> > 
> > Patch was compile tested with ap325rxa_defconfig (implies
> > CONFIG_MTD_NAND_SH_FLCTL=y)
> > 
> > Patch is against 4.1-rc1 (localversion-next is -next-20150501)
> > 
> >  drivers/mtd/nand/sh_flctl.c |    9 ++++++---
> >  1 file changed, 6 insertions(+), 3 deletions(-)
> > 
> > diff --git a/drivers/mtd/nand/sh_flctl.c b/drivers/mtd/nand/sh_flctl.c
> > index c3ce81c..4450864 100644
> > --- a/drivers/mtd/nand/sh_flctl.c
> > +++ b/drivers/mtd/nand/sh_flctl.c
> > @@ -354,6 +354,7 @@ static int flctl_dma_fifo0_transfer(struct sh_flctl
> > *flctl, unsigned long *buf, dma_cookie_t cookie = -EINVAL;
> >  	uint32_t reg;
> >  	int ret;
> > +	unsigned long time_left;
> > 
> >  	if (dir == DMA_FROM_DEVICE) {
> >  		chan = flctl->chan_fifo0_rx;
> > @@ -388,14 +389,16 @@ static int flctl_dma_fifo0_transfer(struct sh_flctl
> > *flctl, unsigned long *buf, goto out;
> >  	}
> > 
> > -	ret =
> > +	time_left =
> >  	wait_for_completion_timeout(&flctl->dma_complete,
> >  				msecs_to_jiffies(3000));
> > 
> > -	if (ret <= 0) {
> > +	if (time_left == 0) {
> >  		dmaengine_terminate_all(chan);
> >  		dev_err(&flctl->pdev->dev, "wait_for_completion_timeout\n");
> > -	}
> > +		ret = -ETIMEDOUT;
> > +	} else
> > +		ret = 1;	/* completion succeeded */
> 
> I'd go one step further and make this function return < 0 on error and 0 on 
> success, like usually done through the kernel API. You could then simplify the 
> code using something like
> 
>         if (!wait_for_completion_timeout(&flctl->dma_complete,
>                                          msecs_to_jiffies(3000))) {
>                 dmaengine_terminate_all(chan);
>                 dev_err(&flctl->pdev->dev, "wait_for_completion_timeout\n");
>                 ret = -ETIMEDOUT;
>         }
> 
> (pre-initializing ret to 0)
> 
> > 
> >  out:
> >  	reg = readl(FLINTDMACR(flctl));
>

ok - given that this needs some fixup anyway due to
not properly reading dmaengine_submit() - I'll fix
things up and re-submit.

thx!
hofrat