[PATCH 3.7] spi/bcm63xx: fix transfer bits_per_words check

[PATCH 3.7] spi/bcm63xx: fix transfer bits_per_words check

[Jonas Gorski]

Transfers often do not have bits_per_words set, so use the spi device's
bits_per_words in this case.

This fixes the driver rejecting valid transfers e.g. generated by
spi_write() or spi_read().

Cc: stable-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Signed-off-by: Jonas Gorski <jonas.gorski-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
---
 drivers/spi/spi-bcm63xx.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/spi/spi-bcm63xx.c b/drivers/spi/spi-bcm63xx.c
index a9f4049..d7c2916 100644
--- a/drivers/spi/spi-bcm63xx.c
+++ b/drivers/spi/spi-bcm63xx.c
@@ -103,7 +103,8 @@ static int bcm63xx_spi_check_transfer(struct spi_device *spi,
 {
 	u8 bits_per_word;
 
-	bits_per_word = (t) ? t->bits_per_word : spi->bits_per_word;
+	bits_per_word = (t && t->bits_per_word) ?
+			t->bits_per_word : spi->bits_per_word;
 	if (bits_per_word != 8) {
 		dev_err(&spi->dev, "%s, unsupported bits_per_word=%d\n",
 			__func__, bits_per_word);
-- 
1.7.10.4


------------------------------------------------------------------------------
Monitor your physical, virtual and cloud infrastructure from a single
web console. Get in-depth insight into apps, servers, databases, vmware,
SAP, cloud infrastructure, etc. Download 30-day Free Trial.
Pricing starts from $795 for 25 servers or applications!
http://p.sf.net/sfu/zoho_dev2dev_nov

Re: [PATCH 3.7] spi/bcm63xx: fix transfer bits_per_words check

[Florian Fainelli]

Le 24/11/2012 14:44, Jonas Gorski a écrit :
> Transfers often do not have bits_per_words set, so use the spi device's
> bits_per_words in this case.
>
> This fixes the driver rejecting valid transfers e.g. generated by
> spi_write() or spi_read().
>
> Cc: stable-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> Signed-off-by: Jonas Gorski <jonas.gorski-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>

Acked-by: Florian Fainelli <florian-p3rKhJxN3npAfugRpC6u6w@public.gmane.org>

> ---
>   drivers/spi/spi-bcm63xx.c |    3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/spi/spi-bcm63xx.c b/drivers/spi/spi-bcm63xx.c
> index a9f4049..d7c2916 100644
> --- a/drivers/spi/spi-bcm63xx.c
> +++ b/drivers/spi/spi-bcm63xx.c
> @@ -103,7 +103,8 @@ static int bcm63xx_spi_check_transfer(struct spi_device *spi,
>   {
>   	u8 bits_per_word;
>
> -	bits_per_word = (t) ? t->bits_per_word : spi->bits_per_word;
> +	bits_per_word = (t && t->bits_per_word) ?
> +			t->bits_per_word : spi->bits_per_word;
>   	if (bits_per_word != 8) {
>   		dev_err(&spi->dev, "%s, unsupported bits_per_word=%d\n",
>   			__func__, bits_per_word);
>


------------------------------------------------------------------------------
Monitor your physical, virtual and cloud infrastructure from a single
web console. Get in-depth insight into apps, servers, databases, vmware,
SAP, cloud infrastructure, etc. Download 30-day Free Trial.
Pricing starts from $795 for 25 servers or applications!
http://p.sf.net/sfu/zoho_dev2dev_nov

Re: [PATCH 3.7] spi/bcm63xx: fix transfer bits_per_words check

[Jonas Gorski]

On 24 November 2012 18:28, Mark Brown
<broonie-yzvPICuk2AATkU/dhu1WVueM+bqZidxxQQ4Iyu8u01E@public.gmane.org> wrote:
> On Sat, Nov 24, 2012 at 02:44:50PM +0100, Jonas Gorski wrote:
>> Transfers often do not have bits_per_words set, so use the spi device's
>> bits_per_words in this case.
>>
>> This fixes the driver rejecting valid transfers e.g. generated by
>> spi_write() or spi_read().
>
> This is now done by the SPI core, there's no need for individual drivers
> to do this.

So that will go to stable kernels, too?

------------------------------------------------------------------------------
Monitor your physical, virtual and cloud infrastructure from a single
web console. Get in-depth insight into apps, servers, databases, vmware,
SAP, cloud infrastructure, etc. Download 30-day Free Trial.
Pricing starts from $795 for 25 servers or applications!
http://p.sf.net/sfu/zoho_dev2dev_nov

Re: [PATCH 3.7] spi/bcm63xx: fix transfer bits_per_words check

[Jonas Gorski]

On 24 November 2012 19:01, Mark Brown
<broonie-yzvPICuk2AATkU/dhu1WVueM+bqZidxxQQ4Iyu8u01E@public.gmane.org> wrote:
> On Sat, Nov 24, 2012 at 06:53:10PM +0100, Jonas Gorski wrote:
>> On 24 November 2012 18:28, Mark Brown
>
>> > This is now done by the SPI core, there's no need for individual drivers
>> > to do this.
>
>> So that will go to stable kernels, too?
>
> No, I hadn't sent it there.  I'd really expect that anyone using a
> release kernel would've noticed this if there were a problem, it's
> pretty obvious when it goes wrong, and I tend to be extremely
> conservative with changes to stable kernels especially framework ones.

Well, I noticed this particular problem on a 3.6 kernel (3.6.7 to be
exact), so by that definition at least one noticed, unless I don't
count as anyone ;-). So, how can I make this driver less broken for
release kernels? I somewhat doubt the framework change will be picked
up by release kernel maintainers.

------------------------------------------------------------------------------
Monitor your physical, virtual and cloud infrastructure from a single
web console. Get in-depth insight into apps, servers, databases, vmware,
SAP, cloud infrastructure, etc. Download 30-day Free Trial.
Pricing starts from $795 for 25 servers or applications!
http://p.sf.net/sfu/zoho_dev2dev_nov

Re: [PATCH 3.7] spi/bcm63xx: fix transfer bits_per_words check

[Florian Fainelli]

On Saturday 24 November 2012 19:19:34 Jonas Gorski wrote:
> On 24 November 2012 19:01, Mark Brown
> <broonie@opensource.wolfsonmicro.com> wrote:
> > On Sat, Nov 24, 2012 at 06:53:10PM +0100, Jonas Gorski wrote:
> >> On 24 November 2012 18:28, Mark Brown
> >
> >> > This is now done by the SPI core, there's no need for individual drivers
> >> > to do this.
> >
> >> So that will go to stable kernels, too?
> >
> > No, I hadn't sent it there.  I'd really expect that anyone using a
> > release kernel would've noticed this if there were a problem, it's
> > pretty obvious when it goes wrong, and I tend to be extremely
> > conservative with changes to stable kernels especially framework ones.
> 
> Well, I noticed this particular problem on a 3.6 kernel (3.6.7 to be
> exact), so by that definition at least one noticed, unless I don't
> count as anyone ;-). So, how can I make this driver less broken for
> release kernels? I somewhat doubt the framework change will be picked
> up by release kernel maintainers.

Considering it is a simple one-liner in spi-bcm63xx and that you don't want to
propagate a framework-level change back to -stable (which makes sense), don't
you want to pick that one?

At least two other users told me about this bug in private conversations.
--
Florian

Re: [PATCH 3.7] spi/bcm63xx: fix transfer bits_per_words check

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 895 bytes --]

On Mon, Nov 26, 2012 at 02:11:20PM +0100, Florian Fainelli wrote:
> On Saturday 24 November 2012 19:19:34 Jonas Gorski wrote:

> > Well, I noticed this particular problem on a 3.6 kernel (3.6.7 to be
> > exact), so by that definition at least one noticed, unless I don't
> > count as anyone ;-). So, how can I make this driver less broken for
> > release kernels? I somewhat doubt the framework change will be picked
> > up by release kernel maintainers.

> Considering it is a simple one-liner in spi-bcm63xx and that you don't want to
> propagate a framework-level change back to -stable (which makes sense), don't
> you want to pick that one?

> At least two other users told me about this bug in private conversations.

Feel free to submit it for stable but it doesn't seem sensible to apply
it to development kernels, we want to remove this code from individual
drivers not add new copies.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]