2012-10-26 11:39:44

by Sekhar Nori

[permalink] [raw]
Subject: Re: [PATCH] davinci: check for presence of channel controller on slot alloc

+ Matt, who is doing the DMA engine conversion for EDMA

On 9/12/2012 11:44 PM, Cyril Chemparathy wrote:
> This patch adds a check for the presence of the channel controller when
> trying to allocate a slot. Without this fix, the kernel panics with a NULL
> pointer dereference when the dma-engine drivers are probed.
>
> Signed-off-by: Cyril Chemparathy <[email protected]>
> ---
> arch/arm/mach-davinci/dma.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/arch/arm/mach-davinci/dma.c b/arch/arm/mach-davinci/dma.c
> index a685e97..2fee31e 100644
> --- a/arch/arm/mach-davinci/dma.c
> +++ b/arch/arm/mach-davinci/dma.c
> @@ -743,6 +743,9 @@ EXPORT_SYMBOL(edma_free_channel);
> */
> int edma_alloc_slot(unsigned ctlr, int slot)
> {
> + if (!edma_cc[ctlr])
> + return -ENODEV;

I am ok with this patch, although I wonder if a better error is -ENXIO
or even -EINVAL (since its most likely the result of a wrong argument).

This patch will conflict with the EDMA movement series that Matt is
doing and he should probably include it in his series to avoid
conflicts. From the description it appears to be not required for v3.7
anyway.

Thanks,
Sekhar


2012-10-26 11:45:26

by Matt Porter

[permalink] [raw]
Subject: Re: [PATCH] davinci: check for presence of channel controller on slot alloc

On Fri, Oct 26, 2012 at 05:09:17PM +0530, Sekhar Nori wrote:
> + Matt, who is doing the DMA engine conversion for EDMA
>
> On 9/12/2012 11:44 PM, Cyril Chemparathy wrote:
> > This patch adds a check for the presence of the channel controller when
> > trying to allocate a slot. Without this fix, the kernel panics with a NULL
> > pointer dereference when the dma-engine drivers are probed.
> >
> > Signed-off-by: Cyril Chemparathy <[email protected]>
> > ---
> > arch/arm/mach-davinci/dma.c | 3 +++
> > 1 file changed, 3 insertions(+)
> >
> > diff --git a/arch/arm/mach-davinci/dma.c b/arch/arm/mach-davinci/dma.c
> > index a685e97..2fee31e 100644
> > --- a/arch/arm/mach-davinci/dma.c
> > +++ b/arch/arm/mach-davinci/dma.c
> > @@ -743,6 +743,9 @@ EXPORT_SYMBOL(edma_free_channel);
> > */
> > int edma_alloc_slot(unsigned ctlr, int slot)
> > {
> > + if (!edma_cc[ctlr])
> > + return -ENODEV;
>
> I am ok with this patch, although I wonder if a better error is -ENXIO
> or even -EINVAL (since its most likely the result of a wrong argument).
>
> This patch will conflict with the EDMA movement series that Matt is
> doing and he should probably include it in his series to avoid
> conflicts. From the description it appears to be not required for v3.7
> anyway.

I'll add this to the am33xx dmaengine series with -EINVAL, thanks.

-Matt