[PATCH v1 1/4] davinci: da8xx/omap-l1: add support for SPI
Michael Williamson
michael.williamson at criticallink.com
Wed Feb 2 08:21:24 EST 2011
On 2/2/2011 8:04 AM, Sergei Shtylyov wrote:
> Hello.
>
> On 02-02-2011 15:55, Michael Williamson wrote:
>
>>>> Add SPI registration routines, clocks, and driver resources for
>>>> DA850/OMAP-L138/AM18x and DA830/OMAP-L137/AM17x platforms.
>
>>> You're only adding clocks for the former platform -- the ones for the latter platfrom are already there...
>
>>>> Signed-off-by: Michael Williamson<michael.williamson at criticallink.com>
>>> [...]
>
>>>> diff --git a/arch/arm/mach-davinci/devices-da8xx.c b/arch/arm/mach-davinci/devices-da8xx.c
>>>> index beda8a4..f421f97 100644
>>>> --- a/arch/arm/mach-davinci/devices-da8xx.c
>>>> +++ b/arch/arm/mach-davinci/devices-da8xx.c
>>>> @@ -725,3 +725,99 @@ int __init da8xx_register_cpuidle(void)
>>>>
>>>> return platform_device_register(&da8xx_cpuidle_device);
>>>> }
>>>> +
>>>> +static struct resource da8xx_spi0_resources[] = {
>>>> + [0] = {
>>>> + .start = 0x01c41000,
>>>> + .end = 0x01c41fff,
>>>> + .flags = IORESOURCE_MEM,
>>>> + },
>>>> + [1] = {
>>>> + .start = IRQ_DA8XX_SPINT0,
>>>> + .end = IRQ_DA8XX_SPINT0,
>>>> + .flags = IORESOURCE_IRQ,
>>>> + },
>>>> + [2] = {
>>>> + .start = EDMA_CTLR_CHAN(0, 14),
>>>> + .end = EDMA_CTLR_CHAN(0, 14),
>>>> + .flags = IORESOURCE_DMA,
>>>> + },
>>>> + [3] = {
>>>> + .start = EDMA_CTLR_CHAN(0, 15),
>>>> + .end = EDMA_CTLR_CHAN(0, 15),
>>>> + .flags = IORESOURCE_DMA,
>>>> + },
>
>>> We have DA830_DMACH_SPI0_[RT]X defined for SPI0 DMA channels...
>
>> Right. The current routines/structures were designed to support either da830 or
>> da850. They work because the interrupts and the DMA channels are fortunately
>> the same for the SPI devices between the platforms. I can use the DA830 enums,
>> but I had preferred the clarity of the EDMA_CTRL_CHAN macro, and it would
>> remove confusion that this might only apply for DA830 and not DA850. It would
>> be nice to rename that DA8XX_DMAC_SPI_[RT]X if the enums were used...
>
>> Can I leave this, or do you really want me to use the enums?
>
> OK, you can leave this as is...
>
>>>> + [4] = {
>>>> + .flags = IORESOURCE_DMA,
>>>> + },
>
>>> BTW, why do you need this placeholder?
>
>> This is for the eventq selection in the edma allocation by the spi driver. When I
>> previously had 0, I should have used EVENTQ_0 (how it was tested). I will add it
>> back in to clarification.
>
> But event queue is not a DMA channel, so can't have just IORESOURCE_DMA in 'flags'. IIUC, there was a patch to add some clarifying flag(s) to IORESOURCE_DMA to support the event queue resource (which I never approved :-).
>
I'd like to get the support in there for the driver that's currently in
the mainline, which needs this resource as it is. I had essentially copied
the setup approach from the dm365.c board file, which is doing the same thing.
It looks like other devices take EVENTQ information as platform data (e.g.,
snd_platform_data takes .asp_chan_q). Would it make sense to update the
spi platform data to accept an EVENTQ parameter as well? I would be willing
to push a follow on patch to sort this out, if it is required.
Thanks.
> WBR, Sergei
More information about the linux-arm-kernel
mailing list