2015-12-01 19:58:56

by Mark Brown

[permalink] [raw]
Subject: Re: [PATCH v5 08/10] spi: expose master transfer size limitation.

On Tue, Dec 01, 2015 at 04:51:06PM -0000, Michal Suchanek wrote:
> On some SPI controllers it is not feasible to transfer arbitrary amount
> of data at once.
>
> When the limit on transfer size is a few kilobytes at least it makes
> sense to use the SPI hardware rather than reverting to gpio driver.

> + /*
> + * on some hardware transfer size may be constrained
> + * the limit may depend on device transfer settings
> + */
> + size_t (*max_transfer_size)(struct spi_device *spi);

Heiner submitted a *very* similar patch just now with a straight
variable plus accessor instead of a function and using a name with _msg.
I'm ambivalent on the implementation but prefer the naming here since
that's more the limitation we're trying to express I think (some
hardware does have limiations about multple transfers too). Can the two
of you come up with something that works for both of you?


Attachments:
(No filename) (893.00 B)
signature.asc (473.00 B)
Download all attachments

2015-12-01 21:08:09

by Heiner Kallweit

[permalink] [raw]
Subject: Re: [PATCH v5 08/10] spi: expose master transfer size limitation.

Am 01.12.2015 um 20:58 schrieb Mark Brown:
> On Tue, Dec 01, 2015 at 04:51:06PM -0000, Michal Suchanek wrote:
>> On some SPI controllers it is not feasible to transfer arbitrary amount
>> of data at once.
>>
>> When the limit on transfer size is a few kilobytes at least it makes
>> sense to use the SPI hardware rather than reverting to gpio driver.
>
>> + /*
>> + * on some hardware transfer size may be constrained
>> + * the limit may depend on device transfer settings
>> + */
>> + size_t (*max_transfer_size)(struct spi_device *spi);
>
> Heiner submitted a *very* similar patch just now with a straight
> variable plus accessor instead of a function and using a name with _msg.
> I'm ambivalent on the implementation but prefer the naming here since
> that's more the limitation we're trying to express I think (some
> hardware does have limiations about multple transfers too). Can the two
> of you come up with something that works for both of you?
>
Sure .. Just one inquiry:
When you say "the naming here" you refer to Michal's or my version?

Actually I like in Michal's hook that it directly takes a struct spi_device.
This saves the caller one level of indirection as the caller usually will
deal with a spi_device and not a spi_master.

If you're fine with Michal's version then this is also fine with me,
especially as the functionality is the same.

2015-12-01 21:32:29

by Mark Brown

[permalink] [raw]
Subject: Re: [PATCH v5 08/10] spi: expose master transfer size limitation.

On Tue, Dec 01, 2015 at 10:07:55PM +0100, Heiner Kallweit wrote:

> Sure .. Just one inquiry:
> When you say "the naming here" you refer to Michal's or my version?

Michael's (transfer).

> Actually I like in Michal's hook that it directly takes a struct spi_device.
> This saves the caller one level of indirection as the caller usually will
> deal with a spi_device and not a spi_master.

> If you're fine with Michal's version then this is also fine with me,
> especially as the functionality is the same.

OK.


Attachments:
(No filename) (514.00 B)
signature.asc (473.00 B)
Download all attachments

2015-12-02 09:56:01

by Michal Suchanek

[permalink] [raw]
Subject: Re: [PATCH v5 08/10] spi: expose master transfer size limitation.

On 1 December 2015 at 20:58, Mark Brown <[email protected]> wrote:
> On Tue, Dec 01, 2015 at 04:51:06PM -0000, Michal Suchanek wrote:
>> On some SPI controllers it is not feasible to transfer arbitrary amount
>> of data at once.
>>
>> When the limit on transfer size is a few kilobytes at least it makes
>> sense to use the SPI hardware rather than reverting to gpio driver.
>
>> + /*
>> + * on some hardware transfer size may be constrained
>> + * the limit may depend on device transfer settings
>> + */
>> + size_t (*max_transfer_size)(struct spi_device *spi);
>
> Heiner submitted a *very* similar patch just now with a straight
> variable plus accessor instead of a function and using a name with _msg.
> I'm ambivalent on the implementation but prefer the naming here since
> that's more the limitation we're trying to express I think (some
> hardware does have limiations about multple transfers too). Can the two
> of you come up with something that works for both of you?

Sorry, missed there is a patch because it shows in the middle of the
discussion for me.

I don't really care which one it is so long as I can get the last
patch in this series based on it.

Thanks

Michal