Re: [PATCH v5 4/4] misc: sram: add support for configurableallocation order

[Philipp Zabel]

Am Mittwoch, den 14.11.2012, 19:15 +0000 schrieb Grant Likely:
> On Thu, 18 Oct 2012 16:27:33 +0200, Philipp Zabel <p.zabel@xxxxxxxxxxxxxx> wrote:
> > From: Matt Porter <mporter@xxxxxx>
> > 
> > Adds support for setting the genalloc pool's minimum allocation
> > order via DT or platform data. The allocation order is optional
> > for both the DT property and platform data case. If it is not
> > present then the order defaults to PAGE_SHIFT to preserve the
> > current behavior.
> > 
> > Signed-off-by: Matt Porter <mporter@xxxxxx>
> > Signed-off-by: Philipp Zabel <p.zabel@xxxxxxxxxxxxxx>
> > ---
> >  Documentation/devicetree/bindings/misc/sram.txt |   12 ++++++++++-
> >  drivers/misc/sram.c                             |   14 ++++++++++++-
> >  include/linux/platform_data/sram.h              |   25 +++++++++++++++++++++++
> >  3 files changed, 49 insertions(+), 2 deletions(-)
> >  create mode 100644 include/linux/platform_data/sram.h
> > 
> > diff --git a/Documentation/devicetree/bindings/misc/sram.txt b/Documentation/devicetree/bindings/misc/sram.txt
> > index b64136c..b1705ec 100644
> > --- a/Documentation/devicetree/bindings/misc/sram.txt
> > +++ b/Documentation/devicetree/bindings/misc/sram.txt
> > @@ -8,10 +8,20 @@ Required properties:
> >  
> >  - reg : SRAM iomem address range
> >  
> > -Example:
> > +Optional properties:
> > +
> > +- alloc-order : Minimum allocation order for the SRAM pool
> 
> Looks okay, but I think the property name is confusing. I for one had
> no idea what 'order' would be and why it was important. I had to read
> the code to figure it out.
> 
> It does raise the question though of what is this binding actually
> for? Does it reflect a limitation of the SRAM? or of the hardware using
> the SRAM? Or is it an optimization? How do you expect to use it?

If I am not mistaken, it is about the expected use case. A driver
allocating many small buffers would quickly fill small SRAMs if the
allocations were of PAGE_SIZE granularity.

I wonder if a common allocation size (say, 512 bytes instead of
PAGE_SIZE) can be found that every prospective user could be reasonably
happy with?

> Assuming it is appropriate to put into the device tree, I'd suggest a
> different name. Instead of 'order', how about 'sram-alloc-align' (in
> address bits) or 'sram-alloc-min-size' (in bytes).

A size in bytes would be the most obvious to me, although that allows to
enter values that are not a power of two.

regards
Philipp


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/