Re: [PATCH] arm: make FORCE_MAX_ZONEORDER configurable for TI AM33XX

Dear Tony



Thank you for applying this patch. I am quite new in such things as is
posting patched to the community, therefore Yegor help me on this manner.



Anyway, I have one more challenging question/request.


I and my colleagues have developed a brand new very enhanced board having
AM35xx CPU, 512MB DDR2, DVI, true dual Ethernet support (not using PRUSS),
CAN, SPI, I2C, 7 USBs, etc…Board is already materialized and operational.


According to our board we have create new machine type in the kernel
(version: linux-3.2-psp04.06.00.07.sdk) and all other required stuff to
support our board, modify some things, like adding support for dual eth,
etc. Our goal is to provide this new board to the market (let say with
large quantity) and therefore I am wondering how to start or if it is even
possible to transfer our kernel change the main stream of this distro?
During this kernel modification process we have been looking to not
changing anything in kernel, what could affect other machine types.



Thank you for you answer, and looking forward to receive your replay.





Best regards,
Dejan


2012/9/25 Yegor Yefremov [off-list ref]

 On 25.09.2012 10:14, Russell King - ARM Linux wrote:

quoted

On Tue, Sep 25, 2012 at 10:05:30AM +0200, Yegor Yefremov wrote:

quoted

How should I change the patch to make it proper? SA1111 is broken

anyway:

quoted

No it isn't.  This is what it produces _today_, and has done for the last
5-10 years without modification

# CONFIG_CLEANCACHE is not set
# CONFIG_FRONTSWAP is not set
CONFIG_FORCE_MAX_ZONEORDER=9
CONFIG_LEDS=y
CONFIG_LEDS_CPU=y

quoted

config FORCE_MAX_ZONEORDER
        int "Maximum zone order" if ARCH_SHMOBILE
        range 11 64 if ARCH_SHMOBILE
        default "9" if SA1111
        default "11"

AFAIK if ARCH_SHMOBILE defines dependency on ARCH_SHMOBILE,

No it doens't.

      int "Maximum zone order" if ARCH_SHMOBILE

is far from being the same as:

      int "Maximum zone order"
      depends on ARCH_SHMOBILE

The former defines a condition upon which the option is offered in GUIs -
or to put it another way, it defines the visibility of the option.

The latter defines a dependency which must be met for the option to be
both visible and appear in the resulting configuration file.

quoted

so SA1111 won't be evaluated (at least if I select SA1111

And did you check that SA1111 remains selected?  I bet you didn't.  Or
maybe you tested your patched version.  Whatever.  The original works,
and has been known to work for years.  Your patch breaks it.  It's
really as simple as that.

Thanks for explanation. I think I've got it now. Please review the v2
version.

Yegor