[PATCH 1/3] mtd: create a partition type device tree binding

Brian Norris computersforpeace at gmail.com
Thu Oct 29 09:29:08 PDT 2015

+ devicetree-spec and others, since other non-driver-specific MTD
partitioning stuff has gotten directed there [1], and this binding
should probably get promoted in short order to the core mtdpart.c

On Thu, Oct 29, 2015 at 01:52:30PM +0100, Linus Walleij wrote:
> The Linux code in drivers/mtd/maps/physmap_of.c will optionally
> look for the linux,part-probe binding for hints on a partition type to
> look for in the MTD. It was added in commit 9d5da3a9b849
> "mtd: extend physmap_of to let the device tree specify the parition probe"
> This solution is too Linux-specific and depend on some
> Linux kernel-internal naming conventions. We need a proper
> way of describing partition types that follow the pattern set by
> other device tree bindings.
> Create a "partition-type" binding for this, and add "my" bindings
> for "arm,arm-flash-structure" as a starter for others to follow.
> A follow-on patch adds support to the Linux kernel to handle this
> binding, with some infrastructure for others to use it too.

Overall, I like this. And thanks for expanding the explanation of DT
partitions vs. parsers. It'd be good if we make parsers like this easier
to integrate, so we see less hard-coded partitions in device tree.

> Cc: devicetree at vger.kernel.org
> Cc: Jason Gunthorpe <jgunthorpe at obsidianresearch.com>
> Cc: Liviu Dudau <Liviu.Dudau at arm.com>
> Reported-by: Liviu Dudau <Liviu.Dudau at arm.com>
> Signed-off-by: Linus Walleij <linus.walleij at linaro.org>
> ---
>  .../devicetree/bindings/mtd/mtd-physmap.txt        |  2 ++
>  .../devicetree/bindings/mtd/partition.txt          | 35 +++++++++++++++++++---
>  2 files changed, 33 insertions(+), 4 deletions(-)
> diff --git a/Documentation/devicetree/bindings/mtd/mtd-physmap.txt b/Documentation/devicetree/bindings/mtd/mtd-physmap.txt
> index 4a0a48bf4ecb..863560bdbb19 100644
> --- a/Documentation/devicetree/bindings/mtd/mtd-physmap.txt
> +++ b/Documentation/devicetree/bindings/mtd/mtd-physmap.txt
> @@ -23,6 +23,8 @@ file systems on embedded devices.
>     unaligned accesses as implemented in the JFFS2 code via memcpy().
>     By defining "no-unaligned-direct-access", the flash will not be
>     exposed directly to the MTD users (e.g. JFFS2) any more.
> + - partition-type : a flash partition type to expect and probe for
> +   on this device. See "partition.txt" for available partition types.
>   - linux,mtd-name: allow to specify the mtd name for retro capability with
>     physmap-flash drivers as boot loader pass the mtd partition via the old
>     device name physmap-flash.
> diff --git a/Documentation/devicetree/bindings/mtd/partition.txt b/Documentation/devicetree/bindings/mtd/partition.txt
> index 8e5557da1955..85d45764a4b3 100644
> --- a/Documentation/devicetree/bindings/mtd/partition.txt
> +++ b/Documentation/devicetree/bindings/mtd/partition.txt
> @@ -1,9 +1,36 @@
>  Representing flash partitions in devicetree
> -Partitions can be represented by sub-nodes of an mtd device. This can be used
> -on platforms which have strong conventions about which portions of a flash are
> -used for what purposes, but which don't use an on-flash partition table such
> -as RedBoot.
> +On-device partition types:
> +
> +It is possible for some drivers to indicate an on-device partition type, i.e.
> +partition tables, footers or other binary pattern in the flash used to
> +define how the flash is partitioned. This can be done in the device tree node
> +of an MTD device by specifying partition-type = "foo"; This tells the operating
> +system to look for the partition type indicated.
> +
> +Required properties:
> +- partition-type : the type of partition. Only one type can be specified.

You're supporting this as a list property (for future expansion,
presumably), so I can only assume the "only one type" is referring to
the number of different parsers available currently, not the behavior of
the property itself? This should probably be worded differently if
that's the intention. Like, "a list of partition types ... ordered by
priority .... We currently support the following: (list only the AFS
type)" (fill in the blanks how you'd like).

Also, this patch will conflict with [1]. I'll probably take [1] soon, so
one of us will have to rebase this.


[1] http://thread.gmane.org/gmane.comp.devicetree.spec/191

> +  Valid types are:
> +  "arm,arm-flash-structure" (also called AFS)
> +
> +Example:
> +
> +flash0 at 40000000 {
> +	/* 2 * 32MiB NOR Flash memory */
> +	compatible = "arm,vexpress-flash", "cfi-flash";
> +	partition-type = "arm,arm-flash-structure";
> +	reg = <0x40000000 0x04000000>;
> +	bank-width = <4>;
> +};
> +
> +
> +Device Tree specified partitions:
> +
> +If there is no specified on-device binary format, partitions can be
> +represented by sub-nodes of an mtd device. This can be used on platforms which
> +have strong conventions about which portions of a flash are used for what
> +purposes.
> +
>  NOTE: if the sub-node has a compatible string, then it is not a partition.
>  #address-cells & #size-cells must both be present in the mtd device. There are
> -- 
> 2.4.3

More information about the linux-arm-kernel mailing list