diff mbox

[v5,2/3] drivers: of: add initialization code for dma reserved memory

Message ID 1376049119-12655-3-git-send-email-m.szyprowski@samsung.com (mailing list archive)
State New, archived
Headers show

Commit Message

Marek Szyprowski Aug. 9, 2013, 11:51 a.m. UTC
Add device tree support for contiguous and reserved memory regions
defined in device tree. Initialization is done in 2 steps. First, the
memory is reserved, what happens very early when only flattened device
tree is available. Then on device initialization the corresponding cma
and reserved regions are assigned to each device structure.

Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Acked-by: Kyungmin Park <kyungmin.park@samsung.com>
Acked-by: Michal Nazarewicz <mina86@mina86.com>
---
 Documentation/devicetree/bindings/memory.txt |  152 ++++++++++++++++++++
 drivers/of/Kconfig                           |    6 +
 drivers/of/Makefile                          |    1 +
 drivers/of/of_reserved_mem.c                 |  197 ++++++++++++++++++++++++++
 drivers/of/platform.c                        |    5 +
 include/linux/of_reserved_mem.h              |   14 ++
 6 files changed, 375 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/memory.txt
 create mode 100644 drivers/of/of_reserved_mem.c
 create mode 100644 include/linux/of_reserved_mem.h

Comments

Rob Herring Aug. 10, 2013, 5:33 p.m. UTC | #1
On Fri, Aug 9, 2013 at 6:51 AM, Marek Szyprowski
<m.szyprowski@samsung.com> wrote:
> Add device tree support for contiguous and reserved memory regions
> defined in device tree. Initialization is done in 2 steps. First, the
> memory is reserved, what happens very early when only flattened device

s/what/which/

> tree is available. Then on device initialization the corresponding cma
> and reserved regions are assigned to each device structure.

What this commit message does not tell me is why does the reservation
have to happen before the fdt is unflattened? It would greatly
simplify the code if it didn't.

> Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
> Acked-by: Kyungmin Park <kyungmin.park@samsung.com>
> Acked-by: Michal Nazarewicz <mina86@mina86.com>
> ---
>  Documentation/devicetree/bindings/memory.txt |  152 ++++++++++++++++++++
>  drivers/of/Kconfig                           |    6 +
>  drivers/of/Makefile                          |    1 +
>  drivers/of/of_reserved_mem.c                 |  197 ++++++++++++++++++++++++++
>  drivers/of/platform.c                        |    5 +
>  include/linux/of_reserved_mem.h              |   14 ++
>  6 files changed, 375 insertions(+)
>  create mode 100644 Documentation/devicetree/bindings/memory.txt
>  create mode 100644 drivers/of/of_reserved_mem.c
>  create mode 100644 include/linux/of_reserved_mem.h
>
> diff --git a/Documentation/devicetree/bindings/memory.txt b/Documentation/devicetree/bindings/memory.txt
> new file mode 100644
> index 0000000..167d013
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/memory.txt
> @@ -0,0 +1,152 @@
> +*** Memory binding ***
> +
> +The /memory node provides basic information about the address and size
> +of the physical memory. This node is usually filled or updated by the
> +bootloader, depending on the actual memory configuration of the given
> +hardware.
> +
> +The memory layout is described by the folllowing node:

typo.

> +
> +memory {
> +       device_type = "memory";
> +       reg =  <(baseaddr1) (size1)
> +               (baseaddr2) (size2)
> +               ...
> +               (baseaddrN) (sizeN)>;
> +};
> +
> +baseaddrX:     the base address of the defined memory bank
> +sizeX:         the size of the defined memory bank
> +
> +More than one memory bank can be defined.
> +
> +
> +*** Reserved memory regions ***
> +
> +In /memory/reserved-memory node one can create additional nodes
> +describing particular reserved (excluded from normal use) memory
> +regions. Such memory regions are usually designed for the special usage
> +by various device drivers. A good example are contiguous memory
> +allocations or memory sharing with other operating system on the same
> +hardware board. Those special memory regions might depend on the board
> +configuration and devices used on the target system.
> +
> +Parameters for each memory region can be encoded into the device tree
> +wit the following convention:
> +
> +[(label):] (name)@(address) {
> +       compatible = "contiguous-memory-region", "reserved-memory-region";
> +       reg = <(address) (size)>;
> +       (linux,default-contiguous-region);
> +};
> +
> +label:         label given to the defined region (optional)
> +name:          an name given to the defined region
> +address:       the base address of the defined region
> +size:          the size of the memory region
> +
> +compatible:    "contiguous-memory-region" - enables binding of this
> +               region to Contiguous Memory Allocator (special region for
> +               contiguous memory allocations, shared with movable system
> +               memory, Linux kernel-specific), alternatively if
> +               "reserved-memory-region" - compatibility is defined, given
> +               region is assigned for exclusive usage for DMA transfers
> +
> +linux,default-contiguous-region: property indicating that the region
> +               is the default region for all contiguous memory
> +               allocations, Linux specific (optional)
> +
> +Each defined region must use unique name. It is optional to specify the
> +base address, so if one wants to use autoconfiguration of the base
> +address, he must specify the '0' as base address in the 'reg' property
> +and assign ann unique name to such regions.
> +
> +
> +*** Device node's properties ***
> +
> +Once the regions in the /memory/reserved-memory node are defined, they
> +can be assigned to device nodes to enable drivers for their special use.
> +The following properties are defined:
> +
> +memory-region = <&phandle_to_defined_region>;
> +
> +This property indicates that the device driver should use the
> +memory region pointed by the given phandle.
> +
> +
> +*** Example ***
> +
> +This example defines a memory consisting of 4 memory banks. 3 contiguous
> +regions are defined for Linux kernel, one default of all device drivers
> +(named contig_mem, placed at 0x72000000, 64MiB), one dedicated to the
> +framebuffer device (labelled display_mem, placed at 0x78000000, 8MiB)
> +and one for multimedia processing (labelled multimedia_mem, placed at
> +0x77000000, 64MiB). 'display_mem' region is then assigned to fb@12300000
> +device for DMA memory allocations (Linux kernel drivers will use CMA is
> +available or dma-exclusive usage otherwise). 'multimedia_mem' is
> +assigned to scaler@12500000 and codec@12600000 devices for contiguous
> +memory allocations when CMA driver is enabled.
> +
> +The reason for creating a separate region for framebuffer device is to
> +match the framebuffer base address to the one configured by bootloader,
> +so once Linux kernel drivers starts no glitches on the displayed boot
> +logo appears. Scaller and codec drivers should share the memory
> +allocations.
> +
> +/ {
> +       /* ... */
> +       memory {
> +               reg =  <0x40000000 0x10000000
> +                       0x50000000 0x10000000
> +                       0x60000000 0x10000000
> +                       0x70000000 0x10000000>;
> +
> +               reserved-memory {
> +                       #address-cells = <1>;
> +                       #size-cells = <1>;
> +
> +                       /*
> +                        * global autoconfigured region for contiguous allocations
> +                        * (used only with Contiguous Memory Allocator)
> +                        */
> +                       contig_region@0 {
> +                               compatible = "contiguous-memory-region";
> +                               reg = <0x0 0x4000000>;
> +                               linux,default-contiguous-region;
> +                       };
> +
> +                       /*
> +                        * special region for framebuffer
> +                        */
> +                       display_mem: region@78000000 {
> +                               compatible = "contiguous-memory-region", "reserved-memory-region";
> +                               reg = <0x78000000 0x800000>;
> +                       };
> +
> +                       /*
> +                        * special region for multimedia processing devices
> +                        */
> +                       multimedia_mem: region@77000000 {
> +                               compatible = "contiguous-memory-region";
> +                               reg = <0x77000000 0x4000000>;
> +                       };
> +               };
> +       };
> +
> +       /* ... */
> +
> +       fb0: fb@12300000 {
> +               status = "okay";
> +               memory-region = <&display_mem>;
> +       };
> +
> +       scaler: scaler@12500000 {
> +               status = "okay";
> +               memory-region = <&multimedia_mem>;
> +       };
> +
> +       codec: codec@12600000 {
> +               status = "okay";
> +               memory-region = <&multimedia_mem>;
> +       };
> +};
> diff --git a/drivers/of/Kconfig b/drivers/of/Kconfig
> index 80e5c13..a83ab43 100644
> --- a/drivers/of/Kconfig
> +++ b/drivers/of/Kconfig
> @@ -80,4 +80,10 @@ config OF_MTD
>         depends on MTD
>         def_bool y
>
> +config OF_RESERVED_MEM
> +       depends on CMA || (HAVE_GENERIC_DMA_COHERENT && HAVE_MEMBLOCK)
> +       def_bool y
> +       help
> +         Initialization code for DMA reserved memory
> +
>  endmenu # OF
> diff --git a/drivers/of/Makefile b/drivers/of/Makefile
> index 1f9c0c4..e7e3322 100644
> --- a/drivers/of/Makefile
> +++ b/drivers/of/Makefile
> @@ -10,3 +10,4 @@ obj-$(CONFIG_OF_MDIO) += of_mdio.o
>  obj-$(CONFIG_OF_PCI)   += of_pci.o
>  obj-$(CONFIG_OF_PCI_IRQ)  += of_pci_irq.o
>  obj-$(CONFIG_OF_MTD)   += of_mtd.o
> +obj-$(CONFIG_OF_RESERVED_MEM) += of_reserved_mem.o
> diff --git a/drivers/of/of_reserved_mem.c b/drivers/of/of_reserved_mem.c
> new file mode 100644
> index 0000000..b256b41
> --- /dev/null
> +++ b/drivers/of/of_reserved_mem.c
> @@ -0,0 +1,197 @@
> +/*
> + * Device tree based initialization code for reserved memory.
> + *
> + * Copyright (c) 2013 Samsung Electronics Co., Ltd.
> + *             http://www.samsung.com
> + * Author: Marek Szyprowski <m.szyprowski@samsung.com>
> + *
> + * This program is free software; you can redistribute it and/or
> + * modify it under the terms of the GNU General Public License as
> + * published by the Free Software Foundation; either version 2 of the
> + * License or (at your optional) any later version of the license.
> + */
> +
> +#include <asm/dma-contiguous.h>
> +
> +#include <linux/memblock.h>
> +#include <linux/err.h>
> +#include <linux/of.h>
> +#include <linux/of_fdt.h>
> +#include <linux/of_platform.h>
> +#include <linux/mm.h>
> +#include <linux/sizes.h>
> +#include <linux/mm_types.h>
> +#include <linux/dma-contiguous.h>
> +#include <linux/dma-mapping.h>
> +#include <linux/of_reserved_mem.h>
> +
> +#define MAX_RESERVED_REGIONS   16
> +struct reserved_mem {
> +       phys_addr_t             base;
> +       unsigned long           size;
> +       struct cma              *cma;
> +       char                    name[32];
> +};
> +static struct reserved_mem reserved_mem[MAX_RESERVED_REGIONS];
> +static int reserved_mem_count;
> +
> +static int __init reserved_mem_fdt_scan(unsigned long node, const char *uname,
> +                                   int depth, void *data)

early_init_dt_scan_reserved_mem would be a more consistent name.

> +{
> +       static int size_cells = OF_ROOT_NODE_SIZE_CELLS_DEFAULT;
> +       static int addr_cells = OF_ROOT_NODE_ADDR_CELLS_DEFAULT;
> +       static int my_depth;
> +       phys_addr_t base, size;
> +       int is_cma, is_reserved;
> +       unsigned long len;
> +       __be32 *prop;
> +
> +       if (depth == 1 && my_depth == 0 && strcmp(uname, "memory") == 0) {
> +               my_depth = depth;
> +               /* scan next node */
> +               return 0;
> +       } else if (depth == 2 && my_depth == 1 &&
> +           strcmp(uname, "reserved-memory") == 0) {
> +               prop = of_get_flat_dt_prop(node, "#size-cells", NULL);
> +               if (prop)
> +                       size_cells = be32_to_cpup(prop);
> +
> +               prop = of_get_flat_dt_prop(node, "#address-cells", NULL);
> +               if (prop)
> +                       addr_cells = be32_to_cpup(prop);

I think we should just require these be the same size as the memory
node which would be dt_root_*_cells.

I'm fine with moving this into drivers/of/fdt.c if that simplifies things.

> +
> +               my_depth = depth;
> +               /* scan next node */
> +               return 0;
> +       } else if (depth != 3 && my_depth != 2) {
> +               /* scan next node */
> +               return 0;
> +       } else if (depth < my_depth) {
> +               /* break scan now */
> +               return 1;
> +       }

This code bothers me and is hard to follow. I don't think trying to
use of_scan_flat_dt is the right approach here. What you really want
here is check for reserved-memory node under the memory node and then
scan each child node. This could all be done from
early_init_dt_scan_memory. u-boot has more rich flat DT functions than
the kernel does since the kernel does very little with the flat DT.
Look at fdt_next_node and it's callers in u-boot for some examples.
But still my first question remains. Can we avoid the flat DT
altogether?


> +       /* now we are scanning a /memory/reserved-memory child */
> +
> +       is_cma = of_flat_dt_is_compatible(node, "linux,contiguous-memory-region");
> +       is_reserved = of_flat_dt_is_compatible(node, "reserved-memory-region");
> +
> +       if (!is_reserved && !(is_cma && IS_ENABLED(CONFIG_CMA))) {
> +               /* ignore node and scan next one */
> +               return 0;
> +       }
> +
> +       prop = of_get_flat_dt_prop(node, "reg", &len);
> +       if (!prop || (len != (size_cells + addr_cells) * sizeof(__be32))) {
> +               pr_err("Reserved mem: node %s, incorrect \"reg\" property\n",
> +                      uname);
> +               /* ignore node and scan next one */
> +               return 0;
> +       }
> +       base = dt_mem_next_cell(addr_cells, &prop);
> +       size = dt_mem_next_cell(size_cells, &prop);
> +
> +       if (!size) {
> +               /* ignore node and scan next one */
> +               return 0;
> +       }
> +
> +       pr_info("Reserved mem: found %s, memory base %lx, size %ld MiB\n",
> +               uname, (unsigned long)base, (unsigned long)size / SZ_1M);
> +
> +       if (reserved_mem_count == ARRAY_SIZE(reserved_mem))
> +               return -ENOSPC;
> +
> +       reserved_mem[reserved_mem_count].base = base;
> +       reserved_mem[reserved_mem_count].size = size;
> +       strlcpy(reserved_mem[reserved_mem_count].name, uname,
> +               sizeof(reserved_mem[reserved_mem_count].name));
> +
> +       if (IS_ENABLED(CONFIG_CMA) && is_cma) {
> +               struct cma *cma;
> +               if (dma_contiguous_reserve_area(size, base, 0, &cma) == 0) {
> +                       reserved_mem[reserved_mem_count].cma = cma;
> +                       reserved_mem_count++;
> +
> +                       if (of_get_flat_dt_prop(node,
> +                                               "linux,default-contiguous-region",
> +                                               NULL))
> +                               dma_contiguous_default_area = cma;
> +               }
> +       } else if (is_reserved) {
> +               if (memblock_remove(base, size) == 0)
> +                       reserved_mem_count++;
> +               else
> +                       pr_err("Failed to reserve memory for %s\n", uname);
> +       }
> +
> +       return 0;
> +}
> +
> +static struct reserved_mem *get_dma_memory_region(struct device *dev)
> +{
> +       struct device_node *node;
> +       const char *name;
> +       int i;
> +
> +       node = of_parse_phandle(dev->of_node, "memory-region", 0);
> +       if (!node)
> +               return NULL;
> +
> +       name = kbasename(node->full_name);
> +       for (i = 0; i < reserved_mem_count; i++)
> +               if (strcmp(name, reserved_mem[i].name) == 0)
> +                       return &reserved_mem[i];
> +       return NULL;

Matching against a struct device_node pointer would be more common way
to match. So it would be good to update reserved_mem with a
device_node ptr when we unflatten the DT.

> +}
> +
> +/**
> + * of_reserved_mem_device_init() - assign reserved memory region to given device
> + *
> + * This function assign memory region pointed by "dma-memory-region" device tree
> + * property to the given device.
> + */
> +void of_reserved_mem_device_init(struct device *dev)
> +{
> +       struct reserved_mem *region = get_dma_memory_region(dev);
> +       if (!region)
> +               return;
> +
> +       if (region->cma) {
> +               dev_set_cma_area(dev, region->cma);
> +               pr_info("Assigned CMA %s to %s device\n", region->name,
> +                       dev_name(dev));
> +       } else {
> +               if (dma_declare_coherent_memory(dev, region->base, region->base,
> +                   region->size, DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE) != 0)
> +                       pr_info("Declared reserved memory %s to %s device\n",
> +                               region->name, dev_name(dev));
> +       }
> +}
> +
> +/**
> + * of_reserved_mem_device_release() - release reserved memory device structures
> + *
> + * This function releases structures allocated for memory region handling for
> + * the given device.
> + */
> +void of_reserved_mem_device_release(struct device *dev)
> +{
> +       struct reserved_mem *region = get_dma_memory_region(dev);
> +       if (!region && !region->cma)
> +               dma_release_declared_memory(dev);
> +}
> +
> +/**
> + * dma_reserved_mem_reserve() - grab memory reserved for device exclusive use
> + *
> + * This function grabs memory from early allocator for device exclusive use
> + * defined in device tree structures. It should be called by arch specific code
> + * once the early allocator (memblock) has been activated and all other
> + * subsystems have already allocated/reserved memory.
> + */
> +void __init dma_reserved_mem_of_reserve(void)
> +{
> +       if (initial_boot_params)
> +               of_scan_flat_dt(reserved_mem_fdt_scan, NULL);
> +}
> diff --git a/drivers/of/platform.c b/drivers/of/platform.c
> index e0a6514..1e4e91d 100644
> --- a/drivers/of/platform.c
> +++ b/drivers/of/platform.c
> @@ -21,6 +21,7 @@
>  #include <linux/of_device.h>
>  #include <linux/of_irq.h>
>  #include <linux/of_platform.h>
> +#include <linux/of_reserved_mem.h>
>  #include <linux/platform_device.h>
>
>  const struct of_device_id of_default_bus_match_table[] = {
> @@ -196,6 +197,7 @@ EXPORT_SYMBOL(of_device_alloc);
>   * Returns pointer to created platform device, or NULL if a device was not
>   * registered.  Unavailable devices will not get registered.
>   */
> +

stray ws change. Please remove.

>  struct platform_device *of_platform_device_create_pdata(
>                                         struct device_node *np,
>                                         const char *bus_id,
> @@ -218,6 +220,8 @@ struct platform_device *of_platform_device_create_pdata(
>         dev->dev.bus = &platform_bus_type;
>         dev->dev.platform_data = platform_data;
>
> +       of_reserved_mem_device_init(&dev->dev);
> +
>         /* We do not fill the DMA ops for platform devices by default.
>          * This is currently the responsibility of the platform code
>          * to do such, possibly using a device notifier
> @@ -225,6 +229,7 @@ struct platform_device *of_platform_device_create_pdata(
>
>         if (of_device_add(dev) != 0) {
>                 platform_device_put(dev);
> +               of_reserved_mem_device_release(&dev->dev);
>                 return NULL;
>         }
>
> diff --git a/include/linux/of_reserved_mem.h b/include/linux/of_reserved_mem.h
> new file mode 100644
> index 0000000..1274946
> --- /dev/null
> +++ b/include/linux/of_reserved_mem.h
> @@ -0,0 +1,14 @@
> +#ifndef __OF_RESERVED_MEM_H
> +#define __OF_RESERVED_MEM_H
> +
> +#ifdef CONFIG_OF_RESERVED_MEM
> +void of_reserved_mem_device_init(struct device *dev);
> +void of_reserved_mem_device_release(struct device *dev);
> +void __init dma_reserved_mem_of_reserve(void);

Declarations don't need __init annotation.

> +#else
> +#define of_reserved_mem_device_init(dev) (void)0
> +#define of_reserved_mem_device_release(dev) (void)0
> +#define dma_reserved_mem_of_reserve() (void)0

static inline is preferred over defines.

Rob
Marek Szyprowski Aug. 12, 2013, 8:34 a.m. UTC | #2
Hello,

On 8/10/2013 7:33 PM, Rob Herring wrote:
> On Fri, Aug 9, 2013 at 6:51 AM, Marek Szyprowski
> <m.szyprowski@samsung.com> wrote:
> > Add device tree support for contiguous and reserved memory regions
> > defined in device tree. Initialization is done in 2 steps. First, the
> > memory is reserved, what happens very early when only flattened device
>
> s/what/which/
>
> > tree is available. Then on device initialization the corresponding cma
> > and reserved regions are assigned to each device structure.
>
> What this commit message does not tell me is why does the reservation
> have to happen before the fdt is unflattened? It would greatly
> simplify the code if it didn't.

Large memory blocks can be RELIABLY reserved only during early boot. This
must happen before the whole memory management subsystem is initialized,
because we need to ensure that the given contiguous blocks are not yet
allocated by kernel. Also it must happen before kernel mappings for the
whole low memory are created, to ensure that there will be no mappings
(for reserved blocks) or mapping with special properties can be created
(for CMA blocks). This all happens before device tree structures are
unflattened, so we need to get reserved memory layout directly from fdt.


> > Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
> > Acked-by: Kyungmin Park <kyungmin.park@samsung.com>
> > Acked-by: Michal Nazarewicz <mina86@mina86.com>
> > ---
> >  Documentation/devicetree/bindings/memory.txt |  152 ++++++++++++++++++++
> >  drivers/of/Kconfig                           |    6 +
> >  drivers/of/Makefile                          |    1 +
> >  drivers/of/of_reserved_mem.c                 |  197 ++++++++++++++++++++++++++
> >  drivers/of/platform.c                        |    5 +
> >  include/linux/of_reserved_mem.h              |   14 ++
> >  6 files changed, 375 insertions(+)
> >  create mode 100644 Documentation/devicetree/bindings/memory.txt
> >  create mode 100644 drivers/of/of_reserved_mem.c
> >  create mode 100644 include/linux/of_reserved_mem.h
> >
> > diff --git a/Documentation/devicetree/bindings/memory.txt b/Documentation/devicetree/bindings/memory.txt
> > new file mode 100644
> > index 0000000..167d013
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/memory.txt
> > @@ -0,0 +1,152 @@
> > +*** Memory binding ***
> > +
> > +The /memory node provides basic information about the address and size
> > +of the physical memory. This node is usually filled or updated by the
> > +bootloader, depending on the actual memory configuration of the given
> > +hardware.
> > +
> > +The memory layout is described by the folllowing node:
>
> typo.
>
> > +
> > +memory {
> > +       device_type = "memory";
> > +       reg =  <(baseaddr1) (size1)
> > +               (baseaddr2) (size2)
> > +               ...
> > +               (baseaddrN) (sizeN)>;
> > +};
> > +
> > +baseaddrX:     the base address of the defined memory bank
> > +sizeX:         the size of the defined memory bank
> > +
> > +More than one memory bank can be defined.
> > +
> > +
> > +*** Reserved memory regions ***
> > +
> > +In /memory/reserved-memory node one can create additional nodes
> > +describing particular reserved (excluded from normal use) memory
> > +regions. Such memory regions are usually designed for the special usage
> > +by various device drivers. A good example are contiguous memory
> > +allocations or memory sharing with other operating system on the same
> > +hardware board. Those special memory regions might depend on the board
> > +configuration and devices used on the target system.
> > +
> > +Parameters for each memory region can be encoded into the device tree
> > +wit the following convention:
> > +
> > +[(label):] (name)@(address) {
> > +       compatible = "contiguous-memory-region", "reserved-memory-region";
> > +       reg = <(address) (size)>;
> > +       (linux,default-contiguous-region);
> > +};
> > +
> > +label:         label given to the defined region (optional)
> > +name:          an name given to the defined region
> > +address:       the base address of the defined region
> > +size:          the size of the memory region
> > +
> > +compatible:    "contiguous-memory-region" - enables binding of this
> > +               region to Contiguous Memory Allocator (special region for
> > +               contiguous memory allocations, shared with movable system
> > +               memory, Linux kernel-specific), alternatively if
> > +               "reserved-memory-region" - compatibility is defined, given
> > +               region is assigned for exclusive usage for DMA transfers
> > +
> > +linux,default-contiguous-region: property indicating that the region
> > +               is the default region for all contiguous memory
> > +               allocations, Linux specific (optional)
> > +
> > +Each defined region must use unique name. It is optional to specify the
> > +base address, so if one wants to use autoconfiguration of the base
> > +address, he must specify the '0' as base address in the 'reg' property
> > +and assign ann unique name to such regions.
> > +
> > +
> > +*** Device node's properties ***
> > +
> > +Once the regions in the /memory/reserved-memory node are defined, they
> > +can be assigned to device nodes to enable drivers for their special use.
> > +The following properties are defined:
> > +
> > +memory-region = <&phandle_to_defined_region>;
> > +
> > +This property indicates that the device driver should use the
> > +memory region pointed by the given phandle.
> > +
> > +
> > +*** Example ***
> > +
> > +This example defines a memory consisting of 4 memory banks. 3 contiguous
> > +regions are defined for Linux kernel, one default of all device drivers
> > +(named contig_mem, placed at 0x72000000, 64MiB), one dedicated to the
> > +framebuffer device (labelled display_mem, placed at 0x78000000, 8MiB)
> > +and one for multimedia processing (labelled multimedia_mem, placed at
> > +0x77000000, 64MiB). 'display_mem' region is then assigned to fb@12300000
> > +device for DMA memory allocations (Linux kernel drivers will use CMA is
> > +available or dma-exclusive usage otherwise). 'multimedia_mem' is
> > +assigned to scaler@12500000 and codec@12600000 devices for contiguous
> > +memory allocations when CMA driver is enabled.
> > +
> > +The reason for creating a separate region for framebuffer device is to
> > +match the framebuffer base address to the one configured by bootloader,
> > +so once Linux kernel drivers starts no glitches on the displayed boot
> > +logo appears. Scaller and codec drivers should share the memory
> > +allocations.
> > +
> > +/ {
> > +       /* ... */
> > +       memory {
> > +               reg =  <0x40000000 0x10000000
> > +                       0x50000000 0x10000000
> > +                       0x60000000 0x10000000
> > +                       0x70000000 0x10000000>;
> > +
> > +               reserved-memory {
> > +                       #address-cells = <1>;
> > +                       #size-cells = <1>;
> > +
> > +                       /*
> > +                        * global autoconfigured region for contiguous allocations
> > +                        * (used only with Contiguous Memory Allocator)
> > +                        */
> > +                       contig_region@0 {
> > +                               compatible = "contiguous-memory-region";
> > +                               reg = <0x0 0x4000000>;
> > +                               linux,default-contiguous-region;
> > +                       };
> > +
> > +                       /*
> > +                        * special region for framebuffer
> > +                        */
> > +                       display_mem: region@78000000 {
> > +                               compatible = "contiguous-memory-region", "reserved-memory-region";
> > +                               reg = <0x78000000 0x800000>;
> > +                       };
> > +
> > +                       /*
> > +                        * special region for multimedia processing devices
> > +                        */
> > +                       multimedia_mem: region@77000000 {
> > +                               compatible = "contiguous-memory-region";
> > +                               reg = <0x77000000 0x4000000>;
> > +                       };
> > +               };
> > +       };
> > +
> > +       /* ... */
> > +
> > +       fb0: fb@12300000 {
> > +               status = "okay";
> > +               memory-region = <&display_mem>;
> > +       };
> > +
> > +       scaler: scaler@12500000 {
> > +               status = "okay";
> > +               memory-region = <&multimedia_mem>;
> > +       };
> > +
> > +       codec: codec@12600000 {
> > +               status = "okay";
> > +               memory-region = <&multimedia_mem>;
> > +       };
> > +};
> > diff --git a/drivers/of/Kconfig b/drivers/of/Kconfig
> > index 80e5c13..a83ab43 100644
> > --- a/drivers/of/Kconfig
> > +++ b/drivers/of/Kconfig
> > @@ -80,4 +80,10 @@ config OF_MTD
> >         depends on MTD
> >         def_bool y
> >
> > +config OF_RESERVED_MEM
> > +       depends on CMA || (HAVE_GENERIC_DMA_COHERENT && HAVE_MEMBLOCK)
> > +       def_bool y
> > +       help
> > +         Initialization code for DMA reserved memory
> > +
> >  endmenu # OF
> > diff --git a/drivers/of/Makefile b/drivers/of/Makefile
> > index 1f9c0c4..e7e3322 100644
> > --- a/drivers/of/Makefile
> > +++ b/drivers/of/Makefile
> > @@ -10,3 +10,4 @@ obj-$(CONFIG_OF_MDIO) += of_mdio.o
> >  obj-$(CONFIG_OF_PCI)   += of_pci.o
> >  obj-$(CONFIG_OF_PCI_IRQ)  += of_pci_irq.o
> >  obj-$(CONFIG_OF_MTD)   += of_mtd.o
> > +obj-$(CONFIG_OF_RESERVED_MEM) += of_reserved_mem.o
> > diff --git a/drivers/of/of_reserved_mem.c b/drivers/of/of_reserved_mem.c
> > new file mode 100644
> > index 0000000..b256b41
> > --- /dev/null
> > +++ b/drivers/of/of_reserved_mem.c
> > @@ -0,0 +1,197 @@
> > +/*
> > + * Device tree based initialization code for reserved memory.
> > + *
> > + * Copyright (c) 2013 Samsung Electronics Co., Ltd.
> > + *             http://www.samsung.com
> > + * Author: Marek Szyprowski <m.szyprowski@samsung.com>
> > + *
> > + * This program is free software; you can redistribute it and/or
> > + * modify it under the terms of the GNU General Public License as
> > + * published by the Free Software Foundation; either version 2 of the
> > + * License or (at your optional) any later version of the license.
> > + */
> > +
> > +#include <asm/dma-contiguous.h>
> > +
> > +#include <linux/memblock.h>
> > +#include <linux/err.h>
> > +#include <linux/of.h>
> > +#include <linux/of_fdt.h>
> > +#include <linux/of_platform.h>
> > +#include <linux/mm.h>
> > +#include <linux/sizes.h>
> > +#include <linux/mm_types.h>
> > +#include <linux/dma-contiguous.h>
> > +#include <linux/dma-mapping.h>
> > +#include <linux/of_reserved_mem.h>
> > +
> > +#define MAX_RESERVED_REGIONS   16
> > +struct reserved_mem {
> > +       phys_addr_t             base;
> > +       unsigned long           size;
> > +       struct cma              *cma;
> > +       char                    name[32];
> > +};
> > +static struct reserved_mem reserved_mem[MAX_RESERVED_REGIONS];
> > +static int reserved_mem_count;
> > +
> > +static int __init reserved_mem_fdt_scan(unsigned long node, const char *uname,
> > +                                   int depth, void *data)
>
> early_init_dt_scan_reserved_mem would be a more consistent name.
>
> > +{
> > +       static int size_cells = OF_ROOT_NODE_SIZE_CELLS_DEFAULT;
> > +       static int addr_cells = OF_ROOT_NODE_ADDR_CELLS_DEFAULT;
> > +       static int my_depth;
> > +       phys_addr_t base, size;
> > +       int is_cma, is_reserved;
> > +       unsigned long len;
> > +       __be32 *prop;
> > +
> > +       if (depth == 1 && my_depth == 0 && strcmp(uname, "memory") == 0) {
> > +               my_depth = depth;
> > +               /* scan next node */
> > +               return 0;
> > +       } else if (depth == 2 && my_depth == 1 &&
> > +           strcmp(uname, "reserved-memory") == 0) {
> > +               prop = of_get_flat_dt_prop(node, "#size-cells", NULL);
> > +               if (prop)
> > +                       size_cells = be32_to_cpup(prop);
> > +
> > +               prop = of_get_flat_dt_prop(node, "#address-cells", NULL);
> > +               if (prop)
> > +                       addr_cells = be32_to_cpup(prop);
>
> I think we should just require these be the same size as the memory
> node which would be dt_root_*_cells.
>
> I'm fine with moving this into drivers/of/fdt.c if that simplifies things.

dt_root_*_cells are global variables, so its not a problem to get access to
them. However I wonder how can we ensure that user/device tree creator will
set #size-cells/#address-cells to the same values as for root memory node?
Would it be enough to state that in binding documentation? If so then the
reserved memory code can skip parsing them and use dt_root_*_cells directly,
what will simplify the code.

> > +
> > +               my_depth = depth;
> > +               /* scan next node */
> > +               return 0;
> > +       } else if (depth != 3 && my_depth != 2) {
> > +               /* scan next node */
> > +               return 0;
> > +       } else if (depth < my_depth) {
> > +               /* break scan now */
> > +               return 1;
> > +       }
>
> This code bothers me and is hard to follow. I don't think trying to
> use of_scan_flat_dt is the right approach here. What you really want
> here is check for reserved-memory node under the memory node and then
> scan each child node. This could all be done from
> early_init_dt_scan_memory.

early_init_dt_scan_memory() is also called from of_scan_flat_dt() and
it also implements similar state machine to parse fdt. The only
difference is the fact that "memory" is a direct child of root node,
so the state machine is much simpler (there is no need to parse
/memory/reserved-memory path).


>   u-boot has more rich flat DT functions than
> the kernel does since the kernel does very little with the flat DT.
> Look at fdt_next_node and it's callers in u-boot for some examples.
> But still my first question remains. Can we avoid the flat DT
> altogether?

I see no such possibility. FDT is unflattened too late to do reliably
all operation required for memory reservation.

> > +       /* now we are scanning a /memory/reserved-memory child */
> > +
> > +       is_cma = of_flat_dt_is_compatible(node, "linux,contiguous-memory-region");
> > +       is_reserved = of_flat_dt_is_compatible(node, "reserved-memory-region");
> > +
> > +       if (!is_reserved && !(is_cma && IS_ENABLED(CONFIG_CMA))) {
> > +               /* ignore node and scan next one */
> > +               return 0;
> > +       }
> > +
> > +       prop = of_get_flat_dt_prop(node, "reg", &len);
> > +       if (!prop || (len != (size_cells + addr_cells) * sizeof(__be32))) {
> > +               pr_err("Reserved mem: node %s, incorrect \"reg\" property\n",
> > +                      uname);
> > +               /* ignore node and scan next one */
> > +               return 0;
> > +       }
> > +       base = dt_mem_next_cell(addr_cells, &prop);
> > +       size = dt_mem_next_cell(size_cells, &prop);
> > +
> > +       if (!size) {
> > +               /* ignore node and scan next one */
> > +               return 0;
> > +       }
> > +
> > +       pr_info("Reserved mem: found %s, memory base %lx, size %ld MiB\n",
> > +               uname, (unsigned long)base, (unsigned long)size / SZ_1M);
> > +
> > +       if (reserved_mem_count == ARRAY_SIZE(reserved_mem))
> > +               return -ENOSPC;
> > +
> > +       reserved_mem[reserved_mem_count].base = base;
> > +       reserved_mem[reserved_mem_count].size = size;
> > +       strlcpy(reserved_mem[reserved_mem_count].name, uname,
> > +               sizeof(reserved_mem[reserved_mem_count].name));
> > +
> > +       if (IS_ENABLED(CONFIG_CMA) && is_cma) {
> > +               struct cma *cma;
> > +               if (dma_contiguous_reserve_area(size, base, 0, &cma) == 0) {
> > +                       reserved_mem[reserved_mem_count].cma = cma;
> > +                       reserved_mem_count++;
> > +
> > +                       if (of_get_flat_dt_prop(node,
> > +                                               "linux,default-contiguous-region",
> > +                                               NULL))
> > +                               dma_contiguous_default_area = cma;
> > +               }
> > +       } else if (is_reserved) {
> > +               if (memblock_remove(base, size) == 0)
> > +                       reserved_mem_count++;
> > +               else
> > +                       pr_err("Failed to reserve memory for %s\n", uname);
> > +       }
> > +
> > +       return 0;
> > +}
> > +
> > +static struct reserved_mem *get_dma_memory_region(struct device *dev)
> > +{
> > +       struct device_node *node;
> > +       const char *name;
> > +       int i;
> > +
> > +       node = of_parse_phandle(dev->of_node, "memory-region", 0);
> > +       if (!node)
> > +               return NULL;
> > +
> > +       name = kbasename(node->full_name);
> > +       for (i = 0; i < reserved_mem_count; i++)
> > +               if (strcmp(name, reserved_mem[i].name) == 0)
> > +                       return &reserved_mem[i];
> > +       return NULL;
>
> Matching against a struct device_node pointer would be more common way
> to match. So it would be good to update reserved_mem with a
> device_node ptr when we unflatten the DT.

I wonder if it really makes sense. To get device_node ptr I will need to
scan /memody/reserved-memory node and match all its children BY NAME
with the structures parsed from FDT (stored in reserved_mem array). Then
I will need to iterate again for each device node with memory-region
property to find the needed entry. Names are unique, IMHO they can serve
as a key for matching structures between FDT and regular, unflattened DT.

> > +}
> > +
> > +/**
> > + * of_reserved_mem_device_init() - assign reserved memory region to given device
> > + *
> > + * This function assign memory region pointed by "dma-memory-region" device tree
> > + * property to the given device.
> > + */
> > +void of_reserved_mem_device_init(struct device *dev)
> > +{
> > +       struct reserved_mem *region = get_dma_memory_region(dev);
> > +       if (!region)
> > +               return;
> > +
> > +       if (region->cma) {
> > +               dev_set_cma_area(dev, region->cma);
> > +               pr_info("Assigned CMA %s to %s device\n", region->name,
> > +                       dev_name(dev));
> > +       } else {
> > +               if (dma_declare_coherent_memory(dev, region->base, region->base,
> > +                   region->size, DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE) != 0)
> > +                       pr_info("Declared reserved memory %s to %s device\n",
> > +                               region->name, dev_name(dev));
> > +       }
> > +}
> > +
> > +/**
> > + * of_reserved_mem_device_release() - release reserved memory device structures
> > + *
> > + * This function releases structures allocated for memory region handling for
> > + * the given device.
> > + */
> > +void of_reserved_mem_device_release(struct device *dev)
> > +{
> > +       struct reserved_mem *region = get_dma_memory_region(dev);
> > +       if (!region && !region->cma)
> > +               dma_release_declared_memory(dev);
> > +}
> > +
> > +/**
> > + * dma_reserved_mem_reserve() - grab memory reserved for device exclusive use
> > + *
> > + * This function grabs memory from early allocator for device exclusive use
> > + * defined in device tree structures. It should be called by arch specific code
> > + * once the early allocator (memblock) has been activated and all other
> > + * subsystems have already allocated/reserved memory.
> > + */
> > +void __init dma_reserved_mem_of_reserve(void)
> > +{
> > +       if (initial_boot_params)
> > +               of_scan_flat_dt(reserved_mem_fdt_scan, NULL);
> > +}
> > diff --git a/drivers/of/platform.c b/drivers/of/platform.c
> > index e0a6514..1e4e91d 100644
> > --- a/drivers/of/platform.c
> > +++ b/drivers/of/platform.c
> > @@ -21,6 +21,7 @@
> >  #include <linux/of_device.h>
> >  #include <linux/of_irq.h>
> >  #include <linux/of_platform.h>
> > +#include <linux/of_reserved_mem.h>
> >  #include <linux/platform_device.h>
> >
> >  const struct of_device_id of_default_bus_match_table[] = {
> > @@ -196,6 +197,7 @@ EXPORT_SYMBOL(of_device_alloc);
> >   * Returns pointer to created platform device, or NULL if a device was not
> >   * registered.  Unavailable devices will not get registered.
> >   */
> > +
>
> stray ws change. Please remove.
>
> >  struct platform_device *of_platform_device_create_pdata(
> >                                         struct device_node *np,
> >                                         const char *bus_id,
> > @@ -218,6 +220,8 @@ struct platform_device *of_platform_device_create_pdata(
> >         dev->dev.bus = &platform_bus_type;
> >         dev->dev.platform_data = platform_data;
> >
> > +       of_reserved_mem_device_init(&dev->dev);
> > +
> >         /* We do not fill the DMA ops for platform devices by default.
> >          * This is currently the responsibility of the platform code
> >          * to do such, possibly using a device notifier
> > @@ -225,6 +229,7 @@ struct platform_device *of_platform_device_create_pdata(
> >
> >         if (of_device_add(dev) != 0) {
> >                 platform_device_put(dev);
> > +               of_reserved_mem_device_release(&dev->dev);
> >                 return NULL;
> >         }
> >
> > diff --git a/include/linux/of_reserved_mem.h b/include/linux/of_reserved_mem.h
> > new file mode 100644
> > index 0000000..1274946
> > --- /dev/null
> > +++ b/include/linux/of_reserved_mem.h
> > @@ -0,0 +1,14 @@
> > +#ifndef __OF_RESERVED_MEM_H
> > +#define __OF_RESERVED_MEM_H
> > +
> > +#ifdef CONFIG_OF_RESERVED_MEM
> > +void of_reserved_mem_device_init(struct device *dev);
> > +void of_reserved_mem_device_release(struct device *dev);
> > +void __init dma_reserved_mem_of_reserve(void);
>
> Declarations don't need __init annotation.
>
> > +#else
> > +#define of_reserved_mem_device_init(dev) (void)0
> > +#define of_reserved_mem_device_release(dev) (void)0
> > +#define dma_reserved_mem_of_reserve() (void)0
>
> static inline is preferred over defines.
>
> Rob
>

Best regards
Rob Herring Aug. 13, 2013, 1 p.m. UTC | #3
On Mon, Aug 12, 2013 at 3:34 AM, Marek Szyprowski
<m.szyprowski@samsung.com> wrote:
> Hello,
>
>
> On 8/10/2013 7:33 PM, Rob Herring wrote:
>>
>> On Fri, Aug 9, 2013 at 6:51 AM, Marek Szyprowski
>> <m.szyprowski@samsung.com> wrote:
>> > Add device tree support for contiguous and reserved memory regions
>> > defined in device tree. Initialization is done in 2 steps. First, the
>> > memory is reserved, what happens very early when only flattened device
>>
>> s/what/which/
>>
>> > tree is available. Then on device initialization the corresponding cma
>> > and reserved regions are assigned to each device structure.
>>
>> What this commit message does not tell me is why does the reservation
>> have to happen before the fdt is unflattened? It would greatly
>> simplify the code if it didn't.
>
>
> Large memory blocks can be RELIABLY reserved only during early boot. This
> must happen before the whole memory management subsystem is initialized,
> because we need to ensure that the given contiguous blocks are not yet
> allocated by kernel. Also it must happen before kernel mappings for the
> whole low memory are created, to ensure that there will be no mappings
> (for reserved blocks) or mapping with special properties can be created
> (for CMA blocks). This all happens before device tree structures are
> unflattened, so we need to get reserved memory layout directly from fdt.
>

Okay. Just making sure.


>> > +       } else if (depth == 2 && my_depth == 1 &&
>> > +           strcmp(uname, "reserved-memory") == 0) {
>> > +               prop = of_get_flat_dt_prop(node, "#size-cells", NULL);
>> > +               if (prop)
>> > +                       size_cells = be32_to_cpup(prop);
>> > +
>> > +               prop = of_get_flat_dt_prop(node, "#address-cells",
>> > NULL);
>> > +               if (prop)
>> > +                       addr_cells = be32_to_cpup(prop);
>>
>> I think we should just require these be the same size as the memory
>> node which would be dt_root_*_cells.
>>
>> I'm fine with moving this into drivers/of/fdt.c if that simplifies things.
>
>
> dt_root_*_cells are global variables, so its not a problem to get access to
> them. However I wonder how can we ensure that user/device tree creator will
> set #size-cells/#address-cells to the same values as for root memory node?
> Would it be enough to state that in binding documentation? If so then the
> reserved memory code can skip parsing them and use dt_root_*_cells directly,
> what will simplify the code.

Yes, just add a note to the binding that the cell sizes are the same
as the memory node.

>> > +
>> > +               my_depth = depth;
>> > +               /* scan next node */
>> > +               return 0;
>> > +       } else if (depth != 3 && my_depth != 2) {
>> > +               /* scan next node */
>> > +               return 0;
>> > +       } else if (depth < my_depth) {
>> > +               /* break scan now */
>> > +               return 1;
>> > +       }
>>
>> This code bothers me and is hard to follow. I don't think trying to
>> use of_scan_flat_dt is the right approach here. What you really want
>> here is check for reserved-memory node under the memory node and then
>> scan each child node. This could all be done from
>> early_init_dt_scan_memory.
>
>
> early_init_dt_scan_memory() is also called from of_scan_flat_dt() and
> it also implements similar state machine to parse fdt. The only
> difference is the fact that "memory" is a direct child of root node,
> so the state machine is much simpler (there is no need to parse
> /memory/reserved-memory path).
>

If you have already found the memory node, then why find it again? I
don't think the existing scan functions handle anything but top-level
nodes very well. So doing something like this from
early_init_dt_scan_memory() is what I was thinking. It is a very rough
sketch:

// find the reserved-memory node
for (ndepth = 0, off = fdt_next_node(blob, parent_offset, &ndepth);
    (off >= 0) && (ndepth > 0);
    off = fdt_next_node(blob, off, &ndepth)) {
    if (ndepth == 1) {
        name = fdt_get_name(blob, off, 0), off);
        if (strcmp(name, "reserved-memory") == 0) {
             parent_offset = off;
             break; // found
    }
}

// find the child nodes
for (ndepth = 0, off = fdt_next_node(blob, parent_offset, &ndepth);
    (off >= 0) && (ndepth == 1);
    off = fdt_next_node(blob, off, &ndepth)) {
    // now handle each child
}

These could probably be further refined with some loop iterator macros.

>> > +       name = kbasename(node->full_name);
>> > +       for (i = 0; i < reserved_mem_count; i++)
>> > +               if (strcmp(name, reserved_mem[i].name) == 0)
>> > +                       return &reserved_mem[i];
>> > +       return NULL;
>>
>> Matching against a struct device_node pointer would be more common way
>> to match. So it would be good to update reserved_mem with a
>> device_node ptr when we unflatten the DT.
>
>
> I wonder if it really makes sense. To get device_node ptr I will need to
> scan /memody/reserved-memory node and match all its children BY NAME
> with the structures parsed from FDT (stored in reserved_mem array). Then
> I will need to iterate again for each device node with memory-region
> property to find the needed entry. Names are unique, IMHO they can serve
> as a key for matching structures between FDT and regular, unflattened DT.

You are iterating multiple times using a string match versus iterating
once more and then doing a pointer match. However, it is not really
performance critical and is fine for now.

Rob
Stephen Warren Aug. 13, 2013, 8:08 p.m. UTC | #4
On 08/09/2013 05:51 AM, Marek Szyprowski wrote:
> Add device tree support for contiguous and reserved memory regions
> defined in device tree. Initialization is done in 2 steps. First, the
> memory is reserved, what happens very early when only flattened device
> tree is available. Then on device initialization the corresponding cma
> and reserved regions are assigned to each device structure.

Hmmm. This seems an awful lot like putting SW configuration/policy
information into DT rather than HW description. This feels like a
slippery slope... Isn't this kind of thing better handled by a kernel
command-line option to set up the CMA size?

> diff --git a/Documentation/devicetree/bindings/memory.txt b/Documentation/devicetree/bindings/memory.txt

> +*** Memory binding ***
> +
> +The /memory node provides basic information about the address and size
> +of the physical memory. This node is usually filled or updated by the
> +bootloader, depending on the actual memory configuration of the given
> +hardware.
> +
> +The memory layout is described by the folllowing node:
> +
> +memory {
> +	device_type = "memory";
> +	reg =  <(baseaddr1) (size1)
> +		(baseaddr2) (size2)
> +		...
> +		(baseaddrN) (sizeN)>;
> +};
> +
> +baseaddrX:	the base address of the defined memory bank
> +sizeX:		the size of the defined memory bank

You probably want to mention that baseaddrX is #address-cells long and
sizeX is #size-cells long. Same for the reserved regions below.

> +*** Reserved memory regions ***
...
> +Parameters for each memory region can be encoded into the device tree
> +wit the following convention:
> +
> +[(label):] (name)@(address) {

That line is DT syntax nothing to do with this binding. I would re-write
this in the more typical DT binding style where the documentation only
specifies the content of the node, not the node itself.

In particular, there's no requirement for a node name to include the
unit address (@address) if it's already unique.

> +	compatible = "contiguous-memory-region", "reserved-memory-region";
> +	reg = <(address) (size)>;
> +	(linux,default-contiguous-region);

(...) isn't a syntax typically used in DT bindings. You'd usually put
that in a a list of "Optional Properties:".

> +Each defined region must use unique name.

Well, DT nodes are supposed to be names based on the type of object they
represent, not by the name/identity of the object they represent.

> +*** Device node's properties ***
> +
> +Once the regions in the /memory/reserved-memory node are defined, they
> +can be assigned to device nodes to enable drivers for their special use.
> +The following properties are defined:
> +
> +memory-region = <&phandle_to_defined_region>;
> +
> +This property indicates that the device driver should use the
> +memory region pointed by the given phandle.

That's quite scary. This is essentially forcing a memory-region property
into every single binding that ever exists. I guess that's not too much
worse than e.g. interrupts/clocks/..., but I think it's worth somehow
requiring bindings to "opt-in" to allowing this property to be part of
their binding rather than just definining the property globally.
Olof Johansson Aug. 16, 2013, 5:25 a.m. UTC | #5
On Tue, Aug 13, 2013 at 02:08:44PM -0600, Stephen Warren wrote:
> On 08/09/2013 05:51 AM, Marek Szyprowski wrote:
> > Add device tree support for contiguous and reserved memory regions
> > defined in device tree. Initialization is done in 2 steps. First, the
> > memory is reserved, what happens very early when only flattened device
> > tree is available. Then on device initialization the corresponding cma
> > and reserved regions are assigned to each device structure.
> 
> Hmmm. This seems an awful lot like putting SW configuration/policy
> information into DT rather than HW description. This feels like a
> slippery slope... Isn't this kind of thing better handled by a kernel
> command-line option to set up the CMA size?

Sorry, you were not part of the in-person discussion since it happened
at Linaro Connect in Dublin. The concern is that we really need a way
to describe some of these _system_ properties. They're not necessarily
hardware properties, but they are well-known and likely properties of
the system that is running.

I.e. say you have a device with a known screen resolution running some
reasonably well-known software stacks/applications that you expect
will use a certain amount of graphics memory. Or you know how large
your camera sensor is, so you know how much carveout the ISP will need
for captures. So defining how large the default carveouts should be
is very much a property of this system. We would previously have put
these settings in a board file and be done, but now we need a way to
describe it.

/chosen was proposed, but that's mostly used for things that the
sysadmin would adjust (traditionally such as firmware stdin/stdout, etc).
/options isn't a good place either.

Having them on the command line doesn't really work well, over time
we'll start to accumulate cruft there. It's bad enough for some of the
devices out there that do it, I'd say.

So, for the specific case of memory carveouts, this was suggested as the best
way forward.


-Olof
Olof Johansson Aug. 16, 2013, 5:32 a.m. UTC | #6
Hi,

On Fri, Aug 09, 2013 at 01:51:58PM +0200, Marek Szyprowski wrote:
> Add device tree support for contiguous and reserved memory regions
> defined in device tree. Initialization is done in 2 steps. First, the
> memory is reserved, what happens very early when only flattened device
> tree is available. Then on device initialization the corresponding cma
> and reserved regions are assigned to each device structure.

Bikeshedding on some wording, mostly in case you have to respin for other
reasons.

> diff --git a/Documentation/devicetree/bindings/memory.txt b/Documentation/devicetree/bindings/memory.txt
> new file mode 100644
> index 0000000..167d013
> --- /dev/null
> +++ b/Documentation/devicetree/bindings/memory.txt
> @@ -0,0 +1,152 @@
> +*** Memory binding ***
> +
> +The /memory node provides basic information about the address and size
> +of the physical memory. This node is usually filled or updated by the
> +bootloader, depending on the actual memory configuration of the given
> +hardware.
> +
> +The memory layout is described by the folllowing node:
> +
> +memory {
> +	device_type = "memory";
> +	reg =  <(baseaddr1) (size1)
> +		(baseaddr2) (size2)
> +		...
> +		(baseaddrN) (sizeN)>;
> +};
> +
> +baseaddrX:	the base address of the defined memory bank
> +sizeX:		the size of the defined memory bank
> +
> +More than one memory bank can be defined.
> +
> +
> +*** Reserved memory regions ***
> +
> +In /memory/reserved-memory node one can create additional nodes

There's really no requirement on naming here. We should instead search all of
/memory for the compatible properties.

> +describing particular reserved (excluded from normal use) memory
> +regions. Such memory regions are usually designed for the special usage
> +by various device drivers. A good example are contiguous memory
> +allocations or memory sharing with other operating system on the same
> +hardware board. Those special memory regions might depend on the board
> +configuration and devices used on the target system.
> +
> +Parameters for each memory region can be encoded into the device tree
> +wit the following convention:
> +
> +[(label):] (name)@(address) {

@(unit-address)

> +	compatible = "contiguous-memory-region", "reserved-memory-region";
> +	reg = <(address) (size)>;
> +	(linux,default-contiguous-region);
> +};
> +
> +label:		label given to the defined region (optional)
> +name:		an name given to the defined region
> +address:	the base address of the defined region

unit-address. It's optional unless there are conflicting node names.

> +size:		the size of the memory region
> +
> +compatible:	"contiguous-memory-region" - enables binding of this
> +		region to Contiguous Memory Allocator (special region for
> +		contiguous memory allocations, shared with movable system
> +		memory, Linux kernel-specific), alternatively if
> +		"reserved-memory-region" - compatibility is defined, given
> +		region is assigned for exclusive usage for DMA transfers

"reserved-memory-region" isn't necessarily just for DMA transfers, it can be
used pretty much for anything. pstore comes to mind.

> +linux,default-contiguous-region: property indicating that the region
> +		is the default region for all contiguous memory
> +		allocations, Linux specific (optional)

Hmm. We handle some of these defaults such as device naming (i2c busses, etc)
through /alias nodes. Maybe it would be more consistent to do that here too,
by providing a phandle there instead.

> +Each defined region must use unique name.

Hmm, this is quite different from how memory nodes work, and it'd
be nice to keep similar semantics. A "name" property would make more
sense for areas that need to be named, but even then, why do names here
matter? References should be by phandle, not by name. So I think it
should likely just be scrapped.

> It is optional to specify the
> +base address, so if one wants to use autoconfiguration of the base
> +address, he must specify the '0' as base address in the 'reg' property
> +and assign ann unique name to such regions.

"an unique"

But I don't think this is a good idea, since it violates basic use of the reg
properties, where you usually don't have overlaps, and no real requirements on
names (as mentioned above).



-Olof
Stephen Warren Aug. 16, 2013, 4:06 p.m. UTC | #7
On 08/15/2013 11:25 PM, Olof Johansson wrote:
> On Tue, Aug 13, 2013 at 02:08:44PM -0600, Stephen Warren wrote:
>> On 08/09/2013 05:51 AM, Marek Szyprowski wrote:
>>> Add device tree support for contiguous and reserved memory regions
>>> defined in device tree. Initialization is done in 2 steps. First, the
>>> memory is reserved, what happens very early when only flattened device
>>> tree is available. Then on device initialization the corresponding cma
>>> and reserved regions are assigned to each device structure.
>>
>> Hmmm. This seems an awful lot like putting SW configuration/policy
>> information into DT rather than HW description. This feels like a
>> slippery slope... Isn't this kind of thing better handled by a kernel
>> command-line option to set up the CMA size?
> 
> Sorry, you were not part of the in-person discussion since it happened
> at Linaro Connect in Dublin. The concern is that we really need a way
> to describe some of these _system_ properties. They're not necessarily
> hardware properties, but they are well-known and likely properties of
> the system that is running.

OK, that seems reasonable enough. It's just the first I heard of this.

I rather suspect that similar arguments will be applied to a bunch of
other data people want to put into DT. I guess we'll just have to wait
and see what gets proposed, and what really is "system data" rather than
policy:-)
Marek Szyprowski Aug. 19, 2013, 2:47 p.m. UTC | #8
Hello,

On 8/13/2013 3:00 PM, Rob Herring wrote:
> On Mon, Aug 12, 2013 at 3:34 AM, Marek Szyprowski
> <m.szyprowski@samsung.com> wrote:
> > Hello,
> >
> >
> > On 8/10/2013 7:33 PM, Rob Herring wrote:
> >>
> >> On Fri, Aug 9, 2013 at 6:51 AM, Marek Szyprowski
> >> <m.szyprowski@samsung.com> wrote:
> >> > Add device tree support for contiguous and reserved memory regions
> >> > defined in device tree. Initialization is done in 2 steps. First, the
> >> > memory is reserved, what happens very early when only flattened device
> >>
> >> s/what/which/
> >>
> >> > tree is available. Then on device initialization the corresponding cma
> >> > and reserved regions are assigned to each device structure.
> >>
> >> What this commit message does not tell me is why does the reservation
> >> have to happen before the fdt is unflattened? It would greatly
> >> simplify the code if it didn't.
> >
> >
> > Large memory blocks can be RELIABLY reserved only during early boot. This
> > must happen before the whole memory management subsystem is initialized,
> > because we need to ensure that the given contiguous blocks are not yet
> > allocated by kernel. Also it must happen before kernel mappings for the
> > whole low memory are created, to ensure that there will be no mappings
> > (for reserved blocks) or mapping with special properties can be created
> > (for CMA blocks). This all happens before device tree structures are
> > unflattened, so we need to get reserved memory layout directly from fdt.
> >
>
> Okay. Just making sure.
>
>
> >> > +       } else if (depth == 2 && my_depth == 1 &&
> >> > +           strcmp(uname, "reserved-memory") == 0) {
> >> > +               prop = of_get_flat_dt_prop(node, "#size-cells", NULL);
> >> > +               if (prop)
> >> > +                       size_cells = be32_to_cpup(prop);
> >> > +
> >> > +               prop = of_get_flat_dt_prop(node, "#address-cells",
> >> > NULL);
> >> > +               if (prop)
> >> > +                       addr_cells = be32_to_cpup(prop);
> >>
> >> I think we should just require these be the same size as the memory
> >> node which would be dt_root_*_cells.
> >>
> >> I'm fine with moving this into drivers/of/fdt.c if that simplifies things.
> >
> >
> > dt_root_*_cells are global variables, so its not a problem to get access to
> > them. However I wonder how can we ensure that user/device tree creator will
> > set #size-cells/#address-cells to the same values as for root memory node?
> > Would it be enough to state that in binding documentation? If so then the
> > reserved memory code can skip parsing them and use dt_root_*_cells directly,
> > what will simplify the code.
>
> Yes, just add a note to the binding that the cell sizes are the same
> as the memory node.
>
> >> > +
> >> > +               my_depth = depth;
> >> > +               /* scan next node */
> >> > +               return 0;
> >> > +       } else if (depth != 3 && my_depth != 2) {
> >> > +               /* scan next node */
> >> > +               return 0;
> >> > +       } else if (depth < my_depth) {
> >> > +               /* break scan now */
> >> > +               return 1;
> >> > +       }
> >>
> >> This code bothers me and is hard to follow. I don't think trying to
> >> use of_scan_flat_dt is the right approach here. What you really want
> >> here is check for reserved-memory node under the memory node and then
> >> scan each child node. This could all be done from
> >> early_init_dt_scan_memory.
> >
> >
> > early_init_dt_scan_memory() is also called from of_scan_flat_dt() and
> > it also implements similar state machine to parse fdt. The only
> > difference is the fact that "memory" is a direct child of root node,
> > so the state machine is much simpler (there is no need to parse
> > /memory/reserved-memory path).
> >
>
> If you have already found the memory node, then why find it again? I
> don't think the existing scan functions handle anything but top-level
> nodes very well. So doing something like this from
> early_init_dt_scan_memory() is what I was thinking. It is a very rough
> sketch:
>
> // find the reserved-memory node
> for (ndepth = 0, off = fdt_next_node(blob, parent_offset, &ndepth);
>      (off >= 0) && (ndepth > 0);
>      off = fdt_next_node(blob, off, &ndepth)) {
>      if (ndepth == 1) {
>          name = fdt_get_name(blob, off, 0), off);
>          if (strcmp(name, "reserved-memory") == 0) {
>               parent_offset = off;
>               break; // found
>      }
> }
>
> // find the child nodes
> for (ndepth = 0, off = fdt_next_node(blob, parent_offset, &ndepth);
>      (off >= 0) && (ndepth == 1);
>      off = fdt_next_node(blob, off, &ndepth)) {
>      // now handle each child
> }
>
> These could probably be further refined with some loop iterator macros.

The above code looks pretty nice, but there are some problems with it:

1. Although it would look nice to call it from early_init_dt_scan_memory()
it won't be possible, because that time it is too early. memblock structures
are initialized after a call to early_init_dt_scan_memory() and until that
no changes to memory layout are easily possible.

2. Currently there are no fdt parsing functions in the kernel. I've tried
to split of_scan_flat_dt() into fdt_next_node() + fdt_get_name() and use
them both in of_scan_flat_dt() and above reserved memory parsing functions.
In the end I got quite a lot of code, which is still quite hard to follow.

Because of the above I decided to resurrect of_scan_flat_dt_by_path()
function from the previous version and use #size-cells/#address-cells
attributes from root node what in the end simplified reserved memory
parsing function. I hope that it can be accepted after such changes
without introducing a burden caused by the whole infrastructure for
manual parsing fdt.

> >> > +       name = kbasename(node->full_name);
> >> > +       for (i = 0; i < reserved_mem_count; i++)
> >> > +               if (strcmp(name, reserved_mem[i].name) == 0)
> >> > +                       return &reserved_mem[i];
> >> > +       return NULL;
> >>
> >> Matching against a struct device_node pointer would be more common way
> >> to match. So it would be good to update reserved_mem with a
> >> device_node ptr when we unflatten the DT.
> >
> >
> > I wonder if it really makes sense. To get device_node ptr I will need to
> > scan /memody/reserved-memory node and match all its children BY NAME
> > with the structures parsed from FDT (stored in reserved_mem array). Then
> > I will need to iterate again for each device node with memory-region
> > property to find the needed entry. Names are unique, IMHO they can serve
> > as a key for matching structures between FDT and regular, unflattened DT.
>
> You are iterating multiple times using a string match versus iterating
> once more and then doing a pointer match. However, it is not really
> performance critical and is fine for now.

Thanks!

Best regards
Rob Herring Aug. 19, 2013, 7:36 p.m. UTC | #9
On Mon, Aug 19, 2013 at 9:47 AM, Marek Szyprowski
<m.szyprowski@samsung.com> wrote:
> Hello,
>
>
> On 8/13/2013 3:00 PM, Rob Herring wrote:
>>
>> On Mon, Aug 12, 2013 at 3:34 AM, Marek Szyprowski
>> <m.szyprowski@samsung.com> wrote:
>> > Hello,
>> >
>> >
>> > On 8/10/2013 7:33 PM, Rob Herring wrote:

[snip]

>> > early_init_dt_scan_memory() is also called from of_scan_flat_dt() and
>> > it also implements similar state machine to parse fdt. The only
>> > difference is the fact that "memory" is a direct child of root node,
>> > so the state machine is much simpler (there is no need to parse
>> > /memory/reserved-memory path).
>> >
>>
>> If you have already found the memory node, then why find it again? I
>> don't think the existing scan functions handle anything but top-level
>> nodes very well. So doing something like this from
>> early_init_dt_scan_memory() is what I was thinking. It is a very rough
>> sketch:
>>
>> // find the reserved-memory node
>> for (ndepth = 0, off = fdt_next_node(blob, parent_offset, &ndepth);
>>      (off >= 0) && (ndepth > 0);
>>      off = fdt_next_node(blob, off, &ndepth)) {
>>      if (ndepth == 1) {
>>          name = fdt_get_name(blob, off, 0), off);
>>          if (strcmp(name, "reserved-memory") == 0) {
>>               parent_offset = off;
>>               break; // found
>>      }
>> }
>>
>> // find the child nodes
>> for (ndepth = 0, off = fdt_next_node(blob, parent_offset, &ndepth);
>>      (off >= 0) && (ndepth == 1);
>>      off = fdt_next_node(blob, off, &ndepth)) {
>>      // now handle each child
>> }
>>
>> These could probably be further refined with some loop iterator macros.
>
>
> The above code looks pretty nice, but there are some problems with it:
>
> 1. Although it would look nice to call it from early_init_dt_scan_memory()
> it won't be possible, because that time it is too early. memblock structures
> are initialized after a call to early_init_dt_scan_memory() and until that
> no changes to memory layout are easily possible.

Okay. So the reserved areas have to be setup after memblock_add is
called. It seems that what exactly early_init_dt_add_memory_arch does
varies by arch. For example, arm64 directly does a memblock_add call
and would probably work as I suggested. The arm code just puts the
information in an intermediate arch specific struct until later. There
is a lot of room for clean-up in the early DT code to reduce the
amount of arch code.

> 2. Currently there are no fdt parsing functions in the kernel. I've tried
> to split of_scan_flat_dt() into fdt_next_node() + fdt_get_name() and use
> them both in of_scan_flat_dt() and above reserved memory parsing functions.
> In the end I got quite a lot of code, which is still quite hard to follow.
>
> Because of the above I decided to resurrect of_scan_flat_dt_by_path()
> function from the previous version and use #size-cells/#address-cells
> attributes from root node what in the end simplified reserved memory
> parsing function. I hope that it can be accepted after such changes
> without introducing a burden caused by the whole infrastructure for
> manual parsing fdt.

This all looks fine at first glance.

Rob
diff mbox

Patch

diff --git a/Documentation/devicetree/bindings/memory.txt b/Documentation/devicetree/bindings/memory.txt
new file mode 100644
index 0000000..167d013
--- /dev/null
+++ b/Documentation/devicetree/bindings/memory.txt
@@ -0,0 +1,152 @@ 
+*** Memory binding ***
+
+The /memory node provides basic information about the address and size
+of the physical memory. This node is usually filled or updated by the
+bootloader, depending on the actual memory configuration of the given
+hardware.
+
+The memory layout is described by the folllowing node:
+
+memory {
+	device_type = "memory";
+	reg =  <(baseaddr1) (size1)
+		(baseaddr2) (size2)
+		...
+		(baseaddrN) (sizeN)>;
+};
+
+baseaddrX:	the base address of the defined memory bank
+sizeX:		the size of the defined memory bank
+
+More than one memory bank can be defined.
+
+
+*** Reserved memory regions ***
+
+In /memory/reserved-memory node one can create additional nodes
+describing particular reserved (excluded from normal use) memory
+regions. Such memory regions are usually designed for the special usage
+by various device drivers. A good example are contiguous memory
+allocations or memory sharing with other operating system on the same
+hardware board. Those special memory regions might depend on the board
+configuration and devices used on the target system.
+
+Parameters for each memory region can be encoded into the device tree
+wit the following convention:
+
+[(label):] (name)@(address) {
+	compatible = "contiguous-memory-region", "reserved-memory-region";
+	reg = <(address) (size)>;
+	(linux,default-contiguous-region);
+};
+
+label:		label given to the defined region (optional)
+name:		an name given to the defined region
+address:	the base address of the defined region
+size:		the size of the memory region
+
+compatible:	"contiguous-memory-region" - enables binding of this
+		region to Contiguous Memory Allocator (special region for
+		contiguous memory allocations, shared with movable system
+		memory, Linux kernel-specific), alternatively if
+		"reserved-memory-region" - compatibility is defined, given
+		region is assigned for exclusive usage for DMA transfers
+
+linux,default-contiguous-region: property indicating that the region
+		is the default region for all contiguous memory
+		allocations, Linux specific (optional)
+
+Each defined region must use unique name. It is optional to specify the
+base address, so if one wants to use autoconfiguration of the base
+address, he must specify the '0' as base address in the 'reg' property
+and assign ann unique name to such regions.
+
+
+*** Device node's properties ***
+
+Once the regions in the /memory/reserved-memory node are defined, they
+can be assigned to device nodes to enable drivers for their special use.
+The following properties are defined:
+
+memory-region = <&phandle_to_defined_region>;
+
+This property indicates that the device driver should use the
+memory region pointed by the given phandle.
+
+
+*** Example ***
+
+This example defines a memory consisting of 4 memory banks. 3 contiguous
+regions are defined for Linux kernel, one default of all device drivers
+(named contig_mem, placed at 0x72000000, 64MiB), one dedicated to the
+framebuffer device (labelled display_mem, placed at 0x78000000, 8MiB)
+and one for multimedia processing (labelled multimedia_mem, placed at
+0x77000000, 64MiB). 'display_mem' region is then assigned to fb@12300000
+device for DMA memory allocations (Linux kernel drivers will use CMA is
+available or dma-exclusive usage otherwise). 'multimedia_mem' is
+assigned to scaler@12500000 and codec@12600000 devices for contiguous
+memory allocations when CMA driver is enabled.
+
+The reason for creating a separate region for framebuffer device is to
+match the framebuffer base address to the one configured by bootloader,
+so once Linux kernel drivers starts no glitches on the displayed boot
+logo appears. Scaller and codec drivers should share the memory
+allocations.
+
+/ {
+	/* ... */
+	memory {
+		reg =  <0x40000000 0x10000000
+			0x50000000 0x10000000
+			0x60000000 0x10000000
+			0x70000000 0x10000000>;
+
+		reserved-memory {
+			#address-cells = <1>;
+			#size-cells = <1>;
+
+			/*
+			 * global autoconfigured region for contiguous allocations
+			 * (used only with Contiguous Memory Allocator)
+			 */
+			contig_region@0 {
+				compatible = "contiguous-memory-region";
+				reg = <0x0 0x4000000>;
+				linux,default-contiguous-region;
+			};
+
+			/*
+			 * special region for framebuffer
+			 */
+			display_mem: region@78000000 {
+				compatible = "contiguous-memory-region", "reserved-memory-region";
+				reg = <0x78000000 0x800000>;
+			};
+
+			/*
+			 * special region for multimedia processing devices
+			 */
+			multimedia_mem: region@77000000 {
+				compatible = "contiguous-memory-region";
+				reg = <0x77000000 0x4000000>;
+			};
+		};
+	};
+
+	/* ... */
+
+	fb0: fb@12300000 {
+		status = "okay";
+		memory-region = <&display_mem>;
+	};
+
+	scaler: scaler@12500000 {
+		status = "okay";
+		memory-region = <&multimedia_mem>;
+	};
+
+	codec: codec@12600000 {
+		status = "okay";
+		memory-region = <&multimedia_mem>;
+	};
+};
diff --git a/drivers/of/Kconfig b/drivers/of/Kconfig
index 80e5c13..a83ab43 100644
--- a/drivers/of/Kconfig
+++ b/drivers/of/Kconfig
@@ -80,4 +80,10 @@  config OF_MTD
 	depends on MTD
 	def_bool y
 
+config OF_RESERVED_MEM
+	depends on CMA || (HAVE_GENERIC_DMA_COHERENT && HAVE_MEMBLOCK)
+	def_bool y
+	help
+	  Initialization code for DMA reserved memory
+
 endmenu # OF
diff --git a/drivers/of/Makefile b/drivers/of/Makefile
index 1f9c0c4..e7e3322 100644
--- a/drivers/of/Makefile
+++ b/drivers/of/Makefile
@@ -10,3 +10,4 @@  obj-$(CONFIG_OF_MDIO)	+= of_mdio.o
 obj-$(CONFIG_OF_PCI)	+= of_pci.o
 obj-$(CONFIG_OF_PCI_IRQ)  += of_pci_irq.o
 obj-$(CONFIG_OF_MTD)	+= of_mtd.o
+obj-$(CONFIG_OF_RESERVED_MEM) += of_reserved_mem.o
diff --git a/drivers/of/of_reserved_mem.c b/drivers/of/of_reserved_mem.c
new file mode 100644
index 0000000..b256b41
--- /dev/null
+++ b/drivers/of/of_reserved_mem.c
@@ -0,0 +1,197 @@ 
+/*
+ * Device tree based initialization code for reserved memory.
+ *
+ * Copyright (c) 2013 Samsung Electronics Co., Ltd.
+ *		http://www.samsung.com
+ * Author: Marek Szyprowski <m.szyprowski@samsung.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License or (at your optional) any later version of the license.
+ */
+
+#include <asm/dma-contiguous.h>
+
+#include <linux/memblock.h>
+#include <linux/err.h>
+#include <linux/of.h>
+#include <linux/of_fdt.h>
+#include <linux/of_platform.h>
+#include <linux/mm.h>
+#include <linux/sizes.h>
+#include <linux/mm_types.h>
+#include <linux/dma-contiguous.h>
+#include <linux/dma-mapping.h>
+#include <linux/of_reserved_mem.h>
+
+#define MAX_RESERVED_REGIONS	16
+struct reserved_mem {
+	phys_addr_t		base;
+	unsigned long		size;
+	struct cma		*cma;
+	char			name[32];
+};
+static struct reserved_mem reserved_mem[MAX_RESERVED_REGIONS];
+static int reserved_mem_count;
+
+static int __init reserved_mem_fdt_scan(unsigned long node, const char *uname,
+				    int depth, void *data)
+{
+	static int size_cells = OF_ROOT_NODE_SIZE_CELLS_DEFAULT;
+	static int addr_cells = OF_ROOT_NODE_ADDR_CELLS_DEFAULT;
+	static int my_depth;
+	phys_addr_t base, size;
+	int is_cma, is_reserved;
+	unsigned long len;
+	__be32 *prop;
+
+	if (depth == 1 && my_depth == 0 && strcmp(uname, "memory") == 0) {
+		my_depth = depth;
+		/* scan next node */
+		return 0;
+	} else if (depth == 2 && my_depth == 1 &&
+	    strcmp(uname, "reserved-memory") == 0) {
+		prop = of_get_flat_dt_prop(node, "#size-cells", NULL);
+		if (prop)
+			size_cells = be32_to_cpup(prop);
+
+		prop = of_get_flat_dt_prop(node, "#address-cells", NULL);
+		if (prop)
+			addr_cells = be32_to_cpup(prop);
+
+		my_depth = depth;
+		/* scan next node */
+		return 0;
+	} else if (depth != 3 && my_depth != 2) {
+		/* scan next node */
+		return 0;
+	} else if (depth < my_depth) {
+		/* break scan now */
+		return 1;
+	}
+
+	/* now we are scanning a /memory/reserved-memory child */
+
+	is_cma = of_flat_dt_is_compatible(node, "linux,contiguous-memory-region");
+	is_reserved = of_flat_dt_is_compatible(node, "reserved-memory-region");
+
+	if (!is_reserved && !(is_cma && IS_ENABLED(CONFIG_CMA))) {
+		/* ignore node and scan next one */
+		return 0;
+	}
+
+	prop = of_get_flat_dt_prop(node, "reg", &len);
+	if (!prop || (len != (size_cells + addr_cells) * sizeof(__be32))) {
+		pr_err("Reserved mem: node %s, incorrect \"reg\" property\n",
+		       uname);
+		/* ignore node and scan next one */
+		return 0;
+	}
+	base = dt_mem_next_cell(addr_cells, &prop);
+	size = dt_mem_next_cell(size_cells, &prop);
+
+	if (!size) {
+		/* ignore node and scan next one */
+		return 0;
+	}
+
+	pr_info("Reserved mem: found %s, memory base %lx, size %ld MiB\n",
+		uname, (unsigned long)base, (unsigned long)size / SZ_1M);
+
+	if (reserved_mem_count == ARRAY_SIZE(reserved_mem))
+		return -ENOSPC;
+
+	reserved_mem[reserved_mem_count].base = base;
+	reserved_mem[reserved_mem_count].size = size;
+	strlcpy(reserved_mem[reserved_mem_count].name, uname,
+		sizeof(reserved_mem[reserved_mem_count].name));
+
+	if (IS_ENABLED(CONFIG_CMA) && is_cma) {
+		struct cma *cma;
+		if (dma_contiguous_reserve_area(size, base, 0, &cma) == 0) {
+			reserved_mem[reserved_mem_count].cma = cma;
+			reserved_mem_count++;
+
+			if (of_get_flat_dt_prop(node,
+						"linux,default-contiguous-region",
+						NULL))
+				dma_contiguous_default_area = cma;
+		}
+	} else if (is_reserved) {
+		if (memblock_remove(base, size) == 0)
+			reserved_mem_count++;
+		else
+			pr_err("Failed to reserve memory for %s\n", uname);
+	}
+
+	return 0;
+}
+
+static struct reserved_mem *get_dma_memory_region(struct device *dev)
+{
+	struct device_node *node;
+	const char *name;
+	int i;
+
+	node = of_parse_phandle(dev->of_node, "memory-region", 0);
+	if (!node)
+		return NULL;
+
+	name = kbasename(node->full_name);
+	for (i = 0; i < reserved_mem_count; i++)
+		if (strcmp(name, reserved_mem[i].name) == 0)
+			return &reserved_mem[i];
+	return NULL;
+}
+
+/**
+ * of_reserved_mem_device_init() - assign reserved memory region to given device
+ *
+ * This function assign memory region pointed by "dma-memory-region" device tree
+ * property to the given device.
+ */
+void of_reserved_mem_device_init(struct device *dev)
+{
+	struct reserved_mem *region = get_dma_memory_region(dev);
+	if (!region)
+		return;
+
+	if (region->cma) {
+		dev_set_cma_area(dev, region->cma);
+		pr_info("Assigned CMA %s to %s device\n", region->name,
+			dev_name(dev));
+	} else {
+		if (dma_declare_coherent_memory(dev, region->base, region->base,
+		    region->size, DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE) != 0)
+			pr_info("Declared reserved memory %s to %s device\n",
+				region->name, dev_name(dev));
+	}
+}
+
+/**
+ * of_reserved_mem_device_release() - release reserved memory device structures
+ *
+ * This function releases structures allocated for memory region handling for
+ * the given device.
+ */
+void of_reserved_mem_device_release(struct device *dev)
+{
+	struct reserved_mem *region = get_dma_memory_region(dev);
+	if (!region && !region->cma)
+		dma_release_declared_memory(dev);
+}
+
+/**
+ * dma_reserved_mem_reserve() - grab memory reserved for device exclusive use
+ *
+ * This function grabs memory from early allocator for device exclusive use
+ * defined in device tree structures. It should be called by arch specific code
+ * once the early allocator (memblock) has been activated and all other
+ * subsystems have already allocated/reserved memory.
+ */
+void __init dma_reserved_mem_of_reserve(void)
+{
+	if (initial_boot_params)
+		of_scan_flat_dt(reserved_mem_fdt_scan, NULL);
+}
diff --git a/drivers/of/platform.c b/drivers/of/platform.c
index e0a6514..1e4e91d 100644
--- a/drivers/of/platform.c
+++ b/drivers/of/platform.c
@@ -21,6 +21,7 @@ 
 #include <linux/of_device.h>
 #include <linux/of_irq.h>
 #include <linux/of_platform.h>
+#include <linux/of_reserved_mem.h>
 #include <linux/platform_device.h>
 
 const struct of_device_id of_default_bus_match_table[] = {
@@ -196,6 +197,7 @@  EXPORT_SYMBOL(of_device_alloc);
  * Returns pointer to created platform device, or NULL if a device was not
  * registered.  Unavailable devices will not get registered.
  */
+
 struct platform_device *of_platform_device_create_pdata(
 					struct device_node *np,
 					const char *bus_id,
@@ -218,6 +220,8 @@  struct platform_device *of_platform_device_create_pdata(
 	dev->dev.bus = &platform_bus_type;
 	dev->dev.platform_data = platform_data;
 
+	of_reserved_mem_device_init(&dev->dev);
+
 	/* We do not fill the DMA ops for platform devices by default.
 	 * This is currently the responsibility of the platform code
 	 * to do such, possibly using a device notifier
@@ -225,6 +229,7 @@  struct platform_device *of_platform_device_create_pdata(
 
 	if (of_device_add(dev) != 0) {
 		platform_device_put(dev);
+		of_reserved_mem_device_release(&dev->dev);
 		return NULL;
 	}
 
diff --git a/include/linux/of_reserved_mem.h b/include/linux/of_reserved_mem.h
new file mode 100644
index 0000000..1274946
--- /dev/null
+++ b/include/linux/of_reserved_mem.h
@@ -0,0 +1,14 @@ 
+#ifndef __OF_RESERVED_MEM_H
+#define __OF_RESERVED_MEM_H
+
+#ifdef CONFIG_OF_RESERVED_MEM
+void of_reserved_mem_device_init(struct device *dev);
+void of_reserved_mem_device_release(struct device *dev);
+void __init dma_reserved_mem_of_reserve(void);
+#else
+#define of_reserved_mem_device_init(dev) (void)0
+#define of_reserved_mem_device_release(dev) (void)0
+#define dma_reserved_mem_of_reserve() (void)0
+#endif
+
+#endif /* __OF_RESERVED_MEM_H */