Skip to contentSkip to conversation

Device Bus support for Marvell EBU SoC

PATCH Superseded Ezequiel Garcia 4 participants 20 messages Apr 8, 2013 – Apr 9, 2013

Subsystems (3)

v4 of 7 Superseded Latest is v7 →

lore
Share & tools
lore
#0 lore raw

v4 cover Device Bus support for Marvell EBU SoC

This patchset continues the effort to add Device Bus support on Marvell EBU SoC.

If you're reading this for the first time you can grab the full details here:

http://www.spinics.net/lists/arm-kernel/msg236151.html

This new version changes the way each chip select is described in the DT. Now there's a separate node for each Device Bus chip select, like this:

	devbus-bootcs at d0010400 {
		compatible = "marvell,mvebu-devbus";
		reg = <0xd0010400 0x8>;
		ranges = <0 0 0xf0000000 0x8000000>; /* CS0 @addr 0xf000000, size 0x8000000 */

		device at 0 {
			reg = <0 0 0x8000000>;
		}
	};

	// ...

	devbus-cs1 at d0010410 {
		compatible = "marvell,mvebu-devbus";
		reg = <0xd0010410 0x8>;
	};

In particular, please note that once the window handling is properly added to the DT, and we remove the mbus usage from this driver, in the above DT nodes the ranges property value will have to be changed, but the ABI will be the same.

This applies on top of v3.9-rc5 with mbus patches applied, which can be found in Jason Cooper's mvebus/soc branch.
If anyone wants to try this, there's a public branch available at:

https://github.com/MISL-EBU-System-SW/mainline-public/tree/mvebu-devbus-v4

How does this version look regarding its inclusion in v3.10? @Jason, @Arnd: if you think this v4 meets your expectation, I'd like to have your Ack so I can ask Jason Cooper to apply this.

If the driver looks good and the temporary window handling is accepted, then I can prepare a rebased patchset on top of Gregory's LPAE work, or on top of any other required patches.

Changes from v3:

* Rename devbus,dev-width to devbus,bus-width and changed the parameter value from bytes to bits: <8>, <16>, ... Suggested by Arnd Bergmann.

* As suggested by Jason Gunthorpe and agreed by Arnd Bergmann. Reworked the whole devbus child chip-select device tree design. Now there's a separate devbus at addr DT node for each controller chip select. The impact of this is:
1. Timing parameters now go into the parent node.
2. There can only be one child per devbus node.

* Remove unneeded bank-width parameter read.

Changes from v2:

* Replace the 'NOR' specific probe by a generic probe. Currently, we don't see anything specific to NOR in this driver. Suggested by Jason Gunthorpe.

* Add a check to detect that every timing parameter is specified in the device tree node.
Suggested by Jason Gunthorpe.

Changes from v1:

* Added a remove() function to unregister childs and remove address windows, as suggested by Thomas Petazzoni.

* Remove unneeded 'inline' prefix on functions as suggested by Thomas Petazzoni.

* Timings parameters are now mandatory and expressed in picoseconds, as requested by Jason Gunthorpe.
This is consistent with OMAP's GPMC dts binding.

* Fix documentation as suggested by Jason Gunthorpe.

Thanks!

Ezequiel Garcia (5):
drivers: memory: Introduce Marvell EBU Device Bus driver
ARM: mvebu: Add Device Bus support for Armada 370/XP SoC
ARM: mvebu: Add support for NOR flash device on Armada XP-GP board
ARM: mvebu: Add support for NOR flash device on Openblocks AX3 board
ARM: mvebu: Add Device Bus and CFI flash memory support to defconfig

 .../bindings/memory-controllers/mvebu-devbus.txt   | 156 ++++++++++
 arch/arm/boot/dts/armada-370-xp.dtsi               |  45 +++
 arch/arm/boot/dts/armada-xp-gp.dts                 |  29 ++
 arch/arm/boot/dts/armada-xp-openblocks-ax3-4.dts   |  29 ++
 arch/arm/configs/mvebu_defconfig                   |   7 +
 drivers/memory/Kconfig                             |  10 +
 drivers/memory/Makefile                            |   1 +
 drivers/memory/mvebu-devbus.c                      | 344 +++++++++++++++++++++
 8 files changed, 621 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/memory-controllers/mvebu-devbus.txt
 create mode 100644 drivers/memory/mvebu-devbus.c
Signature · 1 line
1.8.1.5
#1 lore raw

PATCH v4 1/5 drivers: memory: Introduce Marvell EBU Device Bus driver

Marvell EBU SoCs such as Armada 370/XP, Orion5x (88f5xxx) and Discovery (mv78xx0) supports a Device Bus controller to access several kinds of memories and I/O devices (NOR, NAND, SRAM, FPGA).

This commit adds a driver to handle this controller. So far only Armada 370, Armada XP and Discovery SoCs are supported.

The driver must be registered through a device tree node; as explained in the binding document.

Each Device Bus node in the device tree, represents a chip select. For each node, the driver will:
* set timing parameters
* register one child device
* setup an address decoding window, using the mbus driver

Keep in mind the address decoding window setup is only a temporary hack. This code will be removed from this devbus driver as soon as a proper device tree binding for the mbus driver is added.

---
 .../bindings/memory-controllers/mvebu-devbus.txt   | 156 ++++++++++
 drivers/memory/Kconfig                             |  10 +
 drivers/memory/Makefile                            |   1 +
 drivers/memory/mvebu-devbus.c                      | 344 +++++++++++++++++++++
 4 files changed, 511 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/memory-controllers/mvebu-devbus.txt
 create mode 100644 drivers/memory/mvebu-devbus.c
Documentation/devicetree/bindings/memory-controllers/mvebu-devbus.txt new file +156 -0
new file mode 100644
index 0000000..9ddcd6c
@@ -0,0 +1,156 @@
+Device tree bindings for MVEBU Device Bus controllers
+
+The Device Bus controller available in some Marvell's SoC allows to control
+different types of standard memory and I/O devices such as NOR, NAND, and FPGA.
+The actual devices are instantiated from the child nodes of a Device Bus node.
+
+Required properties:
+
+ - compatible:          Currently only Armada 370/XP SoC are supported,
+                        with this compatible string:
+
+                        marvell,mvebu-devbus
+
+ - reg:                 A resource specifier for the register space.
+                        This is the base address of a chip select within
+			the controller's register space.
+                        (see the example below)
+
+ - #address-cells:      Must be set to 2 to allow memory address translation
+ - #size-cells:         Must be set to 1 to allow CS address passing
+ - ranges:              Must be set up to reflect the memory layout with four
+                        integer values for each chip-select line in use:
+                        <cs-number> 0 <physical address of mapping> <size>
+
+Mandatory timing properties for child nodes:
+
+Read parameters:
+
+ - devbus,turn-off-ps:  Defines the time during which the controller does not
+                        drive the AD bus after the completion of a device read.
+                        This prevents contentions on the Device Bus after a read
+                        cycle from a slow device.
+
+ - devbus,bus-width:    Defines the bus width (e.g. <16>)
+
+ - devbus,badr-skew-ps: Defines the time delay from from A[2:0] toggle,
+                        to read data sample. This parameter is useful for
+                        synchronous pipelined devices, where the address
+                        precedes the read data by one or two cycles.
+
+ - devbus,acc-first-ps: Defines the time delay from the negation of
+                        ALE[0] to the cycle that the first read data is sampled
+                        by the controller.
+
+ - devbus,acc-next-ps:  Defines the time delay between the cycle that
+                        samples data N and the cycle that samples data N+1
+                        (in burst accesses).
+
+ - devbus,rd-setup-ps:  Defines the time delay between DEV_CSn assertion to
+			DEV_OEn assertion. If set to 0 (default),
+                        DEV_OEn and DEV_CSn are asserted at the same cycle.
+                        This parameter has no affect on <acc-first-ps> parameter
+                        (no affect on first data sample). Set <rd-setup-ps>
+                        to a value smaller than <acc-first-ps>.
+
+ - devbus,rd-hold-ps:   Defines the time between the last data sample to the
+			de-assertion of DEV_CSn. If set to 0 (default),
+			DEV_OEn and DEV_CSn are de-asserted at the same cycle
+			(the cycle of the last data sample).
+                        This parameter has no affect on DEV_OEn de-assertion.
+                        DEV_OEn is always de-asserted the next cycle after
+                        last data sampled. Also this parameter has no
+                        affect on <turn-off-ps> parameter.
+                        Set <rd-hold-ps> to a value smaller than <turn-off-ps>.
+
+Write parameters:
+
+ - devbus,ale-wr-ps:    Defines the time delay from the ALE[0] negation cycle
+			to the DEV_WEn assertion.
+
+ - devbus,wr-low-ps:    Defines the time during which DEV_WEn is active.
+                        A[2:0] and Data are kept valid as long as DEV_WEn
+                        is active. This parameter defines the setup time of
+                        address and data to DEV_WEn rise.
+
+ - devbus,wr-high-ps:   Defines the time during which DEV_WEn is kept
+                        inactive (high) between data beats of a burst write.
+                        DEV_A[2:0] and Data are kept valid (do not toggle) for
+                        <wr-high-ps> - <tick> ps.
+			This parameter defines the hold time of address and
+			data after DEV_WEn rise.
+
+ - devbus,sync-enable: Synchronous device enable.
+                       1: True
+                       0: False
+
+An example for an Armada XP GP board, with a 16 MiB NOR device as child
+is showed below. Note that the Device Bus driver is in charge of allocating
+the mbus address decoding window for each of its child devices.
+The window is created using the chip select specified in the child
+device node together with the base address and size specified in the ranges
+property. For instance, in the example below the allocated decoding window
+will start at base address 0xf0000000, with a size 0x1000000 (16 MiB)
+for chip select 0 (a.k.a DEV_BOOTCS).
+
+This address window handling is done in this mvebu-devbus only as a temporary
+solution. It will be removed when the support for mbus device tree binding is
+added.
+
+The reg property implicitly specifies the chip select as this:
+
+  0x10400: DEV_BOOTCS
+  0x10408: DEV_CS0
+  0x10410: DEV_CS1
+  0x10418: DEV_CS2
+  0x10420: DEV_CS3
+
+Example:
+
+	devbus-bootcs at d0010400 {
+		status = "okay";
+		ranges = <0 0 0xf0000000 0x1000000>; /* CS0 @addr 0xf0000000, size 0x1000000 */
+		#address-cells = <2>;
+		#size-cells = <1>;
+
+		/* Device Bus parameters are required */
+
+		/* Read parameters */
+		devbus,bus-width    = <8>;
+		devbus,turn-off-ps  = <60000>;
+		devbus,badr-skew-ps = <0>;
+		devbus,acc-first-ps = <124000>;
+		devbus,acc-next-ps  = <248000>;
+		devbus,rd-setup-ps  = <0>;
+		devbus,rd-hold-ps   = <0>;
+
+		/* Write parameters */
+		devbus,sync-enable = <0>;
+		devbus,wr-high-ps  = <60000>;
+		devbus,wr-low-ps   = <60000>;
+		devbus,ale-wr-ps   = <60000>;
+
+		flash at 0 {
+			compatible = "cfi-flash";
+
+			/* CS0, 16 MiB */
+			reg = <0 0 0x1000000>;
+			bank-width = <2>;
+			#address-cells = <1>;
+			#size-cells = <1>;
+
+			/*
+			 * We split the 16 MiB in two partitions,
+			 * just as an example.
+			 */
+			partition at 0 {
+				label = "First";
+				reg = <0 0x800000>;
+			};
+
+			partition at 800000 {
+				label = "Second";
+				reg = <0x800000 0x800000>;
+			};
+		};
+	};
drivers/memory/Kconfig +10 -0
index 067f311..9a81ab3 100644
@@ -20,6 +20,16 @@ config TI_EMIF
 	  parameters and other settings during frequency, voltage and
 	  temperature changes
 
+config MVEBU_DEVBUS
+	tristate "Marvell EBU Device Bus Controller"
+	default y
+	depends on PLAT_ORION && OF
+	help
+	  This driver is for the Device Bus controller available in some
+	  Marvell EBU SoCs such as Discovery (mv78xx0), Orion (88f5xxx) and
+	  Armada 370 and Armada XP. This controller allows to handle flash
+	  devices such as NOR, NAND, SRAM, and FPGA.
+
 config TEGRA20_MC
 	bool "Tegra20 Memory Controller(MC) driver"
 	default y
drivers/memory/Makefile +1 -0
index 9cce5d7..969d923 100644
@@ -6,5 +6,6 @@ ifeq ($(CONFIG_DDR),y)
 obj-$(CONFIG_OF)		+= of_memory.o
 endif
 obj-$(CONFIG_TI_EMIF)		+= emif.o
+obj-$(CONFIG_MVEBU_DEVBUS)	+= mvebu-devbus.o
 obj-$(CONFIG_TEGRA20_MC)	+= tegra20-mc.o
 obj-$(CONFIG_TEGRA30_MC)	+= tegra30-mc.o
drivers/memory/mvebu-devbus.c new file +344 -0

344 lines not on this page · raw message

Signature · 1 line
1.8.1.5
#2 lore raw

PATCH v4 1/5 drivers: memory: Introduce Marvell EBU Device Bus driver

Looks OK to me, though with mbus dt bindings we'd see some more changes.

Just to note on them for future:

drivers/memory/mvebu-devbus.c line 258

Quoted 5 lines
drivers/memory/mvebu-devbus.c new file +5 -0
+	child = of_get_next_child(node, NULL);
+	if (!child) {
+		dev_err(dev, "%s has no childs\n", node->full_name);
+		return -EINVAL;
+	}

This isn't really necessary anymore, the ranges of the node itself should be parsed to get the window start/end. The driver should never need to look at the children directly..

drivers/memory/mvebu-devbus.c line 280

Quoted 4 lines
drivers/memory/mvebu-devbus.c new file +4 -0
+	/* Get the CS to choose the window string */
+	err = of_property_read_u32(node, "ranges", &cs);
+	if (err < 0)
+		return err;

.. and this is why you've kept the 2 cells address - the top cell is still encoding the target id. This would go away eventually too..

Maybe it would be better in the interm to compute the CS offset from the control register offset?

(reg.start % 0x400)/8 should do the trick?

drivers/memory/mvebu-devbus.c line 300

Quoted 7 lines
drivers/memory/mvebu-devbus.c new file +7 -0
+	devbus->child = of_platform_device_create(child, NULL, &pdev->dev);
+	if (!devbus->child) {
+		dev_warn(dev, "cannot create child device %s\n", child->name);
+		/* Remove the allocated window */
+		mvebu_mbus_del_window(devbus->child_mem.start,
+				      resource_size(&devbus->child_mem));
+	}

This can probably just be of_platform_populate or similar to do all children? For instance, I often use many DT nodes to represent a FPGA, since the my FPGA's tend to have many functionally orthogonal units inside.

Regards,
Jason

#3 lore raw

PATCH v4 1/5 drivers: memory: Introduce Marvell EBU Device Bus driver

Quoted 5 lines · Jason Gunthorpe

drivers/memory/mvebu-devbus.c line 280

Quoted 4 lines
drivers/memory/mvebu-devbus.c new file +4 -0
+     /* Get the CS to choose the window string */
+     err = of_property_read_u32(node, "ranges", &cs);
+     if (err < 0)
+             return err;
Quoted 7 lines

.. and this is why you've kept the 2 cells address - the top cell is still encoding the target id. This would go away eventually too..

Maybe it would be better in the interm to compute the CS offset from the control register offset?

(reg.start % 0x400)/8 should do the trick?

Yes, we came to the axact same conclusion on IRC.

Quoted 8 lines

drivers/memory/mvebu-devbus.c line 300

Quoted 7 lines
drivers/memory/mvebu-devbus.c new file +7 -0
+     devbus->child = of_platform_device_create(child, NULL, &pdev->dev);
+     if (!devbus->child) {
+             dev_warn(dev, "cannot create child device %s\n", child->name);
+             /* Remove the allocated window */
+             mvebu_mbus_del_window(devbus->child_mem.start,
+                                   resource_size(&devbus->child_mem));
+     }
Quoted 4 lines

This can probably just be of_platform_populate or similar to do all children? For instance, I often use many DT nodes to represent a FPGA, since the my FPGA's tend to have many functionally orthogonal units inside.

Good idea, yes.

Arnd

#4 lore raw

PATCH v4 1/5 drivers: memory: Introduce Marvell EBU Device Bus driver

In reply to Jason Gunthorpe

Hi Jason

Quoted 3 lines · Jason Gunthorpe

Looks OK to me, though with mbus dt bindings we'd see some more changes.

Thanks for the review!

Quoted 12 lines

Just to note on them for future:

drivers/memory/mvebu-devbus.c line 258

Quoted 5 lines
drivers/memory/mvebu-devbus.c new file +5 -0
+	child = of_get_next_child(node, NULL);
+	if (!child) {
+		dev_err(dev, "%s has no childs\n", node->full_name);
+		return -EINVAL;
+	}

This isn't really necessary anymore, the ranges of the node itself should be parsed to get the window start/end. The driver should never need to look at the children directly..

You mean the above will not be necessary in the *future*, once we add mbus DT bindings, right?

AFAICT, it's needed for the current workaround to work. FWIW, the whole code below the *FIXME* is part of the temporary to-be-removed thing.

Maybe I'll add a special comment on each chunk that's meant to be removed to avoid confusions.

Quoted 8 lines

drivers/memory/mvebu-devbus.c line 280

Quoted 4 lines
drivers/memory/mvebu-devbus.c new file +4 -0
+	/* Get the CS to choose the window string */
+	err = of_property_read_u32(node, "ranges", &cs);
+	if (err < 0)
+		return err;

.. and this is why you've kept the 2 cells address - the top cell is still encoding the target id. This would go away eventually too..

Quoted 5 lines

Maybe it would be better in the interm to compute the CS offset from the control register offset?

(reg.start % 0x400)/8 should do the trick?

Yes, I'll do this and drop the 2-cell address.

Quoted 4 lines
Quoted 4 lines
drivers/memory/mvebu-devbus.c new file +4 -0
+	devbus->child = of_platform_device_create(child, NULL, &pdev->dev);
+	if (!devbus->child) {
+		dev_warn(dev, "cannot create child device %s\n", child->name);
+		/* Remove the allocated window */
Quoted 9 lines

drivers/memory/mvebu-devbus.c line 300

Quoted 3 lines
drivers/memory/mvebu-devbus.c new file +3 -0
+		mvebu_mbus_del_window(devbus->child_mem.start,
+				      resource_size(&devbus->child_mem));
+	}

This can probably just be of_platform_populate or similar to do all children? For instance, I often use many DT nodes to represent a FPGA, since the my FPGA's tend to have many functionally orthogonal units inside.

In the past I've found it's not possible to use of_platform_populate, altough I can be wrong.

The problem seems to be that through of_platform_populate() it's possible that the child device (cfi-flash, for instance) gets probed before the devbus controller. In other words, the actual parent-child relationship gets lost and it's necesarry to use some child probe deferal mechanism.

This has been recently discussed [1] when implementing GPMC where for simplicity the of_platform_device_create() solution was chosen.
Anyway, I'll try of_platform_populate early tomorrow.

Thanks,

[1] http://www.mail-archive.com/linux-omap at vger.kernel.org/msg85897.html

Signature · 3 lines
Ezequiel Garc?a, Free Electrons
Embedded Linux, Kernel and Android Engineering
http://free-electrons.com
#5 lore raw

PATCH v4 1/5 drivers: memory: Introduce Marvell EBU Device Bus driver

Quoted 15 lines · Ezequiel Garcia
Quoted 5 lines

This can probably just be of_platform_populate or similar to do all children? For instance, I often use many DT nodes to represent a FPGA, since the my FPGA's tend to have many functionally orthogonal units inside.

In the past I've found it's not possible to use of_platform_populate, altough I can be wrong.

The problem seems to be that through of_platform_populate() it's possible that the child device (cfi-flash, for instance) gets probed before the devbus controller. In other words, the actual parent-child relationship gets lost and it's necesarry to use some child probe deferal mechanism.

Quoted 3 lines

This has been recently discussed [1] when implementing GPMC where for simplicity the of_platform_device_create() solution was chosen.
Anyway, I'll try of_platform_populate early tomorrow.

I think the problem mentioned in that thread was that the device would automatically get probed by the top-level of_platform_populate() when you mark the devbus as compatible="simple-bus", which is not what Jason was referring to. Instead you should call of_platform_populate in the probe() function of the devbus device to probe its children after the bus is set up.

Arnd

#6 lore raw

PATCH v4 1/5 drivers: memory: Introduce Marvell EBU Device Bus driver

Hi Arnd,

Quoted 20 lines · Arnd Bergmann
Quoted 15 lines · Ezequiel Garcia
Quoted 5 lines

This can probably just be of_platform_populate or similar to do all children? For instance, I often use many DT nodes to represent a FPGA, since the my FPGA's tend to have many functionally orthogonal units inside.

In the past I've found it's not possible to use of_platform_populate, altough I can be wrong.

The problem seems to be that through of_platform_populate() it's possible that the child device (cfi-flash, for instance) gets probed before the devbus controller. In other words, the actual parent-child relationship gets lost and it's necesarry to use some child probe deferal mechanism.

Quoted 3 lines

This has been recently discussed [1] when implementing GPMC where for simplicity the of_platform_device_create() solution was chosen.
Anyway, I'll try of_platform_populate early tomorrow.

Quoted 7 lines

I think the problem mentioned in that thread was that the device would automatically get probed by the top-level of_platform_populate() when you mark the devbus as compatible="simple-bus", which is not what Jason was referring to. Instead you should call of_platform_populate in the probe() function of the devbus device to probe its children after the bus is set up.

Ah! yes, you're right...

Well, in that case the only issue I can foresee is that if we decide to use of_platform_populate we won't be able to unregister child devices from the remove() callback.

Indeed, the benefits of using of_platform_populate are interesting, but I don't know how much of an issue this represents.

If we can't unregister child devices, we can't remove address windows. Now, this is not a big deal, since we plan to define them statically in the DT anyway.
In that case, it seems we shouldn't allow this driver to be a module, uh?

(actually we currently can't have mvebu-devbus as a module, because mbus API is not exported, but we can fix that if we want).

What do you think?

Signature · 3 lines
Ezequiel Garc?a, Free Electrons
Embedded Linux, Kernel and Android Engineering
http://free-electrons.com
#7 lore raw

PATCH v4 1/5 drivers: memory: Introduce Marvell EBU Device Bus driver

Quoted 9 lines · Ezequiel Garcia

Ah! yes, you're right...

Well, in that case the only issue I can foresee is that if we decide to use of_platform_populate we won't be able to unregister child devices from the remove() callback.

Indeed, the benefits of using of_platform_populate are interesting, but I don't know how much of an issue this represents.

Quoted 7 lines

If we can't unregister child devices, we can't remove address windows. Now, this is not a big deal, since we plan to define them statically in the DT anyway.
In that case, it seems we shouldn't allow this driver to be a module, uh?

(actually we currently can't have mvebu-devbus as a module, because mbus API is not exported, but we can fix that if we want).

I think it's still reasonable to make it a module, but it might need to be one without a module_exit() call to prevent unloading.

We could also try to add the opposite of of_platform_populate to remove an entire subtree.

Arnd

#8 lore raw

PATCH v4 1/5 drivers: memory: Introduce Marvell EBU Device Bus driver

Quoted 15 lines · Arnd Bergmann
Quoted 14 lines · Ezequiel Garcia

Ah! yes, you're right...

Well, in that case the only issue I can foresee is that if we decide to use of_platform_populate we won't be able to unregister child devices from the remove() callback.

Indeed, the benefits of using of_platform_populate are interesting, but I don't know how much of an issue this represents.

If we can't unregister child devices, we can't remove address windows. Now, this is not a big deal, since we plan to define them statically in the DT anyway.
In that case, it seems we shouldn't allow this driver to be a module, uh?

Quoted 9 lines
Quoted 2 lines

(actually we currently can't have mvebu-devbus as a module, because mbus API is not exported, but we can fix that if we want).

I think it's still reasonable to make it a module, but it might need to be one without a module_exit() call to prevent unloading.

We could also try to add the opposite of of_platform_populate to remove an entire subtree.

Yes, good idea.

I'll send a v5 in a minute, with your acked-by, maintaining the of_platform_device_create for now.

I think at this point it's reasonable to add of_platform_populate as a follow-up patch, together with the mbus-related fixes.

Thanks,

Signature · 3 lines
Ezequiel Garc?a, Free Electrons
Embedded Linux, Kernel and Android Engineering
http://free-electrons.com
#9 lore raw

PATCH v4 1/5 drivers: memory: Introduce Marvell EBU Device Bus driver

In reply to Arnd Bergmann
Quoted 2 lines · Arnd Bergmann

I think it's still reasonable to make it a module, but it might need to be one without a module_exit() call to prevent unloading.

Agree, I think it is better to properly handle a broader set of possible DT's than to have module unloading. The DT is supposed to be stable..

Jason

#10 lore raw

PATCH v4 1/5 drivers: memory: Introduce Marvell EBU Device Bus driver

In reply to Ezequiel Garcia
Quoted 13 lines · Ezequiel Garcia
Quoted 12 lines

Just to note on them for future:

drivers/memory/mvebu-devbus.c line 258

Quoted 5 lines
drivers/memory/mvebu-devbus.c new file +5 -0
+	child = of_get_next_child(node, NULL);
+	if (!child) {
+		dev_err(dev, "%s has no childs\n", node->full_name);
+		return -EINVAL;
+	}

This isn't really necessary anymore, the ranges of the node itself should be parsed to get the window start/end. The driver should never need to look at the children directly..

Quoted 2 lines

You mean the above will not be necessary in the *future*, once we add mbus DT bindings, right?

No, you could do this now.

Look at your binding:

+ devbus-bootcs at d0010400 {
+ status = "okay";
+ ranges = <0 0xf0000000 0x1000000>; /* @addr 0xf000000, size 0x1000000 */

Parse that ranges and you get the base and size of the window.

Then 'ranges' is mandatory and 'ranges' must have a single entry.

No need to look at children at all. You can remove all the child parsing code, and then the driver is able to handle a wider range of possible DT uses.

Jason

#11 lore raw

PATCH v4 1/5 drivers: memory: Introduce Marvell EBU Device Bus driver

Hi Jason,

Quoted 5 lines · Jason Gunthorpe

No need to look at children at all. You can remove all the child parsing code, and then the driver is able to handle a wider range of possible DT uses.

Agreed. I just sent a v6 fixing this and using of_platform_populate() to create every child available.

How does it look?

Thanks for the feedback!

Signature · 3 lines
Ezequiel Garc?a, Free Electrons
Embedded Linux, Kernel and Android Engineering
http://free-electrons.com
#12 lore raw

PATCH v4 2/5 ARM: mvebu: Add Device Bus support for Armada 370/XP SoC

In reply to Ezequiel Garcia

Armada 370 and Armada XP SoC have a Device Bus controller to handle NOR, NAND, SRAM and FPGA devices.

This patch adds the device tree nodes to enable the controller, where each node in the device tree, represents a chip select.

---
 arch/arm/boot/dts/armada-370-xp.dtsi | 45 ++++++++++++++++++++++++++++++++++++
 1 file changed, 45 insertions(+)
arch/arm/boot/dts/armada-370-xp.dtsi +45 -0
index 7704829..fe90b64 100644
@@ -176,6 +176,51 @@
 			clocks = <&coreclk 0>;
 			status = "disabled";
 		};
+
+		devbus-bootcs at d0010400 {
+			compatible = "marvell,mvebu-devbus";
+			reg = <0xd0010400 0x8>;
+			#address-cells = <2>;
+			#size-cells = <1>;
+			clocks = <&coreclk 0>;
+			status = "disabled";
+		};
+
+		devbus-cs0 at d0010408 {
+			compatible = "marvell,mvebu-devbus";
+			reg = <0xd0010408 0x8>;
+			#address-cells = <2>;
+			#size-cells = <1>;
+			clocks = <&coreclk 0>;
+			status = "disabled";
+		};
+
+		devbus-cs1 at d0010410 {
+			compatible = "marvell,mvebu-devbus";
+			reg = <0xd0010410 0x8>;
+			#address-cells = <2>;
+			#size-cells = <1>;
+			clocks = <&coreclk 0>;
+			status = "disabled";
+		};
+
+		devbus-cs2 at d0010418 {
+			compatible = "marvell,mvebu-devbus";
+			reg = <0xd0010418 0x8>;
+			#address-cells = <2>;
+			#size-cells = <1>;
+			clocks = <&coreclk 0>;
+			status = "disabled";
+		};
+
+		devbus-cs3 at d0010420 {
+			compatible = "marvell,mvebu-devbus";
+			reg = <0xd0010420 0x8>;
+			#address-cells = <2>;
+			#size-cells = <1>;
+			clocks = <&coreclk 0>;
+			status = "disabled";
+		};
 	};
 };
 
Signature · 1 line
1.8.1.5
#13 lore raw

PATCH v4 3/5 ARM: mvebu: Add support for NOR flash device on Armada XP-GP board

In reply to Ezequiel Garcia

The Armada XP Development Board DB-MV784MP-GP has a NOR flash device connected to the Device Bus. This commit adds the device tree node to support this device.

This SoC supports a flexible and dynamic decoding window allocation scheme; but since this feature is still not implemented we need to specify the window base address in the device tree node itself.

This base address has been selected in a completely arbitrary fashion.

---
 arch/arm/boot/dts/armada-xp-gp.dts | 29 +++++++++++++++++++++++++++++
 1 file changed, 29 insertions(+)
arch/arm/boot/dts/armada-xp-gp.dts +29 -0
index 1c8afe2..39f6a4d 100644
@@ -109,5 +109,34 @@
 				spi-max-frequency = <108000000>;
 			};
 		};
+
+		devbus-bootcs at d0010400 {
+			status = "okay";
+			ranges = <0 0 0xf0000000 0x1000000>; /* CS0 @addr 0xf000000, size 0x1000000 */
+
+			/* Device Bus parameters are required */
+
+			/* Read parameters */
+			devbus,bus-width    = <8>;
+			devbus,turn-off-ps  = <60000>;
+			devbus,badr-skew-ps = <0>;
+			devbus,acc-first-ps = <124000>;
+			devbus,acc-next-ps  = <248000>;
+			devbus,rd-setup-ps  = <0>;
+			devbus,rd-hold-ps   = <0>;
+
+			/* Write parameters */
+			devbus,sync-enable = <0>;
+			devbus,wr-high-ps  = <60000>;
+			devbus,wr-low-ps   = <60000>;
+			devbus,ale-wr-ps   = <60000>;
+
+			/* NOR 16 MiB */
+			nor at 0 {
+				compatible = "cfi-flash";
+				reg = <0 0 0x1000000>;
+				bank-width = <2>;
+			};
+		};
 	};
 };
Signature · 1 line
1.8.1.5
#14 lore raw

PATCH v4 4/5 ARM: mvebu: Add support for NOR flash device on Openblocks AX3 board

In reply to Ezequiel Garcia

The Plat'home Openblocks AX3 has a 128 MiB NOR flash device connected to the Device Bus. This commit adds the device tree node to support this device.

The SoC supports a flexible and dynamic decoding window allocation scheme; but since this feature is still not implemented we need to specify the window base address in the device tree node itself.
This base address has been selected in a completely arbitrary fashion.

---
 arch/arm/boot/dts/armada-xp-openblocks-ax3-4.dts | 29 ++++++++++++++++++++++++
 1 file changed, 29 insertions(+)
arch/arm/boot/dts/armada-xp-openblocks-ax3-4.dts +29 -0
index 3818a82..04576c2 100644
@@ -139,5 +139,34 @@
 		usb at d0051000 {
 			status = "okay";
 		};
+
+		devbus-bootcs at d0010400 {
+			status = "okay";
+			ranges = <0 0 0xf0000000 0x8000000>; /* CS0 @addr 0xf000000, size 0x8000000 */
+
+			/* Device Bus parameters are required */
+
+			/* Read parameters */
+			devbus,bus-width    = <8>;
+			devbus,turn-off-ps  = <60000>;
+			devbus,badr-skew-ps = <0>;
+			devbus,acc-first-ps = <124000>;
+			devbus,acc-next-ps  = <248000>;
+			devbus,rd-setup-ps  = <0>;
+			devbus,rd-hold-ps   = <0>;
+
+			/* Write parameters */
+			devbus,sync-enable = <0>;
+			devbus,wr-high-ps  = <60000>;
+			devbus,wr-low-ps   = <60000>;
+			devbus,ale-wr-ps   = <60000>;
+
+			/* NOR 128 MiB */
+			nor at 0 {
+				compatible = "cfi-flash";
+				reg = <0 0 0x8000000>;
+				bank-width = <2>;
+			};
+		};
 	};
 };
Signature · 1 line
1.8.1.5
#15 lore raw

PATCH v4 5/5 ARM: mvebu: Add Device Bus and CFI flash memory support to defconfig

In reply to Ezequiel Garcia

This patch selects the devbus driver as part of the mvebu default config, along with the required options to detect and support CFI flash memories.

---
 arch/arm/configs/mvebu_defconfig | 7 +++++++
 1 file changed, 7 insertions(+)
arch/arm/configs/mvebu_defconfig +7 -0
index 2ec8119..fe8ba26 100644
@@ -46,6 +46,11 @@ CONFIG_I2C_MV64XXX=y
 CONFIG_MTD=y
 CONFIG_MTD_CHAR=y
 CONFIG_MTD_M25P80=y
+CONFIG_MTD_CFI=y
+CONFIG_MTD_CFI_INTELEXT=y
+CONFIG_MTD_CFI_AMDSTD=y
+CONFIG_MTD_CFI_STAA=y
+CONFIG_MTD_PHYSMAP_OF=y
 CONFIG_SERIAL_8250_DW=y
 CONFIG_GPIOLIB=y
 CONFIG_GPIO_SYSFS=y
@@ -65,6 +70,8 @@ CONFIG_RTC_DRV_S35390A=y
 CONFIG_RTC_DRV_MV=y
 CONFIG_DMADEVICES=y
 CONFIG_MV_XOR=y
+CONFIG_MEMORY=y
+CONFIG_MVEBU_DEVBUS=y
 # CONFIG_IOMMU_SUPPORT is not set
 CONFIG_EXT2_FS=y
 CONFIG_EXT3_FS=y
Signature · 1 line
1.8.1.5
#16 lore raw

v4 cover Device Bus support for Marvell EBU SoC

In reply to Ezequiel Garcia
Quoted 4 lines · Ezequiel Garcia

* Rename devbus,dev-width to devbus,bus-width and changed the parameter value from bytes to bits: <8>, <16>, ... Suggested by Arnd Bergmann.

Quoted 9 lines

* As suggested by Jason Gunthorpe and agreed by Arnd Bergmann. Reworked the whole devbus child chip-select device tree design. Now there's a separate devbus at addr DT node for each controller chip select. The impact of this is:
1. Timing parameters now go into the parent node.
2. There can only be one child per devbus node.

* Remove unneeded bank-width parameter read.

Whole series:

It may be nicer to use #address-cells=<2> for devbus device nodes now that each of them has only a 32 bit address space, but the current code is correct as well using a 64 bit notation.

Arnd

#17 lore raw

v4 cover Device Bus support for Marvell EBU SoC

Quoted 15 lines · Arnd Bergmann
Quoted 4 lines · Ezequiel Garcia

* Rename devbus,dev-width to devbus,bus-width and changed the parameter value from bytes to bits: <8>, <16>, ... Suggested by Arnd Bergmann.

Quoted 9 lines

* As suggested by Jason Gunthorpe and agreed by Arnd Bergmann. Reworked the whole devbus child chip-select device tree design. Now there's a separate devbus at addr DT node for each controller chip select. The impact of this is:
1. Timing parameters now go into the parent node.
2. There can only be one child per devbus node.

* Remove unneeded bank-width parameter read.

Quoted 8 lines

Whole series:

It may be nicer to use #address-cells=<2> for devbus device nodes now that each of them has only a 32 bit address space, but the current code is correct as well using a 64 bit notation.

I'll probably be sending a v5 fixing this later.

@Jason Cooper:
How would you like to handle this, regarding the LPAE series? Do you want me to rebase this devbus on top of that?

Signature · 3 lines
Ezequiel Garc?a, Free Electrons
Embedded Linux, Kernel and Android Engineering
http://free-electrons.com
#18 lore raw

v4 cover Device Bus support for Marvell EBU SoC

Quoted 25 lines · Ezequiel Garcia
Quoted 15 lines · Arnd Bergmann
Quoted 4 lines · Ezequiel Garcia

* Rename devbus,dev-width to devbus,bus-width and changed the parameter value from bytes to bits: <8>, <16>, ... Suggested by Arnd Bergmann.

Quoted 9 lines

* As suggested by Jason Gunthorpe and agreed by Arnd Bergmann. Reworked the whole devbus child chip-select device tree design. Now there's a separate devbus at addr DT node for each controller chip select. The impact of this is:
1. Timing parameters now go into the parent node.
2. There can only be one child per devbus node.

* Remove unneeded bank-width parameter read.

Quoted 8 lines

Whole series:

It may be nicer to use #address-cells=<2> for devbus device nodes now that each of them has only a 32 bit address space, but the current code is correct as well using a 64 bit notation.

Quoted 5 lines

I'll probably be sending a v5 fixing this later.

@Jason Cooper:
How would you like to handle this, regarding the LPAE series? Do you want me to rebase this devbus on top of that?

Let's avoid the dependency and merge them both as is. Once v3.10 drops, you can send a fixup patch to move it to 64bit.

thx,

Jason.

#19 lore raw

v4 cover Device Bus support for Marvell EBU SoC

Quoted 26 lines · Jason Cooper
Quoted 25 lines · Ezequiel Garcia
Quoted 23 lines · Arnd Bergmann
Quoted 4 lines · Ezequiel Garcia

* Rename devbus,dev-width to devbus,bus-width and changed the parameter value from bytes to bits: <8>, <16>, ... Suggested by Arnd Bergmann.

Quoted 9 lines

* As suggested by Jason Gunthorpe and agreed by Arnd Bergmann. Reworked the whole devbus child chip-select device tree design. Now there's a separate devbus at addr DT node for each controller chip select. The impact of this is:
1. Timing parameters now go into the parent node.
2. There can only be one child per devbus node.

* Remove unneeded bank-width parameter read.

Whole series:

It may be nicer to use #address-cells=<2> for devbus device nodes now that each of them has only a 32 bit address space, but the current code is correct as well using a 64 bit notation.

Quoted 8 lines
Quoted 5 lines

I'll probably be sending a v5 fixing this later.

@Jason Cooper:
How would you like to handle this, regarding the LPAE series? Do you want me to rebase this devbus on top of that?

Let's avoid the dependency and merge them both as is. Once v3.10 drops, you can send a fixup patch to move it to 64bit.

After talking with gregory, I'll merge everything as-is, then have him rebase the LPAE changes on top of mvebu/dt.

thx,

Jason.