Skip to contentSkip to conversation

RPi mailbox support

PATCH Applied Andre Heider 2 participants 16 messages Oct 19, 2013 – Oct 22, 2013

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

cover RPi mailbox support

This adds a bcm2835 mailbox driver for the RPi to talk to the on-SoC VideoCore. See [1] for a description.

This patchset uses it to get two properties, the eMMC clock and the memory size.

The RPi backend needs to use this driver early on, namely before the MMU setup. We also want to use the driver in the future after the MMU setup, like setting up simplefb for the kernel. Patch 1 and 2 lay the groundwork for that.

The mailbox driver also needs to handle timouts, so patch 5 and 6 are making the clocksource accessible earlier.

Patch 7 is the driver itself, providing a helper macro for users based on a new common macro from patch 3.

Patch 8 and 9 switch the bc2835_mci driver from a local mailbox implementation to the new driver.

Patch 10 uses the driver to get the proper memory size.

Patch 4 is just a cleanup.

Thanks,
Andre

[1] https://github.com/raspberrypi/firmware/wiki/Mailboxes

Andre Heider (10):
ARM: cache: restore cache functions from the PBL
ARM: cache: do not crash when the MMU isn't yet setup
common: add a macro to align an array on the stack
ARM: bcm2835: cleanup clock registering
ARM: bcm2835: register the clocksource driver earlier
ARM: bcm2835: register the clocksource device earlier
ARM: bcm2835: add a mailbox driver for VideoCore
ARM: rpi: register a clkdev for the eMMC clock
mci: bcm2835: use the registered device clkdev
ARM: rpi: use the proper ARM memory size

 arch/arm/boards/raspberry-pi/rpi.c        |  75 +++++-
 arch/arm/cpu/cache.c                      |  18 +-
 arch/arm/cpu/start.c                      |   4 +-
 arch/arm/mach-bcm2835/Makefile            |   2 +-
 arch/arm/mach-bcm2835/core.c              |  32 +--
 arch/arm/mach-bcm2835/include/mach/mbox.h | 420 ++++++++++++++++++++++++++++++
 arch/arm/mach-bcm2835/mbox.c              | 152 +++++++++++
 drivers/clocksource/bcm2835.c             |   2 +-
 drivers/mci/mci-bcm2835.c                 |  93 ++-----
 drivers/mci/mci-bcm2835.h                 |  48 ----
 include/common.h                          |  11 +
 11 files changed, 700 insertions(+), 157 deletions(-)
 create mode 100644 arch/arm/mach-bcm2835/include/mach/mbox.h
 create mode 100644 arch/arm/mach-bcm2835/mbox.c
Signature · 7 lines
1.8.3.2


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#1 lore raw

PATCH 1/10 ARM: cache: restore cache functions from the PBL

When using CONFIG_MMU_EARLY combined with CONFIG_PBL_IMAGE, the barebox setup reuses the MMU setup from the PBL, but doesn't setup the cache functions.

Set these up to guarantee proper early cache handing before mmu_initcall().

---
 arch/arm/cpu/start.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
arch/arm/cpu/start.c +3 -1
index f8d343f..f0a7df3 100644
@@ -70,7 +70,9 @@ static noinline __noreturn void __start(uint32_t membase, uint32_t memsize,
 		endmem &= ~0x3fff;
 		endmem -= SZ_16K; /* ttb */
 
-		if (!IS_ENABLED(CONFIG_PBL_IMAGE)) {
+		if (IS_ENABLED(CONFIG_PBL_IMAGE)) {
+			arm_set_cache_functions();
+		} else {
 			arm_early_mmu_cache_invalidate();
 			mmu_early_enable(membase, memsize, endmem);
 		}
Signature · 7 lines
1.8.3.2


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#2 lore raw

PATCH 2/10 ARM: cache: do not crash when the MMU isn't yet setup

In reply to Andre Heider

Drivers currently cannot implement explicit cache handling and rely on running the same code before and after mmu_initcall() without crashing.

Depending on the chosen config options, the cache functions are not yet setup and using them early on ends in a null pointer dereference.

The RPi's mailbox driver is such a case; it requires cache handling once the MMU is fully set up and yet the RPi setup needs to use the driver to get the memory size before mem_initcall() and hence mmu_initcall().

Fix this by checking the cache_fns pointer before dereferencing it.

---
 arch/arm/cpu/cache.c | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)
arch/arm/cpu/cache.c +12 -6
index 7aab55b..223e308 100644
@@ -41,32 +41,38 @@ DEFINE_CPU_FNS(v7)
 
 void __dma_clean_range(unsigned long start, unsigned long end)
 {
-	cache_fns->dma_clean_range(start, end);
+	if (cache_fns)
+		cache_fns->dma_clean_range(start, end);
 }
 
 void __dma_flush_range(unsigned long start, unsigned long end)
 {
-	cache_fns->dma_flush_range(start, end);
+	if (cache_fns)
+		cache_fns->dma_flush_range(start, end);
 }
 
 void __dma_inv_range(unsigned long start, unsigned long end)
 {
-	cache_fns->dma_inv_range(start, end);
+	if (cache_fns)
+		cache_fns->dma_inv_range(start, end);
 }
 
 void __mmu_cache_on(void)
 {
-	cache_fns->mmu_cache_on();
+	if (cache_fns)
+		cache_fns->mmu_cache_on();
 }
 
 void __mmu_cache_off(void)
 {
-	cache_fns->mmu_cache_off();
+	if (cache_fns)
+		cache_fns->mmu_cache_off();
 }
 
 void __mmu_cache_flush(void)
 {
-	cache_fns->mmu_cache_flush();
+	if (cache_fns)
+		cache_fns->mmu_cache_flush();
 }
 
 int arm_set_cache_functions(void)
Signature · 7 lines
1.8.3.2


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#3 lore raw

PATCH 3/10 common: add a macro to align an array on the stack

In reply to Andre Heider

The macro can be used for temporary stack buffers which need to meet a minimum alignment requirement.

This will be used by bcm2835 mailbox users, where all buffers need to be aligned.

---
 include/common.h | 11 +++++++++++
 1 file changed, 11 insertions(+)
include/common.h +11 -0
index 0f0ba08..00f1642 100644
@@ -192,6 +192,17 @@ int run_shell(void);
 #define ARRAY_SIZE(arr)		(sizeof(arr) / sizeof((arr)[0]) + __must_be_array(arr))
 #define ARRAY_AND_SIZE(x)	(x), ARRAY_SIZE(x)
 
+/*
+ * The STACK_ALIGN_ARRAY macro is used to allocate a buffer on the stack that
+ * meets a minimum alignment requirement.
+ *
+ * Note that the size parameter is the number of array elements to allocate,
+ * not the number of bytes.
+ */
+#define STACK_ALIGN_ARRAY(type, name, size, align)		\
+	char __##name[sizeof(type) * (size) + (align) - 1];	\
+	type *name = (type *)ALIGN((uintptr_t)__##name, align)
+
 /**
  * container_of - cast a member of a structure out to the containing structure
  * @ptr:	the pointer to the member.
Signature · 7 lines
1.8.3.2


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#4 lore raw

PATCH 4/10 ARM: bcm2835: cleanup clock registering

In reply to Andre Heider

Sync exposed names while at it.

---
 arch/arm/mach-bcm2835/core.c | 31 +++++++------------------------
 1 file changed, 7 insertions(+), 24 deletions(-)
arch/arm/mach-bcm2835/core.c +7 -24
index 906e434..6835960 100644
@@ -32,37 +32,20 @@
 #include <mach/core.h>
 #include <linux/amba/bus.h>
 
-enum brcm_clks {
-	dummy, clk_ref_3, clk_ref_1, clks_max
-};
-
-static struct clk *clks[clks_max];
-
 static int bcm2835_clk_init(void)
 {
-	int ret;
-
-	clks[dummy] = clk_fixed("dummy", 0);
-	clks[clk_ref_3] = clk_fixed("ref3", 3 * 1000 * 1000);
-	clks[clk_ref_1] = clk_fixed("ref1", 1 * 1000 * 1000);
+	struct clk *clk;
 
-	ret = clk_register_clkdev(clks[dummy], "apb_pclk", NULL);
-	if (ret)
-		goto clk_err;
+	clk = clk_fixed("apb_pclk", 0);
+	clk_register_clkdev(clk, "apb_pclk", NULL);
 
-	ret = clk_register_clkdev(clks[clk_ref_3], NULL, "uart0-pl0110");
-	if (ret)
-		goto clk_err;
+	clk = clk_fixed("uart0-pl0110", 3 * 1000 * 1000);
+	clk_register_clkdev(clk, NULL, "uart0-pl0110");
 
-	ret = clk_register_clkdev(clks[clk_ref_1], NULL, "bcm2835-cs");
-	if (ret)
-		goto clk_err;
+	clk = clk_fixed("bcm2835-cs", 1 * 1000 * 1000);
+	clk_register_clkdev(clk, NULL, "bcm2835-cs");
 
 	return 0;
-
-clk_err:
-	return ret;
-
 }
 postcore_initcall(bcm2835_clk_init);
 
Signature · 7 lines
1.8.3.2


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#5 lore raw

PATCH 5/10 ARM: bcm2835: register the clocksource driver earlier

In reply to Andre Heider

RPi's mailbox driver is used early and it needs clock functions to handle timeouts.

Promote to a core_initcall().

---
 drivers/clocksource/bcm2835.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/clocksource/bcm2835.c +1 -1
index d5fe0f0..d1df3d2 100644
@@ -87,4 +87,4 @@ static int bcm2835_cs_init(void)
 {
 	return platform_driver_register(&bcm2835_cs_driver);
 }
-coredevice_initcall(bcm2835_cs_init);
+core_initcall(bcm2835_cs_init);
Signature · 7 lines
1.8.3.2


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#6 lore raw

PATCH 6/10 ARM: bcm2835: register the clocksource device earlier

In reply to Andre Heider

RPi's mailbox driver is used early and it needs clock functions to handle timeouts.

Register the driver straight after its clkdev.

---
 arch/arm/mach-bcm2835/core.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
arch/arm/mach-bcm2835/core.c +2 -1
index 6835960..cba7bf6 100644
@@ -45,6 +45,8 @@ static int bcm2835_clk_init(void)
 	clk = clk_fixed("bcm2835-cs", 1 * 1000 * 1000);
 	clk_register_clkdev(clk, NULL, "bcm2835-cs");
 
+	add_generic_device("bcm2835-cs", DEVICE_ID_SINGLE, NULL, BCM2835_ST_BASE, 0x1C, IORESOURCE_MEM, NULL);
+
 	return 0;
 }
 postcore_initcall(bcm2835_clk_init);
@@ -52,7 +54,6 @@ postcore_initcall(bcm2835_clk_init);
 static int bcm2835_dev_init(void)
 {
 	add_generic_device("bcm2835-gpio", 0, NULL, BCM2835_GPIO_BASE, 0xB0, IORESOURCE_MEM, NULL);
-	add_generic_device("bcm2835-cs", DEVICE_ID_SINGLE, NULL, BCM2835_ST_BASE, 0x1C, IORESOURCE_MEM, NULL);
 	add_generic_device("bcm2835_mci", 0, NULL, BCM2835_EMMC_BASE, 0xFC, IORESOURCE_MEM, NULL);
 	return 0;
 }
Signature · 7 lines
1.8.3.2


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#7 lore raw

PATCH 7/10 ARM: bcm2835: add a mailbox driver for VideoCore

In reply to Andre Heider

This allows exchanging data with the on-SoC GPU.

Based on U-Boot code by Stephen Warren.

---
 arch/arm/mach-bcm2835/Makefile            |   2 +-
 arch/arm/mach-bcm2835/include/mach/mbox.h | 420 ++++++++++++++++++++++++++++++
 arch/arm/mach-bcm2835/mbox.c              | 152 +++++++++++
 3 files changed, 573 insertions(+), 1 deletion(-)
 create mode 100644 arch/arm/mach-bcm2835/include/mach/mbox.h
 create mode 100644 arch/arm/mach-bcm2835/mbox.c
arch/arm/mach-bcm2835/Makefile +1 -1
index 820eb10..940f98c 100644
@@ -1 +1 @@
-obj-y += core.o
+obj-y += core.o mbox.o
arch/arm/mach-bcm2835/include/mach/mbox.h new file +420 -0

420 lines not on this page · raw message

arch/arm/mach-bcm2835/mbox.c new file +152 -0

152 lines not on this page · raw message

Signature · 7 lines
1.8.3.2


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#8 lore raw

PATCH 8/10 ARM: rpi: register a clkdev for the eMMC clock

In reply to Andre Heider

Use the mailbox driver to query the clock frequency and create a clkdev for the bcm2835_mci driver.

---
 arch/arm/boards/raspberry-pi/rpi.c | 40 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 40 insertions(+)
arch/arm/boards/raspberry-pi/rpi.c +40 -0
index 971a603..d3c6b11 100644
@@ -17,11 +17,44 @@
 #include <init.h>
 #include <fs.h>
 #include <linux/stat.h>
+#include <linux/clk.h>
+#include <linux/clkdev.h>
 #include <envfs.h>
 #include <asm/armlinux.h>
 #include <generated/mach-types.h>
 
 #include <mach/core.h>
+#include <mach/mbox.h>
+
+struct msg_get_clock_rate {
+	struct bcm2835_mbox_hdr hdr;
+	struct bcm2835_mbox_tag_get_clock_rate get_clock_rate;
+	u32 end_tag;
+};
+
+static int rpi_register_clkdev(u32 clock_id, const char *name)
+{
+	BCM2835_MBOX_STACK_ALIGN(struct msg_get_clock_rate, msg);
+	struct clk *clk;
+	int ret;
+
+	BCM2835_MBOX_INIT_HDR(msg);
+	BCM2835_MBOX_INIT_TAG(&msg->get_clock_rate, GET_CLOCK_RATE);
+	msg->get_clock_rate.body.req.clock_id = clock_id;
+
+	ret = bcm2835_mbox_call_prop(BCM2835_MBOX_PROP_CHAN, &msg->hdr);
+	if (ret)
+		return ret;
+
+	clk = clk_fixed(name, msg->get_clock_rate.body.resp.rate_hz);
+	if (IS_ERR(clk))
+		return PTR_ERR(clk);
+
+	if (!clk_register_clkdev(clk, NULL, name))
+		return -ENODEV;
+
+	return 0;
+}
 
 static int rpi_mem_init(void)
 {
@@ -40,6 +73,13 @@ static int rpi_console_init(void)
 }
 console_initcall(rpi_console_init);
 
+static int rpi_clock_init(void)
+{
+	rpi_register_clkdev(BCM2835_MBOX_CLOCK_ID_EMMC, "bcm2835_mci0");
+	return 0;
+}
+postconsole_initcall(rpi_clock_init);
+
 static int rpi_env_init(void)
 {
 	struct stat s;
Signature · 7 lines
1.8.3.2


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#9 lore raw

PATCH 9/10 mci: bcm2835: use the registered device clkdev

In reply to Andre Heider

Switch from local mailbox code to using the newly created clock device.

---
 drivers/mci/mci-bcm2835.c | 93 ++++++++++-------------------------------------
 drivers/mci/mci-bcm2835.h | 48 ------------------------
 2 files changed, 19 insertions(+), 122 deletions(-)
drivers/mci/mci-bcm2835.c +19 -74
index 2ffdeec..7d8997c 100644
@@ -28,13 +28,14 @@
  * Author: Wilhelm Lundgren <[email protected]>
  */
 
-#include <asm/mmu.h>
 #include <common.h>
 #include <init.h>
 #include <mci.h>
 #include <io.h>
 #include <malloc.h>
 #include <clock.h>
+#include <linux/clk.h>
+
 #include "mci-bcm2835.h"
 #include "sdhci.h"
 
@@ -469,53 +470,6 @@ int bcm2835_mci_reset(struct mci_host *mci, struct device_d *mci_dev)
 	return bcm2835_mci_wait_command_done(host);
 }
 
-static u32 bcm2835_mci_get_emmc_clock(struct msg_get_clock_rate *clk_data)
-{
-	u32 val;
-	struct bcm2835_mbox_regs __iomem *regs =
-			(struct bcm2835_mbox_regs *) BCM2835_MBOX_PHYSADDR;
-
-	/*Read out old msg*/
-	while (true) {
-		val = readl(&regs->status);
-		if (val & BCM2835_MBOX_STATUS_RD_EMPTY)
-			break;
-		val = readl(&regs->read);
-	}
-
-	/*Check for ok to write*/
-	while (true) {
-		val = readl(&regs->status);
-		if (!(val & BCM2835_MBOX_STATUS_WR_FULL))
-			break;
-	}
-	val = BCM2835_MBOX_PROP_CHAN + ((u32) &clk_data->hdr);
-	dma_flush_range((u32)clk_data, (u32)clk_data + sizeof(*clk_data));
-	writel(val, &regs->write);
-
-	while (true) {
-		/* Wait for the response */
-		while (true) {
-			val = readl(&regs->status);
-			if (!(val & BCM2835_MBOX_STATUS_RD_EMPTY))
-				break;
-		}
-
-		/* Read the response */
-		val = readl(&regs->read);
-		if ((val & 0x0F) == BCM2835_MBOX_PROP_CHAN)
-			break;
-	}
-
-	dma_inv_range((u32)clk_data, (u32)clk_data + sizeof(*clk_data));
-
-	if ((val & ~0x0F) == ((u32) &clk_data->hdr))
-		if (clk_data->get_clock_rate.tag_hdr.val_len
-				& BCM2835_MBOX_TAG_VAL_LEN_RESPONSE)
-			return 1;
-	return 0;
-}
-
 static int bcm2835_mci_detect(struct device_d *dev)
 {
 	struct bcm2835_mci_host *host = dev->priv;
@@ -526,7 +480,22 @@ static int bcm2835_mci_detect(struct device_d *dev)
 static int bcm2835_mci_probe(struct device_d *hw_dev)
 {
 	struct bcm2835_mci_host *host;
-	struct msg_get_clock_rate *clk_data;
+	static struct clk *clk;
+	int ret;
+
+	clk = clk_get(hw_dev, NULL);
+	if (IS_ERR(clk)) {
+		ret = PTR_ERR(clk);
+		dev_err(hw_dev, "clock not found: %d\n", ret);
+		return ret;
+	}
+
+	ret = clk_enable(clk);
+	if (ret) {
+		dev_err(hw_dev, "clock failed to enable: %d\n", ret);
+		clk_put(clk);
+		return ret;
+	}
 
 	host = xzalloc(sizeof(*host));
 	host->mci.send_cmd = bcm2835_mci_request;
@@ -534,31 +503,7 @@ static int bcm2835_mci_probe(struct device_d *hw_dev)
 	host->mci.init = bcm2835_mci_reset;
 	host->mci.hw_dev = hw_dev;
 	host->hw_dev = hw_dev;
-
-	/* Allocate a buffer thats 16 bytes aligned in memory
-	 * Of the 32 bits address passed into the mbox 28 bits
-	 * are the address of the buffer, lower 4 bits is channel
-	 */
-	clk_data = memalign(16, sizeof(struct msg_get_clock_rate));
-	memset(clk_data, 0, sizeof(struct msg_get_clock_rate));
-	clk_data->hdr.buf_size = sizeof(struct msg_get_clock_rate);
-	clk_data->get_clock_rate.tag_hdr.tag = BCM2835_MBOX_TAG_GET_CLOCK_RATE;
-	clk_data->get_clock_rate.tag_hdr.val_buf_size =
-			sizeof(clk_data->get_clock_rate.body);
-	clk_data->get_clock_rate.tag_hdr.val_len =
-			sizeof(clk_data->get_clock_rate.body.req);
-	clk_data->get_clock_rate.body.req.clock_id = BCM2835_MBOX_CLOCK_ID_EMMC;
-
-	if (!bcm2835_mci_get_emmc_clock(clk_data)) {
-		dev_warn(host->hw_dev,
-				"Failed getting emmc clock, lets go anyway with 50MHz\n");
-		host->max_clock = 50000000;
-	} else {
-		host->max_clock = clk_data->get_clock_rate.body.resp.rate_hz;
-		dev_info(host->hw_dev, "Got emmc clock at %d Hz\n",
-				host->max_clock);
-	}
-
+	host->max_clock = clk_get_rate(clk);
 	host->regs = dev_request_mem_region(hw_dev, 0);
 	if (host->regs == NULL) {
 		dev_err(host->hw_dev, "Failed request mem region, aborting...\n");
drivers/mci/mci-bcm2835.h +0 -48

51 lines not on this page · raw message

Signature · 7 lines
1.8.3.2


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#10 lore raw

PATCH 10/10 ARM: rpi: use the proper ARM memory size

In reply to Andre Heider

Use the mailbox driver to query the size. This properly takes the firmware's VideoCore/ARM memory split into account.

Linux can now be booted with more than 128 MiB.

---
 arch/arm/boards/raspberry-pi/rpi.c | 35 +++++++++++++++++++++++++++++++++--
 1 file changed, 33 insertions(+), 2 deletions(-)
arch/arm/boards/raspberry-pi/rpi.c +33 -2
index d3c6b11..06c43f3 100644
@@ -26,12 +26,35 @@
 #include <mach/core.h>
 #include <mach/mbox.h>
 
+struct msg_get_arm_mem {
+	struct bcm2835_mbox_hdr hdr;
+	struct bcm2835_mbox_tag_get_arm_mem get_arm_mem;
+	u32 end_tag;
+};
+
 struct msg_get_clock_rate {
 	struct bcm2835_mbox_hdr hdr;
 	struct bcm2835_mbox_tag_get_clock_rate get_clock_rate;
 	u32 end_tag;
 };
 
+static int rpi_get_arm_mem(u32 *size)
+{
+	BCM2835_MBOX_STACK_ALIGN(struct msg_get_arm_mem, msg);
+	int ret;
+
+	BCM2835_MBOX_INIT_HDR(msg);
+	BCM2835_MBOX_INIT_TAG(&msg->get_arm_mem, GET_ARM_MEMORY);
+
+	ret = bcm2835_mbox_call_prop(BCM2835_MBOX_PROP_CHAN, &msg->hdr);
+	if (ret)
+		return ret;
+
+	*size = msg->get_arm_mem.body.resp.mem_size;
+
+	return 0;
+}
+
 static int rpi_register_clkdev(u32 clock_id, const char *name)
 {
 	BCM2835_MBOX_STACK_ALIGN(struct msg_get_clock_rate, msg);
@@ -58,8 +81,16 @@ static int rpi_register_clkdev(u32 clock_id, const char *name)
 
 static int rpi_mem_init(void)
 {
-	bcm2835_add_device_sdram(0);
-	return 0;
+	u32 size = 0;
+	int ret;
+
+	ret = rpi_get_arm_mem(&size);
+	if (ret)
+		printf("could not query ARM memory size\n");
+
+	bcm2835_add_device_sdram(size);
+
+	return ret;
 }
 mem_initcall(rpi_mem_init);
 
Signature · 7 lines
1.8.3.2


_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#11 lore raw
In reply to Andre Heider

Hi Andre,

Quoted 17 lines · Andre Heider

This adds a bcm2835 mailbox driver for the RPi to talk to the on-SoC VideoCore. See [1] for a description.

This patchset uses it to get two properties, the eMMC clock and the memory size.

The RPi backend needs to use this driver early on, namely before the MMU setup. We also want to use the driver in the future after the MMU setup, like setting up simplefb for the kernel. Patch 1 and 2 lay the groundwork for that.

The mailbox driver also needs to handle timouts, so patch 5 and 6 are making the clocksource accessible earlier.

Patch 7 is the driver itself, providing a helper macro for users based on a new common macro from patch 3.

Quoted 6 lines

Patch 8 and 9 switch the bc2835_mci driver from a local mailbox implementation to the new driver.

Patch 10 uses the driver to get the proper memory size.

Patch 4 is just a cleanup.

I'm in Edinburgh this week and won't have much time. I'll have a loser look next week. This series isn't lost. Thanks for working on this.

Sascha

Signature · 9 lines
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#12 lore raw
Quoted 23 lines · Sascha Hauer

Hi Andre,

Quoted 20 lines · Andre Heider

This adds a bcm2835 mailbox driver for the RPi to talk to the on-SoC VideoCore. See [1] for a description.

This patchset uses it to get two properties, the eMMC clock and the memory size.

The RPi backend needs to use this driver early on, namely before the MMU setup. We also want to use the driver in the future after the MMU setup, like setting up simplefb for the kernel. Patch 1 and 2 lay the groundwork for that.

The mailbox driver also needs to handle timouts, so patch 5 and 6 are making the clocksource accessible earlier.

Patch 7 is the driver itself, providing a helper macro for users based on a new common macro from patch 3.

Patch 8 and 9 switch the bc2835_mci driver from a local mailbox implementation to the new driver.

Quoted 6 lines
Quoted 3 lines

Patch 10 uses the driver to get the proper memory size.

Patch 4 is just a cleanup.

I'm in Edinburgh this week and won't have much time. I'll have a loser look next week. This series isn't lost. Thanks for working on this.

No worries, there's no need to rush :)

Regards,
Andre

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

Applied
#13 lore raw
In reply to Andre Heider
Quoted 17 lines · Andre Heider

This adds a bcm2835 mailbox driver for the RPi to talk to the on-SoC VideoCore. See [1] for a description.

This patchset uses it to get two properties, the eMMC clock and the memory size.

The RPi backend needs to use this driver early on, namely before the MMU setup. We also want to use the driver in the future after the MMU setup, like setting up simplefb for the kernel. Patch 1 and 2 lay the groundwork for that.

The mailbox driver also needs to handle timouts, so patch 5 and 6 are making the clocksource accessible earlier.

Patch 7 is the driver itself, providing a helper macro for users based on a new common macro from patch 3.

Quoted 6 lines

Patch 8 and 9 switch the bc2835_mci driver from a local mailbox implementation to the new driver.

Patch 10 uses the driver to get the proper memory size.

Patch 4 is just a cleanup.

Applied, thanks. I'll probably have a closer look at the caching bits later, but looks good for now.

Sascha

Signature · 9 lines
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#14 lore raw
Quoted 21 lines · Sascha Hauer
Quoted 20 lines · Andre Heider

This adds a bcm2835 mailbox driver for the RPi to talk to the on-SoC VideoCore. See [1] for a description.

This patchset uses it to get two properties, the eMMC clock and the memory size.

The RPi backend needs to use this driver early on, namely before the MMU setup. We also want to use the driver in the future after the MMU setup, like setting up simplefb for the kernel. Patch 1 and 2 lay the groundwork for that.

The mailbox driver also needs to handle timouts, so patch 5 and 6 are making the clocksource accessible earlier.

Patch 7 is the driver itself, providing a helper macro for users based on a new common macro from patch 3.

Patch 8 and 9 switch the bc2835_mci driver from a local mailbox implementation to the new driver.

Quoted 6 lines
Quoted 3 lines

Patch 10 uses the driver to get the proper memory size.

Patch 4 is just a cleanup.

Applied, thanks. I'll probably have a closer look at the caching bits later, but looks good for now.

Thanks Sascha!
Any particular concerns about the caching bits?

Regards,
Andre

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

#15 lore raw
Quoted 24 lines · Andre Heider
Quoted 23 lines · Sascha Hauer
Quoted 22 lines · Andre Heider

This adds a bcm2835 mailbox driver for the RPi to talk to the on-SoC VideoCore. See [1] for a description.

This patchset uses it to get two properties, the eMMC clock and the memory size.

The RPi backend needs to use this driver early on, namely before the MMU setup. We also want to use the driver in the future after the MMU setup, like setting up simplefb for the kernel. Patch 1 and 2 lay the groundwork for that.

The mailbox driver also needs to handle timouts, so patch 5 and 6 are making the clocksource accessible earlier.

Patch 7 is the driver itself, providing a helper macro for users based on a new common macro from patch 3.

Patch 8 and 9 switch the bc2835_mci driver from a local mailbox implementation to the new driver.

Patch 10 uses the driver to get the proper memory size.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

Patch 4 is just a cleanup.

Applied, thanks. I'll probably have a closer look at the caching bits later, but looks good for now.

Thanks Sascha!
Any particular concerns about the caching bits?

I'm concerned about "only invalidate caches if we have functions to do so". This could lead to some quiet bugs when we really ought to invalidate caches but missed to setup the functions early enough.

Sascha

Signature · 9 lines
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox