Skip to contentSkip to conversation

riscv: k210: Enable use of AI ram bank

PATCH Superseded Sean Anderson 2 participants 29 messages Oct 12, 2020 – Nov 3, 2020

Subsystems (4)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover riscv: k210: Enable use of AI ram bank

This ram bank was previously unusable because PLL1 was not started correctly. This series fixes that bug, and allows U-Boot to relocate into the AI ram. This provides an extra 2M of space in which to load payloads.

Second time around and I realized that the bypass clock is no longer necessary (yay). This also necessitates a few more small fixups. I've also added some minor patches to bring the device tree and clock driver closer to what Linux has (or will have). This should be the last round of additions to this series.

Changes in v2:
- Don't re-enable the PLL
- Remove bypass clock, which is no longer necessary
- Simplify PLL instantiation
- Modify clock tree so clint is a child of aclk
- Sync memory dts node with Linux
- Use correct aisram clock

Sean Anderson (16):
clk: k210: Fix PLLs not being enabled
clk: k210: Fix PLL enable always getting taken
clk: k210: Remove bypass clock
clk: k210: Remove k210_register_pll
clk: k210: Move the clint clock to under aclk
clk: Add support for the k210 clock driver pre-relocation
riscv: Enable some devices pre-relocation
lib: fdt: Add fdtdec_setup_mem_size_base_highest
test: Add a test for fdtdec_setup_mem_size_base et al.
ram: Add driver for K210 SRAM
ram: sifive: Default to y only if compiling for fu540
riscv: Probe ram in dram_init
riscv: Enable AI ram on K210
riscv: k210: Rename airam to aisram
riscv: k210: Use AI as the parent clock of aisram, not PLL1
riscv: Don't reserve AI ram in k210 dts

 MAINTAINERS                        |   1 +
 arch/riscv/cpu/generic/dram.c      |  26 +++
 arch/riscv/dts/k210.dtsi           |  22 +--
 arch/sandbox/dts/test.dts          |  12 ++
 board/sipeed/maix/Kconfig          |   2 +
 board/sipeed/maix/maix.c           |  26 ---
 configs/sandbox64_defconfig        |   2 +-
 configs/sandbox_defconfig          |   2 +-
 configs/sandbox_flattree_defconfig |   2 +-
 configs/sipeed_maix_bitm_defconfig |   1 +
 drivers/clk/kendryte/Makefile      |   2 +-
 drivers/clk/kendryte/bypass.c      | 273 -----------------------------
 drivers/clk/kendryte/clk.c         |  61 +++----
 drivers/clk/kendryte/pll.c         |  26 +--
 drivers/ram/Kconfig                |   7 +
 drivers/ram/Makefile               |   1 +
 drivers/ram/kendryte.c             |  56 ++++++
 drivers/ram/sifive/Kconfig         |   2 +-
 include/configs/sipeed-maix.h      |   4 -
 include/fdtdec.h                   |  19 +-
 include/kendryte/bypass.h          |  31 ----
 include/kendryte/pll.h             |   4 -
 lib/fdtdec.c                       |  34 +++-
 test/dm/fdtdec.c                   |  38 ++++
 24 files changed, 230 insertions(+), 424 deletions(-)
 delete mode 100644 drivers/clk/kendryte/bypass.c
 create mode 100644 drivers/ram/kendryte.c
 delete mode 100644 include/kendryte/bypass.h
Signature · 1 line
2.28.0
#1 lore raw

PATCH v2 1/16 clk: k210: Fix PLLs not being enabled

After starting or setting the rate of a PLL, the enable bit must be set.

This fixes a bug where the AI ram would not be accessible, because it requires PLL1 to be running.

---

(no changes since v1)

 drivers/clk/kendryte/pll.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/clk/kendryte/pll.c +2 -0
index ab6d75d585..f198920113 100644
@@ -531,6 +531,7 @@ static int k210_pll_enable(struct clk *clk)
 	k210_pll_waitfor_lock(pll);
 
 	reg &= ~K210_PLL_BYPASS;
+	reg |= K210_PLL_EN;
 	writel(reg, pll->reg);
 
 	return 0;
@@ -550,6 +551,7 @@ static int k210_pll_disable(struct clk *clk)
 	writel(reg, pll->reg);
 
 	reg &= ~K210_PLL_PWRD;
+	reg &= ~K210_PLL_EN;
 	writel(reg, pll->reg);
 	return 0;
 }
Signature · 1 line
2.28.0
#2 lore raw

PATCH v2 1/16 clk: k210: Fix PLLs not being enabled

Quoted 5 lines

After starting or setting the rate of a PLL, the enable bit must be set.

This fixes a bug where the AI ram would not be accessible, because it requires PLL1 to be running.

Quoted 8 lines

---

(no changes since v1)

 drivers/clk/kendryte/pll.c | 2 ++
 1 file changed, 2 insertions(+)
#3 lore raw

PATCH v2 2/16 clk: k210: Fix PLL enable always getting taken

In reply to Sean Anderson

This conditional always evaluated as false, regardless of the value of reg. Fix it so that it properly tests the bits in the PLL register. Also test PLL_EN, now that we set it.

---

Changes in v2:
- New

 drivers/clk/kendryte/pll.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/clk/kendryte/pll.c +2 -1
index f198920113..d46fd0ebbf 100644
@@ -512,7 +512,8 @@ static int k210_pll_enable(struct clk *clk)
 	struct k210_pll *pll = to_k210_pll(clk);
 	u32 reg = readl(pll->reg);
 
-	if ((reg | K210_PLL_PWRD) && !(reg | K210_PLL_RESET))
+	if ((reg & K210_PLL_PWRD) && (reg & K210_PLL_EN) &&
+	    !(reg & K210_PLL_RESET))
 		return 0;
 
 	reg |= K210_PLL_PWRD;
Signature · 1 line
2.28.0
#4 lore raw

PATCH v2 2/16 clk: k210: Fix PLL enable always getting taken

Quoted 6 lines

This conditional always evaluated as false, regardless of the value of reg. Fix it so that it properly tests the bits in the PLL register. Also test PLL_EN, now that we set it.

Quoted 8 lines

---

Changes in v2:
- New

 drivers/clk/kendryte/pll.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
#5 lore raw

PATCH v2 3/16 clk: k210: Remove bypass clock

In reply to Sean Anderson

Now that the PLLs get properly enabled, the bypass clock is no longer necessary.

---

Changes in v2:
- New

 drivers/clk/kendryte/Makefile |   2 +-
 drivers/clk/kendryte/bypass.c | 273 ----------------------------------
 drivers/clk/kendryte/clk.c    |  43 ++----
 include/kendryte/bypass.h     |  31 ----
 4 files changed, 15 insertions(+), 334 deletions(-)
 delete mode 100644 drivers/clk/kendryte/bypass.c
 delete mode 100644 include/kendryte/bypass.h
drivers/clk/kendryte/Makefile +1 -1
index 6fb68253ae..d26bce954f 100644
@@ -1 +1 @@
-obj-y += bypass.o clk.o pll.o
+obj-y += clk.o pll.o
drivers/clk/kendryte/bypass.c deleted +0 -273

273 lines not on this page · raw message

drivers/clk/kendryte/clk.c +14 -29

73 lines not on this page · raw message

include/kendryte/bypass.h deleted +0 -31
deleted file mode 100644
index ab85bbcbfc..0000000000
@@ -1,31 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0+ */
-/*
- * Copyright (C) 2020 Sean Anderson <[email protected]>
- */
-#ifndef K210_BYPASS_H
-#define K210_BYPASS_H
-
-struct clk;
-
-struct k210_bypass {
-	struct clk clk;
-	struct clk **children; /* Clocks to reparent */
-	struct clk **saved_parents; /* Parents saved over en-/dis-able */
-	struct clk *bypassee; /* Clock to bypass */
-	const struct clk_ops *bypassee_ops; /* Ops of the bypass clock */
-	struct clk *alt; /* Clock to set children to when bypassing */
-	size_t child_count;
-};
-
-#define to_k210_bypass(_clk) container_of(_clk, struct k210_bypass, clk)
-
-int k210_bypass_set_children(struct clk *clk, struct clk **children,
-			     size_t child_count);
-struct clk *k210_register_bypass_struct(const char *name,
-					const char *parent_name,
-					struct k210_bypass *bypass);
-struct clk *k210_register_bypass(const char *name, const char *parent_name,
-				 struct clk *bypassee,
-				 const struct clk_ops *bypassee_ops,
-				 struct clk *alt);
-#endif /* K210_BYPASS_H */
Signature · 1 line
2.28.0
#6 lore raw

PATCH v2 4/16 clk: k210: Remove k210_register_pll

In reply to Sean Anderson

This simplifies the PLL creation process, since we don't have to pass all the parameters individually.

---

Changes in v2:
- New

 drivers/clk/kendryte/clk.c | 20 ++++++--------------
 drivers/clk/kendryte/pll.c | 21 ---------------------
 include/kendryte/pll.h     |  4 ----
 3 files changed, 6 insertions(+), 39 deletions(-)
drivers/clk/kendryte/clk.c +6 -14
index e657efc572..88a997e80b 100644
@@ -511,23 +511,15 @@ static int k210_clk_probe(struct udevice *dev)
 	aclk_sels[0] = in0;
 	pll2_sels[0] = in0;
 
-	{
-		const struct k210_pll_params *params = &k210_plls[0];
-
+	pll = k210_create_pll(&k210_plls[0], base);
+	if (pll)
 		clk_dm(K210_CLK_PLL0,
-		       k210_register_pll("pll0", in0, base + params->off,
-					 base + params->lock_off, params->shift,
-					 params->width));
-	}
-
-	{
-		const struct k210_pll_params *params = &k210_plls[1];
+		       k210_register_pll_struct("pll0", in0, pll));
 
+	pll = k210_create_pll(&k210_plls[1], base);
+	if (pll)
 		clk_dm(K210_CLK_PLL1,
-		       k210_register_pll("pll1", in0, base + params->off,
-					 base + params->lock_off, params->shift,
-					 params->width));
-	}
+		       k210_register_pll_struct("pll1", in0, pll));
 
 	/* PLL2 is muxed, so set up a composite clock */
 	mux = k210_create_mux(&k210_muxes[MUXIFY(K210_CLK_PLL2)], base);
drivers/clk/kendryte/pll.c +0 -21
index d46fd0ebbf..184f37aaf2 100644
@@ -578,27 +578,6 @@ struct clk *k210_register_pll_struct(const char *name, const char *parent_name,
 	return clk;
 }
 
-struct clk *k210_register_pll(const char *name, const char *parent_name,
-			      void __iomem *reg, void __iomem *lock, u8 shift,
-			      u8 width)
-{
-	struct clk *clk;
-	struct k210_pll *pll;
-
-	pll = kzalloc(sizeof(*pll), GFP_KERNEL);
-	if (!pll)
-		return ERR_PTR(-ENOMEM);
-	pll->reg = reg;
-	pll->lock = lock;
-	pll->shift = shift;
-	pll->width = width;
-
-	clk = k210_register_pll_struct(name, parent_name, pll);
-	if (IS_ERR(clk))
-		kfree(pll);
-	return clk;
-}
-
 U_BOOT_DRIVER(k210_pll) = {
 	.name	= CLK_K210_PLL,
 	.id	= UCLASS_CLK,
include/kendryte/pll.h +0 -4
index 55a40b9c97..95b8494f40 100644
@@ -55,8 +55,4 @@ extern const struct clk_ops k210_pll_ops;
 
 struct clk *k210_register_pll_struct(const char *name, const char *parent_name,
 				     struct k210_pll *pll);
-struct clk *k210_register_pll(const char *name, const char *parent_name,
-			      void __iomem *reg, void __iomem *lock, u8 shift,
-			      u8 width);
-
 #endif /* K210_PLL_H */
Signature · 1 line
2.28.0
#7 lore raw

PATCH v2 5/16 clk: k210: Move the clint clock to under aclk

In reply to Sean Anderson

No other (real) clocks have the cpu clock as their parent; instead they are children of aclk. Move the clint clock under aclk to match them.

---

Changes in v2:
- New

 drivers/clk/kendryte/clk.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/clk/kendryte/clk.c +1 -1
index 88a997e80b..28e21d0bf9 100644
@@ -625,7 +625,7 @@ static int k210_clk_probe(struct udevice *dev)
 
 	/* The MTIME register in CLINT runs at one 50th the CPU clock speed */
 	clk_dm(K210_CLK_CLINT,
-	       clk_register_fixed_factor(NULL, "clint", "cpu", 0, 1, 50));
+	       clk_register_fixed_factor(NULL, "clint", "aclk", 0, 1, 50));
 
 	return 0;
 }
Signature · 1 line
2.28.0
#8 lore raw

PATCH v2 6/16 clk: Add support for the k210 clock driver pre-relocation

In reply to Sean Anderson

Variables which had previously been stored in .bss are moved to .data. In addition, probed needs to be reset when the clock driver is re-bound post-relocation.

---

(no changes since v1)

 drivers/clk/kendryte/clk.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
drivers/clk/kendryte/clk.c +9 -1
index 28e21d0bf9..3e4818e9ef 100644
@@ -471,7 +471,14 @@ cleanup_mux:
 	return comp;
 }
 
-static bool probed;
+static bool __section(.data) probed;
+
+/* reset probed so we will probe again post-relocation */
+static int k210_clk_bind(struct udevice *dev)
+{
+	probed = false;
+	return 0;
+}
 
 static int k210_clk_probe(struct udevice *dev)
 {
@@ -640,5 +647,6 @@ U_BOOT_DRIVER(k210_clk) = {
 	.id = UCLASS_CLK,
 	.of_match = k210_clk_ids,
 	.ops = &k210_clk_ops,
+	.bind = k210_clk_bind,
 	.probe = k210_clk_probe,
 };
Signature · 1 line
2.28.0
#9 lore raw

PATCH v2 6/16 clk: Add support for the k210 clock driver pre-relocation

Quoted 4 lines

Variables which had previously been stored in .bss are moved to .data. In addition, probed needs to be reset when the clock driver is re-bound post-relocation.

Quoted 8 lines

---

(no changes since v1)

 drivers/clk/kendryte/clk.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
#10 lore raw

PATCH v2 7/16 riscv: Enable some devices pre-relocation

In reply to Sean Anderson

These devices are necessary for the clock driver, which is required by the sram driver, to run pre-relocation.

---

(no changes since v1)

 arch/riscv/dts/k210.dtsi | 4 ++++
 1 file changed, 4 insertions(+)
arch/riscv/dts/k210.dtsi +4 -0
index 7605c01f3c..86d50f14e5 100644
@@ -91,6 +91,7 @@
 			 <&sysclk K210_CLK_SRAM1>,
 			 <&sysclk K210_CLK_PLL1>;
 		clock-names = "sram0", "sram1", "airam";
+		u-boot,dm-pre-reloc;
 	};
 
 	reserved-memory {
@@ -109,6 +110,7 @@
 			compatible = "fixed-clock";
 			#clock-cells = <0>;
 			clock-frequency = <26000000>;
+			u-boot,dm-pre-reloc;
 		};
 	};
 
@@ -503,11 +505,13 @@
 					     "syscon", "simple-mfd";
 				reg = <0x50440000 0x100>;
 				reg-io-width = <4>;
+				u-boot,dm-pre-reloc;
 
 				sysclk: clock-controller {
 					#clock-cells = <1>;
 					compatible = "kendryte,k210-clk";
 					clocks = <&in0>;
+					u-boot,dm-pre-reloc;
 				};
 
 				sysrst: reset-controller {
Signature · 1 line
2.28.0
#12 lore raw

PATCH v2 8/16 lib: fdt: Add fdtdec_setup_mem_size_base_highest

In reply to Sean Anderson

This is very similar to fdtdec_setup_mem_size_base_lowest, except we pick the highest ram bank, instead of the lowest. This is helpful for boards which use separate but contiguous ram banks, as it leaves the most space for loading programs.

---

(no changes since v1)

 include/fdtdec.h | 19 ++++++++++++++++++-
 lib/fdtdec.c     | 34 +++++++++++++++++++++++++++++-----
 2 files changed, 47 insertions(+), 6 deletions(-)
include/fdtdec.h +18 -1
index 62d1660973..6a8e235674 100644
@@ -944,7 +944,7 @@ int fdtdec_setup_mem_size_base(void);
  * gd->ram_start by lowest available memory base
  *
  * Decode the /memory 'reg' property to determine the lowest start of the memory
- * bank bank and populate the global data with it.
+ * bank and populate the global data with it.
  *
  * This function should be called from a boards dram_init(). This helper
  * function allows for boards to query the device tree for DRAM size and start
@@ -956,6 +956,23 @@ int fdtdec_setup_mem_size_base(void);
  */
 int fdtdec_setup_mem_size_base_lowest(void);
 
+/**
+ * fdtdec_setup_mem_size_base_highest() - decode and setup gd->ram_size and
+ * gd->ram_start by highest available memory top
+ *
+ * Decode the /memory 'reg' property to determine the highest end of the memory
+ * bank and populate the global data with it.
+ *
+ * This function should be called from a boards dram_init(). This helper
+ * function allows for boards to query the device tree for DRAM size and start
+ * address instead of hard coding the value in the case where the memory size
+ * and start address cannot be detected automatically.
+ *
+ * @return 0 if OK, -EINVAL if the /memory node or reg property is missing or
+ * invalid
+ */
+int fdtdec_setup_mem_size_base_highest(void);
+
 /**
  * fdtdec_setup_memory_banksize() - decode and populate gd->bd->bi_dram
  *
lib/fdtdec.c +29 -5
index ee1bd41b08..78b34bf4da 100644
@@ -1123,7 +1123,7 @@ int fdtdec_setup_memory_banksize(void)
 	return 0;
 }
 
-int fdtdec_setup_mem_size_base_lowest(void)
+static int fdtdec_setup_mem_size_base_superlative(bool lowest)
 {
 	int bank, ret, reg = 0;
 	struct resource res;
@@ -1131,7 +1131,16 @@ int fdtdec_setup_mem_size_base_lowest(void)
 	phys_size_t size;
 	ofnode mem = ofnode_null();
 
-	gd->ram_base = (unsigned long)~0;
+	/*
+	 * Size must be 1 so we don't underflow when doing the subtraction below
+	 * when lowest = false. Hopefully any real ram banks will have a greater
+	 * size :)
+	 */
+	gd->ram_size = 1;
+	if (lowest)
+		gd->ram_base = (unsigned long)~0;
+	else
+		gd->ram_base = 0;
 
 	mem = get_next_memory_node(mem);
 	if (!ofnode_valid(mem)) {
@@ -1158,17 +1167,32 @@ int fdtdec_setup_mem_size_base_lowest(void)
 		base = (unsigned long)res.start;
 		size = (phys_size_t)(res.end - res.start + 1);
 
-		if (gd->ram_base > base && size) {
+		if (!size)
+			continue;
+
+		if ((lowest && gd->ram_base > base) ||
+		    (!lowest && gd->ram_base + gd->ram_size - 1 < res.end)) {
 			gd->ram_base = base;
 			gd->ram_size = size;
-			debug("%s: Initial DRAM base %lx size %lx\n",
-			      __func__, base, (unsigned long)size);
 		}
 	}
 
+	if (gd->ram_size)
+		debug("%s: Initial DRAM base %lx size %lx\n", __func__,
+		      gd->ram_base, (unsigned long)gd->ram_size);
 	return 0;
 }
 
+int fdtdec_setup_mem_size_base_lowest(void)
+{
+	return fdtdec_setup_mem_size_base_superlative(true);
+}
+
+int fdtdec_setup_mem_size_base_highest(void)
+{
+	return fdtdec_setup_mem_size_base_superlative(false);
+}
+
 #if CONFIG_IS_ENABLED(MULTI_DTB_FIT)
 # if CONFIG_IS_ENABLED(MULTI_DTB_FIT_GZIP) ||\
 	CONFIG_IS_ENABLED(MULTI_DTB_FIT_LZO)
Signature · 1 line
2.28.0
#13 lore raw

PATCH v2 8/16 lib: fdt: Add fdtdec_setup_mem_size_base_highest

Quoted 7 lines

This is very similar to fdtdec_setup_mem_size_base_lowest, except we pick the highest ram bank, instead of the lowest. This is helpful for boards which use separate but contiguous ram banks, as it leaves the most space for loading programs.

Quoted 8 lines

---

(no changes since v1)

 include/fdtdec.h | 19 ++++++++++++++++++-
 lib/fdtdec.c     | 34 +++++++++++++++++++++++++++++-----
 2 files changed, 47 insertions(+), 6 deletions(-)
#14 lore raw

PATCH v2 9/16 test: Add a test for fdtdec_setup_mem_size_base et al.

In reply to Sean Anderson

This adds a test for the various methods of extracting ram_base and ram_size from a device tree.

---

(no changes since v1)

 arch/sandbox/dts/test.dts          | 12 ++++++++++
 configs/sandbox64_defconfig        |  2 +-
 configs/sandbox_defconfig          |  2 +-
 configs/sandbox_flattree_defconfig |  2 +-
 test/dm/fdtdec.c                   | 38 ++++++++++++++++++++++++++++++
 5 files changed, 53 insertions(+), 3 deletions(-)
arch/sandbox/dts/test.dts +12 -0
index c7d1376911..f13bbdca7d 100644
@@ -1221,6 +1221,18 @@
 			compatible = "sandbox,regmap_test";
 		};
 	};
+
+	memory at 0000 {
+		device_type = "memory";
+		reg = <0x1000 0x2000>,
+		      <0x0000 0x1000>;
+	};
+
+	memory at 8000 {
+		device_type = "memory";
+		reg = <0x8000 0x0000>,
+		      <0x4000 0x3000>;
+	};
 };
 
 #include "sandbox_pmic.dtsi"
configs/sandbox64_defconfig +1 -1
index c3ca796d51..6324961cf1 100644
@@ -1,5 +1,5 @@
 CONFIG_SYS_TEXT_BASE=0
-CONFIG_NR_DRAM_BANKS=1
+CONFIG_NR_DRAM_BANKS=4
 CONFIG_ENV_SIZE=0x2000
 CONFIG_PRE_CON_BUF_ADDR=0x100000
 CONFIG_BOOTSTAGE_STASH_ADDR=0x0
configs/sandbox_defconfig +1 -1
index 18f787cb51..2f2333febc 100644
@@ -1,5 +1,5 @@
 CONFIG_SYS_TEXT_BASE=0
-CONFIG_NR_DRAM_BANKS=1
+CONFIG_NR_DRAM_BANKS=4
 CONFIG_ENV_SIZE=0x2000
 CONFIG_PRE_CON_BUF_ADDR=0xf0000
 CONFIG_BOOTSTAGE_STASH_ADDR=0x0
configs/sandbox_flattree_defconfig +1 -1
index dd93167e1b..a4941b54d3 100644
@@ -1,5 +1,5 @@
 CONFIG_SYS_TEXT_BASE=0
-CONFIG_NR_DRAM_BANKS=1
+CONFIG_NR_DRAM_BANKS=4
 CONFIG_ENV_SIZE=0x2000
 CONFIG_BOOTSTAGE_STASH_ADDR=0x0
 CONFIG_DEFAULT_DEVICE_TREE="sandbox"
test/dm/fdtdec.c +38 -0
index 017157a2ec..00ccb8c615 100644
@@ -131,3 +131,41 @@ static int dm_test_fdtdec_add_reserved_memory(struct unit_test_state *uts)
 }
 DM_TEST(dm_test_fdtdec_add_reserved_memory,
 	UT_TESTF_SCAN_PDATA | UT_TESTF_SCAN_FDT | UT_TESTF_FLAT_TREE);
+
+static int _dm_test_fdtdec_setup_mem(struct unit_test_state *uts)
+{
+	ut_assertok(fdtdec_setup_mem_size_base());
+	ut_asserteq(0x1000, gd->ram_base);
+	ut_asserteq(0x2000, gd->ram_size);
+
+	ut_assertok(fdtdec_setup_mem_size_base_lowest());
+	ut_asserteq(0x0000, gd->ram_base);
+	ut_asserteq(0x1000, gd->ram_size);
+
+	ut_assertok(fdtdec_setup_mem_size_base_highest());
+	ut_asserteq(0x4000, gd->ram_base);
+	ut_asserteq(0x3000, gd->ram_size);
+
+	return 0;
+}
+
+/*
+ * We need to wrap the actual test so that we don't overwrite the ram parameters
+ * for the rest of U-Boot
+ */
+static int dm_test_fdtdec_setup_mem(struct unit_test_state *uts)
+{
+	int ret;
+	unsigned long base, size;
+
+	base = gd->ram_base;
+	size = gd->ram_size;
+
+	ret = _dm_test_fdtdec_setup_mem(uts);
+
+	gd->ram_base = base;
+	gd->ram_size = size;
+
+	return ret;
+}
+DM_TEST(dm_test_fdtdec_setup_mem, UT_TESTF_SCAN_FDT | UT_TESTF_FLAT_TREE);
Signature · 1 line
2.28.0
#15 lore raw

PATCH v2 10/16 ram: Add driver for K210 SRAM

In reply to Sean Anderson

This adds a driver to handle enabling the clock for the AI SRAM. This was previously done in board_init, but it needs to happen before relocation now. An alternative would be to move this to board_init_early_f, but by doing it this way we can use clk_bulk.

---

(no changes since v1)

 MAINTAINERS            |  1 +
 drivers/ram/Kconfig    |  7 ++++++
 drivers/ram/Makefile   |  1 +
 drivers/ram/kendryte.c | 56 ++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 65 insertions(+)
 create mode 100644 drivers/ram/kendryte.c
MAINTAINERS +1 -0
index fb9ba37984..166acb30c2 100644
@@ -938,6 +938,7 @@ F:	doc/device-tree-bindings/mfd/kendryte,k210-sysctl.txt
 F:	doc/device-tree-bindings/pinctrl/kendryte,k210-fpioa.txt
 F:	drivers/clk/kendryte/
 F:	drivers/pinctrl/kendryte/
+F:	drivers/ram/kendryte.c
 F:	include/kendryte/
 
 RNG
drivers/ram/Kconfig +7 -0
index a270e13b26..075e6b5cd7 100644
@@ -73,6 +73,13 @@ config IMXRT_SDRAM
 	  to support external memories like sdram, psram & nand.
 	  This driver is for the sdram memory interface with the SEMC.
 
+config K210_SRAM
+	bool "Enable Kendryte K210 SRAM support"
+	depends on RAM
+	help
+	  The Kendryte K210 has three banks of SRAM. This driver does the
+	  necessary initialization.
+
 source "drivers/ram/aspeed/Kconfig"
 source "drivers/ram/rockchip/Kconfig"
 source "drivers/ram/sifive/Kconfig"
drivers/ram/Makefile +1 -0
index 209a78c06f..fdc9bbeb1f 100644
@@ -18,6 +18,7 @@ obj-$(CONFIG_ARCH_ASPEED) += aspeed/
 obj-$(CONFIG_K3_J721E_DDRSS) += k3-j721e/
 
 obj-$(CONFIG_IMXRT_SDRAM) += imxrt_sdram.o
+obj-$(CONFIG_K210_SRAM) += kendryte.o
 
 obj-$(CONFIG_RAM_SIFIVE) += sifive/
 
drivers/ram/kendryte.c new file +56 -0

56 lines not on this page · raw message

Signature · 1 line
2.28.0
#16 lore raw

PATCH v2 10/16 ram: Add driver for K210 SRAM

Quoted 9 lines

This adds a driver to handle enabling the clock for the AI SRAM. This was previously done in board_init, but it needs to happen before relocation now. An alternative would be to move this to board_init_early_f, but by doing it this way we can use clk_bulk.

---

Quoted 9 lines

(no changes since v1)

 MAINTAINERS            |  1 +
 drivers/ram/Kconfig    |  7 ++++++
 drivers/ram/Makefile   |  1 +
 drivers/ram/kendryte.c | 56 ++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 65 insertions(+)
 create mode 100644 drivers/ram/kendryte.c
#17 lore raw

PATCH v2 11/16 ram: sifive: Default to y only if compiling for fu540

In reply to Sean Anderson

Other RISC-V targets should not have RAM_SIFIVE enabled by default.

---

(no changes since v1)

 drivers/ram/sifive/Kconfig | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/ram/sifive/Kconfig +1 -1
index 6aca22ab2a..b24153d971 100644
@@ -1,7 +1,7 @@
 config RAM_SIFIVE
 	bool "Ram drivers support for SiFive SoCs"
 	depends on RAM && RISCV
-	default y
+	default y if TARGET_SIFIVE_FU540
 	help
 	  This enables support for ram drivers of SiFive SoCs.
 
Signature · 1 line
2.28.0
#19 lore raw

PATCH v2 12/16 riscv: Probe ram in dram_init

In reply to Sean Anderson

If CONFIG_RAM is enabled, use the ram device to get the base/size of memory. This provides an easy way for boards/cpus to hook into the dram_init logic, without needing to provide a second SYS_CPU.

---

(no changes since v1)

 arch/riscv/cpu/generic/dram.c | 26 ++++++++++++++++++++++++++
 1 file changed, 26 insertions(+)
arch/riscv/cpu/generic/dram.c +26 -0
index 1dc77efeca..fbcd4ddf5f 100644
@@ -4,15 +4,41 @@
  */
 
 #include <common.h>
+#include <dm.h>
 #include <fdtdec.h>
 #include <init.h>
+#include <log.h>
+#include <ram.h>
 #include <linux/sizes.h>
 
 DECLARE_GLOBAL_DATA_PTR;
 
 int dram_init(void)
 {
+#if CONFIG_IS_ENABLED(RAM)
+	int ret;
+	struct ram_info info;
+	struct udevice *dev;
+
+	ret = uclass_get_device(UCLASS_RAM, 0, &dev);
+	if (ret) {
+		debug("DRAM init failed: %d\n", ret);
+		return ret;
+	}
+
+	ret = ram_get_info(dev, &info);
+	if (ret) {
+		debug("Cannot get DRAM size: %d\n", ret);
+		return ret;
+	}
+
+	gd->ram_base = info.base;
+	gd->ram_size = info.size;
+
+	return 0;
+#else
 	return fdtdec_setup_mem_size_base();
+#endif
 }
 
 int dram_init_banksize(void)
Signature · 1 line
2.28.0
#20 lore raw

PATCH v2 12/16 riscv: Probe ram in dram_init

Quoted 4 lines

If CONFIG_RAM is enabled, use the ram device to get the base/size of memory. This provides an easy way for boards/cpus to hook into the dram_init logic, without needing to provide a second SYS_CPU.

Quoted 8 lines

---

(no changes since v1)

 arch/riscv/cpu/generic/dram.c | 26 ++++++++++++++++++++++++++
 1 file changed, 26 insertions(+)
#21 lore raw

PATCH v2 13/16 riscv: Enable AI ram on K210

In reply to Sean Anderson

CONFIG_SYS_SDRAM_{BASE,SIZE} are no longer necessary. The default get_effective_memsize will also work. The clock driver creates a bunch of devices, so we need to increase the pre-reloc malloc arena.

---

(no changes since v1)

 board/sipeed/maix/Kconfig          |  2 ++
 board/sipeed/maix/maix.c           | 26 --------------------------
 configs/sipeed_maix_bitm_defconfig |  1 +
 include/configs/sipeed-maix.h      |  4 ----
 4 files changed, 3 insertions(+), 30 deletions(-)
board/sipeed/maix/Kconfig +2 -0
index 4c42dd2087..ba61810079 100644
@@ -42,6 +42,8 @@ config BOARD_SPECIFIC_OPTIONS
 	imply CLK_K210
 	imply DM_RESET
 	imply RESET_SYSCON
+	imply RAM
+	imply K210_SRAM
 	imply SYSRESET
 	imply SYSRESET_SYSCON
 	imply PINCTRL
board/sipeed/maix/maix.c +0 -26
index cbcb23cf5c..1b6f8a5115 100644
@@ -9,33 +9,7 @@
 #include <fdt_support.h>
 #include <asm/io.h>
 
-phys_size_t get_effective_memsize(void)
-{
-	return CONFIG_SYS_SDRAM_SIZE;
-}
-
 int board_init(void)
 {
-	int ret, i;
-	const char * const banks[] = { "sram0", "sram1", "airam" };
-	ofnode memory;
-	struct clk clk;
-
-	/* Enable RAM clocks */
-	memory = ofnode_by_compatible(ofnode_null(), "kendryte,k210-sram");
-	if (ofnode_equal(memory, ofnode_null()))
-		return -ENOENT;
-
-	for (i = 0; i < ARRAY_SIZE(banks); i++) {
-		ret = clk_get_by_name_nodev(memory, banks[i], &clk);
-		if (ret)
-			continue;
-
-		ret = clk_enable(&clk);
-		clk_free(&clk);
-		if (ret)
-			return ret;
-	}
-
 	return 0;
 }
configs/sipeed_maix_bitm_defconfig +1 -0
index 459bf0d530..d0100ad9e9 100644
@@ -1,4 +1,5 @@
 CONFIG_RISCV=y
+CONFIG_SYS_MALLOC_F_LEN=0x10000
 CONFIG_TARGET_SIPEED_MAIX=y
 CONFIG_ARCH_RV64I=y
 CONFIG_STACK_SIZE=0x100000
include/configs/sipeed-maix.h +0 -4
index 36ff522e4b..81956cccfd 100644
@@ -14,10 +14,6 @@
 #define CONFIG_SYS_MALLOC_LEN SZ_128K
 #define CONFIG_SYS_CACHELINE_SIZE 64
 
-#define CONFIG_SYS_SDRAM_BASE 0x80000000
-/* Don't relocate into AI ram since it isn't set up yet */
-#define CONFIG_SYS_SDRAM_SIZE (SZ_4M + SZ_2M)
-
 /* For early init */
 #define K210_SYSCTL_BASE 0x50440000
 
Signature · 1 line
2.28.0
#22 lore raw

PATCH v2 13/16 riscv: Enable AI ram on K210

Quoted 7 lines

CONFIG_SYS_SDRAM_{BASE,SIZE} are no longer necessary. The default get_effective_memsize will also work. The clock driver creates a bunch of devices, so we need to increase the pre-reloc malloc arena.

---

Quoted 8 lines

(no changes since v1)

 board/sipeed/maix/Kconfig          |  2 ++
 board/sipeed/maix/maix.c           | 26 --------------------------
 configs/sipeed_maix_bitm_defconfig |  1 +
 include/configs/sipeed-maix.h      |  4 ----
 4 files changed, 3 insertions(+), 30 deletions(-)
#23 lore raw

PATCH v2 14/16 riscv: k210: Rename airam to aisram

In reply to Sean Anderson

This is more consistent with the naming of other ram banks, and matches what Linux is doing.

---

Changes in v2:
- New

 arch/riscv/dts/k210.dtsi | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
arch/riscv/dts/k210.dtsi +2 -2
index 86d50f14e5..05f345f683 100644
@@ -86,11 +86,11 @@
 		reg = <0x80000000 0x400000>,
 		      <0x80400000 0x200000>,
 		      <0x80600000 0x200000>;
-		reg-names = "sram0", "sram1", "airam";
+		reg-names = "sram0", "sram1", "aisram";
 		clocks = <&sysclk K210_CLK_SRAM0>,
 			 <&sysclk K210_CLK_SRAM1>,
 			 <&sysclk K210_CLK_PLL1>;
-		clock-names = "sram0", "sram1", "airam";
+		clock-names = "sram0", "sram1", "aisram";
 		u-boot,dm-pre-reloc;
 	};
 
Signature · 1 line
2.28.0
#25 lore raw

PATCH v2 15/16 riscv: k210: Use AI as the parent clock of aisram, not PLL1

In reply to Sean Anderson

Testing showed that disabling AI while leaving PLL1 enabled disabled the aisram. This suggests that AI is a more appropriate clock for that ram bank.

---

Changes in v2:
- New

 arch/riscv/dts/k210.dtsi | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/riscv/dts/k210.dtsi +1 -1
index 05f345f683..600fce26af 100644
@@ -89,7 +89,7 @@
 		reg-names = "sram0", "sram1", "aisram";
 		clocks = <&sysclk K210_CLK_SRAM0>,
 			 <&sysclk K210_CLK_SRAM1>,
-			 <&sysclk K210_CLK_PLL1>;
+			 <&sysclk K210_CLK_AI>;
 		clock-names = "sram0", "sram1", "aisram";
 		u-boot,dm-pre-reloc;
 	};
Signature · 1 line
2.28.0
#26 lore raw

PATCH v2 15/16 riscv: k210: Use AI as the parent clock of aisram, not PLL1

Quoted 4 lines

Testing showed that disabling AI while leaving PLL1 enabled disabled the aisram. This suggests that AI is a more appropriate clock for that ram bank.

Quoted 9 lines

---

Changes in v2:
- New

 arch/riscv/dts/k210.dtsi | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
#27 lore raw

PATCH v2 16/16 riscv: Don't reserve AI ram in k210 dts

In reply to Sean Anderson

It is no longer necessary to disallow ai ram, since it is enabled by the sram driver.

---

(no changes since v1)

 arch/riscv/dts/k210.dtsi | 12 ------------
 1 file changed, 12 deletions(-)
arch/riscv/dts/k210.dtsi +0 -12
index 600fce26af..66dd438d50 100644
@@ -94,17 +94,6 @@
 		u-boot,dm-pre-reloc;
 	};
 
-	reserved-memory {
-		#address-cells = <1>;
-		#size-cells = <1>;
-		ranges;
-
-		ai_reserved: ai at 80600000 {
-			reg = <0x80600000 0x200000>;
-			reusable;
-		};
-	};
-
 	clocks {
 		in0: osc {
 			compatible = "fixed-clock";
@@ -179,7 +168,6 @@
 			reg = <0x40800000 0xc00000>;
 			interrupts = <25>;
 			clocks = <&sysclk K210_CLK_AI>;
-			memory-region = <&ai_reserved>;
 			status = "disabled";
 		};
 
Signature · 1 line
2.28.0
#28 lore raw

v2 cover riscv: k210: Enable use of AI ram bank

In reply to Sean Anderson
Quoted 8 lines · Sean Anderson

This ram bank was previously unusable because PLL1 was not started correctly. This series fixes that bug, and allows U-Boot to relocate into the AI ram. This provides an extra 2M of space in which to load payloads.

Second time around and I realized that the bypass clock is no longer necessary (yay). This also necessitates a few more small fixups. I've also added some minor patches to bring the device tree and clock driver closer to what Linux has (or will have). This should be the last round of additions to this series.

I discovered that the clock changes I made cause problems with non-default PLL frequencies (specifically the removal of the bypass clock). I've been working on-and-off on a v3 to fix these problems.

--Sean

Quoted 55 lines

Changes in v2:
- Don't re-enable the PLL
- Remove bypass clock, which is no longer necessary
- Simplify PLL instantiation
- Modify clock tree so clint is a child of aclk
- Sync memory dts node with Linux
- Use correct aisram clock

Sean Anderson (16):
clk: k210: Fix PLLs not being enabled
clk: k210: Fix PLL enable always getting taken
clk: k210: Remove bypass clock
clk: k210: Remove k210_register_pll
clk: k210: Move the clint clock to under aclk
clk: Add support for the k210 clock driver pre-relocation
riscv: Enable some devices pre-relocation
lib: fdt: Add fdtdec_setup_mem_size_base_highest
test: Add a test for fdtdec_setup_mem_size_base et al.
ram: Add driver for K210 SRAM
ram: sifive: Default to y only if compiling for fu540
riscv: Probe ram in dram_init
riscv: Enable AI ram on K210
riscv: k210: Rename airam to aisram
riscv: k210: Use AI as the parent clock of aisram, not PLL1
riscv: Don't reserve AI ram in k210 dts

 MAINTAINERS                        |   1 +
 arch/riscv/cpu/generic/dram.c      |  26 +++
 arch/riscv/dts/k210.dtsi           |  22 +--
 arch/sandbox/dts/test.dts          |  12 ++
 board/sipeed/maix/Kconfig          |   2 +
 board/sipeed/maix/maix.c           |  26 ---
 configs/sandbox64_defconfig        |   2 +-
 configs/sandbox_defconfig          |   2 +-
 configs/sandbox_flattree_defconfig |   2 +-
 configs/sipeed_maix_bitm_defconfig |   1 +
 drivers/clk/kendryte/Makefile      |   2 +-
 drivers/clk/kendryte/bypass.c      | 273 -----------------------------
 drivers/clk/kendryte/clk.c         |  61 +++----
 drivers/clk/kendryte/pll.c         |  26 +--
 drivers/ram/Kconfig                |   7 +
 drivers/ram/Makefile               |   1 +
 drivers/ram/kendryte.c             |  56 ++++++
 drivers/ram/sifive/Kconfig         |   2 +-
 include/configs/sipeed-maix.h      |   4 -
 include/fdtdec.h                   |  19 +-
 include/kendryte/bypass.h          |  31 ----
 include/kendryte/pll.h             |   4 -
 lib/fdtdec.c                       |  34 +++-
 test/dm/fdtdec.c                   |  38 ++++
 24 files changed, 230 insertions(+), 424 deletions(-)
 delete mode 100644 drivers/clk/kendryte/bypass.c
 create mode 100644 drivers/ram/kendryte.c
 delete mode 100644 include/kendryte/bypass.h