Skip to contentSkip to conversation

Dragonboard 820c - fix board support

PATCH In review Jorge Ramirez-Ortiz 3 participants 21 messages Apr 7, 2025

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

This series addresses the existing issues with U-Boot support for the DragonBoard 820c.

[PATCH 1/7] board: qualcomm: dragonboard820c: udpate readme [PATCH 2/7] clk/qcom: apq8096: fix set rate for the uart clock [PATCH 3/7] mmc: msm_sdhci: handle bulk clock initialization error [PATCH 4/7] clk/qcom: apq8096: fix the sdhci clock [PATCH 5/7] configs: dragonboard820: enable GPIO
[PATCH 6/7] mach-snapdragon: board interface to enforce fdt [PATCH 7/7] board: qualcommm: dragonboard820c: external fdt

#1 lore raw

PATCH 1/7 board: qualcomm: dragonboard820c: udpate readme

Update build instructions.

---
 board/qualcomm/dragonboard820c/readme.txt | 47 ++++++++++-------------
 1 file changed, 21 insertions(+), 26 deletions(-)
board/qualcomm/dragonboard820c/readme.txt +21 -26
index 966e9520e85..5817c837fdf 100644
@@ -16,10 +16,7 @@
                      Build & Run instructions
 ================================================================================
 
-1) Install mkbootimg and dtbTool from Codeaurora:
-
-   git://codeaurora.org/quic/kernel/skales
-   commit 8492547e404e969262d9070dee9bdd15668bb70f worked for me.
+1) Install mkbootimg
 
 2) Setup CROSS_COMPILE to aarch64 compiler or if you use ccache just do
    CROSS_COMPILE="ccache aarch64-linux-gnu-"
@@ -33,15 +30,15 @@
 
    $ touch rd
 
-5) Generate qualcomm device tree table with dtbTool
+5) Append the dtb to the u-boot binary
 
-   $ dtbTool -o dt.img arch/arm/dts
+   $ gzip u-boot-dtb.bin
+   $ cat u-boot.dtb >> u-boot-dtb.bin.gz
 
 6) Generate Android boot image with mkbootimg:
 
-   $ mkbootimg --kernel=u-boot-dtb.bin             \
+   $ mkbootimg --kernel=u-boot-dtb.bin.gz          \
                --output=u-boot.img                 \
-               --dt=dt.img                         \
                --pagesize 4096                     \
                --base 0x80000000                   \
                --ramdisk=rd                        \
@@ -251,44 +248,42 @@ Wait for 5 seconds before proceeding
 [5300] booting linux @ 0x80080000, ramdisk @ 0x82200000 (0), tags/device tree @ 0x82000000
 [5310] Jumping to kernel via monitor
 
-U-Boot 2017.11-00145-ge895117 (Nov 29 2017 - 10:04:06 +0100)
+U-Boot 2025.04-rc5-00020-g40a61ca0e7eb-dirty (Apr 07 2025 - 09:37:03 +0200)
 Qualcomm-DragonBoard 820C
 
-DRAM:  3 GiB
-PSCI:  v1.0
-MMC:   sdhci@74a4900: 0
+DRAM:  3.5 GiB (effective 3 GiB)
+Core:  136 devices, 18 uclasses, devicetree: board
+MMC:   Bulk clocks not available (-19), trying core clock
+mmc@74a4900: 0
+Loading Environment from EXT4... OK
 In:    serial@75b0000
 Out:   serial@75b0000
 Err:   serial@75b0000
-Net:   Net Initialization Skipped
-No ethernet found.
+Net:   No ethernet found.
 Hit any key to stop autoboot:  0
 switch to partitions #0, OK
 mmc0 is current device
 Scanning mmc 0:1...
 Found /extlinux/extlinux.conf
 Retrieving file: /extlinux/extlinux.conf
-433 bytes read in 71 ms (5.9 KiB/s)
 1:      nfs root
-
+Enter choice: 1:        nfs root
 Retrieving file: /uImage
-19397184 bytes read in 2024 ms (9.1 MiB/s)
-append: root=/dev/nfs rw nfsroot=192.168.1.2:/db820c/rootfs,v3,tcp rootwait ip=dhcp consoleblank=0 console=tty0 console=ttyMSM0,115200n8 earlyprintk earlycon=msm_serial_dm,0x75b0000 androidboot.bootdevice=624000.ufshc androidboot.verifiedbootstate=orange androidboot.ver0
-
+append: root=/dev/nfs rw nfsroot=192.168.1.6:/home/jramirez/Src/qualcomm-lt/db820c/rootfs,v3,tcp rootwait ip=dhcp consoleblank=0 console=tty0 console=ttyMSM0,115200n8 earlyprintk earlyco0
 Retrieving file: /apq8096-db820c.dtb
-38134 bytes read in 37 ms (1005.9 KiB/s)
-
-## Booting kernel from Legacy Image at 95000000 ...
+## Booting kernel from Legacy Image at 155000000 ...
    Image Name:   Dragonboard820c
    Image Type:   AArch64 Linux Kernel Image (uncompressed)
    Data Size:    19397120 Bytes = 18.5 MiB
    Load Address: 80080000
    Entry Point:  80080000
    Verifying Checksum ... OK
-## Flattened Device Tree blob at 93000000
-   Booting using the fdt blob at 0x93000000
-   Loading Kernel Image ... OK
-   Using Device Tree in place at 0000000093000000, end 000000009300c4f5
+## Flattened Device Tree blob at 148600000
+   Booting using the fdt blob at 0x148600000
+Working FDT set to 148600000
+   Loading Kernel Image to 80080000
+   Using Device Tree in place at 0000000148600000, end 000000014860c4f5
+Working FDT set to 148600000
 
 Starting kernel ...
 
Signature · 1 line
2.34.1
Reviewed-by on v1 1/7
#2 lore raw

Re: [PATCH 1/7] board: qualcomm: dragonboard820c: udpate readme

Quoted 94 lines · Jorge Ramirez-Ortiz

Update build instructions.

---
  board/qualcomm/dragonboard820c/readme.txt | 47 ++++++++++-------------
  1 file changed, 21 insertions(+), 26 deletions(-)
board/qualcomm/dragonboard820c/readme.txt +15 -25
index 966e9520e85..5817c837fdf 100644
@@ -16,10 +16,7 @@
                       Build & Run instructions
  ================================================================================
 
-1) Install mkbootimg and dtbTool from Codeaurora:
-
-   git://codeaurora.org/quic/kernel/skales
-   commit 8492547e404e969262d9070dee9bdd15668bb70f worked for me.
+1) Install mkbootimg
 
  2) Setup CROSS_COMPILE to aarch64 compiler or if you use ccache just do
     CROSS_COMPILE="ccache aarch64-linux-gnu-"
@@ -33,15 +30,15 @@
 
     $ touch rd
 
-5) Generate qualcomm device tree table with dtbTool
+5) Append the dtb to the u-boot binary
 
-   $ dtbTool -o dt.img arch/arm/dts
+   $ gzip u-boot-dtb.bin
+   $ cat u-boot.dtb >> u-boot-dtb.bin.gz
 
  6) Generate Android boot image with mkbootimg:
 
-   $ mkbootimg --kernel=u-boot-dtb.bin             \
+   $ mkbootimg --kernel=u-boot-dtb.bin.gz          \
                 --output=u-boot.img                 \
-               --dt=dt.img                         \
                 --pagesize 4096                     \
                 --base 0x80000000                   \
                 --ramdisk=rd                        \
@@ -251,44 +248,42 @@ Wait for 5 seconds before proceeding
  [5300] booting linux @ 0x80080000, ramdisk @ 0x82200000 (0), tags/device tree @ 0x82000000
  [5310] Jumping to kernel via monitor
 
-U-Boot 2017.11-00145-ge895117 (Nov 29 2017 - 10:04:06 +0100)
+U-Boot 2025.04-rc5-00020-g40a61ca0e7eb-dirty (Apr 07 2025 - 09:37:03 +0200)
  Qualcomm-DragonBoard 820C
 
-DRAM:  3 GiB
-PSCI:  v1.0
-MMC:   sdhci@74a4900: 0
+DRAM:  3.5 GiB (effective 3 GiB)
+Core:  136 devices, 18 uclasses, devicetree: board
+MMC:   Bulk clocks not available (-19), trying core clock
+mmc@74a4900: 0
+Loading Environment from EXT4... OK
  In:    serial@75b0000
  Out:   serial@75b0000
  Err:   serial@75b0000
-Net:   Net Initialization Skipped
-No ethernet found.
+Net:   No ethernet found.
  Hit any key to stop autoboot:  0
  switch to partitions #0, OK
  mmc0 is current device
  Scanning mmc 0:1...
  Found /extlinux/extlinux.conf
  Retrieving file: /extlinux/extlinux.conf
-433 bytes read in 71 ms (5.9 KiB/s)
  1:      nfs root
-
+Enter choice: 1:        nfs root
  Retrieving file: /uImage
-19397184 bytes read in 2024 ms (9.1 MiB/s)
-append: root=/dev/nfs rw nfsroot=192.168.1.2:/db820c/rootfs,v3,tcp rootwait ip=dhcp consoleblank=0 console=tty0 console=ttyMSM0,115200n8 earlyprintk earlycon=msm_serial_dm,0x75b0000 androidboot.bootdevice=624000.ufshc androidboot.verifiedbootstate=orange androidboot.ver0
-
+append: root=/dev/nfs rw nfsroot=192.168.1.6:/home/jramirez/Src/qualcomm-lt/db820c/rootfs,v3,tcp rootwait ip=dhcp consoleblank=0 console=tty0 console=ttyMSM0,115200n8 earlyprintk earlyco0
  Retrieving file: /apq8096-db820c.dtb
-38134 bytes read in 37 ms (1005.9 KiB/s)
-
-## Booting kernel from Legacy Image at 95000000 ...
+## Booting kernel from Legacy Image at 155000000 ...
     Image Name:   Dragonboard820c
     Image Type:   AArch64 Linux Kernel Image (uncompressed)
     Data Size:    19397120 Bytes = 18.5 MiB
     Load Address: 80080000
     Entry Point:  80080000
     Verifying Checksum ... OK
-## Flattened Device Tree blob at 93000000
-   Booting using the fdt blob at 0x93000000
-   Loading Kernel Image ... OK

board/qualcomm/dragonboard820c/readme.txt Wait for 5 seconds before proceeding · line 289

Quoted 10 lines
board/qualcomm/dragonboard820c/readme.txt +6 -1
-   Using Device Tree in place at 0000000093000000, end 000000009300c4f5
+## Flattened Device Tree blob at 148600000
+   Booting using the fdt blob at 0x148600000
+Working FDT set to 148600000
+   Loading Kernel Image to 80080000
+   Using Device Tree in place at 0000000148600000, end 000000014860c4f5
+Working FDT set to 148600000
 
  Starting kernel ...
 
#3 lore raw

PATCH 2/7 clk/qcom: apq8096: fix set rate for the uart clock

In reply to Jorge Ramirez-Ortiz

The function should return a valid rate.

---
 drivers/clk/qcom/clock-apq8096.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/clk/qcom/clock-apq8096.c +2 -1
index c77d69128b0..bc00826a5e8 100644
@@ -87,7 +87,8 @@ static ulong apq8096_clk_set_rate(struct clk *clk, ulong rate)
 		return clk_init_sdc(priv, rate);
 		break;
 	case GCC_BLSP2_UART2_APPS_CLK: /*UART2*/
-		return clk_init_uart(priv);
+		clk_init_uart(priv);
+		return 7372800;
 	default:
 		return 0;
 	}
Signature · 1 line
2.34.1
Reviewed-by on v1 2/7
#4 lore raw

Re: [PATCH 2/7] clk/qcom: apq8096: fix set rate for the uart clock

Quoted 12 lines · Jorge Ramirez-Ortiz

The function should return a valid rate.

---
  drivers/clk/qcom/clock-apq8096.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
drivers/clk/qcom/clock-apq8096.c +0 -0
index c77d69128b0..bc00826a5e8 100644

drivers/clk/qcom/clock-apq8096.c static ulong apq8096_clk_set_rate(struct clk *clk, ulong rate) · line 94

Quoted 9 lines
drivers/clk/qcom/clock-apq8096.c +2 -1
@@ -87,7 +87,8 @@ static ulong apq8096_clk_set_rate(struct clk *clk, ulong rate)
  		return clk_init_sdc(priv, rate);
  		break;
  	case GCC_BLSP2_UART2_APPS_CLK: /*UART2*/
-		return clk_init_uart(priv);
+		clk_init_uart(priv);
+		return 7372800;
  	default:
  		return 0;
  	}
#5 lore raw

PATCH 3/7 mmc: msm_sdhci: handle bulk clock initialization error

In reply to Jorge Ramirez-Ortiz

Some boards do not require all clocks to be available (i.e:
dragonboard820c).

This change provides a fallback to the core clock when the bulk cant be retrived.

---
 drivers/mmc/msm_sdhci.c | 27 +++++++++++++++++++++++++--
 1 file changed, 25 insertions(+), 2 deletions(-)
drivers/mmc/msm_sdhci.c +25 -2
index 27bb7052fca..8081330bd25 100644
@@ -9,6 +9,7 @@
 
 #include <clk.h>
 #include <dm.h>
+#include <dm/devres.h>
 #include <malloc.h>
 #include <sdhci.h>
 #include <wait_bit.h>
@@ -56,6 +57,17 @@ struct msm_sdhc_variant_info {
 
 DECLARE_GLOBAL_DATA_PTR;
 
+static int get_core_clock(struct udevice *dev, struct clk_bulk *bulk)
+{
+	bulk->count = 1;
+
+	bulk->clks = devm_kcalloc(dev, 1, sizeof(struct clk), GFP_KERNEL);
+	if (!bulk->clks)
+		return -ENOMEM;
+
+	return clk_get_by_name(dev, "core", &bulk->clks[0]);
+}
+
 static int msm_sdc_clk_init(struct udevice *dev)
 {
 	struct msm_sdhc *prv = dev_get_priv(dev);
@@ -73,8 +85,15 @@ static int msm_sdc_clk_init(struct udevice *dev)
 
 	ret = clk_get_bulk(dev, &prv->clks);
 	if (ret) {
-		log_warning("Couldn't get mmc clocks: %d\n", ret);
-		return ret;
+		log_warning("Bulk clocks not available (%d), trying core clock\n", ret);
+
+		/* Sometimes not all clocks are needed - chainloading uboot */
+		ret = get_core_clock(dev, &prv->clks);
+		if (ret) {
+			log_warning("Core clock not available:(%d)\n", ret);
+			return ret;
+		}
+		n_clks = 1;
 	}
 
 	ret = clk_enable_bulk(&prv->clks);
@@ -83,6 +102,9 @@ static int msm_sdc_clk_init(struct udevice *dev)
 		return ret;
 	}
 
+	if (n_clks == 1)
+		goto set_rate;
+
 	/* If clock-names is unspecified, then the first clock is the core clock */
 	if (!ofnode_get_property(node, "clock-names", &n_clks)) {
 		if (!clk_set_rate(&prv->clks.clks[0], clk_rate)) {
@@ -105,6 +127,7 @@ static int msm_sdc_clk_init(struct udevice *dev)
 		return -EINVAL;
 	}
 
+set_rate:
 	/* The clock is already enabled by the clk_bulk above */
 	clk_rate = clk_set_rate(&prv->clks.clks[i], clk_rate);
 	/* If we get a rate of 0 then something has probably gone wrong. */
Signature · 1 line
2.34.1
#6 lore raw

Re: [PATCH 3/7] mmc: msm_sdhci: handle bulk clock initialization error

Quoted 2 lines · Jorge Ramirez-Ortiz

Some boards do not require all clocks to be available (i.e:
dragonboard820c).

Can you specify which clock isn't available ? Because we have clk-stub for that

Thanks,
Neil

Quoted 75 lines

This change provides a fallback to the core clock when the bulk cant be retrived.

---
  drivers/mmc/msm_sdhci.c | 27 +++++++++++++++++++++++++--
  1 file changed, 25 insertions(+), 2 deletions(-)
drivers/mmc/msm_sdhci.c +25 -2
index 27bb7052fca..8081330bd25 100644
@@ -9,6 +9,7 @@
 
  #include <clk.h>
  #include <dm.h>
+#include <dm/devres.h>
  #include <malloc.h>
  #include <sdhci.h>
  #include <wait_bit.h>
@@ -56,6 +57,17 @@ struct msm_sdhc_variant_info {
 
  DECLARE_GLOBAL_DATA_PTR;
 
+static int get_core_clock(struct udevice *dev, struct clk_bulk *bulk)
+{
+	bulk->count = 1;
+
+	bulk->clks = devm_kcalloc(dev, 1, sizeof(struct clk), GFP_KERNEL);
+	if (!bulk->clks)
+		return -ENOMEM;
+
+	return clk_get_by_name(dev, "core", &bulk->clks[0]);
+}
+
  static int msm_sdc_clk_init(struct udevice *dev)
  {
  	struct msm_sdhc *prv = dev_get_priv(dev);
@@ -73,8 +85,15 @@ static int msm_sdc_clk_init(struct udevice *dev)
 
  	ret = clk_get_bulk(dev, &prv->clks);
  	if (ret) {
-		log_warning("Couldn't get mmc clocks: %d\n", ret);
-		return ret;
+		log_warning("Bulk clocks not available (%d), trying core clock\n", ret);
+
+		/* Sometimes not all clocks are needed - chainloading uboot */
+		ret = get_core_clock(dev, &prv->clks);
+		if (ret) {
+			log_warning("Core clock not available:(%d)\n", ret);
+			return ret;
+		}
+		n_clks = 1;
  	}
 
  	ret = clk_enable_bulk(&prv->clks);
@@ -83,6 +102,9 @@ static int msm_sdc_clk_init(struct udevice *dev)
  		return ret;
  	}
 
+	if (n_clks == 1)
+		goto set_rate;
+
  	/* If clock-names is unspecified, then the first clock is the core clock */
  	if (!ofnode_get_property(node, "clock-names", &n_clks)) {
  		if (!clk_set_rate(&prv->clks.clks[0], clk_rate)) {
@@ -105,6 +127,7 @@ static int msm_sdc_clk_init(struct udevice *dev)
  		return -EINVAL;
  	}
 
+set_rate:
  	/* The clock is already enabled by the clk_bulk above */
  	clk_rate = clk_set_rate(&prv->clks.clks[i], clk_rate);
  	/* If we get a rate of 0 then something has probably gone wrong. */
#7 lore raw

Re: [PATCH 3/7] mmc: msm_sdhci: handle bulk clock initialization error

Quoted 5 lines · On 07/04/2025 11:02, [email protected] wrote:
Quoted 2 lines · Jorge Ramirez-Ortiz

Some boards do not require all clocks to be available (i.e:
dragonboard820c).

Can you specify which clock isn't available ? Because we have clk-stub for that

Enable:
CONFIG_CLK_STUB=y

and change this:

drivers/clk/clk-stub.c +1 -1
index 343fa5cd3fe..c14f5b6e1a6 100644
@@ -14,7 +14,7 @@
 static const struct udevice_id nop_parent_ids[] = {
        { .compatible = "qcom,rpm-proc" },
        { .compatible = "qcom,glink-rpm" },
-       { .compatible = "qcom,rpm-sm6115" },
+       { .compatible = "qcom,glink-smd-rpm" },
        { }
 };
 

And clk_get_bulk() should work!

Neil

Quoted 73 lines

Thanks,
Neil

Quoted 68 lines

This change provides a fallback to the core clock when the bulk cant be retrived.

---
  drivers/mmc/msm_sdhci.c | 27 +++++++++++++++++++++++++--

  1 file changed, 25 insertions(+), 2 deletions(-)
drivers/mmc/msm_sdhci.c +25 -2
index 27bb7052fca..8081330bd25 100644
@@ -9,6 +9,7 @@
   #include <clk.h>
   #include <dm.h>
+#include <dm/devres.h>
   #include <malloc.h>
   #include <sdhci.h>
   #include <wait_bit.h>
@@ -56,6 +57,17 @@ struct msm_sdhc_variant_info {
   DECLARE_GLOBAL_DATA_PTR;
+static int get_core_clock(struct udevice *dev, struct clk_bulk *bulk)
+{
+    bulk->count = 1;
+
+    bulk->clks = devm_kcalloc(dev, 1, sizeof(struct clk), GFP_KERNEL);
+    if (!bulk->clks)
+        return -ENOMEM;
+
+    return clk_get_by_name(dev, "core", &bulk->clks[0]);
+}
+
   static int msm_sdc_clk_init(struct udevice *dev)
   {
       struct msm_sdhc *prv = dev_get_priv(dev);
@@ -73,8 +85,15 @@ static int msm_sdc_clk_init(struct udevice *dev)
       ret = clk_get_bulk(dev, &prv->clks);
       if (ret) {
-        log_warning("Couldn't get mmc clocks: %d\n", ret);
-        return ret;
+        log_warning("Bulk clocks not available (%d), trying core clock\n", ret);
+
+        /* Sometimes not all clocks are needed - chainloading uboot */
+        ret = get_core_clock(dev, &prv->clks);
+        if (ret) {
+            log_warning("Core clock not available:(%d)\n", ret);
+            return ret;
+        }
+        n_clks = 1;
       }
       ret = clk_enable_bulk(&prv->clks);
@@ -83,6 +102,9 @@ static int msm_sdc_clk_init(struct udevice *dev)
           return ret;
       }
+    if (n_clks == 1)
+        goto set_rate;
+
       /* If clock-names is unspecified, then the first clock is the core clock */
       if (!ofnode_get_property(node, "clock-names", &n_clks)) {
           if (!clk_set_rate(&prv->clks.clks[0], clk_rate)) {
@@ -105,6 +127,7 @@ static int msm_sdc_clk_init(struct udevice *dev)
           return -EINVAL;
       }
+set_rate:
       /* The clock is already enabled by the clk_bulk above */
       clk_rate = clk_set_rate(&prv->clks.clks[i], clk_rate);
       /* If we get a rate of 0 then something has probably gone wrong. */
#8 lore raw

Re: [PATCH 3/7] mmc: msm_sdhci: handle bulk clock initialization error

Quoted 6 lines · Neil Armstrong
Quoted 5 lines · On 07/04/2025 11:02, [email protected] wrote:
Quoted 2 lines · Jorge Ramirez-Ortiz

Some boards do not require all clocks to be available (i.e:
dragonboard820c).

Can you specify which clock isn't available ? Because we have clk-stub for that

right, either GCC_SDCC2_AHB_CLK or RPM_SMD_XO_CLK_SRC (we have GCC_SDCC2_APPS_CLK which is the one I needed to initialize and work with MMC).

Quoted 18 lines

Enable:
CONFIG_CLK_STUB=y

and change this:

drivers/clk/clk-stub.c +1 -1
index 343fa5cd3fe..c14f5b6e1a6 100644
@@ -14,7 +14,7 @@
 static const struct udevice_id nop_parent_ids[] = {
        { .compatible = "qcom,rpm-proc" },
        { .compatible = "qcom,glink-rpm" },
-       { .compatible = "qcom,rpm-sm6115" },
+       { .compatible = "qcom,glink-smd-rpm" },
        { }
 };
 
Quoted 1 line

And clk_get_bulk() should work!

ah cool. yes that did work but something else popped up is it worth looking further into this?

U-Boot 2025.04-rc5-00022-gccd064439bc2-dirty (Apr 07 2025 - 11:30:21 +0200) Qualcomm-DragonBoard 820C

DRAM:  3.5 GiB (effective 3 GiB)
Core:  140 devices, 18 uclasses, devicetree: board
MMC:   Couldn't set MMC core clock rate: 0

Couldn't set MMC core clock rate: 0
mmc@74a4900 - probe failed: -22
Couldn't set MMC core clock rate: 0
Loading Environment from EXT4... Couldn't set MMC core clock rate: 0 ** Bad device specification mmc 0 **

In:    serial@75b0000
Out:   serial@75b0000
Err:   serial@75b0000
Net:   No ethernet found.

dragonboard820c =>

Quoted 77 lines

Neil

Quoted 73 lines

Thanks,
Neil

Quoted 68 lines

This change provides a fallback to the core clock when the bulk cant be retrived.

---
  drivers/mmc/msm_sdhci.c | 27 +++++++++++++++++++++++++--

  1 file changed, 25 insertions(+), 2 deletions(-)
drivers/mmc/msm_sdhci.c +25 -2
index 27bb7052fca..8081330bd25 100644
@@ -9,6 +9,7 @@
   #include <clk.h>
   #include <dm.h>
+#include <dm/devres.h>
   #include <malloc.h>
   #include <sdhci.h>
   #include <wait_bit.h>
@@ -56,6 +57,17 @@ struct msm_sdhc_variant_info {
   DECLARE_GLOBAL_DATA_PTR;
+static int get_core_clock(struct udevice *dev, struct clk_bulk *bulk)
+{
+    bulk->count = 1;
+
+    bulk->clks = devm_kcalloc(dev, 1, sizeof(struct clk), GFP_KERNEL);
+    if (!bulk->clks)
+        return -ENOMEM;
+
+    return clk_get_by_name(dev, "core", &bulk->clks[0]);
+}
+
   static int msm_sdc_clk_init(struct udevice *dev)
   {
       struct msm_sdhc *prv = dev_get_priv(dev);
@@ -73,8 +85,15 @@ static int msm_sdc_clk_init(struct udevice *dev)
       ret = clk_get_bulk(dev, &prv->clks);
       if (ret) {
-        log_warning("Couldn't get mmc clocks: %d\n", ret);
-        return ret;
+        log_warning("Bulk clocks not available (%d), trying core clock\n", ret);
+
+        /* Sometimes not all clocks are needed - chainloading uboot */
+        ret = get_core_clock(dev, &prv->clks);
+        if (ret) {
+            log_warning("Core clock not available:(%d)\n", ret);
+            return ret;
+        }
+        n_clks = 1;
       }
       ret = clk_enable_bulk(&prv->clks);
@@ -83,6 +102,9 @@ static int msm_sdc_clk_init(struct udevice *dev)
           return ret;
       }
+    if (n_clks == 1)
+        goto set_rate;
+
       /* If clock-names is unspecified, then the first clock is the core clock */
       if (!ofnode_get_property(node, "clock-names", &n_clks)) {
           if (!clk_set_rate(&prv->clks.clks[0], clk_rate)) {
@@ -105,6 +127,7 @@ static int msm_sdc_clk_init(struct udevice *dev)
           return -EINVAL;
       }
+set_rate:
       /* The clock is already enabled by the clk_bulk above */
       clk_rate = clk_set_rate(&prv->clks.clks[i], clk_rate);
       /* If we get a rate of 0 then something has probably gone wrong. */
#9 lore raw

Re: [PATCH 3/7] mmc: msm_sdhci: handle bulk clock initialization error

Quoted 30 lines · Jorge Ramirez
Quoted 6 lines · Neil Armstrong
Quoted 5 lines · On 07/04/2025 11:02, [email protected] wrote:
Quoted 2 lines · Jorge Ramirez-Ortiz

Some boards do not require all clocks to be available (i.e:
dragonboard820c).

Can you specify which clock isn't available ? Because we have clk-stub for that

right, either GCC_SDCC2_AHB_CLK or RPM_SMD_XO_CLK_SRC (we have GCC_SDCC2_APPS_CLK which is the one I needed to initialize and work with MMC).

Quoted 18 lines

Enable:
CONFIG_CLK_STUB=y

and change this:

drivers/clk/clk-stub.c +1 -1
index 343fa5cd3fe..c14f5b6e1a6 100644
@@ -14,7 +14,7 @@
  static const struct udevice_id nop_parent_ids[] = {
         { .compatible = "qcom,rpm-proc" },
         { .compatible = "qcom,glink-rpm" },
-       { .compatible = "qcom,rpm-sm6115" },
+       { .compatible = "qcom,glink-smd-rpm" },
         { }
  };
 
Quoted 4 lines
Quoted 1 line

And clk_get_bulk() should work!

ah cool. yes that did work but something else popped up is it worth looking further into this?

Yeah look at clk_init_sdc(), it sets SDCC2_AHB & SDCC2_APPS, so you can change next patch to:

drivers/clk/qcom/clock-apq8096.c +1 -1
index c77d69128b0..832cda332dc 100644
@@ -83,7 +83,7 @@ static ulong apq8096_clk_set_rate(struct clk *clk, ulong rate)
        struct msm_clk_priv *priv = dev_get_priv(clk->dev);
 
        switch (clk->id) {
-       case GCC_SDCC1_APPS_CLK: /* SDC1 */
+       case GCC_SDCC2_AHB_CLK: /* SDC2 */
                return clk_init_sdc(priv, rate);
                break;
        case GCC_BLSP2_UART2_APPS_CLK: /*UART2*/

Neil

Quoted 94 lines

U-Boot 2025.04-rc5-00022-gccd064439bc2-dirty (Apr 07 2025 - 11:30:21 +0200) Qualcomm-DragonBoard 820C

DRAM:  3.5 GiB (effective 3 GiB)
Core:  140 devices, 18 uclasses, devicetree: board
MMC:   Couldn't set MMC core clock rate: 0

Couldn't set MMC core clock rate: 0
mmc@74a4900 - probe failed: -22
Couldn't set MMC core clock rate: 0
Loading Environment from EXT4... Couldn't set MMC core clock rate: 0 ** Bad device specification mmc 0 **

In:    serial@75b0000
Out:   serial@75b0000
Err:   serial@75b0000
Net:   No ethernet found.

dragonboard820c =>

Quoted 77 lines

Neil

Quoted 73 lines

Thanks,
Neil

Quoted 68 lines

This change provides a fallback to the core clock when the bulk cant be retrived.

---
   drivers/mmc/msm_sdhci.c | 27 +++++++++++++++++++++++++--
   1 file changed, 25 insertions(+), 2 deletions(-)
drivers/mmc/msm_sdhci.c +25 -2
index 27bb7052fca..8081330bd25 100644
@@ -9,6 +9,7 @@
   #include <clk.h>
   #include <dm.h>
+#include <dm/devres.h>
   #include <malloc.h>
   #include <sdhci.h>
   #include <wait_bit.h>
@@ -56,6 +57,17 @@ struct msm_sdhc_variant_info {
   DECLARE_GLOBAL_DATA_PTR;
+static int get_core_clock(struct udevice *dev, struct clk_bulk *bulk)
+{
+    bulk->count = 1;
+
+    bulk->clks = devm_kcalloc(dev, 1, sizeof(struct clk), GFP_KERNEL);
+    if (!bulk->clks)
+        return -ENOMEM;
+
+    return clk_get_by_name(dev, "core", &bulk->clks[0]);
+}
+
   static int msm_sdc_clk_init(struct udevice *dev)
   {
       struct msm_sdhc *prv = dev_get_priv(dev);
@@ -73,8 +85,15 @@ static int msm_sdc_clk_init(struct udevice *dev)
       ret = clk_get_bulk(dev, &prv->clks);
       if (ret) {
-        log_warning("Couldn't get mmc clocks: %d\n", ret);
-        return ret;
+        log_warning("Bulk clocks not available (%d), trying core clock\n", ret);
+
+        /* Sometimes not all clocks are needed - chainloading uboot */
+        ret = get_core_clock(dev, &prv->clks);
+        if (ret) {
+            log_warning("Core clock not available:(%d)\n", ret);
+            return ret;
+        }
+        n_clks = 1;
       }
       ret = clk_enable_bulk(&prv->clks);
@@ -83,6 +102,9 @@ static int msm_sdc_clk_init(struct udevice *dev)
           return ret;
       }
+    if (n_clks == 1)
+        goto set_rate;
+
       /* If clock-names is unspecified, then the first clock is the core clock */
       if (!ofnode_get_property(node, "clock-names", &n_clks)) {
           if (!clk_set_rate(&prv->clks.clks[0], clk_rate)) {
@@ -105,6 +127,7 @@ static int msm_sdc_clk_init(struct udevice *dev)
           return -EINVAL;
       }
+set_rate:
       /* The clock is already enabled by the clk_bulk above */
       clk_rate = clk_set_rate(&prv->clks.clks[i], clk_rate);
       /* If we get a rate of 0 then something has probably gone wrong. */
#10 lore raw

Re: [PATCH 3/7] mmc: msm_sdhci: handle bulk clock initialization error

Quoted 9 lines · Neil Armstrong
Quoted 8 lines · Jorge Ramirez
Quoted 6 lines · Neil Armstrong
Quoted 5 lines · On 07/04/2025 11:02, [email protected] wrote:
Quoted 2 lines · Jorge Ramirez-Ortiz

Some boards do not require all clocks to be available (i.e:
dragonboard820c).

Can you specify which clock isn't available ? Because we have clk-stub for that

Quoted 7 lines
Quoted 7 lines

right, either GCC_SDCC2_AHB_CLK or RPM_SMD_XO_CLK_SRC (we have GCC_SDCC2_APPS_CLK which is the one I needed to initialize and work with MMC).

Quoted 3 lines

Enable:
CONFIG_CLK_STUB=y

ok, taking this into the db820 defconfig

Quoted 14 lines
Quoted 14 lines
Quoted 7 lines

and change this:

drivers/clk/clk-stub.c +0 -0
index 343fa5cd3fe..c14f5b6e1a6 100644

drivers/clk/clk-stub.c line 19

Quoted 7 lines
drivers/clk/clk-stub.c +1 -1
@@ -14,7 +14,7 @@
  static const struct udevice_id nop_parent_ids[] = {
         { .compatible = "qcom,rpm-proc" },
         { .compatible = "qcom,glink-rpm" },
-       { .compatible = "qcom,rpm-sm6115" },
+       { .compatible = "qcom,glink-smd-rpm" },
         { }
  };

so do we need this change to be persistent? are you commiting it?

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

And clk_get_bulk() should work!

ah cool. yes that did work but something else popped up is it worth looking further into this?

Yeah look at clk_init_sdc(), it sets SDCC2_AHB & SDCC2_APPS, so you can change next patch to:

perfect. all good now.

Quoted 18 lines
drivers/clk/qcom/clock-apq8096.c +1 -1
index c77d69128b0..832cda332dc 100644
@@ -83,7 +83,7 @@ static ulong apq8096_clk_set_rate(struct clk *clk, ulong rate)
        struct msm_clk_priv *priv = dev_get_priv(clk->dev);
 
        switch (clk->id) {
-       case GCC_SDCC1_APPS_CLK: /* SDC1 */
+       case GCC_SDCC2_AHB_CLK: /* SDC2 */
                return clk_init_sdc(priv, rate);
                break;
        case GCC_BLSP2_UART2_APPS_CLK: /*UART2*/

Neil

#11 lore raw

PATCH 4/7 clk/qcom: apq8096: fix the sdhci clock

In reply to Jorge Ramirez-Ortiz

Select the right clock for sdhci.

---
 drivers/clk/qcom/clock-apq8096.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/clk/qcom/clock-apq8096.c +1 -1
index bc00826a5e8..551f52d5197 100644
@@ -83,7 +83,7 @@ static ulong apq8096_clk_set_rate(struct clk *clk, ulong rate)
 	struct msm_clk_priv *priv = dev_get_priv(clk->dev);
 
 	switch (clk->id) {
-	case GCC_SDCC1_APPS_CLK: /* SDC1 */
+	case GCC_SDCC2_APPS_CLK: /* SDC2 */
 		return clk_init_sdc(priv, rate);
 		break;
 	case GCC_BLSP2_UART2_APPS_CLK: /*UART2*/
Signature · 1 line
2.34.1
Reviewed-by on v1 4/7
#12 lore raw

Re: [PATCH 4/7] clk/qcom: apq8096: fix the sdhci clock

Quoted 12 lines · Jorge Ramirez-Ortiz

Select the right clock for sdhci.

---
  drivers/clk/qcom/clock-apq8096.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/clk/qcom/clock-apq8096.c +0 -0
index bc00826a5e8..551f52d5197 100644

drivers/clk/qcom/clock-apq8096.c static ulong apq8096_clk_set_rate(struct clk *clk, ulong rate) · line 89

Quoted 8 lines
drivers/clk/qcom/clock-apq8096.c +1 -1
@@ -83,7 +83,7 @@ static ulong apq8096_clk_set_rate(struct clk *clk, ulong rate)
  	struct msm_clk_priv *priv = dev_get_priv(clk->dev);
 
  	switch (clk->id) {
-	case GCC_SDCC1_APPS_CLK: /* SDC1 */
+	case GCC_SDCC2_APPS_CLK: /* SDC2 */
  		return clk_init_sdc(priv, rate);
  		break;
  	case GCC_BLSP2_UART2_APPS_CLK: /*UART2*/
#13 lore raw

PATCH 5/7 configs: dragonboard820: enable GPIO

In reply to Jorge Ramirez-Ortiz

Configure GPIO.

---
 configs/dragonboard820c_defconfig | 1 +
 1 file changed, 1 insertion(+)
configs/dragonboard820c_defconfig +1 -0
index e29bea7deb2..b5b6bc026e8 100644
@@ -45,4 +45,5 @@ CONFIG_PINCTRL_QCOM_APQ8096=y
 CONFIG_DM_PMIC=y
 CONFIG_PMIC_QCOM=y
 CONFIG_MSM_SERIAL=y
+CONFIG_MSM_GPIO=y
 CONFIG_SPMI_MSM=y
Signature · 1 line
2.34.1
Reviewed-by on v1 5/7
#14 lore raw

Re: [PATCH 5/7] configs: dragonboard820: enable GPIO

Quoted 12 lines · Jorge Ramirez-Ortiz

Configure GPIO.

---
  configs/dragonboard820c_defconfig | 1 +
  1 file changed, 1 insertion(+)
configs/dragonboard820c_defconfig +0 -0
index e29bea7deb2..b5b6bc026e8 100644

configs/dragonboard820c_defconfig CONFIG_PINCTRL_QCOM_APQ8096=y · line 49

Quoted 5 lines
configs/dragonboard820c_defconfig +1 -0
@@ -45,4 +45,5 @@ CONFIG_PINCTRL_QCOM_APQ8096=y
  CONFIG_DM_PMIC=y
  CONFIG_PMIC_QCOM=y
  CONFIG_MSM_SERIAL=y
+CONFIG_MSM_GPIO=y
  CONFIG_SPMI_MSM=y
#15 lore raw

PATCH 6/7 mach-snapdragon: board interface to enforce fdt

In reply to Jorge Ramirez-Ortiz

Some platforms might require the external fdt to boot.

This change provides a mechanism to the board initializaion code to make the requet explicit.

---
 arch/arm/mach-snapdragon/board.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)
arch/arm/mach-snapdragon/board.c +13 -0
index 2ef936aab75..1bca9dce20f 100644
@@ -150,7 +150,13 @@ static void show_psci_version(void)
 /* We support booting U-Boot with an internal DT when running as a first-stage bootloader
  * or for supporting quirky devices where it's easier to leave the downstream DT in place
  * to improve ABL compatibility. Otherwise, we use the DT provided by ABL.
+ *
+ * Alternatively the board can enforce the external fdt.
  */
+bool __weak qcom_board_fdt_external(void)
+{
+	return false;
+}
 int board_fdt_blob_setup(void **fdtp)
 {
 	struct fdt_header *fdt;
@@ -161,6 +167,12 @@ int board_fdt_blob_setup(void **fdtp)
 	external_valid = fdt && !fdt_check_header(fdt);
 	internal_valid = !fdt_check_header(*fdtp);
 
+	if (qcom_board_fdt_external() && external_valid) {
+		debug("Using external FDT\n");
+		*fdtp = fdt;
+		goto out;
+	}
+
 	/*
 	 * There is no point returning an error here, U-Boot can't do anything useful in this situation.
 	 * Bail out while we can still print a useful error message.
@@ -178,6 +190,7 @@ int board_fdt_blob_setup(void **fdtp)
 		*fdtp = fdt;
 	}
 
+out:
 	/*
 	 * Parse the /memory node while we're here,
 	 * this makes it easy to do other things early.
Signature · 1 line
2.34.1
#16 lore raw

Re: [PATCH 6/7] mach-snapdragon: board interface to enforce fdt

Hi,

Quoted 1 line · Jorge Ramirez-Ortiz

Some platforms might require the external fdt to boot.

Why ? we may want to build the fdt in, and discard the external fdt for some reasons.

Quoted 48 lines

This change provides a mechanism to the board initializaion code to make the requet explicit.

---
  arch/arm/mach-snapdragon/board.c | 13 +++++++++++++
  1 file changed, 13 insertions(+)
arch/arm/mach-snapdragon/board.c +13 -0
index 2ef936aab75..1bca9dce20f 100644
@@ -150,7 +150,13 @@ static void show_psci_version(void)
  /* We support booting U-Boot with an internal DT when running as a first-stage bootloader
   * or for supporting quirky devices where it's easier to leave the downstream DT in place
   * to improve ABL compatibility. Otherwise, we use the DT provided by ABL.
+ *
+ * Alternatively the board can enforce the external fdt.
   */
+bool __weak qcom_board_fdt_external(void)
+{
+	return false;
+}
  int board_fdt_blob_setup(void **fdtp)
  {
  	struct fdt_header *fdt;
@@ -161,6 +167,12 @@ int board_fdt_blob_setup(void **fdtp)
  	external_valid = fdt && !fdt_check_header(fdt);
  	internal_valid = !fdt_check_header(*fdtp);
 
+	if (qcom_board_fdt_external() && external_valid) {
+		debug("Using external FDT\n");
+		*fdtp = fdt;
+		goto out;
+	}
+
  	/*
  	 * There is no point returning an error here, U-Boot can't do anything useful in this situation.
  	 * Bail out while we can still print a useful error message.
@@ -178,6 +190,7 @@ int board_fdt_blob_setup(void **fdtp)
  		*fdtp = fdt;
  	}
 
+out:
  	/*
  	 * Parse the /memory node while we're here,
  	 * this makes it easy to do other things early.
#17 lore raw

Re: [PATCH 6/7] mach-snapdragon: board interface to enforce fdt

Quoted 6 lines · On 07/04/25 11:04:30, [email protected] wrote:

Hi,

Quoted 1 line · Jorge Ramirez-Ortiz

Some platforms might require the external fdt to boot.

Why ? we may want to build the fdt in, and discard the external fdt for some reasons.

on the db820c, the device tree that u-boot builds is passed to the LK appendeded with the image via fastboot: the LK, processes it, ammends the device tree and then hands it back to u-boot.

see the boot trace:

[2830] Only one appended non-skales DTB, select it.
[2850] cmdline: androidboot.bootdevice=624000.ufshc androidboot.verifiedbootstate=orange androidboot.veritymode=enforcing androidboot.serialno=2f490ecf androidboot.baseband=apq mdss_mdp0 [2860] Updating device tree: start
[2860] Updating device tree: done
[2870] Return value from recv_data: 14
[2880] RPM GLINK UnInit
[2880] Qseecom De-Init Done in Appsbl
[2880] booting linux @ 0x80080000, ramdisk @ 0x82200000 (0), tags/device tree @ 0x82000000
[2890] Jumping to kernel via monitor

U-Boot 2025.04-rc5-00022-gccd064439bc2 (Apr 07 2025 - 10:06:51 +0200) Qualcomm-DragonBoard 820C

It is this external fdt that we need; so I thought allowing the board to request the external fdt explicitily might make sense.

if you have some other idea/suggestion, please let me know, happy to abide.

Quoted 48 lines
Quoted 47 lines

This change provides a mechanism to the board initializaion code to make the requet explicit.

---
  arch/arm/mach-snapdragon/board.c | 13 +++++++++++++
  1 file changed, 13 insertions(+)
arch/arm/mach-snapdragon/board.c +13 -0
index 2ef936aab75..1bca9dce20f 100644
@@ -150,7 +150,13 @@ static void show_psci_version(void)
  /* We support booting U-Boot with an internal DT when running as a first-stage bootloader
   * or for supporting quirky devices where it's easier to leave the downstream DT in place
   * to improve ABL compatibility. Otherwise, we use the DT provided by ABL.
+ *
+ * Alternatively the board can enforce the external fdt.
   */
+bool __weak qcom_board_fdt_external(void)
+{
+	return false;
+}
  int board_fdt_blob_setup(void **fdtp)
  {
  	struct fdt_header *fdt;
@@ -161,6 +167,12 @@ int board_fdt_blob_setup(void **fdtp)
  	external_valid = fdt && !fdt_check_header(fdt);
  	internal_valid = !fdt_check_header(*fdtp);
+	if (qcom_board_fdt_external() && external_valid) {
+		debug("Using external FDT\n");
+		*fdtp = fdt;
+		goto out;
+	}
+
  	/*
  	 * There is no point returning an error here, U-Boot can't do anything useful in this situation.
  	 * Bail out while we can still print a useful error message.
@@ -178,6 +190,7 @@ int board_fdt_blob_setup(void **fdtp)
  		*fdtp = fdt;
  	}
+out:
  	/*
  	 * Parse the /memory node while we're here,
  	 * this makes it easy to do other things early.
#18 lore raw

Re: [PATCH 6/7] mach-snapdragon: board interface to enforce fdt

In reply to Jorge Ramirez-Ortiz

Hi Jorge,

Quoted 1 line · Jorge Ramirez-Ortiz

Some platforms might require the external fdt to boot.

The better way to do this is to use the u-boot-nodtb.bin to build your boot image rather than the one which includes the internal FDT. If that works I'd prefer it (and it can be mentioned in the build instructions). Otherwise it would be good to know what the issue is here.

Kind regards,>

Quoted 47 lines

This change provides a mechanism to the board initializaion code to make the requet explicit.

---
  arch/arm/mach-snapdragon/board.c | 13 +++++++++++++
  1 file changed, 13 insertions(+)
arch/arm/mach-snapdragon/board.c +13 -0
index 2ef936aab75..1bca9dce20f 100644
@@ -150,7 +150,13 @@ static void show_psci_version(void)
  /* We support booting U-Boot with an internal DT when running as a first-stage bootloader
   * or for supporting quirky devices where it's easier to leave the downstream DT in place
   * to improve ABL compatibility. Otherwise, we use the DT provided by ABL.
+ *
+ * Alternatively the board can enforce the external fdt.
   */
+bool __weak qcom_board_fdt_external(void)
+{
+	return false;
+}
  int board_fdt_blob_setup(void **fdtp)
  {
  	struct fdt_header *fdt;
@@ -161,6 +167,12 @@ int board_fdt_blob_setup(void **fdtp)
  	external_valid = fdt && !fdt_check_header(fdt);
  	internal_valid = !fdt_check_header(*fdtp);
 
+	if (qcom_board_fdt_external() && external_valid) {
+		debug("Using external FDT\n");
+		*fdtp = fdt;
+		goto out;
+	}
+
  	/*
  	 * There is no point returning an error here, U-Boot can't do anything useful in this situation.
  	 * Bail out while we can still print a useful error message.
@@ -178,6 +190,7 @@ int board_fdt_blob_setup(void **fdtp)
  		*fdtp = fdt;
  	}
 
+out:
  	/*
  	 * Parse the /memory node while we're here,
  	 * this makes it easy to do other things early.
Signature · 1 line
Caleb (they/them)
#19 lore raw

Re: [PATCH 6/7] mach-snapdragon: board interface to enforce fdt

Quoted 8 lines · Caleb Connolly

Hi Jorge,

Quoted 1 line · Jorge Ramirez-Ortiz

Some platforms might require the external fdt to boot.

The better way to do this is to use the u-boot-nodtb.bin to build your boot image rather than the one which includes the internal FDT. If that works I'd prefer it (and it can be mentioned in the build instructions).

ah of course, yep, will do that.

#20 lore raw

PATCH 7/7 board: qualcommm: dragonboard820c: external fdt

In reply to Jorge Ramirez-Ortiz

The chainloaded u-boot requires the external fdt.

---
 board/qualcomm/dragonboard820c/dragonboard820c.c | 5 +++++
 1 file changed, 5 insertions(+)
board/qualcomm/dragonboard820c/dragonboard820c.c +5 -0
index d3333a59db0..e7f477b69fe 100644
@@ -94,6 +94,11 @@ static void sdhci_power_init(void)
 			rclk[i].val  << rclk[i].bit);
 }
 
+bool qcom_board_fdt_external(void)
+{
+	return true;
+}
+
 void qcom_board_init(void)
 {
 	sdhci_power_init();
Signature · 1 line
2.34.1