Re: 回复: [PATCH v2] firmware: scmi: Add clock v3.2 CONFIG_SET support

[Marek Vasut]

On 10/29/25 9:54 AM, Alice Guo (OSS) wrote:

Hello Alice,

> I tested this patch on IMX95 and here is the log:
> U-Boot SPL 2026.01-rc1-00001-g9bb6f2fc0f57 (Oct 29 2025 - 16:42:02 +0800)
> SYS Boot reason: por, origin: -1, errid: -1
> Normal Boot
> Trying to boot from MMC2
> Primary set selected
> Load image from MMC/SD 0xd7000
> NOTICE:  BL31: v2.12.0(release):android-16.0.0_1.0.0-rc2-2-ga266ff458
> NOTICE:  BL31: Built : 10:13:07, Oct 17 2025
> 
> 
> U-Boot 2026.01-rc1-00001-g9bb6f2fc0f57 (Oct 29 2025 - 16:42:02 +0800)
> 
> CPU:   NXP i.MX95 Rev2.0 A55 at 1800 MHz
> Model: NXP i.MX95 19X19 board
> DRAM:  15.8 GiB
> Core:  283 devices, 24 uclasses, devicetree: separate
> MMC:   "Synchronous Abort" handler, esr 0x96000006, far 0x0
> elr: 0000000090233fac lr : 0000000090233f60 (reloc)
> elr: 00000000fff4dfac lr : 00000000fff4df60
> x0 : 0000000000000004 x1 : 00000000fdf089c8
> x2 : 000000000000000c x3 : 00000000fdf19f18
> x4 : 000000000000001f x5 : 0000000000000000
> x6 : 00000000000013f0 x7 : 00000000fdf08bb0
> x8 : 00000000000013e0 x9 : 00000000fdf087fc
> x10: 0000000000000003 x11: 00000000000013c0
> x12: 0000000000000000 x13: 00000000fdf08bb0
> x14: 00000000fffffffb x15: 0000000000000002
> x16: 00000000fff6887c x17: 0000000000000000
> x18: 00000000fdf19dd0 x19: 00000000fdf2c640
> x20: 0000000000000001 x21: 0000000000000000
> x22: 00000000fffa7750 x23: 00000000fdf29f30
> x24: 00000000fdf2dd58 x25: 0000000000000000
> x26: 0000000000000000 x27: 0000000000000000
> x28: 0000000000000000 x29: 00000000fdf08950
> 
> Code: f90047e2 f9004be1 9101e3e1 f9004fe0 (b94002a0)
> Resetting CPU ...
> 
> resetting ...
> 
> Store the protocol version in a global variable "ver" during probe, and use it in scmi_clk_gate() instead of accessing priv->version. In this setup, the IMX95 board boots successfully.

No global variables please.

How does this crash even happen on MX95 ? I verified that the private 
data are the same pointer in both probe and scmi_clk_gate(), so is the 
parent device. I think this is either not the case for MX95 (how would 
that happen?) or the private data or device is NULL or something ? Can 
you please apply the diff below and share the output ?

"
diff --git a/drivers/clk/clk_scmi.c b/drivers/clk/clk_scmi.c
index 005968e8b75..dcc71ea4225 100644
--- a/drivers/clk/clk_scmi.c
+++ b/drivers/clk/clk_scmi.c
@@ -137,6 +137,7 @@ static int scmi_clk_get_attibute(struct udevice 
*dev, int clkid, char *name,
  static int scmi_clk_gate(struct clk *clk, int enable)
  {
  	struct scmi_clock_priv *priv = dev_get_priv(clk->dev->parent);
+printf("%s[%d] dev=%px / %px priv=%px\n", __func__, __LINE__, clk->dev, 
clk->dev->parent, priv);
  	struct scmi_clk_state_in_v1 in_v1 = {
  		.clock_id = clk_get_id(clk),
  		.attributes = enable,
@@ -335,6 +336,8 @@ static int scmi_clk_probe(struct udevice *dev)
  	ret = scmi_generic_protocol_version(dev, SCMI_PROTOCOL_ID_CLOCK,
  					    &priv->version);

+printf("%s[%d] dev=%px priv=%px\n", __func__, __LINE__, dev, priv);
+
  	/* register CCF children: CLK UCLASS, no probed again */
  	if (device_get_uclass_id(dev->parent) == UCLASS_CLK)
  		return 0;
"

Thank you !