[PATCH] ASoC: SOF: amd: Load alternatively signed firmware with an "_l" suffix

[PATCH] ASoC: SOF: amd: Load alternatively signed firmware with an "_l" suffix

[Mario Limonciello]

Platforms ship BIOS with either a default or an alternative trust
authority. The PSP validates the signed DSP firmware image so it must
be signed with the matching scheme. Firmware for the alternative scheme
is named with an "_l" suffix before the extension
(e.g. "sof-X-code.bin" becomes "sof-X-code_l.bin").

The PSP reports the signing type in bit 6 of C2PMSG_63
(0 = default, 1 = alternative). Read it over SMN, as is already done
for the other PSP mailbox registers, when building the signed code
image name, and add the "_l" suffix for the alternative scheme.

Reviewed-by: Vijendar Mukunda <Vijendar.Mukunda@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
---
 sound/soc/sof/amd/acp.c | 33 ++++++++++++++++++++++++++++++++-
 sound/soc/sof/amd/acp.h |  2 ++
 2 files changed, 34 insertions(+), 1 deletion(-)

diff --git a/sound/soc/sof/amd/acp.c b/sound/soc/sof/amd/acp.c
index 86e2402b8ffd2..2f1a8be8460d4 100644
--- a/sound/soc/sof/amd/acp.c
+++ b/sound/soc/sof/amd/acp.c
@@ -1212,6 +1212,31 @@ static int amd_sof_sdw_exit(struct snd_sof_dev *sdev)
 }
 #endif
 
+/*
+ * The PSP validates the signed DSP firmware, so it must be signed with the
+ * scheme the platform firmware uses: images for the alternative scheme are
+ * named with an "_l" suffix before the extension, the default ones have none.
+ */
+static int amd_sof_acp_get_fw_suffix(struct snd_sof_dev *sdev, const char **suffix)
+{
+	u32 val;
+	int ret;
+
+	ret = amd_smn_read(0, MP0_C2PMSG_63_REG, &val);
+	if (ret) {
+		dev_err(sdev->dev, "failed to read PSP firmware signing type: %d\n", ret);
+		return ret;
+	}
+
+	/* all ones means the register isn't accessible */
+	if (val == 0xffffffff)
+		return -EIO;
+
+	*suffix = (val & MP0_C2PMSG_63_FW_SIGN_ALT) ? "_l" : "";
+
+	return 0;
+}
+
 int amd_sof_acp_probe(struct snd_sof_dev *sdev)
 {
 	struct pci_dev *pci = to_pci_dev(sdev->dev);
@@ -1448,8 +1473,14 @@ int amd_sof_acp7x_probe(struct snd_sof_dev *sdev)
 	sdev->debug_box.size = BOX_SIZE_1024;
 
 	if (adata->acp_sof_signed_firmware_image) {
+		const char *fw_suffix;
+
+		ret = amd_sof_acp_get_fw_suffix(sdev, &fw_suffix);
+		if (ret)
+			goto free_ipc_irq;
+
 		adata->fw_code_bin = devm_kasprintf(sdev->dev, GFP_KERNEL,
-						    "sof-%s-code.bin", chip->name);
+						    "sof-%s-code%s.bin", chip->name, fw_suffix);
 		if (!adata->fw_code_bin) {
 			ret = -ENOMEM;
 			goto free_ipc_irq;
diff --git a/sound/soc/sof/amd/acp.h b/sound/soc/sof/amd/acp.h
index b17a36b31ee0d..81eeeabc5cf5c 100644
--- a/sound/soc/sof/amd/acp.h
+++ b/sound/soc/sof/amd/acp.h
@@ -97,6 +97,8 @@
 #define ACP_EXT_INTR_ERROR_STAT			0x20000000
 #define MP0_C2PMSG_114_REG			0x3810AC8
 #define MP0_C2PMSG_73_REG			0x3810A24
+#define MP0_C2PMSG_63_REG			0x38109FC
+#define MP0_C2PMSG_63_FW_SIGN_ALT		BIT(6)
 #define MBOX_ACP_SHA_DMA_COMMAND		0x70000
 #define MBOX_ACP_IRAM_DRAM_FENCE_COMMAND	0x80000
 #define MBOX_DELAY_US				1000

---
base-commit: 68e9934bd17af558b62cac8a444de28d4d5cdb9f
change-id: 20261104-superm1-lms-sound-c0f429351fc7

Best regards,
--  
Mario Limonciello <mario.limonciello@amd.com>

Re: [PATCH] ASoC: SOF: amd: Load alternatively signed firmware with an "_l" suffix

[Péter Ujfalusi]

On 04/11/2026 23:53, Mario Limonciello wrote:
> Platforms ship BIOS with either a default or an alternative trust
> authority. The PSP validates the signed DSP firmware image so it must
> be signed with the matching scheme. Firmware for the alternative scheme
> is named with an "_l" suffix before the extension
> (e.g. "sof-X-code.bin" becomes "sof-X-code_l.bin").
> 
> The PSP reports the signing type in bit 6 of C2PMSG_63
> (0 = default, 1 = alternative). Read it over SMN, as is already done
> for the other PSP mailbox registers, when building the signed code
> image name, and add the "_l" suffix for the alternative scheme.
> 
> Reviewed-by: Vijendar Mukunda <Vijendar.Mukunda@amd.com>
> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
> ---
>  sound/soc/sof/amd/acp.c | 33 ++++++++++++++++++++++++++++++++-
>  sound/soc/sof/amd/acp.h |  2 ++
>  2 files changed, 34 insertions(+), 1 deletion(-)
> 
> diff --git a/sound/soc/sof/amd/acp.c b/sound/soc/sof/amd/acp.c
> index 86e2402b8ffd2..2f1a8be8460d4 100644
> --- a/sound/soc/sof/amd/acp.c
> +++ b/sound/soc/sof/amd/acp.c
> @@ -1212,6 +1212,31 @@ static int amd_sof_sdw_exit(struct snd_sof_dev *sdev)
>  }
>  #endif
>  
> +/*
> + * The PSP validates the signed DSP firmware, so it must be signed with the
> + * scheme the platform firmware uses: images for the alternative scheme are
> + * named with an "_l" suffix before the extension, the default ones have none.
> + */
> +static int amd_sof_acp_get_fw_suffix(struct snd_sof_dev *sdev, const char **suffix)
> +{
> +	u32 val;
> +	int ret;
> +
> +	ret = amd_smn_read(0, MP0_C2PMSG_63_REG, &val);
> +	if (ret) {
> +		dev_err(sdev->dev, "failed to read PSP firmware signing type: %d\n", ret);

I would read this in case of error as the signing type is -22 (or
whatever the error code returned is)

> +		return ret;
> +	}
> +
> +	/* all ones means the register isn't accessible */
> +	if (val == 0xffffffff)
> +		return -EIO;

this would also merit an error print?

> +
> +	*suffix = (val & MP0_C2PMSG_63_FW_SIGN_ALT) ? "_l" : "";
> +
> +	return 0;
> +}
> +
>  int amd_sof_acp_probe(struct snd_sof_dev *sdev)
>  {
>  	struct pci_dev *pci = to_pci_dev(sdev->dev);
> @@ -1448,8 +1473,14 @@ int amd_sof_acp7x_probe(struct snd_sof_dev *sdev)
>  	sdev->debug_box.size = BOX_SIZE_1024;
>  
>  	if (adata->acp_sof_signed_firmware_image) {
> +		const char *fw_suffix;
> +
> +		ret = amd_sof_acp_get_fw_suffix(sdev, &fw_suffix);
> +		if (ret)
> +			goto free_ipc_irq;
> +
>  		adata->fw_code_bin = devm_kasprintf(sdev->dev, GFP_KERNEL,
> -						    "sof-%s-code.bin", chip->name);
> +						    "sof-%s-code%s.bin", chip->name, fw_suffix);
>  		if (!adata->fw_code_bin) {
>  			ret = -ENOMEM;
>  			goto free_ipc_irq;
> diff --git a/sound/soc/sof/amd/acp.h b/sound/soc/sof/amd/acp.h
> index b17a36b31ee0d..81eeeabc5cf5c 100644
> --- a/sound/soc/sof/amd/acp.h
> +++ b/sound/soc/sof/amd/acp.h
> @@ -97,6 +97,8 @@
>  #define ACP_EXT_INTR_ERROR_STAT			0x20000000
>  #define MP0_C2PMSG_114_REG			0x3810AC8
>  #define MP0_C2PMSG_73_REG			0x3810A24
> +#define MP0_C2PMSG_63_REG			0x38109FC
> +#define MP0_C2PMSG_63_FW_SIGN_ALT		BIT(6)
>  #define MBOX_ACP_SHA_DMA_COMMAND		0x70000
>  #define MBOX_ACP_IRAM_DRAM_FENCE_COMMAND	0x80000
>  #define MBOX_DELAY_US				1000
> 
> ---
> base-commit: 68e9934bd17af558b62cac8a444de28d4d5cdb9f
> change-id: 20261104-superm1-lms-sound-c0f429351fc7
> 
> Best regards,
> --  
> Mario Limonciello <mario.limonciello@amd.com>

-- 
Péter

Re: [PATCH] ASoC: SOF: amd: Load alternatively signed firmware with an "_l" suffix

[Mark Brown]

[-- Attachment #1: Type: text/plain, Size: 507 bytes --]

On Thu, Nov 05, 2026 at 04:03:32PM +0200, Péter Ujfalusi wrote:
> On 04/11/2026 23:53, Mario Limonciello wrote:

> > +	ret = amd_smn_read(0, MP0_C2PMSG_63_REG, &val);
> > +	if (ret) {
> > +		dev_err(sdev->dev, "failed to read PSP firmware signing type: %d\n", ret);

> > +	/* all ones means the register isn't accessible */
> > +	if (val == 0xffffffff)
> > +		return -EIO;

> this would also merit an error print?

amd_smn_read() will convert this to -ENODEV anyway so that won't be
reached.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH] ASoC: SOF: amd: Load alternatively signed firmware with an "_l" suffix

[Mario Limonciello]

On 11/5/26 08:07, Mark Brown wrote:
> On Thu, Nov 05, 2026 at 04:03:32PM +0200, Péter Ujfalusi wrote:
>> On 04/11/2026 23:53, Mario Limonciello wrote:
> 
>>> +	ret = amd_smn_read(0, MP0_C2PMSG_63_REG, &val);
>>> +	if (ret) {
>>> +		dev_err(sdev->dev, "failed to read PSP firmware signing type: %d\n", ret);
> 
>>> +	/* all ones means the register isn't accessible */
>>> +	if (val == 0xffffffff)
>>> +		return -EIO;
> 
>> this would also merit an error print?
> 
> amd_smn_read() will convert this to -ENODEV anyway so that won't be
> reached.

Oh yeah; that's dead code then, thanks.  I'll respin to drop it.