From: Mario Limonciello <mario.limonciello@amd.com>
To: "Zhang, Yidong (David)" <yidong.zhang@amd.com>,
Lizhi Hou <lizhi.hou@amd.com>
Cc: dri-devel@lists.freedesktop.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] accel/amdxdna: Load the aie4 firmware variant the PSP reports
Date: Thu, 5 Nov 2026 13:38:33 -0600 [thread overview]
Message-ID: <0161e0ed-3b6a-496f-8502-6dcd219f3aca@amd.com> (raw)
In-Reply-To: <2cf0906c-9766-4e3c-bfd0-3390263f5390@amd.com>
On 11/5/26 13:19, Zhang, Yidong (David) wrote:
>
> On 11/5/2026 7:51 AM, Mario Limonciello wrote:
>> Some Platforms ship BIOS with either a default or an alternative trust
>> authority. The PSP validates the NPU firmware, so it must be signed with
>> the matching scheme. Firmware for the alternative scheme is named with
>> an "_l" suffix before the extension. A preproduction part needs firmware
>> of its own, named with a "_p" suffix.
>>
>> The PSP reports the signing type in bit 6 of C2PMSG_63, and in bit 28
>> whether this is a preproduction part. A preproduction part that also
>> reports the alternative scheme is not expected, so warn about it and
>> use the default firmware.
>>
>> Add MODULE_FIRMWARE() entries for the "_l" variants of the npu and cert
>> images of the aie4 devices so that initramfs generators include them.
>>
>> Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
>> ---
>> Changes in v2:
>> - Initialize val in aie4_psp_fw_variant() so a failed SMN read can't
>> leave
>> it uninitialized
>> - Only read the PSP firmware variant after the npufw/certfw path
>> check, so
>> VFs (which have no firmware paths) skip the SMN access
>> - Drop the redundant all-ones check, amd_smn_read() already returns -
>> ENODEV
>> - Link to v1: https://patch.msgid.link/20261104-superm1-lms-npu-v1-1-
>> dbffcc5abc23@amd.com
>> ---
>> drivers/accel/amdxdna/Kconfig | 1 +
>> drivers/accel/amdxdna/aie4_pci.c | 162 ++++++++++++++++++++++
>> ++++------
>> drivers/accel/amdxdna/amdxdna_pci_drv.c | 8 ++
>> 3 files changed, 141 insertions(+), 30 deletions(-)
>>
>> diff --git a/drivers/accel/amdxdna/Kconfig b/drivers/accel/amdxdna/
>> Kconfig
>> index eddff490b4f75..cd17a89cf4146 100644
>> --- a/drivers/accel/amdxdna/Kconfig
>> +++ b/drivers/accel/amdxdna/Kconfig
>> @@ -6,6 +6,7 @@ config DRM_ACCEL_AMDXDNA
>> depends on DRM_ACCEL
>> depends on PCI && HAS_IOMEM
>> depends on X86_64
>> + depends on AMD_NODE
>> depends on AMD_PMF || !AMD_PMF
>> select DRM_SCHED
>> select DRM_GEM_SHMEM_HELPER
>> diff --git a/drivers/accel/amdxdna/aie4_pci.c b/drivers/accel/amdxdna/
>> aie4_pci.c
>> index 02dcb2f813d45..26c6162893760 100644
>> --- a/drivers/accel/amdxdna/aie4_pci.c
>> +++ b/drivers/accel/amdxdna/aie4_pci.c
>> @@ -11,6 +11,8 @@
>> #include <linux/debugfs.h>
>> #include <linux/firmware.h>
>> #include <linux/sizes.h>
>> +#include <linux/string.h>
>> +#include <asm/amd/node.h>
>> #include "aie.h"
>> #include "aie4_msg_priv.h"
>> @@ -558,46 +560,124 @@ static void aie4_classic_hw_stop(struct
>> amdxdna_dev_hdl *ndev)
>> aie4_fw_stop(ndev);
>> }
>> -static int aie4_request_firmware(struct amdxdna_dev_hdl *ndev,
>> - const struct firmware **npufw,
>> - const struct firmware **certfw)
>> +/*
>> + * PSP C2PMSG_63 (SMN) bit 6: 0 = default firmware signing, 1 =
>> alternative
>> + * PSP C2PMSG_63 (SMN) bit 28: 0 = production part, 1 = preproduction
>> part
>> + */
>> +#define AIE4_PSP_C2PMSG_63_SMN_ADDR 0x38109fc
>> +#define AIE4_PSP_FW_SIGN_ALT BIT(6)
>> +#define AIE4_PSP_FW_PREPROD BIT(28)
>> +
>> +enum aie4_fw_variant {
>> + AIE4_FW_DEFAULT,
>> + AIE4_FW_ALT, /* "_l" images */
>> + AIE4_FW_PREPROD, /* "_p" images */
>> +};
>> +
>> +/*
>> + * The PSP only trusts firmware signed with the scheme the platform
>> firmware
>> + * uses and a preproduction part needs firmware of its own. Reading
>> which one is
>> + * in use is optional and only used as a hint for which images to try
>> first:
>> + * anything but a clear answer is treated as default.
>> + */
>> +static enum aie4_fw_variant aie4_psp_fw_variant(struct amdxdna_dev
>> *xdna)
>> {
>> - struct amdxdna_dev *xdna = ndev->aie.xdna;
>> - struct pci_dev *pdev = to_pci_dev(xdna->ddev.dev);
>> - char fw_name[128];
>> + u32 val = 0;
>> int ret;
>> - ret = snprintf(fw_name, sizeof(fw_name), "amdnpu/%04x_%02x/%s",
>> - pdev->device, pdev->revision, ndev->priv->npufw_path);
>> - if (ret >= sizeof(fw_name)) {
>> - XDNA_ERR(xdna, "npu firmware path is truncated");
>> - return -EINVAL;
>> + ret = amd_smn_read(0, AIE4_PSP_C2PMSG_63_SMN_ADDR, &val);
>> + if (ret) {
>> + XDNA_DBG(xdna, "PSP firmware variant not available, ret %d",
>> ret);
>> + return AIE4_FW_DEFAULT;
>> }
>> - ret = request_firmware(npufw, fw_name, &pdev->dev);
>> - if (ret) {
>> - XDNA_ERR(xdna, "failed to request_firmware %s, ret %d",
>> fw_name, ret);
>> - return ret;
>> + if ((val & AIE4_PSP_FW_SIGN_ALT) && (val & AIE4_PSP_FW_PREPROD)) {
>> + XDNA_WARN(xdna,
>> + "Preproduction part with alternative signing, using
>> default firmware");
>> + return AIE4_FW_DEFAULT;
>> }
>> - ret = snprintf(fw_name, sizeof(fw_name), "amdnpu/%04x_%02x/%s",
>> - pdev->device, pdev->revision, ndev->priv->certfw_path);
>> - if (ret >= sizeof(fw_name)) {
>> - XDNA_ERR(xdna, "cert firmware path is truncated");
>> - ret = -EINVAL;
>> - goto release_npufw;
>> + if (val & AIE4_PSP_FW_PREPROD)
>> + return AIE4_FW_PREPROD;
>> +
>> + return (val & AIE4_PSP_FW_SIGN_ALT) ? AIE4_FW_ALT : AIE4_FW_DEFAULT;
>> +}
>> +
>> +/*
>> + * Images for the other variants have a suffix before the ".sbin"
>> extension.
>> + * Returns -ENOENT if @path has no such variant.
>> + */
>> +static int aie4_fw_name(struct amdxdna_dev *xdna, char *name, size_t
>> size,
>> + const char *path, enum aie4_fw_variant variant)
>> +{
>> + struct pci_dev *pdev = to_pci_dev(xdna->ddev.dev);
>> + const char *suffix;
>> + size_t len, base;
>> + int ret;
>> +
>> + ret = snprintf(name, size, "amdnpu/%04x_%02x/%s",
>> + pdev->device, pdev->revision, path);
>> + if (ret >= size) {
>> + XDNA_ERR(xdna, "firmware path %s is truncated", path);
>> + return -EINVAL;
>> }
>> - ret = request_firmware(certfw, fw_name, &pdev->dev);
>> - if (ret) {
>> - XDNA_ERR(xdna, "failed to request_firmware %s, ret %d",
>> fw_name, ret);
>> - goto release_npufw;
>> + switch (variant) {
>> + case AIE4_FW_ALT:
>> + suffix = "_l.sbin";
>> + break;
>> + case AIE4_FW_PREPROD:
>> + suffix = "_p.sbin";
>> + break;
>> + default:
>> + return 0;
>> }
>> + len = ret;
>> + if (len < strlen(".sbin") || strcmp(name + len - strlen(".sbin"),
>> ".sbin"))
>> + return -ENOENT;
>> +
>> + base = len - strlen(".sbin");
>> + if (strscpy(name + base, suffix, size - base) < 0)
>> + return -EOVERFLOW;
>> +
>> return 0;
>> +}
>> +
>> +static int aie4_request_one(struct amdxdna_dev_hdl *ndev, const
>> struct firmware **fw,
>> + const char *path, enum aie4_fw_variant variant)
>> +{
>> + struct amdxdna_dev *xdna = ndev->aie.xdna;
>> + struct pci_dev *pdev = to_pci_dev(xdna->ddev.dev);
>> + char name[128];
>> + int ret;
>> +
>> + ret = aie4_fw_name(xdna, name, sizeof(name), path, variant);
>> + if (ret)
>> + return ret;
>> -release_npufw:
>> - release_firmware(*npufw);
>> + ret = firmware_request_nowarn(fw, name, &pdev->dev);
>> + if (ret)
>> + XDNA_DBG(xdna, "firmware %s not loaded, ret %d", name, ret);
>> +
>> + return ret;
>> +}
>> +
>> +/* The NPU and cert images are validated together, so both use the
>> same variant */
>> +static int aie4_request_firmware(struct amdxdna_dev_hdl *ndev,
>> + const struct firmware **npufw,
>> + const struct firmware **certfw,
>> + enum aie4_fw_variant variant)
>> +{
>> + int ret;
>> +
>> + ret = aie4_request_one(ndev, npufw, ndev->priv->npufw_path,
>> variant);
>> + if (ret)
>> + return ret;
>> +
>> + ret = aie4_request_one(ndev, certfw, ndev->priv->certfw_path,
>> variant);
>> + if (ret)
>> + release_firmware(*npufw);
>> return ret;
>> }
>> @@ -645,18 +725,40 @@ static int aie4_prepare_firmware(struct
>> amdxdna_dev_hdl *ndev,
>> return 0;
>> }
>> +/*
>> + * The firmware can only be started once. It is the PSP that decides
>> whether it
>> + * accepts the images, and it does that when the firmware is started,
>> so there
>> + * is no trying a variant and falling back. Pick the variant that the
>> PSP
>> + * reported (the default one if it can't be read) and use the others
>> only if
>> + * its images are missing.
>> + */
>> static int aie4_load_fw(struct amdxdna_dev_hdl *ndev,
>> void __iomem *tbl[PCI_NUM_RESOURCES])
>> {
>> + struct amdxdna_dev *xdna = ndev->aie.xdna;
>> const struct firmware *npufw, *certfw;
>> - int ret;
>> + enum aie4_fw_variant order[3], hint;
>> + int i, n = 0, ret = -ENOENT;
>> if (!ndev->priv->npufw_path && !ndev->priv->certfw_path)
>> return 0;
>> - ret = aie4_request_firmware(ndev, &npufw, &certfw);
>> - if (ret)
>> + hint = aie4_psp_fw_variant(xdna);
>> + order[n++] = hint;
>> + if (hint != AIE4_FW_DEFAULT)
>> + order[n++] = AIE4_FW_DEFAULT;
>> + if (hint != AIE4_FW_ALT)
>> + order[n++] = AIE4_FW_ALT;
>> +
>> + for (i = 0; i < n; i++) {
>> + ret = aie4_request_firmware(ndev, &npufw, &certfw, order[i]);
>> + if (!ret)
>> + break;
>> + }
>> + if (ret) {
>> + XDNA_ERR(xdna, "failed to load firmware, ret %d", ret);
>> return ret;
>> + }
>> ret = aie4_prepare_firmware(ndev, npufw, certfw, tbl);
>> aie4_release_firmware(ndev, npufw, certfw);
>> diff --git a/drivers/accel/amdxdna/amdxdna_pci_drv.c b/drivers/accel/
>> amdxdna/amdxdna_pci_drv.c
>> index f233eb70a2957..dab8603e2381f 100644
>> --- a/drivers/accel/amdxdna/amdxdna_pci_drv.c
>> +++ b/drivers/accel/amdxdna/amdxdna_pci_drv.c
>> @@ -37,6 +37,14 @@ MODULE_FIRMWARE("amdnpu/17f1_13/npu.sbin");
>> MODULE_FIRMWARE("amdnpu/17f1_13/cert.sbin");
>> MODULE_FIRMWARE("amdnpu/17f2_13/npu.sbin");
>> MODULE_FIRMWARE("amdnpu/17f2_13/cert.sbin");
>> +MODULE_FIRMWARE("amdnpu/17f1_10/npu_l.sbin");
>> +MODULE_FIRMWARE("amdnpu/17f1_10/cert_l.sbin");
>> +MODULE_FIRMWARE("amdnpu/17f2_10/npu_l.sbin");
>> +MODULE_FIRMWARE("amdnpu/17f2_10/cert_l.sbin");
>> +MODULE_FIRMWARE("amdnpu/17f1_13/npu_l.sbin");
>> +MODULE_FIRMWARE("amdnpu/17f1_13/cert_l.sbin");
>> +MODULE_FIRMWARE("amdnpu/17f2_13/npu_l.sbin");
>> +MODULE_FIRMWARE("amdnpu/17f2_13/cert_l.sbin");
>
>
> I tested against both _l, _p and default sbins. All passed.
>
> Reviewed-by: David Zhang <yidong.zhang@amd.com>
>
> Tested-by: David Zhang <yidong.zhang@amd.com>
>
> /David
>
>> /*
>> * 0.0: Initial version
>>
>> ---
>> base-commit: e32e83a6c5ec031a7f59d4794113d8484b220254
>> change-id: 20261104-superm1-lms-npu-6bc3bbe38c29
>>
>> Best regards,
>> --
>> Mario Limonciello <mario.limonciello@amd.com>
applied to drm-misc-next
prev parent reply other threads:[~2026-11-05 19:39 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-05 15:51 Mario Limonciello
2026-11-05 17:47 ` Lizhi Hou
2026-11-05 17:52 ` Mario Limonciello
2026-11-05 17:56 ` Lizhi Hou
2026-11-05 19:19 ` Zhang, Yidong (David)
2026-11-05 19:38 ` Mario Limonciello [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=0161e0ed-3b6a-496f-8502-6dcd219f3aca@amd.com \
--to=mario.limonciello@amd.com \
--cc=dri-devel@lists.freedesktop.org \
--cc=linux-kernel@vger.kernel.org \
--cc=lizhi.hou@amd.com \
--cc=yidong.zhang@amd.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®