Re: [PATCH V1 2/2] accel/amdxdna: Update firmware version check for latest firmware

[Lizhi Hou <lizhi.hou@amd.com>]

On 1/7/26 13:20, Mario Limonciello wrote:
> On 12/18/25 7:43 PM, Lizhi Hou wrote:
>> The latest firmware increases the major version number. Update
>> aie2_check_protocol() to accept and support the new firmware version.
>>
>> Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
>
> I know it's painful and tech debt; but I suggest you add new paths to 
> handle both versions of firmware at least until the next LTS kernel.

With the patch 1, the driver works fine for both current firmware and 
upcoming firmware.

Lizhi

>
>> ---
>>   drivers/accel/amdxdna/aie2_pci.c  | 36 ++++++++-----------------------
>>   drivers/accel/amdxdna/aie2_pci.h  |  5 ++---
>>   drivers/accel/amdxdna/npu1_regs.c |  6 +++---
>>   drivers/accel/amdxdna/npu4_regs.c | 11 +++++-----
>>   drivers/accel/amdxdna/npu5_regs.c |  2 --
>>   drivers/accel/amdxdna/npu6_regs.c |  2 --
>>   6 files changed, 20 insertions(+), 42 deletions(-)
>>
>> diff --git a/drivers/accel/amdxdna/aie2_pci.c 
>> b/drivers/accel/amdxdna/aie2_pci.c
>> index 81a8e4137bfd..181fdbc10dae 100644
>> --- a/drivers/accel/amdxdna/aie2_pci.c
>> +++ b/drivers/accel/amdxdna/aie2_pci.c
>> @@ -56,41 +56,23 @@ struct mgmt_mbox_chann_info {
>>   static int aie2_check_protocol(struct amdxdna_dev_hdl *ndev, u32 
>> fw_major, u32 fw_minor)
>>   {
>>       const struct aie2_fw_feature_tbl *feature;
>> -    struct amdxdna_dev *xdna = ndev->xdna;
>> -
>> -    /*
>> -     * The driver supported mailbox behavior is defined by
>> -     * ndev->priv->protocol_major and protocol_minor.
>> -     *
>> -     * When protocol_major and fw_major are different, it means driver
>> -     * and firmware are incompatible.
>> -     */
>> -    if (ndev->priv->protocol_major != fw_major) {
>> -        XDNA_ERR(xdna, "Incompatible firmware protocol major %d 
>> minor %d",
>> -             fw_major, fw_minor);
>> -        return -EINVAL;
>> -    }
>> +    bool found = false;
>>   -    /*
>> -     * When protocol_minor is greater then fw_minor, that means driver
>> -     * relies on operation the installed firmware does not support.
>> -     */
>> -    if (ndev->priv->protocol_minor > fw_minor) {
>> -        XDNA_ERR(xdna, "Firmware minor version smaller than 
>> supported");
>> -        return -EINVAL;
>> -    }
>> -
>> -    for (feature = ndev->priv->fw_feature_tbl; feature && 
>> feature->min_minor;
>> -         feature++) {
>> +    for (feature = ndev->priv->fw_feature_tbl; feature->major; 
>> feature++) {
>> +        if (feature->major != fw_major)
>> +            continue;
>>           if (fw_minor < feature->min_minor)
>>               continue;
>>           if (feature->max_minor > 0 && fw_minor > feature->max_minor)
>>               continue;
>>   -        set_bit(feature->feature, &ndev->feature_mask);
>> +        ndev->feature_mask |= feature->features;
>> +
>> +        /* firmware version matches one of the driver support entry */
>> +        found = true;
>>       }
>>   -    return 0;
>> +    return found ? 0 : -EOPNOTSUPP;
>>   }
>>     static void aie2_dump_chann_info_debug(struct amdxdna_dev_hdl *ndev)
>> diff --git a/drivers/accel/amdxdna/aie2_pci.h 
>> b/drivers/accel/amdxdna/aie2_pci.h
>> index e1745f07b268..b20a3661078c 100644
>> --- a/drivers/accel/amdxdna/aie2_pci.h
>> +++ b/drivers/accel/amdxdna/aie2_pci.h
>> @@ -237,7 +237,8 @@ enum aie2_fw_feature {
>>   };
>>     struct aie2_fw_feature_tbl {
>> -    enum aie2_fw_feature feature;
>> +    u64 features;
>> +    u32 major;
>>       u32 max_minor;
>>       u32 min_minor;
>>   };
>> @@ -246,8 +247,6 @@ struct aie2_fw_feature_tbl {
>>     struct amdxdna_dev_priv {
>>       const char            *fw_path;
>> -    u64                protocol_major;
>> -    u64                protocol_minor;
>>       const struct rt_config        *rt_config;
>>       const struct dpm_clk_freq    *dpm_clk_tbl;
>>       const struct aie2_fw_feature_tbl *fw_feature_tbl;
>> diff --git a/drivers/accel/amdxdna/npu1_regs.c 
>> b/drivers/accel/amdxdna/npu1_regs.c
>> index ebc6e2802297..6f36a27b5a02 100644
>> --- a/drivers/accel/amdxdna/npu1_regs.c
>> +++ b/drivers/accel/amdxdna/npu1_regs.c
>> @@ -6,6 +6,7 @@
>>   #include <drm/amdxdna_accel.h>
>>   #include <drm/drm_device.h>
>>   #include <drm/gpu_scheduler.h>
>> +#include <linux/bits.h>
>>   #include <linux/sizes.h>
>>     #include "aie2_pci.h"
>> @@ -65,14 +66,13 @@ const struct dpm_clk_freq npu1_dpm_clk_table[] = {
>>   };
>>     static const struct aie2_fw_feature_tbl npu1_fw_feature_table[] = {
>> -    { .feature = AIE2_NPU_COMMAND, .min_minor = 8 },
>> +    { .major = 5, .min_minor = 7 },
>> +    { .features = BIT_U64(AIE2_NPU_COMMAND), .min_minor = 8 },
>>       { 0 }
>>   };
>>     static const struct amdxdna_dev_priv npu1_dev_priv = {
>>       .fw_path        = "amdnpu/1502_00/npu.sbin",
>> -    .protocol_major = 0x5,
>> -    .protocol_minor = 0x7,
>>       .rt_config    = npu1_default_rt_cfg,
>>       .dpm_clk_tbl    = npu1_dpm_clk_table,
>>       .fw_feature_tbl = npu1_fw_feature_table,
>> diff --git a/drivers/accel/amdxdna/npu4_regs.c 
>> b/drivers/accel/amdxdna/npu4_regs.c
>> index a62234fd266d..a8d6f76dde5f 100644
>> --- a/drivers/accel/amdxdna/npu4_regs.c
>> +++ b/drivers/accel/amdxdna/npu4_regs.c
>> @@ -6,6 +6,7 @@
>>   #include <drm/amdxdna_accel.h>
>>   #include <drm/drm_device.h>
>>   #include <drm/gpu_scheduler.h>
>> +#include <linux/bits.h>
>>   #include <linux/sizes.h>
>>     #include "aie2_pci.h"
>> @@ -88,16 +89,16 @@ const struct dpm_clk_freq npu4_dpm_clk_table[] = {
>>   };
>>     const struct aie2_fw_feature_tbl npu4_fw_feature_table[] = {
>> -    { .feature = AIE2_NPU_COMMAND, .min_minor = 15 },
>> -    { .feature = AIE2_PREEMPT, .min_minor = 12 },
>> -    { .feature = AIE2_TEMPORAL_ONLY, .min_minor = 12 },
>> +    { .major = 6, .min_minor = 12 },
>> +    { .features = BIT_U64(AIE2_NPU_COMMAND), .major = 6, .min_minor 
>> = 15 },
>> +    { .features = BIT_U64(AIE2_PREEMPT), .major = 6, .min_minor = 12 },
>> +    { .features = BIT_U64(AIE2_TEMPORAL_ONLY), .major = 6, 
>> .min_minor = 12 },
>> +    { .features = GENMASK_ULL(AIE2_TEMPORAL_ONLY, AIE2_NPU_COMMAND), 
>> .major = 7 },
>>       { 0 }
>>   };
>>     static const struct amdxdna_dev_priv npu4_dev_priv = {
>>       .fw_path        = "amdnpu/17f0_10/npu.sbin",
>> -    .protocol_major = 0x6,
>> -    .protocol_minor = 12,
>>       .rt_config    = npu4_default_rt_cfg,
>>       .dpm_clk_tbl    = npu4_dpm_clk_table,
>>       .fw_feature_tbl = npu4_fw_feature_table,
>> diff --git a/drivers/accel/amdxdna/npu5_regs.c 
>> b/drivers/accel/amdxdna/npu5_regs.c
>> index 131080652ef0..c0a35cfd886c 100644
>> --- a/drivers/accel/amdxdna/npu5_regs.c
>> +++ b/drivers/accel/amdxdna/npu5_regs.c
>> @@ -64,8 +64,6 @@
>>     static const struct amdxdna_dev_priv npu5_dev_priv = {
>>       .fw_path        = "amdnpu/17f0_11/npu.sbin",
>> -    .protocol_major = 0x6,
>> -    .protocol_minor = 12,
>>       .rt_config    = npu4_default_rt_cfg,
>>       .dpm_clk_tbl    = npu4_dpm_clk_table,
>>       .fw_feature_tbl = npu4_fw_feature_table,
>> diff --git a/drivers/accel/amdxdna/npu6_regs.c 
>> b/drivers/accel/amdxdna/npu6_regs.c
>> index 1f71285655b2..1fb07df99186 100644
>> --- a/drivers/accel/amdxdna/npu6_regs.c
>> +++ b/drivers/accel/amdxdna/npu6_regs.c
>> @@ -64,8 +64,6 @@
>>     static const struct amdxdna_dev_priv npu6_dev_priv = {
>>       .fw_path        = "amdnpu/17f0_10/npu.sbin",
>> -    .protocol_major = 0x6,
>> -    .protocol_minor = 12,
>>       .rt_config    = npu4_default_rt_cfg,
>>       .dpm_clk_tbl    = npu4_dpm_clk_table,
>>       .fw_feature_tbl = npu4_fw_feature_table,
>