[PATCH v2 1/2] drm/amdgpu: add debugfs for reset registers list

[Christian König]

Am 11.02.22 um 14:15 schrieb Somalapuram, Amaranath:
>
> On 2/11/2022 5:22 PM, Christian König wrote:
>> Am 11.02.22 um 12:47 schrieb Somalapuram Amaranath:
>>> List of register to be populated for dump collection during the GPU 
>>> reset.
>>>
>>> Signed-off-by: Somalapuram Amaranath <Amaranath.Somalapuram at amd.com>
>>> ---
>>>   drivers/gpu/drm/amd/amdgpu/amdgpu.h         |  5 ++
>>>   drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 61 
>>> +++++++++++++++++++++
>>>   2 files changed, 66 insertions(+)
>>>
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h 
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
>>> index b85b67a88a3d..b90349b86918 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
>>> @@ -813,6 +813,7 @@ struct amd_powerplay {
>>>     #define AMDGPU_RESET_MAGIC_NUM 64
>>>   #define AMDGPU_MAX_DF_PERFMONS 4
>>> +#define AMDGPU_RESET_DUMP_REGS_MAX     128
>>>   struct amdgpu_device {
>>>       struct device            *dev;
>>>       struct pci_dev            *pdev;
>>> @@ -1097,6 +1098,10 @@ struct amdgpu_device {
>>>         struct amdgpu_reset_control     *reset_cntl;
>>>       uint32_t ip_versions[HW_ID_MAX][HWIP_MAX_INSTANCE];
>>> +
>>> +    /* reset dump register */
>>> +    uint32_t reset_dump_reg_list[AMDGPU_RESET_DUMP_REGS_MAX];
>>
>> Using an xarray or just dynamic allocation with krealloc_array would 
>> probably be better.
>>
>> Regards,
>> Christian.
>>
> I can use krealloc_array,
> Then in the second patch i need to use kmalloc (or hard-code values) 
> for this: uint32_t reg_value[AMDGPU_RESET_DUMP_REGS_MAX];
> We actually tried to avoid  dynamic allocation during reset (second 
> patch).

It's correct that you can't allocate any memory during reset, but that's 
also not necessary since you just need to trace one register at a time 
and that uses the trace ring buffer.

> Also AMDGPU_RESET_DUMP_REGS_MAX getting used in trace function.

Exactly that's what I'm trying to avoid here. Big traces have caused us 
quite a bunch of problems already :)

Regards,
Christian.

>
> Regards,
> S.Amarnath
>>> +    int                n_regs;
>>>   };
>>>     static inline struct amdgpu_device *drm_to_adev(struct 
>>> drm_device *ddev)
>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c 
>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
>>> index 164d6a9e9fbb..fb99f3d657a4 100644
>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
>>> @@ -1609,6 +1609,65 @@ DEFINE_DEBUGFS_ATTRIBUTE(fops_ib_preempt, NULL,
>>>   DEFINE_DEBUGFS_ATTRIBUTE(fops_sclk_set, NULL,
>>>               amdgpu_debugfs_sclk_set, "%llu\n");
>>>   +static ssize_t amdgpu_reset_dump_register_list_read(struct file *f,
>>> +                char __user *buf, size_t size, loff_t *pos)
>>> +{
>>> +    struct amdgpu_device *adev = (struct amdgpu_device 
>>> *)file_inode(f)->i_private;
>>> +    char *reg_offset;
>>> +    int i, r, len;
>>> +
>>> +    reg_offset = kmalloc(2048, GFP_KERNEL);
>>> +    memset(reg_offset,  0, 2048);
>>> +    for (i = 0; i < adev->n_regs; i++)
>>> +        sprintf(reg_offset + strlen(reg_offset), "0x%x ", 
>>> adev->reset_dump_reg_list[i]);
>>> +
>>> +    sprintf(reg_offset + strlen(reg_offset), "\n");
>>> +    len = strlen(reg_offset);
>>> +
>>> +    if (*pos >=  len)
>>> +        return 0;
>>> +
>>> +    r = copy_to_user(buf, reg_offset, len);
>>> +    *pos += len - r;
>>> +    kfree(reg_offset);
>>> +
>>> +    return len - r;
>>> +}
>>> +
>>> +static ssize_t amdgpu_reset_dump_register_list_write(struct file 
>>> *f, const char __user *buf,
>>> +        size_t size, loff_t *pos)
>>> +{
>>> +    struct amdgpu_device *adev = (struct amdgpu_device 
>>> *)file_inode(f)->i_private;
>>> +    char *reg_offset, *reg;
>>> +    int ret, i = 0;
>>> +
>>> +    reg_offset = kmalloc(size, GFP_KERNEL);
>>> +    memset(reg_offset,  0, size);
>>> +    ret = copy_from_user(reg_offset, buf, size);
>>> +
>>> +    if (ret)
>>> +        return -EFAULT;
>>> +
>>> +    while ((reg = strsep(&reg_offset, " ")) != NULL) {
>>> +        ret  = kstrtouint(reg, 16, &adev->reset_dump_reg_list[i]);
>>> +        if (ret)
>>> +            return -EINVAL;
>>> +        i++;
>>> +    }
>>> +
>>> +    adev->n_regs = i;
>>> +    kfree(reg_offset);
>>> +
>>> +    return size;
>>> +}
>>> +
>>> +static const struct file_operations amdgpu_reset_dump_register_list 
>>> = {
>>> +    .owner = THIS_MODULE,
>>> +    .read = amdgpu_reset_dump_register_list_read,
>>> +    .write = amdgpu_reset_dump_register_list_write,
>>> +    .llseek = default_llseek
>>> +};
>>> +
>>>   int amdgpu_debugfs_init(struct amdgpu_device *adev)
>>>   {
>>>       struct dentry *root = adev_to_drm(adev)->primary->debugfs_root;
>>> @@ -1672,6 +1731,8 @@ int amdgpu_debugfs_init(struct amdgpu_device 
>>> *adev)
>>>                   &amdgpu_debugfs_test_ib_fops);
>>>       debugfs_create_file("amdgpu_vm_info", 0444, root, adev,
>>>                   &amdgpu_debugfs_vm_info_fops);
>>> +    debugfs_create_file("amdgpu_reset_dump_register_list", 0644, 
>>> root, adev,
>>> +                &amdgpu_reset_dump_register_list);
>>>         adev->debugfs_vbios_blob.data = adev->bios;
>>>       adev->debugfs_vbios_blob.size = adev->bios_size;
>>