[PATCH v9 1/2] drm/amdgpu: add debugfs for reset registers list

Somalapuram, Amaranath Amaranath.Somalapuram at amd.com
Mon Feb 21 14:19:49 UTC 2022


[AMD Official Use Only]


On 2/21/2022 7:09 PM, Christian König wrote:
>
>
> Am 21.02.22 um 14:34 schrieb Somalapuram Amaranath:
>> List of register 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         |  4 +
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c | 96 +++++++++++++++++++++
>>   2 files changed, 100 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h 
>> b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
>> index b85b67a88a3d..6e35f2c4c869 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
>> @@ -1097,6 +1097,10 @@ 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;
>> +    int                             num_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..69c0a28deeac 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c
>> @@ -1609,6 +1609,100 @@ 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[11];
>> +    int i, ret, len = 0;
>> +
>> +    if (*pos)
>> +        return 0;
>> +
>> +    ret = down_read_killable(&adev->reset_sem);
>> +
>> +    if (ret)
>> +        return ret;
>
> We usually don't have an empty line between function call and checking 
> the return code.
>
>> +
>> +    for (i = 0; i < adev->num_regs; i++) {
>> +        down_read(&adev->reset_sem);
>
> That here will just crash because we have already locked the semaphore 
> before the loop.
>
unfortunately it did not crash. Sorry I misunderstood your earlier comments.
>> +        sprintf(reg_offset, "0x%x ", adev->reset_dump_reg_list[i]);
>> +        up_read(&adev->reset_sem);
>> +        ret = copy_to_user(buf + len, reg_offset, strlen(reg_offset));
>> +
>> +        if (ret)
>> +            goto error;
>> +
>> +        len += strlen(reg_offset);
>
> And here the down_read_killable() is missing.
>
>> +    }
>> +
>> +    up_read(&adev->reset_sem);
>> +    ret = copy_to_user(buf + len, "\n", 1);
>> +
>> +    if (ret)
>> +        return -EFAULT;
>> +
>> +    len++;
>> +    *pos += len;
>> +
>> +    return len;
>> +error:
>> +    up_read(&adev->reset_sem);
>> +    return -EFAULT;
>> +}
>> +
>> +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, reg_temp[11];
>> +    uint32_t *tmp;
>> +    int ret, i = 0, len = 0;
>> +
>> +    do {
>> +        reg_offset = reg_temp;
>
> I think you can just drop the reg_offset variable;
>
strsep takes only pointer as input, this is workaround.
>> +        memset(reg_offset, 0, 11);
>> +        ret = copy_from_user(reg_offset, buf + len, min(11, 
>> ((int)size-len)));
>> +
>> +        if (ret)
>> +            goto error_free;
>> +
>> +        reg = strsep(&reg_offset, " ");
>> +        tmp = krealloc_array(tmp, 1, sizeof(uint32_t), GFP_KERNEL);
>
> That must be krealloc_array(tmp, i, ... not krealloc_array(tmp, 1, ... !
I thought it will append (if not it should have crashed or some kernel dump)
> Regards,
> Christian.
>
>> +        ret = kstrtouint(reg, 16, &tmp[i]);
>> +
>> +        if (ret)
>> +            goto error_free;
>> +
>> +        len += strlen(reg) + 1;
>> +        i++;
>> +
>> +    } while (len < size);
>> +
>> +    ret = down_write_killable(&adev->reset_sem);
>> +
>> +    if (ret)
>> +        goto error_free;
>> +
>> +    swap(adev->reset_dump_reg_list, tmp);
>> +    adev->num_regs = i;
>> +    up_write(&adev->reset_sem);
>> +    ret = size;
>> +
>> +error_free:
>> +    kfree(tmp);
>> +    return ret;
>> +}
>> +
>> +
>> +
>> +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 +1766,8 @@ 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;
>


More information about the amd-gfx mailing list