[PATCH] Revert "drm/amd/amdgpu: Fix potential ioremap() memory leaks in amdgpu_device_init()"

Ma, Jun majun at amd.com
Tue Mar 19 08:48:50 UTC 2024


Thanks, I will fix it when push to the next branch

Regards,
Ma Jun

On 3/19/2024 4:22 PM, Christian König wrote:
> Am 19.03.24 um 09:07 schrieb Ma Jun:
>> This patch causes the following iounmap erorr and calltrace
>> iounmap: bad address 00000000d0b3631f
> 
>> So just revert it and amdgpu_device_fini_sw() will cleanup the
>> rmmio mapping.
> 
> I think we can improve the wording here. Something like this:
> 
> The original patch was unjustified because amdgpu_device_fini_sw() will 
> always cleanup the
> rmmio mapping.
> 
>>
>> This reverts commit 923f7a82d2e12a99744a940954f3829ab18a9dc7.
> 
> You Signed-off-by tag is missing.
> 
> With that fixed the patch is Reviewed-by: Christian König 
> <christian.koenig at amd.com>
> 
> Regards,
> Christian.
> 
>> ---
>>   drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 16 ++++++----------
>>   1 file changed, 6 insertions(+), 10 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> index 70261eb9b0bb..3204b8f6edeb 100644
>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> @@ -4042,10 +4042,8 @@ int amdgpu_device_init(struct amdgpu_device *adev,
>>   	 * early on during init and before calling to RREG32.
>>   	 */
>>   	adev->reset_domain = amdgpu_reset_create_reset_domain(SINGLE_DEVICE, "amdgpu-reset-dev");
>> -	if (!adev->reset_domain) {
>> -		r = -ENOMEM;
>> -		goto unmap_memory;
>> -	}
>> +	if (!adev->reset_domain)
>> +		return -ENOMEM;
>>   
>>   	/* detect hw virtualization here */
>>   	amdgpu_detect_virtualization(adev);
>> @@ -4055,7 +4053,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
>>   	r = amdgpu_device_get_job_timeout_settings(adev);
>>   	if (r) {
>>   		dev_err(adev->dev, "invalid lockup_timeout parameter syntax\n");
>> -		goto unmap_memory;
>> +		return r;
>>   	}
>>   
>>   	amdgpu_device_set_mcbp(adev);
>> @@ -4063,12 +4061,12 @@ int amdgpu_device_init(struct amdgpu_device *adev,
>>   	/* early init functions */
>>   	r = amdgpu_device_ip_early_init(adev);
>>   	if (r)
>> -		goto unmap_memory;
>> +		return r;
>>   
>>   	/* Get rid of things like offb */
>>   	r = drm_aperture_remove_conflicting_pci_framebuffers(adev->pdev, &amdgpu_kms_driver);
>>   	if (r)
>> -		goto unmap_memory;
>> +		return r;
>>   
>>   	/* Enable TMZ based on IP_VERSION */
>>   	amdgpu_gmc_tmz_set(adev);
>> @@ -4078,7 +4076,7 @@ int amdgpu_device_init(struct amdgpu_device *adev,
>>   	if (adev->gmc.xgmi.supported) {
>>   		r = adev->gfxhub.funcs->get_xgmi_info(adev);
>>   		if (r)
>> -			goto unmap_memory;
>> +			return r;
>>   	}
>>   
>>   	/* enable PCIE atomic ops */
>> @@ -4347,8 +4345,6 @@ int amdgpu_device_init(struct amdgpu_device *adev,
>>   failed:
>>   	amdgpu_vf_error_trans_all(adev);
>>   
>> -unmap_memory:
>> -	iounmap(adev->rmmio);
>>   	return r;
>>   }
>>   
> 


More information about the amd-gfx mailing list