[PATCH v2 1/6] amdgpu: fix possible resource leakage in kfd_cleanup_nodes()

[Gerry Liu]

> 2025年1月5日 13:22，Shuo Liu <shuox.liu at linux.alibaba.com> 写道：
> 
> Hi Gerry,
> 
> On Sun  5.Jan'25 at 10:45:29 +0800, Jiang Liu wrote:
>> Fix possible resource leakage on error recovery path in function
>> kgd2kfd_device_init().
>> 
>> Signed-off-by: Jiang Liu <gerry at linux.alibaba.com>
>> ---
>> drivers/gpu/drm/amd/amdkfd/kfd_device.c | 9 +++++----
>> 1 file changed, 5 insertions(+), 4 deletions(-)
>> 
>> diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
>> index a29374c86405..fa5054940486 100644
>> --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c
>> +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c
>> @@ -898,15 +898,15 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
>> 		if (kfd->adev->xcp_mgr)
>> 			kfd_setup_interrupt_bitmap(node, i);
>> 
>> +		spin_lock_init(&node->watch_points_lock);
>> +
>> +		kfd->nodes[i] = node;
>> +
>> 		/* Initialize the KFD node */
>> 		if (kfd_init_node(node)) {
>> 			dev_err(kfd_device, "Error initializing KFD node\n");
>> 			goto node_init_error;
>> 		}
>> -
>> -		spin_lock_init(&node->watch_points_lock);
>> -
>> -		kfd->nodes[i] = node;
>> 	}
>> 
>> 	svm_range_set_max_pages(kfd->adev);
>> @@ -921,6 +921,7 @@ bool kgd2kfd_device_init(struct kfd_dev *kfd,
>> 	goto out;
>> 
>> node_init_error:
>> +	i++;
> The err cleanup path can release node itself. So the following
> kfd_cleanup_nodes() may do a double free? 
The code in function kgd2kfd_device_exit() checks flag `kfd->init_complete`.
And if error happens within function kgd2kfd_device_init(), the flag `kfd->init_complete` will be set to false, thus avoid double free.

>> node_alloc_error:
>> 	kfd_cleanup_nodes(kfd, i);
>> 	kfd_doorbell_fini(kfd);
> shuo

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/amd-gfx/attachments/20250105/d0f8f0da/attachment.htm>