On Mon, May 31, 2021 at 06:35:33PM +0800, Dongliang Mu wrote:
On Mon, May 31, 2021 at 5:38 PM Dan Carpenter dan.carpenter@oracle.com wrote:
On Mon, May 31, 2021 at 05:10:49PM +0800, Dongliang Mu wrote:
Hi Dan,
I wonder if we shall split the current patch into two patches, one patch for each memory leak. It is better to satisfy the rule - "one patch only fixes one issue".
We should absolutely fix all these memory leaks. But one patch for two different bugs with different objects and different paths is not very suitable.
I would just send one patch, because I only see it as one bug. But you do what you think is best.
If you think they are the same bug, that's great. Just send and apply only one patch as it is. I will not send version v2.
Sorry for the miscommunication. That's not what I meant at all.
Your patch only introduces one put_device(). We need all five to fix the bug, and we'll have to change the kfree(link_whatever). Use device_unregister() instead put_device(). Include a Reported-by with the correct syzkaller information.
BTW, could you please show me the syzbot link for the bug you tried to resolve? If it is not from syzbot dashboard, can you please show the bug report?
What are you talking about? I'm not trying to fix a syzkaller bug. I'm just reviewing your patch.
regards, dan carpenter