lkml.org 
[lkml]   [2019]   [Sep]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [PATCH v2] drm/amdgpu: fix multiple memory leaks
Date
Am 19.09.19 um 16:28 schrieb Sven Van Asbroeck:
> Hi Christian,
>
> On Thu, Sep 19, 2019 at 4:05 AM Koenig, Christian
> <Christian.Koenig@amd.com> wrote:
>>> +out4:
>>> + kfree(i2s_pdata);
>>> +out3:
>>> + kfree(adev->acp.acp_res);
>>> +out2:
>>> + kfree(adev->acp.acp_cell);
>>> +out1:
>>> + kfree(adev->acp.acp_genpd);
>> kfree on a NULL pointer is harmless, so a single error label should be
>> sufficient.
> That is true, but I notice that the adev structure comes from outside this
> driver:

adev is a very integral part of the driver and always zero initialized:

adev = kzalloc(sizeof(struct amdgpu_device), GFP_KERNEL);

Regards,
Christian.

>
> static int acp_hw_init(void *handle)
> {
> ...
> struct amdgpu_device *adev = (struct amdgpu_device *)handle;
> ...
> }
>
> As far as I can tell, the init() does not explicitly set these to NULL:
> adev->acp.acp_res
> adev->acp.acp_cell
> adev->acp.acp_genpd
>
> I am assuming that core code sets these to NULL, before calling
> acp_hw_init(). But is that guaranteed or simply a happy accident?
> Ie. if they are NULL today, are they likely to remain NULL tomorrow?
>
> Because calling kfree() on a stale pointer would be, well
> not good. Especially not on an error path, which typically
> does not receive much testing, if any.
>
> My apologies if I have misinterpreted this, I am not familiar with
> this code base.
>
> Sven

\
 
 \ /
  Last update: 2019-09-19 18:50    [W:0.196 / U:1.560 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site