Re: [PATCH] perf: fix event leak when perf_event_open() failed to create event_file

[Leon Yu]

On Mon, Mar 21, 2016 at 4:02 PM, Alexander Shishkin
<alexander.shishkin@xxxxxxxxxxxxxxx> wrote:
> Leon Yu <chianglungyu@xxxxxxxxx> writes:
>
>> If something went wrong in anon_inode_getfile, event_file will be set to
>> non-zero error number and able to bypass the NULL test afterward.
>>
>> Consolidate the error path by testing event_file with handly
>> IS_ERR_OR_NULL() helper since we do want to free event in both cases.
>>
>> Signed-off-by: Leon Yu <chianglungyu@xxxxxxxxx>
>> Fixes: 130056275ade ("perf: Do not double free")
>> Cc: <stable@xxxxxxxxxxxxxxx> # v4.5
>> ---
>>  kernel/events/core.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/kernel/events/core.c b/kernel/events/core.c
>> index 6146148..5f2e19f 100644
>> --- a/kernel/events/core.c
>> +++ b/kernel/events/core.c
>> @@ -8617,7 +8617,7 @@ err_alloc:
>>        * If event_file is set, the fput() above will have called ->release()
>>        * and that will take care of freeing the event.
>>        */
>> -     if (!event_file)
>> +     if (IS_ERR_OR_NULL(event_file))
>>               free_event(event);
>
> By this time, we have already checked for IS_ERR(event_file) once, why
> not just fix it up there like so:

My patch is more straightforward, isn't it? :p
Anyway I don't mind which one will get merged as long as the problem is fixed.

(sorry forgot to CC all recipients)

-Leon

> @@ -9414,6 +9414,7 @@ SYSCALL_DEFINE5(perf_event_open,
>                                         f_flags);
>         if (IS_ERR(event_file)) {
>                 err = PTR_ERR(event_file);
> +               event_file = NULL;
>                 goto err_context;
>         }
>
> --
> 2.7.0
>