Re: [PATCH] RDMA/hfi1: fix potential memory leak in setup_base_ctxt()

[Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>]

On 7/18/22 8:30 AM, Leon Romanovsky wrote:
> On Mon, Jul 18, 2022 at 08:11:59AM -0400, Dennis Dalessandro wrote:
>> On 7/18/22 6:39 AM, Leon Romanovsky wrote:
>>> On Mon, Jul 11, 2022 at 07:52:25AM -0400, Dennis Dalessandro wrote:
>>>> On 7/11/22 3:07 AM, Jianglei Nie wrote:
>>>>> setup_base_ctxt() allocates a memory chunk for uctxt->groups with
>>>>> hfi1_alloc_ctxt_rcv_groups(). When init_user_ctxt() fails, uctxt->groups
>>>>> is not released, which will lead to a memory leak.
>>>>>
>>>>> We should release the uctxt->groups with hfi1_free_ctxt_rcv_groups()
>>>>> when init_user_ctxt() fails.
>>>>>
>>>>> Signed-off-by: Jianglei Nie <niejianglei2021@163.com>
>>>>> ---
>>>>>  drivers/infiniband/hw/hfi1/file_ops.c | 4 +++-
>>>>>  1 file changed, 3 insertions(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/drivers/infiniband/hw/hfi1/file_ops.c b/drivers/infiniband/hw/hfi1/file_ops.c
>>>>> index 2e4cf2b11653..629beff053ad 100644
>>>>> --- a/drivers/infiniband/hw/hfi1/file_ops.c
>>>>> +++ b/drivers/infiniband/hw/hfi1/file_ops.c
>>>>> @@ -1179,8 +1179,10 @@ static int setup_base_ctxt(struct hfi1_filedata *fd,
>>>>>  		goto done;
>>>>>  
>>>>>  	ret = init_user_ctxt(fd, uctxt);
>>>>> -	if (ret)
>>>>> +	if (ret) {
>>>>> +		hfi1_free_ctxt_rcv_groups(uctxt);
>>>>>  		goto done;
>>>>> +	}
>>>>>  
>>>>>  	user_init(uctxt);
>>>>>  
>>>>
>>>> Doesn't seem like this patch is correct. The free is done when the file is
>>>> closed, along with other clean up stuff. See hfi1_file_close().
>>>
>>> Can setup_base_ctxt() be called twice for same uctxt?
>>> You are allocating rcd->groups and not releasing.
>>
>> The first thing assign_ctxt() does is a check of the fd->uctxt and it bails with
>> -EINVAL. So effectively only once.
> 
> I'm slightly confused. How will you release rcd->groups?
> 
> assign_ctxt()
>  -> setup_base_ctxt()
>    -> hfi1_alloc_ctxt_rcv_groups()
>       ,,,
>       rcd->groups = kzalloc...
>       ...
>    -> init_user_ctxt() <-- fails and leaves fd->uctx == NULL
> 
> 
> ...
> hfi1_file_close()
>   struct hfi1_ctxtdata *uctxt = fdata->uctxt;
>   ...
>   if (!uctxt)             <-- This is our case
>      goto done; 
>   ...
> 
> done:
>   if (refcount_dec_and_test(&dd->user_refcount))
>      complete(&dd->user_comp);
> 
>   cleanup_srcu_struct(&fdata->pq_srcu);
>   kfree(fdata);
>   return 0;
> 

Looks like this may have been broken with:

e87473bc1b6c ("IB/hfi1: Only set fd pointer when base context is completely
initialized")

The question is does it make more sense to just move the fd->uctxt assignment
up, or call the free directly. I think that might be opening a bigger can of
worms though, as this was part of a larger patch set. Maybe it is best after all
to go with this patch.

Let's add the above as a fixes line and tack on:

Acked-by: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>

It's been like this since 4.14, so no rush to get it in for the ultra late RC.
I'll get it tested as part of the next cycle.

-Denny