[dpdk-dev] [dpdk-stable] [PATCH] test/eventdev: fix sprintf with snprintf

[Ferruh Yigit]

On 3/12/2019 2:44 PM, Aaron Conole wrote:
> "Parthasarathy, JananeeX M" <jananeex.m.parthasarathy at intel.com> writes:
> 
>> Hi
>>
>>> -----Original Message-----
>>> From: Parthasarathy, JananeeX M
>>> Sent: Tuesday, February 19, 2019 6:33 PM
>>> To: Aaron Conole <aconole at redhat.com>; Poornima, PallantlaX
>>> <pallantlax.poornima at intel.com>
>>> Cc: dev at dpdk.org; Pattan, Reshma <reshma.pattan at intel.com>; Rao, Nikhil
>>> <nikhil.rao at intel.com>; stable at dpdk.org
>>> Subject: RE: [dpdk-dev] [PATCH] test/eventdev: fix sprintf with snprintf
>>>
>>>
>>>
>>>> -----Original Message-----
>>>> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Aaron Conole
>>>> Sent: Saturday, February 09, 2019 2:50 AM
>>>> To: Poornima, PallantlaX <pallantlax.poornima at intel.com>
>>>> Cc: dev at dpdk.org; Pattan, Reshma <reshma.pattan at intel.com>; Rao, Nikhil
>>>> <nikhil.rao at intel.com>; stable at dpdk.org
>>>> Subject: Re: [dpdk-dev] [PATCH] test/eventdev: fix sprintf with
>>>> snprintf
>>>>
>>>> Pallantla Poornima <pallantlax.poornima at intel.com> writes:
>>>>
>>>>> sprintf function is not secure as it doesn't check the length of string.
>>>>> More secure function snprintf is used.
>>>>>
>>>>> Fixes: 2a9c83ae3b ("test/eventdev: add multi-ports test")
>>>>> Cc: stable at dpdk.org
>>>>>
>>>>> Signed-off-by: Pallantla Poornima <pallantlax.poornima at intel.com>
>>>>> ---
>>>>>  test/test/test_event_eth_rx_adapter.c | 3 ++-
>>>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>>>
>>>>> diff --git a/test/test/test_event_eth_rx_adapter.c
>>>>> b/test/test/test_event_eth_rx_adapter.c
>>>>> index 1d3be82b5..38f5c039f 100644
>>>>> --- a/test/test/test_event_eth_rx_adapter.c
>>>>> +++ b/test/test/test_event_eth_rx_adapter.c
>>>>> @@ -479,7 +479,8 @@ adapter_multi_eth_add_del(void)
>>>>>  	/* add the max port for rx_adapter */
>>>>>  	port_index = rte_eth_dev_count_total();
>>>>>  	for (; port_index < RTE_MAX_ETHPORTS; port_index += 1) {
>>>>> -		sprintf(driver_name, "%s%u", "net_null", drv_id);
>>>>> +		snprintf(driver_name, sizeof(driver_name), "%s%u", "net_null",
>>>>> +				drv_id);
>>>>>  		err = rte_vdev_init(driver_name, NULL);
>>>>>  		TEST_ASSERT(err == 0, "Failed driver %s got %d",
>>>>>  		driver_name, err);
>>>>
>>>> You call this a fix, but it's not possible for the value of drv_id to
>>>> exceed '32' and the buffer size is plenty accommodating for that.  Did
>>>> I miss something?  What is this fixing?
>>>
>>> It is better practice to use snprintf although in this case buffer will not overflow
>>> as size is big enough to accommodate. The changes were done mainly to
>>> replace sprintf to snprintf. Probably we can remove "fix" line as it is not issue in
>>> this scenario.
>>>
>>> Thanks
>>> M.P.Jananee
>>
>> Please suggest if we can remove "fix" line.
> 
> This is a stylistic change, I don't think it's appropriate to call it a
> fix, so I think you can remove the "Fixes" line.
> 
> On further reflection, I actually think it will still be wrong.  If the
> size buffer is ever changed, what will happen on truncation?  We don't
> get an overflow any longer, but we still pass an invalid argument, so I
> don't think this 'fix' is really even a fix.  It still has a bug -
> albeit not one that immediately triggers SSP exception or stack
> overflow.
> 
> Makes sense?

Hi Aaron,

I see your point and I agree that existing code is not broken, it is functioning
well as it is.

But we are fixing a possible issue, or lets say fixing using less secure API
although it doesn't cause any problem right now. Perhaps we can update the patch
title slightly [1] but I am for keeping the fix and I think it makes sense to
keep "Fixes" tag so that this update can be backported to stable trees.

Thanks,
ferruh

[1]
test/eventdev: fix possible buffer overflow