Re: [dpdk-dev] DPDK Release Status Meeting 14/05/2020

[Kevin Traynor <ktraynor@redhat.com>]

On 14/05/2020 12:17, Ferruh Yigit wrote:
> On 5/14/2020 12:13 PM, Ferruh Yigit wrote:
>> Minutes 14 May 2020
>> -------------------
> 
> <...>
> 
>>   * gcc10
>>     * Akhil's fix for crypto looks good, can replace temporary solution
>>     * Gavin has patch for mbuf, but this stage can be late/risky for complex fix
>>       * At this stage we can go with disabling warning as temporary solution
>>     * Ferruh is getting a warning on hash too, will share it
> 
> Hi David,
> 
> Following is what I am getting [1] with compiler version:
> gcc (GCC) 10.1.1 20200511
> 
> 
> [1]
> .../dpdk/lib/librte_hash/rte_cuckoo_hash.c: In function
> ‘__rte_hash_add_key_with_hash’:
> .../dpdk/lib/librte_hash/rte_cuckoo_hash.c:1104:29:
>     error: ‘ext_bkt_id’ may be used uninitialized in this function
> [-Werror=maybe-uninitialized]
>  1104 |  (h->buckets_ext[ext_bkt_id - 1]).sig_current[0] = short_sig;
>       |                  ~~~~~~~~~~~^~~
> 

The return value of rte_ring_sc_dequeue_elem() is checked, but gcc
seems to find some possibility of ext_bkt_id not being initialized.

It was introduced with commit [1], but I guess the tools wouldn't have
found it at time of commit.

zero is an invalid value for ext_bkt_id, so we can just initialize to
that and check it is overwritten. Sent patch
https://patchwork.dpdk.org/patch/70333/

[1]
commit fbfe568103b046a3c6c69c9ecd6c914b391c6b82
Author: Honnappa Nagarahalli <honnappa.nagarahalli@arm.com>
Date:   Sat Jan 18 13:32:46 2020 -0600

    hash: use 32-bit elements rings to save memory