[PATCH v14 5/6] memarea: support dump API

[fengchengwen]

Hi Anatoly,

  Thanks for your review, a lot of valuable advice.

  PS: This library stays for a long time, want to hear TB's opinion: whether to continue or stop.
      If continue I will push a new version.

Thanks.

On 2023/6/22 23:55, Burakov, Anatoly wrote:
> On 2/9/2023 6:36 AM, Chengwen Feng wrote:
>> This patch supports rte_memarea_dump() API which could be used for
>> debug.
>>
>> Signed-off-by: Chengwen Feng <fengchengwen at huawei.com>
>> Reviewed-by: Dongdong Liu <liudongdong3 at huawei.com>
>> Acked-by: Morten Brørup <mb at smartsharesystems.com>
>> ---
> 
> Provisionally,
> 
> Acked-by: Anatoly Burakov <anatoly.burakov at intel.com>
> 
> As long as below is addressed.
> 
>> +static void
>> +memarea_dump_objects_detail(struct rte_memarea *ma, FILE *f)
>> +{
>> +    struct memarea_objhdr *hdr;
>> +    size_t offset;
>> +    void *ptr;
>> +
>> +    fprintf(f, "  objects:\n");
>> +    TAILQ_FOREACH(hdr, &ma->obj_list, obj_next) {
>> +        if (hdr == ma->guard_hdr)
>> +            break;
>> +        memarea_check_cookie(ma, hdr, 2);
>> +        ptr = RTE_PTR_ADD(hdr, sizeof(struct memarea_objhdr));
>> +        offset = RTE_PTR_DIFF(ptr, ma->area_base);
>> +#ifdef RTE_LIBRTE_MEMAREA_DEBUG
>> +        fprintf(f, "    %p off: 0x%zx size: 0x%zx %s\n",
>> +            ptr, offset, MEMAREA_OBJECT_GET_SIZE(hdr),
>> +            MEMAREA_OBJECT_IS_ALLOCATED(hdr) ? "allocated" : "");
>> +#else
>> +        fprintf(f, "    off: 0x%zx size: 0x%zx %s\n",
>> +            offset, MEMAREA_OBJECT_GET_SIZE(hdr),
>> +            MEMAREA_OBJECT_IS_ALLOCATED(hdr) ? "allocated" : "");
>> +#endif
>> +    }.
>> +}
>> +
>> +int
>> +rte_memarea_dump(struct rte_memarea *ma, FILE *f, bool dump_all)
>> +{
>> +    if (ma == NULL || f == NULL)
>> +        return -EINVAL;
> 
> I feel like the API is inconsistent in this way. I would suggest picking a method of error reporting, and sticking with it. I would suggest returning 0/-1 or ptr/NULL with rte_errno set to indicate error, as that is how most libraries in DPDK behave.
>