On 3/16/2023I 9:19 AM, Dmitry Kozlyuk wrote:
> On Thu, Mar 16, 2023 at 4:11 AM fengchengwen <fengcheng...@huawei.com> wrote:
>> Because we have no API to know the PMD whether impl specific ops, we could
>> only knowed by invoking.
>> Except above impl, I also consider the other two:
>> 1. just invoke rte_eth_dev_priv_dump without previous printf("Device private
>> info") and later error printf.
>> and I think people may curious about the extra output without a prompt
>> just like "Device private info".
>> 2. use fmemopen (the below code), this way will perfect process the PMD
>> which not imp ops.
>> FILE *f = fmemopen(buf, max-size(e.g. 128KB));
>> ret = rte_eth_dev_priv_dump(port_id, f);
>> if (ret == 0) {
>> printf("Device private info:\n");
>> printf("%s", buf);
>> }
>> But the windows platform don't support fmemopen.
>>
>> Hope for more feedback.
>
> What if rte_eth_dev_priv_dump() was a documented no-op when "f == NULL"?
> This can be implemented in ethdev layer:
> 1) if not implemented, return ENOTSUP
> 2) if f == NULL, return 0
> 3) else call PMD
> Technically, even now a null device handle can be used, but this is
> cumbersome and wastes resources for running the API twice.
Not sure about to overload "f == NULL" condition to detect the feature
support.
It may be good to have a generic way to detect the support.
One way is to add new set of APIs just to test the dev_ops, and return
boolean like:
'bool rte_eth_dev_is_priv_dump(uint16_t port_id);'
Another option can be introducing an enum, each enum item can represent
a dev_ops and a single API can be used to detect the support. This
requires more maintenance for long term, as app needs to know more,
not sure if I like this.
If there is no objection to add new APIs, I can own the task, to have
APIs like:
'bool rte_eth_dev_is_xxx(uint16_t port_id);'
