On 29 Apr 2026, at 8:47, Eli Britstein wrote:
> On 28/04/2026 13:06, Eelco Chaudron wrote:
>> External email: Use caution opening links or attachments
>>
>>
>> On 26 Apr 2026, at 7:33, Eli Britstein wrote:
>>
>>> On 15/04/2026 15:48, Eelco Chaudron wrote:
>> [...]
>>
>>>>> +/* Structs. */
>>>>> +
>>>>> +#ifndef NETDEV_DPDK_TX_Q_TYPE
>>>>> +#error "NETDEV_DPDK_TX_Q_TYPE must be defined before" \
>>>>> + "including netdev-dpdk-private.h"
>>>>> +#endif
>>>>> +
>>>>> +#ifndef NETDEV_DPDK_SW_STATS_TYPE
>>>>> +#error "NETDEV_DPDK_SW_STATS_TYPE must be defined before" \
>>>>> + "including netdev-dpdk-private.h"
>>>>> +#endif
>>>>> +
>>>>> +#ifndef NETDEV_DPDK_GLOBAL_MUTEX
>>>>> +#error "NETDEV_DPDK_GLOBAL_MUTEX must be defined before" \
>>>>> + "including netdev-dpdk-private.h"
>>>>> +#endif
>>>> Should the names have some kind of suffix so we can easily see what they
>>>> are? Something like:
>>>>
>>>> #define NETDEV_DPDK_TX_QUEUE_STRUCT
>>>> #define NETDEV_DPDK_SW_STATS_STRUCT
>>>> #define NETDEV_DPDK_GLOBAL_MUTEX_NAME
>>>>
>>>> Open for any other suggestions.
>>> I don't mind about the naming, but taking into consideration the comment on
>>> the next commit about getting rid of them:
>>>
>>> The mutex is there so we can still have "OVS_ACQ_AFTER" and
>>> "OVS_GUARDED_BY". IMO we can drop them in favor of a comment. WDYT?
>>>
>>> The other ones can be void* but then require casting upon use.
>> You can keep the mutex one as it does not actually change the context of the
>> structure, and I like having these checks.
>>
>> Regarding using void *, I'd rather use specific types; see the comments on
>> the other patch.
> Ack. I'll keep the mutex as NETDEV_DPDK_GLOBAL_MUTEX_NAME. Regarding the
> others, I'll use dpdk structs for doca as well, exposed in
> netdev-dpdk-private. it's just few more fields that will not be used.
I'm not sure I fully grasp the last comment, but I'll take another
look when v4 arrives ;)
>>
>> [...]
>>
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
