[dpdk-dev] rte epoll functions

Lukasz Wojciechowski l.wojciechow at partner.samsung.com
Tue Apr 28 03:43:30 CEST 2020


W dniu 27.04.2020 o 18:57, Stephen Hemminger pisze:
> On Mon, 27 Apr 2020 17:01:21 +0200
> David Marchand <david.marchand at redhat.com> wrote:
>
>> On Mon, Apr 27, 2020 at 4:51 PM Stephen Hemminger
>> <stephen at networkplumber.org> wrote:
>>> On Mon, 27 Apr 2020 13:33:27 +0000
>>> Liron Himi <lironh at marvell.com> wrote:
>>>   
>>>> Hi,
>>>>
>>>>
>>>> We noticed that the implementation of rte_epoll_wait doesn't stopped on signal such as SIGINT.
>>>>
>>>> Is anyone know why?
>>>>
>>>> If that on purpose and cannot be changed, how application can exit gracefully when using rx interrupts for example.
>>>>
>>>> Regards,
>>>> Liron
>>>>   
>>> I noticed that, submitted a patch, and it has been unmerged for 9 months
>>> https://protect2.fireeye.com/url?k=b1d79a91-ec04c185-b1d611de-0cc47a31ce4e-7c8024d7bc014fef&q=1&u=https%3A%2F%2Fpatchwork.dpdk.org%2Fpatch%2F57047%2F
>> - Indeed, there was no review nor any comment about the "bien fondé"
>> of this patch for 9 months,
>> - We don't need a new symbol, this should be fixed + symbol versioning
>> to keep current behavior.
>>
>>
> We do need a new symbol. People may want the old behavior.
Do we need new symbol and new functionality?
One can use signalfd to create a file descriptor for set of signals and 
just pass it to epoll.

-- 

Lukasz Wojciechowski
Principal Software Engineer

Samsung R&D Institute Poland
Samsung Electronics
Office +48 22 377 88 25
l.wojciechow at partner.samsung.com



More information about the dev mailing list