[dpdk-dev] [PATCH 1/3] eal: add --dev-hotplug option

David Marchand david.marchand at redhat.com
Wed Jan 2 15:46:06 CET 2019


Hello Jeff,

On Sat, Dec 29, 2018 at 5:06 AM Jeff Guo <jia.guo at intel.com> wrote:

> On 12/17/2018 6:15 PM, David Marchand wrote:
>
>
> On Fri, Dec 14, 2018 at 8:41 AM Jeff Guo <jia.guo at intel.com> wrote:
>
>> This command-line option will enable hotplug event detecting and enable
>> hotplug handling for device hotplug.
>>
>> Signed-off-by: Jeff Guo <jia.guo at intel.com>
>>
>
> Is there a reason why we would want this disabled by default and enabled
> via option ?
>
>
> Before i can give you an answer, let's see what will bring on if enable
> it.
>
> When enable the hotplug will means that it will bring a new netlink socket
> communication
>
> and a sigbus detecting and specific processing. So if user not want to add
> this work load, just
>
> let it to be optional. Do you agree with that? If not please show what is
> your concern. Thanks.
>

If the user does nothing about the sigbus signal handling but the eal
signal handler was not registered, the dpdk app will end up being
terminated by the kernel.
If the user wants to do its own things and don't want the eal to mess with
it... I am under the impression that he can disable the eal sigbus handler
by calling rte_dev_hotplug_handle_disable().
The netlink stuff is handled in the interrupt thread, no impact on the
processing threads and no additional thread afaics.


-- 
David Marchand


More information about the dev mailing list