Re: [dpdk-dev] [dpdk-techboard] [PATCH] config: disable all kmods by default from v20.02

[Kevin Traynor <ktraynor@redhat.com>]

On 17/01/2020 08:29, Maxime Coquelin wrote:
> 
> 
> On 1/16/20 7:43 PM, Jerin Jacob wrote:
>> On Fri, Jan 17, 2020 at 12:04 AM Maxime Coquelin
>> <maxime.coquelin@redhat.com> wrote:
>>>
>>>
>>>
>>> On 1/16/20 2:11 PM, Jerin Jacob wrote:
>>>> Ping.
>>>>
>>>> On Thu, Dec 12, 2019 at 6:42 PM <jerinj@marvell.com> wrote:
>>>>>
>>>>> From: Jerin Jacob <jerinj@marvell.com>
>>>>>
>>>>> Based on the techboard meeting held on 2019-11-06,
>>>>> It's been decided to disable all kmods by default from v20.02.
>>>>>
>>>>> http://mails.dpdk.org/archives/dev/2019-November/151763.html
>>>>>
>>>>> Signed-off-by: Jerin Jacob <jerinj@marvell.com>
>>>>> ---
>>>>>  config/common_linux | 2 --
>>>>>  1 file changed, 2 deletions(-)
>>>>>
>>>>> diff --git a/config/common_linux b/config/common_linux
>>>>> index c5cf3d662..583efadc7 100644
>>>>> --- a/config/common_linux
>>>>> +++ b/config/common_linux
>>>>> @@ -8,9 +8,7 @@ CONFIG_RTE_EXEC_ENV_LINUX=y
>>>>>  CONFIG_RTE_EXEC_ENV_LINUXAPP=y
>>>>>
>>>>>  CONFIG_RTE_EAL_NUMA_AWARE_HUGEPAGES=y
>>>>> -CONFIG_RTE_EAL_IGB_UIO=y
>>>>>  CONFIG_RTE_EAL_VFIO=y
>>>>> -CONFIG_RTE_KNI_KMOD=y
>>>>>  CONFIG_RTE_LIBRTE_KNI=y
>>>>>  CONFIG_RTE_LIBRTE_PMD_KNI=y
>>>>>  CONFIG_RTE_LIBRTE_VHOST=y
>>>>> --
>>>>> 2.24.0
>>>>>
>>>>
>>>
>>> Looks good for make build system, but what about Meson?
>>> as per meson_options.txt, isn't it enabled by default?
>>> $ grep -rn enable_kmods meson_options.txt
>>> 7:option('enable_kmods', type: 'boolean', value: true,
>>
>> + Bruce,
>>
>> Since meson autodetect the presence of Linux kernel dependency for the
>> kmod build,
>> I thought of keeping as it is as there is no harm. I am fine with
>> either way, let me the
>> know the feedback, I will send the v2 based on the comment.
> 
> Actually, it would defeat the main reason I pushed for the kmods to be
> disabled by default. Indeed, even if the kernel dependencies are
> installed, build can break if internal kernel API changes, e.g.:
> 
> commit c128a4e6317d49362e2158edc29887c844067c65
> Author: David Zeng <zengxhsh@cn.ibm.com>
> Date:   Sat Dec 22 00:27:33 2018 +0800
> 
>     kni: fix build on RHEL8 for arm and Power9
> 
>     [ upstream commit 3d5501d568aacbcf71832691278f5656d3a9b649 ]
> 
>     Signed-off-by: David Zeng <zengxhsh@cn.ibm.com>
>     Acked-by: Ferruh Yigit <ferruh.yigit@intel.com>
> 

That's a good point. We take these types of patches into stable, but
there would be a few months delay between a new kernel causing breakage
and the time when stable releases containing the fix are released.
Disabling would at least avoid non-kmod interested user being impacted
by default.