[dpdk-stable] please help backporting some patches to LTS release 17.11.4

[dpdk-stable] please help backporting some patches to LTS release 17.11.4

[Yongseok Koh]

Hi commit authors (and maintainers),

I didn't manage to apply following commits from upstream to stable branch
17.11: conflict happens. I'm wondering can the authors check the following
list and backport those patches belong to you?

FYI, branch 17.11 is located at tree:
   git://dpdk.org/dpdk-stable

It'd be great if you could do that in one or two weeks. Also, please add a
heading line like below before the commit log body:
    [ backported from upstream commit xxx ]

Example: http://dpdk.org/browse/dpdk-stable/commit/?h=16.07&id=c4831394c7d1944d8ec27d52c22997f20d19718e

Also please mention the target LTS in the subject line, as we have more than one
at the same time, for example:

    [PATCH 17.11] foo/bar: fix baz

With git send-email, this can be achieved by appending the parameter:

    --subject-prefix='17.11'

Please let me know if you have any comments, say, need more time, or it's
worthless to packport it. And please send it to "stable@dpdk.org", but not
"dev@dpdk.org".

Thanks.

Yongseok

---
422515b9a  Adrien Mazarguil app/testpmd: fix crash when attaching a device
66f2ac0f8  Adrien Mazarguil net/mlx4: fix minor resource leak during init
0cc1fcdeb  Ajit Khaparde    net/bnxt: check filter type before clearing it
6a9360963  Ajit Khaparde    net/bnxt: fix clear port stats
3c2e87c10  Ajit Khaparde    net/bnxt: fix close operation
90362583e  Ajit Khaparde    net/bnxt: fix incorrect IO address handling in Tx
812bb2087  Andrew Rybchenko net/sfc: handle unknown L3 packet class in EF10 event parser
2a4ed7233  Damjan Marion    net/i40e: do not reset device info data
64cb90f88  Fiona Trahe      crypto/qat: fix checks for 3GPP algo bit params
2a7e3d546  Hyong Youb Kim   net/enic: do not overwrite admin Tx queue limit
e11411b52  Maxime Coquelin  vhost: fix missing increment of log cache count
79d098921  Moti Haimovsky   net/mlx5: fix build with old kernels
c44fbc7cc  Nelio Laranjeiro net/mlx5: clean-up developer logs
179c7e893  Pavan Nikhilesh  mempool/octeontx: fix pool to aura mapping
196e9a486  Qi Zhang         eal: fix hotplug add and remove
9f32c7e7e  Rafal Kozik      net/ena: change memory type
46916aa17  Rafal Kozik      net/ena: check pointer before memset
41e59028d  Rafal Kozik      net/ena: set link speed as none
3abe17195  Rahul Lakkireddy net/cxgbe: fix Rx channel map and queue type
1fc3afdf7  Rasesh Mody      doc: update qede management firmware guide
c6dd1eb8a  Rasesh Mody      net/qede: fix default extended VLAN offload config
a16aef523  Rasesh Mody      net/qede: fix for devargs
e46821e9f  Shahaf Shuler    net/mlx5: separate generic tunnel TSO from the standard one
496f15c2c  Shahed Shaikh    net/qede: fix incorrect link status update
c7641841a  Shahed Shaikh    net/qede: fix unicast MAC address handling in VF
82dbef2c8  Somnath Kotur    net/bnxt: revert reset of L2 filter id
0989161b2  Tiwei Bie        vhost: release locks on RARP packet failure
b2a4a69f7  Xiaoxin Peng     net/bnxt: fix Tx with multiple mbuf

Re: [dpdk-stable] please help backporting some patches to LTS release 17.11.4

[Ilya Maximets]

[keeping in CC only e11411b52 related persons]
[CC: Luca Boccassi]

Hi.
I beleive that I was added to this thread as a reviewer of
"e11411b52  Maxime Coquelin  vhost: fix missing increment of log cache count".

Looking at the current dpdk-stable/17.11 I see that the patch could not be
applied because it fixes issue introduced by another patch
c16915b87109 ("vhost: improve dirty pages logging performance") that doesn't
exist in this branch. OTOH, this patch exists in stable 16.11.
So, I have a question: Is there any reason why c16915b87109 was backported
to 16.11 but not to 17.11 ?

P.S. I don't feel that I should backport e11411b52 myself, but I can do that
     if author/maintainers are too busy. Please, let me know in that case.
     Regarding 16.11, to backport e11411b52 you only need to change the
     destination file,because related code placed in
     lib/librte_vhost/virtio_net.c instead of lib/librte_vhost/vhost.h.

Best regards, Ilya Maximets.

On 27.07.2018 05:40, Yongseok Koh wrote:
> Hi commit authors (and maintainers),
> 
> I didn't manage to apply following commits from upstream to stable branch
> 17.11: conflict happens. I'm wondering can the authors check the following
> list and backport those patches belong to you?
> 
> FYI, branch 17.11 is located at tree:
>    git://dpdk.org/dpdk-stable
> 
> It'd be great if you could do that in one or two weeks. Also, please add a
> heading line like below before the commit log body:
>     [ backported from upstream commit xxx ]
> 
> Example: http://dpdk.org/browse/dpdk-stable/commit/?h=16.07&id=c4831394c7d1944d8ec27d52c22997f20d19718e
> 
> Also please mention the target LTS in the subject line, as we have more than one
> at the same time, for example:
> 
>     [PATCH 17.11] foo/bar: fix baz
> 
> With git send-email, this can be achieved by appending the parameter:
> 
>     --subject-prefix='17.11'
> 
> Please let me know if you have any comments, say, need more time, or it's
> worthless to packport it. And please send it to "stable@dpdk.org", but not
> "dev@dpdk.org".
> 
> Thanks.
> 
> Yongseok
> 
> ---
> 422515b9a  Adrien Mazarguil app/testpmd: fix crash when attaching a device
> 66f2ac0f8  Adrien Mazarguil net/mlx4: fix minor resource leak during init
> 0cc1fcdeb  Ajit Khaparde    net/bnxt: check filter type before clearing it
> 6a9360963  Ajit Khaparde    net/bnxt: fix clear port stats
> 3c2e87c10  Ajit Khaparde    net/bnxt: fix close operation
> 90362583e  Ajit Khaparde    net/bnxt: fix incorrect IO address handling in Tx
> 812bb2087  Andrew Rybchenko net/sfc: handle unknown L3 packet class in EF10 event parser
> 2a4ed7233  Damjan Marion    net/i40e: do not reset device info data
> 64cb90f88  Fiona Trahe      crypto/qat: fix checks for 3GPP algo bit params
> 2a7e3d546  Hyong Youb Kim   net/enic: do not overwrite admin Tx queue limit
> e11411b52  Maxime Coquelin  vhost: fix missing increment of log cache count
> 79d098921  Moti Haimovsky   net/mlx5: fix build with old kernels
> c44fbc7cc  Nelio Laranjeiro net/mlx5: clean-up developer logs
> 179c7e893  Pavan Nikhilesh  mempool/octeontx: fix pool to aura mapping
> 196e9a486  Qi Zhang         eal: fix hotplug add and remove
> 9f32c7e7e  Rafal Kozik      net/ena: change memory type
> 46916aa17  Rafal Kozik      net/ena: check pointer before memset
> 41e59028d  Rafal Kozik      net/ena: set link speed as none
> 3abe17195  Rahul Lakkireddy net/cxgbe: fix Rx channel map and queue type
> 1fc3afdf7  Rasesh Mody      doc: update qede management firmware guide
> c6dd1eb8a  Rasesh Mody      net/qede: fix default extended VLAN offload config
> a16aef523  Rasesh Mody      net/qede: fix for devargs
> e46821e9f  Shahaf Shuler    net/mlx5: separate generic tunnel TSO from the standard one
> 496f15c2c  Shahed Shaikh    net/qede: fix incorrect link status update
> c7641841a  Shahed Shaikh    net/qede: fix unicast MAC address handling in VF
> 82dbef2c8  Somnath Kotur    net/bnxt: revert reset of L2 filter id
> 0989161b2  Tiwei Bie        vhost: release locks on RARP packet failure
> b2a4a69f7  Xiaoxin Peng     net/bnxt: fix Tx with multiple mbuf
> 
> 

Re: [dpdk-stable] please help backporting some patches to LTS release 17.11.4

[Luca Boccassi]

On Fri, 2018-07-27 at 10:42 +0300, Ilya Maximets wrote:
> [keeping in CC only e11411b52 related persons]
> [CC: Luca Boccassi]
> 
> Hi.
> I beleive that I was added to this thread as a reviewer of
> "e11411b52  Maxime Coquelin  vhost: fix missing increment of log
> cache count".
> 
> Looking at the current dpdk-stable/17.11 I see that the patch could
> not be
> applied because it fixes issue introduced by another patch
> c16915b87109 ("vhost: improve dirty pages logging performance") that
> doesn't
> exist in this branch. OTOH, this patch exists in stable 16.11.
> So, I have a question: Is there any reason why c16915b87109 was
> backported
> to 16.11 but not to 17.11 ?
> 
> P.S. I don't feel that I should backport e11411b52 myself, but I can
> do that
>      if author/maintainers are too busy. Please, let me know in that
> case.
>      Regarding 16.11, to backport e11411b52 you only need to change
> the
>      destination file,because related code placed in
>      lib/librte_vhost/virtio_net.c instead of
> lib/librte_vhost/vhost.h.
> 
> Best regards, Ilya Maximets.

Hi Ilya,

Thanks for the hint w.r.t. 16.11, I have now applied and pushed the
patch to dpdk-stable/16.11.

> On 27.07.2018 05:40, Yongseok Koh wrote:
> > Hi commit authors (and maintainers),
> > 
> > I didn't manage to apply following commits from upstream to stable
> > branch
> > 17.11: conflict happens. I'm wondering can the authors check the
> > following
> > list and backport those patches belong to you?
> > 
> > FYI, branch 17.11 is located at tree:
> >    git://dpdk.org/dpdk-stable
> > 
> > It'd be great if you could do that in one or two weeks. Also,
> > please add a
> > heading line like below before the commit log body:
> >     [ backported from upstream commit xxx ]
> > 
> > Example: http://dpdk.org/browse/dpdk-stable/commit/?h=16.07&id=c483
> > 1394c7d1944d8ec27d52c22997f20d19718e
> > 
> > Also please mention the target LTS in the subject line, as we have
> > more than one
> > at the same time, for example:
> > 
> >     [PATCH 17.11] foo/bar: fix baz
> > 
> > With git send-email, this can be achieved by appending the
> > parameter:
> > 
> >     --subject-prefix='17.11'
> > 
> > Please let me know if you have any comments, say, need more time,
> > or it's
> > worthless to packport it. And please send it to "stable@dpdk.org",
> > but not
> > "dev@dpdk.org".
> > 
> > Thanks.
> > 
> > Yongseok
> > 
> > ---
> > 422515b9a  Adrien Mazarguil app/testpmd: fix crash when attaching a
> > device
> > 66f2ac0f8  Adrien Mazarguil net/mlx4: fix minor resource leak
> > during init
> > 0cc1fcdeb  Ajit Khaparde    net/bnxt: check filter type before
> > clearing it
> > 6a9360963  Ajit Khaparde    net/bnxt: fix clear port stats
> > 3c2e87c10  Ajit Khaparde    net/bnxt: fix close operation
> > 90362583e  Ajit Khaparde    net/bnxt: fix incorrect IO address
> > handling in Tx
> > 812bb2087  Andrew Rybchenko net/sfc: handle unknown L3 packet class
> > in EF10 event parser
> > 2a4ed7233  Damjan Marion    net/i40e: do not reset device info data
> > 64cb90f88  Fiona Trahe      crypto/qat: fix checks for 3GPP algo
> > bit params
> > 2a7e3d546  Hyong Youb Kim   net/enic: do not overwrite admin Tx
> > queue limit
> > e11411b52  Maxime Coquelin  vhost: fix missing increment of log
> > cache count
> > 79d098921  Moti Haimovsky   net/mlx5: fix build with old kernels
> > c44fbc7cc  Nelio Laranjeiro net/mlx5: clean-up developer logs
> > 179c7e893  Pavan Nikhilesh  mempool/octeontx: fix pool to aura
> > mapping
> > 196e9a486  Qi Zhang         eal: fix hotplug add and remove
> > 9f32c7e7e  Rafal Kozik      net/ena: change memory type
> > 46916aa17  Rafal Kozik      net/ena: check pointer before memset
> > 41e59028d  Rafal Kozik      net/ena: set link speed as none
> > 3abe17195  Rahul Lakkireddy net/cxgbe: fix Rx channel map and queue
> > type
> > 1fc3afdf7  Rasesh Mody      doc: update qede management firmware
> > guide
> > c6dd1eb8a  Rasesh Mody      net/qede: fix default extended VLAN
> > offload config
> > a16aef523  Rasesh Mody      net/qede: fix for devargs
> > e46821e9f  Shahaf Shuler    net/mlx5: separate generic tunnel TSO
> > from the standard one
> > 496f15c2c  Shahed Shaikh    net/qede: fix incorrect link status
> > update
> > c7641841a  Shahed Shaikh    net/qede: fix unicast MAC address
> > handling in VF
> > 82dbef2c8  Somnath Kotur    net/bnxt: revert reset of L2 filter id
> > 0989161b2  Tiwei Bie        vhost: release locks on RARP packet
> > failure
> > b2a4a69f7  Xiaoxin Peng     net/bnxt: fix Tx with multiple mbuf
> > 
> > 

-- 
Kind regards,
Luca Boccassi

Re: [dpdk-stable] please help backporting some patches to LTS release 17.11.4

[Kevin Traynor]

On 07/27/2018 08:42 AM, Ilya Maximets wrote:
> [keeping in CC only e11411b52 related persons]
> [CC: Luca Boccassi]
> 
> Hi.
> I beleive that I was added to this thread as a reviewer of
> "e11411b52  Maxime Coquelin  vhost: fix missing increment of log cache count".
> 
> Looking at the current dpdk-stable/17.11 I see that the patch could not be
> applied because it fixes issue introduced by another patch
> c16915b87109 ("vhost: improve dirty pages logging performance") that doesn't
> exist in this branch. OTOH, this patch exists in stable 16.11.
> So, I have a question: Is there any reason why c16915b87109 was backported
> to 16.11 but not to 17.11 ?
> 

Hi Ilya, Guessing it was because it was an optimization as opposed to a
functional fix. See below commit wrt clarifying this - ref last para.

Kevin.

commit 56c7ddc1381a56612018b75a848169b917bc33fb
Author: Kevin Traynor <ktraynor@redhat.com>
Date:   Sat Jun 30 00:10:43 2018 +0100

    doc: add default that all fixes should be backported

    Set the starting point that all commits on master branch
    with Fixes tag should be backported to relevant stable/LTS
    branches, and explain that the submitter may indicate it is
    not suitable for backport.

    Of course there will be exceptions that will crop up from time
    to time that need discussion, so also add a sentence for that.

    This is to ensure that there is consistency between what is
    backported to stable/LTS branches, remove some subjectivity
    as to what constitutes "a fix" and avoid possible conflicts
    for future backports.

    Signed-off-by: Kevin Traynor <ktraynor@redhat.com>
    Acked-by: Luca Boccassi <bluca@debian.org>
    Acked-by: Aaron Conole <aconole@redhat.com>
    Acked-by: Thomas Monjalon <thomas@monjalon.net>

> P.S. I don't feel that I should backport e11411b52 myself, but I can do that
>      if author/maintainers are too busy. Please, let me know in that case.
>      Regarding 16.11, to backport e11411b52 you only need to change the
>      destination file,because related code placed in
>      lib/librte_vhost/virtio_net.c instead of lib/librte_vhost/vhost.h.
> 
> Best regards, Ilya Maximets.
> 
> On 27.07.2018 05:40, Yongseok Koh wrote:
>> Hi commit authors (and maintainers),
>>
>> I didn't manage to apply following commits from upstream to stable branch
>> 17.11: conflict happens. I'm wondering can the authors check the following
>> list and backport those patches belong to you?
>>
>> FYI, branch 17.11 is located at tree:
>>    git://dpdk.org/dpdk-stable
>>
>> It'd be great if you could do that in one or two weeks. Also, please add a
>> heading line like below before the commit log body:
>>     [ backported from upstream commit xxx ]
>>
>> Example: http://dpdk.org/browse/dpdk-stable/commit/?h=16.07&id=c4831394c7d1944d8ec27d52c22997f20d19718e
>>
>> Also please mention the target LTS in the subject line, as we have more than one
>> at the same time, for example:
>>
>>     [PATCH 17.11] foo/bar: fix baz
>>
>> With git send-email, this can be achieved by appending the parameter:
>>
>>     --subject-prefix='17.11'
>>
>> Please let me know if you have any comments, say, need more time, or it's
>> worthless to packport it. And please send it to "stable@dpdk.org", but not
>> "dev@dpdk.org".
>>
>> Thanks.
>>
>> Yongseok
>>
>> ---
>> 422515b9a  Adrien Mazarguil app/testpmd: fix crash when attaching a device
>> 66f2ac0f8  Adrien Mazarguil net/mlx4: fix minor resource leak during init
>> 0cc1fcdeb  Ajit Khaparde    net/bnxt: check filter type before clearing it
>> 6a9360963  Ajit Khaparde    net/bnxt: fix clear port stats
>> 3c2e87c10  Ajit Khaparde    net/bnxt: fix close operation
>> 90362583e  Ajit Khaparde    net/bnxt: fix incorrect IO address handling in Tx
>> 812bb2087  Andrew Rybchenko net/sfc: handle unknown L3 packet class in EF10 event parser
>> 2a4ed7233  Damjan Marion    net/i40e: do not reset device info data
>> 64cb90f88  Fiona Trahe      crypto/qat: fix checks for 3GPP algo bit params
>> 2a7e3d546  Hyong Youb Kim   net/enic: do not overwrite admin Tx queue limit
>> e11411b52  Maxime Coquelin  vhost: fix missing increment of log cache count
>> 79d098921  Moti Haimovsky   net/mlx5: fix build with old kernels
>> c44fbc7cc  Nelio Laranjeiro net/mlx5: clean-up developer logs
>> 179c7e893  Pavan Nikhilesh  mempool/octeontx: fix pool to aura mapping
>> 196e9a486  Qi Zhang         eal: fix hotplug add and remove
>> 9f32c7e7e  Rafal Kozik      net/ena: change memory type
>> 46916aa17  Rafal Kozik      net/ena: check pointer before memset
>> 41e59028d  Rafal Kozik      net/ena: set link speed as none
>> 3abe17195  Rahul Lakkireddy net/cxgbe: fix Rx channel map and queue type
>> 1fc3afdf7  Rasesh Mody      doc: update qede management firmware guide
>> c6dd1eb8a  Rasesh Mody      net/qede: fix default extended VLAN offload config
>> a16aef523  Rasesh Mody      net/qede: fix for devargs
>> e46821e9f  Shahaf Shuler    net/mlx5: separate generic tunnel TSO from the standard one
>> 496f15c2c  Shahed Shaikh    net/qede: fix incorrect link status update
>> c7641841a  Shahed Shaikh    net/qede: fix unicast MAC address handling in VF
>> 82dbef2c8  Somnath Kotur    net/bnxt: revert reset of L2 filter id
>> 0989161b2  Tiwei Bie        vhost: release locks on RARP packet failure
>> b2a4a69f7  Xiaoxin Peng     net/bnxt: fix Tx with multiple mbuf
>>
>>

Re: [dpdk-stable] please help backporting some patches to LTS release 17.11.4

[Ilya Maximets]

On 27.07.2018 19:25, Kevin Traynor wrote:
> On 07/27/2018 08:42 AM, Ilya Maximets wrote:
>> [keeping in CC only e11411b52 related persons]
>> [CC: Luca Boccassi]
>>
>> Hi.
>> I beleive that I was added to this thread as a reviewer of
>> "e11411b52  Maxime Coquelin  vhost: fix missing increment of log cache count".
>>
>> Looking at the current dpdk-stable/17.11 I see that the patch could not be
>> applied because it fixes issue introduced by another patch
>> c16915b87109 ("vhost: improve dirty pages logging performance") that doesn't
>> exist in this branch. OTOH, this patch exists in stable 16.11.
>> So, I have a question: Is there any reason why c16915b87109 was backported
>> to 16.11 but not to 17.11 ?
>>
> 
> Hi Ilya, Guessing it was because it was an optimization as opposed to a
> functional fix. See below commit wrt clarifying this - ref last para.

Sure, I understand that. But this doesn't clarify why we have this patch in
16.11 and not in 17.11. From the user point of view this means that upgrade
from 16.11 LTS to 17.11 LTS will significantly slow down live-migration.
Later upgrade to master/future-LTS will speed up it back.

> 
> Kevin.
> 
> commit 56c7ddc1381a56612018b75a848169b917bc33fb
> Author: Kevin Traynor <ktraynor@redhat.com>
> Date:   Sat Jun 30 00:10:43 2018 +0100
> 
>     doc: add default that all fixes should be backported
> 
>     Set the starting point that all commits on master branch
>     with Fixes tag should be backported to relevant stable/LTS
>     branches, and explain that the submitter may indicate it is
>     not suitable for backport.
> 
>     Of course there will be exceptions that will crop up from time
>     to time that need discussion, so also add a sentence for that.
> 
>     This is to ensure that there is consistency between what is
>     backported to stable/LTS branches, remove some subjectivity
>     as to what constitutes "a fix" and avoid possible conflicts
>     for future backports.
> 
>     Signed-off-by: Kevin Traynor <ktraynor@redhat.com>
>     Acked-by: Luca Boccassi <bluca@debian.org>
>     Acked-by: Aaron Conole <aconole@redhat.com>
>     Acked-by: Thomas Monjalon <thomas@monjalon.net>
> 
>> P.S. I don't feel that I should backport e11411b52 myself, but I can do that
>>      if author/maintainers are too busy. Please, let me know in that case.
>>      Regarding 16.11, to backport e11411b52 you only need to change the
>>      destination file,because related code placed in
>>      lib/librte_vhost/virtio_net.c instead of lib/librte_vhost/vhost.h.
>>
>> Best regards, Ilya Maximets.
>>
>> On 27.07.2018 05:40, Yongseok Koh wrote:
>>> Hi commit authors (and maintainers),
>>>
>>> I didn't manage to apply following commits from upstream to stable branch
>>> 17.11: conflict happens. I'm wondering can the authors check the following
>>> list and backport those patches belong to you?
>>>
>>> FYI, branch 17.11 is located at tree:
>>>    git://dpdk.org/dpdk-stable
>>>
>>> It'd be great if you could do that in one or two weeks. Also, please add a
>>> heading line like below before the commit log body:
>>>     [ backported from upstream commit xxx ]
>>>
>>> Example: http://dpdk.org/browse/dpdk-stable/commit/?h=16.07&id=c4831394c7d1944d8ec27d52c22997f20d19718e
>>>
>>> Also please mention the target LTS in the subject line, as we have more than one
>>> at the same time, for example:
>>>
>>>     [PATCH 17.11] foo/bar: fix baz
>>>
>>> With git send-email, this can be achieved by appending the parameter:
>>>
>>>     --subject-prefix='17.11'
>>>
>>> Please let me know if you have any comments, say, need more time, or it's
>>> worthless to packport it. And please send it to "stable@dpdk.org", but not
>>> "dev@dpdk.org".
>>>
>>> Thanks.
>>>
>>> Yongseok
>>>
>>> ---
>>> 422515b9a  Adrien Mazarguil app/testpmd: fix crash when attaching a device
>>> 66f2ac0f8  Adrien Mazarguil net/mlx4: fix minor resource leak during init
>>> 0cc1fcdeb  Ajit Khaparde    net/bnxt: check filter type before clearing it
>>> 6a9360963  Ajit Khaparde    net/bnxt: fix clear port stats
>>> 3c2e87c10  Ajit Khaparde    net/bnxt: fix close operation
>>> 90362583e  Ajit Khaparde    net/bnxt: fix incorrect IO address handling in Tx
>>> 812bb2087  Andrew Rybchenko net/sfc: handle unknown L3 packet class in EF10 event parser
>>> 2a4ed7233  Damjan Marion    net/i40e: do not reset device info data
>>> 64cb90f88  Fiona Trahe      crypto/qat: fix checks for 3GPP algo bit params
>>> 2a7e3d546  Hyong Youb Kim   net/enic: do not overwrite admin Tx queue limit
>>> e11411b52  Maxime Coquelin  vhost: fix missing increment of log cache count
>>> 79d098921  Moti Haimovsky   net/mlx5: fix build with old kernels
>>> c44fbc7cc  Nelio Laranjeiro net/mlx5: clean-up developer logs
>>> 179c7e893  Pavan Nikhilesh  mempool/octeontx: fix pool to aura mapping
>>> 196e9a486  Qi Zhang         eal: fix hotplug add and remove
>>> 9f32c7e7e  Rafal Kozik      net/ena: change memory type
>>> 46916aa17  Rafal Kozik      net/ena: check pointer before memset
>>> 41e59028d  Rafal Kozik      net/ena: set link speed as none
>>> 3abe17195  Rahul Lakkireddy net/cxgbe: fix Rx channel map and queue type
>>> 1fc3afdf7  Rasesh Mody      doc: update qede management firmware guide
>>> c6dd1eb8a  Rasesh Mody      net/qede: fix default extended VLAN offload config
>>> a16aef523  Rasesh Mody      net/qede: fix for devargs
>>> e46821e9f  Shahaf Shuler    net/mlx5: separate generic tunnel TSO from the standard one
>>> 496f15c2c  Shahed Shaikh    net/qede: fix incorrect link status update
>>> c7641841a  Shahed Shaikh    net/qede: fix unicast MAC address handling in VF
>>> 82dbef2c8  Somnath Kotur    net/bnxt: revert reset of L2 filter id
>>> 0989161b2  Tiwei Bie        vhost: release locks on RARP packet failure
>>> b2a4a69f7  Xiaoxin Peng     net/bnxt: fix Tx with multiple mbuf
>>>
>>>
> 
> 
> 

Re: [dpdk-stable] please help backporting some patches to LTS release 17.11.4

[Kevin Traynor]

On 07/27/2018 05:46 PM, Ilya Maximets wrote:
> 
> 
> On 27.07.2018 19:25, Kevin Traynor wrote:
>> On 07/27/2018 08:42 AM, Ilya Maximets wrote:
>>> [keeping in CC only e11411b52 related persons]
>>> [CC: Luca Boccassi]
>>>
>>> Hi.
>>> I beleive that I was added to this thread as a reviewer of
>>> "e11411b52  Maxime Coquelin  vhost: fix missing increment of log cache count".
>>>
>>> Looking at the current dpdk-stable/17.11 I see that the patch could not be
>>> applied because it fixes issue introduced by another patch
>>> c16915b87109 ("vhost: improve dirty pages logging performance") that doesn't
>>> exist in this branch. OTOH, this patch exists in stable 16.11.
>>> So, I have a question: Is there any reason why c16915b87109 was backported
>>> to 16.11 but not to 17.11 ?
>>>
>>
>> Hi Ilya, Guessing it was because it was an optimization as opposed to a
>> functional fix. See below commit wrt clarifying this - ref last para.
> 
> Sure, I understand that. But this doesn't clarify why we have this patch in
> 16.11 and not in 17.11. From the user point of view this means that upgrade
> from 16.11 LTS to 17.11 LTS will significantly slow down live-migration.
> Later upgrade to master/future-LTS will speed up it back.
> 

It's never too late :-) I just checked on 17.11 stable branch,
c16915b87109 applies and e11411b52 applies subsequently. Looks fine,
compile tested only.

Yongseok, do you think you could queue these for the 17.11.4 branch?

>>
>> Kevin.
>>
>> commit 56c7ddc1381a56612018b75a848169b917bc33fb
>> Author: Kevin Traynor <ktraynor@redhat.com>
>> Date:   Sat Jun 30 00:10:43 2018 +0100
>>
>>     doc: add default that all fixes should be backported
>>
>>     Set the starting point that all commits on master branch
>>     with Fixes tag should be backported to relevant stable/LTS
>>     branches, and explain that the submitter may indicate it is
>>     not suitable for backport.
>>
>>     Of course there will be exceptions that will crop up from time
>>     to time that need discussion, so also add a sentence for that.
>>
>>     This is to ensure that there is consistency between what is
>>     backported to stable/LTS branches, remove some subjectivity
>>     as to what constitutes "a fix" and avoid possible conflicts
>>     for future backports.
>>
>>     Signed-off-by: Kevin Traynor <ktraynor@redhat.com>
>>     Acked-by: Luca Boccassi <bluca@debian.org>
>>     Acked-by: Aaron Conole <aconole@redhat.com>
>>     Acked-by: Thomas Monjalon <thomas@monjalon.net>
>>
>>> P.S. I don't feel that I should backport e11411b52 myself, but I can do that
>>>      if author/maintainers are too busy. Please, let me know in that case.
>>>      Regarding 16.11, to backport e11411b52 you only need to change the
>>>      destination file,because related code placed in
>>>      lib/librte_vhost/virtio_net.c instead of lib/librte_vhost/vhost.h.
>>>
>>> Best regards, Ilya Maximets.
>>>
>>> On 27.07.2018 05:40, Yongseok Koh wrote:
>>>> Hi commit authors (and maintainers),
>>>>
>>>> I didn't manage to apply following commits from upstream to stable branch
>>>> 17.11: conflict happens. I'm wondering can the authors check the following
>>>> list and backport those patches belong to you?
>>>>
>>>> FYI, branch 17.11 is located at tree:
>>>>    git://dpdk.org/dpdk-stable
>>>>
>>>> It'd be great if you could do that in one or two weeks. Also, please add a
>>>> heading line like below before the commit log body:
>>>>     [ backported from upstream commit xxx ]
>>>>
>>>> Example: http://dpdk.org/browse/dpdk-stable/commit/?h=16.07&id=c4831394c7d1944d8ec27d52c22997f20d19718e
>>>>
>>>> Also please mention the target LTS in the subject line, as we have more than one
>>>> at the same time, for example:
>>>>
>>>>     [PATCH 17.11] foo/bar: fix baz
>>>>
>>>> With git send-email, this can be achieved by appending the parameter:
>>>>
>>>>     --subject-prefix='17.11'
>>>>
>>>> Please let me know if you have any comments, say, need more time, or it's
>>>> worthless to packport it. And please send it to "stable@dpdk.org", but not
>>>> "dev@dpdk.org".
>>>>
>>>> Thanks.
>>>>
>>>> Yongseok
>>>>
>>>> ---
>>>> 422515b9a  Adrien Mazarguil app/testpmd: fix crash when attaching a device
>>>> 66f2ac0f8  Adrien Mazarguil net/mlx4: fix minor resource leak during init
>>>> 0cc1fcdeb  Ajit Khaparde    net/bnxt: check filter type before clearing it
>>>> 6a9360963  Ajit Khaparde    net/bnxt: fix clear port stats
>>>> 3c2e87c10  Ajit Khaparde    net/bnxt: fix close operation
>>>> 90362583e  Ajit Khaparde    net/bnxt: fix incorrect IO address handling in Tx
>>>> 812bb2087  Andrew Rybchenko net/sfc: handle unknown L3 packet class in EF10 event parser
>>>> 2a4ed7233  Damjan Marion    net/i40e: do not reset device info data
>>>> 64cb90f88  Fiona Trahe      crypto/qat: fix checks for 3GPP algo bit params
>>>> 2a7e3d546  Hyong Youb Kim   net/enic: do not overwrite admin Tx queue limit
>>>> e11411b52  Maxime Coquelin  vhost: fix missing increment of log cache count
>>>> 79d098921  Moti Haimovsky   net/mlx5: fix build with old kernels
>>>> c44fbc7cc  Nelio Laranjeiro net/mlx5: clean-up developer logs
>>>> 179c7e893  Pavan Nikhilesh  mempool/octeontx: fix pool to aura mapping
>>>> 196e9a486  Qi Zhang         eal: fix hotplug add and remove
>>>> 9f32c7e7e  Rafal Kozik      net/ena: change memory type
>>>> 46916aa17  Rafal Kozik      net/ena: check pointer before memset
>>>> 41e59028d  Rafal Kozik      net/ena: set link speed as none
>>>> 3abe17195  Rahul Lakkireddy net/cxgbe: fix Rx channel map and queue type
>>>> 1fc3afdf7  Rasesh Mody      doc: update qede management firmware guide
>>>> c6dd1eb8a  Rasesh Mody      net/qede: fix default extended VLAN offload config
>>>> a16aef523  Rasesh Mody      net/qede: fix for devargs
>>>> e46821e9f  Shahaf Shuler    net/mlx5: separate generic tunnel TSO from the standard one
>>>> 496f15c2c  Shahed Shaikh    net/qede: fix incorrect link status update
>>>> c7641841a  Shahed Shaikh    net/qede: fix unicast MAC address handling in VF
>>>> 82dbef2c8  Somnath Kotur    net/bnxt: revert reset of L2 filter id
>>>> 0989161b2  Tiwei Bie        vhost: release locks on RARP packet failure
>>>> b2a4a69f7  Xiaoxin Peng     net/bnxt: fix Tx with multiple mbuf
>>>>
>>>>
>>
>>
>>

Re: [dpdk-stable] please help backporting some patches to LTS release 17.11.4

[Yongseok Koh]

> On Jul 27, 2018, at 10:24 AM, Kevin Traynor <ktraynor@redhat.com> wrote:
> 
> On 07/27/2018 05:46 PM, Ilya Maximets wrote:
>> 
>> 
>> On 27.07.2018 19:25, Kevin Traynor wrote:
>>> On 07/27/2018 08:42 AM, Ilya Maximets wrote:
>>>> [keeping in CC only e11411b52 related persons]
>>>> [CC: Luca Boccassi]
>>>> 
>>>> Hi.
>>>> I beleive that I was added to this thread as a reviewer of
>>>> "e11411b52  Maxime Coquelin  vhost: fix missing increment of log cache count".
>>>> 
>>>> Looking at the current dpdk-stable/17.11 I see that the patch could not be
>>>> applied because it fixes issue introduced by another patch
>>>> c16915b87109 ("vhost: improve dirty pages logging performance") that doesn't
>>>> exist in this branch. OTOH, this patch exists in stable 16.11.
>>>> So, I have a question: Is there any reason why c16915b87109 was backported
>>>> to 16.11 but not to 17.11 ?
>>>> 
>>> 
>>> Hi Ilya, Guessing it was because it was an optimization as opposed to a
>>> functional fix. See below commit wrt clarifying this - ref last para.
>> 
>> Sure, I understand that. But this doesn't clarify why we have this patch in
>> 16.11 and not in 17.11. From the user point of view this means that upgrade
>> from 16.11 LTS to 17.11 LTS will significantly slow down live-migration.
>> Later upgrade to master/future-LTS will speed up it back.
>> 
> 
> It's never too late :-) I just checked on 17.11 stable branch,
> c16915b87109 applies and e11411b52 applies subsequently. Looks fine,
> compile tested only.
> 
> Yongseok, do you think you could queue these for the 17.11.4 branch?

Thanks, guys.
The two patches have been queued for 17.11.4 successfully.

To Ilya's question.
I picked up candidates for 17.11.4 between v18.05 and v18.08-rc2 because 17.11.3
should have covered up to 18.05.  But c16915b87109 had been merged late in
v18.05-rc6, that might be the reason it's missing in the last minute. You know
18.05 was a huge release :-)

I'll scan v18.05-rcX..v18.05 to see if there's any more missing patches.

Thanks,
Yongseok

>>> 
>>> Kevin.
>>> 
>>> commit 56c7ddc1381a56612018b75a848169b917bc33fb
>>> Author: Kevin Traynor <ktraynor@redhat.com>
>>> Date:   Sat Jun 30 00:10:43 2018 +0100
>>> 
>>>    doc: add default that all fixes should be backported
>>> 
>>>    Set the starting point that all commits on master branch
>>>    with Fixes tag should be backported to relevant stable/LTS
>>>    branches, and explain that the submitter may indicate it is
>>>    not suitable for backport.
>>> 
>>>    Of course there will be exceptions that will crop up from time
>>>    to time that need discussion, so also add a sentence for that.
>>> 
>>>    This is to ensure that there is consistency between what is
>>>    backported to stable/LTS branches, remove some subjectivity
>>>    as to what constitutes "a fix" and avoid possible conflicts
>>>    for future backports.
>>> 
>>>    Signed-off-by: Kevin Traynor <ktraynor@redhat.com>
>>>    Acked-by: Luca Boccassi <bluca@debian.org>
>>>    Acked-by: Aaron Conole <aconole@redhat.com>
>>>    Acked-by: Thomas Monjalon <thomas@monjalon.net>
>>> 
>>>> P.S. I don't feel that I should backport e11411b52 myself, but I can do that
>>>>     if author/maintainers are too busy. Please, let me know in that case.
>>>>     Regarding 16.11, to backport e11411b52 you only need to change the
>>>>     destination file,because related code placed in
>>>>     lib/librte_vhost/virtio_net.c instead of lib/librte_vhost/vhost.h.
>>>> 
>>>> Best regards, Ilya Maximets.
>>>> 
>>>> On 27.07.2018 05:40, Yongseok Koh wrote:
>>>>> Hi commit authors (and maintainers),
>>>>> 
>>>>> I didn't manage to apply following commits from upstream to stable branch
>>>>> 17.11: conflict happens. I'm wondering can the authors check the following
>>>>> list and backport those patches belong to you?
>>>>> 
>>>>> FYI, branch 17.11 is located at tree:
>>>>>   git://dpdk.org/dpdk-stable
>>>>> 
>>>>> It'd be great if you could do that in one or two weeks. Also, please add a
>>>>> heading line like below before the commit log body:
>>>>>    [ backported from upstream commit xxx ]
>>>>> 
>>>>> Example: https://emea01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fdpdk.org%2Fbrowse%2Fdpdk-stable%2Fcommit%2F%3Fh%3D16.07%26id%3Dc4831394c7d1944d8ec27d52c22997f20d19718e&amp;data=02%7C01%7Cyskoh%40mellanox.com%7C523e7982d36c4aa66f5d08d5f3e5c2fc%7Ca652971c7d2e4d9ba6a4d149256f461b%7C0%7C0%7C636683090507046633&amp;sdata=gpjXg6vj8IqDsxj9cMC01OauJpLEP114XtbOSGuFaic%3D&amp;reserved=0
>>>>> 
>>>>> Also please mention the target LTS in the subject line, as we have more than one
>>>>> at the same time, for example:
>>>>> 
>>>>>    [PATCH 17.11] foo/bar: fix baz
>>>>> 
>>>>> With git send-email, this can be achieved by appending the parameter:
>>>>> 
>>>>>    --subject-prefix='17.11'
>>>>> 
>>>>> Please let me know if you have any comments, say, need more time, or it's
>>>>> worthless to packport it. And please send it to "stable@dpdk.org", but not
>>>>> "dev@dpdk.org".
>>>>> 
>>>>> Thanks.
>>>>> 
>>>>> Yongseok
>>>>> 
>>>>> ---
>>>>> 422515b9a  Adrien Mazarguil app/testpmd: fix crash when attaching a device
>>>>> 66f2ac0f8  Adrien Mazarguil net/mlx4: fix minor resource leak during init
>>>>> 0cc1fcdeb  Ajit Khaparde    net/bnxt: check filter type before clearing it
>>>>> 6a9360963  Ajit Khaparde    net/bnxt: fix clear port stats
>>>>> 3c2e87c10  Ajit Khaparde    net/bnxt: fix close operation
>>>>> 90362583e  Ajit Khaparde    net/bnxt: fix incorrect IO address handling in Tx
>>>>> 812bb2087  Andrew Rybchenko net/sfc: handle unknown L3 packet class in EF10 event parser
>>>>> 2a4ed7233  Damjan Marion    net/i40e: do not reset device info data
>>>>> 64cb90f88  Fiona Trahe      crypto/qat: fix checks for 3GPP algo bit params
>>>>> 2a7e3d546  Hyong Youb Kim   net/enic: do not overwrite admin Tx queue limit
>>>>> e11411b52  Maxime Coquelin  vhost: fix missing increment of log cache count
>>>>> 79d098921  Moti Haimovsky   net/mlx5: fix build with old kernels
>>>>> c44fbc7cc  Nelio Laranjeiro net/mlx5: clean-up developer logs
>>>>> 179c7e893  Pavan Nikhilesh  mempool/octeontx: fix pool to aura mapping
>>>>> 196e9a486  Qi Zhang         eal: fix hotplug add and remove
>>>>> 9f32c7e7e  Rafal Kozik      net/ena: change memory type
>>>>> 46916aa17  Rafal Kozik      net/ena: check pointer before memset
>>>>> 41e59028d  Rafal Kozik      net/ena: set link speed as none
>>>>> 3abe17195  Rahul Lakkireddy net/cxgbe: fix Rx channel map and queue type
>>>>> 1fc3afdf7  Rasesh Mody      doc: update qede management firmware guide
>>>>> c6dd1eb8a  Rasesh Mody      net/qede: fix default extended VLAN offload config
>>>>> a16aef523  Rasesh Mody      net/qede: fix for devargs
>>>>> e46821e9f  Shahaf Shuler    net/mlx5: separate generic tunnel TSO from the standard one
>>>>> 496f15c2c  Shahed Shaikh    net/qede: fix incorrect link status update
>>>>> c7641841a  Shahed Shaikh    net/qede: fix unicast MAC address handling in VF
>>>>> 82dbef2c8  Somnath Kotur    net/bnxt: revert reset of L2 filter id
>>>>> 0989161b2  Tiwei Bie        vhost: release locks on RARP packet failure
>>>>> b2a4a69f7  Xiaoxin Peng     net/bnxt: fix Tx with multiple mbuf

Re: [dpdk-stable] please help backporting some patches to LTS release 17.11.4

[Shahaf Shuler]

Hi Koh,

Friday, July 27, 2018 5:40 AM, Yongseok Koh:
> Peng <xiaoxin.peng@broadcom.com>; Yongseok Koh
> <yskoh@mellanox.com>
> Subject: please help backporting some patches to LTS release 17.11.4
> 
> Hi commit authors (and maintainers),
> 
> I didn't manage to apply following commits from upstream to stable branch
> 17.11: conflict happens. I'm wondering can the authors check the following
> list and backport those patches belong to you?
> 
> FYI, branch 17.11 is located at tree:
>    git://dpdk.org/dpdk-stable
> 
> It'd be great if you could do that in one or two weeks. Also, please add a
> heading line like below before the commit log body:
>     [ backported from upstream commit xxx ]
> 
> Example: http://dpdk.org/browse/dpdk-
> stable/commit/?h=16.07&id=c4831394c7d1944d8ec27d52c22997f20d19718e
> 
> Also please mention the target LTS in the subject line, as we have more than
> one at the same time, for example:
> 
>     [PATCH 17.11] foo/bar: fix baz
> 
> With git send-email, this can be achieved by appending the parameter:
> 
>     --subject-prefix='17.11'
> 
> Please let me know if you have any comments, say, need more time, or it's
> worthless to packport it. And please send it to "stable@dpdk.org", but not
> "dev@dpdk.org".
> 
> Thanks.
> 
> Yongseok

The below one is not relevant for 17.11, you can drop it.

> e46821e9f  Shahaf Shuler    net/mlx5: separate generic tunnel TSO from the
> standard one

[dpdk-stable] please help backporting some patches to LTS release 17.11.4

[Yongseok Koh]

Hi commit authors (and maintainers),

I didn't manage to apply following commits from upstream to stable branch
17.11: conflict happens. I'm wondering can the authors check the following
list and backport those patches belong to you?

FYI, branch 17.11 is located at tree:
   git://dpdk.org/dpdk-stable

It'd be great if you could do that in one or two weeks. Also, please add a
heading line like below before the commit log body:
    [ backported from upstream commit xxx ]

Example: http://dpdk.org/browse/dpdk-stable/commit/?h=16.07&id=c4831394c7d1944d8ec27d52c22997f20d19718e

Also please mention the target LTS in the subject line, as we have more than one
at the same time, for example:

    [PATCH 17.11] foo/bar: fix baz

With git send-email, this can be achieved by appending the parameter:

    --subject-prefix='17.11'

Please let me know if you have any comments, say, need more time, or it's
worthless to packport it. And please send it to "stable@dpdk.org", but not
"dev@dpdk.org".

Thanks.

Yongseok

---
22dcd9a4d  Anatoly Burakov  test: parallelize unit tests
33310b592  Beilei Xing      net/i40e: fix device parameter parsing
94110dc0d  Gavin Hu         maintainers: claim maintainership for ARM v7 and v8
9eabcb682  Jananee Parthasarathy test: update autotest list
4278f8df4  Matan Azrad      net/tap: fix zeroed flow mask configurations
ff9fe66c9  Moti Haimovsky   net/mlx4: advertise Rx jumbo frame support
2b4e423fd  Moti Haimovsky   net/mlx4: check RSS queues number limitation
c8dbf6814  Rasesh Mody      net/qede/base: fix to clear HW indication
8224b9d68  Reshma Pattan    latency: free up the memzone