[RFC,00/12] introduce s390x architecture
mbox series

Message ID 20190409190630.31975-1-vivkong@ca.ibm.com
Headers show
Series
  • introduce s390x architecture
Related show

Message

Vivian Kong April 9, 2019, 7:06 p.m. UTC
To build and run DPDK on Linux on IBM Z (s390x), big endian support
is added and s390x vector intrinics are used in the implementation 
of DPDK libraries.

I appreciate your feedback.

Vivian Kong (12):
  mk: introduce s390x architecture
  eal: add support for s390x architecture
  acl: add support for s390x architecture
  lpm: add support for s390x architecture
  examples/l3fwd: add support for s390x architecture
  net/i40e: add support for s390x architecture
  test: add support for s390x architecture
  hash: add support for s390x architecture
  doc: introduce s390x architecture
  ethdev: add cast for bus_device
  mbuf: trivial fix
  test: miscellaneous test fixes

 MAINTAINERS                                   |   9 +
 app/test-acl/main.c                           |   4 +
 app/test/autotest_data.py                     |   2 +-
 app/test/test_cmdline.c                       |   6 +-
 app/test/test_cmdline_ipaddr.c                |  13 +-
 app/test/test_cmdline_num.c                   | 111 +++
 app/test/test_xmmt_ops.h                      |  14 +
 config/defconfig_s390x-native-linux-gcc       |   1 +
 config/defconfig_s390x-native-linuxapp-gcc    |  33 +
 doc/guides/nics/features/i40e.ini             |   1 +
 doc/guides/nics/features/i40e_vec.ini         |   1 +
 doc/guides/rel_notes/release_19_05.rst        |   3 +
 drivers/net/i40e/Makefile                     |   2 +
 drivers/net/i40e/i40e_rxtx_vec_s390x.c        | 631 ++++++++++++++++++
 examples/l3fwd/l3fwd_em.c                     |   8 +
 examples/l3fwd/l3fwd_lpm_s390x.h              | 137 ++++
 examples/l3fwd/l3fwd_s390x.h                  | 259 +++++++
 lib/librte_acl/Makefile                       |   2 +
 lib/librte_acl/acl_bld.c                      |  69 +-
 lib/librte_acl/acl_gen.c                      |   9 +
 lib/librte_acl/acl_run_scalar.c               |   8 +
 lib/librte_acl/rte_acl.c                      |   4 +
 lib/librte_acl/rte_acl.h                      |   1 +
 .../common/arch/s390x/rte_cpuflags.c          |  90 +++
 lib/librte_eal/common/arch/s390x/rte_cycles.c |  11 +
 .../common/arch/s390x/rte_hypervisor.c        |  11 +
 .../common/include/arch/s390x/rte_atomic.h    |  47 ++
 .../common/include/arch/s390x/rte_byteorder.h |  43 ++
 .../common/include/arch/s390x/rte_cpuflags.h  |  42 ++
 .../common/include/arch/s390x/rte_cycles.h    |  44 ++
 .../common/include/arch/s390x/rte_io.h        |  18 +
 .../common/include/arch/s390x/rte_memcpy.h    |  55 ++
 .../common/include/arch/s390x/rte_pause.h     |  22 +
 .../common/include/arch/s390x/rte_prefetch.h  |  40 ++
 .../common/include/arch/s390x/rte_rwlock.h    |  42 ++
 .../common/include/arch/s390x/rte_spinlock.h  |  85 +++
 .../include/arch/s390x/rte_ticketlock.h       |  18 +
 .../common/include/arch/s390x/rte_vect.h      |  33 +
 lib/librte_ethdev/rte_ethdev_pci.h            |   2 +-
 lib/librte_hash/rte_fbk_hash.h                |   7 +
 lib/librte_lpm/Makefile                       |   2 +
 lib/librte_lpm/meson.build                    |   2 +-
 lib/librte_lpm/rte_lpm.h                      |   2 +
 lib/librte_lpm/rte_lpm6.c                     |  26 +-
 lib/librte_lpm/rte_lpm_s390x.h                | 130 ++++
 lib/librte_mbuf/rte_mbuf.h                    |   2 +-
 mk/arch/s390x/rte.vars.mk                     |  16 +
 mk/rte.cpuflags.mk                            |   5 +
 48 files changed, 2099 insertions(+), 24 deletions(-)
 create mode 120000 config/defconfig_s390x-native-linux-gcc
 create mode 100644 config/defconfig_s390x-native-linuxapp-gcc
 create mode 100644 drivers/net/i40e/i40e_rxtx_vec_s390x.c
 create mode 100644 examples/l3fwd/l3fwd_lpm_s390x.h
 create mode 100644 examples/l3fwd/l3fwd_s390x.h
 create mode 100644 lib/librte_eal/common/arch/s390x/rte_cpuflags.c
 create mode 100644 lib/librte_eal/common/arch/s390x/rte_cycles.c
 create mode 100644 lib/librte_eal/common/arch/s390x/rte_hypervisor.c
 create mode 100644 lib/librte_eal/common/include/arch/s390x/rte_atomic.h
 create mode 100644 lib/librte_eal/common/include/arch/s390x/rte_byteorder.h
 create mode 100644 lib/librte_eal/common/include/arch/s390x/rte_cpuflags.h
 create mode 100644 lib/librte_eal/common/include/arch/s390x/rte_cycles.h
 create mode 100644 lib/librte_eal/common/include/arch/s390x/rte_io.h
 create mode 100644 lib/librte_eal/common/include/arch/s390x/rte_memcpy.h
 create mode 100644 lib/librte_eal/common/include/arch/s390x/rte_pause.h
 create mode 100644 lib/librte_eal/common/include/arch/s390x/rte_prefetch.h
 create mode 100644 lib/librte_eal/common/include/arch/s390x/rte_rwlock.h
 create mode 100644 lib/librte_eal/common/include/arch/s390x/rte_spinlock.h
 create mode 100644 lib/librte_eal/common/include/arch/s390x/rte_ticketlock.h
 create mode 100644 lib/librte_eal/common/include/arch/s390x/rte_vect.h
 create mode 100644 lib/librte_lpm/rte_lpm_s390x.h
 create mode 100644 mk/arch/s390x/rte.vars.mk

Comments

Thomas Monjalon April 9, 2019, 8:25 p.m. UTC | #1
09/04/2019 21:06, Vivian Kong:
> To build and run DPDK on Linux on IBM Z (s390x), big endian support
> is added and s390x vector intrinics are used in the implementation 
> of DPDK libraries.
> 
> I appreciate your feedback.

I understand you worked on it during some time and you reveal it just now.
I must say thank you for the work.
It would have been better to notify the community in advance so we would
have discussed about the need for this work before you start.

First question: What is the use case of DPDK on s390x?

Second question: How is it tested? Do you have any CI?

Third question: Do you have time to dedicate to DPDK?
If yes, why not helping with IBM Power?

The reason for this last question is that the support for IBM Power
is still beta. I feel unconfortable merging one more IBM arch while
the first one does not receive enough support and testing.
I think running a CI for IBM Power would be a good first step.
Vivian Kong April 10, 2019, 5:09 p.m. UTC | #2
Hi Thomas,

Thanks for your comments.  Please see my answers to your questions below.

> First question: What is the use case of DPDK on s390x?

IBM would like to use DPDK in our new IBM cloud offering on Z.



> Second question: How is it tested? Do you have any CI?

We ran the tests by "make tests".  We don't have CI set up at the moment
but this is definitely something we are looking into adding.



> Third question: Do you have time to dedicate to DPDK?

After this initial port, we do plan to keep currency base on need but how
frequent will be determined by usage.



> If yes, why not helping with IBM Power?

Our work is specific to Z offering, we'll be interested to run CI for Z
also which will help maintaining our currency.


How is CI set up on the platforms currently supported in DPDK?  It looks
like a build gets run after a patch is submitted.  Wondering if there are
any documentation on how this is set up?

Thanks.


Regards,

Vivian Kong
Linux on IBM Z Open Source Ecosystem
IBM Canada Toronto Lab

Thomas Monjalon <thomas@monjalon.net> wrote on 2019/04/09 04:25:30 PM:

> From: Thomas Monjalon <thomas@monjalon.net>
> To: vivkong@ca.ibm.com
> Cc: dev@dpdk.org
> Date: 2019/04/09 04:25 PM
> Subject: Re: [dpdk-dev] [RFC 00/12] introduce s390x architecture
>
> 09/04/2019 21:06, Vivian Kong:
> > To build and run DPDK on Linux on IBM Z (s390x), big endian support
> > is added and s390x vector intrinics are used in the implementation
> > of DPDK libraries.
> >
> > I appreciate your feedback.
>
> I understand you worked on it during some time and you reveal it just
now.
> I must say thank you for the work.
> It would have been better to notify the community in advance so we would
> have discussed about the need for this work before you start.
>
> First question: What is the use case of DPDK on s390x?
>
> Second question: How is it tested? Do you have any CI?
>
> Third question: Do you have time to dedicate to DPDK?
> If yes, why not helping with IBM Power?
>
> The reason for this last question is that the support for IBM Power
> is still beta. I feel unconfortable merging one more IBM arch while
> the first one does not receive enough support and testing.
> I think running a CI for IBM Power would be a good first step.
>
>
>
Thomas Monjalon April 10, 2019, 5:43 p.m. UTC | #3
10/04/2019 19:09, Vivian Kong:
> Hi Thomas,
> 
> Thanks for your comments.  Please see my answers to your questions below.
> 
> > First question: What is the use case of DPDK on s390x?
> 
> IBM would like to use DPDK in our new IBM cloud offering on Z.
> 
> 
> 
> > Second question: How is it tested? Do you have any CI?
> 
> We ran the tests by "make tests".  We don't have CI set up at the moment
> but this is definitely something we are looking into adding.
> 
> 
> 
> > Third question: Do you have time to dedicate to DPDK?
> 
> After this initial port, we do plan to keep currency base on need but how
> frequent will be determined by usage.
> 
> 
> 
> > If yes, why not helping with IBM Power?
> 
> Our work is specific to Z offering, we'll be interested to run CI for Z
> also which will help maintaining our currency.
> 
> How is CI set up on the platforms currently supported in DPDK?  It looks
> like a build gets run after a patch is submitted.  Wondering if there are
> any documentation on how this is set up?

There is no complete documentation, but we should build one.
That would be great if you could help while riding this path.

Some starting points:
	https://patches.dpdk.org
	http://git.dpdk.org/tools/dpdk-ci/
	http://mails.dpdk.org/listinfo/ci

Please help with IBM Power CI at the same time.
I would like to see some progress on this front
before integrating a new IBM CPU.
Vivian Kong April 10, 2019, 7:15 p.m. UTC | #4
Thanks for the links! I'll go through them.
                                                                                                                                                
 For Power CI, not sure how we can help as it is done by another team.  If s390x is added, we plan on maintaining currency and setting up CI on 
 s390x will help do that.                                                                                                                       
                                                                                                                                                



Regards,

Vivian Kong
Linux on IBM Z Open Source Ecosystem
IBM Canada Toronto Lab

Thomas Monjalon <thomas@monjalon.net> wrote on 2019/04/10 01:43:34 PM:

> From: Thomas Monjalon <thomas@monjalon.net>
> To: Vivian Kong <vivkong@ca.ibm.com>
> Cc: dev@dpdk.org
> Date: 2019/04/10 01:43 PM
> Subject: Re: [dpdk-dev] [RFC 00/12] introduce s390x architecture
>
> 10/04/2019 19:09, Vivian Kong:
> > Hi Thomas,
> >
> > Thanks for your comments.  Please see my answers to your questions
below.
> >
> > > First question: What is the use case of DPDK on s390x?
> >
> > IBM would like to use DPDK in our new IBM cloud offering on Z.
> >
> >
> >
> > > Second question: How is it tested? Do you have any CI?
> >
> > We ran the tests by "make tests".  We don't have CI set up at the
moment
> > but this is definitely something we are looking into adding.
> >
> >
> >
> > > Third question: Do you have time to dedicate to DPDK?
> >
> > After this initial port, we do plan to keep currency base on need but
how
> > frequent will be determined by usage.
> >
> >
> >
> > > If yes, why not helping with IBM Power?
> >
> > Our work is specific to Z offering, we'll be interested to run CI for Z
> > also which will help maintaining our currency.
> >
> > How is CI set up on the platforms currently supported in DPDK?  It
looks
> > like a build gets run after a patch is submitted.  Wondering if there
are
> > any documentation on how this is set up?
>
> There is no complete documentation, but we should build one.
> That would be great if you could help while riding this path.
>
> Some starting points:
>    https://urldefense.proofpoint.com/v2/url?
> u=https-3A__patches.dpdk.org&d=DwICAg&c=jf_iaSHvJObTbx-
>
siA1ZOg&r=t7gXFFEkLvveGPwCXa29Q41okXMOY7sJK43BPFcHaRY&m=VchTLEmGgUOXsp5m5hZPa8aCYMnWpA3V_TmZGGW995I&s=Y9d-

> qW-nAewmuq-Op0l7kKOtq7hzIC3Wm-qcdkqQO8E&e=
>    https://urldefense.proofpoint.com/v2/url?
> u=http-3A__git.dpdk.org_tools_dpdk-2Dci_&d=DwICAg&c=jf_iaSHvJObTbx-
>
siA1ZOg&r=t7gXFFEkLvveGPwCXa29Q41okXMOY7sJK43BPFcHaRY&m=VchTLEmGgUOXsp5m5hZPa8aCYMnWpA3V_TmZGGW995I&s=cP_hEZmUedWHOcXUF-

> TTShw8UprsMFFrYhFyqp_mWHw&e=
>    https://urldefense.proofpoint.com/v2/url?
> u=http-3A__mails.dpdk.org_listinfo_ci&d=DwICAg&c=jf_iaSHvJObTbx-
>
siA1ZOg&r=t7gXFFEkLvveGPwCXa29Q41okXMOY7sJK43BPFcHaRY&m=VchTLEmGgUOXsp5m5hZPa8aCYMnWpA3V_TmZGGW995I&s=WWm6B32WFdHyVc3CAuMwfNsIZEDpPeAhy9MJRKnKXow&e=

>
> Please help with IBM Power CI at the same time.
> I would like to see some progress on this front
> before integrating a new IBM CPU.
>
>
>
Thomas Monjalon April 10, 2019, 7:46 p.m. UTC | #5
10/04/2019 21:15, Vivian Kong:
> Thomas Monjalon <thomas@monjalon.net> wrote on 2019/04/10 01:43:34 PM:
> > Please help with IBM Power CI at the same time.
> > I would like to see some progress on this front
> > before integrating a new IBM CPU.
> 
>  For Power CI, not sure how we can help as it is done by another team.
>  If s390x is added, we plan on maintaining currency and setting up CI on 
>  s390x will help do that.                                                                                                                       

You have the same objective as another team in your company,
so there is probably some room to collaborate.
You are probably in better position to help with CI on Power than
anyone else in the community.
I know it is a different team, but think about other contributors
who are not even in the same company.
David Christensen April 10, 2019, 8:31 p.m. UTC | #6
>>> Second question: How is it tested? Do you have any CI?
>>
>> We ran the tests by "make tests".  We don't have CI set up at the moment
>> but this is definitely something we are looking into adding.
>>
> Please help with IBM Power CI at the same time.
> I would like to see some progress on this front
> before integrating a new IBM CPU.

The IBM Power team is not idle in this area.  There's ongoing work to 
enable DTS on Power systems.

Dave
Vivian Kong April 10, 2019, 8:45 p.m. UTC | #7
Sorry what I meant to say is that we have no control over what Power team
will do in terms on CI.  And yes I will definitely reach out to them to see
what their plans are.

Regards,

Vivian Kong
Linux on IBM Z Open Source Ecosystem
IBM Canada Toronto Lab

Thomas Monjalon <thomas@monjalon.net> wrote on 2019/04/10 03:46:12 PM:

> From: Thomas Monjalon <thomas@monjalon.net>
> To: Vivian Kong <vivkong@ca.ibm.com>
> Cc: dev@dpdk.org
> Date: 2019/04/10 03:46 PM
> Subject: Re: [dpdk-dev] [RFC 00/12] introduce s390x architecture
>
> 10/04/2019 21:15, Vivian Kong:
> > Thomas Monjalon <thomas@monjalon.net> wrote on 2019/04/10 01:43:34 PM:
> > > Please help with IBM Power CI at the same time.
> > > I would like to see some progress on this front
> > > before integrating a new IBM CPU.
> >
> >  For Power CI, not sure how we can help as it is done by another team.
> >  If s390x is added, we plan on maintaining currency and setting up CI
on
> >  s390x will help do that.

>
> You have the same objective as another team in your company,
> so there is probably some room to collaborate.
> You are probably in better position to help with CI on Power than
> anyone else in the community.
> I know it is a different team, but think about other contributors
> who are not even in the same company.
>
>
>
Thomas Monjalon April 10, 2019, 8:46 p.m. UTC | #8
10/04/2019 22:31, David Christensen:
> >>> Second question: How is it tested? Do you have any CI?
> >>
> >> We ran the tests by "make tests".  We don't have CI set up at the moment
> >> but this is definitely something we are looking into adding.
> >>
> > Please help with IBM Power CI at the same time.
> > I would like to see some progress on this front
> > before integrating a new IBM CPU.
> 
> The IBM Power team is not idle in this area.  There's ongoing work to 
> enable DTS on Power systems.

Good to read there is some work in the pipe.
I can remember the first emails (in 2015) talking about some automated
tests for compilation and unit tests on IBM Power.
We are still seeing some compilation issues regularly on Power.
Before going to DTS, I think you should extend your compilation testing,
and basic feature testing with builtin unit tests.
David Christensen April 10, 2019, 9:03 p.m. UTC | #9
>> The IBM Power team is not idle in this area.  There's ongoing work to
>> enable DTS on Power systems.
> 
> Good to read there is some work in the pipe.
> I can remember the first emails (in 2015) talking about some automated
> tests for compilation and unit tests on IBM Power.
> We are still seeing some compilation issues regularly on Power.
> Before going to DTS, I think you should extend your compilation testing,
> and basic feature testing with builtin unit tests.

That's on my ToDo list.  I'm currently investigating some errors from 
app/test/test_barrier.c which fails on weakly ordered memory systems 
such as Power.  I have my eye on several other tests as well that 
definitely need to be validated.

Dave
Pradeep Satyanarayana April 10, 2019, 9:45 p.m. UTC | #10
> > The IBM Power team is not idle in this area.  There's ongoing work to
> > enable DTS on Power systems.
>
> Good to read there is some work in the pipe.
> I can remember the first emails (in 2015) talking about some automated
> tests for compilation and unit tests on IBM Power.

Thomas, Unfortunately we previously had some churn in the teams that worked
on DPDK
on Power. That had a downstream impact. We (Power) should be in better
shape than we
were in the past. For the last several weeks, we have been regularly
running a sub-set of
DTS with unit tests, sample  applications and so on against the master
branch. We are also
periodically running these tests against several of the  stable branches as
well.

> We are still seeing some compilation issues regularly on Power.
> Before going to DTS, I think you should extend your compilation testing,
> and basic feature testing with builtin unit tests.

We have not observed any compilation issues since we started the CI. The
last issue reported
was during 18.11-rc1 that Mellanox fixed. Hopefully, we should catch any
such regressions
quickly in the future, since we now have a CI that runs periodically.

On a different note, I will reach out to Vivian and the s390 team.

Thanks
Pradeep
pradeep@us.ibm.com
Thomas Monjalon April 11, 2019, 6:59 a.m. UTC | #11
10/04/2019 23:45, Pradeep Satyanarayana:
> 
> > > The IBM Power team is not idle in this area.  There's ongoing work to
> > > enable DTS on Power systems.
> >
> > Good to read there is some work in the pipe.
> > I can remember the first emails (in 2015) talking about some automated
> > tests for compilation and unit tests on IBM Power.
> 
> Thomas, Unfortunately we previously had some churn in the teams that worked
> on DPDK
> on Power. That had a downstream impact. We (Power) should be in better
> shape than we
> were in the past. For the last several weeks, we have been regularly
> running a sub-set of
> DTS with unit tests, sample  applications and so on against the master
> branch. We are also
> periodically running these tests against several of the  stable branches as
> well.

Good.
Next step will be to report some test result for each patchset submitted
on the mailing-list, so we can monitor the results from patchwork.

> > We are still seeing some compilation issues regularly on Power.
> > Before going to DTS, I think you should extend your compilation testing,
> > and basic feature testing with builtin unit tests.
> 
> We have not observed any compilation issues since we started the CI. The
> last issue reported
> was during 18.11-rc1 that Mellanox fixed. Hopefully, we should catch any
> such regressions
> quickly in the future, since we now have a CI that runs periodically.

There are some failure when compiling with big endian toolchain.
It seems you are not testing big endian.
If you don't support big endian, please mention it in the doc.

> On a different note, I will reach out to Vivian and the s390 team.

Thank you
David Marchand April 11, 2019, 7:45 a.m. UTC | #12
Hello,

On Tue, Apr 9, 2019 at 9:06 PM Vivian Kong <vivkong@gmail.com> wrote:

> To build and run DPDK on Linux on IBM Z (s390x), big endian support
> is added and s390x vector intrinics are used in the implementation
>

intrinsics*

of DPDK libraries.
>
> I appreciate your feedback.
>

Quick check since this is not 19.05 material.

- New files should be Copyright 2019.
- Avoid copy/paste from Power, noticed a reference to Power in the config
file.
- Configuration files are stacked, only enable/disable the needed parts in
the specific s390 file
- Meson must be added.
- No need to add dead code, remove the commented code like:
  //#define dsb(opt) asm volatile("" : : : "memory")
- Comments using // are banned.
- No need to introduce a RTE_ACL_CLASSIFY_S390X to map to the scalar code.
  Just avoid the SSE code to be selected in the build framework (iiuc).
- Avoid introducing empty lines


Run check scripts from devtools before submitting:

[dmarchan@dmarchan dpdk]$ ./devtools/checkpatches.sh

### mk: introduce s390x architecture

WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
author 'Vivian Kong <vivkong@gmail.com>'

total: 0 errors, 1 warnings, 61 lines checked

### eal: add support for s390x architecture

WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
author 'Vivian Kong <vivkong@gmail.com>'

total: 0 errors, 1 warnings, 601 lines checked

### acl: add support for s390x architecture

WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
author 'Vivian Kong <vivkong@gmail.com>'

total: 0 errors, 1 warnings, 208 lines checked

### lpm: add support for s390x architecture

WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
author 'Vivian Kong <vivkong@gmail.com>'

total: 0 errors, 1 warnings, 203 lines checked

### examples/l3fwd: add support for s390x architecture

WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
author 'Vivian Kong <vivkong@gmail.com>'

total: 0 errors, 1 warnings, 410 lines checked

### net/i40e: add support for s390x architecture

WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
author 'Vivian Kong <vivkong@gmail.com>'

total: 0 errors, 1 warnings, 0 checks, 647 lines checked

### test: add support for s390x architecture

WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
author 'Vivian Kong <vivkong@gmail.com>'

total: 0 errors, 1 warnings, 192 lines checked

### hash: add support for s390x architecture

WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
author 'Vivian Kong <vivkong@gmail.com>'

total: 0 errors, 1 warnings, 16 lines checked

### doc: introduce s390x architecture

WARNING:COMMIT_MESSAGE: Missing commit description - Add an appropriate one

WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
author 'Vivian Kong <vivkong@gmail.com>'

total: 0 errors, 2 warnings, 24 lines checked

### ethdev: add cast for bus_device

WARNING:COMMIT_MESSAGE: Missing commit description - Add an appropriate one

WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
author 'Vivian Kong <vivkong@gmail.com>'

total: 0 errors, 2 warnings, 8 lines checked

### mbuf: trivial fix

WARNING:COMMIT_MESSAGE: Missing commit description - Add an appropriate one

WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
author 'Vivian Kong <vivkong@gmail.com>'

total: 0 errors, 2 warnings, 8 lines checked

### test: miscellaneous test fixes

WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
author 'Vivian Kong <vivkong@gmail.com>'

total: 0 errors, 1 warnings, 32 lines checked

0/12 valid patch
Pradeep Satyanarayana April 11, 2019, 3:57 p.m. UTC | #13
> > > We are still seeing some compilation issues regularly on Power.
> > > Before going to DTS, I think you should extend your compilation
testing,
> > > and basic feature testing with builtin unit tests.
> >
> > We have not observed any compilation issues since we started the CI.
The
> > last issue reported
> > was during 18.11-rc1 that Mellanox fixed. Hopefully, we should catch
any
> > such regressions
> > quickly in the future, since we now have a CI that runs periodically.
>
> There are some failure when compiling with big endian toolchain.
> It seems you are not testing big endian.

Correct, we have been doing all our work on little endian only.

> If you don't support big endian, please mention it in the doc.

Thank you for alerting us about this. Good point that you bring up.
I will need some time to consult and figure out how we approach this.
Let me get back to you on this.

Thanks
Pradeep
pradeep@us.ibm.com
Pradeep Satyanarayana April 18, 2019, 9:06 p.m. UTC | #14
Pradeep Satyanarayana/Beaverton/IBM wrote on 04/11/2019 08:57:05 AM:

> From: Pradeep Satyanarayana/Beaverton/IBM
> To: Thomas Monjalon <thomas@monjalon.net>
> Cc: dev@dpdk.org, David Christensen <drc@linux.vnet.ibm.com>, Vivian
> Kong <vivkong@ca.ibm.com>, David Wilder <wilder@us.ibm.com>
> Date: 04/11/2019 08:57 AM
> Subject: Re: [dpdk-dev] [RFC 00/12] introduce s390x architecture
>
>
> > > > We are still seeing some compilation issues regularly on Power.
> > > > Before going to DTS, I think you should extend your compilation
testing,
> > > > and basic feature testing with builtin unit tests.
> > >
> > > We have not observed any compilation issues since we started the CI.
The
> > > last issue reported
> > > was during 18.11-rc1 that Mellanox fixed. Hopefully, we should catch
any
> > > such regressions
> > > quickly in the future, since we now have a CI that runs periodically.
> >
> > There are some failure when compiling with big endian toolchain.
> > It seems you are not testing big endian.
>
> Correct, we have been doing all our work on little endian only.
>
> > If you don't support big endian, please mention it in the doc.
>
> Thank you for alerting us about this. Good point that you bring up.
> I will need some time to consult and figure out how we approach this.
> Let me get back to you on this.

On the Linux on Power front going forward, there are currently no plans to
support BE,
just LE.

The doc you mention above, is that this one?

https://doc.dpdk.org/guides/platform/index.html

Thanks
Pradeep
pradeep@us.ibm.com
Thomas Monjalon April 18, 2019, 9:18 p.m. UTC | #15
18/04/2019 23:06, Pradeep Satyanarayana:
> 
> Pradeep Satyanarayana/Beaverton/IBM wrote on 04/11/2019 08:57:05 AM:
> 
> > From: Pradeep Satyanarayana/Beaverton/IBM
> > To: Thomas Monjalon <thomas@monjalon.net>
> > Cc: dev@dpdk.org, David Christensen <drc@linux.vnet.ibm.com>, Vivian
> > Kong <vivkong@ca.ibm.com>, David Wilder <wilder@us.ibm.com>
> > Date: 04/11/2019 08:57 AM
> > Subject: Re: [dpdk-dev] [RFC 00/12] introduce s390x architecture
> >
> >
> > > > > We are still seeing some compilation issues regularly on Power.
> > > > > Before going to DTS, I think you should extend your compilation
> testing,
> > > > > and basic feature testing with builtin unit tests.
> > > >
> > > > We have not observed any compilation issues since we started the CI.
> The
> > > > last issue reported
> > > > was during 18.11-rc1 that Mellanox fixed. Hopefully, we should catch
> any
> > > > such regressions
> > > > quickly in the future, since we now have a CI that runs periodically.
> > >
> > > There are some failure when compiling with big endian toolchain.
> > > It seems you are not testing big endian.
> >
> > Correct, we have been doing all our work on little endian only.
> >
> > > If you don't support big endian, please mention it in the doc.
> >
> > Thank you for alerting us about this. Good point that you bring up.
> > I will need some time to consult and figure out how we approach this.
> > Let me get back to you on this.
> 
> On the Linux on Power front going forward, there are currently no plans to
> support BE,
> just LE.
> 
> The doc you mention above, is that this one?
> 
> https://doc.dpdk.org/guides/platform/index.html

Yes, thank you.
David Marchand May 15, 2019, 12:45 p.m. UTC | #16
Hello,

On Thu, Apr 11, 2019 at 9:45 AM David Marchand <david.marchand@redhat.com>
wrote:

> Hello,
>
> On Tue, Apr 9, 2019 at 9:06 PM Vivian Kong <vivkong@gmail.com> wrote:
>
>> To build and run DPDK on Linux on IBM Z (s390x), big endian support
>> is added and s390x vector intrinics are used in the implementation
>>
>
> intrinsics*
>
> of DPDK libraries.
>>
>> I appreciate your feedback.
>>
>
> Quick check since this is not 19.05 material.
>
> - New files should be Copyright 2019.
> - Avoid copy/paste from Power, noticed a reference to Power in the config
> file.
> - Configuration files are stacked, only enable/disable the needed parts in
> the specific s390 file
> - Meson must be added.
> - No need to add dead code, remove the commented code like:
>   //#define dsb(opt) asm volatile("" : : : "memory")
> - Comments using // are banned.
> - No need to introduce a RTE_ACL_CLASSIFY_S390X to map to the scalar code.
>   Just avoid the SSE code to be selected in the build framework (iiuc).
> - Avoid introducing empty lines
>
>
> Run check scripts from devtools before submitting:
>
> [dmarchan@dmarchan dpdk]$ ./devtools/checkpatches.sh
>
> ### mk: introduce s390x architecture
>
> WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
> author 'Vivian Kong <vivkong@gmail.com>'
>
> total: 0 errors, 1 warnings, 61 lines checked
>
> ### eal: add support for s390x architecture
>
> WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
> author 'Vivian Kong <vivkong@gmail.com>'
>
> total: 0 errors, 1 warnings, 601 lines checked
>
> ### acl: add support for s390x architecture
>
> WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
> author 'Vivian Kong <vivkong@gmail.com>'
>
> total: 0 errors, 1 warnings, 208 lines checked
>
> ### lpm: add support for s390x architecture
>
> WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
> author 'Vivian Kong <vivkong@gmail.com>'
>
> total: 0 errors, 1 warnings, 203 lines checked
>
> ### examples/l3fwd: add support for s390x architecture
>
> WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
> author 'Vivian Kong <vivkong@gmail.com>'
>
> total: 0 errors, 1 warnings, 410 lines checked
>
> ### net/i40e: add support for s390x architecture
>
> WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
> author 'Vivian Kong <vivkong@gmail.com>'
>
> total: 0 errors, 1 warnings, 0 checks, 647 lines checked
>
> ### test: add support for s390x architecture
>
> WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
> author 'Vivian Kong <vivkong@gmail.com>'
>
> total: 0 errors, 1 warnings, 192 lines checked
>
> ### hash: add support for s390x architecture
>
> WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
> author 'Vivian Kong <vivkong@gmail.com>'
>
> total: 0 errors, 1 warnings, 16 lines checked
>
> ### doc: introduce s390x architecture
>
> WARNING:COMMIT_MESSAGE: Missing commit description - Add an appropriate one
>
> WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
> author 'Vivian Kong <vivkong@gmail.com>'
>
> total: 0 errors, 2 warnings, 24 lines checked
>
> ### ethdev: add cast for bus_device
>
> WARNING:COMMIT_MESSAGE: Missing commit description - Add an appropriate one
>
> WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
> author 'Vivian Kong <vivkong@gmail.com>'
>
> total: 0 errors, 2 warnings, 8 lines checked
>
> ### mbuf: trivial fix
>
> WARNING:COMMIT_MESSAGE: Missing commit description - Add an appropriate one
>
> WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
> author 'Vivian Kong <vivkong@gmail.com>'
>
> total: 0 errors, 2 warnings, 8 lines checked
>
> ### test: miscellaneous test fixes
>
> WARNING:NO_AUTHOR_SIGN_OFF: Missing Signed-off-by: line by nominal patch
> author 'Vivian Kong <vivkong@gmail.com>'
>
> total: 0 errors, 1 warnings, 32 lines checked
>
> 0/12 valid patch
>

We are now post 19.05.
This is the right time to propose new features/architectures.
If you want to post a new series, please take into account the previous
comments.

Thanks.