RE: [PATCH v3] net/iavf: enable Tx outer checksum offload on avx512

[Zhang, Qi Z]

> -Original Message-
> From: Xu, Ke1 
> Sent: Friday, February 17, 2023 9:51 AM
> To: Zeng, ZhichaoX ; dev@dpdk.org
> Cc: Yang, Qiming ; Zhou, YidingX
> ; Zhang, Qi Z ; Wu, Jingjing
> ; Xing, Beilei ; Richardson,
> Bruce ; Konstantin Ananyev
> 
> Subject: RE: [PATCH v3] net/iavf: enable Tx outer checksum offload on
> avx512
> 
> 
> 
> > -Original Message-
> > From: Zeng, ZhichaoX 
> > Sent: Friday, February 17, 2023 9:49 AM
> > To: dev@dpdk.org
> > Cc: Yang, Qiming ; Zhou, YidingX
> > ; Zhang, Qi Z ; Xu, Ke1
> > ; Zeng, ZhichaoX ; Wu,
> > Jingjing ; Xing, Beilei
> > ; Richardson, Bruce
> > ; Konstantin Ananyev
> > 
> > Subject: [PATCH v3] net/iavf: enable Tx outer checksum offload on
> > avx512
> >
> > This patch is to enable outer checksum offload on avx512 Tx path for
> > tunnel packet by adding Tx path with context descriptor and adjusting
> > path select logic.
> >
> > Signed-off-by: Yiding Zhou 
> > Signed-off-by: Zhichao Zeng 
> 
> V3 patch validated and passed.
> 
> Tested-by: Ke Xu 

replaced V2 in dpdk-next-net-intel.

RE: [PATCH v3] net/iavf: add debug checks for mbuf

[Zhang, Qi Z]

> -Original Message-
> From: Mingjin Ye 
> Sent: Monday, February 13, 2023 4:32 PM
> To: dev@dpdk.org
> Cc: Yang, Qiming ; sta...@dpdk.org; Zhou, YidingX
> ; Ye, MingjinX ; Wu,
> Jingjing ; Xing, Beilei 
> Subject: [PATCH v3] net/iavf: add debug checks for mbuf
> 
> The scalar Tx path would send wrong mbuf that causes the kernel driver to
> fire the MDD event.
> 
> This patch adds mbuf detection in tx_prepare and enables it by defining
> `RTE_ETHDEV_DEBUG_TX` macro to fix this issue. If validation fails, rte_errno

This is a debug utility but not a fix, enable this check will significantly 
downgrade the performance
Removed  "to fix this issue"

> will be set to EINVAL and returned.
> 
> Fixes: 3fd32df381f8 ("net/iavf: check Tx packet with correct UP and queue")
> Fixes: 12b435bf8f2f ("net/iavf: support flex desc metadata extraction")
> Fixes: f28fbd1e6b50 ("net/iavf: check max SIMD bitwidth")

Removed all the fix lines

> Cc: sta...@dpdk.org
> 
> Signed-off-by: Mingjin Ye 

Acked-by: Qi Zhang 

Applied to dpdk-next-net-intel after apply above change.

Thanks
Qi

RE: [PATCH 1/2] net/ice/base: add ability to set markid via switch filter

[Zhang, Qi Z]

> -Original Message-
> From: Yan, Zhirun 
> Sent: Tuesday, February 14, 2023 1:24 PM
> To: dev@dpdk.org; Zhang, Qi Z 
> Cc: Yan, Zhirun 
> Subject: [PATCH 1/2] net/ice/base: add ability to set markid via switch filter
> 
> Support to add large action to set 32 bits markid via switch filter.
> 
> Signed-off-by: Zhirun Yan 

Acked-by: Qi Zhang 

Applied to dpdk-next-net-intel.

Thanks
Qi

RE: [PATCH 2/2] net/ice: support mark action in switch

[Zhang, Qi Z]

> -Original Message-
> From: Yan, Zhirun 
> Sent: Tuesday, February 14, 2023 1:24 PM
> To: dev@dpdk.org; Zhang, Qi Z 
> Cc: Yan, Zhirun 
> Subject: [PATCH 2/2] net/ice: support mark action in switch
> 
> This patch add mark action in switch filter.
> 
> Signed-off-by: Zhirun Yan 

Acked-by: Qi Zhang 

Applied to dpdk-next-net-intel.

Thanks
Qi

Re: Building DPDK with IOVA_AS_VA

[Thomas Monjalon]

09/12/2022 10:33, fengchengwen:
>   The hns3 and idpf both use static_library to build advanced vector (e.g. 
> SVE for hns3 and AVX512 for idpf).
> The static-library is still compiled even the PMD is not compiled, and this 
> lead to the compile fail problem.
> 
>   Although the following could solve the problem, but is there a better plan?
>  +if dpdk_conf.get('RTE_IOVA_AS_PA') == 0
>  +build = false
>  +reason = 'driver does not support disabling IOVA as PA mode'
>  +subdir_done()
>  +endif

Yes there is a better plan: fix drivers/meson.build.
We must move the "continue" to skip the driver as soon as possible.

Re: [PATCH v2] net: not build PMD AVX library when no IOVA as PA

[Thomas Monjalon]

29/01/2023 13:35, Zhang, Qi Z:
> From: Zhang, Qi Z 
> > 
> > PMD not announce pmd_supports_disable_iova_as_pa will not be build
> > when RTE_IOVA_AS_PA is not defined, but some AVX library for vector path
> > is not skipped by the build system which cause compile error.
> > 
> > The patch modify i40e, iavf, ice's meson file to skip AVX library build when
> > RTE_IOVA_AS_PA is not defined.
> > 
> > Cc: sta...@dpdk.org
> > 
> > Signed-off-by: Qi Zhang 
> 
> Move this for next-net review.
> 
> Acked-by: Qi Zhang 

You ack your own patch?

> Applied to dpdk-next-net-intel.

Sorry I don't pull this one, as I think there is a better fix:
we should not handle this option in each driver.
Instead the file drivers/meson.build must be fixed.

Re: [PATCH] net/idpf: fix build option check

[Thomas Monjalon]

17/02/2023 23:34, Thomas Monjalon:
> 13/12/2022 02:52, Zhang, Qi Z:
> > From: Xing, Beilei 
> > > From: Wu, Jingjing 
> > > > When enable_iova_as_pa option is disabled, idpf driver should avoid
> > > > the building in its build file.
> > > >
> > > > Fixes: 5bf87b45b2c8 (net/idpf: add AVX512 data path for single queue
> > > > model)
> > > > Cc: sta...@dpdk.org
> > > >
> > > > Signed-off-by: Jingjing Wu 
> > > > ---
> > > > --- a/drivers/net/idpf/meson.build
> > > > +++ b/drivers/net/idpf/meson.build
> > > > +if dpdk_conf.get('RTE_IOVA_AS_PA') == 0
> > > > +build = false
> > > > +reason = 'driver does not support disabling IOVA as PA mode'
> > > > +subdir_done()
> > > > +endif
> > > 
> > > Acked-by: Beilei Xing 
> > 
> > Applied to dpdk-next-net-intel.
> 
> I think this patch is not needed since we have this check done for all 
> drivers:
> 
> +pmd_supports_disable_iova_as_pa = false
> [...]
> +if dpdk_conf.get('RTE_IOVA_AS_PA') == 0 and not 
> pmd_supports_disable_iova_as_pa and not always_enable.contains(drv_path)
> +build = false
> +reason = 'driver does not support disabling IOVA as PA mode'
> +endif
> 
> I will skip this patch and mark it as Rejected in patchwork.
> Please confirm the decision is OK.

After more thoughts, I think we should only stop compilation in common/idpf:

+if not get_option('enable_iova_as_pa')
+subdir_done()
+endif

No need to give reason or disable build as it will be done anyway in 
drivers/meson.build.

Re: Building DPDK with IOVA_AS_VA

[Thomas Monjalon]

19/02/2023 11:01, Thomas Monjalon:
> 09/12/2022 10:33, fengchengwen:
> >   The hns3 and idpf both use static_library to build advanced vector (e.g. 
> > SVE for hns3 and AVX512 for idpf).
> > The static-library is still compiled even the PMD is not compiled, and this 
> > lead to the compile fail problem.
> > 
> >   Although the following could solve the problem, but is there a better 
> > plan?
> >  +if dpdk_conf.get('RTE_IOVA_AS_PA') == 0
> >  +build = false
> >  +reason = 'driver does not support disabling IOVA as PA mode'
> >  +subdir_done()
> >  +endif
> 
> Yes there is a better plan: fix drivers/meson.build.
> We must move the "continue" to skip the driver as soon as possible.

Scratch that, it's a stupid idea.
The problem is generating sub-libraries inside the driver
in order to map some specific compiler options with some files.
Looks like we must disable the build inside the driver:

+if not get_option('enable_iova_as_pa')
+subdir_done()
+endif

No need to give reason or disable build as it will be done anyway in 
drivers/meson.build.

Re: [PATCH v2] net: not build PMD AVX library when no IOVA as PA

[Thomas Monjalon]

19/02/2023 11:04, Thomas Monjalon:
> 29/01/2023 13:35, Zhang, Qi Z:
> > From: Zhang, Qi Z 
> > > 
> > > PMD not announce pmd_supports_disable_iova_as_pa will not be build
> > > when RTE_IOVA_AS_PA is not defined, but some AVX library for vector path
> > > is not skipped by the build system which cause compile error.
> > > 
> > > The patch modify i40e, iavf, ice's meson file to skip AVX library build 
> > > when
> > > RTE_IOVA_AS_PA is not defined.
> > > 
> > > Cc: sta...@dpdk.org
> > > 
> > > Signed-off-by: Qi Zhang 
> > 
> > Move this for next-net review.
> > 
> > Acked-by: Qi Zhang 
> 
> You ack your own patch?
> 
> > Applied to dpdk-next-net-intel.
> 
> Sorry I don't pull this one, as I think there is a better fix:
> we should not handle this option in each driver.
> Instead the file drivers/meson.build must be fixed.

No sorry, it cannot be handled in drivers/meson.build.

So I suggest disabling the whole driver:

+if not get_option('enable_iova_as_pa')
+subdir_done()
+endif

Note: no need to disable ice, as it is enabled in another commit.

[PATCH] drivers: skip build of sub-libs not supporting IOVA mode

[Thomas Monjalon]

If IOVA as PA is disabled and the driver does not support IOVA as VA,
the build of the driver was disabled.
Unfortunately some drivers were building some sub-libraries
(with specific options for vector paths) which were not disabled.

The build parsing of those drivers need to be skipped earlier
to avoid defining the sub-libraries.

Fixes: a986c2b7973d ("build: add option to configure IOVA mode as PA")
Cc: sta...@dpdk.org

Signed-off-by: Thomas Monjalon 
---
 drivers/common/idpf/meson.build | 4 
 drivers/event/dlb2/meson.build  | 5 -
 drivers/meson.build | 4 ++--
 drivers/net/bnxt/meson.build| 4 
 drivers/net/enic/meson.build| 4 
 drivers/net/hns3/meson.build| 4 +---
 drivers/net/i40e/meson.build| 4 
 drivers/net/iavf/meson.build| 3 +++
 drivers/net/virtio/meson.build  | 4 
 9 files changed, 30 insertions(+), 6 deletions(-)

diff --git a/drivers/common/idpf/meson.build b/drivers/common/idpf/meson.build
index 58059ef443..74c0e92cac 100644
--- a/drivers/common/idpf/meson.build
+++ b/drivers/common/idpf/meson.build
@@ -1,6 +1,10 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2022 Intel Corporation
 
+if not get_option('enable_iova_as_pa')
+subdir_done()
+endif
+
 deps += ['mbuf']
 
 sources = files(
diff --git a/drivers/event/dlb2/meson.build b/drivers/event/dlb2/meson.build
index a2e60273c5..8ae7db0011 100644
--- a/drivers/event/dlb2/meson.build
+++ b/drivers/event/dlb2/meson.build
@@ -1,4 +1,3 @@
-
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2019-2020 Intel Corporation
 
@@ -8,6 +7,10 @@ if not is_linux or not dpdk_conf.has('RTE_ARCH_X86_64')
 subdir_done()
 endif
 
+if not get_option('enable_iova_as_pa')
+subdir_done()
+endif
+
 sources = files(
 'dlb2.c',
 'dlb2_iface.c',
diff --git a/drivers/meson.build b/drivers/meson.build
index 0618c31a69..9c10e2633d 100644
--- a/drivers/meson.build
+++ b/drivers/meson.build
@@ -127,9 +127,9 @@ foreach subpath:subdirs
 # pull in driver directory which should update all the local 
variables
 subdir(drv_path)
 
-if dpdk_conf.get('RTE_IOVA_AS_PA') == 0 and not 
pmd_supports_disable_iova_as_pa and not always_enable.contains(drv_path)
+if not get_option('enable_iova_as_pa') and not 
pmd_supports_disable_iova_as_pa and not always_enable.contains(drv_path)
 build = false
-reason = 'driver does not support disabling IOVA as PA mode'
+reason = 'IOVA as VA not supported'
 endif
 
 # get dependency objs from strings
diff --git a/drivers/net/bnxt/meson.build b/drivers/net/bnxt/meson.build
index 09d494e90f..c9ca09089a 100644
--- a/drivers/net/bnxt/meson.build
+++ b/drivers/net/bnxt/meson.build
@@ -8,6 +8,10 @@ if is_windows
 subdir_done()
 endif
 
+if not get_option('enable_iova_as_pa')
+subdir_done()
+endif
+
 headers = files('rte_pmd_bnxt.h')
 cflags_options = [
 '-DSUPPORT_CFA_HW_ALL=1',
diff --git a/drivers/net/enic/meson.build b/drivers/net/enic/meson.build
index 7131a25f09..f1f7f5432a 100644
--- a/drivers/net/enic/meson.build
+++ b/drivers/net/enic/meson.build
@@ -7,6 +7,10 @@ if is_windows
 subdir_done()
 endif
 
+if not get_option('enable_iova_as_pa')
+subdir_done()
+endif
+
 sources = files(
 'base/vnic_cq.c',
 'base/vnic_dev.c',
diff --git a/drivers/net/hns3/meson.build b/drivers/net/hns3/meson.build
index e1a5afa2ec..743fae9db7 100644
--- a/drivers/net/hns3/meson.build
+++ b/drivers/net/hns3/meson.build
@@ -13,9 +13,7 @@ if arch_subdir != 'x86' and arch_subdir != 'arm' or not 
dpdk_conf.get('RTE_ARCH_
 subdir_done()
 endif
 
-if dpdk_conf.get('RTE_IOVA_AS_PA') == 0
-build = false
-reason = 'driver does not support disabling IOVA as PA mode'
+if not get_option('enable_iova_as_pa')
 subdir_done()
 endif
 
diff --git a/drivers/net/i40e/meson.build b/drivers/net/i40e/meson.build
index e00c1a9ef9..689606a1de 100644
--- a/drivers/net/i40e/meson.build
+++ b/drivers/net/i40e/meson.build
@@ -13,6 +13,10 @@ if arch_subdir == 'riscv'
 subdir_done()
 endif
 
+if not get_option('enable_iova_as_pa')
+subdir_done()
+endif
+
 cflags += ['-DPF_DRIVER',
 '-DVF_DRIVER',
 '-DINTEGRATED_VF',
diff --git a/drivers/net/iavf/meson.build b/drivers/net/iavf/meson.build
index 6df771f917..c95503adda 100644
--- a/drivers/net/iavf/meson.build
+++ b/drivers/net/iavf/meson.build
@@ -1,6 +1,9 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2018 Luca Boccassi 
 
+if not get_option('enable_iova_as_pa')
+subdir_done()
+endif
 
 cflags += ['-Wno-strict-aliasing']
 
diff --git a/drivers/net/virtio/meson.build b/drivers/net/virtio/meson.build
index 0ffd77024e..b8ba1807fa 100644
--- a/drivers/net/virtio/meson.build
+++ b/drivers/net/virtio/meson.build
@@ -7,6 +7,10 @@ if is_windows
 subdir_done()
 endif
 
+if not get_option('enable_iova_as_pa')
+subdir_done()
+endif
+
 sources += files(

Re: [PATCH v2] net: not build PMD AVX library when no IOVA as PA

[Thomas Monjalon]

19/02/2023 12:08, Thomas Monjalon:
> 19/02/2023 11:04, Thomas Monjalon:
> > 29/01/2023 13:35, Zhang, Qi Z:
> > > From: Zhang, Qi Z 
> > > > 
> > > > PMD not announce pmd_supports_disable_iova_as_pa will not be build
> > > > when RTE_IOVA_AS_PA is not defined, but some AVX library for vector path
> > > > is not skipped by the build system which cause compile error.
> > > > 
> > > > The patch modify i40e, iavf, ice's meson file to skip AVX library build 
> > > > when
> > > > RTE_IOVA_AS_PA is not defined.
> > > > 
> > > > Cc: sta...@dpdk.org
> > > > 
> > > > Signed-off-by: Qi Zhang 
> > > 
> > > Move this for next-net review.
> > > 
> > > Acked-by: Qi Zhang 
> > 
> > You ack your own patch?
> > 
> > > Applied to dpdk-next-net-intel.
> > 
> > Sorry I don't pull this one, as I think there is a better fix:
> > we should not handle this option in each driver.
> > Instead the file drivers/meson.build must be fixed.
> 
> No sorry, it cannot be handled in drivers/meson.build.
> 
> So I suggest disabling the whole driver:
> 
> +if not get_option('enable_iova_as_pa')
> +subdir_done()
> +endif
> 
> Note: no need to disable ice, as it is enabled in another commit.

I've sent a patch to better disable more drivers:
https://patches.dpdk.org/project/dpdk/patch/20230219115529.3260580-1-tho...@monjalon.net/

I think more patches would be required to enable more drivers
supporting IOVA as VA. For instance, mempool drivers should be enabled.

RE: [EXT] Re: [PATCH v11 1/4] lib: add generic support for reading PMU events

[Tomasz Duszynski]

>-Original Message-
>From: Konstantin Ananyev 
>Sent: Friday, February 17, 2023 11:15 AM
>To: Tomasz Duszynski ; Konstantin Ananyev 
>;
>dev@dpdk.org
>Subject: RE: [EXT] Re: [PATCH v11 1/4] lib: add generic support for reading 
>PMU events
>
>
>
>> >>
>> >> This is especially useful in cases where CPU cores are isolated i.e
>> >> run dedicated tasks. In such cases one cannot use standard perf
>> >> utility without sacrificing latency and performance.
>> >>
>> >> Signed-off-by: Tomasz Duszynski 
>> >> Acked-by: Morten Brørup 
>> >> ---
>> >>   MAINTAINERS|   5 +
>> >>   app/test/meson.build   |   2 +
>> >>   app/test/test_pmu.c|  62 
>> >>   doc/api/doxy-api-index.md  |   3 +-
>> >>   doc/api/doxy-api.conf.in   |   1 +
>> >>   doc/guides/prog_guide/profile_app.rst  |  12 +
>> >>   doc/guides/rel_notes/release_23_03.rst |   7 +
>> >>   lib/meson.build|   1 +
>> >>   lib/pmu/meson.build|  13 +
>> >>   lib/pmu/pmu_private.h  |  32 ++
>> >>   lib/pmu/rte_pmu.c  | 460 +
>> >>   lib/pmu/rte_pmu.h  | 212 
>> >>   lib/pmu/version.map|  15 +
>> >>   13 files changed, 824 insertions(+), 1 deletion(-)
>> >>   create mode 100644 app/test/test_pmu.c
>> >>   create mode 100644 lib/pmu/meson.build
>> >>   create mode 100644 lib/pmu/pmu_private.h
>> >>   create mode 100644 lib/pmu/rte_pmu.c
>> >>   create mode 100644 lib/pmu/rte_pmu.h
>> >>   create mode 100644 lib/pmu/version.map
>> >>
>> >> diff --git a/MAINTAINERS b/MAINTAINERS index 3495946d0f..d37f242120
>> >> 100644
>> >> --- a/MAINTAINERS
>> >> +++ b/MAINTAINERS
>> >> @@ -1697,6 +1697,11 @@ M: Nithin Dabilpuram 
>> >>   M: Pavan Nikhilesh 
>> >>   F: lib/node/
>> >>
>> >> +PMU - EXPERIMENTAL
>> >> +M: Tomasz Duszynski 
>> >> +F: lib/pmu/
>> >> +F: app/test/test_pmu*
>> >> +
>> >>
>> >>   Test Applications
>> >>   -
>> >> diff --git a/app/test/meson.build b/app/test/meson.build index
>> >> f34d19e3c3..6b61b7fc32 100644
>> >> --- a/app/test/meson.build
>> >> +++ b/app/test/meson.build
>> >> @@ -111,6 +111,7 @@ test_sources = files(
>> >>   'test_reciprocal_division_perf.c',
>> >>   'test_red.c',
>> >>   'test_pie.c',
>> >> +'test_pmu.c',
>> >>   'test_reorder.c',
>> >>   'test_rib.c',
>> >>   'test_rib6.c',
>> >> @@ -239,6 +240,7 @@ fast_tests = [
>> >>   ['kni_autotest', false, true],
>> >>   ['kvargs_autotest', true, true],
>> >>   ['member_autotest', true, true],
>> >> +['pmu_autotest', true, true],
>> >>   ['power_cpufreq_autotest', false, true],
>> >>   ['power_autotest', true, true],
>> >>   ['power_kvm_vm_autotest', false, true], diff --git
>> >> a/app/test/test_pmu.c b/app/test/test_pmu.c new file mode 100644
>> >> index 00..c257638e8b
>> >> --- /dev/null
>> >> +++ b/app/test/test_pmu.c
>> >> @@ -0,0 +1,62 @@
>> >> +/* SPDX-License-Identifier: BSD-3-Clause
>> >> + * Copyright(C) 2023 Marvell International Ltd.
>> >> + */
>> >> +
>> >> +#include "test.h"
>> >> +
>> >> +#ifndef RTE_EXEC_ENV_LINUX
>> >> +
>> >> +static int
>> >> +test_pmu(void)
>> >> +{
>> >> + printf("pmu_autotest only supported on Linux, skipping test\n");
>> >> + return TEST_SKIPPED;
>> >> +}
>> >> +
>> >> +#else
>> >> +
>> >> +#include 
>> >> +
>> >> +static int
>> >> +test_pmu_read(void)
>> >> +{
>> >> + const char *name = NULL;
>> >> + int tries = 10, event;
>> >> + uint64_t val = 0;
>> >> +
>> >> + if (name == NULL) {
>> >> + printf("PMU not supported on this arch\n");
>> >> + return TEST_SKIPPED;
>> >> + }
>> >> +
>> >> + if (rte_pmu_init() < 0)
>> >> + return TEST_SKIPPED;
>> >> +
>> >> + event = rte_pmu_add_event(name);
>> >> + while (tries--)
>> >> + val += rte_pmu_read(event);
>> >> +
>> >> + rte_pmu_fini();
>> >> +
>> >> + return val ? TEST_SUCCESS : TEST_FAILED; }
>> >> +
>> >> +static struct unit_test_suite pmu_tests = {
>> >> + .suite_name = "pmu autotest",
>> >> + .setup = NULL,
>> >> + .teardown = NULL,
>> >> + .unit_test_cases = {
>> >> + TEST_CASE(test_pmu_read),
>> >> + TEST_CASES_END()
>> >> + }
>> >> +};
>> >> +
>> >> +static int
>> >> +test_pmu(void)
>> >> +{
>> >> + return unit_test_suite_runner(&pmu_tests);
>> >> +}
>> >> +
>> >> +#endif /* RTE_EXEC_ENV_LINUX */
>> >> +
>> >> +REGISTER_TEST_COMMAND(pmu_autotest, test_pmu);
>> >> diff --git a/doc/api/doxy-api-index.md b/doc/api/doxy-api-index.md
>> >> index 2deec7ea19..a8e04a195d 100644
>> >> --- a/doc/api/doxy-api-index.md
>> >> +++ b/doc/api/doxy-api-index.md
>> >> @@ -223,7 +223,8 @@ The public API headers are grouped by topics:
>> >> [log](@ref rte_log.h),
>> >> [errno](@ref rte_errno.h),
>> >> [trace](@ref rte_trace.h),
>> >> -  [trace_point](@ref rte_trace_point.h)
>> >>

Re: [PATCH v3] net/iavf: add debug checks for mbuf

[Thomas Monjalon]

19/02/2023 09:30, Zhang, Qi Z:
> From: Mingjin Ye 
> > 
> > The scalar Tx path would send wrong mbuf that causes the kernel driver to
> > fire the MDD event.
> > 
> > This patch adds mbuf detection in tx_prepare and enables it by defining
> > `RTE_ETHDEV_DEBUG_TX` macro to fix this issue. If validation fails, 
> > rte_errno
> 
> This is a debug utility but not a fix, enable this check will significantly 
> downgrade the performance
> Removed  "to fix this issue"
> 
> > will be set to EINVAL and returned.
> > 
> > Fixes: 3fd32df381f8 ("net/iavf: check Tx packet with correct UP and queue")
> > Fixes: 12b435bf8f2f ("net/iavf: support flex desc metadata extraction")
> > Fixes: f28fbd1e6b50 ("net/iavf: check max SIMD bitwidth")
> 
> Removed all the fix lines
> 
> > Cc: sta...@dpdk.org

I think the "stable line" should be removed as well.
It is a new debug feature, we should not try to backport it.

Re: [PATCH v10 2/2] ethdev: add quota flow action and item

[Thomas Monjalon]

02/02/2023 14:47, Gregory Etelson:
> Quota action limits traffic according to pre-defined configuration.
> Quota reflects overall traffic usage regardless bandwidth.
> Quota flow action initialized with signed tokens number value.
> Quota flow action updates tokens number according to
> these rules:
> 1. if quota was configured to count packet length, for each packet
>of size S, tokens number reduced by S.
> 2. If quota was configured to count packets, each packet decrements
>tokens number.
> quota action sets packet metadata according to a number of remaining
> tokens number:
>   PASS - remaining tokens number is non-negative.
>   BLOCK - remaining tokens number is negative.
> 
> Quota flow item matches on that data
> 
> Application updates tokens number in quota flow action
> with SET or ADD calls:
>  SET(QUOTA, val) - arm quota with new tokens number set to val
>  ADD(QUOTA, val) - increase existing quota tokens number by val
> 
> Both SET and ADD return to application number of tokens stored in port
> before update.
> 
> If quota state was BLOCK (negative action tokens number)
> application can change it to PASS after providing enough tokens to
> raise action tokens number to 0 or above.
> 
> Application must create a rule with quota action to mark flow and
> match on the mark with quota item in following flow rule.
> 
> Signed-off-by: Gregory Etelson 
> Acked-by: Ori Kam 

I feel this kind of explanation would have been better in rte_flow.rst
instead of a "doxygen-like" documentation.

Greogory, Ori, feels lucky, you've just got promoted as the rewriters
of the whole rte_flow.rst :)

Re: [PATCH] net/gve: fix Rx no mbufs stats counter update

[Stephen Hemminger]

On Sun, 19 Feb 2023 03:30:59 +0300
Levend Sayar  wrote:

> rx no_mbufs stats counter update is added for another error case.
> 
> Fixes: 4f6b1dd8240c ("net/gve: support basic statistics")
> Cc: junfeng@intel.com
> 
> Signed-off-by: Levend Sayar 
> ---
>  drivers/net/gve/gve_rx.c | 7 ++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/gve/gve_rx.c b/drivers/net/gve/gve_rx.c
> index 66fbcf3930..b0427731f8 100644
> --- a/drivers/net/gve/gve_rx.c
> +++ b/drivers/net/gve/gve_rx.c
> @@ -24,6 +24,7 @@ gve_rx_refill(struct gve_rx_queue *rxq)
>   nmb = rte_pktmbuf_alloc(rxq->mpool);
>   if (!nmb)
>   break;
> +
>   rxq->sw_ring[idx + i] = nmb;
>   }
>   if (i != nb_alloc) {

Looks like accidental whitespace change included in this patch.

> @@ -59,9 +60,13 @@ gve_rx_refill(struct gve_rx_queue *rxq)
>   nmb = rte_pktmbuf_alloc(rxq->mpool);
>   if (!nmb)
>   break;
> +
>   rxq->sw_ring[idx + i] = nmb;
>   }
> - nb_alloc = i;
> + if (i != nb_alloc) {
> + rxq->no_mbufs += nb_alloc - i;
> + nb_alloc = i;
> + }

Would be better to add unlikely() here like:
if (unlikely(i < nb_alloc)) {
rxq->no_mbufs += nb_alloc - i;
nb_alloc = i;
}

Or eliminate conditional branch in hot path completely.
rxq->no_mbufs += nb_alloc - i;
nb_alloc = i;

Or better yet refactor code here to use rte_pktmbuf_alloc_bulk() which
does single ring operation.

>   }
>   rxq->nb_avail -= nb_alloc;
>   next_avail += nb_alloc;

[PATCH v3 1/8] net/null: build null PMD on Windows

[Stephen Hemminger]

Builds fine with current code, no changes needed.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 drivers/net/null/meson.build | 6 --
 1 file changed, 6 deletions(-)

diff --git a/drivers/net/null/meson.build b/drivers/net/null/meson.build
index 4a483955a7a9..a51f8f5211b0 100644
--- a/drivers/net/null/meson.build
+++ b/drivers/net/null/meson.build
@@ -1,11 +1,5 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2017 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files('rte_eth_null.c')
 pmd_supports_disable_iova_as_pa = true
-- 
2.39.1

[PATCH v3 0/8] Enable building more libraries on Windows

[Stephen Hemminger]

While diagnosing some Windows cross build errors;
noticed that lots of important DPDK libraries are not
being built on Windows.

Stephen Hemminger (8):
  net/null: build null PMD on Windows
  net/ring: build on Windows
  lpm: enable on Windows
  reorder: build on Windows
  ip_frag: enable build on Windows
  rib: enable on Windows
  fib: enable on Windows
  pcapng: windows compatibility

v3 - add rte_os_shim.h to rte_ring to fix use of
 strdup() breaking with stricter MS compiler

v2 - fix unnecessary PATH_MAX in net/ring driver

Stephen Hemminger (8):
  net/null: build null PMD on Windows
  net/ring: build on Windows
  lpm: enable on Windows
  reorder: build on Windows
  ip_frag: enable build on Windows
  rib: enable on Windows
  fib: enable on Windows
  pcapng: windows compatibility

 app/test/meson.build|  2 +-
 app/test/test_fib.c | 22 +---
 app/test/test_fib6.c| 24 ++
 app/test/test_fib6_perf.c   | 15 +
 app/test/test_fib_perf.c| 19 ++-
 app/test/test_ipfrag.c  | 12 ---
 app/test/test_lpm.c | 15 +
 app/test/test_lpm6.c| 12 ---
 app/test/test_lpm6_data.h   |  3 +-
 app/test/test_lpm6_perf.c   | 14 +---
 app/test/test_lpm_perf.c| 39 +++---
 app/test/test_reorder.c | 11 --
 app/test/test_rib.c | 19 ---
 app/test/test_rib6.c| 24 ++
 drivers/net/null/meson.build|  6 
 drivers/net/ring/meson.build|  6 
 drivers/net/ring/rte_eth_ring.c |  4 ++-
 lib/fib/meson.build |  6 
 lib/fib/rte_fib.c   |  1 +
 lib/fib/rte_fib6.c  |  1 +
 lib/ip_frag/ip_frag_common.h|  2 ++
 lib/ip_frag/meson.build |  6 
 lib/lpm/meson.build |  6 
 lib/pcapng/meson.build  |  6 
 lib/pcapng/rte_pcapng.c | 59 +++--
 lib/reorder/meson.build |  6 
 lib/reorder/rte_reorder.c   |  1 +
 lib/rib/meson.build |  6 
 lib/rib/rte_rib.c   |  1 +
 lib/rib/rte_rib6.c  |  1 +
 30 files changed, 93 insertions(+), 256 deletions(-)

-- 
2.39.1

[PATCH v3 2/8] net/ring: build on Windows

[Stephen Hemminger]

The command line arguments are stored in node_action_pair
and the name[] was sized to PATH_MAX which does not exist on Windows.
Since the name is either "CREATE" or "ATTACH" it is not
related to PATH_MAX (4096).

With this fix driver builds ok on windows, but need to modify the
test meson build to skip the eventdev test on Windows.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 app/test/meson.build| 2 +-
 drivers/net/ring/meson.build| 6 --
 drivers/net/ring/rte_eth_ring.c | 4 +++-
 3 files changed, 4 insertions(+), 8 deletions(-)

diff --git a/app/test/meson.build b/app/test/meson.build
index f34d19e3c3cb..a713f0382280 100644
--- a/app/test/meson.build
+++ b/app/test/meson.build
@@ -390,7 +390,7 @@ if dpdk_conf.has('RTE_NET_BOND')
 driver_test_names += 'link_bonding_mode4_autotest'
 endif
 endif
-if dpdk_conf.has('RTE_NET_RING')
+if dpdk_conf.has('RTE_LIB_EVENTDEV') and dpdk_conf.has('RTE_NET_RING')
 test_deps += 'net_ring'
 test_sources += 'test_pmd_ring_perf.c'
 test_sources += 'test_pmd_ring.c'
diff --git a/drivers/net/ring/meson.build b/drivers/net/ring/meson.build
index 72792e26b05a..3534a3cc2287 100644
--- a/drivers/net/ring/meson.build
+++ b/drivers/net/ring/meson.build
@@ -1,12 +1,6 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2017 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files('rte_eth_ring.c')
 headers = files('rte_eth_ring.h')
 pmd_supports_disable_iova_as_pa = true
diff --git a/drivers/net/ring/rte_eth_ring.c b/drivers/net/ring/rte_eth_ring.c
index bd5a47dd9017..e8bc9b627102 100644
--- a/drivers/net/ring/rte_eth_ring.c
+++ b/drivers/net/ring/rte_eth_ring.c
@@ -9,6 +9,7 @@
 #include 
 #include 
 #include 
+#include 
 #include 
 #include 
 #include 
@@ -17,6 +18,7 @@
 #define ETH_RING_NUMA_NODE_ACTION_ARG  "nodeaction"
 #define ETH_RING_ACTION_CREATE "CREATE"
 #define ETH_RING_ACTION_ATTACH "ATTACH"
+#define ETH_RING_ACTION_MAX_LEN8 /* CREATE | ACTION */
 #define ETH_RING_INTERNAL_ARG  "internal"
 #define ETH_RING_INTERNAL_ARG_MAX_LEN  19 /* "0x..16chars..\0" */
 
@@ -539,7 +541,7 @@ eth_dev_ring_create(const char *name,
 }
 
 struct node_action_pair {
-   char name[PATH_MAX];
+   char name[ETH_RING_ACTION_MAX_LEN];
unsigned int node;
enum dev_action action;
 };
-- 
2.39.1

[PATCH v3 3/8] lpm: enable on Windows

[Stephen Hemminger]

This builds on Windows without changes.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 app/test/test_lpm.c   | 15 +--
 app/test/test_lpm6.c  | 12 
 app/test/test_lpm6_data.h |  3 ++-
 app/test/test_lpm6_perf.c | 14 +-
 app/test/test_lpm_perf.c  | 39 ---
 lib/lpm/meson.build   |  6 --
 6 files changed, 16 insertions(+), 73 deletions(-)

diff --git a/app/test/test_lpm.c b/app/test/test_lpm.c
index bceb9ae743d7..37b460af3a96 100644
--- a/app/test/test_lpm.c
+++ b/app/test/test_lpm.c
@@ -2,18 +2,6 @@
  * Copyright(c) 2010-2014 Intel Corporation
  */
 
-#include "test.h"
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_lpm(void)
-{
-   printf("lpm not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
 #include 
 #include 
 #include 
@@ -22,6 +10,7 @@ test_lpm(void)
 #include 
 #include 
 
+#include "test.h"
 #include "test_xmmt_ops.h"
 
 #define TEST_LPM_ASSERT(cond) do {\
@@ -1595,6 +1584,4 @@ test_lpm(void)
return global_status;
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(lpm_autotest, test_lpm);
diff --git a/app/test/test_lpm6.c b/app/test/test_lpm6.c
index 1f44392db739..b6b6f8615ea5 100644
--- a/app/test/test_lpm6.c
+++ b/app/test/test_lpm6.c
@@ -10,16 +10,6 @@
 #include 
 
 #include 
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_lpm6(void)
-{
-   printf("lpm6 not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
 #include 
 
 #include "test_lpm6_data.h"
@@ -1803,6 +1793,4 @@ test_lpm6(void)
return global_status;
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(lpm6_autotest, test_lpm6);
diff --git a/app/test/test_lpm6_data.h b/app/test/test_lpm6_data.h
index da9b161f20e5..3a1fa25f8433 100644
--- a/app/test/test_lpm6_data.h
+++ b/app/test/test_lpm6_data.h
@@ -6,6 +6,7 @@
 
 #include 
 #include 
+#include 
 
 struct rules_tbl_entry {
uint8_t ip[16];
@@ -1129,7 +1130,7 @@ static void generate_large_ips_table(int 
gen_expected_next_hop)
 
for (i = 0; i < NUM_IPS_ENTRIES; i++) {
for (j = 0; j < 16; j++)
-   large_ips_table[i].ip[j] = lrand48();
+   large_ips_table[i].ip[j] = rte_rand();
}
 
for (k = j = 0, i = 0; i < NUM_IPS_ENTRIES; i++) {
diff --git a/app/test/test_lpm6_perf.c b/app/test/test_lpm6_perf.c
index aaf2773b6fac..5b684686a687 100644
--- a/app/test/test_lpm6_perf.c
+++ b/app/test/test_lpm6_perf.c
@@ -2,17 +2,6 @@
  * Copyright(c) 2010-2014 Intel Corporation
  */
 
-#include "test.h"
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_lpm6_perf(void)
-{
-   printf("lpm6_perf not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
 
 #include 
 #include 
@@ -24,6 +13,7 @@ test_lpm6_perf(void)
 #include 
 #include 
 
+#include "test.h"
 #include "test_lpm6_data.h"
 
 #define TEST_LPM_ASSERT(cond) do {\
@@ -171,6 +161,4 @@ test_lpm6_perf(void)
return 0;
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(lpm6_perf_autotest, test_lpm6_perf);
diff --git a/app/test/test_lpm_perf.c b/app/test/test_lpm_perf.c
index e858716f909b..e72437ba3850 100644
--- a/app/test/test_lpm_perf.c
+++ b/app/test/test_lpm_perf.c
@@ -3,17 +3,6 @@
  * Copyright(c) 2020 Arm Limited
  */
 
-#include "test.h"
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_lpm_perf(void)
-{
-   printf("lpm_perf not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
 #include 
 #include 
 #include 
@@ -25,7 +14,9 @@ test_lpm_perf(void)
 #include 
 #include 
 #include 
+#include 
 
+#include "test.h"
 #include "test_xmmt_ops.h"
 
 struct rte_lpm *lpm;
@@ -34,8 +25,9 @@ static volatile uint8_t writer_done;
 static volatile uint32_t thr_id;
 static uint64_t gwrite_cycles;
 static uint32_t num_writers;
-/* LPM APIs are not thread safe, use mutex to provide thread safety */
-static pthread_mutex_t lpm_mutex = PTHREAD_MUTEX_INITIALIZER;
+
+/* LPM APIs are not thread safe, use spinlock */
+static rte_spinlock_t lpm_lock = RTE_SPINLOCK_INITIALIZER;
 
 /* Report quiescent state interval every 1024 lookups. Larger critical
  * sections in reader will result in writer polling multiple times.
@@ -267,7 +259,7 @@ static void generate_random_rule_prefix(uint32_t ip_class, 
uint8_t depth)
/* Only generate rest bits except the most significant
 * fixed bits for IP address class
 */
-   start = lrand48() & mask;
+   start = rte_rand() & mask;
ptr_rule = &large_route_table[num_route_entries];
ptr_ldepth_rule = &large_ldepth_route_table[num_ldepth_route_entries];
for (k = 0; k < rule_num; k++) {
@@ -296,7 +288,7 @@ static void insert_rule_in_random_pos(uint32_t ip, uint8_t 
depth)
struct route_rule tmp;
 
  

[PATCH v3 4/8] reorder: build on Windows

[Stephen Hemminger]

This builds on Windows if sys/queue.h is included.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 app/test/test_reorder.c   | 11 ---
 lib/reorder/meson.build   |  6 --
 lib/reorder/rte_reorder.c |  1 +
 3 files changed, 1 insertion(+), 17 deletions(-)

diff --git a/app/test/test_reorder.c b/app/test/test_reorder.c
index f0714a5c18a3..9ebd91477536 100644
--- a/app/test/test_reorder.c
+++ b/app/test/test_reorder.c
@@ -11,16 +11,6 @@
 #include 
 #include 
 #include 
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_reorder(void)
-{
-   printf("reorder not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-#else
-
 #include 
 #include 
 #include 
@@ -395,6 +385,5 @@ test_reorder(void)
return unit_test_suite_runner(&reorder_test_suite);
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
 
 REGISTER_TEST_COMMAND(reorder_autotest, test_reorder);
diff --git a/lib/reorder/meson.build b/lib/reorder/meson.build
index 621c1f350103..03aed53d9007 100644
--- a/lib/reorder/meson.build
+++ b/lib/reorder/meson.build
@@ -1,12 +1,6 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2017 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files('rte_reorder.c')
 headers = files('rte_reorder.h')
 deps += ['mbuf']
diff --git a/lib/reorder/rte_reorder.c b/lib/reorder/rte_reorder.c
index 385ee479da42..00e2cf043e04 100644
--- a/lib/reorder/rte_reorder.c
+++ b/lib/reorder/rte_reorder.c
@@ -3,6 +3,7 @@
  */
 
 #include 
+#include 
 
 #include 
 #include 
-- 
2.39.1

[PATCH v3 5/8] ip_frag: enable build on Windows

[Stephen Hemminger]

This build works on Windows if sys/queue.h is included.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 app/test/test_ipfrag.c   | 12 
 lib/ip_frag/ip_frag_common.h |  2 ++
 lib/ip_frag/meson.build  |  6 --
 3 files changed, 2 insertions(+), 18 deletions(-)

diff --git a/app/test/test_ipfrag.c b/app/test/test_ipfrag.c
index 88cc4cdeea8d..402ce361c1a2 100644
--- a/app/test/test_ipfrag.c
+++ b/app/test/test_ipfrag.c
@@ -10,17 +10,6 @@
 #include 
 #include 
 #include 
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_ipfrag(void)
-{
-   printf("ipfrag not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
 #include 
 #include 
 #include 
@@ -520,6 +509,5 @@ test_ipfrag(void)
return unit_test_suite_runner(&ipfrag_testsuite);
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
 
 REGISTER_TEST_COMMAND(ipfrag_autotest, test_ipfrag);
diff --git a/lib/ip_frag/ip_frag_common.h b/lib/ip_frag/ip_frag_common.h
index 9c0dbdeb6eb9..0f4ad72a90d9 100644
--- a/lib/ip_frag/ip_frag_common.h
+++ b/lib/ip_frag/ip_frag_common.h
@@ -8,6 +8,8 @@
 #include "rte_ip_frag.h"
 #include "ip_reassembly.h"
 
+#include 
+
 /* logging macros. */
 #ifdef RTE_LIBRTE_IP_FRAG_DEBUG
 #defineIP_FRAG_LOG(lvl, fmt, args...)  RTE_LOG(lvl, USER1, fmt, ##args)
diff --git a/lib/ip_frag/meson.build b/lib/ip_frag/meson.build
index 3a252bdaf6a4..ea2de09f7528 100644
--- a/lib/ip_frag/meson.build
+++ b/lib/ip_frag/meson.build
@@ -1,12 +1,6 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2017 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files(
 'rte_ipv4_fragmentation.c',
 'rte_ipv6_fragmentation.c',
-- 
2.39.1

[PATCH v3 6/8] rib: enable on Windows

[Stephen Hemminger]

The RIB library builds on Windows as long as sys/queue.h is included

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 app/test/test_rib.c  | 19 ---
 app/test/test_rib6.c | 24 ++--
 lib/rib/meson.build  |  6 --
 lib/rib/rte_rib.c|  1 +
 lib/rib/rte_rib6.c   |  1 +
 5 files changed, 4 insertions(+), 47 deletions(-)

diff --git a/app/test/test_rib.c b/app/test/test_rib.c
index 06058f8f7c52..65b685641094 100644
--- a/app/test/test_rib.c
+++ b/app/test/test_rib.c
@@ -10,23 +10,6 @@
 #include 
 
 #include 
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_rib(void)
-{
-   printf("rib not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-static int
-test_slow_rib(void)
-{
-   printf("slow_rib not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-#else
-
 #include 
 
 typedef int32_t (*rte_rib_test)(void);
@@ -380,7 +363,5 @@ test_slow_rib(void)
return unit_test_suite_runner(&rib_slow_tests);
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(rib_autotest, test_rib);
 REGISTER_TEST_COMMAND(rib_slow_autotest, test_slow_rib);
diff --git a/app/test/test_rib6.c b/app/test/test_rib6.c
index ba79aedea5f7..336b779d2e1e 100644
--- a/app/test/test_rib6.c
+++ b/app/test/test_rib6.c
@@ -3,32 +3,14 @@
  * Copyright(c) 2019 Intel Corporation
  */
 
-#include "test.h"
-
 #include 
 #include 
 #include 
-
 #include 
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_rib6(void)
-{
-   printf("rib6 not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-static int
-test_slow_rib6(void)
-{
-   printf("slow_rib6 not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-#else
-
 #include 
 
+#include "test.h"
+
 typedef int32_t (*rte_rib6_test)(void);
 
 static int32_t test_create_invalid(void);
@@ -385,7 +367,5 @@ test_slow_rib6(void)
return unit_test_suite_runner(&rib6_slow_tests);
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(rib6_autotest, test_rib6);
 REGISTER_TEST_COMMAND(rib6_slow_autotest, test_slow_rib6);
diff --git a/lib/rib/meson.build b/lib/rib/meson.build
index bda7f576e976..7bacbb453592 100644
--- a/lib/rib/meson.build
+++ b/lib/rib/meson.build
@@ -2,12 +2,6 @@
 # Copyright(c) 2018 Vladimir Medvedkin 
 # Copyright(c) 2019 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files('rte_rib.c', 'rte_rib6.c')
 headers = files('rte_rib.h', 'rte_rib6.h')
 deps += ['mempool']
diff --git a/lib/rib/rte_rib.c b/lib/rib/rte_rib.c
index b0794edf66f5..812a2597d117 100644
--- a/lib/rib/rte_rib.c
+++ b/lib/rib/rte_rib.c
@@ -5,6 +5,7 @@
 
 #include 
 #include 
+#include 
 
 #include 
 #include 
diff --git a/lib/rib/rte_rib6.c b/lib/rib/rte_rib6.c
index 19e4ff97c479..ae44281ae105 100644
--- a/lib/rib/rte_rib6.c
+++ b/lib/rib/rte_rib6.c
@@ -5,6 +5,7 @@
 
 #include 
 #include 
+#include 
 
 #include 
 #include 
-- 
2.39.1

[PATCH v3 8/8] pcapng: windows compatibility

[Stephen Hemminger]

Allow building on Windows, need to provide some comparability
wrappers for writev() and if_nametoindex.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 lib/pcapng/meson.build  |  6 -
 lib/pcapng/rte_pcapng.c | 59 +++--
 2 files changed, 57 insertions(+), 8 deletions(-)

diff --git a/lib/pcapng/meson.build b/lib/pcapng/meson.build
index da938bbcb733..4549925d41b6 100644
--- a/lib/pcapng/meson.build
+++ b/lib/pcapng/meson.build
@@ -1,12 +1,6 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2019 Microsoft Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files('rte_pcapng.c')
 headers = files('rte_pcapng.h')
 
diff --git a/lib/pcapng/rte_pcapng.c b/lib/pcapng/rte_pcapng.c
index 65c8c77fa405..d4b2db773160 100644
--- a/lib/pcapng/rte_pcapng.c
+++ b/lib/pcapng/rte_pcapng.c
@@ -3,15 +3,18 @@
  */
 
 #include 
-#include 
 #include 
 #include 
 #include 
 #include 
-#include 
 #include 
 #include 
 
+#ifndef RTE_EXEC_ENV_WINDOWS
+#include 
+#include 
+#endif
+
 #include 
 #include 
 #include 
@@ -20,6 +23,7 @@
 #include 
 #include 
 #include 
+#include 
 #include 
 #include 
 #include 
@@ -47,6 +51,57 @@ static struct pcapng_time {
struct rte_reciprocal_u64 tsc_hz_inverse;
 } pcapng_time;
 
+
+#ifdef RTE_EXEC_ENV_WINDOWS
+/*
+ * Windows does not have writev() call.
+ * Emulate this by copying to a new buffer.
+ * The copy is necessary since pcapng needs to be thread-safe
+ * and do atomic write operations.
+ */
+
+#define IOV_MAX 128
+struct iovec {
+   void   *iov_base;
+   size_t  iov_len;
+};
+
+static ssize_t writev(int fd, const struct iovec *iov, int iovcnt)
+{
+   size_t bytes = 0;
+   uint8_t *ptr;
+   void *tmp_buf;
+   ssize_t ret;
+   int i;
+
+   for (i = 0; i < iovcnt; i++)
+   bytes += iov[i].iov_len;
+
+   if (unlikely(bytes == 0))
+   return 0;
+
+   tmp_buf = malloc(bytes);
+   if (unlikely(tmp_buf == NULL)) {
+   errno = ENOMEM;
+   return -1;
+   }
+
+   ptr = tmp_buf;
+   for (i = 0; i < iovcnt; i++) {
+   rte_memcpy(ptr, iov[i].iov_base, iov[i].iov_len);
+   ptr += iov[i].iov_len;
+   }
+
+   ret = write(fd, tmp_buf, bytes);
+   free(tmp_buf);
+   return ret;
+}
+
+#define IF_NAMESIZE16
+/* compatiablity wrapper because name is optional */
+#define if_indextoname(ifindex, ifname) NULL
+#endif
+
 static inline void
 pcapng_init(void)
 {
-- 
2.39.1

[PATCH v3 7/8] fib: enable on Windows

[Stephen Hemminger]

The FIB library builds on Windows as long as sys/queue.h is defined.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 app/test/test_fib.c   | 22 +-
 app/test/test_fib6.c  | 24 ++--
 app/test/test_fib6_perf.c | 15 +--
 app/test/test_fib_perf.c  | 19 +++
 lib/fib/meson.build   |  6 --
 lib/fib/rte_fib.c |  1 +
 lib/fib/rte_fib6.c|  1 +
 7 files changed, 9 insertions(+), 79 deletions(-)

diff --git a/app/test/test_fib.c b/app/test/test_fib.c
index a2d1ea8f3abc..eb69d6e2fd4c 100644
--- a/app/test/test_fib.c
+++ b/app/test/test_fib.c
@@ -9,28 +9,10 @@
 
 #include 
 #include 
+#include 
 
 #include "test.h"
 
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_fib(void)
-{
-   printf("fib not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-static int
-test_slow_fib(void)
-{
-   printf("slow_fib not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
-#include 
-
 typedef int32_t (*rte_fib_test)(void);
 
 static int32_t test_create_invalid(void);
@@ -433,7 +415,5 @@ test_slow_fib(void)
return unit_test_suite_runner(&fib_slow_tests);
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(fib_autotest, test_fib);
 REGISTER_TEST_COMMAND(fib_slow_autotest, test_slow_fib);
diff --git a/app/test/test_fib6.c b/app/test/test_fib6.c
index cd971e6ecdfb..15ad09178ae2 100644
--- a/app/test/test_fib6.c
+++ b/app/test/test_fib6.c
@@ -9,29 +9,11 @@
 
 #include 
 #include 
-
-#include "test.h"
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_fib6(void)
-{
-   printf("fib not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-static int
-test_slow_fib6(void)
-{
-   printf("slow_fib not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
 #include 
 #include 
 
+#include "test.h"
+
 typedef int32_t (*rte_fib6_test)(void);
 
 static int32_t test_create_invalid(void);
@@ -442,7 +424,5 @@ test_slow_fib6(void)
return unit_test_suite_runner(&fib6_slow_tests);
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(fib6_autotest, test_fib6);
 REGISTER_TEST_COMMAND(fib6_slow_autotest, test_slow_fib6);
diff --git a/app/test/test_fib6_perf.c b/app/test/test_fib6_perf.c
index 21d2b65318e9..add20c2331b1 100644
--- a/app/test/test_fib6_perf.c
+++ b/app/test/test_fib6_perf.c
@@ -10,21 +10,10 @@
 #include 
 #include 
 #include 
+#include 
 
 #include "test.h"
 
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_fib6_perf(void)
-{
-   printf("fib6_perf not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
-#include 
-
 #include "test_lpm6_data.h"
 
 #define TEST_FIB_ASSERT(cond) do { \
@@ -167,6 +156,4 @@ test_fib6_perf(void)
return 0;
 }
 
-#endif /*ifdef RTE_EXEC_ENV_WINDOWS*/
-
 REGISTER_TEST_COMMAND(fib6_perf_autotest, test_fib6_perf);
diff --git a/app/test/test_fib_perf.c b/app/test/test_fib_perf.c
index 9787874cc9b7..b56293e64f41 100644
--- a/app/test/test_fib_perf.c
+++ b/app/test/test_fib_perf.c
@@ -12,22 +12,11 @@
 #include 
 #include 
 #include 
+#include 
 
 #include "test.h"
 #include "test_xmmt_ops.h"
 
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_fib_perf(void)
-{
-   printf("fib_perf not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
-#include 
-
 #define TEST_FIB_ASSERT(cond) do { \
if (!(cond)) {  \
printf("Error at line %d:\n", __LINE__);\
@@ -246,7 +235,7 @@ static void generate_random_rule_prefix(uint32_t ip_class, 
uint8_t depth)
/* Only generate rest bits except the most significant
 * fixed bits for IP address class
 */
-   start = lrand48() & mask;
+   start = rte_rand() & mask;
ptr_rule = &large_route_table[num_route_entries];
for (k = 0; k < rule_num; k++) {
ptr_rule->ip = (start << (RTE_FIB_MAX_DEPTH - depth))
@@ -265,7 +254,7 @@ static void insert_rule_in_random_pos(uint32_t ip, uint8_t 
depth)
struct route_rule tmp;
 
do {
-   pos = lrand48();
+   pos = rte_rand();
try_count++;
} while ((try_count < 10) && (pos > num_route_entries));
 
@@ -420,6 +409,4 @@ test_fib_perf(void)
return 0;
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(fib_perf_autotest, test_fib_perf);
diff --git a/lib/fib/meson.build b/lib/fib/meson.build
index 9b848d08417c..ddcae0617a73 100644
--- a/lib/fib/meson.build
+++ b/lib/fib/meson.build
@@ -2,12 +2,6 @@
 # Copyright(c) 2018 Vladimir Medvedkin 
 # Copyright(c) 2019 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files('rte_fib.c', 'rte_fib6.c', 'dir24_8.c', 'trie.c')
 heade

Re: [PATCH v11 0/3] Fix cmdline_poll and testpmd signal handling

[Stephen Hemminger]

On Fri,  3 Feb 2023 11:14:06 -0800
Stephen Hemminger  wrote:

> This patchset keeps uncovering bad practices in the cmdline library
> around end of file and signal handling.
> 
> Stephen Hemminger (3):
>   cmdline: make rdline status not private
>   cmdline: handle EOF in cmdline_poll
>   testpmd: cleanup cleanly from signal
> 
>  app/test-pmd/cmdline.c| 29 +
>  app/test-pmd/testpmd.c| 77 ---
>  app/test-pmd/testpmd.h|  1 +
>  lib/cmdline/cmdline.c | 11 +++--
>  lib/cmdline/cmdline.h |  6 +++
>  lib/cmdline/cmdline_private.h |  6 ---
>  6 files changed, 62 insertions(+), 68 deletions(-)
> 

Could this please be merged for 23.03?
There are Ack's.
The only CI failure is a bogus performance test failure.

Re: [PATCH 2/2] net/gve: add extended statistics

[Ferruh Yigit]

On 2/19/2023 12:26 AM, Levend Sayar wrote:
> Ferruh,
> 
> Thanks for this detailed review.
> I am setting superseded this patch and create a new one.
> You’re right at all points.
> For rx.no_mbufs counting, I probably overlooked while rebasing my patch
> and it is mixed with newly came patch.
> 
> When I check ethdev layer again, I noticed that when dev_flags
> has RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS
> Rx/tx queue stats are already added. I am pushing a fresh patch for
> adding rx/tx queue stats.
> 

Hi Levend,

You are right BUT, 'RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS' is a temporary
solution and plan is to remove it [1].

Background is, queue stats in "struct rte_eth_stats" has fixed size and
as number of queues supported by devices increase these fields getting
bigger and bigger, the solution we came was to completely remove these
fields from stats struct and get queue statistics via xstats.

During transition 'RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS' is introduced
until all drivers implement queue stats in xstats. We are not pushing
hard for existing drivers to update but at least requiring new drivers
to implement xstats method.

That is why for net/gve updating queue stats in 'gve_dev_stats_get()'
rejected before, and xstats implementation is requested.


[1]
https://elixir.bootlin.com/dpdk/v22.11.1/source/doc/guides/rel_notes/deprecation.rst#L88


> And also noticed a missing part at rx no_mbufs counting.
> 
> Best,
> Levend
>  
> 
>> On 17 Feb 2023, at 15:34, Ferruh Yigit  wrote:
>>
>> On 2/16/2023 6:58 PM, Levend Sayar wrote:
>>> Google Virtual NIC PMD is enriched with extended statistics info.
>>
>> Only queue stats added to xstats, can you please highlight this in the
>> commit log?
>>
>>> eth_dev_ops callback names are also synched with eth_dev_ops field names
>>>
>>
>> Renaming eth_dev_ops is not related to xstats, and I am not sure if the
>> change is necessary, can you please drop it from this patch?
>>
>>> Signed-off-by: Levend Sayar 
>>> ---
>>> drivers/net/gve/gve_ethdev.c | 152 ++-
>>> drivers/net/gve/gve_rx.c |   8 +-
>>> 2 files changed, 138 insertions(+), 22 deletions(-)
>>>
>>> diff --git a/drivers/net/gve/gve_ethdev.c b/drivers/net/gve/gve_ethdev.c
>>> index fef2458a16..e31fdce960 100644
>>> --- a/drivers/net/gve/gve_ethdev.c
>>> +++ b/drivers/net/gve/gve_ethdev.c
>>> @@ -266,7 +266,7 @@ gve_dev_close(struct rte_eth_dev *dev)
>>> }
>>>
>>> static int
>>> -gve_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info
>>> *dev_info)
>>> +gve_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info
>>> *dev_info)
>>> {
>>> struct gve_priv *priv = dev->data->dev_private;
>>>
>>> @@ -319,15 +319,12 @@ gve_dev_info_get(struct rte_eth_dev *dev,
>>> struct rte_eth_dev_info *dev_info)
>>> }
>>>
>>> static int
>>> -gve_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
>>> +gve_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
>>> {
>>> uint16_t i;
>>>
>>> for (i = 0; i < dev->data->nb_tx_queues; i++) {
>>> struct gve_tx_queue *txq = dev->data->tx_queues[i];
>>> -if (txq == NULL)
>>> -continue;
>>> -
>>
>> I assume check is removed because it is unnecessary, but again not
>> directly related with the patch, can you also drop these from the patch
>> to reduce the noise.
>>
>>> stats->opackets += txq->packets;
>>> stats->obytes += txq->bytes;
>>> stats->oerrors += txq->errors;
>>> @@ -335,9 +332,6 @@ gve_dev_stats_get(struct rte_eth_dev *dev, struct
>>> rte_eth_stats *stats)
>>>
>>> for (i = 0; i < dev->data->nb_rx_queues; i++) {
>>> struct gve_rx_queue *rxq = dev->data->rx_queues[i];
>>> -if (rxq == NULL)
>>> -continue;
>>> -
>>> stats->ipackets += rxq->packets;
>>> stats->ibytes += rxq->bytes;
>>> stats->ierrors += rxq->errors;
>>> @@ -348,15 +342,12 @@ gve_dev_stats_get(struct rte_eth_dev *dev,
>>> struct rte_eth_stats *stats)
>>> }
>>>
>>> static int
>>> -gve_dev_stats_reset(struct rte_eth_dev *dev)
>>> +gve_stats_reset(struct rte_eth_dev *dev)
>>> {
>>> uint16_t i;
>>>
>>> for (i = 0; i < dev->data->nb_tx_queues; i++) {
>>> struct gve_tx_queue *txq = dev->data->tx_queues[i];
>>> -if (txq == NULL)
>>> -continue;
>>> -
>>> txq->packets  = 0;
>>> txq->bytes = 0;
>>> txq->errors = 0;
>>> @@ -364,9 +355,6 @@ gve_dev_stats_reset(struct rte_eth_dev *dev)
>>>
>>> for (i = 0; i < dev->data->nb_rx_queues; i++) {
>>> struct gve_rx_queue *rxq = dev->data->rx_queues[i];
>>> -if (rxq == NULL)
>>> -continue;
>>> -
>>> rxq->packets  = 0;
>>> rxq->bytes = 0;
>>> rxq->errors = 0;
>>> @@ -377,7 +365,7 @@ gve_dev_stats_reset(struct rte_eth_dev *dev)
>>> }
>>>
>>> static int
>>> -gve_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
>>> +gve_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
>>> {
>>> struct gve_priv *priv = dev->data->dev_private;
>>> int err;
>>> @@ -403,20 +391,144 @@ gve_dev_mtu_set(struct rte_eth_dev *dev,
>>> uint16_t mtu)
>>> return 0;
>>> }
>>>
>>> +static int
>>> +gve_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xsta

Re: [PATCH] net/gve: add Rx/Tx queue statistics

[Ferruh Yigit]

On 2/19/2023 12:31 AM, Levend Sayar wrote:
> Google Virtual NIC PMD is enriched with rx/tx queue statistics info.
> 
> Signed-off-by: Levend Sayar 
> ---
>  drivers/net/gve/gve_ethdev.c | 8 
>  1 file changed, 8 insertions(+)
> 
> diff --git a/drivers/net/gve/gve_ethdev.c b/drivers/net/gve/gve_ethdev.c
> index fef2458a16..2a6cb11da3 100644
> --- a/drivers/net/gve/gve_ethdev.c
> +++ b/drivers/net/gve/gve_ethdev.c
> @@ -213,6 +213,7 @@ gve_dev_start(struct rte_eth_dev *dev)
>   }
>   }
>  
> + dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
>   dev->data->dev_started = 1;
>   gve_link_update(dev, 0);
>  
> @@ -331,6 +332,9 @@ gve_dev_stats_get(struct rte_eth_dev *dev, struct 
> rte_eth_stats *stats)
>   stats->opackets += txq->packets;
>   stats->obytes += txq->bytes;
>   stats->oerrors += txq->errors;
> +
> + stats->q_opackets[i] = txq->packets;
> + stats->q_obytes[i] = txq->bytes;
>   }
>  
>   for (i = 0; i < dev->data->nb_rx_queues; i++) {
> @@ -342,6 +346,10 @@ gve_dev_stats_get(struct rte_eth_dev *dev, struct 
> rte_eth_stats *stats)
>   stats->ibytes += rxq->bytes;
>   stats->ierrors += rxq->errors;
>   stats->rx_nombuf += rxq->no_mbufs;
> +
> + stats->q_ipackets[i] = rxq->packets;
> + stats->q_ibytes[i] = rxq->bytes;
> + stats->q_errors[i] = rxq->errors;
>   }
>  
>   return 0;

nack,

as explained in other thread [1], we need to implement xstats in driver
without relying on 'RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS' flag.

[1]
https://inbox.dpdk.org/dev/0123d577-39f5-7213-07b3-04ef3918e...@amd.com/T/#u

Re: [PATCH 2/2] net/gve: add extended statistics

[Levend Sayar]

Hi Ferruh,

Opps, I was not aware of that rejection.
Thanks for notification.
  
Let me supersede this patch.
And create a new one. 

Best,
Levend

> On 19 Feb 2023, at 23:09, Ferruh Yigit  wrote:
> 
> On 2/19/2023 12:26 AM, Levend Sayar wrote:
>> Ferruh,
>> 
>> Thanks for this detailed review.
>> I am setting superseded this patch and create a new one.
>> You’re right at all points.
>> For rx.no_mbufs counting, I probably overlooked while rebasing my patch
>> and it is mixed with newly came patch.
>> 
>> When I check ethdev layer again, I noticed that when dev_flags
>> has RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS
>> Rx/tx queue stats are already added. I am pushing a fresh patch for
>> adding rx/tx queue stats.
>> 
> 
> Hi Levend,
> 
> You are right BUT, 'RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS' is a temporary
> solution and plan is to remove it [1].
> 
> Background is, queue stats in "struct rte_eth_stats" has fixed size and
> as number of queues supported by devices increase these fields getting
> bigger and bigger, the solution we came was to completely remove these
> fields from stats struct and get queue statistics via xstats.
> 
> During transition 'RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS' is introduced
> until all drivers implement queue stats in xstats. We are not pushing
> hard for existing drivers to update but at least requiring new drivers
> to implement xstats method.
> 
> That is why for net/gve updating queue stats in 'gve_dev_stats_get()'
> rejected before, and xstats implementation is requested.
> 
> 
> [1]
> https://elixir.bootlin.com/dpdk/v22.11.1/source/doc/guides/rel_notes/deprecation.rst#L88
> 
> 
>> And also noticed a missing part at rx no_mbufs counting.
>> 
>> Best,
>> Levend
>>  
>> 
>>> On 17 Feb 2023, at 15:34, Ferruh Yigit  wrote:
>>> 
>>> On 2/16/2023 6:58 PM, Levend Sayar wrote:
 Google Virtual NIC PMD is enriched with extended statistics info.
>>> 
>>> Only queue stats added to xstats, can you please highlight this in the
>>> commit log?
>>> 
 eth_dev_ops callback names are also synched with eth_dev_ops field names
 
>>> 
>>> Renaming eth_dev_ops is not related to xstats, and I am not sure if the
>>> change is necessary, can you please drop it from this patch?
>>> 
 Signed-off-by: Levend Sayar 
 ---
 drivers/net/gve/gve_ethdev.c | 152 ++-
 drivers/net/gve/gve_rx.c |   8 +-
 2 files changed, 138 insertions(+), 22 deletions(-)
 
 diff --git a/drivers/net/gve/gve_ethdev.c b/drivers/net/gve/gve_ethdev.c
 index fef2458a16..e31fdce960 100644
 --- a/drivers/net/gve/gve_ethdev.c
 +++ b/drivers/net/gve/gve_ethdev.c
 @@ -266,7 +266,7 @@ gve_dev_close(struct rte_eth_dev *dev)
 }
 
 static int
 -gve_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info
 *dev_info)
 +gve_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info
 *dev_info)
 {
 struct gve_priv *priv = dev->data->dev_private;
 
 @@ -319,15 +319,12 @@ gve_dev_info_get(struct rte_eth_dev *dev,
 struct rte_eth_dev_info *dev_info)
 }
 
 static int
 -gve_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 +gve_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 {
 uint16_t i;
 
 for (i = 0; i < dev->data->nb_tx_queues; i++) {
 struct gve_tx_queue *txq = dev->data->tx_queues[i];
 -if (txq == NULL)
 -continue;
 -
>>> 
>>> I assume check is removed because it is unnecessary, but again not
>>> directly related with the patch, can you also drop these from the patch
>>> to reduce the noise.
>>> 
 stats->opackets += txq->packets;
 stats->obytes += txq->bytes;
 stats->oerrors += txq->errors;
 @@ -335,9 +332,6 @@ gve_dev_stats_get(struct rte_eth_dev *dev, struct
 rte_eth_stats *stats)
 
 for (i = 0; i < dev->data->nb_rx_queues; i++) {
 struct gve_rx_queue *rxq = dev->data->rx_queues[i];
 -if (rxq == NULL)
 -continue;
 -
 stats->ipackets += rxq->packets;
 stats->ibytes += rxq->bytes;
 stats->ierrors += rxq->errors;
 @@ -348,15 +342,12 @@ gve_dev_stats_get(struct rte_eth_dev *dev,
 struct rte_eth_stats *stats)
 }
 
 static int
 -gve_dev_stats_reset(struct rte_eth_dev *dev)
 +gve_stats_reset(struct rte_eth_dev *dev)
 {
 uint16_t i;
 
 for (i = 0; i < dev->data->nb_tx_queues; i++) {
 struct gve_tx_queue *txq = dev->data->tx_queues[i];
 -if (txq == NULL)
 -continue;
 -
 txq->packets  = 0;
 txq->bytes = 0;
 txq->errors = 0;
 @@ -364,9 +355,6 @@ gve_dev_stats_reset(struct rte_eth_dev *dev)
 
 for (i = 0; i < dev->data->nb_rx_queues; i++) {
 struct gve_rx_queue *rxq = dev->data->rx_queues[i];
 -if (rxq == NULL)
 -continue;
 -
 rxq->packets  = 0;
 rxq->bytes = 0;
 rxq->errors = 0;
 @@ -377,7 +365,7 @@ gve_dev_stats_reset(struct rte_eth_dev *dev)
 }
 
 

Re: [PATCH] net/gve: fix Rx no mbufs stats counter update

[Levend Sayar]

> On 19 Feb 2023, at 20:35, Stephen Hemminger  
> wrote:
> 
> On Sun, 19 Feb 2023 03:30:59 +0300
> Levend Sayar mailto:levendsa...@gmail.com>> wrote:
> 
>> rx no_mbufs stats counter update is added for another error case.
>> 
>> Fixes: 4f6b1dd8240c ("net/gve: support basic statistics")
>> Cc: junfeng@intel.com
>> 
>> Signed-off-by: Levend Sayar 
>> ---
>> drivers/net/gve/gve_rx.c | 7 ++-
>> 1 file changed, 6 insertions(+), 1 deletion(-)
>> 
>> diff --git a/drivers/net/gve/gve_rx.c b/drivers/net/gve/gve_rx.c
>> index 66fbcf3930..b0427731f8 100644
>> --- a/drivers/net/gve/gve_rx.c
>> +++ b/drivers/net/gve/gve_rx.c
>> @@ -24,6 +24,7 @@ gve_rx_refill(struct gve_rx_queue *rxq)
>>  nmb = rte_pktmbuf_alloc(rxq->mpool);
>>  if (!nmb)
>>  break;
>> +
>>  rxq->sw_ring[idx + i] = nmb;
>>  }
>>  if (i != nb_alloc) {
> 
> Looks like accidental whitespace change included in this patch.

LS: Right. Let me correct.

>> @@ -59,9 +60,13 @@ gve_rx_refill(struct gve_rx_queue *rxq)
>>  nmb = rte_pktmbuf_alloc(rxq->mpool);
>>  if (!nmb)
>>  break;
>> +
>>  rxq->sw_ring[idx + i] = nmb;
>>  }
>> -nb_alloc = i;
>> +if (i != nb_alloc) {
>> +rxq->no_mbufs += nb_alloc - i;
>> +nb_alloc = i;
>> +}
> 
> Would be better to add unlikely() here like:
>   if (unlikely(i < nb_alloc)) {
>   rxq->no_mbufs += nb_alloc - i;
>   nb_alloc = i;
>   }
> 
> Or eliminate conditional branch in hot path completely.
>   rxq->no_mbufs += nb_alloc - i;
>   nb_alloc = i;
> 
> Or better yet refactor code here to use rte_pktmbuf_alloc_bulk() which
> does single ring operation.
> 
>>  }
>>  rxq->nb_avail -= nb_alloc;
>>  next_avail += nb_alloc;

LS: “unlikely” can be added. You’re right. Code already tries to make a bulk 
allocation first.
If that bulk allocation does not work, it tries to allocate one my one. 

I will supersede this one and create v2.
Thanks Stephen.

Best,
Levend

Re: [PATCH] pipeline: fix IPsec crypto session leak

[Thomas Monjalon]

07/02/2023 19:15, Cristian Dumitrescu:
> In case of rte_ipsec_session_prepare() error, the crypto session was
> not freed, leading to its leakage.
> 
> Fixes: 20777eb5f913 ("pipeline: add IPsec")
> Coverity issue: 383139
> 
> Signed-off-by: Cristian Dumitrescu 

Applied, thanks.

Re: [PATCH V4] pipeline: add RSS support

[Thomas Monjalon]

07/02/2023 17:41, Cristian Dumitrescu:
> Add pipeline support for the Receive Side Scaling (RSS) hashing. While
> the pipeline already supports the stateless hashing schemes, the RSS
> scheme uses a key configured by the control plane and preserved
> between successive RSS hash invocations.
> 
> Signed-off-by: Cristian Dumitrescu 
> Signed-off-by: Kamalakannan R 

Applied, thanks.

Re: [PATCH] pipeline: support conversion between big and small fields

[Thomas Monjalon]

10/02/2023 20:51, Cristian Dumitrescu:
> While the small fields (size <= 64 bits) can be in either network byte
> order (header fields) or host byte order (meta-data fields), the big
> fields (size > 64 bits) are always expected to be in network byte
> order (big endian), both header and metadata fields.
> 
> Previously, a big field could be involved in an assignment only when
> both the destination and the source are big fields of exactly the same
> size. This restriction is now relaxed significantly: in case a big
> field is involved as either destination or source, the other field can
> now be either a big field (of the same or different size) or a small
> header field, so in either case both destination and source are in the
> network byte order.
> 
> Signed-off-by: Cristian Dumitrescu 
> Signed-off-by: Kamalakannan R 

Applied, thanks.

Re: [PATCH] eal/bsd: fix spinlock not unlock in alarm callback

[Thomas Monjalon]

16/02/2023 10:22, Bruce Richardson:
> On Thu, Feb 16, 2023 at 09:28:17AM +0800, fengchengwen wrote:
> > add cc maintainer
> > 
> > On 2023/2/13 20:44, Chengwen Feng wrote:
> > > The spinlock should unlock when clock_gettime() failed. This patch
> > > fixes it by invoking clock_gettime() before lock.
> > > 
> > > Fixes: 26021a715067 ("eal/bsd: support alarm API")
> > > Cc: sta...@dpdk.org
> > > 
> > > Signed-off-by: Chengwen Feng 
> 
> Acked-by: Bruce Richardson 

Applied, thanks.

Re: [PATCH v2] sched: fix for incorrect alignment of bitmap, pipe and queue structs in subport

[Thomas Monjalon]

07/02/2023 07:10, Megha Ajmera:
> Big structures like bitmap, pipes and queues in subport are addressed
> using offset of 'memory' field in subport structures. This means no other
> variable should be added after 'memory' variable or else addressing of such
> structs like bitmap etc. become incorrect.
> 
> Realigned tc_ov_enabled variable in subport structure.
> 
> Fixes: f5e60154ade ("sched: enable traffic class oversubscription 
> conditionally")
> 
> Signed-off-by: Megha Ajmera 
> Acked-by: Cristian Dumitrescu 

I bet the impact is wrongly explained, but I don't want to look deeper.
Applied.

Re: [PATCH v2] sched: fix for incorrect alignment of bitmap, pipe and queue structs in subport

[Thomas Monjalon]

19/02/2023 23:40, Thomas Monjalon:
> 07/02/2023 07:10, Megha Ajmera:
> > Big structures like bitmap, pipes and queues in subport are addressed
> > using offset of 'memory' field in subport structures. This means no other
> > variable should be added after 'memory' variable or else addressing of such
> > structs like bitmap etc. become incorrect.
> > 
> > Realigned tc_ov_enabled variable in subport structure.
> > 
> > Fixes: f5e60154ade ("sched: enable traffic class oversubscription 
> > conditionally")
> > 
> > Signed-off-by: Megha Ajmera 
> > Acked-by: Cristian Dumitrescu 
> 
> I bet the impact is wrongly explained, but I don't want to look deeper.
> Applied.

And I've added Cc: sta...@dpdk.org that David and Cristian asked for.

[PATCH v2] lib/net: add MPLS insert and strip functionality

[Tanzeel-inline]

From: Tanzeel Ahmed 

This patch is new version of [PATCH] lib/net: added push MPLS header API.
I have also added the MPLS strip functionality in the same patch to
address the question asked in last patch.

> You should explain why you add this function.
None of the foundational NICs currently supports MPLS insertion and
stripping, this functionality can help users who rely on MPLS in their
network application.

> I'm not sure it should be inline
I did for perfomance in high-traffic application.

Signed-off-by: Tanzeel Ahmed 

---
v2:
* marked experimental
* coding style fixed
* changed rte_memcpy to memcpy
* mpls header marked as const in parameter
* added MPLS stripping functionality
---
 .mailmap   |  1 +
 lib/net/rte_mpls.h | 97 ++
 2 files changed, 98 insertions(+)

diff --git a/.mailmap b/.mailmap
index a9f4f28..2af4e0d 100644
--- a/.mailmap
+++ b/.mailmap
@@ -1312,6 +1312,7 @@ Takeshi Yoshimura  

 Takuya Asada 
 Tal Avraham 
 Tal Shnaiderman  
+Tanzeel Ahmed 
 Tao Y Yang 
 Tao Zhu 
 Taripin Samuel 
diff --git a/lib/net/rte_mpls.h b/lib/net/rte_mpls.h
index 51523e7..b51520f 100644
--- a/lib/net/rte_mpls.h
+++ b/lib/net/rte_mpls.h
@@ -13,6 +13,8 @@
 
 #include 
 #include 
+#include 
+#include 
 
 #ifdef __cplusplus
 extern "C" {
@@ -36,6 +38,101 @@ struct rte_mpls_hdr {
uint8_t  ttl;   /**< Time to live. */
 } __rte_packed;
 
+#define RTE_MPLS_HLEN 4 /**< Length of MPLS header. */
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Insert MPLS header into the packet.
+ * If it's first MPLS header to be inserted in the packet,
+ *  - Updates the ether type.
+ *  - Sets the MPLS bottom-of-stack bit to 1.
+ *
+ * @param m
+ *   The pointer to the mbuf.
+ * @param mp
+ *   The pointer to the MPLS header.
+ * @return
+ *   0 on success, -1 on error
+ */
+__rte_experimental
+static int
+rte_mpls_push_over_l2(struct rte_mbuf **m, const struct rte_mpls_hdr *mp)
+{
+   struct rte_ether_hdr *oh, *nh;
+   struct rte_mpls_hdr *mph;
+
+   /* Can't insert header if mbuf is shared */
+   if (!RTE_MBUF_DIRECT(*m) || rte_mbuf_refcnt_read(*m) > 1)
+   return -EINVAL;
+
+   /* Can't insert header if ethernet frame doesn't exist */
+   if (rte_pktmbuf_data_len(*m) < RTE_ETHER_HDR_LEN)
+   return -EINVAL;
+
+   oh = rte_pktmbuf_mtod(*m, struct rte_ether_hdr *);
+   nh = (struct rte_ether_hdr *)(void *)
+   rte_pktmbuf_prepend(*m, sizeof(struct rte_mpls_hdr));
+   if (nh == NULL)
+   return -ENOSPC;
+
+   memmove(nh, oh, RTE_ETHER_HDR_LEN);
+
+   /* Copy the MPLS header after ethernet frame */
+   mph = rte_pktmbuf_mtod_offset(*m, struct rte_mpls_hdr*,
+   sizeof(struct rte_ether_hdr));
+   memcpy(mph, mp, RTE_MPLS_HLEN);
+
+   mph->tag_msb = rte_cpu_to_be_16(mp->tag_msb);
+
+   /* If first MPLS header, update ether type and bottom-of-stack bit */
+   if (nh->ether_type != rte_cpu_to_be_16(RTE_ETHER_TYPE_MPLS)) {
+   nh->ether_type = rte_cpu_to_be_16(RTE_ETHER_TYPE_MPLS);
+   mph->bs = 1;
+   } else {
+   mph->bs = 0;
+   }
+
+   return 0;
+}
+
+/**
+ * @warning
+ * @b EXPERIMENTAL: this API may change without prior notice.
+ *
+ * Strips MPLS from the packet. Doesn't update the ether type
+ *
+ * @param m
+ *   The pointer to the mbuf.
+ * @return
+ *   0 on success, -1 on error
+ */
+__rte_experimental
+static inline int
+rte_mpls_strip_over_l2(struct rte_mbuf *m)
+{
+   struct rte_ether_hdr *eh = rte_pktmbuf_mtod(m, struct rte_ether_hdr *);
+   struct rte_mpls_hdr *mph;
+   bool mpls_exist = true;
+
+   if (eh->ether_type != rte_cpu_to_be_16(RTE_ETHER_TYPE_MPLS))
+   return -1;
+
+   /* Stripping all MPLS header */
+   while (mpls_exist) {
+   mph = rte_pktmbuf_mtod_offset(m, struct rte_mpls_hdr*,
+   sizeof(struct rte_ether_hdr));
+   if (mph->bs & 1)
+   mpls_exist = false;
+   memmove(rte_pktmbuf_adj(m, sizeof(struct rte_mpls_hdr)),
+   eh, sizeof(struct rte_ether_hdr));
+   eh = rte_pktmbuf_mtod(m, struct rte_ether_hdr *);
+   }
+
+   return 0;
+}
+
 #ifdef __cplusplus
 }
 #endif
-- 
1.8.3.1

Re: [PATCH] net/gve: fix Rx no mbufs stats counter update

[Stephen Hemminger]

On Sun, 19 Feb 2023 23:43:08 +0300
Levend Sayar  wrote:

> LS: “unlikely” can be added. You’re right. Code already tries to make a bulk 
> allocation first.
> If that bulk allocation does not work, it tries to allocate one my one.

That seems like a unnecessary step and unlikely to help.
Unless the user abuses driver by giving a very small mbuf pool.

Re: [PATCH v3 0/6] dma/ioat: fix issues with stopping and restarting device

[Thomas Monjalon]

> Bruce Richardson (6):
>   dma/ioat: fix device stop if no copies done
>   dma/ioat: fix incorrectly set indexes after restart
>   dma/ioat: fix incorrect error reporting on restart
>   test/dmadev: check result for device stop
>   test/dmadev: create separate function for single copy test
>   test/dmadev: add tests for stopping and restarting dev

Applied, thanks.

[PATCH v4 0/9] Enable building more libraries on Windows

[Stephen Hemminger]

While diagnosing some Windows cross build errors;
noticed that lots of important DPDK libraries are not
being built on Windows.

Stephen Hemminger (8):
  net/null: build null PMD on Windows
  net/ring: build on Windows
  lpm: enable on Windows
  reorder: build on Windows
  ip_frag: enable build on Windows
  rib: enable on Windows
  fib: enable on Windows
  pcapng: windows compatibility

v4 - fix spelling in commnet
 enable build of lib/table as well

v3 - add rte_os_shim.h to rte_ring to fix use of
 strdup() breaking with stricter MS compiler

v2 - fix unnecessary PATH_MAX in net/ring driver

Stephen Hemminger (9):
  net/null: build null PMD on Windows
  net/ring: build on Windows
  lpm: enable on Windows
  reorder: build on Windows
  ip_frag: enable build on Windows
  rib: enable on Windows
  fib: enable on Windows
  pcapng: windows compatibility
  table: enable build on Windows

 app/test/meson.build|  2 +-
 app/test/test_fib.c | 22 +---
 app/test/test_fib6.c| 24 ++
 app/test/test_fib6_perf.c   | 15 +
 app/test/test_fib_perf.c| 19 ++-
 app/test/test_ipfrag.c  | 12 ---
 app/test/test_lpm.c | 15 +
 app/test/test_lpm6.c| 12 ---
 app/test/test_lpm6_data.h   |  3 +-
 app/test/test_lpm6_perf.c   | 14 +---
 app/test/test_lpm_perf.c| 39 +++---
 app/test/test_reorder.c | 11 --
 app/test/test_rib.c | 19 ---
 app/test/test_rib6.c| 24 ++
 drivers/net/null/meson.build|  6 
 drivers/net/ring/meson.build|  6 
 drivers/net/ring/rte_eth_ring.c |  4 ++-
 lib/fib/meson.build |  6 
 lib/fib/rte_fib.c   |  1 +
 lib/fib/rte_fib6.c  |  1 +
 lib/ip_frag/ip_frag_common.h|  2 ++
 lib/ip_frag/meson.build |  6 
 lib/lpm/meson.build |  6 
 lib/pcapng/meson.build  |  6 
 lib/pcapng/rte_pcapng.c | 59 +++--
 lib/reorder/meson.build |  6 
 lib/reorder/rte_reorder.c   |  1 +
 lib/rib/meson.build |  6 
 lib/rib/rte_rib.c   |  1 +
 lib/rib/rte_rib6.c  |  1 +
 lib/table/meson.build   |  6 
 31 files changed, 93 insertions(+), 262 deletions(-)

-- 
2.39.1

[PATCH v4 1/9] net/null: build null PMD on Windows

[Stephen Hemminger]

Builds fine with current code, no changes needed.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 drivers/net/null/meson.build | 6 --
 1 file changed, 6 deletions(-)

diff --git a/drivers/net/null/meson.build b/drivers/net/null/meson.build
index 4a483955a7a9..a51f8f5211b0 100644
--- a/drivers/net/null/meson.build
+++ b/drivers/net/null/meson.build
@@ -1,11 +1,5 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2017 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files('rte_eth_null.c')
 pmd_supports_disable_iova_as_pa = true
-- 
2.39.1

[PATCH v4 2/9] net/ring: build on Windows

[Stephen Hemminger]

The command line arguments are stored in node_action_pair
and the name[] was sized to PATH_MAX which does not exist on Windows.
Since the name is either "CREATE" or "ATTACH" it is not
related to PATH_MAX (4096).

With this fix driver builds ok on windows, but need to modify the
test meson build to skip the eventdev test on Windows.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 app/test/meson.build| 2 +-
 drivers/net/ring/meson.build| 6 --
 drivers/net/ring/rte_eth_ring.c | 4 +++-
 3 files changed, 4 insertions(+), 8 deletions(-)

diff --git a/app/test/meson.build b/app/test/meson.build
index f34d19e3c3cb..a713f0382280 100644
--- a/app/test/meson.build
+++ b/app/test/meson.build
@@ -390,7 +390,7 @@ if dpdk_conf.has('RTE_NET_BOND')
 driver_test_names += 'link_bonding_mode4_autotest'
 endif
 endif
-if dpdk_conf.has('RTE_NET_RING')
+if dpdk_conf.has('RTE_LIB_EVENTDEV') and dpdk_conf.has('RTE_NET_RING')
 test_deps += 'net_ring'
 test_sources += 'test_pmd_ring_perf.c'
 test_sources += 'test_pmd_ring.c'
diff --git a/drivers/net/ring/meson.build b/drivers/net/ring/meson.build
index 72792e26b05a..3534a3cc2287 100644
--- a/drivers/net/ring/meson.build
+++ b/drivers/net/ring/meson.build
@@ -1,12 +1,6 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2017 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files('rte_eth_ring.c')
 headers = files('rte_eth_ring.h')
 pmd_supports_disable_iova_as_pa = true
diff --git a/drivers/net/ring/rte_eth_ring.c b/drivers/net/ring/rte_eth_ring.c
index bd5a47dd9017..e8bc9b627102 100644
--- a/drivers/net/ring/rte_eth_ring.c
+++ b/drivers/net/ring/rte_eth_ring.c
@@ -9,6 +9,7 @@
 #include 
 #include 
 #include 
+#include 
 #include 
 #include 
 #include 
@@ -17,6 +18,7 @@
 #define ETH_RING_NUMA_NODE_ACTION_ARG  "nodeaction"
 #define ETH_RING_ACTION_CREATE "CREATE"
 #define ETH_RING_ACTION_ATTACH "ATTACH"
+#define ETH_RING_ACTION_MAX_LEN8 /* CREATE | ACTION */
 #define ETH_RING_INTERNAL_ARG  "internal"
 #define ETH_RING_INTERNAL_ARG_MAX_LEN  19 /* "0x..16chars..\0" */
 
@@ -539,7 +541,7 @@ eth_dev_ring_create(const char *name,
 }
 
 struct node_action_pair {
-   char name[PATH_MAX];
+   char name[ETH_RING_ACTION_MAX_LEN];
unsigned int node;
enum dev_action action;
 };
-- 
2.39.1

[PATCH v4 3/9] lpm: enable on Windows

[Stephen Hemminger]

This builds on Windows without changes.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 app/test/test_lpm.c   | 15 +--
 app/test/test_lpm6.c  | 12 
 app/test/test_lpm6_data.h |  3 ++-
 app/test/test_lpm6_perf.c | 14 +-
 app/test/test_lpm_perf.c  | 39 ---
 lib/lpm/meson.build   |  6 --
 6 files changed, 16 insertions(+), 73 deletions(-)

diff --git a/app/test/test_lpm.c b/app/test/test_lpm.c
index bceb9ae743d7..37b460af3a96 100644
--- a/app/test/test_lpm.c
+++ b/app/test/test_lpm.c
@@ -2,18 +2,6 @@
  * Copyright(c) 2010-2014 Intel Corporation
  */
 
-#include "test.h"
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_lpm(void)
-{
-   printf("lpm not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
 #include 
 #include 
 #include 
@@ -22,6 +10,7 @@ test_lpm(void)
 #include 
 #include 
 
+#include "test.h"
 #include "test_xmmt_ops.h"
 
 #define TEST_LPM_ASSERT(cond) do {\
@@ -1595,6 +1584,4 @@ test_lpm(void)
return global_status;
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(lpm_autotest, test_lpm);
diff --git a/app/test/test_lpm6.c b/app/test/test_lpm6.c
index 1f44392db739..b6b6f8615ea5 100644
--- a/app/test/test_lpm6.c
+++ b/app/test/test_lpm6.c
@@ -10,16 +10,6 @@
 #include 
 
 #include 
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_lpm6(void)
-{
-   printf("lpm6 not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
 #include 
 
 #include "test_lpm6_data.h"
@@ -1803,6 +1793,4 @@ test_lpm6(void)
return global_status;
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(lpm6_autotest, test_lpm6);
diff --git a/app/test/test_lpm6_data.h b/app/test/test_lpm6_data.h
index da9b161f20e5..3a1fa25f8433 100644
--- a/app/test/test_lpm6_data.h
+++ b/app/test/test_lpm6_data.h
@@ -6,6 +6,7 @@
 
 #include 
 #include 
+#include 
 
 struct rules_tbl_entry {
uint8_t ip[16];
@@ -1129,7 +1130,7 @@ static void generate_large_ips_table(int 
gen_expected_next_hop)
 
for (i = 0; i < NUM_IPS_ENTRIES; i++) {
for (j = 0; j < 16; j++)
-   large_ips_table[i].ip[j] = lrand48();
+   large_ips_table[i].ip[j] = rte_rand();
}
 
for (k = j = 0, i = 0; i < NUM_IPS_ENTRIES; i++) {
diff --git a/app/test/test_lpm6_perf.c b/app/test/test_lpm6_perf.c
index aaf2773b6fac..5b684686a687 100644
--- a/app/test/test_lpm6_perf.c
+++ b/app/test/test_lpm6_perf.c
@@ -2,17 +2,6 @@
  * Copyright(c) 2010-2014 Intel Corporation
  */
 
-#include "test.h"
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_lpm6_perf(void)
-{
-   printf("lpm6_perf not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
 
 #include 
 #include 
@@ -24,6 +13,7 @@ test_lpm6_perf(void)
 #include 
 #include 
 
+#include "test.h"
 #include "test_lpm6_data.h"
 
 #define TEST_LPM_ASSERT(cond) do {\
@@ -171,6 +161,4 @@ test_lpm6_perf(void)
return 0;
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(lpm6_perf_autotest, test_lpm6_perf);
diff --git a/app/test/test_lpm_perf.c b/app/test/test_lpm_perf.c
index e858716f909b..e72437ba3850 100644
--- a/app/test/test_lpm_perf.c
+++ b/app/test/test_lpm_perf.c
@@ -3,17 +3,6 @@
  * Copyright(c) 2020 Arm Limited
  */
 
-#include "test.h"
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_lpm_perf(void)
-{
-   printf("lpm_perf not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
 #include 
 #include 
 #include 
@@ -25,7 +14,9 @@ test_lpm_perf(void)
 #include 
 #include 
 #include 
+#include 
 
+#include "test.h"
 #include "test_xmmt_ops.h"
 
 struct rte_lpm *lpm;
@@ -34,8 +25,9 @@ static volatile uint8_t writer_done;
 static volatile uint32_t thr_id;
 static uint64_t gwrite_cycles;
 static uint32_t num_writers;
-/* LPM APIs are not thread safe, use mutex to provide thread safety */
-static pthread_mutex_t lpm_mutex = PTHREAD_MUTEX_INITIALIZER;
+
+/* LPM APIs are not thread safe, use spinlock */
+static rte_spinlock_t lpm_lock = RTE_SPINLOCK_INITIALIZER;
 
 /* Report quiescent state interval every 1024 lookups. Larger critical
  * sections in reader will result in writer polling multiple times.
@@ -267,7 +259,7 @@ static void generate_random_rule_prefix(uint32_t ip_class, 
uint8_t depth)
/* Only generate rest bits except the most significant
 * fixed bits for IP address class
 */
-   start = lrand48() & mask;
+   start = rte_rand() & mask;
ptr_rule = &large_route_table[num_route_entries];
ptr_ldepth_rule = &large_ldepth_route_table[num_ldepth_route_entries];
for (k = 0; k < rule_num; k++) {
@@ -296,7 +288,7 @@ static void insert_rule_in_random_pos(uint32_t ip, uint8_t 
depth)
struct route_rule tmp;
 
  

[PATCH v4 4/9] reorder: build on Windows

[Stephen Hemminger]

This builds on Windows if sys/queue.h is included.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 app/test/test_reorder.c   | 11 ---
 lib/reorder/meson.build   |  6 --
 lib/reorder/rte_reorder.c |  1 +
 3 files changed, 1 insertion(+), 17 deletions(-)

diff --git a/app/test/test_reorder.c b/app/test/test_reorder.c
index f0714a5c18a3..9ebd91477536 100644
--- a/app/test/test_reorder.c
+++ b/app/test/test_reorder.c
@@ -11,16 +11,6 @@
 #include 
 #include 
 #include 
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_reorder(void)
-{
-   printf("reorder not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-#else
-
 #include 
 #include 
 #include 
@@ -395,6 +385,5 @@ test_reorder(void)
return unit_test_suite_runner(&reorder_test_suite);
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
 
 REGISTER_TEST_COMMAND(reorder_autotest, test_reorder);
diff --git a/lib/reorder/meson.build b/lib/reorder/meson.build
index 621c1f350103..03aed53d9007 100644
--- a/lib/reorder/meson.build
+++ b/lib/reorder/meson.build
@@ -1,12 +1,6 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2017 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files('rte_reorder.c')
 headers = files('rte_reorder.h')
 deps += ['mbuf']
diff --git a/lib/reorder/rte_reorder.c b/lib/reorder/rte_reorder.c
index 385ee479da42..00e2cf043e04 100644
--- a/lib/reorder/rte_reorder.c
+++ b/lib/reorder/rte_reorder.c
@@ -3,6 +3,7 @@
  */
 
 #include 
+#include 
 
 #include 
 #include 
-- 
2.39.1

[PATCH v4 5/9] ip_frag: enable build on Windows

[Stephen Hemminger]

This build works on Windows if sys/queue.h is included.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 app/test/test_ipfrag.c   | 12 
 lib/ip_frag/ip_frag_common.h |  2 ++
 lib/ip_frag/meson.build  |  6 --
 3 files changed, 2 insertions(+), 18 deletions(-)

diff --git a/app/test/test_ipfrag.c b/app/test/test_ipfrag.c
index 88cc4cdeea8d..402ce361c1a2 100644
--- a/app/test/test_ipfrag.c
+++ b/app/test/test_ipfrag.c
@@ -10,17 +10,6 @@
 #include 
 #include 
 #include 
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_ipfrag(void)
-{
-   printf("ipfrag not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
 #include 
 #include 
 #include 
@@ -520,6 +509,5 @@ test_ipfrag(void)
return unit_test_suite_runner(&ipfrag_testsuite);
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
 
 REGISTER_TEST_COMMAND(ipfrag_autotest, test_ipfrag);
diff --git a/lib/ip_frag/ip_frag_common.h b/lib/ip_frag/ip_frag_common.h
index 9c0dbdeb6eb9..0f4ad72a90d9 100644
--- a/lib/ip_frag/ip_frag_common.h
+++ b/lib/ip_frag/ip_frag_common.h
@@ -8,6 +8,8 @@
 #include "rte_ip_frag.h"
 #include "ip_reassembly.h"
 
+#include 
+
 /* logging macros. */
 #ifdef RTE_LIBRTE_IP_FRAG_DEBUG
 #defineIP_FRAG_LOG(lvl, fmt, args...)  RTE_LOG(lvl, USER1, fmt, ##args)
diff --git a/lib/ip_frag/meson.build b/lib/ip_frag/meson.build
index 3a252bdaf6a4..ea2de09f7528 100644
--- a/lib/ip_frag/meson.build
+++ b/lib/ip_frag/meson.build
@@ -1,12 +1,6 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2017 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files(
 'rte_ipv4_fragmentation.c',
 'rte_ipv6_fragmentation.c',
-- 
2.39.1

[PATCH v4 6/9] rib: enable on Windows

[Stephen Hemminger]

The RIB library builds on Windows as long as sys/queue.h is included

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 app/test/test_rib.c  | 19 ---
 app/test/test_rib6.c | 24 ++--
 lib/rib/meson.build  |  6 --
 lib/rib/rte_rib.c|  1 +
 lib/rib/rte_rib6.c   |  1 +
 5 files changed, 4 insertions(+), 47 deletions(-)

diff --git a/app/test/test_rib.c b/app/test/test_rib.c
index 06058f8f7c52..65b685641094 100644
--- a/app/test/test_rib.c
+++ b/app/test/test_rib.c
@@ -10,23 +10,6 @@
 #include 
 
 #include 
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_rib(void)
-{
-   printf("rib not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-static int
-test_slow_rib(void)
-{
-   printf("slow_rib not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-#else
-
 #include 
 
 typedef int32_t (*rte_rib_test)(void);
@@ -380,7 +363,5 @@ test_slow_rib(void)
return unit_test_suite_runner(&rib_slow_tests);
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(rib_autotest, test_rib);
 REGISTER_TEST_COMMAND(rib_slow_autotest, test_slow_rib);
diff --git a/app/test/test_rib6.c b/app/test/test_rib6.c
index ba79aedea5f7..336b779d2e1e 100644
--- a/app/test/test_rib6.c
+++ b/app/test/test_rib6.c
@@ -3,32 +3,14 @@
  * Copyright(c) 2019 Intel Corporation
  */
 
-#include "test.h"
-
 #include 
 #include 
 #include 
-
 #include 
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_rib6(void)
-{
-   printf("rib6 not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-static int
-test_slow_rib6(void)
-{
-   printf("slow_rib6 not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-#else
-
 #include 
 
+#include "test.h"
+
 typedef int32_t (*rte_rib6_test)(void);
 
 static int32_t test_create_invalid(void);
@@ -385,7 +367,5 @@ test_slow_rib6(void)
return unit_test_suite_runner(&rib6_slow_tests);
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(rib6_autotest, test_rib6);
 REGISTER_TEST_COMMAND(rib6_slow_autotest, test_slow_rib6);
diff --git a/lib/rib/meson.build b/lib/rib/meson.build
index bda7f576e976..7bacbb453592 100644
--- a/lib/rib/meson.build
+++ b/lib/rib/meson.build
@@ -2,12 +2,6 @@
 # Copyright(c) 2018 Vladimir Medvedkin 
 # Copyright(c) 2019 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files('rte_rib.c', 'rte_rib6.c')
 headers = files('rte_rib.h', 'rte_rib6.h')
 deps += ['mempool']
diff --git a/lib/rib/rte_rib.c b/lib/rib/rte_rib.c
index b0794edf66f5..812a2597d117 100644
--- a/lib/rib/rte_rib.c
+++ b/lib/rib/rte_rib.c
@@ -5,6 +5,7 @@
 
 #include 
 #include 
+#include 
 
 #include 
 #include 
diff --git a/lib/rib/rte_rib6.c b/lib/rib/rte_rib6.c
index 19e4ff97c479..ae44281ae105 100644
--- a/lib/rib/rte_rib6.c
+++ b/lib/rib/rte_rib6.c
@@ -5,6 +5,7 @@
 
 #include 
 #include 
+#include 
 
 #include 
 #include 
-- 
2.39.1

[PATCH v4 7/9] fib: enable on Windows

[Stephen Hemminger]

The FIB library builds on Windows as long as sys/queue.h is defined.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 app/test/test_fib.c   | 22 +-
 app/test/test_fib6.c  | 24 ++--
 app/test/test_fib6_perf.c | 15 +--
 app/test/test_fib_perf.c  | 19 +++
 lib/fib/meson.build   |  6 --
 lib/fib/rte_fib.c |  1 +
 lib/fib/rte_fib6.c|  1 +
 7 files changed, 9 insertions(+), 79 deletions(-)

diff --git a/app/test/test_fib.c b/app/test/test_fib.c
index a2d1ea8f3abc..eb69d6e2fd4c 100644
--- a/app/test/test_fib.c
+++ b/app/test/test_fib.c
@@ -9,28 +9,10 @@
 
 #include 
 #include 
+#include 
 
 #include "test.h"
 
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_fib(void)
-{
-   printf("fib not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-static int
-test_slow_fib(void)
-{
-   printf("slow_fib not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
-#include 
-
 typedef int32_t (*rte_fib_test)(void);
 
 static int32_t test_create_invalid(void);
@@ -433,7 +415,5 @@ test_slow_fib(void)
return unit_test_suite_runner(&fib_slow_tests);
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(fib_autotest, test_fib);
 REGISTER_TEST_COMMAND(fib_slow_autotest, test_slow_fib);
diff --git a/app/test/test_fib6.c b/app/test/test_fib6.c
index cd971e6ecdfb..15ad09178ae2 100644
--- a/app/test/test_fib6.c
+++ b/app/test/test_fib6.c
@@ -9,29 +9,11 @@
 
 #include 
 #include 
-
-#include "test.h"
-
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_fib6(void)
-{
-   printf("fib not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-static int
-test_slow_fib6(void)
-{
-   printf("slow_fib not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
 #include 
 #include 
 
+#include "test.h"
+
 typedef int32_t (*rte_fib6_test)(void);
 
 static int32_t test_create_invalid(void);
@@ -442,7 +424,5 @@ test_slow_fib6(void)
return unit_test_suite_runner(&fib6_slow_tests);
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(fib6_autotest, test_fib6);
 REGISTER_TEST_COMMAND(fib6_slow_autotest, test_slow_fib6);
diff --git a/app/test/test_fib6_perf.c b/app/test/test_fib6_perf.c
index 21d2b65318e9..add20c2331b1 100644
--- a/app/test/test_fib6_perf.c
+++ b/app/test/test_fib6_perf.c
@@ -10,21 +10,10 @@
 #include 
 #include 
 #include 
+#include 
 
 #include "test.h"
 
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_fib6_perf(void)
-{
-   printf("fib6_perf not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
-#include 
-
 #include "test_lpm6_data.h"
 
 #define TEST_FIB_ASSERT(cond) do { \
@@ -167,6 +156,4 @@ test_fib6_perf(void)
return 0;
 }
 
-#endif /*ifdef RTE_EXEC_ENV_WINDOWS*/
-
 REGISTER_TEST_COMMAND(fib6_perf_autotest, test_fib6_perf);
diff --git a/app/test/test_fib_perf.c b/app/test/test_fib_perf.c
index 9787874cc9b7..b56293e64f41 100644
--- a/app/test/test_fib_perf.c
+++ b/app/test/test_fib_perf.c
@@ -12,22 +12,11 @@
 #include 
 #include 
 #include 
+#include 
 
 #include "test.h"
 #include "test_xmmt_ops.h"
 
-#ifdef RTE_EXEC_ENV_WINDOWS
-static int
-test_fib_perf(void)
-{
-   printf("fib_perf not supported on Windows, skipping test\n");
-   return TEST_SKIPPED;
-}
-
-#else
-
-#include 
-
 #define TEST_FIB_ASSERT(cond) do { \
if (!(cond)) {  \
printf("Error at line %d:\n", __LINE__);\
@@ -246,7 +235,7 @@ static void generate_random_rule_prefix(uint32_t ip_class, 
uint8_t depth)
/* Only generate rest bits except the most significant
 * fixed bits for IP address class
 */
-   start = lrand48() & mask;
+   start = rte_rand() & mask;
ptr_rule = &large_route_table[num_route_entries];
for (k = 0; k < rule_num; k++) {
ptr_rule->ip = (start << (RTE_FIB_MAX_DEPTH - depth))
@@ -265,7 +254,7 @@ static void insert_rule_in_random_pos(uint32_t ip, uint8_t 
depth)
struct route_rule tmp;
 
do {
-   pos = lrand48();
+   pos = rte_rand();
try_count++;
} while ((try_count < 10) && (pos > num_route_entries));
 
@@ -420,6 +409,4 @@ test_fib_perf(void)
return 0;
 }
 
-#endif /* !RTE_EXEC_ENV_WINDOWS */
-
 REGISTER_TEST_COMMAND(fib_perf_autotest, test_fib_perf);
diff --git a/lib/fib/meson.build b/lib/fib/meson.build
index 9b848d08417c..ddcae0617a73 100644
--- a/lib/fib/meson.build
+++ b/lib/fib/meson.build
@@ -2,12 +2,6 @@
 # Copyright(c) 2018 Vladimir Medvedkin 
 # Copyright(c) 2019 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files('rte_fib.c', 'rte_fib6.c', 'dir24_8.c', 'trie.c')
 heade

[PATCH v4 8/9] pcapng: windows compatibility

[Stephen Hemminger]

Allow building on Windows, need to provide some comparability
wrappers for writev() and if_nametoindex.

Signed-off-by: Stephen Hemminger 
Acked-by: Tyler Retzlaff 
---
 lib/pcapng/meson.build  |  6 -
 lib/pcapng/rte_pcapng.c | 59 +++--
 2 files changed, 57 insertions(+), 8 deletions(-)

diff --git a/lib/pcapng/meson.build b/lib/pcapng/meson.build
index da938bbcb733..4549925d41b6 100644
--- a/lib/pcapng/meson.build
+++ b/lib/pcapng/meson.build
@@ -1,12 +1,6 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2019 Microsoft Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files('rte_pcapng.c')
 headers = files('rte_pcapng.h')
 
diff --git a/lib/pcapng/rte_pcapng.c b/lib/pcapng/rte_pcapng.c
index 65c8c77fa405..3c91fc77644a 100644
--- a/lib/pcapng/rte_pcapng.c
+++ b/lib/pcapng/rte_pcapng.c
@@ -3,15 +3,18 @@
  */
 
 #include 
-#include 
 #include 
 #include 
 #include 
 #include 
-#include 
 #include 
 #include 
 
+#ifndef RTE_EXEC_ENV_WINDOWS
+#include 
+#include 
+#endif
+
 #include 
 #include 
 #include 
@@ -20,6 +23,7 @@
 #include 
 #include 
 #include 
+#include 
 #include 
 #include 
 #include 
@@ -47,6 +51,57 @@ static struct pcapng_time {
struct rte_reciprocal_u64 tsc_hz_inverse;
 } pcapng_time;
 
+
+#ifdef RTE_EXEC_ENV_WINDOWS
+/*
+ * Windows does not have writev() call.
+ * Emulate this by copying to a new buffer.
+ * The copy is necessary since pcapng needs to be thread-safe
+ * and do atomic write operations.
+ */
+
+#define IOV_MAX 128
+struct iovec {
+   void   *iov_base;
+   size_t  iov_len;
+};
+
+static ssize_t writev(int fd, const struct iovec *iov, int iovcnt)
+{
+   size_t bytes = 0;
+   uint8_t *ptr;
+   void *tmp_buf;
+   ssize_t ret;
+   int i;
+
+   for (i = 0; i < iovcnt; i++)
+   bytes += iov[i].iov_len;
+
+   if (unlikely(bytes == 0))
+   return 0;
+
+   tmp_buf = malloc(bytes);
+   if (unlikely(tmp_buf == NULL)) {
+   errno = ENOMEM;
+   return -1;
+   }
+
+   ptr = tmp_buf;
+   for (i = 0; i < iovcnt; i++) {
+   rte_memcpy(ptr, iov[i].iov_base, iov[i].iov_len);
+   ptr += iov[i].iov_len;
+   }
+
+   ret = write(fd, tmp_buf, bytes);
+   free(tmp_buf);
+   return ret;
+}
+
+#define IF_NAMESIZE16
+/* compatibility wrapper because name is optional */
+#define if_indextoname(ifindex, ifname) NULL
+#endif
+
 static inline void
 pcapng_init(void)
 {
-- 
2.39.1

[PATCH v4 9/9] table: enable build on Windows

[Stephen Hemminger]

There doesn't seem to be anything keeping this from building on Windows.

Signed-off-by: Stephen Hemminger 
---
 lib/table/meson.build | 6 --
 1 file changed, 6 deletions(-)

diff --git a/lib/table/meson.build b/lib/table/meson.build
index 4f826b4b27e8..f8cef24b5918 100644
--- a/lib/table/meson.build
+++ b/lib/table/meson.build
@@ -1,12 +1,6 @@
 # SPDX-License-Identifier: BSD-3-Clause
 # Copyright(c) 2017 Intel Corporation
 
-if is_windows
-build = false
-reason = 'not supported on Windows'
-subdir_done()
-endif
-
 sources = files(
 'rte_swx_keycmp.c',
 'rte_swx_table_em.c',
-- 
2.39.1

Re: [PATCH 0/2] lib/reorder: fix drain/free issues

[Thomas Monjalon]

07/01/2023 16:19, Volodymyr Fialko:
> This patch address issues with reorder drain/free,
> discovered with enabled `RTE_LIBRTE_MEMPOOL_DEBUG`.
> 
> Volodymyr Fialko (2):
>   reorder: invalidate buf from ready queue in drain
>   test/reorder: fix double free of drained buffers

Applied, thanks.

Re: [PATCH 1/3] reorder: add new drain up to seq number API

[Thomas Monjalon]

We are waiting for a detailed review.
The series could be merged without review.

Please fix cosmetic comments below and send a new version to check CI result.

20/01/2023 11:21, Volodymyr Fialko:
> Introduce new reorder drain API:
> `rte_reorder_drain_up_to_seqn` - exhaustively drain all inserted mbufs
> up to the given sequence number.
> 
> Currently there's no ability to force the drain from reorder buffer,
> i.e. only consecutive ordered or ready packets could be drained.
> New function would give user ability to drain inserted packets, without
> need to wait for missing or newer packets.
> 
> Signed-off-by: Volodymyr Fialko 
[...]
> --- a/lib/reorder/rte_reorder.h
> +++ b/lib/reorder/rte_reorder.h
> @@ -167,6 +167,31 @@ unsigned int
>  rte_reorder_drain(struct rte_reorder_buffer *b, struct rte_mbuf **mbufs,
>   unsigned max_mbufs);
>  
> +/**
> + * @warning
> + * @b EXPERIMENTAL: this API may change without prior notice
> + *
> + * Fetch set of reordered packets up to specified sequence number (exclusive)
> + *
> + * Returns a set of in-order packets from the reorder buffer structure.
> + * Gaps may be present since reorder buffer will try to fetch all possible 
> packets up to given
> + * sequence number.
> + *
> + * @param b
> + *   Reorder buffer instance from which packets are to be drained.
> + * @param mbufs
> + *   Array of mbufs where reordered packets will be inserted from reorder 
> buffer.
> + * @param max_mbufs
> + *   The number of elements in the mbufs array.
> + * @param seqn
> + *   Sequence number up to which buffer will be drained.
> + * @return
> + *   Number of mbuf pointers written to mbufs. 0 <= N < max_mbufs.
> + */
> +__rte_experimental
> +unsigned int
> +rte_reorder_drain_up_to_seqn(struct rte_reorder_buffer *b, struct rte_mbuf 
> **mbufs,
> + unsigned int max_mbufs, rte_reorder_seqn_t seqn);
>  #ifdef __cplusplus

blank line missing

> --- a/lib/reorder/version.map
> +++ b/lib/reorder/version.map
> @@ -16,4 +16,5 @@ EXPERIMENTAL {
>   global:
>  
>   rte_reorder_seqn_dynfield_offset;
> + rte_reorder_drain_up_to_seqn;

Please add a comment "# added in 23.03" as we do in other libs.

Re: [PATCH 3/3] test/reorder: add cases to cover new API

[Thomas Monjalon]

20/01/2023 11:21, Volodymyr Fialko:
> Add new test cases to cover `rte_reorder_drain_up_to_seqn` and
> `rte_reorder_min_seqn_set`.

Instead of doing a separate patch, please squash the tests
when adding the new API functions in patches 1 & 2.

Re: [PATCH v2 0/3] power: remove experimental empty poll API

[Thomas Monjalon]

08/02/2023 11:48, David Hunt:
> The empty poll mechanism was an experimental API added in
> commit id 450f0791312c. It aimed to allow power saving depending
> on the traffic profile. However, it required a training phase
> and required the user to adjust magic numbers depending on
> their workload.
> 
> A new and improved mechanism was added in commit id
> 682a645438c5, also based on empty polls, implemented a
> callback mechanism which added 'monitor', 'pause' and 'scale'
> modes in l3fwd-power. This was and easier mechanism to use,
> so the original empty poll mechanism is no longer needed.
> 
> This patch set removes the experimental empty poll API, the
> empty poll mode from l3fwd-power, and related documentation.
> 
> This is based on a deprecation notice in the previous release.
> 
> [1/3] examples/power: remove empty poll mode from
> [2/3] libs/power: remove experimental empty poll API
> [3/3] doc/power: remove empty poll documentation

Squashed and applied, thanks.

Re: [PATCH] build: fix invalid characters in toolchain definitions

[Thomas Monjalon]

08/02/2023 19:08, Tyler Retzlaff:
> On Tue, Feb 07, 2023 at 03:22:00PM +, Bruce Richardson wrote:
> > When using "icx" (Intel(R) oneAPI DPC++/C++ Compiler) to build DPDK,
> > meson reports the toolchain as "intel-llvm"[1]. This value is used
> > directly to define the RTE_TOOLCHAIN macros, which means that we end up
> > with the invalid macro name "RTE_TOOLCHAIN_INTEL-LLVM", and getting the
> > compiler warning:
> > 
> > ./rte_build_config.h:422:28: warning: ISO C99 requires whitespace after the 
> > macro name [-Wc99-extensions]
> > 
> > This can be fixed, and the macro generation made more robust generally,
> > by adding "underscorify()" on the string. This replaces the "-", and
> > any other invalid characters, with "_" [2].
> > 
> > [1] https://mesonbuild.com/Reference-tables.html#compiler-ids
> > [2] 
> > https://mesonbuild.com/Reference-manual_elementary_str.html#strunderscorify
> > 
> > Fixes: afd18fa21b5e ("build: set toolchain info during meson configure")
> > Cc: sta...@dpdk.org
> > 
> > Reported-by: Keith Wiles 
> > Signed-off-by: Bruce Richardson 
> 
> Acked-by: Tyler Retzlaff 

Applied, thanks.

Re: [PATCH v1 00/13] graph enhancement for multi-core dispatch

[Thomas Monjalon]

This series doesn't look reviewed.
What is the status?

17/11/2022 06:09, Zhirun Yan:
> Currently, rte_graph supports RTC (Run-To-Completion) model within each
> of a single core.
> RTC is one of the typical model of packet processing. Others like
> Pipeline or Hybrid are lack of support.
> 
> The patch set introduces a 'generic' model selection which is a
> self-reacting scheme according to the core affinity.
> The new model enables a cross-core dispatching mechanism which employs a
> scheduling work-queue to dispatch streams to other worker cores which
> being associated with the destination node. When core flavor of the
> destination node is a default 'current', the stream can be continue
> executed as normal.
> 
> Example:
> 3-node graph targets 3-core budget
> 
> Generic Model
> RTC:
> Config Graph-A: node-0->current; node-1->current; node-2->current;
> Graph-A':node-0/1/2 @0, Graph-A':node-0/1/2 @1, Graph-A':node-0/1/2 @2
> 
> + - - - - - - - - - - - - - - - - - - - - - +
> 'Core #0/1/2'
> '   '
> ' ++ +-+ ++ '
> ' | Node-0 | --> | Node-1  | --> | Node-2 | '
> ' ++ +-+ ++ '
> '   '
> + - - - - - - - - - - - - - - - - - - - - - +
> 
> Pipeline:
> Config Graph-A: node-0->0; node-1->1; node-2->2;
> Graph-A':node-0 @0, Graph-A':node-1 @1, Graph-A':node-2 @2
> 
> + - - - - - -+ +- - - - - - + + - - - - - -+
> '  Core #0   ' '  Core #1   ' '  Core #2   '
> '' '' ''
> ' ++ ' ' ++ ' ' ++ '
> ' | Node-0 | ' --> ' | Node-1 | ' --> ' | Node-2 | '
> ' ++ ' ' ++ ' ' ++ '
> '' '' ''
> + - - - - - -+ +- - - - - - + + - - - - - -+
> 
> Hybrid:
> Config Graph-A: node-0->current; node-1->current; node-2->2;
> Graph-A':node-0/1 @0, Graph-A':node-0/1 @1, Graph-A':node-2 @2
> 
> + - - - - - - - - - - - - - - - + + - - - - - -+
> 'Core #0' '  Core #2   '
> '   ' ''
> ' ++ ++ ' ' ++ '
> ' | Node-0 | --> | Node-1 | ' --> ' | Node-2 | '
> ' ++ ++ ' ' ++ '
> '   ' ''
> + - - - - - - - - - - - - - - - + + - - - - - -+
>   ^
>   |
>   |
> + - - - - - - - - - - - - - - - + |
> 'Core #1' |
> '   ' |
> ' ++ ++ ' |
> ' | Node-0 | --> | Node-1 | ' +
> ' ++ ++ '
> '   '
> + - - - - - - - - - - - - - - - +
> 
> 
> The patch set has been break down as below:
> 
> 1. Split graph worker into common and default model part.
> 2. Inline graph node processing and graph circular buffer walking to make
>   it reusable.
> 3. Add set/get APIs to choose worker model.
> 4. Introduce core affinity API to set the node run on specific worker core.
>   (only use in new model)
> 5. Introduce graph affinity API to bind one graph with specific worker
>   core.
> 6. Introduce graph clone API.
> 7. Introduce stream moving with scheduler work-queue in patch 8,9,10.
> 8. Add stats for new models.
> 9. Abstract default graph config process and integrate new model into
>   example/l3fwd-graph. Add new parameters for model choosing.
> 
> We could run with new worker model by this:
> ./dpdk-l3fwd-graph -l 8,9,10,11 -n 4 -- -p 0x1 --config="(0,0,9)" -P
> --model="generic"
> 
> References:
> https://static.sched.com/hosted_files/dpdkuserspace22/a6/graph%20introduce%20remote%20dispatch%20for%20mult-core%20scaling.pdf
> 
> Zhirun Yan (13):
>   graph: split graph worker into common and default model
>   graph: move node process into inline function
>   graph: add macro to walk on graph circular buffer
>   graph: add get/set graph worker model APIs
>   graph: introduce core affinity API
>   graph: introduce graph affinity API
>   graph: introduce graph clone API for other worker core
>   graph: introduce stream moving cross cores
>   graph: enable create and destroy graph scheduling workqueue
>   graph: introduce graph walk by cross-core dispatch
>   graph: enable graph generic scheduler model
>   graph: add stats for corss-core dispatching
>   examples/l3fwd-graph: introduce generic worker model
> 
>  examples/l3fwd-graph/main.c | 218 +--
>  lib/graph/graph.c   | 179 +
>  lib/graph/graph_debug.c |   6 +
>  lib/graph/graph_populate.c  |   1 +
>  lib/graph/graph_private.h   |  44 +++
>  lib/graph/graph_stats.c |  74 +++-
>  lib/graph/meson.build   |  

Re: [PATCH v2] net/ice: fix ice dcf control thread crash

[Thomas Monjalon]

16/02/2023 08:53, Zhang, Ke1X:
> From: Thomas Monjalon 
> > 13/02/2023 08:16, Ke Zhang:
> > > --- a/lib/eal/windows/include/pthread.h
> > > +++ b/lib/eal/windows/include/pthread.h
> > > +static inline void
> > > +pthread_exit(__rte_unused void *__retval) { }
> > 
> > Please don't add more shim layer.
> > There is a new layer rte_thread_* in lib/eal/include/rte_thread.h
> > 
> Thanks for your comments.
> Do I need add a function like rte_thread_exit() in in 
> lib/eal/include/rte_thread.h?

I guess yes.

> There is no function for supporting pthread_exit.

Tyler, how would you achieve the equivalent of pthread_exit?

Re: [PATCH v2] dma/ioat: remove printf within ioat driver

[Thomas Monjalon]

07/02/2023 11:48, Conor Walsh:
> rte_log should be used instead of printf within the drivers. This patch
> changes a printf within the ioat dma driver's start function to an rte log.
> The printfs within ioat's dump function will be retained
> 
> Signed-off-by: Conor Walsh 
> Acked-by: Bruce Richardson 

Applied, thanks.

RE: [PATCH 2/3] net/nfp: get rid of the usage of RTE log level type

[Chaoyong He]

> On 2/17/2023 2:45 AM, Chaoyong He wrote:
> > Register the own RX/TX debug log level type, and get rid of the usage
> > of RTE_LOGTYPE_*. Then we can control the log by a independent switch.
> >
> > Signed-off-by: Chaoyong He 
> > Reviewed-by: Niklas Söderlund 
> > ---
> >  drivers/net/nfp/nfp_logs.c | 10 ++
> > drivers/net/nfp/nfp_logs.h |  8 ++--
> >  2 files changed, 16 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/net/nfp/nfp_logs.c b/drivers/net/nfp/nfp_logs.c
> > index 48c42fe53f..cd58bcee43 100644
> > --- a/drivers/net/nfp/nfp_logs.c
> > +++ b/drivers/net/nfp/nfp_logs.c
> > @@ -5,6 +5,16 @@
> >
> >  #include "nfp_logs.h"
> >
> > +#include 
> > +
> >  RTE_LOG_REGISTER_SUFFIX(nfp_logtype_init, init, NOTICE);
> > RTE_LOG_REGISTER_SUFFIX(nfp_logtype_driver, driver, NOTICE);
> > RTE_LOG_REGISTER_SUFFIX(nfp_logtype_cpp, cpp, NOTICE);
> > +
> > +#ifdef RTE_ETHDEV_DEBUG_RX
> > +RTE_LOG_REGISTER_SUFFIX(nfp_logtype_rx, rx, DEBUG) #endif
> > +
> > +#ifdef RTE_ETHDEV_DEBUG_TX
> > +RTE_LOG_REGISTER_SUFFIX(nfp_logtype_tx, tx, DEBUG) #endif
> > diff --git a/drivers/net/nfp/nfp_logs.h b/drivers/net/nfp/nfp_logs.h
> > index b7632ee72c..315a57811c 100644
> > --- a/drivers/net/nfp/nfp_logs.h
> > +++ b/drivers/net/nfp/nfp_logs.h
> > @@ -15,15 +15,19 @@ extern int nfp_logtype_init;  #define
> > PMD_INIT_FUNC_TRACE() PMD_INIT_LOG(DEBUG, " >>")
> >
> >  #ifdef RTE_ETHDEV_DEBUG_RX
> > +extern int nfp_logtype_rx;
> >  #define PMD_RX_LOG(level, fmt, args...) \
> > -   RTE_LOG(level, PMD, "%s() rx: " fmt "\n", __func__, ## args)
> > +   rte_log(RTE_LOG_ ## level, nfp_logtype_rx, \
> > +   "%s(): " fmt "\n", __func__, ## args)
> >  #else
> >  #define PMD_RX_LOG(level, fmt, args...) do { } while (0)  #endif
> >
> >  #ifdef RTE_ETHDEV_DEBUG_TX
> > +extern int nfp_logtype_tx;
> >  #define PMD_TX_LOG(level, fmt, args...) \
> > -   RTE_LOG(level, PMD, "%s() tx: " fmt "\n", __func__, ## args)
> > +   rte_log(RTE_LOG_ ## level, nfp_logtype_tx, \
> > +   "%s(): " fmt "\n", __func__, ## args)
> >  #else
> >  #define PMD_TX_LOG(level, fmt, args...) do { } while (0)  #endif
> 
> Intention is to replace 'RTE_LOG_DP' with 'PMD_RX_LOG'/'PMD_TX_LOG',
> but these are not exactly same (although difference is minor).
> 
> When 'RTE_ETHDEV_DEBUG_RX' is set, ethdev layer also adds some
> additional load, although I believe that will small comparing to logging in
> driver.
> If 'RTE_LOG_DP' used, the ethdev layer cost can be removed.
> 
> With 'RTE_LOG_DP', log level more verbose than requested won't cause any
> performance impact. Like if ERR level requested, INFO, DEBUG etc logs will
> be compiled out and won't cause any performance impact.
> But with 'RTE_ETHDEV_DEBUG_RX', even log level only request ERR, all
> logging will add cost of at least an if branch (checking log level).
> 
> 
> For many cases I am not sure these differences matters, and already many
> drivers directly uses 'RTE_ETHDEV_DEBUG_RX' as done here. So you may
> prefer to keep as it is.
> 
> But if there is a desire for this fine grain approach, it is possible to add a
> version of 'RTE_LOG_DP' macro that accepts dynamic log type (instead of
> static RTE_LOGTYPE_# type), what do you think?
> 

Thanks for the suggestion.
For now, we prefer to keep as it is.
If we does need the more refined design in the future, we would follow your 
advice here, thanks again.

[Bug 1161] [dpdk-23.03]virtio_user_as_exceptional_path/vhost_exception_path_with_virtio_user: launch testpmd as virtio-user failed

[bugzilla]

https://bugs.dpdk.org/show_bug.cgi?id=1161

lingwei (weix.l...@intel.com) changed:

   What|Removed |Added

 Status|IN_PROGRESS |RESOLVED
 Resolution|--- |FIXED

--- Comment #3 from lingwei (weix.l...@intel.com) ---
Patch have merged into the DPDK main branch.

Verified PASSED.

OS: Ubuntu 22.04.1 LTS/Linux 5.15.45-051545-generic

NIC hardware: Intel Ethernet Controller XL710 for 40GbE QSFP+ 1583

Driver/NIC firmware: i40e-2.22.8/firmware-version: 9.10 0x8000d01e 1.3295.0

commit be26e898ffb3a1719c88bf025d45a847d3a60ffc
Author: Maxime Coquelin 
Date:   Fri Feb 17 13:31:19 2023 +0100

net/virtio-user: fix vhost-kernel initialization

This patch fixes a regression causing devices with
Vhost-kernel backends initialization to fail.

Indeed, Vhost-kernel backend init expects
dev->max_queue_pairs to be set at setup time, while its
assignment was moved after backend setup when adding
control queue support for Vhost-vdpa backends.

It is safe to set dev->max_queue_pairs early for
Vhost-kernel backends as if the backend does not support
multiqueue, its setup will fail.

Bugzilla ID: 1161
Fixes: 7be724856315 ("net/virtio-user: get max number of queue pairs from
device")

Reported-by: Wei Ling 
Signed-off-by: Maxime Coquelin 

-- 
You are receiving this mail because:
You are the assignee for the bug.

回复: [PATCH v5 2/2] eal: add power mgmt support on Arm

[Feifei Wang]

> -邮件原件-
> 发件人: Stephen Hemminger 
> 发送时间: Saturday, February 18, 2023 12:23 AM
> 收件人: Feifei Wang 
> 抄送: Ruifeng Wang ; dev@dpdk.org; nd
> ; David Hunt 
> 主题: Re: [PATCH v5 2/2] eal: add power mgmt support on Arm
> 
> On Wed, 14 Dec 2022 16:14:30 +0800
> Feifei Wang  wrote:
> 
> > +__check_val_size(const uint8_t sz)
> > +{
> > +   switch (sz) {
> > +   case sizeof(uint8_t):  /* fall-through */
> > +   case sizeof(uint16_t): /* fall-through */
> > +   case sizeof(uint32_t): /* fall-through */
> > +   case sizeof(uint64_t): /* fall-through */
> > +   return 0;
> > +   default:
> > +   /* unexpected size */
> > +   return -1;
> > +   }
> > +}
> > +#endif
> 
> One simplification would be to get rid of this function and just check for
> unexpected size in the switch statement in rte_power_monitor().


Thanks for the comments.

__check_val_size API is following intel path. And agree with your comments, 
for arm path, it is unnecessary.

Thus I will delete __check_val_size and simplify the code.

Best Regards
Feifei
> 
> > +   switch (pmc->size) {
> > +   case sizeof(uint8_t):
> > +   __RTE_ARM_LOAD_EXC_8(pmc->addr, cur_value,
> __ATOMIC_RELAXED);
> > +   __RTE_ARM_WFE()
> > +   break;
> > +   case sizeof(uint16_t):
> > +   __RTE_ARM_LOAD_EXC_16(pmc->addr, cur_value,
> __ATOMIC_RELAXED);
> > +   __RTE_ARM_WFE()
> > +   break;
> > +   case sizeof(uint32_t):
> > +   __RTE_ARM_LOAD_EXC_32(pmc->addr, cur_value,
> __ATOMIC_RELAXED);
> > +   __RTE_ARM_WFE()
> > +   break;
> > +   case sizeof(uint64_t):
> > +   __RTE_ARM_LOAD_EXC_64(pmc->addr, cur_value,
> __ATOMIC_RELAXED);
> > +   __RTE_ARM_WFE()
> 
>   default:
>   return -1; /* unexpected size */
> > +   }

RE: [PATCH v2 2/4] config/arm: add NVIDIA BlueField-3 platform

[Ruifeng Wang]

> -Original Message-
> From: Raslan Darawsheh 
> Sent: Thursday, February 16, 2023 10:04 PM
> To: tho...@monjalon.net
> Cc: dev@dpdk.org; Ruifeng Wang ; Bruce Richardson
> 
> Subject: [PATCH v2 2/4] config/arm: add NVIDIA BlueField-3 platform
> 
> Add meson build configuration for NVIDIA BlueField-3 platform with 
> Cortex-A78AE cores.
> 
> Signed-off-by: Raslan Darawsheh 
> ---
>  config/arm/meson.build | 16 
>  1 file changed, 16 insertions(+)
> 
A cross-file 'config/arm/arm64_bluefield3_linux_gcc' is also needed.

Thanks and regards
/Ruifeng

RE: [PATCH] app/testpmd: fix secondary process not forwarding

[Zhang, Yuying]

> -Original Message-
> From: He, ShiyangX 
> Sent: 2023年2月8日 14:39
> To: Zhang, Yuying 
> Cc: dev@dpdk.org; Zhou, YidingX ;
> sta...@dpdk.org; Singh, Aman Deep ;
> Burakov, Anatoly ; Li, Xiaoyun
> ; Alvin Zhang 
> Subject: RE: [PATCH] app/testpmd: fix secondary process not forwarding
> 
> 
> 
> >-Original Message-
> >From: Zhang, Yuying 
> >Sent: Wednesday, February 8, 2023 11:22 AM
> >To: He, ShiyangX 
> >Cc: dev@dpdk.org; Zhou, YidingX ;
> >sta...@dpdk.org; Singh, Aman Deep ;
> Burakov,
> >Anatoly ; Li, Xiaoyun
> >; Alvin Zhang 
> >Subject: RE: [PATCH] app/testpmd: fix secondary process not forwarding
> >
> >Hi Shiyang,
> >
> >> -Original Message-
> >> From: He, ShiyangX 
> >> Sent: 2023年1月13日 17:08
> >> To: He, ShiyangX ; Zhang, Yuying
> >> 
> >> Cc: dev@dpdk.org; Zhou, YidingX ;
> >> sta...@dpdk.org; Singh, Aman Deep ;
> >> Burakov, Anatoly ; Li, Xiaoyun
> >> ; Alvin Zhang 
> >> Subject: RE: [PATCH] app/testpmd: fix secondary process not
> >> forwarding
> >>
> >> @Zhang, Yuying Hi, please take a look at this patch! Are there any
> comments?
> >>
> >> >-Original Message-
> >> >From: He, ShiyangX 
> >> >Sent: Wednesday, January 4, 2023 10:02 AM
> >> >To: Stephen Hemminger 
> >> >Cc: dev@dpdk.org; Zhou, YidingX ;
> >> >sta...@dpdk.org; Singh, Aman Deep ;
> >Zhang,
> >> >Yuying ; Burakov, Anatoly
> >> >; Li, Xiaoyun ;
> >> >Alvin Zhang 
> >> >Subject: RE: [PATCH] app/testpmd: fix secondary process not
> >> >forwarding
> >> >
> >> >>> Under multi-process scenario, the secondary process gets queue
> >> >>> state from the wrong location (the global variable 'ports').
> >> >>> Therefore, the secondary process can not forward since
> >> >>> "stream_init" is
> >not called.
> >> >>>
> >> >>> This commit fixes the issue by calling 'rte_eth_rx/tx_queue_info_get'
> >> >>> to get queue state from shared memory.
> >> >>>
> >> >>> Fixes: a78040c990cb ("app/testpmd: update forward engine
> >> >>> beginning")
> >> >>> Cc: sta...@dpdk.org
> >> >>>
> >> >>> Signed-off-by: Shiyang He 

Acked-by: Yuying Zhang 

> >> >>
> >> >>Would it be possible to fix this the initialization of ports
> >> >>variable, rather than doing a per-state fixup here?
> >> >
> >> >In multi-process scenario, the secondary process does not initialize
> >> >the queue state in the 'ports' variable, and the ethdev's queue
> >> >state may be changed by any other process, which causes 'ports'
> >> >queue state of per-process and ethdev's queue state are
> >> >inconsistent. Therefore, getting the queue state from ethdev is a feasible
> way which I can think of.
> >
> >You should fix the queue state in the 'ports' variable of secondary
> >process in the initialization of ports variable instead of fixup here.
> 
> If the ethdev's queue state is changed by other processes, the queue state in
> the 'ports' variable is inconsistent with ethdev's queue state. Therefore,
> should obtain it from ethdev when accessing the queue state in the 'ports'
> variable.

[PATCH 1/2] eventdev: fix race condition in fast-path set function

[Ashok Kaladi]

If eventdev enqueue or dequeue function is called during
event_dev_fp_ops_set(), it may get pre-empted after setting
the function pointers, but before setting the pointer to port data.
In this case the newly registered enqueue/dequeue function will use
dummy port data and end up in seg fault.

This patch moves the updation of data element to the beginning of
event_dev_fp_ops_set() function.

Fixes: d35e61322de5 ("eventdev: move inline APIs into separate structure")
Cc: sta...@dpdk.org

Signed-off-by: Ashok Kaladi 

diff --git a/.mailmap b/.mailmap
index 5015494210..bbf6744278 100644
--- a/.mailmap
+++ b/.mailmap
@@ -132,6 +132,7 @@ Ashish Jain 
 Ashish Paul 
 Ashish Sadanandan 
 Ashish Shah 
+Ashok Kaladi 
 Ashwin Sekhar T K  
 Asim Jamshed 
 Aviad Yehezkel 
diff --git a/lib/eventdev/eventdev_private.c b/lib/eventdev/eventdev_private.c
index 1d3d9d357e..539aade780 100644
--- a/lib/eventdev/eventdev_private.c
+++ b/lib/eventdev/eventdev_private.c
@@ -107,6 +107,7 @@ void
 event_dev_fp_ops_set(struct rte_event_fp_ops *fp_op,
 const struct rte_eventdev *dev)
 {
+   fp_op->data = dev->data->ports;
fp_op->enqueue = dev->enqueue;
fp_op->enqueue_burst = dev->enqueue_burst;
fp_op->enqueue_new_burst = dev->enqueue_new_burst;
@@ -117,5 +118,4 @@ event_dev_fp_ops_set(struct rte_event_fp_ops *fp_op,
fp_op->txa_enqueue = dev->txa_enqueue;
fp_op->txa_enqueue_same_dest = dev->txa_enqueue_same_dest;
fp_op->ca_enqueue = dev->ca_enqueue;
-   fp_op->data = dev->data->ports;
 }
-- 
2.25.1

[PATCH 2/2] ethdev: fix race condition in fast-path ops setup

[Ashok Kaladi]

If ethdev enqueue or dequeue function is called during
eth_dev_fp_ops_setup(), it may get pre-empted after setting
the function pointers, but before setting the pointer to port data.
In this case the newly registered enqueue/dequeue function will use
dummy port data and end up in seg fault.

This patch moves the updation of each data pointers before updating
corresponding function pointers.

Fixes: c87d435a4d79 ("ethdev: copy fast-path API into separate structure")
Cc: sta...@dpdk.org

Signed-off-by: Ashok Kaladi 

diff --git a/lib/ethdev/ethdev_private.c b/lib/ethdev/ethdev_private.c
index 48090c879a..a0232c669f 100644
--- a/lib/ethdev/ethdev_private.c
+++ b/lib/ethdev/ethdev_private.c
@@ -270,17 +270,17 @@ void
 eth_dev_fp_ops_setup(struct rte_eth_fp_ops *fpo,
const struct rte_eth_dev *dev)
 {
+   fpo->rxq.data = dev->data->rx_queues;
fpo->rx_pkt_burst = dev->rx_pkt_burst;
+   fpo->txq.data = dev->data->tx_queues;
fpo->tx_pkt_burst = dev->tx_pkt_burst;
fpo->tx_pkt_prepare = dev->tx_pkt_prepare;
fpo->rx_queue_count = dev->rx_queue_count;
fpo->rx_descriptor_status = dev->rx_descriptor_status;
fpo->tx_descriptor_status = dev->tx_descriptor_status;
 
-   fpo->rxq.data = dev->data->rx_queues;
fpo->rxq.clbk = (void **)(uintptr_t)dev->post_rx_burst_cbs;
 
-   fpo->txq.data = dev->data->tx_queues;
fpo->txq.clbk = (void **)(uintptr_t)dev->pre_tx_burst_cbs;
 }
 
-- 
2.25.1

RE: [PATCH v4 06/12] app/mldev: add test case to interleave inferences

[Anup Prabhu]

> -Original Message-
> From: Srikanth Yalavarthi 
> Sent: Tuesday, February 7, 2023 9:20 PM
> To: Srikanth Yalavarthi 
> Cc: dev@dpdk.org; Shivah Shankar Shankar Narayan Rao
> ; Jerin Jacob Kollanukkaran
> ; Anup Prabhu ; Prince Takkar
> ; Parijat Shukla 
> Subject: [PATCH v4 06/12] app/mldev: add test case to interleave inferences
> 
> Added test case to interleave inference requests from multiple models.
> Interleaving would load and start all models and launch inference requests
> for the models using available queue-pairs
> 
> Operations sequence when testing with N models and R reps,
> 
> (load + start) x N -> (enqueue + dequeue) x N x R ...
>   -> (stop + unload) x N
> 
> Test can be executed by selecting "inference_interleave" test.
> 
> Signed-off-by: Srikanth Yalavarthi 
> ---
>  app/test-mldev/meson.build |   1 +
>  app/test-mldev/ml_options.c|   3 +-
>  app/test-mldev/test_inference_common.c |  12 +--
>  app/test-mldev/test_inference_interleave.c | 118
> +
>  4 files changed, 127 insertions(+), 7 deletions(-)  create mode 100644
> 
Acked-by: Anup Prabhu 

<>

Re: [PATCH 2/2] ethdev: fix race condition in fast-path ops setup

[fengchengwen]

On 2023/2/20 14:08, Ashok Kaladi wrote:
> If ethdev enqueue or dequeue function is called during
> eth_dev_fp_ops_setup(), it may get pre-empted after setting
> the function pointers, but before setting the pointer to port data.
> In this case the newly registered enqueue/dequeue function will use
> dummy port data and end up in seg fault.
> 
> This patch moves the updation of each data pointers before updating
> corresponding function pointers.
> 
> Fixes: c87d435a4d79 ("ethdev: copy fast-path API into separate structure")
> Cc: sta...@dpdk.org
> 
> Signed-off-by: Ashok Kaladi 
> 
> diff --git a/lib/ethdev/ethdev_private.c b/lib/ethdev/ethdev_private.c
> index 48090c879a..a0232c669f 100644
> --- a/lib/ethdev/ethdev_private.c
> +++ b/lib/ethdev/ethdev_private.c
> @@ -270,17 +270,17 @@ void
>  eth_dev_fp_ops_setup(struct rte_eth_fp_ops *fpo,
>   const struct rte_eth_dev *dev)
>  {
> + fpo->rxq.data = dev->data->rx_queues;
>   fpo->rx_pkt_burst = dev->rx_pkt_burst;
> + fpo->txq.data = dev->data->tx_queues;
>   fpo->tx_pkt_burst = dev->tx_pkt_burst;
>   fpo->tx_pkt_prepare = dev->tx_pkt_prepare;
>   fpo->rx_queue_count = dev->rx_queue_count;
>   fpo->rx_descriptor_status = dev->rx_descriptor_status;
>   fpo->tx_descriptor_status = dev->tx_descriptor_status;
>  
> - fpo->rxq.data = dev->data->rx_queues;
>   fpo->rxq.clbk = (void **)(uintptr_t)dev->post_rx_burst_cbs;
>  
> - fpo->txq.data = dev->data->tx_queues;
>   fpo->txq.clbk = (void **)(uintptr_t)dev->pre_tx_burst_cbs;

Hi Ashok,

The modification is OK for the x86 platform (which has strong memory order, and 
will keep write-after-write order in here, and read-after-read in 
rte_eth_rx/tx_burst),
but for other weak memory order (like ARM platform) will fail.

For the weak memory order, suggest add write-mb in here, and read-mb in 
rte_eth_rx/tx_burst.
But the read-mb in rte_eth_rx/tx_burst will affect performance, especially the 
variable will changes only once when start.

So I suggest use write-mb + delay in here:
   fpo->rxq.data = dev->data->rx_queues;
   fpo->txq.data = dev->data->tx_queues;
   mdelay(5); // delay e.g. 5ms
   fpo->rx_pkt_burst = dev->rx_pkt_burst;
   fpo->tx_pkt_burst = dev->tx_pkt_burst;

And also cc ARMv8 maintainer.

>  }
>  
> 

Re: [PATCH 2/2] ethdev: fix race condition in fast-path ops setup

[fengchengwen]

Sorry resend, because forget one line.

On 2023/2/20 14:08, Ashok Kaladi wrote:
> If ethdev enqueue or dequeue function is called during
> eth_dev_fp_ops_setup(), it may get pre-empted after setting
> the function pointers, but before setting the pointer to port data.
> In this case the newly registered enqueue/dequeue function will use
> dummy port data and end up in seg fault.
> 
> This patch moves the updation of each data pointers before updating
> corresponding function pointers.
> 
> Fixes: c87d435a4d79 ("ethdev: copy fast-path API into separate structure")
> Cc: sta...@dpdk.org
> 
> Signed-off-by: Ashok Kaladi 
> 
> diff --git a/lib/ethdev/ethdev_private.c b/lib/ethdev/ethdev_private.c
> index 48090c879a..a0232c669f 100644
> --- a/lib/ethdev/ethdev_private.c
> +++ b/lib/ethdev/ethdev_private.c
> @@ -270,17 +270,17 @@ void
>  eth_dev_fp_ops_setup(struct rte_eth_fp_ops *fpo,
>   const struct rte_eth_dev *dev)
>  {
> + fpo->rxq.data = dev->data->rx_queues;
>   fpo->rx_pkt_burst = dev->rx_pkt_burst;
> + fpo->txq.data = dev->data->tx_queues;
>   fpo->tx_pkt_burst = dev->tx_pkt_burst;
>   fpo->tx_pkt_prepare = dev->tx_pkt_prepare;
>   fpo->rx_queue_count = dev->rx_queue_count;
>   fpo->rx_descriptor_status = dev->rx_descriptor_status;
>   fpo->tx_descriptor_status = dev->tx_descriptor_status;
>  
> - fpo->rxq.data = dev->data->rx_queues;
>   fpo->rxq.clbk = (void **)(uintptr_t)dev->post_rx_burst_cbs;
>  
> - fpo->txq.data = dev->data->tx_queues;
>   fpo->txq.clbk = (void **)(uintptr_t)dev->pre_tx_burst_cbs;

Hi Ashok,

The modification is OK for the x86 platform (which has strong memory order, and 
will keep write-after-write order in here, and read-after-read in 
rte_eth_rx/tx_burst),
but for other weak memory order (like ARM platform) will fail.

For the weak memory order, suggest add write-mb in here, and read-mb in 
rte_eth_rx/tx_burst.
But the read-mb in rte_eth_rx/tx_burst will affect performance, especially the 
variable will changes only once when start.

So I suggest use write-mb + delay in here:
   fpo->rxq.data = dev->data->rx_queues;
   fpo->txq.data = dev->data->tx_queues;
   wmb();
   mdelay(5); // delay e.g. 5ms
   fpo->rx_pkt_burst = dev->rx_pkt_burst;
   fpo->tx_pkt_burst = dev->tx_pkt_burst;

And also cc ARMv8 maintainer.

>  }
>  
> 

[Bug 1159] lib/vhost build failure with clang 3.4.2

[bugzilla]

https://bugs.dpdk.org/show_bug.cgi?id=1159

David Marchand (david.march...@redhat.com) changed:

   What|Removed |Added

   Assignee|dev@dpdk.org|david.march...@redhat.com
 CC||david.march...@redhat.com
 Status|UNCONFIRMED |RESOLVED
 Resolution|--- |FIXED

--- Comment #1 from David Marchand (david.march...@redhat.com) ---
Thanks for reporting Ali.
This should be fixed with: bb9d6b230a4a ("disable lock annotation with clang
3.4.2").

-- 
You are receiving this mail because:
You are the assignee for the bug.

Re: [PATCH] net/hns3: support disable IOVA as PA mode

[Thomas Monjalon]

16/02/2023 09:36, Ruifeng Wang:
> From: Chengwen Feng 
> > Subject: [PATCH] net/hns3: support disable IOVA as PA mode

Could we change the title to "support IOVA as VA" ?

> > 
> > Claim PMD supports pmd_supports_disable_iova_as_pa.
> > 
> > Signed-off-by: Chengwen Feng 
> > ---
> >  drivers/net/hns3/hns3_rxtx_vec_neon.h |  2 +-  
> > drivers/net/hns3/hns3_rxtx_vec_sve.c  | 13
> > +
> >  drivers/net/hns3/meson.build  |  7 +--
> >  3 files changed, 11 insertions(+), 11 deletions(-)
> > 
> Reviewed-by: Ruifeng Wang 

Re: [PATCH] eal: use same atomic intrinsics for gcc and clang

[Thomas Monjalon]

18/02/2023 02:58, Honnappa Nagarahalli:
> The size generic atomic intrinsics generate the same
> code as the size specific intrinsics for gcc. Use size
> generic intrinsics for both gcc and clang.
> 
> Fixes: 7bdccb93078e ("eal: fix ARM build with clang")
> Cc: sta...@dpdk.org
> Cc: pbhagavat...@marvell.com
> 
> Signed-off-by: Honnappa Nagarahalli 
> Reviewed-by: Ola Liljedahl 
> Reviewed-by: Ruifeng Wang 
> Acked-by: Morten Brørup 
> Acked-by: Tyler Retzlaff 

Applied, thanks.