[dpdk-dev] [PATCH] net/octeontx: register fpa as platform HW mempool

[dpdk-dev] [PATCH] net/octeontx: register fpa as platform HW mempool

[Pavan Nikhilesh]

Register octeontx-fpavf as platform HW mempool when net/octeontx pmd is
used.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
---

 This patch depends on http://dpdk.org/dev/patchwork/patch/34239 patchset.

 drivers/net/octeontx/octeontx_ethdev.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/net/octeontx/octeontx_ethdev.c b/drivers/net/octeontx/octeontx_ethdev.c
index adca3435e..2b4a07d2b 100644
--- a/drivers/net/octeontx/octeontx_ethdev.c
+++ b/drivers/net/octeontx/octeontx_ethdev.c
@@ -16,6 +16,7 @@
 #include <rte_dev.h>
 #include <rte_kvargs.h>
 #include <rte_malloc.h>
+#include <rte_mbuf_pool_ops.h>
 #include <rte_prefetch.h>
 #include <rte_bus_vdev.h>

@@ -1326,6 +1327,7 @@ octeontx_probe(struct rte_vdev_device *dev)
 		res = -ENOTSUP;
 		goto parse_error;
 	}
+	rte_mbuf_set_platform_mempool_ops("octeontx_fpavf");
 	probe_once = 1;

 	return 0;
--
2.16.0

Re: [dpdk-dev] [PATCH] net/octeontx: register fpa as platform HW mempool

[Ferruh Yigit]

On 1/22/2018 3:45 PM, Pavan Nikhilesh wrote:
> Register octeontx-fpavf as platform HW mempool when net/octeontx pmd is
> used.
> 
> Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
> ---
> 
>  This patch depends on http://dpdk.org/dev/patchwork/patch/34239 patchset.

This patch was waiting dependent patch, which seems merged now.

But now because of "__rte_experimental" tag in
rte_mbuf_set_platform_mempool_ops() that this patch uses getting build errors [1].

Need to add a special note to pmd makefile to allow experimental API usage:
CFLAGS += -DALLOW_EXPERIMENTAL_API



[1]
...dpdk/drivers/net/octeontx/octeontx_ethdev.c:1330:2: error:
'rte_mbuf_set_platform_mempool_ops' is deprecated: Symbol is not yet part of
stable ABI [-Werror,-Wdeprecate
d-declarations]


        rte_mbuf_set_platform_mempool_ops("octeontx_fpavf");
        ^

...dpdk/x86_64-native-linuxapp-clang/include/rte_mbuf_pool_ops.h:37:5: note:
'rte_mbuf_set_platform_mempool_ops' has been explicitly marked deprecated here
int __rte_experimental

    ^


...dpdk/x86_64-native-linuxapp-clang/include/rte_compat.h:107:16: note: expanded
from macro '__rte_experimental'

__attribute__((deprecated("Symbol is not yet part of stable ABI"), \
               ^

1 error generated.

[dpdk-dev] [PATCH v2] net/octeontx: register fpa as platform HW mempool

[Pavan Nikhilesh]

Register octeontx-fpavf as platform HW mempool when net/octeontx pmd is
used.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
---

 v2 Changes:
 - fixed compilation issues caused by using experimental API in both meson and
 make.

 drivers/net/octeontx/Makefile          | 2 ++
 drivers/net/octeontx/meson.build       | 2 ++
 drivers/net/octeontx/octeontx_ethdev.c | 2 ++
 3 files changed, 6 insertions(+)

diff --git a/drivers/net/octeontx/Makefile b/drivers/net/octeontx/Makefile
index b47bf7a40..3e4a10662 100644
--- a/drivers/net/octeontx/Makefile
+++ b/drivers/net/octeontx/Makefile
@@ -43,6 +43,8 @@ else
 CFLAGS_octeontx_rxtx.o += -O3 -Ofast
 endif

+CFLAGS_octeontx_ethdev.o += -DALLOW_EXPERIMENTAL_API
+
 LDLIBS += -lrte_eal -lrte_mbuf -lrte_mempool -lrte_ring
 LDLIBS += -lrte_ethdev -lrte_net -lrte_kvargs
 LDLIBS += -lrte_mempool_octeontx
diff --git a/drivers/net/octeontx/meson.build b/drivers/net/octeontx/meson.build
index ef789eaa9..135e63e2f 100644
--- a/drivers/net/octeontx/meson.build
+++ b/drivers/net/octeontx/meson.build
@@ -8,6 +8,8 @@ sources = files('octeontx_rxtx.c',
 		'octeontx_ethdev.c'
 		)

+cflags += '-DALLOW_EXPERIMENTAL_API'
+
 deps += ['mempool_octeontx', 'eventdev']

 includes += include_directories('base')
diff --git a/drivers/net/octeontx/octeontx_ethdev.c b/drivers/net/octeontx/octeontx_ethdev.c
index 06aa0a2c9..b739c0b39 100644
--- a/drivers/net/octeontx/octeontx_ethdev.c
+++ b/drivers/net/octeontx/octeontx_ethdev.c
@@ -16,6 +16,7 @@
 #include <rte_dev.h>
 #include <rte_kvargs.h>
 #include <rte_malloc.h>
+#include <rte_mbuf_pool_ops.h>
 #include <rte_prefetch.h>
 #include <rte_bus_vdev.h>

@@ -1326,6 +1327,7 @@ octeontx_probe(struct rte_vdev_device *dev)
 		res = -ENOTSUP;
 		goto parse_error;
 	}
+	rte_mbuf_set_platform_mempool_ops("octeontx_fpavf");
 	probe_once = 1;

 	return 0;
--
2.16.0

Re: [dpdk-dev] [PATCH v2] net/octeontx: register fpa as platform HW mempool

[santosh]

On Sunday 04 February 2018 09:03 PM, Pavan Nikhilesh wrote:
> Register octeontx-fpavf as platform HW mempool when net/octeontx pmd is
> used.
>
> Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
> ---

Acked-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>

Re: [dpdk-dev] [PATCH v2] net/octeontx: register fpa as platform HW mempool

[Ferruh Yigit]

On 2/4/2018 3:33 PM, Pavan Nikhilesh wrote:
> Register octeontx-fpavf as platform HW mempool when net/octeontx pmd is
> used.
> 
> Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
> ---
> 
>  v2 Changes:
>  - fixed compilation issues caused by using experimental API in both meson and
>  make.
> 
>  drivers/net/octeontx/Makefile          | 2 ++
>  drivers/net/octeontx/meson.build       | 2 ++

Meson patches are still going through the Bruce's tree. Can you please make this
two separate patches (not patchset), I can get the driver one and Makefile,
Bruce can get meson one.

Thanks,
ferruh

Re: [dpdk-dev] [PATCH v2] net/octeontx: register fpa as platform HW mempool

[Pavan Nikhilesh]

On Mon, Feb 05, 2018 at 10:34:00AM +0000, Ferruh Yigit wrote:
> On 2/4/2018 3:33 PM, Pavan Nikhilesh wrote:
> > Register octeontx-fpavf as platform HW mempool when net/octeontx pmd is
> > used.
> >
> > Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
> > ---
> >
> >  v2 Changes:
> >  - fixed compilation issues caused by using experimental API in both meson and
> >  make.
> >
> >  drivers/net/octeontx/Makefile          | 2 ++
> >  drivers/net/octeontx/meson.build       | 2 ++
>
> Meson patches are still going through the Bruce's tree. Can you please make this
> two separate patches (not patchset), I can get the driver one and Makefile,
> Bruce can get meson one.

Just a heads up doing so would break meson build on next-net branch.
Will split the patches and resend them.

>
> Thanks,
> ferruh

Thanks,
Pavan.

Re: [dpdk-dev] [PATCH v2] net/octeontx: register fpa as platform HW mempool

[Bruce Richardson]

On Mon, Feb 05, 2018 at 10:34:00AM +0000, Ferruh Yigit wrote:
> On 2/4/2018 3:33 PM, Pavan Nikhilesh wrote:
> > Register octeontx-fpavf as platform HW mempool when net/octeontx pmd is
> > used.
> > 
> > Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
> > ---
> > 
> >  v2 Changes:
> >  - fixed compilation issues caused by using experimental API in both meson and
> >  make.
> > 
> >  drivers/net/octeontx/Makefile          | 2 ++
> >  drivers/net/octeontx/meson.build       | 2 ++
> 
> Meson patches are still going through the Bruce's tree. Can you please make this
> two separate patches (not patchset), I can get the driver one and Makefile,
> Bruce can get meson one.
> 
Hi Ferruh,

I disagree with this. Since the meson build is now merged into mainline,
many patches will have meson file changes to keep things building. These
patches should not go in the next-build tree but should be part of the
tree the rest of the patch targets. It should not be permitted to apply
a patch to next-net which breaks the meson build.

Only patches which add new capabilities or drivers to the build system
will go in my tree.

Regards,
/Bruce

Re: [dpdk-dev] [PATCH v2] net/octeontx: register fpa as platform HW mempool

[Ferruh Yigit]

On 2/5/2018 11:46 AM, Bruce Richardson wrote:
> On Mon, Feb 05, 2018 at 10:34:00AM +0000, Ferruh Yigit wrote:
>> On 2/4/2018 3:33 PM, Pavan Nikhilesh wrote:
>>> Register octeontx-fpavf as platform HW mempool when net/octeontx pmd is
>>> used.
>>>
>>> Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
>>> ---
>>>
>>>  v2 Changes:
>>>  - fixed compilation issues caused by using experimental API in both meson and
>>>  make.
>>>
>>>  drivers/net/octeontx/Makefile          | 2 ++
>>>  drivers/net/octeontx/meson.build       | 2 ++
>>
>> Meson patches are still going through the Bruce's tree. Can you please make this
>> two separate patches (not patchset), I can get the driver one and Makefile,
>> Bruce can get meson one.
>>
> Hi Ferruh,
> 
> I disagree with this. Since the meson build is now merged into mainline,
> many patches will have meson file changes to keep things building. These
> patches should not go in the next-build tree but should be part of the
> tree the rest of the patch targets. It should not be permitted to apply
> a patch to next-net which breaks the meson build.
> 
> Only patches which add new capabilities or drivers to the build system
> will go in my tree.

OK, so I will merge this with original patch.

> 
> Regards,
> /Bruce
> 

Re: [dpdk-dev] [PATCH v2] net/octeontx: register fpa as platform HW mempool

[Bruce Richardson]

On Sun, Feb 04, 2018 at 09:03:09PM +0530, Pavan Nikhilesh wrote:
> Register octeontx-fpavf as platform HW mempool when net/octeontx pmd is
> used.
> 
> Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
> ---
> 
>  v2 Changes:
>  - fixed compilation issues caused by using experimental API in both meson and
>  make.
> 
>  drivers/net/octeontx/Makefile          | 2 ++
>  drivers/net/octeontx/meson.build       | 2 ++
>  drivers/net/octeontx/octeontx_ethdev.c | 2 ++
>  3 files changed, 6 insertions(+)
> 
> diff --git a/drivers/net/octeontx/Makefile b/drivers/net/octeontx/Makefile
> index b47bf7a40..3e4a10662 100644
> --- a/drivers/net/octeontx/Makefile
> +++ b/drivers/net/octeontx/Makefile
> @@ -43,6 +43,8 @@ else
>  CFLAGS_octeontx_rxtx.o += -O3 -Ofast
>  endif
> 
> +CFLAGS_octeontx_ethdev.o += -DALLOW_EXPERIMENTAL_API
> +
>  LDLIBS += -lrte_eal -lrte_mbuf -lrte_mempool -lrte_ring
>  LDLIBS += -lrte_ethdev -lrte_net -lrte_kvargs
>  LDLIBS += -lrte_mempool_octeontx
> diff --git a/drivers/net/octeontx/meson.build b/drivers/net/octeontx/meson.build
> index ef789eaa9..135e63e2f 100644
> --- a/drivers/net/octeontx/meson.build
> +++ b/drivers/net/octeontx/meson.build
> @@ -8,6 +8,8 @@ sources = files('octeontx_rxtx.c',
>  		'octeontx_ethdev.c'
>  		)
> 
> +cflags += '-DALLOW_EXPERIMENTAL_API'
> +

Rather than having each driver modify cflags, I've set up the variable
"allow_experimental_apis" in the top-level meson.build file. So just put

allow_experimental_apis = true

in your driver file.

/Bruce

[dpdk-dev] [PATCH v3] net/octeontx: register fpa as platform HW mempool

[Pavan Nikhilesh]

Register octeontx-fpavf as platform HW mempool when net/octeontx pmd is
used.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
Acked-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>
---

  v3 Changes:
  - spilt meson specific changes to a separate patch.

  v2 Changes:
  - fixed compilation issues caused by using experimental API in both meson and
  make.

 drivers/net/octeontx/Makefile          | 2 ++
 drivers/net/octeontx/octeontx_ethdev.c | 2 ++
 2 files changed, 4 insertions(+)

diff --git a/drivers/net/octeontx/Makefile b/drivers/net/octeontx/Makefile
index b47bf7a40..3e4a10662 100644
--- a/drivers/net/octeontx/Makefile
+++ b/drivers/net/octeontx/Makefile
@@ -43,6 +43,8 @@ else
 CFLAGS_octeontx_rxtx.o += -O3 -Ofast
 endif

+CFLAGS_octeontx_ethdev.o += -DALLOW_EXPERIMENTAL_API
+
 LDLIBS += -lrte_eal -lrte_mbuf -lrte_mempool -lrte_ring
 LDLIBS += -lrte_ethdev -lrte_net -lrte_kvargs
 LDLIBS += -lrte_mempool_octeontx
diff --git a/drivers/net/octeontx/octeontx_ethdev.c b/drivers/net/octeontx/octeontx_ethdev.c
index 06aa0a2c9..b739c0b39 100644
--- a/drivers/net/octeontx/octeontx_ethdev.c
+++ b/drivers/net/octeontx/octeontx_ethdev.c
@@ -16,6 +16,7 @@
 #include <rte_dev.h>
 #include <rte_kvargs.h>
 #include <rte_malloc.h>
+#include <rte_mbuf_pool_ops.h>
 #include <rte_prefetch.h>
 #include <rte_bus_vdev.h>

@@ -1326,6 +1327,7 @@ octeontx_probe(struct rte_vdev_device *dev)
 		res = -ENOTSUP;
 		goto parse_error;
 	}
+	rte_mbuf_set_platform_mempool_ops("octeontx_fpavf");
 	probe_once = 1;

 	return 0;
--
2.14.1

Re: [dpdk-dev] [PATCH v3] net/octeontx: register fpa as platform HW mempool

[Ferruh Yigit]

On 2/5/2018 10:55 AM, Pavan Nikhilesh wrote:
> Register octeontx-fpavf as platform HW mempool when net/octeontx pmd is
> used.
> 
> Signed-off-by: Pavan Nikhilesh <pbhagavatula@caviumnetworks.com>
> Acked-by: Santosh Shukla <santosh.shukla@caviumnetworks.com>

Applied to dpdk-next-net/master, thanks.